\n \n | Afrotheria (common name) | \n Non-afrotherian analogues (common name) | \n Basis for analogy | \n
\n \n | Hemicentetes semispinosus (streaked tenrec), Setifer setosus (greater hedgehog tenrec), Tenrec ecaudatus (tailless tenrec) | \n Erinaceus europaeus (European hedgehog) | \n Overall appearance, body shape and size, foraging behaviour, terrestrial locomotion with climbing ability. | \n
\n \n | Limnogale mergulus (web-footed tenrec) | \n Ornithorhynchus anatinus (duck-billed platypus), Sorex araneus (Eurasian shrew) | \n Morphology (S. araneus), semi-aquatic lifestyle and habitat (O. anatinus) | \n
\n \n | Orizoryctes tetradactylus (four-toed rice tenrec) | \n Talpa europaea (European mole) | \n Overall appearance, fossorial locomotion, | \n
\n \n | Potamogale velox (otter shrew) | \n Lutra lutra (Eurasian otter), Ornithorhynchus anatinus (duck-billed platypus) | \n Semi-aquatic lifestyle, overall appearance (L. lutra), foraging behaviour | \n
\n \n | Amblysomus hottentotus (Hottentot golden mole), Chrysochloris asiatica (Cape golden mole) | \n Cryptomys hottentotus (common mole-rat), Talpa europaea (European mole) | \n Overall morphology, body size, fossorial locomotion, and subterranean habitat | \n
\n \n | Elephantulus rufescens (East African long-eared elephant shrew), | \n Microtus arvalis (common vole), Mus musculus (house mouse), Rattus rattus (black rat), | \n Overall body shape and size (except R. rattus), fast, terrestrial locomotion | \n
\n
\n\n \n | Elephantulus rupestris (Western rock elephant shrew), Macroscelestes proboscideus (round-eared elephant shrew) | \n Sorex araneus (Eurasian shrew) | \n (although no living “microcursorial” mammals other than elephant shrews exist) | \n
\n \n | Petrodromus tetradactylus (four-toed sengi), Rhynchocyon cirnei (checkered sengi) | \n Oryctolagus cuniculus (European rabbit), Rattus rattus (black rat), Tragulus javanicus (Javan mouse deer) | \n Fast, terrestrial locomotion, relative hindlimb size (O. cuniculus, T. javanicus) | \n
\n \n | Orycteropus afer (aardvark) | \n Manis tricuspis (tree pangolin), Tolypeutes matacus (southern three-banded armadillo) | \n Fossorial/burrowing locomotion, foraging behaviour | \n
\n \n | Heterohyrax brucei (bush hyrax), Procavia capensis (rock hyrax) | \n Marmota marmota (alpine marmot), Ochotona rufescens (Afghan pika) | \n Rocky/open/mountainous habitat, terrestrial and climbing locomotion | \n
\n \n | Dendrohyrax arboreus (tree hyrax) | \n Didelphis sp. (opossum) Tamandua mexicana (northern tamandua) | \n Arboreality, climbing locomotion | \n
\n \n | Dugong dugon (dugong), Trichechus sp. (manatee) | \n Odobenus rosmarus (walrus), Phoca vitulina (harbour seal), Sotalia fluviatilis (tucuxi/gray dolphin) | \n Aquatic lifestyle, overall body size and shape (O. rosmarus), foraging behaviour (O. rosmarus) | \n
\n \n | Elephas maximus (Asian elephant) | \n n/a | \n n/a | \n
\n
\n\nAll afrotherian “orders” are represented. Higher taxonomic groups of the analogues are as follows: Xenarthra (Tamandua mexicana, Tolypeutes matacus), Carnivora (Lutra lutra, Phoca vitulina, Odobenus rosmarus), Pholidota (Manis tricuspis), Cetartiodactyla (Tragulus javanicus, Sotalia fluviatilis), Eulipotyphla (Erinaceaeus europaeus, Sorex araneus, Talpa europaea), Rodentia (Mus musculus, Rattus rattus, Microtus arvalis, Marmota marmota, Cryptomys hottentotus), Lagomorpha (Oryctolagus cuniculus, Ochotona rufescens), Marsupialia (Didelphis sp.), and Monotremata (Ornithorhynchus anatinus).\n\nContextual data\n\nRelevant data on positional behaviour (locomotion and posture) and certain aspects of ecology were collected from the literature for each species. Data that directly measure auditory capacity could unfortunately not be included because common parameters of hearing ability (e.g., minimum/maximum perceived frequency, or frequency of greatest hearing sensitivity) were\nunknown for many species in the sample. However, habitat type (e.g., aquatic, subterranean, or ground-dwelling) carries information about the medium through which species detect sound and may thus indirectly also reflect basic aspects of hearing. Multiple aspects of ecology and positional behaviour were captured by 11 ordinal variables, presented in Table 2. Locomotion and postures of mammals vary extensively and they have been studied in varying levels of detail for the species in our sample. For the purpose of this study, positional behaviour was measured on an ordinal scale. Species’ mean body mass correlates with both ecology and locomotor behaviour across mammals, and so it was included as an additional contextual variable, measured on an interval scale.\n\nTable 2 The 12 contextual variables of body size, ecology, and positional behaviour used in this study.\n\n\n \n | Type | \n Variables \\( p \\) | \n Measurement levels | \n Notes | \n
\n \n | Body size | \n \\( p = 1 \\) (body mass) | \n Continuous (in g) | \n | \n
\n \n | Food acquisition | \n \\( p = 1 \\) (pursuit of moving prey) | \n 1 = never or rarely,
2 = mixed diet / sometimes,
3 = often or always | \n See text for further explanation. | \n
\n \n | Habitat type | \n \\( p = 3 \\)
- aquatic;
- ground-dwelling;
- arboreal | \n 1 = never or rarely,
2 = sometimes,
3 = often or always.
For aquatic only:
1 = never or rarely,
2 = moves freely on land and in water,
3 = mainly aquatic, only rests on land,
4 = strictly aquatic. | \n Subterranean habitats are captured by a fossorial mode of locomotion (see below). | \n
\n \n | Mode of locomotion | \n \\( p = 4 \\)
- scansorial/climbing;
- cursorial/running;
- leaping/jumping/hopping;
- fossorial/burrowing | \n 1 = typical/frequent,
2 = occasional,
3 = atypical/rare. | \n | \n
\n \n | Dimensionality of movement | \n \\( p = 1 \\) (moves three-dimensionally) | \n 1 = never or rarely,
2 = sometimes,
3 = often or always. | \n Refers to whether animals move in 3D at any given point in time. | \n
\n \n | Agility | \n \\( p = 1 \\) (agility) | \n 1 = extra slow,
2 = slow, | \n Capturing both speed and manoeuvrability. | \n
\n
\n\n \n | # | \n Type | \n Definition | \n
\n \n | Anatomical landmarks | \n
\n | 1 | | ASC ampulla to vestibule (midline) |
\n | 2 | | ASC ampulla to slender part of ASC (midline) |
\n | 3 | | CC to ASC & PSC (midline) |
\n | 4 | | CC to vestibule (midline) |
\n | 5 | | Slender part of PSC to ampulla (midline) |
\n | 6 | | PSC ampulla to vestibule (midline) |
\n | 7 | | Slender part of LSC to vestibule (midline) |
\n | 8 | | Slender part of LSC to ampulla (midline) |
\n | 9 | | LSC ampulla to vestibule (midline) |
\n | 10 | | Helicotrema |
\n | 11 | | Centre of the round window |
\n | 12 | | Posterior extremity of the axis of maximum elongation of the oval window |
\n | 13 | | Anterior extremity of the axis of maximum elongation of the oval window |
\n \n | Semilandmarks | \n
\n | 1-20 | SL (20) | Midline curve of the ASC: LM 1 → 2 → 3 (2 SL: ampulla, 18 SL: slender part) |
\n | 21-40 | SL (20) | Midline curve of the PSC: LM 6 → 5 → 3 (2 SL: ampulla, 18 SL: slender part) |
\n | 41-45 | SL (5) | Midline curve of the CCR: LM 3 → 4 |
\n | 46-65 | SL (20) | Midline curve of the LSC: LM 9 → 8 → 7 (2 SL: ampulla, 18 SL: slender part) |
\n | 66-105 | SL (40) | Midline curve of the cochlea: apical side of the RW → LM 10 |
\n | 106-111 | SL (6) | Outline of the oval window: LM 13 → 12 (inferior), then 12 → 13 (superior) |
\n
\n\nLandmarks 1-10 and all semilandmarks are defined as in Gunz et al. (2012), except for the different numbers of semilandmarks along each curve. For the semilandmarks, we indicate the order in which they are placed between anatomical landmarks. LM, anatomical landmark, SL, semilandmark. See Fig. 2 for visualisation.\nData analysis\n\nIn order to standardise for overall location, scale, and orientation of the landmark configurations, we performed a Generalised Procrustes Analysis (GPA) with mirroring allowed because the sample comprised both left and right ears (Rohlf & Slice, 1990; Mitteroecker & Gunz, 2009). We chose the minimum bending energy as a criterion to slide semilandmarks (Gunz et al., 2005; Gunz and Mitteroecker, 2013). Due to the huge variation in cochlear morphology, the magnitude of sliding per iteration was reduced by a factor of 0.1 and the number of iterations was increased accordingly.\n\nIn the case of adaptive convergence, a given afrotherian species is expected to share more similarities in labyrinth morphology with analogues than non-analogues. In order to test this, we computed pairwise Procrustes distances between taxa and compared average distances for pairs of analogues and for pairs of non-analogues. Because elephants do not have a living analogue, we excluded this species in the computations. We also computed the average Procrustes distance for pairs of afrotherians, to compare the signal of convergent evolution with that of phylogenetic history. Significance tests of the H0 of equal average distances were based on permutation tests that permuted the affiliation of species across Afrotheria-analogue pairs.\n\nTo explore the morphological diversity of the bony labyrinth across species, we performed a principal component analysis (PCA) of the Procrustes shape coordinates (results are shown in the Supplementary Information). We also performed a two-block partial least squares (2B-PLS) analysis to investigate the association between bony labyrinth shape and the 12 contextual variables (Rohlf & Corti, 2000; Bookstein et al., 2003; Mitteroecker & Bookstein, 2007). The contextual variables were mean-centred and scaled to unit variance. Additionally, we conducted a phylogenetic 2B-PLS, based on phylogenetic generalised least squares (PGLS; Adams & Felice, 2014) and the phylogeny in Fig. 1, in order to assess whether the associations between bony labyrinth shape and the contextual variables were mediated through the species’ phylogenetic relationships (Figs. S3-4). PCA, 2B-PLS, and shape visualisations were performed using the software Wolfram Mathematica 12. Scatter plots were generated using the packages ggplot2 3.4.2 (Wickham, 2016) and ggrepel 0.9.3 (Slowikowski, 2023) of the software R version 4.2.2 (R Development Core team, 2022).\n\nFinally, to evaluate to what extent the associations between bony labyrinth shape, ecology and positional behaviour are phylogenetically patterned, we quantified the phylogenetic signal of the\nPLS dimensions by means of Blomberg’s \\( K \\) (Blomberg et al. 2003). Phylogenetic signal, or the tendency for closely related species to resemble each other in ecology and/or morphology, is expected to be high when close relatives are more similar to each other than they are to distant relatives. Conversely, phylogenetic signal is low when trait values vary randomly across the phylogenetic tree or when distant relatives tend to resemble each other more than closely related species, as in evolutionary convergence (e.g. Blomberg et al. 2003; Revell et al. 2008). To compute Blomberg's \\( K \\) for the shape and context scores along the PLS dimensions we used the function ‘phylosig’ from the package phytools (Revell 2012) and the phylogenetic tree, with branch lengths scaled to time, in Fig. 1.\n\nData availability\n\nAll data are available in an online data repository (https://osf.io/9mtwh/?view_only=88a550a637f84615b5666d69278309d4).\n\nAcknowledgements\n\nP.M., N.D.S.G., C.P. and A.L.M. disclose support for the research of this work from the Austrian Science Fund [FWF grant number P33736]. The California Academy of Sciences, Museum of Vertebrate Zoology at Berkeley, as well as Museum of Comparative Zoology, Harvard University, and Peabody Museum, Yale University provided access to some CT-scans, the collection of which was funded by oVert TCN. Robert Asher provided access to data originally appearing in Asher et al. 2017 (http://dx.doi.org/10.1080/02724634.2017.1317638). Deborah Bird provided access to data originally appearing in Bird et al. 2018 (https://doi.org/10.1098/rspb.2018.0100), the collection of which was funded by NSF Graduate Research Fellowship Program grant DGE-1144087. Ted Macrini provided access to other data, with data collection funded by NSF DEB-0309369 and data upload to MorphoSource funded by DBI-1902242. All the files were downloaded from www.MorphoSource.org, Duke University.\n\nReferences\n\nAdams, D. C. & Felice R. N. Assessing trait covariation and morphological integration on phylogenies using evolutionary covariance matrices. *PLoS ONE* **9**, e94335 (2014).\nAltenberg, L. In Modularity: Understanding the Development and Evolution of Complex Natural Systems (eds Callebaut, W. & Rasskin-Gutman, D.) 99-128 (MIT Press, 2005).\nAnthwal, N. & Thompson, H. The development of the mammalian outer and middle ear. J. Anat. **228(2)**, 217–232 (2016).\nAsher, R. J., Novacek, M. J. & Geisler J. H. Relationships of Endemic African Mammals and Their Fossil Relatives Based on Morphological and Molecular Evidence. *J. Mamm. Evol.* **10(1)**, 131-194 (2003).\nBenoit, J., Crumpton, N., Merigeaud, S., & Tabuce, R. Petrosal and bony labyrinth morphology supports paraphyly of *Elephantulus* within Macroscelididae (Mammalia, Afrotheria). *J. Mamm. Evol.*, **21(2)**, 173–193 (2014).\nBenoit, J., et al. A test of the lateral semicircular canal correlation to head posture, diet and other biological traits in \"ungulate\" mammals. *Sci. Rep.*, **10(1)**, 19602 (2020).\nBerlioz, E., Cornette, R., Lenoir, N., Santin, M. D. & Lehmann, T. Exploring the ontogenetic development of the inner ear in Aardvarks. *J. Anat.* **238(5)**, 1128-1142 (2021).\nBillet, G., Germain, D., Ruf, I., de Muizon, C., & Hautier, L. The inner ear of *Megatherium* and the evolution of the vestibular system in sloths. *J. Anat.*, **223(6)**, 557–567 (2013).\nBininda-Emonds, O. R. P. et al. The delayed rise of present-day mammals. *Nature* **446**, 507-512 (2007).\nBlomberg, S. P., Garland, T. Jr, Ives, A. R. Testing for phylogenetic signal in comparative data: behavioral traits are more labile. *Evolution* **57(4)**, 717-745 (2003).\nBookstein, F. L. et al. Cranial integration in *Homo*: singular warps analysis of the midsagittal plane in ontogeny and evolution. *J. Hum. Evol.* **44(2)**, 167–187 (2003).\nBoucherie, A., Polet, C., Lefèvre, P. & Vercauteren, M. Sexing the bony labyrinth: A morphometric investigation in a subadult and adult Belgian identified sample. *J. Forensic Sci.* **66(3)**, 808-820 (2021).\nBraga, J. et al. Cochlear shape reveals that the human organ of hearing is sex-typed from birth. *Sci. Rep.* **9(1)**, 10889 (2019).\nCosteur, L., et al. The bony labyrinth of toothed whales reflects both phylogeny and habitat preferences. *Sci. Rep.*, **8(1)**, 7841 (2018).\nCosteur, L., Mennecart, B., Müller, B., & Schulz, G. Prenatal growth stages show the development of the ruminant bony labyrinth and petrosal bone. *J. Anat.*, **230(2)**, 347-353 (2017).\nCrumpton, N., Kardjilov, N., & Asher, R. J. Convergence vs. Specialization in the ear region of moles (mammalia). *J. Morphol.*, **276(8)**, 900–914 (2015).\nDella Santina, C. C., Potyagaylo, V., Migliaccio, A. A., Minor, L. B. & Carey, J. P. Orientation of human semicircular canals measured by three-dimensional multiplanar CT reconstruction. JARO **6**, 191-206 (2005).\n\ndel Rio, J., et al. Allometry, function and shape diversification in the inner ear of platyrrhine primates. *J. Mamm. Evol.* **28**, 135-143 (2021).\n\nDooling, R. J., Fay, R. R. & Popper, A. N. *Comparative Hearing: Birds and Reptiles* (Springer, New York, 2000).\n\ndos Reis, M., et al. Phylogenomic datasets provide both precision and accuracy in estimating the timescale of placental mammal phylogeny. *Proc. R. Soc. B.* **279**, 3491–3500 (2012).\n\nEkdale, E. G. Comparative anatomy of the bony labyrinth (inner ear) of placental mammals. *PLoS ONE* **8(6)**, e66624 (2013).\n\nEkdale, E. G. Form and function of the mammalian inner ear. *J. Anat.* **228(2)**, 324–337 (2016).\n\nEkdale, E. G. & Racicot, R. A. Anatomical evidence for low frequency sensitivity in an archaeocete whale: comparison of the inner ear of *Zygohriza kochii* with that of crown Mysticeti. *J. Anat.* **226**, 22–39 (2015).\n\nFleischer, G. Hearing in extinct cetaceans as determined by cochlear structure. *J. Paleontol.* **50**, 133–152 (1976).\n\nFritzsch, B., Pan, N., Jahan, I., Elliot K. L. Inner ear development: building a spiral ganglion and an organ of Corti out of unspecified ectoderm. *Cell Tissue Res.* **361**, 7–24 (2015).\n\nFuchs, J. C. & Tucker, A. S. Development and Integration of the Ear. *Curr. Top. Dev. Biol.* **115**, 213–32 (2015).\n\nGleich O. & Manley G. A. In *Comparative Hearing: Birds and Reptiles* (eds Dooling, R. J, Fay, R. R. & Popper, A. N.) 70-138 (Springer, 2000).\n\nGonzales, L. A., Malinzak, M. D. & Kay, R. F. Intraspecific variation in semicircular canal morphology —A missing element in adaptive scenarios? *Am. J. Phys. Anthropol.* **168(1)**, 10-24 (2019).\n\nGrohé, C., Tseng, Z. J., Lebrun, R., Boistel, R. & Flynn, J. J. Bony labyrinth shape variation in extant Carnivora: a case study of Musteloidea. *J. Anat.* **228(3)**, 366-383 (2016).\n\nGunz, P. & Mitteroecker, P. Semilandmarks: a method for quantifying curves and surfaces. *Hystrix* **24(1)**, 103-109 (2013).\n\nGunz, P., Mitteroecker, P. & Bookstein, F. L. In *Modern morphometrics in physical anthropology* (ed Slice, D. E.) 73-98. (Plenum, 2005).\nGunz, P., Ramsier, M., Kuhrig, M., Hublin, J. J. & Spoor, F. (2012). The mammalian bony labyrinth reconsidered, introducing a comprehensive geometric morphometric approach. J. Anat. **220**(6), 529-543.\n\nHansen, T. F., Pélabon, C. & Houle, D. Heritability is not Evolvability. *Evol. Biol.* **38**, 258-277 (2011).\n\nHeffner, R. S. & Heffner, H. E. In *The Evolutionary Biology of Hearing* (eds Webster, D. B., Fay, R. R. & Popper, A. N.), 691–715 (Springer, 1992).\n\nHoule D. Comparing evolvability and variability of quantitative traits. *Genetics* **130**(1), 195-204 (1992).\n\nHoyte, D. A. N. (1961). The postnatal growth of the ear capsule in the rabbit. Am. J. Anat., 108(1), 1–16.\n\nJones, K. E. et al. PanTHERIA: a species-level database of life history, ecology, and geography of extant and recently extinct mammals: Ecological Archives E090-184. *Ecology* **90**, 2648-2648 (2009).\n\nKöppl, C. & Manley, G. A. A functional perspective on the evolution of the cochlea. *Cold Spring Harbor Persp. Med.* **9**(6), a033241 (2018).\n\nLange, S., Stalleicken, J., & Burda, H. Functional morphology of the ear in fossorial rodents, *Microtus arvalis* and *Arvicola terrestris*. *J. Morphol.*, **262**(3), 770-779 (2004).\n\nLe Maître, A., Grunstra, N. D. S., Pfaff, C. & Mitteroecker, P. Evolution of the Mammalian Ear: An Evolvability Hypothesis. *Evol. Biol.* **47**, 187–192 (2020).\n\nLe Maître, A., Guy, F., Merceron, G. & Kostopoulos, D. S. Morphology of the Bony Labyrinth Supports the Affinities of *Paradolichopithecus* with the Papionina. *Int. J. Primatol.*, **44**(1), 209–236 (2023).\n\nLebrun, R., Perier, A., Masters, J., Marivaux, L. & Couette, S. Lower levels of vestibular developmental stability in slow-moving than fast-moving primates. *Symmetry* **13**(12), 2305 (2021).\n\nLovegrove, B. G. & Mowoe, M. O. The evolution of micro-cursoriality in mammals. *J. Exp. Biol.* **217**, 1316-1325 (2014).\n\nLuo, Z.-X. Developmental patterns in mesozoic evolution of mammal ears. *Annu. Rev. Ecol. Evol. Syst.* **42**, 355–380 (2011).\n\nLuo, Z.-X., Schultz, J. A. & Ekdale, E. In *Evolution of the Vertebrate Ear: Evidence from the Fossil Record* (eds Clack, J.A., Fay, R. R. & Popper, A. N.) 139-174 (Springer, 2016).\n\nMacPhee, R. D. E. & Novacek, M. J. In *Mammal Phylogeny: Placentals* (eds Szalay, F. S., Novacek, M. J. & McKenna, M. C.) 13–31 (Springer, 1993).\n\nManley, G. A. Evolutionary paths to mammalian cochleae. *JARO* **13**(6), 733–743 (2012).\nManoussaki, D. et al. The influence of cochlear shape on low-frequency hearing. Proc. Natl. Acad. Sci. U. S. A. **105**, 6162–6166 (2008).\n\nMason, M. J., Cornwall, H. L. & Smith, E. S. J. Ear structures of the naked mole-rat, *Heterocephalus glaber*, and its relatives (Rodentia: Bathyergidae). Plos ONE, **11(12)**, e0167079 (2016).\n\nMcVean, A. Are the semicircular canals of the European mole, *Talpa europaea*, adapted to a subterranean habitat? Comp. Bioch. Physiol. A: Molec. Int. Physiol. **123(2)**, 173–178 (1999).\n\nMennecart, B., et al. Bony labyrinth morphology clarifies the origin and evolution of deer. Sci. Rep., **7**, (13176). (2017). https://doi.org/10.1038/s41598-017-12848-9\n\nMeredith, R. W. et al. Impacts of the Cretaceous Terrestrial Revolution and Kpg Extinction on Mammal Diversification. Science **334**, 521-524 (2013).\n\nMittermeier, R. A. & Wilson, D. E. Handbook of the Mammals of the World, Vol. 4: Sea Mammals (Lynx Edicions, Barcelona, 2014).\n\nMittermeier, R. A. & Wilson, D. E. Handbook of the Mammals of the World, Vol. 5: Monotremes and Marsupials (Lynx Edicions, Barcelona, 2015).\n\nMittermeier, R. A. & Wilson, D. E. Handbook of the Mammals of the World, Vol. 8: Insectivores, Sloths and Colugos (Lynx Edicions, Barcelona, 2018).\n\nMitteroecker, P. The developmental basis of variational modularity: insights from quantitative genetics, morphometrics, and developmental biology. Evol. Biol. **36**, 377-385 (2009).\n\nMitteroecker, P. & Bookstein, F. The conceptual and statistical relationship between modularity and morphological integration. Syst. Biol. **56(5)**, 818–836 (2007).\n\nMitteroecker, P. & Gunz, P. Advances in geometric morphometrics. Evol. Biol. 36, 235-247 (2009).\n\nMoore, A. M., Hartstone-Rose, A. & Gonzalez-Socoloske, D. Review of sensory modalities of sirenians and the other extant Paenungulata clade. Anat. Rec, **305(3)**, 715-735 (2021).\n\nMourlam, M. J. & Orliac, M. J. Infrasonic and ultrasonic hearing evolved after the emergence of modern whales. Curr. Biol. **27(12)**, 1776-1781 (2017).\n\nMuller, M. Size limitations in semicircular duct systems. J. Theor. Biol., **198(3)**, 405–437 (1999).\n\nMuller, M., & Verhagen, J. H. G. Optimization of the Mechanical Performance of a Two-duct Semicircular Duct System - Part 2: Excitation of Endolymph Movements. J. Theor. Biol., **216(4)**, 425–442 (2002).\n\nOrliac, M. J., Mourlam, M. J., Boisserie, J.-R., Costeur, L., & Lihoreau, F. Evolution of semiaquatic habits in hippos and their extinct relatives: insights from the ear region. Zool. J. Linn. Soc. (2023). https://doi.org/10.1093/zoolinnean/zlac112\nPark, T., Mennecart, B., Costeur, L., Grohé, C., & Cooper, N. Convergent evolution in toothed whale cochleae. BMC Evol. Biol. **19(1)**, 195 (2019).\n\nPavlicev, M. & Wagner, G. P. A model of developmental evolution: selection, pleiotropy and compensation. Trends Ecol. Evol. **27(6)**, 316-22 (2012).\n\nPfaff, C., Martin, T. & Ruf, I. Bony labyrinth morphometry indicates locomotor adaptations in the squirrel-related clade (Rodentia, Mammalia). Proc. Royal Soc. B: Biol. Sci. **282(1809)**, 20150744 (2015).\n\nR Core Team (2022). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/\n\nRevell, L. “phytools: An R package for phylogenetic comparative biology (and other things).” Methods Ecol. Evol. **3**, 217-223 (2012).\n\nRevell, L. J., Harmon, L. J. & Collar, D. C. Phylogenetic Signal, Evolutionary Process, and Rate, *Syst. Biol.* **57(4)**, 591–601 (2008).\n\nRichard, C., et al. Inner ear ossification and mineralization kinetics in human embryonic development-microtomographic and histomorphological study. *Sci. Rep.* **7(1)**, 4825 (2017).\n\nRohlf, F. J. & Corti, M. Use of two-block partial least-squares to study covariation in shape. *Syst. Biol.* **49(4)**, 740–753 (2000).\n\nRohlf, F. J. & Slice, D. Extensions of the Procrustes method for the optimal superimposition of landmarks. *Syst. Biol.* **39(1)**, 40-59 (1990).\n\nRuf, I., Luo, Z.-X. & Martin, T. Re-investigation of the basicranium of *Haldanodon exspectatus* (Docodonta, Mammaliaformes). *J. Vert. Paleontol*. 33, 382–400 (2013).\n\nSchmitt, A. & Gheerbrant, E. The ear region of earliest known elephant relatives: new light on the ancestral morphotype of proboscideans and afrotherians. *J. Anat.* **228(1)**, 137–152 (2016).\n\nSchwab, J. A., Kriwet, J., Weber, G. W. & Pfaff, C. Carnivoran hunting style and phylogeny reflected in bony labyrinth morphometry. *Sci. Rep.* **9(1)**, 1–9 (2019).\n\nSienknecht, U. J. Developmental origin and fate of middle ear structures. *Hear. Res.* **301**, 19–26 (2013).\n\nSlowikowski, K. ggrepel: Automatically Position Non-Overlapping Text Labels with 'ggplot2'. R package version 0.9.3 (2023), https://CRAN.R-project.org/package=ggrepel.\n\nSpoor, F., Bajpai, S., Hussain, S. T., Kumar, K. & Thewissen, J. G. Vestibular evidence for the evolution of aquatic behaviour in early cetaceans. *Nature* **417(6885)**, 163–166 (2002).\n\nSpoor, F. et al. The primate semicircular canal system and locomotion. *Proc. Nat. Acad. Sci. U.S.A.* **104(26)**, 10808-10812 (2007).\nSpringer, M. S. Afrotheria. Curr. Biol. **32(5)**, R205-210 (2022).\nStanhope, M. J. et al. Highly Congruent Molecular Support for a Diverse Superordinal Clade of Endemic African Mammals. Mol. Phylogenet. Evol. **9(3)**, 501-508 (1998).\nTabuce, R, Asher, R. J. & Lehmann, T. Afrotherian mammals: a review of current data. Mammalia **72**, 2-14 (2008).\nUpham, N. S., Esselstyn, J. A. & Jetz, W. Inferring the mammal tree: species-level sets of phylogenies for questions in ecology, evolution, and conservation. *PLOS Biol.* (2019). https://doi.org/10.1371/journal.pbio.3000494\nWagner, G. P. & Altenberg, L. Perspective: Complex adaptations and the evolution of evolvability. *Evolution* **50(3)**, 967-976 (1996).\nWang, H., Meng, J. & Wang, Y. Cretaceous fossil reveals a new pattern in mammalian middle ear evolution. *Nature* **576(7785)**, 102–105 (2019).\nWebster, D. B. Ear structure and function in modern mammals. *Am. Zool.* **6**, 451–466 (1966).\nWest, C. D. The relationship of the spiral turns of the cochlea and the length of the basilar membrane to the range of audible frequencies in ground dwelling mammals. *J. Acoust. Soc. Am.* **77**, 1091–1101 (1985).\nWickham, H. ggplot2: Elegant Graphics for Data Analysis. (Springer-Verlag New York, 2016).\nWilson, D. E., Lacher, T. E. & Mittermeier, R. A. *Handbook of the Mammals of the World, Vol. 6: Lagomorphs and Rodents I* (Lynx Edicions, Barcelona, 2016).\nWilson, D. E. & Mittermeier, R. A. *Handbook of the Mammals of the World, Vol. 1: Carnivores* (Lynx Edicions, Barcelona, 2009).\nWilson, D. E. & Mittermeier, R. A. *Handbook of the Mammals of the World, Vol. 2: Hoofed Mammals* (Lynx Edicions, Barcelona, 2011).\nWilson, D. E., Mittermeier, R. A. & Lacher, T. E. *Handbook of the Mammals of the World, Vol. 7: Rodents II* (Lynx Edicions, Barcelona, 2017).\nYamada, M. & Yoshizaki, F. Osseous labyrinth of Cetacea. *Sci. Rep. Whales Res. Inst.* **14**, 291–304 (1959).\nZachos, F. E. Mammalian Phylogenetics: In *Mammals of Europe - Past, Present, and Future* (eds Hackländer, K. & Zachos, F. E.) 31-48 (Springer, 2020).\nSupplementary Files\n\nThis is a list of supplementary files associated with this preprint. Click to download.\n\n• GrunstraetalSupplementaryInfo.docx"}
{"paper_id": "prb69", "title": "", "text": "Environment-induced heritable variations are common\n\nYuan-Ye Zhang\nzhangyuanye@xmu.edu.cn\n\nXiamen University https://orcid.org/0000-0002-8636-2242\nXiaohe Lin\nXiamen University\nJunjie Yin\nXiamen University\nYifan Wang\nXiamen University\nJing Yao\nXiamen University\nQingshun Q. Li\nGraduate College of Biomedical Sciences, Western University of Health Sciences, Pomona, CA91766, USA\nVit Latzel\nAcademy of Sciences of the Czech Republic\nOliver Bossdorf\nUniversity of Tübingen https://orcid.org/0000-0001-7504-6511\n\nArticle\n\nKeywords:\n\nPosted Date: October 20th, 2023\n\nDOI: https://doi.org/10.21203/rs.3.rs-3416426/v1\n\nLicense: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License\n\nAdditional Declarations: There is NO Competing Interest.\nVersion of Record: A version of this preprint was published at Nature Communications on May 30th, 2024. See the published version at https://doi.org/10.1038/s41467-024-49024-3.\nEnvironment-induced heritable variations are common\n\nXiaohe Lin1†, Junjie Yin1†, Yifan Wang1†, Jing Yao1, Qingshun Q. Li1,2, Vit Latzel3, Oliver Bossdorf4, Yuan-Ye Zhang1*\n\n1Key Laboratory of the Ministry of Education for Coastal and Wetland Ecosystems, College of the Environment and Ecology, Xiamen University, Xiamen, Fujian, China\n2Biomedical Sciences, College of Dental Medicine, Western University of Health Sciences, Pomona, California, USA\n3Institute of Botany of the CAS, Zamek 1, 252 43 Pruhonice, Czech Republic\n4Institute of Ecology & Evolution, University of Tübingen, Auf der Morgenstelle 5, 72076 Tübingen, Germany\n† These authors contributed equally to this work.\n* Correspondence: zhangyuanye@xmu.edu.cn (Y.-Y.Z.)\nAbstract\nParental or ancestral environments can induce heritable phenotypic changes, but whether such environment-induced heritable changes are a common phenomenon remains unexplored. Here, we subjected 14 genotypes of Arabidopsis thaliana to 10 different environmental treatments and observed phenotypic and genome-wide gene expression changes over successive generations. We found that all environmental treatments caused heritable phenotypic and gene expression changes, with a substantial proportion stably transmitted over all observed generations. Intriguingly, the susceptibility of a genotype to environmental inductions could be predicted based on the transposon abundance in the genome. Our study thus challenges the classic view that the environment only participates in the selection of heritable variation and suggests that the environment can broadly contribute to the generation of heritable variation.\nIntroduction\n\nHeritable variation is the foundation of biodiversity and adaptation to the changing climate\\(^{1,2}\\), ultimately generated through mutations and moulded by forces of environmental selection—a tenet well established in textbooks\\(^3\\). Several pioneering studies suggested that heritable variation is also directly introduced by parental or ancestral environments\\(^{4-7}\\), causing stably transmitted phenotypic and gene expression changes that potentially enhance offspring resistance to environmental stress\\(^{4,7,8}\\). However, whether such environment-induced heritable changes are a common phenomenon remains unexplored, and it is also unknown whether the phenomenon is an occasional exception or a fundamental principle. Therefore, exploring the prevalence of environment-induced heritable changes will challenge the textbook knowledge of the origin of heritable variation and provide critical insights for predicting the adaptive response to global environmental changes.\n\nThe phenomenon in which the parental or ancestral environment induces offspring phenotypic changes is conventionally referred to as the parental or transgenerational effect\\(^4\\), and the major challenge in understanding the prevalence of this phenomenon could be the scale of the experiments\\(^9\\). Expanding the scale of these experiments to include different genetic backgrounds has revealed that such environment-induced changes are genotype specific\\(^{10,11}\\) and predictable based on the climate of the genotype’s origin\\(^{12,13}\\), while this predictability decreases when sacrificing the number of genotypes to include various environmental treatments\\(^9\\). Nevertheless, these studies span only one or two offspring generations, and it is unclear whether environment-induced changes are heritably stable. A large-scale experiment encompassing multiple genotypes, environments, and generations is necessary to reveal the prevalence and predictability of environment-induced heritable changes.\n\nTo address the prevalence and predictability of environment-induced heritable changes, we subjected 14 natural accessions (genotypes) of *Arabidopsis thaliana* to 10 ancestral environmental treatments and assessed environment-induced phenotypic changes over four offspring generations (Test I, Fig. 1, Extended Data Fig. 1a, Supplementary Table 1). To explore the reproducibility of environment-induced\nchanges, we planted all four offspring generations together under uniform conditions to eliminate any heterogeneity between generations in the growth conditions (Test II, Fig. 1). Besides performing phenotypic observations, we analysed the genome-wide gene expression of samples from Test II to investigate whether heritable phenotypic changes were predicted by heritable gene expression changes and whether the induced expression changes were nonrandom with respect to environments.\n\nResults\n\nPrevalence of environment-induced heritable phenotypic changes\n\nIn the ancestral generation, we subjected A. thaliana genotypes to the control environment and 10 treatment environments: jasmonic acid addition (JA), control for JA, nutrient addition, nutrient depletion, leaf removal, drought, low and high concentrations of cadmium, and low and high concentrations of salinity. The statistics showed that the genotype, environmental treatment and their interaction significantly affected all phenotypes measured (except for the interaction effect on flowering time, Supplementary Table 2). Among the specific treatments, nutrient addition significantly increased aboveground biomass, and jasmonic acid addition slightly enhanced fruit production, representing benign environments (Extended Data Fig. 2a). Other treatments significantly decreased fruit production and thus represented stressful environments (Extended Data Fig. 2a). High concentrations of cadmium and salinity generated more stressful conditions than the corresponding low-concentration treatments, resulting in either smaller plants with fewer fruits or shorter plants with fewer fruits (Extended Data Fig. 2a). Therefore, our experiment provided a gradient of benign and stressful environments to assess their effects on offspring phenotypes.\n\nIn the offspring generations, we found little evidence of significant main effects of the ancestral environments on phenotypes but significant genotype-by-ancestral_environment interactions (G × A.E) on all phenotypes measured except for fruit number (Extended Data Fig. 2b, Supplementary Table 2). More importantly, the G × A.E interactions remained significant in the four successive offspring generations. These results demonstrated that the effects of the ancestral environments on offspring phenotypes (i.e., transgenerational effects) depended on the genotypes and persisted for\nfour offspring generations. To understand the stable transmission of such genotype-specific changes caused by ancestral environments, we estimated the effect sizes of changes for different genotypes and environments and plotted them in the order of generations (Fig. 2a). The plot showed considerable variation in the magnitude and direction of the phenotypic changes. In some notable cases, such as the response of genotype Abd-0 to high cadmium and the response of TRE-1 to high salinity, we found that the environment-induced phenotypic changes were stably transmitted over four generations of offspring (Fig. 2a). Therefore, our study demonstrates that environment-induced phenotypic changes are genotype-specific and can be stably transmitted (i.e., are heritable) for at least four offspring generations.\n\nWhile several cases of environment-induced heritable changes have been identified, a fundamental question remains: how prevalent are these changes across different environments? To answer this question, we transformed the quantitative estimates of effect sizes into qualitative assessments, defining an environment-induced change as occurring when the effect size significantly deviated from zero. The occurrence data revealed environment-induced phenotypic changes in each genotype and environment (Fig. 2b). Importantly, a substantial proportion (36.8% 10.4%, and 7.5%) of these changes occurred in two, three, and four offspring generations, suggesting that these environment-induced changes were heritable. Furthermore, we observed heritable changes (changes occurring in at least two generations) in all environments and persistent changes (changes occurring in three generations and remaining significant in the fourth generation) in seven environments (Fig. 2b). Consequently, these findings demonstrate that environment-induced heritable changes commonly occur across various environments.\n\nPredictability and reproducibility of environment-induced phenotypic changes\n\nThe qualitative data revealed that the genotype, ancestral treatment, offspring generation, and phenotype significantly affected the occurrence of environment-induced phenotypic changes (Fig. 3a). The frequency of occurrence of fruit number was lowest among phenotypes and was marginally significantly lower than those of flowering time and plant height. This frequency also varied across different offspring\ngenerations, but no consistent decline was observed over successive generations. In general, the magnitude of the ancestral treatment effect did not predict the frequency of occurrence. For instance, the leaf removal treatment, affecting only fruit number, appeared to be much less stressful than the low-salinity treatment (affecting four phenotypes), but the occurrence frequencies associated with these two treatments were similar (Extended Data Fig. 2a, Fig. 3a). Only within the same treatment type was a higher intensity of treatment associated with an increase in the occurrence frequency of induced phenotypic changes (high vs. low concentrations of cadmium/salinity, Fig. 3a).\n\nAmong the factors affecting the occurrence of environment-induced changes, genotype emerged as the primary explanatory variable. This factor accounted for the largest proportion of the variance in occurrence frequency (Fig. 3a), and removing the three genotypes with the highest frequencies (TRE-1, Abd-0, Sap-0) rendered the effect of ancestral environments on occurrence insignificant (Extended Data Fig. 3a). To explore which factor predicted the susceptibility of a genotype to environmental induction, we regressed the occurrence frequency of genotypes against different explanatory variables (climate of origin, methylation levels, and transposon abundance) and found that the abundance of transposons in the Gypsy (r = -0.840) and MuDR (r = 0.815) superfamilies explained a significant proportion of the variation (Fig. 3b, Extended Data Fig. 3b). Therefore, these findings suggest that genotype may be a primary predictor of the occurrence of environment-induced changes, and this property of a genotype can be predicted by the number of transposons in its genome.\n\nAlthough the occurrence of environment-induced changes is predictable to some extent, predicting the effect sizes of these changes proved challenging. Consistent with a previous study9, the responses of the ancestors showed little, if any, predictability related to the effect sizes of offspring (Fig. 3c). Nevertheless, it was encouraging to find that the effect sizes of offspring could exceed the range of ancestral responses, as observed for flowering time (Fig. 3c). For this trait, the fixed G × A.E effect also explained a greater proportion of the total variation than G × E (Extended Data Fig. 3c). Compared to ancestor-offspring correlations, the coefficients estimated between offspring generations were significantly greater and consistently significantly deviated\nfrom zero, except for fruit number, for which the coefficient reached 0.682 (median = 0.474) (Fig. 3d, Extended Data Fig. 4a-b). The coefficient increased to 0.857 (median = 0.614) when including only significant effect sizes in the analysis and declined substantially and insignificantly deviated from zero for most cases when excluding significant effect sizes from the analysis (Extended Data Fig. 4c). A similar pattern of correlation coefficients was also observed when comparing effect sizes between Test I and Test II (Fig. 3e, Extended Data Fig. 4a, please also see below for Test II). Intriguingly, when we scaled down the experiment to include only one genotype (but multiple treatments) or one treatment (but multiple genotypes), we found that the chances of obtaining significantly correlated effect size correlations between generations declined substantially (Extended Data Fig. 4d). Therefore, our study demonstrates that the effect sizes of environment-induced phenotypic changes are reproducible between offspring generations and tests, and conducting large-scale experiments to obtain abundant significant effect sizes is crucial for achieving this reproducibility.\n\nDespite the findings regarding the predictability and reproducibility of environment-induced changes, some unresolved questions remained: why did some significant phenotypic changes disappear in one generation but reappear in the next (Fig. 2b), and why did the occurrence frequency of induced changes vary between offspring generations (Fig. 3a)? A potential explanation is that the growth conditions differed between offspring generations, as evidenced by significant phenotypic variation between generations in the control group (Extended Data Fig. 2c, Supplementary Table 3). Such variation may also contribute to the inconsistent effect sizes observed between generations (significant G × A.E × F, Supplementary Table 4). To eliminate the influence of variable growth conditions, we conducted a second test of ancestral treatment effects (Test II), where we selected a subset of 10 genotypes and five treatments and grew F1 to F4 plants together under the same conditions (Fig. 1). Consistent with the results of Test I, Test II revealed that the effect of ancestral environments depended on the genotypes (Extended Data Fig. 5a, Supplementary Table 5); that environment-induced heritable changes were common across environments\n(Extended Data Fig. 6); that the occurrence frequency of these changes varied significantly between genotypes (Extended Data Fig. 7a); and that the variation among genotypes could be predicted by the abundance of several transposon superfamilies, including HAT (\\( r = 0.893, P = 0.041 \\)), Mariner (\\( r = 0.873, P = 0.053 \\)), and MuDR (\\( r = 0.846, P = 0.071 \\)) (Extended Data Fig. 7b-c).\n\nSurprisingly, the repeatability of the effect sizes between generations in Test II did not increase compared to that in Test I (Fig. 3d, Extended Data Fig. 7d). We still found significant phenotypic variations between generations in the control group (Extended Data Fig. 5b, Supplementary Table 6), and the genotype-specific ancestral treatment effects varied significantly between generations for flowering time and aboveground biomass (significant G \\( \\times \\) A.E \\( \\times \\) F, Supplementary Table 7). As all plants in Test II were grown under the same conditions, the observed generation effects could only be introduced by different growth conditions of previous generations in Test I. Taken together, these results suggest that both the current conditions and conditions of previous generations can affect the measurements of environment-induced changes, and conducting large-scale experiments under well-controlled conditions is the key to achieving high reproducibility and an in-depth understanding of environment-induced changes.\n\nEnvironment-induced heritable gene expression changes\n\nIn Test II, we conducted transcriptomic analysis of offspring generations (F1-F4) to investigate genome-wide gene expression changes induced by ancestral environments (Fig. 1). The patterns of genome-wide gene expression were primarily determined by genotypes (Extended Data Fig. 8a), with noticeable treatment-specific differentiation observed only in the high-cadmium treatment of Abd-0 and drought treatment of TRE-1 (Extended Data Fig. 8b). The number of differentially expressed genes (DEGs) induced by the ancestral treatment ranged from 146 to 3694 (Fig. 4a), which varied significantly between genotypes but not between environmental treatments or generations (Fig. 4b). Importantly, a significant proportion (15.5%, 3.43%, and 1.54%) of these genes exhibited differential expression in two, three, and four offspring generations (Fig. 4c), indicating that the gene expression changes were\nheritable. We considered those genes that were differentially expressed for at least two generations to be heritable DEGs, and the number of heritable DEGs ranged from 69 to 2668. Importantly, we identified heritable DEGs in each genotype and environment; heritable DEGs that remained differentially expressed over four offspring generations were also found in each environment (Fig. 4c). The number of heritable DEGs for each genotype and treatment was significantly related to the occurrence frequency of induced phenotypic changes (Extended Data Fig. 9a) and was also significantly explained by the number of transposons in several superfamilies (Fig. 4d, Extended Data Fig. 9b). Therefore, these results demonstrate that environment-induced heritable gene expression changes were common across different environments, and the number of genes showing heritable expression changes can be predicted by the abundance of transposons in the genomes.\n\nAn important unanswered question is whether the induced gene expression changes are directional or unidirectional. If these changes are directional, we expect the DEGs induced by different environments to have distinct functions. To test this hypothesis, we conducted GO and KEGG functional enrichment analyses of heritable DEGs and found that 80.2% (73 out of 91) of enriched GO terms and 58% (27 out of 46) of enriched KEGG pathways were shared between treatments (Fig. 5a-b, Extended Data Fig. 9c-d). The overlapping functions appeared to be primarily associated with the stress response (Fig. 5a) and metabolic pathways (Extended Data Fig. 9c). These results indicated that, contrary to our expectations, heritable DEGs induced by different environmental treatments were functionally similar, which may be associated with these treatments (drought and low and high concentrations of cadmium) generating similar stresses (i.e., inducing hypoxia in the root system).\n\nIntriguingly, in addition to being shared between different environmental treatments, a substantial proportion of terms/pathways were also shared between genotypes (Fig. 5b, Extended Data Fig. 9d). The shared terms/pathways likely originated from heritable DEGs that were shared between treatments and/or genotypes (Fig. 5b-c, Extended Data Fig. 9d-e). As the heritable expression changes in these genes were repeatedly induced in two or more treatments or genotypes, these genes could be\nconsidered susceptible to environmental induction. We hypothesize that the susceptibility of these genes is explained by certain genomic features, such as the abundance of transposons in their promoter and gene body regions. To test this hypothesis, we identified all heritable DEGs shared between genotypes or treatments (Extended Data Fig. 10a) and found that the abundance of members of several superfamilies, including Gypsy, HAT, Mariner, and Helitron, upstream and body of these susceptible genes was significantly higher than that of background genes (Extended Data Fig. 10b). Therefore, these results suggest that environment-induced changes may be nonrandom in the sense of genomic features.\n\nTo investigate whether the heritable phenotypic changes were associated with heritable gene expression changes, we focused on the phenotype of flowering time, which has a comparatively well-characterized molecular basis\\(^{14-16}\\). Among the 79 genes underlying flowering time changes identified in the literature, 16 were differentially expressed in at least one genotype or treatment, and seven were heritable DEGs (Extended Data Fig. 10c). The heritable gene expression changes of six genes co-occurred with heritable flowering time changes, including the early flowering of Abd-0 in response to high cadmium and drought-induced late flowering in TRE-1 in response to drought (Fig. 5d, Extended Data Fig. 6). In Abd-0, ancestral treatment with high cadmium caused downregulation of *FLC* and *MAF4* and upregulation of *FT* and *SOC1*, promoting early flowering (Fig. 5d). In TRE-1, ancestral drought treatment caused the upregulation of *MAF3* and *MAF4* and downregulation of *FT* and *SOC1*, which was associated with delayed flowering time (Fig. 5d). *FLC* is the primary determinant of the vernalization requirement and negatively regulates the expression of *FT* \\(^{17,18}\\). *MAFs* are paralogues of *FLC*, which act redundantly with *FLC* as floral repressors\\(^{17,19}\\). Therefore, these results suggest that heritable expression changes in key genes underlie environmentally induced heritable phenotypic changes.\n\n**Discussion**\n\nHere, we conducted a large-scale multigenerational experiment, and the main finding of our study was that environment-induced heritable changes were caused by each ancestral environmental treatment. These results extend the understanding of the\nphenomenon of environment-induced heritable changes beyond occasional findings under specific environmental conditions\\(^{4,5,13}\\) and suggest that this phenomenon could be a fundamental principle of genetics. Our work challenges the textbook understanding that environments are only involved in the selection of heritable variation\\(^{3}\\) and suggests that environments are broadly involved in the generation of heritable variation.\n\nThe scale of our experiment allowed us to explore the stochasticity of environment-induced heritable changes and reveal genotypes with a higher abundance of transposons that are more susceptible to environmental inductions. Furthermore, our study revealed that genes with more transposons are also prone to environmental induction. These results indicate that transposons play a central role in mediating environment-induced changes, which could occur through epigenetic modifications of transposons (i.e., DNA methylation\\(^{20,21}\\) and RNA interference\\(^{22}\\)) that simultaneously regulate the expression of downstream genes or through *de novo* transposon insertions that alter gene expression patterns\\(^{23,24}\\). The rates of *de novo* transposon insertions have been found to increase under both biotic and abiotic stresses\\(^{25}\\), and these rates varied by orders of magnitude between genotypes\\(^{26}\\). A comprehensive analysis that connects environment-induced variations with the mechanisms controlling transposon activity is an essential next step to revolutionize our fundamental understanding of the origins of genetic variation.\n\nThe finding that environment-induced heritable changes are common across environments also indicates that, besides standing genetic variation within populations\\(^{27,28}\\), *de novo* heritable variations caused by environmental changes can serve as another primary source for rapid adaptation. These *de novo* heritable variations may arise from the mechanisms discussed earlier, including epigenetic changes (DNA methylation\\(^{29}\\), histone modifications\\(^{30}\\), and small interfering RNAs\\(^{6}\\)) or *de novo* insertion of transposons\\(^{23,24}\\). Therefore, investigating environment-induced heritable changes not only expands our knowledge in the context of basic research but also contributes to assessing the evolutionary potential and consequences of populations under global environmental changes.\nFig. 1 | Design of the experiment. The experiment subjected 14 genotypes of Arabidopsis thaliana to one control and ten environmental treatments and lasted seven generations. Plants were grown for two generations before launching the “ancestral generation” to eliminate variation due to seed collection or storage. The ten ancestral environmental treatments included jasmonic acid addition (JA), control for JA, nutrient addition, nutrient depletion, leaf removal, drought, low cadmium, high cadmium, low salinity and high salinity. The control and each treatment group consisted of 12 and six replicates, resulting in a total of 1008 lines. To test the ancestral environmental effect on offspring, we established Test I to grow all these lines in the control environment over four successive generations (F1 to F4). To explore the reproducibility and assess the gene expression changes, we established Test II, where we selected ten genotypes and six environments and planted all four offspring generations together in the control environment.\nFig. 2 | Prevalence of environment-induced heritable phenotypic changes. a, The effect sizes of environment-induced phenotypic changes measured in Test I. b, The occurrence of significant and marginally significant environment-induced changes and this occurrence summed over phenotypes. Environment-induced heritable changes (significant changes in at least two generations) marked with dash lines are found in each environment.\nFig. 3 | Predictability and reproducibility of environment-induced phenotypic changes. a, The occurrence frequency of significant environment-induced changes varied between genotypes, ancestral environmental treatments, offspring generations and phenotypes. b, The occurrence frequency of different genotypes could not be predicted from the climate at origin or methylation, but from the abundance of transposons in certain superfamilies. c, The effect sizes of induced changes show limited correlations with ancestor response. d, The effect sizes are quantitatively reproducible between offspring generations. e, The effect sizes are quantitatively reproducible between Test I and Test II. Solid circles represent significant effect sizes in either the generation or test, and transparent circles represent nonsignificant effect sizes in both generations and tests. FT: Flowering time; PH: Plant height; AB: Aboveground biomass; RD: Rosette diameter; FN: Fruit number.\nFig. 4 | Environment-induced heritable gene expression changes are found in each ancestral environmental treatment, and the predictability of environment-induced gene expression changes. **a**, The number of differentially expressed genes (DEGs) for each genotype, ancestral environment treatment and offspring generation. **b**, The number of DEGs varied significantly between different genotypes, but not between treatments and generations. The error bars indicate the standard error. **c**, The number of heritable DEGs (genes significantly differentially expressed in two, three or four offspring generations) for each genotype and treatment. **d**, The number of heritable DEGs for different genotypes could be predicted from the abundance of transposons in several superfamilies.\nFig. 5 | Heritable gene expression changes induced by various environments were functionally similar and explained the induced phenotype changes. a, Gene ontology (GO) enrichment results of heritable differentially expressed genes (DEGs) show that induced gene expression changes are functionally similar between environmental treatments and genotypes. b, Venn plot shows the numbers of shared terms between environmental treatments and genotypes. c, Venn plot shows the number of heritable DEGs within the shared terms. d, The heritable DEGs explained the environment-induced heritable changes in flowering time.\nMethods\n\nPlant materials and the experimental design\n\nArabidopsis thaliana is a model plant species for molecular biology research and is increasingly used to address questions in ecology and evolution31. It is a self-pollinating annual weed that originates from North Africa32 and is broadly distributed across Europe, Asia, and North America. Seeds collected from different geographic sites, known as natural accessions or ecotypes, are preserved at the Arabidopsis Biological Resource Centre (ABRC). As this species is predominantly selfing, different accessions are also taken as different genotypes11,13. Here, we adopted 14 A. thaliana genotypes ordered from the ABRC, including Abd-0 (from Aberdeen, UK), Ang-0 (Namur, Belgium), Col-0 (Columbia, US), Dja-1 (Chui, Kyrgyzstan), En-1 (Frankfurt, Germany), Fei-0 (Aveiro, Portugal), Kar-1 (Suusamyr, Kyrgyzstan), Ler-0 (Munich, Germany), Olympia-2 (Olympia, Greece), Sap-0 (Hokkaido, Japan), Tol-0 (Ohio, USA), TRE-1 (Marne, France), Tri-0 (Sevilla, Spain) and Ws-2 (Belarus, Belarus) (Extended Data Fig. 1a, Supplementary Table 1). Four seeds per genotype were planted in a common environment for two generations to eliminate the variation due to seed collection or storage. In the second generation, seeds of the four resulting plants were pooled for each genotype to establish the ancestral generation.\n\nTo investigate whether environmental changes caused heritable changes in offspring (i.e., transgenerational effects), we subjected the ancestral generation to different environmental treatments and assessed the offspring phenotypic changes over four generations in a common environment. The ancestral environments included the control and ten treatments: jasmonic acid addition (JA), control for JA, nutrient addition, nutrient depletion, leaf removal, drought, low cadmium, high cadmium, low salinity and high salinity (Fig. 1). Each treatment included six replicate plants per genotype, and the control included 12 plants per genotype, resulting in 1008 plants grown or lines established in the ancestral generation. Following the single-seed descendant approach, the offspring of all lines were grown generation-by-generation in the control environment for four generations (F1-F4) (Test I), and altogether, 4032 plants were grown for Test I.\nThe growth room contained four shelves, each with four layers (Extended Data Fig. 1b). In each layer, we organized the plants into six rows and 14 columns, and this organization allowed one replicate per treatment/genotype to be planted in two layers. For one replicate of the ancestral generation, treatments were randomly assigned to rows in two layers, and genotypes were distributed randomly among columns in a row (Extended Data Fig. 1c). For the offspring of this replicate, we assigned them to two random layers and distributed the plants randomly among different positions (Extended Data Fig. 1c). All randomization was conducted using the “RAND” function in Excel.\n\nTo eliminate the potential influence of growth conditions that varied between generations, we planted all four offspring generations together in the control environment (Test II). Due to limited space in the growth room, Test II included a subset of ten genotypes (Abd-0, Col-0, Ang-0, Dja-1, Fei-0, Ler-0, Sapporo-0, Tol-0, TRE-1 and Tri-0) and six different environments (the control, drought, low cadmium, high cadmium, low salinity and high salinity) and a total of 1440 plants. We adopted a completely random design, assigning the 1440 plants randomly into positions among the 16 layers, with six rows and 15 columns each (Extended Data Fig. 1c). Due to the insufficient storage of seeds, especially for the F1 generation, Test II eventually involved 1252 plants. The randomization was conducted using the “RAND” function in Excel.\n\nGrowth conditions, phenotype measurements, and sample collection\n\nThe experiment was conducted by four graduate students who assisted each other (Extended Data Fig. 1d). Prior to sowing, we placed 10-15 seeds in a 2 mL centrifuge tube, added 1 mL of sterilized distilled water, and kept the tubes in a 4 °C freezer in the dark for 7 days. The soaked seeds were then evenly distributed into pots (5.5 × 5.5 × 6.0 cm) filled with a mixture of 1:1 sterilized soil and vermiculite. The plants were subjected to a photoperiod of 16 hours light (7:00 am to 11:00 pm) and 8 hours dark (11:00 pm to 7:00 am), with temperatures ranging between 22 and 24 °C. On the tenth day, one A. thaliana seedling was randomly selected and retained. Lines that failed to germinate were transplanted by seedling of a randomly selected replicate from the same genotype and treatment. For the control environment, we watered the plants from the\nbottom every four days and substituted water with half Hoagland solution every 16 days. For the jasmonic acid treatment, we sprayed the plants with 100 \\( \\mu \\)mol/L jasmonic acid solution (dissolved in 0.5% ethanol) every four days and sprayed 0.5% ethanol solution to plants in the JA control treatment. For the nutrient addition treatment, we watered plants with half Hoagland solution every four days and never gave Hoagland solution to plants in the nutrient depletion treatment. For the leaf removal treatment, we removed half of the total number of leaves 30 days after germination. For the drought treatment, plants were watered every eight days. We dissolved CaCl\\(_2\\) in water or half Hoagland solution to form 500 \\( \\mu \\)mol/L and 1500 \\( \\mu \\)mol/L CdCl\\(_2\\) solutions and watered the plants with these solutions in low- and high-cadmium treatments. For the low- and high-salinity treatments, we watered the plants with 50 mmol/L NaCl solutions. These treatments started 25 days after sowing and lasted for 32 days. Seeds for the next generation were collected 50-60 days after sowing. In the ancestral generation, 26 out of 1008 (2.58%) lines died before seed production, and their offspring generation was replaced with seeds of a randomly selected replicate from the same genotype and treatment.\n\nWe collected phenotypic data from the ancestral and offspring generations in Test I and Test II. Flowering time was checked daily and recorded as the days from sowing to the opening of the first flower. Fruit number was the number of siliques for a plant, used as a proxy for plant fitness. Rosette diameter was measured as the length of the most extended rosette cross centre. Plant height was the height of a plant when pulled straight. Aboveground biomass was the weight of the aboveground parts after drying in an oven at 80 \\( ^\\circ \\)C for 48 h.\n\n**Statistical analysis of phenotypic data**\n\nTo assess the effects of the environmental treatment and genotype on the ancestral generation, we applied a linear mixed-effect model to each phenotype (flowering time, plant height, aboveground biomass, rosette leaf diameter and fruit number). The model included genotype (G), environment (E), and their interaction (G \\( \\times \\) E) as fixed effects and the layer (A.Layer) and row (A.Row) as random effects, which was coded as \\( Y_{Anc} \\sim G + E + G:E + (1 \\mid A.\\text{Layer}) + (1 \\mid A.\\text{Row}) \\) (1). Note that the row number was unique\nand differed between layers.\n\nTo assess the effects of ancestral treatment and genotype on offspring generations, we applied a linear mixed-effect model to each phenotype measured in Tests I and II. The model included offspring layer (O. Layer), genotype (G), ancestral environment (A.E), and their interaction (G × A.E) as fixed effects and ancestor layer (A. Layer) and row (A. Row) as random effects and was coded as \\( Y_{\\text{Off}} \\sim O.\\text{Layer} + G + A.E + G:A.E + (1 | A.\\text{Layer}) + (1 | A.\\text{Row}) \\) (2). We also integrated all four offspring generations in a model. The model included offspring layer (O. Layer), genotype (G), ancestral environment (A.E), and offspring generation (F) and their two and three interactions as fixed effects and line (Line), ancestral layer (A.Layer) and row (A.Row) as random effects and was coded as \\( Y_{\\text{Off}} \\sim O.\\text{Layer} + G + A.E + F + G:A.E + G:F + A.E: F + G:A.E:F + (1 | \\text{Line}) + (1 | A.\\text{Layer}) + (1 | A.\\text{Row}) \\) (3). In Test I, generation was nested in the offspring layer, removing the effect of which also removed that of generation, and Model (3) ~.- F was applied to Test I. In addition, we applied a generalized linear mixed-effect model to the subsets of data, including only the offspring of the control environment. This model included the genotype (G) and offspring generation (F) and their interaction as fixed effects and was coded as \\( Y \\sim G + F + G:F+ (1 | A.\\text{Layer}) + (1 | A.\\text{Row}) \\) (4).\n\nThese models were formulated with function *lmer* in the R ver.3.6.2 package “lme4”\\( ^{33} \\). Statistics of fixed effects were estimated with the *Anova* function in the “car” package\\( ^{34} \\), and those of random effects were estimated with the *ranova* function in the “lmerTest” package\\( ^{35} \\). The effect sizes of ancestral responses to environmental treatments for different genotypes were extracted from model (1), and those of induced changes in offspring generations for different genotypes were extracted from model (2) with the *summary* function following Groot et al\\( ^{13} \\). For the JA treatment, the effect sizes were estimated by applying the models to a subset of data consisting of JA and JA control, and for other treatments, the effect sizes were estimated by applying the models to the respective treatment and the control. The significance of effect sizes was obtained with the *summary* function in the “lmerTest” package. According to these significances, we converted the effect sizes to qualitative occurrence data, where significant effect\nsizes were assigned a value of one, and nonsignificant effect sizes were assigned a value of zero.\n\nTo estimate the variance explained by different variables in mixed-effects models (1) and (2), we converted these models into full random effect models and applied the function VarCorr (package “lme4”) to the models. Model (1) was converted to \\( Y_{Anc} \\sim 1 + (1 | G) + (1 | E) + (1 | G \\times E) + (1 | A.Layer) + (1 | A.Row) \\), and Model (2) was converted to \\( Y_{Off} \\sim 1 + O.Layer + (1 | G) + (1 | A.E) + (1 | G \\times A.E) + (1 | A.Layer) + (1 | A.Row) \\). The estimated variance was visualized with the function geom_scatterpie in the R package “scatterpie” (available at https://github.com/GuangchuangYu/scatterpie).\n\nPredictability and reproducibility analysis\n\nTo estimate whether the genotype, treatment, generation, and phenotype significantly affected the occurrence of transgenerational effects, we applied generalized linear mixed-effect models to the qualitative occurrence data of Test I and Test II. The model included the genotype, treatment, generation and trait as fixed effects and an additional factor group as a random effect. As effect sizes of different traits were estimated from the same group of six replicate plants per genotype/treatment/generation, we defined the six replicates as one group, which was assigned a group ID. The model was formulated with function glmer in the R ver.3.6.2 package “lme4” and was coded as \\( Y \\sim G + A.E + F + Trait + (1 | group.id) \\) (5). Statistics of fixed effects were estimated with the Anova function in the “car” package. Furthermore, we used the glht function in the “multcomp” package\\(^{36}\\) to further investigate the differences in occurrence frequency among different ancestral treatments, generations, and phenotypes.\n\nTo explore which factors could predict the genotypic variation in the occurrence of transgenerational effects, we related the occurrence probability of genotypes to climates at the origin sites, the DNA methylation level, and the number of transposons in the genomes. We extracted climate data (Bio1-Bio19) from WorldClim (https://worldclim.org), including monthly climate data between 1901 and 2000. The data were extracted with the coordinates of genotypes from a 30-second spatial resolution (~1 km2) by the R package “raster” (Supplementary Table 1). The raw\nmethylation data of 14 genotypes were obtained from published sources37. Quality control was performed using Trimmomatic v0.3638 with the following parameters: ILLUMINA_CLIP: TruSeq3-PE.fa:2:30:10\" LEADING:20 TRAILING:20 SLIDINGWINDOW:4:20 MINLEN:50. We aligned the reads against the TAIR10 reference with bwa-meth39 with the default parameters. The aligned reads with mapping quality higher than 30 were obtained with SAMtools v 1.740 with the -q 30 option, and duplicates were marked and removed by Picard v2.18.10 (http://broadinstitute.github.io/picard) and SAMtools (-F 1024). The methylation level of each cytosine for different contexts (CG, CHG, and CHH) was extracted with MethylDackel (https://github.com/dpryan79/MethylDackel). We excluded cytosines with a read depth lower than 5, and the average methylation level per cytosine was estimated for each context (Supplementary Table 1). Furthermore, we extracted the number of transposable elements from published sources25. The original file contained copy number estimation based on read coverage for the 317 TE families analysed across 211 A. thaliana accessions. We summarized the total number of TEs according to superfamilies and then used log-transformed counts for further analysis. These data were available for seven genotypes, including Abd-0, Ang-0, Col-0, Dja-1, Sap-0, Tol-0, and Ws-2 (Supplementary Table 1). The transposon data of Col-0 were not included in the subsequent analysis because Col-0 is the reference genome and shows biased (much more detailed) annotation of transposons compared to other genotypes. We applied arcsine square root transformation to the occurrence frequency and log transformation to the number of transposons in different superfamilies, regressed the transformed occurrence frequency per genotype against the climate data, average methylation levels and transformed transposon number, and extracted the correlation coefficient and \\( P \\) value from the regression.\n\nTo investigate whether the effect size of environment-induced changes can be quantitatively predicted from the ancestor’s responses, we calculated the pairwise Pearson correlation coefficients between the ancestor’s responses and the effect sizes of different offspring generations. To assess the reproducibility of effect sizes between offspring generations and tests, we also computed pairwise Pearson correlation\ncoefficients between offspring generations for each test and between the tests. These correlation coefficients were estimated for subsets of data, including significant effects in either generation only or excluding these significant effects. In addition, for Test I, we used subsets of data, including the effect sizes of only one genotype or only one treatment, to calculate correlation coefficients between generations. Visualization was conducted using the ggplot function in the R package “ggplot2”.\n\nGenome-wide gene expression analyses\n\nTo assess patterns of gene expression changes caused by ancestral treatments, we collected fresh leaf samples from plants in Test II and conducted transcriptomic analysis. This analysis included four genotypes (Abd-0, Ang-0, Tol-0, TRE-1), four environments (the control, drought, low cadmium, and high cadmium), four generations (F1-F4), each with three replicates, and a total of 144 samples. These genotypes and environments were carefully selected to include genotypes and treatments representing a high/low occurrence probability of transgenerational effects. For instance, Abd-0 and TRE-1 represent genotypes sensitive to environmental induction, whereas Ang-0 and Tol-0 represent genotypes resistant to environmental induction. We sampled two to three rosette leaves per plant, immediately froze the samples in liquid nitrogen, and stored them at −80 °C before RNA extraction. We extracted total RNA with TRIzol reagent (Invitrogen)41. The samples of total RNA were subjected to quality assessment with an Agilent 2100 Bioanalyzer. The transcriptome sequencing library was prepared with the NEBNext® UltraTM RNA Library Prep Kit for Illumina® (NEB, USA). The library was sequenced with the Illumina NovaSeq platform to produce 150 bp paired-end reads. Quantity assessment, library preparation and sequencing were conducted by Novogene (Beijing, China). Sequencing yielded 19.5 to 29.4 million pairs of reads (average: 22.6) per sample (Supplementary Table 8).\n\nTo improve the quality of reads, we trimmed and filtered the reads with Trimmomatic v0.36 with the following parameters: “ILLUMINA_CLIP: TruSeq3-PE.fa:2:30:10” LEADING:20 TRAILING:20 SLIDINGWINDOW:4:20 MINLEN:50. Reads with improved quality were aligned to the TAIR10 reference genome using STAR v 2.7.942 with the default parameters and a TAIR10 gff-converted gtf-format file\n(parameter ‘-sjdbGTFfile’) (Supplementary Table 8). The conversion was conducted with gffread v.0.12.743. We extracted the gene expression matrix using featureCounts v1.22.2 44, with the ‘-p’ option to count the meta-features of genes using the TAIR10 gtf file. Genes were included in subsequent analysis if the sum CPM (counts per million) over all samples for the treatment and control per genotype was greater than one. The retained gene and the expression were subjected to cluster analysis of different samples with the plotPCA function in the DESeq2 v1.36.0 package45, and cluster analysis results were visualized using the R package “ggplot2”. We subjected the retained genes and their expression to differential expression analysis with the R package DESeq2 v1.36.0 (these genes were also used for random sampling, please see below). For each genotype, the difference in gene expression was estimated by comparing the expression of a particular gene between the ancestral treatment and the control. The expression matrix was normalized by function vst in package “DESeq2”. The difference in gene expression was estimated with function DESeq, and the differentially expressed genes (DEGs) were filtered by | log2FoldChange | >1.5 and adjusted \\( P < 0.05 \\). A gene was considered to be a heritable DEG if it was significantly differentially expressed for at least two generations. To explore the factors that explained the number of heritable DEGs, we regressed the log-transformed number per genotype and treatment against the arcsine-transformed occurrence frequency per genotype and treatment as well as the log-transformed number of transposons per genotype and extracted the correlation coefficient and \\( P \\) value from the regression.\n\nTo investigate the function of DEGs, we focused on genes that were differentially expressed in at least two offspring generations (heritable DEGs) and conducted KEGG and Gene Ontology (GO) enrichment analysis. The enrichment analysis was conducted with functions enrichKEGG and enrichGO in the “clusterProfiler v4.4.4” package46. Enrichment analysis was conducted separately for four genotypes and three environments, which resulted in 12 sets of enrichment results (Supplementary Table 9). For the KEGG results, we took the union of significantly enriched pathways of the 12 sets, clustered the pathways according to the maps (https://www.genome.jp/kegg/pathway.html) and visualized the results. For the GO\nresults, we removed duplicated terms for each set with function simplify in the package (Supplementary Tables 10). Then, we filtered out terms with less than five genes enriched and selected the top 15 terms with the smallest \\( P \\) values for each set, took their union, and visualized the results. KEGG and GO visualization was conducted with the R package “pheatmap v1.0.12” (https://github.com/raivokolde/pheatmap). To facilitate visualization, we characterized the terms into functions of response to stimulus, biosynthetic and metabolic processes, cellular process, signalling, biological regulation, rhythmic process, and developmental process. Then, we extracted the mapped heritable DEGs in pathways and terms and summarized their overlapping patterns between genotypes and environments.\n\nWe found 2669 heritable DEGs shared between genotypes or environments, which are considered genes sensitive to environmental induction. We hypothesized that their sensitivity was related to the enrichment of transposons in their regulatory and gene body regions. To test this hypothesis, we randomly sampled 2669 genes from the genes subjected to the differential expression analysis and estimated the sum of the number of transposons per superfamily overlapping the 2 kb upstream and gene body regions. Through random sampling 10000 times, we obtained the null distribution of transposon abundance upstream of genes for different superfamilies. Then, we compared the observed transposon abundance upstream of the 2669 heritable DEGs against this distribution. The \\( P \\) value was estimated as the proportion of randomly sampled values greater than the observed value, and significant \\( P \\) values indicate significant enrichment of transposons upstream of heritable DEGs compared with the null expectation. We adopted the transposon annotation of TAIR10 and performed the analysis using a customized script coded with R package “GenomicRanges” v.1.42.0\\(^{47}\\) (available at https://github.com/Xiaohe-Lin/Heritable_variation/). Notably, our study adopted Col-0 as the reference, and genotype-specific genome assembly is necessary for a comprehensive understanding of this nonrandom phenomenon.\n\nTo investigate whether the heritable phenotypic changes were caused by heritable gene expression changes, we focused on flowering time and identified 79 flowering time-related genes from the literature\\(^{14-16}\\) (Supplementary Table 11). We visualized the\nexpression patterns of a subset of genes that were DEGs for at least one generation using the R package “pheatmap”.\n\nData availability\n The phenotypic data of the ancestral generation, Test I and Test II are available on GitHub (https://github.com/Xiaohe-Lin/Heritable_variation). Raw RNA-seq data are available from the NCBI SRA database with accession ID PRJNA997595.\n\nCode availability\n Codes for data processing are available on GitHub (https://github.com/Xiaohe-Lin/Heritable_variation).\nReferences\n\n1 Hoffmann, A. A. & Sgrò, C. M. Climate change and evolutionary adaptation. Nature 470, 479-485 (2011).\n2 Scheffers, B. R. et al. The broad footprint of climate change from genes to biomes to people. Science 354, aaf7671 (2016).\n3 Futuyma, D. J. & Kirkpatrick, M. Evolution (Oxford University Press, 2017).\n4 Agrawal, A. A., Laforsch, C. & Tollrian, R. Transgenerational induction of defences in animals and plants. Nature 401, 60-63 (1999).\n5 Klosin, A., Casas, E., Hidalgo-Carcedo, C., Vavouri, T. & Lehner, B. Transgenerational transmission of environmental information in C. elegans. Science 356, 320-323 (2017).\n6 Rechavi, O. et al. Starvation-induced transgenerational inheritance of small RNAs in C.elegans. Cell 158, 277-287 (2014).\n7 Yin, J., Zhou, M., Lin, Z., Li, Q. Q. & Zhang, Y.-Y. Transgenerational effects benefit offspring across diverse environments: a meta-analysis in plants and animals. Ecol. Lett. 22, 1976-1986 (2019).\n8 Galloway, L. F. Parental environmental effects on life history in the herbaceous plant Campanula americana. Ecology 82, 2781-2789 (2001).\n9 Latzel, V. et al. Parental environmental effects are common and strong, but unpredictable, in Arabidopsis thaliana. New Phytol. 237, 1014-1023 (2023).\n10 Colicchio, J. Transgenerational effects alter plant defence and resistance in nature. J. Evol. Biol. 30, 664-680 (2017).\n11 Alvarez, M., Bleich, A. & Donohue, K. Genotypic variation in the persistence of transgenerational responses to seasonal cues. Evolution 74, 2265-2280 (2020).\n12 Munch, S. B. et al. A latitudinal gradient in thermal transgenerational plasticity and a test of theory. Proc. Royal Soc. B 288, 20210797 (2021).\n13 Groot, M. P. et al. Transgenerational effects of mild heat in Arabidopsis thaliana show strong genotype specificity that is explained by climate at origin. New Phytol. 215, 1221-1234 (2017).\nHe, L. et al. DNA methylation-free Arabidopsis reveals crucial roles of DNA methylation in regulating gene expression and development. Nat. Commun. 13, 1335 (2022).\n\nGaudinier, A. & Blackman, B. K. Evolutionary processes from the perspective of flowering time diversity. New Phytol. 225, 1883-1898 (2020).\n\nRoux, F., Touzet, P., Cuguen, J. & Le Corre, V. How to be early flowering: an evolutionary perspective. Trends Plant Sci. 11, 375-381 (2006).\n\nKim, D. H., Doyle, M. R., Sung, S. & Amasino, R. M. Vernalization: winter and the timing of flowering in plants. Annu. Rev. Cell Dev. Biol. 25, 277-299 (2009).\n\nSearle, I. et al. The transcription factor FLC confers a flowering response to vernalization by repressing meristem competence and systemic signaling in Arabidopsis. Genes Dev. 20, 898-912 (2006).\n\nRatcliffe, O. J., Nadzan, G. C., Reuber, T. L. & Riechmann, J. L. Regulation of flowering in Arabidopsis by an FLC homologue. Plant Physiol. 126, 122-132 (2001).\n\nMorgan, H. D., Sutherland, H. G. E., Martin, D. I. K. & Whitelaw, E. Epigenetic inheritance at the agouti locus in the mouse. Nat. Genet. 23, 314-318 (1999).\n\nOng-Abdullah, M. et al. Loss of Karma transposon methylation underlies the mantled somaclonal variant of oil palm. Nature 525, 533-537 (2015).\n\nIto, H. et al. An siRNA pathway prevents transgenerational retrotransposition in plants subjected to stress. Nature 472, 115-119 (2011).\n\nHof, A. E. v. t. et al. The industrial melanism mutation in British peppered moths is a transposable element. Nature 534, 102-105 (2016).\n\nSchmidt, J. M. et al. Copy number variation and transposable elements feature in recent, ongoing adaptation at the Cyp6g1 locus. PLoS Genet. 6, e1000998 (2010).\n\nBaduel, P. et al. Genetic and environmental modulation of transposition shapes the evolutionary potential of Arabidopsis thaliana. Genome Biol. 22, 138 (2021).\n\nHo, E. K. H. et al. Engines of change: Transposable element mutation rates are high and variable within Daphnia magna. PLoS Genet. 17, e1009827 (2021).\nBarrett, R. D. H. & Schluter, D. Adaptation from standing genetic variation. Trends Ecol. Evol. **23**, 38-44 (2008).\n\nBitter, M. C., Kapsenberg, L., Gattuso, J. P. & Pfister, C. A. Standing genetic variation fuels rapid adaptation to ocean acidification. Nat. Commun. **10**, 5821 (2019).\n\nWeaver, I. C. G. *et al.* Epigenetic programming by maternal behavior. Nat. Neurosci. **7**, 847-854 (2004).\n\nWan, Q.-L. *et al.* Histone H3K4me3 modification is a transgenerational epigenetic signal for lipid metabolism in *Caenorhabditis elegans*. Nat. Commun. **13**, 768 (2022).\n\nPigliucci, M. Ecology and evolutionary biology of *Arabidopsis*. *Arabidopsis Book* **1**, e0003 (2002).\n\nDurvasula, A. *et al.* African genomes illuminate the early history and transition to selfing in *Arabidopsis thaliana*. Proc. Natl. Acad. Sci. U.S.A. **114**, 5213-5218 (2017).\n\nBates, D., Mächler, M., Bolker, B. & Walker, S. Fitting linear mixed-effects models using lme4. *J. Stat. Softw.* **67**, 1-48 (2015).\n\nFox, J. & Weisberg, S. *An R Companion to Applied Regression*. (SAGE Publications, 2019).\n\nKuznetsova, A., Brockhoff, P. B. & Christensen, R. H. B. lmerTest package: Tests in linear mixed effects models. *J. Stat. Softw.* **82**, 1-26 (2017).\n\nHothorn, T., Bretz, F. & Westfall, P. Simultaneous inference in general parametric models. *Biom. J.* **50**, 346-363 (2008).\n\nKawakatsu, T. *et al.* Epigenomic diversity in a global collection of *Arabidopsis thaliana* accessions. *Cell* **166**, 492-505 (2016).\n\nBolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence data. *Bioinformatics* **30**, 2114-2120 (2014).\n\nPedersen, B. S., Eyring, K. R., De, S., Yang, I. V. & Schwartz, D. A. Fast and accurate alignment of long bisulfite-seq reads. Preprint at *arXiv* https://doi.org/10.48550/arXiv.1401.1129 (2014).\nLi, H. et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics **25**, 2078-2079 (2009).\n\nRio, D. C., Ares, M., Jr., Hannon, G. J. & Nilsen, T. W. Purification of RNA using TRIzol (TRI reagent). Cold Spring Harb. Protoc. **2010**, pdb.prot5439 (2010).\n\nDobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics **29**, 15-21 (2013).\n\nPertea, G. & Pertea, M. GFF Utilities: GffRead and GffCompare. *F1000Res* **9**, 304-304 (2020).\n\nLiao, Y., Smyth, G. K. & Shi, W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics **30**, 923-930 (2014).\n\nLove, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. **15**, 550 (2014).\n\nYu, G. C., Wang, L. G., Han, Y. Y. & He, Q. Y. clusterProfiler: an R package for comparing biological themes among gene clusters. Omics **16**, 284-287 (2012).\n\nLawrence, M. et al. Software for computing and annotating genomic ranges. PLoS Comput. Biol. **9**, e1003118 (2013).\nAcknowledgements: We thank Wenjing Hong, Gang Wang, and Rong Wang for their constructive suggestions. We thank Ming Zhou, Xianwen Ma, Jiayi Ding, and Xinwei Wang for their help in the growth room and Xiuxiu Wang and Haidong Qu for their technical assistance.\n\nAuthor contributions: Y-Y.Z., O.B. and V.L. conceived the idea of this project; Y-Y.Z. and JJ.Y. designed the experiments; Y-Y.Z. supervised and Q.Q.L. co-supervised the project; JJ.Y., Y.W., J. Y, and X. L carried out the experiments; JJ.Y., Y.W., and X. L analysed the data; X. L wrote the first draft of this manuscript, and all authors contributed to the revision of the manuscript.\n\nFunding: This study was supported by the National Natural Science Foundation of China (32371558, 32071485) and the Fundamental Research Funds for the Central Universities of China (20720210075).\n\nCompeting interests: The authors declare no competing interests.\nSupplementary Files\n\nThis is a list of supplementary files associated with this preprint. Click to download.\n\n• supportingdata.xlsx\n• Extendeddatafigures.pdf"}
{"paper_id": "prb7", "title": "Measurement of the 2S1/2,F=0 ↔ 2P1/2,F=1 transition in Muonium", "text": "Measurement of the 2S1/2,F=0 ↔ 2P1/2,F=1 transition in Muonium\n\nGianluca Janka\nETH Zurich https://orcid.org/0000-0002-9147-6885\n\nBen Ohayon\nETH Zurich\n\nIrene Cortinovis\nETH Zurich\n\nZak Burkley\nETH Zurich\n\nLucas de Sousa Borges\nETH Zurich\n\nEmilio Depero\nETH Zurich\n\nArtem Golovizin\nP.N. Lebedev Physical Institute https://orcid.org/0000-0002-2747-1297\n\nXiaojie Ni\nPaul Scherrer Institute\n\nZaher Salman\nPaul Scherrer Institute https://orcid.org/0000-0002-3431-8135\n\nAndreas Suter\nPaul Scherrer Institut https://orcid.org/0000-0002-9214-6474\n\nThomas Prokscha\nPaul Scherrer Institute https://orcid.org/0000-0002-7643-4695\n\nPaolo Crivelli (paolo.crivelli@cem.ch)\nInstitute for Particle Physics, ETH Zurich https://orcid.org/0000-0001-5430-9394\n\nArticle\n\nKeywords:\n\nPosted Date: June 22nd, 2022\n\nDOI: https://doi.org/10.21203/rs.3.rs-1718825/v1\nLicense: This work is licensed under a Creative Commons Attribution 4.0 International License.\nRead Full License\n\nAdditional Declarations: There is NO Competing Interest.\n\nVersion of Record: A version of this preprint was published at Nature Communications on November 25th, 2022. See the published version at https://doi.org/10.1038/s41467-022-34672-0.\nMeasurement of the \\(2S_{1/2}, F=0 \\rightarrow 2P_{1/2}, F=1\\) transition in Muonium\n\nG. Janka,\\textsuperscript{1} B. Ohayon,\\textsuperscript{1} I. Cortinovis,\\textsuperscript{1} Z. Burkley,\\textsuperscript{1} L. de Sousa Borges,\\textsuperscript{1} E. Depero,\\textsuperscript{1} A. Golovizin,\\textsuperscript{2} X. Ni,\\textsuperscript{3} Z. Salman,\\textsuperscript{3} A. Suter,\\textsuperscript{3} T. Prokscha,\\textsuperscript{3} and P. Crivelli\\textsuperscript{1,*}\n(Mu-MASS collaboration)\n\n\\textsuperscript{1}Institute for Particle Physics and Astrophysics, ETH Zürich, CH-8093 Zürich, Switzerland\n\\textsuperscript{2}P.N. Lebedev Physical Institute, 53 Leninsky prospect., Moscow 119991, Russia\n\\textsuperscript{3}Laboratory for Muon Spin Spectroscopy, Paul Scherrer Institute, CH-5232 Villigen PSI, Switzerland\n(Dated: June 2, 2022)\n\nMuons are puzzling physicists since their discovery when they were first thought to be the meson predicted by Yukawa to mediate the strong force. The recent results at Fermilab on the muon g-2 anomaly puts the muonic sector once more under the spotlight and calls for new measurements with this fascinating particle. Here we present the results of the first measurement of the \\(2S_{1/2}, F=0 \\rightarrow 2P_{1/2}, F=1\\) transition in Muonium, the hydrogen-like bound state of a positive muon and an electron. The measured value of 580.6(68) MHz is in agreement with the theoretical calculations. From this measurement a value of the Lamb shift of 1045.5(68) MHz is extracted, compatible with previous experiments. We also determine for the first time the \\(2S\\) hyperfine splitting in Muonium to be 559.6(72) MHz. The measured transition being isolated from the other hyperfine levels holds the promise to provide an improved determination of the Muonium Lamb shift at a level where bound state QED recoil corrections not accessible in hydrogen could be tested. Such a measurement will also be sensitive to new physics in the muonic sector, e.g. to new bosons which might provide an explanation of the g-2 muon anomaly or Lorentz and CPT violation. We also present the first observation of Muonium in the \\(n = 3\\) excited state opening up the possibility of new precise microwave measurements as realized in hydrogen.\n\nI. INTRODUCTION\n\nThe discovery of muons has an intriguing history. As in the case of the positron, they were first detected in cosmic radiation by Anderson and Neddermayer in 1936 [1]. Interestingly, the first hints of muons were actually already seen in 1933 by Kunze [2] in his Wilson chamber, but he was not confident enough about his results to claim the discovery of a new particle.\n\nThe muon was first thought to be the pion predicted by Yukawa (1935) [3], the heavy quantum (meson) responsible of mediating the nuclear (strong) force in analogy to the light quantum (photon) for the electromagnetic interaction. The pion based on the range of the nuclear force should have had a mass of 100 to 200 times the mass of the electron and should be both positively and negatively charged. The muon (207 times the electron mass) seemed to have just the expected value. However, in 1946 an experiment of Conversi et al. [4] showed that their interaction with nuclei was too weak to be attributed to pions. They observed that the negative mesotron would decay instead of being absorbed by carbon after having formed a pionic atom as predicted by Tomonaga-Araki [5]. This was the first indication of the formation of muonic atoms as it was realized a few years later. Moreover, the lifetime of the muon was of the order of \\(10^{-6}\\) s which is \\(10^{12}\\) times longer than expected by strong interaction processes. Finally in 1947, Powell et al. [6] detected the pion using photographic emulsions at high altitudes and verified that it decays into a muon and a neutral particle that was identified to be a neutrino. The possibility that this could be a photon was eliminated by the non-observation of pair production as expected in this case. This led Isidor Rabi to come up with his famous quote: “Who ordered that?”.\n\nThe fascinating history of the muon continues to this day. The recent results at Fermilab [7] confirming that the measured muon anomalous magnetic moment (g-2 muon) deviates from the standard model prediction by 4.2 standard deviations calls for further scrutiny. Muonium (M), the bound state of the positive muon (\\(\\mu^+\\)) and an electron is an ideal system to study the muon properties and hunt for possible new effects. Due to its lack of sub-structure it is free from finite size effects and is therefore an excellent candidate to probe bound-state QED [8] and search for new physics beyond the Standard Model [9–11].\n\nPrecise measurements of the ground-state hyperfine structure (HFS) [12] and the \\(1S - 2S\\) transition [13] in Muonium were performed, with improvements proposed by MuSEUM (HFS, [14]) and Mu-MASS (\\(1S - 2S\\), [15]) ongoing. All the measurements so far of the M Lamb shift (LS) are limited by statistics [16–18]. Only recently, the formation of an intense metastable M(2S) beam was demonstrated by the Mu-MASS collaboration at the low-energy muon beamline (LEM) at PSI [19], opening up the possibility for a new generation of precision measurements of the M LS.\n\nA first measurement at the LEM beamline resulted in a M LS value of 1047.2(25) MHz [20], representing an order of magnitude improvement upon the previous measurements. With the muCool beamline [21] and (if approved)\nthe high intensity muon beam (HiMB) upgrade at PSI [22], the \\( \\mu^+ \\) beam quality and flux will further improve, making it feasible to reach statistical uncertainties of the order of a few tens of kHz within a few days of beamtime. This would allow to probe QED corrections enhanced in the Muonium system such as the Barker-Glover effect (160\\(\\mathrm{kHz}\\)) or the nucleus self-energy (40\\(\\mathrm{kHz}\\)), which are still not in reach for hydrogen [23].\n\nThe prospect of not being limited by statistics calls for a systematically more robust method to extract the LS. Following the example of the most precise measurements of the hydrogen Lamb shift [24–26], extracting the LS from the isolated hyperfine transition \\(2S_{1/2}, F=0 \\rightarrow 2P_{1/2}, F=1\\) is preferred over the other two allowed transitions (see Fig. 1). Due to the reduced influence of other nearby transitions, the systematic uncertainties related to line-pulling would become negligible.\n\nWe report here the first measurement of the isolated hyperfine transition, from which we extract the M LS. Combining this result with our previous measurement [20], we determine the 2S hyperfine splitting in Muonium for the first time. Additionally, we present the first observation of Muonium in the \\(n = 3\\) excited state.\n\nII. EXPERIMENTAL METHOD\n\nThe LEM beamline at PSI provides low (1-30 keV) energy muons with a rate of up to 20 kHz using a solid neon moderator [27]. In our experiment, we set the energy of the \\( \\mu^+ \\) to 7.5 keV in order to maximize the number of M(2S) available for the measurement [19]. The beam transport is optimized with several lenses along the beamline, eventually focusing the muons onto a carbon foil at the entrance of the setup shown in Fig. 2.\n\nIn contrast to the LS measurements with protons, the flux of muons is around ten orders of magnitude smaller, which makes it essential to have an efficient background suppression without sacrificing efficiency. Therefore, the incoming muons are tagged on an event by event basis to be able to discriminate their times-of-flight (TOFs). When impinging onto the 10 nm thin carbon foil, a \\( \\mu^+ \\) releases secondary electrons. These electrons are detected by a microchannel plate (Tag MCP) to give the start signal of the TOF measurement. Upon leaving the foil, the tagged \\( \\mu^+ \\) has a probability of picking up an electron and form M, primarily in the ground state. From measurements with hydrogen [28, 29], 5% to 10% is expected to be formed in \\( n = 2 \\) and 2% in \\( n = 3 \\). This expectation was confirmed with a measurement of 11(4)% [19] for M in the metastable state. The M(2S) lifetime is limited by the decay time of the muon itself (\\( \\tau_{M(2S)} = 2.2 \\) \\(\\mu\\)s). M(3S) is unstable (\\( \\tau_{M(3S)} = 158 \\) ns) and will relax back to the ground state via the intermediate 2P state (\\( \\tau_{M(2P)} = 1.6 \\) ns).\n\nThe formed beam passes first through a transmission line (TL) which depopulates unwanted hyperfine states (HFS TL). As established by the most precise H LS measurements, this reduces the background and at the same time narrows and simplifies the overall line-shape to be fitted. By setting the HFS TL at a frequency of 1140 MHz and with an average power of roughly 29 W, according to our simulation, we depopulate the \\( F=1, m_F = \\pm 1 \\) states by 88% and the \\( F=1, m_F = 0 \\) state by 16%. The transition of interest \\( 2S_{1/2}, F=0 \\rightarrow 2P_{1/2}, F=1 \\) at 583 MHz remains almost unaffected, where its initial state \\( F=0, m_F = 0 \\) is only depopulated by 4%. With a second TL (Scanner TL) we scan the line-shape of this transition. Seven frequency points are measured in the range of 200 MHz to 800 MHz with an average power of 22.3 W. An additional data point is taken with the Scanner TL turned off. The output powers of both TLs are continuously monitored with power meters to ensure the stability of the measurement.\n\nDue to the short lifetime of the 2P states, the atoms excited by the microwave relax to the ground state before reaching the detection setup. The remaining excited states are quenched by applying an electrical field of the order of 250 V cm\\(^{-1}\\) between the two grids mounted in front of the coated Ly\\(\\alpha\\)-MCPs (see Fig.2) and relax back to the ground state with the emission of UV photons (2P Ly\\(\\alpha\\): 121.5 nm, 3P Ly\\(\\beta\\): 102.5 nm). These photons interact with the coating and release single electrons, which are in turn detected by the Ly\\(\\alpha\\)-MCPs. The beam then continues onto the Stop-MCP, giving the stop time for the TOF measurement. An event is recorded only if a coincidence signal between the Tag- and Stop-MCP is seen above threshold (double coincidence). The time window for a valid event is set to 10 \\(\\mu\\)s and multiple hits in the detectors are recorded. The signature for the detection of M 2S atoms is defined as the double coincidence and additionally the Ly\\(\\alpha\\)-MCP signal in the time region of interest (triple coincidence, see also Fig. 2).\n\nIn Fig. 3, the measured TOFs between Tag- and Stop-MCP (x-axis) are correlated with the TOFs between Tag- and Ly\\(\\alpha\\)-MCP (y-axis). The energy distribution of the M atoms after the carbon foil is well-known from previous measurements at the LEM beamline and reproduced by simulation [30]. The most-probable energy of the \\( \\mu^+ \\) after the foil is 5.7(2) keV. The expected TOF of an M atom from the Tag- to the Stop-MCP ranges from 90 ns to 135 ns, with a most-probable TOF of\n\n\n\nFIG. 1. The hyperfine transitions of the M 2S_{1/2} – 2P_{1/2} levels, including the 2S and 2P hyperfine splittings.\nFIG. 2. Schematic view of the experimental setup. The signal signature consists of a coincidence signal between Tag, Lyα and Stop-MCP. The coloured line-shapes above the TLs correspond to the observable transitions (1326 MHz (green), 1140 MHz (orange), 583 MHz (blue) and the combined \\(3S_{1/2} - 3P_{1/2}\\) (yellow) contribution). The time scale is given for the most-probable Muonium atom with an energy of 5.7 keV.\n\n117 ns. Applying this time cut, we extract the detected amount of M and \\( \\mu^+ \\), which serves as normalization factor \\( S_{Norm} = N_M + N_{\\mu^+} \\), accounting for variations in the beam intensity. The TOF between the tagging and the emission of a Lyα photon is calculated to be in the range of 30 ns to 75 ns. Applying both of these cuts, we identify our signal region, portrayed in Fig. 3. The amount of signal events are extracted for each frequency point (\\( S_{Lyα}(f) \\)). The normalized signals per frequency point are then calculated as \\( S(f) = S_{Lyα}(f)/S_{Norm} \\).\n\nIn region A of Fig. 3, events with longer TOFs than expected are summarized. Those events might be caused by either convoy electrons created at the foil [31] or by ionization products of a \\( \\mu^+ \\) interacting with the residual gas. The region A is leaking into the signal region, contributing to the measured background. The region B contains the events where the TOFs between the Lyα- and Stop-MCP are almost zero. These MCPs are close to each other, which leads to the risk of one detector picking up a noise signal when the other one is firing or vice versa. This noise signal can have a large amplitude with an intense after-ringing, which is mistaken as an additional signal that contributes to the events in region C.\n\nThe average power inside the TLs is frequency-dependent. To be able to fit a line-shape, all frequency points are corrected to the same “global” power of 22.5 W, obtaining the corrected signal \\( S_c \\). The applied correction is given by:\n\n\\[\nS_c(f) = (S(f) - S_{BKG}) \\cdot C(f) + S_{BKG},\n\\]\n\nwhere \\( S_{BKG} \\) is the background level. The correction factors \\( C(f) \\) are extracted by simulating the scaling ratio between the excitation probabilities at the effectively measured power and the global power for each frequency.\n\nFIG. 3. Correlation plot of the TOF between the Tag- and Stop-MCP and the TOF between the Tag- and Lyα-MCP. The signal region of Lyα events is circled. The regions labelled A,B,C are explained in the main text.\n\nBecause of the linear polarization of the field in the TLs, we can approximate our system to two levels and use the relevant optical Bloch equations [32] in the simulation. Single atom trajectories through both TLs reaching the Stop-MCP are simulated. While being in one of the TLs, the optical Bloch equations are numerically solved by an adaptive stepsize Runge-Kutta algorithm\n[33]. To reproduce the field inside the TLs, for each TL a field map is generated in SIMION, following the procedure of Lundeen and Pipkin [25]. The numbers of excited and ground state atoms detected in the Stop-MCP are counted and from that the average excitation probability calculated. As input parameter to the simulation we assign to each atom an initial state with its specific resonance frequency, and the power for both TLs operating at random phases of the fields. The momentum and initial position of the specific particle from the LEM is simulated with the musrSIM package [34] beforehand, from which the simulation randomly draws an atom.\n\nThe individual line-shape \\( P^{(i)} \\), where \\( i \\) stands for the assigned resonance frequency in MHz, is constructed by simulating the excitation probability for each initial state with large statistics over a frequency range between 200 MHz to 2000 MHz in steps of 1 MHz. Combining all relevant transitions, a global line-shape \\( P_n \\) for a \\( n \\) state is obtained:\n\n\\[\nP_{n=2} = 0.5 \\cdot P^{(1140)} + 0.25 \\cdot P^{(583)} + 0.25 \\cdot P^{(1326)}\n\\]\n\\[\nP_{n=3} = 0.5 \\cdot P^{(339)} + 0.25 \\cdot P^{(174)} + 0.25 \\cdot P^{(394)},\n\\]\n\nwhere the constant factors are coming from spin statistics.\n\nThe fitting function is constructed with the simulated line-shapes:\n\n\\[\nS_c = S_{BKG} + \\sum_n \\left[ B_n \\cdot P_n (f - f_{offset}) \\right],\n\\]\n\nwhere \\( B_n \\) is scaling the line-shape \\( P_n \\) for a specific \\( n \\) state and \\( f_{offset} \\) is introduced to allow for a global frequency offset compared to the theoretical resonance frequency used in the simulation.\n\n\n \n | Motif equivalence class | \n % of paternal | \n % of maternal | \n Odds ratio maternal/paternal (\\( P \\)) | \n
\n \n | AAC | \n 10.7% | \n 7.6% | \n 0.68[0.52;0.90] (7.6 · 10^{-3}) | \n
\n \n | AAG | \n 2.1% | \n 2.7% | \n 1.28[0.77;2.05] (0.36) | \n
\n \n | AAT | \n 80.9% | \n 84.2% | \n 1.26[1.03;1.55] (0.02) | \n
\n \n | ACC | \n 0.3% | \n 0.0% | \n No maternal mDNMs. | \n
\n \n | ACG | \n 0.0% | \n 0.0% | \n No mDNMs | \n
\n \n | ACT | \n 1.6% | \n 0.8% | \n 0.49[0.20;1.04] (7.2 · 10^{-2}) | \n
\n \n | AGC | \n 0.6% | \n 1.4% | \n 2.14[0.99;4.50] (5.1 · 10^{-2}) | \n
\n \n | AGG | \n 0.2% | \n 0.2% | \n 1.56[0.19;8.48] (0.64) | \n
\n \n | ATC | \n 3.4% | \n 2.1% | \n 0.63[0.37;1.02] (7.2 · 10^{-2}) | \n
\n \n | CCG | \n 0.3% | \n 1.0% | \n 3.39[1.28;9.20] (2.0 · 10^{-2}) | \n
\n \n | Motif equivalence class | \n % of paternal | \n % of maternal | \n Enrichment | \n
\n \n | AC | \n 88.7% | \n 79.8% | \n 0.50[0.46;0.55] (2.9 · 10^{-48}) | \n
\n \n | AG | \n 4.6% | \n 7.4% | \n 1.65[1.43;1.88] (4.5 · 10^{-12}) | \n
\n \n | AT | \n 6.6% | \n 12.7% | \n 2.05[1.84;2.29] (3.6 · 10^{-35}) | \n
\n \n | CG | \n 0.04% | \n 0.1% | \n 3.04[0.88;9.76] (6.2 · 10^{-2}) | \n
\n
\n\nTable S10 Motif equivalence class odds ratios for di- and trinucleotide microsatellites. Maternal mDNMs are more common at AAT, CCG, AG and AT motif class microsatellites while paternal mDNMs are more common at AAC and AC motif class microsatellites. Bold represents significant enrichment (p<0.05).\n\n \n | RRT length | \n Purity effect (Ice/UKBB) | \n \\( P \\) (Ice/UKBB) | \n
\n \n | 11-20 | \n 0.39/0.23 | \n < \\( 1.0 \\cdot 10^{-320} \\)/< \\( 1.0 \\cdot 10^{-320} \\) | \n
\n \n | 21-30 | \n 0.71/0.67 | \n < \\( 1.0 \\cdot 10^{-320} \\)/< \\( 1.0 \\cdot 10^{-320} \\) | \n
\n \n | 31-40 | \n 0.87/0.85 | \n < \\( 1.0 \\cdot 10^{-320} \\)/< \\( 1.0 \\cdot 10^{-320} \\) | \n
\n \n | 41-50 | \n 1.17/1.15 | \n < \\( 1.0 \\cdot 10^{-320} \\)/< \\( 1.0 \\cdot 10^{-320} \\) | \n
\n \n | 51-60 | \n 1.17/1.18 | \n < \\( 1.0 \\cdot 10^{-320} \\)/< \\( 1.0 \\cdot 10^{-320} \\) | \n
\n \n | 61-70 | \n 1.10/1.13 | \n < \\( 1.0 \\cdot 10^{-320} \\)/< \\( 1.0 \\cdot 10^{-320} \\) | \n
\n \n | 71-80 | \n 1.05/1.08 | \n < \\( 1.0 \\cdot 10^{-320} \\)/< \\( 1.0 \\cdot 10^{-320} \\) | \n
\n \n | 81-90 | \n 0.96/1.07 | \n 3.9 \\cdot 10^{-141}/6.8 \\cdot 10^{183} | \n
\n \n | >90 | \n 0.76/0.83 | \n 5.4 \\cdot 10^{56}/6.1 \\cdot 10^{57} | \n
\n
\n\nTable S19 Linear regression coefficients and p-values for the effect of repeat purity on expected heterozygosity within ten base pair bins of RRT length for both the Icelandic and UKB datasets.\n\n \n | Name | \n Sequence | \n A*02:01 Accession | \n Protein Name | \n Intensity Ratio (ESK1/MAGE) | \n -log(p) | \n Esk1 JNFAT signal | \n
\n \n | ILF2127-135 | \n KILPTLEAV | \n 0.051 Q12905 | \n Interleukin enhancer-binding factor 2 | \n 64.0 | \n 4.59 | \n + | \n
\n \n | RBM4B59-67 | \n KLHGVNINV | \n 0.03 Q9BQ04 | \n RNA-binding protein 4B | \n 84.4 | \n 2.94 | \n ++ | \n
\n \n | RLS117-125 | \n KIYEGQVEV | \n 0.004 P46777 | \n 60S ribosomal protein L5 | \n 4.6 | \n 2.67 | \n - | \n
\n \n | SEPT2121-129 | \n YIDEQFERY | \n 1.77 Q15019 | \n Septin-2 | \n 4.3 | \n 2.19 | \n - | \n
\n \n | SET150-158 | \n KVLSKEFHL | \n 0.324 Q01105 | \n Protein SET | \n 48.5 | \n 3.76 | \n - | \n
\n \n | SF3B478-86 | \n KLYGKPIRV | \n 0.004 Q15427 | \n Splicing factor 3B subunit 4 | \n 238.9 | \n 1.93 | \n + | \n
\n \n | SHC1469-477 | \n RVPPPPPQSV | \n 0.183 P29353 | \n SHC-transforming protein 1 | \n 5.3 | \n 2.7 | \n ++ | \n
\n \n | ARHGFEF26863-871 | \n RLLLGLETNV | \n 0.059 Q96DR7 | \n Rho guanine nucleotide exchange factor 26 | \n 181.0 | \n 1.8 | \n ++ | \n
\n \n | CRYL1130-138 | \n KLFAGLVHV | \n 0.004 Q9Y2S2 | \n Lambda-crystallin homolog | \n 256.0 | \n 3.36 | \n ++ | \n
\n \n | MSMO1164-172 | \n RIKYIHKV | \n 0.03 Q15800 | \n Methylsterol monooxygenase 1 | \n 13.0 | \n 2.94 | \n - | \n
\n \n | CYP2C859-67 | \n KVYGPVFTV | \n 0.011 P10632 | \n Cytochrome P450 2C8 | \n 22.6 | \n 3.6 | \n ++ | \n
\n \n | USP9Y1674-1682 | \n RLWGEVPVNL | \n 0.014 O00507 | \n Prob. ubiquitin carboxyl-terminal hydrolase FAF-Y | \n 27.9 | \n 3.95 | \n ++ | \n
\n \n | PLG92-100 | \n VLFEEKVYL | \n 0.008 P00747 | \n Plasminogen | \n 4.9 | \n 2.67 | \n - | \n
\n \n | MAF116-124 | \n ALISNSHQL | \n 0.038 Q75444 | \n Transcription factor Maf | \n 3.0 | \n 3 | \n - | \n
\n \n | ALDOA216-224 | \n ALSDDHIYL | \n 0.011 P04075 | \n Fructose-bisphosphate aldolase A | \n 3.5 | \n 2 | \n - | \n
\n \n | HNRNPA115-23 | \n KLFIGGLSF | \n 1.625 P09651 | \n Heterogeneous nuclear ribonucleoprotein A1 | \n 5.3 | \n 3.57 | \n - | \n
\n \n | EIF3F348-357 | \n IALNEKLVNL | \n 0.613 O00303 | \n Eukaryotic translation initiation factor 3 subunit F | \n 0.4 | \n 1.99 | \n - | \n
\n
\nTable 2: Change of the solvent-accessible surface area (ΔSASA) of the RMF peptide in the absence of the ESK1 Fab (calculated from the X-ray crystal structure with PDB ID 4WUU) and comparison to the confirmed cross-reactive peptides, ordered by their signal in the JNFAT activation assay. Bold font, red: Amino acid identical to respective residue in RMF peptide. Grey background: HLA anchor residues.\n\n\n \n | Peptide Sequence | \n ESK1 binding | \n Nature | \n Protein origin | \n Expression in the liver | \n
\n \n | RMFPNAPYL | \n yes | \n control | \n sp|P19544|WT1_HUMAN | \n Not detected | \n
\n \n | RMFWNAPYL | \n yes | \n control | \n sp|P19544|WT1_HUMAN, 1 aa substitution | \n na | \n
\n \n | RMFGNAPYL | \n yes | \n control | \n sp|P19544|WT1_HUMAN, 1 aa substitution | \n na | \n
\n \n | QLQNPSYDK | \n no | \n control | \n no match | \n | \n
\n \n | RAFPNAPYV | \n no | \n predicted | \n sp|P19544|WT1_HUMAN, 2 aa substitution | \n na | \n
\n \n | RLLPDFFGV | \n yes | \n predicted | \n sp|Q96S15|WDR24_HUMAN | \n Hepatocytes: medium, Cholangiocytes: low | \n
\n \n | RLFPLASLV | \n yes | \n predicted | \n sp|Q9NY72|SCN3B_HUMAN | \n Not detected | \n
\n \n | RVLPNLVAV | \n yes | \n predicted | \n sp|Q8IY22|CMIP_HUMAN | \n Hepatocytes: low, Cholangiocytes: low | \n
\n \n | RMFPNAFLY | \n yes | \n predicted | \n sp|P19544|WT1_HUMAN, 1 aa substitution | \n Not detected | \n
\n \n | RVMPCLAEV | \n yes | \n predicted | \n sp|Q6ZMD2|SPNS3_HUMAN | \n Not detected | \n
\n \n | RLFPLAWTV | \n yes | \n predicted | \n sp|P20827|EFNA1_HUMAN | \n Not detected | \n
\n \n | RLLPLPLGV | \n yes | \n predicted | \n sp|Q8NDA2|HMCN2_HUMAN | \n Not detected | \n
\n \n | RLLPLLNSV | \n yes | \n predicted | \n sp|Q8IU68|TMC8_HUMAN; sp|Q5STAT6|CODA1_HUMAN; | \n Hepatocytes: medium | \n
\n \n | RLLPLLRTV | \n yes | \n predicted | \n sp|P52798|EFNA4_HUMAN | \n Hepatocytes: Low | \n
\n \n | RLWPSLAVS | \n yes | \n predicted | \n no match | \n | \n
\n \n | RVMPSSFFL | \n yes | \n predicted | \n sp|O75663|TIPRL_HUMAN | \n Hepatocytes: low | \n
\n \n | RMFFNAPYL | \n yes | \n predicted | \n sp|P19544|WT1_HUMAN, 1 aa substitution | \n na | \n
\n \n | RLWPQGRGV | \n yes | \n predicted | \n sp|Q9UQQ1-6|NALDL_HUMAN | \n Not detected | \n
\n \n | RVLPAAFLV | \n yes | \n predicted | \n tr|AOA075B7D9.1|TAF15_HUMAN | \n Hepatocytes: high; Cholangiocytes: medium | \n
\n \n | RILPLFQGV | \n yes | \n predicted | \n sp|Q9UN76|S6A14_HUMAN | \n Not detected | \n
\n \n | RIIPYLIHV | \n yes | \n predicted | \n sp|P56645|PER3_HUMAN | \n Hepatocytes: high; Cholangiocytes: medium | \n
\n \n | RLLPALASV | \n yes | \n predicted | \n sp|Q03154|ACY1_HUMAN | \n Hepatocytes: high; Cholangiocytes: low | \n
\n \n | KLMGAISFFI | \n yes | \n predicted | \n sp|Q8NF4M|ADCY4_HUMAN | \n Hepatocytes: low; Cholangiocytes: low | \n
\n \n | MNLNVIAIV | \n no | \n predicted | \n sp|Q8IZF0|NALCN_HUMAN | \n Not detected | \n
\n \n | RVIMPCNWVV | \n yes | \n predicted | \n sp|Q9H2S6|TNMD_HUMAN | \n Not detected | \n
\n \n | RLLGGMALLL | \n yes | \n predicted | \n sp|Q9NX94-2|WBP1L_HUMAN | \n Not detected | \n
\n \n | RLMAMVDVLV | \n yes | \n predicted | \n sp|Q8NFP9|NBEA_HUMAN | \n Not detected | \n
\n \n | KTDPTVLLFV | \n no | \n predicted | \n sp|Q9H3R1|NDST4_HUMAN | \n Not detected | \n
\n \n | RMFSGVGVYL | \n yes | \n predicted | \n sp|Q969Y0|NXPE3_HUMAN | \n Not detected | \n
\n \n | ALLTAALWYI | \n no | \n predicted | \n sp|P17813|EGLN_HUMAN | \n Not detected | \n
\n \n | YIHFTPNFLL | \n no | \n predicted | \n sp|Q9UQC9|CLCA2_HUMAN; sp|Q9Y6N3|CLCA3_HUMAN | \n Not detected | \n
\n
"}
{"paper_id": "prb76", "title": "", "text": "Transgenic T cell therapy targeting the glioblastoma stem cell antigen PTPRZ1 with a vaccine-induced, patient-derived T cell receptor\n\nLukas Bunse\nl.bunse@dkfz.de\n\nGerman Cancer Research Center https://orcid.org/0000-0002-4490-7574\n\nYu-Chan Chih\nGerman Cancer Research Center\n\nAmelie Christina Dietsch\nGerman Cancer Research Center\n\nPhilipp Koopmann\nGerman Cancer Research Center\n\nXiujuan Ma\nGerman Cancer Research Center\n\nDennis Alexander Agardy\nGerman Cancer Research Center\n\nBinghao Zhao\nGerman Cancer Research Center\n\nAlice De Roia\nGerman Cancer Research Center\n\nAlexandros Kourtessakis\nGerman Cancer Research Center\n\nMichael Kilian\nAnn Romney Center for Neurologic Diseases, Brigham and Women's Hospital, Harvard Medical School\n\nChristopher Krämer\nGerman Cancer Research Center\n\nAbigail K. Suwala\nInstitute for Pathology, Department of Neuropathology, Heidelberg University\n\nMiriam Stenzinger\nInstitute for Clinical Transfusion Medicine and Cell Therapy\n\nHalvard Bonig\nGoethe University\n\nAgnieszka Blum\nArdigen, ul. Podole 76\nVictor Pieńkowski\nArdigen, ul. Podole 76\nKuralay Aman\nGerman Cancer Research Center\nJonas P Becker\nGerman Cancer Research Center\nHenrike Feldmann\nDepartment of Neurology, Medical Faculty Mannheim, Mannheim Center for Translation Neuroscience (MCTN)\nTheresa Bunse\nGerman Cancer Research Center\nRichard Harbottle\nGerman Cancer Research Center\nAngelika B. Riemer\nGerman Cancer Research Center\nHai-Kun LIU\nGerman Cancer Research Center https://orcid.org/0000-0003-3338-4066\nNima Etminan\nDept. of Neurosurgery, University Hospital Mannheim\nFelix Sahm\nDepartment of Neuropathology, Heidelberg University Hospital, and CCU Neuropathology, German Consortium for Translational Cancer Research (DKTK), German Cancer Research Center (DKFZ), Heidelberg https://orcid.org/0000-0001-5441-1962\nMiriam Ratliff\nUniversity Hospital Mannheim https://orcid.org/0000-0002-9464-6544\nWolfgang Wick\nUniversity of Heidelberg https://orcid.org/0000-0002-6171-634X\nMichael Platten\nGerman Cancer Research Center https://orcid.org/0000-0002-4746-887X\nEdward Wilhelm Green\nGerman Cancer Research Center\n\nArticle\n\nKeywords:\n\nPosted Date: August 9th, 2024\n\nDOI: https://doi.org/10.21203/rs.3.rs-4782908/v1\nLicense: This work is licensed under a Creative Commons Attribution 4.0 International License.\nRead Full License\n\nAdditional Declarations: Yes there is potential Competing Interest. The identified reactive TCR was patented by Y-C.C., M.K., C.K., E.W.G., W.W., M.P. and L.B., “T cell receptor derived binding polypeptides” (WO 2023/213904). E.W.G. and M.P. are founders of TCellTech.\n\nVersion of Record: A version of this preprint was published at Nature Communications on February 1st, 2025. See the published version at https://doi.org/10.1038/s41467-025-56547-w.\nTransgenic T cell therapy targeting the glioblastoma stem cell antigen PTPRZ1 with a vaccine-induced, patient-derived T cell receptor\n\nAuthors: Yu-Chan Chih1,2,3,4, Amelie C. Dietsch1,2, Philipp Koopmann1,2, Xiujian Ma2,5*, Dennis A. Agardy1,2,3,4, Binghao Zhao1,2, Alice De Roia1,2,3,4,6, Alexandros Kourtseakis2,7,8, Michael Kilian1,2,4,9, Christopher Krämer1,2,4, Abigail K. Suwala2,10,11, Miriam Stenzinger1,2,13, Halvard Bonig14,15, Agnieszka Blum16, Victor Murcia Pienkowski16, Kuralay Aman1,2, Jonas P. Becker2,17,18, Henrike Feldmann3,4, Theresa Bunse1,2, Richard Harbottle2,6, Angelika B. Riemer2,17,18, Hai-Kun Liu2,5, Nima Etminan19, Felix Sahm2,10,11, Miriam Ratliff19, Wolfgang Wick2,7,8, Michael Platten1,2,4,20,21,22, Edward W. Green1,2, Lukas Bunse1,2,4,22\n\nAffiliations:\n1Clinical Cooperation Unit (CCU) Neuroimmunology and Brain Tumor Immunology, German Cancer Research Center (DKFZ), Heidelberg, Germany\n2German Cancer Consortium (DKTK), DKFZ, core center Heidelberg, Germany\n3Faculty of Biosciences, Heidelberg University, Heidelberg, Germany\n4Department of Neurology, Medical Faculty Mannheim, Mannheim Center for Translation Neuroscience (MCTN), Heidelberg University, Mannheim, Germany\n5Division of Molecular Neurogenetics, DKFZ, DKFZ-ZMBH alliance, Heidelberg, Germany\n6DNA Vector Laboratory, DKFZ, Heidelberg, Germany\n7Neurology Clinic, Heidelberg University Hospital, Heidelberg, Germany\n8CCU Neurooncology, DKFZ, Heidelberg, Germany\n9Ann Romney Center for Neurologic Diseases, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA\n10Institute for Pathology, Department of Neuropathology, Heidelberg University, Heidelberg, Germany\n11CCU Neuropathology, DKFZ, Heidelberg, Germany\n12Institute for Clinical Transfusion Medicine and Cell Therapy, Heidelberg, Germany\n13Institute for Immunology, Heidelberg University Hospital, Heidelberg, Germany\n14Faculty of Medicine, Goethe University, Frankfurt a.M., Germany\n15Institute for Transfusion Medicine and Immunohematology, German Red Cross Blood Service Baden-Württemberg-Hessen, Frankfurt a.M., Germany\n16Ardigen, ul. Podole 76, Kraków, Poland\n17Division of Immunotherapy and Immunoprevention, DKFZ, Heidelberg, Germany\n18Molecular Vaccine Design, German Center for Infection Research (DZIF), partner site Heidelberg, Germany\n19Department of Neurosurgery, University Hospital Mannheim, Mannheim, Germany\n20Immune Monitoring Unit, National Center for Tumor Diseases (NCT), NCT Heidelberg, a partnership between DKFZ and Heidelberg University Hospital, Heidelberg, Germany\n21Helmholtz Institute for Translational Oncology Mainz (HI-TRON Mainz) – A Helmholtz Institute of the DKFZ, Mainz, Germany\n22DKFZ Hector Cancer Institute at the University Medical Center Mannheim, Mannheim, Germany\n\nABSTRACT\n\nResponse to CAR-T cell therapy in glioblastoma (GB) patients is limited, particularly because of the paucity of cell surface immunotherapeutic targets or low antigen sensitivity. Moreover, tonic CAR signaling leads to T cell dysfunction eventually resulting in non-durable responses. T cell receptor-engineered T (TCR-T) cell therapy circumvents the latter limitation by allowing safe and ubiquitous targeting of the GB-derived peptideome. Protein tyrosine phosphatase receptor type Z1 (PTPRZ1), a GB antigen associated with GB cell stemness, was previously targeted in multipeptide vaccination trials. Here, we identify a therapeutic HLA-A*02-restricted PTPRZ1-reactive TCR retrieved from a vaccinated GB patient. Single-cell sequencing of primary brain tumor samples showed *PTPRZ1* overexpression exclusively in malignant cells and most pronounced in GB stem cells and astrocyte (AC)- or oligodendrocyte-progenitor (OPC)-like cells. The validated vaccine-induced TCR recognized the endogenously processed antigen without cross-reactivity to *in silico* predicted low-risk off-targets. PTPRZ1-specific TCR-engineered human primary T cells (PTPRZ1-TCR-T) killed target cells antigen-specifically. Intravenous adoptive cell transfer against experimental brain tumors alone was not efficacious, but the addition of intracerebroventricular administration of PTPRZ1-TCR-T\nresulted in potent tumor rejection. PTPRZ1-TCR-T maintained stem cell memory phenotype *in vitro* and *in vivo*. PTPRZ1-TCR-T lysed all examined HLA-A*02+ primary GB cell lines with a preference for slow-cycling GB stem cells (GSCs) and AC-like cells. To further develop PTPRZ1-TCR-T as off-the-shelf therapy for HLA-A*02 GB patients, a phase I first-in-human clinical trial INVENT4GB is in preparation.\n\n**Main text:**\n\n**INTRODUCTION**\n\nEngineered T cell therapy has constituted a great success in combating hematopoietic cancers in the last decade with more than half of the patients demonstrating clinical responses leading to long-term survival in many studies\\(^{1-3}\\). On the other hand, its efficacy in treating solid tumors is still limited\\(^4\\). This is largely due to poor T cell infiltration, intratumoral T cell dysfunction and limited T cell persistence. These challenges are particularly pronounced in glioblastoma (GB), where the tumor is compartmentalized by the blood-brain barrier (BBB) excluding even unleashed peripheral tumor-reactive T cells, making GB one of the least responsive tumors to immunotherapies\\(^{5-8}\\).\n\nGB carries relatively few mutations\\(^9\\). Given the paucity of tumor-specific cell surface antigens, only a limited number of candidate targets can be exploited. Several chimeric antigen receptor (CAR)-T cell therapy clinical trials have been launched to target GB-associated surface antigens, e.g., IL13Ra2, HER2 and GD2, with fewer focusing on the GB-specific but subclonal antigen, EGFRvIII\\(^{10,11}\\). Primarily driven by sophisticated synthetic improvements of cellular approaches, some patients showed response to CAR-T cell therapy\\(^{12}\\). In addition, combinatorial treatments are under exploitation\\(^{13,14}\\). Unlike CARs, T cell receptors (TCRs) enable targeting both intracellular and extracellular antigens that are processed and loaded onto the major histocompatibility complex (MHC)\\(^{15}\\). This nature of antigen recognition by TCRs conceivably broadens the range of targetable antigens. Moreover, in comparison to CARs, TCRs display superior antigen sensitivity, reduced tonic signaling, and engage differential proinflammatory intracellular pathways\\(^{16,17}\\). Encouraging results have been demonstrated in Phase I and II clinical trials using TCR-engineered T cells (TCR-T) to treat solid tumors\\(^{18}\\). Despite current progress in advancing cell therapy, TCR-T cell therapy has not yet been investigated in GB patients. Previously, we have isolated and validated vaccine-induced glioma-reactive TCRs targeting the tumor-associated antigen, NLGN4X\\(^{19}\\), and neoantigens, IDH1R132H, H3K27M and CICR215W/Q\\(^{20-22}\\). In a multinational European GB immunotherapy clinical trial, Glioma Actively Personalized Vaccine Consortium 101 (GAPVAC-101), newly diagnosed GB patients were immunized with a warehouse of unmuted GB-associated peptides (APVAC1), and a warehouse of individual-specific mutated peptides (APVAC2)\\(^{23}\\). 92% of APVAC1-vaccinated patients developed MHC class I (MHC-I)-dependent immunogenicity. PTPRZ1\\(^{1814-1822}\\), an HLA-A*02-restricted epitope derived from the GB-associated antigen protein tyrosine phosphatase receptor type Z1, PTPRZ1, was administered to four patients within APVAC1 and elicited 100% immunogenicity\\(^{23,24}\\). Multimer-sorted PTPRZ1\\(^{1814-1822}\\)T cells demonstrated cytotoxicity against HLA-A*02+ GB cell lines with PTPRZ1 expression.\n\nPTPRZ1 is involved in central nervous system development\\(^{25,26}\\). Binding to its ligands, PTN and MK, leads to oligomerization and inactivation of the phosphatase receptor, resulting in sustained phosphorylation of its substrates, such as β-catenin and ALK\\(^{27,28}\\). Its absence leads to early onset of oligodendrocyte differentiation and disruption of the perineuronal net\\(^{25,29}\\).\nWhile its expression is very limited across adult tissues, it is essential for gliomagenesis. In gliomas, PTPRZ1 takes part in cancer cell proliferation, migration, and invasiveness30-32. In addition, it contributes to angiogenesis and tumor radioresistance33,34. Notably, depletion of PTPRZ1 results in significant impairment of glioma cell sphere formation *in vitro* and delays tumor growth *in vivo*, indicating a strong associated of PTPRZ1 with glioma cell stemness35,36. Recent studies have employed PTPRZ1 along with other markers to define GB stem cells (GSCs) in single-cell transcriptomic and flow cytometric analyses37,38. Here, we performed a comprehensive multimodal assessment of PTPRZ1 in GB, demonstrated that it is a highly attractive immunotherapeutic target, and developed an off-the-shelf TCR-T therapy, in principle, applicable for all HLA-A*02+ glioma patients.\n\nRESULTS\n\nPTPRZ1 is exclusively overexpressed and presented in glioma cells, and associated with distinct cellular states and stemness in GB\n\nTo evaluate the extent of PTPRZ1 association with gliomas, we first examined its expression at transcript and protein levels in human GB tissues. In TCGA bulk RNA sequencing (RNA-seq) data, *PTPRZ1* was significantly upregulated in GB compared to adjacent normal (Fig. 1A). In addition, *PTPRZ1* was overexpressed in low-grade glioma (LGG) (Fig. S1A), and its expression correlated positively with ABSOLUTE tumor purity (Fig. S1B,C)39. In order to detail *PTPRZ1* at cellular resolution, we interrogated publicly available GB and isocitrate dehydrogenase (IDH)-mutant glioma single-cell RNA-seq (scRNA-seq) datasets40,41, and found *PTPRZ1* to be highly expressed in malignant cells but not in immune cells such as T cells and macrophages, nor in normal glial cells, oligodendrocytes and microglia (Fig. 1B and Fig. S1D-H). In a paired glioma and stromal cell scRNA-seq dataset (n(GB)=16, n(IDH-mut glioma)=3, and n(pediatric high-grade glioma)=1), we independently validated malignant cell-type specificity of *PTPRZ1* expression (Fig. 1C,D and Fig. S1I,J). Interindividual levels of *PTPRZ1* overexpression varied across patients but were found in all gliomas (Fig. S1K,L). To evaluate protein abundance, we analyzed n=20 matched primary and recurrent GB samples and found patient-individual PTPRZ1 levels without temporal alterations between primary and recurrent tumors (Fig. 1E,F). In concordance with transcriptomic data using GFAP to depict tumor purity, abundance of PTPRZ1+ cells correlated positively with GFAP+ cells (Fig. 1G). To address whether PTPRZ1 upregulation leads to PTPRZ1-derived epitope presentation, we performed untargeted human leukocyte antigen (HLA) ligandomics. PTPRZ1-derived ligands were found across examined primary GB cell lines, and PTPRZ11814-1822, in particular, was among the overlapping PTPRZ1-derived ligands (Fig. 1H and Fig. S1I).\n\nDistinct cellular states were demonstrated in GB (Fig. S2A)41, and *PTPRZ1* expression was found to be enriched in astrocyte-like (AC-like) and oligodendrocyte-progenitor-like (OPC-like) GB cells and positively correlated with their module scores (Fig. S2B,C). Similarly, in our own cohort, we found higher *PTPRZ1* expression levels in AC-like and OPC-like GB cells (Fig. 1I-K). Comparable to previous results41, the distribution of cell states differed across patients (Fig. S2D). To assess correlation with GB stemness, a GSC score of each cell was derived from a previously defined gene set38. Indeed, *PTPRZ1* was positively correlated with GSC score in both cohorts (Fig. 1L and Fig. S2E). Additionally, we grouped TCGA-GB tumors according to dominant GB cell states (Fig. S2F), and concluded again that *PTPRZ1* was positively correlated with an AC-like state and GSC (Fig. S2G,H). Collectively, we found that *PTPRZ1* expression correlates positively with AC-like and OPC-like GB states and GB stemness, and that its robust expression results in a high MHC ligand load in GB.\nA vaccine-induced T cell receptor binds intracellularly processed PTPRZ1^{1814-1822} without evidence of off-target reactivity\n\nTo discover therapeutic PTPRZ1-reactive TCRs, peripheral blood mononuclear cells (PBMCs) were retrieved from a female HLA-A*02+ GAPVAC-101 patient, pt.16, who had undergone 11 APVAC1 vaccinations at the time of sample collection with a favorable clinical course, showing immune responses to the examined PTPRZ1 peptides (Fig. 2A)^{23}. Reactive T cells were sorted following in vitro restimulation with PTPRZ1^{1347-1355} or PTPRZ1^{1814-1822} peptide and then subjected to scTCR-seq. The TCR repertoire of the sorted T cells following in vitro restimulation with PTPRZ1^{1347-1355} peptide was polyclonal while the one following in vitro restimulation with PTPRZ1^{1814-1822} peptide was oligo-clonal, with the top CDR3 taking up 75.09% and the second top 19.76% of the repertoire (Fig. 2A). Interestingly, the top CDR3 was constituted of one β chain paired with different α chains. Subsequently, the dominant TCRs for both peptides were cloned into a non-viral, non-integrating episomal scaffold matrix attachment region (S/MAR) DNA vector^{42}. Constant regions of TCR α and β chains were murinized to avoid mispairing with endogenous human TCRs^{43}. For subsequent TCR validation, TCRs were electroporated into Jurkat cells carrying NFAT, AP-1, and NF-κB reporters or electroporated along with a plasmid encoding a reporter into Jurkat cells (Fig. 2B, Fig. S3A). TCR-Jurkat cells were then cocultured with peptide-loaded presenter cells for 24 hours. Out of the top two TCRs for PTPRZ1^{1814-1822}, sharing the same β chain but paired with different α chains, one demonstrated strong reactivity against peptide-loaded target cells (Fig. 2C). None of the tested dominant TCRs for PTPRZ1^{1347-1355} were reactive to peptide-loaded target cells (Fig. S3B,C). Thus, the identified reactive TCR for PTPRZ1^{1814-1822} was subsequently further examined.\n\nExogenous peptides do not necessarily induce immunogenicity against naturally processed and presented antigens^{44}. Therefore, we next assessed whether the peptide vaccine-induced TCR was indeed reactive to the endogenously processed and presented antigen PTPRZ1^{1814-1822}. Since the complete open reading frame (ORF) of *PTPRZ1* is large (~7kbp), we generated a well-characterized HLA-A*02+ GB cell line (U87) stably expressing a tandem minigene (TMG) encoding several antigens derived from the GAPVAC-101 and IMA950 trials, including the PTPRZ1 antigens of interest (Fig. S4A,B and Table. S1)^{23,45}. Upon coculture with U87 TMG cells, TCR-Jurkat cells also showed strong reporter activity (Fig. 2D), indicating that the identified TCR is able to recognize the intracellularly processed and presented antigen PTPRZ1^{1814-1822}. Next, we aimed to evaluate the safety profile of the PTPRZ1^{1814-1822} TCR. *PTPRZ1* is, if at all, barely expressed across adult tissues^{24}, and the TCR was isolated from a vaccinated patient without notable adverse events^{23}; hence, we hypothesized that the TCR had undergone thymic selection, making both off-target and on-target off-tumor toxicities unlikely. Nevertheless, to experimentally assess potential off-targets, we applied ARDitox, an *in silico* artificial intelligence (AI)-based prediction tool for off-target TCR binding^{46}. A panel of potential off-targets for the PTPRZ1^{1814-1822} TCR was generated with various off-target-specific safety and presentation scores (Fig. 2E, Table S2). No high-risk (safety score < 3) potential off-targets were predicted while 12 low-risk (safety score > 3) off-targets with relevant presentation probabilities on HLA-A*02 were predicted. In a subsequent Jurkat reporter assay, we did not find reactivity against any of these 12 potential off-targets (Fig. 2F). Together, these data reveal a patient-derived vaccine-induced PTPRZ1^{1814-1822}-reactive TCR that binds to both exogenous and intracellularly processed and presented antigen on MHC-I and does not react against AI-predicted low-risk off-targets.\nPTPRZ1^{1814-1822} TCR engineered primary human T cells exert effector functions and maintain T_{SCM} phenotype in vitro\n\nWe next endeavored to explore whether the PTPRZ1^{1814-1822} TCR elicits effector functions in primary human T cells. We employed a GMP-compatible retroviral system commonly used for CAR-T cell manufacturing to transduce the constant region-murinized PTPRZ1-TCR. Assessed by flow cytometric analysis of murine constant TCR β chain surface abundance and representative for multiple experiments, transduction efficiency was >85% for PTPRZ1-TCR and negative control Influenza (Flu)-TCR, targeting the HLA-A*02-restricted epitope GILGFVFTL (Fig. S5A). Without further enrichment, the TCR surface expression was maintained without a decrease in frequency 20 days post transduction in vitro (Fig. 3A). Of note, our culture conditions using IL-7 and IL-15 for stimulation and maintenance instead of conventional IL-2 stimulation favored the expansion of CD8^{+} T cells (Fig. 3B)^{47}. Subsequent longitudinal subtyping of CD8^{+} TCR-T cell into stem cell memory (T_{SCM}, CD45RA^{+} CD62L^{+}), central memory (T_{CM}, CD45RA^{+} CD62L^{+}), effector memory (T_{EM}, CD45RA^{+} CD62L^{-}), and CD45RA^{+} effector memory (T_{EMRA}, CD45RA^{+} CD62L^{-}) revealed a long-lasting and predominant adaptation to a T_{SCM} phenotype over the course of three weeks in vitro regardless of TCR expression (Fig. 3C). In contrast to previous studies, we did not enrich for naïve T cells yet still were able to generate T_{SCM}-abundant engineered T cell products^{47}. The generated TCR-T cells were then cocultured with various target cells. Only upon contact with cognate peptide-loaded target cells or U87 TMG, CD8^{+} TCR-T cells were activated (Fig. 3D). Importantly, the primary HLA-A*02^{+} GB cell line P3, endogenously expressing PTPRZ1^{23}, also activated CD8^{+} PTPRZ1^{1814-1822} TCR-T cells (Fig. 3D). Moreover, the frequencies of effector cytokine- and cytolytic protein- expressing CD8^{+} TCR-T cells were increased in an antigen-specific fashion (Fig. 3E and Fig. S5B-D). PTPRZ1^{1814-1822} TCR-T cells (PTPRZ1-TCR-T) demonstrated antigen-specific dose-dependent cytotoxicity with an optimal E:T ratio of 2:1 (Fig. 3F,G).\n\nA recent study suggests that cytotoxic CD4^{+} CAR-T cells facilitate long-term tumor control^{48}. Our TCR-T cell manufacturing process transduced both CD4^{+} and CD8^{+} T cells, and we observed that CD4^{+} TCR-T cells, even without the co-receptor CD8, were activated in an antigen-dependent manner (Fig. S6A-C). Moreover, additional irradiation of tumor cells in vitro resulted in increased lysis of target cells likely due to enhanced activation of PTPRZ1-TCR-CD4^{+}-T but not PTPRZ1-TCR-CD8^{+}-T (Fig. 3D,E, Fig. S5E-H and Fig. S6C-F). Although cytolytic proteins were elevated in PTPRZ1-TCR-CD4^{+}-T, it remained unclear if PTPRZ1-TCR-CD4^{+}-T encountered target cells directly to execute cytotoxicity or to support neighboring cytotoxic PTPRZ1-TCR-CD8^{+}-T. Thus, we further unraveled the role of CD4^{+} T cells engineered with the CD8-restricted PTPRZ1^{1814-1822} TCR on tumor cell lysis in vitro through enriching CD4^{+} and CD8^{+} T cells via magnetic-activated cell sorting (MACS) post transduction (Fig. 3H). As expected, serial dilution of PTPRZ1-TCR-CD4^{+}-T in comparison to that of PTPRZ1-TCR-CD8^{+}-T revealed that CD4^{+} T cells play a minor role in target cell killing (Fig. 3I,J). However, we found a moderate increase in cytotoxicity when both CD4^{+} and CD8^{+} T cells carried the PTPRZ1^{1814-1822} TCR compared to the combination of PTPRZ1^{1814-1822}-reactive CD8^{+} T cells with non-target-specific CD4^{+} T cells (Fig. 3K). Thus, our data suggest a differential but synergistic role for CD4^{+} and CD8^{+} T cells following antigen-specific activation via PTPRZ1^{1814-1822} TCR.\n\nPTPRZ1-TCR-T is efficacious in experimental flank and brain tumors\n\nTo investigate the therapeutic potential of the PTPRZ1^{1814-1822} TCR in vivo, we first inoculated U87 TMG cells subcutaneously (s.c.) in the flank of immunodeficient mice, followed by two\ndoses of intravenous (i.v.) TCR-T cell administration (Fig. 4A). Following adoptive cell transfer (ACT) with PTPRZ1-TCR-T, flank tumors regressed over time while control mice without treatment or treated with an irrelevant TCR-T product showed sustained tumor growth and met termination criteria by day 40 post tumor inoculation (Fig. 4B and Fig.S7A-D). Of note, after initial tumor regression, some PTPRZ1-TCR-T-treated animals experienced tumor recurrence starting from day 42 onwards, yet by the predefined experimental endpoint, 33% (3 out of 9) of PTPRZ1-TCR-T cell-treated animals remained tumor-free, resulting in prolonged survival (Fig. S7C,D).\n\nSubsequently, we advanced to investigate PTPRZ1-TCR-T in experimental brain tumors. A recent study has demonstrated that repeated systemic i.v. CAR-T delivery does not result in GB control in a phase I trial, not even in a combinatorial treatment with immune checkpoint blockade14. At least in part, this might be attributed to the BBB resulting in the exclusion of T cells infiltrating via the blood stream into the CNS5,6. Indeed, when we delivered PTPRZ1-TCR-T into mice bearing intracranial U87 TMG tumors i.v. (Fig. S8A), no therapeutic effect was observed (Fig. S8B). Hence, we adapted our ACT regime to two doses of intracerebroventricular (i.cv.) ACT following one dose of i.v. ACT (Fig. 4C). Notably, by combined i.v. and i.cv. PTPRZ1-TCR-T, mice showed preclinical response (Fig. 4D,E). 5 out of 7 (71.4%) of PTPRZ1-TCR-T-treated tumor-bearing mice responded radiographically to ACT while tumors of control-treated mice continued to grow (Fig. 4E). Only PTPRZ1-TCR-T-treated mice (4 out of 7; 57.1%) survived till the experimental endpoint (Fig. 4F and Fig. S9A-C). One out of the survivors remained tumor-free macroscopically while the others experienced tumor recurrence at experimental endpoint (Fig. S9C). To assess persistence of the TCR-T cell in vivo, we monitored T cell engraftment in the blood. Three weeks post first ACT, TCR-T cells were detected in all mice receiving ACT (Fig. 4G). Adoptive transfer of different CD8+ T cell subsets is known to affect T cell engraftment and T cell effector functions. Paradoxically, ACT with more stem-like or central memory T cells outperforms other CD8+ subsets as they retain self-renewal capacity but nonetheless, have less cytotoxic and cytokine-releasing capacity49-51. Hence, we characterized engrafted T cells by flow cytometry. Up to three weeks following first ACT, we found the T_{SCM} phenotype to remain dominant in vivo with negligible inter-individual differences (Fig. 4H,I). No difference in population frequencies was observed when comparing therapeutic PTPRZ1-TCR-T- with control TCR-T-treated mice (Fig. 4I), suggesting that the engraftment and CD8+ T cell subset maintenance were independent of cognate antigen encounter in vivo.\n\nTo decipher if in recurrent tumors, either dormant GB cells regain proliferative capacity when reactive TCR-T cells are no longer present within the GB microenvironment52,53, or MHC and/or antigen loss occurs54,55, we analyzed post-mortem tumors reaching preclinical termination criteria or experimental endpoint. We found that transferred human T cells in the tumor could only be detected in mice treated with PTPRZ1-TCR-T (Fig. 4J,K), even after 76 days following the last ACT (Fig. S9D). In addition, MHC1 expression was maintained in recurrent tumors (Fig. 4J). We thereby examined antigen expression with RNAscope™ and found that TMG transcripts were greatly diminished in the recurrent tumors (Fig. 4L,M). These data demonstrate the on-target activity of PTPRZ1-TCR-T and highlights the ACT route-dependent efficacy in experimental GB. Intriguingly, the reactive TCR-T cell product persists in experimental brain tumors, and the relapse is driven by downregulation and/or loss of the tandem antigens in this model.\n\nPTPRZ1-TCR-T shows cytotoxicity against patient-derived HLA-A*02+ GB spheroid cell lines and preferentially targets slow-cycling cells\nTo assess PTPRZ1-TCR-T efficacy on primary GB, we established patient-derived primary GB cell lines following previously described protocols of tumor cell isolation via FACS or MACS (Fig. S11)56. The established primary GB cell lines were maintained in serum-free medium and cultured in spheroids to maintain stemness. Benchmarking cytotoxicity, we cocultured PTPRZ1-TCR-T with primary HLA-A*02+ GB cell lines, D170_44 and P3, and found comparable cytotoxicities (Fig. 5A). Again, both CD8+ and CD4+ PTPRZ1-TCR-T cells were activated by the primary GB cell line D170_44 (Fig. S10A,B).\n\nSlow-cycling GB cells are considered stem-like as they have the potential for tumor initiation, therapy resistance, and generation of large number of progenies37,57,58. As PTPRZ1 was defined as a GSC marker and found to be highly associated with GSC scores, we hypothesized that PTPRZ1-TCR-T preferentially targeted stem-like slow-cycling cells (SCC). To identify SCC, primary GB cell lines were labeled with fluorescent dye, and the top 10% dye-retaining cells after expansion were considered SCC while the rest were defined as fast-cycling cells (FCC). Upon coculture with PTPRZ11814-1822 TCR-T cells, dye-retaining SCC and dye-losing FCC were enumerated by flow cytometry to assess differential cytotoxicity (Fig. 5B). Indeed, in the first five hours, PTPRZ11814-1822 TCR-T cells preferentially killed SCC while after 24 hours, both SCC and FCC were lysed (Fig. 5C and Fig. S10C), suggesting a preferential anti-tumor activity on stem-like SCC by PTPRZ1-TCR-T.\n\nFurthermore, to demonstrate HLA-A*02-dependency and assess the breadth of anti-tumor activity by PTPRZ11814-1822 TCR, we established five additional primary GB cell lines, of which MA120, MA140, and D170_108 were HLA-A*02+, and MA108 and MA118 were HLA-A*02- (Fig. 5D). Upon coculture, PTPRZ1-TCR-T exclusively killed HLA-A*02+ lines (Fig. 5E). Primary HLA-A*02+ GB cell lines led to activation of CD8+ PTPRZ1-TCR-T at various degrees and additionally activated CD4+ PTPRZ1-TCR-T in a moderate fashion. Conversely, HLA-A*02- GB cell lines did not elicit any activation nor were lysed (Fig. 5E-H and Fig. S10D). Together, these data demonstrate that PTPRZ11814-1822 TCR can broadly and specifically target HLA-A*02+ GB primary cell lines (5 out of 5) with a preference for stem-like SCC.\n\nPTPRZ1-TCR-T impacts GB cell states and targets GB stemness in individual patient tumor organoids (IPTOs)\n\nAs previously shown, *PTPRZ1* is higher expressed in AC-like and OPC-like cells and associated with GB stemness (Fig. 1L,M). Therefore, we evaluated whether PTPRZ1-enriched subsets of malignant cells were targeted preferentially by PTPRZ1-TCR-T. Individual patient tumor organoids (IPTOs) maintain tumor multicellular characteristics and are thereby suitable to study PTPRZ1-TCR-T-mediated alterations in a realistic multicellular-orchestrated microenvironment (Fig. 6A). IPTOs from three HLA-A*02+ patients were treated with TCR-T cells and interrogated by scRNA-seq. GB cells and immune cells were identified based on the absence of EGFP and presence of B2M, respectively (Fig. 6B, Fig. S11A,B), as host feeder organoid cells are engineered to be EGFP-expressing and B2M-deficient. The GB cell cluster expressed high levels of *PTPRZ1* (Fig. 6B). Upon PTPRZ1-TCR-T treatment, malignant cell frequency was decreased (Fig. 6C), and average *PTPRZ1* expression in malignant cells was lowered (Fig. 6D). To understand the underlying mechanism of lowered *PTPRZ1* expression, we aimed to decipher cell state distribution in post-treatment IPTOs. Assessing cell states, we again observed that *PTPRZ1* was associated with AC-like tumor cells (Fig. 6E), but upon PTPRZ1-TCR-T cell treatment, AC-like score and AC-like tumor cell frequency were reduced\nsignificantly (Fig. 6F,G). Corroborating our previous findings, a positive correlation of GSC score and *PTPRZ1* was identified (Fig. 6H), and PTPRZ1-TCR-T coculture led to a drop of GSC cell frequency (Fig. 6I). Collectively, the results confirm that *PTPRZ1* is highly expressed in malignant brain tumor cells, most abundantly in GSCs, and that AC-like cells and GSCs are primarily targeted by PTPRZ1*1814-1822* TCR-T cells.\n\nDISCUSSION\n\nIn this study, we profiled the expression of PTPRZ1, a GB-associated antigen and GSC marker, across bulk and scRNA-seq datasets in publicly available and own cohorts, and took advantage of a previous multinational European GB vaccination trial, enabling us to retrieve an HLA-A*02-restricted PTPRZ1-reactive TCR in a reverse translation research paradigm. The TCR was comprehensively examined *in vitro* and *in vivo* for its anti-tumor activity against experimental flank and brain tumors, primary GB cell lines, and IPTOs. Besides the validated TCR reactivity, we identified and excluded low-risk off-targets. In addition to the robust *in vivo* efficacy of PTPRZ1-TCR-T, we found universal targeting of HLA-A*02* primary GB cell lines and malignant cells in IPTOs with a preference for stem-like SCC and AC-like tumor cells.\n\nPTPRZ1 is essential during neurodevelopment but not in adults\\(^{25,26}\\); however, its expression re-emerges and is required for gliomagenesis and tumor progression. Whereas the mechanistic relevance of PTPRZ1 in GB remains elusive, its functions and contributions to disease malignancy are well-demonstrated, ranging from cancer cell proliferation and stemness to angiogenesis and therapy resistance\\(^{30-34}\\). Here, we explored treatment with the identified PTPRZ1*1814-1822* TCR, yet therapies for non-HLA-A*02* patients and combinatorial therapies should be investigated further, particularly when taking all the variety of PTPRZ1-derived MHCI-ligands into consideration (Fig. 1H). Apart from PTPRZ1 upregulation in GB, we also found robust overexpression of PTPRZ1 in IDH-mutant gliomas. Further investigations are required to examine whether PTPRZ1-TCR-T can be applied to oligodendrogliomas and astrocytomas or other cancer entities that display elevated *PTPRZ1* expression\\(^{59}\\). Moreover, our results further support PTPRZ1 as a GSC marker, not only from the bioinformatic analyses but also from *in vitro* assays showing preferential killing of stem-like SCCs. Therefore, from a neuro-oncological perspective, targeting a cell population that drives disease initiation, progression, and treatment resistance, supports future clinical investigations of PTPRZ1-TCR-T.\n\nA major advantage of TCR-T therapy is its capability to target all proteins, including intracellular neoantigens, as long as they are processed and presented on MHC. As studied by others, PTPRZ1, a transmembrane receptor, can be, in principle, targeted by CAR-T cells\\(^{60}\\). However, protein glycosylation is often altered in cancers, including brain tumors\\(^{61,62}\\). Such aberrant post-translational modification may impact CAR-T cell therapy efficacy\\(^{63}\\), and increasing evidence is emerging suggesting TCR but not tonic CAR signaling cascades to be associated with favorable intratumoral T cell fate\\(^{16,17}\\). Conversely, limitations of TCR-T are HLA-type dependency and the prerequisite of MHC expression. Despite HLA restriction which is inherently associated with the need of therapy individualization, our TCR covers a common HLA and targets the intracellular domain of PTPRZ1 presented on MHCI. As MHC antigen processing and presentation machinery is rarely altered in GB\\(^{55}\\), TCR-T cell therapy presents a highly promising approach. Of note, in our glioma model overexpressing but not depending on TMG *in vivo*, we found loss of the cognate antigen (Fig .4L). It remains speculative, if this is a unique feature of our model as *PTPRZ1* expression remained stable during disease\nprogression in patients (Fig. 1F). Nevertheless, more GB-reactive TCRs should be developed to expand the HLA coverage and to allow multivalent cell therapy, preventing escape through antigen loss.\n\nIn our study, the presence of CD4+ T cells within PTPRZ1-TCR-T was required for optimal cytotoxicity in vitro. In line with this, the relevance of the CD4+::CD8+::APC immune triad for solid tumor eradication has been highlighted recently64. In PTPRZ1-TCR-T, CD4+ T cells also produced effector proteins and cytokines, suggesting a dual action of moderate direct cytotoxicity and paracrine reprogramming of CD8+ T cells. Whether CD4+ T cells engineered with CD8-restricted TCR can hijack tumor cells to form triad to restore and improve CD8+ T cell functions will require further experimental investigations. Notably, CD4+ CAR-T cell persistence was associated with long-term survival in leukemia patients, but on the other hand, CD4+ CAR-T cells are implicated in cytokine release syndrome. Here, fine tuning of CD4+::CD8+ T cell ratio without impacting anti-tumor immunity is required48,65.\n\nEncouraged by the preclinical results, we are initiating a phase I clinical trial, Intraventricular T cell receptor transgenic T cell therapy to treat glioblastoma (INVENT4GB)66, to assess the feasibility and safety of CD4+/CD8+ PTPRZ1-TCR-T cell therapy in patients with recurrent GB. It will mark the first-in-human TCR-T cell therapy against GB and highlight the here proposed “bedside to bench to bedside” approach.\n\nMATERIALS AND METHODS\n\nCell lines\nU87 was cultured in DMEM supplemented with 10% fetal bovine serum (FBS) and 1% Penicillin/Streptomycin (P/S). U87 TMG cell line was generated by transfection with pMXs-IRES-PuroR plasmid encoding TMG as illustrated in Fig. S2C. Transfection was performed with Fugene HD transfection reagent (#E2312, Promega) following the manufacturer instructions. Briefly, U87 cells were seeded at a density of 3 x 10^5 cells per well in a 6-well plate and rested for 24h. On the next day, the medium was replenished, and the cells were transfected with 2 μg of the DNA plasmid and rested for another 48h. Cells stably expressing TMG were then selected with 2 μg/ml Puromycin. T2 cells were kindly provided by Dr. Angelika Riemer, Division of Immunotherapy and Prevention, DKFZ, and cultured in RPMI-1640 supplemented with 20% FBS and 2% L-Glutamine.\nHEK293 cells were cultured in IMDM supplemented with 10% FBS.\nJurkat TCR deficient cells were purchased from ATCC, and Jurkat 76-Triple parameter reporter (J76-TPR) was kindly provided by Prof. Peter Steinberger, Division for Immune Receptors and T Cell Activation, Institute of Immunology, Medical University of Vienna67. Jurkat cells were cultured in RPMI-1640 supplemented with 10% FBS and 1% P/S.\nPrimary brain tumor samples were collected at the University hospital of Mannheim. All patients have provided written signed informed consent in accordance to the WMA Declaration of Helsinki principles. Ethical approval for the isolation of brain tumor tissue and analyses was obtained from the Mannheim Medical Faculty Ethics Committee (2017-589N-MA, 608-22, 574-23). The harvested tumors were processed with tumor dissociation kit (#130-095-929, Miltenyi Biotec) and further enriched for tumor cells with tumor cell isolation kit (#130-108-339, Miltenyi Biotec) following the manufacture instructions. The isolated tumor cells and previously established primary GB cell lines were cultured in DMEM-F12 supplemented with B27 (#17504044, Thermo Fisher), 5 μg/ml Insulin (#I9278, Sigma-Aldrich), 5 μg/ml Heparin (#H4784, Sigma-Aldrich), 20 ng/ml epidermal growth factor (EGF, #AF-100-15, Peprotech) and 20 ng/ml fibroblast growth factor (FGF, #100-18b, Peprotech)56.\nAll cell lines were incubated at 37°C and 5% CO₂ with saturated humidity.\n\nMice\nNOD-Prkdcscid-Il2rgTm1/Rj (NXG) mice were purchased from Janvier labs. NOD.Cg-Prkdcscid H2-K1lmtBpe H2-Ab1lem1Mvw H2-D1lmtBpe Il2rgtm1Wjl/SzJ (NSG MHC KO) mice were bred at the DKFZ animal facility. Mice were housed under Specific and Opportunistic Pathogen Free (SOPF) conditions and under 12h day/night cycle with water and chow ad libitum. All animal procedures were conducted in compliance with the institutional laboratory animal research guidelines and were approved by the governmental institutions (Regional Administrative Authority Karlsruhe, Germany, file number: G-263/18 and G-130/23).\n\nIPTO generation and culture\nInduced pluripotent stem cells (iPSCs; AICS-0036-006, Institute for Cell Science) expressing enhanced green fluorescent protein (EGFP) were utilized to generate cerebral organoids. Organoid generation adhered to a published method⁶⁸, beginning with the seeding of dissociated iPSCs in 96-well round bottom ultra-low attachment plates with previously described hESC medium⁶⁸, supplemented with 4 ng/ml basic fibroblast growth factor (bFGF; #PeproTech, 100-18B), and 50 μM ROCK inhibitor (#72304, Stemcell Technologies). Formed embryoid bodies were subsequently transferred to 24-well ultra-low attachment plates and 6-well ultra-low attachment plates in previously described Neural Induction Medium⁶⁸, followed by a transition to improved differentiation medium -A and improved differentiation medium +A⁶⁹, with agitation introduced from day 18. Resected tumor tissues were processed and dissected into small explants, which were subsequently inserted into incised cerebral organoids and embedded in Matrigel. Generated IPTOs were then cultured in improved differentiation medium +A on an orbital shaker in an incubator (75 rpm, 37°C). After 10 days of incubation, the IPTOs were ready for TCR-T testing.\n\nElectroporation of Jurkat cells\nNeon transfection system (Thermo Fisher) was employed to deliver TCR and reporters. For transfection of Jurkat cells without reporters, 2 x 10⁶ Jurkat cells were resuspend in 100 μl R buffer with 5 μg of TCR and 5 μg of reporter plasmids. For transfection of J76-TPR, 2 x 10⁶ cells were resuspend in 100 μl R buffer with 5 μg of TCR plasmid. Electroporation was then performed at 1325 V, 3 pulses, 10 ms.\n\nTransduction of primary human T cells\nRetroviral transduction was performed as previously described⁷⁰. Briefly, TCR were inserted into SFG-IRES-GFP retroviral vector (kind provision from Dr. Martin Pule, Addgene #22493; RRID:Addgene_22493) with In-Fusion Cloning (#638947, Takara). 3 x 10⁶ HEK293 cells were seeded a day before transfection in 10ml IMDM supplemented with 10% FBS in 60.1 cm² dish. On the day of transfection, 3.75 μg TCR-SFG along with 3.75 μg PeqPam and 2.5 μg RD114 packaging plasmids were resuspended in 470 ml IMDM and 30 μl Fugene HD transfection reagent. PeqPam and RD114 packaging plasmids were kindly gifted by Dr. Tim Sauer. After 10-minute incubation, cells were transfected with the DNA-Fugene HD mix and incubated for 48h. The viral supernatant was then collected and strained through 0.45 μm strainers. Human T cells were obtained from healthy donors via density gradient separation followed by MACS with Pan T cell isolation kit (#130-096-535, Miltenyi) in accordance with manufacturer instructions. Isolated T cells were activated in CTL medium (45% RPMI-1640, 45% Click’s medium, 10% FBS, 10 ng/ml IL-7 [#200-7, Peprotech] and 5 ng/ml IL-15 [#200-15, Peprotech]) with T Cell TransAct (#130-111-160, Miltenyi) for 48h at a concentration of 1 x 10⁶ cells/ml. The strained viral supernatant was plated 0.5 ml per well in a non-tissue culture\ntreated 24-well plate (#351147, Falcon) precoated o/n with 0.5 ml of 7 μg/ml RetroNectin (#T100B, Takara) and centrifuged at 2000g for 90 minutes at 4°C. Later, the supernatant was removed and activated T cells were seeded at a concentration of 0.5x10^6 cell/ml in 1 ml CTL medium per well and centrifuged at 500g for 5 minutes. After 4 days of incubation, TCR-T cells were ready for expression analysis or coculture assays. TCR-T cells were maintained in culture with CTL medium over the course of described time, and the medium was refreshed every 3-4 days.\n\nSubcutaneous tumor inoculation\nU87 TMG cell line was prepared at a concentration of 4 x 10^6 cells/ml in PBS-Matrigel mixture at a ratio of 1:1. Immunodeficient mice were shaved at the flank site before injecting subcutaneously 200 μl of prepared tumor cells (8 x 10^5 cells) with 27G needle slowly and steadily into it. Tumor-bearing mice were then monitored routinely for tumor-related symptoms and measured for tumor growth with a caliper. Upon termination criteria or the experimental endpoint, mice were sacrificed with anesthesia and organs of interest were harvested for downstream analyses.\n\nIntracranial tumor inoculation\nU87 TMG cell line was resuspended at a concentration of 75 x 10^6 cells/ml in PBS and 2 μl of which, namely 1.5 x 10^5 cells, was stereotactically implanted into the right hemisphere of immunodeficient mice with the following coordinates: 2 mm right lateral of the bregma and 1 mm anterior to the coronal suture at the depth of 3 mm below the dural surface. A 10-μl Hamilton micro-syringe driven by a fine-step stereotactic device (Stoelting) was employed for injection. The surgery was conducted under anesthesia (Ketamin, 100 mg/kg i.p. and Xylazin, 10 mg/kg i.p.) and analgesia (Carprofen, 5 mg/kg s.c.). Mice continued to receive analgesia for 3 days post surgery and was checked daily for tumor-related symptoms. Upon termination criteria or the experimental endpoint, mice were sacrificed with anesthesia and organs of interest were harvested for downstream analyses.\n\nIntravenous adoptive cell transfer\nTCR-T cells generated as described above were resuspended at a concentration of 25 x 10^6 cells/ml in PBS. Mice were shortly warmed with red-light lamp before intravenously receiving ACT of 5 x 10^6 cells in 200 μl PBS with 27G needle. On the day of and the day after ACT, mice were given 5 x 10^4 units of IL-2 i.p. in 100 μl PBS.\n\nIntracerebroventricular adoptive cell transfer\n5 x 10^6 TCR-T cells were resuspended in 4 μl PBS and stereotactically injected into the cerebral ventricle of the mice with the following coordinates: 0.5 mm left lateral to the bregma at the depth of 1.8 mm below the dural surface. A 10-μl Hamilton micro-syringe driven by a dine step stereotactic device (Stoelting) was employed for injection. The surgery was conducted under anesthesia (Ketamin, 100 mg/kg i.p. and Xylazin, 10 mg/kg i.p.) and analgesia (Carprofen, 5 mg/kg s.c.). Mice continued to receive analgesia for 3 days post surgery.\n\nAdoptive cell transfer on IPTO\nUpon the establishment of IPTOs, 150 x 10^3 TCR-T cells were injected in 3 μl with a 10-μl Hamilton micro-syringe. After 3-day incubation, feeder cells were first macroscopically removed with a scalpel, and the remained tumor chunk was processed with tumor dissociation kit (#130-095-929, Miltenyi Biotec).\n\nMagnetic resonance imaging\nMRI was conducted by the small animal imaging core facility at DKFZ with a Bruker BioSpec 3Tesla (Ettlingen, Germany) with Para Vision software 360 V1.1. Mice were anesthetized with 3.5% sevoflurane in air, and the imaging was performed with a T2 TurboRARE sequence: TE = 48 ms, TR = 3350 ms, FOV 20x20 mm, slice thickness 1 mm, averages = 3, Scan Time = 3m21s, echo spacing 12 ms, rare factor 8, slices 20, image size 192 x 192. Tumor volume was assessed by manual segmentation using Bruker Para Vision software 6.0.1.\n\nUntargeted ligandomics\nImmunoprecipitation of HLA class I:peptide complexes was performed as previously described with additional steps for the forced oxidation of methionine using H2O2 and reduction and alkylation of cysteine using tris(2-carboxyethyl)phosphine and iodoacetamide71. Lyophilized peptides were dissolved in 12 μl of 5% ACN in 0.1% TFA and spiked with 0.5μl of 100 fmol/μl Peptide Retention Time Calibration Mixture and 10 fmol/μl JPTRT 11, a subset of peptides from the Retention Time Standardization Kit (JPT), and transferred to QuanRecovery Vials with MaxPeak HPS (Waters, Milford, MA, USA). All samples were analyzed using an UltiMate 3000 RSLCnano system coupled to an Orbitrap Exploris 480 equipped with a FAIMS Pro Interface (Thermo Fisher). For chromatographic separation, peptides were first loaded onto a trapping cartridge (Acclaim PepMap 100 C18 μ-Precolumn, 5μm, 300 μm i.d. x 5 mm, 100 Å; Thermo Fisher) and then eluted and separated using a nanoEase M/Z Peptide BEH C18 130A 1.7μm, 75μm x 200mm (Waters). Total analysis time was 120 min, and separation was performed using a flow rate of 0.3 μl/min with a gradient starting from 1% solvent B (100% ACN, 0.1% TFA) and 99% solvent A (0.1% FA in H2O) for 0.5 min. Concentration of solvent B was increased to 2.5% in 12.5 min, to 28.6% in 87 min and then to 38.7% in 1.4 min. Subsequently, concentration of solvent B was increased to 80% in 2.6 min and kept at 80% solvent B for 5 min for washing. Finally, the column was re-equilibrated at 1% solvent B for 11 min. The LC system was coupled on-line to the mass spectrometer using a Nanospray-Flex ion source (Thermo Fisher), a SimpleLink Uno liquid junction (FossillOnTech) and a CoAnn ESI Emitter (Fused Silica 20 μm ID, 365 μm OD with orifice ID 10 μm; CoAnn Technologies). The mass spectrometer was operated in positive mode and a spray voltage of 2400 V was applied for ionization with an ion transfer tube temperature of 275 °C. For ion mobility separation, the FAIMS module was operated with standard resolution and a total carrier gas flow of 4.0 l/min. Each sample was injected twice using either a compensation voltage of -50 V or -65 V for maximal orthogonality and thus increased immunopeptidome coverage. Full Scan MS spectra were acquired for a range of 300 – 1650 m/z with a resolution of 120,000 (RF Lens 50%, AGC Target 300%). MS/MS spectra were acquired in data-independent mode using 44 previously determined dynamic mass windows optimized for HLA class I peptides with an overlap of 0.5 m/z. HCD collision energy was set to 28% and MS/MS spectra were recorded with a resolution of 30,000 (normalized AGC target 3000%). MS raw data was analyzed using Spectronaut software (version 17.6, Biognosys)72, and searched against the UniProtKB/Swiss-Prot database (retrieved: 21.10.2021, 20387 entries). Search parameters were set to non-specific digestion and a peptide length of 7 -15 amino acids. Carbamidomethylation of cysteine and oxidation of methionine were included as variable modifications. Results were reported with 1% FDR at the peptide level. Peptides identified by Spectronaut were further analyzed using NetMHCpan 4.173. Predicted non-binders were excluded from the analysis.\n\nRNAscope™\nTarget RNA transcripts on slides were detected with RNAscope 2.5 HD Duplex Kit (#322435, Advanced Cell Diagnostics) following manufacture instructions. Briefly, slides with sectioned frozen murine brains were removed from -80°C and immediately fixated with pre-chilled 10% neutral buffered formalin at 4°C. To dehydrate the slides, they were then serially immersed in\n50%, 70% and 100% EtOH at RT. After another immersion in fresh 100% EtOH, slides were either immediately used or stored in 100% EtOH at -20°C for up to a week. To stain the slides, they were first air-dried and pre-treated with RNAscope Hydrogen Peroxide. After a quick wash with PBS, they were then pre-treated with RNAscope Protease IV. Slides were subsequently quickly rinsed with PBS. For target detection, custom probes were manufactured, and the probe mix was applied on slides and incubated in HybEZ Oven for 2h at 40°C. Next, the slides were washed with 1X Wash Buffer and kept in 5X SSC at RT o/n. On the next day, slides were washed with 1X Wash Buffer and underwent a series of amplification steps with Amp 1-6 following the manual. First probe was then detected with FastRed. After more washes with 1X Wash Buffer, slides underwent another series of amplification steps with Amp 7-10 following the instructions. Second probe was then detected with FastGreen. After more washes with 1X Wash Buffer, slides were counterstained with 50% Hematoxylin staining solution. Next, slides were immediately rinsed with tap water, followed by drying in HybEZ Oven at 60°C. Once the slides cooled down, they were briefly dipped in fresh Xylene and mounted with VectaMount Mounting Medium (#H-5000, Vector Labs).\n\nARDitox off-target prediction\nOff-targets were predicted as described previously46. Briefly, 9-mer peptides of human proteome that share at least 5 amino acids with the target epitope were shortlisted. Next, epitopes derived from frequent single nucleotide polymorphisms with a frequency more than 1% were included. High-affinity presented epitopes predicted with ARDisplay were selected74. Safety score compared the physico-chemical properties of probable TCR-facing amino acids.\n\nTCGA and publicly available dataset analysis\nTCGA data were downloaded through the package TCGAbiolinks (2.28.4)75. Sc GB and glioma datasets were downloaded according to the instructions40,41. The data were handled with Seurat (5.0.3)76. Cell state scores were directly used if specified in the dataset; otherwise, they were calculated based on the defined gene sets with AddModuleScore function41. GSC scores were calculated with the published gene set38.\n\nSingle-cell RNA-seq and analysis\nIsolated primary tumor cells and non-tumor cells from brain tumor samples and IPTO cells were resuspended in 0.04% BSA in PBS; Up to 20 x 10^3 cells were loaded for 5’ single cell sequencing (#1000263, 10x Genomics), and the libraries were prepared according to the manufacturer protocol. Single-cell data were aligned with cellranger (7.0.0) and handled with Seurat (5.0.3). Cells expressing few transcripts or genes were excluded before normalization. Doublets were identified and excluded with scDblFinder (1.14.0)77. Harmony (1.0.3) was subsequently used to integrate datasets78. Canonical markers were employed to identify and annotate the cell type. Module gene sets were derived from previous studies and their scores were calculated using AddModuleScore function in Seurat. Plots were made with ggplot2 (3.5.0) and SCpubr (2.0.2)79,80.\n\nStatistical analysis and Figures\nData are presented as individual values or as mean ± SEM unless stated otherwise. Applied statistical tests are indicated in Figure legends. GraphPad Prism 9.0 was used for statistical tests and plots. Some Figures were created with BioRender.com.\n\nREFERENCES\n1 Hirayama, A. V. et al. High rate of durable complete remission in follicular lymphoma after CD19 CAR-T cell immunotherapy. Blood 134, 636-640, doi:10.1182/blood.2019000905 (2019).\n2 Schuster, S. J. et al. Long-term clinical outcomes of tisagenlecleucel in patients with relapsed or refractory aggressive B-cell lymphomas (JULIET): a multicentre, open-label, single-arm, phase 2 study. Lancet Oncol 22, 1403-1415, doi:10.1016/S1470-2045(21)00375-2 (2021).\n3 Cappell, K. M. & Kochenderfer, J. N. Long-term outcomes following CAR T cell therapy: what we know so far. Nat Rev Clin Oncol 20, 359-371, doi:10.1038/s41571-023-00754-1 (2023).\n4 Albelda, S. M. CAR T cell therapy for patients with solid tumours: key lessons to learn and unlearn. Nat Rev Clin Oncol 21, 47-66, doi:10.1038/s41571-023-00832-4 (2024).\n5 Bausart, M., Preat, V. & Malfanti, A. Immunotherapy for glioblastoma: the promise of combination strategies. J Exp Clin Cancer Res 41, 35, doi:10.1186/s13046-022-02251-2 (2022).\n6 Strickler, J. H., Hanks, B. A. & Khasraw, M. Tumor Mutational Burden as a Predictor of Immunotherapy Response: Is More Always Better? Clin Cancer Res 27, 1236-1241, doi:10.1158/1078-0432.CCR-20-3054 (2021).\n7 Reardon, D. A. et al. Effect of Nivolumab vs Bevacizumab in Patients With Recurrent Glioblastoma: The CheckMate 143 Phase 3 Randomized Clinical Trial. JAMA Oncol 6, 1003-1010, doi:10.1001/jamaoncol.2020.1024 (2020).\n8 Omuro, A. et al. Radiotherapy combined with nivolumab or temozolomide for newly diagnosed glioblastoma with unmethylated MGMT promoter: An international randomized phase III trial. Neuro Oncol 25, 123-134, doi:10.1093/neuonc/noac099 (2023).\n9 Tian, R., Basu, M. K. & Capriotti, E. Computational methods and resources for the interpretation of genomic variants in cancer. BMC Genomics 16 Suppl 8, S7, doi:10.1186/1471-2164-16-S8-S7 (2015).\n10 Wang, G. & Wang, W. Advanced Cell Therapies for Glioblastoma. Front Immunol 13, 904133, doi:10.3389/fimmu.2022.904133 (2022).\n11 Bunse, L., Bunse, T., Kramer, C., Chih, Y. C. & Platten, M. Clinical and Translational Advances in Glioma Immunotherapy. Neurotherapeutics 19, 1799-1817, doi:10.1007/s13311-022-01313-9 (2022).\n12 Choi, B. D. et al. Intraventricular CARv3-TEAM-E T Cells in Recurrent Glioblastoma. N Engl J Med 390, 1290-1298, doi:10.1056/NEJMoa2314390 (2024).\n13 An, Z. et al. Antitumor activity of the third generation EphA2 CAR-T cells against glioblastoma is associated with interferon gamma induced PD-L1. Oncoimmunology 10, 1960728, doi:10.1080/2162402X.2021.1960728 (2021).\n14 Bagley, S. J. et al. Repeated peripheral infusions of anti-EGFRvIII CAR T cells in combination with pembrolizumab show no efficacy in glioblastoma: a phase 1 trial. Nat Cancer, doi:10.1038/s43018-023-00709-6 (2024).\n15 Tsimberidou, A. M. et al. T-cell receptor-based therapy: an innovative therapeutic approach for solid tumors. J Hematol Oncol 14, 102, doi:10.1186/s13045-021-01115-0 (2021).\n16 Wu, L. et al. Targeting CAR to the Peptide-MHC Complex Reveals Distinct Signaling Compared to That of TCR in a Jurkat T Cell Model. Cancers (Basel) 13, doi:10.3390/cancers13040867 (2021).\n17 Barden, M. et al. CAR and TCR form individual signaling synapses and do not cross-activate, however, can co-operate in T cell activation. Front Immunol 14, 1110482, doi:10.3389/fimmu.2023.1110482 (2023).\n18 Baulu, E., Gardet, C., Chuvin, N. & Depil, S. TCR-engineered T cell therapy in solid tumors: State of the art and perspectives. Sci Adv 9, eadf3700, doi:10.1126/sciadv.adf3700 (2023).\n19 Kramer, C. et al. NLGN4X TCR transgenic T cells to treat gliomas. Neuro Oncol 26, 266-278, doi:10.1093/neuonc/noad172 (2024).\n20 Kilian, M. et al. T-cell Receptor Therapy Targeting Mutant Capicua Transcriptional Repressor in Experimental Gliomas. Clin Cancer Res 28, 378-389, doi:10.1158/1078-0432.CCR-21-1881 (2022).\n21 Boschert, T. et al. H3K27M neoepitope vaccination in diffuse midline glioma induces B and T cell responses across diverse HLA loci of a recovered patient. Sci Adv 10, eadi9091, doi:10.1126/sciadv.adi9091 (2024).\n22 Platten, M. et al. A vaccine targeting mutant IDH1 in newly diagnosed glioma. Nature 592, 463-468, doi:10.1038/s41586-021-03363-z (2021).\n23 Hilf, N. et al. Actively personalized vaccination trial for newly diagnosed glioblastoma. Nature 565, 240-245, doi:10.1038/s41586-018-0810-y (2019).\n24 Dutoit, V. et al. Exploiting the glioblastoma peptidome to discover novel tumour-associated antigens for immunotherapy. Brain 135, 1042-1054, doi:10.1093/brain/aws042 (2012).\n25 Eill, G. J., Sinha, A., Morawski, M., Viapiano, M. S. & Matthews, R. T. The protein tyrosine phosphatase RPTPzeta/phosphacan is critical for perineuronal net structure. J Biol Chem 295, 955-968, doi:10.1074/jbc.RA119.010830 (2020).\n26 Johnson, K. G. & Van Vactor, D. Receptor protein tyrosine phosphatases in nervous system development. Physiol Rev 83, 1-24, doi:10.1152/physrev.00016.2002 (2003).\n27 Maeda, N., Nishiwaki, T., Shintani, T., Hamanaka, H. & Noda, M. 6B4 proteoglycan/phosphacan, an extracellular variant of receptor-like protein-tyrosine phosphatase zeta/RPTPbeta, binds pleiotrophin/heparin-binding growth-associated molecule (HB-GAM). J Biol Chem 271, 21446-21452, doi:10.1074/jbc.271.35.21446 (1996).\n28 Perez-Pinera, P., Zhang, W., Chang, Y., Vega, J. A. & Deuel, T. F. Anaplastic lymphoma kinase is activated through the pleiotrophin/receptor protein-tyrosine phosphatase beta/zeta signaling pathway: an alternative mechanism of receptor tyrosine kinase activation. J Biol Chem 282, 28683-28690, doi:10.1074/jbc.M704505200 (2007).\n29 Kuboyama, K. et al. Protein tyrosine phosphatase receptor type z negatively regulates oligodendrocyte differentiation and myelination. PLoS One 7, e48797, doi:10.1371/journal.pone.0048797 (2012).\n30 Ulbricht, U., Eckerich, C., Fillbrandt, R., Westphal, M. & Lamszus, K. RNA interference targeting protein tyrosine phosphatase zeta/receptor-type protein tyrosine phosphatase beta suppresses glioblastoma growth in vitro and in vivo. J Neurochem 98, 1497-1506, doi:10.1111/j.1471-4159.2006.04022.x (2006).\n31 Bourgonje, A. M. et al. Intracellular and extracellular domains of protein tyrosine phosphatase PTRZ-B differentially regulate glioma cell growth and motility. Oncotarget 5, 8690-8702, doi:10.18632/oncotarget.2366 (2014).\n32 Qin, E. Y. et al. Neural Precursor-Derived Pleiotrophin Mediates Subventricular Zone Invasion by Glioma. Cell 170, 845-859.e819, doi:10.1016/j.cell.2017.07.016 (2017).\n33 Papadimitriou, E. et al. Pleiotrophin and its receptor protein tyrosine phosphatase beta/zeta as regulators of angiogenesis and cancer. Biochim Biophys Acta 1866, 252-265, doi:10.1016/j.bbcan.2016.09.007 (2016).\n34 Lacore, M. G. et al. The Glycoprotein M6a Is Associated with Invasiveness and Radioresistance of Glioblastoma Stem Cells. Cells 11, doi:10.3390/cells11142128 (2022).\n35 Fujikawa, A. et al. Targeting PTPRZ inhibits stem cell-like properties and tumorigenicity in glioblastoma cells. Sci Rep 7, 5609, doi:10.1038/s41598-017-05931-8 (2017).\n36 Shi, Y. et al. Tumour-associated macrophages secrete pleiotrophin to promote PTPRZ1 signalling in glioblastoma stem cells for tumour growth. Nat Commun 8, 15080, doi:10.1038/ncomms15080 (2017).\n37 Yang, C. et al. Slow-Cycling Cells in Glioblastoma: A Specific Population in the Cellular Mosaic of Cancer Stem Cells. Cancers (Basel) 14, doi:10.3390/cancers14051126 (2022).\n38 Patel, A. P. et al. Single-cell RNA-seq highlights intratumoral heterogeneity in primary glioblastoma. Science 344, 1396-1401, doi:10.1126/science.1254257 (2014).\n39 Carter, S. L. et al. Absolute quantification of somatic DNA alterations in human cancer. Nat Biotechnol 30, 413-421, doi:10.1038/nbt.2203 (2012).\n40 Venteicher, A. S. et al. Decoupling genetics, lineages, and microenvironment in IDH-mutant gliomas by single-cell RNA-seq. Science 355, doi:10.1126/science.aai8478 (2017).\n41 Neftel, C. et al. An Integrative Model of Cellular States, Plasticity, and Genetics for Glioblastoma. Cell 178, 835-849 e821, doi:10.1016/j.cell.2019.06.024 (2019).\n42 Bozza, M. et al. A nonviral, nonintegrating DNA nanovector platform for the safe, rapid, and persistent manufacture of recombinant T cells. Sci Adv 7, doi:10.1126/sciadv.abf1333 (2021).\n43 Cohen, C. J., Zhao, Y., Zheng, Z., Rosenberg, S. A. & Morgan, R. A. Enhanced antitumor activity of murine-human hybrid T-cell receptor (TCR) in human lymphocytes is associated with improved pairing and TCR/CD3 stability. Cancer Res 66, 8878-8886, doi:10.1158/0008-5472.CAN-06-1450 (2006).\n44 Viner, N. J., Nelson, C. A., Deck, B. & Unanue, E. R. Complexes generated by the binding of free peptides to class II MHC molecules are antigenically diverse compared with those generated by intracellular processing. J Immunol 156, 2365-2368 (1996).\n45 Migliorini, D. et al. Phase I/II trial testing safety and immunogenicity of the multipeptide IMA950/poly-ICLC vaccine in newly diagnosed adult malignant astrocytoma patients. Neuro Oncol 21, 923-933, doi:10.1093/neuonc/noz040 (2019).\n46 Pienkowski, V. M. et al. ARDitox: platform for the prediction of TCRs potential off-target binding. bioRxiv, 2023.2004.2011.536336, doi:10.1101/2023.04.11.536336 (2023).\n47 Meyran, D. et al. T(STEM)-like CAR-T cells exhibit improved persistence and tumor control compared with conventional CAR-T cells in preclinical models. Sci Transl Med 15, eabk1900, doi:10.1126/scitranslmed.abk1900 (2023).\n48 Melenhorst, J. J. et al. Decade-long leukaemia remissions with persistence of CD4(+) CAR T cells. Nature 602, 503-509, doi:10.1038/s41586-021-04390-6 (2022).\n49 McLellan, A. D. & Ali Hosseini Rad, S. M. Chimeric antigen receptor T cell persistence and memory cell formation. Immunol Cell Biol 97, 664-674, doi:10.1111/imcb.12254 (2019).\n50 Hinrichs, C. S. et al. Human effector CD8+ T cells derived from naive rather than memory subsets possess superior traits for adoptive immunotherapy. Blood **117**, 808-814, doi:10.1182/blood-2010-05-286286 (2011).\n51 Berger, C. et al. Adoptive transfer of effector CD8+ T cells derived from central memory cells establishes persistent T cell memory in primates. *J Clin Invest* **118**, 294-305, doi:10.1172/JCI32103 (2008).\n52 Phan, T. G. & Croucher, P. I. The dormant cancer cell life cycle. *Nat Rev Cancer* **20**, 398-411, doi:10.1038/s41568-020-0263-0 (2020).\n53 Min, H. Y. & Lee, H. Y. Cellular Dormancy in Cancer: Mechanisms and Potential Targeting Strategies. *Cancer Res Treat* **55**, 720-736, doi:10.4143/crt.2023.468 (2023).\n54 Jhunjhunwala, S., Hammer, C. & Delamarre, L. Antigen presentation in cancer: insights into tumour immunogenicity and immune evasion. *Nat Rev Cancer* **21**, 298-312, doi:10.1038/s41568-021-00339-z (2021).\n55 Dhatchinamoorthy, K., Colbert, J. D. & Rock, K. L. Cancer Immune Evasion Through Loss of MHC Class I Antigen Presentation. *Front Immunol* **12**, 636568, doi:10.3389/fimmu.2021.636568 (2021).\n56 Ratliff, M. et al. Patient-Derived Tumor Organoids for Guidance of Personalized Drug Therapies in Recurrent Glioblastoma. *Int J Mol Sci* **23**, doi:10.3390/ijms23126572 (2022).\n57 Hoang-Minh, L. B. et al. Infiltrative and drug-resistant slow-cycling cells support metabolic heterogeneity in glioblastoma. *EMBO J* **37**, doi:10.15252/embj.201798772 (2018).\n58 Deleyrolle, L. P. et al. Evidence for label-retaining tumour-initiating cells in human glioblastoma. *Brain* **134**, 1331-1343, doi:10.1093/brain/awr081 (2011).\n59 Xia, Z. et al. The Expression, Functions, Interactions and Prognostic Values of PTPRZ1: A Review and Bioinformatic Analysis. *Journal of Cancer* **10**, 1663-1674, doi:10.7150/jca.28231 (2019).\n60 Martinez Bedoya, D. et al. PTPRZ1-targeting RNA CAR-T cells exert antigen-specific and bystander antitumor activity in glioblastoma. *bioRxiv*, 2023.2012.2023.573190, doi:10.1101/2023.12.23.573190 (2023).\n61 Rosa-Fernandes, L., Oba-Shinjo, S. M., Macedo-da-Silva, J., Marie, S. K. N. & Palmisano, G. Aberrant Protein Glycosylation in Brain Cancers, with Emphasis on Glioblastoma. *Adv Exp Med Biol* **1382**, 39-70, doi:10.1007/978-3-031-05460-0_4 (2022).\n62 Tokumura, K. et al. The Bioinformatics Identification of Potential Protein Glycosylation Genes Associated with a Glioma Stem Cell Signature. *BioMedInformatics* **4**, 75-88 (2024).\n63 Heard, A. et al. Antigen glycosylation regulates efficacy of CAR T cells targeting CD19. *Nat Commun* **13**, 3367, doi:10.1038/s41467-022-31035-7 (2022).\n64 Espinosa-Carrasco, G. et al. Intratumoral immune triads are required for immunotherapy-mediated elimination of solid tumors. *Cancer Cell*, doi:10.1016/j.ccell.2024.05.025 (2024).\n65 Bove, C. et al. CD4 CAR-T cells targeting CD19 play a key role in exacerbating cytokine release syndrome, while maintaining long-term responses. *J Immunother Cancer* **11**, doi:10.1136/jitc-2022-005878 (2023).\n66 Bunse, L. et al. P06.05.A INTRACEREBROVENTRICULAR TCR-TRANSGENIC T CELL THERAPY TO TREAT GLIOBLASTOMA (INVENT4GB). *Neuro-Oncology* **25**, ii46-ii46, doi:10.1093/neuonc/noad137.145 (2023).\n67 Rosskopf, S. et al. A Jurkat 76 based triple parameter reporter system to evaluate TCR functions and adoptive T cell strategies. Oncotarget 9, 17608-17619, doi:10.18632/oncotarget.24807 (2018).\n68 Lancaster, M. A. & Knoblich, J. A. Generation of cerebral organoids from human pluripotent stem cells. Nature Protocols 9, 2329-2340, doi:10.1038/nprot.2014.158 (2014).\n69 Lancaster, M. A. et al. Guided self-organization and cortical plate formation in human brain organoids. Nature Biotechnology 35, 659-666, doi:10.1038/nbt.3906 (2017).\n70 Sauer, T. et al. CD70-specific CAR T cells have potent activity against acute myeloid leukemia without HSC toxicity. Blood 138, 318-330, doi:10.1182/blood.2020008221 (2021).\n71 Chong, C. et al. High-throughput and Sensitive Immunopeptidomics Platform Reveals Profound Interferongamma-Mediated Remodeling of the Human Leukocyte Antigen (HLA) Ligandome. Mol Cell Proteomics 17, 533-548, doi:10.1074/mcp.TIR117.000383 (2018).\n72 Bruderer, R. et al. Extending the limits of quantitative proteome profiling with data-independent acquisition and application to acetaminophen-treated three-dimensional liver microtissues. Mol Cell Proteomics 14, 1400-1410, doi:10.1074/mcp.M114.044305 (2015).\n73 Reynisson, B., Alvarez, B., Paul, S., Peters, B. & Nielsen, M. NetMHCpan-4.1 and NetMHCIIpan-4.0: improved predictions of MHC antigen presentation by concurrent motif deconvolution and integration of MS MHC eluted ligand data. Nucleic Acids Res 48, W449-W454, doi:10.1093/nar/gkaa379 (2020).\n74 Mazzocco, G. et al. AI Aided Design of Epitope-Based Vaccine for the Induction of Cellular Immune Responses Against SARS-CoV-2. Front Genet 12, 602196, doi:10.3389/fgene.2021.602196 (2021).\n75 Colaprico, A. et al. TCGAbiolinks: an R/Bioconductor package for integrative analysis of TCGA data. Nucleic Acids Res 44, e71, doi:10.1093/nar/gkv1507 (2016).\n76 Hao, Y. et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat Biotechnol 42, 293-304, doi:10.1038/s41587-023-01767-y (2024).\n77 Germain, P. L., Lun, A., Garcia Meixide, C., Macnair, W. & Robinson, M. D. Doublet identification in single-cell sequencing data using scDbIfinder. F1000Res 10, 979, doi:10.12688/f1000research.73600.2 (2021).\n78 Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat Methods 16, 1289-1296, doi:10.1038/s41592-019-0619-0 (2019).\n79 Wickham, H. ggplot2: Elegant Graphics for Data Analysis. (Springer-Verlag New York, 2016).\n80 Blanco-Carmona, E. Generating publication ready visualizations for Single Cell transcriptomics using SCpubr. bioRxiv, 2022.2002.2028.482303, doi:10.1101/2022.02.28.482303 (2022).\n\nAuthor contributions\nY.C.C. conceptualized the study, designed and performed experiments, analyzed and interpreted data, and wrote the manuscript. A.C.D., D.A.A., A.D.R., A.K., M.K. and C.K. performed in vitro and in vivo experiments and edited the manuscript. X.M. and H-K.L. established and processed organoids and edited the manuscript. P.K. and B.Z. analyzed patient and organoid scRNA-seq data and edited the manuscript. A.K.S. and F.S. collected GB histology samples and edited the manuscript. M.S., H.B., T.B., R.H. and W.W. interpreted data and edited the manuscript. A.B. and V.M.P. predicted off-targets. K.A., J.P.B. and A.B.R. performed the ligandomics analysis and edited the manuscript. H.F. and N.E. collected patient\ntumor samples and edited the manuscript. M.R. established primary GB cell lines and edited the manuscript. M.P. supervised the study, interpreted the data, provided funding and edited the manuscript. E.W.G. provided cloning methodology, retrieved the TCR and edited the manuscript. L.B. conceptualized and supervised the study, interpreted the data, and wrote the manuscript.\n\nAcknowledgements and funding\nWe acknowledge the support of the DKFZ Light Microscopy Facility, Genomics and Proteomics Core Facility, Center for Preclinical Research, Flow Cytometry Core Facility and single-cell Open Lab. We also would like to highlight and acknowledge the technical support from Viktoria Sachs, Chris Pagel, Kristine Jähne and Rebecca Köhler. This study was supported by grants from the Swiss Cancer Foundation (Swiss Bridge Award), the Else Kröner Fresenius Foundation (2019_EKMS.49), the University Heidelberg Foundation (Hella Bühler Award), the DFG (German Research Foundation), project 404521405 (SFB1389 UNITE Glioblastoma B03), the DKFZ Hector institute (T-SIRE), the Hertie Foundation, and the University of Heidelberg, ExploreTech! Grant to L.B.; the DKTK Joint Funding AMI2GO, the Rolf Schwiete Foundation (2021-009) to M.P. and L.B.; the HI-TRON strategy project PACESSETTING and the DKTK Joint Funding Program INNOVATION INVENT4GB to W.W., M.P., and L.B. The DFG, project 404521405 (SFB1389 UNITE Glioblastoma B01) to T.B. and M.P. Y-C.C. received an international fellowship from the Helmholtz International Graduate School at DKFZ and was supported by RTG 2099 and UNITE SFB1389. A.C.D. was supported by Mildred-Scheel scholarship. P.K. and D.A.A. were funded by RTG 2099 and UNITE SFB1389.\n\nCompeting interests\nThe identified reactive TCR was patented by Y-C.C., M.K., C.K., E.W.G., W.W., M.P. and L.B., “T cell receptor derived binding polypeptides” (WO 2023/213904). E.W.G. and M.P. are founders of TCellTech.\nFIGURES\n\nFig. 1. PTPRZ1 is upregulated in GB, particularly in GSCs and AC-like GB.\n(A) *PTPRZ1* expression in primary tumor compared to adjacent normal tissues in TCGA-GB dataset with t-test. (B) *PTPRZ1* expression in single cells of GB from previously published dataset\\(^{41}\\). Cell type annotation is in Fig. S1D. (C) UMAP of scRNA-sequenced glioma and GB samples from Fig. S1I with cell type annotation. (D) Gene expression of *PTPRZ1* for (C). More cell type-defining genes are shown in Fig. S1J. (E) & (F) Protein expression of PTPRZ1 and GFAP from 20 primary and 20 recurrent GB matched samples. Each dot is the average of all tumor pieces of a patient. (G) Correlation of PTPRZ1\\(^+\\) and GFAP\\(^+\\) cell frequencies. Each dot represents a tumor piece. (H) Untargeted ligandomics of primary GB cell lines with overlapping of PTPRZ1 ligands across primary GB cell line. The highlighted overlapped peptide is PTPRZ1\\(^{1814-1822}\\). (I) & (J) Cellular states of cancer cells in (C) and *PTPRZ1* expression. (K) & (L) *PTPRZ1* expression across distinct cellular states and its correlation with GSC score. (K) was analyzed with one-way ANOVA multiple comparison corrected with Holm-Šidák method. (L) was analyzed with Spearman correlation.\nFig. 2. A PTPRZ1-reactive TCR was identified from a vaccinated GB patient.\n(A) Schematic workflow depicting the sorting and sequencing of PTPRZ1-reactive T cells with CDR3 frequency plots showing dominant TCRs. The later identified reactive TCR is highlighted. (B) Validation of dominant TCR clonotypes from (A) using Jurkat cells transfected with a TCR and a reporter plasmid (top) or Jurkat reporter cells transfected with a TCR plasmid (bottom). (C) Luminescence reporter signal of TCR-Jurkat cells upon overnight coculture with peptide-loaded target cells. (D) Luminescence reporter signal following overnight coculture with peptide-loaded or antigen-expressing target cells. (E) in silico AI-predicted off-targets of the identified PTPRZ1-reactive TCR with presentation probability indicated in blue circles and safety score indicated in red squares. Presentation probability predicts whether the peptide is presented on MHC, and lower safety score denotes a higher likelihood of cross-reactivity. (F) Fluorescence reporter signal of TCR-Jurkat cells upon overnight coculture with various peptide-loaded target cells.\nFig. 3. PTPRZ1^{1814-1822} TCR-T product was persistent, T_{SCM}-abundant and efficient in target cell killing.\n(A) TCR expression on primary human T cells over 3-week period. (B) Percentage of CD4^{+} and CD8^{+} T cells in TCR-T cell products over 3 weeks. (C) CD8^{+} T cell subsets on the day of PBMC isolation and after transduction with longitudinal monitoring on the right. (D) Percentage of surface activation marker-positive, CD137^{+}, cells in CD8^{+} TCR-T cells upon 24h coculture with target cells. (E) Percentage of effector protein-positive, Granzyme B^{+}, cells in CD8^{+} TCR-T cells upon 24h coculture with target cells. (F) Detection of LDH released into the medium after 24h coculture of TCR-T cells and target cells. (G) Titration of E:T ratio with TCR-T cells and target cells, measured with LDH release. 1 unit is 75 x 10^{3} cells. Specific OD is OD_{PTPRZ1}-OD_{Flu}. (H) Schematic of isolating CD4^{+} and CD8^{+} TCR-T cells with MACS and subjecting them to various cocultures. (I) Cytotoxicity upon serial diluting CD4^{+} T cells. (J) Cytotoxicity upon serial diluting CD8^{+} T cells. (I) and (J) were measured with LDH release, and 1 unit equals to 75 x 10^{3} cells in. (K) Cytotoxicity of coculture using reactive CD8^{+} T cells with the target cell and either reactive or non-reactive CD4^{+} T cells, measured with cell counting through flow cytometry. (D)-(F) were analyzed with two-way ANOVA multiple comparison corrected with Holm-Šidák method. (K) was analyzed with paired t-test.\nFig. 4. Intravenous and intracerebroventricular delivery of PTPRZ1-TCR-T is efficacious in experimental flank and brain tumors\n\n(A) Workflow of i.v. ACT on s.c. tumor model. (B) Longitudinal s.c. tumor growth monitoring. (C) Workflow of i.v. and i.cv. ACT on i.c. tumor model. (D) i.c. tumor imaging with preclinical MRI. (E) Longitudinal monitoring of i.c. tumor size with MRI and assessment of radiographic response upon ACT treatment. (F) T cell engraftment validation with cheek blood after ACT, measured with cell counting through flow cytometry. (G) Contour plot of CD8+ TCR-T cell subsets of transferred T cells in the host. (H) Percentages of different CD8+ T cell subsets of transferred T cells. (I) Overall survival of i.c. tumor-bearing mice treated with i.v. and i.cv. ACT. (J) Immunofluorescent staining of grafted i.c. tumor cells, MHC I+, and transferred T cells, CD3+ MHC I+. (K) Transferred T cell numbers in tumoral and non-tumoral tissues in (J) upon ACT treatment. (L) RNAscope™ identifying TMG expression in i.c. tumors. (M) H score analysis of TMG expression in (L). (B) and (E) were analyzed with nonlinear regression. (I) was analyzed with Log-rank test. (F) and (K) were analyzed with one-way and two-way ANOVA multiple comparison, respectively, corrected with Holm-Šidák method. (M) was analyzed with nonparametric t-test.\nFig. 5. PTPRZ1^{1814-1822} TCR-T cells broadly lyse HLA-A*02+ GB primary cells, particularly stem-like SCCs.\n(A) Cytotoxicity measured with LDH release of primary cell line upon coculture with TCR-T cells. (B) Experimental design to assess preferential killing of TCR-T cells on dye-retaining SCCs or dye-losing FCCs. (C) Assessment of specific lysis of SCCs and FCCs upon short-term, 5h, or long-term, 24h, coculture with TCR-T cells, measured with cell counting by flow cytometry. (D) HLA-A2 typing of established GB primary cell lines. (E) Assessment of specific lysis by PTPRZ1^{1814-1822} TCR-T cells normalized to Flu TCR-T cell-treated HLA-A*02+ or HLA-A*02- tumors. (F) & (G) Activation of CD8+ TCR-T cells after 24h coculture with various GB primary cell lines that are HLA-A*02+ or HLA-A*02-. (H) Activation of CD4+ TCR-T cells after 24h coculture with GB primary cell lines. All analyses were performed with two-way ANOVA multiple comparison corrected with Holm-Šidák method.\nFig. 6. IPTOs reveal predominant targeting of AC-like cells and GSCs by PTPRZ1^{1814-1822} TCR-T cells.\n(A) IPTO generation from 3 HLA-A*02+ GB samples and treatment with TCR-T cells. (B) UMAP of malignant cells and T cells derived from Fig. S11A with their marker expression on the right. More canonical markers are visualized in Fig. S11B. (C) Frequency of cell content in IPTOs upon TCR-T cell treatment. (D) *PTPRZ1* expression after TCR-T cell treatment. (E) Cellular states of generated IPTOs with *PTPRZ1* expression. (F) & (G) AC-like module score and frequency of distinct cellular states after ACT. (H) & (I) Correlation of *PTPRZ1* expression and GSC module score in malignant cells in IPTOs and frequency of GSC upon TCR-T cell treatment. Cells score over 0.15 GSC module score are defined as GSCs. (D) & (F) were analyzed with one-way ANOVA multiple comparison corrected with Holm-Šídák method. (C), (G) & (I) were analyzed with Fisher’s exact test. (H) were analyzed with spearman correlation.\nFigures\n\nFigure 1\n\nPTPRZ1 is upregulated in GB, particularly in GSCs and AC-like GB. (A) PTPRZ1 expression in primary tumor compared to adjacent normal tissues in TCGA-GB dataset with t-test. (B) PTPRZ1 expression in single cells of GB from previously published dataset41. Cell type annotation is in Fig. S1D. (C) UMAP of scRNA-sequenced glioma and GB samples from Fig. S1I with cell type annotation. (D) Gene expression of PTPRZ1 for (C). More cell type-defining genes are shown in Fig. S1J. (E) & (F) Protein expression of PTPRZ1 and GFAP from 20 primary and 20 recurrent GB matched samples. Each dot is the average of all tumor pieces of a patient. (G) Correlation of PTPRZ1+ and GFAP+ cell frequencies. Each dot represents a tumor piece. (H) Untargeted ligandomics of primary GB cell lines with overlapping of PTPRZ1 ligands across primary GB cell line.\nThe highlighted overlapped peptide is PTPRZ11814-1822. (I) & (J) Cellular states of cancer cells in (C) and PTPRZ1 expression. (K) & (L) PTPRZ1 expression across distinct cellular states and its correlation with GSC score. (K) was analyzed with one-way ANOVA multiple comparison corrected with Holm–Šídák method. (L) was analyzed with Spearman correlation.\n\nFigure 2\n\nA PTPRZ1-reactive TCR was identified from a vaccinated GB patient. (A) Schematic workflow depicting the sorting and sequencing of PTPRZ1-reactive T cells with CDR3 frequency plots showing dominant TCRs. The later identified reactive TCR is highlighted. (B) Validation of dominant TCR clonotypes from (A) using Jurkat cells transfected with a TCR and a reporter plasmid (top) or Jurkat reporter cells transfected with a TCR plasmid (bottom). (C) Luminescence reporter signal of TCR-Jurkat cells upon overnight coculture with peptide-loaded target cells. (D) Luminescence reporter signal following overnight coculture with peptide-loaded or antigen-expressing target cells. (E) in silico AI-predicted off-targets of the identified PTPRZ1-reactive TCR with presentation probability indicated in blue circles and safety score indicated in red squares. Presentation probability predicts whether the peptide is presented on MHC, and lower safety score denotes a higher likelihood of cross-reactivity. (F) Fluorescence reporter signal of TCR-Jurkat cells upon overnight coculture with various peptide-loaded target cells.\nFigure 3\n\nPTPRZ11814-1822 TCR-T product was persistent, TSCM-abundant and efficient in target cell killing. (A) TCR expression on primary human T cells over 3-week period. (B) Percentage of CD4+ and CD8+ T cells in TCR-T cell products over 3 weeks. (C) CD8+ T cell subsets on the day of PBMC isolation and after transduction with longitudinal monitoring on the right. (D) Percentage of surface activation marker-positive, CD137+, cells in CD8+ TCR-T cells upon 24h coculture with target cells. (E) Percentage of effector protein-positive, Granzyme B+, cells in CD8+ TCR-T cells upon 24h coculture with target cells. (F) Detection of LDH released into the medium after 24h coculture of TCR-T cells and target cells. (G) Titration of E:T ratio with TCR-T cells and target cells, measured with LDH release. 1 unit is 75 x 103 cells. Specific OD is ODPTPRZ1-ODFlu. (H) Schematic of isolating CD4+ and CD8+ TCR-T cells with MACS and subjecting them to various cocultures. (I) Cytotoxicity upon serial diluting CD4+ T cells. (J) Cytotoxicity upon serial diluting CD8+ T cells. (I) and (J) were measured with LDH release, and 1 unit equals to 75 x 103 cells in.\n(K) Cytotoxicity of coculture using reactive CD8+ T cells with the target cell and either reactive or non-reactive CD4+ T cells, measured with cell counting through flow cytometry. (D)-(F) were analyzed with two-way ANOVA multiple comparison corrected with Holm-Šídák method. (K) was analyzed with paired t-test.\n\nFigure 4\n\nIntravenous and intracerebroventricular delivery of PTPRZ1-TCR-T is efficacious in experimental flank and brain tumors (A) Workflow of i.v. ACT on s.c. tumor model. (B) Longitudinal s.c. tumor growth monitoring. (C) Workflow of i.v. and i.cv. ACT on i.c. tumor model. (D) i.c. tumor imaging with preclinical MRI. (E) Longitudinal monitoring of i.c. tumor size with MRI and assessment of radiographic response upon ACT treatment. (F) T cell engraftment validation with cheek blood after ACT, measured with cell counting through flow cytometry. (G) Contour plot of CD8+ TCR-T cell subsets of transferred T cells in the host. (H) Percentages of different CD8+ T cell subsets of transferred T cells. (I) Overall survival of i.c. tumor-\nbearing mice treated with i.v. and i.cv. ACT. (J) Immunofluorescent staining of grafted i.c. tumor cells, MHC I+, and transferred T cells, CD3+ MHC I+. (K) Transferred T cell numbers in tumoral and non-tumoral tissues in (J) upon ACT treatment. (L) RNAscope™ identifying TMG expression in i.c. tumors. (M) H score analysis of TMG expression in (L). (B) and (E) were analyzed with nonlinear regression. (I) was analyzed with Log-rank test. (F) and (K) were analyzed with one-way and two-way ANOVA multiple comparison, respectively, corrected with Holm–Šidák method. (M) was analyzed with nonparametric t-test.\n\nFigure 5\n\nPTPRZ11814-1822 TCR-T cells broadly lyse HLA-A*02+ GB primary cells, particularly stem-like SCCs. (A) Cytotoxicity measured with LDH release of primary cell line upon coculture with TCR-T cells. (B) Experimental design to assess preferential killing of TCR-T cells on dye-retaining SCCs or dye-losing FCCs. (C) Assessment of specific lysis of SCCs and FCCs upon short-term, 5h, or long-term, 24h, coculture with TCR-T cells, measured with cell counting by flow cytometry. (D) HLA-A2 typing of established GB primary cell lines. (E) Assessment of specific lysis by\nPTPRZ11814-1822 TCR-T cells normalized to Flu TCR-T cell-treated HLA-A*02+ or HLA-A*02- tumors. (F) & (G) Activation of CD8+ TCR-T cells after 24h coculture with various GB primary cell lines that are HLA-A*02+ or HLA-A*02-. (H) Activation of CD4+ TCR-T cells after 24h coculture with GB primary cell lines. All analyses were performed with two-way ANOVA multiple comparison corrected with Holm–Sidak method.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nFigure 6\n\nIPTOs reveal predominant targeting of AC-like cells and GSCs by PTPRZ11814-1822 TCR-T cells. (A) IPTO generation from 3 HLA-A*02+ GB samples and treatment with TCR-T cells. (B) UMAP of malignant cells and T cells derived from Fig. S11A with their marker expression on the right. More canonical markers are visualized in Fig. S11B. (C) Frequency of cell content in IPTOs upon TCR-T cell treatment. (D) PTPRZ1 expression after TCR-T cell treatment. (E) Cellular states of generated IPTOs with PTPRZ1 expression. (F) & (G) AC-like module score and frequency of distinct cellular states after ACT. (H) & (I) Correlation of PTPRZ1 expression and GSC module score in malignant cells in IPTOs and frequency of GSC upon TCR-T cell treatment. Cells score over 0.15 GSC module score are defined as GSCs. (D) & (F) were analyzed with one-way\nANOVA multiple comparison corrected with Holm-Šidák method. (C), (G) & (I) were analyzed with Fisher’s exact test. (H) were analyzed with spearman correlation.\n\nSupplementary Files\n\nThis is a list of supplementary files associated with this preprint. Click to download.\n\n• SFigure1.pdf\n• SFigure2.pdf\n• SFigure3.pdf\n• SFigure4.pdf\n• SFigure5.pdf\n• SFigure6.pdf\n• SFigure7.pdf\n• SFigure8.pdf\n• SFigure9.pdf\n• SFigure10.pdf\n• SFigure11.pdf\n• TransgenicTcelltherapytargetingtheglioblastomastemcellantigenPTPRZ1withavaccineinducedpatientderivedTcellreceptorSupplementary.pdf"}
{"paper_id": "prb77", "title": "Divergently Access N-center 5,6,7-Perifused Cycle", "text": "# Abstract\n\nNitrogen-containing heterocycles are the key components in many pharmaceuticals and functional materials. In this study, we reported a transition metal-catalyzed high-order reaction sequence for synthesizing a structurally unique N-center 5,6,7-perifused cycle (NCPC). The key characteristics included the formation of a seven-membered ring by the 8π electrocyclization of various alkenes and aromatic heterocycles as π-components, in which metal carbene species generated that further induced the cleavage of the α-C-H or -C-C bond. Specifically, the latter could react with various nucleophilic reagents containing -O, -S, -N, and -C. The stereo-controlled late-stage modification of some complicated pharmaceuticals indicated the versatility of this protocol.\n\n[Physical sciences/Chemistry/Organic chemistry/Synthetic chemistry methodology](/browse?subjectArea=Physical%20sciences%2FChemistry%2FOrganic%20chemistry%2FSynthetic%20chemistry%20methodology) [Physical sciences/Chemistry/Organic chemistry/Reaction mechanisms](/browse?subjectArea=Physical%20sciences%2FChemistry%2FOrganic%20chemistry%2FReaction%20mechanisms) [Physical sciences/Chemistry/Medicinal chemistry/Drug discovery and development](/browse?subjectArea=Physical%20sciences%2FChemistry%2FMedicinal%20chemistry%2FDrug%20discovery%20and%20development)\n\n# Introduction\n\nNatural products are often used as the lead compounds\n \n | Term | \n Definition | \n
\n \n | i | \n Sample, \\( i = 1, \\ldots, n \\). | \n
\n \n | j | \n Taxon, \\( j = 1, \\ldots, d \\). | \n
\n \n | \\( A_{ij} \\) | \n The absolute abundance in the ecosystem for taxon \\( j \\) in sample \\( i \\). | \n
\n \n | \\( O_{ij} \\) | \n The observed abundance for taxon \\( j \\) in sample \\( i \\). | \n
\n \n | \\( S_i \\) | \n Sample-specific sampling fraction. | \n
\n \n | \\( C_j \\) | \n Taxon-specific sampling efficiency. | \n
\n \n | \\( E_{ij} \\) | \n Random error for taxon \\( j \\) in sample \\( i \\). | \n
\n \n | \\( a_{ij} \\) | \n \\( \\log A_{ij} \\). | \n
\n \n | \\( o_{ij} \\) | \n \\( \\log O_{ij} \\). | \n
\n \n | \\( s_i \\) | \n Sample-specific sampling fraction in log scale. | \n
\n \n | \\( c_j \\) | \n Taxon-specific sequencing efficiency in log scale. | \n
\n \n | \\( e_{ij} \\) | \n Random error for taxon \\( j \\) in sample \\( i \\) in log scale. | \n
\n
\n\nTable 1. Definitions of key notations\n\nThe absolute abundances \\( A_{ij} \\) in the ecosystem are non-negative. Let \\( \\mathcal{T} \\) be the set of all possible configurations of presence or absence of a taxon, then\n\n\\[\n\\mathcal{T} = \\left\\{ (1, 1, \\ldots, 1), \\right. \\\\\n\\left. (0, 1, \\ldots, 1), \\ldots, (1, 1, \\ldots, 0) \\right. \\\\\n\\vdots \\\\\n\\left. (0, 0, \\ldots, 1), \\ldots, (1, 0, \\ldots, 0) \\right\\}\n\\]\n\nwhere 0 and 1 correspond to the cases where a taxon is missing (zero) or present in a sample, respectively. We represent each event in \\( \\mathcal{T} \\) by \\( T \\) and it is indexed by \\( t = 1, \\ldots, 2^d - 1 \\). For example, \\( T_1 \\) corresponds to the case where all taxa are present in a sample, and \\( T_{2^{d-1}} \\) is the case where only the first taxon is present, and all remaining taxa are zeros.\n\nWe further assume that for each sample, the pattern of taxon presence is generated by independent r.v. \\( M_{ij}, 1 \\leq i \\leq n, 1 \\leq j \\leq d \\) such that\n\n\\[\nP(M_{i} = T_t) = \\delta_t^j, 0 \\leq \\delta_t^j \\leq 1,\n\\]\n\nwhere \\( M_i = (M_{i1}, \\ldots, M_{id})^T \\).\n\nWe propose that the observed abundance is generated by the following multiplicative model:\n\n\\[\nO_{ij} = S_i C_j A_{ij} E_{ij}.\n\\]\nFor \\( A_{ij} \\) with \\( M_{ij} = 1 \\), i.e., for the non-zero absolute abundance, perform the log transformation, the above multiplicative model can be transformed into an additive model,\n\n\\[\no_{ij} = s_i + c_j + a_{ij} + e_{ij}.\n\\]\n\nAssume that conditional on \\( M_i = 1 \\), the absolute abundance has finite second moment:\n\n\\[\nE(a_{ij}) = \\alpha_j, Var(a_{ij}) = \\sigma^0_{jj}.\n\\]\n\nAdditionally, we assume that:\n\n\\( s_i \\) is a sample-specific constant,\n\\( c_j \\) is a taxon-specific constant,\n\\( e_{ij} \\sim i.i.d.\\ F_e \\) with \\( E(e_{ij}) = 0, Var(e_{ij}) = \\sigma^2_e \\),\n\\( a_{ij} \\perp \\perp e_{ij} \\).\n\nWe define two concepts of correlation coefficients in this section. One quantifies the strength of linear relationship between a pair of taxa by modifying Pearson and Spearman correlation coefficients for sparse high-dimensional compositional data. The second quantifies the strength of nonlinear relationship between a pair of taxa by modifying distance correlation coefficient. We provide both measures.\n\nRemark 1. Since distance correlation coefficient is a measure of nonlinear relationship between a pair of taxa, it does not have a sign and is always non-negative. On the other hand, the measure of linear correlation coefficient described below can take positive or negative values depending upon whether they are positively or negatively linearly correlated.\n\nRemark 2. A distance correlation coefficient of zero between two taxa implies that the taxa are independent. It is well-known that if Pearson or Spearman correlation coefficients between two taxa is zero then it only suggests a lack of linear relationship but not independence.\n\nRemark 3. Our proposed SECOM methodology outputs both linear as well as distance correlation coefficients between pairs of taxa. If the relationship between a pair is nonlinear then we expect the distance correlation between the two taxa to be nonzero whereas the linear correlation could be zero. On the other hand if the relationship is linear then the linear correlation coefficient is expected to be nonzero and sign of the coefficient is positive or negative depending upon whether the two taxa are positively or negatively associated. When the linear correlation coefficient is nonzero then we expect the distance correlation coefficient also to be nonzero. This is because linear correlation coefficient is a special case of nonlinear correlation coefficient. In such cases, we recommend the user to use the value provided by the linear correlation coefficient. However, there may be situations where the estimated linear correlation is nonzero but the distance correlation is zero. This can happen due sample size and power issues and the sparsity constraints. In such cases again we recommend the researcher to use linear correlation coefficient value.\n\nEstimation of linear correlation coefficients\nIn this section we develop methodology for estimating linear correlation coefficient between the abundances of pairs of taxa \\( l \\) and \\( m, 1 \\leq l,m \\leq d \\), while noting that for each subject \\( i \\), the abundances \\( a_{il} \\) and \\( a_{im} \\) are not observable. The only observable values are the observed counts, e.g. OTUs, \\( o_{il} \\) and \\( o_{im} \\). According to our hypothesized model, the actual abundance \\( a_{ij} \\) of the \\( j^{th} \\) taxon in a unit volume of the specimen from the \\( i^{th} \\) subject, in expectation, is a function of observed counts \\( o_{ij} \\), and unknown nuisance parameters \\( s_i \\) and \\( c_j \\). Therefore, to obtain an approximate unbiased estimator of \\( a_{ij} \\) using the observable \\( o_{ij} \\), we need to estimate and nuisance parameters \\( s_i \\) and \\( c_j \\) and eliminate the bias introduced by them. Once that is done, we can define a Pearson or Spearman correlations using the modified observed counts \\( o_{ij} \\) to derived bias-corrected estimates of the true correlation coefficient between the abundance of a pair of taxa. In the following we now describe the methodology to eliminate the nuisance parameters \\( s_i \\) and \\( c_j \\).\n\nDenote the covariance matrix of \\( a_{ij} \\) as \\( \\Sigma^0 = [\\sigma^0_{lm}]_{l,m=1,\\ldots,d} \\), and the Pearson correlation coefficient matrix as \\( R^0 = [\\rho^0_{lm}]_{l,m=1,\\ldots,d} \\). We make the following sparsity assumption regarding the covariance matrix.\n\nAssumption 0.1 (Approximately sparse covariance matrices).\n\n\\[\n\\mathcal{U}(K) = \\{ \\Sigma : \\sigma_{jj} < K, \\frac{1}{d^2} \\sum_{l \\neq m} \\sigma_{lm} = o(1) \\}.\n\\]\nTwo examples of matrices which satisfy this condition are (1) a d-diagonal matrix, and (2) the AR(1) covariance matrix.\n\nFrom (2), (3), (4), and (5), by centering the data suitably, we eliminate the effect of taxon-specific sampling efficiency \\( c_j \\) as follows:\n\n\\[\no_{ij} - \\bar{o}_j = (s_i - \\bar{s}_.) + (a_{ij} - \\bar{a}_j) + (e_{ij} - \\bar{e}_j) = \\textcircled{1} + \\textcircled{2} + \\textcircled{3},\n\\]\n\nwhere, for the \\( j^{th} \\) taxon and \\( i = 1, 2, ..., n \\), \\( \\bar{s}_. = \\frac{1}{n} \\sum_{i=1}^n s_i \\), \\( \\bar{a}_j = \\frac{1}{n} \\sum_{i=1}^n a_{ij} \\), \\( \\bar{e}_j = \\frac{1}{n} \\sum_{i=1}^n e_{ij} \\). Note that:\n\n(1) We call \\( \\textcircled{1} \\) the \"sampling fraction difference\", which is a sample-specific bias term,\n\n(2) For \\( \\textcircled{2} \\): \\( E(a_{ij} - \\bar{a}_j) = 0, Var(a_{ij} - \\bar{a}_j) = (1 - \\frac{1}{n}) \\sigma_{jj}^0 \\),\n\n(3) For \\( \\textcircled{3} \\): \\( E(e_{ij} - \\bar{e}_j) = 0, Var(e_{ij} - \\bar{e}_j) = (1 - \\frac{1}{n}) \\sigma_e^2 \\).\n\nDefine the set of non-zero taxa in sample \\( i \\) by \\( d(i) = \\{ j : M_{ij} = 1, 1 \\leq j \\leq d \\} \\). Then by on Assumption 0.1, we have\n\n\\[\n\\frac{1}{|d(i)|} \\sum_{j \\in d(i)} \\left\\{ (o_{ij} - \\bar{o}_j) - \\textcircled{1} \\right\\} = \\frac{1}{|d(i)|} \\sum_{j \\in d(i)} \\textcircled{2} + \\frac{1}{|d(i)|} \\sum_{j \\in d(i)} \\textcircled{3} \\longrightarrow_p 0 \\text{ as } |d(i)| \\to \\infty.\n\\] (6)\n\nTherefore, a quick and simple estimator of the sampling fraction difference is the arithmetic mean of the difference between observed abundances, i.e. \\( \\textcircled{1} = \\bar{s}_i - \\bar{s}_. = \\frac{1}{|d(i)|} \\sum_{j \\in d(i)} (o_{ij} - \\bar{o}_j) \\). Taking into account the variance of residuals \\( o_{ij} - \\bar{o}_j \\), a weighted average is proposed as follows:\n\n\\[\n\\textcircled{1} = \\bar{s}_i - \\bar{s}_. = \\sum_{j \\in d(i)} w_j (o_{ij} - \\bar{o}_j),\n\\] (7)\n\nwhere \\( w_j = \\frac{1}{Var(o_{ij} - \\bar{o}_j)} / \\sum_{l \\in d(i)} \\frac{1}{Var(o_{lj} - \\bar{o}_l)} \\).\n\nDenote the centered absolute abundance by \\( x_{ij} = a_{ij} - \\bar{a}_j, 1 \\leq i \\leq n, 1 \\leq j \\leq d \\). Then\n\n\\[\nCov(x_{il}, x_{im}) = (1 - \\frac{1}{n}) \\sigma_{lm}^0, \\quad Var(x_{ij}) = (1 - \\frac{1}{n}) \\sigma_{jj}^0, \\quad Corr(x_{il}, x_{im}) = \\rho_{lm}^0.\n\\]\n\nSimilarly, define the centered absolute abundance with noise by\n\n\\[\ny_{ij} := (a_{ij} - \\bar{a}_j) + (e_{ij} - \\bar{e}_j) = x_{ij} + (e_{ij} - \\bar{e}_j) = (o_{ij} - \\bar{o}_j) - (s_i - \\bar{s}_.),\n\\]\n\nand denote \\( Cov(y_{il}, y_{im}) := \\sigma_{lm}, Corr(y_{il}, y_{im}) := \\rho_{lm}, \\Sigma = [\\sigma_{lm}]_{l,m=1,...,d}, R = [\\rho_{lm}]_{l,m=1,...,d} \\). Then the off diagonal and diagonal elements of \\( \\Sigma \\) are, respectively,\n\n\\[\n\\sigma_{lm} = (1 - \\frac{1}{n}) \\sigma_{lm}^0, \\quad \\sigma_{ll} = (1 - \\frac{1}{n})(\\sigma_{ll}^0 + \\sigma_e^2).\n\\]\n\nWe make the following assumption that noise to signal ratio is very small.\n\nAssumption 0.2 (Small noise-to-signal ratio).\n\n\\[\n\\sigma_e^2 \\ll \\sigma_{ll}^0.\n\\]\n\nThen we have:\n\n\\[\n\\rho_{lm} = \\frac{\\sigma_{lm}}{\\sqrt{\\sigma_{ll} \\sigma_{mm}}} = \\frac{\\sigma_{lm}^0}{\\sqrt{(\\sigma_{ll}^0 + \\sigma_e^2)(\\sigma_{mm}^0 + \\sigma_e^2)}} \\approx \\frac{\\sigma_{lm}^0}{\\sqrt{\\sigma_{ll}^0 \\sigma_{mm}^0}} = \\rho_{lm}^0,\n\\]\n\nwhich means the Pearson correlation coefficient among \\( y'_{ij}'s \\) provides a good approximation to the Pearson correlation coefficient among \\( x'_{ij}'s \\) as well as among \\( a'_{ij}'s \\). Since \\( y_{ij} = (o_{ij} - \\bar{o}_j) - (s_i - \\bar{s}_.) \\). From (7), we estimate \\( y_{ij} \\) by \\( \\hat{y}_{ij} = (o_{ij} - \\bar{o}_j) - \\bar{s}_i - \\bar{s}_. \\). Thus, from (6),\n\n\\[\n\\hat{y}_{ij} \\longrightarrow_p (o_{ij} - \\bar{o}_j) - (s_i - \\bar{s}_.) = y_{ij}, \\text{ as } |d(i)| \\to \\infty.\n\\] (8)\nIn the sequel, for all \\(1 \\leq l,m \\leq d\\), denote \\(n(l) = \\{i : M_{il} = 1, 1 \\leq i \\leq n\\}\\) and \\(n(l,m) = \\{i : M_{il} = M_{im} = 1, 1 \\leq i \\leq n\\}\\), where \\(n(l)\\) is the number of samples where \\(l^{th}\\) taxon is observed, and \\(n(l,m)\\) represents the number of samples where both \\(l^{th}\\) and \\(m^{th}\\) are observed. For each \\(1 \\leq l,m \\leq d\\), let\n\n\\[\n\\hat{\\mu}_l = \\frac{1}{|n(l)|} \\sum_{i \\in n(l)} \\hat{y}_{il},\n\\]\n\\[\n\\hat{\\sigma}_{lm} = \\frac{1}{|n(l,m)|} \\sum_{i \\in n(l,m)} (\\hat{y}_{il} - \\hat{\\mu}_l)(\\hat{y}_{im} - \\hat{\\mu}_m),\n\\]\n\\[\n\\Sigma_n = [\\hat{\\sigma}_{lm}]_{l,m=1,\\ldots,d},\n\\]\n\\[\nR_n = [\\hat{\\rho}_{lm}]_{l,m=1,\\ldots,d} = \\left[ \\frac{\\hat{\\sigma}_{lm}}{\\sqrt{\\hat{\\sigma}_{ll} \\hat{\\sigma}_{mm}}} \\right]_{l,m=1,\\ldots,d}.\n\\]\n\nTo prove the consistency of the proposed covariance estimator, we make the following mild Assumption 0.3 that each taxon (Assumption 0.3(i)) or each pair of taxa (Assumption 0.3(ii)) is present with a nonzero probability (\\(\\delta_{min}\\)).\n\n**Assumption 0.3 (Minimal probability). Define**\n\n\\[\nC_l(l) = \\{t : \\text{taxon } l \\text{ is present in } T_t \\text{ in } M_t, 1 \\leq t \\leq 2^d - 1\\},\n\\]\n\\[\nC_l(l,m) = \\{t : \\text{taxa } l \\text{ and } m \\text{ are present in } T_t \\text{ in } M_t, 1 \\leq t \\leq 2^d - 1\\}.\n\\]\n\n*There exists a constant \\(\\delta_{min} > 0\\) such that for any \\(1 \\leq l,m \\leq d\\),\n\n\\[(i) \\frac{1}{n} \\sum_{l=1}^n \\sum_{t \\in C_l(l)} \\delta_l^t = \\delta_l > \\delta_{min}, \\quad (ii) \\frac{1}{n} \\sum_{l=1}^n \\sum_{t \\in C_l(l,m)} \\delta_l^t = \\delta_{l,m} > \\delta_{min}.\\]*\n\n**Theorem 0.1.** *Suppose \\(\\Sigma^0\\) satisfies Assumption 0.1 and Assumption 0.3 is valid, then*\n\n\\[\n\\left\\| R_n - R \\right\\|_\\infty = O_p\\left( \\sqrt{\\frac{\\log d}{n}} \\right).\n\\]\n\nThus, the empirical correlation matrix \\(R_n\\) based on \\(\\hat{y}_{ij}'s\\) consistently estimates the correlation matrix \\(R\\) of \\(y_{ij}\\) in the *Sup* norm. Under the Assumption 0.2, \\(R\\) is a reasonable approximation to the true correlation matrix \\(R^0\\). The details of the proof are provided in the Supplementary Notes.\n\nInstead of Pearson correlation coefficient, if the Spearman correlation coefficient is of interest, then we convert \\(x_{ij}, y_{ij}, \\hat{y}_{ij}\\) into ranks \\(r_{ij}^0, r_{ij}, \\hat{r}_{ij}\\). Clearly, for \\(r_{ij}^0, r_{ij}, \\hat{r}_{ij} \\in \\{1,2,\\ldots,n\\}\\),\n\n\\[\n\\hat{r}_{ij} \\longrightarrow_p r_{ij}, \\text{ as } |d(i)| \\to \\infty,\n\\]\n\\[\nr_{ij} \\approx r_{ij}^0 \\text{ as long as Assumption 0.2 holds.}\n\\]\n\nWith a slight abuse of notations, denote \\(Corr(r_{il}, r_{im}) = s_{lm}\\), \\(Corr(\\hat{r}_{il}, \\hat{r}_{im}) = \\hat{s}_{lm}\\), \\(S = [s_{lm}]_{l,m=1,\\ldots,d}\\), \\(S_n = [\\hat{s}_{lm}]_{l,m=1,\\ldots,d}\\). Then we have the following result.\n\n**Theorem 0.2.** *Suppose \\(\\Sigma^0\\) satisfies Assumption 0.1 and Assumption 0.3 is true, then:*\n\n\\[\n\\left\\| S_n - S \\right\\|_\\infty = O_p\\left( \\sqrt{\\frac{\\log d}{n}} \\right).\n\\]\n\nThe proof of this result relies on the observation that standardized rank \\(\\zeta_{il} = \\frac{\\hat{r}_{il} - E(\\hat{r}_{il})}{\\sqrt{Var(\\hat{r}_{il})}}\\) is naturally bounded, which allows the applicability of the Hoeffding’s inequality. The details of the proof are deferred to the Supplementary Notes.\n\nWhen computing correlations of high-throughput sequencing data, it is reasonable to assume that only a small fraction of features (genes/proteins/taxa etc.) are correlated with each other, i.e. the correlation matrix is sparse. Inspired by the Sparse Estimation of the Correlation matrix (SEC) approach\\(^{43}\\), we estimate the sparse correlation matrix by solving\n\n\\[\n\\hat{R} = \\arg\\min_R \\frac{1}{2} \\|R - R_n\\|_F^2 + \\lambda \\|R\\|_1, \\text{ s.t. } R \\succeq \\varepsilon I, R_{jj} = 1, j = 1,\\ldots,d.\n\\]\n\nwhere\n(1) \\( \\| \\cdot \\| \\) is the Frobenius norm,\n\n(2) \\( R \\succeq \\varepsilon I \\) means that \\( R - \\varepsilon I \\) is semipositive definite,\n\n(3) \\( \\lambda \\) is the regularization parameter for the \\( l_1 \\) norm.\n\nDenote the non-diagonal support of \\( R \\) as \\( A_0 = \\{l, m : l \\neq m, \\rho_{lm} \\neq 0\\} \\) and its cardinality as \\( q \\). The convergence of \\( \\hat{R} \\) to \\( R \\) in Frobenius norm is established in the following theorem.\n\n**Theorem 0.3.** *Suppose \\( \\Sigma^0 \\) satisfies Assumption 0.1 and Assumption 0.3 is valid, and \\( \\lambda = M \\sqrt{\\frac{\\log d}{n}} \\) for some constant \\( M \\), we have the estimator defined in (11) such that*\n\n\\[\n\\| \\hat{R} - R \\|_F^2 = O_p(q \\frac{\\log d}{n})\n\\]\n\nA sketch of the proof is provided in the Supplementary Notes.\n\nFrom the computational point of view, the penalized likelihood problem stated in (11) can be replaced with the soft thresholding operator\n\n\\[\n\\hat{R} = \\text{sign}(R_n)((|R_n| - \\lambda)_+,\n\\]\n\nwhere the tuning parameter \\( \\lambda \\) is selected by cross-validation \\( \\lambda = \\min_\\lambda \\| \\hat{R} - R_n \\|_F \\).\\footnote{23}\n\nSince the solution \\( \\hat{R} \\) may not be symmetric in general, the final estimate of \\( \\hat{R} \\) is obtained by forcing \\( \\hat{\\rho}_{ll} = 1 \\) and choosing \\( \\hat{\\rho}_{lm} = \\hat{\\rho}_{lm} I(\\hat{\\rho}_{lm} \\leq \\hat{\\rho}_{ml}) + \\hat{\\rho}_{ml} I(\\hat{\\rho}_{lm} > \\hat{\\rho}_{ml}) \\). We denote the \\( \\hat{R} \\) obtained from the thresholding procedure as \"SECOM (Thresholding)\".\n\nThe sparse matrix \\( \\hat{R} \\) can also be obtained by thresholding on the basis of on p-values. Let \\( p_{lm} \\) be the corresponding p-value for \\( \\hat{\\rho}_{lm} \\), then \\( \\hat{R} \\) can be defined as\n\n\\[\n\\hat{R} = [\\hat{\\rho}_{lm} I(p_{lm} < \\alpha)]_{l,m=1,\\ldots,d},\n\\]\n\nwhere \\( \\alpha \\) is a pre-specified significance level, e.g. \\( \\alpha = 0.005 \\). We denote the \\( \\hat{R} \\) obtained based on p-values as \"SECOM (Filtering)\".\n\n**Estimation of nonlinear correlation coefficient**\nIn addition to linear correlation, we are also interested in describing nonlinear correlations among taxa within an ecosystem (e.g., gut or oral environments) or across ecosystems. Therefore, the concept of distance correlation\\textsuperscript{36} was adopted to quantify general dependence between a pair of taxa.\n\n**Definition 0.1** (Distance correlation coefficient). Let \\( X \\) and \\( Y \\) denote two random variables. Then the distance covariance and correlation coefficient between \\( X \\) and \\( Y \\) are defined as following\n\n(1) \\( dCov^2(X,Y) = \\frac{1}{\\pi^2} \\iint \\frac{\\| f_{X,Y}(t,s) - f_X(t)f_Y(s) \\|^2}{t^2 s^2} dt ds \\),\n\n(2) \\( dVar^2(X) = \\frac{1}{\\pi^2} \\iint \\frac{\\| f_{X,X}(t,s) - f_X(t)f_X(s) \\|^2}{t^2 s^2} dt ds \\),\n\n(3) \\( dCor(X,Y) = \\frac{dCov(X,Y)}{\\sqrt{dVar(X)dVar(Y)}} \\).\n\nwhere \\( \\| \\cdot \\| \\) denotes Euclidean norm, and\n\n(1) \\( 0 \\leq dCor(X,Y) \\leq 1 \\),\n(2) \\( dCor(X,Y) = 0 \\) if and only if \\( X \\) and \\( Y \\) are independent,\n(3) \\( dCor(X,Y) = 1 \\) implies that dimensions of the linear subspaces spanned by \\( X \\) and \\( Y \\) samples respectively are almost surely equal.\n\nEstimator of the distance correlation coefficient based on random samples \\( X = (X_1, X_2, \\ldots, X_n)' \\) and \\( Y = (Y_1, Y_2, \\ldots, Y_n)' \\) is derived as follows. Define pairwise distances as \\( d_{ij} = |x_i - x_j|, e_{ij} = |y_i - y_j| \\), and doubly centered distances as \\( D_{ij} = d_{ij} - d_i - d_j + d \\), \\( E_{ij} = e_{ij} - \\bar{e}_i - \\bar{e}_j + \\bar{e} \\), \\( i \\neq j \\), the empirical estimates of distance covariance and correlation can be obtained by\n\n(1) \\( dCov_n^2(X,Y) = \\frac{1}{n^2} \\sum_{i=1}^n \\sum_{j=1}^n D_{ij} E_{ij}, \\)\n(2) \\( dVar_n^2(X,Y) = \\frac{1}{n^2} \\sum_{i=1}^n \\sum_{j=1}^n D_{ij}^2. \\)\n\nNote that the rank-transformed data also preserve the general dependence in the original data. We illustrate this using both linear and quadratic examples as shown in Supplementary Fig. 5a and b, respectively. In Supplementary Fig. 5a, \\( y = 2x + e, x \\sim U[-2,2], e \\sim N(0,1) \\); whereas in Supplementary Fig. 5b, \\( y = x^2 + e, x \\sim U[-2,2], e \\sim N(0,1) \\). It is easy to see from the figure that even though the shape of \\((X,Y)\\) is not perfectly recovered using ranks (denoted as \\((r(X), r(Y))\\)), distance correlations computed using the original data and the rank-transformed data (denoted as \\(dCor(X,Y)\\) and \\(dCor(r(X), r(Y))\\), respectively) are close enough to each other with a minimal numerical difference, which means the distance correlation computed from the rank-transformed data provides a reasonably good approximation for the distance correlation computed from the original data. Therefore, researchers can use either original abundances or the rank-transformed abundances to calculate distance correlations.\n\nSimilar to the procedure shown in the estimation of linear correlation coefficients, define\n\n\\[\n\\begin{align*}\nd_{lm} &:= dCor(y_{il}, y_{jm}) \\quad \\text{(or } dCor(r_{il}, r_{jm})) \\\\\n\\hat{d}_{lm} &:= dCor(\\hat{y}_{il}, \\hat{y}_{jm}) \\quad \\text{(or } dCor(\\hat{r}_{il}, \\hat{r}_{jm})) \\\\\nD &= [d_{lm}]_{l,m=1,\\ldots,d}, \\\\\nD_n &= [\\hat{d}_{lm}]_{l,m=1,\\ldots,d},\n\\end{align*}\n\\]\n\nthe sparse distance correlation matrix estimation can be obtained by thresholding or p-values:\n\n\\[\n\\begin{align*}\n\\hat{D} &= sign(D_n)(|D_n| - \\lambda)_+ \\text{, or} \\\\\n\\hat{D} &= [\\hat{d}_{lm} I(p_{lm} < \\alpha)]_{l,m=1,\\ldots,d}.\n\\end{align*}\n\\]\n\nData availability\nSimulation data can be found under https://github.com/FrederickHuangLin/SECOM-Code-Archive. The NoMIC data used in this study are not publicly available but may be obtained by contacting Dr. Merete Eggesbø (Merete.Eggesbo@fhi.no).\n\nCode availability\nAll analyses can be found under https://github.com/FrederickHuangLin/SECOM-Code-Archive.\n\nReferences\n1. Mainali, K., Bewick, S., Vecchio-Pagan, B., Karig, D. & Fagan, W. F. Detecting interaction networks in the human microbiome with conditional granger causality. PLoS computational biology **15**, e1007037 (2019).\n2. Turnbaugh, P. J. *et al.* A core gut microbiome in obese and lean twins. *nature* **457**, 480–484 (2009).\n3. Gevers, D. *et al.* The treatment-naive microbiome in new-onset crohn’s disease. *Cell host & microbe* **15**, 382–392 (2014).\n4. Lozupone, C. A. *et al.* Alterations in the gut microbiota associated with hiv-1 infection. *Cell host & microbe* **14**, 329–339 (2013).\n5. Weiss, S. *et al.* Normalization and microbial differential abundance strategies depend upon data characteristics. *Microbiome* **5**, 1–18 (2017).\n6. Lin, H. & Peddada, S. D. Analysis of microbial compositions: a review of normalization and differential abundance analysis. *NPJ biofilms microbiomes* **6**, 1–13 (2020).\n7. Lin, H. & Peddada, S. D. Analysis of compositions of microbiomes with bias correction. *Nat. communications* **11**, 1–11 (2020).\n8. Morton, J. T. *et al.* Establishing microbial composition measurement standards with reference frames. *Nat. communications* **10**, 1–11 (2019).\n9. Mandal, S. *et al.* Analysis of composition of microbiomes: a novel method for studying microbial composition. *Microb. ecology health disease* **26**, 27663 (2015).\n10. Gloor, G. B. & Reid, G. Compositional analysis: a valid approach to analyze microbiome high-throughput sequencing data. *Can. journal microbiology* **62**, 692–703 (2016).\n11. Gloor, G. B., Wu, J. R., Pawlowsky-Glahn, V. & Egozcue, J. J. It’s all relative: analyzing microbiome data as compositions. Annals epidemiology 26, 322–329 (2016).\n\n12. Gloor, G. B., Macklaim, J. M., Pawlowsky-Glahn, V. & Egozcue, J. J. Microbiome datasets are compositional: and this is not optional. Front. microbiology 8, 2224 (2017).\n\n13. Morton, J. T. et al. Balance trees reveal microbial niche differentiation. MSystems 2, e00162–16 (2017).\n\n14. Friedman, J. & Alm, E. J. Inferring correlation networks from genomic survey data. PLOS computational biology 8, e1002687 (2012).\n\n15. Lovell, D., Pawlowsky-Glahn, V., Egozcue, J. J., Marguerat, S. & Bähler, J. Proportionality: a valid alternative to correlation for relative data. PLoS computational biology 11, e1004075 (2015).\n\n16. Kurtz, Z. D. et al. Sparse and compositionally robust inference of microbial ecological networks. PLoS computational biology 11, e1004226 (2015).\n\n17. Morton, J. T. et al. Learning representations of microbe–metabolite interactions. Nat. methods 16, 1306–1314 (2019).\n\n18. McLaren, M. R., Willis, A. D. & Callahan, B. J. Consistent and correctable bias in metagenomic sequencing experiments. Elife 8, e46923 (2019).\n\n19. Anderson, T. W. & Olkin, I. Maximum-likelihood estimation of the parameters of a multivariate normal distribution. Linear algebra its applications 70, 147–171 (1985).\n\n20. Serra, A., Coretto, P., Fratello, M. & Tagliaferri, R. Robust and sparse correlation matrix estimation for the analysis of high-dimensional genomics data. Bioinformatics 34, 625–634 (2018).\n\n21. Bickel, P. J. & Levina, E. Regularized estimation of large covariance matrices. The Annals Stat. 36, 199–227 (2008).\n\n22. Bickel, P. J. & Levina, E. Covariance regularization by thresholding. The Annals Stat. 36, 2577–2604 (2008).\n\n23. Rothman, A. J., Levina, E. & Zhu, J. Generalized thresholding of large covariance matrices. J. Am. Stat. Assoc. 104, 177–186 (2009).\n\n24. Cai, T. & Liu, W. Adaptive thresholding for sparse covariance matrix estimation. J. Am. Stat. Assoc. 106, 672–684 (2011).\n\n25. Cai, T. T. & Yuan, M. Adaptive covariance matrix estimation through block thresholding. The Annals Stat. 40, 2014–2042 (2012).\n\n26. Lam, C. & Fan, J. Sparsistency and rates of convergence in large covariance matrix estimation. Annals statistics 37, 4254 (2009).\n\n27. Cai, T. T. & Zhou, H. H. Optimal rates of convergence for sparse covariance matrix estimation. The Annals Stat. 40, 2389–2420 (2012).\n\n28. Yi, F. & Zou, H. Sure-tuned tapering estimation of large covariance matrices. Comput. Stat. & Data Analysis 58, 339–351 (2013).\n\n29. Maurya, A. A joint convex penalty for inverse covariance matrix estimation. Comput. Stat. & Data Analysis 75, 15–27 (2014).\n\n30. Paulson, J. N., Stine, O. C., Bravo, H. C. & Pop, M. Differential abundance analysis for microbial marker-gene surveys. Nat. methods 10, 1200–1202 (2013).\n\n31. Dahl, C. et al. Preterm infants have distinct microbiomes not explained by mode of delivery, breastfeeding duration or antibiotic exposure. Int. journal epidemiology 47, 1658–1669 (2018).\n\n32. Iszatt, N. et al. Environmental toxicants in breast milk of norwegian mothers and gut bacteria composition and metabolites in their infants at 1 month. Microbiome 7, 1–14 (2019).\n\n33. Sugihara, G. et al. Detecting causality in complex ecosystems. science 338, 496–500 (2012).\n\n34. Hoffmann, A. R., Proctor, L., Surette, M. & Suchodolski, J. The microbiome: the trillions of microorganisms that maintain health and cause disease in humans and companion animals. Vet. pathology 53, 10–21 (2016).\n\n35. Hansen, T. H., Gøbel, R. J., Hansen, T. & Pedersen, O. The gut microbiome in cardio-metabolic health. Genome medicine 7, 1–16 (2015).\n\n36. Székely, G. J., Rizzo, M. L., Bakirov, N. K. et al. Measuring and testing dependence by correlation of distances. The annals statistics 35, 2769–2794 (2007).\n\n37. Aitchison, J. Principles of compositional data analysis. Lect. Notes-Monograph Ser. 73–81 (1994).\n38. Williamson, B. D., Hughes, J. P. & Willis, A. D. A multi-view model for relative and absolute microbial abundances. Biometrics (2019).\n\n39. Vandeputte, D. et al. Quantitative microbiome profiling links gut community variation to microbial load. Nature **551**, 507–511 (2017).\n\n40. Granger, C. L. et al. Maternal breastmilk, infant gut microbiome and the impact on preterm infant health. Acta Paediatr. **110**, 450–457 (2021).\n\n41. Ho, N. T. et al. Meta-analysis of effects of exclusive breastfeeding on infant gut microbiota across populations. Nat. communications **9**, 1–13 (2018).\n\n42. Sagheddu, V., Patrone, V., Miragoli, F., Puglisi, E. & Morelli, L. Infant early gut colonization by lachnospiraceae: high frequency of ruminococcus gnavus. Front. pediatrics **4**, 57 (2016).\n\n43. Cui, Y., Leng, C. & Sun, D. Sparse estimation of high-dimensional correlation matrices. Comput. Stat. & Data Analysis **93**, 390–403 (2016).\n\nAcknowledgements\nWe are grateful to all HUMIS cohort participants. The research of SDP and HL was funded in part by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), National Institutes of Health, Bethesda, MD. The research of ME was funded in part by the Research Council of Norway, NEVRINOR [grant agreement no. 226402].\n\nAuthor contributions\nTheory and Methodology were conceived and developed by HL and SDP. Illustration section was developed by HL, ME and SDP. Interpretation of the results from the analysis of NoMIC study were performed by ME and SDP. All numerical works were carried out by HL and ME. All three authors contributed towards writing the paper.\n\nCompeting interests\nAuthors declare no conflicts of interest.\nFigure 1. A toy example showing the difference between correlation methods. a A total of 100 taxa were generated, where the true abundances in the ecosystem were generated using a Negative-Binomial distribution. Taxa T1 and T3 are perfectly linearly correlated, and taxa T2 and T4 are quadratically correlated. The remaining taxa are uncorrelated. Only the first 5 taxa (T1 to T5) are provided for the simplicity of exposition. The lower triangular part of the matrix shows the pairwise scatter plots, the upper triangular part of the matrix shows the corresponding pairwise Pearson correlation coefficients, and the diagonal of the matrix shows the density plots for each taxon. b Observed abundances were generated by incorporating sample-specific sampling fractions and taxon-specific sequencing efficiencies to the true abundances. The lower triangular part of the matrix shows the pairwise scatter plots, the upper triangular part of the matrix shows the corresponding pairwise Pearson correlation coefficients, and the diagonal of the matrix shows the density plots for each taxon. c, d, e Pairwise correlation coefficients returned from the standard Pearson correlation coefficients, SparCC, and SECOM based on the observed abundances, respectively. The correlation coefficient ranges from -1 to 1, color coded by blue to red, correspondingly.\nFigure 2. Comparisons of estimation accuracy and sparsity recovery for linear relationships within an ecosystem. The average relative norm loss (Frobenius and Spectral) and sparsity recovery (FPR/TPR) of various correlation methods are shown in a and b, respectively. Synthetic data were generated from Negative-Binomial (NB) distributions. The X-axis denotes the simulation settings, which are a combination of sample size \\( n \\), number of taxa \\( d \\), and the dispersion parameter \\( \\alpha \\) of the corresponding NB distribution. The number of iterations for each simulation setting = 100. Data are represented by the mean of all iterations and error bars represent standard errors. Color and the name of the corresponding correlation methods are shown at the bottom within the graph. The results demonstrate that SECOM (Thresholding and Filtering) and SPIEC-EASI (MB and GL) outperformed all existing methods not only in terms of estimation accuracy, but also in terms of uniformly small FPR, and comparable TPR.\nFigure 3. Comparisons of sparsity recovery for nonlinear relationships within an ecosystem. Synthetic data were generated from Negative-Binomial (NB) distributions. The X-axis denotes the simulation settings, which are a combination of sample size \\( n \\), number of taxa \\( d \\), and the dispersion parameter \\( \\alpha \\) of the corresponding NB distribution. The number of iteration for each simulation setting = 100. Data are represented by the mean of all iterations and error bars represent standard errors. Color and the name of the corresponding correlation methods are shown at the bottom within the graph. Results showed that only SECOM controlled the FPR while maintaining high TPR. Other than SECOM, none of existing methods identify nonlinear relationships.\n\n\nFigure 4. Comparisons of estimation accuracy and sparsity recovery for linear relationships between two ecosystems. The relative norm loss (Frobenius and Spectral) and sparsity recovery (FPR/TPR) of various correlation methods are shown in a and b, respectively. Synthetic data were generated from Negative-Binomial (NB) distributions. The X-axis denotes the simulation settings, which are the combinations of sample size \\( n \\), number of taxa \\( d \\), and the dispersion parameter \\( \\alpha \\) of the corresponding NB distribution. Data from different ecosystems have different sampling fractions and sequencing efficiencies in each simulation run. The number of iterations for each simulation setting = 100. Data are represented by the mean of all iterations and error bars represent standard errors. Color and the name of the corresponding correlation methods are shown at the bottom within the graph. Results demonstrated that SECOM methods, namely, SECOM(Thresholding) and SECOM(Filtering) were the only methods successfully quantified microbial correlations between ecosystems.\nFigure 5. Temporal patterns of bias corrected abundance (a) and relative abundances (b) for Norwegian infant gut microbiome data among families. The x-axis represents days (30, 120, and 365 days) and y-axis denotes either the bias corrected abundances or relative abundances. The top 10 most abundant taxa up to 365 days were selected for the visualization purpose. Families with prevalence across samples < 10% were not shown separately and grouped into \"Others\". Different taxa were coded by different colors as shown on the legends.\nFigure 6. Temporal patterns of correlations for Norwegian infant gut microbiome data among families. The correlation metrics (Linear and Distance) are represented by rows and days (30, 120, and 365 days) are represented by columns, respectively. The top 10 most abundant families up to 365 days were initially selected for the visualization purpose. Families with prevalence across samples < 10% were discarded and left blank in the heatmap. The linear correlation coefficient ranges from -1 to 1, color coded by blue to red, respectively.\nSupplementary Files\n\nThis is a list of supplementary files associated with this preprint. Click to download.\n\n• SECOMSupplementaryInformation.pdf\n• suppfig1.pdf\n• suppfig2.pdf\n• suppfig3.pdf\n• suppfig4.pdf\n• suppfig5.pdf"}
{"paper_id": "prb81", "title": "AC amplification gain in organic electrochemical transistors for impedance-based single cell sensors", "text": "# Abstract\n\nResearch on water-gated and organic electrochemical transistor (OECT) architectures is motivated by the prospect of a highly biocompatible interface capable of amplifying bioelectronic signals at the site of detection. Despite many demonstrations in these directions, a quantitative model for OECTs as impedance biosensors is still lacking. We overcome this issue by introducing a model experiment where we simulate the detection of a single cell by the impedance sensing of a dielectric microparticle. The highly reproducible experiment allows us to study the impact of transistor geometry and operation conditions on device sentivity. With the data we rationalize a mathematical model that provides clear guidelines for the optimization of OECTs as single cell sensors, and we verify the quantitative predictions in an in-vitro experiment. In the optimized geometry, the OECT-based impedance sensor allows to record single cell adhesion and detachment transients, showing a maximum gain of (20.2±0.9) dB with respect to a single electrode based impedance sensor.\n\n# Introduction\n\nIn recent years water gated and electrochemical transistor architectures have found broad attention in biosensor research.¹,² The interest is motivated by the possibility to combine electrochemical transduction mechanisms known from metallic electrodes with the intrinsic amplification properties of a transistor structure.³ The amplification is highly wanted to improve signal to noise ratio in challenging sensor applications aiming for instance at biochemical detection in complex mixtures⁴ or single cell bioelectronic monitoring.⁵ Significant progress on this concept has been made by the introduction of organic or carbon-based semiconducting materials with high stability in water such as carbon nanotubes,⁶,⁷ graphene,⁸,⁹ or organic semiconductors.¹⁰ When patterned into a semiconducting channel connected to a source and drain electrode, these materials enable transistor-like structures to replace single metallic working electrodes. Since the semiconducting channel is in direct contact with the aqueous electrolyte, a strong capacitive coupling results between the channel’s electronic carrier concentration and the electrochemical potential in the solution.¹¹ Events that alter the electrochemical potential and impact on the ionic distribution at the channel-electrolyte interface thus gate the sensor’s semiconducting channel conductivity. Even small perturbations in the ionic distribution can cause a large variation in the number of electronic carriers flowing through the channel from source to drain electrode, hence leading to an amplification effect.¹² This qualitative argument is demonstrated by improved sensitivity in several applications realized with water gated transistor structures.¹³ However, to achieve optimized sensor devices, a quantitative understanding of such signal gain is needed. The goal of this work is to derive a quantitative model that relates transistor amplification gain to semiconductor material properties and water gated transistor architecture.\n\nA widely explored class of water gated transistors are organic electrochemical transistors (OECTs).¹⁴ OECTs exploit organic mixed ionic and electronically conductors such as poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS).² The high chemical stability, the facile, solvent-based processing and the high biocompatibility, make PEDOT:PSS a material of choice for healthcare applications.¹⁵ In PEDOT:PSS, electronic transport is achieved by mobile hole charges present in the oxidized semiconducting polymer PEDOT. The positive charge of the holes is counterbalanced by fixed, negative ionic charges of the polyanion PSS.¹⁶ Oxidized organic semiconductor and ionic polyanion form a nanophase separated network that generates close electrostatic interaction between the two phases combined with efficient electronic as well as ionic charge transport.¹⁵,¹⁷ Consequently, high electronic carrier mobility (µ > 10 cm² V⁻¹ s⁻¹)¹⁸ is combined with a strong volumetric capacitive coupling between the ionic phase and the electronic phase (c > 30 F/cm³).¹⁹ In OECTs this properties combination gives rise to large values of transconductance gm = ∂Ich/∂Vg that express how small changes in the electrostatic potential ∂Vg of the electrolyte gate the electronic channel current Ich.²⁰ Relying on the large transconductance combined with the biocompatible material properties, many research works propose OECTs as amplifying transistor to be integrated in electrochemical and bioelectronic sensors for healthcare applications.²¹ Established examples regard potentiometric sensors used to quantify the concentration of ionic or redox active analytes²² and to record electrophysiological signals.²³ In this case, the electrical circuit involving the OECT must operate ultimately as a voltage amplifier, that produces an amplified output voltage ready for digitization. Gain is then expressed as the ratio between the original potentiometric signal and the output voltage. Depending on the circuit design, DC amplification gains reaching 30 V/V have been demonstrated in OECT based potentiometric sensors.²⁴,³\n\nA second, emerging class of biosensors that takes advantage of OECT amplification regards impedance based sensors for monitoring cellular adhesion and cell layer barrier properties as quantified by the transepithelial electrical resistance (TEER).²⁵,⁴ Monitoring such cellular properties is of importance in studies of wound healing, cancer development, toxicity assays or recognition processes in the immune system.²⁵ Impedance based biosensors quantify the ionic current that passes through a cellular layer when an AC voltage is applied.²⁶ Changes in cellular adhesion or intercellular barrier properties change the measured current signal thus providing a simple means for real-time monitoring.²⁷ By replacing the metallic working electrode of a traditional impedance sensor with a water gated OECT, amplification of the current signal is achieved thus increasing sensitivity and reducing possible noise pickup.²⁸ Current amplification becomes particularly important in high impedance applications as encountered when the sensor size is scaled down to micrometric scales matching cellular dimensions. Such downscaling opens the opportunity to translate impedance based cellular monitoring to the single cell level.²⁹ This ultimate sensor resolution is highly desired in biomedical research as the importance of single cell phenotyping is increasingly recognized for the study of cell development and physiology, as well as for research on cellular pathologies such as cancer.³⁰ An important step in this direction was recently achieved by Hempel et al. by demonstrating single cell sensitivity in an OECT enabled impedance sensor. The authors found that significant differences in the sensors transfer function are caused by cells adhering to the transistor surface. Such changes in transfer function are usually quantified as the transistor bandwidth (or response time) and equivalent circuit models were developed to explain the observed frequency response and its relation to impedances at the cell/PEDOT:PSS interface.⁵\n\nDespite the promising results, a quantitative study on the transistor amplification gain in the frequency spectrum relevant for cellular impedance sensing is still lacking. Different studies demonstrate that the high OECT transconductance is limited to the low frequency regime and strongly depends on transistor geometry and materials properties.²⁰ Consequently, there is a need for a quantitative model that relates impedance sensor gain to OECT transconductance and other device properties in order to enable a rational optimization of OECT based impedance sensors. A clear understanding of gain is also desired to define when OECT amplification has significant advantages over one-terminal, low-impedance microelectrode-based sensors offering simpler fabrication and electrical operation. To overcome the issue, we introduce in this work a model experiment that allows a quantitative analysis of amplification gain in OECT based impedance sensors. As cellular in-vitro experiments are inherently difficult to control we substitute the cell by a dielectric microparticle of similar dimensions. We control the position of the microparticle on top of the microscale impedance sensors with an AFM and achieve highly reproducible measurements that enable to compare the current output of OECT based sensors with equivalent microelectrode sensors. To rationalize the findings, we develop an analytical model that describes the gain as a function of the applied frequency, the device geometry and PEDOT:PSS materials properties. Relying on this model, we design an optimized device and demonstrate its efficiency by measuring the transients of single cell adhesion and detachment in in-vitro experiments. Noteworthy, we observe a significant AC gain reaching values of (20.2±0.9) dB for the transistor structure, thereby demonstrating the advantages arising from the OECT amplification in single-cell impedance sensing experiments.\n\n# Results\n\nImpedance sensing of a dielectric microparticle: The first objective of our work is to introduce a novel experiment to quantify the sensitivity of electrochemical impedance sensors operated in OECT or microelectrode configuration. To this end we realized the experimental setup shown in Figure 1a and 1b. The setup contains a dielectric microparticle (with diameter of 50 µm) attached to the bottom part of an AFM cantilever to have micrometric control of its position in the 3 spatial directions (Figure 1c). Once the microparticle is finely aligned with the x-y coordinates on the center of the sensor surface, we use the z-stage of the microscope to control the particle-sample distance d. Contact of the microparticle with the sensor surface is determined by the onset of a repulsive force acting on the AFM cantilever. The electric circuit to operate the OECT impedance sensor contains a Ag/AgCl wire that is used as the gate electrode, controlling the electrical potential of the aqueous electrolyte solution (0.1 M PBS). A DC voltage VD,DC is applied between the source (S) and drain (D) electrodes of the OECT to drive the electronic current ID,DC in the PEDOT:PSS channel. The measured transfer and output characteristics of a typical OECT (Figure 1d), demonstrate that the gate voltage effectively modulates the channel current. For impedance sensing we superimpose a small sinusoidal oscillation signal VG,AC (with amplitude 10 mV and angular frequency ω) on the gate bias VG,DC. This leads to an AC current in the PEDOT:PSS layer, which is measured by a lock-in amplifier connected to the source contact (IS,AC).\n\nIn the microelectrode configuration (Figure 1b) the circuit is simplified, as source and drain electrodes are in short circuit and are jointly connected to the lock-in amplifier. Therefore, no OECT channel current is present, and all the electric current measured during impedance sensing is the gate current IG,AC, flowing from the electrolyte into the PEDOT:PSS layer. All other components are identical to the OECT configuration to permit a direct comparison.\n\nIn Figure 1e we show the results of a typical microparticle distance - AC current experiment conducted at 1.17 kHz excitation frequency. The amplitudes of the AC currents in OECT and microelectrode configuration are plotted as a function of time. During the experiment, the microparticle is approached and retracted from the device channel for three consecutive times. For both configurations, the current amplitude follows the motion of the microparticle in a highly reproducible manner over consecutive cycles, highlighting the stability of the characterization method. In Figure 1f, the same data is plotted as a function of the distance between microparticle and sensor surface. Both types of devices produce a reversible, linear response in which the approach leads to a reduction in AC amplitude. Qualitatively, this response is expected, as the microsphere represents a barrier for the ionic current in the electrolyte: when it is close to the sensor surface, the half space through which ions can approach the active layer is reduced, thus increasing the effective impedance of the electrolyte Zel. Consequently, upon approach, the interfacial impedance measured with the sensor increases and the AC current amplitude drops. We note that in first order approximation a similar response is expected when a biological cell adheres to the sensor surface.\n\nThe results are crucial for our goal as they permit the quantitative assessment of the sensitivity of the impedance sensor. For the case of a high sensitivity, small changes in the impedance Zel cause large variation in AC current amplitude. Therefore, we define the sensitivity as s = ∂IAC/∂Zel. In our experiment, ∂Zel is directly related to the microparticle displacement ∂Zel = p*∂d. The proportionality constant p is independent on the sensor configuration (OECT vs microelectrode) and we obtain its numerical value by fitting the microelectrode impedance spectra (see Supp. Inf S2). Accordingly, the sensitivity is given by the slope of the approach curves shown in Figure 1e. We obtain numerical values for this particular case (WxL = 100x100 µm) of sOECT = (0.059±0.002) nA/Ω and sµE = (0.023±0.006) nA/Ω. The values show a greater sensitivity in the OECT device with respect to the microelectrode, due to the contribution of OECT channel current to the AC response.\n\nQuantitative model for PEDOT:PSS-based impedance sensors: We developed an analytical model to express the impedance sensitivity s as a function of the sensor operation conditions, material properties and geometry. Objective is a quantitative understanding of the factors that increase sensitivity in OECT configuration with respect to PEDOT:PSS microelectrodes. Many studies decouple charge transport in OECTs in an electronic and an ionic circuit. A schematic of this representation is reported in Figure 2a, where the components of the electronic and the ionic circuit are indicated in blue and orange, respectively. Electronic charge carriers (holes) are driven by the drain voltage VD,DC and carry the channel current in an OECT, while ionic charge carriers are driven by the gate voltage VG = VG,DC + VG,AC and modulate the concentration of holes and, consequently, the electronic conductivity of the transistor channel. The limited conductivity of the electrolyte as well as the presence of dielectric objects close to the sensor surface generate an impedance Zel which causes a potential drop in the electrolyte, and the voltage at the electrolyte/channel interface VG* can be considered as the effective gate voltage that acts on the channel and determines the drain current. For the impedance sensing we are interested in the AC response of the transistor and we express the AC current flowing in the OECT channel as Ich,AC = gm,AC*VG*,AC. Following Bernards model, gm,AC can be expressed as gm,AC^lin = -W/L µp cv t VD,DC and gm,AC^sat = -W/L µp cv (VG,DC - Vt) in linear or saturation conditions. In these expressions, W, L and t indicate the width, length and thickness of the sensor channel, µp the holes mobility, cv the volumetric capacitance of PEDOT:PSS and Vt the OECT threshold voltage.\n\nTo derive the overall AC current, it is important to note that in AC transport conditions the source (and drain) current signals are composed of two contributions:\n$$\n{I}_{S,AC }= {I}_{ch,AC } + {f}_{OECT}\\\\bullet {I}_{G,AC } \\quad (1)\n$$\nThe first (Ich,AC) originates from the channel current, whereas the second (IG,AC) is due to the gate current and regards the capacitive current that has increasing importance at higher frequencies. Its value is given by IG,AC = VG,AC / ZG in which ZG = Zel + Zch is the overall impedance of the sensor given by the series combination of the electrolyte impedance Zel and the impedance related to the PEDOT:PSS channel capacitance Zch = 1/(iω Cch). The channel capacitance can further be related to the geometry and the volumetric capacitance of the PEDOT:PSS layer: Cch = cv*W*L*t. Possible contributions due to parasitic capacitances are neglected for simplicity. The factor fOECT in eqn. 1 determines how the gate current is distributed between the source and the drain terminal and is typically assumed to be equal to 0.5. Several studies demonstrate that the fOECT factor is slightly dependent on VD,DC, and VG,DC as well as on channel geometry. For this reason, fOECT was not assumed as constant in this experiment, but its value was set for each sensor in order to best fit the experimental data with the model.\n\nThe figure of merit of the OECT as impedance sensor (the sensitivity sOECT) indicates its capability to transduce a variation of Zel in a current output. This can be calculated from the model by differentiating eq. (1):\n$$\n{s}_{OECT }= \\\\left|\\\\frac{\\\\partial ({I}_{ch,AC}+{f}_{OECT}{I}_{G,AC})}{\\\\partial {Z}_{el}}\\\\right|=|{s}_{ch}+{f}_{OECT}\\\\bullet {s}_{\\\\mu E}| \\quad (2)\n$$\nAfter inserting the expressions for the two AC current contributions and differentiation we obtain for the channel sensitivity:\n$$\n{s}_{ch}=\\\\frac{{g}_{m,AC}}{{Z}_{G}}\\\\left(1-\\\\frac{{Z}_{el}}{{Z}_{G}}\\\\right){V}_{G,AC}\n$$\nand the sensitivity of the microelectrode is:\n$$\n{s}_{\\\\mu E}= \\\\frac{{V}_{G,AC}}{{Z}_{G}^{2}} \\quad (4)\n$$\nThe suitability of this simple approach to model the AC response of an OECT is demonstrated in Figure 2b and c. Figure 2b compares the frequency response of an OECT and of a microelectrode with the model predictions. The PEDOT:PSS channel width and length are W = 100 µm and L = 100 µm, respectively. The channel capacitance and the electrolyte impedance Zel were extracted for each device geometry by fitting the microelectrode impedance spectrum. The average volumetric capacitance of PEDOT:PSS resulted to be cv = (30.0 ± 1.3) F/cm³, obtaining a result consistent with literature findings. The OECT device shows a significantly higher current in the low frequency domain. Here the electronic channel current Ich prevails, and the transistor demonstrates clear amplifying properties. Then, above a cutoff frequency fc = 645 Hz, the transistor response is limited by the slow ionic transport between the channel and the electrolyte. At the same time the microelectrode’s response increases with frequency until a current limitation is reached due to the electrolyte impedance. As a consequence, in the high frequency limit both impedance sensor configurations yield the same current response. Importantly, the cutoff frequency that determines the OECT amplification is determined by the channel geometry as demonstrated in Figure 2c. The plot of the current amplitude versus frequency for OECTs with different channel sizes clearly shows that with increasing channel area and length a strong reduction in fc is observed.\n\nFinally, we systematically study the OECTs sensitivity towards electrolyte impedance changes with the microsphere experiment introduced above. Figure 2d shows the measured values for sOECT obtained for three different channel geometries at different AC frequencies. Eq. 2 is in excellent agreement with the frequency dependence of the measured data. In the low-frequency range the sensitivity shows a linear increase with frequency until it reaches a sensitivity maximum sOECT^max. Beyond the maximum, a smaller decrease in sensitivity is observed until it settles to a constant value for high frequencies. It is important to highlight that the position of sOECT^max corresponds to the device cutoff frequency fc (see Supp. Inf. S3 for the full mathematical treatment), and hence is a geometry-dependent parameter.\n\nIn Figure 2e we compare the sensitivity of a microelectrode and an OECT with the same dimensions (WxL = 200x50 µm). The transistor amplification, which is significant at low frequencies, has a relevant impact on the sensitivity. However, at high frequencies the response of both devices is limited by the electrolyte resistance and no significant differences are present. Such an observation is reflected by a frequency dependent OECT gain, which can be directly calculated with our model from eq. 2 and 4:\n$$\n{gain }_{OECT}= 20\\\\bullet {{log}}_{10}\\\\left(\\\\frac{{s}_{OECT}}{{s}_{microel}}\\\\right) \\quad (5)\n$$\nThe OECT gain is highest in the low-frequency regime, but is still significant in the 0.1-10 kHz range, where the impedance of the cell layers is typically measured. This justifies the use of a transistor structure for high precision bioelectronic impedance sensing experiments. Figure 2f demonstrates that the OECT gain is a geometry-dependent parameter. The smallest device (WxL = 50x50 µm) shows the highest gain, while a rectangular channel geometry is preferable for OECTs with the same area, since the gain is proportional to W/L ratio.\n\nSingle cell impedance sensor experiment: We demonstrated the value of the mathematical model here proposed for the optimization of a PEDOT:PSS-based single cell impedance sensor by monitoring single cell adhesion and detachment in an in-vitro experiment, simultaneously measuring the impedance changes with both an OECT and a microelectrode. According to the model prediction and the AFM experiment, we patterned the device channels with a 200x50 µm rectangular geometry, which provides the best performances in terms of sensitivity. The T98G cell line cultured in Minimum Essential Medium was diluted to have a final density of 1 x 10³ cells/cm³ and poured on the surface of the impedance sensors (see the Methods section for full details). After seeding, the cells reached the underlying substrate by gravity. We microfabricated a linear array of 10 PEDOT:PSS channels (see Supp. Inf. S4) to largely increase the probability of a single cell settling onto a sensor. An optical image of the final experimental configuration is reported in Figure 3a, showing a single T98G cell positioned at the center of the PEDOT:PSS channel. The encapsulation of the metallic electrodes with negative photoresist insulates the device from all the remaining cells which are not lying in the PEDOT:PSS active layer. We acquired the current spectra of both the OECT and the microelectrode at consecutive time intervals to make a real-time detection of the cell adhesion process. To stress the full consistency of the measurements acquired with the OECT and the microelectrode, we plot in Figure 3b the current amplitudes measured at 625 Hz as a function of time. The sensing frequency was selected in correspondence to the OECT cutoff (measured at time t=0), where the model indicates the maximum sensitivity. The signals acquired at the beginning of the experiment are stable around a maximum value. Afterwards, at time t=20 min the current start to decrease, indicating the beginning of the cell adhesion process. This produces a rapidly varying response until t=60 min, when the decrease becomes slower, and the current stabilizes around a minimum value. At t=200 min, we used a cell dissociation agent (trypsin) to completely remove the cell from the sample surface, and devices recovered their original current amplitude.\n\nWe report in Figure 3c the full current spectra acquired before and after the treatment with trypsin (t=180 min and t=240 min, respectively). The cell adhesion produced a large shift of the OECT low-pass cutoff towards smaller frequency. In parallel, the current spectrum of a control device placed in the same reservoir, but with no cell seeded on the PEDOT:PSS layer, remained unaltered (see Supp. Inf. S5). After trypsinization, the initial cutoff frequency is fully recovered. These combined observations clearly demonstrate that the cutoff shift is only caused by the single cell adhesion on the PEDOT:PSS layer. The same considerations can be extended to the PEDOT:PSS microelectrode. Here, the cell adhesion process is revealed by a decrease in the gate current amplitude, which reaches its minimum at t=180 min, coherently with the OECT measurements. After the cell detachment (t=240 min), the current increases to its original values.\n\nTo provide a direct comparison between the sensing performances, we averaged the current amplitudes acquired when the cell is detached (t<30 min and t=240 min) and attached (120