Antigen reactivity defines tissue-resident memory and exhausted T cells in tumours

Antigen reactivity defines tissue-resident memory and exhausted T cells in tumours | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results Antigen reactivity defines tissue-resident memory and exhausted T cells in tumours Thomas N. Burn , Jan Schröder , Luke C. Gandolfo , Maleika Osman , Elanor N. Wainwright , Enid Y. N. Lam , Keely M. McDonald , Rachel B. Evans , Shihan Li , Daniel Rawlinson , Lachlan Dryburgh , Ali Zaid , Zoltan Maliga , Dominick Schienstock , Philippa Meiser , Hyun Jae Lee , Hongjin Lai , Marcela L. Moreira , Pirooz Zareie , Louis H-Y. Lee , Lutfi Huq , Susan N. Christo , Justine J. W. Seow , Keith A. Ching , Stéphane M Guillaume , Kathy Knezevic , Simone L. Park , Maximilien Evrard , Jason Waithman , Thomas Gebhardt , Scott N. Mueller , Georgina E. Riddiough , Marcos V. Perini , Simon C. H. Tsao , Terence P. Speed , Peter K. Sorger , Sherene Loi , Francis R. Carbone , Stephanie Gras , Timothy S. Fisher , Bas J. Baaten , Mark A. Dawson , Laura K. Mackay doi: https://doi.org/10.1101/2025.07.23.666465 Thomas N. Burn 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jan Schröder 2 Computational Science Initiative, Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Luke C. Gandolfo 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia 3 Bioinformatics Division, Walter and Eliza Hall Institute of Medical Research , Melbourne, VIC, Australia 4 School of Mathematics and Statistics, University of Melbourne , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Maleika Osman 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Elanor N. Wainwright 5 Peter MacCallum Cancer Centre , Melbourne, VIC, Australia 6 The Sir Peter MacCallum Department of Medical Oncology, University of Melbourne , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Enid Y. N. Lam 5 Peter MacCallum Cancer Centre , Melbourne, VIC, Australia 6 The Sir Peter MacCallum Department of Medical Oncology, University of Melbourne , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Keely M. McDonald 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Rachel B. Evans 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Shihan Li 2 Computational Science Initiative, Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Daniel Rawlinson 2 Computational Science Initiative, Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Lachlan Dryburgh 2 Computational Science Initiative, Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Ali Zaid 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Zoltan Maliga 7 Laboratory of Systems Pharmacology, Harvard Medical School , Boston, Massachusetts, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Dominick Schienstock 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Philippa Meiser 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Hyun Jae Lee 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Hongjin Lai 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia 8 Department of Thoracic Surgery and Institute of Thoracic Oncology, West China Hospital, Sichuan University , Chengdu, China Find this author on Google Scholar Find this author on PubMed Search for this author on this site Marcela L. Moreira 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Pirooz Zareie 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Louis H-Y. Lee 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Lutfi Huq 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Susan N. Christo 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Justine J. W. Seow 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Keith A. Ching 9 Oncology Research Unit, Pfizer Inc , San Diego, California, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Stéphane M Guillaume 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Kathy Knezevic 5 Peter MacCallum Cancer Centre , Melbourne, VIC, Australia 6 The Sir Peter MacCallum Department of Medical Oncology, University of Melbourne , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Simone L. Park 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Maximilien Evrard 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jason Waithman 10 School of Biomedical Sciences, The University of Western Australia , Perth, WA, Australia 11 Telethon Kids Institute, The University of Western Australia , Perth, WA, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Thomas Gebhardt 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Scott N. Mueller 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Georgina E. Riddiough 12 Department of Surgery, The University of Melbourne , Heidelberg, VIC, Australia 13 HPB & Liver Transplant Unit , Austin Health, Heidelberg, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Marcos V. Perini 12 Department of Surgery, The University of Melbourne , Heidelberg, VIC, Australia 13 HPB & Liver Transplant Unit , Austin Health, Heidelberg, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Simon C. H. Tsao 12 Department of Surgery, The University of Melbourne , Heidelberg, VIC, Australia 14 School of Natural Sciences, Macquarie University , Sydney, NSW, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Terence P. Speed 3 Bioinformatics Division, Walter and Eliza Hall Institute of Medical Research , Melbourne, VIC, Australia 4 School of Mathematics and Statistics, University of Melbourne , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Peter K. Sorger 7 Laboratory of Systems Pharmacology, Harvard Medical School , Boston, Massachusetts, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Sherene Loi 5 Peter MacCallum Cancer Centre , Melbourne, VIC, Australia 6 The Sir Peter MacCallum Department of Medical Oncology, University of Melbourne , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Francis R. Carbone 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Stephanie Gras 15 Infection and Immunity Program, La Trobe Institute for Molecular Science (LIMS) , Bundoora, VIC, Australia 16 Department of Biochemistry and Chemistry, La Trobe University , Bundoora, VIC, Australia 17 Department of Biochemistry and Molecular Biology, Monash University , Clayton, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Timothy S. Fisher 9 Oncology Research Unit, Pfizer Inc , San Diego, California, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Bas J. Baaten 9 Oncology Research Unit, Pfizer Inc , San Diego, California, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Mark A. Dawson 5 Peter MacCallum Cancer Centre , Melbourne, VIC, Australia 6 The Sir Peter MacCallum Department of Medical Oncology, University of Melbourne , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Laura K. Mackay 1 Department of Microbiology and Immunology, The University of Melbourne at the Peter Doherty Institute for Infection and Immunity , Melbourne, VIC, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: lkmackay{at}unimelb.edu.au Abstract Full Text Info/History Metrics Preview PDF Abstract CD8 + T cells are a key weapon in the therapeutic armamentarium against cancer. While CD8 + CD103 + T cells with a tissue-resident memory T (T RM ) cell phenotype have been favourably correlated with patient prognoses 1 – 6 , the tumour microenvironment also contains dysfunctional exhausted T (T EX ) cells that exhibit a myriad of T RM -like features, leading to conflation of these two populations. Here, we deconvolute T RM and T EX cells within the intratumoural CD8 + CD103 + T cell pool across human cancers, ascribing markers and gene signatures that distinguish these CD8 + populations and enable their functional distinction. We found that while T RM cells exhibit superior functionality and are associated with long-term survival post-tumour resection, they are not associated with responsiveness to immune checkpoint blockade. Deconvolution of the two populations showed that tumour-associated T EX and T RM cells are clonally distinct, with the latter comprising both tumour-independent bystanders and tumour-specific cells segregated from their cognate antigen. Intratumoural T RM cells can be forced towards an exhausted fate when chronic antigen stimulation occurs, arguing that the presence or absence of continuous antigen exposure within the microenvironment is the key distinction between respective tumour-associated T EX and T RM populations. These results suggest unique roles for T RM and T EX cells in tumour control, underscoring the need for distinct strategies to harness these T cell populations in novel cancer therapies. Main text T cell-mediated tumour control is a key facet of cancer immunotherapy and pinpointing the most effective T cell subtypes for therapeutic targeting is a critical area of ongoing research. Two subsets of significant interest to the cancer immunotherapy field are tissue-resident memory T (T RM ) cells and exhausted T (T EX ) cells. CD8 + T RM cells are a non-recirculating memory T cell population that reside long-term in every organ examined 7 – 9 . Canonical T RM cells develop following the resolution of infection or inflammation, where they can provide rapid, localised immune protection 10 , 11 . In contrast, CD8 + T EX cells form in the context of chronic infection and cancer, driven by persistent antigen recognition and inflammation 12 – 14 . T cell exhaustion is characterised by the loss of proliferative capacity and function such that restoration of T cell activity forms the basis of successful immune checkpoint blockade (ICB) therapies. Despite this, while T EX cells may be temporarily reinvigorated following ICB, their long-term fate remains largely unaltered 15 . Most recently, T EX -phenotype cells have been shown to become resident in tumours, likely sharing some aspects of the T RM transcriptional program to limit recirculation 16 . Because many studies use genetic signatures to identify T cell subtypes, their transcriptional similarities have resulted in T RM and T EX cells being conflated in the literature. While tumour-associated T RM cells may exist as a potential antipode to the dysfunctional T EX population, their identification within tumours and contribution to cancer control has not been adequately addressed. T EX cells share the T RM gene signature Experimental systems such as parabiosis or transplantation have demonstrated the non-circulating behaviour of T RM cells 10 , 11 , 17 . Such experiments in mice led to the identification of T RM -associated markers including CD69 and CD103 that distinguish T RM cells from circulating memory T cells (T CIRCM ), with markers partially cross-validated in humans by their expression on donor-derived T cells present within transplanted organs 18 – 20 . T RM cells defined by CD69 or CD103 expression are transcriptionally distinct from T CIRCM in humans 7 , 21 , 22 and mice 8 , 23 . At the core of the T RM gene signature exists a transcriptional program designed to halt T cell migration, which includes downregulated expression of key regulators of T cell recirculation such as KLF2 , S1PR1 , and S1PR5 24 , 25 . However, it has recently been shown that T EX cells also cease to migrate and become resident in tumours 16 . Thus, we hypothesised that T EX cells utilise a common transcriptional program to T RM cells to inhibit migration, resulting in considerable transcriptional overlap between these populations. Indeed, we revealed this is the case, with the T RM transcriptional profile derived from acute viral infection models (lymphocytic choriomeningitis virus (LCMV) and herpes simplex virus (HSV)) 8 significantly correlated to the T EX transcriptional profile derived from chronic versus acute LCMV infection 26 ( Fig. 1a ) . Given this overlap between T RM and T EX cells, we reasoned that T RM gene signatures would identify T EX cells in single-cell RNA sequencing (scRNAseq) datasets. To test this, we utilised published data of CD8 + T cells from spleens of mice infected with acute or chronic LCMV 27 . Strikingly, we found the T RM gene signature was most enriched in terminal T EX cells during chronic infection ( Fig. 1b-d , Extended Data Fig 1a ). Further, we analysed CD8 + T cells from murine breast cancer (BC) and adjacent tissue 16 , finding the T RM gene signature highly enriched in tumour-specific T EX cells isolated from tumours, similar to virus-specific cells in adjacent tissue ( Fig. 1e-f ). Thus, the utilisation of T RM gene signatures to identify intratumoural T RM cells results in aberrant T EX cell identification. Download figure Open in new tab Figure 1: Canonical T RM proteins and gene signatures do not deconvolute T RM and T EX cells. a , Log2 fold-change (FC) of differentially expressed T RM 8 and T EX 26 cell genes compared to circulating memory T cells (T CIRCM ). T RM signature genes 8 are highlighted (green=up, blue=down). p-value indicates Fisher’s exact test for association. b-d , scRNA-seq data from LCMV-specific T cells after acute (Arm) or chronic (Cl13) infection isolated from spleens at d8, d15, or d30 post-infection 27 . b , UMAP projection of scRNA-seq data annotated by infection and timepoint, c, quantification of T RM module score 8 and d , overlay of core T RM 8 and Exh-Term 27 transcriptional signatures on respective populations. e-f , UMAP projection ( e ) and quantification of T RM module score 8 ( f ) on LCMV-specific P14 (virus-sp) or tumour-specific OT-I T cells from CITEseq analysis of the respective organs following EO771-OVA BC tumours 16 . g-h , Flow cytometry of CD8 + T cells isolated from BC tumours or breast tissue from N=5 BC patients. g , Representative plots and h , summary data for %CD69 + CD103 + CCR7 - CD45RA - of CD8 + T cells. i-o , CITEseq of CD3 + CD8 + T cells from primary BC tumours (N=5) and non-cancerous breast tissue (N=8). i, Schematic. j, Data was Harmony-integrated, and unified protein and RNA-seq data were represented on weighted nearest neighbours UMAP and coloured by cluster. k , Expression of respective cell surface proteins (αCD103, αCD69, αCD49a) and transcripts ( ITGAE, ZNF683, CXCR6, KLF2) across annotated clusters. l-m , CD8 + T cells segregated by tissue of origin ( l ), and relative cluster composition of CD103 + resident T cells isolated from BC tumours or tissue ( m ). n , PCA of pseudobulked clusters annotated in (j). o, average module scores of published T RM 8 , 21 , 28 and T EX 28 gene signatures by annotated subsets. Many studies have also identified T RM cells in tumours via CD69 and CD103 co-expression 1 – 6 . CD69 + CD103 + CD8 + T cells are present in both non-cancerous tissue and tumours, the latter of which likely encompasses a mixed population of T RM and T EX cells ( Fig. 1g-h , Extended Data Fig. 1b ). Thus, CD69, CD103, and T RM gene signatures all appear insufficient to distinguish T RM and T EX cells should they coexist. Given this overlap in tissue-residency features, we set out to differentiate T RM and T EX populations in human tumours by formulating two testable assumptions. First, cells enriched for expression of core-residency genes and gene signatures in healthy tissue in the absence of overt infection or inflammation are predominantly bona fide T RM cells, while tumour-derived cells expressing these signatures comprise both T RM and T EX cells. Second, that tumour-derived cells expressing core-residency gene signatures can be further segregated into T EX and T RM cells by the relative presence or absence of ‘exhaustion’ gene-signatures. If there are bona fide T RM cells in tumours, they would be expected to be transcriptionally similar T RM cells to those in associated healthy tissue. To this end, we performed single-cell cellular indexing of transcriptomes and epitopes by sequencing (CITEseq) with TCR profiling on CD3 + T cells from human BC tumours and normal breast tissue ( Fig. 1i ). CD8 + T cells from BC tumours and breast tissue were distributed over 15 clusters, classified into 5 major T cell subsets (T EMRA , T EM , MAIT, ψ8 T cells, and CD103 + resident cells) based on protein and transcriptional profiles ( Fig. 1j-k , Extended Data Fig. 1c ). CD103 + “resident T cells” shared expression of canonical residency genes including ZNF683 (HOBIT) and CXCR6 , and downregulation of KLF2 , which controls major tissue egress-promoting gene products 24 ( Fig. 1k ). As per above, we reasoned that tumour-derived CD103 + resident T cells would include both T RM and T EX populations while healthy tissue-derived cells would primarily contain T RM cells. Accordingly, we defined two CD103 + clusters (c0, c5) as T RM cells, based on their over-representation (>90%) within healthy tissue, while clusters (c7, c12, c15), primarily found in tumours and largely absent from healthy tissue, were classified as T EX cells ( Fig. 1l-m ). Pseudobulk PCA analysis confirmed the distinction between T RM and T EX clusters, whilst highlighting their similarity in the PC1 axis, driven predominantly by the downregulation of genes associated with T cell egress including KLF2 ( Fig. 1n , Extended Data Fig. 1d ). Supporting our annotations, published T RM gene signatures 8 , 21 , 28 were enriched within both T RM and T EX populations while T EX gene signatures 28 were selectively enriched within T EX cells ( Fig. 1o ). Segregation of T EX and T RM cell populations revealed that while they share expression of CD103, CD69, and ZNF683 (HOBIT), and the downregulation of KLF2 , T EX cells expressed higher levels of CD38, CD39, PD-1, CTLA-4, TIGIT and HAVCR2 (TIM3) ( Extended Data Fig. 1e-i ). These data show that T EX cells within human tumours co-opt a residency program, and that commonly utilised T RM cell-associated proteins or gene signatures cannot distinguish between T RM and T EX populations. T RM and T EX cells exhibit disparate functional capacities To disentangle tumour-associated T RM and T EX cells, we developed gene signatures from our BC dataset that could accurately distinguish these populations. Genes were included in the T RM gene signature based on differential expression (DE) in T RM cells compared to other T cell subsets, followed by successive DE analysis between T RM and T EX metaclusters. An analogous approach was used to derive the T EX gene signature ( Fig. 2a-c , Extended Data Fig. 2a-b ). While both T RM and T EX cells were defined by CD103 expression and KLF2 downregulation ( Fig. 2d , Extended Data Fig. 2c ) , they were further distinguished by expression of markers including CD94, CD161, CD73, CD38, CD101, CD39, GNLY , and PD-1 ( Fig. 2e , Extended Data Fig. 2d ). This enabled reliable discrimination of the two populations via cyclic immunofluorescence microscopy (CycIF 29 , 30 ) and flow cytometry, allowing examination of their intratumoural location and functional properties. Download figure Open in new tab Figure 2: Spatial and functional characterisation of CD103 + T RM and T EX cells in tumours. a, Volcano plots showing differential expression between T RM , T EX , and all other subsets in BC dataset. b-c, Enrichment of T RM ( b ) and T EX gene signatures ( c ) on labelled subsets. d-e, Flow cytometry, concatenated from N=13 donors (N=11 with tumour sample, and N=10 with healthy tissue). d, expression of CD69 and CD103 on CCR7 - CD45RA - CD8 + T cells. e, expression of T RM and T EX -defining proteins on CD69 + CD103 + T cells. f-h, CycIF imaging of BC tumours. f, representative image of tumour section showing panCK (tumour) and aSMA (stroma) expression, and relative location of respective annotated cell types. Distance to nearest panCK + cell ( g ), and frequency in bins (10μm) segregated by distance of respective cell types to panCK + cells, pooled from N=7 donors (g analysed by t-test, ***p<0.001). Expression of TNF, IFNψ ( i ), IL-2 and CD107a ( j ) on T EX and T RM cells from PMA-ionomycin stimulated CD8 + T cells from BC tumours or tissue as per clusters in Extended Data Fig. 3b . k . Summary of i-j. Donors contributing a minimum of 10 cells within a cluster (T RM or T EX isolated from either tumour or healthy tissue) were enumerated and plotted (N=4 T EX , N=10 T RM ), analysed by t-test, *p<0.05, **p<0.01. l, survival of BC patients from the METABRIC dataset 50 with the highest (top 25%) T RM or T EX gene signature score enrichments compared to patients with lowest (bottom 25%) gene enrichment scores, plotted on Kaplan-Meier curves with log-rank test. m, receiver operating characteristic (ROC) curves from BC patients, either control (Ctrl N=210) or treated with ICB (pembrolizumab/anti-PD-1 N=69) from the iSPY trial 31 . Clinical response to ICB associated with T RM or T EX gene signatures respectively. Using a 47-marker CycIF panel, we identified CD103 + KLF2 - CD8 + T cells across 7 BC patients ( Extended Data Fig. 2e-f ) . These cells were stratified based on the relative expression of GNLY and PD-1 into GNLY Hi PD-1 Lo and GNLY Lo PD-1 Hi populations, approximating T RM and T EX cells, respectively ( Extended Data Fig. 2e-g ) . GNLY Lo PD-1 Hi cells expressed higher levels of CD39 and LAG3, consistent with an exhausted phenotype ( Extended Data Fig. 2h ) . Unbiased clustering of CD103 + KLF2 - CD8 + T cells reinforced this distinction: GNLY Hi PD-1 Lo cells were enriched in cluster c3, which expressed T RM -associated markers including CD94, CD7, and NKG2A, while GNLY Lo PD-1 Hi cells dominated cluster c1 with increased TIM3, LAG3, and CD39 expression ( Extended Data Fig. 2i-k ) , supporting our in situ gating strategy. Spatial analysis revealed that both populations localised near panCK + tumour regions ( Fig. 2f , Extended Data Fig. 2l ) . However, GNLY Lo PD-1 Hi (approximating T EX ) cells were on average significantly closer to panCK + tumour cells, suggesting an increased potential for direct tumour interaction ( Fig. 2g, h ) . Beyond phenotypic and spatial differences, we also observed distinct functional capacities between T RM and T EX populations. Upon ex vivo restimulation, T RM cells exhibited higher production of IFNψ, TNF, and IL-2 and showed elevated expression of granulysin ( GNLY), while T EX cells expressed more granzyme A ( GZMA ) and granzyme K ( GZMK ) ( Fig. 2i-k , Extended Data Fig. 3a-e ) . Moreover, deconvolution of these populations in transcriptomic datasets revealed that enrichment for the T RM gene signature was associated with improved overall survival in BC patients, whereas T EX gene signature enrichment correlated with poorer outcomes ( Fig. 2l ). This association appeared BC-subtype specific, given that triple negative BC (TNBC) survival was best predicted by total CD103 + (both T RM and T EX ) cells, consistent with our prior work ( Extended Data Fig. 3f-h ) 1 . Further, we tested the association of the T RM and T EX signatures with responses to ICB (pembrolizumab/αPD-1) in the I-SPY2 trial 31 , revealing that patients with higher T EX gene signature expression were more likely to achieve a pathologic complete response (pCR) following ICB, whereas elevated T RM signature expression was inconsequential to ICB responsiveness ( Fig. 2m ). These data suggest that while T RM cells are associated with positive prognoses in breast cancer, current ICB therapies exclusively target and enhance T EX cell-mediated anti-tumour responses. T RM and T EX gene signatures delineate T RM cells across tumours To determine the applicability of these BC T RM and T EX gene signatures across tumour types, we next developed signatures from CD8 + T cells isolated from liver tumours for comparison. To this end we performed CITEseq on CD8 + T cells isolated from liver metastases from colorectal cancer patients, paired non-cancerous liver tissue, and liver tissue from cancer-free donors ( Fig. 3a ). Two CD103 + resident (CD69 + , CD103 + , KLF2 - ) T cell clusters were identified and denoted as T EX and T RM populations as described above, with T EX cells mostly present in tumour-derived tissue and T RM clusters present in both tumour and non-cancerous tissue ( Extended Data Fig. 4a-d ). In line with our findings in BC, we found that both T RM and T EX populations were enriched for published T RM but not T EX gene signatures ( Extended Data Fig. 4e ). Importantly, liver T RM and T EX signatures could accurately identify BC T RM and T EX cells ( Fig. 3b ), and similarly, BC T RM and T EX signatures identified liver T RM and T EX cells, respectively ( Extended Data Fig. 4f ). Overall, BC T RM signature genes were highly enriched in liver T RM cells (and vice versa) which was also true for the respective T EX gene signatures highlighting shared gene expression across tumour-associated T RM and T EX cells from disparate tumour types ( Fig. 3c , Extended Data Fig. 4g ). Download figure Open in new tab Figure 3: Specific T RM and T EX gene signatures enable T RM and T EX cell identification across tumours. a, CITEseq of CD3 + CD8 + T cells isolated from secondary liver tumours (colorectal cancer patients, N=4) and non-cancerous liver tissue (N=6). Liver T RM vs. T EX gene signatures were acquired as described for BC signatures. b, Overlay of liver T RM and T EX signature module scores on BC dataset. c, Gene-set enrichment analysis of BC T RM vs. liver T RM signatures and vice versa. d-g, scRNA-seq of tumour-associated CD8 + T cells from various cancers. d, Cancers from pan-cancer atlas 28 . e, Overlay of refined T EX (top) and T RM (bottom) gene signatures on CD8 + T cells on UMAP of pan-cancer scRNA-seq dataset 28 . f, Tumour-derived (top) and healthy tissue-derived (bottom) CD8 + T cells, and annotated clusters. g, Relative frequencies of different CD8 + T cell subsets by cancer 28 . h, Log2 ratio of T EX to T RM across different cancers 28 , 40 , 51 – 55 . i, T EX /T RM ratio association with median tumour mutational burden (TMB) 32 . p-value indicates that the slope of the regression line departs from zero. BC (breast), BCC (basal cell carcinoma, BCL (B cell lymphoma), CRC (colorectal cancer), ESCA (esophageal), FTC (fallopian tube), LC (lung cancer), MEL (melanoma), MM (multiple myeloma), OV (ovarian), PACA (pancreatic), RC (renal), THCA (thyroid), UCEC (uterine corpus endometrial). While the collection of genes within T RM and T EX signatures strongly correlated across BC and liver tumours, the expression of individual genes and surface proteins was not universally consistent. For example, CD94, CD101, CD161 and CD73 were specifically enriched in T RM cells in BC ( Fig. 2e ) but not liver-derived tumours ( Extended Data Fig. 4h ). Therefore, caution must be applied when extrapolating individual genes and proteins across T cell populations in distinct settings. Nonetheless, by focusing on the leading-edge genes, those contributing to the enrichment in each comparison ( Fig. 3c , Extended Data Fig. 4g ), we defined broadly applicable pan-cancer T RM and T EX signatures ( Extended Data Fig. 5a,b ). Using these gene signatures, we could distinguish T RM and T EX cells across a range of tumours from multiple datasets, including a pan-cancer atlas 28 and additional studies 48 – 53 ( Fig. 3d-g , Extended Data Fig. 5c ). A composite of the samples from the pan-cancer atlas showed that T EX cells were predominantly detected in tumour-derived tissue and largely absent from non-cancerous, healthy tissue ( Fig. 3e-f ). When analysed by tumour type, the relative abundance of T EX and T RM cells varied, with melanoma (MEL), oesophageal cancer (ESCA), and B cell lymphoma (BCL) displaying the highest T EX to T RM ratio among the cancers examined ( Fig. 3g-h ). Strikingly, this ratio correlated positively with tumour mutational burden (TMB) 32 , suggesting that neoantigen load may preferentially promote T EX cell formation ( Fig. 3i ). Overall, these data demonstrate our ability to deconvolute T RM and T EX cells across various tumour settings, facilitating a detailed investigation of their functional and developmental differences. Tumour-associated T RM are clonally distinct and enriched for virus-specific cells Given the observed correlation between T EX cell abundance and TMB, we hypothesised that T EX and T RM populations may be clonally distinct, potentially reflecting differences in antigen specificity. To investigate this, we examined clonal overlap among the top 100 expanded TCR clones across T cell subsets in both our BC dataset and the pan-cancer atlas. In both datasets, T RM and T EX cells displayed limited clonal overlap with each other and instead showed greater clonal overlap with T EM cells ( Fig. 4a ). High Jaccard dissimilarity scores supported this, indicating that TCR repertoires of tumour-derived T RM , T EX and T EM cells were largely distinct ( Fig. 4b ). Within our BC dataset, expanded clones were occasionally shared across subsets, with most sharing occurring between T RM and T EM or T EX and T EM , rather than directly between T RM and T EX ( Fig. 4c ) . Since no public TCRs were present across donors in our BC dataset, liver dataset, or the pan-cancer atlas, we pooled the TCR data for further examination. We assessed clonotype sharing among T RM , T EX , and T EM cells at various thresholds based on overlapping cell numbers for each clonotype, and found that most clonal overlap among T RM , T EX , and T EM cells was due to a single shared cell despite significant numbers of expanded clones ( Fig. 4d , Extended Data Fig. 6a ). These data indicate that while a single clone can adopt T RM , T EX , or T EM phenotypes, there is preferential development of one subset for a given TCR. Download figure Open in new tab Figure 4: Tumour-associated T RM and T EX cells are clonally distinct with discrete specificities. a, Relative frequency of the top 100 expanded TCR clones within each metacluster from the BC dataset (top) or pan-cancer atlas (bottom) 28 . b, Jaccard dissimilarity index scores (1 – Jaccard index) for expanded (minimum 2 cells) tumour-derived T cell clones showing dissimilarity between respective populations from BC dataset or pan-cancer atlas, analysed by t-test. c, Venn diagram indicating clonal overlap between expanded T RM , T EM , or T EX clonotypes from BC dataset. d, UpSet plot showing the overlap among T EM , T RM , and T EX cell types from pooled BC, liver, and pan-cancer datasets, with minimum overlap cutoffs of 1, 2, or 3 cells. e , Violin plots representing TCRdist analysis of pooled BC, liver, and pan-cancer datasets, showing the distance between T RM or T EX with respective subsets, where lower values indicate greater similarity. f-g, Distribution, and enumeration of cells expressing virus-specific TCRs as determined by VDJdb 35 – 37 within the BC dataset ( f ) and pan-cancer dataset ( g ). Individual dots indicate virus-specific cells, uniquely coloured by clones ( f ) or phenotype ( g ). h, Frequency of virus-specific cells within each clonotype that adopt T RM or T EX cell phenotypes, analysed using a chi-squared test to determine the relative frequency of a virus-specific T cell adopting each phenotype. i, frequency of virus-specific T cells of total CD8 + T cells in respective datasets. j-l, Clonal sharing between tumour and healthy tissue-derived CD8 + T cells 40 . j, Clonal sharing between subsets detected across tumour vs healthy tissue-derived CD8 + T cells, filtered on clones identified in both tissues. Heatmap scaled by % sharing in each row, numbers indicate shared clones pooled from CRC (colorectal), UCEC (uterine corpus endometrial), LC (lung) and RC (renal) cancers 40 . k, Circos plot indicating selected clones from distinct cancers, and number of cells from each clonotype occupying each tissue and phenotype. l , summary of clonal sharing between tumour T EX and respective phenotype in healthy tissue, split by cancer. n.s. p>0.05, *p<0.05, ****p<0.0001. Structurally similar TCRs are predicted to recognise similar epitopes. Using TCRdist 33 , 34 , we identified clear structural segregation between T RM and T EX cell TCRs, with significant structural similarity predicted only among cells of the same phenotype. This finding suggests that T RM and T EX populations possess distinct epitope specificities ( Fig. 4e ). To investigate further, we integrated TCR sequences and HLA allele expression with VDJdb 35 – 37 to predict viral reactivity of CD8 + T cells in both our BC dataset and the pan-cancer atlas. Predicted virus-specific clonotypes were predominantly associated with a T RM phenotype, while virus-specific T EX cells were exceedingly rare ( Fig. 4f-h ). Interestingly, the propensity to adopt a T RM cell phenotype varied depending on viral specificity. Predicted influenza A-specific cells most frequently exhibited a T RM phenotype, while EBV-specific T RM -like cells were rarely identified ( Extended Data Fig. 6b-c ). Critically, only a very small frequency of the total CD8 + T cells in both datasets was predicted to be virus-specific ( Fig. 4i ) , yet these data indicate that tumour-antigen-independent cells are enriched for the T RM gene signature. In contrast, previous studies have shown that tumour-antigen-specific cells express markers such as CD39 38 , 39 , which are characteristic of T EX cells in our classification. Based on this, we hypothesised that if tumour-derived clones expressing the T EX gene signature are also found in healthy tissue, where tumour antigen is presumably absent, they may instead show enrichment of the T RM gene signature. Although clonal sharing between tumour and healthy-tissue-derived cells in our BC and pan-cancer datasets was limited, likely due to inadequate sampling, we leveraged a dataset in which T cell clones were identified in both tumour and adjacent healthy tissue 40 . We extracted CD8 + T cell clones shared across sites and annotated them based on our T RM and T EX gene signatures. Consistent with our hypothesis, there was significant clonal sharing between tumour-derived T EX and healthy tissue-derived T RM cells ( Fig. 4j-k , Extended Data Fig. 6d-e ) . Indeed, when clonal overlap was observed, tumour-derived T EX clonotypes were more frequently shared with healthy T RM than with other T cell types ( Fig. 4l ) . Together, these data suggest that the T RM gene signature is enriched in predicted tumour-antigen-independent memory T cells, and tumour-specific cells in healthy non-cancerous tissues. Antigen drives the distinction between T RM and T EX cells in tumours Given that putative tumour-antigen-specific T cells preferentially adopt a T RM phenotype in surrounding tissues where tumour antigen may be absent, we hypothesised that tumour-antigen reactivity drives the divergence between tumour-associated T RM and T EX cells. To test whether tumour-specific T RM can form in tumours, we utilised an orthotopic murine BC (AT3) model engineered to express the model antigen ovalbumin (OVA). High-dimensional flow cytometry of tumour-infiltrating T cells (TILs) revealed 4 clusters enriched for tumour-specific OT-I T cells expressing CD69 and PD-1, and differing in markers such as CD39, CD103, Tcf1, Tox and Tim3 ( Fig. 5a-f , Extended Data Fig. 7a ) . We annotated these subsets based on phenotypic and functional characteristics. First, we defined resident cells as non-migratory. In lieu of definitive migration assays (e.g., parabiosis), we used indirect indicators of tissue residency, including intravenous (IV) labelling and FTY720 treatment. Over 90% of TILs were IV-negative, and CD69 + TILs showed reduced IV-staining relative to CD69 - TILs ( Extended Data Fig. 7b ) . TIL frequencies were also unaffected by FTY720 treatment ( Extended Data Fig. 7c ) , supporting the notion that the majority of TIL are resident. Download figure Open in new tab Figure 5: Low-avidity and bystander CD8 + T cells preferentially adopt a T RM phenotype in tumours. a-f, Mice received 1×10⁴ naïve OT-I T cells and were challenged with AT3-OVA; tumour-infiltrating T cells were analysed on day 22. a, Experimental schematic. b, UMAP of CD8⁺ TILs (defined by markers in Extended Data Fig. 7a ) showing OT-I cells and FlowSOM clusters; Cluster 7 was stratified by Tox expression (into c7a (Tox Lo ) and c7b (Tox Hi )). c, Distribution of OT-I cells across FlowSOM clusters. d, Relative marker expression by cluster. e-f, Representative gating (e) and marker expression (f) from tumour-derived OT-I cells, Endo. CD8: endogenous CD8⁺ T cells. g, Cytokine expression post-PMA/ionomycin restimulation. h, OT-I frequencies at d 14 and 22. i, T RM cell number per gram tumour vs. tumour mass at day 14. j, Proportion of T RM and Tox Hi OT-I vs. %GFP-OVA + AT3 at d 22. k-l, Effector OT-I (strong TCR signaling) and OT-3 (weak TCR signaling) T cells were co-transferred into tumour-bearing mice (day 10); intratumoural T cells were analysed on day 28. k, Schematic. l, Frequency of CD69⁺ cells and PD-1 expression. m-n, LCMV-immune mice (receiving 5×10⁴ P14 prior to infection) were challenged with AT3-OVA >100d post-LCMV infection; intratumoural T cells analysed on day 23 post-tumour inoculation. m, Schematic. n, Frequencies of CD45.1⁺ LCMV-specific (P14) vs. SIINFEKL-tetramer⁺ tumour-specific cells among CD8⁺CD69⁺ TILs. o-r, scRNA-seq of bystander (OT-I) and Ag-specific (gBT-I) CD8⁺ T cells co-transferred into mice with B16-gB melanoma tumours. o, Schematic. p, UMAP of sorted cells coloured by transgenic T cell. q-r, Overlay of BC T EX (q) and T RM (r) gene signatures. p-values from Wilcoxon signed-rank test. Statistics: b-j,l, pooled from 2 independent experiments; n, representative of 2 experiments; b-f, representative of >8 total experiments, minimum N=5 mice/group/experiment. g: one-way repeated-measures ANOVA with Tukey post-test; h: two-way ANOVA with Sidak test; i,j: r 2 indicates fit of linear regression line and p-value indicates slope departs from zero. l,n: two-way repeated-measures ANOVA with Sidak test. l,n: connected points from individual mice. n.s., p > 0.05; *p<0.05, **p < 0.01; ***p < 0.001; ****p < 0.0001. Second, we considered cells exhausted if they exhibited functional impairment, and memory cells as those capable of persisting without ongoing antigen stimulation. Accordingly, CD39 + CD103 + cells were annotated as T RM cells based on superior functionality, while CD39 + CD103 - Tox Hi cells represented the most dysfunctional, bona fide exhausted population ( Fig. 5g , Extended Data Fig. 7d-e ) . A separate CD69 + CD39 - CD103 - population expressed Tcf1, PD-1, and Tox, consistent with progenitor exhausted T cells (T PROG ), which were selectively depleted when Tcf7 expression was ablated ( Fig. 5e-f , Extended Data Fig. 8a-d ) . Strikingly, we found that the number of T RM phenotype cells inversely correlated with tumour size at early timepoints (d14) ( Fig. 5i ) , and acquisition of the T RM phenotype correlated with the loss of OVA-GFP expression from tumour cells over time (d23) ( Fig. 5j ) . In contrast, Tox Hi exhausted cells were enriched in tumours that maintained high tumour-antigen load, consistent with the role of antigen in driving T cell exhaustion ( Fig. 5j ) . Inducible TCR depletion in tumour-specific T cells increased both the frequency and number of CD103 + T cells ( Extended Data Fig. 8e-h ) , further supporting that CD103 expression marks memory T cells in this model, and that tumour-specific T RM cells preferentially form or persist when cognate antigen is absent, consistent with classical T cell memory paradigms. Given that the presence or absence of intratumoural TCR signalling appeared to influence T EX versus T RM cell fate, we next assessed whether reducing TCR signal strength would also influence this distinction. Indeed, OT-3 T cells, which have reduced TCR signalling towards OVA-peptide compared to OT-I T cells, displayed increased CD103 and decreased PD-1 expression ( Fig. 5k-l ) , indicating that low-avidity TCR signalling promotes T RM cell differentiation within tumours. Consistent with the T RM phenotype of virus-specific T cells in human tumours, virus-specific memory T cells generated >100 days after LCMV infection infiltrated AT3-OVA tumours and adopted a T RM phenotype ( Fig. 5m, n ). Furthermore, transfer of tumour-specific (OT-I) and non-specific bystander (gBT-I) effector T cells into tumour-bearing mice showed that bystander T cells more readily acquired a T RM phenotype ( Extended Data Fig. 8i-j ). These bystander T cells expressed significantly lower levels of T EX -related proteins such as PD-1, underscoring phenotypic differences between tumour-specific and bystander-derived T RM populations ( Extended Data Fig. 8j ). The development of intratumoural bystander T RM cells required intrinsic TGFβ-signaling ( Extended Data Fig. 8k-l ), and their efficient tumour entry was dependent on CXCR3 and CXCR6 expression ( Extended Data Fig. 8m-n ) . Given the apparent correlation between antigen presence and T RM cell biasing, we speculated that the distinct transcriptional T EX and T RM cell signatures derived from patient samples above may similarly correlate with tumour-antigen reactivity. To assess this, we performed scRNAseq on tumour-specific (gBT-I) and bystander (OT-I) T cells derived from B16-gB tumours and observed discrete clustering of these populations ( Fig. 5o,p ). Notably, tumour-specific T cells displayed enrichment of the human BC T EX gene signature ( Fig. 5q ), whereas bystander T cells exhibited high expression of the BC T RM gene signature ( Fig. 5r ). Consistently, bystander T cells from E0771 BC tumours 16 expressed the T RM gene signature and tumour-specific T cells expressed the T EX gene signature ( Extended Data Fig. 8o-q ). Thus, the relative presence or absence of intratumoural TCR signalling drives the distinct transcriptional profiles of tumour-associated T RM and T EX cells. Altogether, these data indicate that strong and persistent TCR signalling is antithetical to T RM cell development, suggesting that bona fide T RM cells within human tumours are likely tumour-agnostic bystanders, low-affinity tumour-specific T cells, or T cells not in contact with their cognate antigen. Tumour antigen drives T RM cells towards a T EX cell fate The observation that T RM and T EX cells are clonally distinct in human tumours likely reflects intrinsic differences in antigen reactivity, rather than indicating that these populations derive from separate precursor cells. Notably, this clonal distinction does not preclude the presence of tumour-specific T RM cells, nor the possibility of clonal overlap with T EX cells. Indeed, our analyses suggested that such clonal overlap could occur when tumour-specific T RM cells are present in tumours or in tissues where cognate antigen is presumably absent. As such, these results did not resolve whether tumour-specific T RM and T EX cells arise from a shared progenitor or from distinct developmental lineages. To address this, we employed the SPLINTR barcoding system 41 to introduce unique barcodes into mono-specific naïve or effector OT-I T cells before adoptive transfer into AT3-OVA tumour-bearing mice ( Fig. 6a ). To generate naïve barcoded T cells, we transduced hematopoietic stem cells from OT-I donor mice with SPLINTR-encoding lentivirus and performed intra-thymic injections into sub-lethally irradiated recipients. 8 weeks later, we pooled naïve T cells from 20 chimeric donor mice to increase barcode diversity and transferred either 2×10 3 or 1×10 4 naïve OT-I T cells into recipient mice subsequently inoculated with AT3-OVA. 24 days after tumour inoculation, we sorted SPLINTR-barcoded OT-I T cells from spleens and tumour populations, including CD69 - , T PROG , T RM , and the heterogeneous CD39 + CD103 - population we refer to as “T EX -like”. Barcode distribution was assessed by DNA sequencing. Download figure Open in new tab Figure 6: Tumour-associated T RM cells can be driven towards exhaustion. a-d, SPLINTR barcode-seq of naïve (T N ) or effector (T EFF ) OT-I T cells transferred into mice AT3-OVA tumour-bearing mice. a, Schematic. TIL populations as defined in Fig. 5e . T EX -like population includes Tox Lo and Tox Hi populations. b, Barcode representation in a T N transferred mouse sorted by total pooled barcode where bubble size reflects clone proportion in sample. c-d, Pearson correlation of barcodes identified in respective populations from representative mouse (heatmaps) and all repeats compared to T RM population (line-plot) from T N transfer (c) or T EFF transfer (d). e-i, Sort-retransfer of respective OT-I populations from AT3-OVA mice. e, Schematic. Sorted cells were co-transferred intratumourally into recipient mice. f-g , % of populations isolated from AT3 bearing recipient mice, split by input cell phenotype, and output phenotype in recipient mouse, showing independent mice (f) and summary (g). h-i , % of populations isolated from AT3-OVA bearing recipient mice, split by input cell phenotype, and output phenotype in recipient mouse, showing independent mice (h) and summary (i). Statistics: b-c, representative of 8-independent biological replicates from 3 experiments; 1-way repeated measures ANOVA with Tukey post-test. d, representative of 2-independent biological repeats each with technical replicates (not shown); 1-way repeated-measures ANOVA with test for linear trend. f-i, pooled from 2-independent experiments, N=19 total mice; 2-way repeated measures ANOVA with Sidak post-test n.s. p>0.05, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Minimal barcode sharing was observed between mice receiving 2×10 3 OT-I T cells, and sharing was only detectable in mice that received 1×10 4 cells ( Extended Data Fig. 9a ) , indicating that most barcodes were unique and minimising the possibility of PCR artifacts. To avoid the use of highly duplicated barcodes in the original naïve pool, we excluded any barcodes identified across multiple mice before assessing barcode sharing across splenic and tumour populations within individual mice. These analyses showed that antigen-specific, tumour-derived T PROG , T EX -like, and T RM cells were more similar to each other, and distinct from CD69 - and spleen-derived populations ( Fig. 6b-c ) . We conducted a parallel experiment using effector OT-I T cells transduced with SPLINTR-encoding retrovirus, transferred into AT3-OVA-bearing mice. Barcode diversity was again determined across spleen and tumour populations, confirming that barcoded OT-I library pools were unique ( Extended Data Fig. 9b ). As per the naïve T cell experiment, barcode distribution reinforced that tumour-derived T PROG , T EX -like, and T RM cells displayed a high degree of barcode overlap and were distinct from CD69 - and spleen-derived populations ( Fig. 6d , Extended Data Fig. 9c ). Altogether, these data indicate that tumour-specific T RM cells do not arise from a distinct T cell lineage but instead share common progenitors with other intratumoural populations such as T EX -like and T PROG cells. The developmental association between each cell type left unresolved whether tumour-antigen-specific T RM cells that develop after antigen loss or following spatial segregation from cognate antigen can transition into T EX cells upon antigen re-encounter. To investigate this, we isolated intratumoural OT-I T cells exhibiting T PROG , T EX -like, or T RM phenotypes from AT3-OVA tumours and re-transferred them intravenously into secondary recipient mice bearing AT3-OVA tumours ( Extended Data Fig. 9d ). Among these populations, T EX -like OT-I T cells were significantly less efficient at repopulating tumours compared to T PROG or T RM cells ( Extended Data Fig. 9e-f ). Due to the relatively low recovery of transferred cells, we performed composite protein expression profiling from each transferred group by experiment, and compared these profiles to T PROG , T EX -like, and T RM OT-I T cells isolated from concurrently analysed primary tumours ( Extended Data Fig. 9g-h ). This analysis revealed that all re-transferred populations in secondary tumours most closely resembled the T EX -like phenotype observed in primary tumours. To determine whether differences in trafficking influenced the ability of cells to repopulate tumours and adopt distinct phenotypes, we sorted congenically distinct T EX -like and T RM OT-I T cells from primary AT3-OVA tumours and co-transferred them directly into secondary tumours that either expressed or lacked OVA ( Fig. 6e ) . Notably, direct intratumoural transfer eliminated the repopulation advantage previously observed for T RM cells, potentially reflecting their enhanced ability to repopulate distal sites following intravenous transfer ( Extended Data Fig. 9i ) . In the absence of cognate antigen, transferred T RM cells largely retained their phenotype, whereas approximately 50% of the heterogenous T EX -like population (input, comprising Tox Lo and Tox Hi ) could adopt the T RM phenotype ( Fig. 6f-g ) . Strikingly, the dysfunctional Tox Hi population was absent within AT3 tumours lacking antigen, consistent with earlier data showing these cells are sustained in the presence of antigen ( Fig. 6f-g ) . Conversely, when the same populations were transferred into AT3-OVA tumours, a fraction of T RM cells maintained their phenotype; however, the majority transitioned to a T EX -like state, including a significant proportion acquiring the Tox Hi phenotype ( Fig. 6h-i ) . Thus, T RM cells can be driven towards terminal exhaustion within the antigen-rich environment of secondary tumours. Overall, these data indicate that, unlike non-tumour reactive bystander T RM cells, tumour-specific T RM and T EX populations can arise from a common origin, with tumour-specific T RM cells having the capacity to be driven towards exhaustion upon chronic antigen re-encounter. Discussion Our study for the first time reconciles two critical T cell subsets associated with tumour control, namely CD8 + T RM and T EX cells. We show that these subsets are distinct T cell populations that have been conflated in the literature. This conflation has arisen due to T EX cells engaging a residency program that relies on the same transcriptional machinery used by T RM cells to inhibit tissue egress. Thus, T RM gene signatures developed without T EX cell consideration cannot disentangle these two populations. This study resolves this issue by establishing broad gene signatures that reliably distinguish these subsets across various human tumours. The deconvolution of tumour-associated T RM and T EX cells highlights their potential to play distinct roles in anti-tumour immunity and their differential impact on ICB responses. High T EX gene signature expression in BC correlates with positive ICB outcomes reflecting the therapy’s aim to reinvigorate T EX cells through blockade of inhibitory receptors like PD-1 and CTLA4 which are expressed at much higher levels in T EX than T RM cells. The association of high TMB with increased T EX cell frequency suggests that increased novel tumour epitopes enable T EX cell development. T EX cells, as defined in this study, consistently co-express CD39 and CD103, markers known to enrich tumour-specific CD8 + T cells 38 , 39 , 42 . High TMB is generally associated with better ICB responses, including BC subtypes such as triple-negative breast cancer (TNBC), where TMB predicts favourable outcomes to therapies like pembrolizumab 43 – 46 . The observed link between high T EX gene signature scores, TMB, and ICB efficacy underscores the critical role of T EX cells in tumour control. We found that most tumour-associated T RM cells identified in humans are clonally and developmentally distinct from T EX cells. Tumour-antigen agnostic T cells, including T cells specific for previous infections, can adopt a residency phenotype within tumours. Consistently, we show that TCR signalling is antithetical to bona fide T RM cell formation in tumours, aligning with the conceptual notion that immunological memory can only develop following clearance of cognate antigen. However, we also reveal that tumour-antigen-specific T RM cells can develop in settings of reduced TCR signaling and diminished antigen sensing. Tumour-specific T RM cells likely exist in lower numbers compared to tumour-specific T EX cells since it is expected that they have encountered less antigen, and therefore had significantly reduced proliferative bursts, accounting for the limited clonal sharing that we observe between T RM and T EX cells in human tumours. While T RM gene signatures are associated with overall survival in BC patients, the mechanisms underlying T RM -mediated tumour control remain unclear. It is possible that the accumulation of these T RM cells may coincide with other features of tumour control, such as tumour antigen clearance. However, our analyses suggest that tumour-specific T RM cells exist in healthy tissues surrounding tumours, raising the possibility that they contribute to long-term immune surveillance and protection against tumour recurrence. Thus, the clinical benefit of tumour-specific T RM cells may lie in their ability to maintain equilibrium with residual cancerous cells, thereby preventing tumour recurrence, or in their potential prophylactic use following vaccination 47 . Given that these cells develop after clearance of their cognate antigen and have an increased ability to traffic and repopulate distal sites, the presence of T RM cells in human tumours may reflect an effective immune response against tumour antigen that can lead to lasting protection both at the primary tumour location, and potential sites of metastasis. Finally, significant populations of bona fide, non-exhausted, tumour-agnostic T RM cells can be identified within diverse tumours. These cells maintain superior functionality in humans and murine models but are not targeted by current immunotherapies. Thus, approaches to activate bystander T RM cells via TCR-independent pathways or administration of viral peptides 39 , 48 , 49 , or to redistribute functional tumour-specific T RM cells, combined with current T EX cell-targeted ICB therapies, could raise the ceiling of effective anti-tumour responses. Extended Data Download figure Open in new tab Extended Data Figure 1: Identification of CD8 + T cell populations in tumours. a, T RM gene signature expression on LCMV-specific CD8 + T cell clusters from scRNA-seq of the respective dataset 27 (Related to Fig. 1b-d ). b , Flow cytometry of CD8 + T cells isolated from liver tumours and liver tissue from N=4 colorectal cancer patients with liver metastases. Representative plots and summary data for CD69 + CD103 + CCR7 - CD45RA - CD8 + T cells, analysed by Mann-Whitney test, *p<0.05. c, Expression of respective lineage-defining genes and cell surface proteins from BC-CITEseq data ( Fig. 1i-j ) displayed on heatmap by cluster. d, pathway enrichment of downregulated loading genes associated with PC1 from PCA plot in Fig. 1n . Red bars indicate pathways related to chemokine receptor signaling and trafficking e, Schematic showing metacluster annotations. f-g , Expression of selected shared and distinct T RM and T EX genes ( f ) or cell surface proteins ( g ) across metaclusters from BC dataset. h-i , heatmaps showing top DE genes ( h ) and cell surface proteins ( i ) from BC dataset. Download figure Open in new tab Extended Data Figure 2: Lineage-defining genes and proteins enable in situ localisation of T RM and T EX cells. a, T RM and T EX signature genes. Shared genes indicate genes are up (or down)-regulated in the T RM vs all and T EX vs. all comparisons ( Fig. 2a ). T RM ( b ) and T EX ( c ) gene signature module scores displayed on BC UMAPs for summary module score data ( Fig. 2b-c ). c-d , Expression of respective genes and cell surface proteins displayed on BC UMAPs. e-l, CycIF imaging of BC tumours related to Fig. 2f-h . e , representative image of BC tumour using CycIF approach showing respective stains. f , Histocytometric gating strategy showing CD8 + T cells (CD3 + CD8 + CD4 - ) concatenated from CycIF images from 7 donors showing CD103, KLF2, GNLY, and PD-1 expression on respective populations. g, number of CD8 + CD103 + KLF2 - T cells identified in each slide from respective donors, broken down by GNLY and PD-1 expression based on gates in (f). h , Expression of CD39 and LAG3 proteins on respective populations, pooled from 7 donors. Mann-Whitney test ****p<0.001. i , UMAP projection and clustering of CD8 + CD103 + KLF2 - T cells, coloured by gates defined in (f). j , number of cells from respective gates identified within each cluster. k , relative expression of respective proteins in each cluster. l, representative image of tumour section showing panCK and aSMA expression, and relative location of respective annotated cell types. Download figure Open in new tab Extended Data Figure 3: Functional assessment and survival associations of BC T RM and T EX cells. a-e, Flow cytometry of PMA-ionomycin restimulated BC and breast tissue-derived CD103 + resident-phenotype T cells related to Fig. 2d-k . a , Schematic. b-c, CD69 + CD103 + CD45RA - CCR7 - CD8 + T cells represented in UMAP space based on the expression of CD69, CD101, CD39, CD94, CD73, CD103, CD45RA, CCR7, CD38 showing CD101 and CD39 expression ( b ) and annotated T RM and T EX populations and relative contribution by cells derived from BC tumours or tissue ( c ). d, Gating of TNF, IFNψ, IL-2, and CD107a for Fig. 2i-k based on unstimulated control. e, Expression of granulysin ( GNLY ), granzyme A ( GZMA ) and granzyme K ( GZMK ) on T RM and T EX populations from BC CITEseq dataset. f-h, Survival of BC patients from the dataset with the highest (top 25%) T RM ( f ), T EX ( g ), or combined T RM + T EX union ( h ) gene signature score enrichments compared to patients with lowest (bottom 25%) gene enrichment scores and plotted on Kaplan-Meier curves with log-rank test, as in ( Fig. 2l ), segregated by BC subtypes (as annotated). Download figure Open in new tab Extended Data Figure 4: Deconvolution of CD103 + T RM and T EX cells in liver metastases a-d, CITEseq of CD3 + CD8 + CD4 - non-MAIT cells isolated from secondary liver tumours (colorectal cancer patients, N=4) and non-cancerous liver tissue (N=6) as depicted in Fig. 3a . a, Data was Harmony integrated, and unified protein and RNA-seq data represented on weighted nearest neighbours UMAP, coloured by clusters that were annotated based on expression of lineage-defining cell surface proteins and genes. b, Expression of selected cell surface proteins (αCD69, αCD103,) and genes ( ITGAE, ZNF683, CXCR6, KLF2 ) on respective clusters. c-d, CD8 + T cells segregated by tissue of origin ( c ), and relative cluster composition of CD103 + resident T cells (c9+c10) isolated from BC tumours or tissue ( d ). e, Average module scores of published T RM 8 , 21 , 28 and T EX 28 cell gene signatures by annotated clusters. f, Module score overlays of relevant signatures on liver CITEseq dataset. g, gene-set enrichment analysis of BC T EX vs liver T EX signatures and vice versa. h, Expression of respective cell surface proteins across annotated clusters. Download figure Open in new tab Extended Data Figure 5: Pan-cancer T RM and T EX signature genes. a-b , Relative expression of the top genes from pan-cancer T RM ( a ) and T EX ( b ) gene signatures across respective CD8 + T cell populations in the pan-cancer atlas 28 . c , Relative frequencies of different CD8 + T cell subsets by cancer from respective datasets 48 – 53 . Download figure Open in new tab Extended Data Figure 6: Assessment of virus-specific clones and TCR sharing across tissues. a, Clone size distribution across respective datasets. b-c , Frequency and count of cells from BC dataset (b) and pan-cancer atlas (c) split by respective virus-specificity. d-e, Clonal sharing between tumour and healthy tissue-derived CD8 + T cells 40 related to Fig. 4j-l . Clonal sharing between subsets detected across tumour vs healthy tissue-derived CD8 + T cells, filtered on clones identified in both tissues. d, Heatmap scaled by % sharing in each row, split by respective cancers 40 . e, Circos plots indicating selected clones from respective cancers, and number of cells from each clonotype occupying each tissue and phenotype. Download figure Open in new tab Extended Data Figure 7: Phenotypic and functional assessment of AT3-OVA CD8 + TIL populations. a , Expression of respective proteins on CD8 + T cells isolated from AT3-OVA tumours projected in UMAP space, related to Fig. 5a-d . b , Gating and intravenous (IV) labelling of respective CD8 + T cell populations from spleen and tumour at d24 post-inoculation. c , FTY720 treatment of AT3-OVA bearing mice showing relative cell number (normalised to median of control group in each experiment) and frequency of CD69 + PD-1 + cells within CD44 + CD8 + T cells from tumours, and total CD8 + T cells from the blood. d, representative gating of IFNψ, TNFα, and IL-2 on respective populations (as in Fig. 5e ). Gates based on unstimulated control. e , % expression of IFNψ, TNFα, and IL-2 on respective endogenous CD8 + (top) and OT-I (bottom) T cell populations. Statistics: a-e, pooled from 2 independent experiments, minimum N=5 mice/group/experiment. b , one-way repeated-measures ANOVA with Holm-Sidak test, connected points from individual mice. c , Two-way ANOVA with Tukey’s test. e , one-way repeated-measures ANOVA with Tukey post-test. n.s., p > 0.05; *p<0.05, **p < 0.01; ***p < 0.001; ****p < 0.0001. Download figure Open in new tab Extended Data Figure 8: Mechanistic assessment of factors controlling formation of TIL populations. a-d , CRISPR/Cas9 depletion of Tcf7 from naïve OT-I T cells on the development of T cell populations in AT3-OVA tumours harvested 24 d post-inoculation. a , TCF1 expression on co-transferred sg CD19 (control) and sg Tcf7 treated OT-I T cells isolated from tumours. b , Representative gating of CD69 + OT-I T cells. c , frequency and number of CD69 + OT-I T cells from respective gates as in (b). d , Log2 fold-change of cell numbers of respective populations. e-f , effect of TCRα depletion on TIL populations. Congenically distinct TCRα +/+ Cre ERT2 and TCRα fl/fl Cre ERT2 mice were infected with LCMV-OVA, SIINFEKL-tetramer + CD8 + T cells were sorted from spleens at d8 post-infection and co-transferred into AT3-OVA bearing mice 10 d post-tumour inoculation. Mice were then treated with Tamoxifen (Tam) or vehicle control on d16-20 post-tumour, and tumour-derived cells isolated on d27. e , schematic. f , TCRβ staining on respective populations. g , % CD103 expression of respective populations. h , ratio of cell counts for respective populations. i-j , Effector OT-I (Ag-specific) and gBT-I (bystander) CD8 + T cells were co-transferred into mice 10 d post-AT3-OVA inoculation and intratumoural T cells analysed 7 d later. i , Schematic. j , Frequency and PD-1 expression of respective CD69 + populations. k-l , Effector control (TGFβRII +/+ ) and TGFβRII -/- bystander OT-I T cells were transferred into AT3 tumour-bearing mice 10 d post inoculation. Representative gating (k) and cell counts (l) of respective populations harvested from tumours 6 d following transfer. m-n , Effector control (sg CD19 ) and CXCR3 (m, sg CXCR3 ) or CXCR6 (n, sgCXCR6 ) deficient gBT-I T cells were co-transferred into AT3-OVA tumour bearing mice 10 d post-inoculation and analysed in spleens and tumours 7 d later. Expression of CXCR3 or CXCR6, respective cell counts, and ratios of cells are shown. o-q, Analysis of CITEseq of LCMV-specific P14 or tumour-specific OT-I T cells isolated from EO771-OVA BC tumours 16 . o, Unified cell-surface protein and RNA-seq expression data represented on weighted nearest neighbours UMAP coloured by T cell transgenic. p-q, quantification of BC T EX ( p ) and T RM ( q ) signature module scores, p-value calculated by Wilcoxon signed-rank test. Statistics: a-d , i-l, pooled from N=2 independent experiments; m-n, representative of N=2 independent experiments with a minimum of 5 mice per group. c,g,j,m,n, linked symbols indicate cells from the same mouse, 2-way repeated-measures ANOVA with Sidak post-test. d,h, 1-way ANOVA repeated-measures ANOVA with Tukey post-test. l, 2-way ANOVA with Sidak post-test. p-q, Wilcoxon signed-rank test. p>0.05, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Download figure Open in new tab Extended Data Figure 9: Developmental relationship between intratumoural T RM and T EX cells. a-c, Related to Fig. 6a-d . a , heatmap showing counts of individual barcodes (y-axis) across different replicate mice and sorted populations from naïve OT-I SPLINTR experiments. b , Total barcode-seq counts of each identified barcode from two (M1 vs M2) effector OT-I SPLINTR experimental replicates. c, Barcode-seq data from representative effector OT-I SPLINTR experiment (M1) sorted by frequency in the tumour T RM sample, where bubble size reflects clone proportion in sample. d-h , Sort and intravenous retransfer of AT3-OVA intratumoural populations. d, Schematic. e, Enumeration of transferred populations 28 d post-tumour inoculation as the frequency of input cell number. f , Cell numbers of respective populations isolated from spleen and tumours, normalised to input frequency, isolated 28 d post-tumour inoculation. g, Relative expression of proteins on respective populations of primary transferred (Input) and retransferred (Output) cells 28 d post-tumour inoculation determined by flow cytometry with unbiased hierarchical clustering indicated and h, PCA indicating expression of proteins in ( g ), each dot being concatenated samples from independent experiments. i , related to Fig. 6e-i , showing total cell counts isolated from tumours at d 22 post-tumour inoculation. Statistics: a, pooled from 3 independent experiments, with 10 total replicates (i.e. M1-10). b, M1 and M2 indicate independent experiments, each with technical replicates (not shown). d-h, Data pooled from 3 independent experiments, N=11-23 mice per group. n.s. p>0.05, *p<0.05, ***p<0.001. i , Data representative of 2-independent experiments, linked symbols indicate cells isolated from the same mouse, N=9 mice per group, analysed by 2-way repeated-measures ANOVA with Sidak post-test. Methods Mice C57BL/6, gBT-I:CD45.1.2, gBT-I:CD45.1.1, gBT-I:CD90.1.2, P14:CD45.1.2, OT-I:CD45.1.2, OT-I:CD90.1.2; OT-3:CD45.1.2:TCRα -/- 56 , OT-I:TGFβRII fl/fl :Cre dLck :CD45.1(TGFβRII -/- ) and TCRα fl/fl Cre ERT2 female mice were bred and maintained in the Department of Microbiology and Immunology, University of Melbourne under a 12h/12h light/dark cycle, at 19-22°C and 40-70% humidity. All experiments were approved by the University of Melbourne Animal Ethics Committee (ID nos. 21651 and 21938). All mice were between 6-14 weeks of age at the beginning of the experiments. Human studies This project was approved by the Human Research Ethics Committee of the University of Melbourne (ID nos. 13009 and 14517). All participating patients provided written informed consent. Isolation of lymphocytes from human tumours and tissues Following tumour excision, a representative tumour fragment was processed to generate single-cell suspensions as previously described 1 . Briefly, tumour or healthy tissues were finely diced in RPMI1640 containing 10% FCS and 0.5mg/ml collagenase D (Worthington Biochemical, Lakewood, NJ) and were incubated for 30 min at 37°C. Digested pieces were mashed through 70-μm strainers and washed with RPMI1640 with 10% FCS. Lymphocytes from liver tumours and healthy liver tissues were enriched through density gradient centrifugation (500 g , 20min at 25°C) on a 44% / 70% isotonic Percoll gradient (GE Healthcare, diluted in HBSS, and cells at the solution interphase were isolated and washed with HBSS. Blood was diluted 1:1 in HBSS, overlaid on Ficoll-Paque PLUS (Sigma Aldrich), centrifuged (400g, 15min, 25°C), and peripheral blood mononuclear cells (PBMCs) isolated from interphase. Cells were either utilised immediately for flow cytometry, or frozen in 10% DMSO: 90% FCS freezing media (breast, breast tumour, and blood samples) or Cryostor CS10 Freeze Media (StemCell Technologies, #07930; liver and liver tumour samples) for CITEseq and restimulation assays. Murine tumour and infection models The murine TNBC cell line, AT3-OVA was provided by Prof. Phillip Darcy (Peter MacCallum Cancer Centre, Melbourne, VIC, Australia) 57 , 58 . AT3-OVA cells were cultured with complete DMEM (DMEM, 10% FCS, 2 mM L-glutamine, 100U ml -1 penicillin, 100mg ml -1 streptomycin). 5×10 5 AT3-OVA cells in exponential growth phase (∼70-80% confluency) were injected orthotopically into the 4 th mammary fat pad in a total volume of 50µl HBSS. The murine melanoma B16F1-gB.GFP (B16-gB) cell line was provided by Jason Waithman (University of Western Australia) 47 . B16-gB cells were cultured and passaged in complete RPMI (RPMI1640, 10% FCS, 2mM L-glutamine, 100U ml -1 penicillin, 100mg ml -1 streptomycin, 50mM 2-mercaptoethanol) at 37°C and 5% CO 2 . B16-gB was inoculated epicutaneously as previously described 47 . Tumours were measured every 2-3 days following the development of palpable tumours using vernier callipers, and tumour volume was calculated (length × width 2 )/2. Mice were euthanised when tumours reached an ethical limit of 1.0 cm 3 . Effector gBT-I or OT-I T cells were transferred into mice following the observation of tumour growth, between 14-20 d following tumour inoculation, and then harvested at tumour endpoint (when tumours reached a maximum of 1000mm 3 ) which occurred 14-20 d following T cell transfer. LCMV infection was performed by intraperitoneal (i.p.) injection of 2×10 5 pfu of the Armstrong strain of LCMV. T cell transfer For the adoptive transfer of naïve T cells, 1-5×10 4 transgenic (P14 or OT-I) T cells were transferred intravenously to recipient mice 1-2 d before infection with LCMV or inoculation with AT3-OVA. For effector T cell transfer, transgenic (P14, OT-I, OT-3, or gBT-I) T cells were activated in culture for 5 d with gp 33-41 (KAVYNFATM – P14; Auspep), OVA 257-264 (SIINFEKL – OT-I/OT-3; Auspep), or gB 498-505 (SSIEFARL – gBT-I; Auspep) peptide-pulsed splenocytes, in the presence of recombinant human IL-2 (25 U ml -1 ; Peprotech) in complete RPMI (as above) at 37°C and 5% CO 2 . T cells were split 1:1 with fresh media and IL-2 on days 2-4. T cells were resuspended in 200µl HBSS for intravenous transfers. Unless stated otherwise, 1×10 4 effector OT-I, 5×10 6 gBT-I, or 1×10 6 OT-3 were transferred 10 d post-AT3-OVA inoculation. In vivo treatments For in vivo intravascular staining, 3μg of αCD90.2-biotin was injected i.v. in 200μL PBS, 3min before euthanasia. To inhibit S1P-signaling pathways, mice were administered FTY720 (Cayman Chemical) diluted in 2% (2-Hydroxypropyl)-β-cyclodextrin (Sigma-Aldrich) or vehicle daily via i.p. injection (1μg/g) for the indicated times in the figure legend ( Extended Data Fig. 7c ). For tamoxifen treatment, mice were administered 2 mg of tamoxifen (or ethanol as vehicle control) diluted in sunflower seed oil (both from Sigma) i.p. daily for a total of five injections. Isolation of lymphocytes from mouse tissues Lymphocytes from spleens and lymph nodes were isolated by grinding through 70-μm strainers. AT3-OVA tumours and MFPs (avoiding the inguinal lymph node) were collected into collagenase III solution (3mg ml -1 ; Worthington) containing DNase I (2.5 mg ml -1 ; Sigma), chopped into fine pieces and incubated for 60 min at 37°C. B16-gB tumours were collected into liberase TL research grade solution (0.25 mg ml -1 ; Roche), chopped into fine pieces, and incubated for 60 min at 37°C. Spleens and tumours were passed through a 70-mm strainer and erythrocytes were lysed in red cell lysis buffer (eBioscience) prior to staining for flow cytometry. Cells isolated from B16-gB tumours were cryopreserved in 10% DMSO:90% FCS for scRNAseq. Flow cytometry and cell sorting Single-cell suspensions were stained with fluorescently conjugated antibodies at 4°C for 30-45 min in FACS buffer (1% BSA and 0.05M EDTA in PBS). Dead cells were excluded by staining with Zombie Aqua or Near Infrared dyes (Biolegend). For transcription factor and cytokine staining, samples were fixed and permeabilised using the Foxp3/transcription-factor-staining buffer kit (eBioscience) according to the manufacturer’s instructions and stained with fluorescent antibodies against intracellular proteins in permeabilisation buffer (containing 2% rat and mouse serum (eBioscience)). Cells were enumerated using SPHERO calibration particles (BD Biosciences). Fluorescently labelled cells were acquired on a Cytek Aurora, unmixed with SpectroFlo® software and analysis was performed using FlowJo (v.10.10.0; Treestar), or OMIQ for high-dimensional flow cytometry analysis. For cell sorting experiments, cells were sorted using a BD FACS Aria, using a 100 μm nozzle. For human cell sorting experiments for CITEseq, CD3 + Zombie NIR - , cells were sorted into 50% FCS in RPMI before downstream processing. For B16-gB sorting experiments, DAPI - OT-I or gBT-I cells were sorted into 50% FCS in RPMI before downstream processing. For mouse AT3-OVA sort-transfer experiments, mice were treated with anti-ARTC2 nanobody intravenously (50 mg per mouse; S+16a, Biolegend) 10 minutes before organ collection, stained as above, and respective populations (see figure legend) sorted into 50% FCS in RPMI before washing with HBSS and transfer into recipient mice (8×10 4 cells in 200 µl intravenously). In vitro stimulation assays To assess cytokine production capacity by T cells, cells were stained with surface stain antibodies (as above), then incubated with phorbol myristate acetate (PMA; 50 ng ml -1 ; Sigma-Aldrich), Ionomycin (1 mg ml -1 ), Brefeldin A (10 mg ml -1 , Sigma-Aldrich), GolgiStop (1:1500, BD) in complete RPMI (as above) for 4-5 h before intracellular staining and flow cytometry. Unstimulated controls were included to confirm that stimulation did not alter cell surface staining and to serve as negative controls for staining. Bulk RNA-seq analysis (For Fig. 1a ) Scatter plot of logFCs (log2-fold-changes) was produced using raw RNA-seq count data from GEO for Man et al. (accession GSE84820 26 ) and Mackay et al. (accession GSE70813 8 ). From the Man data, wild-type day 30 chronic and acute LCMV samples were selected for analysis, while from the Mackay data, the following samples were selected for three separate analyses: (1) Skin T RM , T CM , and T EM HSV samples; (2) Gut T RM , T CM , and T EM LCMV samples; and (3) Liver T RM , T CM , and T EM LCMV samples. T CM and T EM samples were subsequently treated as a single ’T CIRC ’ group. For each analysis, genes were annotated with information from NCBI and those with obsolete symbols or annotated as rRNA were removed, as were genes that failed to achieve a count above 10 in all samples in at least one group. Each dataset was further processed by applying the imputation strategy published previously 59 , 60 . Counts-per-million values were calculated using the edgeR package 61 , together with scaling factors derived from the TMM method 62 , log2 transformed with a prior count of 1, followed by application of the normalisation method RUV-III 63 with biological replicates nominated as replicates, mouse housekeeping genes 56 nominated as ‘negative control’ genes, and k=1 factors of unwanted variation. Normalisation success was assessed with relative log expression plots 64 , PCA plots 65 , and histograms of p-values. The limma package 66 was used to fit gene-wise linear models for the given group structure with the output from RUV-III as an additional model covariate. LogFC estimates from the Mackay Skin, Gut, and Liver analyses were averaged to create a ’pooled T RM ’ expression profile and then plotted against the estimates from the Man analysis. A p-value was calculated by constructing a two-way contingency table, counting the number of genes with concordant/discordant logFCs, and applying Fisher’s exact test for association. Core T RM gene signature was obtained as described previously 67 . Single-cell CITE/RNA sequencing library preparation Following sorting of cell populations, sorted cells were stained with TotalSeq-C Universal Cocktail V1.0 (Biolegend; Human CITEseq experiments) or TotalSeq-C Hashtag antibodies (Biolegend: mouse B16-gB scRNA-seq experiment) according to manufacturer’s instructions. Cells were filtered through a 40-mm Flowmi cell strainer, loaded onto a 10X chromium controller, and prepared for sequencing using a Chromium Next GEM Single Cell 5’ kit with feature barcoding and immune receptor mapping (v2, Dual Index) and VDJ enrichment kits for mouse or human TCR from 10X. Libraries were generated according to the manufacturer’s instruction. Libraries were profiled on an Agilent Tapestation and quantified using a Qubit before sequencing on an Illumina NextSeq 2000 P2 or P3 kit. Single-cell CITE/RNA sequencing analysis Sequencing reads of three separate lanes were aligned to the hg38 reference genome and T cell receptor reference (VDJ) and counted with cell ranger-6.1.2. CITEseq antibodies were aligned to their custom reference sequences from BioLegend. Patient samples were demultiplexed into genotypic donors using vireo 68 on aligned BAM files from three lanes. Donors were matched to genotypes using known cell frequencies in each sample. Three batches of single-cell data were merged and processed using Seurat 69 (version 4.3.0). Specifically, cells were filtered if they contained fewer than 500 genes, more than 5% mitochondrial RNA, and were annotated to have more than 1 beta chain in the VDJ assay or two genotypes in the vireo analysis (cell doublets). Then, for the RNA assay, we used NormalizeData to normalise the counts data and determine the top 2000 variable genes using FindVariableFeatures. We excluded T cell receptor components (^TRA/^TRB), mitochondrial genes (^MT), and ^HLA from the variable genes as unwanted factors of variance and performed principal component analysis (RunPCA). The CITEseq assay was processed similarly with CLR normalisation and margin=2. We removed the individual effects between the donors using Harmony 70 on the RNA and CITEseq assays individually and then combining the correction reductions using the FindMultiModalNeighbors functions (using 25 dimensions from the RNA reduction and 18 from CITEseq). UMAP reductions and cell neighbours were calculated using 25 dimensions from the weighted nearest neighbour reduction. Clusters were detected with a resolution of 3. We then determined T cell subsets within the data by assessing cluster markers (FindAllMarkers) and their overall expression and comparing them to published markers in the literature (merging clusters as need be). We subset the data to CD8+ T cells only at this point by excluding other T cell populations and contaminating immune populations. All further analyses and plots were generated in R (version 4.2) using tidyverse 71 functions and ShinyCell 72 . Heatmaps were created with the pheatmap package (version 1.0.12). Subset signature expression levels were calculated using the AddModuleScore function. Pseudobulk analysis was performed using the edgeR package 61 : counts of reads were summed per subset. Pseudobulk samples from the periphery of the data (gdT, MAIT) are filtered from the analysis and lowly expressed genes removed using filterByExpr(). The samples were then normalised using TMM and the calcNormFactors() function. Dimensional reduction was performed using the plotMDS function. Other single-cell datasets, such as Giles 2022 27 , or subsets of the data (tumour only) were analysed analogously to the methods above (leaving out CITEseq and TCR where appropriate). TCR analysis TCRs from cellranger outputs were paired based on cell barcodes and merged with gene expression data. In cases where multiple contigs were detected the contigs with the highest UMI was kept. TCR clonotype was defined by the joined alpha and beta CDR3 nucleotide sequences, and expanded TCR clonotypes were determined by filtering the list of clonotypes to those that are found in 2 or more cells. Antigen specificity of cells was based on TCR clonotype and donor HLA identity. HLA type was determined with ArcasHLA using the ‘—single’ flag 73 . Reads covering the genome co-ordinates chr6:28510120-33480577 were extracted into a separate fastq file per donor and processed individually. Plots of clonotype diversity and similarity within/across celltypes were calculated using djvdj ( https://rnabioco.github.io/djvdj/ version 0.1.0) and plotted with ComplexHeatmap 74 . For clonotype sharing plots, expanded clonotypes within each tissue were filtered to only include those which are shared across more than one subset. The number of shared clonotypes was quantified for each subset pairing, and the total number of cells contributing to clonotype sharing from each subset was counted. Visualisations were generated with custom R scripts using the ggraph package (version 2.0.5). For iterations of the sharing plots with filtering, input data were refined so that a shared TCR was removed from the plot if one of the celltypes in the pairing contributed just 1 cell (for n>1 filtering) or 2 or fewer cells (for n>2 filtering). Other visualizations of clonotype cell fate and cluster sharing were generated using custom R scripts, as provided. Comparisons between TCR sequences was performed using TCRdist3 33 , 34 using default weights with the distance matrix calculated as the sum pairwise distance for the alpha and beta chains. To determine potential virus-recognising clones, published TCRs with known viral HLA-peptide specificities were obtained from VDJdb 35 – 37 following which edit distances were calculated for each clone against reference CDR3a/b peptide sequences. The lowest edit distance was obtained for each clone, with distances less than one considered viral-associated. Pan-cancer atlas processing 28 Preprocessed Seurat objects were obtained 28 . SCTransform was used to normalise, scale and regress pre-filtered dataset. Pre-computed scores for dissociation induced genes(DIG), Malat1, mitochondrial percentage and cell-cycle(G2M and S scores) were used for regression. T-cell receptor variable/joining (^TR[A|B|G|D][V|J]), B-cell receptor variable/joining (^IG[H|L|K][V|J|D]) DIG(^HSP/^DNAR), cell-cycle, DIG(^HSP/^DNAJ) and ribosomal (^RP([0-9]+-|[LS])) genes were removed from VariableFeatures gene list priors to calculating principal components(PC). Subsequent UMAP and FindNeighbours commands performed with 15 PCs. T RM and T EX scores were calculated with the AddModuleScore command using respective genesets. T EM and T EMRA scores were calculated using differentially expressed genes calculated from the BC dataset. Signature acquisition and Module Scoring T RM and T EX expression signatures were derived by summing raw counts, gene-wise, over each donor within each cluster, to create ‘pseudo-bulk’ samples for each donor/cluster combination. Samples composed of less than 20 cells were removed, and samples from the same meta-cluster were subsequently treated as belonging to the same group. Genes annotated as being of ’HLA’, ’TRB’, ’TRA’, ’TRG’, or ’TRD’ type were removed, as were genes that failed to achieve a count above 5 in at least 2 samples in at least one group (breast) or a count above 3 in at least 4 samples in at least one group (liver). Counts were log2 transformed with a prior count of 1/2 (breast) or 1 (liver), followed by application of the normalisation method RUV-III with samples from the same group nominated as replicates, human housekeeping genes nominated as ‘negative control’ genes, and k=20 (breast) or k=15 (liver) factors of unwanted variation. Normalisation success was assessed as above. The edgeR package [*] was used fit gene-wise negative binomial generalized linear models for the given group structure with the output from RUV-III as additional model covariates, with a prior count of 1/2 (breast) or 1 (liver). Likelihood ratio tests were employed to test for differential expression (DE), where a gene was judged to be DE if the Benjamini and Hochberg 75 adjusted p-value < 0.05. For breast, the T RM signature was defined by a contrast between the T RM group and the average of all other groups; the T EX signature was defined similarly. For liver, the T RM signature was defined by a contrast between the CD103 + T RM group and the average of all other groups except the CD103 - T RM group; the T EX signature was defined similarly. The breast T RM ‘union’ signature was derived using the same steps for breast above, except that samples obtained from the T RM or T EX clusters were combined into a single group. Signature scores were calculated from the SCTransformed assay using the ’AddModuleScore’ function from the Seurat package 69 for up and down genes separately and combined by averaging the scores for the up genes and the sign reversed scores for the down genes. These scores were overlayed onto UMAP plots using the FeaturePlot function from Seurat. To generate the pan-cancer T RM signature we focused on the leading-edge genes contributing to the enrichment in the barcode plots ( Fig. 3c , Extended Data Fig. 4g ). Specifically, the leading-edge genes contributing to T RM signature associations between the BC and liver datasets were refined by intersecting the genes in the breast and liver T RM signatures (described above), then subsetting to genes which are either (1) concordantly DE in both breast and liver or (2) DE in one with a concordant absolute logFC > 0.5 in the other. The pan-cancer T EX signature was refined similarly ( Extended Data Fig. 4g ). Barcode enrichment plots were generated using limma , and gene set enrichment p-values were calculated using the camera function on the log fold-changes for populations of interest against the background calculated with FindMarkers. Survival Analysis (For Fig. 2l , Extended Data Fig. 3f-h ) Clinical information and normalised microarray data for the study 50 , 76 were downloaded from the cBioPortal ( https://www.cbioportal.org ). Signature scores for each patient were calculated using the ‘sig.score’ function from the genefu package 77 . Kaplan-Meier curves were calculated using the R package survival and were plotted, with the log-rank test p-value, using the R package survminer . (For Fig. 2m ) Clinical information and microarray data for the iSPY study 31 were downloaded from supplementary material and GEO (accession GSE194040), respectively. Probes targeting the same gene were averaged. If a gene had missing values, these were imputed using the average of all non-missing values across the gene. For each platform batch, relative log expression 64 values were computed, the median value for each sample was calculated, then samples were ranked by this median value: the bottom, middle, and top third rankings defined 3 separate ‘pseudo-batches’ of samples. For each PAM50 breast cancer expression subtype in each pseudo-batch, one of the following was performed: if there were between 5-10 samples, 5 samples were randomly selected and averaged, gene-wise, to create one ‘pseudo-sample’; or if there were >10 samples, the bottom 5 ranked samples and the top 5 ranked samples were averaged, gene-wise, the create 2 pseudo-samples. These pseudo-samples were used as pseudo-replicates in the RUV-III-PRPS normalisation methodology 78 , with all genes nominated as ‘negative control’ genes, and k=7 factors of unwanted variation. Normalisation success was assessed with relative log expression and PCA plots 65 . Based on this normalised data, signature scores for each patient were calculated using the ‘sig.score’ function from the genefu package. ROC curves and p-values were calculated using the R packages pROC 79 and verification , respectively. Cyclic immunofluorescence (CycIF) of human breast tumours Selection of breast cancer samples FFPE-embedded tumour tissues from 7 patients (6 female, 1 male) were purchased from the Cooperative Human Tissue Network (CHTN) based on histologic criteria, for invasive carcinoma (ductal/lobular) and 5 µm thick sections were cut on Superfrost Plus histology slides (Fisher scientific) at the BWH Histopathology core, as previously described 29 . CycIF staining FFPE tissues were deparaffinized, rehydrated, subjected to antigen retrieval on a Leica BondRx and characterised by cyclic immunofluorescence (CycIF) imaging, as previously described 29 . Briefly, tissues were stained overnight at 4°C with primary antibodies or for 1 hour at RT for secondary antibody conjugates in a dark, humidified chamber with antibodies diluted in Superblock (Thermo-Fisher) supplemented with 1 mg/ml Hoechst 33258 (Bio-Rad Laboratory) rinsed for 30 minutes in PBS (RT) and mounted with 50% glycerol in PBS using a 24×60mm coverslip. All samples were stained and imaged in together to reduce batch effects using the antibody panel. Stained tissues were imaged using a Cytefinder II HT (Rarecyte) automated slide scanning fluorescence microscope using a 20x (0.6 NA) objective. After imaging, mounted slides were soaked in PBS (RT) to detach coverslips, then immersed in PBS supplemented with 4.5 % hydrogen peroxide and 24 mM sodium hydroxide and exposed to LED light for 1 hour. Slides were then rinsed twice in PBS in preparation for the next staining cycle. Image processing, assembly, segmentation and single-cell quantification were performed using MCMICRO 30 . CycIF image analysis CycIF Images were processed with Cecelia 80 . Individual channels were denoised and segmented with Cellpose 81 using a radius of 10 and 8 μm, respectively. T cell (CD3e) and cancer cell (panCK) segmentation was based on their respective markers in combination with the DNA (Hoechst-stained nuclei) channel from the third imaging cycle, the same cycle as CD3e. Cell populations were gated using Cecelia’s histocytometry module and spatial interactions analysed. Cell distances were extracted using K-nearest neighbour from the DBSCAN library 82 . Clusters within the CD8 + CD103 + KLF2 - T cells were defined using the Seurat package. We first removed the individual donor effects and performed principal component analysis using a set of 22 features (CD39, CD45RO, CD3E, CD25, CD73, CD8A, GZMB, CD103, LAG3, PD1, TIM3, CD69, TCF1, FOXP3, GNLY, CD4, pJUN, pERK, KLF2, CD7, CD94, NKG2A) (RunPCA). Shared neighbour graphs (FindNeighbors) and UMAP reduction (RunUMAP) were performed using the first 10 dimensions. Clusters were detected using a resolution of 0.8 (FindClusters). CRISPR/Cas9 editing of CD8 + T cells CRISPR/Cas9 editing of naïve CD8 + T cells was performed as previously described 83 . Single guide RNAs (sgRNA) targeting: Tcf7 (5′-UCUGCUCAUGCCCUACCCAC-3′, 5′-AGCUGGGGGACGCCAUGUGG-3′, 5′-UGUGCACUCAGCAAUGACCU-3′), Cxcr3 (5’-GAACAUCGGCUACAGCCAGG-3’, 5’-UGAGGGCUACACGUACCCGG-3’), Cxcr6 (5’-UCUGUACGAUGGGCACUACG-3’, 5’-UGUGCCAAAGACCCACUCAU-3’), and Cd19 (5′-AAUGUCUCAGACCAUAUGGG-3′, 5’-GAGAAGCUGGCUUGGUAUCG-3’) were purchased from Synthego (CRISPRevolution sgRNA EZ Kit). sgRNA/Cas9 RNPs were formed by incubating 0.3 nmol of each sgRNA with 0.6 μl Alt-R S.p. Cas9 nuclease V3 (10 mg ml −1 ; Integrated DNA Technologies) for 10 min at room temperature. Naïve CD8 + T cells were negatively enriched from spleen and lymph nodes of gBT-I or OT-I mice by incubating cell suspensions with anti-CD4, anti-CD11b anti-F4/80, anti-Ter119 and anti-I-A/I-E monoclonal antibodies, followed by incubation with goat anti-rat IgG-coupled magnetic beads (Qiagen) before removing bead-bound cells. 1×10 7 enriched T cells were resuspended in 20 μl of P3 (P3 Primary Cell 4D-Nucleofector X Kit; Lonza), mixed with sgRNA/Cas9 RNP and electroporated using a Lonza 4D-Nucleofector system (DN100). Cells were rested for 30 min in a 96-well plate before direct transfer into recipient mice (co-transfer of 1:1 ratio of sg Cd19 :sg Tcf7 CRISPR-edited naïve gBT-I cells, total of 2 × 10 4 cells per mouse: Extended Data Fig. 8a-d ) or activation in culture for 5 days with peptide pulsed splenocytes (gB 498-505 (SSIEFARL)) in the presence of IL-2 (25 U ml −1 , Peprotech) at 37 °C, 5% CO 2 , before transfer into recipient mice (co-transfer of 1:1 ratio of sg Cd19 :sg Cxcr3 or sg Cd19 :sg Cxcr6 CRISPR-edited naïve gBT-I cells, total of 1×10 6 cells per mouse: Extended Data Fig. 8m-n ). SPLINTR Methods Production of libraries To construct the SPLINTR barcoding systems, Violet-light excited GFP (VEX) was used to replace NGFR in MSCV-IRES-NGFR 84 using the NcoI and BamH1 restriction sites, or eGFP into the SPLINTR lentivirus vector 41 . A semi-random oligonucleotide library synthesised by Integrated DNA Technologies with the following pattern (NNSWSNNWSW) 6 was amplified by eight cycles of PCR. The barcode library was cloned into the 3’UTR of VEX using BamH1 and MfeI restriction sites at a 50:1 insert:vector ratio. 10 side-by-side ligation reactions were pooled and purified using two Monarch PCR & DNA Clean up columns (NEB) in a volume of 6 µl per column. The ligation reactions were pooled and split across two 25 µl aliquots of Endura Electrocompetent cells (Lucigen). Cells were recovered for 1 hr post-electroporation, pooled and grown in 500 mL of LB supplemented with ampicillin (100 µg ml -1 ) overnight at 37°C. The plasmid library was extracted using NucleoBond Xtra kit (Macherey-Nagel). SPLINTR libraries were sequenced to a depth of 100 million paired-end reads per technical duplicate and reference libraries used for downstream analysis were generated as described previously 41 . SPLINTR retrovirus and lentivirus VEX library represents a highly diverse barcode library containing 3.2×10 6 unique barcodes or 1.3×10 6 barcodes, respectively. Generation of SPLINTR + OT-I T cells Naïve (T N ) OT-I T cells were barcoded via transduction of hematopoietic stem and progenitor cells (HSPC) and intra-thymic transfer into sublethally irradiated chimeric mice. Briefly, bone marrow was isolated from OT-I mice, and lineage(lin)-positive cells were depleted using the EasySep™ Mouse Hematopoietic Progenitor Cell Isolation Kit (Stem Cell Technologies). Lin-negative HSPCs were plated in fibronectin-coated 12-well plates containing polyvinyl alcohol (PVA)-based media: 1× Ham’s F-12 Nutrient Mix liquid media (Gibco) supplemented with 10 mM HEPES, 1× P/S/G, 1 ITSX, 1 mg/mL PVA along with TPO (100 ng/µL) and mouse SCF (10 ng/µL). For barcoding, 20 pools of 1×10 6 cells (cultured for 6 d) were transferred to 24-well fibronectin-coated plates in 250 μL of PVA media. Viral barcoding vectors were titrated to achieve 10% reporter expression (0.1 MOI) to minimize multiple integrations. Cells were transduced via spinfection at 2000g for 2 h (no break). After transduction, media was refreshed and cells were cultured for a further 48 h before VEX-positive and Lin-cells were sorted (ARIA Fusion, BD). The 20 pools of 3.5-5.5×10 5 cells were individually plated into 48-well plates and expanded for 1-week. Recipient C57Bl/6 mice were then irradiated 4Gy and HSPCs were transplanted intrathymically such that each mouse received an independent barcode pool. 8 weeks later, chimeric mice were bled, equal numbers of VEX + OT-I T cells from each mouse were pooled, and cells were transferred into C57Bl/6 mice that were then inoculated with AT3-OVA tumours. Effector (T EFF ) OT-I T cells were barcoded as follows. Naïve OT-I T cells were isolated from spleens and LNs and enriched via negative selection as for CRISPR/Cas9 experiments. Enriched OT-I T cells were activated with anti-CD3 (2μg ml -1 , 145-2C11, BioXCell) and anti-CD28 (1μg ml -1 , 37.51, BioXCell) for 24 hours before ‘spinfection’ with a pre-titrated volume of SPLINTR retrovirus to ensure <5% (0.05 MOI) transduction efficiency to limit multiple barcode integrations in a single cell, in 24-well plates, pre-coated with Retronectin (32 ug ml - 1 , Takara). Transduced cells were sorted 24 hours later and transferred into mice bearing AT3-OVA tumours. At the endpoint, transduced OT-I T cell populations were sorted from spleens and tumours and lysed in Viagen lysis buffer with 0.5 μg/ml proteinase K (Invitrogen) for DNA barcode sequencing. Sequencing and analysis Barcode sequences were amplified from genomic DNA using primers flanking the constant region of the barcode before adding i5 and i7 indexes compatible with NGS sequencing 41 . Libraries were sequenced on an Illumina NextSeq2000 using 100 bp single end chemistry targeting 2 million reads per sample. The BARtab pipeline ( https://github.com/DaneVass/BARtab ) 85 was used to map the sequencing reads to a barcode reference library, perform qc analysis, and generate a barcode counts table. The bartools R package 85 was used to collapse PCR replicates and generate barcode abundance bubble plots and correlation heatmaps for data visualisation. Scatterplots comparing barcode abundance between two samples were generated using the R package ggplot2 . Data and material availability statement All data are available from the corresponding author upon reasonable request. The single-cell CITE and RNA-sequencing data generated from this study will be deposited in the GEO before publication. Source data are provided with this paper. Code availability The code generated and used for the analysis of single-cell CITE and RNAseq data is provided with this paper: https://github.com/CSI-Doherty/Burn2025 . Ethics approval All animal experiments were approved by The University of Melbourne Animal Ethics Committee (ID nos. 21651 and 21938). All study on human specimens was approved by the Human Research Ethics Committee of the University of Melbourne (ID nos. 13009 and 14517). All participating patients provided written informed consent. Ethics declarations The authors declare no competing interests. Acknowledgements We thank the Flow Cytometry Unit and Bioresources Facility at the Doherty Institute (University of Melbourne) for technical assistance. We thank all Pfizer Inc. Cancer Immunology team members, the laboratory of L.K.M, Professor William Heath, and Dr. Yannick Alexandre (University of Melbourne) for comments and discussion. This work was supported by Pfizer Inc., Cancer Council Victoria Grants-in-Aid, and National Health and Medical Research Council (NHMRC) to L.K.M. L.K.M. is a Senior Medical Research Fellow supported by the Sylvia and Charles Viertel Charitable Foundation. Z.M. is supported by NCI grant R50-CA252138. Histopathology was supported by NIH core grant P30-CA06516. References ↵ Savas , P. et al. Single-cell profiling of breast cancer T cells reveals a tissue-resident memory subset associated with improved prognosis . Nature Medicine 24 , 986 – 993 ( 2018 ). OpenUrl CrossRef PubMed Edwards , J. et al. CD103+ Tumor-Resident CD8+ T Cells Are Associated with Improved Survival in Immunotherapy-Naïve Melanoma Patients and Expand Significantly During Anti–PD-1 Treatment . Clinical Cancer Research 24 , 3036 – 3045 ( 2018 ). OpenUrl Abstract / FREE Full Text Luoma , A. M. et al. Tissue-resident memory and circulating T cells are early responders to pre-surgical cancer immunotherapy . Cell 185 , 2918 – 2935.e2929 ( 2022 ). OpenUrl CrossRef PubMed Anadon , C. M. et al. Ovarian cancer immunogenicity is governed by a narrow subset of progenitor tissue-resident memory T cells . Cancer Cell 40 , 545 – 557.e513 ( 2022 ). OpenUrl CrossRef PubMed Corgnac , S. et al. CD103+CD8+ TRM Cells Accumulate in Tumors of Anti-PD-1-Responder Lung Cancer Patients and Are Tumor-Reactive Lymphocytes Enriched with Tc17 . Cell Reports Medicine 1 , 100127 ( 2020 ). OpenUrl PubMed ↵ Clarke , J. et al. Single-cell transcriptomic analysis of tissue-resident memory T cells in human lung cancer . Journal of Experimental Medicine 216 , 2128 – 2149 ( 2019 ). OpenUrl Abstract / FREE Full Text ↵ Poon , M. M. L. et al. Tissue adaptation and clonal segregation of human memory T cells in barrier sites . Nat Immunol 24 , 309 – 319 ( 2023 ). OpenUrl PubMed ↵ Mackay , L. K. et al. Hobit and Blimp1 instruct a universal transcriptional program of tissue residency in lymphocytes . Science 352 , 459 – 463 ( 2016 ). OpenUrl Abstract / FREE Full Text ↵ Wijeyesinghe , S. et al. Expansible residence decentralizes immune homeostasis . Nature 592 , 457 – 462 ( 2021 ). OpenUrl CrossRef PubMed ↵ Gebhardt , T. et al. Memory T cells in nonlymphoid tissue that provide enhanced local immunity during infection with herpes simplex virus . Nature Immunology 10 , 524 – 530 ( 2009 ). OpenUrl CrossRef PubMed Web of Science ↵ Jiang , X. et al. Skin infection generates non-migratory memory CD8+ TRM cells providing global skin immunity . Nature 483 , 227 – 231 ( 2012 ). OpenUrl CrossRef PubMed Web of Science ↵ Philip , M. & Schietinger , A . CD8+ T cell differentiation and dysfunction in cancer . Nature Reviews Immunology 22 , 209 – 223 ( 2022 ). OpenUrl CrossRef PubMed Gebhardt , T. , Park , S. L. & Parish , I. A . Stem-like exhausted and memory CD8+ T cells in cancer . Nature Reviews Cancer 23 , 780 – 798 ( 2023 ). OpenUrl CrossRef PubMed ↵ Giles , J. R. , Globig , A. M. , Kaech , S. M. & Wherry , E. J . CD8(+) T cells in the cancer-immunity cycle . Immunity 56 , 2231 – 2253 ( 2023 ). OpenUrl CrossRef PubMed ↵ Pauken , K. E. et al. Epigenetic stability of exhausted T cells limits durability of reinvigoration by PD-1 blockade . Science 354 , 1160 – 1165 ( 2016 ). OpenUrl Abstract / FREE Full Text ↵ Gavil , N. V. , et al. Chronic antigen in solid tumors drives a distinct program of T cell residence . Sci Immunol 8 , eadd5976 ( 2023 ). OpenUrl CrossRef PubMed ↵ Masopust , D. et al. Dynamic T cell migration program provides resident memory within intestinal epithelium . Journal of Experimental Medicine 207 , 553 – 564 ( 2010 ). OpenUrl Abstract / FREE Full Text ↵ Bartolomé-Casado , R. et al. Resident memory CD8 T cells persist for years in human small intestine . Journal of Experimental Medicine 216 , 2412 – 2426 ( 2019 ). OpenUrl Abstract / FREE Full Text Pallett , L. J. et al. Longevity and replenishment of human liver-resident memory T cells and mononuclear phagocytes . Journal of Experimental Medicine 217 ( 2020 ). ↵ Snyder , M. E. et al. Generation and persistence of human tissue-resident memory T cells in lung transplantation . Science Immunology 4 , eaav5581 ( 2019 ). OpenUrl ↵ Kumar , B. V. et al. Human Tissue-Resident Memory T Cells Are Defined by Core Transcriptional and Functional Signatures in Lymphoid and Mucosal Sites . Cell Rep 20 , 2921 – 2934 ( 2017 ). OpenUrl CrossRef PubMed ↵ Cheuk , S. et al. CD49a Expression Defines Tissue-Resident CD8(+) T Cells Poised for Cytotoxic Function in Human Skin . Immunity 46 , 287 – 300 ( 2017 ). OpenUrl CrossRef PubMed ↵ Mackay , L. K. et al. The developmental pathway for CD103(+)CD8+ tissue-resident memory T cells of skin . Nat Immunol 14 , 1294 – 1301 ( 2013 ). OpenUrl CrossRef PubMed ↵ Skon , C. N. et al. Transcriptional downregulation of S1pr1 is required for the establishment of resident memory CD8+ T cells . Nature Immunology 14 , 1285 – 1293 ( 2013 ). OpenUrl CrossRef PubMed ↵ Evrard , M. et al. Sphingosine 1-phosphate receptor 5 (S1PR5) regulates the peripheral retention of tissue-resident lymphocytes . Journal of Experimental Medicine 219 ( 2022 ). ↵ Man , K. et al. Transcription Factor IRF4 Promotes CD8(+) T Cell Exhaustion and Limits the Development of Memory-like T Cells during Chronic Infection . Immunity 47 , 1129 – 1141 e1125 ( 2017 ). OpenUrl CrossRef PubMed ↵ Giles , J. R. et al. Shared and distinct biological circuits in effector, memory and exhausted CD8(+) T cells revealed by temporal single-cell transcriptomics and epigenetics . Nat Immunol 23 , 1600 – 1613 ( 2022 ). OpenUrl CrossRef PubMed ↵ Zheng , L. et al. Pan-cancer single-cell landscape of tumor-infiltrating T cells . Science 374 , abe6474 ( 2021 ). OpenUrl CrossRef PubMed ↵ Lin , J. R. , Fallahi-Sichani , M. , Chen , J. Y. & Sorger , P. K. Cyclic Immunofluorescence (CycIF), A Highly Multiplexed Method for Single-cell Imaging . Current Protocols in Chemical Biology 8 , 251 – 264 ( 2016 ). OpenUrl PubMed ↵ Schapiro , D. et al. MCMICRO: a scalable, modular image-processing pipeline for multiplexed tissue imaging . Nature Methods 19 , 311 – 315 ( 2022 ). OpenUrl PubMed ↵ Wolf , D. M. et al. Redefining breast cancer subtypes to guide treatment prioritization and maximize response: Predictive biomarkers across 10 cancer therapies . Cancer Cell 40 , 609 – 623.e606 ( 2022 ). OpenUrl CrossRef PubMed ↵ Chalmers , Z. R. et al. Analysis of 100,000 human cancer genomes reveals the landscape of tumor mutational burden . Genome Med 9 , 34 ( 2017 ). OpenUrl CrossRef PubMed ↵ Dash , P. et al. Quantifiable predictive features define epitope-specific T cell receptor repertoires . Nature 547 , 89 – 93 ( 2017 ). OpenUrl CrossRef PubMed ↵ Mayer-Blackwell , K. et al. TCR meta-clonotypes for biomarker discovery with tcrdist3 enabled identification of public, HLA-restricted clusters of SARS-CoV-2 TCRs . Elife 10 ( 2021 ). ↵ Bagaev , D. V. et al. VDJdb in 2019: database extension, new analysis infrastructure and a T-cell receptor motif compendium . Nucleic Acids Res 48 , D1057 – D1062 ( 2020 ). OpenUrl CrossRef PubMed Goncharov , M. et al. VDJdb in the pandemic era: a compendium of T cell receptors specific for SARS-CoV-2 . Nat Methods 19 , 1017 – 1019 ( 2022 ). OpenUrl CrossRef PubMed ↵ Shugay , M. et al. VDJdb: a curated database of T-cell receptor sequences with known antigen specificity . Nucleic Acids Res 46 , D419 – D427 ( 2018 ). OpenUrl CrossRef PubMed ↵ Duhen , T. et al. Co-expression of CD39 and CD103 identifies tumor-reactive CD8 T cells in human solid tumors . Nat Commun 9 , 2724 ( 2018 ). OpenUrl CrossRef PubMed ↵ Simoni , Y. et al. Bystander CD8(+) T cells are abundant and phenotypically distinct in human tumour infiltrates . Nature 557 , 575 – 579 ( 2018 ). OpenUrl CrossRef PubMed ↵ Wu , T. D. et al. Peripheral T cell expansion predicts tumour infiltration and clinical response . Nature 579 , 274 – 278 ( 2020 ). OpenUrl CrossRef PubMed ↵ Fennell , K. A. et al. Non-genetic determinants of malignant clonal fitness at single-cell resolution . Nature 601 , 125 – 131 ( 2022 ). OpenUrl CrossRef PubMed ↵ Lee , Y. J. , et al. CD39(+) tissue-resident memory CD8(+) T cells with a clonal overlap across compartments mediate antitumor immunity in breast cancer . Sci Immunol 7 , eabn8390 ( 2022 ). OpenUrl PubMed ↵ Chowell , D. et al. Improved prediction of immune checkpoint blockade efficacy across multiple cancer types . Nat Biotechnol 40 , 499 – 506 ( 2022 ). OpenUrl CrossRef PubMed Subbiah , V. , Solit , D. B. , Chan , T. A. & Kurzrock , R . The FDA approval of pembrolizumab for adult and pediatric patients with tumor mutational burden (TMB) >/=10: a decision centered on empowering patients and their physicians . Ann Oncol 31 , 1115 – 1118 ( 2020 ). OpenUrl CrossRef PubMed Barroso-Sousa , R. et al. Prevalence and mutational determinants of high tumor mutation burden in breast cancer . Ann Oncol 31 , 387 – 394 ( 2020 ). OpenUrl CrossRef PubMed ↵ Loi , S. , et al. Association Between Biomarkers and Clinical Outcomes of Pembrolizumab Monotherapy in Patients With Metastatic Triple-Negative Breast Cancer: KEYNOTE-086 Exploratory Analysis . JCO Precision Oncology 7 ( 2023 ). ↵ Park , S. L. et al. Tissue-resident memory CD8+ T cells promote melanoma–immune equilibrium in skin . Nature 565 , 366 – 371 ( 2019 ). OpenUrl CrossRef PubMed ↵ Rosato , P. C. et al. Virus-specific memory T cells populate tumors and can be repurposed for tumor immunotherapy . Nat Commun 10 , 567 ( 2019 ). OpenUrl CrossRef PubMed ↵ Ning , J. et al. Functional virus-specific memory T cells survey glioblastoma . Cancer Immunol Immunother 71 , 1863 – 1875 ( 2022 ). OpenUrl PubMed ↵ Curtis , C. et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups . Nature 486 , 346 – 352 ( 2012 ). OpenUrl CrossRef PubMed Web of Science ↵ Lambrechts , D. et al. Phenotype molding of stromal cells in the lung tumor microenvironment . Nature Medicine 24 , 1277 – 1289 ( 2018 ). OpenUrl CrossRef PubMed Zilionis , R. et al. Single-Cell Transcriptomics of Human and Mouse Lung Cancers Reveals Conserved Myeloid Populations across Individuals and Species . Immunity 50 , 1317 – 1334 .e1310 ( 2019 ). OpenUrl CrossRef PubMed ↵ Guo , X. et al. Global characterization of T cells in non-small-cell lung cancer by single-cell sequencing . Nature Medicine 24 , 978 – 985 ( 2018 ). OpenUrl CrossRef PubMed Sade-Feldman , M. et al. Defining T Cell States Associated with Response to Checkpoint Immunotherapy in Melanoma . Cell 175 , 998 – 1013.e1020 ( 2018 ). OpenUrl CrossRef PubMed ↵ Yost , K. E. et al. Clonal replacement of tumor-specific T cells following PD-1 blockade . Nature Medicine 25 , 1251 – 1259 ( 2019 ). OpenUrl CrossRef PubMed Methods References 56. ↵ Enouz , S. , Carrié , L. , Merkler , D. , Bevan , M. J. & Zehn , D . Autoreactive T cells bypass negative selection and respond to self-antigen stimulation during infection . Journal of Experimental Medicine 209 , 1769 – 1779 ( 2012 ). OpenUrl Abstract / FREE Full Text 57. ↵ Virassamy , B. et al. Intratumoral CD8(+) T cells with a tissue-resident memory phenotype mediate local immunity and immune checkpoint responses in breast cancer . Cancer Cell 41 , 585 – 601 e588 ( 2023 ). OpenUrl CrossRef PubMed 58. ↵ Stewart , T. J. & Abrams , S. I . Altered Immune Function during Long-Term Host-Tumor Interactions Can Be Modulated to Retard Autochthonous Neoplastic Growth . The Journal of Immunology 179 , 2851 – 2859 ( 2007 ). OpenUrl PubMed 59. ↵ Christo , S. N. et al. Discrete tissue microenvironments instruct diversity in resident memory T cell function and plasticity . Nat Immunol 22 , 1140 – 1151 ( 2021 ). OpenUrl CrossRef PubMed 60. ↵ Fonseca , R. et al. Runx3 drives a CD8(+) T cell tissue residency program that is absent in CD4(+) T cells . Nat Immunol 23 , 1236 – 1245 ( 2022 ). OpenUrl CrossRef PubMed 61. ↵ Robinson , M. D. , Mccarthy , D. J. & Smyth , G. K. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data . Bioinformatics 26 , 139 - 140 ( 2010 ). OpenUrl CrossRef PubMed Web of Science 62. ↵ Robinson , M. D. & Oshlack , A . A scaling normalization method for differential expression analysis of RNA-seq data . Genome Biology 11 , R25 ( 2010 ). OpenUrl CrossRef PubMed 63. ↵ Molania , R. , Gagnon-Bartsch , J. A. , Dobrovic , A. & Speed , T. P . A new normalization for Nanostring nCounter gene expression data . Nucleic Acids Research 47 , 6073 – 6083 ( 2019 ). OpenUrl CrossRef PubMed 64. ↵ Gandolfo , L. C. & Speed , T. P . RLE plots: Visualizing unwanted variation in high dimensional data . PLOS ONE 13 , e0191629 ( 2018 ). OpenUrl CrossRef PubMed 65. ↵ Risso , D. , Schwartz , K. , Sherlock , G. & Dudoit , S . GC-Content Normalization for RNA-Seq Data . BMC Bioinformatics 12 , 480 ( 2011 ). OpenUrl CrossRef PubMed 66. ↵ Ritchie , M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies . Nucleic Acids Research 43 , e47 – e47 ( 2015 ). OpenUrl CrossRef PubMed 67. ↵ Park , S. L. et al. Divergent molecular networks program functionally distinct CD8(+) skin-resident memory T cells . Science 382 , 1073 – 1079 ( 2023 ). OpenUrl CrossRef PubMed 68. ↵ Huang , Y. , Mccarthy , D. J. & Stegle , O . Vireo: Bayesian demultiplexing of pooled single-cell RNA-seq data without genotype reference . Genome Biology 20 ( 2019 ). 69. ↵ Stuart , T. et al. Comprehensive Integration of Single-Cell Data . Cell 177 , 1888 – 1902.e1821 ( 2019 ). OpenUrl CrossRef PubMed 70. ↵ Korsunsky , I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony . Nature Methods 16 , 1289 – 1296 ( 2019 ). OpenUrl CrossRef PubMed 71. ↵ Wickham , H. et al. Welcome to the Tidyverse . Journal of Open Source Software 4 , 1686 ( 2019 ). OpenUrl CrossRef 72. ↵ Ouyang , J. F. , Kamaraj , U. S. , Cao , E. Y. & Rackham , O. J. L . ShinyCell: simple and sharable visualization of single-cell gene expression data . Bioinformatics 37 , 3374 – 3376 ( 2021 ). OpenUrl CrossRef PubMed 73. ↵ Orenbuch , R. et al. arcasHLA: high-resolution HLA typing from RNAseq . Bioinformatics 36 , 33 – 40 ( 2020 ). OpenUrl CrossRef PubMed 74. ↵ Gu , Z. , Eils , R. & Schlesner , M . Complex heatmaps reveal patterns and correlations in multidimensional genomic data . Bioinformatics 32 , 2847 – 2849 ( 2016 ). OpenUrl CrossRef PubMed 75. ↵ Benjamini , Y. & Hochberg , Y . Royal Statistical Society Publications . Journal of the Royal Statistical Society: Series B (Methodological ) 57 ( 1995 ). 76. ↵ Pereira , B. et al. The somatic mutation profiles of 2,433 breast cancers refine their genomic and transcriptomic landscapes . Nature Communications 7 , 11479 ( 2016 ). OpenUrl PubMed 77. ↵ Gendoo , D. M. A. et al. Genefu: an R/Bioconductor package for computation of gene expression-based signatures in breast cancer . Bioinformatics 32 , 1097 – 1099 ( 2016 ). OpenUrl CrossRef PubMed 78. ↵ Molania , R. et al. Removing unwanted variation from large-scale RNA sequencing data with PRPS . Nature Biotechnology 41 , 82 – 95 ( 2023 ). OpenUrl CrossRef PubMed 79. ↵ Robin , X. et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves . BMC Bioinformatics 12 , 77 ( 2011 ). OpenUrl CrossRef PubMed 80. ↵ Schienstock , D. , Hor , J. L. , Devi , S. & Mueller , S. N . Cecelia: a multifunctional image analysis toolbox for decoding spatial cellular interactions and behaviour . Nature Communications 16 ( 2025 ). 81. ↵ Stringer , C. & Pachitariu , M . Cellpose3: one-click image restoration for improved cellular segmentation . Nature Methods 22 , 592 – 599 ( 2025 ). OpenUrl PubMed 82. ↵ Hahsler , M. , Piekenbrock , M. & Doran , D. dbscan: Fast Density-Based Clustering with R . Journal of Statistical Software 91 , 1 – 30 ( 2019 ). OpenUrl 83. ↵ Nüssing , S. et al. Efficient CRISPR/Cas9 Gene Editing in Uncultured Naive Mouse T Cells for In Vivo Studies . The Journal of Immunology 204 , 2308 – 2315 ( 2020 ). OpenUrl PubMed 84. ↵ Izon , D. J. et al. Notch1 Regulates Maturation of CD4+ and CD8+ Thymocytes by Modulating TCR Signal Strength . Immunity 14 , 253 – 264 ( 2001 ). OpenUrl CrossRef PubMed Web of Science 85. ↵ Holze , H. et al. Analysis of synthetic cellular barcodes in the genome and transcriptome with BARtab and bartools . Cell Rep Methods 4 , 100763 ( 2024 ). OpenUrl PubMed View the discussion thread. Back to top Previous Next Posted July 23, 2025. Download PDF Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Antigen reactivity defines tissue-resident memory and exhausted T cells in tumours Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Antigen reactivity defines tissue-resident memory and exhausted T cells in tumours Thomas N. Burn , Jan Schröder , Luke C. Gandolfo , Maleika Osman , Elanor N. Wainwright , Enid Y. N. Lam , Keely M. McDonald , Rachel B. Evans , Shihan Li , Daniel Rawlinson , Lachlan Dryburgh , Ali Zaid , Zoltan Maliga , Dominick Schienstock , Philippa Meiser , Hyun Jae Lee , Hongjin Lai , Marcela L. Moreira , Pirooz Zareie , Louis H-Y. Lee , Lutfi Huq , Susan N. Christo , Justine J. W. Seow , Keith A. Ching , Stéphane M Guillaume , Kathy Knezevic , Simone L. Park , Maximilien Evrard , Jason Waithman , Thomas Gebhardt , Scott N. Mueller , Georgina E. Riddiough , Marcos V. Perini , Simon C. H. Tsao , Terence P. Speed , Peter K. Sorger , Sherene Loi , Francis R. Carbone , Stephanie Gras , Timothy S. Fisher , Bas J. Baaten , Mark A. Dawson , Laura K. Mackay bioRxiv 2025.07.23.666465; doi: https://doi.org/10.1101/2025.07.23.666465 Share This Article: Copy Citation Tools Antigen reactivity defines tissue-resident memory and exhausted T cells in tumours Thomas N. Burn , Jan Schröder , Luke C. Gandolfo , Maleika Osman , Elanor N. Wainwright , Enid Y. N. Lam , Keely M. McDonald , Rachel B. Evans , Shihan Li , Daniel Rawlinson , Lachlan Dryburgh , Ali Zaid , Zoltan Maliga , Dominick Schienstock , Philippa Meiser , Hyun Jae Lee , Hongjin Lai , Marcela L. Moreira , Pirooz Zareie , Louis H-Y. Lee , Lutfi Huq , Susan N. Christo , Justine J. W. Seow , Keith A. Ching , Stéphane M Guillaume , Kathy Knezevic , Simone L. Park , Maximilien Evrard , Jason Waithman , Thomas Gebhardt , Scott N. Mueller , Georgina E. Riddiough , Marcos V. Perini , Simon C. H. Tsao , Terence P. Speed , Peter K. Sorger , Sherene Loi , Francis R. Carbone , Stephanie Gras , Timothy S. Fisher , Bas J. Baaten , Mark A. Dawson , Laura K. Mackay bioRxiv 2025.07.23.666465; doi: https://doi.org/10.1101/2025.07.23.666465 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Immunology Subject Areas All Articles Animal Behavior and Cognition (7617) Biochemistry (17633) Bioengineering (13856) Bioinformatics (41841) Biophysics (21399) Cancer Biology (18529) Cell Biology (25422) Clinical Trials (138) Developmental Biology (13352) Ecology (19860) Epidemiology (2067) Evolutionary Biology (24281) Genetics (15582) Genomics (22461) Immunology (17700) Microbiology (40295) Molecular Biology (17140) Neuroscience (88413) Paleontology (666) Pathology (2823) Pharmacology and Toxicology (4813) Physiology (7632) Plant Biology (15107) Scientific Communication and Education (2042) Synthetic Biology (4284) Systems Biology (9808) Zoology (2267)