Newborn and child-like molecular signatures in older adults stem from TCR shifts across human lifespan

CD8+ T cells provide robust antiviral immunity, but how epitope-specific T cells evolve across the human lifespan is unclear. Here we defined CD8+ T cell immunity directed at the prominent influenza epitope HLA-A*02:01-M158–66 (A2/M158) across four age groups at phenotypic, transcriptomic, clonal and functional levels. We identify a linear differentiation trajectory from newborns to children then adults, followed by divergence and a clonal reset in older adults. Gene profiles in older adults closely resemble those of newborns and children, despite being clonally distinct. Only child-derived and adult-derived A2/M158+CD8+ T cells had the potential to differentiate into highly cytotoxic epitope-specific CD8+ T cells, which was linked to highly functional public T cell receptor (TCR)αβ signatures. Suboptimal TCRαβ signatures in older adults led to less proliferation, polyfunctionality, avidity and recognition of peptide mutants, although displayed no signs of exhaustion. These data suggest that priming T cells at different stages of life might greatly affect CD8+ T cell responses toward viral infections.

Ex vivo epitope-specific, single-cell paired TCRαβ analyses across age groups remain rare and were performed by us for influenzaspecific 16 and SARS-CoV-2-specific CD8 + T cells 13 ; however, not across the human lifespan.Highly functional public (shared) clonotypes dominated adult A2/M1 58 -specific TCRαβ repertoires 16,26,27 , whereas older adults had prominent private clonotypes (not shared) 16 .Children had fewer expanded SARS-CoV-2-specific clonotypes compared to adults, despite common TCRαβ motifs 13 .Underlying mechanisms and functional consequences of age-related changes within epitope-specific TCR repertoire across the human lifespan remain unresolved.Limited epitope-specific T cell single-cell RNA-sequence (scRNA-seq) studies focus on adult versus older bulk T lymphocytes in healthy individuals 28,29 or those with COVID-19 (ref.30), but not across the human lifespan.
We defined epitope-specific CD8 + T cell immunity across the human lifespan ex vivo in newborns, children, adults and older adults.We incorporated single-cell transcriptome and paired TCRαβ analyses to define epitope-specific T cells directed at the prominent and conserved HLA-A*02:01-restricted M1 58-66 peptide derived from influenza A viruses (A2/M1 58 ) (refs.11,16,31-33).Public clonotypes present across different individuals 16,26,27,34 allowed us to track numerical, phenotypic, functional and molecular changes within public A2/M1 58 + CD8 + TCRαβ clonotypes across human lifespan.We identified age-related TCR repertoire shifts within older epitope-specific CD8 + T cells, stemming from newborn/child-like molecular signatures detected in older adults.Our findings have implications for rationally designed T cell-targeted vaccines and immunotherapies across age groups.
Open symbols indicate <10 A2/M1 58 + CD8 + T cells counted, which were not used for phenotypic analyses.h,i, Frequency of naive and memory subsets within the total CD8 + T cell (h) or A2/M1 58 + CD8 + T cell populations (i) across all age groups.j-m, Frequencies of CD57 and PD-1 expression on CD8 + T cells (j and l, respectively) and A2/M1 58 + CD8 + T cells (k and m, respectively) per age group.responses were not modulated by concurrent CMV infection, conversely to total CD8 + T cells (Extended Data Fig. 2).Overall, A2/M1 58 -specific CD8 + T cell phenotypes change across the human lifespan, but these changes are distinct from total CD8 + T cells, with absence of terminally differentiated A2/M1 58 + CD8 + T cells in older adults.

+ CD8 + T cell clusters across human lifespan
We defined molecular gene signatures in healthy HLA-A*02:01expressing newborns, children, adults and older adults (n = 3 per group), selected to reflect phenotypic heterogeneity, including dispersion of heterogenicity within T CM and T naive populations in children and older adults, and lack of T EMRA in older adults (Extended Data Fig. 3a).Ex vivo-isolated A2/M1 58 + CD8 + T cells were index-sorted for single-cell transcriptome analysis.A total of 793 cells across all age groups passed quality control for analyses.
Overall, scRNA-seq combined with protein expression phenotypes revealed differential molecular signatures of A2/M1

Young gene profiles resemble gene profiles for older adults
To identify age group-distinctive gene signatures, we performed differential gene expression analysis between cells stratified by age and investigated heterogenicity within A2/M1 58 + CD8 + T cells in children and older adults, and lack of exhaustion and terminal differentiation in older A2/M1 58 + CD8 + T cells.Newborn A2/M1 58 + CD8 + T cells expressed naive gene signatures, with high expression of CCR7, SELL and TNFAIP3, which became less prominent in children and adults, but increased in older adults.Adult A2/M1 58 + CD8 + T cells expressed effector-memory phenotype profiles, including LITAF, KLRG1, CXCR4 (contributing to homeostasis self-renewal and homing) 39 and NK-like signatures, KLRK1 encoding NKG2D (involved in stress-induced cytotoxic response) 40 , less prominent in children and older adults (Fig. 2d).Mixed naive/memory phenotypes in children and older adults (Fig. 1i) were verified by shared mixed gene expression profiles (Fig. 2d).
As TCR signatures are important in antiviral immune responses, we investigated TCR-associated genes across age groups.TRBV19 and TRAV27 gene expression, key features of highly functional A2/ M1 58 + CD8 + T cells 26,27,38 , increased from children to adults, whereas TRBV27 expression was higher in newborns and older adults (Fig. 2d).Age-specific TCR changes suggest potential shifts in dominance of key TCR clonotypes across the human lifespan.Overall, differentially expressed gene profiles across the human lifespan revealed clear distinctions between naive A2/M1 58 + CD8 + T cells in newborns, mixed naive/memory profiles in children, cytotoxic-effector-memory profiles in adults and inversion toward naive profiles in older adults, without evidence for exhaustion and terminal differentiation, agreeing with phenotypic data (Fig. 1i-m).

Age-specific molecular changes stem from distinct clonotypes
UMAP analysis demonstrated heterogeneity among A2/M1 58 + CD8 + T cells, with clusters segregating cells based on age group and phenotype, with newborn and older adults sharing common gene and phenotypic signatures (Fig. 2).We thus asked whether distinct lineages can explain molecular and phenotypic differences observed across the human lifespan.We hypothesized two scenarios (1) A2/M1 58 -specific TCR clonotypes are shared between children, adults and older adults, with older TCR clonotypes reverting to a quiescent gene profile similar to newborns and children; and (2) older adults have distinct A2/ M1 58 -specific TCR clonotypes compared to adults, with gene expression profiles similar to those detected in newborn and children.To evaluate the relationship between differentiation and fate of the clonal lineage of A2/M1 58 + CD8 + T cells across age groups, we performed pseudotime trajectory analysis using partition-based graph abstraction (PAGA) based on gene expression data, combined with clonotype information to infer the connections and order of differentiation throughout T cell states (Fig. 3).
Similar to UMAP, PAGA analysis confirmed that A2/M1 58 + CD8 + T cells separate into clusters based on age group and phenotype with additional segregation of newborns and older adults into two distinct clusters, clusters 3 and 4 (Fig. 3a,b).TCR mapping identified highly functional TRAV27-TRBV19-CDR3-RS TCR features dominated cluster 0 and 2, whereas TRBV27-expressing clonotypes dominated the older adult cluster 4 (Fig. 3c).
We established the A2/M1 58 + CD8 + T cell pseudotime trajectory across five PAGA clusters.Each cell was assigned a state based on dimensionality reduction or clustering in PAGA.Pseudotime values were inferred for each cell, allowing cells to be ordered along a trajectory across which they can be considered as proxy of cell lineage differentiation (Fig. 3d).Cluster 3, consisting of predominantly naive T cells from newborns, was the logical choice for the differentiation root (Fig. 3b,d).A trajectory was identified by connecting root cluster 3 with clusters 1 and 0, and terminating in cluster 2 (trajectory 1) (Fig. 3a).This trajectory largely correlated with clonal expansion and increased usage of TRAV27-TRBV19-CDR3-RS features (Fig. 3c).Older TRBV27-expressing A2/M1 58 + CD8 + T cells dominating cluster 4 (Fig. 3c), suggested deviation, branching off from cluster 0 and terminating in cluster 4 (trajectory 2), indicating distinct lineage differentiation in older adults.
To quantify variation of gene and phenotypic signatures along the pseudotime (Fig. 3d), we performed Loess smoothing fit.T cell phenotype analysis showed rapid declining growth rates of naive A2/ M1 58 + CD8 + T cells along the pseudotime, whereas growth rates of T CM and T EM A2/M1 58 + CD8 + T cells increased, the latter largely consisting of cytotoxic effectors from adults displaying TRBV19-CDR3-RS features (Fig. 3a-c).Indeed, gene expression analysis demonstrated a rapid increase of TRAV27 and TRBV19, along with NKG7, GNLY and GZMA expression, representing increased cytotoxicity profiles along trajectory 1. Conversely, naive-associated genes, including CD27 and transcription factors FOS and JUNB declined over the pseudotime, whereas TCR signaling genes (LCK) remained highly expressed.
Overall, our trajectory analysis demonstrated nonlinear differentiation of A2/M1 58 + CD8 + T cells across the human lifespan.Trajectory 1 consisted of newborns, children and adults along a differentiation branch toward the effector cytotoxic A2/M1 58 + CD8 + T cells associated with optimal TCR features dominating adult A2/M1 58 + CD8 + repertoires.Trajectory 2 was dominated by older less-differentiated T cells, encompassing a distinct clonal lineage.
Overall, A2/M1 58 + CD8 + TCRαβ repertoires are highly diverse in newborns, greatly cluster in children and adults, before diversifying in older adults.These changes are attributed to both TCR α-chains and β-chains.

Young public TCRs replaced by private TCRs in older adults
Circos analysis of pooled TRAV and TRBV sequences was performed to understand how changes in TCRαβ diversity related to gene segment usage (Fig. 4d and Extended Data Fig. 4b).Newborn A2/M1 58 + CD8 + TCRαβ repertoires were highly diverse, although newborns expressed TRBV19 bias 10,11 .Clonally expanded TRBV19-expressing TCRs became more prevalent in children and adults.Strong TRAV27-TRBV19 associations were observed in children (10 out of 12 children, 82 out of 714  TCRαβ clonotypes) and became more pronounced in adults (8 out of 8 adults, 96 out of 409 TCRαβ clonotypes).Consistent with kPCA and neighbor distance distribution analyses, higher TRAV and TRBV diversity was observed among older adults.TRBV19 became less prevalent in six out of ten older adults and the TRBV19-TRAV27 association was observed in four out of ten older donors (33 out of 409 TCRαβ clonotypes).Instead, older adults displayed large clonal expansions expressing other TRAV and TRBV gene segments.We defined distribution of high-prevalent public (shared clonotypes, detected at least twice within each individual), low-prevalent public (shared clonotypes, detected only once within each individual) and high-prevalent private clonotypes (not shared, detected at least Public Private twice in a single individual).High-prevalent public TCRαβ clonotypes were observed during childhood, peaked in adults, before decreasing in older adults, particularly the previously identified full public TCRαβ clonotype (TRAV27/TRAJ42, CDR3α-GAGGGSQGNLIF, TRBV19/ TRBV2-7 and CDR3β-CASSIRSSYEQYF) associated with optimal immunity in adults 16,26,27,34 (Fig. 4e and Extended Data Fig. 4c).Loss of public TCRαβ clonotypes in older adults was associated with increased prevalence of private TCR clonotypes (Fig. 4e and Extended Data Fig. 4d).
Our data suggest that repeated influenza virus exposures expand public TCRαβ clonotypes in children and adults, which are replaced by private TCRαβ clonal expansions in older adults.

Young public CDR3αβ-motifs are less frequent in older adults
As hypervariable CDR3α and CDR3β regions predominantly mediate fine pHLA-I specificity, we dissected CDR3αβ regions by analyzing length and amino acid sequence to identify TCR motifs.Conversely to diverse lengths observed for A2/M1 58 + CD8 + CDR3α regions across age groups, newborn CDR3β sequences were predominantly 8-10 amino acids in length, whereas an eight-amino acid length dominated in children, adults and older adults (Extended Data Fig. 5a,b and Supplementary Table 2).
We identified CDR3 motif similarities to highlight key conserved residues driving A2/M1 58 -specific TCR recognition either within or between age groups.No CDR3α motif was identified in newborns, attributed to high TCR α-chain diversity.A single top-scoring, TRAV27-TRAJ42-associated, glycine-rich, CDR3α-(CA)GGGSQG(NLI) motif was identified in children, adults and older adults (Fig. 5a).A single glycine was enriched above background levels, suggesting limited involvement in specific pHLA interaction 44 .Clonotypes expressing the full public TRAV27-TRAJ42-associated CDR3α-GAGGGGSQGNLIF or shorter variants, including GGGSQG, GGG or GG were shared between age groups with frequencies increasing in children, peaking in adults and decreasing in older adults (Fig. 5b,c), confirming our finding that TCRα chains contributed to diversifying the older TCR repertoire (Fig. 4b).
Generally, 'IV'-expressing CDR3β clonotypes were detected at a low frequency in children, adults and older adults, except in child TN022.No 'IV'-expressing clonotypes were shared between age groups (Fig. 5d,e and Supplementary Table 2).
Both children and older adults expressed less-prominent CDR3β-'IY'-motifs with highest frequencies in children (Fig. 5a,e).'IY'-expressing CDR3β motifs were associated with TRBV19 and variable TRBJ gene usage, resulting in limited sharing among age groups (Fig. 5a,d).CDR3β-'(I)F'-motif expressing clonotypes were identified in newborns, children and older adults, and were associated with TRBV19, TRBV27 and variable TRBJ segments (Fig. 5a,e).Despite strong enrichment of the 'F' residue, CDR3β-(I)F-expressing clonotypes were detected at low frequencies and were not shared between age groups, with the exception of DMC19 dominated by the highly frequent CDR3β-'F'-expressing clonotype (71.9%) (Fig. 5d,e and Supplementary Table 2).
Overall, we identified public-associated CDR3α and CDR3β motifs, with CDR3α 'GGGSQG' and CDR3β-'RS' motifs being most prominent.Public-associated CDR3α and CDR3β motifs became less frequent in older adults and were replaced with high-frequency private CDR3 motifs uniquely identified in a single individual.CDR3β sequence, the thickness is proportional to the number of TCR clones with the respective pair.The number of sequences considered for each circos plot is shown at the right bottom.e, Frequency of high-prevalent (>2 similar TCRs within a single individual) public (shared) and private (not shared) clonotypes across different age groups.Dark red represents high-prevalent public TCR (TRAV27, TRAJ42, CDR3α GAGGGSQGNLIF, TRBV19, TRBV2-7 and CDR3β CASSIRSSYEQYF), whereas the light red are clonotypes expressing the full public TCRβ chain (TRBV19, TRBV2-7 and CDR3β CASSIRSSYEQYF) but the TCR α-chain could not be identified.Numbers in the graph represent the number of donors in which this specific high-prevalent clonotype was identified.Statistical analysis was performed using a two-sided Kruskal-Wallis test with Dunn's correction for multiple tests.P values are indicated above the graphs.N, newborn; C, children; A, adult; OA, older adult.
Thus, our observations of age-dependent prevalence of public-associated CDR3β motifs were confirmed in independent large cohorts of aging donors.

Higher probability of generation underpins public TCRs
To understand why A2/M1 58 + CD8 + public clonotypes were prominent in children and adults, and declined in older adults, we estimated the probability of generation (P gen ) of A2/M1 58 + CD8 + TCR α-chains and β-chains using TCRdist 44 .Larger P gen values are associated with easier-to-generate clonotypes, whereas smaller P gen values indicate harder-to-generate (rarer) clonotypes.
TCR α-chain P gen values were similar for all ages (Fig. 6c).Significantly greater P gen values were observed within adult TCR β-chains compared to children and older adults (Fig. 6d), supported by fewer inferred TCRβ N insertions in adult A2/M1 58 -specific TCR β-chains (Fig. 6e).The number of inferred exonuclease deletions were similar across age groups (Fig. 6f), suggesting that relatively easy-to-generate TCR β-chain formation in adults may be driven by reduced terminal deoxynucleotidyl transferase (TdT) activity during V(D)J recombination following early childhood.Indeed, TdT expression decreases between birth and adulthood 48,49 , coinciding with increased public A2/M1 58 + CD8 + cells after childhood.
Overall, adult TRBV19-expressing clonotypes had a larger P gen compared to children and older adults, likely driven by expanded full public clonotypes.Child and adult public-associated TRBV19-expressing clonotypes had larger P gen values compared to other TRBV-expressing clonotypes (Fig. 6g).Conversely, highly prevalent non-TRBV19-expressing TCR clonotypes in children and older adults had significantly lower P gen values compared to newborns (Fig. 6g).Patterns for total TCR clonotypes were complemented by TCR logo analyses depicting V and J gene frequencies, CDR3 amino acid sequences and inferred rearrangement structures of grouped TCRs (Extended Data Fig. 5c,d).TCR clusters expressing public-associated CDR3α-'(G)GGSQG' or CDR3β-'RS' and -'IG' had higher probabilities of generation in children, adults and older adults (Extended Data Fig. 5c,d).TCR clusters expressing private-associated CDR3β-'(I)F' had lower probability of TCR recombination (Extended Data Fig. 5d).
Overall, child and adult A2/M1 58 + CD8 + TCRs, often expressing TRBV19 public-associated CDR3α-'GGGSQG' and/or CDR3β-'RS' motifs, are easier to generate, explaining why they are shared between HLA-A*02:01-expressing individuals.Conversely, large clonal expansions in older adults, often associated with non-TRBV19 non-public CDR3α and/or CDR3β sequences, with increased numbers of insertions and deletions, are harder to generate and less likely to be shared.
Overall, highly prevalent clonotypes expressing public features in children and adults were associated with robust A2/M1 58 + CD8 + T cell proliferation.These data demonstrated the importance of TCRαβ diversity, as ex vivo low-prevalent TCRs also proliferated upon in vitro stimulation.Further studies are needed to demonstrate whether broad-spectrum TCR proliferation is also observed following influenza infection.
Overall, children's proliferating A2/M1 58 + CD8 + T cells had the highest polyfunctionality whereas newborns had the lowest polyfunctionality followed by older adults, in line with their lower cytotoxic gene expression profiles ex vivo.
Overall, CD8 co-receptors enhanced binding avidity of age-specific private TCRs representing newborn, child, one adult and one older TCR, but had less impact on weak-binding older TCRs.
Overall, ex vivo A2/M1 58 -specific TCRs from older adults had reduced activation capacity compared to public and other age-specific TCRs.The CD8 co-receptor decreased the public, newborn (pN1) and adult (A1) TCR activation threshold, potentially resulting from improved binding avidity (Fig. 8b-f).
Overall, age-specific TCRs display reduced ability to recognize M1 58-66 peptide variants.Ex vivo older private TCRs displayed unique A2/M1 58 binding profiles, underpinning their reduced binding capacity, avidity, functionality and proliferating capacity compared to other prominent age-specific TCRs, especially TCRs with public-associated features found at high frequency in children and adults.

Discussion
We linked age-specific single-cell molecular gene profiles with phenotypes, functionality and paired single-cell TCRαβ repertoires of influenza-specific HLA-A*02:01/M1 58-66 -specific CD8 + T cells.Unexpectedly, older A2/M1 58 + CD8 + T cells did not reach terminally differentiated or exhausted end points.Instead, reduced functionality was associated with loss of highly functional public TCRαβ clonotypes dominating younger TCRαβ repertoires.Conversely, large clonal expansions of less-functional private TCRαβs dominated older TCRαβ repertoires.Age-specific transcriptomes supported a linear differentiation trajectory from newborns to children, then adults, whereas suboptimal clonal resets in older adults were associated with newborn/child-like molecular signatures.
Age-specific A2/M1 58 + CD8 + T cells transcriptomes corroborated age-specific phenotypic profiles, including mixed naive/memory phenotypes in children and older adults.Newborn naive T cells uniquely expressed TLB.The role of TLB, encoding TNF-C, remains ill defined.Cytotoxic and TRAV27/TRBV19 genes dominated child and adult A2/ M1 58 -specific transcriptomes and their public TCR features resulted in higher proliferative capacity and polyfunctionality, compared to newborns and older adults.Older A2/M1 58 -specific transcriptomes, dominated by CXCR4 (ref.39), KLF2, SELL, TXNIP, PIK3IP1, CD37 and TRBV27, displayed less-differentiated cell states, lacked exhaustion genes, expressed AP-1 transcription factors FOS and JUN (progenitor of exhaustion in acute infection 53 ), C-JUN (resistance to exhaustion 43 ) and distinct clonal lineages.Trajectory analysis supported lack of a terminally differentiated end stage and suboptimal clonal reset in older adults.Suboptimal clonotypes expressing older private features were detected at a low frequency in younger age groups.
During infections, 'best-fit' high-avidity clonotypes are selected from naive TCR repertoires and expand following subsequent encounters 54,55 .We demonstrate that 'best-fit' high-avidity public clonotypes peak in adults and are gradually replaced by low-avidity clonotypes in older adults.TRBV19/CDR3β-RS-expressing clonotypes dominate over TRBV19-expressing clonotypes with other public CDR3β-associated features, possibly because CDR3β-RS only requires two amino acids to bind the A2/M1 58 complex regardless of the TCR α-chain 27,38 ; however, TCR repertoire diversity remains important to protect against escape variants 5,27 .How this delicate balance between expanded best-fit clonotypes and TCR diversity is maintained following repeated infections remains unexplored.Children and adults maintain diverse TCR repertoires during in vitro expansion.Reduced public TCRαβ clonotypes and TCRαβ diversity within older TCR repertoires explains why older adults, in the absence of pre-existing antibodies, are at higher risk of severe disease during Article https://doi.org/10.1038/s41590-023-01633-8influenza epidemics and pandemics.Conversely, highly functional and diverse public TCRαβ repertoires in children clarifies their relative superiority in fighting influenza infections.Understanding how we can preserve this delicate balance between expansion of 'best-fit' TCRs while maintaining TCR diversity may be the Holy Grail in defining how we can maintain optimal immunity across the human lifespan through vaccination and/or immunotherapies.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material.If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.© The Author(s) 2 02 3 1 Department of Microbiology and Immunology, University of Melbourne at the Peter Doherty Institute for Infection and Immunity, Melbourne, Victoria, Australia. 2Department of Hematopoiesis, Sanquin Research and Landsteiner Laboratory, Amsterdam UMC, University of Amsterdam, Amsterdam, The Netherlands. 3 Department of Immunology, St. Jude Children's Research Hospital, Memphis, TN, USA. 4 School of Medical Sciences and The Kirby Institute, UNSW Sydney, Sydney, New South Wales, Australia. 5Viral and Structural Immunology Laboratory, Department of Biochemistry and Chemistry, La Trobe Institute for Molecular Science, La Trobe University, Bundoora, Victoria, Australia. 6Immunity Program and Department of Biochemistry and Molecular Biology, Biomedicine Discovery Institute, Monash University, Clayton, Victoria, Australia. 7School of Health Sciences and School of Medicine, University of Tasmania, Launceston, Tasmania, Australia. 8Victorian Infectious Diseases Reference Laboratory, The Royal Melbourne Hospital at The Peter Doherty Institute for Infection and Immunity, Melbourne, Victoria, Australia. 9Obstetrics, Nutrition and Endocrinology Group, Department of Obstetrics and Gynaecology, University of Melbourne, Melbourne, Victoria, Australia. 10Deepdene Surgery, Deepdene, Victoria, Australia. 11Institute of Infection and Immunity, Cardiff University School of Medicine, Cardiff, UK. 12 School of Health and Biomedical Science, RMIT University, Melbourne, Victoria, Australia.13   Tasmanian Vaccine Trial Centre, Clifford Craig Foundation, Launceston General Hospital, Launceston, Tasmania, Australia. 14These authors contributed equally: Fabio Luciani, Katherine Kedzierska.e-mail: kkedz@unimelb.edu.auArticle https://doi.org/10.1038/s41590-023-01633-8

Study participants and ethics
Donors were selected from a large HLA-typed randomly recruited lifespan cohort n ≥ 500, consisting of 154 newborns, 30 children, 360 adults/older adults, based on their expression of HLA-A*02:01.Overall, 11 HLA-A*02:01 + newborns (0 years), 12 HLA-A*02:01 + children (median 9 years, range 3-16), 20 HLA-A*02:01 + adults (median 37 years, range 18-58) and 18 HLA-A*02:01 + older adults (median 72 years, range 63-88)  were included in the study (Supplementary Table 1).Adults and older adults were recruited via the University of Melbourne (UoM), Deepdene Medical Clinic and the Australian Red Cross Lifeblood.Children were recruited via the Launceston General Hospital and St Jude Children's Research Hospital.Umbilical cord blood, reflecting newborn's blood, was obtained via Mercy Hospital for Women.Peripheral blood was collected before the COVID-19 pandemic.All participants or their guardians provided informed written consent.Participants of the study did not receive any compensation.PBMCs were isolated using Ficoll-Paque (GE HealthCare) gradient centrifugation, and then cryopreserved in liquid nitrogen until required.HLA class I and II molecular genotyping was performed from genomic DNA by the Australian Red Cross Lifeblood.CMV status was determined as described previously 56

Single-cell RNA sequencing
Three HLA-A*02:01-expressing donors from each age group (Supplementary Fig. 2a) were selected for scRNA-seq analysis.A2/M1 58 + CD8 + T cells were TAME-enriched and then individually (index-) sorted into chilled 96-well twin.tecPCR plates (Eppendorf) containing lysis buffer (1 µl RNase inhibitor and 19 µl Triton X-100) after TAME and on a BD Aria III sorter.Libraries were generated as described previously 61,62 .A Nextera XT DNA Library Prep kit was used for the generation of sequencing libraries and sequencing performed on a NextSeq500 platform with 150-bp high-output paired-end chemistry for 901 A2/ M1 58 -tetramer + CD8 + T cells/donor and 101 controls.

scRNA-seq quality control, normalization and batch correction
Downstream analysis was performed in R using packages downloaded from Bioconductor v.3.10.Cells were removed from each batch if they did not meet these criteria: less than 40% reads aligned to mitochondrial genes, number of detected genes more than 400.The bulk samples and genes expressed in zero cells were removed.Normalization was performed using the NormalizeData function from Seurat (v.4.1.0).To minimize potential batch effects, the experiment was designed to distribute donors from each age group to be included across separate experiments.In addition, donors were distributed over across plates within each experiment.Batch effect was tested using the FindIntegra-tionAnchors and IntegrateData functions from Seurat 67 .For scRNA-seq bioinformatics analysis no correction for batch effect was performed because of poor evidence of bias and also for sufficient mix of cells among clusters based on experimental design (plates and experiment number).The observed segregation based on age group and phenotype was an expected biological bias given the experimental design.

Analysis of scRNA-seq data
Dimensionality reduction and clustering were also performed in Seurat using the normalized matrix.PCA was performed on the normalized data using the 3,000 most variable genes.Clustering was performed using the shared nearest neighbor modularity optimization-based clustering algorithm (FindClusters(resolution = 0.8, algorithm = 'louvain')) as implemented in Seurat.
Differential gene expression was performed using MAST (v.1.16.0) implemented in the function FindMarkers in Seurat v.4 and using two-sided P values to report the results REF_SEURAT 67 .Notably, the test used was MAST and batch was used as a latent variable with a log(FC) threshold of 0.3.Other parameters were kept as default.Signature scores were computed from the normalized single cell transcriptomic matrix as the average log(FPKM + 1) of all genes in the signature.Differential expression output across all the analyses is reported in Supplementary  Gene set enrichment analysis GSEA was performed using the R package fsgea (v.1.16.0).Normalized enrichment scores were assessed using the fgsea(…, maxSize = 500, nperm = 10,000) function across the curated Molecular Signatures Database (MSigDB) Hallmark, C2 curated gene sets consisting of canonical gene sets PID).Customized gene signatures for T cell phenotypes are reported in Supplementary Table 6, which were prepared by manually curating published data.

Trajectory inference
PAGA analysis was performed through Scanpy (v.1.7.1) (ref.68) with parameters as recommended 22 .The FPKM matrix following normalization (with Seurat), along with the top 20 PCAs (previously generated with Seurat) were used and visualization was performed using sc.pp.neighbors(n_neighbors = 8, n_pcs = 20) and a coarse-grained and simplified graph using sc.tl.paga.Clusters were calculated using sc.tl.louvain(resolution = 0.8) and visualization was performed using sc.pl.paga and sc.pl.draw_graph.Pseudotime analysis was performed using the diffusion map algorithm (sc.tl.dpt) by manually assigning an initial iroot value.Scaled pseudotime were used with Loess smoothing and were calculated as uniformly distributed mapping of the diffusion pseudotime values to preserve the cell order and account for heterogeneous distribution of gaps between pseudotime values.The growth rate with which T cell phenotypic/age group subsets change along the inferred pseudotime trajectories were calculated as the ratio between the difference in cell numbers and the scaled pseudotime values over a window of size 0.05.These values were then plotted using the geom_smooth R function with default parameters.

Single-cell RT-PCR and paired TCRαβ sequencing
A2/M1 58 + CD8 + T cells were TAME-enriched and subsequently individually (index-) sorted into chilled 96-well twin.tecPCR plates (Eppendorf) and immediately stored at −80 °C until required.Single-cell paired CDR3α and CDR3β regions were analyzed by multiplex-nested PCR with reverse transcription and followed by sequencing of the CDRα and CDRβ products, essentially as described previously 6,26,69 , except for using double amounts of reaction mix in the complementary DNA step for older adult samples.Sequences were analyzed with FinchTV.V-J regions were identified by IMGT query (www.imgt.org/IMGT_vquest).TCR sequences were parsed using the TCRdist analytical pipeline 44 .Clonotypes were defined as single-cell TCRαβ pairs that exhibit the same V, J and CDR3 regions.

A2/M1 58 -specific TCR motifs in publicly available datasets
A2/M1 58 -specific TCRβ sequences identified in our study were further verified in two publicly available bulk TCRβ datasets from independent cohorts of healthy individuals 8,47 .TCR was considered matched when it had the same CDR3β and genomic V segment.For all the donors in the cohorts, we calculated the total frequency of matched TCRs within each CDR3β-motif.As conventional HLA typing was not available for all donors in the independent cohorts, we divided them into HLA-A*

+ CD8 + T cell proliferation assay
PBMCs from HLA-A*02:01 + donors (~21 × 10 6 ) were pre-incubated with cell trace violet (Violet Proliferation Dye 450, BD Horizon) according to the manufacturer's instructions before generating A2/M1 58 -specific CD8 + T cell lines 6 .Briefly, one-third of the labeled PBMCs were pulsed with 10 µM M1 58-66 peptide (GILGFVFTL), or DMSO as unstimulated control, for 60 min at 37 °C, washed twice with RPMI and incubated with the remaining two-thirds of the non-peptide-pulsed autologous PBMCs in cRPMI at a final concentration of non-pulsed cells at 1 × 10 6 per well for each day of the proliferation assay.Cells were cultured in a 48-wells plate for 10 d at 37 °C and 5% CO 2 .Cultures were supplemented on day 4 with 20 U ml −1 rIL-2 (Roche) and were maintained with fresh medium containing 10 U ml −1 IL-2 when needed.On day 3, 4, 5, 6, 7, 9 (ICS) and 10, respective wells were collected, counted and washed once in MACS buffer and incubated with anti-human FcR block (20 µl per 1 × 10 7 cells) (Miltenyi Biotec) for 15 min on ice before staining with PE-streptavidin-conjugated tetramers (1:100 dilution in MACS buffer) for 1 h at room temperature.After one wash, cells were incubated for 30 min on ice with the same surface stain as described for the TAME, except without anti-CD71-BV421.Cells were subsequently washed once and resuspended in MACS buffer for single cell-(index)-sorting using a BD FACSAria III for subsequent TCRαβ sequencing (days 3, 5, 6, 7 and 10) or fixed with 1% PFA for acquiring on an LSR Fortessa II (day 4), followed by analysis using FlowJo software (v.10.8.1).Proliferation due in combination with A2/M1 58 tetramer staining of CD8 + T cells was used to track proliferating A2/M1 58 -specific CD8 + T cells over time.Ex vivo numbers of A2/M1 58 + CD8 + T cells on day 0 were based on the frequencies obtained from TAME analysis, as frequencies were relatively low.

Fig. 1 |
Fig. 1 | Age-related changes in A2/M1 58 + CD8 + T cell frequencies and phenotypes.a, 'Lifespan' HLA-A*02:01-positive cohort, median age and number of donors per age category.b, Age distribution within the HLA-A*02:01-expressing lifespan cohort.c, Representative FACS panels and gating strategy for A2/M1 58 + CD8 + T cells in the enriched fraction and phenotypic Horizontal bars indicate the median, dots represent individual donors, with n = 11 newborns, n = 12 children, n = 20 adults and n = 18 older adults (b-h,j,l) and n = 10 newborns, n = 12 children, n = 20 adults and n = 16 older adults (i,k,m).Black line is a locally estimated scatter-plot smoothing) Loess trend line with error bands shaded in gray representing 95% confidence interval (CI) (e,g).Technical replicates were not performed due to limited samples.Statistical analysis was performed using a two-sided Kruskal-Wallis with Dunn's correction for multiple tests.P values are indicated above the graphs.N, newborn; C, children; A, adult; OA, older adult.

Fig. 3 |
Fig. 3 | Single-cell transcriptomics shows evolutionary trajectory across human life.a, PAGA analysis of single cells (n = 793) identified five clusters (left) and their relative connectivity (middle) and colored based on the four UMAP clusters (right).b, PAGA analysis colored by age (left), phenotype (right).Single-cell phenotypes were obtained via index sorting.c, PAGA analysis colored by public TCR features (left), TRBV27 expression (middle) and clone size (right).Bar charts represent respective age, phenotype and TCR features distributions in each PAGA cluster.d, Trajectory derived from scRNA-seq data with colored

Fig. 4 |
Fig. 4 | Age-related changes in A2/M1 58 + CD8 + TCRαβ repertoire.a-e, A2/ M1 58 + CD8 + T cells were enriched by TAME followed by single-cell sorting for TCRαβ analysis.a, A 2D kernel principal-component analysis (PCA) projection of the A2/M1 58 + CD8 + TCR landscape colored by Vα, Jα, Vβ and Jβ gene usage (left to right) for all four age groups generated by TCRdist.Encoding clone size indicated by symbol size.b, TCRdiv diversity measures of the TCRα, TCRβ or paired TCR αβ-chains.c, smoothed density profiles of neighbor distance distribution are shown for each age group.A lower distribution peak indicates more clustered A2/M1 58 + CD8 + single TCRα, TCRβ or paired TCRαβ repertoire, average distance values for each age group are depicted within the plot.PDF, probability density function.d, TRAV and TRBV clonotype pairing per age group illustrated by circos plots.Left arch segment colors indicate TRAV usage, right outer arch colors depict TRBV usage.Connecting lines indicated TRAV-TRBV gene pairing and are colored based on their TRAV usage and segmented based on their CRD3α and

Fig. 5 |
Fig. 5 | Age-related changes within the A2/M1 58 + CD8 + CDR3αβ-motifs.a, The top-scoring A2/M1 58 + CD8 + CDR3α (left TCR logo) and CDR3β (right TCR logo) sequence motifs for each age group.Each logo depicts the V (left side) and J (right side) gene frequencies with the CDR3 amino acid sequence in the middle with the full height (top) and scaled (bottom) by per-residue reparative entropy to background frequencies derived from TCRs with matching gene-segment composition to highlight motif positions under selection.The middle section indicates the inferred rearrangement structure by source region (light gray for V-region, dark gray for J, black for D and red for N insertions) of the grouped receptors.b, Persistence of TCRα clonotypes expressing selected prominent CDR3α motifs across different age groups.Colors identify the most prominent

Fig. 7 |
Fig. 7 | Proliferation and polyfunctionality of A2/M1 58 + CD8 + T cells across human life.a,b, Total numbers (a) and fold increase (b) of A2/M1 58tetramer + CD8 + T cells from day 0 (ex vivo tetramer enrichment, due to low frequency), day 3, 4, 5, 6, 7 and 10 following in vitro M1 58-66 peptide stimulation of newborn, child, adult and older adult peripheral blood mononuclear cells (PBMCs) (no previous enrichment).c, Representative FACS plots indicating the gating strategy used to characterize dividing (red) and undivided (blue) A2/M1 58 + CD8 + T cells by the loss of cell trace violet over a 10-d expansion of representative donors.Gray dots represent total CD8 + T cells.d, Persistence of TCRα clonotypes expressing selected prominent CDR3α motifs across a 10-d expansion in each age group.Shared TCRα clonotypes are connected by colored lines.e, Persistence of TCRβ clonotypes expressing selected prominent CDR3β

58 + 8 Extended Data Fig. 2 |
CD8 + T cells (d-top panel) and number of A2/M1 58 + CD57 + CD8 + T cells (d-middle panel) or A2/M1 58 + PD-1 + CD8 + T cells (d-bottom panel) per 10 6 CD8 + T cells across all age groups.Co-expression of CD38 and HLA-DR in total CD8 + T cells (e) and A2/M1 58 + CD8 + T cells (f).Numbers of A2/M1 58 + phenotype + CD8 + T cells per 10 6 CD8 + T cell data were right shifted by 0.001 (that is absence of A2/M1 58 events in a specific phenotypic population are displayed as 0.001) (b,d-middle and bottom panel).Only samples with 10 or more total A2/M1 58 + events were included for phenotype analysis (b-f).Horizontal bars indicate the median, dots represent individual donors, with n = 10 newborns, n = 12 children, n = 20 adults and n = 16 older adults in b, d and f and n = 11 newborns, n = 12 children, n = 20 adults and n = 18 older adults in c and e. Technical replicates were not performed due to limited samples.Statistical analysis was performed using a two-sided Kruskal-Wallis with Dunn's correction for multiple tests (b, d-middle and bottom panel) or a two-sided Tukey's multiple comparisons test (c, d-top panel, e, f).Exact significant p-values are indicated above the graphs.N, newborns; C, children; A, adults; OA, older Adults.Article https://doi.org/10.1038/s41590-023-01633-CMV-relatedchanges in A2/M1 58 + CD8 + T cell phenotypes.Frequency of naïve, memory, CD57 or PD-1 expressing subsets within the total CD8 + T cell (a) or A2/M1 58 + CD8 + T cell populations (b) across all age groups split based on their CMV status.Horizontal bars indicate the median, dots represent individual donors.Horizontal bars indicate the median, dots represent individual donors, with n = 3 CMV − , n = 4 CMV + and n = 4 CMV unknown newborns, n = 6 CMV − , n = 5 CMV + and n = 1 CMV unknown children, n = 7 CMV − , n = 4 CMV + and n = 9 CMV unknown adults and n = 2 CMV − , n = 13 CMV + and n = 3 CMV unknown older adults in a and n = 3 CMV − , n = 4 CMV + and n = 3 CMV unknown newborns, n = 6 CMV − , n = 5 CMV + and n = 1 CMV unknown children, n = 7 CMV − , n = 4 CMV + and n = 9 CMV unknown adults and n = 1 CMV -, n = 12 CMV + and n = 3 CMV unknown older adults in b.Statistical analysis was performed between donors with a known positive or negative CMV status within each age group using a two-sided Mann-Whitney U-test.Exact significant p-values are indicated above the graphs.

8 Extended Data Fig. 3 | 8 Extended Data Fig. 6 |
Articlehttps://doi.org/10.1038/s41590-023-01633-Molecular and phenotypic differentiation within A2/M1 58 + CD8 + T cells across human lifespan.a, Specification selected donors for single-cell multi-omic analysis.Phenotype frequencies were obtained via flow cytometry protein expression data.b, Dot plot of selected transcription factors grouped by UMAP clusters.Dot size represents the proportion with non-zero expression from each age group.The color represents average mean expression.c, Heatmap of enriched pathways identified from GSEA using differentially expressed genes between each UMAP cluster.All pathways shown have Benjamini-Hochberg adjusted p-values < 0.05 in at least one cluster.NES: Normalized enrichment score.d, Dimensionality reduction (UMAP) and clustering of scRNAseq data excluding TCR genes colored by clusters (top), age groups (bottom).e, Heatmap of enriched pathways identified from GSEA using differentially expressed genes between each age group.All pathways shown have Benjamini-Hochberg adjusted p-values < 0.05 in at least one age group.NES: Normalized enrichment score.f, Dot plot of selected transcription factors grouped by age group.Dot size represents the proportion with non-zero expression from each age group (N, newborns; C, children; A, adults; OA, older Adults).The color represents average mean expression.g, Volcano plot of a pairwise comparison without correction for multiple testing of differentially expressed genes between children (blue) and older adults (pink) with a fold change |log2(FC)| >0.3 and p-value < 0.05.Article https://doi.org/10.1038/s41590-023-01633-Dividedand undivided A2/M1 58 + CD8 + TCRαβ repertoires across age groups.TRAV and TRBV clonotype pairing for pooled donors within each age group, newborns (a), children (b), adults (c) and older adults (d), illustrated by circos plots for fast, slow, undivided and total A2/ M1 58 + CD8 + T cells.Left arch segment color indicates TRAV usage, right outer arch color depicts TRBV usage.Connecting lines indicate TRAV-TRBV gene pairing and are colored based on their TRAV usage and segmented based on their CRD3α and CDR3β sequence.The thickness is proportional to the number of TCR clones with the respective pair.The number of sequences considered for each circos plot is shown at the right bottom.Extended Data Fig. 8 | Transient and stable expression of age-specific TCR.a, Gating strategy transient transfection age-specific TCRs in HEK293T cells.b, Median MFI A2/M1 58 tetramer-PE staining of transiently expressed TCRs in HEK293T cells, dotted line indicates MFI threshold set by the parental cell line expressing no TCR (n = 3 independent experiments, median and IQR).c, Gating strategy and A2/M1 58 tetramer-PE staining with a normal (A2/M1 58 -WT, lightly shaded), knockout CD8-binding site (A2/M1 58 -KO open) and enhanced CD8binding site (A2/M1 58 -Enh, closed) tetramer of SKW-3-CD3 + and SKW-3-CD3 + CD8 + TCR-expressing cell lines.d, Median MFI of A2/M1 58 tetramer-PE staining with a normal (A2/M1 58 -WT, lightly shaded), knockout CD8-binding site (A2/M1 58 -KO open) and enhanced CD8-binding site (A2/M1 58 -Enh, closed) tetramer of SKW-3-CD3 + TCR-expressing cell lines (n = 2 independent experiments), dotted line newborns, n = 12 children, n = 8 adults and n = 10 older adults, with clone size indicated by symbol size.Statistical analysis of P gen and for the number of insertions and deletions between age groups utilized a two-sided mixed-effects model with donor encoded as a random effect, as described in Methods.P values were adjusted (P adj ) for multiple testing with the Benjamini- . Experiments conformed to the Declaration of Helsinki Principles and the Australian National Health and Medical Research Council Code of Practice.The study was approved by the Human Research Ethics Committee of the UoM (ethics IDs 24567, 13344 and 23852), Australian Red Cross Lifeblood (ID 2015 8), St Jude Children's Research Hospital (XPD12-089 IIBANK), Mercy Hospital for Women (R14-25) and Tasmanian Health and Medical Human Research Ethics Committee (ID H0017479).