p16INK4a Regulates Cellular Senescence in PD-1-Expressing Human T Cells

T-cell dysfunction arising upon repeated antigen exposure prevents effective immunity and immunotherapy. Using various clinically and physiologically relevant systems, we show that a prominent feature of PD-1-expressing exhausted T cells is the development of cellular senescence features both in vivo and ex vivo. This is associated with p16INK4a expression and an impaired cell cycle G1 to S-phase transition in repeatedly stimulated T cells. We show that these T cells accumulate DNA damage and activate the p38MAPK signaling pathway, which preferentially leads to p16INK4a upregulation. However, in highly dysfunctional T cells, p38MAPK inhibition does not restore functionality despite attenuating senescence features. In contrast, p16INK4a targeting can improve T-cell functionality in exhausted CAR T cells. Collectively, this work provides insights into the development of T-cell dysfunction and identifies T-cell senescence as a potential target in immunotherapy.


INTRODUCTION
The capacity of conventional T cells to control or eradicate pathogens and transformed cells is compromised following repeated stimulations, owing to the development of T-cell dysfunction. The cornerstone of most current immunotherapies is to use the remarkable potential of T cells by either interfering with processes associated with T-cell dysfunction in vivo or manipulating T cells ex vivo. According to the prevailing model, T-cell differentiation is dictated by the strength and duration of Tcell stimulation (1). As such, repeated antigenic encounters in the setting of chronic infections or cancer, as well as the extensive culture required for the generation of sufficient cell numbers for treatment in certain adoptive cell therapy (ACT) settings, are conducive to the induction of T-cell terminal differentiation, exhaustion, senescence or apoptosis (2). Whether these cellular states are distinct, partially overlapping, co-regulated or part of a continuum remains incompletely understood. Moreover, several gradations within these categories exist as exemplified by variable responses of dysfunctional T cells following the blockade of the exhaustion-associated immune checkpoint PD-1 in cancer therapy (3,4).
Senescent immune cells accumulate with organismal aging (5). Cellular senescence can be triggered by various forms of cellular stress and is characterized by prolonged cell cycle arrest as well the acquisition of several features including morphological changes, chromatin remodeling, metabolic reprogramming, and secretion of multiple factors collectively called the senescence-associated secretory phenotype (SASP) (6). The most extensively studied pathways controlling cellular senescence triggered by DNA damage involve p53/p21 Cip1 and/ or p16 INK4a /retinoblastoma (Rb) regulators (7)(8)(9)(10). While the p53/p21 Cip1 pathway seems to be important at the initiation of the senescence process, p16 INK4a is, in addition, associated with the maintenance of the cell cycle arrest associated to senescence (11,12). Indeed, it has been suggested that the p53/p21 Cip1mediated cell cycle arrest is temporary in cases of low to moderate amounts of DNA damage (13)(14)(15). However, a prolonged arrest may lead to the upregulation of the cyclindependant kinase inhibitor p16 INK4a , which will activate the transcriptional regulator Rb, resulting in a permanent cell cycle arrest in G1 phase (16,17).
The p38 mitogen-activated protein kinase (MAPK) pathway is also actively linked to the establishment of cellular senescence (18)(19)(20)(21). Genotoxic stress leading to histone modifications and the recruitment of DNA damage sensors can trigger p38 phosphorylation and mediate its nuclear translocation (22)(23)(24). Then, activated p38 can arrest cell division by activating p53 or by directly enhancing the activity of molecules such as p16 INK4a to block cell cycling at the G2/M or G1/S checkpoints, respectively (20,25,26). A form of DNA damage-associated senescence has also been linked to p38MAPK signaling in both circulating T cells and antigen-experienced tumor infiltrating lymphocytes (TILs) (27,28). However, the mechanisms downstream of T-cell activation and p38MAPK activation leading to cellular senescence are still elusive. Here, we show that p16 INK4a expression associated with cellular senescence is a cardinal feature of PD-1-expressing T cells following repeated Tcell activation in multiple experimental and clinically relevant settings and as such, represents a target to reverse T-cell dysfunction in immunotherapy.

Study Approval
This study was approved by the local Hô pital Maisonneuve-Rosemont research ethics authorities and participants' informed consent was obtained (CÉR2020-2141, HQ2017-004 and CÉR13030). All animal protocols (2016-OC-024 and 2017-AV-010) were likewise approved by the local Animal Care Committee in accordance with the Canadian Council on Animal Care guidelines.

Mouse
C57BL/6 mice were purchased from The Jackson Laboratory. Mice were maintained in a specific pathogen-free environment at the Hô pital Maisonneuve-Rosemont Research Center. Female 6-12 week-old mice were infected with 2 x 10 5 FFU (Focus Forming Units) of LCMV Armstrong intraperitoneally (i.p.) (acute infection) or 2 x 10 6 FFU of LCMV clone 13 intravenously (i.v.) (chronic infection). The spleens were harvested 8 or 30 days post-infection for analysis.

T-Cell Line Generation
For polyclonal activation, human T cells were enriched from PBMCs using the EasySep ™ Human T Cell Enrichment Kit (StemCell) and pulsed at a 1:5 ratio (beads:cells) with anti-CD3/ CD28 magnetic beads (Gibco ™ Dynabeads ™ Human T-Activator). Cells were cultured in G-Rex bioreactors (Wilson Wolf Manufacturing) in T-cell medium (Advanced RPMI 1640, 10% human serum, 1X L-glutamine). Cell number and viability was evaluated with a Typan Blue exclusion assay and live cells concentration was adjusted to 0.5 x 10 6 cells/mL and restimulated weekly with beads.
Spleens from LCMV clone 13 infected mice were harvested at day 8 and 30 post-infection and mechanically dissociated. CD8 + CD44 + Tet-gp33 + cells were sorted and RNA was extracted with TRIzol reagent then further purified using RNeasy columns (Qiagen). Library preparation was done with the KAPA mRNAseq stranded library preparation kit (Roche, after mRNA capture with Dynabeads ® mRNA DIRECTTM Purification Kit). The data can be found in the Gene Expression Omnibus (GSE132989).

Gene-Set Enrichment Analysis of RNA-Seq Data
Human gene symbols were ranked by the fold changes of the gene expression as profiled by RNA-seq. Then, gene-set enrichment was analyzed using GSEA 3.0 software (http://software. broadinstitute.org/gsea/downloads.jsp) (32). GSEA enrichment table files were loaded in the Enrichment Map plugin from Cytoscape software v3.2.1 (33) and filtered for significance according to the p-value (0.001) and FDR thresholds (0.01). Overlap between significant gene-sets was computed according to the Jaccard+overlap combined coefficient.

Quantitative PCR
Cells were recovered after sorting on a FACS Aria III instrument (BD Biosciences). RNA extraction was performed by a two-step approach using Trizol (Invitrogen) and the RNeasy micro kit (Qiagen) according to the manufacturer's instructions. After DNAse treatment (Ambion, Life technologies), RNA was reverse transcribed with random primers using the High Capacity cDNA Reverse Transcription Kit (Life Technologies) as described by the manufacturer. Real time qPCR reactions were performed using TaqMan Advanced Fast Universal PCR Master Mix (Life Technologies). The Viia7 qPCR instrument (Life Technologies) was used to detect amplification levels. Relative expression (RQ = 2 -DDCT ) was calculated using the Expression Suite software (Life Technologies), and normalization was done using HPRT, GAPDH and ACTB.

Telomere Measurement
Telomeres were measured by quantitative PCR as described before (34). Briefly, genomic DNA was isolated with the DNeasy kit (Qiagen). For telomere amplification (T Primer pair), the sequences are (written 5′!3′): Relative telomere/single copy gene (T/S) ratios reflect relative length differences in telomeric DNA.
Weekly anti-CD3/CD28 stimulated T cells were transduced with retroviruses for shRNA delivery 24 hours after the third stimulation and subsequently stimulated as described above. For the transduction, cells were resuspended in viral supernatant supplemented with 10% fetal bovine serum, transferred in retronectin-coated plates (Takara) and spinoculated 90 minutes at 2250 rpm at 20°C.

Chimeric Antigen Receptor (CAR) T-Cell Generation
Second generation B-cell maturation antigen (BCMA)-specific CAR construct [on the CD28z-CAR backbone and expressing the NGFR marker as described in (39,40)] is a kind gift from Jonathan Bramson (McMaster University, Hamilton, Canada). Lentiviral transduction of T cells was performed 24 hours after stimulation with CD3/CD28-coated beads. Cells were restimulated at day 7 and 14 with beads. Cells were then transduced with shRNA-containing retroviruses as described above and co-cultured weekly with human BCMA-expressing KMS-11 cells (also a kind gift from J. Bramson) until day 35.

Statistical Analysis
Statistical analyses were performed with R statistical language or GraphPad Prism v8.3.0. Statistical details of experiments, including number of independent donors, statistical test (depending if normal distribution was assumed according to Shapiro-Wilk testing) and statistical significance (p-value) are reported in the figure legends. Samples were paired when indicated. For gene expression, statistics were calculated with DESeq2 and adjusted p-values were used. P-values of less than 0.05 were considered significant.

Repeated Ex Vivo Stimulations Recapitulate Phenotypic and Functional T-Cell Exhaustion
To define the molecular events leading to T-cell dysfunction following repeated stimulations, we exposed primary human T cells to anti-CD3/CD28 antibody-coated beads weekly in culture ( Figure 1A). Despite substantial donor-dependent variation in the magnitude of total T-cell proliferation, we consistently observed a strong initial expansion phase, followed by a cessation of T-cell growth typically after 5 to 7 rounds of stimulation ( Figure 1B). T cells were characterized after a single stimulation (day 7) and compared with cells harvested after two consecutive weeks of halted proliferation, occurring between 35-49 days. As anticipated, serial stimulations promoted the accumulation of more differentiated effector memory (TEM) and effector (TEFF) phenotype T cells relative to central memory (TCM) phenotype T cells in both CD4 + and CD8 + subtypes ( Figure 1C and Supplementary Figure 1). We next focused on CD8 + T cells to compare our findings to a large body of previous data on CD8 + T-cell differentiation and exhaustion. We noted a marked decrease in CCR7 expression as well as co-expression of multiple inhibitory receptors as the cultures progressed over time ( Figures 1D-F). In contrast, costimulatory molecules CD27 and CD28 expression decreased as expected for dysfunctional T cells (41) (Figure 1G). In addition, the percentage of cytokineproducing T cells declined between the week prior to culture termination (day 28) and the end of the culture (day 35) ( Figure 1H). Thus, repeated ex vivo CD3/CD28 stimulations induce classical T-cell differentiation and exhaustion features.

Repeated Stimulations Program an Exhaustion Transcriptome
We then proceeded to transcriptional profiling after a single stimulation (day 7) and repeatedly stimulated dysfunctional CD8 + T cells (day 35). Principal component analysis showed a clear segregation between time points and revealed a total of 476 gene transcripts differentially regulated between single and repeated stimulations (Figures 2A, B). By focusing on T-cell activation genes (GO:0042110), we readily identified the downregulation of TCM-associated (e.g. CCR7 and SELL) and the upregulation of exhaustion-associated (e.g. EOMES and HAVCR2) mRNAs in repeatedly stimulated T cells relative to day 7 cells (Supplementary Figure 2). Integrated networks of gene set enrichment analysis pointed to several cell cycle-related pathways in dysfunctional T cells, but also the upregulation of genes in the T-cell receptor and death receptor signaling pathways ( Figure 2C). The TOX mRNA expression, which encodes the master transcription factor regulating T-cell exhaustion fates, was slightly higher in repeatedly stimulated CD8 + T cells relative to day 7 T cells, which led to a significant increase of the three isoforms of its target transcription factors NR4A (42, 43) ( Figures 2D, E). Along the same lines, dysfunctional CD8 + T cells lost the expression of TCF7 mRNA (encoding TCF1), suggesting a state compatible with terminal exhaustion (42, 44) ( Figure 2F). Collectively, these results confirm the development of T-cell exhaustion in repeatedly stimulated T cells ex vivo.

Repeatedly Stimulated CD8 + T Cells Display a Strong Senescence-Associated Gene Signature
Given that serial stimulations efficiently generated late-stage exhausted CD8 + T cells, we sought to identify the driver mechanisms responsible for their arrested expansion. Our transcriptome analysis revealed that prominent characteristics of repeatedly stimulated T cells were associated with cellular senescence such as the downregulation of genes involved in DNA repair processes (45), cell cycle and E2F target genes (37), as well as upregulated senescence-associated transcripts (46,47) (Figures 2C, 3A and Supplementary Figure 2C).
More specifically, transcript levels of p16 INK4a and its related gene p15 INK4b , central mediators of cellular senescence, were upregulated ( Figure 3B). Due to the critical role of the p53 tumor suppressor pathway in cellular senescence, we also sought for evidence of contribution of this pathway upon repeated stimulations. Despite a decrease of TP53 transcripts and the absence of induction of its classical downstream gene target CDKN1A in repeatedly stimulated CD8 + T cells, we observed the up-regulation of numerous p53 target genes, such as BTG2, YPEL3, GADD45A/G, and PLK2, which can contribute to cellular senescence (48) ( Figure 3C). Other transcriptional features of cellular senescence were also found such as increased CAV1 and the loss of LMNB1 mRNAs (49,50), coding for major structural proteins of the plasma membrane and nuclear lamina, respectively ( Figure 3D). In addition, dysfunctional CD8 + T cells displayed elevated levels of transcripts related to the senescence-associated secretory phenotype (SASP) and a decreased expression of several genes required for DNA repair processes, which are features of cell senescence (45,47,51) (Figures 3E, F). Furthermore, when analyzing a previously published dataset (31) of circulating T cells from acute myeloid leukemia (AML) patients, a condition known to be associated with immune exhaustion (31,52,53), we noted high expression of transcripts encoding PD-1 along with p16 INK4a , p21 Cip1 , and senescence-associated b-galactosidase (SA-b-Gal) (54) in T cells from patients relative to normal donors (Supplementary Figure 3). This adds to our findings inferring that exhaustion-and senescence-associated transcripts are co-regulated in dysfunctional T cells.

Cellular Senescence Features Are Restricted to PD-1-Expressing T Cells
Consistent with growth arrest, the proportion of CD8 + T cells expressing the proliferation marker Ki67 was lower following repeated stimulations ( Figure 4A). We next evaluated the telomere repeat copy number to single copy gene (T/S) ratio and confirmed that serially stimulated CD8 + T cells do not have critically short telomeres, hinting at a process independent of their replicative potential ( Figure 4B). However, repeatedly stimulated CD8 + T cells accumulated histone H2AX phosphorylation indicative of DNA damage, another classical feature of cell senescence (55-57) ( Figure 4C). We then interrogated whether exhaustion and cellular senescence were developing in the same T cells. Proliferation arrest was predominant among CD8 + PD1 + T cells, which were found to be arrested in the G0/G1 phases of the cell cycle in greater proportion at the end of the culture than after a single stimulation ( Figure 4D). We then compared PD-1expressing and PD-1-negative CD8 + T cells. Remarkably, repeated stimulations led to an increase in SA-b-Gal activity following the third week (day 21), exclusively in PD-1 + cells ( Figures 4E, F). Consistent with the data obtained with anti-CD3/CD28 coated beads, repeated stimulations with the HLA-A0201-restricted Epstein-Barr virus (EBV) epitope LMP2 426-434 also led to a robust induction of SA-b-Gal in LMP2-specific PD-1-expressing T cells, confirming that the co-occurrence of cell senescence features and PD-1 expression is observed across several models (Supplementary Figure 4). When focussing on gene expression between repeatedly stimulated PD-1-expressing and PD-1negative CD8 + T cells, both CDKN2A (p16 INK4a ) and CDKN1A (p21 Cip1 ) were upregulated in the PD-1 + fraction, without any change in TP53 (p53) expression ( Figure 4G). Moreover, cyclindependent kinase 6 (CDK6) and E2F3 transcription factor were negatively impacted in PD-1-expressing cells. These latter are both implicated in the progression into the S-phase of the cell cycle and mainly controlled by p16 INK4a (58,59). Other senescenceassociated gene transcripts, such as SERPINE1 and PML, were also increased in PD-1 + T cells (48, 60-62) ( Figure 4G). Altogether, these results show that transcriptomic, phenotypic and functional features of cellular senescence are restricted to repeatedly stimulated PD-1-expressing exhausted CD8 + T cells.

Exhausted Antigen-Specific T Cells Exhibit Senescence-Associated Features In Vivo
Long term culture can impose a selection pressure on ex vivo expanded T cells and bias their differentiation state (63). We thus used the well-described mouse T-cell exhaustion model of lymphocytic choriomeningitis virus (LCMV) infection to determine whether our findings extended to this in vivo model. We first compared virus-specific CD8 + T cells in early and late chronic LCMV clone 13 (cl13) infection and noted that exhausted GP33-specific CD8 + T cells share a similar senescence-associated mRNA expression pattern when compared to dysfunctional human CD3/CD28 stimulated CD8 + T cells (Figures 5A, B). Indeed, CDKN2A and CDKN2B along with SERPINE1 were likewise upregulated, while cyclins (e.g. CCNE1, CCNB1 and CCNA1) were downregulated ( Figure 5B). In addition, these exhausted T cells showed a downregulation of gene transcripts required for DNA damage repair processes as seen for human dysfunctional T cells ( Figure 5C). Moreover, chronic infection promoted the accumulation of a higher proportion of GP33-specific cells expressing PD-1 and increased SA-b-Gal activity as compared to an acute infection ( Figures 5D-F), confirming that CD8 + Tcell dysfunction is associated with features of senescence in exhausted PD-1 + cells in vivo.

Inhibition of p38MAPK Limits DNA Damage and T-Cell Senescence
The p38MAPK pathway is closely related to cellular senescence development, and is active in terminally differentiated T cells (27,28). Likewise, repeatedly stimulated CD8 + T cells ex vivo expressed a transcriptomic signature linked to the p38MAPK signaling cascade ( Figure 6A). Given that short term p38MAPK inhibition with 500 nM of the p38MAPK inhibitor BIRB796 ex vivo has been previously associated with improved T-cell function (27,28), we sought to evaluate whether it may also reverse established cellular senescence features in serially activated dysfunctional T cells. To this end, we added BIRB796 (hereafter referred to as p38i) to the cell cultures following the third CD3/CD28 stimulation ( Figure 6B), corresponding to the timing of increased SA-b-Gal activity detection in PD-1 + cells ( Figure 4F). While p38i did not impact overall T-cell recovery by the end of the culture, nor the percentage of PD-1-expressing CD8 + T cells, it led to a greater accumulation of Ki67 + cells (Figures 6B-D). When focussing specifically on PD-1-expressing CD8 + T cells, we noted that p38i reduced the expression of CDKN2A (p16 INK4a ), TP53 and PML ( Figure 6E). It also correlated with a decreased percentage of PD-1-expressing cells displaying DNA damage and with SA-b-Gal activity ( Figures 6F, G). It is noteworthy to point out that despite the mitigation of senescence features, this specific dose of p38 inhibitor did not improve, but rather decreased the IFNg-secreting potential of T cells ( Figure 6H).
Thus, the p38MAPK signaling pathway is implicated in T-cell dysfunction and p16 INK4a upregulation. However, despite the positive effects of p38i on certain senescence-associated features, the pharmacological blockade of p38MAPK after several rounds of stimulation failed to restore T-cell functionality.

Inhibition of p16 INK4a Expression Reinvigorates Dysfunctional T Cells
Since pharmacologic p38i can downregulate both p16 INK4a and p53 transcripts in PD-1-expressing exhausted CD8 + T cells, we sought to gain more insights into which pathway downstream of p38MAPK is responsible for conferring senescence features to dysfunctional T cells. We thus performed knockdown of p16 INK4a and p53 with validated shRNA vectors (37,38). Knockdown performed following the third CD3/CD28 stimulation did not significantly affect global expansion nor PD-1 expression (as for p38i) in repeatedly stimulated cultures ( Figures 7A-C). However, p16 INK4a , but not p53 targeting, successfully limited the development of CD8 + T cells with increased SA-b-Gal activity ( Figure 7D). We next used a clinically relevant approach to investigate cell fate in context of initial CD3/CD28 stimulation followed by serial antigenic encounters. We generated B-cell maturation antigen (BCMA)specific CAR T cells (40) that were activated and expanded for three weeks with anti-CD3/CD28 beads and subsequently transduced with p16 INK4a -targeting or non-targeting shRNAs ( Figures 7E, F). While exposure to repetitive stimulations with BCMA-expressing human KMS-11 multiple myeloma cells (64) led to the expression of inhibitory receptors such as PD-1 and KLRG1 (Supplementary Figure 5), p16 INK4a knockdown increased the fraction of Ki67 + cells and restricted the development of SA-b-Gal-expressing CAR T cells ( Figures 7G, H). In contrast with data obtained with p38i, directly modulating p16 INK4a expression increased the proportion of cytokine-secreting T cells in this setting ( Figure 7I). Thus, our results in ex vivo repeatedly beadstimulated T cells dovetail with findings in several other experimental systems and establish p16 INK4a as a key mediator of activation-induced T-cell senescence and dysfunctionality.

DISCUSSION
The loss of T-cell function following repeated stimulations is a major limitation for the control of chronic viral infections and cancer. Modern cancer immunotherapy hinges on the prevention or reversal of T-cell dysfunction, either through immune checkpoint blockade or adoptive cell therapy. In the latter situation, ex vivo manipulations prior to T-cell infusion, as well as repeated stimulations after adoptive transfer in vivo, often induce T-cell dysfunction, which severely impedes the therapeutic potential of this approach. However, the processes articulating the development and maintenance of T-cell dysfunction remain elusive (65). Among T-cell dysfunctional states, exhaustion occurs when the native T-cell or synthetic (i.e. CAR) receptor is strongly and repeatedly engaged, in presence of various inflammatory cytokines. Exhausted T cells then progressively show diminished proliferative and functional capacities while expressing high levels of immune checkpoints such as PD-1 (66,67). In line with this, we found that repeatedly stimulated T cells ex vivo displayed a gradual accumulation of CD8 + T cells co-expressing multiple inhibitory receptors along with a reduced cytokine-secretion capacity and a marked growth arrest. In addition, our transcriptional profile showed the co-occurrence of a T-cell exhaustion and senescence program among repeatedly stimulated CD8 + T cells. We further established the strong relationship between PD-1 expression and senescence features in individual T cells in several experimental systems in human and mouse. Whether such co-occurrence depends on a co-regulation or distinct mechanisms acting in parallel remains unclear. However, our data argue against the notion that exhaustion and senescence are mutually exclusive pathophysiologically.
We extend these findings showing that growth arrest in ex vivo stimulated T cells is not related to abnormal telomere length, but rather associated with activation-induced genotoxic stress and DNA damage. DNA double strand breaks can initiate p38MAPK signaling cascade. When phosphorylated, p38 mediates the activation of several transcription factors and inflammatory cytokines (68). It has also been reported that p38MAPK can induce p16 INK4a transcription as well as activate p53 following direct phosphorylation on Ser33 to mediate cell cycle arrest (12,13,69). The inhibition of p38MAPK in the context of a brief ex vivo restimulation of terminally differentiated and gH2AX-expressing T cells was shown to reverse senescence features (27,28). Given the pleiotropic effects of p38MAPK signaling, it was unclear whether p38i would reverse T-cell senescence features and promote functionality in the context of ongoing stimulations once senescence features become conspicuous. In repeatedly stimulated T cells, inhibiting p38MAPK signaling after the third T-cell stimulation could limit the development of senescence features by reducing the expression of p16 INK4a , limiting DNA damage and SA-b-Gal activity in PD-1expressing cells. Although it also reduced the expression of p53, p38i treatment promoted the increase of p21 Cip1 , suggesting that an alternative senescence mechanism from the p53/p21 Cip1 axis may still take place. Consistent with previous reports showing a correlation between p38MAPK inhibition and decreased IFNg expression (70), exposing T cells to p38i during repeated TCR stimulations failed to restore T-cell functionality in terms of pro-inflammatory cytokine secretion despite improvements in senescence-associated features. Although we cannot exclude the possibility that modulating the dosage or method of p38MAPK inhibition may result in different outcomes, p38i in the context of serially activated T cells could alleviate DNA damage and senescence features but this did not result in functional improvements, suggesting that timing of p38MAPK modulation in the context of immunotherapy must be carefully planned.
Our data indicate that T-cell cellular senescence in the context of chronic activation is maintained by p16 INK4a regulation as CDKN2A silencing after the third stimulation could attenuate senescence features. We further show that CAR T cells serially stimulated with cancer cells bearing the targeted antigens develop dysfunctionality in terms of dual expression of exhaustion and senescence features, which can also reflect the fate of adoptively transferred cells. Whether tonic signaling associated with CAR T-cell exhaustion (71,72) also results in direct induction of senescence features is currently unknown but may represent a factor impeding CAR T-cell efficacy. Similar to our CD3/CD28 stimulation model, the targeting of p16 INK4a downstream of p38MAPK likewise reversed senescence features, yet it could also improve cytokine secretion capacity among serially activated CAR T cells. This enhancement in cytokine secretion was not accompanied by decreased PD-1 expression or an increased expansion, which suggest the contribution of other growth regulatory mechanisms. Despite p16 INK4a being predominantly activated, we cannot rule out the potential contribution of the p53/p21 Cip1 pathway. Indeed, we observed elevated mRNA levels of p21 Cip1 in PD-1-expressing dysfunctional CD8 + T cells, arguing in favor of its potential role in early senescence initiation (11,12,73). This may explain why certain senescence features are not reversed following p16 INK4a or p38MAPK targeting. This is reassuring as interfering with p16 INK4a expression in senescent cells raises the concern of inducing malignant transformation. It was previously shown that p16 INK4a -deficient mouse T cells do not undergo cancerization (74), perhaps through the induction of such compensatory mechanisms. Finally, in the context where senescent cells can exhibit signs of resistance to checkpoint blockade (75,76), p16 INK4a may be considered as an actionable target to alleviate senescence features in repeatedly stimulated PD-1-expressing T cells while also preserving cytokine-secretion capacity for therapeutic purposes.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi. nlm.nih.gov/geo/, GSE132727; https://www.ncbi.nlm.nih.gov/ geo/, GSE132989.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Hô pital Maisonneuve-Rosemont research ethics authorities. The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by Hopital Maisonneuve-Rosemont Animal Care Committee in accordance with the Canadian Council on Animal Care guidelines.

AUTHOR CONTRIBUTIONS
VJ and J-SD designed the study. VJ performed experiments and analyzed the data. MN did the shRNA constructions. M-ÈL performed LCMV mouse infections. DDS and SB did the mouse RNA-Seq. LD did the EBV-specific T-cell cultures. CM provided technical help. CC and SL helped with conceptualization. VJ and J-SD wrote the manuscript. NL, HM, FM, and J-SD supervised the study. All authors contributed to the article and approved the submitted version. expert flow cytometry and telomere assay advice, respectively, Jonathan L. Bramson and Claude Perreault for helpful discussions, Héma-Québec for leukoreduction system chambers, the genomic platform of the Institute for Research in Immunology and Cancer (IRIC) for RNA sequencing and Raphaëlle Lambert for qPCR, and Patrick Gendron for advice regarding the bioinformatics analysis. J-SD and VJ hold senior clinician researcher career award and post-doctoral fellowship from the Fonds de Recherche du Quebec -Santé(FRQS), respectively. MN is the recipient of post-doctoral fellowships from the FRQS and the Cole Foundation. FM holds the Canada Research Chair in Epigenetics of Aging and Cancer. J-SD, FM, and HM are Cole Foundation Early Career Transition Award laureates. J-SD is a member of the TheĆell network and of the Canadian Donation and Transplant Research Program.