{"paper_id":"162be388-67f7-4d20-a93a-75bcc0f9e05b","body_text":"An early mTOR-dependent window during human T cell activation programs T \ncell state \n \n \nM Valeria Lattanzio 1,2,*, Anouk P Jurgens 1,2,*, Arie J Hoogendijk 3, Carmen van der Zwaan 3, \nLeyma Wardak1,2, Kaspar Bresser1,2,†  & Monika C Wolkers1,2,†  \n \n1 Sanquin Blood Supply Foundation, Department of Research, and Amsterdam UMC, Landsteiner \nLaboratory, and Amsterdam Institute for Infection & Immunity, Cancer Center Amsterdam, Cancer \nImmunology, Amsterdam, The Netherlands  \n \n2 Oncode Institute, Utrecht, The Netherlands \n \n3 Sanquin Blood Supply Foundation, Department of Research, Mass Spectrometry Laboratory, \nAmsterdam, The Netherlands, (and) Amsterdam UMC location University of Amsterdam, Landsteiner \nLaboratory; Amsterdam, The Netherlands \n \n*  These authors contributed equally \n† Correspondence; m.wolkers@sanquin.nl, k.bresser@sanquin.nl.  \n \nABSTRACT \nT cell activation results in profound proteome remodeling that programs T cells into distinct \ncellular states. The mechanistic target of rapamycin (mTOR) biases T cell differentiation toward \na cytotoxic fate at the expense of memory-precursor formation, making mTOR inhibition an \nattractive strategy to boost T cell memory during vaccination. Here, we used matched time-\nresolved mRNA sequencing and quantitative mass spectrometry to define how the human T cell \nproteome is remodeled during the first 24 hours of activation. We found that human T cells \nrapidly remodel their proteome in distinct, temporally ordered modules that drive translation and \nproliferation while promoting a cytotoxic T cell state. Notably, mTOR inhibition during the first 24 \nhours of T cell activation perturbed these protein modules. Strikingly, transient mTOR inhibition \nlimited to the first 16 hours of T cell priming was sufficient to imprint a memory-like T cell state, \nwhile preserving the capacity to produce inflammatory cytokines and mediate target cell killing. \nTogether, these findings indicate that mTOR activity dictates stable functional trajectories during \nearly T cell activation, revealing a therapeutic window to refine vaccination responses. \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\nINTRODUCTION  \nT cells are critical players in adaptive immune responses against pathogens and malignancies. \nTo become proficient effector cells, T cells first undergo rapid and profound changes, both \nduring naïve T cell priming and during activation of differentiated T cell subsets.  Specifically, T \ncell receptor (TCR) engagement triggers a cascade of signaling events that promote cell-cycle \nprogression, induce extensive metabolic rewiring, and ultimately lead to lineage differentiation \nand acquisition of effector functions\n1–3.  \n \nTo meet the demands of T cell expansion and differentiation, protein production is substantially \nincreased\n4–6 and the composition of the proteome is altered in a highly dynamic fashion 7–9. A \nstraightforward mechanism to facilitate this enhanced protein production is to increase the \navailability of mRNA templates for translation. However, recent evidence indicates that this is \nnot the primary driver for protein production. For example, activation of naïve T cells results in a \n~1.4-fold increase in total mRNA abundance but a ~5-fold increase in total protein abundance 7. \nFurthermore, the reported correlation coefficient between protein and mRNA abundance in T \ncells is within the range of r = 0.41-0.65\n7,10,11. These findings highlight that, in addition to \ntranscriptional regulation, post-transcriptional events substantial contribute to proteome upon T \ncell priming and activation\n12–15. Despite these efforts, the precise temporal coordination of gene \nand protein expression during the early events of human T cell activation remains incompletely \ndefined.  \n \nThe kinase mechanistic target of rapamycin (mTOR) is a central regulator of T cell activation. \nEngagement of the TCR and co-stimulatory molecules trigger mTOR signaling 16,17 leading to \nmetabolic reprogramming 18,19, proliferation 20, differentiation 21–23, survival and protein \ntranslation24,25.  mTOR is the core component of two complexes, mTORC1 and mTORC2, which \nexert different functions 26,27: Whereas mTORC1 enhances cap-dependent translation through \nphosphorylation of 4E-BP and LARP1 28,29, mTORC2 regulates cellular survival and \nproliferation30,31. Coordinated mTORC1 and mTORC2 signaling promotes T cell differentiation, \ndriving naïve CD8 + T cells toward short-lived effector cells 22, while directing CD4 + T cells into \ndistinct helper T cell subsets 17,23. Because mTOR inhibition promotes the generation of \npredominantly memory-precursor T cells during immune responses, it has been proposed to \ninclude mTOR inhibition in vaccination strategies to enhance memory formation 21,32. However, \nthe precise timeline during which mTOR activity instructs cell state during T cell activation \nremains incompletely understood.  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \nIn this study, we measured the core transcriptional and proteomic changes in human effector T \ncells during the first 24 hours of TCR engagement. Time-resolved mRNA sequencing and mass \nspectrometry analysis revealed that proteome remodeling during T cell activation occurs in \nsequential phases, with tightly instructed timelines for specific gene classes. mTOR signaling \nspecifically instructs the programs for proliferation, translation, and effector function. Importantly, \ntransient inhibition of mTOR during T cell priming was sufficient to induce a memory-precursor-\nlike T cell phenotype while preserving the T cell effector function. Collectively, our findings \nprovide insights in how mTOR modulates the proteome in activated T cells, which should help \nguide the rational design of future vaccination strategies.  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\nRESULTS \n \nProteome remodeling during T cell activation occurs in a staged manner  \nWe first studied the kinetics of mRNA and protein expression during the first 24 hours of human \neffector T cell (Teff) activation. Teff cells were generated by activating peripheral blood-derived \nCD3+ T cells with anti-CD3/CD28 antibodies for three days, followed by a 9-day culture period \nin the absence of TCR triggering (see Methods). Next, Teff cells were re-activated with anti-\nCD3/CD28 antibodies for 2, 4, 6, or 24 hours, and subjected to mRNA-sequencing and data-\nindependent acquisition (DIA) LC-MS/MS ( Figure 1a, Supplementary Figure 1, \nSupplementary Figure 2a-b ). As T cell activation results in increased cellular abundance of \nboth mRNA and protein\n7,8, all samples obtained across the different time points were \nnormalized, enabling comparison of relative mRNA and protein abundances. \n \nWe first asked how relative protein abundances changed during the first 24 hours of activation. \nTo visualize this, we clustered all differentially expressed proteins (adjusted P value < 0.05; \n2,375 out of 6,750 detected proteins) into 5 expression modules.  Each module captured a \ndistinct kinetic pattern (Figure 1b, Supplementary Table 1), including proteins that were rapidly \ninduced (early up), proteins that progressively increased (gradual up), proteins induced only at \n24 hours (late up), proteins that decreased progressively (gradual down), and proteins \ndecreased only at 24 hours (late down). Intriguingly, these 5 protein modules were enriched for \ndistinct functional groups of proteins ( Figure 1c-d): (1) early up included cell cycle related \nproteins (e.g., MYC and CDK2) and key regulators of effector function JUN and TBX21; (2) \ngradual up was enriched for ribosomal RNA processing (e.g., NOP56 and UTP20) and proteins \ninvolved in ATP synthesis (e.g., MTX1 and ATP5PF); (3) late up contained proteins involved in \nrRNA processing and epigenetic regulation (e.g., EED and SUZ12); (4) gradual down was \nmarked by components involved in immune-regulation and TCR-signaling components (e.g., \nLEF1 and LAT); and (5) late down was enriched for proteins involved in cytokine signaling (e.g., \nMIF and S100B) and programmed cell death (e.g., BCL2 and DFFA). \nIn summary, and consistent with previous reports\n8,13, T cell activation induces proteome \nremodeling in a highly staged process, with cell cycle components—and master regulator \nMYC—induced within 2 hours, followed by increased expression of the translation machinery at \n6 hours, and epigenetic and transcriptional programs within 24 hours. \n \nTranscriptome and proteome remodeling follow distinct kinetics during activation \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\nThe abundance and availability of mRNA templates are key determinants for protein production. \nThese determinants are defined by transcriptional and post-transcriptional events 7–9. Therefore, \nwe performed a correlation analysis to assess how transcriptome and proteome kinetics relate \nduring the first 24 hours of T cell activation. Interestingly, changes in mRNA and protein \nabundance correlated only mildly at early timepoints (0-6h, spearman r ≈  0.27) but gradually \nbecame more aligned at later timepoints (reaching spearman r = 0.47 at 24h; Figure 1e). This \nincreased alignment could, in part, be explained by delayed protein expression, as early \nchanges in mRNA abundance showed a higher correlation with protein abundance changes at \nlater timepoints ( Figure 1f-g). In line with this observation, the gene set ‘ribosomal RNA \nprocessing’, which was enriched in the ‘gradual up’ and ‘late up’ protein expression modules, \nalready peaked at 4 hours post-activation at the mRNA level (Supplementary Figure 2c-d).  \n \nTo better understand the relationship between mRNA and protein abundance, we calculated \nSpearman’s rank correlation coefficients (\nρ ) between mRNA and protein levels for genes within \neach of the five protein expression modules (Figure 1b), classifying genes as positively \nassociated (\nρ  > 0.5), inversely associated ( ρ  < −0.5), or not associated (remaining genes; \nFigure 1h). In addition, a time-lagged Spearman correlation (i.e., correlating mRNA abundance \nat earlier timepoints with protein abundance at later timepoints; see Methods ) was used to \naccount for delayed protein expression ( Figure 1h-I, Supplementary Table 3 ). The majority of \nthe positively associated genes belonged to the ‘early up’ protein module ( Figure 1j) and were \nstrongly enriched for several pathways involved in the cell cycle ( Supplementary Figure 3 ), \nindicating that a tight relationship between transcription and translation contributes to these \nearly kinetics. Genes with a delayed positive association primarily belonged to protein modules \nwith gradual kinetics (e.g., the translation regulators BSM1 and BOP1; Figure 1h-i). Almost all \ngenes that were negatively associated belonged to the ‘late down’ protein module (e.g., cytokine \nsignal transducers CRKL and JAK1 ; Figure 1h-i ). Interestingly, the largest group of genes \n(1,147 out of 2,375) did not display an obvious association between RNA and protein \nabundance (e.g. metabolic enzymes GAPDH and GPI ; Figure 1j ). In conclusion, the relative \nRNA and protein abundances follow distinct kinetics upon TCR triggering, pointing to the pivotal \nrole of post-transcriptional mechanisms during these early stages of T cell state wiring.  \n \nmTOR instructs proteome remodeling during early T cell activation \nmTOR is involved in several (post-)translational regulatory mechanisms, including the promotion \nof translation\n11,33, and mTOR has been shown to boost translation during early T cell \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\nactivation7,34. Therefore, we asked if mTOR activity is involved in the kinetics of the identified \nprotein expression modules. To this end, Teff cells (matched donor pools to Figure 1) were re-\nactivated with anti-CD3/CD28 antibodies for 0, 6, and 24 hours in the presence of the \nmTORC1/2 inhibitor Torin-1 or DMSO alone. Non-stimulated T cells (0h) were incubated with \nthe inhibitor or DMSO alone for 6 hours ( Figure 2a, Supplementary Figure 4a-e ). Torin-1 \ntreatment during T cell activation efficiently suppressed phosphorylation of the mTOR targets \n4E-BP (mTORC1) and AKT (mTORC2; Supplementary Figure 4f-g), confirming that mTOR \nactivity was perturbed.  \n \nWe first studied which of the protein expression modules (as defined in Figure 1b) were \nspecifically sensitive to mTOR blockade. Because the relationship between protein abundance \nand time may not be well described by simple statistical models, we applied a tree-based \nmachine learning approach. In short, we trained an XGBoost model\n35 on the control dataset \n(Figure 1) to learn the characteristic kinetic protein expression profile of each of the 5 defined \nmodules and used it to classify the module identity for each protein in the Torin-1/DMSO-treated \nsamples (Figure 2b, Supplementary Figure 5a ). This analysis revealed that Torin-1 treatment \nstrongly affected 4 out of 5 protein modules ( Figure 2c, Supplementary Figure 5b), indicating \nthat mTOR signaling broadly regulates protein expression kinetics during the first 24 hours of T \ncell activation. In line with that finding, Torin-1 treatment globally reduced protein translation in \nactivated T cells, as measured by the puromycin incorporation assay\n36 (Figure 2d-e). Moreover, \nLC-MS analysis revealed that the abundance of proteins related to translation initiation was \nstrongly reduced in Torin-1 treated Teff cells ( Figure 2f-g). Despite this general perturbation of \ntranslation, Torin-1 treatment did not block T cell activation per se. In fact, the abundance of \nproteins that are induced downstream of TCR signaling, including BCL2L1, CDK4, IL2RA, \nJUNB, and TNFRSF4, was upregulated normally (Figure 2h-i). Thus, although mTOR regulates \nearly proteome remodeling across all the protein expression modules, its regulation is \nconcentrated on a distinct set of proteins.  \n \nmTOR inhibition disrupts distinct pathways  \nWe next investigated if specific pathways were affected by mTOR inhibition, by clustering the \nTorin-1 sensitive proteins across all timepoints ( Figure 3a, Supplementary Table 4 ) and \nassessing these clusters for enriched pathways ( Figure 3b-c ). Torin-1 treatment strongly \nrepressed the induction of proteins involved in protein translation (RPS15, EEF1G; cluster C1) \nand to a lesser extent cell cycle (CDK4, CCNB1; cluster C2), demonstrating that these known \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\ndownstream effects of mTOR inhibition can already be observed during the first 24 hours of \nTCR triggering. Torin-1 treatment also blocked the repression of a large group of proteins \n(cluster C3). These proteins are associated with several pathways, including components of \ncytokine signaling (IFIT5, JAK1), neutrophil degranulation (CSTB, ADAM10), and transcriptional \nregulation (TCF1, LEF1; Figure 3a-c ). Finally, Torin-1 treatment induced the expression of \nseveral proteins of which the expression was stable in the DMSO control during the first 24 \nhours of activation (cluster C4, Figure 3a). No pathway enrichment was detected in cluster C4. \nNevertheless, this cluster included key regulators of T cell function, including the transcription \nfactors ARID5A and GATA3, the catalytic subunit of PI3K (PIK3CG), activator of the NF-\nκ B \npathway IKK-β  (IKBKB), and cytokine signaling modulator CBLB (Figure 3c).  \n \nNext, we assessed how mTOR instructs these Torin-1 sensitive protein clusters during T cell \nactivation. Ribosome biogenesis and protein translation pathways were strongly repressed \nduring mTOR inhibition (cluster C1, Figure 3b). Furthermore, mTOR is known to promote the \ntranslation of transcripts that contain 5’UTR TOP-motifs, which is mediated through \nphosphorylation of the translation repressors 4E-BP and LARP\n25,29. Consistent with our \nobservation that Torin-1 treatment effectively reduces p4E-BP ( Supplementary Figure 5 ), the \nprotein output of TOP-motif transcripts was perturbed ( Figure 3d). Moreover, TOP-motif \ntranscripts were strongly enriched in cluster C1 ( Figure 3e ), highlighting this pathway as a \nmechanism driving the Torin-1 induced repression of proteins in cluster C1.  \n \nPerturbation of mTOR signaling also affected the abundance of proteins involved in proliferation \n(C2, Figure 3b). Consequently, cell cycle progression was affected in activated T cells treated \nwith Torin-1, as evidenced by the reduced expression levels of Ki67 ( Figure 3f-g) and a \ndecreased percentage of activated T cells progressing through the S-phase ( Figure 3h ). In \ncancer cells, mTORC1-mediated phosphorylation stabilizes SKP2, an E3 ubiquitin ligase that \npromotes proliferation by degrading the cell cycle inhibitor p27\n37,38. Consistent with this \nmechanism, SKP2 protein abundance increased upon activation of human T cells, coinciding \nwith reduced p27 abundance ( Figure 3i ). Importantly,\n this alteration in the abundance ratio \nbetween SKP2 and p27 was reduced upon Torin-1 treatment ( Figure 3i-j). These data suggest \nthat mTOR signaling contributes to the proliferation program in activated Teff cells, at least in \npart, by regulating the abundance of SKP2 protein. \n \nmTOR instructs early activation-induced T cell differentiation \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\nUpon activation, T cells differentiate into highly active cytotoxic or long-lived memory-precursor \ncells1,3,39. This process is coordinated by the transcription factors TCF1 \n(quiescence/multipotency) and T-bet (effector function). Long-term treatment with rapamycin (>7 \ndays) or T-cell-specific ablation of mTOR complex components in murine infection skews T cells \ntoward the TCF1\nHIGH memory-precursor cell state 21,22,40. In human Teff, the abundance of \nproteins involved in T cell differentiation was altered upon Torin-1 treatment (C3; Figure 3a-b). \nIntriguingly, flow cytometric analysis showed that a 24-hour treatment with Torin-1 during re-\nactivation of Teff cells was sufficient to retain high levels of TCF1 ( Figure 4a-b). Concordantly, \nTorin-1 treatment substantially blocked the induction of T-bet, a major driver of the effector T cell \nprogram ( Figure 4c-d). In addition, gene-set enrichment analysis of the LC-MS data showed \nthat proteins associated with various T cell memory subsets were enriched in Torin-1 treated \nsamples ( Figure 4e ). This included increased expression of several hallmark proteins \nassociated with T cell memory, such as BCL2, SELL, GATA3 and CD27, and downregulation of \nthe effector T cell related proteins EOMES, ZEB2, and EZH2 (Figure 4f).  \n \nWe next studied how mTOR inhibition favors the multipotency/memory trajectory during Teff \nactivation. Previous work in mice showed that the transcription factor FOXO1 controls the \nexpression of TCF1. FOXO1 is directly phosphorylated through the mTOR-AKT axis\n22,31, which \nexcludes FOXO1 from the nucleus and prohibits its transcriptional activity 41. In line with these \nprior studies, Torin-1 treatment of human Teff limited AKT phosphorylation upon T cell activation \n(Supplementary Figure 4 ). Importantly, Torin-1 treatment also resulted in increased nuclear \naccumulation of FOXO1 after 24 hours of TCR stimulation compared to DMSO controls ( Figure \n4g-h). Combined, these data imply that the induction of a more multipotent T cell state by Torin-\n1 during Teff activation is driven by sustained nuclear retention of FOXO1 through the mTOR-\nAKT axis. \n \nmTOR inhibition during priming imprints T cell state   \nHaving established that Torin-1 treatment during T cell activation maintains Teff cells in a more \nmultipotent state, we next asked if mTOR inhibition would yield similar effects during naïve T cell \npriming, and whether this multipotent state remains ‘imprinted’. Therefore, we activated naïve \nCD8\n+ T cells for 48 hours using anti-CD3/CD28 beads in the presence or absence of Torin-1 \n(Figure 5a). After 48 hours, T cells were removed from activation, washed to eliminate Torin-1 \nfrom the cultures, and subsequently maintained under identical culture conditions ( Figure 5a). \nAt day 7 post activation, Torin-1 treated T cells had expanded significantly less than control \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\ntreated T cells (5-fold versus 30-fold expansion; Figure 5b). Consistent with our observations in \nTeff cells ( Figure 4), Torin-1 treatment during T cell priming resulted in a substantially higher \nabundance of TCF1 HIGH T cells ( Figure 5c). Moreover, mTOR inhibition substantially increased \nthe expression of the lymph node homing receptor CCR7 and the co-stimulatory protein CD28 \n(Figure 5d-e ), whereas expression of the inhibitory receptors CD39 and TIM3 was markedly \nreduced (Figure 5d-f). Memory-precursor-like T cells generally produce less effector molecules \nthan their more differentiated counterparts (e.g., short-lived effectors) 42. Strikingly, Torin-1 \ntreated T cells displayed only a slight reduction in the production of TNF after 3 hours of re-\nactivation, and no significant differences in the production of IFN γ , or IL2 (Figure 5g). Together, \nthese data indicate that Torin-1 treatment during T cell priming skews T cells towards a more \nmultipotent T cell state.  \n \nTransient mTOR inhibition during priming is sufficient to imprint a multipotent T cell state \nTreatment with mTOR inhibitors has been shown to skew T cell responses to a multipotent \nmemory-precursor fate in mouse models\n21,22. In line with this, recent clinical studies revealed \nthat patients receiving continuous mTOR inhibition generated enhanced vaccine-induced T cell \nmemory pools\n43–45. However, long-term treatment with immune-suppressive mTOR inhibition \ncan lead to adverse effects 46,47, impeding the broad use of mTOR inhibitors to improve vaccine \nefficacy. Because we found that a 48-hour Torin-1 treatment imprints a multipotent state in \nprimed naïve T cells, we questioned whether transient mTOR inhibition during the initial stage of \nT cell priming would be sufficient achieve a similar outcome. To test this, naïve T cells were \nprimed with anti-CD3/CD28 beads in the presence or absence of Torin-1 for 16 hours ( Figure \n6a). T cells were then washed to remove Torin-1, a and T cell activation was continued for the \nremaining 32 hours (total 48 hours). This transient Torin-1 treatment impacted the population \nexpansion less than the continuous treatment (10-fold versus 5-fold expansion; Figure 6b). \nRemarkably, transient Torin-1 treatment was sufficient to imprint a more multipotent T cell state, \nas T cells expressed significantly higher levels of TCF1 at day 7 post-activation ( Figure 6c ). \nThis was matched by enhanced expression of CCR7 and CD28 and reduced expression of \nCD39 and TIM3 ( Figure 6d-e ). Importantly, the transient Torin-1 treatment did not affect the \nproduction of cytokines during a 3-hour re-stimulation assay (Figure 6f-g). Yet, expression of \nthe cytolytic protein granzyme B was markedly reduced ( Figure 6h ), which is a feature of \nmultipotent T cells\n48. To test whether the reduced Granzyme B expression resulted in reduced T \ncell functionality, we measured the killing capacity of Torin-1 and control -treated T cells in a co-\nculture assay. To test this independent of TCR-specificity we used ‘universal target’ cells that \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\nexpress a membrane-tethered single-domain anti-CD3 antibody ( Figure 6i, see Methods ). \nWhereas the production of TNF was unaltered, T cells produced slightly less IFN γ  and IL2 after \n16 hours of co-culture (Figure 6j). Interestingly, T cell activation markers (CD25, CD137, CD69) \nwere induced to a similar degree in T cells primed with or without Torin-1 ( Figure 6k ), \nsuggesting that the response to TCR triggering was equal. In line with this observation, Torin-1 \ntreated T cells were equally capable of killing target cells as the control T cells, shown by equal \nnumbers of remaining Mel888 cells after a 48-hour co-culture ( Figure 6l ). In conclusion, \ntransient Torin-1 treatment during the first 16 hours of T cell priming is sufficient to imprint T cell \ndifferentiation toward more multipotent T cell state. Importantly, despite lacking many effector-\nassociated characteristics, Torin-1 treated treatment generates functionally competent, cytotoxic \nT cells.  \n  \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\nDISCUSSION \n \nIn this study, we present a time-resolved dissection of matched transcriptome and proteome \ndynamics in activated human T cells. Our data show that TCR stimulation induces a highly \nordered remodeling of the protein networks and places mTOR activity as a master regulator of T \ncell differentiation during early activation. Specifically, mTOR directs the choice between effector \nand multipotent T cell states within 24 hours of TCR stimulation by promoting translation \n(transcripts containing TOP motifs), proliferation (maintenance of SKP2), and differentiation \n(nuclear exclusion of FOXO1). These findings highlight that proteome remodeling during early T \ncell activation proceeds through tightly regulated molecular stages that are directly orchestrated \nthrough mTOR.  \n \nOur quantification of transcript–protein relationships during the first 24 hours of T cell activation \nuncovered that mRNA and protein abundance are often decoupled, with the majority of genes \npresenting limited concordance between mRNA and protein abundance. Interestingly, our \nanalyses indicate that delayed protein expression accounts for a substantial fraction of the \nobserved mRNA–protein associations. Nevertheless, even after accounting for this delay, \napproximately 50% of all differentially expressed proteins show no association with their \ncorresponding mRNA levels. The most notable exception to this finding is cell-cycle–associated \nproteins, which display a strong positive correlation between mRNA and protein abundance, \nsuggesting that their protein abundance is mainly driven by transcriptional control. Additionally, \nand in line with previous studies in static T cell states\n10, our temporally resolved data identify \ngene class-specific mRNA and protein expression patterns under dynamic conditions, strongly \nsuggesting a dominant contribution of post-transcriptional and translational regulation. The \nmolecular mechanisms governing these levels of regulation, although emerging approaches are \nnow allowing their systematic investigation\n49. Irrespective of the regulatory pathways are yet to \nbe defined, our study provides a valuable resource to study the kinetics of mRNA and protein \nexpression during early T cell activation.  \n \nmTOR inhibition during T cell activation modulates the abundance of specific groups of proteins \nin a target-specific manner. For example, mTOR inhibition in murine T cells increased the \nexpression of the CDK inhibitor p27, thereby inhibiting the cell cycle by altering the ratio of p27 \nand cyclin proteins\n8. Notably, we identify the ubiquitin ligase SKP2 as a likely effector linking \nmTOR signaling to cell-cycle-related protein ex pression during T cell activation. SKP2 has been \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\ndescribed as a direct target of mTOR in gastric cancer, where mTOR activity promotes p27 \ndegradation via SKP2 upregulation 37. In agreement with this mechanism, Torin-1 treatment \nsuppresses SKP2 accumulation and increased ex pression of p27 during T cell activation, \nleading to an impairment of the G1/S transition. Provided the marked reduction of SKP2 protein \nabundance following mTOR inhi bition, it is tempting to speculat e that mTOR directly stabilizes \nSKP2 through phosphorylation, as was reported for cancer cells 37. Overall, our data point to \nSKP2 as a putative mediator through which mT OR coordinates activa tion-induced cell-cycle \nprogression in human T cells.  \n \nOur findings support a staged temporal model of early T cell activation in which: (i) cell-cycle \noccurs rapidly (minutes - 2 hours), followed by (ii) translational capacity build-up (2-6 hours). \nSimilarly to cell cycle genes, mTOR inhibition selectively represses protein modules involved in \ntranslation, including TOP-motif-containing transcripts\n8,25. Yet, proteins induced by T cell \nactivation were largely unaffected by Torin-1 treatment, indicating that the enhanced translation \nof these proteins by TCR signaling pathways does not rely on mTOR activity. This is consistent \nwith literature showing that NFAT and AP-1 activity is largely mTOR-independent\n50,51. Lastly, we \nshow that mTOR (iii) instruct s the commitment to differentiation at later stages of T cell \nactivation (6-24 hours). At this time point, mTOR inhibition promotes the expression of the \ntranscription factor TCF1, which mirrors the transcriptional and functional profile of memory-\nprecursor cells observed in long-term rapamycin-treated murine models\n21,22,27,52. Mechanistically, \nwe present that this shift of phenotype towards memory-like cells aligns with the nuclear \ntranslocation of FOXO1, which is prevented by mTORC2-mediated phosphorylation of AKT\n22. \nBuilding on this observation, we demonstrate that transient mTOR inhibition during T cell \npriming, limited to 16-hour treatment, is sufficient to durably skew T cell fate toward a phenotype \nresembling long-lived memory-precursor cells, while retaining their cytotoxic capacity. Thus, \ntransient mTOR inhibition enhances T cell multipotency without compromising functional \npotency. Previous pre-clinical and clinical studies 32,43–45 have shown that mTOR modulation \nduring vaccination strategies enhances the frequency and longevity of memory-like T cells. \nHowever, the broad immunosuppressive effects associated with continuous mTOR inhibition \nstrongly reduce the applicability of such drugs in a clinical setting. Our findings provide a \nrationale to overcome this limitation, proposing that short-term co-administration of mTOR \ninhibition during vaccination strategies should be sufficient  to enhance the durability of vaccine-\ninduced immunity.  \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\nMETHODS  \nCell culture  \nPeripheral blood mononuclear cells (PBMCs) from anonymized healthy donors were used in \naccordance with the Declaration of Helsinki (Seventh Revision, 2013) after written informed \nconsent (Sanquin). PBMCs were isolated through Lymphoprep density gradient separation \n(Stemcell Technologies). For matched mass spectrometry and RNA-sequencing analysis, \nPMBCs from 3 pools of 40 donors were used. For flow cytometry analysis of Teff, PBMCs were \npooled from 5 donors.  Cells were used after cryopreservation. T cells were activated in 24-well \nplates were pre-coated overnight at 4°C with 2 µg/mL rat a-mouse IgG2a (MW1483, Sanquin) in \nphosphate-buffered saline (PBS). Plates were washed with PBS and coated for >3 h with 1 \nµg/mL \nα CD3 (HIT3a, Biolegend) at 37°C. 1.3x10 6 PBMCs/well were seeded with 1 µg/mL \nsoluble α CD28 (CD28.2, Biolegend) in 1 mL culture medium (Iscove's Modified Dubecco's \nMedium,  IMDM) supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, 100 \nµg/mL streptomycin, and 2 mM L-glutamine. After 48h of activation at 37°C, 5% CO2, cells were \nharvested and cultured in standing T25/T75 tissue culture flasks (Thermo Scientific) at a density \nof 0.8x10 6/mL in culture medium supplemented with 100 IU/mL recombinant human (rh)IL2 \n(Protech) and 10 ng/mL rhIL-15 (Protech). Culture media was refreshed every 2-3 days.  Teff \ncells were cultured for 9 days without TCR triggering before re-activation and time-course \nsampling. \nNaïve CD8\n+ T cells were isolated from PBMCs of individual donors using the BD IMag™ Human \nNaïve CD8+ T Cell Enrichment Set (BD Biosciences) following the manufacturers protocol. Cells \nwere cultured in RPMI (Gibco) supplemented with 10% FBS, 100 U/mL penicillin, 100 µg/mL \nstreptomycin, 50 IU/ml rhIL2 (Protech), 5 ng/ml rhIL15 (Gibco) and 10 ng/ml IL7 (Gibco). \nIsolated cells were activated using Dynabeads™ Human T-Activator CD3/CD28 for T Cell \nExpansion and Activation (Gibco) at a 1:1 T-cell-to-bead ratio, in the presence of 250 nM Torin-1 \n(dissolved in DMSO) or an equal v/v percentage of DMSO (0.05%). Torin-1 and DMSO was \nremoved from cell cultures at the indicated timepoints by 3 sequential washes in complete \nculture media. Next, cells were pelleted and re-seeded in culture medium. CD8\n+ T cell cultures \nwere passaged every 2-3 days and were kept at a density of 0.4-1.0x106 cells/mL.   \n \nT cell activation and time course sampling \nTeff cells were restimulated with 1  μ g/ml soluble anti-CD3 ε (Pelicluster CD3, Sanquin) and 1 \nμ g/mL anti-CD28 (clone 28.2 biolegend) for the indicated durations. For mTOR inhibition, \n250nM Torin-1, or an equal v/v percentage of DMSO (0.05%), was added. For intracellular \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\nprotein measurements, 1 μ g/mL brefeldin A (BD Bioscience) and 1 μ g/mL Monensin (Invitrogen) \nwere added during the last two hours of activation. CD107a expression was measured by \nadding anti-CD107a (H4B3, BD Bioscience) from the start of activation. To measure translation \nefficiency, T cells were incubated with 5 μ g/ml Puromycin dihydrochloride (Sigma) in culture \nmedia for 10 min at 37°C. \n \nRNA sequencing  \n1 x10\n6 T cells were used for RNA sequencing (Genewiz; Azenta Life Sciences). During RNA \nisolation and library preparation, ribosomal RNA (rRNA) was depleted. Approximately 30x10 6 \nreads were obtained per sample. Sequencing adaptor trimming and low-quality read removal \nwas performed using fastp 53 (default settings). As rRNA depletion is not absolute, obtained \nreads were first mapped to a custom rRNA library (deposited to Zenodo) using STAR 54. \nUnmapped reads were subsequently aligned to the GRCh38 primary assembly using the \ngencode.v47 basic annotation (both files downloaded from \nhttps://www.gencodegenes.org/human\n). Aligned reads were subsequently quantified using \nhtseq-count55. Differential gene-expression analysis between indicated samples was performed \nusing the R package limma56. \n \nMass spectrometry data acquisition  \nFor mass spectrometry, samples were lysed in 1% sodium deoxycholate (Bioworld), 10 /i8 mM \nTCEP (Thermo Scientific), 40 /i8 mM chloroacetamide (Sigma-Aldrich), 100 /i8 mM Tris-HCl pH 8.0 \n(Gibco). Lysates were incubated for 5/i8 minutes at 95/i8 °C, sonicated for 10/i8 minutes in a sonifier \nbath (Branson model 2510), and digested overnight with 250 ng Trypsin/Lys-C (Promega) at 25 \n°C. Tryptic digests were desalted on an Assaymap BRAVO (Agilent) using C18 cartridges (5ul, \nAgilent) according to manufactures instructions. Concentrations were determined with a Pierce \nQuantitative Fluorometric Peptide Assay (Thermo). Peptides (500 ng) were loaded on Evotip \nPure (Evosep) tips according to manufacturer’s guidelines and separated on a Performance \nColumn (EV1115, EvoSep) using the 60 samples per day gradient. Buffer A consisted of 0.1% \nformic acid, buffer B of 0.1% formic acid in acetonitrile (Biosolve). Data was acquired on a \ntimsTOF HT mass spectrometer (Bruker Daltonics) operated in diaPASEF mode, using a MS1 \nscan range of 100-1700 m/z. MS2 acquisition was performed using 32 pyDIAID-optimized\n57 \nmass and ion mobility windows, ranging from 400.2 to 1500.8 m/z and 0.70 to 1.50 1/k0 with a \ncycle time of 1.80 seconds. A collision energy of 20.00 eV at 0.6 1/k0 and 59 eV at 1.60 1/k0 \nwas used. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \nMass spectrometry analysis  \nRaw mass spectrometry files were processed with DIA-NN 1.8.1, using the reviewed human \nproteome database (Swiss-ProtDatabase, 20,370 entries, release 2024.04.22). Standard \nsettings and a generated library-based spectra search were used. Label free quantification \nvalues were log2 transformed. Missing values were imputed by a normal distribution (width=0.3, \nshift = 1.8), assuming these proteins were close to the detection limit. Differential abundance of \nproteins was tested using the R package limma\n56. \n \nProtein expression clusters and correlation to mRNA abundance \nProteins exhibiting differential expression in any comparison between time-points were selected. \nCutoffs for differential expression were set at a Benjamini-Hochberg adjusted P < 0.05 and an \nabsolute fold-change > 1.27, resulting in 2,650 proteins for further analysis. Protein expression \nwas subsequently scaled across samples and divided into 5 clusters using kmeans clustering. \nUpon visual inspection of the expression profiles, clusters were defined as “early up\", \"gradual \nup\", \"late up\", \"gradual down\" and \"late down\". Full list of classification in Supplementary Table \n3.  Enriched Reactome pathways in clusters were tested with the compareCluster function from \nthe ClusterProlifer R package. \nThe within-gene correlation between protein and mRNA abundance was calculated as \nSpearman's rank correlation coefficient. To account for the potentially delayed protein \nexpression, ‘delayed correlations’ were calculated by shifting the time points obtained by RNA \nsequencing by 1 or 2 positions. This effectively tests if mRNA abundance is predictive of protein \nlevels at later timepoints.  Proteins and their relation to mRNA abundance were subsequently \nclassified as “Positive association\" (r > 0.5), \"Delayed positive association\" (r > 0.5 at time-point \nshifts of 1 or 2), \"Negative association\" (r < -0.5), \"No association\" (r > -0.5 and r < 0.5). Full list \nof classification in Supplementary Table 9. \n \nXGBoost modelling of protein expression clusters \nTo test the effects of Torin-1 treatment on the protein expression clusters, an XGBoost model \nwas trained to classify cluster identity, using the protein abundance of the 2,650 differentially \nexpressed proteins from Figure 1b. Before model training abundance values were scaled across \nsamples. Model training was performed using the caret\n58 and xgboost35 R packages. Repeated \ncross-validation was applied during training (10 cross-validations, 2 repeats). Settings for the \nXGBoost model were: number of rounds = 1000, maximum tree depth = 1, subsample ratio of \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\ncolumns = 0.5, learning rate = 0.3, minimum loss reduction = 1, minimum sum of instance \nweight = 0.9, subsample ratio of the training instances = 1. Training objective was set to \n\"multi:softprob\" and the target metric was “Accuracy”. The resulting model had an accuracy of \n0.901 and a kappa of 0.870. The obtained XGBoost model was subsequently used to predict \nclass identity in the LC-MS data obtained in Figure 2 using the predict function implemented in \nthe caret R package.  \n \nImmunoblotting  \nCell lysates (1×10\n6 cells/sample) were prepared by standard procedures using RIPA lysis buffer \n(Thermo) supplemented with protease and phosphatase inhibitors (Thermo). Lysates were run \non 4–12% SDS-PAGE (Thermo). SDS-PAGE gels were directly transferred onto nitrocellulose \nmembranes (iBlot2, Thermo). Membranes were blocked with 5% BSA TBST solution (Fraction \nV, Sigma). And incubated with \nα -SKP2 (Proteintech), or α -RhoGDI (MAB9959, Abnova), \nfollowed by α -Rabbit (4050-05, Southern Biotech) HRP-conjugated secondary antibodies.  \n \nFlow cytometry and intracellular staining.  \nT cells were washed with FACS buffer (PBS with 1% FBS and 2 mM EDTA) and labeled for 20 \nminutes at 4°C with α -CD4 (SK3, BD Horizon), α -CD8 (SK1, BD Horizon), α -CD69 (FN50, BD \nHorizon), α -TIM3 (73D, BD Horizon), α -CD137(4B4-1, BioLegend), α -CD27 (O323, Biolegend), \nα -CCR7 (3D12, BD Pharmigen), α -CD28 (CD28.2, BioLegend), α -CD39 (AI, BD Horizon), α -\nCD25 (BC96, BioLegend). Dead cells were excluded with Near-IR (Life Technologies). For \nintracellular staining, cells were fixed and permeabilized with Cytofix/Cytoperm kit (BD \nBiosciences) and stained with α -IFN-γ  (4S.B3, BD Bioscience), α -TNF (MAb11, BD Bioscience), \nα -IL2 (MQ1-17H12, Biolegend) α -Puromycin (12D10, Merk). For transcription factor expression, \ncells were fixed and permeabilized with eBiosci ence™ Foxp3/ Transcription Factor Staining \nBuffer Set (Invitrogen) prior to staining with α -TCF1 (C63D9, Santa Cruz Biotechnology), and α -\nTbet (4B10, Biolegend). For phospho-flow cytometry, cells were fixed with IC Fixation Buffer \n(eBioscience) for 10 mi nutes at 37°C and permeabilized with 90% ice-cold methanol for 30 \nminutes at 4°C. Cells were stained with α -4E-BP1 (V3NTY24, eBioscience) and α -AKT \n(SDRNR, eBioscience) for 1 hour at room temperature. For cell cycle staining, cells were fixed \nand permeabilized with eBioscience™ Foxp3/ Tr anscription Factor Staining Buffer Set \n(Invitrogen) prior to staining with \nα -Ki67(B56, BD Bioscience). Cells were then treated with 100 \nμ g/ml of RNAse for 10 minutes at room temperature. Prior to acquisition, cells were stained with \nPropidium Iodide (Invitrogen). Acquisition was performed using BD LSRFortessa Cell Analyzer \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n(BD Bioscience) or FACS Symphony A5 Cell Analyzer (BD Bioscience). Data were analyzed \nwith FlowJo (BD Biosciences, version 10.8.1). \nFor quantification of FOXO1 nuclear localization, cells were labelled for 20 minutes at 4°C with \nRY586 mouse α -human CD8 + (568115, BD Biosciences). Cells were fixed and permeabilized \nusing the eBioscience™ Foxp3/Transcription Factor Staining Buffer Set (Invitrogen) prior to \nstaining for 1 hour at RT with \nα -FOXO1 (C29H4, Santa Cruz Biotechnology). Prior to \nacquisition, cells were stained with DAPI (62248, Thermo Scientific). Acquisition was performed \nusing ImageStream MK II (Cytek Amnis). Data were analyzed with IDEAS (Amnis/ Luminex). \n \nCo-culture experiments \nFor co-culture experiments, a modified Mel888 line was used expressing a membrane-tethered \nanti-CD3 single-chain variable fragment (scFv) of the OKT3 clone, which was generated as \npreviously described 59, following the approach of Leitner et al 60. 15,000 Mel888 cells were \nseeded per well on a flat-bottom 96-well tissue-culture plate. After 4 hours, 15,000 T cells were \nadded to each well, resulting in a 1:1 effector:target ratio. T cells were seeded in the presence of \n50 U/ml rhIL2. Plates were centrifuged at 10 × g for 1 minute (without brake) and incubated at \n37C for 30 minutes. Next, Brefeldin A (BD Biosciences) was added to the culture, following \nmanufacturers protocol. The co-culture was incubated at 37 \noC for 16 hours. Co-cultures were \nharvested, and cytokine production was assessed using the BD Cytofix/Cytoperm™ \nFixation/Permeabilization Kit (BD Biosciences). For measurements of T cell activation and \nmelanoma cell killing, co-cultures were in cubated for 48 hours befor e analysis. For these \nanalyses, co-cultures were harvested and stained for the indicated activation markers. Before \nflow cytometry analysis, 5,000 Precision Count Beads™ (Biolegend) were added to allow for the \ncalculation of surviving melanoma cells. \n \nData Visualization and statistical analysis \nStatistical analysis was performed using the rstatix R package, using two-tailed paired Student’s \nt-tests followed by Benjamini-Hochberg correction. P values are indicated in the figures. Data \nwere visualized with ggplot2\n61 (version 3.4.2). Heatmaps were generated using the \nComplexHeatmap R package. \n \nRESOURCE AVAILABILITY  \nLEAD CONTACT \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\nInformation and requests for resources and reagents should be directed to and will be fulfilled \nby the lead contacts Monika C Wolkers (m.wolkers@sanquin.nl) and Kaspar Bresser \n(k.bresser@sanquin.nl)  \n \nMATERIALS AVAILABILITY \nThis study did not generate new unique reagents.  \n \nDATA CODE AND AVAILABILITY \n• The MS data have been deposited to the ProteomeXchange Consortium via the PRIDE 62  \npartner repository with the dataset identifier PXD073124 \n• The RNA sequencing data have been deposited to DANS Data Station Life Sciences \nand are available at https://doi.org/10.17026/LS/PAYFTR \n• All codes used in the analyses described in this study are available at \nhttps://github.com/kasbress/Tcell_Activation_mTOR \n• Any additional information required to reanalyze the data reported in this paper is \navailable from the lead contact upon request \n \nACKNOWLEDGMENTS \nW e  t h a n k  E r i k  M u l  f r o m  t h e  S a n q u i n  C e n t r a l  F a c i l i t y  f o r  a n a l y s i s  o f  t h e  F O X O 1  n u c l e a r  l o c a l i z a t i o n  d a t a .  \nThis research was supported by Oncode Institute, the European Research Council consolidator award \nPRINTERS 817533, and Landsteiner Foundation for Blood T ransfusion (LSBR) R esearch grant 2202 to \nM.C. W .; and the Cancer Center Amsterdam (CCA) research grant CCA2023-9-91 to K.B. \n \nAUTHOR CONTRIBUTIONS \nConceptualization, A.P . J, M. V .L., K.B., M.C. W .; Methodology , A.P . J, M. V .L., K.B., and M.C. W .; Investigation , \nA.P . J, M. V .L., L. W ., K.B., C. v .d.Z., A. J.H.; Formal Analysis, A.P . J, M. V .L., K.B., A. J.H.; V alidation, A.P .J, M. V .L., \nK.B..; W riting-Original Draft, A.P . J, M. V .L., K.B. W riting – Review & Editing, A.P .J, M. V .L., K.B. and M.C. W .; \nSupervision, K.B., and M.C. W .; Funding Acquisition, M.C. W and K.B..  \n \nDECLARATION OF INTEREST \nThe authors declare no competing interests. \n \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\nREFERENCES \n \n1. Gerlach, C. et al. The Chemokine Receptor CX3CR1 Defines Three Antigen-\nExperienced CD8 T Cell Subsets with Distinct Roles in Immune Surveillance and \nHomeostasis. Immunity 45, 1270–1284 (2016). \n2. Jaccard, A. et al. Reductive carboxylation epigenetically instructs T cell \ndifferentiation. Nature 621, 849–856 (2023). \n3. Akondy, R. S. et al. Origin and differentiation of human memory CD8 T cells after \nvaccination. Nature 552, 362–367 (2017). \n4. Jurgens, A. P ., Popovi\nć , B. & Wolkers, M. C. T cells at work: How post-transcriptional \nmechanisms control T cell homeostasis and activation. Eur J Immunol 51, 2178–\n2187 (2021). \n5. Sinclair, L. V. & Cantrell, D. A. Protein Synthesis and Metabolism in T Cells. Annual \nReview of Immunology 43, 343–366 (2025). \n6. Marchingo, J. M. & Cantrell, D. A. Protein synthesis, degradation, and energy \nmetabolism in T cell immunity. Cell Mol Immunol 19, 303–315 (2022). \n7. Wolf, T. et al. Dynamics in protein translation sustaining T cell preparedness. Nat \nImmunol 21, 927–937 (2020). \n8. Howden, A. J. M. et al.  Quantitative analysis of T cell proteomes and environmental \nsensors during T cell differentiation. Nat Immunol 20, 1542–1554 (2019). \n9. Araki, K. et al. Translation is actively regulated during the differentiation of CD8+ \neffector T cells. Nat Immunol 18, 1046–1057 (2017). \n10. Nicolet, B. P. & Wolkers, M. C. The relationship of mRNA with protein expression in \nCD8+ T cells associates with gene class and gene characteristics. PLoS ONE 17, \ne0276294 (2022). \n11. Hukelmann, J. L. et al. The cytotoxic T cell proteome and its shaping by the kinase \nmTOR. Nat Immunol 17, 104–112 (2016). \n12. Turner, M. & Petkau, G. RNA-binding proteins and ribonucleoproteins as \ndeterminants of immunity. Nat Rev Immunol https://doi.org/10.1038/s41577-025-\n01254-2 (2026) doi:10.1038/s41577-025-01254-2. \n13. Weerakoon, H. et al. Integrative temporal multi-omics reveals uncoupling of \ntranscriptome and proteome during human T cell activation. NPJ Syst Biol Appl 10, \n21 (2024). \n14. Nicolet, B. P., Zandhuis, N. D., Lattanzio, V. M. & Wolkers, M. C. Sequence \ndeterminants as key regulators in gene expression of T cells. Immunol Rev 304, 10–\n29 (2021). \n15. Salerno, F. et al. Translational repression of pre-formed cytokine-encoding mRNA \nprevents chronic activation of memory T cells. Nat Immunol 19, 828–837 (2018). \n16. Myers, D. R., Norlin, E., Vercoulen, Y . & Roose, J. P . Active Tonic mTORC1 Signals \nShape Baseline Translation in Naive T Cells. Cell Reports 27, 1858-1874.e6 (2019). \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n17. Zeng, H. et al. mTORC1 couples immune signals and metabolic programming to \nestablish Treg-cell function. Nature 499, 485–490 (2013). \n18. Michalek, R. D. et al. Estrogen-related receptor-α  is a metabolic regulator of effector \nT-cell activation and differentiation. Proc. Natl. Acad. Sci. U.S.A. 108, 18348–18353 \n(2011). \n19. Yang, K., Neale, G., Green, D. R., He, W. & Chi, H. The tumor suppressor Tsc1 \nenforces quiescence of naive T cells to promote immune homeostasis and function. \nNat Immunol 12, 888–897 (2011). \n20. Pollizzi, K. N. et al. mTORC1 and mTORC2 selectively regulate CD8+ T cell \ndifferentiation. J. Clin. Invest. 125, 2090–2108 (2015). \n21. Araki, K. et al. mTOR regulates memory CD8 T-cell differentiation. Nature 460, 108–\n112 (2009). \n22. Zhang, L. et al. Mammalian Target of Rapamycin Complex 2 Controls CD8 T Cell \nMemory Differentiation in a Foxo1-Dependent Manner. Cell Reports 14, 1206–1217 \n(2016). \n23. Yang, K. et al. T Cell Exit from Quiescence and Differentiation into Th2 Cells Depend \non Raptor-mTORC1-Mediated Metabolic Reprogramming. Immunity 39, 1043–1056 \n(2013). \n24. Ricciardi, S. et al. The Translational Machinery of Human CD4+ T Cells Is Poised for \nActivation and Controls the Switch from Quiescence to Metabolic Remodeling. Cell \nMetabolism 28, 895-906.e5 (2018). \n25. Thoreen, C. C. et al. A unifying model for mTORC1-mediated regulation of mRNA \ntranslation. Nature 485, 109–113 (2012). \n26. Saxton, R. A. & Sabatini, D. M. mTOR Signaling in Growth, Metabolism, and \nDisease. Cell 168, 960–976 (2017). \n27. Pollizzi, K. N. & Powell, J. D. Regulation of T cells by mTOR: the known knowns and \nthe known unknowns. Trends in Immunology 36, 13–20 (2015). \n28. Tcherkezian, J. et al. Proteomic analysis of cap-dependent translation identifies \nLARP1 as a key regulator of 5′ TOP mRNA translation. Genes Dev. 28, 357–371 \n(2014). \n29. Philippe, L., van den Elzen, A. M. G., Watson, M. J. & Thoreen, C. C. Global \nanalysis of LARP1 translation targets reveals tunable and dynamic features of 5’ \nTOP motifs. Proc Natl Acad Sci U S A 117, 5319–5328 (2020). \n30. García-Martínez, J. M. & Alessi, D. R. mTOR complex 2 (mTORC2) controls \nhydrophobic motif phosphorylation and activation of serum- and glucocorticoid-\ninduced protein kinase 1 (SGK1). Biochemical Journal 416, 375–385 (2008). \n31. Sarbassov, D. D., Guertin, D. A., Ali, S. M. & Sabatini, D. M. Phosphorylation and \nRegulation of Akt/PKB by the Rictor-mTOR Complex. Science 307, 1098–1101 \n(2005). \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n32. Lin, X. et al. Sustained mTORC1 activation in activated T cells impairs vaccine \nresponses in older individuals. Sci. Adv. 11, eadt4881 (2025). \n33. Jurgens, A. P . et al. mTOR signaling during T cell activation promotes cytokine \nproduction in T cells through 3′  UTR-mediated translation control. Molecular Cell 85, \n4452-4462.e5 (2025). \n34. Salerno, F., Paolini, N. A., Stark, R., Von Lindern, M. & Wolkers, M. C. Distinct PKC-\nmediated posttranscriptional events set cytokine production kinetics in CD8+ T cells. \nProc. Natl. Acad. Sci. U.S.A. 114, 9677–9682 (2017). \n35. Chen, T. & Guestrin, C. XGBoost: A Scalable Tree Boosting System. in Proceedings \nof the 22nd ACM SIGKDD International Conference on Knowledge Discovery and \nData Mining 785–794 (ACM, San Francisco California USA, 2016). \ndoi:10.1145/2939672.2939785. \n36. Argüello, R. J. et al. SunRiSE – measuring translation elongation at single-cell \nresolution by means of flow cytometry. Journal of Cell Science 131, jcs214346 \n(2018). \n37. Geng, Q. et al. Phosphorylation by mTORC1 stablizes Skp2 and regulates its \noncogenic function in gastric cancer. Mol Cancer 16, 83 (2017). \n38. Shapira, M., Kakiashvili, E., Rosenberg, T. & Hershko, D. D. The mTOR inhibitor \nrapamycin down-regulates the expression of the ubiquitin ligase subunit Skp2 in \nbreast cancer cells. Breast Cancer Res 8, R46 (2006). \n39. Gerlach, C. et al. Heterogeneous differentiation patterns of individual CD8+ T cells. \nScience 340, 635–639 (2013). \n40. Pearce, E. L. et al. Enhancing CD8 T-cell memory by modulating fatty acid \nmetabolism. Nature 460, 103–107 (2009). \n41. Michelini, R. H., Doedens, A. L., Goldrath, A. W. & Hedrick, S. M. Differentiation of \nCD8 memory T cells depends on Foxo1. Journal of Experimental Medicine 210, \n1189–1200 (2013). \n42. Amoah, S., Yammani, R. D., Grayson, J. M. & Alexander-Miller, M. A. Changes in \nFunctional but Not Structural Avidity during Differentiation of CD8+ Effector Cells In \nVivo after Virus Infection. The Journal of Immunology 189, 638–645 (2012). \n43. Netti, G. S. et al. mTOR inhibitors improve both humoral and cellular response to \nSARS-CoV-2 messenger RNA BNT16b2 vaccine in kidney transplant recipients. \nAmerican Journal of Transplantation 22, 1475–1482 (2022). \n44. Perkins, G. B. et al. Mechanistic Target of Rapamycin Inhibitors and Vaccine \nResponse in Kidney Transplant Recipients. JASN \nhttps://doi.org/10.1681/ASN.0000000716 (2025) doi:10.1681/ASN.0000000716. \n45. Withers, H. G. et al. mTOR inhibition modulates vaccine-induced immune responses \nto generate memory T cells in patients with solid tumors. J Immunother Cancer 13, \ne010408 (2025). \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n46. Pallet, N. & Legendre, C. Adverse events associated with mTOR inhibitors. Expert \nOpinion on Drug Safety 12, 177–186 (2013). \n47. Nguyen, L. S. et al. Sirolimus and mTOR Inhibitors: A Review of Side Effects and \nSpecific Management in Solid Organ Transplantation. Drug Saf \nhttps://doi.org/10.1007/s40264-019-00810-9 (2019) doi:10.1007/s40264-019-00810-\n9. \n48. Delpoux, A. et al. FOXO1 constrains activation and regulates senescence in CD8 T \ncells. Cell Reports 34, 108674 (2021). \n49. Nicolet, B. P., Jurgens, A. P ., Bresser, K., Guislain, A. & Wolkers, M. C. Learning the \nSequence Code for mRNA and Protein Abundance in Human Immune Cells. \nhttp://biorxiv.org/lookup/doi/10.1101/2023.09.01.555843 (2023) \ndoi:10.1101/2023.09.01.555843. \n50. Lee, A. S., Kranzusch, P . J., Doudna, J. A. & Cate, J. H. D. eIF3d is an mRNA cap-\nbinding protein that is required for specialized translation initiation. Nature 536, 96–\n99 (2016). \n51. Gamper, C. J. & Powell, J. D. All PI3Kinase signaling is not mTOR: dissecting \nmTOR-dependent and independent signaling pathways in T cells. Front. Immun. 3, \n(2012). \n52. Delgoffe, G. M. et al. The kinase mTOR regulates the differentiation of helper T cells \nthrough the selective activation of signaling by mTORC1 and mTORC2. Nat \nImmunol 12, 295–303 (2011). \n53. Chen, S. fastp 1.0: An ultra\n‐ fast all‐ round tool for FASTQ data quality control and \npreprocessing. iMeta 4, e70078 (2025). \n54. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 \n(2013). \n55. Putri, G. H., Anders, S., Pyl, P . T., Pimanda, J. E. & Zanini, F. Analysing high-\nthroughput sequencing data in Python with HTSeq 2.0. Bioinformatics 38, 2943–\n2945 (2022). \n56. Smyth, G. K. Linear Models and Empirical Bayes Methods for Assessing Differential \nExpression in Microarray Experiments. Statistical Applications in Genetics and \nMolecular Biology 3, 1–25 (2004). \n57. Skowronek, P . et al. Rapid and In-Depth Coverage of the (Phospho-)Proteome With \nDeep Libraries and Optimal Window Design for dia-PASEF. Molecular & Cellular \nProteomics 21, 100279 (2022). \n58. Kuhn, M. Building Predictive Models in R Using the caret Package. J. Stat. Soft. 28, \n(2008). \n59. Lattanzio, M. V. et al. Mapping the dynamic RNA binding proteome in human effector \nT cells identifies differentiation and cytotoxicity regulators. Preprint at \nhttps://doi.org/10.64898/2025.12.05.692493 (2025). \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n60. Leitner, J. et al. T cell stimulator cells, an efficient and versatile cellular system to \nassess the role of costimulatory ligands in the activation of human T cells. Journal of \nImmunological Methods 362, 131–141 (2010). \n61. Hadley Wickham. Ggplot2: Elegant Graphics for Data Analysis. (Springer-Verlag \nNew York, 2016). \n62. Perez-Riverol, Y. et al. The PRIDE database resources in 2022: a hub for mass \nspectrometry-based proteomics evidences. Nucleic Acids Research 50, D543–D552 \n(2022). \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \nFigure 1. Proteome and transcriptome remodeling during early T cell activation. ( a) Experimental\ndesign. (b) Differentially expressed proteins (n = 2,375, FDR < 0.05) were clustered into five temporal\nexpression modules using hierarchical clustering. Grey lines indicate individual proteins; colored lines\nindicate medians. Example proteins from each module are indicated beside the plots. ( c) Enrichment of\nReactome pathways in each module. A maximum of 5 pathways is shown for each cluster. See\nSupplementary Table 2 for full list. (d) Selected proteins for each protein module. Dots indicate individual\ndonors. (e) Relationship between protein- and mRNA-level fold changes comparing each time-point to the\ntal \nral \nes \n of \nee \nal \nhe \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\npre-activation baseline. Dots indicate genes, colored contours indicate a 2D kernel density estimation. \nSpearman correlation coefficient is indicated in the plots. ( f) Number of significant (adjusted P < 0.05) \ntranscripts and proteins at each indicated timepoint compared to the pre-activation baseline. ( g) \nSpearman correlations between the fold-changes of protein and mRNA across all activation time-points \ncompared to the pre-activation baseline (0 hours). Correlations on the diagonal correspond to the plots \nshown in panel (e). ( h) Examples scatterplots for the within-gene mRNA and protein associations. See \nmethods for definition of associations. (i) Number of genes within each association. ( j) Contribution of all \npositive and negative associations (left) and contribution of associations to all genes (right) indicated in \n(b). Displayed LC-MS data obtained from 3 pools of 40 donors. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \nFigure 2. mTOR instructs proteome remodeling during early T cell activation. (a) Experimental\ndesign. (b) LC-MS data from Figure 1 was used to train a XGBoost classifier for the protein expression\nmodules. This model was then used to classify the DMSO/Torin-1 LC-MS data. ( c) Model performance\nplotted as area under the ROC curves (AUC). Black lines connect matched donor pools. ( d-e) Puromycin\nincorporation assay of DMSO and Torin-1 treated CD8 + T cells at indicated time- points after CD3/CD28\ncross-linking. Histograms display puromycin incorporation in a representative donor (d), barplots display\nthe median fluorescence intensity for all donors (e). Black dashed line indicates the maximum density of\nfluorescence for the no puromycin control. ( f-g) Expression kinetics of proteins from the Reactome\npathway ‘translation initiation’ (f), and selected examples (g). Grey lines indicate individual proteins;\ncolored lines indicate medians (f). Black lines connect matched donor pools (g). ( h-i) Expression kinetics\nof proteins fro m the Reactome pathway ‘T cell activation’ (h), and selected examples (i). Grey lines\nindicate individual proteins; colored lines indicate medians (h). Black lines connect matched donor pools\n(i). LC-MS data is obtained from 3 pools of 40 donors. Panel (d) and (e) display 3 pools of 5 donors and\nare representative of 2 independent experiments. P values indicated in (e) were calculated using a two -\ntailed paired Student’s t-test followed by Benjamini-Hochberg correction.  \n \ntal \non \nce \nin \n28 \nay \n of \ne \ns; \nics \nes \nls \nnd \n-\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \nFigure 3. mTOR instructs early activation-induced T cell differentiation. (a) Clustering analysis of all\nTorin-1 dependent differentially expressed proteins from the LC- MS data. Euclidean distance is used for\nthe rows; spearman correlation is used for the columns. ( b) Enrichment analysis for Reactome pathways\nin each cluster using a hypergeometric model. A maximum of 5 pathways is shown for each cluster. See\nSupplementary Table 5 for full list. (c) Representative proteins from each Torin-1 sensitive cluster. Black\nlines connect matched donor pools. ( d) Expression kinetics of proteins translated from TOP- motif\ncontaining transcripts. Grey lines indicate individual proteins; colored lines indicate medians. ( e)\nPercentage of TOP-motif containing transcripts within each Torin-1 dependent cluster (panel a). ( f-g)\nPropidium-iodide-based cell cycle analysis of CD8 + T cells at 24 hours post-activation (anti- CD3/CD28).\n(f) Representative scatterplots with 2D kernel density estimation overlay to illustrate gating strategy.\nBoxes indicate definition of cell cycle phases. ( g) Median fluorescence intensity of Ki67 staining. Black\nlines connect individual donors. ( h) Percentage of cells in each phase of the cell cycle. Dots indicate\nindividual donors; bars represent group means. ( i-j) Expression kinetics (LC- MS) of SKP2 and p27 (i),\nand relative SKP2/p27 abundance ratios (j). Black lines connect matched donor pools. LC- MS data is\nobtained from 3 pools of 40 donors. Panel (f- h) display 3 pools of 5 donors and are representative of 2\nindependent experiments. P values indicated in (g-j) were calculated using a two-tailed paired Student’s t-\ntest followed by Benjamini-Hochberg correction. \nall \nfor \nys \nee \nck \ntif \n) \n) \n8). \ny. \nck \nte \n(i), \n is \nf 2 \n-\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \nFigure 4. mTOR instructs early activation-induced T cell differentiation. (a-d) Intracellular staining for\nTCF1 (a-b) and TBET (c-d) protein expression of CD8 + T cells at 24 hours post-activation (anti -\nCD3/CD28), measured by flow-cytometry. Representative histograms for single donors are shown (a, c).\nDashed lines indicate the maximum density of fluorescence for the non-a ctivated control (DMSO). Strip\ncharts display percentage of TCF1 positive cells (b) or median fluorescence intensity of TBET (d). Black\nlines connect individual donors. ( e) Gene-set enrichment analysis comparing Torin- 1 and DMSO treated\nsamples. Running enrichment score of selected pathways is shown, see Supplementary Table 6  for full\nresults. ( f) Clustering analysis of selected genes representative of T cell differentiation states. ( g-h)\nIntracellular localization analysis of FOXO1 and nucleic acid dye DAPI of CD8 + T cells at 24 hours post -\nactivation. Representative images (g) and quantifications for 3 donors (h) are shown. Overlap between\nDAPI and FOXO1 signals is scaled between 0 and 1 to visualize the amount of FOXO1 outside of the\nnucleus. LC-MS data is obtained from 3 pools of 40 donors. Panel (b, d) display 3 pools of 5 donors and\nare representative of 2 independent experiments. P values indicated in (b, d) were calculated using a two-\ntailed paired Student’s t-test followed by Benjamini-Hochberg correction. \n \n \nfor \n-\nc). \nrip \nck \ned \null \n) \n-\nen \nhe \nnd \n-\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \nFigure 5. mTOR inhibition during T cell priming imprints T cell state. (a) Experimental setup. Naïve\nCD8+ T cells were activated in the presence of Torin- 1 or DMSO for 48 hours and subsequently cultured\nfor 7 days in the absence of both. Read-outs are performed at day 7 post-activation. ( b) Fold-expansion\nof T cell cultures between day 0-7. (c) TCF1 protein expression, measured by intracellular flow-cytometry\nat day 7 post- activation. Representative histograms of TCF1 fluorescence intensity (left) and percentage\nof TCF1-positive T cells (right) are shown. Dashed line indicates the threshold for positive cells. ( d-e)\nSurface staining for indicated proteins, measured by flow- cytometry. Representative histograms (d) and\nmedian fluorescence intensity (MFI) of indicated proteins (e). Dashed lines (d) indicate the maximum\ndensity of fluorescence for the DMSO controls, solid lines (e) connect individual donors. ( f) Surface\nstaining for indicated proteins, measured by flow- cytometry. Scatterplots with 2D kernel density contours\nfor representative samples (left) and summarized percentages (right) are shown. ( g) Intracellular staining\nfor indicated cytokines after a 3-hour re-stimulation within anti- CD3/CD28 antibodies in the presence of\nbrefeldin A. Solid lines connect individual donors. Displayed data is compiled from of 2 independent\nexperiments comprising 5-6 donors. P values indicated in (c, e-g) were calculated using a two- tailed\npaired Student’s t-test followed by Benjamini-Hochberg correction. \n \n \nve \ned \non \ntry \nge \n) \nnd \nm \nce \nrs \nng \n of \nnt \ned \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \nFigure 6. Transient mTOR inhibition during T cell priming imprints T cell state. (a) Experimental\nsetup. Naïve CD8+ T cells were activated in the presence of Torin-1 or DMSO. After 16 hours Torin-1 and\nDMSO was removed, read-outs were performed at day 7 post-activation. ( b) Fold-expansion of T cell\ncultures between day 0-7. ( c) TCF1 protein expression, measured by intracellular flow- cytometry.\nRepresentative histograms of TCF1 protein staining (left) and percentage of TCF1- positive cells (right)\nare shown. Dashed line indicates the threshold for positive cells. ( d) Surface staining for indicated\nproteins, measured by flow-cytometry. Median fluorescence intensity (MFI) of indicated proteins is shown.\nSolid lines connect individual donors. ( e) Surface staining for indicated proteins, measured by flow -\ncytometry. Scatterplots with 2D kernel density contours for representative samples (left) and summarized\npercentages (right) are shown. Dashed lines indicate threshold for positive cells and solid lines connect\nindividual donors. (f) Intracellular staining for indicated cytokines after a 3-hour re-stimulation within anti -\nCD3/CD28 antibodies in the presence of brefeldin A. Scatterplots with 2D kernel density contours for\nrepresentative samples (left) and summarized percentages (right) are shown. ( g) Median fluorescence\nintensity of cytokine positive populations shown in (f). ( h) Granzyme B expression 7 days post-activation.\nRepresentative histograms (left) and summarized MFI (right) are shown. ( i) Experimental setup for co -\nculture assay of T cells with mel888 melanoma cells that express a membrane tethered single- domain\nantibody directed against CD3. (j) Intracellular cytokine staining after a 16-hour co-culture in the presence\nof brefeldin A (added 15 minutes after start co-culture). Dots indicate individual donors. ( k) Percentage\npositive T cells for indicated surface markers of activation after a 48-hour co-culture. Dots indicate\nindividual donors. ( l) Percentage of viable mel888 cells after a 48-hour co- culture. Solid lines connect\nindividual donors, and dashed lines indicate 100% survival relative to wild type mel888 cells. Displayed\ntal \nnd \nell \nry. \nht) \ned \nn. \n-\ned \nct \n-\nfor \nce \nn. \n-\nin \nce \nge \nte \nct \ned \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\ndata is compiled from of 2 independent experiments comprising 3-6 donors. P values indicated in (c-h) \nwere calculated using a two-tailed paired Student’s t-test followed by Benjamini-Hochberg correction. \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \nSupplementary Figure 1. LC-MS dataset of T cell activation time course ( a-b) Number of identified\nproteins (a) and intensity value (LFQ) (b) at indicated activation time points of CD3/CD28-activated T cells\nfrom LC-MS data.  (c) Principal component analysis of LC-MS data. Treatment conditions are indicated by\ncolor, donor pools (experimental replicates) by shapes. ( d) Clustering analysis of all differentially\nexpressed proteins from the LC-MS data. ( e) Differentially expressed proteins of LC-MS data, comparing\neach activation time point (2, 4, 6, and 24 hours) to the 0- hour time point. Colored numbers indicate the\nnumber of proteins that pass the significance threshold (adjusted P < 0.05). \n \n \ned \nlls \nby \nlly \nng \nhe \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \nSupplementary Figure 2. mRNA-seq dataset of T cell activation time-course. (a) Clustering analysis\nof all differentially expressed genes of activation time points (0, 2, 3, 4, 6 and 24 hours) from the RNA-seq\ndata. (b) Differentially expressed genes of RNA- seq data, comparing the activated time points (2, 4, 6,\nand 24 hours) to the 0-hour time point. Colored numbers indicate the number of transcripts that pass the\nsignificance threshold (adjusted P < 0.05). (c) Differentially expressed transcripts (n = 1,487; FDR < 0.05)\nwere clustered into five temporal expression modules using hierarchical clustering. See Supplementary\nTable 7 for cluster annotation.  Grey lines indicate individual transcripts; colored lines indicate medians.\n(d) Enrichment of Reactome pathways in each module. A maximum of 4 pathways is shown for each\ncluster. See Supplementary Table 8 for full list. \n \n \n \n \nis \neq \n 6, \nhe \n5) \nry \ns. \nch \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \nSupplementary Figure 3. Pathway analysis of mRNA-protein association clusters.  E nrichment of\nReactome pathways in each cluster defined in Figure 1h-i. A maximum of 15 pathways is shown for each\ncluster. See Supplementary Table 9 for full list. \n \nof \nch \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \nSupplementary Figure 4. LC-MS dataset of T cell activation with mTOR inhibition. (a-b) Number of\nidentified proteins (a) and corresponding LFQ intensity value (b) of Torin- 1 treated T cells, at indicated\ntime points from LC-MS data.  (c) Principal component analysis of LC-MS data. Treatment conditions are\nindicated by color, time points by shapes. ( d) Clustering analysis of all differentially expressed proteins\nfrom the LC-MS data, comparing Torin-1 treated (purple) with the DMSO control (grey) samples. ( e)\nDifferentially expressed proteins of LC-MS data, comparing activation time points (6 and 24 hours) to the\n0-hour time point within each treatment condition. ( f-g) Teff cells re-activated with anti- CD3/CD28 for 24\nhours in the presence of Torin-1 or DMSO control. ( f) Intracellular staining for p-4E-BP1 ( Thr36/Thr45\nphosphorylation site), measured by flow-cytometry. Representative histograms of anti-p-4E- BP1\nfluorescence intensity (a) and median fluorescence intensity (b) are shown. Dashed lines indicate the\nmaximum density of fluorescence for the DMSO controls and solid lines connect individual donors. ( g)\nIntracellular staining of p-AKT (Ser473 phosphorylation site), measured by flow-cytometry. Representative\nhistogram of anti-p- AKT fluorescence intensity (c) and median fluorescence intensity (d) are shown.\nDashed lines indicate the maximum density of fluorescence for the DMSO controls and solid lines\nconnect individual donors. Displayed data in (f- g) is representative of 2 independent experiments\ncomprising 3 pools of 5 donors. P values indicated in (f-g) were calculated using a two- tailed paired\nStudent’s t-test followed by Benjamini-Hochberg correction. \n \n \n \nof \ned \nre \nns \n) \nhe \n24 \n45 \nP1 \nhe \n) \nve \nn. \nes \nts \ned \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint \n\n \n \nSupplementary Figure 5. Modeling of protein expression modules using XGBoost.  LC- MS data\nfrom Figure 1 were used to train a XGBoost classifier for the 5 indicated protein expression modules. This\nmodel was then used to classify the DMSO/Torin-1 LC-MS data (Figure 2). ( a) XGBoost model\nperformance metrics during resampling. ( b) Mode l performance as receiver operating characteristic\n(ROC) curves. Diagonal back lines denote random predictions. \n \n \n \n \n \n \n \n \n \n \n \nta \nis \nel \ntic \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 2, 2026. ; https://doi.org/10.64898/2026.01.29.702520doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}