Autoregulated splicing of TRA2β programs T cell fate in response to antigen-receptor stimulation

Abstract

T cell receptor (TCR) sensitivity to peptide-MHC dictates T cell fate. Canonical models of TCR sensitivity cannot be fully explained by transcriptional regulation. We demonstrated a post-transcriptional regulatory mechanism of TCR sensitivity through an evolutionarily ultraconserved poison exon (PE) in the RNA-binding protein (RBP) TRA2β that guided alternative splicing of TCR signaling transcripts. TRA2β-PE splicing, seen during cancer and infection, was required for TCR-induced effector T cell expansion and function. Tra2β-PE skipping enhanced T cell response to antigen by increasing TCR sensitivity. As antigen levels decrease, Tra2β-PE re-inclusion allowed T cell survival. Finally, we found TRA2β-PE was first included in the genome of jawed vertebrates capable of TCR gene rearrangements. We propose that TRA2β-PE splicing acts as a gatekeeper of TCR sensitivity to shape T cell fate.

One-Sentence Summary:

T cells use splicing of TRA2β’s ultraconserved poison exon as a gatekeeper of TCR sensitivity to produce durable immunity.

Main Text:

During immune responses, T cell clonal expansion, effector differentiation, and survival is induced in response to antigen-derived peptides, presented by major histocompatibility complex (MHC) molecules, that are recognized by the T cell receptor (TCR). This foundational immune defense process uses signal transduction, gene transcription, and regulation of translation as instrumental steps to enact the consequences of TCR stimulation (1, 2). How T cells regulate TCR sensitivity remains an open question. TCR sensitivity is defined by antigen affinity, duration of antigen exposure, and the intensity of stimulation required to execute specific T cell functions. TCR stimulation is relayed through signaling and changes in gene expression that influence functional T cell responses. Although post-transcriptional mechanisms have been linked to controlling T cells (3), their role in regulating TCR responses is poorly understood

Recent work points to a fundamental role for post-transcriptional regulation by RNA-binding proteins (RBPs) as regulators of T cell function (4–6). While RBPs are ubiquitous across tissues, expression differences and isoform-level regulation cause tissue-specific changes that remain an intensive area of study (7). Our recent findings showed that costimulatory signals induce RBPs to carry out post-transcriptional alternative splicing to generate specific RNA isoform changes that optimize CD8⁺ T cell effector function (8). Interruption of RBP-induced alternative splicing disrupts T cell activation: thus, TCR sensitivity might be regulated post-transcriptionally by RBP-mediated RNA splicing.

RBPs are among the most highly conserved genes across 240 mammalian species (9). Many RBPs use a self-regulatory machinery termed ultraconserved poison exon (PE) elements. Ultraconserved PE elements are defined as >200bp regions of 100% sequence identity among human, mouse, and rat (10). PEs are included via alternatively splicing to introduce a stop-codon that causes premature translation termination and nonsense-mediated decay of their RBP transcript (11). PE inclusion thus prevents RBP protein expression, while PE skipping allows RBP protein expression. Despite their ultraconservation warranting mechanistic investigation, our current understanding of PE function is extremely limited. PE elements were only recently shown to affect cerebral cortex formation (12) and impact RNA splicing during tumorigenesis (13–15). The role of ultraconserved PE elements in the immune response is unknown. Therefore, we investigated the role of RBPs in T cell activation, and focused our study on TRA2β, which controls its own transcript levels using a spliced ultraconserved PE.

Results:

CD8⁺ T cell activation induces splicing of ultraconserved poison exons in RBPs

To understand the function of RBPs in regulating CD8⁺ T cell responses, we used an established in vivo CRISPR/Cas9 screen (8, 16) to knockout 10 RBPs we previously found (8) upregulated in effector CD8⁺ T cells from mice receiving costimulation with anti-OX40 and anti-41BB, stimuli that augment the function of effector T cells and are pathways targeted in human cancer immunotherapy (17). Naïve CD8⁺ T cells were electroporated with CRISPR guide RNA, adoptively transferred into recipient mice, activated with antigen and costimulation, and collected after 4 days (Fig. 1A). Unlike the other five RBPs that suppressed the expansion of T cells in our screen (Fig. 1B), Tra2β and Srsf7 control their transcript levels through an ultraconserved PE (18) that introduces a premature stop codon via alternative splicing to trigger nonsense-mediated decay (NMD) and reduce protein levels when included (Fig. 1C).

Fig. 1. CD8⁺ T cell activation induces differential splicing of ultraconserved poison exon elements in RBPs.

(A) Model of in vivo CRISPR/Cas9 mediated RBP knockout and adoptive transfer of CD8⁺ T cells after receiving antigen (SIINFEKL) with costimulation (anti-OX40, anti-41BB).

(B) Percent OT-I, CD45.1⁺ cells of CD8⁺ T cell from spleen after 4 days. Data were pooled from two independent experiments. Each data point represents data from an individual mouse, n≥3 per group.

(C) Splicing of a poison exon (PE) element in an SR RBP.

(D) Representative read alignment plot of Tra2β’s PE from naïve (no stimulus) or effector (antigen with costimulation) CD8⁺ T cells (GSE200236; RNA-seq samples described in (8), n≥5 per group).

(E) Percent spliced in (PSI) of Tra2β’s and Srsf7’s PE’s between naïve and effector cells determined by rMATS analysis (GSE200236, GSE200237; RNA-seq samples described in (8)). Each dot represents RNA from CD8⁺ T cells from one mouse, n≥5/group.

(F) PSI of Tra2β’s and Srsf7’s PE’s across CD8⁺ T cells collected from mice over the course an acute Listeria infection (GSE89307). Each dot represents RNA from one mouse, n=3 per group.

(G) PSI of TRA2β’s and SRSF7’s PE’s between human naïve (up 13 years post vaccination), effector (14 days post vaccination), and memory (4–13 years post-vaccination) CD8⁺ T cells after immunization with Yellow Fever Virus (YFV) vaccine determined by rMATS analysis (GSE100745), n>3 per group where each dot represents data from an individual.

(H) RT-PCR of TRA2β’s and SRSF7’s PE’s naïve and effector (CD3/CD28 stimulated for 48 hours) CD8⁺ T cells from human PBMCs. SRSF7 contains multiple open reading frames that lead to production of several transcripts including its PE (61).

(I) PSI of Tra2β’s and Srsf7’s PE’s between T cells from patients with acute lymphocytic leukemia (ALL) and CD3⁺ T cell from peripheral blood of healthy controls. Each dot represents RNA from n = 3 patients (GSE139622).

(J) Mean fluorescence intensity of TRA2β protein between naïve and effector CD8⁺ T cells from human PBMCs.

For (B, F, G) bars show mean values +/− SD using an Ordinary one-way ANOVA with Dunnett (B) or Sidak’s correction (F,G) for multiple comparisons. For (E, I, J), bars show mean values +/− SD using an unpaired two-sided t test. *P <0.05, **P <0.005.

Given that ultraconserved PE elements have high conservation, control transcript levels, and regulate RBPs that suppressed T cell clonal expansion, we hypothesized that splicing of ultraconserved PE elements was a crucial mechanism governing the duration and intensity of TCR signals. Thus, we analyzed two published RNA-seq datasets (8) of ex-vivo naïve and effector CD8⁺ T cells specifically for splicing changes in these PE regions (Fig. 1D, E, Data File S1). Effector cells showed differential skipping of PEs from Srsf7 and Tra2β, among five other SR RBP family members (Fig. 1D, E, Fig. S1A). Furthermore, the PEs of Srsf7 and Tra2β were both differentially skipped in effector CD8⁺ T cells in RNA-seq data analyzed from mice with acute Listeria infection at day 5 (19), but re-included in day 7 effector and memory CD8⁺ T cells (Fig. 1F, Data File S2). The same pattern of splicing was observed in RNA-seq data comparing human naïve, effector CD8⁺ T cells, and memory CD8⁺ T cells after vaccination with Yellow Fever Virus (YFV) vaccine (Fig. 1G, Data File S3) (20). These data demonstrated that YFV effector CD8⁺ T cells (14 days post vaccination) skipped PEs of Srsf7 and Tra2β compared to naïve cells, while both PEs were re-included in antigen-specific YFV memory cells 4–13 years after vaccination.

In order to test whether TCR engagement of human T cells induced differential PE splicing, we stimulated human PBMCs with anti-CD3/CD28 and analyzed splicing of PE elements by semi-quantitative PCR and RNA-seq to test whether TCR engagement of T cells induced differential PE splicing. Skipping of PE elements in effector CD8⁺ T cells occurred in SRSF7 and TRA2β, as well as in five other SR family members (Fig. 1H, Fig. S1B). To test if splicing of the PE occurs in human disease, we compared PE splicing in T cells of patients with acute lymphocytic leukemia (ALL) to normal CD3⁺ cells from healthy individuals (21). We found skipping of PE elements in TRA2β and SRSF7 in active ALL (Fig. 1I, Data File S4). These data suggest that early effector T cells in humans and mice skip PEs during physiological and pathological responses.

We additionally re-analyzed published RNA-seq data from control and T cells lacking Srsf1 (Srsf1-KO) after LCMV infection (22) and found an increase in inclusion of Tra2β-PE in Srsf1-KO T cells (Fig. S1C, Data File S5). Given that Srsf1 is known to bind to Tra2β-PE (14) this difference during LCMV infection indicated that Srsf1 was required for Tra2β-PE skipping in T effector cells.

As RBPs Srsf1 and Celf2 have already been implicated in T cell function (23–25), we chose to focus on the function of Tra2β’s PE (Tra2β-PE), whose immunological role was unknown.

TRA2β-PE skipping is required for activation and effector function of CD8⁺ T cells

Based on our findings of increased TRA2β-PE splicing and TRA2β protein expression in human and mouse effector CD8⁺ T cells (Fig. 1E, Fig. 1J, Fig. S1C, Fig. S1D), we hypothesized that the skipping of PE elements was required for T cell activation. Hence, we blocked splicing of TRA2β-PE element in human CD8⁺ T cells (Fig. 2A) using a specific anti-sense oligonucleotide (ASO) targeting the upstream intronic silencer sequence (ISS) of the TRA2β pre-mRNA transcript. This approach prevents skipping of the PE, as previously described in cancer models (14).

Fig. 2. TRA2β poison exon inclusion suppresses blasting, activation, and effector function of CD8⁺ T cells.

(A) Favoring inclusion of TRA2β’s poison exon element (TRA2β-PE) using an antisense oligonucleotide to the intronic splicing silencer region.

(B) RT-PCR of RNA after treatment of human CD8⁺ T cells (sorted from a culture of PBMCs) stimulated with CD3/CD28 with control scrambled or TRA2β-PE targeting antisense oligonucleotide (ASO) at 200nM, 100nM, and 50nM doses, n=2 experiments.

(C) Percent spliced in (PSI) of TRA2β-PE between human naïve CD8⁺ T cells and effector CD8⁺ T cells treated with control scrambled or TRA2β-PE ASO determined by rMATS analysis of RNA-seq data, n=4 human subjects per group (GSE267609).

(D) TRA2β protein expression by mean fluorescence intensity (MFI) after ASO 200nM treatment in human CD8⁺ T cells 48 hours after stimulation, n=3 donor samples/group. Parent population of live, CD8⁺ T cells (naïve or effector by activation marker).

(E, F) Mean fluorescence intensity (MFI) of forward scatter and side scatter representative of CD8⁺ T cell size and granularity two days after treatment with TRA2β-PE targeting ASO.

(G) % CD25 expression on CD8⁺ T cells over two days after treatment with 200nM Tra2β PE targeting ASO.

(H) MFI of CD25 as in E.

(I) % PD1 expression on CD8+ T cells over two days after treatment with TRA2-PE targeting ASO.

(J) % CD69 expression on CD8⁺ T cells over two days after treatment with Tra2β PE targeting ASO.

(E-J) were all conducted with 200nM ASO with an n=7 representative of 3 independent experiments.

(K,L) Relative expression of IFNγ and TNFα after restimulation of 20,000 CD8⁺ T cells treated with TRA2β-PE targeting ASO (200nM dose) or scrambled control for 48 hours with PMA and ionomycin (PMAi), n= 3 donor samples/group.

For (C, E-J) bars show mean values +/− SD using an Ordinary one-way ANOVA with Sidak’s correction for multiple comparisons For (D, K-L), bars show mean values +/− SD using an unpaired two-sided t test. *P <005, **P <0.005

We stimulated ASO-treated healthy donor-derived PBMCs with anti-CD3/CD28 and IL-2, and after 48 h, sorted live, CD8⁺ T cells. Semi-quantitative PCR and RNA-seq showed suppression of TRA2β-PE skipping by the ASO compared to scrambled control in a dose dependent manner (Fig. 2B, C, Data File S6) correlating with decreased abundance of TRA2β protein (Fig. 2D, S2A). These changes took place without a difference in total transcript between control and TRA2β-PE ASO effector cells, as determined from RNA-seq data (Fig. S2B).

The TRA2β-PE ASO reduced T cell activation markers (CD25), inhibited blasting (T cell size) and granularity (side scatter area) (Fig. 2E, F). The percent of CD25⁺ CD8⁺ T cells as well as the CD25 mean fluorescence intensity (MFI) was reduced after promoting PE inclusion at both 24- and 48-h (Fig. 2G, H). In addition, the upregulation of other markers of T cell activation, PD1 and CD69, were impaired 48-h after activation (Fig. 2I, J). Production of IFNγ and TNFα by activated CD8⁺ T cells following restimulation was almost completely suppressed in cells treated with TRA2β-PE ASO (Fig. 2K, L). Similar results were seen for human CD4⁺ T cells (Fig. S3). Pathway analysis of transcripts different between control and TRA2β-PE ASO treated effector CD8⁺ T cells revealed that the TCR signaling pathway was among the top 10 pathways differentially expressed between the two groups (Fig. S2C). These data indicated that skipping of TRA2β-PE was required for production of TRA2β protein and the activation, expansion, and cytokine production by human CD8⁺ and CD4⁺ T cells.

TCR-triggered mTOR signaling induces skipping of Tra2β-PE in CD8⁺ T effector cells

To understand how TCR-activation could lead to skipping of Tra2β-PE, we considered mTOR, which is regulated by TCR signal strength (26, 27). OT-I CD8⁺ T cells express an engineered TCR that responds to the ovalbumin peptide (SIINFEKL, also known as N4), and derivative variants, and provides an experimental system to test responses to different strengths of antigen binding (28). The G4 variant (SIIGFEKL), is a very low affinity antigen for the OT-I TCR, and generally elicits weaker effector T cell responses compared to high affinity antigens, such as N4 (29). We tested TCR stimulation of antigen-specific OT-I CD8⁺ T cells to the very low affinity G4 and high affinity N4 antigens to examine Tra2β-PE skipping. We found that Tra2β-PE was skipped by 12 hours post TCR induction (Fig. S4A). Inhibition of mTOR with either Torin or rapamycin suppressed S6 (pS6) phosphorylation (Fig. S4B), and suppressed skipping of Tra2β-PE 12 hours after T cell activation (Fig. S4C) correlating with decreased production of TNFα and IFNγ (Fig. S4D, E). We confirmed that Tra2β-PE was more included in gene transcripts after Torin inhibition using RNA-sequencing (Fig. S4F, Data File S7). Moreover, the spliceosome was the 7^th most differentially expressed pathway between control and Torin treated CD8⁺ T cells (Fig. S4G, Data File S8). Our data, thus, suggest that TCR-triggered mTOR signaling induced Tra2β-PE skipping.

Tra2β regulates T effector cell expansion through spliced transcripts encoding proteins involved in T cell signaling

To determine the genes that may be controlled by TRA2β protein in CD8⁺ T cells during immune responses in vivo, we analyzed of Tra2β knockout CD8⁺ T cells following adoptive transfer (Fig. S5A, as in Fig. 1B). Consistent with our findings of T cells activated in vitro, IFNγ production was suppressed in antigen-specific Tra2β knockout CD8⁺ T cells (Fig. 3A). RNA-seq analysis with Whippet identified 1559 differentially spliced sites dependent on Tra2β, with 822 alternative splicing (AS) changes (Fig. 3B, Data File S9) (30). AS changes were primarily composed of 720 cassette exon splicing (CE) changes in 547 genes, making up 88% of total events (Fig. 3C).

Fig. 3. Tra2β regulates T effector cell expansion and cytokine production through alternative splicing of TCR signaling pathway genes.

(A) IFNγ levels of mouse effector CD8⁺ T cells by ELISA 20 h after restimulation with PMAi between Ctrl and Tra2β KO groups, n = 7 mice/group, each data point shows data from one mouse.

(B) Deep RNA-seq of analyzed with Whippet of Control and Tra2β KO CD8⁺ T cells receiving antigen with costimulation (anti-OX40, anti-41BB) (Ag+Co) that were collected after 3.5 days in vivo. Filtering included probability >0.9, Percent-spliced in (Psi) difference >0.05 or <−0.05; total reads >50; 100 splicing events shown, n ≥2 mice/group, one line shows data from one mouse (GSE267609).

(C) Splicing events from (B) split by type.

(D) PSI levels of Slamf6 exon 7 spliced between naïve, effector control (Ctrl gRNA), and effector Tra2β KO CD8⁺ T cells, n≥2 mice/group, individual data points show data from a single mouse (GSE267609).

(E) RT-PCR validation of Slamf6 exon 7 splicing in D. Cut image between lane 8 and 9 due to gel size, please see Fig. S18A.

(F) Model of in vivo co-transfer CRISPR/Cas9 mediated Slamf6 exon 7 isoform-specific knockout and adoptive transfer of CD8⁺ T cells. Mice received antigen with costimulation (anti-OX40, anti-41BB) (Ag+Co) and T cells were tracked over 7 days in vivo.

(G) Percent OT-I, CD45.1⁺ (KO of Slamf6 Exon 7) vs CD45.1⁺/CD45.2⁺ (Control) cells as a proportion of CD8⁺ T cell from blood 4 and 6 days after receiving Ag+Co. Each data point represents a biological replicate, n=9 mice per group, 2 independent experiments.

(H) Validation of Slamf6 exon 7 CRISPR/Cas9-induced KO from splenic CD8⁺ T cells at day 7 after transfer, n=2 experiments.

(I) IFNγ levels of mouse CD8⁺ T cells by ELISA 20 h after restimulation with PMAI between Ctrl and Slamf6 Exon 7 KO groups, n = 9/ group, 2 independent experiments.

(J) CRISPR screen of 23 candidate Tra2β-dependent splice sites with Tra2β CLIP-seq binding sites in their genes from effector CD8⁺ T cells. OT-I CD8⁺ T cells receiving either scrambled control or gRNA specific for the splice site exon indicated were co-cultured with either high affinity SIINFEKL(N4) or low affinity SIITFEKL (T4) peptide for 72 hours. Control and knockout cells were distinguished by the expression of CD45.1⁺ or CD45.2⁺. △% expansion values are indicative of a delta change between end point (72h effector cells) and start point (0h naïve) as well as a delta from knockout to control. For example, if knockout-to-control ratio was 50–50 at 0h, then became 40–60 at 72h, this is expressed as a difference of −20%. TNFα and IFNγ values are representative of a delta from knockout to control of %TNFα⁺ and %IFNγ⁺ CD8⁺ T cells after 4h restimulation with PMAi, n=1 experiment. Source Data are in Data File S11.

(K, L) Pathway analysis of Tra2β CLIP-seq binding sites from naïve (K) or effector (L) CD8⁺ T cells using the strict threshold set by Eclipse Bio of log2 fold change ≥ 3 and −log10(p-value) ≥ 3, n=1 experiment (GSE267609). Source Data are in Data File S12.

(M) TCR signaling pathway genes with Tra2β CLIP-seq binding sites from naïve (K) or effector (L) CD8⁺ T cells. Source Data are in Data File S12.

(N) Top 3 binding motifs ranked by p value determined by HOMER from Tra2β CLIP-seq of naïve (K) or effector (L) CD8⁺ T cells (33).

For (D, G), bars show mean values +/− SD using an Ordinary one-way (D) or two-way (G) ANOVA with Sidak’s correction for multiple comparisons. For (A,I), bars show mean values +/− SD using an unpaired two-sided t test. *P <0.05, **P <0.005

In order to identify splicing events induced by TCR signaling and costimulation as well as regulated by Tra2β, we compared the 720 Tra2β KO CE changes with 404 CEs that were differentially spliced between naïve and control effector CD8⁺ T cells (Fig. S5B). This revealed 15 genes with CE changes in both groups (Fig. S5C). We further validated cassette exon splicing changes using a second algorithm, rMATS (31). On filtering, we detected 62 differentially spliced cassette exons in 57 genes regulated by TRA2β (Fig. S6A, B, Data File S10). Moreover, 9 of the 15 events from Whippet analysis (Fig. S5C) were also found in the 62 events from rMATS analysis (Fig. S6A, B).

One of these 9 spliced exons, Slamf6 exon 7, was preferentially skipped in the Tra2β knockout CD8⁺ T cells (Fig. 3D), which we validated by RT-PCR (Fig. 3E). This isoform missing Slamf6 exon 7, also known as Ly108-H1, was shown to inhibit CD4⁺ T cell proliferation while ameliorating systemic lupus erythematosus (32). To determine if Tra2β-dependent splicing of Slamf6 exon 7 impacted activation of CD8⁺ T cells, we co-transferred CD8⁺ T cells that were pretreated with Cas9 complexed with control gRNA or gRNA specific for exon 7 (Fig. 3F). Slamf6 exon 7 knockout CD8⁺ T cells stimulated with antigen plus anti-OX40/−41BB reduced cell numbers on days 4 and 6 compared to control (Fig. 3G), mirroring the decrease in expansion of Tra2β KO in mouse T cells (Fig. 1B). Isoform specific knockout cells had 40% less Slamf6 exon 7-containing transcript (Fig. 3H) and reduced surface protein expression of Slamf6 (not isoform-specific) (Fig. S7A). Restimulation of day 7 effector CD8⁺ T cells with PMAi ex vivo showed no difference in IFNγ production (Fig. 3I). Similarly, in vitro data showed reduced cell accumulation beginning at 48 hours of Slamf6 exon 7 knockout CD8⁺ T cells (Fig. S7B, C), and restimulation of these cells with PMAi also did not affect intracellular production of IFNγ (Fig. S7D). Thus, Slamf6 exon 7 deletion inhibited expansion but not IFNγ production, which we attributed to a different function of Tra2β.

To investigate whether the other candidate cassette exon splicing events regulated T cell effector function, we conducted an isoform-specific CRISPR screen. We used a co-culture assay to compare OT-I CD8⁺ T cells receiving either scrambled control gRNA or gRNA that knocked out a specific transcript isoform containing one of 61 candidate spliced exons. To assess TCR sensitivity, co-cultured T cells were stimulated with either high affinity N4 or low affinity (SIITFEKL) T4 peptides and harvested after 72 hours. Unlike the very low affinity G4 peptide used previously for short time course stimulation (Fig. S4A), T4 peptide elicits a stronger expansion over several days allowing for measurable T cell responses in culture, albeit still less than the N4 peptide.

A range in responses was evident depending on the splicing event, but the majority of the 61 candidates affected CD8⁺ T cell expansion and/or IFNγ and TNFα cytokine production upon restimulation with PMAi (Fig. 3J, Data File S11). To investigate if these spliced events resulted from direct Tra2β-binding, CLIP-seq analysis of Tra2β in effector CD8⁺ T cells was performed. We found 23 out of the 56 genes (in addition to Slamf6) with Tra2β-dependent spliced exons contained Tra2β binding peaks (Fig. S6A, B red, C, D, Data File S12). Among the 23 genes that had CLIP peaks, Lrmp-exon 8 and Ikzf1-exon 4 knockout enhanced expansion the most (22%) while Ptprc-exon 6 knockout suppressed expansion (−42.8%) compared to control CD8⁺ T cells (Fig. 3J). Importantly, many isoform changes also had specific effects depending on peptide affinity and cytokine released. Ikzf1-exon 4 skipping enhanced IFNγ production but only in response to lower T4 peptide affinity, while Tcf7-exon 5 skipping suppressed TNFα production only in the stronger N4 condition (Fig. 3J).

To understand the role of TRA2β immediately before antigen stimulation, we conducted CLIP-seq against TRA2β in ex vivo naïve mouse CD8⁺ T cells. Unlike effector T cells, naïve cells exist in a relatively quiescent state poised to be triggered by specific antigen through their TCR. We, thus, tested whether their TCR signaling pathways were influenced by TRA2β prior to specific antigen exposure. Transcripts that bound to TRA2β in naïve cells were enriched in those associated with the TCR signaling pathway (Fig. 3K). TRA2β bound to TCR signaling factors essential for early activation, namely Cd3, Cd8a, Zap70, Lat, Lck and 18 others (Fig. 3M, top circle, Data File S12). Tra2β-bound transcripts from effector T cells were also enriched in TCR signaling components (Fig. 3L). Notably, there were 14 shared TCR signaling genes with TRA2β binding sites in both naïve and effector CD8⁺ T cells, including Nfkb1, Bcl10, and Ptprc (Fig. 3M, intersection). However, some transcripts associated with downstream TCR signaling were only associated with TRA2β in effector T cells: these included transcription factors like Nfatc1, Kras, and Jun among 14 others (Fig. 3M bottom circle, Data File S12). These data suggest that TRA2β may control genes of the TCR pathway both at the instant of antigen-TCR stimulation and during further CD8⁺ T cell activation and proliferation.

A dual role in controlling early and late genes of the TCR pathway might be partially explained by a change of binding motif. Using HOMER to identify motifs in CLIP-seq peaks enriched for TRA2β binding revealed the same top enriched binding motif in both naïve and effector CLIP-seq groups (GANGANGA) (33). However the second most enriched motif (UGAACUGGAUGA naïve, vs. GGAGCAGCUC effector) and the third most enriched motif (CUUCUGGA naïve, vs. CUUCAGGAGG) differed (Fig. 3N).

Altogether, our data suggest that, in CD8⁺ T cells, TRA2β protein regulated gene transcripts involved in the TCR signaling pathway through splicing (Fig. 3J) and direct binding of RNA (Fig. 3K,L,M).

Tra2β-PE skipping alters CD8⁺ T cell responses to transient and prolonged antigen availability

Our data suggested that skipping of Tra2β-PE in the transcript was important for increasing TRA2β protein expression upon TCR stimulation of naïve T cells, which was required for T cell activation and differentiation in response to antigen. Conversely, when Tra2β-PE was included in the mRNA, the production of TRA2β protein was reduced and TCR responses were impaired.

However, we previously observed that Tra2β-PE was re-included in later stage effector cells and in memory CD8⁺ T cells responding to Listeria infection (Fig. 1F). We reasoned that if splicing of Tra2β-PE in T effector cells was antigen dependent, re-inclusion of Tra2β-PE should be required for long-term survival of activated T cells after antigen has been cleared. To test this, we employed two separate models of antigen triggering which differed in the duration that antigen was present after the initial stimulation and, thus, altered the total number of surviving T cells over time as measured by clonal expansion and cell accumulation (Fig. 4A). In these settings, we compared control or Tra2β-PE-deficient (PE-KO) OT-I CD8⁺ T cells, which increased TRA2β expression (Fig. S8A), that were co-transferred in equal numbers into the same recipient mice. By deleting the PE, we could thus inhibit re-inclusion once antigen-mediated TCR signaling was lost.

Fig. 4. Tra2β-PE dictates response to antigen sensing in CD8⁺ T cells.

(A) Control or Tra2β-PE knockout (PE-KO) CD8⁺ T cells were adoptively transferred into recipient mice at a 1:1 ratio and either received SIINFEKL antigen and costimulation (anti-OX40, anti-41BB) (antigen-bolus model) or OVA-X31 Influenza virus (antigen-prolonged model). Expansion of total antigen-specific CD8⁺ T cells in the blood after adoptive transfer in the antigen-bolus model, n≥7, mice/group, n=3 experiments. Expansion of total antigen-specific CD8⁺ T cells in the blood after adoptive transfer in the antigen-prolonged model, n=10 mice/group, n=2 experiments.

(B) Splicing of Tra2β-PE from naïve to day 11 effector CD8⁺ T cells as measured by RT-PCR in the antigen-bolus model, n= 2 experiments.

(C) Relative expansion of Control and PE-KO antigen-specific CD8⁺ T cells in the blood in the antigen-bolus model, n≥7 mice/group, n=3 experiments.

(D) Relative expansion of Control and PE-KO CD8⁺ T cells in the spleen the axillary and inguinal (draining) lymph nodes in the antigen-bolus model including both high (N4) and low (T4) affinity ligand experiments as well as naïve CD8⁺ T cells, n= 8 mice/group for N4/T4, n≥4 mice/group for naïve.

(E) Relative expansion of Control and PE-KO CD8⁺ T cells in the blood in the antigen-prolonged model, n≥7 mice/group, n=3 experiments.

(F, G, H) Relative expansion of CD8⁺ T cells in mediastinal (draining) lymph nodes, lungs, and peripheral (non-draining) lymph nodes at day 16, 27, and 47 post-infection from the antigen-prolonged model, n≥8 mice/group, n=2 experiments. Parent population of live, CD8⁺ antigen-specific, congenically marked T cells.

(I) Percent spliced in (PSI) of TRA2β-PE in CD8+ T cells gathered from peripheral blood of PD1 immunotherapy-responding and nonresponding patients with melanoma (GSE171256). The prognosis model to predict response of patients to immune checkpoint inhibitor therapy was built using PD-1 and the top 20 genes of the PD-1 co-expression genes (21 predictor genes in total) that were most correlated (most significant P values) with PD-1 indicative of an exhausted state, n=13+ human subjects/group.

(J) Relative expansion of Control and PE-KO CD8⁺ T cells in the blood in low affinity (T4) antigen-bolus model, n=8 mice/group, n=2 experiments.

In (C), (D) and (J) 50% expansion means the total antigen-specific population is composed of 50% Control and 50% PE-KO CD8⁺ T cells.

For (A, C-H, J) bars show mean values +/− SD using an Ordinary one-way (A) or two-way (C-H, J) ANOVA with Sidak’s correction for multiple comparisons. For (I), bars show mean values +/− SD using an unpaired, nonparametric two-sided Mann-Whitney t test. *P <0.05, **P <0.005

Following immunization of mice with SIINFEKL and co-stimulation, Tra2β-PE was skipped in control CD8 T cells at the early effector stage (day 5), but was present in the transcript by day 11 post-stimulation (Fig. 4B). During the early effector response, both control and Tra2β-PE-deficient CD8 T cells expanded proportionally (Fig. 4C). However, at day 11 there were proportionally fewer PE-KO T cells in the blood (Fig. 4C) which correlated with re-expression of the Tra2β-PE in control cells (Fig. 4B). At this time point, antigen had been cleared and, therefore, T cells are less likely to be exposed to ligands that would induce TCR signaling. At day 11 post-stimulation, the proportion of PE-KO CD8⁺ T cells in the spleen and draining lymph nodes was reduced compared to control (Fig. 4D, see N4 spleen, N4 lymph node), however, these differences were dependent on the strength of the initial T cell stimulation. Indeed, we found no reduction in the proportion of naive PE-KO CD8⁺ T cells compared to naïve control CD8⁺ T cells undergoing homeostatic proliferation (Fig. 4D, see Naïve spleen, Naïve lymph node), indicating that Tra2β-PE control of survival was likely dependent on specific antigen even in cases without overt T cell activation.

Thus, during initial activation and the expansion phase when antigen-specific T cells increase in number, Tra2β-PE was skipped. During the contraction phase of the immune response, when effector T cells start to die which coincides with antigen being cleared, Tra2β-PE is included in the transcript again. This suggests a differential regulation of splicing during different phases of the immune response. When Tra2β-PE is forcibly removed, antigen-specific CD8⁺ T cells have a more prominent contraction phase. This suggested that CD8⁺ PE-KO cells could be more dependent on antigen for their survival because they could not re-include Tra2β-PE.

In certain cases, mice infected with influenza A virus exhibit weight loss and prolonged circulating antigen (34, 35). We observed weight loss and detected influenza antigen up to 27 days post infection with X31-OVA influenza A virus encoding the SIINFEKL peptide (Fig. S8B–D). There were proportionally more PE-KO CD8⁺ T cells in the blood 12–16 days post infection (Fig. 4E). This difference was also observed in mediastinal draining lymph nodes as well as in the lung resident CD8⁺ T cells (Fig. 4F, 4G). However, the proportion of control and PE-KO CD8⁺ T cells was similar in peripheral inguinal lymph nodes which are non-antigen draining (Fig. 4H). Furthermore, PE-KO CD8⁺ T cells in lung and mediastinal lymph nodes were proportionally higher at days 27 and 47 (Fig. 4F, G). Thus, initial Tra2β-PE skipping programs the cell for TCR sensing of antigen that subsequently allows expansion. Re-inclusion of the PE into the Tra2β mRNA once antigen is cleared facilitates nonsense mediated decay of its transcript. We propose that the subsequent reduction of TRA2β protein expression results in a new signature of isoform changes and RBPs that reprogram the effector T cell for survival. In cases where antigen never induced T cell activation, as in the antigen-inexperienced T cells seen in the non-draining lymph nodes, inclusion of the PE may not be required for short-term survival.

In situations where antigen is cleared quickly, not being able to re-include the PE is detrimental to survival of activated T cells. However, in circumstances of prolonged antigen availability, skipping of the PE appears to augment T cell numbers for longer. A potential explanation for this is that the boost of Tra2β-PE skipping in responses to antigen might generally be advantageous, which could be maintained after antigen clearance. Overall, the data from these models suggested that T cell persistence, proliferation, and survival in response to antigen availability was controlled by the inclusion and exclusion of Tra2β-PE.

Cells exhibiting an exhausted phenotype are associated with situations of chronic antigen stimulation. We examined published sequencing data of human CD8⁺ T cells isolated from the blood of non-responder (to PD-1 immunotherapy) melanoma patients (36). CD8⁺ T cells from individuals that responded to PD-1 therapy had a higher percentage of TRA2β-PE included in the message (Fig. 4I and Data File S13). Chronic restimulation of human CD8⁺ and CD4⁺ T cells in vitro resulted in greater skipping of TRA2β-PE in exhausted cells (as defined by high TIM3 protein expression) compared to effector T cells, with the exhausted T cells also producing less IFNγ (Fig. S8E, F, G). Thus, in the context of exhaustion, TRA2β-PE skipping is associated with impaired T cell responses such as those seen in immunotherapy non-responder patients.

Tra2β-PE skipping augments the response of T cells to low affinity TCR ligands

Our experiments in the antigen-bolus model indicated that the numbers of T cells that persisted in circulation and tissues after antigen had cleared was influenced by Tra2β-PE inclusion (Fig. 4C,D). Therefore, we examined the kinetics of the expansion of control and PE-KO OT-I CD8⁺ T cells to low affinity SIITFEKL (T4) and high affinity SIINFEKL (N4) antigens. Peak expansion - when the highest frequency of antigen-specific CD8⁺ T cells was detected at day 6 - was lower in mice that received T4 (Fig. S8H) vs. mice that received N4 (Fig. 4A). CD8⁺ T cells without Tra2β’s PE, however, expanded more at days 4 and 6 compared to control CD8⁺ T cells in T4 immunized mice (Fig. 4J). By day 9, however, once antigen disappears, PE-KO CD8⁺ T cells returned to day 0 levels, similar to the contraction seen in N4 models (Fig. 4J). Thus, skipping of Tra2β-PE raises antigen sensitivity to low affinity TCR ligands.

We then investigated how Tra2β transcript splicing was influenced by TCR stimulation by ligands of different affinity using RNA-seq data from mouse CD8⁺ T cells stimulated in vivo. Tra2β-PE was more skipped in CD8⁺ T cells from mice stimulated by N4 peptide compared to T4 peptide (Fig. S8I, Data File S14). This suggested that Tra2β-PE skipping was promoted by increased TCR signaling strength.

We additionally performed pathway analysis on differentially expressed genes between N4 and T4 stimulated CD8⁺ T cells. Genes involved in oxidative phosphorylation and the spliceosome were among those most upregulated in N4-stimulated T cells compared to T4-stimulated T cells (Fig. S8J). The majority of genes encoding RBPs decreased in expression in the T4-stimulated T cells compared to N4-stimulated T cells (Fig. S8J). There was a clear overlap between gene expression changes of spliceosomal RBPs between N4 and T4 stimulation of T cells and N4-stimulated T cells with and without Torin-mediated inhibition of mTORC1 (compare Fig. S8J with S4G). This could suggest an overlapping mechanism. In total, these data indicated that increased TCR antigen sensitivity – defined by the proportional T cell response to antigen affinity and duration of antigen exposure – is programmed by Tra2β-PE skipping during the CD8⁺ T cell response.

Tra2β-PE skipping regulates gene expression associated with TCR and JAK-STAT signaling during immune responses in vivo

To understand how skipping of the PE element in Tra2β promoted increased T cell number in response to antigen, we conducted RNA-seq of control versus PE-KO OT-I CD8⁺ T cells that were adoptively transferred and sorted from the lung interstitium at day 16 post-infection with X31-OVA influenza A. We confirmed that the PE element in Tra2β was excluded to a greater extent in PE-KO (20% PSI) cells compared to control (38% PSI) (Fig. S9A, left panel). Tra2β is known to regulate splicing of Tra2α’s (Tra2-alpha) PE element – accordingly, PE-KO T cells showed greater inclusion of Tra2α’s PE element (20% PSI) compared to control (10% PSI) (Fig. S9A, right panel) (37). We found that 905 genes were differentially expressed, with 806 genes downregulated and 99 genes upregulated in PE-KO T cells compared to controls (Fig. S9B, Data File S15). We observed that the TCR alpha constant chain (Trac) (Fig. 5A) and key TCR signaling mediators (Fig. 5B) were downregulated in PE-KO cells exposed to prolonged antigen stimulation, which was consistent with our findings that TRA2β bound to genes linked to TCR signaling. Pathway analysis between control and PE-KO groups showed that the JAK-STAT pathway, a cornerstone mediator of TCR regulation and activation, was the most enriched pathway for downregulated genes in PE-KO T cells (Fig. 5C) – this included inflammatory mediators, Il13ra1, Ifnγr2 and Jak2 (Fig. 5D) (38). We also observed downregulation of chemokine receptors such as Ccr7 that define central memory cells (Fig. 5E). Genes that were upregulated in PE-KO CD8⁺ T cells were enriched for autophagy, a key survival pathway, as the top pathway (Fig. S9C, D).

Fig. 5. Prolonged skipping of the conserved Tra2β-PE element suppresses TCR and JAK-STAT signaling but enhances cell cycle gene splicing.

(A) Transcript expression levels of key TCR constant chain genes from RNA-seq done on Control and PE-KO lung CD8⁺ T cells from Fig 4E (antigen-prolonged model), n= 4 mice/group where each point is RNA from CD8⁺ T cells from one mouse (GSE267609).

(B) Transcript expression levels of TCR signaling pathway genes as in A, n= 4 mice/group where each point is RNA from CD8⁺ T cells from one mouse (GSE267609).

(C) Pathway analysis of top 806 genes downregulated in PE-KO between control and PE-KO lung CD8⁺ T cells from Fig 4G (GSE267609).

(D) Key downregulated JAK-STAT pathway genes from C, n= 4 mice/group where each point is RNA from CD8⁺ T cells from one mouse.

(E) Key downregulated chemokine signaling pathway genes from C, n= 4 mice/group where each point is RNA from CD8⁺ T cells from one mouse.

(F) Filtering process for splicing events from control and PE-KO CD8⁺ T cells from Fig. 4G.

(G) Percent spliced in of cell cycle regulating RNAs from control and PE-KO CD8⁺ T cells from Fig 4G, n= 4 mice/group where each point is RNA from CD8⁺ T cells from one mouse (GSE267609).

(H) Percent spliced in levels of TRA2β-PE analyzed from human thymic subsets (GSE151081), n=2 humans/group where each point is RNA from the indicated thymic subset from one human.

(I) Pathway analysis using Pathfinder of genes downregulated in ISP thymocytes compared to DP CD3⁺ thymocytes.

(J) MYC RNA expression across human thymic subsets, n=2 humans/group where each point is RNA from the indicated thymic subset from one human.

(K) Transcript expression levels of TCR alpha constant chain (TRAC) in human thymic subsets (GSE151081), n=2 humans/group where each point is RNA from the indicated thymic subset from one human.

(L) Transcript expression levels of recombinase activating gene 1 and 2 (RAG1) in human thymic subsets (GSE151081), n=2 humans/group where each point is RNA from the indicated thymic subset from one human.

(M) Transcript expression levels of recombinase activating gene 2 (RAG2) in human thymic subsets (GSE151081), n=2 humans/group where each point is RNA from the indicated thymic subset from one human.

(N) Percent conservation (determined by BLAST) to human DNA of TRA2β-PE and TRA2β’s exon 4.

(O) Alignment of TRA2β-PE visualized using the UCSC Genome Browser and corresponding TRA2β protein coding element alignment.

For (A,B, D,E), bars show mean values +/− SD using multiple unpaired t tests with Holm-Sidak’s multiple comparison. For (F,H, J-M) bars show mean values +/− SD using an Ordinary one-way ANOVA with Sidak’s correction for multiple comparisons. *P <0.05, **P <0.005, ***P <0.0005, ****P <0.00005 Source data are in Data File S19.

Tra2β-PE skipping regulates cell cycle genes during immune responses in vivo

Our pathway analysis of PE-KO cells did not reveal a change in transcript expression of cell cycle genes. However, we hypothesized that antigen-associated expansion of these CD8⁺ T cell occurred through Tra2β-mediated alternative splicing events. We identified 21 overlapping core exon splicing changes (Fig. 5F, Fig. S10A,B), 4 of which occurred in cell proliferation regulating genes Ccne1, Med24, Ltk, and Rcc1 (Fig. 5G). Another differentially spliced gene was Bclaf1, an apoptotic regulator (Fig. S10C, Data File S9, Data File S18), that was found to be differentially spliced in Tra2β-KO CD8⁺ T cells (Data File S10). These findings suggest that splicing events in genes that control cell cycle in PE-KO CD8⁺ T cells might control proliferation and expansion of CD8⁺ T cells in a Tra2β-dependent manner.

Tra2β-PE skipping correlates with TCR-dependent and proliferative stages of T cell development in the thymus

As Tra2β controlled the differentiation of effector T cells, we hypothesized that this mechanism could be involved in T cell development, where T cells undergo rapid changes in cell cycle and antigen-based selection through TCR signaling. We analyzed RNA-seq data from human thymic subsets and cord-blood (CB)-derived precursor cells (HPCs) and found that Tra2β-PE was differentially skipped during the double negative (DN) stages and was re-included upon reaching the double positive (DP) and single positive stages (SP), with the intermediate single positive (ISP) phase being the inflection point (Fig. 5H, Data File S16) (39). The PE of SRSF7 was also skipped in DN stages and reincluded in DP (Fig. S11A, Data File S16). JAK-STAT and TCR signaling genes were among the most downregulated in ISP, where the PE was most skipped, compared to DP thymocytes, where the PE was included (Fig. 5I). This mirrored our findings with enforced PE skipping in the PE-KO CD8⁺ T cells after influenza infection (Fig. 5C). A difference between the thymic DN phase (PE skipped) and DP phases (PE included) is a switch from general hematopoietic development to TCR-based clonal selection (40). Indeed, pathway analysis for upregulated genes in the ISP (PE skipped) over DPCD3⁺ (PE included) – which defines T cells poised to undergo positive and negative selection through their TCR – determined cell cycle genes among the top 5 upregulated pathways (Fig. S11B) with higher levels of MYC and NOTCH1 expression (Fig. 5J, Fig. S11C). Therefore, skipping of the PE element correlated with stages of prolonged antigen signaling and cell cycling. Indeed, MYC was previously shown to directly correlate with TRA2β protein expression in breast cancer models (41). Notably in human thymic samples, the DP phase also corresponds with expression of TRAC, RAG1, and RAG2, genes essential for TCR diversity that, in turn, determine the magnitude of T cell expansion in response to antigen (Fig. 5K–M).

Analyzing the occurrence of splicing events using single cell RNA-seq data of thymic development also indicated that the PE elements of Tra2β and Srsf7 were more skipped during the early DP (P) phase of T cell development, but were included during late DP (Q) and SP phases (Fig. S12A–J). PE element inclusion similarly negatively correlated with gene expression associated with metaphase signaling and cell cycle control at the single-cell level (Fig. S12K).

Tra2β-PE appeared in the genome in jawed vertebrates

The 276 base pair PE sequence of Tra2β is among the most conserved genomic sequences in evolution (10, 11) that we now show controls TCR sensing of antigen (Fig. 4). Hence, we asked when during evolution an ultraconserved PE sequence was integrated into the pre-existing Tra2β gene. BLAST analysis showed Tra2β-PE was present in Tra2β gene across a breadth of vertebrate genomes species, except in jawless fish (lamprey) (Fig. 5 N, O, Data File S19.) In contrast to the absence of the PE in lamprey genome, coding exons 1, 4, 5, 6, 7, and 9 of TRA2β were still present and conserved (Fig. 5 N, O, Fig. S13A). Amino acid conservation of the entire region showed that TRA2β homologues can be traced back as far as Drosophila (Fig. S13B). Further analysis with Vast-DB, a database of conserved transcript isoforms, showed that Tra2β-PE is present in vertebrates down to zebrafish but not present in Drosophila, agreeing with our genomic findings of the regions after Drosophila but before zebrafish (Fig. S13C) (42). On the contrary, Tra2β coding exon 4 was present in all species including Drosophila (Fig. S13D), confirming that Tra2β-PE was integrated into the pre-existing Tra2β gene. Notably, jawless vertebrates such as lamprey do not have Rag1/Rag2 genes nor have a classical T cell receptor, while jawed vertebrates do (43). Thus, these evolutionary conservation data suggest TRA2β-PE was inserted in the genome as the T cell-mediated adaptive immune system matured.

Discussion:

Our data suggest that the evolutionarily ultraconserved poison exon in Tra2β is skipped in response to antigen stimulation which subsequently increases the expression of TRA2β protein to mediate splicing of TCR signaling genes. We thus propose that TRA2β-PE splicing acts as a gatekeeper of TCR signaling that can control the extent of T cell responses. Specifically, Tra2β-PE skipping (gate opening) allows activation and increases T cell sensitivity to TCR ligands to enhance T cell responses. On the other hand, Tra2β-PE inclusion (gate closing) allows for long-term T cell survival after resolution of the response. While the focus of our study was effector CD8⁺ T cells, we also found that TRA2β-PE skipping was required for human CD4⁺ T cell activation and proliferation.

Canonically, specific antigen recognition by the TCR (signal 1) and costimulation (signal 2) initiate T cell expansion through general transcription factors including Jun, Fos, NF-kB, Notch, Myc, Nfatc1 and others. We previously discovered that TCR and costimulation-induced RNA-binding proteins are required for optimal RNA splicing and effector response (8). Our present data suggest that post-transcriptional control through poison exon elements regulates RBP expression and T cell expansion.

Differential binding of TCR pathway transcripts in naive compared to effector CD8⁺ T cells suggests that in addition to splicing, TCR signaling could alter the interaction between TRA2β and specific RNA motifs. Therefore, it is possible TRA2β undergoes a structural change or acquires new binding partners that allows for a shift in binding sites in both old and newly transcribed genes.

In vivo, we found that enforced skipping of Tra2β-PE augmented CD8⁺ T cell expansion in response to prolonged antigen. However, the same cells decrease in abundance compared to cells that can regulate Tra2β splicing in the antigen-bolus model where antigen is cleared. Further, when triggered with low affinity T4 antigen, PE-KO CD8⁺ T cells expand much more compared to control CD8⁺ T cells. The PE must be reincluded, regardless of antigen affinity, for responding T cells to survive once antigen dissipates. The closest parallel, though transcriptionally regulated, is recently identified signaling checkpoint and CAR-T therapy candidate RASA2, a TCR stimulation-dependent antigen sensitivity attenuator (1).

Chronic antigen stimulation can lead to T cell exhaustion. Based on our analysis of exhausted T cells generated in vitro, and analysis of data from immunotherapy-treated melanoma patients, we propose that PE splicing could be an indicator for T cell exhaustion during long-term T effector responses. Tra2β-PE was more skipped in exhausted, immunotherapy non-responder T cells compared to immunotherapy responder T cells. Moreover, control of Tra2β-PE splicing could contribute to an optimal effector response. Thus, PE splicing might be leveraged to relieve T cell exhaustion.

Our data, therefore, suggest that the Tra2β-PE splicing controls fundamental T cell functions in different ways depending on the duration and strength of antigen-mediated TCR stimulation. During transient antigen exposure, PE skipping favored cytokine production, but in the context of prolonged antigen exposure, PE skipping correlated with suppressed cytokine signaling and promoted pathways associated with proliferation and cell survival.

More broadly, PE elements are required for cell viability and proliferative control in cancers (13, 14). While prior studies showed that many ultraconserved elements have no effect on organismal viability or function, we highlight that PEs are essential facilitators of T cell functional states.

The canonically diverse TCR response appeared between jawless and jawed vertebrates in evolution, replacing the more primitive Variable Lymphocyte Receptors (VLRs) (44). While they can mount an adaptive immune defense, jawless vertebrates such as lampreys have no MHCI/MHCII that mediate antigen-presentation, do not use Rag1/Rag2 to generate antigen receptor diversity, and do not harbor the complex cytokine and interferon orthologs such as IL-2, IL-4, IL-7, and IFNγ (43). Based on our data, Tra2β-PE appears at this same inflection point between jawless and jawed vertebrates.

Tra2β-PE is only one of at least nine serine/arginine-rich (SR) proteins with differentially spliced PEs on TCR stimulation. Alternatively spliced ultraconserved elements are also found in the hnRNP RBP family, which contains essential exon elements (EE) that, when included, produce a translated isoform (18). We found EEs in several hnRNP genes differentially spliced on T cell activation as well as in T-ALL. Several of the SR and hnRNP RBPs are known to control T cell development, effector function, and, as shown here, expansion. As each RBP controls a network of splicing isoform changes, multiple PE or EE elements could be required for control of T effector responses. One key example is that of Srsf7, which, like Tra2β, appears differentially spliced in multiple T cell states. Another intriguing example is that of Srsf1, which is known to control thymic T cell development and autoimmunity, but has an ultraconserved element in its 3’UTR – unlike in other SR genes, inclusion of this element leads to expression of a translated isoform (23, 24). Here, we show that Srsf1 is required for skipping of Tra2β-PE in T effector cells.

We also found that TCR-induced mTOR signaling is required for Tra2β-PE skipping. Prior studies have shown that mTOR signaling regulates SRSF1 through phosphorylation to induce its activity and translocation to the nucleus (45). Thus, we postulate that mTOR-induced regulation of SRSF1 could be a mechanism for Tra2β-PE skipping. Whether there is a master regulator for PE splicing, which isoforms are regulated by other RBPs, and what key T cell functions are controlled by other PE splicing events remains to be investigated.

Understanding regulation of T cell expansion and effector function through RBP regulators such as PE elements during infection and disease could lead to clinical applications. The approaches used in this study, antisense oligonucleotides, CRISPR-based splice switching, or isoform-specific targeting, could be applied to adoptive T cell therapies and mRNA therapeutics used in cancer treatments (1, 46–49). Modifying specific elements and isoforms without permanently impacting overall gene expression would, thus, create tunable therapeutic T cell responses.

Materials and Methods

Mice

All animal studies were performed in accordance with UConn Health (Farmington, CT) Institutional Animal Care and Use Committee (IACUC) regulations and were approved by the committee. Mice were maintained at a maximum of 5 mice per cage, fed regular chow, had nesting material and igloos, and were on a 12 h dark/light cycle at 30–70% humidity and 20–26.1 °C. C57BL/6J CD45.2 (WT) mice were purchased from The Jackson Laboratory (Stock #000664) (Bar Harbor, ME). Ova (SIINFEKL^257–264)-specific OT-I TCR transgenic, recombination activating 1-deficient (Rag1^−/−) or (Rag2^−/−) mice that were WT on C57BL/6J CD45.1 or C57BL/6J CD45.1/2 Het background were bred in house (OT-I CD45.1, OT-I CD45.1/CD45.2) or purchased from Taconic (OT-I CD45.2 Model #2334, Rensselaer, NY). All mice were maintained in the UConn Health Animal Facility in accordance with National Institutes of Health guidelines. Animals were euthanized in accordance with American Veterinary Medical Association (AVMA) and UConn Health IACUC protocol guidelines in a CO2 EuthanEx chamber and by cervical dislocation. Death was confirmed by observation of respiration cessation as well as lack of response to stimuli. All mice used in these studies were between 1.5 and 8 months of age including both female and male sex.

Adoptive transfer and immunizations

Naïve splenic OT-I cells (6 × 10⁵ viable CD8⁺ TCR Vα2⁺/Vβ5⁺ CD45.1⁺ or CD45.1⁺/CD45.2⁺) were intravenously transferred into WT C57BL/6J CD45.2⁺ recipients. For co-transfer PE-KO experiments with antigen-bolus, 6 × 10⁵ of each CD45.1⁺ and CD45.1⁺/CD45.2⁺ OT-I cells were transferred. Recipient mice were injected intraperitoneally with 100 μg SIINFEKL^257–264 peptide (InvivoGen, San Diego, CA) or with 100ug SIITFEKL^257–264 peptide (Anaspec, Fremont, CA) and with agonist anti-CD134 antibody (20ug) (Clone OX-86 mAb; Bio X Cell, Lebanon NH) and agonist anti-CD137 antibody (10ug) (Clone 3H3; Bio X Cell) (Ag+Costim group) 4–6 h after adoptive transfer. Flow cytometry was used to calculate percentage of CD8⁺ CD45.1⁺, Vα2⁺, Vβ5⁺ of total CD8⁺ T cell from spleen and lymph nodes prior to sorting. Cells from spleen and lymph nodes were sorted for CD8⁺ CD45.1⁺, TCR Vα2⁺, Vβ5⁺ markers 3 days and 12 h after stimulation by flow cytometry for >95% purity prior to use in downstream assays.

For influenza co-transfer experiments, X31-OVA virus stocks were grown in fertilized chicken eggs (Charles River, Wilmington MA). Naïve splenic OT-I cells (5 × 10³ viable CD8⁺ TCR Vα2⁺ Vβ5⁺ of each CD45.1⁺ and CD45.1⁺/CD45.2⁺) were intravenously transferred into WT C56BL/6J recipients and 24 hours after transfer, mice were infected with 2000 plaque-forming units (PFU) X31-OVA. Tail vein blood, mediastinal lymph nodes, and lungs were obtained as described in figure legends. Mice were injected with intravenous-anti-CD8b-PE antibody (Thermo Fisher, Catalog #12–0083-82, Clone H35–17.2, Waltham, MA) just before harvest at days 16, 27, and 47 to distinguish between resident and circulating populations.

Flow cytometry

Surface staining: naïve OT-I splenocytes for adoptive transfer were identified as: LIVE/DEAD Fixable Blue Dead (ThermoFischer, Catalog #L23105, Waltham, MA), CD8⁺ (BioLegend, Catalog #100712, Clone 53–6.7, San Diego, CA), CD4⁻ (BioLegend, Catalog #100566, Clone RM4–5, San Diego, CA), CD45.1⁺ (BioLegend, Catalog #110730, Clone A20, San Diego, CA), CD45.2⁺ (BioLegend, Catalog#109806, Clone 104, San Diego, CA), Vα2⁺ (Thermo Fisher, Catalog #46–5812-80/B20.1, Waltham, MA), Vβ5⁺ (BioLegend, Catalog #139504/MR9–4). Analysis of cells was conducted using LIVE/DEAD Fixable, CD4 (BioLegend), CD45.1 (BioLegend), CD45.2 (BioLegend), Vα2 (Thermo Fisher), Vβ5 (BioLegend), CD8 (BioLegend). LSR Aria II (BD Biosciences) was used for acquisition. Slamf6⁺ (BioLegend, Catalog #134610, Clone 330-AJ, San Diego, CA) was used for Slamf6 exon 7 KO experiments. Viable cell gate is representative of a size gate, single-cell gate, and viability gate. All sorting was done using FACSAria II (BD Biosciences).

For intracellular staining, cells were incubated for the last 4–5h in the presence of GolgiPlug (BD Biosciences), surface stained, fixed with 1.5% PFA, permeabilized with 1% Saponin, and stained at 4 °C overnight with anti-IFNγ antibody (BD Biosciences, Catalog# 554412, Clone XMG1.2, Franklin Lake, NJ) and anti-TNFα antibody. CellTrace Violet (Thermo Fisher, Catalog #C34557, Waltham, MA) experiments were performed according to the manufacturer’s protocol. The acquisition was performed by LSRII.

Human PBMCs were identified: CD8⁺ (BD Biosciences, Catalog #341051, Clone SK1, Franklin Lake, New Jersey), CD4⁺ (BD Biosciences, Catalog #561840, Clone RPA-T4,Franklin Lake, New Jersey), CD25⁺ (BD Biosciences, Catalog #565106, Clone 2A3, Franklin Lake, New Jersey), CD69⁺ (BD Biosciences, Catalog #555530, Clone FN50, Franklin Lake, New Jersey, PD1⁺ (Thermo Fisher, Catalog #25–9985-82, Clone J43, Waltham, MA), Tra2β⁺ (Santa Cruz Biotechnology, Catalog #sc-166829, Clone D-2, Dallas, TX). All flow cytometry data were gathered using BD FACSDiva 9.0 and analyzed with FlowJo 10.6.1 (Tree Star, Ashland, OR).

In vitro stimulation and mTOR Inhibition

OT-I CD8⁺ T cells from mice were plated at a density of 100,000 cells/200ul media and stimulated with either 100nM SIIGFEKL^257–264 peptide (Anaspec, Fremont, CA) or 100nM SIINFEKL peptide for 4 or 12 hours. Cells were collected after each timepoint and RNA was extracted for RT-PCR.

For mTOR inhibition experiments, OT-I CD8+ T cells from mice were plated at a density of 100,000 cells/200ul media and stimulated with 100nM SIINFEKL peptide after treatment with either DMSO, (0.01%) Torin (2μM) (Cat#S2827, Selleckchem, Houston TX), or (1μM) Rapamycin (cat#HY-10219, MedChemExpress, Monmouth Junction, NJ). After 20 minutes, phospho-S6 was measured by intracellular staining and flow cytometry using anti-pS6 antibody (Cat#12–9007-42, Invitrogen, Grand Island, NY). After 12 hours, cells were collected, and RNA was extracted for RNA-sequencing.

ELISA

All cells were restimulated for 20 h with phorbol 12-myristate 13-acetate (PMA; 50 ng/ml Calbiochem, Darmstadt, Germany) + ionomycin (1 μg/ml Invitrogen) prior to ELISA. Secreted IFNγ and TNFα from CD8⁺ T cell culture supernatants was analyzed by ELISA (human IFNγ Catalog # DIF50C, R&D Systems Minneapolis, MN, mouse IFNγ Catalog #555138, human TNFα Catalog #5239869, and mouse TNFα Catalog #555268, BD Pharmingen, Franklin Lake, NJ).

RT-PCR

RNA was extracted using a miRNEasy Micro kit (QIAGEN, Catalog #217084) without DNaseI treatment. 250–500ng of RNA was reverse transcribed using Superscript III reverse transcriptase (Invitrogen, Catalog #18080044). Semiquantitative PCR was used to amplify 10–50ng of cDNA with Phusion hot start II DNA polymerase (Thermo Fisher, Catalog #F549S) and primers listed in Data File S20. PCR products were separated in a 2% agarose gel stained with EtBR and imaged using Azure Biosystems. PCR bands were quantified using ImageJ software. Percent spliced in (PSI) ratio of each exon-containing transcript was calculated as exon-included isoform band intensity divided by the intensity of included and skipped isoform bands.

Influenza detection

Detection of IAV NP gene segments from the lung of infected mice was performed as previously described (50). Briefly, tissue was resected from the left lobe and 20mg was finely chopped in 100uL of sterile PBS before using a Bio-Gen PRO200 Homogenizer (Pro Scientific, Oxford, CT, Cat # 01–01200). RNA was extracted using the RNEasy Mini Kit (QIAGEN, Catalog #74104) and a total of 1μg of lung RNA was synthesized into cDNA using the PrimeScript^™ Reverse Transcription (RT) reagent Kit (TaKaRa Bio, Inc, Cat# RR037A). For synthesis of initial PCR products, 1μL of the cDNA reaction mixture served as the template using the Phusion Hot Start II High-Fidelity DNA polymerase kit (Thermo Fisher Scientific, Cat #F-549L). Forward and reverse primers in the first (outer) PCR: F 5’ (GATGGCATGCCATTCTGCCG); R 5’ (GTACTCCTCTGCATTGTCTC). For the second (inner) PCR, forward and reverse primers: F 5’ (GAACAACCGTTATGGCAGCA); R 5’ (CATGTCAAAGGAAGGCACGA). PCR cycles were followed as specified in the Phusion Hot Start II High-Fidelity DNA polymerase kit without modifications. The final PCR products were visualized on a 2% agarose gel with SYBR Safe staining.

Conservation

TRA2β’s human ultraconserved element and protein coding exon sequences were identified using UCSC Genome Browser. MultiZ vertebrate alignment was used to visualize relative conservation for 11 other species and identify UCE/coding exon sequences. Each sequence was then individually run against the human UCE (or coding exon) sequence using NCBI BLAST to assign an individual conservation score.

For Vast-DB analysis , a database of alternative splicing changes over evolution, coordinates were entered for TRA2β-PE (chr3:185931577–185931852) and for TRA2β exon 4 (chr3:185922011–185925626) to compare conservation across organisms (42).

Bulk RNA-seq

For each group, 10⁴–10⁶ cells were lysed in Trizol lysis buffer. RNA was extracted and processed using miRNEasy Extraction Kit (QIAGEN, Hilden, Germany). RNA quantification, Library preparation and sequencing was conducted by the Whitehead Institute Genome Technology Core (Boston, MA). RNA quantification was done by NanoDrop and subsequent TapeStation analysis. RNA library construction was done using a SMART-Seq v.4 Ultra-low input kit (Takara Bio, Shiga, Japan) and cDNA was converted to sequencing library using NexteraXT DNA Library Prep Kit with indexing primers (Illumina, San Diego, CA). For deep-sequencing, libraries were sequenced for paired end 2 × 100bp or 2 × 150 bp reads for 50–100 million total reads/sample on a NOVASeq 6000 (Illumina). Quality controlled reads (fastq) were aligned to mouse genome (mm39) or human genome (Hg38) using STAR and BAM file conversion, sorting, and indexing were done with Samtools (51, 52). Read counts were obtained from resulting BAM files using Stringtie (53). For transcriptomics analysis, differentially expressed genes were identified using DESeq2 (FDR < 0.05) (54). FPKM values from Stringtie of most differentially expressed transcripts by FDR were used to create heatmaps of significant DEGs. Pathway analysis was conducted using PathfindR (55).

For RNA splicing analysis, raw reads (fastq) were processed using RMATS (in Fig. 1, Fig. 2, Fig. 3, Fig. 5, Fig. S1, Fig. S4, Fig. S6, Fig. S8, Fig. S9, Fig. S10, Fig. S12) and Whippet (in Fig. 3, Fig. 5, Fig. S5, and Fig. S12) for alternative splicing, alternative polyadenylation, or alternative TSS events (30, 31). Significantly differentially spliced events using Whippet analysis were identified using probability > 0.9, Total Reads >50, and percent spliced in (PSI) differential of >0.1 or <−0.1 (Fig. 5, Fig. S12) or PSI differential of >0.05 or <−0.05 (Fig. 3, Fig. S5) for our screen. Significant differentially spliced events using RMATS analysis were identified using an FDR<.05 and a PSI differential of >0.1 or <−0.1, (Fig. 1, Fig. 2, Fig. 5, Fig. S1, Fig. S4, Fig. S8, Fig. S9, Fig. S12) or a PSI differential of >0.025 or <−0.025 and total read count per sample of > 5 for our CRISPR screen (Fig. 3, Fig. S6). Integrative Genomics Viewer (Broad Institute, Cambridge MA) was used to create splice graph images (56). PE coordinates are provided in Data S20.

Single Cell Smart-seq2 processing

Processing followed steps directly from Psix on Github (57). In brief, FASTQ files were obtained from SRA and individually aligned using STAR mapping for transcript counts as well as splice junctions (splice junctions can be found in the SJ.out.tab file output with all reads using the parameter --outSJfilterReads All during STAR alignment) (57). Ready-to-use cassette exon annotations were obtained for mouse and human from Psix annotations on Github. TPM matrix was created using RSEM version 1.2.31 while dimensionality reduction to achieve a low-dimensional cell space was done using the PCA function in R (58). Ingenuity Pathway Analysis (QIAGEN, Hilden, Germany) was used to analyze pathways enriched after Pearson correlation (59).

Human cell samples and PBMC stimulation

PBMCs from healthy male and female donors between 18 and 55 years of age (Stem Cell Technologies, Catalog #70025, Vancouver, Canada; informed consent: “This consent form provides information about the research study. You will be asked to read this consent form and the study staff will review the consent form with you and answer any questions you may have. If you decide to participate in the study, you will be asked to sign and date this consent form. You will be given a copy of this signed and dated consent form. You should not join this research study until all of your questions are answered.”) were obtained from Stem Cell Technologies under a protocol approved by the Western Institutional Review Board (WIRB). PBMCs were cultured at 1 × 10⁶ cells/ml in Complete T cell Media (RPMI 1640 (ATCC, Catalog #30–2001, Manassas, VA), 2 mM L-glutamine, 10 mM HEPES, 1 mM sodium pyruvate, 4500 mg/L glucose, and 1500 mg/L sodium bicarbonate, 1× Antibiotic-antimycotic (Gibco, Waltham MA, Catalog #15240062), 5% Human AB Serum (Sigma-Aldrich, St. Louis, MO, Catalog #H4522), 1× Non-essential amino acids (Gibco, Catalog #11140050) and 10 ng/mL recombinant human IL-2 (R&D Systems, Minneapolis, MN, Catalog #202-IL-010). Cells were stimulated with plate bound anti-CD3 (BioLegend, San Diego,CA, catalog #317347, Clone OKT3) and soluble anti-CD28 (BD Pharmingen, Franklin Lake, NJ , Catalog #555725) + IL-2. Cells were restimulated for 20 h with phorbol 12-myristate 13-acetate (PMA; 50 ng/ml, Calbiochem, Darmstadt, Germany) + ionomycin (1 μg/ml Invitrogen). For cytokine analysis experiments, cells were stimulated with CD3 + CD28 beads (Gibco, Waltham MA, Catalog #11131D) for 4 days at a 1:2 ratio of cells:beads and restimulated every 2 days for 8 days, replacing media at day 4.

Antisense oligonucleotide (ASO)-targeting

PBMCs from healthy donors (as described above) were cultured in Complete T cell media and electroporated with 50nM of 2′-O-methoxyethyl-phosphorothioated TRA2β-targeting ASO-1570 (Integrated DNA Technologies, Coralville, IA, sequence in Data File S20) or non-targeting control (60) (sequence in Data File S20). All ASOs were purified using RNase free high performance liquid chromatography. Cells were then stimulated with plate-bound anti-CD3 (BioLegend) and soluble anti-CD28 (BD) + IL-2 for 48 hours. CD8⁺ CD25⁺ Live cells were then sorted using FACS for >95% purity prior to use for RT-PCR as described above in “Flow cytometry”.

CRISPR/Cas9 in vivo

CRISPR/Cas9 experiments in vivo followed the protocol as previously described by Nussing et al (16). Briefly, splenocytes from naïve CD45.1⁺ and CD45.1⁺/CD45.2⁺ OT-I mice were enriched for CD8⁺ T cells and separated into 5–10 million cells per group. Control gRNA or gRNA against target RBP or Tra2β-PE (see Data File S20 for gRNA list) (Synthego, Menlo Park, CA) were complexed with Cas9 enzyme and electroporated into CD8⁺ OT-I cells. After recovery, cells were adoptively transferred into C57BL/6J mice. Recipient mice were injected intraperitoneally with 100 μg SIINFEKL^257–264 peptide (InvivoGen) or with 100ug SIITFEKL^257–264 peptide (Invivogen), agonist anti-CD134 antibody (20μg) (Bio X Cell) and agonist anti-CD137 antibody (10μg) (Bio X Cell) 4–6 h after adoptive transfer. For influenza infections, mice were infected 24 hours after transfer with 2000 plaque-forming units (PFU) X31-OVA (H3N2) (Charles River) as described above. Cells were sorted for CD8⁺ CD45.1⁺, CD45.2⁺ markers 3 days and 12 h after stimulation by flow cytometry for >95% purity prior to use in downstream assays. Viable cell gate is representative of a size gate, single-cell gate, and LIVE/Dead negative gate. When targeting gRNA against RBPs, using either CD45.1⁺ or CD45.1⁺/CD45.2⁺ OT-I cells resulted in identical outcomes. In PE-KO experiments, using TRA2β-PE flanking gRNA in CD45.1⁺ OT-I T cells or TRA2β-PE flanking gRNA in CD45.1⁺/CD45.2⁺ OT-I T cells resulted in identical outcomes for both antigen-bolus and antigen-prolonged models.

CRISPR/Cas9 in vitro splice site screen

CRISPR/Cas9 experiments in vitro followed the protocol for electroporation as described in vivo described by Nussing et al (16) without adoptive transfer. Briefly, splenocytes from naïve CD45.1⁺, CD45.1⁺/CD45.2⁺ OT-I mice were enriched for CD8⁺ T cells and separated into 250,000 cells per group. Control gRNA was complexed with Cas9 enzyme and electroporated into CD45.1⁺ OT-I CD8⁺ T cells. Target gRNA against isoform-specific exons (see Data File S20 for gRNA list) (Synthego, Menlo Park, CA) were complexed with Cas9 enzyme and electroporated into CD45.1⁺/CD45.2⁺ OT-I CD8⁺ T cells. 125,000 Control CD45.1⁺ and 125,000 Target CD45.1⁺/CD45.2⁺ were co-cultured into individual wells of two 96 well plates. A sample was taken for analysis of relative expansion by flow cytometry prior to stimulation. Each group of cells received SIINFEKL (100nM) + IL-2 (10U/ml) + anti-CD28 antibody (100ng/ml) (clone 37–51, cat#14–0281-85, ebioscience, San Diego CA) in one plate and SIITFEKL (100nM) + IL-2 (10U/ml) in the second plate. After 72 hours, cells were analyzed by flow cytometry for relative expansion. Further, both plates were restimulated with PMAi for 4 hours to analyze for intracellular IFNγ and TNFα as described above.

eCLIP sequencing and analysis

eCLIP sequencing was conducted on naïve OT-I CD8⁺ T cells pooled from 50 mice (total of 100 million cells) and on effector OT-I CD8⁺ T cells pooled from 30 recipient mice (100 million T cells) after adoptive transfer and stimulation with 100ug SIINFEKL + agonist anti-CD134 antibody (20ug) and agonist anti-CD137 antibody (10ug) as described above. Cells were UV crosslinked on ice at 150 mJ/cm² using a Stratagene UV Stratalinker 2400 and sent to Eclipse Bioinnovations Inc. for processing using their RBP-eCLIP kit and protocol (RBP-eCLIP Protocol v2.01 R). Immunoprecipitation was performed using the anti-Tra2β rabbit polyclonal antibody (Santa Cruz Biotechnology, Dallas TX, cat. #sc-166829) and anti-IgG rabbit antibody (Eclipse Bioinnovations Inc., San Diego CA). Libraries were sequenced by Eclipse Bioinnovations Inc. for single end sequencing (1×72bp). Sequences were analyzed using the original eCLIP analysis pipeline (Eclipse Data Analysis Protocol). Significant eCLIP peaks (Tra2β binding events) were identified by Clipper using the strict threshold set by Eclipse Bioinnovations of log2 fold change ≥ 3 and −log10(p-value) ≥ 3. Peaks are available in Data File S12. HOMER was used for identification of top binding motifs by Eclipse Bioinnovations (33).

Statistics

Unless otherwise indicated, dots represent an individual biological replicate, and summary graphs represent means ± SD. Ordinary one-way or two-way ANOVA with Sidak’s correction for multiple comparisons and unpaired two-sided t test (single or multiple) were used for comparisons. Analyses were performed with GraphPad Prism version 9 (GraphPad Software). All comparisons and tests are noted in the figure legend in which they appear.

Data and materials availability

All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials. RNA-seq and CLIP-seq data generated for this study are available on GEO at GSE267609. Sequencing data from mouse Naïve and Effector CD8+ T cell can be found at GSE200236, GSE200237. Sequencing data from mouse Listeria infection can be found at GSE89307. Sequencing data from human Yellow Fever Virus can be found at GSE100745. Sequencing data from T-ALL data can be found at GSE139622. Sequencing data from human thymus can be found at GSE151081. Single cell sequencing thymic data can be found at GSE109774.

All gating strategies are provided in Fig. S14–16. All raw gels are provided in Fig. S17–S18.