Regulation of CD8 T Cell Differentiation by the RNA-Binding Protein DDX5

CD8⁺ T cells are a key component of the adaptive immune response. Activated CD8⁺ T cells differentiate into heterogeneous cell populations that provide efficient pathogen clearance and durable protection against reinfection. Terminal effector (TE) cells are responsible for mediating pathogen clearance through the production of inflammatory cytokines and cytolytic granules, whereas memory cells provide potent recall responses in the event of reinfection (1). The circulating memory cell population is comprised of several distinct subsets, including central memory T (T_CM), effector memory T (T_EM), and terminal T_EM (t-T_EM) cells, that possess unique functional, migratory, and survival properties (2–5). T_CM cells express key transcription factors, such as Tcf7, Eomes, and Foxo1, that enable their stem cell–like capacity to expand, differentiate, and self-renew (6). T_EM and t-T_EM cells are characterized by potent effector function upon infectious rechallenge.

RNA-binding proteins are critical regulators of cellular differentiation due to their broad influence on RNA metabolism and gene regulation (7, 8). However, the understanding of their impact on T cell differentiation is still developing. We previously reported that the RNA-binding protein, DEAD-box protein 5 (DDX5), regulates the differentiation of an effector-like tissue-resident memory cell population (9). These findings, coupled with recent evidence suggesting a role for DDX5 in regulating the effector function of Th17 cells (10), raised the possibility that DDX5 might also regulate differentiation of circulating effector and memory CD8⁺ T cell subsets.

DDX5 is a member of the DEAD-box family of RNA helicases that are named for a conserved DEAD (Asp-Glu-Ala-Asp) motif that is important for ATP binding, hydrolysis, and RNA helicase activities (11, 12). DDX5 has been shown to be crucial for cell development, proliferation, and organ maturation, and it has been implicated in the tumorigenesis of many cancer types (13). DDX5 has been reported to influence nearly every modality of gene regulation, including chromatin organization and epigenetic remodeling, transcriptional coactivation and corepression, and alternative splicing of RNA transcripts and RNA metabolism (11). Although DDX5 is a well-known polyfunctional regulator of gene expression, its role in CD8⁺ T cell biology has not been extensively explored.

In this study, we report the consequences of T cell–specific deletion of DDX5 in the differentiation and function of CD8⁺ T cells responding to lymphocytic choriomeningitis virus (LCMV) infection. DDX5 deficiency reduced the proportions of KLRG1^hiCD127^lo TE cells and increased the proportions of the KLRG1^loCD127^hi memory precursor (MP) cell population at day 7 postinfection (p.i.). Furthermore, DDX5 deficiency reduced the proportions of T_EM and t-T_EM cells and increased the proportion of T_CM cells at day 30 p.i. DDX5-deficient CD8⁺ T cells exhibited decreased expression of genes associated with terminally differentiated effector cells, including Klrg1, Zeb2, and Cx3cr1, and increased expression of genes associated with T_CM cell differentiation, including Tcf7 and Eomes. Ectopic expression of DDX5 increased the proportions of T_EM and t-T_EM cells and decreased the proportion of T_CM cells, and was associated with reduced expression of Tcf7 and Eomes. Taken together, these findings suggest that DDX5 may regulate CD8⁺ T cell differentiation by virtue of repressing genes that promote the generation of T_CM cells.

All mice were housed under specific pathogen–free conditions in an American Association of Laboratory Animal Care–approved facility at the University of California San Diego (UCSD), and all procedures were approved by the UCSD Institutional Animal Care and Use Committee. C57BL/6J (CD45.1.2⁺ or CD45.2⁺) and P14 TCR transgenic (CD45.1⁺ or CD45.1.2⁺, maintained on a C57BL/6J background) mice were bred at UCSD or purchased from The Jackson Laboratory. Ddx5^fl/fl mice were obtained from Dr. Frances Fuller-Pace’s laboratory (University of Dundee) and have been previously described (14). To generate congenically distinct P14 Ddx5^fl/fl Cd4-Cre⁺ and P14 Ddx5^fl/fl Cd4-Cre⁻ mice, Ddx5^fl/fl mice were crossed to P14 Cd4-Cre⁺ mice (either CD45.1⁺ or CD45.2⁺). All mice were used from 6 to 9 wk of age, male mice were used as recipients, and male or female mice were used as donors in adoptive transfer experiments. No randomization of blinding was used in infection experiments.

For isolation of CD8⁺ T cells from the spleen and peripheral lymph nodes, tissues were collected and filtered through 70-μm cell strainers to yield a single-cell suspension before treatment with RBC lysing buffer Hybri-Max (Sigma-Aldrich) for 5 min. CD8⁺ T cells were then isolated using the CD8a⁺ T cell isolation kit and LS MACS columns (Miltenyi Biotec) according to the manufacturer’s protocol.

Surface proteins were stained for 10 min on ice in HBSS (Corning) supplemented with 1% FBS with the following Abs: Vα2 (B20.1), CD8α (53-6.7), CD45.1 (A20), CD45.2 (104), KLRG1 (2F1/KLRG1), CD127 (A7R34), CD27 (LG.3A10), CX3CR1 (SA011F11), and CD62L (MEL-14) (all were purchased from BioLegend). Samples were then stained in fixable viability dye eFluor 780 (Thermo Fisher Scientific) at 1:1000 on ice for 10 min. Cells were then fixed in resuspended in eBioscience fixation/permeabilization diluent (eBioscience) for 30 min at room temperature. After fixation, intracellular targets were stained in eBioscience permeabilization buffer (eBioscience) for 8 h at 4°C with the following intracellular Abs: TCF1/7 (C63D9), EOMES (Dan11mag), Ki67 (16A8), T-bet (4B10), and granzyme B (GB11). For assessment of cytokine production, cells were cultured in the presence of LCMV gp_33–41 peptide (GenScript) and protein transport inhibitor cocktail (Thermo Fisher Scientific) for 3 h at 37°C. After cell surface and viability staining, cells were fixed and permeabilized using BD Cytofix/Cytoperm (BD Biosciences) for 30 min at room temperature before staining with anti–IFN-γ (XMG1.2), TNF-α (MP6-XT22), and IL-2 (JES6-5H4) Abs (all from BioLegend) for 30 min on ice.

For DDX5 staining, after surface marker staining and fixation using an eBioscience transcription factor fixation/permeabilization kit, isolated lymphocytes were stained with anti-DDX5 (JM52-30) at 1:200 to 1:400 in 1× permeabilization buffer, at room temperature for 2 h, followed by staining with Alexa Fluor 488 donkey anti-rabbit IgG (minimal x-reactivity) Ab (BioLegend) at 1:100 at room temperature for 1–2 h.

Total RNA was extracted with the RNeasy kit (Qiagen) and reverse transcribed using the iScript select cDNA synthesis kit (Bio-Rad Laboratories). Real-time RT-PCR was performed using iTaq Universal SYBR Green supermix (Bio-Rad Laboratories). DDX5 expression data were normalized to Gapdh mRNA levels. hDdx5 primer sequences were as follows: forward, 5′-ATGTCGGGTTATTCGAGTGACC-3′; reverse, 5′-ACTTCCTCCAAATCGAGGTGC-3′. Gapdh primer sequences were as follows: forward, 5′-TGAGTATGTCGTGGAGTCTAC-3′; reverse, 5′-TGGACTGTGGTCATGAGCC-3′.

For naive CD8⁺ T cell transfers, splenocytes were collected from naive congenically distinct Ddx5^fl/fl-Cd4-Cre⁺ P14⁺ and Ddx5^fl/fl-Cd4-Cre⁻ P14⁺ mice. To quantify the number of Vα2⁺CD8⁺CD45.1⁺ cells, an aliquot of the isolated lymphocytes was stained with Abs against Vα2, CD8α, and CD45.1 and analyzed using flow cytometry. Vα2⁺CD8⁺CD45.1⁺ cells (1 × 10⁵) were adoptively transferred into CD45.2⁺ WT recipients 1 d before i.p. infection with 2 × 10⁵ PFU of LCMV-Armstrong. At days 7 and 30 p.i., donor mice were sacrificed, their spleens were harvested, and the donor cells were analyzed using flow cytometry.

For induced Ddx5 deletion, Vα2⁺CD8⁺CD45.1⁺ P14 cells from naive congenically distinct Ddx5^fl/fl-Ert2-Cre⁺ P14⁺ and Ddx5^fl/fl-Ert2-Cre⁻ P14⁺ mice were mixed 1:1 and adoptively transferred into CD45.2⁺ host mice before subsequent LCMV infection. To induce Ddx5 deletion, recipient mice were treated i.p. with 1 mg of tamoxifen diluted in sunflower oil on days 4–7 or 30–33 following infection.

For forced DDX5 (pMThy1.1_hDDX5_FL) experiments, retroviral particles were generated using platinum E cells grown in 10-cm plates with full selection media (DMEM, 10% FBS [v/v], 2 mM l-glutamine, 100 U/ml penicillin-streptomycin, 1 μg/ml puromycin, and 10 μg/ml blasticidin). Eighteen hours before transfection, selection media were replaced with antibiotic-free media (DMEM, 10% FBS [v/v], 2 mM l-glutamine). For each 10-cm plate, 10 μg of each construct and 5 μg of pCL-Eco helper plasmids were mixed in Opti-MEM (Thermo Fisher Scientific) to a volume of 700 μl. This was combined with 45 μl of TransIT-LT1 reagent (Mirus Bio) and 655 μl of Opti-MEM for 20 min at room temperature. The mixture was then added dropwise to each 10-cm plate. Twelve hours later, media were replaced with fresh antibiotic-free media supplemented with cholesterol at 30 μM and ViralBoost (1:500; ALSTEM Cell Advancements). The viral supernatant was subsequently harvested at 48 and 72 h. Retroviral supernatant was filtered through a 0.45-μm syringe filter and stored at −80°C. Naive WT P14 CD8⁺ T cells were plated at a density of 1 × 10⁶ cells/ml in 24-well plates precoated with 5 μg/ml each of anti-CD3 (clone 3C11) and anti-CD28 (clone 37.51) with T cell media supplemented with IL-2 (100 U/ml). The cells were then centrifuged for 90 min at 2000 rpm. Retroviral supernatant was removed and replaced with stored T cell media and incubated at 37°C for 2 h before adoptive transfer.

For cotransfer experiments, cells of each construct were mixed at a 1:1 ratio 5 × 10⁵ donor cells/mouse were adoptively transferred into CD45.2⁺ male recipient mice. One hour later, mice were infected with 2 × 10⁵ PFU of LCMV. Seven and 30 d p.i., mice were euthanized, spleens were harvested, and the donor cells were analyzed by flow cytometry.

For ex vivo restimulation assays, P14 CD8⁺ T cells were plated in a U-bottom 96-well plate at 10–15 × 10⁶ cells/well in the presence of 1 ng/μl LCMV gp_33–41 peptide (GenScript) and 1× protein transport inhibitor cocktail (eBioscience) for >3 h at 37°C. Cells were then fixed and permeabilized using BD Cytofix/Cytoperm (BD Biosciences) and stained for IFN-γ (XMG1.2), TNF-α (MP6-XT22), and IL-2 (JES6-5H4) Abs (all purchased from BioLegend) for 30 min on ice. Flow cytometry was performed on a NovoCyte (ACEA Biosciences). FACS sorting of cells was done on a FACSAria Fusion or FACSAria II (BD Biosciences). FlowJo software (BD Biosciences) was used for analysis of flow cytometry data.

WT CD45.1 and Ddx5^fl/flCd4-Cre⁺ CD8⁺ CD45.1.2 P14 cells were adoptively cotransferred into congenically distinct hosts. The host mice were infected with 2 × 10⁵ PFU of LCMV 24 h after transfer and sacrificed on days 4, 7, and 30 p.i. Circulating lymphocytes were isolated from the spleen and resuspended with Ab staining buffer containing TruStain FcX PLUS (anti-mouse CD16/32; BioLegend). Cells were incubated with TotalSeq-A Ab panel (Supplemental Table I) and fluorescently labeled Abs targeting CD8, Vα2, CD45.1, and CD45.2 for 30 min at 4°C. Stained cells were washed with Ab staining buffer three times before sorting for P14 CD8⁺ T cells of each genotype. About 10,000 cells per sample were loaded into single-cell G chips (10x Genomics) and partitioned into gel bead-in-emulsions in a Chromium controller (10x Genomics). Single-cell RNA libraries were prepared according to the 10x Genomics Chromium Next GEM Single Cell 3′ v3.1 (dual index) user guide and sequenced on a HiSeq 4000 (Illumina).

Reads from cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq) were aligned to mm10 using Cell Ranger (v6.0.1). The unique molecular identifier count matrix was used for downstream analysis using Seurat (v4.1.1). In R, samples were subsampled to 2000 cells per sample. Cells were further filtered to exclude cells with <300 genes per cell and <5% mitochondrial genes. Standard parameters were used for normalization, filtering, principal component analysis reduction, and uniform manifold approximation and projection (UMAP) creation. The samples were clustered with Louvain clustering using the FindClusters function with a resolution of 1. Differential expression analysis was completed between clusters and conditions using the FindMarkers function. Data are available at Gene Expression Omnibus under GSE230028.

To assess phenotypic changes in circulating CD8⁺ T cell populations resulting from deletion of DDX5, we adoptively cotransferred P14 CD8⁺ T cells, which express a transgenic TCR that recognizes the immunodominant epitope of LCMV, from mice with a T cell–specific deletion of Ddx5 (Ddx5^fl/fl Cd4-Cre⁺ [Ddx5^cKO]) and congenically distinct control mice (Ddx5^fl/fl Cd4-Cre⁻ [WT]) into CD45.2⁺ recipient mice. Relatively high numbers of P14 cells were transferred (2 × 10⁵ CD8⁺ P14 cells total), which may alter the kinetics and phenotype of the ensuing CD8 T cell response, representing a limitation of the current study (15). One day after cell transfer, recipient mice were infected with the Armstrong strain of LCMV. At days 7 and 30 p.i., lymphocytes were isolated from the spleens of recipient mice and donor cells were analyzed by flow cytometry (Fig. 1A). DDX5-deficient CD8⁺ T cells were reduced compared with their control counterparts at days 7 and 30 p.i., suggesting a possible role for DDX5 in cellular proliferation and/or survival (Fig. 1B). Furthermore, the altered frequencies of DDX5-deficient CD8⁺ T cells compared with control cells remained relatively consistent through the duration of acute infection (Fig. 1C, 1D). Consistent with the notion of DDX5 repressing cellular proliferation, DDX5-deficient CD8⁺ T cells expressed higher levels of Ki67, a marker of cellular proliferation (Fig. 1E, 1F), compared with control cells. To further elucidate the role of DDX5 in proliferation, we stained WT and DDX5-deficient CD8⁺ T cells with the fluorescent dye, CFSE, which dilutes as cells divide, allowing for efficient tracing of cell proliferation patterns. CFSE-labeled WT and DDX5-deficient cells were mixed at a 1:1 ratio and adoptively transferred into congenically distinct recipients. Forty-eight hours after LCMV infection, recipient mice were sacrificed, and donor cells were analyzed by flow cytometry. DDX5-deficient cells exhibited greater CFSE dilution than did control cells (Fig. 1G, 1H, Supplemental Fig. 1A, 1B). Lastly, we observed that DDX5-deficient CD8⁺ T cells expressed higher levels of annexin V, a membrane-bound apoptotic marker, at 48 h p.i. (Fig. 1I, 1J). Having shown that conditional knockout of Ddx5 accelerated proliferation and led to an early death phenotype, we next investigated whether deletion of DDX5 after the initial proliferative burst of early infection also affected CD8⁺ T cell survival. Importantly, inducible deletion of Ddx5 during the peak of infection (days 4–7 p.i.) did not result in a numerical deficit in DDX5-deficient cells at days 9 and 30 p.i. (Supplemental Fig. 1C, 1D), suggesting that the DDX5 may only play a role in proliferation and survival during the first few days of infection. Taken together, these results indicate that DDX5 may have an important role in inhibiting apoptosis, leading to the observed decrease of DDX5-deficient cells on days 7 and 30 p.i.

We next examined the consequences of T cell–specific deletion of Ddx5 (Ddx5^fl/fl Cd4-Cre⁺ [Ddx5^cKO]) on circulating effector and memory cell populations (Fig. 1A). Ddx5 deletion resulted in reduced proportions and absolute numbers of KLRG1^hiCD127^lo TE-phenotype cells and increased proportions of KLRG1^loCD127^hi MP-phenotype cells at day 7 p.i. (Fig. 2A, Supplemental Fig. 1E). Furthermore, DDX5 deletion resulted in increased proportions of CD62L^hiCD127^hi T_CM cells at day 30 p.i. (Fig. 2B, Supplemental Fig. 1F). Conversely, Ddx5^cKO memory cells displayed reduced proportions and absolute numbers of CD62L^loCD127^hi T_EM cells and CD62L^loCD127^lo t-T_EM cells at day 30 p.i. (Fig. 2B, 2C). Moreover, compared with control cells, DDX5-deficient memory cells exhibited increased expression of CD62L at day 30 p.i. (Supplemental Fig. 2A). Consistent with the observed increase in T_CM cell differentiation, CD127 (Il7r) and CD122 (Il15rb) expression levels were both increased in the absence of DDX5, suggesting that these cells may be better able to compete for IL-7 and IL-15 (Supplemental Fig. 2B, 2C). Analyses of functional parameters demonstrated that compared with control cells, DDX5-deficient CD8⁺ T cells displayed equivalent frequencies of IFN-γ⁺TNF⁺ cells, whereas the frequency of IL-2⁺ DDX5-deficient CD8⁺ T cells was increased compared with controls upon ex vivo restimulation with gp_33–41 peptide (Fig. 2D, 2E). Moreover, IFN-γ, TNF, and IL-2 expression and IFN-γ⁺IL-2⁺ coexpression were unaffected in Ddx5^cKO cells compared with controls (Supplemental Fig. 2D–G). Importantly, analysis of pretransfer WT and Ddx5^cKO cells showed similar frequencies of naive CD44^loCD62L^hi CD8⁺ T cells (Supplemental Fig. 3A). In addition, naive pretransfer WT and Ddx5^cKO CD8⁺ T cells exhibited similar expression of the transcription factor TCF1, which is expressed by all naive CD8⁺ T cells, suggesting that the phenotypes observed in Ddx5^cKO cells were not a consequence of altered thymic selection or homeostasis (Supplemental Fig. 3B, 3C). Taken together, these data demonstrate that deletion of DDX5 changes the composition of the Ag-experienced CD8⁺ T cell pool.

Because DDX5 deletion appeared to inhibit TE cell differentiation and promote the generation of T_CM cells, we next assessed the functional consequences of ectopic expression of DDX5. We isolated and activated congenically distinct CD8⁺ P14 T cells, retrovirally transduced the cells with empty vector (EV) or DDX5 ectopic expression (Ddx5^OE) constructs, mixed them at a 1:1 ratio, and adoptively cotransferred the cells into CD45.2⁺ mice (Fig. 2F). Recipient mice were sacrificed at days 7 and 30 p.i. and donor cells were analyzed using flow cytometry. In contrast to decreased proportions of TE cells and increased proportions of T_CM cells resulting from Ddx5 deletion, ectopic expression resulted in increased proportions and absolute numbers of TE cells and reduced proportions of MP cells at day 7 p.i., along with decreased proportions of T_CM cells and increased proportions and absolute numbers of T_EM and t-T_EM cells at day 30 p.i. (Fig. 2G, 2H). Furthermore, Ddx5^OE memory cells exhibited reduced CD62L and CD127, whereas CD122 expression was unaffected (Supplemental Fig. 2H–J). In contrast to conditional knockout of Ddx5, a greater proportion of Ddx5^OE cells expressed TNF and IFN-γ compared with EV controls, and TNF expression was increased among Ddx5^OE cells (Fig. 2K, Supplemental Fig. 2K, 2L). Additionally, while the proportions of IL-2–producing or IL-2 and IFN-γ–coproducing memory cells were not affected by ectopic expression of Ddx5, the expression of IL-2 was enhanced in Ddx5^cKO CD8⁺ T cells (Fig. 2L, Supplemental Fig. 2M, 2N). Taken together, these findings indicate that DDX5 promotes the differentiation of effector-like subsets of CD8⁺ T cells.

Having shown that DDX5 plays a role in the generation of memory cells, we next investigated whether DDX5 also plays a role in established memory cells. We therefore used an inducible Cre recombinase model in which the deletion of Ddx5 is temporally regulated by administration of tamoxifen. Congenically distinct Ddx5^fl/fl Ert2-Cre⁻ (WT) and Ddx5^fl/fl Ert2-Cre⁺ (Ddx5^iKO) CD8⁺ P14 T cells were adoptively cotransferred into CD45.2⁺ recipients and infected with LCMV 24 h later. Recipient mice received tamoxifen on days 30–33 p.i. and donor cells were analyzed at day 40 p.i. by flow cytometry (Fig. 3A). A reduction in DDX5 protein and mRNA expression were confirmed by flow cytometry and quantitative real-time PCR of sorted WT and Ddx5^iKO CD8⁺ T cells (Supplemental Fig. 3D–F). Induced deletion of Ddx5 in established circulating memory cells resulted in increased absolute numbers of total DDX5-deficient memory cells relative to control cells (Fig. 3B). Among memory T cell subsets, Ddx5 deletion resulted in increased proportions and absolute numbers of T_CM cells and reduced proportions of t-T_EM cells (Fig. 3C). Furthermore, compared with control cells, reduced proportions of DDX5-deficient CD8⁺ T cells could produce IFN-γ and TNF (Fig. 3D). Notably, the production levels of IFN-γ and TNF were largely similar among Ddx5^iKO and control cells among the circulating memory cell subsets, except for reduced expression of IFN-γ in the Ddx5^iKO T_CM cell population (Fig. 3E, 3F). The frequency of IL-2–producing Ddx5^iKO memory cells was also reduced compared with WT control cells, whereas IL-2 production was only significantly decreased within the t-T_EM cell population and increased in the T_CM cell population (Fig. 3H). Taken together, these findings suggest a role for DDX5 in promoting the maintenance of t-T_EM cells.

To investigate potential mechanisms by which DDX5 regulates CD8⁺ T cell differentiation, we employed CITE-seq, which allows for measurements of the transcriptome and selected proteins in the same single cells (16). We adoptively cotransferred control WT and Ddx5^cKO CD8⁺ P14 T cells into congenically distinct recipient mice that were infected with LCMV 1 d after transfer (Fig. 4A). At days 4, 7, and 30 p.i., splenic CD8⁺ P14 T cells were isolated by FACS and processed for CITE-seq using the 10x Genomics platform. Abs targeting cell surface molecules, some of which have been previously implicated in CD8⁺ T cell activation and differentiation, were selected for inclusion in the CITE-seq Ab panel (Supplemental Table I).

UMAP analyses revealed that CD8⁺ T cells clustered distinctly at all time points p.i. (Fig. 4B), indicating that T cells exhibit distinct gene expression patterns during differentiation, as previously demonstrated (9, 16–18). Notably, WT and Ddx5^cKO CD8⁺ T cells clustered distinctly at days 7 and 30 p.i. (Fig. 4B). Ddx5^cKO CD8⁺ T cells exhibited increased expression of Sell and Il7r and decreased expression of genes associated with TE differentiation, such as Zeb2, Klrg1, and Bhlhe40, consistent with the observed bias toward T_CM cell differentiation in the absence of DDX5 (Supplemental Fig. 4A). Moreover, pathway analyses revealed an enrichment of genes regulating cytokine production and cytotoxicity in WT cells relative to Ddx5^cKO cells, as well as of processes related to ribosome biogenesis and cytoplasmic translation, consistent with the known involvement of DDX5 in ribosome assembly and function (19) (Supplemental Fig. 4B).

We next evaluated circulating memory cell heterogeneity at day 30 p.i. Clusters 0, 9, 15, and 19 were primarily comprised of day 30 p.i. WT CD8⁺ T cells, whereas clusters 2 and 12 were primarily comprised of Ddx5^cKO CD8⁺ T cells (Fig. 4C). Comparative Gene Ontology analyses among the six day 30 clusters revealed that the Ddx5^cKO clusters were more similar to each other than to the WT clusters (Supplemental Fig. 4C). Application of T_CM, T_EM, and t-T_EM transcriptional signatures to WT and DDX5-deficient memory cells revealed that the four WT-predominant clusters (0, 9, 15, and 19) tended to score more highly for the T_EM and t-T_EM transcriptional signatures, whereas the two Ddx5^cKO-predominant clusters (2, 12) tended to score more highly for the T_CM gene signature (Fig. 4D–F). Furthermore, application of these transcriptional signatures to WT and DDX5-deficient memory cells within each cluster demonstrated that WT cells scored higher for the T_EM and t-T_EM gene signatures, whereas DDX5-deficient cells scored higher for the T_CM gene signature (Fig. 4G–I). Analyses focused on specific genes revealed that compared with the four WT clusters (0, 9, 15, and 19), the two Ddx5^cKO clusters (2, 12) exhibited increased expression of transcriptional regulators associated with T_CM cells, including Bcl2, Tcf7, Lef1, Eomes, Foxo1, and Bach2 (Fig. 4J, 4K). In contrast, the two Ddx5^cKO clusters (2, 12) exhibited reduced expression of genes associated with effector cell differentiation and function, including Zeb2, Tbx21, Gzma, Gzmb, Prf1, Cx3cr1, and Klrg1, compared with the WT clusters (0, 9, 15, and 19) (Fig. 4J, 4K). Using protein expression derived from the CITE-seq protein panel, we observed that the WT clusters (0, 9, 15, and 19) tended to express higher levels of molecules associated with TE cells, including CX3CR1 and KLRG1, whereas the two Ddx5^cKO clusters (2, 12) tended to express higher levels of molecules associated with T_CM cells, such as CD62L and CD127 (Fig. 4L, Supplemental Fig. 4D). Taken together, these findings raised the possibility that DDX5 functions to repress the expression of T_CM cell–associated genes to promote the differentiation of effector-like subsets of CD8⁺ T cells.

The CITE-seq analyses indicated that a major consequence of DDX5 deletion was increased expression of genes that promote T_CM cell generation. We therefore sought to confirm some of the key transcriptional findings with independent flow cytometry experiments. Indeed, greater proportions of DDX5-deficient TE and MP cells expressed TCF1, and DDX5-deficient TE cells displayed increased expression of TCF1 compared with control cells at day 7 p.i. (Fig. 5A, 5B); conversely, ectopic expression of Ddx5 resulted in decreased proportions of TE and MP cells expressing TCF1 and reductions of TCF1 expression in the Ddx5^OE TE and MP cell populations (Fig. 5C, 5D). Analyses of TCF1 expression among DDX5-deficient circulating memory CD8⁺ T cell subsets revealed increased expression of TCF1 at day 30 p.i. (Fig. 5E, Supplemental Fig. 4E). The observed increase in TCF1 expression among circulating memory CD8⁺ T cells was also evident within the T_CM, T_EM, and t-T_EM cell populations, and DDX5-deficient CD8⁺ T cells displayed increased proportions of T_CM, T_EM, and t-T_EM cells expressing TCF1 (Fig. 5F, Supplemental Fig. 4F, 4G). Conversely, forced expression of DDX5 resulted in decreased expression of TCF1 among circulating memory CD8⁺ T cells (Fig. 5G), which was significant within the T_CM and T_EM cell subsets (Fig. 5H). Importantly, forced expression of Ddx5 also resulted in decreased proportions of T_CM, T_EM, and t-T_EM cells expressing TCF1 (Supplemental Fig. 4H, 4I). Similarly, Eomes expression was increased in DDX5-deficient CD8⁺ T cells as well as within the T_CM and t-T_EM cell populations (Fig. 5I, 5J); moreover, increased proportions of DDX5-deficient t-T_EM cells expressed Eomes (Supplemental Fig. 4J–L). Conversely, ectopic expression of Ddx5 resulted in reduced proportions of cells expressing Eomes in the total circulating memory CD8⁺ T cell pool and among T_CM, T_EM, and t-T_EM cell subsets (Fig. 5K, 5L). Moreover, ectopic expression of Ddx5 resulted in reduced proportions of Eomes^hi T_CM, T_EM, and t-T_EM cells (Supplemental Fig. 4M, 4N). Furthermore, DDX5-deficient CD8⁺ T cells exhibited increased coexpression of Eomes and TCF1 (Supplemental Fig. 4O). Taken together, these data suggest that DDX5 may promote the differentiation and maintenance of t-T_EM cells by repressing genes associated with the generation of T_CM cells.

RNA-binding proteins are important regulators of cellular differentiation due to their unique ability to operate at multiple levels of gene regulation, but their role in immune cell biology has not been extensively explored (19, 20). In this study, we identify the RNA-binding protein DDX5 as a regulator of effector and circulating memory CD8⁺ T cell differentiation in response to microbial infection.

Although DDX5 has been implicated in nearly every aspect of genetic regulation from chromatin organization to RNA metabolism and is suggested to broadly regulate cell differentiation, the role of DDX5 in T cells during acute infection is only beginning to be explored (21–23). We show in the current study that DDX5-deficient CD8⁺ T cells exhibited accelerated proliferation and reduced survival compared with controls, implicating DDX5 in the regulation of CD8⁺ T cell proliferation. However, the substantial CFSE dilution was accompanied by a small difference in the proportion of Ki67⁺ cells, suggesting that factors other than cell cycle entry might be contributing to the observed phenotype. Furthermore, consistent with our previously reported role for DDX5 in regulating the differentiation of an effector-like subset of tissue-resident memory cells (9), we demonstrate that deletion of DDX5 resulted in a reduction in the proportions of TE, T_EM, and t-T_EM cells, along with increased proportions of MP and T_CM cells. In established memory cells, Ddx5 deletion promoted maintenance of the T_CM population while simultaneously limiting the t-T_EM population. Some observed differences between WT and Ddx5^iKO cells were modest, representing an important limitation of the current study. Furthermore, DDX5-deficient CD8⁺ T cells exhibited increased expression of key genes known to be involved in T_CM cell differentiation, such as Tcf7 and Eomes. Taken together, these findings raise the possibility that DDX5 regulates differentiation of effector-like CD8⁺ T cell subsets by virtue of repressing genes that promote T_CM cell generation.

There are several potential mechanisms by which DDX5 may repress expression of genes associated with T_CM cells. Multiple TCF1 isoforms have been identified in T cells (24, 25); the long isoforms possess a β-catenin binding domain and are dispensable for TE cell differentiation, but they are necessary for the generation of T_CM cells (26). Our data suggest that DDX5 may repress TCF1 expression, raising the possibility that DDX5 could regulate the differentiation of TE and t-TEM cells through alternative splicing of TCF1 and shifting the composition of the TCF1 isoform pool to favor the short isoforms (27–29). Such a possibility will be investigated in future studies. Furthermore, prior studies have described the intrinsic histone-modifying capabilities of TCF1 and its homolog, LEF1 (30), suggesting the possibility that repression of TCF1 expression by DDX5 may shift the epigenetic landscape to favor expression of genes associated with TE cell differentiation. Another potential mechanism by which DDX5 could repress expression of genes associated with T_CM cells is through an association with the polycomb repressive complex 2 (PRC2), thereby mediating the subsequent silencing of memory genes through deposition of repressive H3K27me3 marks. Indeed, recent reports have posited a role for long noncoding RNAs (lncRNAs) in enabling targeted PRC2-mediated gene silencing (27, 28). Recently published global RNA interactions with DNA by deep sequencing (GRID-seq) data revealed that the lncRNA Malat1 interacts at genomic loci associated with H3K27me3 deposition in TE cells; furthermore, short hairpin RNA–mediated knockdown of Malat1 inhibited TE and t-T_EM cell differentiation (31, 32). Coupled with the well-established capacity of DDX5 to bind lncRNAs, these findings raise the possibility that lncRNAs, such as Malat1, may guide DDX5/PRC2 activity for sequence-specific silencing of T_CM cell–associated transcripts in TE and t-T_EM cells. Future studies exploring the putative relationships between DDX5, PRC2, and lncRNAs may reveal previously unknown mechanisms of sequence-specific PRC2-directed CD8⁺ T cell differentiation.

The possibility of DDX5-mediated repression of the T_CM cell–associated transcriptional program may be especially pertinent in the context of chronic infections and cancer. In the current study, deletion of DDX5 appeared to derepress Tcf7 expression, potentially allowing for the preferential differentiation of T_CM cells. The ability of TCF1 to preserve cellular plasticity has recently garnered much attention, as many groups have shown that this property is essential for maintaining a stem-like population of CD8⁺ T cells that may play a critical role in chronic infections and cancer (33–36). Furthermore, high expression of DDX5 has been correlated with aggressive tumor features, including treatment resistance, and has been shown to be predictive of poor clinical outcomes in many cancer types (37, 38). Taken together, these insights suggest that targeting DDX5 in cancer may inhibit DDX5-driven tumorigenesis and drug resistance within the tumor, as well as potentially skewing the immune response toward TCF1-expressing stem-like CD8⁺ T cells. Future studies evaluating the efficacy of DDX5-deficient CD8⁺ T cells in controlling tumor growth may provide support for DDX5 as a target for cancer therapy.

Overall, our study has identified a role for DDX5 in promoting the differentiation of TE, T_EM, and t-T_EM cells, potentially through repression of T_CM cell–associated gene expression. These findings advance our understanding of the role of RNA-binding proteins in regulating CD8⁺ T cell differentiation and serve as the basis for future studies aimed at improving CD8⁺ T cell responses in infection and cancer.

J.T.C. reported grants from Takeda and Eli Lilly outside the submitted work. The other authors have no financial conflicts of interest.