Stem-like memory and precursors of exhausted T cells share a common progenitor defined by ID3 expression

Abstract

Stem-like T cells are attractive immunotherapeutic targets in cancer patients given their ability to proliferate and differentiate into effector progeny. Thus, identifying T cells with enhanced stemness and understanding their developmental requirements is of broad clinical and therapeutic interest. Here, we demonstrate that during acute infection, the transcriptional regulator ID3 identifies stem-like T cells that are uniquely adapted to generate precursor of exhausted ‘Tpex’ cells in response to chronic infection or cancer. Expression of ID3 itself enables Tpex cells to sustain T cell responses in chronic infection or cancer, while loss of ID3 results in impaired maintenance of CD8⁺ T cell immunity. Further, we demonstrate that IL-1 family members including IL-36β and IL-18 promote the generation of ID3⁺ T cells that mediate superior tumor control. Overall, we identify ID3 as a common denominator of stem-like T cells in both acute and chronic infections that is specifically required to sustain T cell responses to chronic stimulation.

Introduction

CD8⁺ T cells are an essential part of the adaptive immune response against intracellular pathogens and tumors. Their central role is supported by their ability to rapidly undergo clonal expansion and produce cytokines such as IFN-γ and TNF, and cytotoxic molecules such as granzymes and perforin. Most effector T cells are short-lived and undergo apoptosis once an infection has been cleared, leaving only a small fraction of memory T cells behind (1, 2). Memory T cells are quiescent, long-lived, and, in some instances, retain stem-like properties, allowing them to vigorously proliferate and differentiate into large numbers of effector T cells after reinfection while also giving rise to another generation of long-lived pluripotent memory T cells (3, 4). Given sufficient time between challenges, memory T cells retain stemness and can cycle between rapid cell division and re-formation of quiescent memory T cells that can last even beyond their organismal lifespan (5). Memory T cells are heterogenous and organized in a developmental hierarchy based on stem-like potential and tissue distribution. Central memory T cells predominantly reside in lymphoid organs and are equipped with superior proliferative potential compared to effector memory T cells that circulate through the periphery or tissue-resident memory T cells that reside in non-lymphoid tissues (2). Furthermore, stem-like memory T cells have been identified in circulation in humans and mice (4, 6, 7); however, their precise identity and function have remained controversial (6, 8).

In contrast to acutely resolved infections, cancer and chronic infections such as those by Human Immunodeficiency Virus (HIV) or Hepatitis B and C Virus (HBV/HCV) induce an alternative T cell differentiation pathway that leads to the functional exhaustion of T cells. Exhausted T cells are defined by sustained expression of inhibitory receptors, such as PD-1, and an impaired ability to secrete effector cytokines (9–11). To sustain functionality in response to ongoing stimulation, exhausted CD8⁺ T cells adapt by preserving their cellular metabolism (12–14) and undergo substantial transcriptional and epigenetic changes. Exhausted T cells are heterogenous, consisting of two major cell subsets: exhausted effector T (Tex) and precursors of exhausted T (Tpex) cells. While Tex cells, which can further be segregated into CX3CR1⁺ effector-like and CD101⁺ terminally differentiated Tex cells (15, 16), retain some level of effector function, Tpex cells largely lack cytotoxic function but retain high proliferative and developmental potential (10, 17). These stem-like features allow Tpex cells to continuously self-renew while also replenishing the pool of Tex cells (18–22), and mediate the response to PD-1 immune checkpoint inhibition (20–23). Thus, Tpex cells adopt a distinct molecular profile, combining transcriptional, epigenetic and metabolic features of both memory and exhausted T cells (10, 11, 17). While they display hallmarks of exhaustion, they simultaneously express key transcriptional regulators associated with memory T cells, including TCF1, FOXO1 and BCL6, which play a critical role in the generation and maintenance of functional Tpex cells (10, 17, 18, 20–22, 24, 25). Overall, a decade of research has shown that stem-like memory and Tpex cells play critical roles in maintaining T cell immunity to ongoing infection and cancer (1, 10, 11, 17, 26) and are key targets of cancer immunotherapy (8, 27, 28); however, their developmental relationships are still poorly understood. Using single cell RNA sequencing (scRNAseq), here we explore the differentiation of memory and Tpex cells that arise in response to different viral infections. We identify a lineage of precursor and mature memory T cells in acute infection, marked by the expression of the transcriptional regulator ID3 and the surface marker Ly108, that closely resembles Tpex cells that arise during chronic infection. In contrast to conventional ID3^- memory T cells that are selectively adapted to acute responses, ID3⁺ T cells retain higher multipotency and are capable of mounting efficient recall T cell responses to both acute and chronic stimulation. ID3 was required to sustain a protracted CD8⁺ T cell response by promoting the ongoing proliferative potential of Tpex cells. Finally, we show that IL-1 family cytokines, such as IL-36β or IL-18, promoted the generation of ID3⁺ T cells in vitro, which showed superior stemness and proliferative capacity when exposed to chronic infection or tumor challenge. Thus, our data reveal that stem-like memory T cells and Tpex cells share a common progenitor that arises at early stages in both acute and chronic infections and is characterized by the expression of ID3.

Results

ID3 identifies stem-like memory precursor cells in acute infection

To investigate the relationship between memory and Tpex cells, we performed single-cell RNA sequencing (scRNAseq) of T cell receptor (TCR) transgenic P14 CD8⁺ T cells, which are specific for the Lymphocytic Choriomeningitis Virus (LCMV)-derived gp33 epitope. We compared P14 cells isolated on day 8 or 21 from mice infected with acute LCMV Armstrong with P14 cells isolated from chronic LCMV clone-13 on day 21. As expected, uniform manifold approximation and projection (UMAP) separated P14 cells largely according to the type of infection (Fig. 1, A and B). Cells from chronically infected mice expressed elevated levels of exhaustion-related genes such as Pdcd1 (PD-1), Lag3 and Tox (fig. S1, A to C), and could be separated into three main clusters: Tpex (Tcf7^hi), effector-like (Cx3cr1^hi) and terminally differentiated Tex (Cd101^hi) cells (Fig. 1C; fig. S1, D to F). P14 cells from an acute LCMV infection clustered mainly into effector (Teff) and memory T cells based on expression of Klrg1 and Cx3cr1 or Il7r and Tcf7, respectively (Fig. 1, A to C; fig. S1, D, E, G and H).

Fig. 1. ID3 identifies stem-like memory precursors in acute infection.

(A to E) P14 cells were isolated 8 (acute only) or 21 (acute and chronic) days post acute (Armstrong) or chronic (clone-13) LCMV infection for scRNAseq. (A) UMAP plot of 2,916 Armstrong d8-derived (light green), 5,853 Armstrong d21-derived (dark green) and 6,605 clone-13 derived P14 cells (black). Cells were pooled from 1-2 experiments with 5 mice each. (B) UMAP plot highlighting individual P14s from d8 (left) and d21 acute (middle) and chronic infection (right). (C) UMAP colored by annotated T cell subsets. (D) Normalized expression of Tcf7 and Il7r among P14s from d8 acute infection. MP, memory precursor. (E) Volcano plot showing differentially expressed genes (FDR <0.05) between MP1 and MP2. (F to K) 5,000 Id3^GFP/+ P14s were transferred into congenically marked naïve mice, infected with acute (Armstrong, F to K) and chronic (Docile, K) LCMV, and analyzed on day 8. (F) TCF1 or CD127 versus KLRG1 expression. (G) ID3 or Ly108 versus CXCR6 expression among CD127⁺KLRG1^- P14s derived from peripheral lymph nodes (LN), spleen or blood, and frequencies of ID3⁺ MP cells across all three organs. GFP, green fluorescent protein. (H) TCF1, CD127, CD95 and CXCR3 expression among ID3⁺ and ID3^- MP cells as well as KLRG1⁺ effector T (Teff) cells as depicted in F and G derived from spleen. (I and J) Frequencies (I) and numbers (J) of CD62L⁺ T cells among ID3⁺ and ID3^- MP cells as well as KLRG1⁺ Teff cells. (K) PD-1 and TOX expression of ID3⁺ and CXCR6⁺ MP cells compared to ID3⁺ Tpex cells obtained from chronic (Docile) LCMV infected. Symbols (G, I and K) represent individual mice. Data are combined (G, H to J) or representative (H and K) of at least two independent experiments with at least 4 mice. Statistical analyses were performed using one-way ANOVA and Tukey’s multicomparison test (H to K).

Notably, P14 cells from day 8 post acute LCMV infection contained two populations of memory precursor (MP) cells based on elevated expression of Tcf7 and Il7r (Fig. 1, A to D). While one Tcf7⁺ population (MP1) clustered closely to Klrg1⁺ effector T cells, the other Tcf7⁺ population (MP2) clustered within Tpex cells derived from chronic LCMV (Fig. 1, A to D). In line with our results, published scRNAseq data from T cells isolated at day 7 post LCMV Armstrong infection (29) could similarly be segregated into two MP populations, with one population clustering closely to Tpex cells (fig. S1I). Gene expression comparison between both Tcf7⁺ MP subsets revealed 246 differentially expressed genes (data file S1) including memory-associated genes Sell (CD62L), Ccr7 and Id3, which were expressed higher in the MP2 population, whereas Ccr2, Cxcr6 and Id2 were expressed higher in the MP1 subset (Fig. 1E). Consistent with our scRNAseq data, flow cytometry analysis of adoptively transferrred Id3 reporter P14 cells (24, 30) from mice infected with acute LCMV 8 days before confirmed that CD127⁺KLRG1^- T cells could be segregated into two populations based on the expression of ID3 or Ly108 and CXCR6 (Fig. 1, F and G). While conventional MP1 cells were ID3^lo, Ly108^lo and CXCR6^hi and dominated in the spleen and blood, MP2 cells were ID3^hi, Ly108^hi, and negative for CXCR6 and were predominantly found in lymph nodes (Fig. 1G). ID3⁺ MP2 cells compared to ID3^- MP1 cells expressed higher levels of TCF1, CD127 (IL-7R) and CD95 but similar levels of CXCR3 (Fig. 1H). While ID3⁺ MP2 cells contained a higher fraction of CD62L⁺ cells (Fig. 1I), the number of CD62L⁺ T cells was highest among KLRG1⁺ effector T cells, highlighting that CD62L alone was not a suitable marker to distinguish different memory precursor and effector subsets in the spleen (Fig. 1J). ID3⁺ MP2 compared to ID3^- MP1 cells expressed modestly elevated levels of exhaustion-related molecules such as PD-1 or TOX; however, both populations showed low expression compared to Tpex cells derived from day 8 chronic infection (Fig. 1K). Thus, our data demonstrate that during a primary T cell response to acute LCMV, a subset of ID3⁺ memory precursor cells develops, which resembles Tpex cells found in chronic infection.

ID3⁺ precursor cells can give rise to CXCR6⁺ memory T cells

To understand the relationship between CXCR6⁺ID3^- MP1 and ID3⁺CXCR6^- MP2, we isolated both subsets as well as KLRG1⁺ Teff cells on day 8 post acute LCMV (Armstrong) infection and transferred them into time-matched acute LCMV infected mice (Fig. 2A). ID3⁺ Tpex cells isolated on day 8 post chronic LCMV (Docile) infection and transferred into time-matched acute LCMV infected mice served as an additional control. Two weeks post-transfer, we obtained larger numbers of cells derived from CXCR6⁺ MP1 cells compared to ID3⁺ MP2, KLRG1⁺ Teff or Tpex cells (Fig. 2B). While both MP subsets and Tpex cells maintained high CD127 expression, KLRG1⁺ Teff remained CD127^- (Fig. 2C). MP and Tpex cells gave rise to CD62L⁺ memory cells, with the highest fraction of CD62L⁺ cells among cells that derived from ID3⁺ MPs (Fig. 2D). Notably, only memory T cells derived from ID3⁺ MP and Tpex cells expressed ID3 (Fig. 2, E and F). In contrast, the progeny of CXCR6⁺ MP1 cells retained high CXCR6 but remained negative for ID3 expression. Critically, some ID3⁺ MP2 or Tpex-derived cells also expressed CXCR6 (Fig. 2, E and G). These observations indicate that ID3 expression defines a multipotent precursor population that is found in both acute and chronic infection, can maintain their population and give rise to memory cells defined by elevated expression of CXCR6.

Fig. 2. ID3⁺ precursor cells give rise to CXCR6+ memory T cells.

(A) Experimental scheme. 10,000 Id3^GFP/+ P14 cells were transferred into congenically marked naïve mice prior to infection with acute (Armstrong) or chronic (Docile) LCMV. 8 days post-infection, ID3⁺CXCR6^- and CXCR6⁺ID3^- MP cells as well as KLRG1⁺ effector T cells and ID3⁺ Tpex cells were sorted and 5×10⁵ cells transferred into time-matched LCMV Armstrong infected mice. Two weeks post-transfer, donor-derived P14 cells were isolated from the spleen and analyzed. (B) P14 numbers in the spleen. (C) CD127 versus KLRG1 expression among transferred P14 cells on day 22 and frequencies of CD127⁺ memory T cells (right). (D) CD127 versus CD62L expression among CD127⁺ T cells and frequencies of CD62L⁺ memory T cells in all mice (right). (E to G) ID3 versus CXCR6 expression among CD127⁺ T cells (E) and frequencies of total ID3⁺ (F) and CXCR6⁺ID3^- (G) memory T cells in all mice. GFP, green fluorescent protein. Symbols (B to D, F and G) represent individual mice. Data are either combined (F and G) or representative (B to D) of two independent experiments with at least 4 mice. Statistical analyses were performed using one-way ANOVA and Tukey’s multicomparison test (B to G).

ID3 defines memory T cells preadapted to chronic stimulation

Next, we examined among our scRNA data (Fig. 1) the distribution of the transcriptional signature of ID3⁺ MP2 cells (data file S2) within the established memory and exhausted T cells derived 21 days post-infection. Notably, together with Tpex cells, a small subset of memory T cells from day 21 acute LCMV showed enrichment of the ID3⁺ MP2 cell signature (Fig. 3A), with Id3 and Cxcr6 being among the genes that distinguished this subset of memory T cells (Fig. 3B). Flow cytometry analysis confirmed that CD62L⁺ memory T cells from acutely resolved (day 24) LCMV infection could be segregated into two populations based on the expression of ID3. ID3⁺ memory cells were predominantly found in lymph nodes, present in spleen but absent in peripheral blood (Fig. 3C). Similar to ID3⁺ MP2 cells (Fig. 1), ID3⁺ memory cells expressed higher levels of Ly108 while lacking CXCR6 expression compared to ID3^- or CD62L^- memory T cells (fig. S2A). Similar observations were made when examining memory T cells that arose following bacterial infection with Listeria monocytogenes in mice (fig. S2, B to G). Human CD8⁺ memory T cells could also be segregated based on ID3 and CXCR6 expression as re-analysis of scRNAseq data from SARS-Cov-2-specific memory T cells (31) revealed the presence of Sell⁺ cells that could be segregated based on Id3 and Cxcr6 expression (fig. S2, H to K). Consistent with our observations in mice, human Id3⁺ memory T cells showed an enrichment of the MP2⁺ cell signature (data file S2) compared to Cxcr6⁺ cells (fig. S2L).

Fig. 3. ID3 defines memory T cells preadapted to chronic stimulation.

(A) Enrichment of ‘ID3⁺ MP2 signature’ (data file S2) across scRNAseq data described in Fig. 1. (B) Normalized gene expression of Id3 and Cxcr6. Dotted area highlights MP2-enriched cells among acute d21-derived cells. (C) 50,000 Id3^GFP/+ P14s were transferred into congenically marked naïve mice, infected with acute (Armstrong) LCMV, and analyzed on day 24. ID3 versus CD62L expression among P14s derived from peripheral lymph nodes (LN), spleen or blood, and frequencies of ID3⁺ (dark green), ID3^- (light green) and CD62L^- (grey) T cells across all three organs. GFP, green fluorescent protein. (D) Experimental scheme for (E to N). ID3⁺CD62L⁺, ID3^-CD62L⁺ and ID3^-CD62L^-memory P14 cells, as depicted in C, were sorted from day 28 acute (Armstrong) LCMV infected mice, and 6,000 transferred into naïve mice prior to either acute (Armstrong; E to I) or chronic (Docile; J to N) LCMV challenge. (E and J) P14 frequencies among CD8⁺ T cells in the blood on day 7 post acute (E) or day 7 (J) and 28 (K) post chronic infection. (F and K) P14 numbers in the spleens >4 weeks post-infection with either acute (F) or chronic (K) LCMV challenge. (G) CD62L expression among transferred P14s on day 28 post acute rechallenge. (H and I) P14s from acutely infected mice were ex vivo restimulated with gp33 peptide. (H) Frequencies of TNF⁺IFN-γ⁺ double producers (H). (I) Expression and frequencies of IL-2 producers among TNF⁺IFN-γ⁺ double producers. (L) ID3 versus TCF1 expression among transferred P14s on day 28 post chronic LCMV challenge. GFP, green fluorescent protein. (M) Numbers of ID3^-TCF1^- Tex cells. (N) CX3CR1 versus granzyme B (GzmB) expression among Tex of transferred P14s on day 28. Symbols (C and E to N) represent individual mice. Data are combined of 2-3 independent experiments with at least 4 mice. Statistical analyses were performed using one-way ANOVA and Tukey’s multicomparison test (C and E to N).

To test the developmental potential of both CD62L⁺ memory T cell subsets, we isolated CD62L⁺ P14 cells that were ID3⁺ or ID3^- as well as CD62L- memory P14 cells from mice that had been infected with acute LCMV and transferred equal numbers into naïve mice followed by challenge with either acute or chronic LCMV (Fig. 3D). As expected, CD62L^- memory cells in comparison to both CD62L⁺ memory subsets showed the lowest capacity to re-expand in response to acute infection (Fig. 3E). ID3⁺ and ID3^- CD62L⁺ memory cells demonstrated similar re-expansion capacity as well as formation of secondary memory T cells following re-challenge with acute LCMV (Fig. 3, E to G, fig. S3). The progeny of ID3⁺ memory cells, however, showed the highest ability to produce IL-2 compared to the ID3^- memory cells (Fig. 3, H and I). In contrast to acute re-challenge, ID3+ memory cells showed a distinct ability to sustain T cell responses in chronic infection compared to ID3^- cells (Fig. 3, J and K). Moreover, only ID3⁺ memory cells, but not ID3^- memory cells, were able to generate ID3⁺TCF1⁺ Tpex cells in chronic infection whereas the formation of CX3CR1⁺ Tex cells was intact in all three re-expanding subsets (Fig. 3, L to N). Together, our data show that ID3 identifies stem-like memory T cells with the highest proliferative and developmental potential that are specifically able to generate Tpex cells to maintain the response to prolonged infections.

ID3 is required to sustain the CD8⁺ T cell response to chronic infection

To better understand the role of ID3 in CD8⁺ T cell responses, we utilized Id3 reporter mice in which Gfp replaces the Id3 allele (24, 30). Homozygote Id3^GFP/GFP (referred to as ID3 knockout, ID3KO) and control (Id3^GFP/+) P14 cells were adoptively co-transferred into congenically marked naïve mice prior to infection with either acute or chronic LCMV (Fig. 4A). In line with previous findings in acute LCMV infection (32, 33), ID3KO P14 cells showed a modest reduction compared to control P14 cells at the peak of infection (day 8) and during the memory phase (Fig. 4, B and C). The formation of memory precursor cells in acute infection, including MP1 and MP2 cells, was more strongly impacted by the absence of ID3 than the formation of KLRG1⁺ Teff cells (fig. S4). During chronic infection, ID3KO P14 cells showed severely impaired population expansion compared to control cells and their abundance was reduced ∼60-fold by day 21 post-infection (Fig. 4, B and C). While ID3KO P14 cells were able to generate TCF1⁺ Tpex cells, they were >10-fold reduced compared to control Tpex cells five days post-infection (Fig. 4, D and E). In contrast, TCF1^- Tex cells showed only a three-fold reduction among ID3-deficient P14 cells compared to controls (Fig. 4E). By day 21 post-infection, however, both ID3KO Tpex and Tex cells were dramatically reduced compared to control P14 cells (Fig. 4F). ID3KO cells showed minor alterations in their PD-1 and TOX expression and ability to produce IFN-γ and TNF (Fig. 4, G to I), suggesting that exhaustion per se was not impacted by ID3 expression.

Fig. 4. ID3 is required to sustain the CD8⁺ T cell response to chronic infection.

(A) Experimental scheme. 2,500 Id3^GFP/+ (control, ctrl) and 2,500 Id3^GFP/GFP (ID3KO) P14 cells were co-transferred into congenically marked naïve mice prior to infection with acute (Armstrong) or chronic (Docile) LCMV. (B) Absolute numbers of control (green or black) and ID3KO (orange) P14s obtained from the spleen of mice infected with acute or chronic LCMV. (C) Fold change in control-to-ID3KO P14 ratio in chronic (black) versus acute (green) LCMV. (D) TCF1 versus TIM-3 expression among control (black) and ID3KO (orange) P14s on day 5 post-infection with chronic LCMV. Graphs show representative flow plots (left) and frequency of TCF1⁺ P14s for all mice (right). (E) Control (CD45.1/2) versus ID3KO (CD45.2) P14 ratio of TCF1⁺ Tpex cells and TIM-3⁺TCF1^- Tex cells and absolute numbers of ctrl and ID3KO Tpex and Tex cells on day 5 in all mice (right). (F) Tpex and Tex numbers of ctrl and ID3KO P14s on day 21. (G and H) PD-1 (G) and TOX (H) expression of control (black) and ID3KO (orange) Tpex (solid line) and Tex (dashed line) cells on day 35 post-infection with chronic LCMV. Graphs show representative histograms (left, host CD8⁺ T cells in grey) as well as calculated mean fluorescent index (MFI, right). (I) Frequencies of IFN-γ⁺ (left) and TNF⁺ among IFN-γ⁺ (right) TCF1⁺ Tpex and TIM-3⁺ Tex cells after ex vivo gp33-peptide re-stimulation. Symbols in (B and C) represent the means of 10 mice and in (D to I) individual mice; the lines connect P14 cells within the same host. Data are combined of 2-3 (B to F) or representative of at least two independent experiments with 3 mice (G to I). Statistical analyses were performed using either two-way ANOVA and Sidak’s multicomparison test (B and C) or paired two-tailed Student’s t-test (D to I).

To further explore the role of ID3 in CD8⁺ T cell responses, we retrovirally overexpressed ID3 (ID3OE) in P14 cells and adoptively transferred them together with control cells into mice prior to challenge with acute or chronic LCMV (fig. S5A). We found that ID3OE favored population expansion of T cells responding to chronic compared to acute infection (fig. S5, B and C). This increased T cell expansion in chronic infection was due to an enhanced generation of Tex cells in ID3OE P14 cells whereas Tpex cell formation was unaltered (fig. S5D). By day 20 post-infection, ID3OE P14 cells showed similar maintenance compared to control cells (fig. S5, E and F), indicating that overexpression of ID3 cannot further promote long-term T cell responses in chronic infection. ID proteins including ID3 negatively regulate DNA-binding activity of E-protein transcription factors such as E2A (34). To directly test the contribution of E2A activity on T cell expansion, we retrovirally overexpressed E2A (E47OE) in P14 cells and adoptively transferred them into mice infected with chronic LCMV one day earlier (fig. S5G). In agreement with our ID3KO data, overexpression of E2A limited the expansion of T cells, with both Tpex and Tex populations being impacted (fig. S5, H and I). Together, these data show that ID3 is crucial for the expansion of CD8+ T cells, particularly those responding to chronic infection.

ID3 is required for the formation of functional Tpex cells

To unravel the molecular mechanisms controlled by ID3, we isolated ID3KO as well as control Tpex and Tex cells on day 5 post-infection and performed RNA sequencing. As expected, lack of ID3 had a greater impact on Tpex cells than on Tex cells with 545 and 84 differentially expressed genes, respectively (Fig. 5, A and B; fig. S6A and data file S3). ID3-deficient Tpex cells showed downregulation of transcripts associated with cell proliferation and enhanced expression of genes associated with inflammation and apoptosis (Fig. 5C). Upregulated genes within ID3-deficient Tpex cells included genes associated with stemness and CD62L⁺ Tpex cells, such as Myb, Ccr7 and Sell (CD62L), while genes encoding TCR-responsive genes such as Ikzf2 (HELIOS), Irf4, and Pdcd1 (PD-1) were downregulated (Fig. 5B). Moreover, Kit, encoding the receptor tyrosine kinase cKIT that is expressed by hematopoietic stem cells (35) and CD62L^- Tpex cells (23), was downregulated in ID3KO Tpex cells (Fig. 5B). Flow cytometric analysis confirmed that ID3KO Tpex cells contained an increased frequency of CD62L⁺ Tpex cells while cKIT⁺ Tpex cells were reduced (Fig. 5, D and E). Critically, however, all three Tpex cell subsets (Fig. 5D) were numerically decreased throughout infection in the absence of ID3 (Fig. 5F, fig. S6, B and C), indicating that ID3 is required for the optimal expansion of all Tpex populations.

Fig. 5. ID3 is required for the formation of functional Tpex cells.

(A to C) ID3⁺ (Tpex) and TIM-3⁺ (Tex) control and ID3KO P14 cells were purified for RNAseq analysis on day 5 post chronic LCMV (Docile) infection. (A) Venn diagram showing differentially expressed genes between control and ID3KO Tpex or Tex cells. Genes upregulated in control groups are indicated in red, genes upregulated in ID3KO groups are indicated in blue. (B) Volcano plot showing differentially expressed genes (FDR <0.05) between control and ID3KO Tpex cells. (C) Differentially expressed hallmark gene sets between control and ID3KO Tpex cells. (D to F) 5,000 Id3^GFP/+ (ctrl) and 5,000 Id3^GFP/GFP (ID3KO) P14s were co-transferred into congenically marked naïve mice prior to chronic LCMV (Docile) infection. (D) CD62L versus cKIT expression among ctrl and ID3KO Ly108⁺ Tpex cells. (E) Frequencies of cKIT⁺ cells among ctrl and ID3KO Tpex cells. (F) Absolute numbers of CD62L+ ctrl and ID3KO Tpex cells. (G) Experimental scheme for H to N. ID3KO or control Tpex cells were sorted on day 5 post chronic LCMV and 10⁵ Tpex cells transferred into infection-matched mice and analyzed 7 days post-transfer. (H) Frequencies of ctrl and ID3KO cells among total P14 splenocytes. (I) TCF1 versus Ly108 expression among donor-derived P14s and frequencies of TCF1^-Ly108^- Tex cells. (J) Donor-derived ctrl and ID3KO Tpex and Tex numbers as depicted in I. (K) CXCR3 expression among Tpex cells. (L) Ki67 expression among co-transferred ctrl and ID3KO Tpex cells. (M) CD62L versus cKIT expression among Tpex cells. (N) CX3CR1 versus Ki67 expression among Tex cells. RNAseq was performed with two experimental replicates (each n=10). Symbols (D and E) represent the means of 10 mice and in (F and H to N) individual mice, the lines connect P14s within the same host. Data are representative of (K) or combined (D to J and L to N) of two independent experiments with at least 3 mice. Statistical analyses were performed using paired, two-tailed Student’s t-test (D to N).

To directly test the proliferative and developmental potential of ID3-deficient Tpex cells, we isolated ID3KO or control Tpex cells on day 5 post chronic LCMV infection and transferred equal numbers into infection-matched mice (Fig. 5G). 7 days post-transfer, transferred ID3KO Tpex cells showed impaired expansion compared to control Tpex cells, giving rise to lower numbers of both Tpex and Tex cells (Fig. 5, H to J). Donor-derived ID3-deficient and control Tpex cells expressed similarly high levels of CXCR3 (Fig. 5K), which is critical for Tpex differentiation and function (36). In contrast, ID3-deficient Tpex cells showed low expression of Ki67 compared to control Tpex cells (Fig. 5L), indicating impaired Tpex cell proliferation in the absence of ID3.

ID3-deficient compared to control Tpex cells also maintained a higher proportion of CD62L⁺ Tpex versus cKIT⁺ Tpex cells (Fig. 5M). In contrast, ID3KO Tex cell progeny showed unimpaired Ki67 expression and modestly increased differentiation of CX3CR1⁺ cells (Fig. 5K), indicating that proliferation and differentiation of Tex cells is ID3-independent.

To evaluate the contribution of ID3 within individual Tpex subsets, we performed RNAseq on CD62L⁺ and CD62L^-cKIT^- Tpex cells from ID3KO and control P14 cells on day 12 post chronic LCMV infection (fig. S6D). ID3 deficiency affected a total of 171 genes across both subsets, with 61 genes being shared between the two cell types, including Id3, Cxcl10, Ifit3, Gzmk and Dnase1l3 (fig. S6, E to G, and data file S4). Differentially expressed genes downregulated in ID3KO cells related to pathways linked to proliferation and metabolic status (fig. S6, F and G). Indeed, both CD62L⁺ and CD62L^-cKIT^- ID3KO Tpex cells showed decreased expression of Ki67 compared with their WT counterparts (fig. S6, H and I). Overall, our data show that ID3 is essential to ensure the proliferative capacity and subsequent self-renewal of Tpex cells responding to chronic infections.

cKIT⁺ Tpex cells exihibit low stem-like potential

Our data so far showed that ID3 regulates Tpex cell proliferation and differentiation of cKIT⁺ Tpex cells. To better understand the identity of cKIT⁺ Tpex cells, we isolated CD62L⁺, cKIT⁺ and CD62L^-cKIT^- Tpex cells and performed RNA sequencing on day 5 post-infection. CD62L⁺ and cKIT⁺ Tpex cells showed the strongest transcriptional differences as evidenced by 572 differentially expressed genes (Fig. 6, A and B; fig. S7 and data file S5) with CD62L^-cKit^- Tpex cells exhibiting an intermediate transcriptional profile (Fig. 6B). cKIT⁺ Tpex cells, compared to both CD62L⁺ and CD62L^-cKit^- Tpex cells, were enriched for genes involved in cell cycling and anabolic metabolism (Fig. 6C), suggesting that they were in a more activated state. To directly compare the developmental potential of each Tpex cell subset, we isolated CD62L⁺, CD62L^-cKit^- and cKIT⁺ Tpex cells on day 12 post-infection and transferred equal numbers into infection-matched recipients (Fig. 6D). CD62L⁺ Tpex cells showed the overall highest capacity to generate progeny, while cKIT⁺ Tpex cells showed the lowest ability to sustain P14 cells in chronic infection (Fig. 6E). In line with their high stem-like capacity, CD62L⁺ Tpex cells maintained overall higher numbers of TCF1⁺ Tpex cells while also giving rise to higher numbers of CX3CR1⁺ Tex cells compared to transferred cKIT⁺ Tpex cells (Fig. 6, F to H). CD62L⁺ Tpex cells were able to self-renew and give rise to CD62L^- Tpex cell progeny (Fig. 6I), whereas both CD62L^-cKit^- and cKIT⁺ Tpex populations primarily formed CD62L^-cKit^- Tpex cells (Fig. 6, I and J). Taken together, our data indicate that cKIT expression on Tpex cells is transient while demarcating Tpex cells with the lowest stem-like potential.

Fig. 6. cKit⁺ Tpex cells exihibit low stem-like potential.

(A to E) 10,000 Id3^GFP/+ P14 cells were transferred into congenically marked naïve mice prior to infection with chronic LCMV (Docile). 5 days post-infection, CD62L⁺ (purple), CD62L^-cKit^- (pink) and cKIT⁺ Tpex cells (blue) were purified for RNA-seq analysis. (A) Scheme showing differentially expressed genes between all three subsets. (B) Principal component analysis (PCA) of all RNA samples. (C) Differentially expressed hallmark gene sets enriched in cKIT⁺ Tpex cells compared to CD62L⁺ (left) or CD62L^- cKit^- Tpex cells (right). (D) Experimental scheme for E to I. 5,000 Id3^GFP/+ P14 cells were transferred into congenically marked naïve mice prior to infection with chronic LCMV (Docile). 12 days post-infection, CD62L⁺ (purple), CD62L^-cKIT^- (pink) and cKIT⁺ Tpex cells (blue) were sorted, and equal numbers (2.5-5×10⁶ cells) transferred into time-matched LCMV Docile infected mice. Two weeks post-transfer, donor-derived P14 cells were isolated from the spleen and analyzed. (E) P14 numbers in the spleen. (F) TCF1 versus CX3CR1 expression among transferred P14 cells. (G) Frequencies of CX3CR1⁺ Tex cells among P14 cells in all mice. (H) Numbers of TCF1⁺ Tpex (left) and CX3CR1⁺ Tex cells (right). (I) CD62L versus cKIT expression among donor-derived Tpex cells. (J) Frequencies of CD62L⁺, CD62L^-cKit^- and cKIT⁺ Tpex cells among donor-derived Tpex cells. RNA-seq analysis was performed with two experimental replicates. Symbols (E, G, H and J) represent individual mice. Data are combined of two (E to J) independent experiments with at least 3 mice. Statistical analyses were performed using one-way ANOVA and Tukey’s multicomparison test (E, G, H and J).

IL-1 family members such as IL-36β promote the formation of ID3⁺ T cells

Given the importance of ID3 in sustaining the CD8⁺ T cell response to chronic infection, we next explored potential modulators of ID3 expression. To this end, we performed an in vitro cytokine screen by expanding Id3^GFP P14 cells with gp33-peptide in the presence of 21 different cytokines (Fig. 7A). We first classified the impact of different cytokines on population expansion and the generation of multipotent T cells based on TCF1 expression. Among all conditions, T cell stimulation in the presence of IL-36β, IL-7, IL-15, IL-18, IL-2, IL-36α or IL-33 was most potent in promoting the generation of TCF1⁺ T cells (fig. S8, A and B). However, only IL-1 cytokine family members, including IL-36β, IL-18, IL-36α and IL-33, promoted the generation of T cells expressing ID3 in addition to TCF1 (Fig. 7, B and C; fig. S8, C and D). Increased expansion of T cells in the presence of IL36β or IL-18 was dependent on ID3 as ID3KO P14 cells showed impaired proliferation in the presence of these cytokines, whereas T cell expansion in the presence of IL-2 was independent of ID3 (Fig. 7D). IL-36β stimulation compared to IL-2 resulted in significantly enhanced expression of the TCR-induced transcriptional regulator Nur77 (Fig. 7E), which is in line with the role of strong TCR signals driving expression of ID3 (37). To evaluate the ability of in vitro activated ID3⁺ T cells to respond to chronic infection, we transferred either IL-2- or IL-36β-stimulated P14 cells into mice that were infected 6 days earlier with chronic LCMV (Fig. 7F). Compared with IL-2, IL-36β-stimulated T cells showed increased expansion of both Tpex and Tex cells (Fig. 7, G to I), indicating that T cell activation in the presence of IL-1 family members, such as IL-36β, can promote the formation of ID3⁺ T cells with enhanced capacity to sustain CD8⁺ T cell responses in chronic infection.

Fig. 7. IL-1 family members such as IL-36β promote the formation of ID3⁺ T cells.

(A) Experimental scheme for (B to D). Id3^GFP/+ (control, ctrl) or Id3^GFP/GFP (ID3KO) P14 cells were in vitro activated with gp33-peptide in the presence of different cytokines as indicated and analyzed 5 days post-activation. (B) ID3 versus Ly108 expression on activated P14 cells. GFP, green fluorescent protein. (C) Numbers of ID3^lo (white) and ID3^hi (blue) P14 cells 5 days post-activation. (D) Numbers of ctrl and ID3KO P14 cells expanded with gp33-peptide in the presence of IL-2 (left), IL-36β (middle) or IL-18 (right) following in vitro activation. (E) Nur77^Tempo-P14 cells were in vitro activated with gp33-peptide in the presence of IL-2 (light red), IL-36β (blue) or without cytokine (grey; no cytokine, NC). Shown is representative histogram of Nur77 expression (left) and quantified MFI (right). (F) Experimental scheme for G to I. Congenically marked P14 cells in vitro primed with IL-2 or IL-36β (as in A) and 5×10⁵ P14 cells were transferred into day 6 chronically LCMV (Docile) infected mice and analyzed 9 days post-transfer. (G) Frequency (left) and absolute numbers (right) of total P14 cells primed with IL-2 (light red) or IL-36β (blue) on day 15 post-infection. (H) Frequency of IL-2- or IL-36β-primed TCF1⁺ Tpex cells on day 15 post-infection. (I) Frequency and numbers (right) of granzyme B (GzmB)⁺ Tex cells among IL-2- or IL-36β-primed cells. Data are representative of three experiments (B to E) or combined (G to I) of two independent experiments with 4-5 mice. Statistical analyses were performed using unpaired, two-way ANOVA and Sidak’s multicomparison test (D), two-tailed Student’s t-test (G to I) or one-way ANOVA and Tukey’s multicomparison test (E).

IL-36β- and IL-18-induced ID3⁺ T cells mediate superior tumor control

To test the role of ID3 in regulating CD8⁺ T cell responses to cancer, we transferred ID3KO or control P14 cells into mice bearing B16F10-gp33 tumors (Fig. 8A). In line with previous work (38), ID3KO P14 cells showed an impaired ability to control tumor growth (Fig. 8, B and C). Considering that IL-36β treatment promoted the expression of ID3 in CD8⁺ T cells and their subsequent expansion in chronic infection (Fig. 7), we next evaluated the ability of IL-1 family cytokine-elicited T cells to mediate protection in a tumor model. To this end, we expanded OT-I transgenic T cells specific for the ovalbumin (OVA)-derived SIINFEKL peptide in the presence of different cytokines, including IL-1 family cytokines, and then transferred equal numbers into B16F10-OVA-bearing mice (Fig. 8D). Notably, mice that received OT-I cells primed in the presence of IL-36β or IL-18 were most efficient in controling tumor growth (Fig. 8E; fig. S9, A and B), resulting in significantly prolonged survival (Fig. 8F and fig. S9C). While overall tumor infiltration was similar in all conditions (Fig. 8G; fig. S9, D and E), both IL-36β- and IL-18-stimulated OT-I cells showed an increased frequency in the tumor-draining lymph nodes (Fig. 8, H and I). Further, IL-18- and IL-36β-stimulated cells displayed enhanced formation of Tpex cells in the tumor compared to IL-2-stimulated T cells, while still sustaining cytotoxic potential based on granzyme B expression (Fig. 8, J and K). IL-36β- and IL-18-stimulated OT-I cells also showed increased tendencies to proliferate based on Ki67 expression (Fig. 8L). Thus, in vitro activated T cells primed in the presences of IL-1 family cytokines such as IL-36β and IL-18 elicited superior tumor control, along with increased accumulation of T cells in tumor-draining lymph nodes and enhanced preservation of Tpex cells within the tumor.

Fig. 8. IL-36β- and IL-18-induced ID3⁺ T cells mediate superior tumor control.

(A) Experimental scheme for B and C. 8×10⁴ Id3^GFP/+ (ctrl) or Id3^GFP/GFP (ID3KO) P14 cells were transferred into naïve mice before B16F10-gp33 inoculation with cells. (B) B16F10-gp33 growth in mice. Shown are the mean growth and all individual mice. (C) Survival of tumor-bearing mice after engraftment with ctrl or ID3KO P14s. (D) Experimental scheme for E to L. OT-I T cells were in vitro activated with SIINFEKL peptide in the presence of either IL-2, IL-18 or IL-36β before injecting 5×10⁶ cells into B16F10-Ova bearing mice. (E and F) B16F10-Ova growth in mice (E) and overall survival (F) compared to mice that did not receive OT-Is. (G to I) Frequencies and numbers of IL-2 (light red), IL-18 (dark blue) and IL-36β (blue) in vitro primed OT-Is in tumors 13 days post-transfer (G) or lymph nodes (LN) 6 (H) and 13 (I) days post OT-I transfer. (J) TCF1 versus granzyme B (Gzmb) expression of IL-2, IL-18 and IL-36β primed OT-I cells in the tumor on day 13. Graphs show representative flow plots gated on CD44⁺PD-1⁺ OT-I cells (left) and frequencies of TCF1+ Tpex cells (right) in the tumor. (K) Frequencies and numbers of GzmB⁺ T cells in the tumor. (L) TCF1 versus Ki67 expression of OT-Is in the tumor on day 13 post OT-I transfer. Symbols represent individual mice (G to L); lines represent either individual (B and E) or the means of 6-13 mice per group or individual mice (C and F). Data are representative of two (B and C) or three experiments (E and F), or combined (G to L) of at least two independent experiments with 2-5 mice. Statistical analyses were performed using two-way ANOVA and Dunnett’s multicomparison test (B), one-way ANOVA and Tukey’s multicomparison test (G to L) or log rank (Mantel-Cox) test (C and F).

Discussion

CD8⁺ T cell responses to chronic infections and cancer are sustained by a specialized subset of precursors of exhausted T (Tpex) cells that retain self-renewal capacity while also giving rise to exhausted effector T (Tex) cells (10, 17, 21, 22). Tpex cells combine critical transcriptional and epigenetic features of exhausted and memory T cells (10, 17), and their generation and maintenance depend on transcriptional regulators associated with memory T cells including transcription factors such as TCF1, FOXO1 and BCL6 (18, 19, 21, 22, 25), as well as factors that mediate stemness such as MYB (23). Here, we identify in different acute infections a T cell subpopulation within the precursor-memory lineage defined by the expression of ID3 that displays characteristics of Tpex cells and retains high developmental potential. In contrast to ID3^- memory T cells, ID3⁺ memory T cells were exclusively able to generate Tpex cells in response to chronic infection and thereby showed superior ability to sustain a T cell response. We show that ID3 was required for sustained proliferation of Tpex cells in chronic infection, including the formation of cKIT⁺ Tpex cells. Finally, we demonstrate that IL-1 family members, including IL-36β and IL-18, specifically promoted the expansion of ID3-expressing T cells in vitro that mediated superior T cell responses in chronic infection and elicited improved tumor control. Thus, we here identify ID3 as a common denominator of stem-like T cells that arise in response to different infections, and show that expression of ID3 plays a critical role in maintaining long-lasting CD8⁺ T cell immunity in response in chronic infection and cancer.

Stem-like memory T cells are considered to be superior in maintaining long-term CD8⁺ T cell immunity to infection and have been promoted as prime targets in cancer immunotherapy (1, 4, 8). However, their precise identity and developmental requirements have remained poorly understood. In contrast, our understanding of the mechanisms that maintain long-term exhausted T cell responses have substantially improved over the last decade (10, 17, 28). Here, we explored the developmental relationship between memory T cells that develop in response to acute infection and Tpex cells that arise in response to chronic infection. Using scRNAseq, we found that memory precursor (MP) cells that arose in response to acute infection, traditionally identified as CD127⁺TCF1⁺KLRG1^- cells (39–41), could be further segregated into CXCR6⁺ID3^- cells and ID3⁺Ly108⁺CXCR6^- MP cells, the latter resembling Tpex cells found in response to chronic infection. However, in line with previous data (24), ID3⁺ MP cells in acute infection expressed lower levels of PD-1 and TOX compared to Tpex cells, suggesting they are functionally less constrained. Importantly, we show that ID3⁺ MP cells can give rise to CXCR6⁺ memory T cells whereas CXCR6⁺ MP cells can only maintain themselves. Similar to memory precursors, we show that CD62L⁺ memory T cells, usually termed central memory T cells (42), contained ID3⁺ and ID3^- cells. While both CD62L⁺ memory populations showed similar capacity to re-expand in response to an acute infection, ID3⁺ T cells were exclusively able to generate ID3⁺ progeny. Furthermore, ID3⁺ memory T cells were superior in their response to chronic infection and were exclusively able to generate Tpex cells. Thus, independent of the type of infection, antigen-experienced T cells differentiate into effector and precursor cells, with ID3-expressing precursor cells having the highest developmental potential and the distinct ability to sustain CD8⁺ T cell immunity in response to protracted infections.

We and others have shown that ID3 faithfully identifies Tpex cells in chronic infection and cancer (18, 24, 38, 43); however, the function of ID3 in Tpex cell biology has remained poorly understood. To address this knowledge gap, we examined the response of ID3-deficient and control T cells to LCMV infection. ID3-deficient T cells showed substantially impaired population expansion in response to chronic LCMV. This was in contrast to acute infection, during which ID3-deficient T cells, consistent with previous data (32, 33), showed a modestly reduced population expansion. This suggested that ID3 was specifically required for T cells in the context of chronic but not acute stimulation. Transcriptional and flow cytometric profiling of ID3-deficient and control Tpex cells revealed increased expression of molecules associated with stemness, including Sell, Ccr7, and Myb in ID3-deficient T cells (4, 23, 44, 45), while genes linked to the activation of Tpex cells, such as Kit, were reduced. Indeed, ID3-deficient Tpex cells showed substantially reduced proliferation and expression of cell cycle genes, explaining the impaired population expansion of ID3KO T cells in chronic infection. The absence of ID3 may lead to elevated E2A activity in Tpex cells (30), a conclusion consistent with our data showing that overexpression of E2A reduced T cell numbers in chronic infection. Overall, these results indicate that ID3 sustains T cell responses to chronic infection by regulating proliferation and differentiation within the Tpex cell compartment.

Our data suggest that the decreased representation of cKIT⁺ Tpex cells in the absence of ID3 reflects the dampened proliferative capacity of these cells. This conclusion is consistent with our previous observations showing that cKIT expression is elevated in hyperactivated Tpex cells in chronic infection, for example following PD-1 or MYB deletion (23). Indeed, transcriptional profiling of cKIT⁺ Tpex cells compared to CD62L⁺ and CD62L^-cKIT^- Tpex cells revealed that cKIT⁺ Tpex cells had higher expression of transcripts linked to activation, cell cycling and response to cytokines. cKIT⁺ Tpex cells showed the lowest developmental and proliferative potential compared to CD62L⁺ or CD62L^-cKit^- Tpex cells based on their ability to sustain a T cell response in chronic infection. Notably, cKIT⁺ Tpex cells could self-renew while also giving rise to cKIT^- Tpex cells, indicating that cKIT expression is transient and might identify cycling CD62L^- Tpex cells.

Consistent with our earlier findings that formation of ID3⁺ T cells can be induced by strong TCR signaling (24), we found that ID3⁺ T cells in both acute and chronic infection express elevated levels of PD-1 and TOX, suggesting that they may develop in response to strong TCR signals. This idea is supported by recent work demonstrating that early memory precursors in acute infection expressed elevated levels of ID3, showed evidence of stronger TCR signaling and elevated expression of inhibitory receptors compared to their effector counterparts (46). Overall, these observations indicate that ID3, independent of the type of infection, is induced following strong TCR stimulation and marks cells with the highest developmental potential.

In addition to TCR signaling, cytokines play an important role in shaping acute and chronic T cell responses (11). However, the role of cytokines in promoting stem-like T cell differentiation is insufficiently known. Here, we show that IL-1 family members, including IL-36β, IL-36α, IL-33 and IL-18 promoted the formation of ID3⁺ precursor cells in vitro. This may be mediated by increasing TCR signaling pathways, that upregulate ID3 (37), as IL-36β-primed compared to IL-2-primed T cells showed enhanced expression of the TCR-induced transcriptional regulator Nur77 (NR4A1). Indeed, Nur77 has been shown to directly drive an exhaustion program including the formation of Tpex cells (47, 48). ID3⁺ P14 cells activated in the presence of IL-1 family cytokines were able to mount a superior T cell response compared to T cells primed in the presence of IL-2 when transferred into chronically infected mice. Similarly, IL-1 cytokine family-induced T cells mediated a prolonged anti-tumor response when transferred into tumor-bearing mice. This included increased expansion of T cells in tumor-draining lymph nodes, which is linked to superior tumor control (49, 50), and sustained proliferative capacity and stemness within the tumor microenvironment. While little is known about the role of IL-36β in T cells (51), enhanced IL-18 signaling has been shown to promote the formation of TCF1⁺ Tpex cells and mediate better tumor control in a murine melanoma model (52). Similarly, IL-33, another IL-1 family member (53, 54), has recently been shown to promote the generation of Tpex cells in chronic infection (55). Notably, all these cytokines signal through Myd88 (56), which has been shown to be required for efficient T cell proliferation in chronic infection (57). Our data extend these observations and show that IL-1 cytokine family members can act by promoting ID3 expression in activated T cells and thus promote superior developmental potential of chronically stimulated CD8⁺ T cells.

Overall, our data identify ID3⁺ T cells as stem-like T cells that retain the highest developmental potential in both acute and chronic infection and are preadapted to sustain T cell responses to protracted or chronic challenges. Moreover, our data show that the transcriptional regulator ID3 is required for sustained T cell responses in chronic infection and cancer by promoting ongoing proliferation and differentiation of Tpex cells. Therefore, ID3⁺ memory cells and Tpex cells are promising therapeutic targets for T cell-focused immunotherapies in chronic infection and cancer.

Materials and Methods

Study design

The aim of this study was to investigate the developmental relationship of Tpex cells in chronic infection with T cells derived from acute infection, which identified ID3 as a critical regulator. For this, we utilized Id3 reporter mice in which GFP replaces ID3 crossed to T cell receptor transgenic P14 mice specific for the LCMV-derived gp33 epitope. While heterozygote (Id3^GFP/+) cells faithfully report ID3 expression, homozygous Id3^GFP/GFP cells lack ID3 and thus served as knockout mice. We compared T cell responses in in vivo mouse model of chronic infection with Lymphocytic choriomeningitis virus (LCMV) and compared to responses in mice infected with acute LCMV. We performed adoptive transfer experiments of individual knock out or control T cell subsets to specifically delineate the impact of ID3 in different populations. We utilized flow cytometry to track and characterize T cells which were complemented by bulk and single cell RNA sequencing. Sampling replicates and number of animals are indicated in figure legends. Finally, we used preclinical tumor models to evaluate the anti-tumor response of in vitro activated T cells in the presence of different cytokines.

Mice

Wildtype C57BL/6JArc mice on a Ly5.1 or Ly5.2 background were obtained from the Australian Resources Centre or WEHI institute. Id3^GFP (30) or Nur77-reporter (Tempo) mice (58) were crossed to include the P14 TCR transgene (JAX: Tg(TcrLCMV)327Sdz). OT-I mice (JAX: Tg(TcraTcrb)1100Mjb/J) were bred on a Ly5.1 background. Mice were housed under specific pathogen-free conditions (SPF) and used for experiments at the age of 6-10 weeks. Experimental groups were age and sex matched male or female mice. All mice were maintained and used in accordance with the guidelines of the University of Melbourne and the Peter MacCallum Cancer Centre Animal Ethics Committees.

LCMV and Listeria infections

LCMV Armstrong, clone-13 and Docile strains were propagated and titrated on Vero African green monkey kidney cells according to an established protocol (59). Frozen stocks were diluted in PBS and 2×10⁵ plaque forming units (PFU) LCMV Armstrong were injected intraperitoneally and 2×10⁶ PFU LCMV clone-13 or Docile were injected intravenously. Recombinant L. monocytogenes strain expressing gp33 (60) was grown in Brain Heart infusion (BHI) to an optical density (OD) of ∽0.1. For Lm-Gp33 challenges, 3,000 colony forming units (CFU) were injected intravenously.

B16F10 tumor model

B16F10-Ova or B16F10-gp33 expressing cancer cells were cultured at 37°C and 10% CO₂ in DMEM supplemented with 10% FCS, 2 mmol/l GlutaMax and 50 U/ml penicillin/streptomycin. Mice were shaved on the right flank prior to injecting 2×10⁵ B16F10-Ova or 3×10⁵ B16F10-gp33 cells subcutaneously. Mice were randomized into treatment groups 7 days after tumor inoculation.

Tumors were measured three times weekly in two dimensions (a = length; b = width) with calipers and tumor volume (V) was determined by the following equation: V = ab2/2.

T cell purification and adoptive transfer

To obtain single-cell suspensions, spleens or lymph nodes were mashed through a 70 μm nylon cell strainer (BD) followed by red blood cell lysis with a hypotonic ammonium chloride-potassium bicarbonate (ACK) buffer. P14 isolation was performed using the mouse CD8⁺ T cell enrichment kit (Miltenyi Biotech), and ctrl or ID3KO P14 cells were intravenously injected into naive congenically marked recipients.

For adoptive transfer experiments or RNAseq, P14 cells were pre-enriched by positive selection using magnetic columns (Miltenyi Biotech). In brief, total splenocytes were first stained with biotin-labeled anti-Ly5.1 or -Ly5.2 antibodies in DMEM supplemented with 10% fetal bovine serum, followed by secondary staining with anti-biotin beads. Pre-enriched samples obtained from magnetic columns were then stained (table S1) prior to cell subset isolation. Individual cell subsets were isolated by via FACS for bulk RNA sequencing or adoptive transfer experiments. At least 100,000 P14 cells were sorted for RNA sequencing and 5,000-100,000 P14 cells were adoptively transferred per experiment.

In vitro T cell activation and adoptive transfer

For in vitro cytokine screening, whole spleens were dissociated through a 70 μm nylon cell strainer and red cells lysed with ACK buffer. Splenocytes were subsequently washed and resuspended in RPMI media supplemented with 10% FCS, 2 mmol/l GlutaMax, 50 U/ml penicillin/streptomycin, 1x nonessential amino acids, 1 mmol/l sodium pyruvate, 10 mmol/l HEPES, and 50 μmol/l 2b-mercaptoethanol. Naïve Id3^GFP P14 splenocytes were isolated using a negative selection kit (Stem Cell Technologies) and 5×10⁴ P14 cells were combined with 3.5×10⁵ C57/BL6 splenocytes and 50 ng/ml gp33 peptide before transfer in 200 μl to a 96 well plate. Cell mixes were supplemented with recombinant human IL-2 (100 IU/ml; Biolegend) or various recombinant mouse cytokines (50 ng/ml; Biolegend) and cultured for three days at 37°C with 5% CO₂ before being passaged into fresh media (cytokine only, no peptide) and cultured for an additional two days prior to analysis.

For adoptive transfer, purified OT-I or P14 single-cell suspensions were diluted (5-10×10⁴ naive T cells) in supplemented RPMI media and T cells expanded with SIINFEKL peptide (20 ng/ml) or gp33 peptide (100 ng/ml) and selected cytokines for three days, before passaging into fresh cytokine-containing media and cultured for an additional two days prior to analysis and adoptive transfer. For adoptive transfer experiments, activated OT-I or P14 cells were harvested from culture, washed in PBS and injected intravenously into either tumor-bearing (5×10⁶ OT-I cells) or LCMV infected recipients (5×10⁵ P14 cells).

Isolation of tumor-infiltrating lymphocytes

6 and 13 days post OT-I transfer, tumor-bearing mice were euthanized via CO₂ asphyxiation and cardiac perfused with 10 ml of cold PBS prior to excision. Excised tumors were weighed, finely chopped using scissors in 1 ml RPMI, and transferred into 20 ml of prewarmed digestion solution consisting of 100 U/ml Type-I collagenase (Worthington), 10% FCS, 10 mmol/l HEPES, 1 μmol/l MgCl₂, 1 μmol/l CaCl₂ in RPMI base media. Tumors were dissociated at 37°C for 40 min and then further mechanically disassociated and filtered through a 70 μm nylon cell strainer. Extracted tissue was pelleted and resuspended in 44% isotonic Percoll solution (in Hank’s buffered salt solution) and underlaid with 77% isotonic Percoll solution before density gradient centrifugation (2000 rpm, 20 min at room temp). Cells at the interphase were collected and filtered through a 40 μm nylon cell strainer prior to subsequent analysis.

Surface and Intracellular Antibody Staining of Mouse Cells

Surface staining was performed for 30 min at 4°C in PBS supplemented with 2% fetal calf serum (cell staining buffer) after 10 min incubation with purified anti-mouse CD16/32 Ab (FcgRII/III block; 2.4G2, 1:1000) and (fixable) viability dye (1:1000, ThermoFisher). For intracellular cytokine staining, splenocytes were ex vivo re-stimulated with gp33 peptide (5 mM) for 5 h in the presence of Brefeldin A (Sigma) for the last 4.5 h, fixed and permeabilized using the Cytofix/Cytoperm (BD) or transcription factor staining kit (ThermoFisher). Intracellular transcription factor staining was performed with a transcription factor staining kit (ThermoFisher). For GFP stain, cells were fixed for 10 min with 4% formaldehyde prior using the transcription factor staining kit. Used antibodies are listed in table S1.

ID3 and E2A overexpressing vector and retroviral transduction

To produce retroviral supernatant, HEK293T human embryonic kidney cells were transfected with a retroviral expression pMSCV plasmid containing either murine Thy1.1 and E2A–IRES–Thy1.1 or Id3−IRES−mCherry and mCherry expression cassettes, as previously described (61, 62).

For ID3 overexpression, P14 cells were activated for approximately 40 h with LCMV-derived gp33 peptide and 50 U/ml IL-2 prior to spin infection with viral supernatant supplemented with 50 μM 2-mercaptoethanol and 5 μg/ml of polybrene for 90 min at 37 °C. Transduced P14 cells were subsequently incubated for 1 h at 37 °C in RPMI medium before adding interleukin-7 (10 ng/ml). Transduced cells (mCherry⁺) were sorted 24 hrs later and transferred into naïve mice before infection with LCMV.

For E2A overexpression, P14 cells were isolated using the naïve CD8⁺ T Cell Isolation Kit (Stem Cell Technologies) and activated for approximately 20 hours with 2 μg/ml aCD28 (Invitrogen) and 100 U/ml IL-2 in plates pre-coated with 5 μg/ml aCD3 (Invitrogen). The next day, activated P14 cells were transduced via spin infection with viral supernatant supplemented with 8 μg/ml polybrene at 2,000 g for 1 h at 32°C and 5×10⁴ P14 cells injected into mice that were infected one day earlier with 4×10⁶ PFU of LCMV clone-13.

Single cell RNA sequencing

Sorted P14 splenocytes from LCMV clone-13 and Armstrong infected mice were incubated with TruStain FcX PLUS (Cat# 101319) for 10 min on ice before TotalSeq hashtag antibody cocktail (BioLegend, Cat# 155831 and 155835) was added and incubated for additional 30 mins on ice. Single cell RNAseq libraries were made using the Chromium Single cell 3’ library and gel bead kit (10x Genomics) according to the manufacturer’s instructions. Briefly, hashtag-labelled P14 cells were pooled and loaded into a 10x Chromium controller. Generated GEMs were incubated to produce full-length cDNA and barcoded DNA from cell surface protein feature barcode. cDNA molecules and barcodes were amplified by PCR (Run #1, table S2) to generate the downstream library construction. Amplified cDNA was then subjected to enzymatic fragmentation, end-repaired and A-tailing, and SPRIselect purification (Beckman Coulter). Sample indexes (Dual Index Plate TT Set A, Cat# PN-3000431) and protein feature barcode libraries (Dual Index Plate NT Set A, Cat# PN-3000483) were sequentially added by PCR amplification (Run #2, table S2) followed by SPRIselect purification. Libraries were assessed using Agilent Tape station and quantified using a QubitTM 3.0 Fluometer (ThermoFisher). Pooled 3’ GEX and cell surface libraries at 10 nM concentration were sequenced using NextSeq2000 by the Illumina platform at Genomics Research Core Facility at WEHI.

Bioinformatic analyses scRNAseq

Bioinformatic analyses of single cell and bulk RNA sequencing is described within supplementary data.

Data analyses, quantification, and statistical analyses

Flow cytometry was performed using a BD Fortessa and LSR II or Cytek Aurora. Sort-purification was performed on a BD FACSAria Fusion or CytoFLEX SRT (Beckman Coulter). Data were analyzed using FlowJo 10 (BD); graphs prepared with Prism v10 (GraphPad Software). Depending on experimental setup, paired or unpaired Student’s t test (two-tailed), or one-way ANOVA was used to assess significance. Highly skewed data (p < 0.05 F-test) was log transformed before statistical analysis. The exact significance values are stated in all graphs and the number of biological or technical replicates (n) is stated in figure legends.

Materials And Methods

Bioinformatic analyses scRNAseq

Sequencing reads were aligned to the mm10 reference genome and counted with cellranger-6.1.2. Hashtag antibodies were demultiplexed with the cellranger multifunction specifying antibodies as “Multiplexing capture”. Two batches of single cell data were merged and processed using Seurat (version 4.3.0)(63–65). Specifically, cells were filtered for >800 genes and less than 7% mitochondrial RNA before SCTransform was used to normalize the counts data. We mitigated the batch effect (different sex mice between batch 1 and batch 2) by using RPCA integration from the Seurat package. UMAP reductions and cell neighbors were calculated using 25 dimensions from the PCA. Clusters were detected with a resolution of 0.8. A small cluster of low ribosomal contents was removed from further analysis. Differential gene signatures between MP1 and MP2 cells were calculated using the FindMarkers function and applied to the data using AddModuleScore. Plots were generated using ggplot2 and ShinyCell (66).

For human scRNA data, published clusters (31) were reproduced using the postprocessing script obtained from their github page (https://github.com/pogorely/COVID_vax_CD8). Data were processed using R version 4.2.0 and Seurat version 4.3.0 and gene signature expression analyzed using AddModuleScore.

Bulk RNA sequencing and bioinformatic analyses

RNA-seq datasets from sorted control and ID3KO Tpex (ID3GFP⁺) and Tex (TIM-3⁺) P14 cells were aligned to the mouse reference genome GRCm38/mm10 using the Subread aligner (Rsubread version 2.10.2)(67). Gene-level counts of mapped reads were obtained using featureCounts (68, 69) and the inbuilt Rsubread annotation that is a modified version of NCBI RefSeq mouse (mm10) genome annotation build 38.1 (70). Genes that had a CPM (counts per million reads) <0.5 in all libraries were excluded from further analysis. Read counts were converted to log2-CPM, quantile normalized, and precision weighted with the voom function of the limma package (71, 72). A linear model was fitted to each gene, and empirical Bayes moderated t-statistics were used to assess differences in expression (73). Genes were called differentially expressed if they achieved a false discovery rate (FDR) of less than 0.15. Enrichment of Hallmark gene sets from the Molecular Signature Database (MSigDB version 7.5.1)(74) was tested using roast method (75) within the limma package. Testing of enrichment of gene ontology (GO) terms and KEGG pathways was performed using the goana and kegga functions respectively from the limma package.

RNA sequencing reads of Tpex subsets were aligned to the GRCm39 reference genome and transcriptome (version 105) using the STAR aligner (version 2.7.8a)(76) and transcript counts established using featureCounts from the subread package (version 2.0.0)(67). The data was then analyzed using R (version 4.2.0), limma (version 4.54.1)(71, 72), RUV (version 0.9.7.1)(77) and plots generated with ggplot2. Differentially expressed genes were obtained by filtering out low or non-expressed genes (filterByExpr function) as well as T cell receptor genes (^Tr), immunoglobin genes (^Ig), class 2 antigen genes (^H2), Jchain, and normalising counts by TMM. From the initial MDS plots, we identified a batch effect between the replicates, which we removed using RUVIII (for visualization) and RUV4 (for covariate estimation – k=3, all genes as controls). We estimated group-wise dispersion and mean-variance trends with the voom function and fit a linear model using lmFit. We moderated the model parameters using eBayes and obtained differential genes for each of the pairwise contrasts between the experimental groups (FDR 0.05, Benjamini-Hochberg correction). Gene set enrichment was performed using ROAST on the MSigDB Hallmark 50 mouse gene sets.

One Sentence Summary.

ID3 defines stem-like T cells with high developmental potential that sustain T cell responses to chronic infection and cancer.

Data and materials availability

Sequencing data generated for this study have been deposited in the Gene Expression Omnibus database with accession code GSE224367, GSE226030 and GSE277451. Custom code has been deposited at https://doi.org/10.5281/zenodo.14538162. Tabulated data underlying the figures are provided in Data file S6. All other data needed to support the conclusions of the paper are available in the paper or the Supplementary Materials.