C57BL/6 wild-type (WT) and CD39 knockout (CD39KO) mice (6 to 10 weeks) were housed at the animal facility of the CIBICI-CONICET/Facultad de Ciencias Quimicas-UNC. All animal studies were conducted in compliance with institutional guidelines and approved by the Institutional Animal Care and Use Committee of the Facultad de Ciencias Químicas, Universidad Nacional de Córdoba, Argentina (RD-2019-1689-E-UNC-DEC#FCQ, RD-2023-1938-E-UNC-DEC#FCQ). The Entpd1 null mice (CD39KO) mice were kindly provided by Dr. Simon C. Robson (Beth Israel Deaconess Medical Center, Boston MA) (26).

MC38 was kindly provided by Dr. Clothilde Théry. B16F10-OVA cell line was purchased from ATCC. Both cell lines were maintained in DMEM (GIBCO) supplemented with 10% fetal bovine serum (FBS; Natocor), 1 mM L-glutamine, 25 mM HEPES, and 40 µg/mL gentamicin. In both MC38 and B16F10-OVA tumor-bearing mice, neither CD45+ nor CD45- cells express CD39 within TME after engraftment in CD39KO mice (Supplementary Figure 1A), confirming that MC38 and B16F10-OVA tumor cell lines do not express CD39.

Male C57BL/6 or CD39KO mice were anesthetized with ketamine/xylazine (100/10 mg/kg body weight) administered intraperitoneally (i.p.), and subsequently inoculated subcutaneously (s.c.) with either 5 × 10⁵ MC38 cells or 1 × 10⁶ B16F10-OVA cells. Tumors and draining lymph nodes (dLNs) were collected on day 17 post-injection (p.i.). Tumor volume was measured on 7–17 days p.i. using a manual caliper to determine the longest (D) and shortest (d) tumor diameters. Volume was calculated using the formula: tumor volume = (d × D²) × 0.5. Tumors were disaggregated mechanically and enzymatically with 2 mg/mL collagenase IV and 50 U/mL DNase I (Roche). In experiments assessing effector function, tumor-infiltrating mononuclear cells were enriched by density gradient centrifugation using Ficoll-Hypaque or Percoll (GE Healthcare). All procedures were performed following protocols previously established in our laboratory (12, 27).

Single cell suspensions were stained with fluorochrome-conjugated monoclonal antibodies against mouse (fluorochromes and clones, Supplementary Table S1). Gating strategies for cellular populations from tumors and dLNs are shown in Supplementary Figure 1B.

For ex vivo intracellular cytokines, TFs, and CD107a, mononuclear cells were stained using protocols previously established in our laboratory (12, 27).

OVA-specific CD8⁺ T cells were detected with a PE-conjugated OVA^257–264 H-2Kb dextramer (Immudex) according to the manufacturer´s instructions. Samples were acquired using a BD LSRFortessa flow cytometer and analyzed using FlowJo software.

C57BL/6 WT and CD39KO mice were inoculated with B16F10-OVA tumor cells as previously described. Treatment administration began on day 5 p.i., when tumors became palpable. The CD39 inhibitor POM-1 (TOCRIS) was injected i.p. in 200 µL PBS containing 250 µg of POM-1 on days 5, 7, 10, and 13 p.i.

For IL-2cx treatment, mice received daily i.p. injections of 200 µL IL-2cx from day 7 to day 12 p.i. The complex was prepared by mixing 15,000 IU of recombinant human IL-2 (Clinigen Healthcare Ltd) with 4.5 µg of anti-human IL-2 monoclonal antibody (Mab602, Bio-Techne), followed by incubation at 37°C for 30 minutes before injection. For the POM-1/IL-2cx combined treatment, mice received co-injections of both compounds according to the individual schedules used for IL-2cx and POM-1 monotherapies.

GraphPad Prism 9.0.0 software was used for statistical analysis. Prior to statistical testing, data were evaluated for outliers using the ROUT method, which were excluded if identified, and assessed for normality to determine the appropriate statistical test. Samples from two different experimental groups were considered unpaired, whereas samples involving cell populations obtained from the same individual were considered paired. The statistical test used for each comparison is detailed in each figure and was chosen according to the software’s recommendations. P-values ≤0.05 were considered statistically significant.

We first evaluated the impact of CD39 deficiency on tumor progression. WT or CD39KO mice were injected with the highly immunogenic MC38 tumor cells. From day 12 post-injection (p.i.) until the experimental endpoint at day 17 p.i., CD39KO mice exhibited significantly reduced tumor volumes compared to their WT counterparts (Figure 1A). Consistently, tumor weight at endpoint was also lower in CD39KO mice (Figure 1B). Flow cytometric analysis revealed a higher frequency of tumor-infiltrating (T-I) CD45⁺ leukocytes in CD39KO mice than in WT mice (Figure 1C). However, no significant differences were observed in the composition of CD45⁺ immune cell subpopulations between the two groups.

Given the critical role of CD8⁺ T cells in tumor control, we next focused on this T-I T cell population. We assessed their activation and differentiation status and found that, in both WT and CD39KO mice, the majority of CD8⁺ T cells exhibited an effector memory (EM) phenotype (CD62L^-CD44⁺), while central memory (CM) cells (CD62L⁺CD44⁺) accounted for around 20% and naïve cells (CD62L⁺CD44^-) were nearly absent (Figure 1D). Importantly, no significant differences were observed in the relative frequencies of these subsets between groups.

We then analyzed iRs and transcription factors (TFs) associated with CD8⁺ T cell activation and exhaustion. CD39KO mice showed a higher frequency of CD8⁺ T cells expressing TIM-3 and 2B4, but not of LAG-3 and PD-1, compared to WT mice (Figure 1E). Additionally, CD39KO mice showed higher frequencies of CD8⁺ T cells co-expressing 4 iRs than WT mice (Figure 1F; Supplementary Figure 1C). Moreover, T-I CD8⁺ T cells from CD39KO mice displayed elevated expression level (MFI) of the TFs Eomes, T-bet, IRF4, and TOX, while TCF-1 levels were comparable to those in WT mice (Figure 1G). Accordingly, CD39KO mice exhibited a higher percentage of TOX-expressing CD8⁺ T cells and similar frequencies of TCF-1⁺ cells compared to WT mice (Supplementary Figure 1D).

Together, these results show that CD39 deficiency limits MC38 tumor growth and is associated with increased tumor infiltration by CD8⁺ T cells displaying an exhausted phenotype characterized by elevated expression of multiple iRs and TFs.

Since the previous analysis showed that T-I CD8⁺ T cells from CD39KO mice express multiple iRs and exhaustion-associated TFs, we next investigated the heterogeneity within this population, focusing on subsets defined by PD-1 expression levels. A hallmark of T cell exhaustion is its heterogeneity (8), and previous studies have demonstrated that different levels of PD-1 expression can distinguish between subsets of exhausted T cells (28, 29). Specifically, high PD-1 expression characterizes terminally exhausted cells, whereas intermediate expression defines pre-exhausted subsets.

To explore this heterogeneity among T-I exhausted CD8⁺ T cells from WT and CD39KO mice, we classified them into PD-1^High, PD-1^Int, and PD-1^Neg populations. As shown in Figure 2A, all three subsets were identified in MC38 tumor-bearing mice of both WT and CD39KO. Comparative analysis revealed that CD39KO mice displayed a significantly higher proportion of PD-1^High CD8⁺ T cells and a lower frequency of PD-1^Int cells compared to WT mice.

Further analysis of iRs and inhibitory molecules expression revealed that T-I PD-1^High CD8⁺ T cells, in WT as well as in CD39KO mice, presented the highest frequencies of cells co-expressing 4 inhibitory molecules (PDL-1, 2B4, TIM-3, and LAG-3), compared to PD-1^Int or PD-1^Neg populations (Figure 2B). In particular, TIM-3, a marker associated with terminal exhaustion (30), was predominantly expressed in the PD-1^High population (Supplementary Figure 2A).

We also investigated exhaustion-related TFs, including TOX and TCF-1 (31, 32). PD-1^High CD8⁺ T cells exhibited the highest frequency and mean fluorescence intensity (MFI) of TOX expression, compared to the other subsets (Figure 2C; Supplementary Figure 2B). Conversely, TCF-1 expression was largely restricted to PD-1^Int and PD-1^Neg cells and was almost absent in the PD-1^High CD8⁺ T cells (Figure 2D; Supplementary Figure 2C).

Importantly, we complemented these phenotypic analyses with functional assays. Upon in vitro PMA/Ionomycin stimulation, PD-1^High cells from both WT and CD39KO mice showed the highest frequency of IFN-γ–producing cells (Figure 2E). Additionally, analysis of molecules related to cytotoxic function within T-I CD8⁺ T cell population, revealed that PD-1^High population from WT and CD39KO mice presented higher frequencies of Perforin and Granzyme B (GzmB) producing cells than PD-1^Int or PD-1^Neg CD8⁺ T cells (Figure 2E). Importantly, within this PD-1^High compartment, CD39 KO mice display higher frequencies of GzmB⁺ and Perforin⁺ CD8⁺ T cells compared to WT mice (Supplementary Figure 2D).

Moreover, we observed a positive correlation (WT, r=0.9652 and P ≤0.01; CD39KO, r=0.88 and P ≤0.05) between PD-1 expression levels and the proportion of polyfunctional CD8⁺ T cells co-expressing CD107a, Perforin, and GzmB (Figure 2F).

Collectively, these results indicate that the absence of CD39 favors the accumulation of cytotoxic PD-1^High CD8⁺ T cells with features of terminal exhaustion. Furthermore, the enhanced cytotoxic profile of PD-1^High CD8⁺ T cells in CD39KO mice is consistent with their improved MC38 tumor control.

To extend our observations from the MC38 model, we next analyzed the impact of CD39 deficiency in the less immunogenic B16F10-OVA melanoma model, which is widely employed to study antigen-specific CD8⁺ T cell responses and provides a model to determine impact of checkpoint blockade on immunogenicity. This approach allowed us to determine whether the absence of CD39 similarly affected tumor growth and to further dissect the dynamics of OVA-specific CD8⁺ T cells within the tumor and draining lymph nodes (dLNs).

In these studies, using B16F10-OVA, we observed that CD39KO recipient mice exhibited similar tumor growth kinetics and tumor weight at the experimental endpoint (17 days p.i.), when compared to WT mice (Figures 3A, B).

Next, we analyzed various immune cell populations in tumors and dLNs. CD39KO mice exhibited higher frequencies of CD8⁺ T cells in both tissues compared to WT mice (Figure 3C), accompanied by a relative reduction in B cells in dLNs, with no notable differences in the total frequency of CD45⁺ cells (Figure 3C; Supplementary Figure 3A).

We then assessed the activation and differentiation status of CD8⁺ T cells in tumors and dLNs using CD44 and CD62L expression. The absence of CD39 promoted the accumulation of EM T-I CD8⁺ T cells at the expense of the CM population (Figure 3D). As in MC38 tumors, naïve CD8⁺ T cells were virtually absent. In dLNs, CD39KO mice displayed higher frequencies of CM and lower frequencies of naïve CD8⁺ T cells, with no significant changes in EM cells compared to WT mice.

Analysis of antigen-specific CD8⁺ T cells revealed that CD39KO mice harbored a higher frequency of OVA-specific T-I CD8⁺ T cells compared to WT mice (Figure 3E).

We further analyzed the distribution of exhausted CD8⁺ T cell subsets. As observed in the MC38 model, CD8⁺ TILs from both CD39KO and WT B16F10-OVA tumor–bearing mice were classified into PD-1^High, PD-1^Int, or PD-1^Neg populations based on PD-1 expression levels (Figure 3F). Notably, CD39KO mice exhibited a higher frequency of PD-1^Int CD8⁺ T cells than WT mice. We also demonstrated that PD-1^High CD8⁺ T cells in both groups correspond to terminally exhausted cells, as they co-express multiple iRs, including high levels of TIM-3 (Supplementary Figures 3B, C). Accordingly, PD-1^High CD8⁺ T cells displayed the highest expression of TOX and the lowest levels of TCF-1, compared to PD-1^Int and PD-1^Neg subsets (Supplementary Figures 3D, E). In line with previous findings (9), terminally exhausted CD8⁺ T cells (PD-1^High) exhibited the highest frequencies of IFN-γ–producing cells (Supplementary Figure 3F) and expression of cytotoxic molecules, including CD107a, GzmB, and Perforin (Supplementary Figure 3G).

In contrast, PD-1^Int CD8⁺ T cells maintained TCF-1 expression (Supplementary Figure 3E) and the ability to produce IFN-γ (Supplementary Figure 3F), consistent with a pre-exhausted phenotype.

Interestingly, CD39KO mice showed similar proportions of OVA-specific CD8⁺ T cells within both the pre- and terminally exhausted populations, whereas WT mice exhibited a higher proportion of OVA-specific CD8⁺ T cells among terminally exhausted cells compared to the pre-exhausted and PD-1^Neg subsets (Figure 3G). Moreover, pre-exhausted OVA-specific CD8⁺ T cells were significantly more abundant in CD39KO mice than in WT mice.

Taken together, these findings demonstrate that, the absence of CD39 promotes the accumulation of effector and tumor antigen-specific CD8⁺ T cells within the TME, although not impacting tumor growth in this specific B16F10-OVA model.

Since our results in the B16F10-OVA model showed that CD39 deficiency favors the accumulation of pre-exhausted CD8⁺ T cells, which are key targets of immune ICB therapy (9), we reasoned that targeting CD39 might synergize with immunotherapies that boost CD8⁺ T cell responses. In this context, IL-2cx redirects IL-2 activity to CD8⁺ T and NK cells and has been shown to enhance antitumor CD8⁺ T cell responses, leading to improved tumor control in B16F10 tumor-bearing mice (25). We therefore evaluated the effect of a combined therapeutic approach, blocking CD39 activity with the pharmacological inhibitor POM-1 together with IL-2cx administration.

Following tumor engraftment, B16F10-OVA tumor-bearing WT mice were injected with PBS as control, or treated with POM-1, IL-2cx, the combination of POM-1 and IL-2cx, as depicted in Figure 4A. We found that CD39 inhibition with POM-1 alone had no effect on tumor progression compared to the control group (Figure 4B). In line with previous reports (25), mice treated with IL-2cx exhibited delayed tumor growth relative to controls. At the experimental endpoint (15 days p.i.), mice receiving the POM-1/IL-2cx combined therapy displayed significantly reduced tumor volume compared to control, POM-1, and IL-2cx groups. Moreover, combination-treated mice showed slower tumor growth, with differences becoming evident by day 12 p.i. compared to all other groups (Supplementary Figure 4A).

Analysis of tumor-infiltrating leukocytes showed that mice treated with POM-1 alone exhibited lower frequencies of CD45⁺ cells compared to IL-2cx or the combination, while no other significant differences were detected (Figure 4C). Mice treated with the POM-1/IL-2cx combined therapy exhibited higher frequencies of tumor-infiltrating CD8⁺ T cells than the control group (Figure 4D). No differences were observed in the frequency of Ki-67⁺ CD8⁺ T cells among the treated groups (Figure 4E). When focusing on OVA-specific CD8⁺ T cells, all treated groups (IL-2cx, POM-1, or the combination) showed reduced frequencies compared to PBS controls (Figure 4F; Supplementary Figure 4B). We also examined PD-1 and TOX expression on OVA-specific CD8⁺ T cells across all treatment groups. More than 80 percent of OVA-specific CD8⁺ T cells displayed high PD-1 expression (PD-1^High) in all conditions, including PBS controls, and TOX expression showed a similar pattern (Supplementary Figure 4C).

We next assessed cytotoxic molecules and found that mice treated with the POM-1/IL-2cx combined therapy exhibited higher frequencies of T-I CD8⁺ T cells expressing GzmB and CD107a compared to all other groups (Figure 4G). In addition, the frequency of Perforin⁺ CD8⁺ T cells was increased in the POM-1/IL-2cx combined therapy group relative to the control group. However, no significant differences were observed in the frequencies of IFN-γ–producing CD8⁺ T cells among the groups.

In addition, we analyzed CD8⁺ T cells in dLNs and found that the treatments did not affect the frequency of CD45⁺ cells, CD8⁺ T cells, the expression of cytotoxic molecules, or the proportion of IFN-γ–producing CD8⁺ T cells (Supplementary Figures 4D, E). Moreover, none of the treatment strategies resulted in detectable OVA-specific CD8⁺ T cells in the dLNs (data not shown).

Together, these results demonstrate that CD39 inhibition with pharmacological antagonist potentiates the efficacy of IL-2cx therapy, leading to enhanced CD8⁺ T cell infiltration and cytotoxic activity within the TME.

Because the balance between progenitor and terminally exhausted T cells can influence therapeutic outcomes (9), we next evaluated whether the treatments affected the frequency or functional status of pre-exhausted and terminally exhausted CD8⁺ T cells in B16F10-OVA tumor–bearing mice. To this end, we analyzed the distribution of T-I CD8⁺ T cell subsets defined by PD-1 expression. No significant differences were observed in the frequencies of PD-1^Neg, PD-1^Int or PD-1^High CD8⁺ T cells among the treated and control groups (Figure 5A). However, we found that POM-1/IL-2cx combined therapy increased the frequency of TOX⁺ cells within the pre-exhausted PD-1^Int subset compared to all other groups, while simultaneously reducing the frequency of TCF-1⁺ cells in this population relative to the PBS group (Figure 5B). Additionally, PD-1^Int CD8⁺ T cells from POM-1/IL-2cx treated group exhibited a reduced proliferation capacity, as evidenced by lower frequencies of Ki-67⁺ cells compared to the POM-1 or PBS-treated group (Figure 5C).

Notably, the combined therapy resulted in higher frequencies of GzmB⁺ PD-1^Int CD8⁺ T cells compared to the other treatments evaluated (Figure 5D). Furthermore, treatment with POM-1, IL-2cx, or their combination increased the % Perforin⁺ PD-1^Int CD8⁺ T cells compared to the control group, although no significant differences were observed in the frequencies of CD107a and IFN-γ-expressing cells. Analysis of the PD-1^High CD8⁺ T cells showed that the tested treatments did not increase the expression of cytotoxic molecules (GzmB, Perforin, CD107a) or the proportion of IFN-γ-expressing cells (Supplementary Figure 5A).

Overall, these data suggest that POM-1/IL-2cx combined treatment modulates the functional profile of pre-exhausted CD8⁺ T cells by increasing GzmB expression, which may contribute to their cytotoxic potential.

We next evaluated the effect of the combined therapy on other immune cell populations involved in either promoting or inhibiting tumor growth, including NK cells, conventional CD4⁺ T cells (Tconv), Tregs, and myeloid cells (gating strategy in Supplementary Figures 6A-C). As shown in Figure 6A, the frequencies of NK cells did not differ among the treated groups. However, mice treated with IL-2cx or the POM-1/IL-2cx combination exhibited increased frequencies of activated NK cells (KLRG1⁺) compared to the POM-1 or control groups. Despite these phenotypic changes, we did not detect an increase in NK cell effector function, as measured by the production of GzmA, Perforin, and CD107a in mice treated with the POM-1/IL-2cx combination (Figure 6B).

We next examined the CD4⁺ T cell compartment and found that none of the treatments augmented the recruitment of either Tconv (CD4⁺FOXP3^-) or Treg (CD4⁺FOXP3⁺) cells to the tumor site (Figure 6C). Our previous work showed that T-I CD39⁺ CD4⁺ Tconv cells exhibit cytotoxic potential (27). Therefore, we evaluated the presence of CD39-expressing Tconv cells within the TME. Although the frequency of CD39⁺ T-I CD4⁺ Tconv cells remained unchanged across treatment groups, mice receiving the POM-1/IL-2cx combined therapy showed higher expression of GzmB in this CD4⁺ T cell compartment compared to controls or monotherapy with POM-1 or IL-2cx (Figures 6D, E).

Finally, given the pivotal role of the myeloid compartment in tumor progression, we analyzed T-I monocytic myeloid-derived suppressor cells (M-MDSCs), total and M2-like macrophages. POM-1 treatment did not modify the frequency of T-I M-MDSCs (CD11b⁺ F4/80^-Ly6G^-Ly6C⁺), whereas IL-2cx significantly reduced this population compared to the control group. A similar significant decrease was observed in mice receiving the combined POM-1/IL-2cx treatment (Figure 6F). Interestingly, analysis of molecules associated with the immunosuppressive adenosine pathway (33) revealed that mice receiving the POM-1/IL-2cx combined treatment had a lower proportion of CD39⁺ M-MDSCs compared with the control or IL-2cx-treated groups, as well as reduced frequencies of CD38⁺ and CD73⁺ M-MDSCs compared with all treated groups or with the PBS group, respectively (Figure 6G). No differences were observed in the frequency of macrophages (CD11b⁺F4/80⁺Ly6G^-Ly6C^-) or M2-like macrophages (CD11b⁺Ly6G^-Ly6C^-F4/80⁺CD206⁺) (Supplementary Figure 6D).

Altogether, these results indicate that POM-1/IL-2cx combined therapy not only enhances the expression of cytotoxic molecules on pre-exhausted CD8⁺ T cells but also reshapes the tumor immune landscape by modulating the CD4⁺ T cell compartment and reducing the immunosuppressive profile of M-MDSCs.