MAP4K4 negatively regulates CD8 T cell–mediated antitumor and antiviral immunity

Map4k4 loss suppresses tumor growth in a CD8 T cell–dependent manner

Previously, we revealed that Map4k4 iKO mice had impaired growth of both KPP1 pancreatic and TC1 lung tumors with a modest reduction on vascular function (26). We expand this finding to additional models and showed that Map4k4 iKO hosts also reduced tumor growth of the MC38 colorectal and KPR3070 pancreatic tumor models (Fig. 1A and figs. S1A and S2E). Given that MAP4K4 is highly expressed in immune cells (fig. S3) (28), we evaluated the immune compartment’s contribution to antitumor efficacy. Increased numbers of intratumoral CD3 and CD3/CD8 T cells were seen in KPP1 (Fig. 1, B and C), TC1 (fig. S1B), and MC38 (Fig. 1D) tumors grown in Map4k4 iKO hosts. CD3 T cells appear to be distributed throughout the tumor and show no bias toward tumor vasculature detected by MECA-32 staining in KPP1 tumors (Fig. 1B). To understand the contribution of mature lymphocytes to reduced tumor growth in Map4k4 iKO hosts, we bred the Map4k4-floxed allele into the Rag2^−/− background. MC38 tumors grew at similar rates in Rag2^−/− Map4k4 iKO mice compared with Rag2^−/− Map4k4 wild-type (WT) hosts, revealing that T and B cells are required for the Map4k4 deletion phenotype (Fig. 1E and fig. S1C). Depleting CD8 T cells, but not CD4 T cells, reversed inhibition of tumor growth for Map4k4 iKO hosts (Fig. 1F and fig. S1D), indicating a requisite role of CD8 T cells in mediating the Map4k4 iKO phenotype. To understand whether Map4k4^−/− T cells are sufficient to suppress tumor growth, we adoptively transferred CD3 T cells isolated from Map4k4 WT and iKO spleens into Rag2^−/− hosts. Although Map4k4^−/− T cell counts in the blood of reconstituted mice were slightly less than WT T cells (fig. S1E), they were more effective than WT T cells in reducing MC38 tumor growth and increasing tumor necrosis (Fig. 1, G to I, and fig. S1F). Because more Map4k4^−/− CD8 T cells, but not CD4 T cells, accumulated to higher numbers in tumors than their WT counterparts (fig. S1G), we also reconstituted Rag2^−/− mice with only CD8 T cells from Map4k4 WT and iKO spleens and confirmed comparable T cell reconstitution in peripheral organs between both genotypes (fig. S1H). Similarly, Map4k4^−/− CD8 T cells were better at reducing tumor growth than WT CD8 T cells (Fig. 1J and fig. S1I). Next, we analyzed tumor-infiltrating lymphocytes (TILs) to assess whether CD8 T cell functionality is affected. Among freshly isolated MC38 CD8 TILs from immune competent Map4k4^−/− hosts without any in vitro stimulation, more cells expressed the cytolytic protein granzyme B (GZMB) (Fig. 1K) and fewer expressed negative regulators, programmed cell death protein–1 (PD-1), cytotoxic T lymphocyte–associated protein–4 (CTLA-4), and CD39 (fig. S2, A to D). These cells also produced more cytokines after ex vivo anti-CD3/28 stimulation (Fig. 1L). Together, these data suggest that Map4k4^−/− CD8 T cells are more functional and are both necessary and sufficient to inhibit tumor growth.

• Download high-res image
• Open in new tab
• Download Powerpoint

Given the above findings, we asked whether inhibition of PD-1/PD-L1 signaling, a pathway that causes T cell dysfunction induced by chronic antigen stimulation (29), can further improve the antitumor activity afforded by Map4k4 loss. Loss of Map4k4 and anti–PD-L1 treatment alone showed comparable efficacy in the pancreatic KPR3070 tumor model, and the combination of these two treatments demonstrated an additive effect (fig. S2E), indicating that MAP4K4 could be a potential target to combine with PD-L1/PD-1 inhibitors. The combination only resulted in tumor stasis and not tumor regression, but this is not surprising because tumor cell proliferation rates are generally higher in implanted mouse tumor models than in human cancers; therefore, achieving tumor regression in such models is challenging. Retrospective analysis of approved drugs and their preclinical performances showed that reducing tumor growth rates by ~60% in preclinical models predicted objective clinical responses in patients (30). Mean tumor growth rates are reduced by 60% in the iKO group, 57% in the anti–PD-L1 group, and 88% in the combination group relative to the control, suggesting that this combination might translate into objective response in the clinic.

Map4k4 loss augments activation of naïve CD8 T cells

Because CD8 T cell activation is a multistep process, alteration of any step could account for the observed phenotype with Map4k4-deleted CD8 T cells described above. To determine which step during T cell activation is affected by Map4k4 loss, we analyzed the early response to tumor challenge in draining LNs from mice harboring MC38 tumors for 10 days, a time point when tumor sizes were not substantially different between genotypes. Map4k4^−/− hosts had enlarged draining LNs (Fig. 2A), increased frequency of CD8 T cells recognizing the tumor-specific retroviral antigen p15e (Fig. 2B), and Ki67⁺ CD8 T cells (Fig. 2C) in the draining LNs. Despite increased Ki67⁺ CD8 TILs at this time point, it had not yet led to increased TIL counts (Fig. 2D). We then investigated whether these early changes eventually led to enhanced T cell responses within tumors by examining TILs 16 days after tumor implantation. Total CD8 TILs (Fig. 2E) as well as tumor-specific p15e⁺ CD8 TILs and MC38 neoantigen-specific CD8 TILs (Fig. 2F) were increased in Map4k4^−/− hosts at this time point. To assess the cytokine secretion potential of CD8 TILs, we restimulated MC38 tumor single-cell suspensions ex vivo with a combination of MC38 tumor-specific peptides. Increased numbers of interferon-γ–positive (IFN-γ⁺) CD8 TILs were detected, and these cells secreted more IFN-γ in Map4k4^−/− hosts (Fig. 2G). Together, these data suggest that Map4k4 loss enhanced the early antitumor response in draining LNs that precedes the intratumoral phenotype.

• Download high-res image
• Open in new tab
• Download Powerpoint

The enhanced early antitumor immune response upon Map4k4 deletion implies its role in T cell priming. To evaluate this possibility, we examined the outcome of Map4k4 deletion in naïve CD8 T cells. We first confirmed MAP4K4 protein loss in Map4k4 iKO CD8 T cells by Western blot (fig. S4A). Map4k4^−/− splenic naïve CD8 T cells proliferated more in response to stimulation with anti-CD3/CD28 antibodies assessed by increased live cell counts and 5-bromo-2′-deoxyuridine (BrdU) incorporation (Fig. 3A). In addition, more Map4k4^−/− CD8 T cells expressed IFN-γ and tumor necrosis factor–α (TNF-α) as well as secreted more IFN-γ and interleukin-2 (IL-2) in response to anti-CD3/28 stimulation (Fig. 3B and fig. S4B).

• Download high-res image
• Open in new tab
• Download Powerpoint

Consistent with increased proliferation and IFN-γ/TNF-α frequency, more iKO CD8 T cells displayed activation markers, CD69 and CD44, after 48 hours of stimulation (fig. S4, C and D). The frequency of early T cell activation marker ICAM-1 and both frequency and mean fluorescence intensity (MFI) of NUR77 were also elevated in Map4k4^−/− CD8 T cells after 4 hours of anti-CD3/28 stimulation (fig. S4E). Because NUR77 is not only an early activation marker but also an indicator of TCR signal strength (31, 32), increased NUR77 supports elevated TCR signaling in Map4k4^−/− CD8 T cells. We then stimulated CD8 T cells with lower concentrations of anti-CD3 along with a fixed concentration of anti-CD28 to examine whether these cells have a lower activation threshold. At a lower concentration of anti-CD3 (0.5 μg/ml), WT CD8 T cells lacked IFN-γ secretion, whereas iKO CD8 T cells readily secreted IFN-γ, produced more IL-2, and increased CD69⁺ frequency (Fig. 3C and fig. S4F). These data suggest that Map4k4 deletion enhances activation of naïve splenic CD8 T cells in response to polyclonal stimulation. To determine whether kinase activity is required for enhanced activation, WT naïve CD8 T cells were exposed to GNE-220, a MAP4K4 kinase inhibitor (26), or to the vehicle control, dimethyl sulfoxide (DMSO), during anti-CD3/28 stimulation. Mouse splenic naïve CD8 T cells treated with GNE-220 secreted more IFN-γ and IL-2 and had a higher frequency of CD69 positivity and more live cell counts indicative of proliferation (fig. S4, G to I). We also stimulated human peripheral CD8 T cells with low or high concentrations of anti-CD3/28 in the presence of 1 μM GNE-220, which increased IFN-γ and IL-2 secretion compared with control (fig. S4J). These data indicate that inhibition of MAP4K4 kinase activity accounts for the effect of Map4k4 deletion.

Map4k4 loss enhances antigen-driven CD8 T cell activation

To determine how Map4k4^−/− CD8 T cells respond to antigen-driven stimulation, we crossed Map4k4^fl/fl:Rosa26-CreERT2 (fl, floxed) mice to OT-1 TCR transgenic (Tg) mice. OT-1 CD8 T cells specifically recognize the ovalbumin-derived SIINFEKL peptide (33). Naïve Map4k4^−/− OT-1 CD8 T cells secreted more IFN-γ and IL-2 when stimulated with SIINFEKL-pulsed APCs (Fig. 3D and fig. S5A). Preactivated Map4k4^−/− OT-1 CD8 T cells killed more ovalbumin-overexpressing MC38 tumor cells in vitro (fig. S5B). We also evaluated in vivo priming of splenic CD8 T cells 18 hours after injection with different doses of SIINFEKL peptide (0, 0.1, 1, and 10 μg per mouse) in Map4k4 WT and iKO OT-1 Tg mice. More Map4k4^−/− CD8 T cells were positive for CD69, GZMB, and Ki67, even at lower peptide doses (Fig. 3E), demonstrating that Map4k4-deleted CD8 T cells have enhanced activation at both lower and higher antigen challenge.

Because both the abundance and TCR ligand strength determines the quality of T cell responsiveness, we investigated whether MAP4K4 inhibition affects naïve CD8 T cell priming in response to varying strengths of antigens by using different affinity peptides (34, 35) at a range of concentrations. MAP4K4 inhibition enhanced CD25 and CD69 expression (fig. S6, A and B) and IFN-γ and IL-2 secretion (fig. S6, C and D) of naïve CD8 T cells after they were cocultured with DCs pulsed with high-affinity SIINFEKL (fig. S6, A to D, first column) or medium-affinity SIITFEKL (fig. S6, A to D, T4, second column) peptides in a concentration-dependent manner. However, MAP4K4 inhibition had a weak or negligible effect on CD69 expression or cytokine production after stimulation with low-affinity peptide SIIGFEKL, although it had a modest effect on CD25 expression with peptide (100 μg/ml; fig. S6, A to D, G4, third column). These data suggest that MAP4K4 inhibition enables CD8 T cells to respond to low antigen density and further enhances their activation at high antigen density, but does not alter their sensitivity to antigen affinity.

We then validated the role of MAP4K4 in the early CD8 T cell antigen response in the LCMV–Armstrong strain (LCMV-ARM) infection model that heavily depends on rapid CD8 T cell priming and activation in an antigen-dependent manner (36). We infected Map4k4 WT and iKO mice with LCMV-ARM and analyzed spleens at 5, 7, 8, and 15 days after infection. Spleen weights and CD8 T cell counts were increased in iKO mice at days 7 and 8 (Fig. 4, A and B, and fig. S7, A and B). We analyzed other major cell types in the spleens at day 7 and found no alterations except increased neutrophil numbers (fig. S7, C, D, and F). At day 8, there were more LCMV-antigen–specific (GP33 and NP396) (Fig. 4C) and IFN-γ⁺ CD8 T cells (Fig. 4D) in the iKO spleens without ex vivo stimulation. In addition, more iKO CD8 T cells produced IFN-γ in response to NP396 peptide restimulation ex vivo (fig. S7E). We also analyzed cytolytic activity of CD8 T cells by CD107a surface expression that is associated with release of lytic granule proteins perforin and granzymes (37). MFI of CD107a in the positive population was increased in iKO CD8 T cells (Fig. 4E). On day 15 after infection, we found no difference between Map4k4 WT and iKO spleens (Fig. 4, A and B), indicating that the action of MAP4K4 occurred during the early steps of the antiviral response. Together, these results suggest that antigen-specific acute activation and expansion of CD8 T cells are enhanced upon Map4k4 deletion in the LCMV-ARM infection model.

• Download high-res image
• Open in new tab
• Download Powerpoint

MAP4K4 regulates LFA-1 and ERM activation during CD8 T cell priming

Next, we evaluated whether MAP4K4 controls IS formation between CD8 T cells and DCs during priming. We found that Map4k4-deleted or MAP4K4-inhibited OT-1 CD8 T cells formed more conjugates with SIINFEKL-pulsed WT DCs compared with WT OT-1 CD8 T cells (Fig. 5A and fig. S8, A and B) and LFA-1⁺ T cell–DC contact areas were increased with Map4k4^−/− CD8 T cells (Fig. 5B). On resting T cells, LFA-1 exists in a low-affinity conformation. Upon TCR stimulation, LFA-1 switches to a high-affinity (HA) state that strengthens the contact between T cells and APCs (38). We asked whether Map4k4 deletion affects LFA-1 status. More Map4k4^−/− splenic CD8 T cells displayed HA–LFA-1 on the surface after 30 min (fig. S9A) or 48 hours of anti-CD3/28 stimulation than WT cells (Fig. 5C), with total surface LFA-1 levels being comparable between genotypes (fig. S9C). HA–LFA-1 MFI was also higher among HA–LFA-1⁺ Map4k4^−/− CD8 T cells at 48 hours after stimulation (fig. S9B). More naïve and activated Map4k4^−/− splenic CD8 T cells adhered to mouse ICAM-1–Fc protein–coated plates than WT CD8 T cells (Fig. 5D), whereas adhesion to plates coated with an irrelevant mouse immunoglobulin G (IgG)–Fc protein did not differentiate between T cell genotypes (fig. S9D). These data suggest that MAP4K4 regulates LFA-1 status during priming of CD8 T cells.

• Download high-res image
• Open in new tab
• Download Powerpoint

To explore what acts downstream of MAP4K4, we examined ERM protein phosphorylation (pERM) in CD8 T cells because they are reportedly a direct substrate of MAP4K4 (19) and their activation is linked to T cell priming (17, 18). As expected, the fraction of pERM-high CD8 T cells decreased after 5 min of stimulation with anti-CD3/28 antibodies or specific antigen, and Map4k4 loss enhanced this effect by ~25 to 30% (Fig. 5, E and F). Total ERM levels were comparable between Map4k4 WT and iKO CD8 T cells with and without stimulation (fig. S9E). Because MSN (gene that encodes moesin) is the most abundantly expressed ERM member in CD8 T cells and mediates the inhibitory effect of MAP4K4 on endothelial cell integrin activation (26), we knocked down MSN and MAP4K4 in Jurkat cells, a commonly used T cell line (fig. S9F), and cocultured them with unpulsed or superantigen-pulsed Raji cells. Under both conditions, knocking down MSN was sufficient to increase HA–LFA-1 on Jurkat cell surfaces (fig. S9G) and enhanced adhesion to human ICAM-1–Fc similar to knocking down MAP4K4 (fig. S9H), whereas MSN or MAP4K4 knockdown did not alter adhesion to the negative control human IgG1-Fc protein (fig. S9I). We confirmed ICAM-1–Fc binding results with MSN or MAP4K4 knockdown in human peripheral blood CD8 T cells (Fig. 5G). Conversely, overexpression of constitutively active forms of MAP4K4 and moesin in CD8a-positive TALL-1 cells or Jurkat cells reduced adhesion to human ICAM-1–Fc (Fig. 5H and fig. S9J). Therefore, moesin and MAP4K4 are necessary and sufficient for regulating LFA-1 status and interaction with its ligand ICAM-1. Last, we knocked down MAP4K4 and MSN together in primary human CD8 T cells and Jurkat cells to assess additive effects on adhesion and observed equivalent adhesion to single knockdown (Fig. 5G and fig. S9K), indicating that they most likely function in a linear pathway.

The activity of MAP4K4 in CD8 T cells requires LFA-1

We next asked whether LFA-1 activity mediates the phenotype observed with Map4k4 loss. Elevated conjugation of Map4k4^−/− or MAP4K4-inhibited OT-1 CD8 T cells with SIINFEKL-loaded WT APCs was reversed with increasing concentrations of anti–LFA-1 blocking antibody (Fig. 6A). Blocking LFA-1 during antigenic stimulation of CD8 T cells with SIINFEKL-loaded DCs also diminished the effect of Map4k4 iKO on IFN-γ and IL-2 secretion (Fig. 6B). Increased adhesion of MAP4K4- or MSN-knockdown Jurkat cells on human ICAM-1–Fc–coated surfaces was also reversed by a small-molecule inhibitor that blocks LFA-1–ICAM-1 interactions (Fig. 6C). Furthermore, blocking LFA-1 during in vivo priming with SIINFEKL peptide attenuated activation markers CD69, GZMB, and Ki67 in Map4k4^−/− splenic OT-1 CD8 T cells (Fig. 6D). Last, blocking LFA-1 during LCMV-ARM infection restored Map4k4^−/− splenic NP396-tetramer⁺ CD8 T cell counts and IFN-γ frequency among the NP396-tetramer⁺ CD8 T cells to near WT levels (Fig. 6E). Thus, LFA-1 function is necessary for the phenotypes associated with Map4k4 loss in CD8 T cells, supporting the notion that LFA-1 acts downstream of MAP4K4. To strengthen this conclusion, we tested whether the MAP4K4 inhibitor and LFA-1 agonistic antibody would have similar effects on CD8 T cell activation. We stimulated mouse splenic CD8 T cells with anti-CD3 alone or together with an LFA-1 cross-linking antibody (TS1/22) in the presence of DMSO or GNE-220. Cross-linking LFA-1 and GNE-220 treatment had comparable effects on T cell activation and IFN-γ expression (Fig. 6F). Except for a slight increase in CD25 frequency, MAP4K4 inhibition did not further enhance the impact of cross-linking LFA-1.

• Download high-res image
• Open in new tab
• Download Powerpoint

We then examined whether LFA-1 mediates the antitumor effect in Map4k4 iKO hosts. Consistent with the finding that LFA-1–deficient mice were not primed by tumor antigens and failed to reject immunogenic tumors (39), anti–LFA-1 treatment increased MC38 tumor growth rates and abolished the differences between Map4k4 iKO and WT hosts (Fig. 7A), indicating that LFA-1 is required for MAP4K4 function. Map4k4 loss increased CD8 TIL reactivity to the MC38 antigen p15e or a combination of MC38 neoantigens that was abrogated upon LFA-1 blockade (Fig. 7, B and C). To assess the cytolytic potential of autologous CD8 TILs, MC38 tumor single-cell suspensions were restimulated ex vivo with a combination of MC38 tumor-specific peptides, and membrane CD107a on CD8 TILs was measured. Map4k4^−/− CD8 TILs increased the percentage and MFI of membrane CD107a to a greater extent than WT CD8 TILs, an effect that was disrupted by LFA-1 blockade (Fig. 7D). Last, we examined whether inhibition of MAP4K4 during ex vivo restimulation is sufficient to enhance CD107a membrane expression. MAP4K4 inhibitor GNE-220 treatment increased the percentage and MFI of CD107a on CD8 TILs above vehicle control upon tumor antigen restimulation, and the effect of GNE-220 was abolished by the anti–LFA-1 antibody (Fig. 7E). Although anti–LFA-1 increased tumor growth in Map4k4 WT hosts, it did not consistently affect CD8 TILs in the ex vivo studies. We hypothesize that this difference likely reflects functions of LFA-1 in other cell type(s) with antitumor activity such as natural killer cells and CD4 T cells. The role of MAP4K4 in other type(s) of intratumoral leukocytes remains to be determined in future studies.

• Download high-res image
• Open in new tab
• Download Powerpoint

MAP4K4 is down-regulated upon TCR engagement

Because MAP4K4 negatively regulates CD8 T cell priming and activation, we examined how MAP4K4 itself is regulated during IS formation. Phosphorylation of a critical residue in the MAP4K4 kinase domain, threonine-181 (T181), was rapidly reduced upon anti-CD3/28 stimulation in Jurkat cells and primary human CD8 T cells. The kinetics of MAP4K4 dephosphorylation mirrored the rapid phosphorylation of proximal TCR components such as SLP76, PLCγ, ERK, ZAP70, and NF-κB p65 (fig. S10, A and B). Total MAP4K4 protein levels decreased gradually after activation in both Jurkat and primary human CD8 T cells (fig. S10, A and B). MAP4K4 was detected in the cortex of Jurkat cells and OT-1 CD8 T cells and at the synapse between pairs of Jurkat-Raji cells or OT-1 CD8-DCs (fig. S11, A and D). When Jurkat cells were cocultured with superantigen (SEE)–loaded Raji cells, or OT-1 CD8 T cells were cocultured with SIINFEKL-loaded DCs for 60 min, MAP4K4 levels were reduced at the synapse and the cortex (fig. S11, A, B, D, and E). Random contacts between T cells and unpulsed APCs did not reduce MAP4K4 levels compared with isolated single T cells (fig. S11, A, B, D, and E), suggesting that this process is dependent on TCR engagement. Quantification of Jurkat-Raji and OT-1 CD8-DC doublets in the presence of antigen showed that co-occurrence of CD3 and MAP4K4 was reduced at the synapse upon IS formation (fig. S11, C and F). Together, these data indicate that MAP4K4 is down-regulated upon TCR engagement, and this mode of regulation is consistent with MAP4K4 being a negative regulator of CD8 T cell priming and activation.

Pan-tissue deletion of Map4k4 is tolerated in adult mice

As MAP4K4 is broadly expressed and regulates different biological processes in many cell types (20–24, 26, 27), we evaluated the consequences of deleting Map4k4 in adult mice under normal physiological conditions 5 weeks or 5 months after tamoxifen administration. Hematological analyses of WT and iKO tissues showed no or modest alterations (fig. S12, A and B) at 5 weeks after tamoxifen injection. Because platelet counts were slightly decreased in iKO mice, we performed a panel of coagulation measurements. Map4k4^−/− mice exhibited a slightly enhanced coagulation potential indicated by a decrease in activated partial thromboplastin time (fig. S12C); however, this was not associated with evidence of overall hypercoagulability/disseminated intravascular coagulation as noted by the comparable fibrinogen levels, prothrombin time, and previous platelet counts between the genotypes. Therefore, the significance of these marginal changes is unclear. Furthermore, platelet counts of Map4k4 iKO mice went back to WT levels at 5 months (fig. S12D). At this time point, flow cytometric (FCM) analysis of thymus showed no alterations in the major cell populations (fig. S12E). FCM analysis of spleen and mesenteric LNs showed a modestly decreased frequency of naïve (CD62L⁺ CD44⁻) CD8 T cells with a slightly increased frequency of effector/memory (CD44⁺ CD62L⁻) CD8 T cells in Map4k4 iKO mice (fig. S12, F and G). Therefore, MAP4K4 deletion does not lead to overt abnormalities in mice.

Higher MAP4K4 expression correlates with poorer outcome in cancer patients

The relationship between MAP4K4 expression reportedly correlates negatively with patient outcomes in several cancer types (24, 40, 41). We substantiated these findings with data from six clinical trials that evaluated the effects of the anti–PD-L1 antibody atezolizumab: four trials for lung cancer (FIR, POPLAR, BIRCH, and OAK) and two urothelial bladder cancer trials (IMvigor210 and IMvigor211). In three lung cancer trials, high MAP4K4 expression correlated with poorer overall survival irrespective of treatment (fig. S13, A to C). A much smaller lung cancer trial showed a similar trend (fig. S13D, FIR). In two urothelial bladder cancer trials, a similar correlation was seen in IMvigor210, but not in IMvigor211 (fig. S13, E and F). Our data, along with published literature, are consistent with MAP4K4 being a negative regulator of several anticancer biological processes.