Antibody Blockade of Semaphorin 4D Promotes Immune Infiltration into Tumor and Enhances Response to Other Immunomodulatory Therapies

Tumor growth depends on dynamic interactions within a complex ecosystem, including the proliferating neoplastic cells, immune cells, and stroma that constitute the tumor microenvironment (TME; refs. 1, 2). There is increasing evidence that effective immunotherapies can harness the immune system to promote durable tumor regression and prolong survival. To lyse neoplastic cells, leukocytes must become activated by antigen-presenting cells (APC) and gain entry into the TME, while avoiding escape mechanisms induced within the TME. Such escape mechanisms, including repulsive proteins and anergic signals, together with the architecture of the TME, modulate T-cell traffic and activation. Coordinated Th1 and cytotoxic T cell (CTL) immune infiltration into the central core and the invasive margins of human colorectal tumors (deemed “immunoscore”) has been correlated with a favorable clinical outcome (3–5); however, entry of these effectors into TME is often restricted (1, 6). Increased immune infiltration into TME may enable more effective antitumor responses.

Semaphorins consist of a family of soluble and transmembrane proteins, originally defined as axonal-guidance factors (7) that also induce cytoskeletal changes in immune, endothelial, and tumor cells and guide their migration (8–10) in the TME. For example, associations of Sema3a and Neuropilin-1 have been reported to regulate movement of tumor-associated macrophages (TAM) in hypoxic regions (11). Herein, we describe a novel role for the regulation of leukocyte infiltration and tumor growth by semaphorin 4D (SEMA4D, CD100).

SEMA4D has been reported to inhibit immune-cell migration (12), and its expression may be regulated by hypoxia (13–15) and other cues within the TME. TAMs have been reported to be a necessary source of SEMA4D within the TME (16). The membrane-bound form of SEMA4D is expressed as a disulfide-linked homodimer. Upon cellular activation, the extracellular portion can be proteolytically cleaved to generate a 240-kDa soluble homodimer (17, 18). Both soluble and cell-associated forms of SEMA4D are biologically active (19). Immunohistochemical (IHC) analysis of several tumor types revealed that both SEMA4D and its receptors are overexpressed on various human and murine tumors, including human head and neck, prostate, colon, breast, and lung cancers (20). Moreover, elevated expression correlates with invasive disease and poor prognosis in human cancer (21–25).

Three cellular receptors have been identified for SEMA4D: Plexin-B1 (PLXNB1), Plexin-B2 (PLXNB2), and CD72. PLXNB1, the highest-affinity SEMA4D receptor (K_D = 1 nmol/L), is expressed on APCs, as well as in nonlymphoid tissues, including some tumor, endothelial, and neural cells (26, 27). SEMA4D engagement with PLXNB1 has been shown to induce activation and migration of endothelial cells and regulate migration of tumor cells (26, 28). PLXNB1 binds to and can transactivate oncogenic membrane tyrosine kinases, such as cMET and ERBB2 (21, 22, 26, 28), when cross-linked by SEMA4D. PLXNB2, whose best characterized ligand is SEMA4C, also binds SEMA4D with intermediate affinity. A recent report indicates that PLXNB2 is expressed on keratinocytes and can activate SEMA4D-positive γδ T cells to contribute to epithelial wound repair (29). CD72 is a relatively low-affinity (K_D = 300 nmol/L) SEMA4D receptor (2) that is expressed primarily on B cells, APCs, and platelets (30).

Our understanding of the complex role of SEMA4D and its receptors in the process of tumor growth and metastasis is evolving. In this report, we describe the effects of antibody blockade of SEMA4D on modulation of the stroma, focusing on tumor-directed immune responses in two syngeneic tumor models. Because TAMs have been reported as a stromal source of SEMA4D (16), and the balance of M1/M2 macrophage has been identified as a prognostic factor in colon cancer (31), we selected a murine colon cancer model. In addition, given the reported role of PLXNB1 and ERBB2 interactions in oncogenic signaling (21), we also investigated the effect of anti-SEMA4D in an ERBB2-dependent mammary tumor model. We further investigate the mechanism of action by comparing and contrasting the effect of anti-SEMA4D antibody in combination with other immunomodulatory therapies.

Colon26 cells were purchased from the NCI (DTP, DCTD Tumor Repository, Frederick, MD) and were cultured in RPMI-1640 with 20% FBS and 0.05 mmol/L 2-mercaptoethanol. Tubo cells, a generous gift from Guido Forni (University of Torino, Italy), were cultured in DMEM with 20% FBS. Clone Tubo.A5 was isolated by FACS sorting from the Tubo cell line as a stable clone expressing high levels of rat Neu. All cells were confirmed to be Mycoplasma free; authentication (32) confirmed cell line identification and relation of clones to parental lines.

All mouse experiments were carried out according to the guidelines of the NIH and the University Committee on Animal Resources (UCAR) at the University of Rochester (Rochester, NY) for the care and use of laboratory animals. Female mice, 8 to 10 weeks of age, were used for all studies—BALB/cJ were purchased from Charles River or The Jackson Laboratories, and CB17 SCID were from Charles River. Some experiments were conducted at Charles River Discovery Research Services (Morrisville, NC).

Anti-SEMA4D monoclonal antibodies (MAb) were generated following immunization of SEMA4D^−/− mice with rSEMA4D-his. Splenocytes were fused with Sp2/0-Ag14 mouse lymphoblast cells to generate hybridomas; hybridoma clone 67-2 was selected on the basis of affinity, specificity, and functional ability to block the binding of SEMA4D to its receptors. Anti–CTLA-4 clone 4F10-UC10-11 and anti-human CD20 clone 2B8.1E7 hybridomas were obtained from the American Type Culture Collection (ATCC). 2B8.1E7, which is specific for human CD20 and does not cross-react with murine B cells, was used as an irrelevant control IgG1, isotype-matched with MAb67. Antibodies were purified from hybridoma supernatants using ProG affinity column and ion exchanges, and were confirmed to have <0.5 EU/mg endotoxin levels, ≥95% purity, and ≤5% high-molecular-weight species. Other antibody clones, including anti–PD-1/MAb RMP1-14, anti-CD8/MAb 2.43, and anti-CD4/MAb GK1.5, were purchased from BioXCell and were verified to have <2 EU/mg endotoxin levels, ≥95% purity, and ≤5% high-molecular-weight species before in vivo use.

Colon26 cells (500,000) were implanted s.c. into Balb/c mice. Tubo cells (30,000) were implanted s.c. into the mammary fat pad of Balb/c mice. Similar growth kinetics were observed with 5-fold fewer Colon26 (100,000) or 2-fold fewer Tubo.A5 (15,000) cells implanted into SCID mice. Generally, 20 mice per group were implanted; mice were excluded from the analysis if sacrificed due to tumor excoriation or unexplained death before reaching endpoint tumor volume. Cyclophosphamide (Baxter) was diluted with sterile saline on each treatment day from the stock solution to provide a 50 mg/kg dose. Tumor growth was monitored using calipers two to three times per week and tumor volume (mm³) was calculated using the formula for elliptical spheroid: (w² × l)/2.

Paraformaldehyde -fixed, paraffin-embedded sections (5 μm) were stained with the following antibodies: polyclonal anti-CD8 (AbBiotec), anti-CD86 (BioSS), anti-CD20 (Epitomics), anti-FoxP3 (FJK-16s biotin-conjugated; ebioscince), anti-F4/80 (BM8; eBioscience), anti–CD206-MMR (CO68C2; BioLegend), and anti-CD11c (AP-MAB0806; Novus). Slides were imaged at ×20 magnification using a Retiga QICAM-12-bit camera (brightfield) or QICAM Aqua-16 bit (Fluorescent) coupled to an Olympus Ix50 microscope. Additional details are included in Supplementary Methods.

Cohorts of tumor-bearing mice were sacrificed at a predetermined endpoint, based on mean tumor volume (MTV) of the control group. Tumor-infiltrating leukocytes (TIL) were isolated by dissociating tissue into small pieces and digestion with 2 mg/mL collagenase D and 0.1 mg/mL DNase I for 20 minutes at 37°C. Digested tumors were then clarified by passing through a 0.07-mm cell strainer, followed by RBC lysis (Bioscience) for 5 minutes at room temperature. TILs were further enriched from tumor suspensions using Lympholyte-M (maximum of 25 × 10⁶ cells with 5 mL Lympholyte-M), or alternatively were mixed with αCD45 microbeads (Miltenyi Biotec) and enriched into CD45⁺ leukocytes by positive selection on an AUTOMACS column. Spleens were dissociated into small pieces and the tissue homogenate was pressed through a 0.07-mm cell strainer, followed by RBC lysis (Bioscience) for 5 minutes at room temperature.

Splenocytes and TILs were assayed by Cytometric Bead Array (CBA) cytokine analysis. Cells (200,000) were cultured in triplicate. On day 2, culture supernatants were collected. Fifty microliters of supernatant was analyzed using BD's mouse Inflammation CBA kit on a FACSCanto II flow cytometer. CBA data analysis was completed using FCAP Array software.

Colon26 splenocytes and TILs were assayed for tumor-specific CD8⁺ responders using Mouse IFNγ ELISPOT Ready-SET-GO (Affymetrix), according to the manufacturer's instruction. Cells (5,000–10,000 TILs/well or 250,000 splenocytes/well) were plated in triplicate on MAIPS-4510 Immobilon—P membrane (Millipore) ELISPOT plates, and incubated for 18 hours at 37°C, in the presence or absence of MHC-I/H-2L^d–restricted immunodominant AH1 peptide of gp70 (0.01 mmol/L; ref. 33). The plates were developed using AEC (Sigma) substrate solution and analyzed by Cellular Technology Ltd.

Fluorescent-conjugated rat or hamster MAbs to CD3 (clone: 17A-2), CD8a (clone: 53-6.7), CD11b (clone: M1/70), Gr-1 (clone: RB6-8C5), CD69 (clone: H1.2F3), CD25 (clone: PC61) with appropriate isotype controls were purchased from BioLegend. Samples were run on a FACSCanto II flow cytometer and analyzed using the FlowJo software.

GraphPad was used for statistical and graphical analysis. Prism reports results as statistically nonsignificant (ns) at P > 0.05, statistically significant at 0.01 < P ≤ 0.05 (symbolized as *), very significant (**) at 0.001 < P ≤ 0.01, and extremely significant (***) at P ≤ 0.001. All error bars on graphs depict SEM. Additional information is included in Supplementary Data.

Complex interactions between the tumor and stroma, including lymphocytes and abundantly present macrophages, influence tumor growth and invasion. SEMA4D and its receptors are expressed on both tumors (Supplementary Fig. S1) and stromal cells. To evaluate the effects of SEMA4D in this complex environment, we first assessed SEMA4D expression in the TME in situ. Although expression of SEMA4D on Colon26 cells cultured in vitro was determined to be low by FACS analysis (data not shown), SEMA4D expression is strongly induced in vivo, forming a gradient with particularly high expression at the invasive margin of the tumor (Fig. 1 and Supplementary Fig. S2A). This area of high SEMA4D expression demarcated a striking pattern of mononuclear phagocytic cell distribution, with F4/80⁺ cells restricted to the inner core and CD11c⁺ cells and CD206⁺ M2 macrophage restricted to the outer stromal edge. Anti-SEMA4D antibody blockade with murine MAb67 disrupted this gradient and was associated with increased penetration of double-positive F4/80⁺/CD11c⁺ inflammatory APCs and an apparent decrease in the density of protumorigenic CD206⁺ M2 TAMs at the leading edge (Fig. 1A–D).

Concomitantly with the altered distribution of APCs, we observed increased infiltration of CD8⁺ staining T cells into the tumor leading edge (Fig. 2A). CD45⁺ TILs were isolated from dissociated tumors, and an increase in CD8⁺ TILs was confirmed by FACS analysis (Fig. 2B). Functional activity of CD8⁺ TILs was evaluated using ELISPOT to determine tumor-specific CTL activity in response to a previously identified Colon26 MHC-I–restricted immunodominant peptide, AH1 (34). Tumor peptide–specific CD8⁺ T cells secreting IFNγ were significantly increased in tumor following in vivo treatment with MAb67 (Fig. 2C).

To evaluate cytokine activity associated with an increase in activated APCs following anti-SEMA4D MAb67 antibody treatment, total TILs and splenocytes were isolated from Colon26 tumor–bearing mice. TILs from mice treated with MAb67 produced higher levels of proinflammatory Th1 cytokines, IFNγ and TNFα, with a decrease in the immune-suppressive cytokine MCP-1 (Fig. 2D; ref. 35). In contrast, no statistically significant changes in cytokine levels were observed in the spleens of mice treated with MAb67 (Fig. 2E). Collectively, these data demonstrated that MAb67 enhanced APC and T-cell infiltration and promoted a Th1 cytokine milieu in the TME. The effects appear to be localized to the TME, as there was no evidence of similar activity in peripheral lymphoid tissue.

In Balb/c mice, MAb67 delayed Colon26 tumor growth and increased survival by as much as 29% (P < 0.0001; Fig. 3A and B). Significant tumor-growth delay (TGD), ranging from 8% to 29%, and complete tumor regression (CR) of approximately 8% were reproducible effects observed in many independent experiments performed in the Colon26 Balb/c tumor model (Table 1), both in our laboratory and as determined independently at Charles River Discovery Services. Mice that underwent complete tumor regression remained tumor free for at least 20 weeks and subsequently rejected rechallenge with homologous Colon26 tumor (Table 1). It was, therefore, striking that depletion of CD8⁺ T cells completely abrogated antitumor effects of MAb67 treatment (Fig. 3C), indicating that CTLs are the main effectors of tumor inhibition in this model.

Despite the evidence for a changed bias in the distribution of immune subsets in TME, the effect of anti-SEMA4D/MAb67 on Colon26 tumor growth was relatively modest. We considered the possibility that the modest effect of anti-SEMA4D on tumor growth might be due to relatively low immunogenicity and a corresponding paucity of immune cells that might benefit from enhanced tumoral infiltration mediated by anti-SEMA4D antibody. In view of recent findings of durable antitumor effects of blocking immune checkpoint signals, as well as data that immunomodulatory therapies can be greatly improved when combinations are used (36, 37), we hypothesized that enhanced immune infiltration into tumor induced by anti-SEMA4D antibody may be particularly effective in combination with inhibitors of immune checkpoint blockade.

The combination of anti-SEMA4D and anti–CTLA-4 antibodies was evaluated in Colon26 tumor–bearing mice. To determine additive effects in this model, a suboptimal anti–CTLA-4 regimen was empirically determined (data not shown). As shown in Fig. 4A, under these conditions, anti–CTLA-4 as single agent resulted in 3% TGD (P = 0.0706). Anti-SEMA4D significantly delayed tumor growth by 9% (P = 0.0064) and tumors regressed in 10% of mice. Strikingly, the combination of anti–CTLA-4 with anti-SEMA4D resulted in maximum possible TGD of 114% and 85% CR (P < 0.0001 compared with control Ig or monotherapy treatments). It has been reported that combination with anti–PD-1 can also significantly improve the response to anti–CTLA-4 in this model (38). As a benchmark, we compared the combination of anti–CTLA-4 with anti–PD-1, resulting in 60% CR (P < 0.0001). The combination of anti-SEMA4D with anti–CTLA-4 was the most effective therapy in this model.

To determine whether these dramatic combination effects would translate to other immunotherapies, we tested combinations of anti-SEMA4D with either anti–PD-1 or cyclophosphamide in the Colon26 model. Figure 4B demonstrates increased efficacy when anti-SEMA4D and anti–PD-1 antibodies are combined. Using a dose of anti–PD-1 consistent with reported efficacy in this model (39), anti–PD-1 monotherapy resulted in 6% TGD (P = 0.0483) and 5% CR, while anti-SEMA4D/MAb67 treatment resulted in 9% TGD (P = 0.0025) and 5% CR. The combination of anti-SEMA4D with anti–PD-1 significantly improved the response to 67% TGD (P < 0.0001 compared with control and P = 0.0025 compared with anti–PD-1 monotherapy) and 35% CR (P = 0.003 compared with control and P = 0.04 compared with anti–PD-1 monotherapy). The additive benefit of this combination, however, contrasts with the more dramatic effect of anti-SEMA4D in combination with anti–CTLA-4. As discussed below, this is consistent with the different locus of activity of each of these agents.

Cyclophosphamide has been shown to mitigate Treg-mediated immunosuppression in rodents and in patients with cancer (34), as well as to promote immunogenic cell death and T cell–mediated immunity (40). To determine combination effects with anti-SEMA4D, a suboptimal dose of cyclophosphamide was administered 1 day before immunotherapy. Both MAb67 and cyclophosphamide significantly improved survival, with 10% and 15% CR, respectively, while the combination improved the TGD to 256% and significantly increased regression frequency to 40% CR (P < 0.001 for treatment effect and survival; P = 0.0033 for CR; Fig. 4C). Although the survival difference between cyclophosphamide monotherapy and the combination of cyclophosphamide plus MAb67 was not statistically significant (P = 0.1), these results suggest a trend to improved response and warrant additional studies with this and other chemotherapeutic combinations.

To further investigate the mechanism of action for the highly effective combination of anti-SEMA4D plus anti–CTLA-4, a series of experiments were conducted to assess effects on immune-cell composition and activity in tumor. Colon26 tumor–bearing mice were sacrificed before tumor rejection, at a time when tumor volumes of treated mice were comparable (Supplementary Fig. S3). Combination therapy significantly enhanced the percentage of CD8⁺ cells in TILs (Fig. 5A). Functional activity of these T cells was evaluated using ELISPOT (Fig. 5B). Similar to previous results, anti-SEMA4D significantly increased the frequency of tumor peptide–specific CD8⁺ T cells secreting IFNγ within the tumor, either as a single agent or in combination with anti–CTLA-4. Although, as expected, anti–CTLA-4 antibody increased the frequency of tumor-reactive CTLs in both tumors and spleens, there was not a further increase in activated CD8⁺ cells in the spleens of mice treated with the antibody combination, suggesting that the activity mediated by anti-SEMA4D treatment is localized to the tumor, even when combined with anti–CTLA-4 treatment. The important role of CD8⁺ T cells in protective immunity induced by treatment with the combination of anti-SEMA4D and anti–CTLA-4 is further supported by the demonstration that in mice that have rejected tumor following such combination treatment, depletion of CD8⁺ cells abrogates the specific immunologic memory response to a subsequent challenge with homologous tumor (Supplementary Fig. S4).

Anti–CTLA-4 therapy has been reported to increase the ratio of effector cytotoxic T cells (Teff) to suppressor regulatory T cells (Treg). Anti-SEMA4D and anti–CTLA-4 demonstrated independent and combinatorial effects on Teff to Treg ratios as determined by FACS analysis (Teff = CD8⁺/CD69⁺; Treg = CD4⁺/CD25⁺; Fig. 5C and Supplementary Fig. S5A). In support of this observation, IHC analysis using an alternate set of markers (CD8 and FoxP3) also demonstrated that the combination of anti–CTLA-4 with anti-SEMA4D significantly increased the Teff:Treg ratio compared with either monotherapy.

We also evaluated the balance of other immune cells and cytokines in this model. Anti-SEMA4D treatment increased the percentage of activated monocytes (CD86⁺) in the tumor, while combination with anti–CTLA-4 resulted in a further and significant increase in CD86 expression relative to anti–CTLA-4 alone (Fig. 5D). Moreover, we observed a significant increase in proinflammatory cytokines, IFNγ, TNFα, and IL6, following combination therapy (Fig. 5E). Conversely, combination therapy resulted in a significant decrease in immunosuppressive cytokines MCP-1 and IL10, which were not reduced by anti–CTLA-4 therapy alone. These data suggest an orchestrated shift in the balance of pro- and anti-inflammatory signals, leading to active tumor rejection following treatment with anti-SEMA4D and anti–CTLA-4.

Collectively, these data demonstrate a novel immunomodulatory mechanism of action for SEMA4D that enhanced the activity of other immunomodulatory agents and resulted in durable tumor rejection and resistance to subsequent tumor challenge (Table 1).

Previous studies (21) have demonstrated an important role for the high-affinity PLXNB1 receptor of SEMA4D in the ERBB2 signaling pathway. It was, therefore, of interest to determine how antibody blockade of SEMA4D would affect growth in a tumor expressing ERBB2 and PLXNB1. The orthotopic Tubo.A5 murine mammary carcinoma model afforded an opportunity to investigate the interaction between immune modulation and interference with the ERBB2 signaling pathway. Strikingly, treatment of Tubo.A5 with MAb67 as a single agent resulted in maximal TGD (Fig. 6A and B and Table 1) and 86% CR, with no recurrences up to 150 days, while the median time to endpoint of the control group was 38 days. At 90 days after implant, only one mouse treated with MAb67 was sacrificed because of tumor burden, another tumor had shrunk from a maximum of 718 to 214 mm³, one mouse had never developed measurable tumor, and 12 additional mice were tumor-free regressors. Similar effects were observed at doses as low as 3 mg/kg of MAb67 anti-SEMA4D antibody (Supplementary Fig. S6B), which corresponds to serum drug levels of 0.003 to 0.005 mg/mL and is also the minimal drug concentration required to fully saturate SEMA4D on circulating T cells (Supplementary Fig. S6C). This suggests that antibody saturation of SEMA4D on circulating T cells may serve as a useful PD marker in further clinical development.

To investigate immunologic memory, 12 tumor-free regressor mice were rechallenged with viable Tubo.A5 tumor in the contralateral breast tissue on day 90 (Fig. 6B). All of the MAb67-treated regressor mice rejected subsequent challenge with Tubo.A5. In other experiments, Tubo.A5 regressor mice challenged with Colon26 were unable to reject that challenge (data not shown), demonstrating that MAb67 monotherapy promoted a tumor-specific memory response (see also Table 1).

As in the Colon26 model, neutralization of a strong SEMA4D gradient at the invasive tumor front (Supplementary Fig. S2B) was associated with a shift in the balance of proinflammatory monocyte infiltration and activity, as well as increased T-cell penetration in Tubo.A5 tumors. A significant increase in F4/80⁺/CD11c⁺ APCs (Fig. 6C) and a concomitant significant decrease in CD11b⁺Gr1⁺ immunosuppressive myeloid-derived suppressor cells (MDSC; P < 0.0001) and MCP-1 cytokine secretion (P = 0.0079) were observed in Tubo.A5 TILs (Fig. 6D). In addition, a significant increase in the frequency of CD3⁺ T cells (P = 0.0012; Fig. 6E) and a trend toward an increase in the ratio of Teffs and Tregs following treatment were also observed (Supplementary Fig. S5B).

In striking contrast to treatment of Colon26, anti-SEMA4D as a single agent resulted in maximum possible TGD and average of 61% complete tumor regression in the Tubo.A5 model. The kinetics of regression suggest that several weeks were required for an effective immune response to be induced in this model, during which time the development of an aggressive phenotype may have been held in check by blockade of the ERBB2 pathway. Accordingly, immune checkpoint blockade was found to be relatively ineffective by itself in the Tubo model (Supplementary Fig. S7A and Table 1). Anti–PD-1 was minimally effective in Tubo.A5 at doses up to 0.25 mg when administered either prophylactically at 6 dpi (data not shown) or therapeutically when tumor volume reached approximately 100 mm³. Anti–CTLA-4 treatment was only marginally effective (23% CR, ns), and combination of the two immune checkpoint antibodies did not significantly improve responses in the Tubo model. This suggests that, in the absence of blocking the ERBB2 signaling pathway, immune resistance was not sufficient to inhibit tumor growth. The presence of an immune-independent mechanism in Tubo.A5 is further supported by the observation that MAb67 treatment induced modest, but significant, TGD in Tubo.A5 grown in SCID mice. This contrasts with results in the ERBB2-negative Colon26 model, where effects of MAb67 were completely abrogated in immune-deficient SCID mice (Supplementary Fig. S7B and S7C). Most importantly, however, the failure of tumor rejection in SCID mice highlights the requirement for lymphocytes as effectors in both models.

These data demonstrate that SEMA4D expression at the invasive tumor edge creates a barrier to immune infiltration and biases the balance of regulatory and effector immune cells and signals. In effect, antibody-mediated SEMA4D blockade “opens the gates” to the tumor and shifts the balance toward antitumorigenic immune activity within the tumor. In two tumor models, we have described MAb67-mediated tumor rejection, corresponding with increased immune activity in the TME. Anti-SEMA4D treatment elicits a resilient antitumor response with immunologic memory, as demonstrated by durable regressions, lack of primary regrowth, or metastasis up to 17 weeks after tumor rejection, and resistance to rechallenge with the same, but not different, tumor. SEMA4D blockade represents a novel immunomodulatory mechanism to promote durable tumor rejection.

The ability of SEMA4D blockade to enhance leukocyte infiltration is consistent with reports that SEMA4D and SEMA3A can inhibit spontaneous and chemokine-induced migration of monocytes and B cells (12). The precise molecular mechanism controlling TIL recruitment in different tumor models remains to be determined, but may be explained by several reported activities for SEMA4D, including (i) inhibition of lymphocyte migration (12), (ii) regulation of stromal components and fibrotic “wound repair” mechanisms (29), (iii) the SEMA4D–PLXNB1–RhoA signaling axis on endothelial cells regulating extravasation and/or angiogenesis (41), and (iv) the association of PLXNB1 with receptor tyrosine kinases, including ERBB2 and MET (26, 42), affecting neoplastic survival and perhaps increasing immunogenicity of the tumor.

Following SEMA4D blockade, we observed significantly reduced levels of MCP-1, which can enhance recruitment of immature monocytes into the tumor, where they may differentiate into TAMs depending on the microenvironment (43). We also observed SEMA4D-mediated effects on IL10, which can promote polarization toward a CD206⁺ M2 phenotype and decrease expression of CD86 within the TME (35). Indeed, along with MAb67-induced reduction in IL10, we observed an increase in activated CD86⁺ APCs with a concomitant decrease in CD11b⁺Gr1⁺ MDSCs and in the density of CD206⁺ M2 macrophages following SEMA4D blockade. Tumor-associated M2 macrophages can interact with lymphocytes to promote tumor growth; for example, TAMs can regulate the influx of CD4⁺ Tregs and can inhibit cytotoxic CD8⁺ T-cell responses through several mechanisms (44). MCP-1 and IL10 can also affect Teff:Treg ratios and reduce IFNγ production (45, 46). By reducing these suppressive signals, we revealed an important shift toward functional tumor-specific, IFNγ-secreting, CD8⁺ cytotoxic T cells and an increase in the ratio of Teffs to Tregs. Increased T-cell activity also correlated with increased frequency of CD11c⁺F4/80⁺ APCs, which represent potent APCs that can cross-present tumor antigens and locally activate tumor-specific T cells (47). Coordinated infiltration of CD8⁺ and Th1 cytokine–secreting cells suggests that SEMA4D antibody blockade increases immunoscore.

The observation that anti-SEMA4D promotes immune infiltration into the TME suggested that this therapeutic might act in synergy with other immunomodulatory therapies. Indeed, anti-SEMA4D enhanced activity of several therapies reported to promote antitumor immune responses through diverse modes of action. It was striking that combination of anti-SEMA4D with anti–CTLA-4 proved more effective than combination with another immune checkpoint blockade inhibitor, anti–PD-1. Both CTLA-4 and PD-1 are upregulated on activated T cells and promote T-cell hyporesponsiveness; however, each plays a nonredundant role in modulating immune responses. CTLA-4 plays a pivotal role in attenuating the early activation of naïve and memory T cells in draining lymph nodes where the corresponding ligands, CD80 (B7-1) and CD86 (B7-2), are highly expressed. In contrast, PD-1 is primarily involved in modulating T-cell activity in tissues via its interaction with locally induced PD-L1 and PD-L2. Because anti–CTLA-4 enables greater expansion of tumor-specific T cells in draining lymph nodes in the presence of cognate ligands, B7-1 and B7-2, anti-SEMA4D antibody can act synergistically to facilitate infiltration of these expanded T cells into the tumor. In contrast, to the extent that anti–PD-1 prevents PD-L1 expressed by tumor from inactivating tumor-specific T cells that have already penetrated into the tumor environment (48), the added benefit of anti-SEMA4D in facilitating tumor infiltration may be more limited. Other immunotherapeutic interventions that activate antitumor immune responses in the periphery, such as costimulatory factors and cancer vaccines, may also benefit from disruption of immune barriers by combination with anti-SEMA4D treatment and warrant further investigation.

In the Tubo.A5 model, potent single-agent efficacy of anti-SEMA4D treatment contrasts with the need for combination therapy in the Colon26 model. Tubo.A5, unlike Colon26, expresses both PLXNB1 and ERBB2. Cross-linking of PLXNB1 by SEMA4D has been reported to transactivate ERBB2, and it has been recently reported that the ERBB2 signaling pathway must pass through PLXNB1 (42). Interference with this oncogenic pathway by anti-SEMA4D might well enhance the independent immunologic activities of this antibody by either increasing tumor immunogenicity or delaying the development of an aggressive tumor phenotype long enough for an effective immune response to be induced. The specific mechanisms of tumor growth inhibition by MAb67 in the Tubo.A5 model are currently under investigation.

The effects of anti-SEMA4D antibody treatment appear to be localized to the TME, as evidenced by changes in T-cell activity in TILs but not in the spleen, even where treatment with anti–CTLA-4 was shown to increase tumor-specific splenic T cells. This is important because, although SEMA4D is expressed on most immune cells throughout the body, our data suggest that the antibody acts locally within the tumor rather than generating systemic immunomodulatory effects, which could lead to toxicity. We have determined and report elsewhere that, while anti-SEMA4D antibody treatment can effectively saturate SEMA4D⁺ peripheral blood mononuclear cells, it does not significantly affect peripheral peptide-specific T-cell or antibody responses, nor interfere with viral clearance (49). This is in contrast with the genetic ablation of SEMA4D early in development, which has been reported to give rise to immunologic defects (30). A simple explanation for the localized effect of anti-SEMA4D may be that the SEMA4D gradient established in tumor is not replicated in most normal tissues. This warrants further investigation. The lack of systemic immune effects by SEMA4D blockade is consistent with a favorable safety profile, which is supported by results of 6-month high-dose nonclinical toxicology in rodents and monkeys and phase I clinical testing of a humanized anti-SEMA4D antibody, VX15/MAb2503 (50).

SEMA4D blockade represents an immunotherapeutic option, with a novel and independent mechanisms of action (Supplementary Fig. S8). Other immunotherapeutic agents, including those targeting CTLA-4 and PD-1, are proving to be very effective treatment options for some cancer, resulting in complete and durable responses. As shown here, anti-SEMA4D therapy can also be effective in some tumors that do not respond to immune checkpoint blockade, such as Tubo.A5, and, importantly, can combine synergistically with immunotherapies that either reduce negative feedback, such as checkpoint blockade, or enhance immune responses, such as costimulatory activity, vaccines, or chemotherapies that induce immunogenic cell death. Results of experiments reported here demonstrate the ability of anti-SEMA4D antibody to induce a significant and durable antitumor immune response and have motivated a phase I clinical trial of a humanized anti-SEMA4D antibody, VX15/2503, in patients with solid tumors. Results of this clinical study will be reported elsewhere.

M. Zauderer has ownership interest (including patents) in Vaccinex, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E.E. Evans, A.S. Jonason Jr, T.L. Fisher, E.S. Smith, M. Zauderer

Development of methodology: E.E. Evans, A.S. Jonason Jr, H. Bussler, J. Veeraraghavan, M.A. Doherty, L.A. Winter, C. Mallow, S. Giralico, K. Klimatcheva, T.L. Fisher, M. Paris

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.E. Evans, A.S. Jonason Jr, H. Bussler, S. Torno, J. Veeraraghavan, C. Reilly, M.A. Doherty, J. Seils, L.A. Winter, S. Giralico, M. Scrivens, K. Klimatcheva, W.J. Bowers, M. Paris

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.E. Evans, A.S. Jonason Jr, J. Veeraraghavan, C. Reilly, M.A. Doherty, J. Seils, L.A. Winter, C. Mallow, S. Giralico, M. Scrivens, K. Klimatcheva, W.J. Bowers, M. Paris, E.S. Smith, M. Zauderer

Writing, review, and/or revision of the manuscript: E.E. Evans, A.S. Jonason Jr, C. Reilly, T.L. Fisher, W.J. Bowers, M. Paris, E.S. Smith, M. Zauderer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.E. Evans, A.S. Jonason Jr, H. Bussler, J. Veeraraghavan, C. Reilly, J. Seils, R. Kirk, A. Howell, K. Klimatcheva

Study supervision: E.E. Evans, A.S. Jonason Jr, H. Bussler, W.J. Bowers, E.S. Smith

This study was financially supported by Vaccinex, Inc.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.