The role of vascular endothelial growth factor in the hypoxic and immunosuppressive tumor microenvironment: perspectives for therapeutic implications

VEGF plays an important role in vascular development and in diseases involving abnormal growth of blood vessels. Since discovery of its dual roles in endothelial proliferation and vascular permeability [18, 19], VEGF has been considered a key mediator of neovascularization in tumors (Fig. 1). Elevated VEGF levels are associated with poor clinical outcomes in numerous tumors including GBs [2, 20]. In addition to angiogenic effects, VEGF suppresses the anti-tumor immune response [15]. VEGF inhibits maturation of dendritic cells (DCs), resulting in inactivation of CTLs [21]. VEGF also induces an immunosuppressive TME by strongly inducing Tregs, TAMs, and MDSCs [22]. Furthermore, VEGF enhances the expression of PD-1 on CD8+ CTLs and Tregs in a VEGFR2-dependent manner (Fig. 1) [23]. Tumor-derived VEGF, interleukin (IL)-10, and prostaglandin E3 cooperatively induced Fas ligand expression in endothelial cells, leading to exhaustion of CTL but not Tregs (Fig. 1) [24].

VEGF is induced by hypoxia via activation of the transcription factor, HIF-1, which could play an important role in triggering tumor angiogenesis [25, 26]. The TME is mainly altered by hypoxia and acidosis [27]. Hypoxia supports the escape of tumor cells from immune surveillance by recruiting TAMs, Tregs, and MDSCs into the TME directly or through upregulation of VEGF [28].

Macrophages often behave as immunosuppressive cells and contribute to inflammatory diseases. Macrophages express different functional programs in response to microenvironmental signals, defined as M1/M2 polarization [29]. M2 macrophages produce growth factors and anti-inflammatory cytokines to suppress the host immune response, resulting in tumor progression [29].

TAMs typically behave as M2 macrophages [30], playing a pivotal role in the TME [28, 30]. The interaction between the tumor cells and TAMs is promoted via macrophage colony-stimulating factor and its receptor [29]. TAMs induce various growth factors such as basic fibroblast growth factor, epidermal growth factor, hepatocyte growth factor, and platelet-derived growth factor [29]. In addition, the lack of arginine by arginase I, and IL-10, transforming growth factor (TGF)-β, and prostaglandin F2 produced by TAMs suppress effector T cells [29]. TAMs produce VEGF and matrix metalloproteinase 9, which promote angiogenesis, invasion, and metastasis [29]. Although the existence of TAMs is still controversial as a prognostic biomarker for cancer patients, most reports have demonstrated that increased TAMs are related to poor prognosis in GBs; breast, esophagus, and liver cancers; and malignant lymphoma [31]. Interestingly, TAMs infiltration is associated with the resistance to anti-angiogenic therapy through downregulation of macrophage migration inhibitory factor in GBs [32]. VEGF in the hypoxic TME is a key factor for transitioning from the M1 to M2 macrophage phenotype [33]. Furthermore, HIF-1α promotes the migration and differentiation of TAMs from immature myeloid cells via VEGF exposure [17].

Tregs (CD4 + CD25 + Foxp3+) play an active and significant role in the progression of tumors, and play an important role in suppressing tumor-specific immunity [34, 35]. Tregs suppress T cell-mediated immune responses via TGF-β, IL-10, and IL-35. Tregs induce the apoptosis of effector T cells through granzyme B. Consumption of IL-2 by Tregs inhibits effector T cells proliferation. Furthermore, Tregs directly suppress DCs [35]. Increased Tregs in tumor tissue or peripheral blood are significantly associated with shorter survival in most cancer patients including those with GBs [36]. Tregs may be critical for evaluation of the clinical significance of immunosuppressive effects after anti-angiogenic therapy. To detect Tregs more precisely, effector Tregs are defined as CD45RA-Foxp3^highCD4+ which may reflect an immune suppressive function. Non Tregs (CD45RA-Foxp3^lowCD4+), which produce inflammatory cytokines such as interferon-γ, are included in the CD4 + CD25 + Foxp3+ cell population [36].

The role of HIF-1α has also been implicated in direct regulation of the differentiation of Tregs and in promotion of the recruitment of Tregs to the TME via overexpression of CC chemokine ligand 22 and 28 [37]. HIF-1α induces VEGFR2-expressing Tregs through VEGF production [37].

MDSCs are a type of myeloid cells that can differentiate into macrophages, DCs and granulocytes, thus suppressing CTLs. In humans, MDSCs are CD11b + CD33 + Cd14 + HLA-DR- [38]. CD115, CD124, and VEGFR were also identified in MDSCs [39]. MDSCs are induced by IL-6, VEGF, and prostaglandin E2, which are produced by tumor cells. In addition, MDSCs are activated by interferon-γ, IL-4, IL-13, and TGF-β produced by CTLs and tumor stromal cells [38]. MDSCs also produce TGF-β, IL-10, and metalloproteinase 9 [40]. Interestingly, recent findings suggest that natural killer T cells activated by α-GalCer-loaded CD11b + Gr1 + MDSC can acquire the ability to convert immunosuppressive MDSCs into immunity-promoting antigen-presenting cells. Reprogramed MDSCs show upregulation of expression of CD11b, CD11c, and CD86, which support immunity by antigen-specific CTLs without increasing Tregs [22, 41]. GB patients with a more favorable prognosis exhibit decreased MDSCs and increased DCs compared to those with a worse prognosis [42]. Interactions between MDSCs and glioma stem cells via migration inhibitory factor enhances the function of MDSCs, which could be targeted to reduce the growth of GBs [42]. In addition, MDSCs directly promotes angiogenesis, which is associated with refractoriness to anti-angiogenic therapy [43]. VEGF is strongly associated with MDSC accumulation in GBs [44, 45]. HIF-1α has also been implicated in direct regulation of the function and differentiation of MDSCs in the hypoxic TME [46].

PD-1 is expressed on CD8+T cells and Tregs. PD-L1 is expressed on tumor cells in numerous malignant tumors including GBs and binds to PD-1 to negatively regulate the immune response of CD8+T cells [6, 7]. A recent study showed that PD-1 is expressed on TAMs and correlates negatively with phagocytic potency, demonstrating its relevance to tumor immunity [47]. PD-L1 protein expression was identified in 61 to 88% of patients with GBs [48]. PD-1/PD-L1 expression is associated with poor prognosis for patients with GBs [49, 50]. Garber et al. demonstrated that GB specimens have a higher frequency of PD-1+ tumor-infiltrating lymphocytes (TILs) compared with lower-grade gliomas, whereas PD-L1 expression does not significantly differ among malignant grades [51]. Isocitrate dehydrogenase (IDH)-wild-typed GBs are more “immunologically active” than IDH-mutated GBs. GBs with wild-type IDH display a higher number of TILs and elevated expression of PD-L1 compared with IDH-mutant GBs. Therefore, IDH-wild-type GBs may be more readily targeted by PD-1/PD-L1 checkpoint blockade [52]. In particular, PD-L1 is strongly expressed in the mesenchymal subgroup of GBs [53].

A previous study demonstrated that PD-1 expression on CD8+T cells and Tregs is induced by VEGF [23]. Similarly, PD-L1 is upregulated via VEGF exposure in some types of tumors including GBs [53]. Immunologically “cold” tumors including GBs are not good targets for immune checkpoint inhibitors [16]. Anti-angiogenic therapy has the possibility to change “cold” tumors into “hot” tumors with a favorable microenvironment. Therefore, combination therapy with an immune checkpoint inhibitor and VEGF-targeted therapy is expected to be an efficient treatment strategy [17]. In addition, hypoxia directly causes upregulation of PD-L1 and CTLA-4 on MDSCs, TAMs, DCs, and tumor cells through HIF-1α [54]. HIF-2α is also associated with PD-L1 expression on tumor cells in metastatic renal cell carcinoma [55]. Blocking HIF-1α by agonizing nitric oxide signaling decreased PD-L1 expression and promotes CTL-mediated lysis [56].

VEGF promotes the accumulation of TAMs, Tregs, and MDSCs in tumor tissue and secondary lymphoid organs [10]. In addition, HIF-1α activation is also a major component of the hypoxic TME. HIF-1α upregulation directly drives recruitment of immature myeloid cells and Tregs, and fosters their phenotypic conversion into highly suppressive MDSCs and TAMs [58]. Recent studies demonstrated functional cross talk among TAMs, Tregs, and MDSCs that was strongly associated with hypoxia-induced VEGF production [43, 57, 59]. Tregs modify the phenotype of TAMs to express inhibitory B7-H molecules. Tregs depletion significantly downregulates the expression of immune suppressive molecules such as B7-H1 on MDSCs and TAMs, and also reduces tumor growth. In addition, Tregs produce IL-10, IL-4, and IL-13, and induce monocyte differentiation toward TAMs [43, 57, 59]. MDSCs are reported to differentiate into TAMs in the hypoxic microenvironment, which is regulated by STAT3 activity [60]. Since these immunosuppressive cells and immune checkpoint molecules show cross talk in the hypoxic TME, anti-VEGF/VEGFR therapy can induce tumor oxygenation [61], resulting in an immune-supportive TME.

Alteration of immunosuppressive cells and immune checkpoint molecules has been investigated utilizing clinical samples before and after chemotherapy [36, 49, 62‐103], but results have been inconsistent (Tables 1 and 2) [36, 49, 62‐103]. Metronomic cyclophosphamide decreases Tregs [62, 66‐68, 83], and gemcitabine decreases both Tregs and MDSCs [35, 65, 103]. Tregs and MDSCs are not increased when chemotherapeutic agents are co-administered with immunomodulatory agents such as vaccination and IL-2 (Table 1) [49, 63, 75, 78]. However, most chemotherapeutic agents including temozolomide tend to increase Tregs and MDSCs [71, 72, 74, 79, 84]. In addition, chemotherapy also enhances PD-1/PD-L1 via TGF-β induced epithelial-mesenchymal transition [69].

In contrast, combinational chemotherapy with VEGF/VEGFR-targeted agents could make the TME immune supportive [23, 96]. VEGF/VEGFR-targeted therapy such as Bev, sorafenib, and sunitinib reduces the population of Tregs in the peripheral blood and tumor tissue in GBs, metastatic colorectal cancer, hepatocellular carcinoma, and renal cell carcinoma (Table 2) [86‐103]. However, some studies using paired peripheral blood before and after VEGFR-targeted therapy demonstrated that Tregs and PD-1 were significantly increased [104]. The effect of anti-angiogenic therapy has been highly controversial regarding MDSCs [91, 105]. Serum VEGF-A levels are correlated with the population of MDSCs, and Bev may decrease MDSCs [90, 94, 95]. Sunitinib and sorafenib may also decrease the population of MDSCs and recover the Th1 reaction in the patients of hepatocellular carcinoma and metastatic kidney cancer [93, 105]. However, the population of MDSCs does not change following Bev and axitinib, despite decreasing in serum VEGF of patients with kidney cancer [106]. The status of MDSCs following anti-VEGF/VEGFR therapy is inconsistent [88, 90, 93‐95, 97, 98, 104].

Furthermore, in the recurrent stage after anti-angiogenic therapy, the status of immunosuppressive cells and immune checkpoint molecules is also highly controversial (Table 2). VEGF-targeted therapy can lead to either tumor oxygenation or tumor hypoxia [14, 61, 107]. Therefore, both an immune-supportive TME in normoxic conditions and a hypoxia-induced immunosuppressive TME have been reported after VEGF-targeted therapy [14, 61, 107]. Tregs in the peripheral blood are increased in patients with recurrent GBs after development of resistance to VEGFR inhibitors [89]. An increased level of PD-1 expression on CD4 and CD8 T cells was reported in patients with GBs or metastatic renal cell carcinoma that is refractory to VEGFR-targeted therapy [89, 109]. Previous studies have demonstrated changes in the TME using tumor specimens resected under and after Bev therapy. Bev downregulates the expression of PD-1 and PD-L1 immune checkpoint molecules, suppresses the infiltration of TAMs and Tregs, and increases CTL infiltration. Importantly, the conditions are sustained during long-term Bev usage [14]. VEGF persistently contributes to tumor growth, even if secondary signaling pathways are upregulated. These findings support the concepts of continuous usage as Bev beyond progression [110]. The reason for the discrepancy in the status of immunosuppressive cells and molecules at the recurrent stage among relevant studies remains unclear. The reason may be the difference between peripheral blood and tumors, the difference in the target of inhibition, or the response rate of targeted therapies. Tregs, PD-L1, and TAMs are upregulated in patients showing partial response [101]. Tada et al. suggested that analyses of TILs and immune cells using tumor specimens are more important than analyses of peripheral blood for investigation of cancer immunology [36]. Further investigation using tumor samples as well as peripheral blood may be required for monitoring immunosuppressive cells and molecules and seeking for predictable and prognostic biomarkers.

VEGF plays a key role in the development of the immunosuppressive TME by inhibition of DC maturation and enhancement of immunosuppressive cells and molecules [21, 23]. Immunologically “cold” tumors are unresponsive to immunotherapies including immune checkpoint inhibitors. Anti-angiogenic therapy may change the TME into an immunological favorable “hot” microenvironment. Bev suppresses immunosuppressive cells including TAMs, Tregs, and MDSCs, and improves the migratory capacity of CTLs [14]. Theoretically, Bev could enhance the effect of PD-1/PD-L1 inhibitors (Fig. 2). Therefore, many basic research studies have demonstrated that combination therapy of VEGF/VEGFR inhibitors and PD-1/PD-L1 inhibitors induces a synergistic effect on several types of tumors including GBs, melanoma, lung cancer, and hepatocellular carcinoma [92, 96, 111]. Patients with metastatic renal cell carcinoma show improvement in antigen-specific T cell migration after combination usage of anti-VEGF and anti-PD-L1 antibodies [111]. Currently, clinical trials testing theses combination therapies are being conducted for several types of tumors such as recurrent GBs, renal cell carcinoma, colorectal cancer, and ovarian cancer (NCT03024437, NCT02659384, NCT02873962, NCT02017717) [16]. This treatment strategy may lead to promising results for those malignant refractory tumors. However, high dose and long-term usage of anti-VEGF/VEGFR therapy is associated with hypoxia which is one mechanism of resistance to this combination therapy [108]. Further investigations are warranted to determine the adequate dosage and duration of anti-angiogenic agents for the combination usage with immune checkpoint inhibitors. Because immune checkpoint inhibitors and anti-VEGF/VEGFR treatment are immunological “break off” strategies, other immunotherapies such as DCs-based immunotherapy [112], tumor vaccine therapy [113], and chimeric antigen receptor T-cell therapy [114] should be added to them as “acceleration on” strategies to improve therapeutic efficacy. Furthermore, hypoxia-targeted therapy is another treatment strategy to overcome the mechanisms of resistance to immunotherapy, via suppressing Tregs, TAMs, and MDSCs [115].

As described above, reprogramming the TME to an immune-supportive microenvironment improves cancer immunotherapy [36, 49, 62‐103]. To conquer therapeutic refractoriness to immunotherapy, the cross talk of immunosuppressive cells and molecules in the TME must be comprehensively regulated.