Autocrine VEGF signalling on M2 macrophages regulates PD‐L1 expression for immunomodulation of T cells

Abstract M2‐polarized macrophages, on one hand, can promote tumour vascularization by producing proangiogenic factors, such as vascular endothelial growth factor (VEGF). On the other hand, the expression of VEGF receptors (VEGFR) in this cell lineage was also reported. Although the function of VEGF/VEGFR axis plays a pivotal role in macrophages infiltration and angiogenesis, however, there is still lack of the direct evidence to show the role of VEGF as an autocrine operating in M2 macrophages, particularly for immunomodulation. In our study, we surprisingly discovered that M2 macrophages polarized by baicalin can simultaneously express VEGF and its receptors. Taking advantage of this unique culture system, we were able to investigate the biological activity of M2 macrophages in response to the autocrine VEGF milieu. Our results showed that the expression of programmed death‐ligand 1 (PD‐L1) on M2 macrophages was significantly up‐regulated in autocrine VEGF milieu. Through the blockade of autocrine VEGF signalling, PD‐L1 expression on M2 macrophages was dramatically down‐regulated. Furthermore, transplantation of PD‐L1+ M2 macrophage stimulated by autocrine VEGF into allogeneic mice significantly suppressed host CD4+/CD8+ T cells in the peripheral blood and increased CD4+ CD25+ regulatory T cells in the bone marrow. In conclusion, our findings provide a novel biological basis to support the current successful strategy using combined VEGF/PD‐1 signalling blockade in cancer therapy.

IL-10 and glucocorticoids (M2c). 6 M2d macrophages are the other subtype of M2 macrophages induced by costimulation with TLR and adenosine A2A receptor agonists. 7 Besides, M2d macrophages treated with the combination of lipopolysaccharide (LPS) and 5′-N-ethylcarboxamidoadenosine expressed high levels of vascular endothelial growth factor (VEGF), IL-10, and iNOS, low levels of TNF-α and IL-12, and mildly elevated levels of arginase-1. 8 Previous study investigated that additional VEGF-induced angiopoietin-2 (ANGPT2) can up-regulate programmed death-ligand 1 (PD-L1) expression on M2polarized macrophages derived from human peripheral blood mononuclear cells (PBMC). 9 Another study showed that a fraction of monocytes/macrophages in the peri-tumoural stroma, while expressing surface PD-L1 molecules, effectively suppressed tumour-specific T cell immunity and contributed to the tumours progression in vivo; the effect could also be reversed by blocking PD-L1 on those monocytes. 10 To the best of the current findings, whether VEGF not only acts as a paracrine, but also an autocrine for M2 macrophage to regulate PD-L1 expression and T cell immunomodulation remains unknown.
Recently, the large-scale randomized clinical trials demonstrated excitingly that anti-VEGF enhanced the efficacy of PD-L1 blockade in cancer therapy. For instance, simultaneous blockade of programmed cell death protein 1 (PD-1) and VEGF receptor 2 (VEGFR2) dramatically decreases the tumour size and tumour neovascularization in vivo. 11 PD-1 blockade combined with ANGPT2 and VEGF-A blockade improves the anti-tumoural activity by decreasing the mean tumour volume and weight. 12 Furthermore, in the experimental mice, the combination of anti-angiogenic/anti-PD-L1 therapy impaired tumour regrowth and resulted in a low tumour burden. 13 Although a possible affiliation of VEGF and PD-L1 was implied by its synergistic effect in tumour progression, the biological interpretations were restricted, mainly addressing on the independent role of VEGF and PD-L1 in angiogenesis and T cell suppression, respectively.
In our present study, we surprisingly found that baicalin, one of flavone glucoside extracted from the roots of Scutellaria baicalensis, not only can polarize macrophages toward VEGF-secreting M2d macrophages, but also promote their expression of VEGF receptor simultaneously. We therefore were capable of creating an isolated milieu to investigate the biological activity of M2 macrophages in response to autocrine VEGF, especially for immunomodulation.

| Mice
Six-to 10-week-old adult female C57BL/6J mice were purchase from National Laboratory Animal Center (Taipei, Taiwan) and housed in a clean conventional animal facility at 22°C with 12-h light/dark cycle.    Table 1.

| Gene expression
Ensured all reaction components properly thawed and mixed.

| Enzyme-linked immunosorbent assay (ELISA)
Cell-free supernatants were collected and stored at −20°C until

PD-L1 + M2 macrophages in vivo
Nine to 12 weeks old female C57BL/6 mice were divided into two groups. For PD-L1 hi group, mice were injected through retro-orbital plexus with 1 × 10 6 /100 μL RAW 264.7-derived M2 macrophages polarized by 50 μmol/L baicalin (48 hours) on day 0, day 7, and day 14, respectively. In PD-L1 lo/− group, RAW cells without stimulation by baicalin were injected via the same way. The mice were killed on day 19 after transplantation and the mononuclear cells from spleen, peripheral blood, and bone marrow were collected for flow cytometry analysis of T cells composition.

| Collection of mononuclear cells from spleen, peripheral blood, and bone marrow in mice
The whole spleen was squeezed by glass grinder and the dissociated cells were filtered through a 40 μm cell strainer (BD Bioscience). 50 μL heparinized blood was lysed with 3 mL RBC lysis buffer, shaking at 37°C, 250 rpm for 15 minutes to remove red blood cells. Total bone marrow cells were collected from femur and tibia of mice by flushing with 1 mL PBS twice through 25G syringe.

| Statistical analysis
All results were collected from at least three independent experiments and data were presented as mean ± SEM. Statistical significance of pairwise differences among three or more groups were determined using one-way analysis of variance (ANOVA) followed by LSD test. Statistical analysis was performed with SPSS for Windows T A B L E 1 Primer sequence

Primer
Sequencing References

| M2 macrophages polarization by baicalin
To access the polarization of M2 macrophages by baicalin, we used RAW 264.7 macrophage cell line. We further characterized the phenotype of macrophages induced by baicalin vs LPS. The results showed the significant differences in morphology between M1 and M2 macrophages polarized by LPS and baicalin, respectively, after 24 hours ( Figure 1A). Macrophages polarized by baicalin showed atypical round shape and larger size than that induced by LPS. Gene expression profiling showed that M1 macrophage phenotypes were significantly downregulated ( Figure 1B), and M2 macrophage phenotypes were correspondingly up-regulated after the treatment of baicalin ( Figure 1C).
Interestingly, M1 macrophages polarized by LPS can easily be switched to M2 macrophages following baicalin induction ( Figure 1C). Our results were in parallel to the previous study. 14 Meanwhile, FACS analysis of CD206, a typical M2 macrophages marker, was not expressed dominantly on M2 macrophages polarized by baicalin compared with that polarized by IL-4 ( Figure 1D). Thus, according to CD206 and CD86 expression, it was so firmly to explain that M2 macrophages polarized by baicalin were not belonging to M2b character. 7 its receptors VEGFR1 and VEGFR2 were found ( Figure 2E,F). We also confirmed the similar results in bone marrow derived M2 macrophages polarized by baicalin ( Figure S2).

| Autocrine VEGF milieu of M2 macrophages
By the simultaneous expression of VEGF-A and its receptors,  Figure 3B). Therefore, the effect of axitinib on untreated control cells seems not be curious to be explored.

| Autocrine VEGF signalling up-regulated PD-L1 expression in M2 macrophages
Owing to the prior study that additional VEGF- Furthermore, PD-L1 surface protein expression was also up-regulated on bone marrow derived CD11b + M2 macrophages in baicalininduced VEGF culture milieu ( Figure S2).

| Allostimulatory activity of PD-L1 + M2 macrophages in vivo
According to Sakhno et al's study, B7-H1 + (PD-L1)-M2 macrophages prohibited allogeneic T cell proliferating activity in mix lymphocyte culture that was associated with the higher numbers of apoptotic T cells through PD-1/PD-L1 pathway. 16 To verify in vivo functionality of PD-L1 + M2 macrophages stimulated by autocrine VEGF, we focused on their immunomodulatory capacity in allogeneic mice model. Flow cytometry analysis firmly showed that by transplantation of~84% vs 4% PD-L1 + M2 macrophages, the cell frequency of host CD8 + and CD4 + T cells in peripheral blood were significantly reduced 19 days post-transplantation ( Figure 5A,B). Interestingly, a group of CD4 + CD25 + regulatory T cells was significantly increased in bone marrow after transplantation ( Figure 5C). According to previous literature, 17 it is worth noting that cell frequency of CD8 + and CD4 + cells in the peripheral blood of normal mice were 7%-10% and 8%-12%, respectively. Our data in parallel showed that CD3 + total T cell was less than 26.6% in the peripheral blood of mice before transplantation ( Figure 5D). However, the cell frequency of CD8 + and CD4 + T cells in the peripheral blood of host mice in PD-L1 lo/− group were slightly higher (~16% and 19.7%, respectively) than normal. This result indicated that a gentle graft vs host (GVH) immune-response by transplantation of total 1 × 10 6 RAW 264.7 cells (Balb/c) into C57BL/6J mice was possibly provoked. In contrast with PD-L1 lo/− group, the reactivity of GVH was significantly alleviated in PD-L1 hi group (CD8, 12%; CD4, 17%). Therefore, the regulation of T cells functions upon the interaction with PD-L1 + M2 macrophages stimulated by autocrine VEGF was truly confirmed.

| DISCUSSIONS
According to our results, we attempt to orchestrate a triangle relationship among M2 macrophages, autocrine VEGF/VEGFR and PD-L1 expression for their role in immunomodulation (Figure 6). At present, the evidence were merely restricted by their bilateral relationship, respectively; and in most of the circumstances, VEGF was identified to play a role as a paracrine factor or stimulus for M2 macrophages' development. In order to strengthen our finding that

| Corelation of VEGF and PD-L1
There are few but important direct evidence to support the co-relation between VEGF and PD-L1. The recent study on clear cell renal cell carcinoma showing PD-L1 expression by immunohistochemistry staining was associated with VEGF expression that makes adverse pathological features in patients. 18 Another study emphasized that added ANGPT2 promoted PD-L1 expression on CSF1, IL-10, and IL-4 activated M2 macrophages. 9 Although these findings strongly suggested that tumour cells or M2 macrophages were capable of receiving VEGF signalling for PD-L1 up-regulation. However, the results were unable to elucidate whether M2 macrophages can deliver VEGF autonomously for their regulation of PD-L1 expression. In contrast, our study successfully indicated that M2 macrophages were proportions of IFNɤ + T and NK cells. 12 This was also in parallel with another study showing that blockade of PD-1 and VEGFR2 decreased the tumour size and tumour neovascularization in rodent. 11 Furthermore, PD-L1 was up-regulated in mouse tumours relapsing from antiangiogenic therapy. 13

| Corelation of M2 macrophages and PD-L1
Tumour-associated macrophages (TAM) of the M2 phenotype expressing PD-L1 are also one of the most inspiring topics that has been discussing among researchers. A previous study showed that PD-L1 + monocytes were accumulated in the peritumoural stroma area of cancers and increased with tumour progression. These activated PD-L1 + monocytes suppressed tumour-specific T cell proliferation, cytokine production, and cytotoxic potential in vitro and also fostered tumour growth in NOD/SCID mice bearing human tumours. 10 Corresponding to TAM study showed that the tumourconditioned media strongly induced PD-L1 expression on bone marrow-derived monocytes mediated by TNF-α. 21 Besides that, a recent study showed that CD14 + CD68 hi CD163 hi intratumoural monocyte/ on CD8 + T cells. 23 For cancer cells that express PD-L1 to affect macrophage activity, Gordon and the team found that PD-1 expression on TAMs correlates with phagocytosis inhibition and total in vivo phagocytosis levels. 24 The other study has demonstrated that CD163 + M2-like macrophage infiltration is highly associated with PD-L1 expression in gastric adenocarcinomal cells. 25

| Corelation of M2 macrophages and VEGF
The previous study proved that VEGF can polarize THP-1-derived macrophages toward the M2 phenotype and enhanced macrophage migration. 26 The similar result highlighting a novel function of both recombinant VEGF-C protein and tumoural VEGF-C could efficiently enhance migration of murine macrophages RAW 264.7 cell.
Tumoural VEGF-C also acted paracrinely to induce macrophage recruitment, and resultantly promoted clinical nonsmall cell lung cancer cell metastasis. 27 Moreover, a decrease in TAM (CD45 + , CD11b + , F4/80 + ) was observed upon axitinib treatment in both subcutaneous MC38 and LLC1 tumour cells. 28 On the other side, a study identified that baicalin can increase VEGF expression through the activation of the ERRα pathway in U251 human glioma cells and implicated the participation of macrophages in angiogenesis. 29 Again, TGF-β1 promotes VEGF secretion in bone marrow derived macrophages and in oral squamous cell carcinoma TAM was reported. 30 We address the function of baicalin for initiating M2 macrophages polarization and autocrine VEGF operation. The previous study showing that human PBMC treated with 1 mM baicalin can significantly enhance the IFN-γ secretion in the cell culture supernatants. 31 Additionally, PBMC from ulcerative colitis patients intervened by baicalin were obviously elevated the levels of IL-10. 32 In our study, we also found an increased level of IL-10 expression in RAW 264.7 and in bone marrow derived M2 macrophages (data not shown). Besides, the previous study suggested that increased levels of TNF-α, IL-1β, and IL-6 was reversed by baicalin treatment in damaged colon tissues. 33 Upon on those findings, we hypothesize that autocrine VEGF operation by transient activation of monocytes/ macrophages highly depends on the soluble factors, including IFN-γ, IL-10 and the proportion of proinflammatory cytokines in the milieu.
In combination with our finding and the evidence demonstrated that elevated PD-L1 expression on TAM can be inhibited by anti-IL-10 antibody, 34 therefore, the PD-L1 expression on TAM regulated by autocrine VEGF potentially existed.
Hereby, our findings enhance overall understanding of autocrine VEGF that participates in the regulation of PD-L1 expression on M2 macrophages for immunomodulation. We not only construct the actual trilateral relations among M2 macrophages, autocrine VEGF and PD-L1 expression for their role of immunomodulation, but also support a theory in biological aspect to explain how anti-VEGF signalling turns to be a main character to enhance the effect of PD-1 blockade in cancer therapy. Additionally, it highlights the value for the development of inhibitors targeting hypoxia-inducible factor 1 and 2 in TAMs. 35 Further investigation is needed to examine whether PD-L1 + M2 macrophages stimulated by autocrine VEGF can be potent for immunotherapy of autoimmune diseases.

ACKNOWLEDG EMENTS
This study was supported by the Ministry of Science and Technology in Taiwan (grant MOST 106-2314-B-020-001, MOST 104-2314-B-020-MY3, and MOST 106-2731-M-020-001). We thank for assistance from Core Facilities-Flow Cytometry Unit at National Pingtung University of Science and Technology in Taiwan. We also thank Mr.
Renanda Baghaz Dzulhamdhani Surya Putra for the technique support on cell culture.

CONFLI CT OF INTEREST
The authors have no conflicts of interests to declare.