Background
Targeting angiogenesis has been and continues to be an attractive therapeutic modality in both newly diagnosed and recurrent glioblastoma (GBM) patients. Vascular endothelial growth factor (VEGF) is the most abundant and important mediator of angiogenesis in GBM [
1]. Anti-angiogenesis with Bevacizumab (Bev), an anti-VEGF monoclonal antibody [
2], has been used for devascularization to limit the growth of malignant glioma. However, Bev antibody failed to prolong overall survival despite the extension of progression-free survival in three randomized phase III trials [
3‐
5]. These transient benefits are due to elusive mechanisms underlying resistance to the anti-angiogenic therapy [
6]. The role of VEGF blocking agents and how to best incorporate them into the treatment paradigm for GBM should evolve if the understanding of their effects and that of subsequent resistance mechanisms improves.
Cancer is closely associated with inflammation. The GBM tumor microenvironment contains innate immune cells in addition to the cancer cells and their surrounding stroma [
7]
. Microglia/macrophages were assumed as one reason for the poor beneficial effect of anti-angiogenic therapy. However, if literature evidences the effects of VEGF on GBM [
8], the underlying mechanisms and their impact on microglia/macrophages are not clarified sufficiently and some data are contradictory. VEGF is able to mobilize blood monocytes and microglia cell lines in vitro [
9,
10], and microglia/macrophages themselves produce VEGF [
11,
12]. Some studies report that anti-angiogenic therapy led to an increase in the amount of microglia/macrophages that conduce to resistance development [
13‐
15]; however, this increase is not documented in terms of kinetics or quantitative data on cell subsets.
In an earlier study [
16], we developed an orthotopic GBM model by grafting U87 in nude mice and recapitulating the biophysical constraints normally governing tumor invasion. This model suitable for intravital multiphoton microscopy allowed to repeatedly imaged tumor cells and blood vessels during GBM development in control and Bev treated mice. The treatment massively reduced tumoral microvessel densities but only transiently reduced tumor growth rate [
17]. Altogether our results supported the view that GBM growth is not directly related to blood supply but, as proposed by others [
18], that tumor angiogenesis and tumor growth could be promoted by inflammation.
In the brain, differential contributions of infiltrating versus resident myeloid populations have been demonstrated in the pathogenesis of GBM. In order to gain insight in the respective involvement of resident microglia and circulating leucocytes across the different stages of tumor development, we devised a clinically relevant syngenic GBM model suitable for intravital dynamic multiphoton imaging by grafting the murine DsRed-GL261 cell line in C57BL/6 multicolor Thy1-CFP//LysM-EGFP//CD11c-EYFP fluorescent reporter mice [
19]. In these animals, CFP expression occurs in subpopulations of neurons; EGFP in peripheral myelomonocytic cells including neutrophils, infiltrating monocytes and their progeny; and EYFP in a subset of microglia. They are particularly appropriate for long-term tracking of different types of immune cells in vivo. We showed that invasion of the tumor by microglial CD11c-EYFP
+ cells dominated early stages of tumor development, then followed by a massive recruitment of circulating LysM-EGFP
+ cells.
In the present study, we used the above mouse GBM model to assess, by in vivo two-photon imaging combined to immunochemistry and multiparametric cytometry (FACS), how Bev therapy influenced the inflammatory landscape at two critical times of tumor development and to evaluate whether it reprograms the tumor immune microenvironment. Besides uncovering some specific features of the spatio-temporal distribution of recruited subsets of immune cells, our findings strongly support that VEGF blockade has an effect on blood vessels, levels of monocytes traveling in the blood vessels, and the density of myeloid recruited cells. Importantly, Bev modifies the ratios between subsets of DCs and the number of MHCII expressing cells thus possibly the way in which innate response controls the adaptive response.
Discussion
VEGF is an important factor in tumor vascularization and used as target for anti-angiogenic treatment strategies in GBM. Based on promising phase II trials, Bev, an anti-VEGF monoclonal antibody, was approved by the Food and Drug Administration as monotherapy for GBM. However, three randomized phase III trials showed that Bev, despite extending progression-free survival failed to prolong patients’ overall survival [
6]. The underlying mechanisms of acquired resistance to anti-angiogenic therapy remain obscure. In part, this is a result of the limited understanding of the effects of anti-VEGF therapy on the microenvironment of the tumor. Detailed information on how Bev affects GBM is important not only to understand the success or failure of such treatment, but also to provide educated advice on how combination therapies should be optimized.
In a recent study using the GL261 GBM mouse model, Turkowski
et al. [
8] reported that VEGF is a modulator of the innate immune response with suppressive effects on the immunologic and pro-angiogenic function of microglia/macrophages. They showed that high level of VEGF expressed by engineered tumor cells led to threefold enlarged tumor volumes and a pronounced remodeling of the vascular structure along with a reduced infiltration of microglia/macrophages by approximately 50%. Here, we investigated the reverse situation by depleting VEGF. We used multicolor fluorescent reporter mice and in vivo spectral two-photon microscopy to gain insight on the effects of the depletion at two critical stages of tumor development in individual animals. These observations were combined with multiparametric cytometry and immunohistochemistry to link information on cell distribution with precise phenotypes and frequencies of innate cell subsets with a focus on DCs. In addition, to describe dynamic effects on cell subsets, our approaches uncovered so far undescribed spatio-temporal distributions of innate cell subpopulations in the tumor microenvironment and potential routes entering and/or exiting the brain parenchyma.
We started Bev treatment at day 14 after tumor grafting. As expected and already described [
17,
19,
25], Bev persistently normalized intra-tumoral vessels and transiently slowed down tumor growth rate but the tumor then escaped the treatment as shown by comparing progression in between day 21 and day 28. Nonetheless, the effect on tumor size of VEGF blockade is less drastic than that of VEGF over-expression [
8]. Recurrent two-photon imaging of GL261-DsRed grafted mice after injection of QDots 705 to highlight functional blood vessels also revealed a persistent treatment effect on vessel density and tortuosity especially inside the tumor. These observations confirmed observations we made in a U87-MG model [
17] that tumor growth can be sustained without an increase in blood vessel density suggesting that GBM growth is rather governed either by vessels and/or stromal properties.
Intravital imaging and time-lapse recordings of fluorescent cells (Additional files
1 and
2: Videos S1 and S2) revealed their high association to the tumor vasculature. Importantly, we observed that under Bev treatment, the number of LysM-EGFP
+ cells trafficking inside the vessels was significantly higher than in untreated mice. A possibility is that anti-VEGF treatment results in an increased blood flow that might occur during the tumor vessel normalization window [
31]. At both day 21 and day 28 LysM-EGFP
+ density in the tumor and its environment is more than twofold higher in Bev-treated mice than in controls. We cannot infer, however, that a change in the properties of tumor blood vessels and their normalization induced by Bev [
32] is directly related to this increase. Indeed, both in control and treated mice, we never observed extravasation of these cells from tumor blood vessels. Interestingly, imaging revealed a gradient of LysM-EGFP
+ cells from the meninges towards the tumor core. This supports the view that meningeal vessels are potential gateways for immune cells recruited into the brain [
33]. In addition, a recent publication [
34] points to a direct local interaction between the brain and the skull bone marrow through the meninges. The authors determined the origin of leukocytes that were recruited to cerebral inflamed tissues and found that skull bone marrow myeloid cells migrate towards the inflamed brain through microscopic channels that directly connect the skull marrow cavities with the dura. The spatial distribution revealed by brain clearing (Fig.
6b) would also support such an origin as we observed a gradient of fluorescent cells in the parenchyma, potentially recruited towards the tumor.
The FACS data (Fig.
2a and b) indicated that at an early stage (day 21) neutrophils outnumbered monocytes among LysM-EGFP
+ cells. At day 28, LysM-EGFP
+ cells are essentially monocytes at stage P1. All our data (FACS, imaging) converge to support a decrease in recruitment of neutrophils and the increase of the recruitment of monocytes at stage P1 as a consequence of VEGF blockade. Our previous observations [
19,
29] indicated that monocytes differentiate in situ in the pathologic CNS into either moDC (P2–P3) or macrophages (moMac P4–P5). MoMac are very scarce and we could not determine whether Bev significantly modifies their levels. During these dynamic differentiation processes monocytes start to express CD11c-EYFP and moDC are co-expressing the two fluorophores. In agreement with the increased recruitment of monocytes, and in support of this differentiation process, VEGF blockade increases the density of moDC with twice as many intra-tumoral double-labeled cells in Bev-treated compared to untreated mice (Fig.
4, right panel).
Taken together, our data indicated that in our model, LysM-EGFP
+ cells recruitment starts at a rather precise time during the development of the tumor (around days 20–22 after tumor grafting) with recruitment of neutrophils, which then stopped, concomitant or followed by recruitment of monocytes that continues thereafter. The progressive wavelike shift in the phenotypes of infiltrated monocytes corresponds to the refractoriness of the tumor to treatment. As already reported [
19], after day 21, microglia representation significantly decreases among the innate cell subsets and the EYFP
+ population. This dynamic change is in agreement with the view that resident microglia rather than peripheral leucocytes promote vascularization in brain tumors [
11]. Bev appears to further decrease the number of activated microglial cells in the brain both at an early stage (day 21) and later stage (day 28) of tumor development (Fig.
2). Whereas FACS analysis concerned the tumor and the surrounding brain parenchyma, the observation is further supported by immunolabeling of intra-tumoral CD11c-EYFP
+/TMEM119
+/Iba1
+ cells at day 28 (Fig.
5c, Additional file
3: Figure S3A).
In their study, Turkowski
et al. [
8] observed that VEGF over-expression resulted in a reduced expression of microglia/macrophages by approximately 50%, whereas several published data report an increase in the amount of these cells elicited by anti-angiogenic treatments. It should be pointed out that most of the published data are almost exclusively performed on the entire myeloid cell fraction because the lack of lineage markers prevented their classification into subsets. The authors infer that the increase in the amount of innate immune cells conduces to resistance development in tumors but pro- and antitumoral characteristics were not assigned to the evident subpopulations [
14,
15].
In an effort towards this goal, we deciphered precisely the key populations, i.e., DC subsets and controlling immune responses, that are affected by VEGF blockade. We report that the blockade elicited changes in the ratios of the different subsets present during the escape phase of the tumor. At day 28 among the EYFP+ subsets, besides the 6–7% of cells identified as activated microglia, we uncovered a large increase of moDC, a threefold decrease of cDC1 and a twofold increase of cDC2. A trend towards this modification of proportions between subsets was already detectable at day 21, i.e., after 7 days of Bev treatment.
Bev decreases cDC1 density, a subset necessary for the development of anti-tumor immunity. These cells excel at cross-presentation [
35], are migratory and the main interaction partners of antigen-specific CD8
+ T cells [
36] and have been reported to induce protective immunity in cancers. It is thus possible that their decrease induced by Bev participates to the escape of the tumor treatment by lowering the anti-tumor T cell adaptive response. The maturation of P1 monocytes towards moDC could differ as we observed more LysM-EGFP
+/CD11c-EYFP
+ double-labeled cells in treated tumors. moDC, also expressing MHCII, are increased in Bev-treated tumors. Notably, other tumor models with a high moDC content also harbor relatively more MHCII
+ cells, suggesting that the microenvironment in these tumors favors the differentiation of infiltrating monocytes towards these MHCII
+ DCs exhibiting immunosuppressive capacities [
37]. Notably, the use of moDC to promote protective immunity in patients suffering from infections or cancer have shown moDC limited efficacy, owing to their short life, poor migratory properties, and recirculation to lymph nodes.
Surprisingly, our observations on tissue sections and brain clearing revealed that the EYFP
+ population migrates in the parenchyma following specific routes such as the corpus callosum. There are two non-exclusive possibilities that would deserve further studies. One is that the population of EYFP fluorescent DCs, never seen in the blood vessels, enters the parenchyma
via the choroid plexuses [
38] and then follows this route to finally accumulate in the lower parts of the tumor (Fig.
5b). Another one is that migratory cDCs use this route to reach the choroid plexus then join the lymph nodes.
It is conceivable that this dynamic adjustment of DC subsets induced by Bev treatment we distinctively visualize in vivo reflects several modifications of physical and chemical communication between innate immune cells and tumor cells, favoring the escape of the tumor to VEGF blockade [
39]
.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.