Deletion of neuropilin 1 from microglia or bone marrow-derived macrophages slows glioma progression

Abstract

Glioma-associated microglia and macrophages (GAM) which infiltrate high-grade gilomas represent constitute a major cellular component of these lesions. GAM behavior is influenced by tumor-derived cytokines that suppress initial anti-tumorigenic properties, causing them to support tumor growth and to convert and suppress adaptive immune responses to the tumor. Mice which lack the transmembrane receptor neuropilin-1 (Nrp1), which modulates GAM immune polarization exhibit a decrease in glioma volumes and neoangiogenesis and an increase in anti-tumorigenic GAM infiltrate. Here we show that replacing the peripheral macrophage populations of wild type mice with Nrp1-depleted bone marrow-derived macrophages (BMDM) confers resistance to the development of glioma. This resistance occurred in a similar fashion seen in mice in which all macrophages lacked Nrp1 expression. Tumors had decreased volumes, decreased vascularity, increased CTL infiltrate, and Nrp1-depleted BMDM adopted a more anti-tumorigenic phenotype relative to wild type GAM within the tumors. Mice with Nrp1-deficient microglia and wild type peripheral macrophages showed resistance to glioma development and had higher microglial infiltrate than mice with wild type GAM. Our findings show how manipulating Nrp1 in either peripheral macrophages or microglia reprograms their phenotype and relieve their pathogenic roles in tumor neovascularization and immunosuppression.

Introduction

Gliomas are the most common primary malignancies of the central nervous system (CNS), making up a total of 80% of all malignant CNS tumors that are diagnosed in the USA [1, 2]. The NCI estimated that CNS malignancies constituted a disease burden of 23,800 cases with 16,700 deaths attributable to these diseases per year in 2017. The incidence is relatively similar world-wide with a marginally higher rate of diagnosis in men [3].

The standard of therapy for high-grade glioma (HGG) consists of resection, when possible, followed by fractionated radiation and chemotherapy with temozolomide (TMZ). This therapy and its dosing regimen were implemented in 2005 and have yet to be revised [4]. Median time of survival after GBM diagnosis, regardless of treatment strategy, averages between only 12 to 15 months [5]. Resection is frequently incomplete as HGGs are highly invasive and often located in vital brain regions. Patients incur frequent complications of the disease and its treatment, including seizures, neurological disorders, hydrocephalus, and the adverse effects of chemotherapy. There is a substantial need to identify more effective and specific targets for HGG. Immunomodulating therapies are becoming realistic strategies for the management of glioma since the innate and adaptive systems play crucial roles in cancer progression and patient survival [6].

Gliomas are comprised of various cell populations including tumor stem cells, stromal cells, blood vessels, infiltrating monocytic populations, and resident immune cells. Microglia are the resident macrophages of the CNS and are responsible for protecting it from infection and responding to injury. Microglia derive from yolk-sac progenitor cells and migrate to the CNS early on in development. After populating the CNS, the resident pool of microglia is replenished over time by a CNS-specific pool of stem cells, rather than via the bone marrow, as all other macrophages are [7]. In certain disease states it is believed that local microgliosis is the primary contributor to macrophage accumulation in the CNS [8]. However under certain circumstances BMDMs have been shown to infiltrate and differentiate into microglia in the CNS, leading to some ambiguity surrounding the origin of adult populations of these cells in the CNS [9, 10].

In HGG, bone marrow derived monocytes (BMDMs) are attracted to and migrate into the gliomas following the secretion of chemotactic factors such as colony stimulating factor 1 (CSF1) and CCL2 by tumor cells [11]. Once recruited to the glioma, BMDMs develop immunosuppressive phenotypes and secrete growth factors and cytokines, which help to nurture the growth and spread of the tumor [12]. Due to similarities in the genes expressed by microglia and macrophages, it is often difficult to differentiate whether microglia or BMDMs constitute the bulk of glioma associated macrophages (GAMs).

Previously, our group showed that GAMs are vital to the progression of HGG in a murine model of the disease as focal elimination of the cells from the tumors resulted in significantly lower disease burden [13]. Aside from the total eradication of GAMs from the tumor microenvironment, their activity can potentially be pharmacologically or genetically manipulated.

Neuropilin 1 (Nrp1) is a cell surface receptor expressed by various cells throughout the body including endothelial cells, subsets of T cells, subsets of DCs, subsets of myeloid-derived cells, and microglia [14–16]. Nrp1 can complex with other receptors, including vascular endothelial growth factor receptor 2 (VEGFR2), transforming growth factor β receptor I/II (TGFβRI/II), and hepatocyte growth factor receptor (cMET) and amplify signaling pathways associated with these receptors [17–19]. It has been shown that mice with Nrp1 ablation from LysM-Cre expressing microglia and macrophages are resistant to pathological neovascularization in retinal sclerosis [20]. These same mice with Nrp1-deficient macrophages were shown to have increased survival in orthotopic breast and pancreatic cancer models due to poor vascularization and increased infiltration of their tumors by anti-tumorigenic macrophages [21]. Zhang et al. reported that certain types of HGGs are populated by GAMs that are characterized by significantly elevated Nrp1 expression [22].

We showed that knocking out Nrp1 in Csf1R expressing microglia and macrophages caused them to adopt a more anti-tumorigenic phenotype and have reductions in the capacity to signal through immunosuppressive TGFβ-dependent SMAD2/3 pathway. Gliomas in mice with Nrp1-depleted GAMs (Nrp1^MgKO) had reduced tumor vascularity, slower disease progression, and significantly longer survival [23]. However, we were interested in whether the increased longevity of the animals was impacted by differences in adaptive immune responses to the tumors. Moreover, we were interested whether resident Nrp1-deficient microglia or peripherally-derived, Nrp1-deficient BMDMs were primarily responsible for the phenomenon we observed.

Here, we show that tumors in mice with Nrp1-deficient GAMs have significantly increased T cell infiltrate, relative to that in gliomas of normal mice. We then show that replacing the peripheral macrophages of wild type mice with Nrp1-depleted BMDMs confers resistance to the development of glioma in a similar fashion that is seen in Nrp1^MgKO mice. Tumors had decreased volumes, decreased vascularity, increased CTL infiltrate, and Nrp1^MgKO BMDMs adopted a more anti-tumorigenic phenotype relative to other macrophages populating the tumors. Finally, we show that mice with Nrp1-deficient microglia and wild type peripheral macrophages show resistance to glioma development in a similar fashion, but exhibit higher microglial infiltrate than mice which have wild type microglia and macrophages. These results support the notion that manipulation of Nrp1 in either peripheral macrophages or microglia allows them to adopt a more anti-tumorigenic, less pro-angiogenic phenotype that decreases neoangiogenesis and alters the adaptive immune response to the glioma.

Materials and Methods

Animals

Mice were bred under maximum isolation conditions with a 12:12 hour light: dark cycle with food ad libitum. Littermates from Nrp1^fl/fl × Nrp1^fl/fl Csf1R-cre breeding pairs were used for experiments [24, 25]. Mice expressing cre recombinase (Nrp1^MgKO) effectively lack Nrp1 in microglia and macrophages. Littermates lacking cre recombinase are considered ‘wild-type’. Csf1R-GFP mice express eGFP under the Csf1R promoter, but otherwise, are ‘wild type’ in phenotype [26]. All mice were bred on the C57BL/6 background.

Genotyping

Genotyping was performed as previously described [23]. Briefly, tissue was acquired from mice by tail snip. Crude DNA was extracted by boiling the tissue in 50mM NaOH for 10 minutes followed by pH neutralization via the addition of 1M Tris-HCl (pH 8.0). Presence of Csf1R-cre and/or the floxed Nrp1 alleles was confirmed using Taq PCR (New England Biotechnologies) with primers designed against their appropriate gene sequences [23]. All mice were genotyped prior to use.

Glioma-derived cell line

The GL261 cell line was previously purchased from the American Tissue Culture Collection and is derived from a chemically-induced C57BL/6 murine astroglioma [27]. This cell line was expanded in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum (FBS) and 1× penicillin/streptomycin (P/S). The cell line was passaged no more than 3 times prior to use for experiments.

Murine glioma model

Gliomas were established in mice as previously described [13]. Briefly, mice were anesthetized using atropine (0.6 mg/kg, i.p.) and 2.5% avertin (0.02 mg/g, i.p.). A midline incision was made on the scalp and a small burr hole was drilled in the skull at stereotactic coordinates of bregma, −1 mm anteroposterior and +2 mm mediolateral. 3 × 10⁴ GL261-eGFP cells suspended in 1μL of PBS were injected in anesthetized mice over two minutes at a depth of 3mm. After completion of the procedure, the incision was sutured and the mice were placed on a heated surface until fully recovered from anesthesia. Buprenorphine was administered to mice for pain management if signs of distress were observed. If mice were found to have lost more than 30% of their initial body weight over the disease course, they were euthanized. All animal procedures were approved by the Stony Brook University Institutional Animal Care and Use Committee (IACUC). Mice were approximately 3–4 months old if not receiving allografts and 5 months old if having received an allograft. Males and females were equally distributed across separate cohorts.

Bone Marrow Allografts

2 month old Nrp1^fl/fl, Nrp1^fl/fl Csf1R Cre, or Csf1R-GFP mice were used for these experiments. Csf1R-GFP mice are also on a C57/Bl6 background. One day prior to irradiation, the water of these mice was supplemented with trimethoprim-sulfamethoxazole (400mg/L) and mice were maintained on this water for the rest of the duration of experiments. Donor mice were killed by CO₂ asphyxiation and their femurs and tibiae were dissected and flushed with 1×HBSS supplemented with 1% FBS. Marrow was then resuspended, filtered, and kept on ice. Recipient mice were irradiated with lead head protection with a dose of 1100 rads to deplete their bone marrow over the course of 20 minutes. Immediately after, recipient mice were injected with 10–20 million donor marrow cells via tail vein injection. Donor-recipient pairs were performed in such a way that either recipient or donor marrow Csf1R-expressing cells would be GFP+ in order to differentiate between donor and recipient macrophages. Mice were allowed to recover for 12 weeks prior to use in further experiments. The spleens of mice were analyzed after mice were killed in order to determine the successful engraftment of donor marrow.

Imaging

Imaging was performed on either a Nikon Eclipse E600 Microscope (Nikon), a Zeiss LSM 510 confocal microscope (Zeiss), or a Leica SP8X confocal microscope (Leica).

Immunohistochemistry

Tissue fixation and staining was performed as previously described [28]. Briefly, for evaluation of tumor tissue at different time points, mice were deeply anesthetized with 2.5% avertin i.p. and transcardially perfused with 50 mL 1× PBS followed by 50 mL 4% paraformaldehyde (PFA) in 1× PBS. Brains were removed and post-fixed in 4% PFA/PBS for 12 hours followed by 30%w/v sucrose in PBS at 4°C. The tissue was then embedded in optimal cutting temperature compound (Tissue-Tek), and sectioned using a Leica cryostat (Nussloch, Germany). 50-μm thick coronal sections were taken throughout all tumor containing tissue and suspended in 1×PBS at 4°C for analysis. Additional 20-μm thick coronal sections were collected onto slides for further analysis.

For evaluation of tumor sizes, three serial sections directly beneath the injection track were selected from each animal and total tumoral area was measured from each image and averaged using NIS-Elements software (Nikon Instruments Inc.). As tumors were spherical in shape, an average radius was calculated from these images and the average volume of tumors was calculated using the formula 4/3πr³.

For evaluation of tumor vascularity, serial sections were stained with anti-CD-31 (1:500) (Millipore). Photomicrographs, including 10 random 40× images within tumors were captured and tumor vascularity for each specimen was calculated by thresholding images in ImageJ for CD31+ staining and dividing the total pixel count by the total area of the images.

For evaluation of intra-tumoral inflammation, sections were stained for iba1 (1:1000, Wako) to evaluate GAM populations. Three random, 40× fields of view within the border of the rim of each tumor were used to evaluate GAM density and the relative ratios of GFP+ macrophages to total Iba1+ macrophages. Co-staining with iba1 and CD86 (1:50) (Millipore) or 1:50 CD206 (BioRad) was used to evaluate GAM polarization. Four random, 40× images were taken for each section and Image J (NIH) was used to calculate Mander’s co-localization coefficients between each inflammation marker for GFP+ macrophages and GFP- macrophages within each image. The density of CTLs within tumors was assessed by staining tumor sections for CD8-α (eBioscience) and using four random, 40× images through tumor tissue and counting the total CD8- α+ cells.

For evaluation of GAM proliferation, 20-μm thick serial tissue sections from either wild type or Nrp1^MgKO mice, which received either wild type or Nrp1^MgKO GFP⁺ BMDMs were stained for ki67 (1:500 Invitrogen, Clone SolA15), and Iba1 (1:500 Wako). Random 40× confocal images were taken along the tumor border and within the tumor and the number of ki67+, ki67+ Iba1+ GFP+, and ki67+ Iba1+ GFP- cells were quantified.

Flow Cytometry For Tumor Evaluation

Mice were heavily anesthetized with avertin and transcardially perfused with 50mL ice cold PBS. Brains were extracted and tumor tissue was micro-dissected and placed in ice cold FACS buffer (pH 7.40, 0.1M PBS, 1mM EDTA, 1%BSA, 50U/mL DNase I). Tissue was then subjected to digestion in papain supplemented with 50U/mL DNase I for 30 minutes followed by gentle trituration. Tissue homogenates were passed through 40μM filters and resuspended in 30% percoll in 1× HBSS. Samples were spun for 15 minutes at 1500rpm after which the pelleted fractions were resuspended in FACS buffer. Samples were washed and stained with CD3-Pacific Blue, CD4-PE, CD8-APC, and TNFα-PE antibodies in different combinations for 1 hour (all 1:100, Biolegend). Samples were then washed in FACS buffer thrice, fixed in 1%PFA w/v and cell staining was analyzed on a BD FACS Calibur using BD FACSDiva software (BD Biosciences). Samples were permeabilized with 0.1% Triton-X100 after fixation for intracellular staining.

For cell sorting experiments, brain hemispheres containing tumors were collected from mice 20 days post glioma induction and prepared for flow cytometry as above. Samples were then stained with CD11b-APC and CD45-PerCP-Cy5.5 for 1 hour. Samples were sorted and collected based as CD11b-APC positivity and relative CD45-PerCP-Cy5.5 expression and used for RNA isolation.

RNA isolation, cDNA synthesis and quantitative real-time PCR

Total RNA was collected from the CD11b⁺ CD45⁺ sorted samples using RNA-Bee following the manufacturer’s protocol. A cDNA library was created using the Applied Biosystems High capacity cDNA Reverse Transcriptase kit (Ref# 4368814). cDNA was loaded onto SABiosciences Chemokines and Receptors RT² Profiler PCR Array (PAMM-022A) using Fast SYBR Green Master Mix reporter (Ref# 4385612) and run on an Applied Biosystems QuantStudio 3 thermocycler. Data was analyzed using the Qiagen online data analysis center per the ΔΔC_T method, using 2 reference genes for normalization. Fold-Change is the normalized gene expression of the test sample divided by the normalized gene expression of the control sample. Fold-Regulation represents fold-change results in a biologically meaningful way. Fold-change values greater than one indicate a positive- or an up-regulation, and the fold-regulation is equal to the fold-change. Fold-change values less than one indicate a negative or down-regulation. The fold-regulation is the negative inverse of the fold-change.

Statistical Analysis

Data comparing two population means was analyzed using Student’s T-tests. Data assessing differences between multiple population means was compared by one-way ANOVA with Tukey’s post-hoc test analysis. A p-value < 0.05 was accepted as designating a significant difference between two population means with 95% confidence. Differences in survival between populations were analyzed by the Mantel-Cox log rank test. Statistical analysis was performed using GraphPad Prism (GraphPad Software Inc., La Jolla, CA).

RESULTS

Nrp1-Deficient Bone Marrow Derived Monocytes (BMDM) Slow Disease Progression in Wild Type Hosts

We previously showed that animals with Nrp-1-deficient microglia and macrophages (MG/MP) developed smaller gliomas, survived longer than wild-type mice, and that these Nrp-1-deficient MG/MPs polarized to an anti-tumorigenic phenotype [23]. To examine whether the anti-tumorigenic phenotype was more prominent in MG or in MPs, we transferred Nrp1-depleted BMDMs to wild type animals and assessed disease progression. Csf1R-GFP mice were lethally irradiated and received bone marrow (BM) transplants from either wild type or Nrp1^MgKO mice. Mice were allowed to recover for 8 weeks, and then they were implanted orthotopically with 3×10⁴ GL261 glioma cells. At 15 and 20 days post-disease induction, mice which had received Nrp1^MgKO marrow allografts had smaller tumors with significantly lower vessel density than those observed in mice which had received wt marrow (Figure 1A–D).

Figure 1. Mice that receive Nrp1-depleted BMDM allografts maintain their own microglial populations and exhibit slower glioma growth.

A–B. Representative images (A) and quantification of the volumes (B) of gliomas induced in Csf1R-GFP (wt) mice which received wild type (wt) or Nrp1^MgKO (KO) bone marrow allografts at 15 and 20 days post-disease induction. C–D. Representative images (C) and quantification of the luminal area of CD31+ tumor vasculature (D) of gliomas induced in Csf1R-GFP (wt) mice which received wild type (wt) or Nrp1^MgKO (KO) bone marrow allografts at 15 and 20 days post-disease induction. E–F. Representative images of Iba1 staining in splenic tissue (E.) and brain tissue (F.) derived from Csf1R-GFP mice which were lethally irradiated and immediately injected with marrow from either wt or Nrp1^MgKO (KO) donor mice. Donor mouse-derived BMDMs are GFP- G. Quantification of the number of splenic macrophages in mice which were GFP+ and, hence, derived from the host mouse (n=4 per group, NSD = no significant difference, * = p<0.05, ** = p<0.01).

To determine the macrophage composition of tumors, the microglia/macrophage marker Iba1 was used. Since almost all microglia seen in the contralateral hemisphere of mice expressed GFP and almost all the splenic macrophage populations did not express GFP (Figure 1E–G), it was assumed that nearly all GFP+Iba1+ cells within the tumor were derived from the host’s microglial pool, while all GFP- Iba+ cells were donor BMDMs which had infiltrated the tumor. Tumors in mice that had received wt or Nrp1^MgKO marrow were predominantly populated by BMDMs rather than microglia (Figure 2A). Moreover, microglia often accumulated around the border and periphery of tumors, but were rather sparse in the central core of tumors (Figure 2B, C, D). When analyzing the phenotypic polarization of the microglia and BMDMs populating tumors in both groups, BMDMs from wt or Nrp1^MgKO donors were found to express higher levels of CD86 relative to the wt microglia (Figure 3A) and lower levels of CD206 (Figure 3B). When comparing between the two groups however, Nrp1^MgKO BMDMs had higher expression of CD86 than wt BMDMs in the tumors of other animals (Figure 3C). Additionally, Nrp1^MgKO BMDMs had lower expression of CD206 than wt microglia (Figure 3B, D). Sorting GAM subpopulations based on CD45 expression (Supplemental Figure S1), we were able to analyze differentially expressed genes in cells present in the tumor microenvironment. Through transcriptomic analysis of Nrp1-depleted CD45-high expressing cells, we found that they express lower levels of HIF1α, CCR8, and IL1A (Figure 3E), which are factors that contribute to the tumor inflammatory milieu and neoangiogenesis [29–31]. Although a greater number of downregulated transcripts were observed in the Nrp1-depleted CD45-low expressing (microglial) populations (Supplemental Table S1), there were no pronounced gene expression changes, compared to those observed in CD45-high expressing cells (macrophages), suggesting that infiltrating BMDMs may be more plastic phenotypically than their resident microglial counterparts.

Figure 2. Tumors in mice with wild type microglia are predominantly populated by BMDM-derived GAMs.

A. Schematic of disease induction in Csf1R-GFP mice which have GFP-expressing microglia but have donor derived-peripheral BMDMs. B. Representative images of Iba1-stained macrophage infiltrate into gliomas at 15 and 20 days post-disease induction. Cells labeled with GFP and Iba1 are host-derived microglia while macrophages only labeled by Iba1 are derived from either a wt or Nrp1^MgKO (KO) donor mouse. C. Representative images of Iba1+ cells accumulated around the border and periphery of tumors D. Quantification of the host-derived microglia vs BMDM-derived GAMs (WT or KO, depending on the donor) present in tumoral tissue (n=4 per group. * = p<0.05, ***=p<0.01).

Figure 3. Nrp1-depleted BMDM-derived GAMs exhibit greater anti-tumorigenic polarization than wild type microglia and are associated with increased tumor T cell infiltrate.

A–B. Representative images of CD86 (A) and CD206 (B) staining of GAMs in Csf1R-GFP mice which received wild type (WT) or Nrp1^MgKO donor BMDMs. GFP+, Iba1+ cells are derived from host, WT microglia in both conditions while GFP-, Iba1+ cells are derived from the donor BMDMs (WT or KO). C–D. Meander’s colocalization coefficients for CD86 (C) and CD206 (D) overlap with Iba1 staining from donor derived BMDMs (either WT or KO) or Iba1 staining from GFP+, WT microglial populations (WT Host) E. Scatter plot of fold change values of differentially regulated genes in Nrp1^MgKO BMDMs. F–G. Representative images (F) and quantification (G) of the density of CD8+ CTL infiltrate into tumors at 15 and 20 days post-disease induction in Csf1R-GFP (wt) mice which received either wt or Nrp1^MgKO donor marrow (n=4 per group, * = p<0.05, ** = p<0.01, ****= p<0.0001).

We investigated the nature of the T cell infiltrate during glioma progression in Nrp1^MgKO mice. GL261-containing tumor tissue was isolated from wild type and Nrp1^MgKO mice and subjected to flow cytometric analysis for CD3, CD4, CD8, and TNFα. CD3+CD4+ and CD3+CD8+ T cells were found to make up a significantly higher percentage of cells isolated from tumors in Nrp1^MgKO mice. Additionally, CD3+ T cells isolated from tumors in Nrp1^MgKO mice were found to express significantly higher TNFα, relative to those isolated from tumors in wild type mice (Supplementary Figure S2). Similarly, we assessed the tumors populated with Nrp1^MgKO BMDMs or wt BMDMs to examine whether their CTL infiltrate correlated with the more anti-tumorigenic populations of GAMs. When stained for CD-8α, tumors from mice which had received Nrp1^MgKO BMDMs were found to have significantly higher CTL density within tumors at 15 and 20 days than in those receiving wt BMDMs (Figure 3F, G).

Mice with Nrp1-Deficient Microglia Alone Exhibit Slower Disease Progression

To investigate whether Nrp1^MgKO microglia alone could confer resistance to the growth of glioma, mice which were either wild type or Nrp1^MgKO were lethally irradiated and injected with marrow from Csf1R-GFP mice. The marrow of these mice was allowed to reconstitute for 12 weeks after which GL261-based gliomas were established in the brains of these mice.

In the contralateral brain side, GFP- populations of Iba1+ cells were prevalent, indicating that native host populations of microglia or microglial-specific progenitors survived the irradiation in the brains of each host. However, the spleens of each animal were predominantly populated by GFP⁺Iba1⁺ cells, indicating that peripheral macrophage populations were derived from the Csf1R-GFP donor mice (Supplementary Figure S3). Thus, these animals were chimeric having host-derived microglia, and donor-derived peripheral macrophages.

Post-glioma induction, Nrp1^MgKO mice (which had Nrp1-depleted MGs but wt BMDMs) were found to have significantly decreased tumor volumes at 15 days relative to those observed in wt mice, which received wt BMDM allografts (Figure 4A). At 20 days, the tumors trended to be smaller in Nrp1^MgKO host mice (p = 0.0586, Figure 4B). However at 15 and 20 days tumors in Nrp1^MgKO host mice, which received wt BMDMs had significantly decreased vascularity and significantly increased CTL infiltrate relative to tumors in wt host mice, which had received wt BMDMs (Figure 4C–D, E–G). These results indicated that Nrp1-depleted microglia alone had the potential to slow glioma progression and allow for an increase in immune cell infiltrate, indicating that microglia may play an anti-tumorigenic role within the tumor microenvironment.

Figure 4. Mice with Nrp1-deficient microglia and normal BMDMs exhibit slower glioma disease progression, correlated with increased CTL infiltrate.

A. Schematic of glioma induction in wt or Nrp1^MgKO mice, which have GFP^- microglia but have GFP⁺, donor-derived BMDMs. B–D. Representative images of tumors (B), CD31+ tumor vasculature (C), and CD8+ CTL infiltrate (D) in wild type (WT) or Nrp1^MgKO (KO) mice which received donor wt BMDMs from Csf1R-GFP mice. E–G. Quantification of tumor volumes (E), CD31+ area (F), and CTL infiltrate (G) at 15 and 20 days post-disease induction (n = 4–5 per group, * = p<0.05, ** = p<0.01, ***=p<0.001, **** = p<0.0001).

Next, we assessed whether there were differences in the composition of microglia and BMDMs constituting the total GAM infiltrate of the tumors. Mice, which had wt or Nrp1^MgKO microglia and wt BMDMs were evaluated. Gliomas in Nrp1^MgKO mice with wt BMDMs were found to have significantly elevated numbers of microglia which constituted a higher percentage of total GAM infiltrate, relative to wt mice with wt BMDMs (Figure 5A–D). Likewise, BMDMs constituted a significantly lower fraction of total GAM infiltrate into gliomas in the mice with Nrp1-depleted microglia (Figure 5E). However, total GAM infiltrates into tumors were not found to be different between the two groups (Figure 5F). Staining the tumors for the proliferation marker Ki67 revealed no significant difference in Ki67 levels across WT or Nrp1-depleted Iba-1+ cells, supporting the idea of greater migration towards and infiltration into tumors by Nrp1-depleted microglia as a more probable mechanism explaining their observed higher density in tumors. BMDMs across both groups showed similar levels of Ki67 as well (Supplementary Figure S4).

Figure 5. Gliomas in mice with Nrp1-depleted microglia exhibit higher numbers of microglia rather than BMDMs constituting the GAM infiltrate.

A. Representative GAM infiltrate in wild type (WT host) or Nrp1^MgKO (KO host) mice which received donor wt BMDMs from Csf1R-GFP mice. Iba1+, GFP- cells are host-derived microglia (WT or KO) while Iba1+, GFP+ cells are donor-derived, wt BMDMs. B–E. Quantification of the microglia (B) and BMDM-derived (C) infiltrate into tumors and the percentage of total GAMs that microglial (D) and BMDM-derived (E) infiltrate make up within tumors in mice with WT or Nrp1^MgKO (KO) microglia, but wt BMDMs. F. Quantification of the total GAM infiltrate into tumors (n = 4–5 per group, * = p<0.05, *** = p<0.001).

As there were differences in the ratio of microglia to BMDMs constituting the tumors depending on whether the microglia were Nrp1^MgKO, we also assessed whether there were differences in the polarization state of Nrp1^MgKO microglia relative to wt microglia and wt BMDMs that were infiltrating tumors. Nrp1^MgKO microglia exhibited significantly higher expression of CD86 and significantly lower expression of CD206, relative to wt microglia in the tumors of the wt BMDM allografted mice (Figure 6A–D). These results indicate that Nrp1-depletion from local microglial populations has the potential to alter their behavior, a property distinct from other GAMs.

Figure 6. Nrp1-depleted microglia exhibit more anti-tumorigenic polarization than wt microglia and BMDMs.

A–B. Representative CD86 (A) and CD206 (B) staining of GAMs populating tumors in mice with wt microglia (WT Host) and Nrp1^MgKO microglia (KO host), donor derived BMDMs are wt, GFP+. C–D. Quantification of the expression of CD86 (C) and CD206 (D) by host-derived microglia and donor-derived BMDMs in the tumors of mice which were wt or Nrp1^MgKO (KO) for Nrp1 expression for their microglia (n = 4 per group, * = p<0.05, ***=p<0.001).

Discussion

Our results reveal that depending on the manipulation of Nrp1 in microglia or BMDMs, the immune polarization of each of these cell populations can be altered and have the potential to slow glioma progression by impacting neovascularization, tumor growth, and altering the adaptive immune response, regardless of the activity of the other population (Figure 7). While tumors in wild type animals seem to be infiltrated by BMDMs that make up the majority of the GAM population, removing Nrp1 from microglia caused this cell population to predominate the tumor microenvironment. Based on measuring no proliferative changes in either cell type within the glioma microenvironment, independent of Nrp1 expression (Supplementary Figure S4), we speculate that differences in cell numbers were primarily due to differential migration and recruitment into tumors. Moreover, microglia are typically considered to have greater longevity than BMDMs, which may have also contributed to this observation [32]. These data are in agreement with prior work by Casazza et al. using Nrp1-depleted macrophages in other tumor models in which tumors had higher densities of infiltrating macrophages due to altered Nrp1-dependent chemotaxis in these cells [21].

Figure 7. Manipulation of Nrp1 expression in microglia or peripheral macrophages slows the progression of glioma.

Schematic of the glioma disease course in mice, which have Nrp1-depleted BMDMs or Nrp1-depleted microglia. Tumors in wild type mice are highly vascularized and predominately populated by pro-tumorigenic BMDM-derived GAMS and pro-tumorigenic microglia-derived GAMs to a lesser extent. Tumors exhibit relatively low CTL infiltrate. Ablation of Nrp1 from BMDM populations alters the disease course by decreasing tumor vascularity and increasing CTL infiltrate. BMDMs still make up a higher percentage of total GAMs but have increased anti-tumorigenic properties relative to tumor infiltrating microglia which still have pro-tumorigenic polarization and make up a minority population. When Nrp1 is selectively ablated from microglia only, anti-tumorigenic microglia infiltrate tumors in much greater numbers while BMDMs make up a smaller fraction of the total GAM infiltrate. These tumors also exhibit decreased vascularity and increased CTL infiltrate as well.

Studies have shown that BMDMs appear to be the predominant population contributing to GAM infiltrate, which aligns with our results [6, 33]. Our finding that WT BMDMs are the major GAM population in tumors of mice with WT microglia differs slightly from reports that earlier in the disease course microglia almost exclusively make up the GAM population, although marked increases in BMDM-derived macrophage infiltration were observed later in the disease [34, 35]. As we initially delivered more cells to induce the glioma than other studies, thus expediting tumor growth, it is quite possible that microglia predominated in the tumors early, but BMDM infiltrate increased with the rapid neovascularization of the tumors. We observed that Nrp1 depletion from microglia resulted in microglia predominating in tumors, possibly causing reductions in angiogenesis due to decreased expression of factors such as HIF1α, IL1A, and CCR8, thus preventing BMDMs from infiltrating the tissue, and limiting peripheral access into the tumors. In either case, depletion of Nrp1 from microglia or BMDMs made these cells more anti-tumorigenic and elicited enhanced CTL infiltration of the tumoral tissue, showing that manipulating either cell population has the potential to elicit more effective trafficking of CTLs. In the future we will profile the proteome of CTLs that populate the altered tumor microenvironment.

Tumor associated macrophages play diverse roles in the tumor microenvironment and this holds true for their role in gliomas as well. As they constitute up to 30% of the cellular content of many HGGs, they also are significant constituents of tumors [36]. We have shown here that the complexity of glioma increases as resident microglia and invading BMDMs play slightly different roles in tumor progression. Manipulation of Nrp1 in these cell populations has the potential to alter their behavior in the glioma microenvironment. Moreover, Nrp1’s removal from either microglial populations or just BMDMs can slow the disease progression, although not to the extent that is seen in animals, which carry Nrp1 deletion from both macrophage populations [23]. It would be meaningful to explore whether sustained pharmacological dosing with Csf1R inhibitors could lead to the preferential depletion of host microglia. The Csf1R inhibitor BLZ945 in a murine model of glioma does not entirely deplete infiltrating GAMs, but rather, causes surviving macrophages to adopt a more anti-tumoral phenotype associated with tumor regression [37]. This has been reported in other cancer types and associated with enhanced CTL infiltration into tumors as well [38]. Combinatorial Nrp1 and Csf1R inhibition, in that case, may prove more efficacious and worthwhile to explore [39].

Nrp1 is highly expressed by endothelial cells, many glioma-derived cells of HGGs, and subsets of immunosuppressive T_regs. It functions to support immunosuppression and increased tumoral expansion and, thus, its global inhibition within the tumor microenvironment is favorable for multiple reasons. If a patient’s innate immune cells could be manipulated in a way to enhance anti-tumoral immune responses in a similar manner that other autologous immune cells are modified and returned to a patient, this could have additional therapeutic potential. While T cells are widely regarded as the most important cell population to manipulate to trigger an anti-tumoral immune response, the importance of the microenvironment and the macrophages that often are highly immunosuppressive cannot be disregarded. Exploring Nrp1 inhibition in combination with well-defined immunotherapeutic approaches, such as PD-1 blockade, could prove fruitful in finding synergism between the therapies.