Class A1 scavenger receptor modulates glioma progression by regulating M2-like tumor-associated macrophage polarization

Macrophages enhance glioma development and progression by shaping the tumor microenvironment. Class A1 scavenger receptor (SR-A1), a pattern recognition receptor primarily expressed in macrophages, is up-regulated in many human solid tumors. We found that SR-A1 expression in 136 human gliomas was positively correlated with tumor grade (P<0.01), but not prognosis or tumor recurrence. SR-A1-expressing macrophages originated primarily from circulating monocytes attracted to tumor tissue, and were almost twice as numerous as resident microglia in glioma tissues (P<0.001). The effects of SR-A1 on glioma proliferation and invasion were assessed in vivo using an SR-A1-deficient murine orthotopic glioma model. SR-A1 deletion promoted M2-like tumor-associated macrophage (TAM) polarization in mice by activating STAT3 and STAT6, which resulted in robust orthotopic glioma proliferation and angiogenesis. Finally, we found that HSP70 might be an endogenous ligand that activates SR-A1-dependent anti-tumorigenic pathways in gliomas, although its expression does not appear informative for diagnostic purposes. Our findings demonstrate a relationship between TAMs, SR-A1 expression and glioma growth and provide new insights into the pathogenic role of TAMs in glioma.


INTRODUCTION
Despite significant advances in neurosurgery and chemo-radiotherapy, malignant gliomas, the most frequent primary tumors of the central nervous system, remain resistant to conventional therapeutic strategies. Glioblastoma (GBM) patients (WHO grades III and IV) have an average survival time of approximately 1.5 years after diagnosis [1,2]. Glioma lethality is attributable to its unchecked proliferation and invasive nature cultivated largely by the glioma microenvironment. This microenvironment is a complex structure of malignant cancer cells embedded in the vasculature and surrounded by a dynamic tumor stroma, which consists of various nonmalignant cells including fibroblasts and myeloid cells [3,4]. Macrophages, an important cell component of the tumor stroma, are recruited and mobilized by tumor-derived factors such as IL-4, IFN-γ and IL-13 and then induced to differentiate into pro-tumorigenic tumor-associated macrophages (TAMs). TAMs can promote tumor angiogenesis, proliferation and invasion [1,2,5], and TAM infiltration is often associated with poor prognosis. Inhibition of TAM infiltration leads to glioma shrinkage in animal models [6,7].
At the initial phase of tumor progression, or in non-progressing or regressing tumors, TAMs assume a classically activated phenotype characterized by a proinflammatory response and active antigen presentation [8]. They create an inflammatory environment that is promutagenic and pro-growth. As tumors progress to a more advanced and thus more invasive stage, TAMs resemble alternatively activated macrophages (M2-like), which stimulate angiogenesis, tumor cell intra/extravasations and proliferation [9]. At metastatic sites, TAMs promote tumor cell settlement and subsequent outgrowth of metastatic lesions. Recent profiling studies showed that TAMs that arise from CCR2 + bone marrow-derived cells could be distinguished from resident macrophages by VCAM-1

Research Paper
expression. The gene expression profiles of TAMs and conventional alternatively activated macrophages largely do not overlap [10][11][12][13]. Despite rigorous attempts, the precise ontogeny and function of TAMs in solid tumors, especially in gliomas, are not well understood.
Class A1 scavenger receptor (SR-A1; also known as MSR1 and CD204) is a pattern recognition receptor involved in the progression of multiple tumors [6,7,14]. It is a marker for alternatively activated macrophages/ microglia in glioma progression [15]. SR-A1 suppresses lung cancer progression by inhibiting MMP-9 production [16,17] and modulates macrophage polarization in cardiovascular inflammatory microenvironments [18][19][20]. The present study demonstrates that SR-A1 inhibits glioma growth, invasion and angiogenesis by preventing M2-like TAM infiltration and differentiation. HSP70 is identified as a potential endogenous ligand for activating the SR-A1-linked STAT3 and STAT6 signaling cascades in the glioma microenvironment. Our results provide mechanistic insights and novel rationale for targeting the HSP70/SR-A1 axis in glioma intervention.

SR-A1 expression in macrophages/microglia is associated with glioma malignancy and prognosis
Immunohistochemical (IHC) analysis in 136 human glioma samples showed that SR-A1 expression increased with glioma grade (Figure 1Aa & 1Ab), suggesting that SR-A1 might be associated with malignant glioma initiation and progression. Kaplan-Meier analysis revealed that GBM patients with higher SR-A1 levels had a slightly better survival rate than those with lower levels ( Figure  1B). Glioma recurrence in the SR-A1 low expression group was more frequent than in SR-A1 high expression group for patients with the same GBM grade (Table 1). FACS analysis showed that 4.24% of cells in grade III gliomas and 10.2% of cells in grade IV gliomas were CD11b + macrophages. However, SR-A1 high /CD11b + macrophages accounted for 60.8% of SR-A1 + /CD11b + macrophages in grade III gliomas but only 56.1% in grade IV gliomas, reflecting decreased SR-A1 high /CD11b + macrophages in advanced malignant glioma ( Figure 1C). These results suggest that SR-A1 expression is inversely associated with glioma malignance.
We also evaluated the influence of SR-A1 on macrophage polarization in vitro. SR-A1 deficiency resulted in up-regulation of M2-like markers (Mmp-2, Tgf-β, Mrc-2, Mgl-1, Fizz-1), but no significant changes in M1-like markers (Tnf-a) in the presence of GL261 cells ( Figure 7A). These results were corroborated by FACS analysis of the macrophage subtypes ( Figure 7B). These data support the hypothesis that BMDM-specific SR-A1 might inhibit glioma progression by preventing M2-like polarization.
Finally, we explored the potential signaling cascades through which SR-A1 may regulate M2-like BMDM polarization. We found that co-culture with glioma cells induced STAT3 and STAT6 phosphorylation, two important signaling molecules controlling M2-like macrophage differentiation in BMDMs [25]. SR-A1 deletion in BMDMs increased this effect ( Figure 7C & 7D), suggesting that STAT3 and STAT6 signaling contribute to SR-A1 mediated BMDM polarization in the glioma microenvironment. www.impactjournals.com/oncotarget
Macrophage cytokine levels were measured to assess SR-A1 activation. qPCR results showed that treatment with fraction 3 (40% sucrose) or fraction 7 (30% sucrose) elevated M1-like cytokine production and suppressed that of M2-like cytokines in Sr-a1 +/+ BMDMs compared with Sr-a1 −/− BMDMs ( Figure 8A). The protein components of these two fractions were identified by SDS-PAGE coupled with liquid chromatography-mass spectrometry (LC-MS). Heat shock protein 70 (HSP70), which has been reported as a ligand for SR-A1, was detected among 47 identified proteins [26]. HSP70 expression in glioma was also confirmed by IHC analysis and western blot (Supplementary Figure S7).

DISCUSSION
Of the multiple unique stromal cell types commonly found in solid tumors, TAMs are essential in fostering tumor progression [27,28]. TAMs function in angiogenesis and produce soluble mediators to support tumor cell proliferation, survival and invasion [24]. Mechanisms of TAM communication with the tumor microenvironment remain poorly understood. This study suggests that HSP70/SR-A1-mediated signaling connects . Representative STAT6 C. and STAT 3 D. western blots in murine BMDMs (1×10 6 ) with equal numbers of GL261 cells in the transwell co-culture system for 4 or 6 h. Changes in target protein levels were compared with total STAT6 or STAT3 (Sr-a1 +/+ group and Sr-a1 −/− group was separated by irrelevant protein sample). TAM polarization to glioma growth, invasion and angiogenesis.
TAM infiltration is correlated with enhanced glioma malignancy and poor prognosis [29][30][31], and approximately one third of all cells in glioma biopsies are labeled by macrophage markers [32,33]. Macrophages and microglia are especially numerous in high-grade tumors [34], and ramified microglia (with preserved longer, perpendicularly branching cell processes) are abundant in gemistocytic astrocytomas [35]. A positive correlation was found between the number of CR3 (CD11b)-positive macrophages/microglia in gliomas and tumor proliferation rate [32]. Another striking feature, especially of high-grade gliomas, is the large number of immune cells (microglia and macrophages) that accumulate in the tumor mass (tumor-associated myeloid cells; TAMs). In GBM, TAMs can constitute up to 30% of the tumor mass [36] and, consistent with our observation that SR-A1 expression increased with glioma grade, are reportedly less abundant in lower grade gliomas. However, improved survival and reduced tumor recurrence were observed in patients with higher SR-A1 levels. We suggest that increased SR-A1 expression might be caused by TAM infiltration. Our studies showed increased SR-A1 + /CD11b + and decreased SR-A1 high macrophages during glioma development and progression in both human and murine gliomas. These findings indicate that SR-A1 may serve not only as a marker but also as a negative regulator of TAMs during glioma progression. SR-A1 may serve as a tumor suppressor during glioma progression, but might be suppressed by the glioma microenvironment, leading to its down-regulation in BMDMs co-cultured with glioma cells.
The origin of TAMs is heavily debated. TAMs may be derived from circulating monocytes attracted to the tumor tissue. Alternatively, TAMs can be converted from resident microglia under the instruction of the tumorigenic microenvironment [37]. As gliomas progress, abundant BMDMs are recruited to the tumor tissue and become the dominant macrophages possessing hypertrophic (M2-like) characteristics [38]. Our observations (Figure 4) indicate that BMDMs are a primary source of SR-A1 in the glioma microenvironment. In addition, bone marrow transplantation experiments ( Figure 5) confirm that SR-A1 in BMDMs is largely, if not completely, accountable for its anti-tumorigenic activities. SR-A1 expression in tumor tissue TAMs, mostly recruited macrophages, correlates inversely with glioma malignancy in humans ( Figure 1). This is confirmed by the observation that SR-A1 deficiency increases hypoxia (Figure 3), which is consistent with previous reports that local hypoxia supports angiogenesis and inflammation and contributes to tumor metastasis and progression [39].
The dynamic nature of TAMs during tumor progression is influenced in part by local concentrations of cytokines and chemokines, as well as varied interactions of TAMs with normal and malignant cells. Our study demonstrates for the first time that SR-A1 could inhibit TAM differentiation toward an M2-like phenotype by suppressing STAT3 and STAT6 signaling in murine gliomas. SR-A1 deletion increased the observed frequency of the TAM M2 subtype. Polarized TAMs influence tumor progression by releasing multiple cytokines, such as VEGF and MMP-9, to sustain angiogenesis and proliferation [40]. In agreement with our previous studies in lung cancer [16], we found that SR-A1 deletion was associated with overproduction of MMP-9, VEGF and other tumorigenic and angiogenic factors (Figure 3 & Supplementary Figure S3) The interplay between tumor cells and stromal cells determines the glioma tissue phenotype [41]. Many glioma-related genes are expressed comparably in both tumor cells and stromal cells, although they usually function differently in different cell types. For example, while TLRs in tumor cells facilitate evasion of immune surveillance and thus promote tumor growth, they polarize macrophages to an anti-tumor M1-like phenotype [28,42]. This dichotomy makes targeting these molecules for cancer treatment challenging [43]. SR-A1 is expressed only in TAMs, making it an ideal protein for stromal cell-specific tumor therapy. Our data highlight a role for SR-A1 in the regulation of a dynamic interplay between tumor cells and stromal cells. On one hand, tumor cells promote BMDM recruitment into glioma tissues ( Figure 4E & 4F). SR-A1 deficiency seems to amplify this effect, likely as a consequence of enhanced p38 phosphorylation [44]. Reciprocally, infiltrated BMDMs induce tumor cells to alter their migratory and invasive behaviors. SR-A1 deficiency appears to enhance this effect by activating STAT3/ STAT6 signaling and routing macrophages toward an M2-like phenotype. This type of cell-cell interaction appears to be context-dependent as we observed that SR-A1 influenced macrophage infiltration at the tumor center but not around the edges ( Figure 4H). SR-A1 may serve a key gatekeeper, blocking communication between tumor cells and stromal cells and preventing glioma deterioration.
SR-A1 appeared to influence macrophage proliferation, infiltration and differentiation simultaneously. Robbins, et al. reported that SR-A1 is necessary for promoting resident macrophage proliferation in mouse atherosclerotic lesions, while others have shown the SR-A1 deficiency enhances macrophage recruitment in different diseases [45][46][47][48]. We recently showed that SR-A1 deficiency could enhance RAGE expression and function [49], and others showed that RAGE expression in TAMs could enhance glioma progression by promoting angiogenesis [50]. Thus, SR-A1-dependent macrophage infiltration may alter the local microenvironment, which in turn may impact cell proliferation and/or death, shaping a pro-M1 or pro-M2 setting to promote differentiation.
Finally, our results demonstrate that HSP70 could be an endogenous ligand that activates SR-A1dependent anti-tumorigenic pathways in gliomas. HSPs are over-expressed in a wide range of human cancers and are implicated in tumor cell proliferation, differentiation, invasion, metastasis, death and recognition by the immune system [51,52]. Our results show that HSP70, but not HSP40 or 60, is specifically involved in glioma progression, although its expression does not appear informative for diagnostic purposes. Elevated HSP70 expression correlates with poor prognosis in breast, endometrial, uterine, cervical and bladder carcinomas [53]. However, its expression predicts improved response to chemotherapy in osteosarcomas and glioma [53]. This likely reflects the complex and unique nature of the tumor microenvironment in different tissues and organs.
Although HSP70 by itself is sufficient to induce an M1-like phenotype in macrophages in vitro, its precise role in glioma pathogenesis is unclear. HSP70-induced phenotypic alterations of SR-A1-null macrophages suggest that HSP70 has SR-A1 independent functions yet to be determined. Additionally, Neyen, et al. identified several SR-A1 ligands that do not completely overlap with those found in the present study in models of ovarian and pancreatic cancer [54]. This apparent discrepancy highlights the importance of the tumor microenvironment in SR-A1 activity.
In summary, our findings demonstrate a relationship between TAMs, SR-A1 expression and glioma growth and provide new insights into the pathogenic role of TAMs in glioma. The HSP70/SR-A1 pathway may inhibit STAT3/6 signaling in TAMs to slow glioma progression. Our present study suggests that inhibiting macrophage recruitment and influencing macrophage polarization (away from an M2-like phenotype) with SR-A1 ligands such as HSP70 could reduce glioma development and progression.

Samples
We evaluated specimens resected from primary glioma patients between 2010 and 2014 at the Brain Hospital Affiliated with Nanjing Medical University ( Table 2). Informed written consent was obtained from all patients under protocols approved by the Institutional Review Board of Nanjing Medical University. Two pathologists provided histological diagnoses according to the revised 2008 WHO classification [55]. Patients were divided into SR-A1 high and low expression groups according to median SR-A1 expression level.

Animals, tumor implantation and bone marrow transplantation
All animal protocols were reviewed and approved by the intramural Ethics Committee on Humane Treatment of Experimental Animals. SR-A1 +/+ and GFP + C57BL6 mice (6-8 weeks old) were obtained from the animal colony of Nanjing Medical University. SR-A1 −/− C57BL6 mice were obtained from the Jackson Laboratory (Stock number: 006096) and generated in Nanjing Medical University. Mice were bred and maintained under pathogen-free conditions with a 12:12-h light: dark cycle and regular chow diet and water.

Magnetic resonance imaging and evaluation of tumor volume
On days 14 and 21 after tumor implantation, animals were anesthetized and glioma images were acquired as described previously [56].

Histopathology and immunohistochemistry
Tumors were fixed in 4% paraformaldehyde overnight. Samples used for paraffin sections were transferred to 10% neutral-buffered formalin and embedded in paraffin wax. Sections (4 μm) were prepared, stained with hematoxylin and eosin (H&E) and examined microscopically. Tumors used for cryosections were washed with phosphate-buffered saline (PBS) after fixation and placed in 30% sucrose at 4°C before being transferred into optimal cutting temperature-embedding compound.

Bone marrow derived macrophage isolation
Bone marrow cells were harvested and cultured as described previously [57].

Flow cytometry
Whole brain hemispheres were broken into single cells using a syringe plunger and filtrated through a 70-μm cell strainer. Erythrocytes were lysed using the erythrolyse solution (Sigma-Aldrich, USA), and the remaining cells were resuspended in PBS (10 6 /100 μL PBS). Before staining, cells were permeabilized using the Leucoperm kit (Serotec). Cells were then incubated with the primary antibody in the dark for 45 min at room temperature, washed with 5 mL PBS and resuspended in 400 μL 2% paraformaldehyde in PBS. Fluorescence-labeled cells were counted using the FACS Calibur cytometer connected to CellQuest software (BD Biosciences). Primary

Total RNA purification and real-time PCR analysis
Total RNA was isolated from 100 mg of tumor tissue using the RNeasy extraction kit (Takara, Japan). RNA yield and purity were determined by spectrophotometer. cDNA synthesis was done with 500 ng total RNA using the cDNA synthesis kit (Takara). qRT-PCR was done using SYBR Green ER qPCR Super mix (Roche, Swiss) and an AB7500 system (Seegene, USA). Primers were designed by Takara. Expression of each target gene was normalized to that of the control gene, β-actin or GAPDH.
Macrophage and glioma cell co-culture system GL261 cells were plated with murine primary BMDMs. In brief, primary Sr-a1 +/+ and Sr-a1 −/− macrophages were isolated and re-suspended at a density of 5×10 4 cells/ml in DMEM with 1% FBS. After 5 days, medium was removed and 2×10 4 GL261 cells were seeded on top of the BMDMs in DMEM with 10% FBS, and cells were co-cultured for 4, 6 and 12 h. As controls, primary macrophages were switched to the same medium without GL261 cells. Macrophage RNA was isolated and quantitatively assessed via qRT-PCR. The experiments were repeated three times with duplicate samples per group. Segregated macrophage-glioma co-cultures were prepared as follows: 5×10 4 Sr-a1 −/− or Sr-a1 +/+ macrophages were seeded on 0.4 μm inserts (Millicell, USA) in DMEM with 1% FBS. After 5 days, inserts were moved to 24-well plates containing 5,000 GL261 cells/well in DMEM with 10% FBS. Empty inserts with the same medium were used as controls. GL261 proliferation was measured every other day. Experiments were repeated three times with duplicate samples per group.

Cell proliferation assay
The CCK-8 kit (Beyotime, China) was used to measurement cell proliferation. In brief, equal numbers of GL261 cells were seeded in 96-well or 24-well plates (co-culture) with respective treatments (duplicate samples for each treatment). At each time point, the medium was changed to include a 1:10 dilution of CCK-8. After 2 h at 37°C, absorbance was read at 450 nm. Absorbance values at each time point were normalized using those at day 0.

Preparation and LC-MS analysis of brain lysate
Mice were perfused with PBS transcardially. The brain glioma was removed, homogenized with RPMI-1640 and centrifuged at 15,000 rpm for 5 min. The supernatant was made up to 1 ml with RPMI-1640 and used as the brain lysate. For digestion assays, brain lysate was incubated with pronase (1-10 U/ml, Roche, USA) or DNase I (50 μg/ml, Roche) at 37°C for 1 h. For the sucrose density gradient centrifugation and LC-MS analysis, the brain lysate was ultracentrifuged at 47,000 rpm for 1 h. The supernatant was applied to DEAE sepharose fast flow columns (GE Healthcare, USA), and the flow-through was condensed by ultrafiltration using an Amicon Ultra-4 centrifugal filter unit with Ultracel-10 membrane (Millipore, USA). Four hundred microliters of condensed solution was layered on a 1-ml 10-40% (w/w) linear sucrose gradient in PBS and centrifuged at 40,000 rpm for 12 h. Sucrose was depleted by ultrafiltration from each of the sucrose gradient fractions. We added the sucrose-gradient fractions to cultures of SR-A1 +/+ and SR-A1 −/− macrophages to examine induction of macrophage polarization cytokine expression. LC-MS analysis of sucrose gradient fractions 1-7 was performed after trypsinization using a Qstar-XL mass spectrometer (Applied Biosystems, USA).

Generation of recombinant protein
HSP40, 60 and 70 recombinant proteins with GST tags were kindly provided by Dr. Su Chuan from Nanjing Medical University. Murine IL-4 (100 ng/ml), IL-13 (100 ng/ml) and LPS (50 ng/ml) recombinant proteins were acquired from Sigma-Aldrich. Recombinant proteins were co-cultured with SR-A1 +/+ and SR-A1 −/− macrophages and recombinant GST protein was used as a negative control for cytokine induction in macrophages.

Assays
TGF-β, MCP-1 and IL-10 levels in tumor homogenates and plasma were determined using mouse TGF-β, mouse MCP-1 and mouse IL-10 ELISA Kits according to the manufacturer's protocols (Excell, China). Plates were read on a Bio-Rad (USA) Model 680 microplate reader at 415 nm. Protein antibody array data were provided by Ray Biotech Inc, China.
Tumor size was calculated in T1-weighted image sets obtained 10 min after contrast injection. Slices showing enhanced areas were analyzed by measuring the area of a region of interest around the section of enhancement. Volume was calculated by multiplying by the slice thickness. This procedure was repeated for all slices showing enhancement and the areas were summed to determine a total volume.

Statistical analysis
All experiments were performed at least three times. In Kaplan-Meier curves, survival differences were compared by log-rank analysis. All statistical analyses were performed using Statistical Package for Social Science (SPSS) 13.0 software. Statistical significance was assessed by Student's t test (unpaired two-tailed). P<0.05 (*), 0.01 (**) or 0.001 (***) was considered statistically significant. www.impactjournals.com/oncotarget