CCR5-Mediated Signaling is Involved in Invasion of Glioblastoma Cells in Its Microenvironment

The chemokine CCL5/RANTES is a versatile inflammatory mediator, which interacts with the receptor CCR5, promoting cancer cell interactions within the tumor microenvironment. Glioblastoma is a highly invasive tumor, in which CCL5 expression correlates with shorter patient survival. Using immunohistochemistry, we identified CCL5 and CCR5 in a series of glioblastoma samples and cells, including glioblastoma stem cells. CCL5 and CCR5 gene expression were significantly higher in a cohort of 38 glioblastoma samples, compared to low-grade glioma and non-cancerous tissues. The in vitro invasion of patients-derived primary glioblastoma cells and glioblastoma stem cells was dependent on CCL5-induced CCR5 signaling and is strongly inhibited by the small molecule CCR5 antagonist maraviroc. Invasion of these cells, which was enhanced when co-cultured with mesenchymal stem cells (MSCs), was inhibited by maraviroc, suggesting that MSCs release CCR5 ligands. In support of this model, we detected CCL5 and CCR5 in MSC monocultures and glioblastoma-associated MSC in tissue sections. We also found CCR5 expressing macrophages were in close proximity to glioblastoma cells. In conclusion, autocrine and paracrine cross-talk in glioblastoma and, in particular, glioblastoma stem cells with its stromal microenvironment, involves CCR5 and CCL5, contributing to glioblastoma invasion, suggesting the CCL5/CCR5 axis as a potential therapeutic target that can be targeted with repositioned drug maraviroc.


Introduction
Glioblastoma is one of the most aggressive brain tumors and poorly responsive malignancies to treatment with among the shortest survival rates of all cancers [1]. Patients' 5-year survival rate is less than 5% [2], regardless of novel modalities in surgery, irradiation, and chemotherapy [3,4]. Table 1. Immunohistochemical analyses of CCL5 and CCR5 expression in glioblastoma and noncancerous tissues.

Figure 1.
Brain tissue sections immunolabeling for CCL5 and CCR5. Immunohistochemical localization of CCL5 and CCR5 in glioblastoma and non-cancerous tissue (NB1 and NB2) sections was performed as described in Materials and Methods. Cell nuclei were counterstained by hematoxylin (blue). CCR5 epitope blocking peptide (P) was used (in CCR5+P images) as a control for specific binding of the primary antibody. Scale bar represents 100 µm. Black arrows indicate examples of CCL5 and CCR5 positive cells. Microscopy was carried out at 20 × objective magnification. *Survival: from the date of the first operation until death, GB-glioblastoma. **Necrosis and angiogenesis were analyzed as »yes« or »no« by observation of the obtained glioblastoma tissue sample, before processing of the sample. ***Karnofsky score (at the time of the first operation) was determined by the clinician. Patient's functional impairment; 80-100: normal activity, able to work, Figure 1. Brain tissue sections immunolabeling for CCL5 and CCR5. Immunohistochemical localization of CCL5 and CCR5 in glioblastoma and non-cancerous tissue (NB1 and NB2) sections was performed as described in Materials and Methods. Cell nuclei were counterstained by hematoxylin (blue). CCR5 epitope blocking peptide (P) was used (in CCR5+P images) as a control for specific binding of the primary antibody. Scale bar represents 100 µm. Black arrows indicate examples of CCL5 and CCR5 positive cells. Microscopy was carried out at 20× objective magnification. * Survival: from the date of the first operation until death, GB-glioblastoma. ** Necrosis and angiogenesis were analyzed as »yes« or »no« by observation of the obtained glioblastoma tissue sample, before processing of the sample. *** Karnofsky score (at the time of the first operation) was determined by the clinician. Patient's functional impairment; 80-100: normal activity, able to work, no special care needed; 50-70: unable to work, able to live at home, a varying amount of assistance needed; 0-40: unable to care for self, hospital care. **** Glioblastoma subtypes, mesenchymal (MES), proneural (PN) classical (CL), and MIX were determined based on the pattern of mRNA expression levels of selected genes, according to Behnan et al. (2017). n.a.: not available, ND: not determined. ***** IDH = Isocitrate dehydrogenase enzyme mutations were determined at the Pathology; wt-wild type, non-mutated.

Expression of CCL5/CCR5 Axis in Primary Glioblastoma Cells and Glioblastoma Stem Cells
To further investigate the cellular origin of CCL5 and CCR5 in glioblastoma tissues, using IHC we screened for the expression of CCL5 and CCR5 in primary differentiated glioblastoma cells and glioblastoma stem cells (GSCs) that were cultured from patients' tumors. Brain tissue samples from glioblastoma patients were obtained from the Department of Neurosurgery of the University Medical Centre, University of Ljubljana. These tumor samples were either used for the generation of primary glioblastoma cells and GSC or were frozen upon tumor removal for RNA extraction. GSC cells and the two previously established CD133+ GSC lines, NCH644, and NCH421k were grown as spheroid suspensions in serum-free, complete Neurobasal Medium as described by Podergajs [30]. Spheroids were fluorescently labeled for CCL5 and CCR5 expression as described in Materials and Methods. The ICC analyses are shown in Figure 2. The quantification scoring analyses revealed that the chemokine expression seems to be higher compared to the receptor and that very low or no CCL5 nor CCR5 expression was observed in patient Nb. 3 and in the U373 cell line (Table 3). In the three glioblastoma stem cells (GSCs) spheroids, high CCR5 protein expression was seen, but CCL5 could not be detected even using more sensitive detection by immunofluorescence ( Figure 3). Normal astrocytes do not express CCL5 nor CCR5 ( Figure S2). no special care needed; 50-70: unable to work, able to live at home, a varying amount of assistance needed; 0-40: unable to care for self, hospital care. ****Glioblastoma subtypes, mesenchymal (MES), proneural (PN) classical (CL), and MIX were determined based on the pattern of mRNA expression levels of selected genes, according to Behnan et al. (2017). n.a.: not available, ND: not determined. *****IDH = Isocitrate dehydrogenase enzyme mutations were determined at the Pathology; wt-wild type, non-mutated.

Expression of CCL5/CCR5 Axis in Primary Glioblastoma Cells and Glioblastoma Stem Cells
To further investigate the cellular origin of CCL5 and CCR5 in glioblastoma tissues, using IHC we screened for the expression of CCL5 and CCR5 in primary differentiated glioblastoma cells and glioblastoma stem cells (GSCs) that were cultured from patients' tumors. Brain tissue samples from glioblastoma patients were obtained from the Department of Neurosurgery of the University Medical Centre, University of Ljubljana. These tumor samples were either used for the generation of primary glioblastoma cells and GSC or were frozen upon tumor removal for RNA extraction. GSC cells and the two previously established CD133+ GSC lines, NCH644, and NCH421k were grown as spheroid suspensions in serum-free, complete Neurobasal Medium as described by Podergajs [30]. Spheroids were fluorescently labeled for CCL5 and CCR5 expression as described in Materials and Methods. The ICC analyses are shown in Figure 2. The quantification scoring analyses revealed that the chemokine expression seems to be higher compared to the receptor and that very low or no CCL5 nor CCR5 expression was observed in patient Nb. 3 and in the U373 cell line (Table 3). In the three glioblastoma stem cells (GSCs) spheroids, high CCR5 protein expression was seen, but CCL5 could not be detected even using more sensitive detection by immunofluorescence ( Figure 3). Normal astrocytes do not express CCL5 nor CCR5 ( Figure S2). Immunocytochemical localization of CCL5 and CCR5 in primary glioblastoma cells. ICC localization of CCL5 and CCR5 in primary glioblastoma cells isolated from patients' tumors and the Figure 2. Immunocytochemical localization of CCL5 and CCR5 in primary glioblastoma cells. ICC localization of CCL5 and CCR5 in primary glioblastoma cells isolated from patients' tumors and the glioblastoma cell line U373 as performed as described in Materials and Methods. Cell nuclei were counterstained by hematoxylin (blue). CCR5 epitope blocking peptide (P) was used (in CCR5 + P images) as a control. Negative control staining was performed in the absence of the primary antibody. Scale bar represents 50 µm. Microscopy was carried out at 20× objective magnification.

Mesenchymal Stem Cells In Vitro and In Vivo Tumor Sections Express CCL5
The tumor progression of glioblastoma induces a host response, which is associated with the infiltration of stromal cells, e.g., bone-marrow-derived mesenchymal stem cells (MSC) and hematopoietic stem cells (HSC) and their progenitors, comprising various mature lymphocytes, macrophages [12,31]. Previous studies have shown that MSCs, homing to glioblastoma can de-differentiate to other stromal cells via paracrine effectors, such as immunomodulatory cytokines, or by direct interactions with GB cells [32]. Moreover, we have demonstrated that human MSCs exploit the immune response mediating chemokines to impact the phenotype of glioblastoma [32,33] and later described complex mechanisms of their indirect [14,34] and/or direct cross-talk [35,36]. Here, we were interested if MSC were alone and when in glioblastoma microenvironment express CCL5 and CCR5. Therefore, we have immunolabeled bone marrow-derived MSCs in monocultures by CCL5 and CCR5 antibodies and demonstrated high expression of both antigens in MSCs ( Figure S1). Furthermore, labeling these antigens of tissue sections from 3 patients, (Nb. 8, Nb. 12, and Nb. 5), we found that CCL5 expression was co-localized with MSC marker CD105 (Figure 4).

CCR5 Is Expressed in Glioblastoma-Associated Macrophages in Tumors
Cross-talk of glioblastoma cells with microglia and infiltrating macrophages occurs through the release of cytokines, which promote tumor growth [14,35,37]. Tumor-associated macrophages represent about 40% of all cells in a glioblastoma specimen [38] and microglia-mediated immunosuppression may involve CCL5/CCR5 via CCR5 signaling on macrophages to induce their activation and polarization [39]. Fluorescence immunohistochemical staining of tissue sections of 3 patients (Nb. 8, Nb. 12, and Nb. 5) revealed the expression of CCR5 in glioblastoma-associated macrophages, labeled by the antibody specific marker CD68 ( Figure 5). Furthermore, labeling these antigens of tissue sections from 3 patients, (Nb. 8, Nb. 12, and Nb. 5), we found that CCL5 expression was co-localized with MSC marker CD105 (Figure 4.).

CCR5 is Expressed in Glioblastoma-Associated Macrophages in Tumors
Cross-talk of glioblastoma cells with microglia and infiltrating macrophages occurs through the release of cytokines, which promote tumor growth [14,35,37]. Tumor-associated macrophages represent about 40% of all cells in a glioblastoma specimen [38] and microglia-mediated immunosuppression may involve CCL5/CCR5 via CCR5 signaling on macrophages to induce their activation and polarization [39]. Fluorescence immunohistochemical staining of tissue sections of 3 patients (Nb. 8, Nb. 12, and Nb. 5) revealed the expression of CCR5 in glioblastoma-associated macrophages, labeled by the antibody specific marker CD68 ( Figure 5).

CCR5 is Expressed in Glioblastoma-Associated Macrophages in Tumors
Cross-talk of glioblastoma cells with microglia and infiltrating macrophages occurs through the release of cytokines, which promote tumor growth [14,35,37]. Tumor-associated macrophages represent about 40% of all cells in a glioblastoma specimen [38] and microglia-mediated immunosuppression may involve CCL5/CCR5 via CCR5 signaling on macrophages to induce their activation and polarization [39]. Fluorescence immunohistochemical staining of tissue sections of 3 patients (Nb. 8, Nb. 12, and Nb. 5) revealed the expression of CCR5 in glioblastoma-associated macrophages, labeled by the antibody specific marker CD68 ( Figure 5). Macrophages were immunolabeled, using an antibody against the specific marker CD68. Nuclei were stained with DAPI (blue), CD68 with Alexa Fluor 546 (red), and CCR5 with Alexa Fluor 488 (green) dye. Merged images represent colocalization (yellow color) of CD68 and CCR5. Microscopy was carried out at 20× objective magnification. Scale bar represents 100 µm.

CCL5 and Mesenchymal Stem Cells Enhance the Invasion of Primary Glioblastoma Cells and Glioblastoma Stem Cells
Glioblastoma cell invasion that has been characterized as a single cell infiltration into the brain parenchyma is crucially supported by its microenvironment comprised of stromal cells. Previously, we have mostly studied mesenchymal stem cells (MSCs), affecting glioblastoma cell phenotype via paracrine interactions, secreting cytokines as demonstrated by Motaln et al. [34] and Breznik et al. [36]. Here, we focused on revealing the functional significance of the CCL5/CCR5 axis in paracrine, i.e., indirect MSC-GB cell interaction, such as invasion. Transwell chambers invasion assays were used to investigate primary GB cells from patient Nb. 2, which expressed high CCR5, but very low CCL5 antigens ( Figure 2).
After the cells were stimulated with recombinant human chemokine CCL5, added to the lower chamber, Matrigel invading cells were quantified as described in Methods. The invasion was inhibited by synthetic CCR5 antagonist maraviroc added to the cells in the upper chamber ( Figure 6A). When CCL5-expressing MSCs ( Figure S1) were added to the lower transwell compartment as a chemoattractant, GB Nb.2 cells' invasion was significantly enhanced and was also inhibited by maraviroc ( Figure 6B). Noteworthy, GB Nb.2 cell viability was not affected by maraviroc even at higher concentrations ( Figure S3). Further, we validated the functional role of the CCL5/CCR5 axis in the invasion of established GSC line NCH644, which expresses CCR5 but low or no CCL5 (Figure 3). GSCs were stimulated with recombinant human chemokine CCL5 and their invasion in the presence or absence of maraviroc was quantified as above ( Figure 6C). When CCL5-expressing MSCs were used as a chemoattractant in the lower chamber, GSCs invasion was significantly enhanced but was remarkably inhibited by the CCR5 inhibitor maraviroc ( Figure 6D). The significance was not reached here, due to the higher variance among biological repetitive experiments. Each value represents mean ± SD (n = 3). ★ p < 0.05, ★★ p < 0.01, ★★★ p < 0.001 vs. control group (t-test).

CCL5 and CCR5 mRNA Levels are Increased in High-Grade Gliomas
To determine, if the high protein levels of CCL5/CCR5 in GB tissues result from increased gene expression, i.e., transcriptional activity, we determined the mRNA levels of CCL5 and CCR5 in the tissues of normal and malignant specimens: non-cancerous brain tissues (n = 16), glioma I-II-low-
CCL5 and CCR5 mRNA levels were significantly higher in GB and GB rec samples compare to non-cancerous brain tissues ( Figure 7A,B). Primary glioblastoma cells expressed higher levels of CCL5 mRNA, compared to GSCs what also correlated with protein levels seen in Figures 2 and 3.  CCR5 (B) at mRNA was determined in glioma, non-cancerous brain tissues and glioblastoma cells analyzed by RT-qPCR. mRNA values were normalized to housekeeping genes HPRT1 and GAPDH and analyzed with quantGenius software [40] as described in Materials and Methods. n-number of samples; N-non-cancerous brain tissues; glioma I-II-lowgrade gliomas: pilocytic astrocytoma, astrocytoma, oligodendroglioma; glioma III-anaplastic astrocytoma, anaplastic oligodendroglioma, and anaplastic mixed oligoastrocytoma; GBglioblastoma; GB rec-recurrent glioblastoma; GB cells-primary glioblastoma cells; GSC-glioblastoma stem cells isolated from patient tumor samples. ★★ p < 0.01, ★★★ p < 0.001 versus the non-cancerous brain tissues.

CCL5 and CCR5 mRNA Levels Differ Among Glioblastoma Subtypes
We analyzed the CCL5 and CCR5 mRNA levels in four GB subtypes, MES (Mesenchymal), PN (Proneural), CL (Classical), and MIX, based on the expression values of 12 genes from Behnan et al. (2016) [41]. PN subtype was classified with expression levels of P2RX7, STMN4, SOX10, and ERBB3 genes. CL subtype was classified with expression levels of ACSBG1 and KCNF1 and MES subtype with expression levels of S100A, DAB2, TGFB1, THBS1, COL1A2, and COL1A1, as described in Materials and Methods. Cl subtype exhibited the highest level of both CCL5 and CCR5 mRNA expressions, while MES expressed the lowest (Figure 8).  CCR5 (B) at mRNA was determined in glioma, non-cancerous brain tissues and glioblastoma cells analyzed by RT-qPCR. mRNA values were normalized to housekeeping genes HPRT1 and GAPDH and analyzed with quantGenius software [40] as described in Materials and Methods. n-number of samples; N-non-cancerous brain tissues; glioma I-II-low-grade gliomas: pilocytic astrocytoma, astrocytoma, oligodendroglioma; glioma III-anaplastic astrocytoma, anaplastic oligodendroglioma, and anaplastic mixed oligoastrocytoma; GB-glioblastoma; GB rec-recurrent glioblastoma; GB cells-primary glioblastoma cells; GSC-glioblastoma stem cells isolated from patient tumor samples.

CCL5 and CCR5 mRNA Levels Differ among Glioblastoma Subtypes
We analyzed the CCL5 and  [41]. PN subtype was classified with expression levels of P2RX7, STMN4, SOX10, and ERBB3 genes. CL subtype was classified with expression levels of ACSBG1 and KCNF1 and MES subtype with expression levels of S100A, DAB2, TGFB1, THBS1, COL1A2, and COL1A1, as described in Materials and Methods. Cl subtype exhibited the highest level of both CCL5 and CCR5 mRNA expressions, while MES expressed the lowest (Figure 8).
(2016) [41]. PN subtype was classified with expression levels of P2RX7, STMN4, SOX10, and ERBB3 genes. CL subtype was classified with expression levels of ACSBG1 and KCNF1 and MES subtype with expression levels of S100A, DAB2, TGFB1, THBS1, COL1A2, and COL1A1, as described in Materials and Methods. Cl subtype exhibited the highest level of both CCL5 and CCR5 mRNA expressions, while MES expressed the lowest (Figure 8).

Discussion
The CCL5/CCR5 axis has been reported as a mechanism of tumor progression in pancreatic [19], gastric [23], and breast cancer [42]. The CCL5-receptors' signaling can favor cancer progression, directly affecting proliferation, migration, and cell survival of cancer cells by autocrine signaling, or indirectly by paracrine signaling recruiting pro-tumor and/or anti-inflammatory effector cells into the tumor microenvironment (TME) [43].
The basic question when investigating chemokine autocrine signaling in cancer, such as presented by CCL5/CCR5 axis, is "what activates what," whereas in paracrine signaling in heterogeneous cancers the question is "what attracts what." Autocrine signaling means that GB cells express both ligand and receptor, and thus activate the pathways downstream of CCR5 in a cell-autonomous manner. Therefore, we first need to reveal the CCL5/CCR5 distribution in patients' tissues with respect to glioma stage and glioblastoma subtype, and secondly CCL5 and CCR5 relative expressions in the isolated primary glioblastoma cell lines. By analyzing CCL5/CCR5 mRNA and protein expression in glioma tissues in a larger cohort of 65 patients, we confirmed that both, CCL5 and CCR5 genes are increasingly expressed in advanced glioma ( Figure 7). Moreover, higher CCL5 and CCR5 were detected in secondary, recurrent glioblastoma as compared to the primary glioblastoma. As the recurrences are known to be more aggressive [44], we suggest that CCL5 and CCR5 autocrine signaling is playing a significant role in glioblastoma progression. Secondly, we detected higher CCL5 and CCR5 proteins in glioblastoma tissues and cells than in non-malignant brain tissues and normal astrocytes (Figures 1 and 2).
Our results are consistent with studies that showed increased CCL5 and CCR5 expression in human glioma tissues [45,46] and glioblastoma cells [47], compared to normal counterparts. Our finding of increased CCR5 expression in worse prognosis glioma are consistent with reports that high CCR5 levels correlate with shorter survival [28]. Similarly, CCL5 was demonstrated as a bad prognostic marker for survival of patients with various types of cancer [19]. As in low-grade gliomas, for which we have shown low expression of both CCL5 and CCR5, we may speculate that the progression probably relays more on stromal cells' supportive chemokine stimulation. Whereas in glioblastoma, high levels of CCL5/CCR5 enable an autocrine chemokine activation, resulting in increased tumor cell proliferation and invasion [45,46] that is becoming independent of stromal cells.
This is leading to a lower survival rate of glioblastoma patients, as suggested by Pan et al. [45]. These authors showed that CCL5 established autocrine signaling in high-grade glioma by affecting growth regulatory circuit that was critical in particular for mesenchymal (MES) glioblastoma subtype.
In contrast to expected, in our group of patient-derived GB tissues, CCL5 and CCR5 proteins were not correlating, as each may be present in some specimen, but absent in others, and they were found in a different subcellular compartment. This indicates on non-exclusive partnering, but also the promiscuous binding of CCL5 and CRR5 in tumors in vivo [45,47,48].
As mentioned above, recurrent glioblastoma supposedly acquires aggressive mesenchymal phenotype, being possibly induced by irradiation [44] via epithelial to mesenchymal transition (EMT). Such phenotype is more invasive and tends to express higher stemness-related genes (Majc et al., accepted, 2020) [49]. The dilemma of autocrine signaling in glioblastoma is therefore also related to glioblastoma cell subtypes. However, exploring the gene expression distribution of CCR5 and CCL5 among different genetic subtypes, we found the highest levels of both genes in the CL-glioblastoma subtype and the lowest in MES-glioblastoma subtype (Figure 8). This contrasts to the results by Pan et al. [45], demonstrating the highest CCL5 (gene) expression in MES-glioblastoma and the lowest in PN-glioblastoma. This discrepancy could be due to the low number of samples in each subtype group in our study, and using smaller [41] vs. larger panel of gene fingerprints defining the subtype [7] by us than by Pan et al. [45]. Moreover, in MES-glioblastoma, the autocrine CCL5-dependent activation loop has also been proven by adding exogenous CCL5, and because no further activation was achieved, it was concluded that CCL5 promotes survival and proliferation of the cells in a cell-autonomous manner. Noteworthy, MES-subtypes characteristically express the CD44 a non-conventional CCL5 receptor, also a stemness marker [50]. As CCL5 is a promiscuous ligand, binding to more than one receptor [46], several receptors need to be blocked to inhibit CCL5 driven axis processes in brain tumors.
There are three major reasons for poor survival: (1) increased tumor cell invasion, (2) the abundance of more aggressive glioblastoma stem cells (GSCs), and (3) supportive stromal cells in TME. Firstly, glioblastoma invasion is characterized by extensive single-cell infiltration into healthy brain tissue, preventing total tumor removal during surgery [51]. Increased invasion of glioblastoma cells could also be activated by CCL5/CCR5 signaling the migratory downstream pathways through αvβ3 integrin, PI3K/Akt kinases, NF-κB pathways [52], and proteases such as matrix metalloproteases (MMPs) [53]. However, protease inhibitors, such as MMPs and cathepsin inhibitors failed to inhibit invasive cancer spread in clinical trials. Thus CCL5/CCR5 axis blocking agents were suggested as efficient anti-invasive therapeutics [54]. Both ligand CCL5 and its receptor CCR5 have been suggested as potential therapeutic targets in various cancers, including glioblastoma, breast and prostate cancer, and impairing disease progression [28,55]. Most promising is CCR5 blocking by synthetic drug maraviroc, an allosteric inverse CCR5 agonist [56], which has been proven to significantly inhibit proliferation, colony formation, and migration of several carcinomas, including breast [42] and prostate cancer [55]. Maraviroc has very recently been reported also in metastasis of breast cancer cells xenografts [18,20,22,42]. Here, we demonstrated that CCR5-expressing primary glioblastoma cells and glioblastoma stem cells (GSC) invasion, when enhanced by recombinant CCL5, was also significantly inhibited by adding maraviroc.
Secondly, we demonstrated that maraviroc inhibited glioblastoma stem cell (GSCs) invasion. This is an important novelty of this research, as GSCs are recognized as a key target of therapy in glioblastomas and all other cancers, as these are cancer stem cells (CSCs) and are highly resistant to irradiation and chemotherapy. As CSCs represent the tumor-initiating cells, i.e., the seed of primary and the secondary tumors metastases, these are the cells that need to be eradicated by a novel kind of therapy. High levels of GSCs in glioblastoma were observed in more aggressive tumors vs. low-grade glioma, as reported by us and others [57,58] and their abundance is related to prognosis.
These cells are trafficking within the tissues into and out their niches [12] and presumably invade into the brain parenchyma, based on the chemoattraction among the tumor and stromal cells, as has been demonstrated for CXCR12/SDF-1α [31,59]. As maraviroc significantly inhibited CCL5-induced GSC invasion (Figure 6), we propose targeting CCL5/CCR5 signaling as novel glioblastoma therapeutics, as initially suggested by Kast et al. [60]. Moreover, we show that GSCs express only CCR5, but not CCL5 (Figure 3), indicating that only paracrine signaling would stimulate GSC invasion. This may be occurring in vivo, as dormant GSCs reside in glioblastoma tissue niches and are presumably activated in a paracrine cross-talk by stromal cells, infiltrating the niche to migrate out of the niches [12,59], expressing CCL5 and CCR5.
Thirdly, glioblastoma TME consists beside brain tissue astrocytes and microglia, also from infiltrating immune cells, macrophages, lymphocytes, neutrophils, and mesenchymal stem cells (MSCs) that interact in complex networks of molecular signals [43,61], where chemokines are the key molecules for directing the cells to move along a chemical gradient towards the tumor [62]. We are still far from understanding complex multiple interactions under in vivo conditions, however, by categorically studying bilateral ligand and receptor expressions by selected cell types, their specific mechanisms in CCL5/CCR5 signaling in glioblastoma may be elucidated. Here, we focused on MSCs, proven as glioblastoma-infiltrating cells, recruited from bone marrow or brain tissues, and also present in GSC niches, where MSCs may also affect glioblastoma cell differentiation and proliferation as well as invasion, as proven by us and others [12,31,36,63].
We demonstrated that paracrine MSC-glioblastoma and GSC cell interactions enhance invasion, maintained by CCR5 receptors, as it was inhibited when maraviroc was added to the system ( Figure 6). Our extensive previous research [35,36,64], provided sufficient evidence by quantifying a set of chemokines released from bone-marrow MSCs in indirect co-cultures and glioblastoma cells. MSC have been demonstrated to secrete among other chemokines, also CCL5, which interacts with specific cytokine receptors such as CCR1, CCR3, and CCR5. CCL5 paracrine signaling was found to promote the migratory, invasive, and metastatic properties of breast cancer cells [24]. Similar was later confirmed by Choi et al. [65] demonstrating that also adipose MSCs target brain tumor-initiating cells from glioblastoma, medulloblastoma, and ependymoma, by releasing potential cytokines, including CXCR4/SDF-1alpha, CCR5/RANTES, IGF1R/IGF-1, IL6R/IL-6, and IL8R/IL-8.
Complementary to this, we demonstrated here the bilateral ligand and receptor expression on glioblastoma tissue sections using the specific markers of CD105 for MSCs and CD68 for macrophages. We showed that MSCs express ligand CCL5 (Figure 4) and macrophages receptor CCR5 ( Figure 5). These results further suggest that the CCL5/CCR5 axis may mediate cellular cross-talk between MSCs, macrophages, and GSCs by attracting them to peri-vascular tumor niches, that are populated by MSCs. The involvement of the CCL5/CCR5 axis in MSC-GB cell interactions has not been known so far, in comparison to the well-known pro-migratory role of macrophage-secreted CCL5 [53]. Finally, we hypothesize that MSC-secreted CCL5 maintains the interactions between MSCs and GSCs, and targeting the CCL5/CCR5 axis with maraviroc may become effective anti-invasive therapy preventing invasive GSCs migration out of their niches to spread to brain parenchyma.
In conclusion, we have demonstrated the heterogeneous tissue/cellular distribution and subcellular expression of CCL5 and CCR5 in glioblastoma. Using the CCR5 antagonist, maraviroc, we have shown CCL5 and CCR5 drive primary glioblastoma (GB) cells and glioblastoma stem cells (GSCs) invasion and their interactions with stromal MSCs and can be used as repositioned drug for novel clinical trials in glioblastoma. These results suggest paracrine and autocrine CCL5/CCR5 axis-dependent signaling in a lower grade (gliomas) vs. higher grade glioblastoma invasion. The potential role of CCL5/CCR5 in paracrine GSC niche interactions warrants further investigations. Culture Collection (ATTC, Manassas, VA, USA) and were grown in DMEM high glucose medium (GE Healthcare, Il, Chicago, IL, USA), supplemented with 10% (v/v) FBS, 2 mM L Glutamine, 100 IU/mL penicillin and 100 µg streptomycin, as described in Kološa et al. [63] and Breznik et al. [36]. Glioblastoma stem cell lines, NCH644 and NCH421k were obtained from CLS (Cell Lines Service GmbH, Eppelheim, Germany) and grown as spheroid suspensions in complete Neurobasal Medium (Invitrogen, Life Technologies, Carlsbad, CA, USA) containing 2 mM L-glutamine, 1 × penicillin/streptomycin, 1 × B-27 (Invitrogen, Life Technologies, Carlsbad, CA, USA), 1 U/mL heparin (Sigma-Aldrich, St. Louis, MO, USA), 20 ng/mL bFGF and EGF (both from Invitrogen, Life Technologies, Carlsbad, CA, USA). All cell lines were maintained at 37 • C with 5% CO 2 and 95% of humidity. All cell cultures were tested for mycoplasma contamination using MycoAlert Mycoplasma Detection Kit (Lonza, Basel, Switzerland).

Glioblastoma Samples from Patients
Glioma biopsies were obtained from 65 patients that operated at the Department of Neurosurgery, University Medical Centre of Ljubljana, Slovenia. Tumor tissue samples were snap-frozen in liquid nitrogen and stored in the liquid nitrogen for RNA/DNA analyses. The study was approved by the National Medical Ethics Committee of the Republic of Slovenia (approval no. 0120-179 190/2018/4). Patients with glioblastoma (glioma grade IV) were selected for this study ( Table 2). The clinical parameters and tumor characteristics were provided by the Department of Neurosurgery and Institute of Pathology at medical faculty in Ljubljana (Table 2). Formalin-fixed, paraffin-embedded tissues were prepared at the Institute of Pathology and were used for immunohistochemical analyses. Non-cancer brain samples (NB1 and NB2) were also obtained from the Institute of Pathology, from patients who were brain cancer-free.

Establishment of Primary Glioblastoma and Glioblastoma Stem Cell Lines
Fresh glioblastoma tumor tissue samples were minced by scalpels in DMEM/high glucose cell culture media supplemented with 10% FBS, 2 mM L-glutamine, and penicillin-streptomycin and seeded in 6 well plates. Outgrowing cells were detached with 0.25% trypsin-EDTA solution (Sigma-Aldrich, St. Louis, MO, USA) and transferred to T25 cell culture flasks. Cells were collected by low-speed centrifugation (1000 rpm for 60 s). After centrifugation 2-3 times, the cells were transferred to T75 culture flasks and expanded for subsequent analyses.
Cells' solution was further filtered through Nylon mash 40 µm pores (BD Falcon cell strainer, Nylon). Single cells were collected and resuspended in stem cell media, Neurobasal Medium (Invitrogen, Life Technologies, Carlsbad, CA, USA) containing 2 mM L-glutamine, 1 × penicillin/streptomycin, 1 × B-27 (Invitrogen, Life Technologies), 1 U/mL heparin (Sigma-Aldrich, St. Louis, MO, USA), 20 ng/mL bFGF and EGF (both from Invitrogen, Life Technologies, Carlsbad, CA, USA) and cultured on agar coated T25 flasks until spheres with a diameter of 200 µm were formed. Healthy spheres were frozen in stem cell media with 10% DMSO for further analysis. GSCs were authenticated for stem cell marker CD133 and SOX2 expression using immunofluorescence.

Immunohistochemistry and Immunocytochemistry
Immunohistochemistry (IHC) analyses were performed using antibodies against CCR5 (ab65850, Abcam, Cambridge, UK), CCR5 peptide (ab192862, Abcam, Cambridge, UK), and CCL5-RANTES (ab189841, Abcam, Cambridge, UK). After fixation, tumor sections (4 µm thick) were deparaffinized in xylene and rehydrated in ethanol. Antigen retrieval was carried out in 10 mM sodium citrate buffer (pH 6.0) at 95 • C for 20 min followed by 20-min cooling on ice. The sections were treated with 100% methanol (Merck, Kenilworth, NJ, USA) containing 0.3% H 2 O 2 (Merck, Kenilworth, NJ, USA) for 10 min to block endogenous peroxidase activity to reduce non-specific background staining, followed by a washing step in distilled water. Non-specific binding sites were blocked with 1% bovine serum albumin with 2% goat serum in PBS before incubation with antibodies overnight in the fridge. The sections were incubated with biotinylated secondary antibody followed by horseradish peroxidase-conjugated streptavidin (Cell Signaling Technology, Danvers, MA, USA). The sections were further incubated with the 2,4-diaminobenzidine substrate and counterstained with hematoxylin. Immunocytochemistry was performed as described without the deparaffinization and antigen retrieval. To achieve high antibody specificity, we used CCR5 blocking peptide (CCR5 P) that binds specifically to the target antibody epitope in 10 times higher concentration as the primary CCR5 antibody. IHC scoring was performed by a pathologist using a semi-quantitative grading system; immunostaining intensity: +++ strong, ++ moderate, + weak or no expression, and the abundance of stained cells percentage positive cells: 0, 1-33% = 1, 33-66% = 2, 66-100% = 3. The intracellular localization was evaluated as m = membrane, n = nuclear, c = cytoplasmic; e = extracellular.

Immunofluorescence of Glioblastoma Stem Cell Spheroids
The 3D GSC spheroids were washed with PBS, fixed in ice-cold methanol (Sigma-Aldrich, St. Louis, MO, USA) for 15 min at room temperature, and incubated for 15 min in 0.1% Triton X-100/1% FBS/PBS at room temperature for membrane permeabilization. The spheroids were stained for 30 min at room temperature with the following antibodies: CCR5 (ab65850, Abcam, Cambridge, UK ) and CCL5-RANTES (ab189841, Abcam, Cambridge, UK). Negative control staining was performed in the absence of the primary antibodies. Spheroids were stained with an Alexa Fluor 488 ® -and Alexa Fluor 546 ® -conjugated secondary antibody (1:200; Invitrogen, Life Technologies, Carlsbad, CA, USA ) for 30 min at room temperature. For nuclear staining, the spheroids were incubated with the Hoechst 33342 dye (1:1000, Invitrogen, Life Technologies), for 5 min at room temperature. The spheroids were then mounted in AntiFade reagent (Invitrogen, Life Technologies, Carlsbad, CA, USA) and analyzed with a confocal microscope (Leica DFC 7000 T, Wetzlar, Germany).

Immunofluorescence of Glioblastoma Tumor Tissue Sections
Tumor sections, prepared at the Institute of Pathology, Medical Faculty, were deparaffinized in xylene and rehydrated in ethanol. Following rehydration, antigen retrieval was carried out in 10 mM sodium citrate buffer (pH 6.0) at 95 • C for 20 min followed by 20-min cooling on ice.

Invasion Assay
Primary glioblastoma cell (GB) and glioblastoma stem cell (GSC) invasion was measured using 24-well Transwell units with 6.5 mm inserts and 8 µm pores (Corning, New York, NY, USA). Primary GB from patient 2 (10,000/insert) and GSC (NCH644, 80,000/insert), were seeded in the upper compartment, which was coated with 0.5 mg/mL Matrigel (Becton Dickinson, Franklin Lakes, NJ, USA) in serum-free medium. The lower compartment was seeded with MSC (20,000/insert) in MSC media containing 10% FBS or with recombinant CCL5/RANTES peptide (R&D, 278-RN-050, Minneapolis, MN, USA) (300 ng/mL). Maraviroc (MVR, Selleckchem, S200, Houston, TX, USA) in a final concentration of 10 µM was added into the upper chamber to GB and to GSCs. Cells were allowed to invade at 37 • C in 5% CO2 for 48 h. Non-invading cells were removed from the upper surface of the membrane using a cotton swab. The lower surface of the membrane was fixed in 4% PFA, stained with 0.1% crystal violet, and stained cells were counted using the Nikon Eclipse Ti-inverted microscope (Nikon Instruments, Melville, NY, USA) at 4× magnification. Three biological experiments with two separate membranes for each condition were analyzed.

Gene Expression Analysis
Total RNA from glioblastoma tissues and cells was isolated using AllPrep DNA/RNA/Protein Mini Kit (Qiagen, MD, USA) according to the manufacturer's instruction. 1 µg of RNA was reverse transcribed using a High-Capacity cDNA Reverse Transcription Kit (Thermo Fischer Scientific, Waltham, MA, USA). High-throughput RT-qPCR was used to measure CCL5, CCR5 expression. RT-qPCR was performed with FAM-MGB probes with Fluidigm BioMark HD System RT-PCR (Fluidigm Corporation, San Francisco, CA, USA) using 48.48 Dynamic Arrays IFC [66], where 42 samples and 24 assays (probes) were mixed pairwise in nanoliter chambers to enable parallel analysis of 2304 reactions.
Visualization and analysis of qPCR results were done using the Fluidigm RT-qPCR analysis software and quantGenius software [40]. Relative copy numbers of mRNA were normalized to housekeeping genes HPRT1 and GAPDH. Assays are described in Table S1.

Glioblastoma Subtyping
Firstly, we assessed whether the expression profiles of 12 selected genes (COL1A2, COL1A, TGFB1, THBS1, DAB2, S100A4, P2RX7, STMN4, SOX10, ERBB3, ACSBG1, KCBF1) from 4 sample types (GB-glioblastoma; GB rec-recurrent glioblastoma; GB cells-primary glioblastoma cells; GSC-glioblastoma stem cells) are suitable markers for GB subtype distinction into mesenchymal (MES), proneural (PN), classical (CL) subtype and finally the subtype combination (MIX). Since the number of subtypes (clusters) was known in advance, we used k-means clustering to partition the expression profiles of the selected genes in one of the four subtypes. K-means clustering partitions each gene to the subtype (cluster) with the nearest mean. Data was first standardized. We used two clustering techniques, k-means, and PAM (partition around medoids).
The difference between the two is that k-means uses artificially calculated means, while PAM uses the so-called medoids, which are actual dataset values. PAM is also more robust. The cluster (subtype) assignment for each gene was then compared and the method which shows more concordance with the subtype assignment from clinical data (EGFRIII mut, IDH mut, PFGFR, p53 status) was selected (in our case this was the k-means clustering). When the analyses were repeated by removing genes with extreme values (only 2 such genes), the results did not significantly change. All analyses were done in R version 3.6.1 and its libraries fpc (used for visualizations) [67] and cluster (used for PAM clustering) [68].

Differentially Expressed Genes among Tissues and Glioblastoma Subtypes
We analyzed the differences in the expression of CCL5 and CCR5 among sample types; N-non-cancerous brain tissues; glioma I-II-low-grade gliomas; glioma III; GB-glioblastoma; GB recrecurrent glioblastoma; GB cells; GSCs in the first analysis and between previously defined subtypes (mesenchymal-MES, proneural-PN, classical-CL subtype and finally the subtype combination-MIX) in the second analysis. To minimize the effect of genes with a low expression we first removed them from the analysis by replacing the Ct values > 40 as zero. We then assessed the overall similarity of genes/samples (sample types in the first analysis and subtypes in the second one) by hierarchical clustering and plotted the heatmaps for visual inspection of the results. The differential expression analyses were done using linear models and the empirical Bayes method to moderate the standard errors of the log of fold changes that were estimated with the linear model. In the first set of analyses we tested which genes differentially expressed when glioma or glioblastoma samples were compared with normal tissue when the gene expression in glioblastoma cells and GSCs was compared to normal astrocytes (NAS) when gene expression in glioblastoma samples was compared with N, glioma I and II, glioma III, and GB-rec and finally when gene expression in GSCs was compared with NAS and GB-cells. In the second experiment, samples were categorized according to their subtypes and the difference in gene expression between every pair of subtypes was tested. To enable an easier estimation of genes that were differentially expressed in two or more analyses we used Venn diagrams. All analyses were done in R version 3.6.1 and its libraries HTqPCR (used for data preprocessing and visualizations) [69], limma (used for differential expression calculations and Venn diagram visualizations) [70].