Ccl5 establishes an autocrine high-grade glioma growth regulatory circuit critical for mesenchymal glioblastoma survival

Glioblastoma (GBM) is the most common malignant brain tumor in adults, with a median survival of 15 months. These poor clinical outcomes have prompted the development of drugs that block neoplastic cancer cell growth; however, non-neoplastic cell-derived signals (chemokines and cytokines) in the tumor microenvironment may also represent viable treatment targets. One such chemokine, Ccl5, produced by low-grade tumor-associated microglia, is responsible for maintaining neurofibromatosis type 1 (NF1) mouse optic glioma growth in vivo. Since malignant gliomas may achieve partial independence from growth regulatory factors produced by non-neoplastic cells in the tumor microenvironment by producing the same cytokines secreted by the stromal cells in their low-grade counterparts, we tested the hypothesis that CCL5/CCL5-receptor signaling in glioblastoma creates an autocrine circuit important for high-grade glioma growth. Herein, we demonstrate that increased CCL5 expression was restricted to both human and mouse mesenchymal GBM (M-GBM), a molecular subtype characterized by NF1 loss. We further show that the NF1 protein, neurofibromin, negatively regulates Ccl5 expression through suppression of AKT/mTOR signaling. Consistent with its role as a glioblastoma growth regulator, Ccl5 knockdown in M-GBM cells reduces M-GBM cell survival in vitro, and increases mouse glioblastoma survival in vivo. Finally, we demonstrate that Ccl5 operates through an unconventional CCL5 receptor, CD44, to inhibit M-GBM apoptosis. Collectively, these findings reveal an NF1-dependent CCL5-mediated pathway that regulates M-GBM cell survival, and support the concept that paracrine factors important for low-grade glioma growth can be usurped by high-grade tumors to create autocrine regulatory circuits that maintain malignant glioma survival.


INTRODUCTION
Malignant gliomas (glioblastoma or GBM) are the most common brain tumor in adults, with approximately over 13,000 individuals in the United States dying from their cancer each year. After initial diagnosis, the median survival is only 9 months [1,2], which improves to ~15 months following surgical resection, radiation and chemotherapy [3]. In an effort to improve clinical outcomes for people with this deadly cancer, numerous conventional and biologically-based anti-neoplastic treatments have been evaluated that specifically target the rapidly dividing cancer cells. Unfortunately, these newer therapies have resulted in only modest increases in overall survival, highlighting the urgent need to identify more effective therapies for these brain malignancies.
Over the past decades, numerous reports have described an instructive role for the tumor microenvironment in glioma formation and maintenance [4]. In this regard, glioblastomas contain a rich nonneoplastic component, including endothelial cells and immune system cells (e.g., microglia and macrophages), whose function is dictated by small molecules responsible for their recruitment and activation. One class of these proteins, chemokines, is critical for microglia and macrophage infiltration and activation through the Research Paper expression of chemokine receptors on their plasma membranes. Consistent with an ecosystem paradigm in which non-neoplastic cells attracted by cancer cells in turn support neoplastic cell growth, inhibition of chemokine expression or chemokine receptor function has been shown to attenuate both low-grade and high-grade glioma growth in numerous experimental preclinical model systems in vivo. For example, CCL2 recruits regulatory T cells and myeloid-derived suppressor cells to increase GBM growth [5], as well as indirectly increases glioma invasiveness by stimulating IL-6 secretion from tumor-associated microglia (TAM) [6]. In addition, CXCL12 and its receptors CXCR4/CXCR7 function by maintaining GBM stem-like cell survival and increasing glioma cell invasion [7,8]. Similarly, colony stimulating factor-1 (CSF-1) creates a supportive glioma microenvironment, such that silencing CSF-1-receptor signaling on macrophages is sufficient to reduce high-grade glioma growth in mice [9,10]. Together, these experimental observations establish a fundamental role for chemokines in glioma biology.
In addition to the chemokines mentioned above, we have recently discovered that Ccl5 produced by tumorassociated microglia is a potent paracrine regulator of neurofibromatosis type 1 (NF1)-associated murine lowgrade optic glioma growth [11]. In this experimental system, Ccl5 was sufficient to stimulate Nf1-deficient optic nerve glial cell proliferation in vitro, whereas Ccl5 antibody-mediated inhibition attenuated mouse optic glioma proliferation in vivo. Interestingly, CCL5 has recently emerged as a key member of an autocrine cytokine circuit critical for KRAS-driven lung cancer growth [12]. Since the NF1 protein, neurofibromin, functions as a negative regulator of RAS, specifically K-RAS in astroglial cells [13,14], and NF1 mutation/ loss characterizes the mesenchymal GBM (M-GBM) molecular subtype [15], we sought to determine whether Ccl5 expression creates a unique autocrine circuit for mesenchymal glioblastoma growth.
In this report, we describe for the first time, that increased CCL5 expression is enriched in both human and mouse M-GBM with NF1 loss. Leveraging Nf1-deficient murine M-GBM models, we further demonstrate that Ccl5 shRNA-mediated knockdown increases glioblastoma cell apoptosis in vitro, as well as extends mouse glioblastoma survival in vivo. Moreover, neurofibromin negatively regulates Ccl5 expression through RAS/AKT/mTOR pathway activation. Lastly, Ccl5 produced by the M-GBM cells acts in a cell autonomous fashion to increase M-GBM cell survival using a non-conventional CCL5 receptor, CD44, to suppress programmed cell death. Collectively, these findings broaden the concept that growth regulatory factors produced by non-neoplastic cells in the low-grade glioma microenvironment are employed to establish autocrine circuits in high-grade gliomas, thus facilitating a state of relative stromal independence.

Ccl5 functions in a cell-autonomous fashion to increase M-GBM survival
Based on our previous studies examining CCL5 expression and function in low-grade glioma, we first sought to determine whether CCL5 expression was higher in human GBM tumors relative to non-neoplastic brain. Using the GSE16011 [16] dataset, for which normal brain was available for comparison, CCL5 mRNA levels were higher in GBM samples relative to their non-neoplastic counterpart ( Figure 1A). We next analyzed CCL5 mRNA expression in the four GBM molecular subtypes using data from the Cancer Genome Atlas (TCGA, provisional), and found that mesenchymal GBM (M-GBM) exhibited the highest levels of CCL5 mRNA expression, while proneural GBM (PN-GBM) expressed the lowest ( Figure 1B).
This relative enrichment of CCL5 expression in the mesenchymal subtype, which frequently exhibits loss of function mutations in the NF1 tumor suppressor gene [15], prompted us to focus on murine models characterized by Nf1 loss. Using two representative mouse M-GBM cell lines (1861 and 4622), which lack Nf1 and Tp53 expression [17], and two representative mouse proneural GBM cell lines ((2)61 and (5)54), which retain neurofibromin expression ( Figure 1C) [18], we found that Ccl5 levels were elevated in the culture medium of mouse M-GBM cells relative to PN-GBM cells, as well as to wild-type primary mouse astrocytes ( Figure 1D).
To determine whether M-GBM cell growth was dependent on Ccl5 expression, 1861 (M-GBM) cells were grown in the presence of mouse Ccl5 (mCcl5). mCcl5 treatment increased cell growth, as revealed by ~14% increase in BrdU incorporation relative to the vehicle treatment ( Figure 2A). Next, we reduced Ccl5 levels in 1861 cells by shRNA-mediated knockdown (KD). Using two independent constructs (shCcl5-1 and shCcl5-2), Ccl5 protein levels were reduced to ~20% of control levels (LacZ shRNA; Figure 2B), resulting in reduction in cell number ( Figure 2C) and cell growth (BrdU incorporation; Figure 2D) as a consequence of increased apoptosis (increased cleaved caspase-3 expression and %TUNEL + cells; Figure 2E-2G). Similar results were obtained using another M-GBM cell line (4622; Supplementary Figure 1A-1E). Importantly, treating PN-GBM cells with mCcl5 did not affect cell growth, suggesting that Ccl5 is not a regulator for PN-GBM growth (Supplementary Figure 1F).
Given the role of Ccl5 as a chemokine, it may recruit monocytes to the implantation site to facilitate tumor growth; however, we found no change in the number of recruited monocytes or chemokine expression within the tumor between the control and Ccl5 KD groups (data not shown). In addition, we sought to determine whether Ccl5 produced by non-neoplastic cells contributes to M-GBM tumor growth by comparing M-GBM cell growth in wild-type versus Ccl5−/− mouse striata. Mouse survival and tumor proliferation were similar between the two groups, arguing against a major stromal source for Ccl5 in mediating the observed M-GBM survival effects (Supplementary Figure 3). Collectively, these observations demonstrate that Ccl5 positively regulates M-GBM growth in vivo and in vitro through an autocrine stimulatory mechanism.
Neurofibromin regulates glioma cell growth through the AKT/mTOR pathway [20,21], such that Nf1−/− astrocytes exhibit AKT/mTOR hyperactivation relative to Nf1+/+ (wild-type) astrocytes (Supplementary Figure 4A). Consistent with this neurofibromin regulatory function, AKT/mTOR activity in 1861 cells was reduced following NF1-GRD expression (Supplementary Figure 4C). Moreover, AKT pharmacological inhibition (MK2206) of M-GBM cells similarly reduced AKT activity ( Figure 4C, 4D) and Ccl5 expression ( Figure 4E). In addition, AKT inhibition in 1861 cells reduced mTOR activity ( Figure 4F, 4G), and mTOR inhibition using rapamycin ( Figure 4F, 4G) reduced Ccl5 expression ( Figure 4H). The concentrations and durations of inhibitor treatments were specifically selected so that cell growth was not affected (Supplementary Figure 4D), arguing that any potential changes in Ccl5 expression were a secondary effect of suppressed cell growth. As a complementary approach, we employed shRNA knockdown of Raptor, a molecule critical for the function of mTOR complex-1 ( Figure 4I). Similar to rapamycin treatment, Raptor KD reduced Ccl5 expression ( Figure 4J). However, in these experiments, cell growth was inhibited following Raptor KD (Supplementary Figure 4E), requiring that the amount of Ccl5 present in the medium be normalized to the total cell number.
Since MEK/ERK activation is also elevated following NF1 loss [22], we analyzed Ccl5 expression following PD-0325901 (PD901, MEK inhibitor) treatment. In these experiments, MEK inhibition did not change Ccl5 expression (Supplementary Figure 5). Collectively, these experiments establish RAS-GAP/AKT/mTOR signaling as the primary mechanism underlying neurofibromin regulation of Ccl5 expression.

Ccl5 controls M-GBM survival through CD44
CCL5 functions by binding to G-protein coupled receptors, including CCR (C-C chemokine receptor) 1, 3 and 5. Surprisingly, Ccr1, Ccr3 or Ccr5 mRNA levels were nearly undetectable in M-GBM cells (Supplementary Figure 6), prompting us to investigate non-conventional CCL5 receptors (GPR75, CXCR4 and CD44) as potential binding partners [23,24]. In these experiments, only Cd44 mRNA was detected in 1861 and 4622 cells ( Figure 5A), which also observed at the protein level ( Figure 5B). Importantly, CD44 mRNA expression in human GBM subtypes was highest in the mesenchymal group, and was lowest in the proneural group ( Figure 5C).
To determine whether CD44 was necessary for M-GBM cell survival, we reduced Cd44 expression in 1861 cells using shRNA. Following Cd44 KD, there was increased apoptosis (cleaved caspase-3 expression and %TUNEL + cells) relative to controls (LacZ KD) ( Figure 5D, 5E). In order to demonstrate that CD44 functions as the receptor for Ccl5 in 1861 cells, Cd44 KD cells were treated with mCcl5. Following Cd44 KD, mCcl5 could no longer increase 1861 cell growth ( Figure 6A). Similarly, Ccl5 KD in Cd44 KD cells ( Figure 6B) did not change cell growth ( Figure 6C), establishing that the growth-promoting effects of Ccl5 require CD44 expression on M-GBM cells. In order to test whether other CD44 ligands function in a similar fashion as Ccl5, we reduced Cd44 using shRNA in Ccl5 KD cells (shCcl5 + shCd44), and compared cell proliferation to that observed in Ccl5 KD (shCcl5) and control (shLacZ) cells. It should be noted that the shLacZ and shCcl5 cells were further infected with virus carrying the shLacZ construct as an internal control for a second round of viral infection. We observed no further reduction of cell proliferation in Ccl5 KD cells following Cd44 knockdown ( Figure 6D-6F), suggesting that this prosurvival pathway is specific to the interaction between Ccl5 and CD44, and is not regulated by other CD44 ligands. Taken together, these data support a model in which Nf1 loss in M-GBM cells leads to hyperactivation of the AKT/mTOR pathway, which positively regulates Ccl5 expression and increases M-GBM cell survival through CD44 signaling ( Figure 6G).

DISCUSSION
Chemokines were originally described as key signaling molecules in cancer biology by virtue of their ability to recruit stromal cells (e.g., macrophages, endothelial cells) to the developing tumor [25,26]. While this is one important function for these immune modulatory proteins, emerging evidence suggests that chemokines and cytokines may create autocrine growth regulatory circuits that facilitate tumor expansion. In this regard, CXCR2 signaling has been shown to regulate KRAS-induced pancreatic cancer growth [27], whereas IL-8 serves as an autocrine growth factor for malignant mesothelioma maintenance [28]. In the context of glioma, CXCL12/CXCL12-receptor (CXCR4/CXCR7) autoregulatory signaling has been demonstrated to be important for glioblastoma survival [8,29,30] and angiogenesis [31], while CX3CL1/CX3CR1 axis signaling in glioma cells controls glioma cell invasion [32].
The emergence of such autoregulatory chemokine/ chemokine receptor loops relevant to glioma growth prompted us to examine a recently-identified paracrine signal (CCL5), produced by low-grade glioma-associated microglia [11], as a potential autocrine signaling molecule in high-grade glioma. While low-grade gliomas are dependent on growth-facilitatory signals from the tumor microenvironment, we hypothesize that their malignant counterparts achieve partial stromal independence by producing the very growth factors elaborated by nonneoplastic cells in low-grade gliomas. In this study, we demonstrate that M-GBM cells express high levels of CCL5 that provides an auto-stimulatory signal to increase cell survival. Similarly, in other studies, M-GBM cells express high levels of CXCL12, which forms another autoregulatory loop with its receptor CXCR4/CXCR7, to increase GBM cell proliferation [8,29,30]. In NF1associated low-grade gliomas, CXCL12 also serves as a paracrine factor that drives the growth of neoplastic Nf1−/− astrocytes [33]. These findings support the idea that highgrade glioma cells achieve certain degrees of stromal independence by producing growth factors previously provided by non-neoplastic stromal cells in the low-grade glioma tumor microenvironment.
CCL5 is a potent chemokine operative in several types of cancers, including lymphoma, melanoma, prostate cancer and breast cancer [34]. However, while CCL5 function in glioma has not been fully elucidated, several reports have suggested that it might be an important chemokine in this setting. First, one of the main CCL5 receptors, CCR5, is expressed in U87 and U251 glioblastoma cell lines, such that its knockdown inhibits U87 growth in nude mice [35]. Second, CCL5 mRNA and protein expression are elevated in both tumor tissue and serum from patients with highgrade glioma [36]. Lastly, CCL5 is produced by microglia in the microenvironment of murine Nf1 optic glioma, and is increased in low-grade gliomas (pilocytic astrocytoma) relative to non-neoplastic tissue [11]. In this latter study, CCL5 treatment increased optic nerve glial proliferation in vitro, whereas antibody-mediated CCL5 inhibition reduced optic glioma growth in vivo.
CCL5 typically functions by binding to its receptors CCR1, CCR3 and/or CCR5. Surprisingly these receptors were not detectable in M-GBM cells. Instead, M-GBM cells express the non-conventional CCL5 receptor, CD44, and use this receptor to maintain tumor cell survival. CD44 is a cell surface glycoprotein that can function as a co-receptor for extracellular matrix components, growth factors and chemokines. CD44 plays several critical roles in GBM and drives GBM cell invasion, proliferation and drug resistance. CD44 expression is increased in GBM compared to low-grade gliomas and high levels of CD44 are associated with worse survival of GBM patients [37]. Inhibiting CD44 function using monoclonal antibodies or RNA silencing reduced GBM growth in vivo and prolonged mouse survival [38,39]. The interaction between CD44 and its ligands (e.g., hyaluronic acid and osteopontin) modulates multiple aspects of glioma biology. Activation of CD44 by hyaluronic acid was shown to increase glioma cell invasion [40][41][42][43]. In addition, osteopontin-CD44 interaction induces cleavage of the intracellular domain of CD44, increases a stem cell phenotype, and increases the growth of PN-GBM cells [44]. In this study, we found that the interaction between Ccl5 and CD44 enhances cell survival in M-GBM cells. This pro-survival effect is specific to Ccl5, and does not appear to involve other CD44 ligands, since Cd44 knockdown in Ccl5 KD cells does not further reduce cell growth. Our study expands our understanding of the potential mechanisms by which CD44 regulates GBM maintenance.
CCL5 employs CD44 as an alternative receptor to facilitate the entry of HIV virus [24], where it forms a complex with CD44 and Src kinases to activate ERK signaling. In our studies, ERK activity in M-GBM cells was not reduced following Ccl5 KD (data not shown), suggesting that a distinct signaling pathway is activated following Ccl5 binding to CD44 in M-GBM cells. Previous studies have demonstrated increased cell survival following CD44 activation in several different cancers. For example, CD44 protects chronic lymphocytic leukemia cells from spontaneous and induced cell death by increasing expression of the MCL-1 pro-survival   [45], while CD44 activation in breast cancer cells induces c-Jun mediated survival [46]. It has been suggested in GBM that CD44 also inhibits the Hippomediated apoptotic signaling pathway to achieve resistance to reactive oxygen species-or cytotoxic agent-induced stress [39]. Further work will be required to determine how CD44 regulates cell survival in M-GBM cells.
GBM tumors are genetically heterogeneous and can be divided into mesenchymal, classical, neural and proneural subtypes. Herein, we focused on studying the behavior of mesenchymal GBM since Ccl5 has the highest expression in this group. One common feature of the M-GBM subtype is NF1 loss/mutation, which results in impaired neurofibromin RAS suppression. Consistent with this observation, we demonstrated that neurofibromin directly regulates CCL5 expression through RAS signaling. First, we show that ectopic expression of a functional, but not mutant, NF1-GAP-related domain reduced Ccl5 expression. Second, Nf1 loss in primary astrocytes resulted in increased Ccl5 expression. Third, pharmacological inhibition of RAS downstream effector activation (AKT/ mTOR) suppressed Ccl5 expression in M-GBM cells. Collectively, these results elucidate one mechanism by which Ccl5 expression is controlled by RAS.
The Ccl5-mediated autocrine circuit identified in this study may not apply to other GBM subtypes, given that they do not express high levels of either CCL5 or CD44. In fact, we have shown that treating murine PN-GBM cells with Ccl5 did not increase cell growth. While multiple signaling pathways contribute to GBM tumorigenesis, we report one in which a chemokine-driven autocrine pathway supports GBM cell survival. The regulation of chemokines by oncogenic signaling pathways suggests that other brain cancer-associated mutations might similarly modulate chemokine production. Further dissection of the chemokine/cytokine networks regulated by these tumor suppressor/oncogenic pathways might uncover unique immune modulatory programs with differential impact not only on tumor cell growth, but also on the stromal cell types in the tumor microenvironment.
Lastly, while Ccl5 silencing results in ~20% reduction in cell growth in vitro, there was a ~70% reduction in tumor cell growth following Ccl5 KD in vivo. The greater cell survival benefit in vivo raises the possibility that CCL5 may not solely function in an autocrine manner, but may also recruit monocytes (macrophages/microglia) to the tumor and enhance tumor growth in a paracrine manner. However, we found no difference in the percent of intratumoral Iba1 + monocytes in mice bearing Ccl5 KD versus control (LacZ KD) M-GBM cells (data not shown). Since it is possible that attenuated Ccl5 expression alters the functional status of tumor-associated macrophages/microglia, current studies are focused on identifying potential changes in the tumor microenvironment conferred by reduced glioblastoma tumor cell Ccl5 production in vivo.
In summary, we demonstrate, for the first time, that CCL5 functions in an autocrine growth-promoting circuit, and establish a new receptor responsible for CCL5 function in mesenchymal glioblastoma cells. We further show that key oncogenic signaling pathways, like RAS, operate to modulate chemokine expression. As such, these studies expand the role of chemokines in dictating glioma biology, and suggest that paracrine growth regulatory signals produced by non-neoplastic cells in the low-grade glioma microenvironment may be converted to autocrine tumor maintenance circuits in high-grade malignancies relevant to mesenchymal glioblastoma survival in vitro and in vivo.

Mice
4-week-old male C57BL/6 mice were purchased from Taconic, and used in accordance with an approved Animal Studies Committee protocol at Washington University.

BrdU incorporation
Cells were plated in 96-well plates, and treated with 250 ng/mL mCcl5 (R&D systems, 478-MR-025) or 0.5 % BSA in PBS (vehicle control) for 4 hours at 37°C. BrdU was added to the cell culture 2 hours before fixing the cells and performing the BrdU ELISA assay (Roche Life Science, 11647229001) according the manufacturer's protocol.

Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL)
Cells maintained in 0.5% fetal bovine serum for 24 hours were fixed with 4% PFA, and TUNEL + cells detected using the fluorescence-based in situ cell death detection kit (Roche Life Science, 11684795910). The number of TUNEL + cells was normalized to the total cell number (DAPI + nuclei) using NIH ImageJ software.

Western blotting
1861 cells expressing shRNAs were maintained in 0.5% fetal bovine serum for 24 hours prior to lysis and Western blotting with appropriate primary antibodies (Supplementary Table 2) [14,52]. Quantification of western blots were performed using ImageStudio (LI-COR). AKT and S6 activities were quantified as signal intensity of phospho-proteins versus signal intensity of total proteins.

Mouse chemokine enzyme-linked immunosorbent assay (ELISA)
Cells were serum starved for 24 hours, and the culture medium collected for mouse Ccl5 ELISA (R&D systems, MMR00). The amount of secreted Ccl5 determined by ELISA was normalized to total cell number.

Quantitative real-time PCR
qPCR was performed as described previously [11] using validated primers (Supplementary Table 3).

Intracranial injections
5 × 10 4 1861 cells were implanted into the striatum of 4-week-old C57BL/6 mice as previously described [49]. Mice were monitored daily, and euthanized when moribund. Kaplan-Meier survival curves were generated, and differences between control (shLacZ) and Ccl5 KD (shCcl5-1 and shCcl5-2) groups analyzed using the Logrank test. For mCcl5 ELISA and Western blotting of the implant, tissue around the tumor implantation site was dissected 6-weeks post-implantation, followed by lysing the tissue as described previously [14,52].

Immunohistochemistry
Implanted mouse brains were fixed at the time of death or deemed moribund by the veterinary staff, embedded in paraffin and processed as described previously [53], followed by performing immunohistochemistry using primary and secondary antibodies (Supplementary Table 2). Ccl5 signal was amplified using tyramide signal amplification as described previously [11].

Statistical analyses
Data analysis was performed using Prism GraphPad. Unpaired two-tailed Student's t tests were used to determine the difference between two groups. All data were presented as mean values with standard error of mean (SEM). All in vitro experiments were repeated at least three times with similar results.