Heparan Sulfate Synthesized by Ext1 Regulates Receptor Tyrosine Kinase Signaling and Promotes Resistance to EGFR Inhibitors in GBM

The malignant behavior of glioblastoma (GBM) is dependent, in part, on signaling from multiple receptor tyrosine kinases (RTK). Among the most commonly altered RTKs is EGFR, which is amplified and/or mutated in over 50% of these tumors (1, 2). Despite its importance in the malignant biology of GBM, attempts to inhibit EGFR signaling have failed to improve outcomes for patients with GBM (3). One explanation for this failure is RTK functional redundancy, which has shown that alternative RTKs, such as platelet-derived growth factor receptor α (PDGFRA), insulin-like growth factor receptor (IGFR1), FGFRs, and c-Met, can compensate for reduced RTK signaling when EGFR is being therapeutically targeted (4, 5).

An aspect of RTK signaling that has experienced only limited investigation in GBM, as well as in the response to RTK targeted therapy, is growth factor bioavailability (6, 7). Heparan sulfate proteoglycans (HSPG), present on cell surfaces and in the extracellular matrix (ECM), bind to diverse partners, including soluble growth factors, membrane proteins, and components of the ECM (8, 9). Thus, HSPGs play an important role in regulating ligand bioavailability and RTK signaling. Although HSPG function depends on both its core protein and covalently attached heparan sulfate (HS) glycosaminoglycan chains, the HS chains are primarily responsible for determining ligand specificity and binding. A critical step essential for HS biosynthesis is performed by the HS-polymerizing enzymes EXT1 (exostosin 1) and EXT2 (exostosin 2) that encode glycosyltransferases that elongate HS chains (10, 11). Conditional disruption of EXT1 has established its essential role in mammalian brain development (12), and Ext1^−/− and Ext2^−/− mice have profound developmental defects resulting in embryonic lethality (13, 14). In the embryonic spinal cord, HS 6-O-sulfation levels, regulated by the extracellular sulfatases (Sulfs), create a morphogen gradient of Shh necessary for the normal motor neuron to oligodendrocyte progenitor cell switch (15, 16). In the adult brain, HS continues to regulate cell signaling. Alterations in HS structure promote glial scar formation after central nervous system injury (17) and mediate tau localization in neurodegenerative disease (18).

In GBM, HS is an important component of the cellular microenvironment and promotes PDGFRA signaling, angiogenesis, and tumor cell invasion (19–22). Data from in vitro studies and model organisms have indicated several potential mechanisms by which HS can regulate ligand-mediated signaling, including: direct binding to and sequestration of soluble ligands; localization of ligand to specific cell surface regions; and acting as a coreceptor on the cell surface (8, 23–25).

Given the importance of HS in regulating growth factor–mediated signaling in the brain, we investigated HS regulation of multiple RTKs in GBM and its contribution to innate resistance to RTK inhibitor therapy. We have pursued this in three ways: (i) By generation of tumor-prone murine neural progenitor cells (mNPC) with Ext1knockout, therefore preventing biosynthesis of HS; (ii) use of explant cultures derived from patient-derived xenografts enzymatically cleared of cell surface HS; and (iii) investigation of in vivo tumor growth in different tumor microenvironments. Loss of cell surface HS decreased the activity of multiple RTKs, reduced cell surface–bound PDGF-BB, and decreased tumor cell growth in vitro and in vivo. Our studies highlight the importance of HS-mediated regulation of growth factor signaling in the biology of GBM and its contribution to RTK inhibitor resistance.

MNPCs, including both untransduced cells and tumor-prone cells containing EGFRvIII (mNPC-E*), and patient-derived GBM tumorspheres were cultured as described previously using minimal essential media under nonadherent conditions (20, 22, 26). Patient-derived GBM tumorsphere lines GBM43 (from C.D. James), GBM5 (from C.D. James), and GBM1 (SF9487 from UCSF), established and demonstrated to have tumor-generating capacity, were maintained under nonadherent conditions as tumorspheres (20, 26–28). The three human lines were selected based on their molecular and transcriptional properties as follows: GBM43 has moderate expression of cell surface HS (20), expression of EGFR with response to erlotinib, and expression of PDGFRA with a robust signaling response to PDGF-BB (20); GBM1 also has moderate expression of cell surface HS and expression of EGFR with response to erlotinib (20); and GBM5 has high levels of cell surface HS expression (20). All cell lines were analyzed before using and each subsequent year by short tandem repeat analysis to ensure the identity and validity of cells and confirm by PCR that they are mycoplasma negative. Cells were passaged no more than 15 to 20 times and then a new frozen aliquot of cells was used. Unless otherwise specified media and supplements were purchased from Thermo Fisher Scientific. Invasion assays use Corning Matrigel Growth Factor Reduced (GFR) Basement Membrane Matrix (Corning Inc.). The EGFR tyrosine kinase inhibitor, erlotinib (#OSI-744), was purchased from Selleckchem. Cell viability was assessed using CellTiter-Glo 3D Reagent (Promega) or DyLight800 (Thermo Fisher Scientific). To measure self-renewal (sphere-forming ability) and to assess sphere size, neurospheres were dissociated by trituration, then replated at clonal density in round-bottom 96-well ultralow attachment plates (Corning). Wells with more than one cell per well were eliminated by visual inspection, and secondary neurospheres were counted and imaged on days 4 and 6 (DFC3000G, Leica). Sphere size was quantified using ImageJ (RRID:SCR_003070).

Mice harboring loxP-flanked Ext1 (Ext1^f/f) (strain B6;129-Ext1<tm1Vcs>/J, #009326) were purchased from The Jackson Laboratory and backcrossed a minimum of 10 generations into FVB/N Ink4a/Arf^−+/−. All experiments were executed in compliance with institutional guidelines and regulations and after approval from the appropriate Institutional Review Board (UCSF Office of Research Institutional Animal Care and Use Program, #AN169893). mNPCs were isolated from the subventricular zone (SVZ) of 4-week-old male Ink4a/Arf^−/−;Ext1^fl/f mice as described previously (22, 26, 29). Neural progenitor cells maintained as neurospheres were transduced with EGFRvIII by retrovirus and Cre recombinase by adenovirus (Ad-GFP-2A-iCre, #1772, Vector Biolabs) sequentially. The pBabe vector based on the Moloney Murine Leukemia Virus (MoMuLV) containing EGFRvIII was used to generate retrovirus (21, 25, 28). After expanding, HS-negative cells (HS-null) were enriched by cell sorting (SH800, Sony Biotechnology Inc.) following immunostaining with anti-HS antibody (10E4 epitope) and Alexa488-labeled anti-mouse IgM antibody. Isogenic control cells were treated in parallel and were transduced with GFP (Ad-CMV-GFP, #1768, Vector Biolabs), expanded, and sorted for GFP-positive cells (HS-wt) by cell sorting. Cell surface HS levels were re-evaluated every 6 months and in tumor cells cultured from in vivo tumors and remained stable.

Antibodies used for flow cytometry, Western blotting, and immunostaining included mouse anti-HS (clone 10E4, US Biological Cat #H1890, RRID:AB_10013601); mouse anti-HS (clone F69–3G10, AMSBIO Cat# 370260–1, RRID:AB_10892311); mouse anti-chondroitin sulfate (clone CS-56, Sigma-Aldrich Cat #C8035, RRID:AB_476879); rabbit anti-Ki-67 (clone 30–9, Ventana Medical Systems Cat #790–4286, RRID:AB_2631262); mouse antiphosphotyrosine (Millipore Cat #05–321, RRID:AB_309678), rabbit antiphosphorylated PDGFRα (Y742; R&D Systems Cat# AF2114, RRID:AB_416551); rabbit anti-PDGFRα (Millipore Cat #07–276, RRID:AB_310485); mouse anti-GAPDH antibody (Millipore Cat #MAB374, RRID:AB_2107445); and from Cell Signaling Technology rabbit anti-phosphorylated AKT (S473; Cat #4060, RRID:AB_2315049), rabbit anti-AKT (Cat #9272, RRID:AB_329827), rabbit antiphosphorylated ERK1/2 (T202/T204; Cat #9101, RRID:AB_331646), mouse anti-ERK1/2 Cat #9107, RRID:AB_10695739), and rabbit anti–β-actin antibody (Cat #4967, RRID:AB_330288). Secondary antibodies included Alexa488-labeled goat anti-rabbit IgG antibody (Cat #A11008, RRID:AB_143165), Alexa488-labeled anti-mouse IgM (Cat #A-11001, RRID:AB_2534069), and Alexa568-labeled goat anti-mouse IgG antibody (Cat #A11004, RRID:AB_2534072) from Thermo Fisher Scientific. For immunostaining of formalin-fixed paraffin-embedded sections, the 3G10 neoepitope was generated by digestion of HS by heparitinase I and II for 4 hours at 37°C following proteinase K antigen retrieval. The negative control is staining in the absence of enzyme.

Cells were dissociated by treatment with Accutase (Sigma) and incubated with or without heparitinase I, II, and III from Flavobacterium heparinum as described previously (ref. 20; kind gift from Dr. Kuberan Balagurunathan, University of Utah; Sigma) for 1 hour at 37°C. Cells (5 × 10⁵) were washed with PBS, blocked in PBS plus 5% goat serum on ice for 15 minutes, washed twice with ice-cold PBS plus 1% BSA, incubated on ice for 1 hour with an anti-HS antibody diluted in PBS plus 1% BSA, and detected with Alexa488-labeled secondary antibodies. Cells were analyzed by FACS Canto II Flow Cytometer (BD Bioscience).

Protein lysates from mNPC-E* were analyzed by Mouse Phospho-RTK Array Kit (R&D Systems) per the manufacturer's protocol (300 μg) or by Western blotting (20 μg total protein) using an Odyssey detection system (LI-COR). The murine RTK array contains 39 RTK capture and 4 control antibodies spotted in duplicate on a membrane. The human RTK array contains 49 RTK capture antibodies spotted in duplicate in addition to the control antibodies.

All experiments were executed in compliance with institutional guidelines and regulations and after approval from the appropriate institutional review board (UCSF Office of Research Institutional Animal Care and Use Program, #AN169893). Cell suspensions of HS-null and HS-wt mNPC-E* (1 × 10⁷ cells/mL in PBS) were diluted with equal volume of matrigel (8.4 mg/mL; BD). Under inhalation anesthesia with isoflurane, 1 × 10⁶ cells were injected subcutaneously in 4-week-aged male athymic nude mice (Charles River Laboratories, RGD Cat #2312499, RRID:RGD_2312499; 2 μL per injection site). Tumor volumes were calculated as major axis × minor axis × height. For inhibition of EGFR, when tumor volume reached 200 mm³ (designated day 0 of treatment), erlotinib (150 mg/kg) was administered intraperitoneal in 15% Captisol (Ligand Pharmaceuticals, Inc.) every other day for a total of three doses. Mice were weighed, and subcutaneous tumor volumes were measured daily. For intracranial allografts, 1 × 10⁵ cells diluted in sterile saline were injected using a stereotactic device over 5 minutes as described previously (21). Mice were weighed daily and monitored for signs of tumor development. Numbers of mice used in each experiment are listed in corresponding figure legend.

Cells were washed twice, incubated on ice for 1 hour with biotinylated-human PDGF-BB (Fluorokine Biotinylated-human PDGF-BB Kit; R&D Systems), and bound ligand was detected using a FITC-labeled avidin probe and a FACS Canto II Flow Cytometer (BD). Cell treatment with heparitinase I, II, and III, Chondroitinase ABC, or heparin was performed prior to incubation with biotinylated-human PDGF-BB.

All statistics were performed using Graph Pad Prism 6.0 software (Graph Pad Software). All data are presented as mean ± SD. One-way ANOVA, the Student t test analysis, and Mann–Whitney test were performed at confidence interval levels of 95%. Log-rank (Mantel–Cox) test was used for survival analysis.

Carbohydrates on the cell surface and in the ECM, including glycosaminoglycans, O-glycans, and glycolipids, influence cell signaling by modulating RTK function (21, 22, 30, 31). To investigate the role for HS glycosaminoglycans in ligand-mediated signaling in GBM, we developed a murine model for GBM in which HS biosynthesis could be knocked out. Exostosin-1 (Ext1) encodes an enzyme that elongates HS chains and is essential for HS biosynthesis and for embryonic viability (13, 14). Ext1^f/f/Ink4a/Arf^−/− mice were generated from mice harboring loxP-flanked Ext1 (Ext1^f/f) and Inka4a/Arf-null (Ink4a/Arf^−/−) alleles. Neural progenitor cells were isolated from the subventricular zone (SVZ) and transduced with EGFRvIII, a constitutively active and oncogenic form of EGFR. These cells were then transduced with Cre-expressing adenovirus, then flow sorted to identify cells lacking surface HS (HS-null; Fig. 1A and B). Biosynthetic functions of EXT1 could then be investigated using these HS-null tumor-prone progenitor cells. Control cells were transduced with GFP-expressing adenovirus and sorted for GFP expression (HS-wt).

HS-null cells demonstrated a significant reduction in cell growth as determined by assessment of sphere size at days 4 and 6 after plating (Fig. 1C; P < 0.05) and cell growth in bulk cultures at day 3 (Supplementary Fig. S1C; P < 0.01). Secondary neurosphere assays demonstrated a similar reduction in sphere size in HS-null cells (HS-null sphere area normalized to HS-wt 0.75 ± 0.0, P < 0.01, n = 5 per condition). The overall ability of HS-null and HS-wt cells to form spheres from single cells (self-renewal), however, was not altered in primary or secondary neurosphere assays (Supplementary Fig. S1A–S1B). Exogenous heparin rescued cell growth in HS-null cells (Fig. 1C; P < 0.01 day 4 and P < 0.005 day 6). HS-null cells also demonstrated decreased invasion in 3D matrigel invasion assays relative to HS-wt at 24 hours (Fig. 1D; P < 0.01).

Ext1^−/− (HS-null) cells showed decreased total tyrosine phosphorylation levels and a corresponding decrease in activation of downstream signaling pathways, as indicated by reduced levels of phosphorylated Akt and Erk1/2 (Fig. 2A and B). To identify specific signaling pathways affected by loss of cell surface HS, we performed a comprehensive analysis of RTK phosphorylation in HS-null and control cells. HS-null cells showed reduced phosphorylation of several RTKs. RTKs with the greatest reduction included fibroblast growth factor receptors (FGFR3, FGFR4), insulin receptor family (InsulinR, IGF1R), platelet-derived growth factor receptors (PDGFRα, PDGFRβ), receptor tyrosine-protein kinase Erb-B2 (ErbB2), and receptor tyrosine-protein kinase UFO (Axl; Fig. 2C; Supplementary Fig. S2).

HS-wt cells readily formed tumors in athymic nude mice. In contrast, cells deficient for HS resulted in a delay in allograft establishment and reduced rate of tumor growth upon establishment (Fig. 2D). This decreased growth was associated with a decrease in proliferating tumor cells and reduced vascular density in HS-null tumors relative to HS-wt tumors (Fig. 2E and F). HS-null tumors had a near-complete absence of HS in tumors (Fig. 2G and H). Thus, HS deficiency in a murine model for GBM resulted in decreased RTK activation, decreased tumor cell proliferation, and decreased tumor growth in vivo.

Noncell autonomous HS, in the form of heparin, rescued cell growth in HS-null cells (Fig. 1C) and promoted growth of HS-wt cells largely via FGF2-mediated signaling (Fig. 3A). Thus, both cell autonomous and noncell autonomous HS can promote ligand-mediated signaling in GBM. In the hypodermis, HS-null allografts grow in the absence of tumor cell invasion and maintain a relative HS-null phenotype (Fig. 2H). In contrast, HS is abundant in the brain (Fig. 3B and C), and intracerebral allografts exhibit single-cell tumor invasion (Fig. 3D and E). Thus, HS-null cells are surrounded by an HS-rich microenvironment (Fig. 3F and G). To investigate the role for both cell autonomous and noncell autonomous HS in GBM, HS-null and HS-wt cells were implanted orthotopically in HS-wt mice. Intracerebral allografts of HS-null cells trended toward slowed growth relative to HS-wt cells (Fig. 3J; P < 0.08). At the time of signs or symptoms of tumor, however, the HS-null and HS-wt tumors had similar tumor vascular density (relative vascular density 78 ± 22 vs. 100 ± 13, respectively). Comparison of tumor size, extent of invasion, extent of necrosis, and proliferation (Ki-67 immunostaining) revealed similar patterns between HS-null and HS-wt tumors. Although prolonged growth of HS-null tumor cells in the brain may select for cells with HS expression, immunostaining (Fig. 3H and I) failed to identify robust expression of HS on HS-null tumor cells as compared with HS-wt cells. In addition, flow cytometry on cells isolated from tumors at 42.5 days after implantation revealed robust HS expression on HS-wt cells and no detectable HS expression on HS-null cells (Supplementary Fig. S3). These data suggest in the brain, both cell autonomous and noncell autonomous HS contribute to GBM growth.

We next investigated the role for HS in RTK regulation in GBM43, a patient-derived GBM tumorsphere line with moderate cell surface HS content and expression of both EGFR and PDGFRA (20). Enzymatic removal of cell surface HS, via enzymatic clearing with heparitinase I, II, III (ΔHS; Fig. 4A, Supplementary Fig. S4A), resulted in reduced phosphorylation of several RTKs. RTKs with the greatest reduction included insulin receptor family (InsulinR, IGF1R) and PDGFRα (Fig. 4B). PDGFRα signaling is important in GBM, and we have previously identified a role for HS structure in PDGFRα signaling in GBM (22, 32). To investigate potential mechanisms by which HS may influence RTK signaling, we examined effects of HS deficiency with respect to cell surface PDGF-BB ligand and PDGFRα signaling. Enzymatic removal of cell surface HS resulted in a 70% reduction of cell surface–bound PDGF-BB (Fig. 4C and D; P < 0.01). Removal of the glycosaminoglycan chondroitin sulfate had no effect on cell surface–bound PDGF-BB (Supplementary Fig. S4B–S4C). Enzymatic removal of cell surface HS was also associated with reduced phosphorylation of PDGFRα and ERK1/2 as compared with control-treated cells (Fig. 4E and F; P < 0.01 and P < 0.05, respectively). To investigate long-term effects of HS on cells, exogenous heparin, a highly sulfated form of the glycosaminoglycan HS, was added to GBM43. Heparin promoted cell viability in a dose-dependent manner (Fig. 4G). In contrast, the glycosaminoglycan chondroitin sulfate A had no effect on cell growth. Heparin-induced growth of GBM43 was associated with increased phosphorylation of PDGFRα and ERK1/2 (Fig. 4H). GBM5 is a patient-derived GBM tumorsphere line with high cell surface levels of HS (20). Similar to GBM43, enzymatic removal of cell surface HS from GBM5 resulted in reduced phosphorylation of several RTKs with reduced phosphorylation of Axl being most striking (46% of control treated cells, Supplementary Fig. S4E). In PDGF-BB binding assays, enzymatic removal of cell surface HS in GBM5 resulted in a 60% reduction of cell surface–bound PDGF-BB, while removal of the positively charged glycosaminoglycan chondroitin sulfate had no effect (Supplementary Fig. S4A–S4D).

To investigate how HS sulfation influences cell surface PDGF-BB binding, GBM43 cells were acutely coincubated with a series of structurally defined polymeric heparin mimetics or fully sulfated heparin (NS2S6S; Supplementary Figs. S5A–S5B and S6). Relative to control-treated cells, all mimetics competed with cell surface HS and decreased PDGF-BB binding, with the most potent being the trisulfated form, NS2S6S.

The extracellular enzyme SULF2 removes 6O-sulfation on HSPGs and is known to affect HS binding to multiple growth factors such as stromal cell-derived factor 1 (SDF1) and VEGF (33). We asked whether exogenous SULF2 could alter cell surface–bound PDGF-BB in an experiment in which conditioned media containing SULF2 or inactive SULF2 (S2ΔCC; ref. 34) were harvested from 293T cells that had been modified to express wild-type or mutant SULF2. Production of active SULF2 enzyme was confirmed using a fluorogenic substrate for SULF2 (Supplementary Fig. S6B), and by mass-spectrometry analysis of conditioned media that revealed decreased 6-O-sulfation from trisulfated heparin (Supplementary Fig. S6C). Addition of SULF2-conditioned media resulted in a small but significant reduction in cell surface–bound PDGF-BB relative to S2ΔCC-conditioned media (Supplementary Fig. S6D; P < 0.01). Taken together, these data suggest alterations in HS sulfation, including those mediated by the extracellular enzyme SULF2, alter ligand availability at the cell surface and thereby alter PDGFRA signaling.

HSPGs consist of a core protein and covalently attached glycosaminoglycan chains. Several protein cores exhibit increased expression in GBM (19). In GBM43, we identified several potential HS-containing proteins using the 3G10 antibody. Flow cytometry for GPC1 and flow cytometry and immunofluorescence for SDC1 demonstrated robust expression of both proteoglycans on the cell surface of GBM43 (Supplementary Fig. S7).

In diverse cancers, ligand-mediated activation of alternative RTKs with functional redundancy to the therapeutically targeted RTK is a mechanism of innate and acquired RTK inhibitor resistance (6, 7). That is RTK ligands present in the tumor microenvironment can reduce the effectiveness of targeted kinase inhibitors. Using two distinct patient-derived GBM tumorsphere lines that express EGFR and respond to EGFR inhibition (20), we investigated the ability of four different growth factors relevant in GBM to confer resistance to EGFR inhibition. In GBM43, exogenous FGF2 and PDGF-BB could partially rescue cell viability upon EGFR inhibition (Fig. 5A–C). In GBM1, partial rescue was conferred by addition of FGF2 or IGF-1 and not by PDGF-BB (Fig. 5D).

We next investigated whether HS contributes to ligand-dependent tumor resistance to RTK inhibition. To do this, we focused on the role for HS in FGF2-mediated resistance to EGFR inhibition. In both patient-derived GBM43 and murine GBM, FGF2 partially rescued cell viability in HS-deficient versus control cells upon EGFR inhibition (Fig. 6A and B). As early as 24 hours after drug treatment, HS-null cells exhibited increased cell death at both 0.5 μmol/L and 5 μmol/L erlotinib relative to HS-wt (Fig. 6C). To determine whether HS mediates resistance to EGFR inhibition in vivo, HS-null and HS-wt cells were engrafted into mice and EGFR inhibitor therapy was initiated when tumors reached an equivalent size. Following completion of treatment, HS-null tumors were significantly smaller than HS-wt tumors (Fig. 6D; day 5 after treatment, P < 0.05).

Glycosylation of cell surface proteins is an important posttranslational modification that affects cell–cell communication, cell signaling pathway activity, and cell adhesion (35). In GBM, alterations in HS proteoglycans, present on the cell surface and in the ECM, promote oncogenic signaling and invasion (19, 20, 22). By genetically preventing HS chain biosynthesis in a murine model for GBM and enzymatically clearing cell surface HS from patient-derived cell lines, we demonstrate the importance of HS in growth factor driven glioma progression as well as response to targeted therapies. Tumor cells lacking HS had decreased proliferation, increased length of time for establishment as an allograft and decreased allograft growth rate subsequent to establishment, and decreased activation of multiple RTKs. We have shown that HS regulates PDGFRA signaling pathway activity and that decreased pathway activity is associated with reduced cell surface–bound PDGF-BB ligand. HS influence of ligand-mediated signaling may be particularly important in contributing to tumor adaptive response to RTK inhibition.

Ext1 is an essential enzyme for HS biosynthesis. Using neural progenitors isolated from Ext1^fl/fl Ink4a/Arf^−/− mice, we generated isogenic cell pairs (22, 26) that differed only in their ability to synthesize HS glycosaminoglycan chains. Causing tumor cell HS deficiency in this manner mitigated tumor malignancy both in vitro and in vivo. HS on the cell surface and in the tumor microenvironment contributes to ligand-mediated signaling via several mechanisms (reviewed in refs. 19, 36), including the sequestration of growth factors (37). Growth factors demonstrated to bind to HS include FGF2 (17, 38, 39), platelet-derived growth factor (40, 41), hepatocyte growth factor (scatter factor; ref. 42), and VEGF (21, 33). HS can also serve as a coreceptor for ligand-mediated signaling. Indeed, HSPGs stabilize FGF ligand-receptor complex to promote FGF2 signaling (23, 43, 44).

In GBM, oncogenic activation of PDGFRA is common and regulated in part via HSPGs (22, 32, 45). Investigating how HS alters PDGF-BB–mediated signaling using patient-derived GBM lines, we found that enzymatic clearing of cell surface HS reduced cell surface–bound PDGF-BB ligand, reduced activation of PDGFRα, and reduced downstream signaling pathway activity. Thus, HS may regulate local PDGF-BB bioavailability.

The sulfation pattern of the HS chains is a major determinant of the specificity and the affinity of HS-ligand interactions, particularly the 6O-sulfate (6OS) of glucosamine (24, 33, 37, 46). Regulation of 6OS levels occurs during biosynthesis and postsynthetically by the extracellular sulfatases (SULF). Using polymeric heparin mimetics differing in their sulfation status combined with exogenous SULF2 enzyme, we demonstrated cell surface PDGF-BB binding is dependent on the sulfation status of the HS chains. Binding also has specificity for HS as reducing cell surface chondroitin sulfate glycosaminoglycans had no impact on cell surface–bound ligand.

Ligand-mediated activation of RTKs with functional redundancy to a specific RTK being therapeutically targeted is one mechanism of resistance to RTK inhibitors in diverse cancers (6, 7). In patient-derived GBM lines and a murine model for GBM, we demonstrate HS is a major factor regulating growth factor bioavailability and promotes resistance to targeted RTK inhibition. In the absence of cell surface HS, FGF2-mediated resistance to EGFR inhibition was diminished. Importantly, inhibition of EGFR in vivo was more efficacious in HS-null than HS-wt tumors (Fig. 6D).

HS acts in both a cell autonomous and a noncell autonomous manner. In the HS-low microenvironment of the hypodermis, HS-null allografts grow as relatively circumscribed masses and exhibit profound growth defects. This is likely due to a combination of decreased growth factor signaling in tumor cells, as evidenced by decreased RTK signaling and tumor cell proliferation, and decreased tumor angiogenesis, as evidenced by decreased vascular density, in HS-null tumors. HS binds vascular endothelial growth factor and promotes signaling in normal tissues and in GBM (21). Decreased tumor angiogenesis is likely one factor in the apparent growth plateau of the HS-null cells in vivo (Fig. 2). In the brain, however, HS is abundant in the microenvironment and tumor cells exhibit diffuse single cell invasion. Thus, HS-null tumor cells invade and proliferate within an HS-rich microenvironment. In this context, HS-null cells exhibited a less severe growth defect. Comparing orthotopic HS-null and HS-wt tumors at late time points, tumor growth and angiogenesis were similar suggesting HS in the tumor microenvironment rescues the profound growth defects observed in the hypodermis. Although HS-null tumor cells appeared to lack robust HS expression (Fig. 3I), it is possible that even low-level HS expression may contribute to tumor growth in the brain. As HS-null mice are not viable, we could not directly test the role for noncell autonomous HS in tumor growth and invasion. Although this is a limitation of our study, our in vitro and in vivo data, comparing two different tumor compartments, suggest HS on cell surfaces and in the brain tumor microenvironment contribute to tumor growth in GBM.

Our data with FGF2, IGF-1, and PDGF-BB suggest HS promotes ligand-mediated signaling in GBM via multiple mechanisms that collectively contribute to tumor malignancy and progression. As HS regulates multiple ligand-mediated signaling pathways, inhibitors to HS biosynthesis and function are currently being investigated and in clinical trials (47–50). In combination with diverse RTK inhibitors, the modulation of HS function may be a promising therapeutic strategy in GBM and in other therapy-resistant neoplasms. This includes diffuse lower grade glioma, a tumor that may be particularly amenable to strategies that broadly target oncogenic signaling pathways and angiogenesis.

A. Wade reports grants from NIH during the conduct of the study, that contributions to the manuscript derive from work conducted from 2012-2015 while a postdoc in Joanna J. Phillips' lab, and is currently employed at Navire Pharma and CoA Therapeutics; Navire Pharma is developing a SHP2 inhibitor for use in tumors resistant to targeted therapies. A. Kumar reports grants, personal fees, and other from NCI during the conduct of the study. J.J. Phillips reports grants from NIH/NCI, NIH/NINDS, T.J. Martell Foundation, Gerson and Barbara Bakar Philanthropic Fund, and Sence Foundation during the conduct of the study. No disclosures were reported by the other authors.

Y. Ohkawa: Conceptualization, data curation, investigation, methodology, writing-original draft, writing-review and editing. A. Wade: Conceptualization, data curation, formal analysis, investigation, writing-review and editing. O.R. Lindberg: Data curation, investigation, writing-review and editing. K.Y. Chen: Data curation, formal analysis, investigation, writing-review and editing. V.M. Tran: Conceptualization, data curation, formal analysis, investigation, writing-review and editing. S.J. Brown: Resources, writing-review and editing. A. Kumar: Formal analysis, investigation, writing-review and editing. M. Kalita: Conceptualization, resources, investigation, writing-review and editing. C.D. James: Resources, formal analysis, writing-review and editing. J.J. Phillips: Conceptualization, resources, data curation, formal analysis, funding acquisition, investigation, writing-original draft, project administration, writing-review and editing.

This study was supported in part by NIH/NCI U01 CA229345 to J.J. Phillips, NIH/NINDS R01 NS081117 to J.J. Phillips, NIH/NCI U01 CA168878 to J.J. Phillips, and NIH/NCI P50CA221747 to C.D. James. We acknowledge and thank the T.J. Martell Foundation, Gerson and Barbara Bakar Philanthropic Fund, and the Sence Foundation for their support of this work. We are grateful and thank Dr. Kuberan Balagurunathan for sharing his expertise, providing us with heparitinase I, II, III, and providing us with polymeric heparin mimetics. We also acknowledge the UCSF Brain Tumor SPORE Biorepository (NIH/NCI P50CA097257) for providing histology services, NIH Grant PO1-HL107152 to Balagurunathan Kuberan for providing essential resources, the UCSF Helen Diller Family Comprehensive Cancer Center Laboratory for Cell Analysis (P30CA082103) for microscopy services, and the UCSF Mass Spec Facility (NIH NCRR P41RR001614) for use of their mass spectrometer. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.