Leukocyte Differentiation by Histidine-Rich Glycoprotein/Stanniocalcin-2 Complex Regulates Murine Glioma Growth through Modulation of Antitumor Immunity

Histidine-rich glycoprotein (HRG) is a cation- and heparin-binding, 75-kDa plasma glycoprotein produced by liver hepatocytes, present in blood at a concentration of 150 μg/mL (2 μmol/L; ref. 1). HRG comprises two N-terminal cystatin domains, and a C-terminal His/Pro-rich repeat region (2). The plasma levels of HRG in healthy adults are stable, although the protein is continuously taken up and turned over by leukocytes (3). HRG binds heparan sulfate with high affinity (4). It also binds to several components of the complement system (5) and to immunoglobulins (6), thereby enhancing macrophage phagocytosis of necrotic material and immune complexes (5, 6) in a manner dependent on Zn²⁺ binding to the His/Pro region (7). HRG has also been shown to possess antimicrobial activity (8).

HRG has been implicated in cancer growth and spread by influencing endothelial cell adhesion and migration, through interactions with integrin adhesion complexes (9, 10). Moreover, HRG affects cancer inflammation and immunity by direct gene regulatory effects on tumor-infiltrating leukocytes (11). In several mouse tumor models (4T1 breast cancer, Panc02 pancreatic cancer, and T241 fibrosarcoma), HRG therapy promotes M1 polarization of tumor-associated macrophages (TAM) in vitro and in vivo (12). The M1 polarization is accompanied by reduced production of proangiogenic and protumoral stimulation, resulting in normalization of tumor vessel morphology and perfusion, reduced metastatic spread, and enhanced chemotherapy delivery and efficacy. In addition, HRG promotes antitumor immunity via increased dendritic cell (DC), natural killer (NK), and T-cell tumor infiltration as a consequence of specific cytokine production by HRG-treated M1 macrophages (11). Conversely, the Hrg knockout mouse displays accelerated tumor growth and spread and marked upregulation of M2 markers such as IL10, in peritoneal macrophages (13). The naïve Hrg knockout mouse presents with enhanced coagulation and fibrinolysis but otherwise lacks an overt phenotype (14).

We have previously shown that iodinated HRG becomes enriched on spleen leukocytes when administered to the circulation in mice and that cultured inflammatory cells specifically bind HRG (3). Using a recently developed protein interaction–decoding technology, TRICEPS-based ligand receptor capture (LRC) technology (15, 16), we presently identify Stanniocalcin-2 (STC2) as an HRG interacting partner, on the cell surface of inflammatory cells. STC2 is a heparin-binding, secreted homodimeric glycoprotein produced by several different cell types. Moreover, STC2 is implicated in negative regulation of tissue growth as Stc2-deficient mice show 10% to 15% increased weight and faster growth than wild-type litter mates (17). STC2 expression correlates with malignancy and spread in different human cancer forms (18, 19). STC2 is upregulated in T cells developing a Th2 response (20) and it is a metagene, that is, a gene with broad effects on biological processes that negatively correlate with immune-related metagenes in breast cancer transcript arrays (21).

Primary tumors of the central nervous system (CNS) represent a particularly complex target for therapeutic intervention highlighted by the lack of significant changes in 5-year survival rate in the past decades (22). The anatomic location severely restricts surgical intervention and, even when possible, the invasive growth pattern of gliomas rapidly results in tumor recurrence within the resection cuff. Despite the aggressive combination of radio- and chemotherapy, grade 4 gliomas remain almost invariably fatal. Recently, the anti-VEGF–neutralizing antibody Bevacizumab has been added to the therapeutic regiment for glioblastoma, and although 6-month progression-free survival rates have increased, overall survival remains largely unchanged (23). There remains an urgent need for novel and/or neoadjuvant therapies.

Here, we delivered HRG, using a nonreplicating adenovirus-based vector, to orthotopically growing GL261 gliomas. The growth-suppressive effects of HRG on gliomas were paralleled by infiltration of STC2-expressing inflammatory cells. Moreover, HRG and STC2 individually and in complex regulated immune gene expression in inflammatory and immune cells. Thus, depending on the molecular configuration, HRG and STC2 may serve essential roles in steering inflammatory cell polarity to positively or negatively regulate immune and inflammatory pathways in cancer.

See Supplementary Material for methodology not described below.

Subcutaneous Panc02 (24) tumor studies were carried out in 8- to 10-week-old male C57Bl/6 mice. Panc02 cells were cultured in DMEM GlutaMAX medium (Gibco), 10% FBS and used for injection on the back of mice (2.5 × 10⁵ cells/mouse). When tumors reached 50 mm³ in size, Ad-Empty, Ad-HRG, or Ad-HRG-Luc vectors (5 × 10⁸ ffu; ref. 25) were injected intratumorally. Blood was sampled over 21 days by submandibular vein bleeding (26) into ethylenediaminetetraacetic acid (EDTA)–coated tubes and snap-frozen. In addition, mice were injected with d-luciferin intraperitoneally (100 mg/kg) and imaged on the IVIS live-imaging system.

To establish intracerebral glioma (27) in 8- to 10-week-old female C57Bl/6 mice, GL261 cells, cultured in DMEM GlutaMAX medium (Gibco), 10% FBS, were used for injection according to Lal and colleagues (28). Briefly, an implantable guide screw was placed in the skull at coordinates dorsal to the caudate nucleus (+1 mm anteroposterior, −2.3 medial-lateral from bregma). One week later (day 1), GL261 cell inoculation was carried out by injecting 7,500 cells in 0.5 μL, at a depth of 4.5 mm from the top of the screw (corresponding to −2.5 mm from the dorsal cortical surface). Six days later (day 7), a small incision above the guide screw was made and adenovirus vectors (7 × 10⁷ ffu) were injected at coordinates corresponding to the inoculation site, followed, 6 days later (day 13), by a second injection. Twenty-one days after GL261 cell inoculation, animals were given an intravenous injection of biotinylated-tomato lectin (4 mg/kg, Vector Labs), and circulated for 10 minutes followed by terminal anesthesia with xylazine/ketamine and transcardiac perfusion with PBS and 4% paraformaldehyde/PBS. Brains were dissected, postfixed in paraformaldehyde for 4 hours at 4°C, and incubated overnight in 30% sucrose, 4°C. Prepared brains were vibratome-sectioned into 100 μm coronal sections, mounted onto superfrost glass, and distributed in 6 series covering the entire tumor. For tumor size estimation, one series was stained with H&E, images acquired on a Nikon Eclipse 80i, and the tumor area/section determined by manual identification and automated quantification using the ImageJ analysis software. Final tumor area for each animal was determined by the sum of tumor area from each section in a single series.

For chemotherapy treatment, temozolomide was given by oral gavage in PBS/0.6% DMSO. A dose–response curve was established by treating tumor-bearing mice with temozolomide at a concentration range of 1.12 to 28 mg/kg/week. For cotreatment with Ad5, mice were given a temozolomide dose corresponding to the IC₂₀ (1 mg/kg/treatment corresponding to 3 mg/kg/week) from days 10 to 12 and 16 to 18.

For colony-stimulating factor-1 (CSF1)–neutralizing antibody treatment, mice were treated by intraperitoneal injection with 1 mg antibody per mouse, either anti-CSF1 (Clone 5A1, Bio X Cell) or isotype control (BE0088, Bio X Cell).

In vivo animal experiments were carried out in strict accordance with the ethical permit provided by the Committee on the Ethics of Animal Experiments of the University of Uppsala (Permit Numbers: C192/12, and C232/12). The Swedish Work Environment Authority has approved the work with replication-defective adenoviral vectors encoding immunomodulatory transgenes (ID number 202100–2932 v66a16 and v76a5). All experiments regarding adenoviral vectors were conducted under Biosafety level 2.

The human monoblastic U937 cell line that differentiates toward monocyte/macrophage lineage in response to vitamin D₃ (29) was a kind gift from Prof. Helena Jernberg Wiklund, Uppsala University (Uppsala, Sweden). Cells were tested negative for Mycoplasma but was otherwise not authenticated.

Statistical analyses were done using one-way ANOVA followed by different tests (Fisher LSD, Tukey, and Bonferroni) depending on sample properties for comparison between multiple groups. Student t test or χ² test was used for comparison between two groups. All experiments were repeated at least three times in an independent manner.

Replication-deficient Ad5 vectors (i) lacking a transgene element (Ad5-Empty); (ii) encoding murine HRG (Ad5-HRG); or (iii) encoding HRG and luciferase separated by a T2A auto-protease linker (Ad5-HRG-Luc) were tested for their properties by infection of pancreatic adenocarcinoma-derived Panc02 cells. HRG expression was evident from 24 hours postinfection (hpi) in Ad5-HRG-Luc and Ad5-HRG–infected cell culture medium while absent from the Ad5-Empty infected control. HRG expression peaked at 48 hpi in Panc02 cells, corresponding to 600 ng HRG/5 × 10⁴ cells (Supplementary Fig. S1A). Luciferase expression was present in Ad5-HRG-Luc–infected cells only (Supplementary Fig. S1B). Expressed HRG protein harvested from the culture medium bound to heparin and inhibited the directed migration of endothelial cells, as expected (Supplementary Fig. S1C and S1D, respectively). Thus, HRG was robustly expressed from cells infected with Ad5-HRG-Luc or Ad5-HRG and, importantly, retained properties typical for HRG.

To examine the expression of Ad5 transgenes in vivo, subcutaneous Panc02 tumors were injected intratumorally with a single dose of 1 × 10⁸ ffu Ad5-Empty or Ad5-HRG-Luc. Luciferase activity was assayed by in vivo live luminescence imaging and systemic blood HRG levels assayed by ELISA. Luciferase expression peaked at 2 days postinfection (dpi) and dropped off exponentially thereafter, but was evident until the experimental endpoint at 21 dpi (Fig. 1A). Consistent with the known circulating levels of HRG in murine plasma, HRG was present at 150 μg/mL in peripheral blood prior to virus treatment. The HRG concentration in Ad5-HRG lysates from perfused Panc02 tumors showed a 3- to 6-fold increase compared with controls, PBS, or Ad5-empty Panc02 tumors (Fig. 1B). Tumor HRG expression in Ad5-HRG but not Ad5-Empty injected mice was verified by immunoblotting (Supplementary Fig. S2).

To examine the potential for the HRG adenovirus to suppress the growth of established tumors, subcutaneous Panc02 tumors were injected with buffer or 1 × 10⁸ ffu Ad5-Empty or Ad5-HRG. Injections started on firmly established tumors (50 mm³) and were carried out every 4 days for a total of 4 injections. As shown in Fig. 1C and D, treatment with Ad5-HRG significantly reduced established Panc02 tumor growth and mean tumor volume at the experimental endpoint.

To assess the potential therapeutic consequence of the Ad5-HRG treatment on orthotopic glioma, we employed the guide screw system (28), which allows for repeated delivery of therapeutics to brain tumors. A total of 7,500 GL261 glioma cells were injected via the guide cannula at coordinates corresponding to the caudate nucleus. GL261 is known to replicate phenotypic characteristics of the human glioblastoma such as invasive growth (27). Six days after implantation, buffer or virus (Ad5-empty or Ad5-HRG) was administered to the coordinates used for GL261 cell inoculation. A second treatment was delivered 6 days after the first injection, and the animals were sacrificed on day 22 (Fig. 2A). Ad5-HRG treatment demonstrated a significant inhibition of GL261 growth compared with Ad5-Empty and buffer-treated animals (Fig. 2B). This pattern was seen robustly over three independent experiments using two independently produced and titered Ad5 batches.

HRG has previously demonstrated antitumor effects in different mouse tumor models, correlating with changes in vascular and immune cell parameters (11). Consistent with previous observations, Ad5-HRG treatment resulted in a significant increase in the total number and area of lectin-positive and therefore perfused vessels, compared with both Ad5-Empty and buffer-treated animals (Fig. 2C). The known role of CD45⁺ cells in HRG antitumor activity prompted the examination of changes in leukocyte numbers in treated tumors. Indeed, CD45⁺ cells were specifically increased upon Ad5-HRG treatment (Fig. 2D).

CSF-1 neutralization is an effective strategy for reducing macrophage recruitment though the downregulation of macrophage cell survival, proliferation, and differentiation (30). To determine the contribution of CSF1 receptor (CSF1R)–expressing macrophages/monocytes to CD45⁺ cells for the antitumor effect of HRG on glioma, we treated glioma-bearing animals with a neutralizing CSF1 antibody (Clone 5A1) in combination with intracerebral Ad5 vector injection. The αCSF1 Ab was delivered at a dose previously demonstrated to cause the specific loss of CSFR1⁺ cells and macrophages (3), one day after glioma cell inoculation and at days 6, 10, and 14 (Fig. 3A). Treatment with Ad5-HRG combined with control immunoglobulin (IgG) resulted in a 50% reduction in tumor size compared with the corresponding Ad5-Empty–treated animals (Fig. 3B). Importantly, treatment with Ad5-HRG/αCSF1 Ab did not further reduce glioma growth compared with Ad5-Empty/αCSF1 Ab or Ad5-Empty/control IgG (Fig. 3B) (31). Moreover, treatment with the αCSF1 Ab efficiently suppressed infiltration of CD45⁺ cells by Ad5-HRG (Fig. 3C). That HRG failed to further suppress glioma growth when combined with the CSF1-neutralizing antibody is consistent with an essential role for CD45⁺/CSF1R⁺ macrophages in the mechanism of action of HRG. In addition, CD8a⁺ T cells were reduced in Ad5-HRG/αCFS1 Ab–treated tumors (Fig. 3D). Although the changes in CD8a⁺ cell infiltration did not reach significant differences, the trends followed the same pattern as the effects of HRG on CD45⁺ cell infiltration, suggesting a possible relationship between these.

To assess the potential for cotreatment of HRG with chemotherapy, the DNA alkylating chemotherapeutic temozolomide, which is used routinely in the clinic for the treatment of glioblastoma multiforme, was tested in combinatorial treatment with Ad5-HRG. The estimated IC₅₀ for temozolomide was 9.58 mg/mL in the GL261 model (Supplementary Fig. S3A and S3B). Next, glioma-bearing mice were treated with Ad5-Empty or Ad5-HRG at days 7 and 13 in combination with a suboptimal IC₂₀ dose of temozolomide. Temozolomide was delivered in 2 cycles, 1 mg/kg/day for days 10 to 12 and days 16 to 18 (Supplementary Fig. S4A). This schedule was chosen to avoid, as much as possible, the induction of cytotoxicity in Ad5-transfected cells within the 48-hour period after Ad5 injection.

Consistent with previous results, Ad5-HRG treatment significantly attenuated tumor growth compared with Ad5-Empty–treated tumors (Supplementary Fig. S4B). Temozolomide treatment in Ad5-Empty–treated animals demonstrated no significant suppression of tumor growth, as expected given the suboptimal dose. However, temozolomide treatment negated the beneficial effect of Ad5-HRG treatment. Thus, the combination of Ad5-HRG and temozolomide resulted in the same extent of tumor growth as for Ad5-Empty–treated animals cotreated with DMSO/PBS vehicle (Supplementary Fig. S4B). Ad5-HRG treatment alone consistently resulted in a significant increase in both CD45⁺ cell infiltration (Supplementary Fig. S4C) and CD8a⁺ T cells (Supplementary Fig. S4D). The increase in both these cell populations in the tumor, essential for HRG's suppressive effect on tumor growth, was lost when combining Ad5-HRG with temozolomide treatment. These data further support an essential role for CD45⁺ cells in HRGs mechanism of action.

The mechanism underlying the highly reproducible effect of HRG on the immunomodulatory profile of inflammatory cells remains poorly understood. To identify HRG's potential interaction partners on inflammatory cells, we employed the novel protein interaction–decoding technology, TRICEPS-based LRC technology (15, 16). The TRICEPS cross-linker is equipped with three orthogonal functionalities; one arm for conjugation of the orphan ligand, in this case, HRG, a second arm for capturing oxidized carbohydrates on the putative receptor/secreted protein on living cells, and a third arm carrying a biotin tag for purification of captured receptors for subsequent analysis by quantitative mass spectrometry (MS; Fig. 4A). We applied TRICEPS-based LRC to the monocytic cell line, U937, whose differentiation to CD14⁺ macrophages is induced by Vitamin D₃ (VitD₃; ref. 29). TRICEPS-LRC identified a single interaction candidate for HRG on the cell surface of U937 cells, namely STC2 (Fig. 4B; see Table 1 for MS identification of all identified HRG-captured proteins). In addition, we identified HRG in the HRG-capture reaction. As expected, the insulin receptor was enriched in the insulin-captured control reaction but not in the HRG capture.

To validate the identified interaction between HRG and STC2, we used HEK293T cells endogenously expressing STC2, either lacking murine HRG expression (293WT) or overexpressing murine HRG (293HRG). The presence of HRG–STC2 complexes was examined by coimmunoprecipitation in cell culture lysate and medium. In addition, 6 hours prior to harvest, the cell culture medium was supplemented with recombinant STC2, or not, to enrich for the interaction. Anti-HRG immunoblotting of anti-STC2 immunoprecipitates demonstrated the presence of HRG in complex with STC2 in the medium from 293 HRG cultures. The HRG–STC2 interaction was enhanced by the addition of recombinant STC2 to the cell culture medium (Fig. 4C). Cell lysates treated in an identical manner demonstrated a similar, but nonsignificant, increase in HRG in STC2 immunoprecipitates. As HEK293 cells do not express HRG endogenously (9), these data suggest that the HRG–STC2 complex is established on the cell surface and then released into the medium.

To understand the relevance of STC2 expression in the glioma models, we immunostained tissue for the presence of murine STC2 in addition to CD45⁺ cells. STC2 positivity could be seen diffusely associated with glioma cells (Fig. 5A), consistent with its secreted nature, while more intense staining was associated with CD45⁺ cells (Fig. 5B). In anti-CSF1 antibody–treated gliomas, loss of CD45⁺ cells (see Fig. 3C) was accompanied by a loss in STC2 expression in the tumor (Fig. 5A; quantification).

To examine the consequence of the HRG-STC2 interaction on leukocytes, we utilized the U937 cell differentiation assay, on which TRICEPS-based LRC was also performed (see Fig. 4B). In this assay, U937 cells differentiate toward the monocyte/macrophage lineage as assessed by CD14 expression. In agreement with previous findings (13), the addition of HRG further potentiated VitD₃-induced differentiation (Supplementary Fig. S5) in a manner dependent on the VitD₃ dose. When differentiated at low concentrations of VitD₃ (5 nmol/L), STC2 dose dependently suppressed CD14 expression (Fig. 5C) as assessed by the number of CD14⁺ cells and the mean fluorescence intensity of positive cells. Importantly, when included with STC2 during cell differentiation, HRG completely negated the suppressive effect of STC2. These data demonstrate that the HRG–STC2 interaction is of functional consequence in modulating leukocyte gene expression.

Microarray transcript analysis on U937 cells exposed to the different treatments showed overlapping and distinct sets of regulated genes in the four conditions: (i) VitD₃ alone to induce monocytic differentiation of U937 cells; (ii) combined with HRG; (iii) combined with STC2; or (iv) combined with HRG + STC2, compared with undifferentiated control cells. The number of regulated genes in the treatment groups was (i) VitD₃ group: 538; (ii) VitD₃ + HRG: 538; (iii) VitD₃ + STC2: 595; and (iv) VitD₃ + HRG/STC2: 473 (Supplementary Fig. S6). Gene ontology (GO) analysis using ToppGene focused on GO categories differently regulated in the treatment groups (Supplementary Table S1). Further GO analysis was performed on the uniquely regulated genes in treatment groups 2 to 4 relative to VitD₃ alone (Supplementary Table S2A and S2B).

The differentially expressed VitD₃ + HRG genes relative to VitD₃ alone (group 2 vs. 1) had a limited number of GO annotations (Supplementary Table S2A, showing loss-of-function GO:s and S2B, showing gain-of-function GO:s). The differentially expressed VitD₃ + STC2 genes (group 3 vs. 1) had the largest number of GO annotations, indicating a broader repertoire of gene regulation by STC2 than HRG on inflammatory cells. Many of the STC2-regulated genes were related to inflammation, innate immune response, or acquired immune responses (51%). Surprisingly, both loss and gain of expressed genes in the VitD₃ + STC2 samples relative to VitD₃ (groups 3 vs. 1) had a strong immune component (Supplementary Table S2A and S2B). Importantly, inclusion of HRG in the VitD₃ + STC2 treatment (group 4) reduced the number of GO annotations compared with that in group 3, but the remaining gene regulation was still strongly enriched for processes related to inflammation or the immune systems (59%). Table 2 provides a summary of GO biological function described in detail in Supplementary Tables S1 and S2.

In conclusion, STC2 appeared to have a more dramatic effect than HRG on gene expression in U937 cells after VitD₃ differentiation; treatment with STC2 in the presence of HRG eliminated certain of these gene expression changes. Many of these gene regulatory events, both quantitative and qualitative, related to the immune system.

To validate the relevance of the HRG/STC2–regulated genes described above for tumor immune responsiveness, real-time qPCR was performed on GL261 tumors harvested after treatment with buffer, Ad5-Empty, or Ad5-HRG (Fig. 6A). Ad5-Empty treatment induced the expression of CD11b, CXCR4, IL6, and IFNγ. Ad5-HRG decreased gene expression of the M1 markers IL6 and IFNγ, whereas M2 markers were not statistically different (relative expression of Ad5-HRG–treated tumor arginase-1 mRNA was 1.25 ± 0.28 and MRC1 1.78 ± 0.44 compared with buffer only–treated tumors, respectively), suggesting that there is no M1 skewing of the immune system in this scenario. CD11b is a marker for myeloid cells (32), suggesting recruitment of macrophages to the tumor as a consequence of virus administration. CXCR4 is a receptor for CXCL-12 involved in immune cell retention in numerous settings (33). IL6 and IFNγ are proinflammatory cytokines, suggesting that Ad5 primed an immune response from CXCR4-positive cells and/or myeloid cells. Treatment with Ad5-HRG suppressed the gene regulatory effects to the levels seen in buffer-treated glioma, and therefore, these cytokines are likely not directly involved in HRG's tumor-suppressive effects but rather reflect a suppression of a general immune response against adenoviruses.

In the absence of data supporting an M1 skewing of the innate immune response, we postulated that HRG effects on CD45/CD8 infiltration and myeloid cell differentiation in the GL261 glioma counteracted tumor-induced immunosuppression. Absence of effects of HRG on gene expression of the immune checkpoint inhibitors PD-L1, PD-1, and CTLA-4 relative tumors exposed to buffer treatment only (relative expression of Ad5-HRG–treated tumors relative buffer only-treated tumors was 0.7 ± 0.5 for PD-L1, 1.2 ± 0.8 for PD-1 and 0.9 ± 0.4 for CTLA-4) targeted our focus to T regulatory cells (Treg) as possible mediators. Tregs are known to provide an immunosuppressive tumor environment, promoting tumor growth (34, 35). We first noted that cells could be detected in the tumors that coexpressed the two IL35 components, Ebi-3 and IL12a (Fig 6B). These were too infrequent to allow quantification by immunofluorescence staining of sections, and instead, FACS analysis was employed for accurate assessment of their presence. Figure 6C describes the gating strategy after staining dispersed GL261 gliomas for CD4, FoxP3, Ebi3, and IL12a. As shown in Fig. 6D, HRG-treated gliomas exhibited fewer IL35-expressing Tregs, providing an explanation for the increased infiltration of CD45/CD8⁺ cells and reduced tumor growth.

The mechanistic understanding of HRG's antitumor effects, underlying the exploration of its usefulness as a cancer therapeutic, derives from studies on peripheral, non-CNS tumor models in mice. The effect of HRG in tumor growth and spread has been studied by delivering purified recombinant protein, HRG-derived peptides, or by expressing HRG in tumor cells (9, 11, 13, 36). One study demonstrated that HRG expression in an orthotopic murine glioma model (RCAS/TV-A) resulted in a reduction in tumor incidence and, strikingly, the absence of grade 4 lesions, as assessed by mitotic index and vessel density (37). This indicated that HRG may be therapeutically beneficial also for tumors in the CNS. Herein, we have developed an adenovirus-based vector encoding HRG to allow for local delivery of HRG to the tumor site, thereby circumventing the need for systemic delivery. HRG gene therapy demonstrated significant and robust inhibition of tumor growth. HRG-treated gliomas displayed an increased number of perfused vessels, in line with previous observations of the ability of HRG to normalize the tumor vasculature (11). CD45⁺ and T-cell numbers were increased in HRG-treated tumors compared with controls. The immunoregulatory cytokine profile in HRG-treated gliomas showed that HRG suppressed immunomodulatory effects seen in gliomas treated with the Ad5 virus alone. Changes in Ad5-Empty–treated gliomas were compatible with inflammatory cell recruitment. In contrast, Ad5-HRG administration reduced expression of Tregs expressing the immunosuppressive dimeric cytokine IL35, composed of IL12a and Ebi3/IL27β (38). Tregs expressing this cytokine have in many instances been shown to promote tumor expansion (34, 35). The related cytokine IL12, composed of IL12α and IL12β on the other hand, is a potent inducer of antitumor immunity (39). It was recently shown that oncolytic virus delivery of IL12 to murine glioma results in enhanced T effector–mediated antitumor response and suppressed Treg infiltration (40). Moreover, the combination of oncolytic IL12 delivery with neutralization of checkpoint inhibitors PD-1 and CTLA-4 dramatically blocks murine glioma growth (41). The strong preclinical effects of checkpoint inhibitors have promoted clinical trials for treatment of glioblastoma multiforme either as monotherapy or in a wide range of combinatorial treatments (42). Treg-suppressive treatment, in the longer perspective possibly based on a HRG mimetic, may have a very good potential to be efficacious in such combinatorial glioma therapy.

Ad5-HRGs mechanism of action in gliomas replicated that reported previously for peripheral tumors as depletion of CSFR1⁺ cells negated the beneficial effect of HRG (3, 11). The use of the chemotherapeutic temozolomide in combination with HRG also antagonized the effects of HRG, likely through a similar mechanism, that is, the depletion of CSFR1⁺ cells. Indeed, a number of murine studies have demonstrated the significant depletion of monocytes, and consequently macrophages, following temozolomide treatment, even at subtherapeutic concentration (43). Although this study highlights the potential utility of HRG for glioma treatment, it also identifies potential challenges when combined with monocyte/macrophage cell targeting therapeutics, particularly of note because temozolomide is the first-line grade 4 glioma chemotherapeutic.

Despite the robust effects of HRG on tumors both peripherally and in the CNS, there still remains a lack of understanding of exactly how HRG exerts its effects on inflammatory cells. To address this important question, we used the recently developed protein interaction decoding technology, TRICEPS-based LRC (15, 16, 18), to identify HRG interaction partners on the cell surface of an inflammatory cell model. TRICEPS is a trivalent linker molecule simultaneously binding the orphan ligand, the N-glycosylated putative receptor/interaction partner at the cell surface, as well as streptavidin for purification. HRG-coupled TRICEPS followed by MS identified STC2 as a previously unknown interaction partner candidate of HRG in differentiated U937 cells.

The interaction between HRG and STC2 was further validated by coimmunoprecipitation. In VitD₃-treated U937 cells, STC2 negatively regulated CD14⁺ cell differentiation, in line with the described ability of peptides derived from STC1 and STC2 to directly affect inflammatory cell function (44). Crucially, transcript analyses in U937 cells treated with HRG or STC2 individually or in combination indicated that HRG modulated certain immunoregulatory changes seen in STC2-treated cells. Whether the HRG/STC2 complex neutralizes STC2 directly or whether HRG modulates STC2′s biology, for example, by regulation of a potential STC2 receptor on inflammatory cells, is unknown at this point. It is noteworthy that although HRG can be retained by binding to cell surface–expressed heparan sulfate (4), specific, saturable binding of HRG occurs to inflammatory cells, independently of heparan sulfate (3).

Taken together, our data suggest that STC2 and HRG can directly interact and modulate the phenotype of inflammatory cells. Another possible mechanism of action for the HRG–STC2 complex may involve the reported cytoprotective effects of STC2 (45). Thus, HRG might sensitize tumor cells to ER stress by neutralizing STC2. HRG may also synergize with other small-molecule effectors of ER stress, for example, small-molecule plant compounds that bind to GRP78 (46). An interesting compound in this regard is Honokiol, which confers immunogenicity by regulating calreticulin exposure, activating ER stress, and inhibiting epithelial-to-mesenchymal transition (47).

In vivo, the significance of the HRG/STC2 interaction remains to be explored fully. The observation that both tumor cells and CD45⁺ cells expressed STC2 in the glioma model shows (i) that STC2 is present in the tumor microenvironment where HRG is known to act and, moreover, (ii) that STC2 is expressed by inflammatory cells, which HRG has been shown to directly influence. Thus, HRG may steer tumor development by direct interaction with STC2 on or proximal to inflammatory cells, modulating their phenotype.

Here, we have demonstrated the applicability of HRG therapy in an adenovirus gene therapy–based system, showing that local HRG delivery to the tumor site is a feasible approach. We have successfully applied this to an in vivo glioma model and demonstrated that the mechanism of action is in part distinct relative HRG's effects in peripheral tumor models. Although HRG promotes efficient tumor vessel perfusion and infiltration of CD8⁺ cells in both situations, the M₁–M₂ polarization noted in peripheral tumors (11) is not established in gliomas. Instead, in gliomas, HRG acts to balance the immunomodulatory effects of STC2, correlating with a reduction in Treg infiltration. It remains to be assessed as to whether HRG therapy, in gliomas, acts through both microglia and infiltrating macrophages and whether gene expression changes in these two distinct populations differentially contribute to HRGs effects. This study also highlights the potential antagonism between HRG and therapies that target macrophage/myeloid cell populations while simultaneously potentially promoting immune checkpoint therapy.

Cancer is a leading cause of death in the industrialized world and to an increasing extent a cause of death in developing countries. Although our understanding of common features of signaling and metabolic aberrations in cancer is increasing, it is clear that cancer is highly heterogenous and adapts in response to therapies. Novel therapies, possibly administered in a sequential manner and acting to induce tissue homeostasis, are therefore urgently needed (48). HRG gene therapy in the CNS offers novel and distinct advantages above current therapeutics. The ability to normalize the glioma vasculature has the potential to enhance therapeutic delivery similar to the effect of the anti-VEGF antibody bevacizumab. In addition, the suppression of immunosuppressive TAMs or Tregs in the tumor may enhance antitumor immune response, which may be particularly relevant in combination with immunostimulatory antibody therapeutics. Finally, the identification of STC2 as a partner for HRG in regulation of CD45⁺ cell phenotype presents a new concept where serum proteins appearing in different constellations serve as endogenous rheostats to steer the tumor inflammation profile in pro- or antitumor directions.

N. Sobotzki is an investment associate at M Ventures B.V. B. Wollscheid and N. Sobotzski have ownership interest (including stock, patents, etc.) in and B. Wollscheid is a consultant/advisory board member for Dualsystems Biosystems Biotech AG. No potential conflicts of interest were disclosed by the other authors.

Conception and design: F.P. Roche, I. Pietilä, E.O. Sjöström, N. Sobotzki, M. Welsh, L. Claesson-Welsh

Development of methodology: F.P. Roche, E.O. Sjöström, O. Noguer, M. Essand, M. Welsh

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F.P. Roche, I. Pietilä, H. Kaito, E.O. Sjöström, N. Sobotzki, O. Noguer, T.P. Skare, B. Wollscheid, M. Welsh

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F.P. Roche, I. Pietilä, H. Kaito, N. Sobotzki, T.P. Skare, B. Wollscheid, M. Welsh, L. Claesson-Welsh

Writing, review, and/or revision of the manuscript: F.P. Roche, I. Pietilä, H. Kaito, N. Sobotzki, O. Noguer, T.P. Skare, B. Wollscheid, M. Welsh, L. Claesson-Welsh

Study supervision: L. Claesson-Welsh

Other (performed, analyzed, and provided the TRICEPS-related experiments and data): N. Sobotzki, B. Wollscheid

We thank Marie Hedlund, Uppsala University, for expert technical assistance with the glioma model. We also thank Pernilla Martinsson, Uppsala University for expert technical assistance on biochemical analyses and Berith Nilsson, Uppsala University, for expert technical assistance in the production of recombinant adenoviruses. We also acknowledge the contribution of the Array and Analysis Facility, Uppsala University. In particular, expert evaluation by Dr. Pascal Pucholt, Dept. Medical Sciences and Dr. Lei Lui Conze, Dept. Immunology, Genetics and Pathology, is greatly appreciated. This study was supported by grants from the Swedish Cancer Society (16 0585), the Swedish Research Council (2015-02375_3) and a Wallenberg Scholar grant from the Knut and Alice Wallenberg foundation to L. Claesson-Welsh, from the Swedish Cancer Society (16 0520), and the Swedish Research Council (2016-01085) to M. Welsh, and funding from the Swiss National Science Foundation (#31003A_160259) to B. Wollscheid.

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.