Overexpression of HOXC10 promotes angiogenesis in human glioma via interaction with PRMT5 and upregulation of VEGFA expression.

High levels of angiogenesis are associated with poor prognosis in patients with gliomas. However, the molecular mechanisms underlying tumor angiogenesis remain unclear.
Methods: The effect of homeobox C10 (HOXC10) on tube formation, migration, and proliferation of human umbilical vein endothelial cells (HUVECs) and on chicken chorioallantoic membranes (CAMs) was examined. An animal xenograft model was used to examine the effect of HOXC10 on xenograft angiogenesis or the effect of bevacizumab, a monoclonal antibody against vascular endothelial growth factor A (VEGFA), on HOXC10-overexpressing xenografts. A chromatin immunoprecipitation assay was applied to investigate the mechanism in which HOXC10 regulated VEGFA expression.
Results: Overexpressing HOXC10 enhanced the capacity of glioma cells to induce tube formation, migration and proliferation of HUVECs, and neovascularization in CAMs, while silencing HOXC10 had the opposite result. We observed that CD31 staining was significantly increased in tumors formed by HOXC10-overexpressing U251MG cells but reduced in HOXC10-silenced tumors. Mechanistically, HOXC10 could transcriptionally upregulate VEGFA expression by binding to its promoter. Strikingly, treatment with bevacizumab, a monoclonal antibody against VEGFA, significantly inhibited the growth of HOXC10-overexpressing tumors and efficiently impaired angiogenesis. Protein arginine methyltransferase 5 (PRMT5) and WD repeat domain 5 (WDR5), both of which regulate histone post-translational modifications, were required for HOXC10-mediated VEGFA upregulation. Importantly, a significant correlation between HOXC10 levels and VEGFA expression was observed in a cohort of human gliomas. Conclusions: This study suggests that HOXC10 induces glioma angiogenesis by transcriptionally upregulating VEGFA expression, and may represent a potential target for antiangiogenic therapy in gliomas.

Introduction

Gliomas represent the most common primary tumor of the central nervous system 1, accounting for 30% of all brain and central nervous system tumors, and 80% of all malignant brain tumors 2, 3. Despite advances in surgical techniques and therapeutic strategies over the last two decades, the outcomes of patients harboring malignant gliomas remain very poor, with a cumulative one-year survival rate below 30% 4, 5. Glioblastoma multiforme (GBM), a grade IV glioma, is the most aggressive type of primary brain tumor 6, with an average survival time following diagnosis of 12 to 15 months 5. This poor prognosis is largely attributable to aberrant angiogenesis, high invasiveness and therapeutic resistance 1, 7. Uncontrolled proliferation of glioma cells requires angiogenesis; among different types of cancers, glioma displays the highest degree of vascularization 1, 8, 9. Increased tumor microvascular density (MVD) has been identified as a hallmark in gliomas with a poor patient prognosis 10, 11. Therefore, a better understanding of the molecular mechanisms underlying glioma angiogenesis could contribute to the identification of novel therapeutic targets to treat glioma and eventually improve prognosis.

Angiogenesis, which refers to the formation of new blood vessels from pre-existing vasculature, occurs in many physiological and pathological processes, such as development, wound healing, and cancer. Abnormal angiogenesis is necessary for the growth, survival, and metastasis of most solid tumors, and involves the coordinated roles of many bioactive molecules 12. For instance, an imbalance between pro-angiogenic factors, such as the vascular endothelial growth factor (VEGF) family, and inhibitors of angiogenesis, including endostatin, angiostatin, and other related molecules, results in abnormal angiogenesis 13. The VEGF family proteins, which include VEGFA, VEGFB, VEGFC, VEGFD, and placental growth factor, are the best-characterized regulators of angiogenesis and metastatic growth in human carcinogenesis 14-16. However, the precise molecular mechanisms underlying tumor angiogenesis are not completely understood.

There are 39 class I homeobox (HOX) genes in the human genome and they are distributed in four different clusters (A, B, C, and D), located on four chromosomes (2, 7, 12, and 17). HOX genes are evolutionarily conserved genes that make important contributions to cell differentiation and embryonic development 17-19. Previous studies reported that HOX genes have various influences on tumor cell proliferation, apoptosis, stem cell renewal, differentiation, motility, and angiogenesis 19-22. HOXC10, a member of HOX gene family, encodes a transcription factor containing a conserved DNA binding homeodomain and is highly expressed in many cancers 23-25. Feng et al. reported that HOXC10 expression is appreciably increased in human thyroid cancer tissues compared to that in normal human thyroid tissues 24. Zhai and colleagues reported that elevated HOXC10 is associated with increased invasiveness of cervical cancer-derived cell lines and that knockdown of HOXC10 markedly impaired cervical cancer cell invasiveness 23. Moreover, HOXC10 expression is dramatically elevated in lung adenocarcinoma compared to normal lung tissues and upregulation of HOXC10 correlated with unfavorable prognosis in patients with lung cancer 25. Moreover, upregulated HOXC10 expression promotes proliferation, migration, and invasion in gliomas26. However, the precise biological role of HOXC10 in glioma angiogenesis remains largely undefined.

In the present study, we reported that HOXC10 was markedly overexpressed in glioma cells and clinical samples compared to primary normal human astrocytes (NHAs) and normal human brain tissues, respectively. Overexpression of HOXC10 significantly promoted, whereas knockdown of HOXC10 impaired, glioma angiogenesis in vitro and in vivo. Furthermore, we demonstrated that HOXC10 upregulated VEGFA expression via directly binding to the VEGFA promoter, eventually promoting angiogenesis in glioma. Therefore, our results establish a vital role for HOXC10 in regulating glioma angiogenesis, suggesting that HOXC10 might serve as a potential target for glioma therapy.

Methods

Cell culture and reagents

Primary NHAs were purchased from Sciencell Research Laboratories (Carlsbad, CA, USA). Glioma cell lines A-172, U87, LN229, and LN-18, were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and U251MG cells were obtained from ECACC (Salisbury, UK). All cells were cultured under the conditions recommended by the manufacturer. Cell lines were authenticated using STR method. Bevacizumab (avastin) was purchased from Genentech (San Francisco, CA, USA). OICR-9429 was purchased from Selleck (Houston, TX, USA).

Patient information and tissue specimens

A total of 94 paraffin-embedded and archived glioma samples, including pilocytic astrocytoma, diffuse fibrillary astrocytoma, anaplastic astrocytomas and glioblastoma multiforme, which were histopathologically and clinically diagnosed at the First Affiliated Hospital of Sun Yat-sen University, from January 2008 to December 2014, were examined in this study. Prior informed patient consent and approval from the Institutional Research Ethics Committee were obtained for the use of these clinical materials for research purposes.

Vectors, retroviral infection, and transfection

The human HOXC10 gene and its truncations were amplified from cDNA using PCR and cloned into the pSin-EF2-puro retroviral vector. HOXC10-targeting short hairpin RNA (shRNA) oligonucleotides sequences were cloned into a pSuper-retro-puro vector to generate pSuper-retro-HOXC10-RNAi(s). Transfection of plasmids or siRNAs was conducted using Lipofectamine 3000 reagent (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions. Stable cell lines expressing HOXC10 or HOXC10 shRNA were selected using 0.5 μg/mL puromycin for 10 days from 48 h after infection.

Western blotting analysis

Western blotting (WB) was performed as previously described 27, using anti-HOXC10, anti-Flag, anti-protein arginine methyltransferase 5 (PRMT5), anti-glutathione-s-transferase (GST), anti-hemagluttinin (HA), anti-VEGFA and anti-WD repeat domain 5 (WDR5) antibodies (Abcam, Cambridge, MA, USA). Membranes were stripped and re-probed with an anti-α-tubulin antibody (Sigma-Aldrich, St Louis, MO, USA) as a protein loading control.

Human umbilical vein endothelial cells (HUVECs) tubule formation assay

Matrigel (200 µL; BD Biosciences, Mountainview, CA, USA) was added into wells of a 24-well plate and polymerized for 30 min at 37 °C. HUVECs (2×104) in 200 µL of conditioned medium were added to each well and incubated at 37 °C in 5% CO2 for 20 h. Images were captured under bright-field illumination using a Zeiss Axio Imager A1 microscope (Carl Zeiss, Jena, Germany) at 100× final magnification. The degree of tube formation in the current study was verified as compared to the formation of cell strings.

Chicken chorioallantoic membrane (CAM) assay

To evaluate the direct effect on angiogenesis, a CAM assay was performed at the 8th day of development of fertilized chicken eggs, as previously described 28. A 1-cm diameter window was opened in the shell of each egg with an 8-day-old chicken embryo (Yueqin Breeding Co. Ltd, Guangdong, China). The surface of the dermic sheet on the floor of the air sac was removed to expose the CAM. A 0.5 cm diameter filter paper was first placed on top of the CAM, and 100 μL of conditioned medium harvested from transduced glioma cells was added onto the center of the paper. After the window was closed with sterile adhesive tape, the eggs were incubated at 37 °C under 80-90% relative humidity for 48 h. Following fixation with stationary solution (methanol: acetone = 1:1) for 15 min, the CAM was cut and harvested, and gross photos of each CAM were taken using a digital camera (Panasonic, Osaka, Japan). The effect of conditioned media was evaluated by the number of second- and third-order vessels in comparison with that treated with the medium harvested from the vector-control group. Statistical analysis of second- and third-order vessels was carried out using a two-tailed Student's t-test.

Gene set enrichment analysis (GSEA)

Gene set enrichment analysis was performed as previously described 29. The public microarray data (GSE7696) from the GEO database (https://www.ncbi.nlm.nih.gov/gds) was downloaded and HOXC10 expression was treated as a numeric variable. Then, a continuous-type CLS file of the HOXC10 profile to phenotype labels in GSEA was applied. The metric for ranking genes in GSEA was set as “Pearson”, and the other parameters were set to their default values. GSEA was performed using GSEA 2.0.9 software (http://www.broadinstitute.org/gsea/).

Animal xenografts

Five-week-old female BALB/c nude mice were purchased from Beijing Vital River (Beijing, China). Briefly, 5×105 of the indicated glioma cells stably expressing luciferase were stereotactically injected into the right cerebrum of mice (n = 6/group). Five days after implantation, mice were intraperitoneally treated with IgG or Bevacizumab (10 mg/kg) every three days. Intracranial tumor growth was monitored in vivo using bioluminescence IVIS imaging (Xenogen, Almeda, CA) equipped with Living Image software (Xenogen). Briefly, mice were anesthetized with isofluorane and subsequently received intraperitoneal injection with luciferin (75 μg/g body weight). Images were acquired 5-10 min after luciferin administration. A photographic image of the animal was taken in the chamber under dim illumination, followed by acquisition and overlay of the pseudocolor image representing the spatial distribution of photon counts produced by active luciferase within the animal. A signal averaging time of 1 min was used for luminescence image acquisition. Signal intensity was quantified as the sum of all detected photon counts (photons per second) within a region of interest prescribed over the tumor site using living image software Xenogen. Three brains of each group of mice were harvested on day 21 and subjected to immunohistochemistry (IHC) and hematoxylin and eosin (H&E) staining. IHC was performed using anti-HOXC10 polyclonal antibody (Abcam, ab153904), anti-CD31 polyclonal antibody (Abcam, ab28364), and anti-VEGFA monoclonal antibody (Abcam, ab1316, [VG-1]). Quantification of microvessel density was performed in accordance with a previous study 30. Briefly, areas of greatest vessel density were then examined under higher magnification (×100) and counted. Any distinct area of positive staining for CD31 was counted as a single vessel. Results were expressed as the mean number of vessels ± SD per high-power field (×100). Experiments associated with animals complied with the Laboratory Animal Care and Use Guide and the Institutional Research Ethics Committee approved these experiments.

3-(4, 5-Dimethyl-2-thiazolyl)-2, 5-diphenyl-2H- tetrazolium bromide (MTT) assay

Cells (2×103) were seeded into 96-well plates and stained at the indicated time point with 100 μL of sterile MTT (Sigma-Aldrich) dye (0.5 mg/mL) for 4 h at 37 °C, followed by removal of the culture medium and the addition of 150 μL of dimethyl sulfoxide (Sigma-Aldrich). The absorbance was measured at 570 nm, with 655 nm used as the reference wavelength.

Transwell migration assay

Cells (2×104) were plated on the top side of a polycarbonate Transwell filter in the upper chamber of a BioCoat Invasion Chamber (BD Biosciences) and incubated at 37 °C for 22 h. The cells remaining on the upper surface were removed with cotton swabs. Cells that had migrated to the lower membrane surface were fixed in 1% paraformaldehyde, stained with crystal violet, and counted under an optical microscope (×200 magnification). Cell counts are expressed as the mean number of cells from 10 random fields per well.

Luciferase reporter assay

Cells were seeded in 48-well plates (3×103 cells/well) in triplicate and allowed to settle for 24 h. Luciferase reporter plasmids (100 ng) with 1 ng of pRL-TK Renilla plasmid (Promega, Madison, WI, USA) were transfected into glioma cells using Lipofectamine 3000 reagent. Luciferase and Renilla signals were determined 24 h after transfection using a Dual Luciferase Reporter Assay Kit (Promega).

Chromatin immunoprecipitation assay

Cells (2×106) plated in a 100 mm culture dish were treated with 1% formaldehyde to cross-link proteins to DNA. The cell lysates were sonicated to shear the DNA into 300-1000 bp lengths. Aliquots containing equal amounts of chromatin supernatants were incubated on a rocking bed at 4 °C overnight with either 1 μg of anti-Flag antibody (Sigma-Aldrich), anti-polymerase II antibodies (Abcam), or 1 μg anti-IgG antibody (Abcam) as a negative control. Following reverse cross-linking of protein-DNA complexes to free the DNA, real-time PCR was carried out. The primers used in this study are listed in the supplementary materials.

Co-immunoprecipitation analysis

The indicated protein in the cell lysates was immunoprecipitated using antibodies including anti-Flag (Sigma-Aldrich), anti-HA (Sigma-Aldrich), and anti-HOXC10 antibody (Novus Biologicals, Littleton, CO, USA) antibodies. The immunocomplexes were then examined using antibodies targeting Flag (Sigma-Aldrich), PRMT5 (Abcam), HOXC10 (Novus Biologicals), and HA (Sigma-Aldrich), and visualized using an ECL detection system.

Far-western blotting analysis

The assay was performed using a standard protocol, according to a previous study 31. Briefly, a plasmid encoding Flag-tagged HOXC10 was transfected into U251MG cells, and 48 h later, U251MG cells were immunoprecipitated using Flag-tag affinity gel (Sigma-Aldrich). Membranes were incubated with 5 μg/mL of recombinant GST-PRMT5. After TBS-T washes, the polyvinylidene fluoride (PVDF) membrane was incubated with anti-GST antibodies. After development and imaging, the anti-GST antibody was stripped and membranes were incubated with anti-Flag antibodies.

Statistical analysis

Data were evaluated using Fisher's exact test, the log-rank test, the Chi-squared test, and Student's two-tailed t-test. Statistical analyses were performed using the SPSS 21.0 statistical software package (IBM Corp, Armonk, NY, USA). Data are represented as the mean ± SD. P < 0.05 was considered statistically significant.

Results

HOXC10 plays an important role in glioma progression

It has been reported that elevated expression of HOXC10 induced glioma proliferation, migration and invasion26. Therefore, we verified whether HOXC10 was upregulated in gliomas via analyzing public glioma datasets from The Cancer Genome Atlas (TCGA) and the Gene Expression Omnibus (GEO; GSE13276 and GSE4290). As expected, HOXC10 was significantly elevated in gliomas, especially in glioblastoma, compared to non-tumor brain tissues and normal brain tissues, respectively (Figure ). Furthermore, Kaplan-Meier analysis of data from the TCGA and HG-UG133A dataset indicated that GBM or low-grade gliomas (LGG) patients with higher HOXC10 expression presented shorter overall survival than those with lower HOXC10 expression (Figure ). Moreover, we found that HOXC10 expression was not only correlated with poorer survival of patients with astrocytoma but also correlated with poorer survival of patients with oligodendroglioma analyzed in TCGA dataset, which indicated that HOXC10 is involved in the progression of both astrocytomas and oligodendrogliomas (Figure ).

To further determine whether HOXC10 expression was correlated with aggressive clinicopathological characteristics of gliomas, we performed western blotting analysis and real-time PCR analysis. The results indicated that the expression levels of HOXC10 in 12 human glioma tissues were higher than those in four normal brain tissues (Figure ). We also performed western blotting analysis and real-time PCR analysis in NHAs and multiple glioma cell lines, including A-172, U87, LN229, T98G, LN-18, and U251MG. The glioma cell lines showed higher protein levels of HOXC10 than the NHAs (Figure ). Next, we performed IHC analysis to examine the level of HOXC10 in 94 paraffin-embedded, archived glioma tissues and normal brain tissues (Table ). As shown in Figure , the average staining intensity of HOXC10 in Grade I to IV tumors was noticeably stronger than that in normal brain tissue. Quantitative analysis of the mean optical density (MOD) indicated that the HOXC10 level increased in parallel with disease grade in glioma, revealing a positive correlation between HOXC10 level and glioma progression. Furthermore, upregulation of HOXC10 correlated significantly with advanced disease grade and poor survival in patients with glioma (Table ). Based on Kaplan-Meier analysis, higher HOXC10 expression predicted shorter survival of patients with all WHO Grade, WHO Grade I-II, and WHO Grade III-IV gliomas (Figure ). However, we did not observe HOXC10 level significantly correlated with IDH and 1p/19q status (Table and Table ). Importantly, univariate and multivariate analysis indicated that HOXC10 might serve as a prognostic factor in gliomas (Table ). Taken together, these results strongly suggested that HOXC10 might play pivotal roles in glioma progression.

Dysregulation of HOXC10 alters the ability of gliomas to induce angiogenesis

A wealth of literature supports the view that homeobox genes are highly expressed in several types of cancer and play a crucial role in angiogenesis 32-34. To validate the effect of HOXC10 on angiogenesis of intracranial xenograft, U251MG and LN229 glioma cell lines stably expressing HOXC10 or HOXC10-targeting shRNA were established (Figure ). Glioma cells, transduced with the control vector, the HOXC10-overexpression vector, the control shRNA vector, or the HOXC10-sh#1 vector, were stereotactically implanted into mouse brains (n = 6/group). Tumors formed by HOXC10-overexpressing U251MG cells were significantly larger than those formed by the control cells, as indicated by luminescence and H&E staining (Figure ). Conversely, silencing HOXC10 yielded smaller tumors than the control shRNA vector (Figure ). Furthermore, overexpression of HOXC10 shortened, while silencing HOXC10 prolonged the survival of tumor-bearing mice compared to the control shRNA group (Figure ). Moreover, we found that microvascular density was efficiently enhanced in tumors formed by HOXC10-overexpressing tumors and reduced in HOXC10-silenced tumors, as shown by CD31 staining (Figure ).

Consistent with the in vivo results, overexpressing HOXC10 significantly promoted, but silencing HOXC10 prominently inhibited, the capability of tube formation, migration ability, and proliferation of HUVECs (Figure and Figure ) and the formation of second- and third-order vessels in the chicken CAM assay (Figure ). Gene set enrichment analysis (GSEA) results also revealed that angiogenesis gene signatures (LU_TUMOR_ ANGIOGENESIS_UP, ANGIOGENESIS, HALLMARK_ANGIOGENESIS, GO_ ANGIOGENESIS) were significantly elevated in the higher HOXC10 expression group compared to the lower HOXC10 expression group (Figure ). Collectively, our results suggested that HOXC10 overexpression contributes to angiogenesis in gliomas.

HOXC10 modulates angiogenesis through regulating VEGFA expression

Previous studies have suggested that the expression of VEGFA is biologically and clinically involved in the angiogenesis of tumor cells 35. Interestingly, three potential HOXC10 binding sites were predicted in the promoter of VEGFA in the JASPAR database (Figure ). As expected, the expression of VEGFA, at both mRNA and protein levels, was consistently upregulated in HOXC10-overexpressing glioma cells, and downregulated in HOXC10-silenced glioma cells compared to control cells (Figure ). Importantly, silencing VEGFA or treatment with bevacizumab, a monoclonal antibody against VEGFA, could reverse the effects of HOXC10 overexpression-mediated tube formation, migration, and proliferation of HUVECs (Figure ). Importantly, we found that tumors formed by VEGFA-silenced, HOXC10-overexpressing cells were significantly smaller than those formed by control-shRNA-transduced cells (Figure ). Mice bearing VEGFA-silenced U251MG/HOXC10 cells had longer survival than those bearing control-shRNA-transduced cells (Figure ). The microvascular density was significantly reduced in tumors formed by VEGFA-silenced, HOXC10-overexpressing U251MG cells (Figure ). Strikingly, treatment with bevacizumab significantly inhibited the growth of HOXC10-overexpressing tumor and efficiently impaired angiogenesis (Figure ). These results provide strong evidence that VEGFA is the key effector of HOXC10-induced angiogenesis and bevacizumab might have potential therapeutic value for the treatment of HOXC10-induced glioma.

HOXC10 regulates VEGFA promoter activity in glioma cells

A luciferase reporter assay was performed and showed that overexpression of HOXC10 promoted, while silencing HOXC10 repressed, the luciferase activity of the VEGFA promoter (Figure ). To identify the site that HOXC10 binds to in the VEGFA promoter, fragments of the VEGFA promoter, namely P1(nucleotides -2056 to +176), P2 (nucleotides -1751 to +176), P3 (nucleotides -931 to +176), P4 (nucleotides -226 to +176), P5 (nucleotides -2056 to -907) and P6 (nucleotides -732 to -208), were inserted into the luciferase reporter pGL3 vector, respectively (Figure ). We observed that overexpression of HOXC10 increased, while silencing of HOXC10 expression decreased, the luciferase activity of VEGFA promoter fragments P1, P2, P3, P6, with the luciferase signals of P4 and P5 being largely unaffected (Figure ). These results suggest that the HOXC10 binding site may be located in nucleotides -292 to -283 (Figure ). Meanwhile, mutation of nucleotides -289 to -286 reversed the effect of overexpression or knockdown of HOXC10 on the luciferase activity of P1 (Figure ). Moreover, a chromatin immunoprecipitation (CHIP) assay showed that HOXC10 bound to region 3 within the VEGFA promoter region (Figure ), which agreed with the luciferase assay results and suggested that HOXC10 upregulates VEGFA expression by binding to the VEGFA promoter. In addition, we found that the recruitment of RNA polymerase II (Pol II) to the VEGFA promoter was enhanced in HOXC10-transduced cells, but impaired in HOXC10-silenced cells (Figure ). Taken together, our results provide corroborating evidence that HOXC10 transcriptionally upregulates VEGFA expression.

HOXC10 upregulates VEGFA expression by interacting with PRMT5

To further study the mechanism by which HOXC10 upregulates VEGFA expression, systematic ChIP was performed to probe histone post-translational modifications, including H3R2me1, H3R2me2s, H3R8me2s, H2R3me2s, and H4R3me2s. Interestingly, the enrichment of H3R2me1 and H3R2me2s, both of which are PRMT5-catalyzed histone post-translational modifications, in VEGFA promoter was markedly increased in HOXC10-transduced cells, but strikingly decreased in HOXC10-silenced cells (Figure ). Furthermore, we found that recruitment of PRMT5 to the VEGFA promoter was significantly enhanced in HOXC10-overexpressing glioma cells, but markedly suppressed in HOXC10-silenced glioma cells (Figure ). Moreover, knockdown of PRMT5 strongly reversed the HOXC10-overexpressing effect on VEGFA enrichment (Figure ). These results indicate that PRMT5 might play a role in HOXC10-mediated VEGFA upregulation.

To verify the hypothesis that HOXC10 induced VEGFA expression via forming a complex with PRMT5, co-immunoprecipitation (co-IP) assay was carried out and the results demonstrated that HOXC10 formed a complex with PRMT5 in U251MG and LN229 cells (Figure ). Far-western blotting further demonstrated that HOXC10 directly interacted with PRMT5 (Figure ). To determine the fraction of HOXC10 responsible for their interaction, several HOXC10 truncations were generated (Figure ). Co-IP assays showed that PRMT5 interacted with HOXC10-FL (Full Length), HOXC10-F1 and HOXC10-F4, suggesting that a region between aa 77 and aa 197 is required for HOXC10-PRMT5 interaction (Figure ). Overexpressing HOXC10-FL and HOXC10-F1 markedly promoted the luciferase activity of VEGFA, while overexpressing HOXC10-F2, HOXC10-F3, HOXC10-F4, all of which lacked the PRMT5 binding site or homeobox, could not activate the VEGFA promoter-dependent luciferase activity (Figure ). Consistently, VEGFA expression, at both mRNA and protein levels, was significantly increased in the HOXC10-FL- and HOXC10-F1-overexpressing cells but decreased in the cells transfected with HOXC10-F2, or HOXC10-F3 (Figure ). Consistent with the inhibitory effect of HOXC10-F4 on the interaction of HOXC10 and PRMT5, we also found that overexpression of HOXC10-F4 peptide significantly abrogated HOXC10-induced VEGFA upregulation and reduced the ability of HOXC10-transduced cells to induce tube formation and proliferation of HUVECs, which provided further evidence that HOXC10 activates VEGFA transcription by interacting with PRMT5 (Figure ). These results suggested that PRMT5 is required for HOXC10-mediated upregulation of VEGFA.

PRMT5 and WDR5 are required for HOXC10-mediated VEGFA transcription

Next, we further investigated the effect of PRMT5 on HOXC10-mediated VEGFA upregulation. As shown in Figure and Figure , knockdown of PRMT5 significantly reduced VEGFA expression and inhibited the occupancy of H3R2me1 and H3R2me2s on VEGFA in HOXC10-overexpressing glioma cells, with no change in control marker H3. PRMT5-methylated H3R2me1 may lead to WDR5-induced H3K4 methylation and transcriptional activation of target genes 36. To investigate whether WDR5 and H3K4me3 are involved in HOXC10-mediated VEGFA transcriptional regulation, ChIP-qPCR for H3K4me3, Polymerase II, and WDR5 was conducted. As expected, overexpressing HOXC10 strongly increased, while silencing HOXC10 prominently decreased, the abundance of H3K4me3, Polymerase II, and WDR5 on the VEGFA promoter in U251MG and LN229 cells (Figure ). Moreover, occupancy of H3K4me3, Polymerase II, and WDR5 on the VEGFA promoter were dramatically attenuated in PRMT5-silenced, HOXC10-transduced glioma cells (Figure and Figure ). Furthermore, silencing WDR5 profoundly impaired the stimulatory effect of overexpressing HOXC10 (Figure ). Importantly, OICR-9429, an inhibitor of WDR5 interaction with the H3 tail, markedly decreased the expression of VEGFA and the occupancy of H3K4me3 and Polymerase II on the VEGFA promoter (Figure ). These results suggested that HOXC10 interacts with PRMT5, which promotes H3R2me1 and H3R2me2s enrichment on the VEGFA promoter and recruits WDR5 for subsequent methylation of H3K4 at the VEGFA promoter, eventually resulting in VEGFA transcription.

Clinical relevance of HOXC10 and VEGFA expression in human glioma

Finally, we determined whether the HOXC10/VEGFA axis was clinically relevant. As shown in Figure , IHC analysis showed that HOXC10 levels were positively correlated with the expression of VEGFA in human glioma specimens (P < 0.01). Consistent with the IHC results, western blotting analysis demonstrated that levels of HOXC10 also correlated strongly with the levels of VEGFA (r = 0.755, P = 0.012) in 10 freshly collected clinical glioma samples (Figure ). Furthermore, statistical analysis revealed that HOXC10 expression was significantly correlated with VEGFA expression in the TCGA-LGG group (n = 516, r = 0.260, P < 0.01). Although HOXC10 expression in the entire TCGA-GBM group was not significantly correlated with VEGFA expression (n = 156, r = 0.146, P = 0.069), we found that HOXC10 expression in the PRMT5 higher-GBM subgroup was significantly correlated with VEGFA levels (n = 78, r = 0.279, P = 0.014), which further supported that PRMT5 contributes to HOXC10-mediated VEGFA upregulation (Figure ). These results strongly supported the notion that overexpression of HOXC10 transcriptionally upregulates VEGFA, which promotes angiogenesis and ultimately leads to poor survival of patients with glioma.

Discussion

Glioblastoma, the most frequent type of primary brain tumor, is one of the most aggressive and deadly human cancers 37. Angiogenesis has been reported to be closely involved in glioma growth, invasion, metastasis, and decreased patient survival 38, 39. Therefore, inhibiting the angiogenic process might be used as a potential therapy to prevent tumor growth and metastasis. However, the molecular mechanisms involved in regulating glioma angiogenesis remain largely unclear. The present study revealed that higher expression of HOXC10 was significantly correlated with a poorer prognosis of patients harboring gliomas. Additionally, our results indicated that upregulated HOXC10 promotes angiogenesis in human gliomas in vitro and in vivo via upregulating VEGFA expression.

Homeobox genes are highly expressed in several types of cancer and play a crucial role in angiogenesis 32-34. Seki et al. reported that HOXB9-positive breast tumors present a significant increase in vasculature, and Kaplan-Meier analysis revealed that HOXB9-positive breast cancer patients have poor prognosis 32. Zhu and colleagues reported that HOXB13 expression is correlated positively with tumor VEGF levels and microvessel density in hepatocellular carcinoma (HCC), and univariate and multivariate analysis showed HCC patients with higher HOXB13 expression exhibits poorer overall survival and disease-free survival 33. Additionally, Storti et al. reported that HOXB7 is overexpressed in more than 40% myeloma cell lines compared to normal plasma cells, and knockdown of HOXB7 expression in multiple myeloma cells represses the production of angiogenic factors and impairs their pro-angiogenic properties 34. In this study, we demonstrated that higher expression of HOXC10 correlated significantly with advanced grade and poorer prognosis in patients with gliomas. In addition, HOXC10 modulated the VEGFA signaling pathway to mediate an angiogenesis phenotype in human gliomas, further indicating that HOXC10 might represent a promising therapeutic target for anti-angiogenic strategies against gliomas.

Angiogenesis consists of multiple processes, including endothelial cell proliferation, migration, and alignment to form capillary-like structures. It has been reported that anti-angiogenic therapy holds promise for treating malignancies, including gliomas 40. Although initial responses to anti-angiogenic therapy are often significant, the benefits are transitory and are followed by a restoration of tumor growth and progression41. Two large randomized clinical trials of bevacizumab therapy had no significant effect on prolonging survival as part of first line therapy for glioblastomas in humans 42, 43, which suggested that the established glioblastoma cell lines and the employed mouse models might not represent the native characteristics of glioblastoma cells and native environment of glioblastomas in human brain. Indeed, the high passage established glioblastoma cell lines showed substantial alterations of morphology, genome, and functional characteristics compared to the glioblastoma cells present in the glioblastoma mass 44. It has been reported that the original low-passaged U-251 cells maintained a DNA copy number resembling a typical GBM profile, whereas all long-term subclones lost the typical glioblastoma profile, demonstrated by array comparative genomic hybridization (aCGH). In addition, long-term subclones exhibited variations in phenotypic marker expression and increased growth rate in vitro and a more aggressive growth in vivo 45. However, glioma primary cultures showed genetic make-ups that were similar to those of patient tumors. For instance, Balbous et al. reported that COL1A1 and IFITM1 genes identified using gliomaspheres derived from glioma patients could be considered for use in stratifying patients with glioblastoma into subgroups for risk of recurrence at diagnosis and prognostic evolution 46. These results suggested that patient-derived primary glioma cells are more appropriate to be used as models for studies of glioblastoma biology. Meanwhile, the human cell line-derived xenografts showed a level of angiogenesis and circumscribed growth with variable extents of collective tissue invasion, which exhibited limited single cell infiltration in the brain and lacked necrosis and microvascular abnormalities, both of which were characteristics of human glioblastoma 47. Moreover, glioblastoma PDX models predict poor prognosis of patients with glioblastoma and the model can be used to identify the gene signatures and pathways signatures associated with the clinical aggressiveness of glioblastomas 48. Therefore, patient-derived glioblastoma cells or patient-derived glioblastoma xenograft models need to be used to further examine the promotory effect of HOXC10 overexpression on angiogenesis and VEGFA upregulation.

The PRMT5 gene encodes an arginine methyltransferase, which catalyzes the transfer of methyl groups to the amino acid arginine in target proteins, including histones, and promotes target gene transcription. A wealth of literature supports the view that PRMT5 plays a vital role in regulating cancer progression via histone methylation 49-51. PRMT5 promotes the cancer stem cell function of breast cancer by enhancing the enrichment of H3K4me3 on the FOXP1 promoter and subsequently driving FOXP1 transcription and expression 49. SHARPIN augments invasiveness of lung cancer by interacting with PRMT5, which methylates histones H3 and H4, subsequently upregulating metastasis-related gene expression 50. Furthermore, PRMT5 is recruited through interaction with Sp1 and activates the androgen receptor by symmetrically dimethylating H4R3, which significantly enhanced the growth of prostate xenograft tumors in mice 51. Our results showed that HOXC10 interacted with PRMT5, leading to enrichment of H3R2me1, H3R2me2s, and H3K4me3 on the VEGFA promoter, which eventually resulted in increased VEGFA expression and enhances growth of gliomas. Thus, our results provide an additional molecular mechanism in angiogenesis and show that targeting PRMT5 might be an efficient therapeutic approach.

VEGF-A has been demonstrated to show prominent activity with vascular endothelial cells, mainly through its interactions with the VEGFR1 and -R2 receptors. However, bevacizumab, a recombinant humanized monoclonal antibody that blocks angiogenesis by inhibiting secreted VEGF-A, did not improve survival in humans with glioblastoma. Interestingly, it has been recently also reported that the expression levels of different vascular endothelial growth factor A (VEGF) isoforms were associated with angiogenesis and patient prognosis in human cancers 52-54. Two novel splicing VEGF isoforms L-VEGF144 and L-VEGF138, which were transcripted at the first upstream CUG codon, had nuclear localization ability and were also involved in angiogenic processes in glioblastoma53, 54. Herein, we demonstrated that HOXC10 transcriptionally upregulates VEGFA expression by binding to its promoter and silencing HOXC10, which significantly decreased VEGFA expression and angiogenesis, implicating that HOXC10 may regulate expression of various splicing VEGF isoforms, which needs to be further investigated in future studies.

In conclusion, our findings suggest that HOXC10 is overexpressed in malignant glioma tissues and correlates with aggressive clinicopathological characteristics. Overexpression of HOXC10 enhances, while downregulation of HOXC10 impairs, pro-angiogenesis activity in gliomas. Importantly, expression of HOXC10 correlated strongly with the level of VEGFA in clinical glioma samples. Taken together, these findings provide a new strategy to treat malignant glioma, which might lead to improved prognosis in patients with glioma.

Supplementary figures and tables.

Click here for additional data file.