Phospholipase D1 is upregulated by vorinostat and confers resistance to vorinostat in glioblastoma

Abstract Glioblastoma (GBM) is an aggressive brain tumor and drug resistance remains a major barrier for therapeutics. Epigenetic alterations are implicated in GBM pathogenesis, and epigenetic modulators including histone deacetylase (HDAC) inhibitors are exploited as promising anticancer therapies. Here, we demonstrate that phospholipase D1 (PLD1) is a transcriptional target of HDAC inhibitors and confers resistance to HDAC inhibitor in GBM. Treatment of vorinostat upregulates PLD1 through PKCζ‐Sp1 axis. Vorinostat induces dynamic changes in the chromatin structure and transcriptional machinery associated with PLD1 promoter region. Cotreatment of vorinostat with PLD1 inhibitor further attenuates invasion, angiogenesis, colony‐forming capacity, and self‐renewal capacity, compared with those of either treatment. PLD1 inhibitor overcomes resistance to vorinostat in GBM cells intracranial GBM tumors. Our finding provides new insight into the role of PLD1 as a target of resistance to vorinostat, and PLD1 inhibitor might provide the basis for therapeutic combinations with improved efficacy of HDAC inhibitor.

. Various HDAC inhibitors such as vorinostat (suberoylanilide hydroxamic acid, SAHA) and valproic acid are currently being tested in clinical trials on GBM (Chinnaiyan et al., 2012;Friday et al., 2012;Galanis et al., 2009;Moroni et al., 2013). Although the use of HDAC inhibitor as monotherapy in the clinic has been validated in cutaneous T-cell lymphoma, they are less effective against solid tumors (Lee, Kim, Kummar, Giaccone, & Trepel, 2008;Rasheed, Johnstone, & Prince, 2007). However, the mechanism of HDAC inhibitor resistance in solid tumors is not well-elucidated and a better understanding will improve their clinical efficacy (Fantin & Richon, 2007). Accordingly, elucidation of resistance markers and its molecular mechanisms can lead to strategies to maximize the therapeutic efficacy of HDAC inhibitor by combining agents that target factor(s) associated with resistance. Phospholipase D (PLD) hydrolyzes phospholipid to generate phosphatidic acid, a lipid second messenger, and two isoforms of phosphatidylcholinespecific PLD, PLD1 and PLD2 have been identified (Frohman, 2015). PLD is upregulated in various cancers and implicated in tumor malignancy, maintenance of self-renewal of cancer stem cells, and resistance to radiotherapy and chemotherapy (Brown, Thomas, & Lindsley, 2017;Cheol Son et al., 2013;Kang, Choi, & Min, 2014;Kang et al., 2015;Kang, Lee, Hwang, 2017). PLD is known to increase the invasive migration and proliferation of GBM (Bruntz, Taylor, Lindsley, & Brown, 2014;O'Reilly et al., 2013;Sayyah et al., 2014). However, it is unknown whether PLD confers chemoresistance to GBM. In the present study, our goal was to investigate the effect of PLD1 on resistance to vorinostat in GBM and how vorinostat is responsible for the upregulation of PLD1.

| Transfection and luciferase reporter assays
Following the manufacturer's instructions, luciferase reporter of PLD1 promoter (pGL4-PLD1; Kang et al., 2008) and expression plasmids were transiently transfected into cells with Lipofectamine 3000 (Invitrogen) reagents. Relative luciferase activity was obtained by normalization of firefly and Renilla luciferase activity. Dual-luciferase assay kits (E1910) were purchased from Promega.

| In vitro limiting dilution assay (LDA)
To determine the number of sphere-forming units (SFU), in vitro LDA was performed as previously described (Kang et al., 2015).
The average number of SFU counted upon replating of 10 LDAs derived from single spheres constituted the in vitro self-renewal assay.

| Colony-forming assay
For colony formation assays, the cells were seeded into six-well plates (2.5 × 10 4 cells per well) and treated with the indicated agents in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. After 14 days, the cells were fixed in 4% paraformaldehyde in phosphate-buffered saline for 10 min at room temperature and stained with 0.5% crystal violet in 20% methanol for 20 min. Images were captured using a flatbed scanner, and the cells were dissolved with 20% acetic acids in 20% methanol for 30 min. 550 | 2.7 | PLD activity assay PLD activity was assessed by measuring the formation of [ 3 H]phosphatidylbutanol, the product of PLD-mediated transphosphatidylation, in the presence of 1-butanol as previously described (Kang, Lee, Hwang, et al., 2017).

| Invasion assay
Invasion assays were performed as described previously (Kang et al., 2011). The extent of invasion, which was defined as movement of cells from the upper chamber to the lower chamber, was expressed as an average number of cells per microscopic field.

| Apoptosis assay
Apoptotic cell death was measured by APC-conjugated anti-Annexin V Apoptosis Detection Kit I (550474; BD Bioscience). The terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) assay was performed using In Situ Cell Death Detection Kit, POD (Roche), according to the manufacturer's protocol.

| Subcutaneous xenograft and intracranial tumor formation
Xenograft tumors were generated by subcutaneous injection of 1 × 10 7 U87 cells. Tumors were measured with calipers to estimate their volumes. GBM cells were injected intracranially using a stereotactic device at a depth of 3 mm into the right forebrains of 9-10 weeks old athymic nude mice or syngeneic C57BL/6 mice (5 × 10 5 cells/mouse). The mice were anesthetized with tribromoethanol (250 mg/kg; Sigma-Aldrich) intraperitoneally. The mice were injected intraperitoneally with vorinostat (SML0061, 20 mg/kg; Sigma-Aldrich), PLD1 inhibitor (VU0155069, 13206, 10 mg/kg; Cayman Chemical), or vorinostat/PLD1 inhibitor three times per week for 4 weeks. The protocol and procedures for animal studies were ethically reviewed and approved by the Institutional Animal Care Committee of Pusan National University.

| Statistical analysis
Data were analyzed using Student's t test, and correlation coefficients were calculated using Spearman's r. Survival probability of mice bearing intracranial GBM cell lines, defined as the time from brain resection to death, was analyzed using Kaplan-Meier, and differences were evaluated using the log-rank test. Statistical analysis was performed using GraphPad Prism 5.0 (GraphPad Software).

| Vorinostat upregulates the expression of PLD1 via PKCζ signaling pathway
We investigated whether HDAC inhibitors affect the expression of PLD. Treatment with vorinostat, trichostatin A (TSA), or sodium butyrate (NaB) in U87MG cells, increased expression of PLD1 but not PLD2, as analyzed by quantitative PCR and western blot ( Figure 1a).
As a control, HDAC inhibitors increased the expression of p21 and acetylated histone 4 (Ac-H4) (Figure 1a). HDAC inhibitor-induced PLD1 upregulation was also observed in various human or murine GBM cells ( Figure S1a 3.2 | Vorinostat induces marked Sp1 phosphorylation dependent on PKCζ, and Sp1 is required for vorinostat induction of PLD1 Sp1-dependent gene activation via HDAC inhibition has been observed (Gui, Ngo, Xu, Richon, & Marks, 2004;Yokota et al., 2004). Therefore, we examined whether vorinostat induction of PLD1 is mediated by Sp1 transcriptional factor using the pharmacologic inhibitor, mithramycin (MTM), which interferes with the binding of Sp1 to GC-rich promoters, only exists in the cytosol, but is also present in the nucleus, strengthening the concept that Sp1 could be a nuclear target of PKC (Zhang, Liao, & Dufau, 2006). Moreover, Sp1 interacts with PKCζ, and vorinostat increased the association between Sp1 and PKCζ ( Figure 2f). We found two putative Sp1-binding sites in the PLD1 promoter ( Figure 2g). The reporter gene assay using a series of 5΄-deletion constructs of the PLD1 promoter showed that the region from −1,887 to −1,290, which contains two putative Sp1-binding sites, is involved in vorinostat-mediated PLD1 promoter activation ( Figure S2). As a positive control, the expression of Sp1 reporter gene was enhanced by vorinostat. Moreover, mutation of two Sp1-binding sites, Sp1-A or Sp1-B significantly attenuated vorinostat-induced PLD1 promoter activity, respectively ( Figure 2g).
Taken together, these results suggest the functional importance of the PKCζ-Sp1 signaling axis in the vorinostat-induced induction of PLD1.

| Combination of vorinostat with depletion or inhibition of PLD1 promotes cell death of GBM
We next examined whether vorinostat-induced PLD1 expression is responsible for increased PLD activity. As shown in Figure 4a, vorinostat stimulated the enzymatic activity of PLD, which was inhibited by PLD1 depletion using two kinds of short hairpin RNA. Moreover, vorinostat-induced PLD activation was decreased by PLD1 inhibitor, VU0155069 (Figure 4b). Thus, it is suggested that vorinostat-induced PLD1 expression is associated with increased PLD activity. As PLD is known to protect anticancer drug-induced cell death (Kang et al., 2014), we further examined whether vorinostat-induced PLD1 expression is associated with chemoresistance. Vorinostat below 2 μM did not affect the viability of U87 cells but vorinostat above 3 μM reduced the viability of the glioma cells (Figure 4c). A combination of varinostat with depletion or inhibition of PLD1 significantly decreased the viability of U87 cells, compared with that of either treatment (Figure 4c,d). Moreover, a combinational treatment of ChIP assays for binding of the indicated proteins to PLD1 promoters in U87 cells treated with vorinostat (2 μM) for 6 hr, after which a single or double ChIP assay was performed using the indicated antibodies. The GC-rich region of the PLD1 promoter was used as a control. The intensity of the indicated bands was normalized to the intensity of their respective α-tubulin bands and quantified against each other. Results are representative of at least three independent experiments, and shown as the mean ± SEM. *p < .05; **p < .01; ***p < .001. CBP; CREB-binding protein; ChIP, chromatin immunoprecipitation; HAT, histone acetyltransferase; HDAC, histone deacetylase; PCAF, p300/CBP-associated factor; PLD1, phospholipase D1; n.s., not significant; SEM, standard error of the mean alone inhibited the growth of tumors (Figure 4g). The combination of the two drugs was more effective at reducing tumor formation than being used alone. Furthermore, the combined treatment in the mice promoted apoptosis as analyzed by TUNEL assay (Figure 4h). Collectively, these results suggest that combined treatment of vorinostat with PLD1 inhibition promotes cell death of GBM.

| Combinational treatment of vorinostat with PLD1 inhibitor further suppresses invasion and angiogenesis
We further investigated the combinational therapeutic effect of vorinostat and PLD1 inhibitor against invasion and angiogenesis. (g) Athymic nude mice were injected subcutaneously with U87 cells (n = 7/group). Mice were subjected to intraperitoneal injection with vehicle, VU0155069 (10 mg/kg) alone, vorinostat (20 mg/kg) alone, or in combination three times a week for 27 days. The tumor volume of mice was measured with vernier calipers every 3 days. (h) The paraffin-embedded tumor sections were analyzed by TUNEL assay. Results are representative of at least three independent experiments, and shown as the mean ± SEM. *p < .05; **p < .01; ***p < .001. GBM, glioblastoma; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; n.s., not significant; PLD1, phospholipase D1; SEM, standard error of the mean; shRNA, short hairpin RNA; TUNEL, terminal deoxynucleotidyl transferase dUTP nick-end labeling F I G U R E 5 Combinational treatment of vorinostat with PLD1 inhibitor suppresses invasion and angiogenesis. (a) The cells were seeded in matrigel-coated invasion chambers and treated with vorinostat (2 μM) and/or VU0155069 (10 μM) for 24 hr. The extent of invasion was expressed as an average number of cells per microscopic field. The cells were treated with the indicated drug (s) for 36 hr. Conditioned medium was collected and applied to HUVEC, and then migration (b) and tube formation (c) were measured. (d) U87 cells were treated with the indicated drug(s) for 36 hr and then secretion of VEGF was quantified by ELISA. (e) Inhibitory effects of PLD1 inhibitor and vorinostat on angiogenesis in U87 and U251 CAM-implanted tumors. After the cells were loaded (1.5 × 10 6 cells/CAM) onto CAMs, the indicated drug(s) were administered at the time of implantation. Five days after implantation, CAM were resected and imaged under the microscope. Tumor vasculature and the number of vessels were analyzed. The data represent the mean ± SEM of at least six chick embryos. Results are representative of at least three independent experiments, and shown as the mean ± SEM. *p < .05; **p < .01; ***p < .001. ELISA, enzyme-linked immunosorbent assay; HUVEC, human umbilical vein endothelial cells; PLD1, phospholipase D1; SEM, standard error of the mean; VEGF, vascular endothelial growth factor 556 | significantly suppressed by treatment with PLD1 inhibitor. Vorinostat had a marginal effect on the tumor-induced neovascularization.
The combined treatment significantly suppressed neovascularization when compared with monotherapy ( Figure 5e). Therefore, the potential anticancer efficacy of the combined treatment with vorinostat and PLD1 inhibitor regimens is linked to inhibitory effects against invasion, migration, and angiogenesis.
3.6 | Combinational therapy of vorinostat with PLD1 inhibitor efficiently attenuates the tumorigenic potential of GBM Combination with alkylating drug, TMZ, and ionizing radiation (IR) is currently used as a standard treatment for GBM. We tried to investigate whether PLD1 inhibitor or vorinostat affect the standard chemoradiotherapies in GBM and TMZ-resistant GBM. As analyzed by colony-forming capacities, U251-TMZ-R cells showed resistance to TMZ, TMZ/IR, or TMZ/IR/vorinostat, relative to U251 cells ( Figure 6a,b). PLD1 inhibitor only treatment significantly suppressed colony-forming capacities in both U251 and U251-TMZ-R, and combination of TMZ/IR/vorinostat with PLD1 inhibitor markedly reduced the colony-formation, compared with that of TMZ/IR/vorinostat in U251-TMZ-R (Figure 6a,b). We further investigated the therapeutic effect in patient-derived GBM cell lines. Genome-wide transcriptome analyses have suggested that GBM can be divided into four clinically relevant subtypes: classic, mesenchymal (MES), neural and proneural (PN) GBM (Phillips et al., 2006;Verhaak et al., 2010).
We examined the colony-forming capacities using MES and PN subtype of GBM. GBM-PN-528 (PN subtype GBM) and GBM-MES-83 (MES subtype GBM) showed more resistance to TMZ at higher concentration (50 and 100 μM) (Figure 6c,d). Combination of TMZ/IR or TMZ/IR/vorinostat showed resistance in GBM-MES-83, relative to GBM-PN-528. Actually, MES subtype of GBM is known to be more aggressive and radio-resistant than PN subtype of GBM (Mao et al., 2013). PLD1 inhibitor alone significantly suppressed the colony formation, and a combination of TMZ/IR/vorinosta with PLD1 inhibitor dramatically abolished the colony-forming capacities compared with that of TMZ/IR/vorinostat (Figure 6c

| DISCUSSION
In the present study, we demonstrate that PLD1 acts as a novel transcriptional target of HDAC inhibitors and confers resistance to vorinostat in GBM. GBM, a very aggressive brain tumor remains one of the deadliest of malignancies, with limited treatment options and a high rate of recurrence, and thus represents an urgent unmet medical need (Wen & Kesari, 2008). GBM recurrence is linked to the epigenetic mechanisms and cellular pathways (Esteller, 2008).
Consequently, multidisciplinary research efforts, including epigenetic modalities-HDAC inhibitors such as vorinostat, are certainly needed. HDAC inhibitors are epigenetic agents that target the aberrant epigenetic characteristics of the tumor cells. However, molecular determinants of resistance to HDAC inhibitors are poorly understood. A better understanding of the mechanisms that determine resistance to HDAC inhibitors would provide the basis for therapeutic combinations with improved clinical efficacy. PLD has been reported to be intimately associated with the signaling pathways modified in GBM (Bruntz et al., 2014;Colman et al., 2003;Kang et al., 2014;Mathews et al., 2015). Moreover, PLD1 inhibitor exhibits potent anticancer activity in a patient-derived xenograft model harboring APC tumor suppressor and PI3KCA mutation, which results in hyperactivation of the mitogenic Wnt/β-catenin and PI3K/Akt signaling pathways , suggesting that inhibition of PLD1 might overcome limited clinical benefit due to drug resistance. Therefore, our finding of HDAC- group) were intracranially transplanted into the brains of immunocompromised mice and then treated intraperitoneally with vehicle, VU0155069 (10 mg/kg) alone, vorinostat (20 mg/kg) alone, or in combination three times a week for 4 weeks. Representative images of H&E-stained sections of mouse brains. (g) Survival of mice was evaluated (n = 6/group, Kaplan-Meier model with two-sided log-rank test). **p < .01; ***p < .001. Results are representative of at least three independent experiments, and shown as the mean ± SEM. **p < .01; ***p < .001. GBM, glioblastoma; H&E, hematoxylin and eosin; IR, ionizing radiation; MES, mesenchymal; PLD1, phospholipase D1; SEM, standard error of the mean; TMZ, temozolomide cancer and treatment responsiveness is undoubtedly complicated. PLD1 inhibition augments the efficacy of anticancer regimens via facilitation of autophagic pathways (Jang et al., 2014). The control of autophagy might also be used as a therapeutic strategy to treat cancer cells that are resistant to cell death. Our findings show that vorinostat-induced PLD1 upregulation plays a pivotal role in protection from apoptosis. Furthermore, combination of the drugs significantly suppressed invasion, angiogenesis, colony formation, self-renewal capacity of GBM, and intracranial GBM tumor formation. PLD1 inhibition overcame resistance to conventional therapeutic treatment of GBM. As cancer stem cells contributes to drug resistance, targeting PLD1 effectively might overcome GBMmediated therapeutic resistance. As PLD1 is a new target of vorinostat resistance, and combinational therapy of PLD1 inhibitor with vorinostat might be a potential therapeutic strategy against GBM tumorigenesis, it would be interesting to know whether it is possible to develop some biomarkers of therapeutic efficacy that could facilitate a more precise selection of the most suitable candidates for innovative combination therapy.

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author D.S.M. upon reasonable request.