TWIST1-EP300 Expedites Gastric Cancer Cell Resistance to Apatinib by Activating the Expression of COL1A2

The association between collagen type I alpha (COL1A) and chemoresistance has been verified in cancers. However, the specific role of COL1A2 in gastric cancer (GC) cell resistance to apatinib, a highly selective small-molecule inhibitor of vascular endothelial growth factor receptor 2, has not been investigated before. The purpose of this study was to explore the potential factors associated with COL1A2 regulation on GC cell apatinib resistance in vitro. With the aid of the Oncomine database and integrated bioinformatics methods, we identified COL1A2 overexpression in GC and its prognostic value. Mechanistically, the COL1A2 promoter has a distinct H3K27ac modification site and that E1A binding protein p300 (EP300) and twist family bHLH transcription factor 1 (TWIST1) can bind to the COL1A2 promoter, which in turn transcriptionally activated COL1A2 expression. In addition, overexpression of COL1A2 significantly promoted resistance to apatinib in GC cells, but knockdown of EP300 or TWIST1 remarkably inhibited COL1A2 expression and promoted sensitivity of GC cells to apatinib. Our findings demonstrated that the combination of EP300 and TWIST1 has a synergistically regulatory effect on COL1A2 expression, thus contributing to apatinib resistance in GC cells.


Introduction
In China, gastric cancer (GC) ranks the second place in the incidence and mortality, just behind lung cancer [1]. Patients with locally advanced or metastatic diseases have a somber prognosis, with a median overall survival of 10-12 months, and chemotherapy is the corner stone of treatments [2]. The regimen combining a fluoropyrimidine + a platinum-based agent + an anthracycline or docetaxel is a standard first-line therapy, and second lines of therapy may include irinotecan and taxanes [3]. After failure of second-line treatment, the results of further chemotherapies are disappointing, and no chemotherapeutic agents showed a clear benefit in prolonging survival. Therefore, the discovery of novel and effective treatment options is clinically needed for metastatic GC.
Apatinib mesylate, with a chemical name of N-[4(1cyano-cyclopentyl)phenyl]-2-(4-pyridylmethyl)amino-3-pyridine carboxamide mesylate, is an orally administered small-molecule inhibitor that targets vascular endothelial growth factor receptor 2 (VEGFR2) [4]. With a favorable side effect profile and improved outcomes, apatinib has demonstrated a considerable potential to enhance therapeutic options in various tumor types [5]. More importantly, apatinib improved the median overall survival of patients with advanced GC after two or more lines of prior chemotherapy of approximately 3 months [6]. Unfortunately, the antiangiogenic strategy has a major practical limitation: resistance inevitably develops [7]. However, the underlying molecular mechanism of apatinib resistance in GC remains largely unclear. Collagen type I (COL1) is the most abundant matrix protein in the cancer stroma, and COL1 could promote tumor progression by facilitating processes, such as cancer cell growth, invasion, and metastasis, and by supporting anticancer drug resistance [8]. Interestingly, COL1A2 was found as a gene with greater than a 50-fold increase in expression in cisplatin-, paclitaxel-, doxorubicin-, topotecan-, vincristine-, and methotrexate-resistant ovarian cancer cells [9]. Consistently, COL1A2 was identified as one of the 21 upregulated genes in 3 datasets between GC tissues and adjacent normal tissues [10]. However, there is little information regarding the regulatory role of COL1A2 in apatinib resistance. Mechanistically, T-box transcription factor 3 was revealed to bind to and activate the COL1A2 promoter in fibrosarcoma and chondrosarcoma [11]. Therefore, we suspected that there is also a transcription factor regulating the COL1A2 expression in GC as well. In the present study, we explored the detailed molecular mechanism of COL1A2 in apatinib-resistant GC. Our study might provide new insights into the chemoresistance of GC at the molecular level and improve the outcomes of GC patients by exploring potential therapeutic targets to overcome apatinib resistance.

Cell
2.6. Colony Formation Assay. The AGS/MKN-45 cells were dispersed into a single cell suspension and seeded into 6well plates (400 cells/2 mL/well) and gently rotated to distribute the cells evenly. After maintaining at 37°C and 5% CO 2 for 24 h, the cells were incubated with RPMI-1640 containing 10% FBS and 1 μM apatinib for 2 weeks. When colonies were visible in the Petri dish with the naked eyes, the cells were carefully washed with phosphate-buffered saline (PBS) twice and fixed with pure methanol for 20 min. Olympus Optical Co., Ltd., Tokyo, Japan), and cells were counted in five randomly selected microscopic fields.
2.9. Chromatin Immunoprecipitation (ChIP). ChIP was performed using the Magna ChIP™ A/G One-Day ChIP Kit (Catalog #17-10085; Merck) according to the manufacturer's instructions. Cross-linking of chromatin was achieved by using 1% formaldehyde for 10 min at 37°C and neutralizing with glycine for 5 min at room temperature. All GC cells were washed with cold 1 mL PBS + proteinase inhibitor (1 mM phenylmethanesulfonyl fluoride, 1 mg peptidase, and 1 mg gastrase inhibitor A). The cells were centrifuged at 716 × g for 5 min at 4°C and disrupted using sodium dodecyl sulfate (SDS) lysis buffer (1% SDS, 10 mM ethylenediaminetetraacetic acid and 50 mM Tris-Hcl pH = 8:0). Subsequent sonication at 150 Hz was performed with four sets of 10 s pulse shears on ice using a high-intensity ultrasonic processor (Cole-Parmer). Equal amounts of chromatin were immunoprecipitated with antibodies targeting TWIST1 (1 : 500, #GTX60776, GeneTex) or H3K27ac (1 : 500, #GTX50903, GeneTex) overnight at 4°C with anti-mouse IgG (1 : 500, ab18413; Abcam) as an isotype control. The total chromatin was used as input. Immunoprecipitation products were collected after incubation with magnetic beads. Beads were washed using a magnetic separation rack, and bound chromatin was eluted in ChIP elution buffer with a proteinase K mixer according to the manufacturer's instructions.
2.10. Dual-Luciferase Reporter Assay. The targeting relationship between TWIST1 and COL1A2 was verified using a dual-luciferase reporter gene assay. The pGL3-COL1A2 wild-type (WT) luciferase reporter gene vector was designed and synthesized based on the binding sequence of the COL1A2 promoter region to TWIST1. The pGL3-COL1A2 plasmid was then cotransfected with TWIST1 overexpression vectors at different doses into HEK293T cells (ACS-4500, American Type Culture Collection, Manassas, VA, USA). The cells were lysed after 48 h. The supernatant was obtained by centrifugation at 12,000 rpm for 1 min. Luciferase activities were measured using the Dual-Luciferase® Reporter Assay System (E1910, Promega Corp., Madison, WI, USA), with renilla luciferase as an internal control.
2.12. Data Analysis. In this study, data were analyzed with SPSS software (version 21.0, IBM Corp., Armonk, NY, USA). All experiments were repeated at least three times, and the results were showed as the mean ± standard deviation (SD). Unpaired t-test and one-way or two-way analysis of variance with Tukey's correction were applied to analyze the statistics. When p value was less than 0.05, the differences between each group were considered significant.

COL1A2 Is Highly Expressed in GC Patients and
Associated with Poor Prognosis and Apatinib Resistance. In order to explore the pathogenesis and treatment of GC, we first analyzed the differentially expressed genes in Chen Gastric, Cho Gastric, Cui Gastric, Deng Gastric, and Wang Gastric datasets by cancer vs. normal analysis in the Oncomine database (Figure 1(a)). We screened out that COL1A2 was the gene with the most pronounced expression differences. We further found in TCGA-STAD database that the expression of COL1A2 was significantly increased in GC tissues and was associated with poor prognosis of patients (Figures 1(b) and 1(c)). Subsequently, we further observed from the HPA website (https://www.proteinatlas.org/) that the intensity of COL1A2 staining was much stronger in GC tissues than in paracancerous tissues ( Figure 1(d)). Subsequently, we exposed AGS and MKN-45 cells to graded concentrations of apatinib for the construction of cell lines resistant to apatinib. After 6 months of continuous culture, we examined the IC50 values of parental and resistant cells to apatinib using the CCK-8. We found that the sensitivity of resistant cells to apatinib was significantly reduced (p < 0:01) (Figure 1(e)), indicating that the resistant cell line was successfully constructed. In addition, a significant increase in COL1A2 expression at mRNA and protein levels was observed in drug-resistant cells (all p < 0:01) (Figures 1(f) and 1(g)).

TWIST1 Is Positively Correlated with COL1A2 and
Transcriptionally Activates COL1A2. To explore the upstream regulatory mechanism of COL1A2, we conducted a series of prediction and assays. Firstly, we queried the promoter of COL1A2 located in Chr7:94394582~94394928 (Figure 2(a)) from the Ensembl website. Subsequently, we further analyzed the top 30 genes with the highest association with COL1A2 in the datasets of Chen Gastric, Cho Gastric, Cui Gastric, Deng Gastric, and Wang Gastric by coexpression analysis in the Oncomine website. We screened out TWIST1 (Figure 2(b)) from them. Moreover, in TCGA-STAD database, a positive correlation between TWIST1 and COL1A2 was identified using Spearman's correlation analysis (Figure 2(c)). Also, we found that the expression of TWIST1 in drug-resistant cells was also significantly higher than that in parental cells (all p < 0:01) (Figures 2(d) and 2(e)).
Thus, we performed ChIP-qPCR using antibody against TWIST1 in order to verify whether TWIST1 can bind to the promoter of COL1A2 as a transcription factor, and we observed that anti-TWIST1 can enrich more COL1A2 promoter fragment relative to IgG (all p < 0:01) (Figures 2(f) and 2(g)). To further verify that TWIST1 can transcriptionally activate COL1A2 expression, we designed pGL3-Luc luciferase reporter vector containing COL1A2 promoter sequence (Figure 2(h)) and cotransfected it into 293T cells together with different concentrations of TWIST1 overexpression plasmids. It was observed that as the concentration of TWIST1 increased, the luciferase activity in the cells had a significant increase (p < 0:01) (Figure 2(i)). The above results indicated that TWIST1 can transcriptionally activate the expression of COL1A2.

The COL1A2 Promoter Has a Distinct Enhancer-Like
Signature. We further found a significant H3K27ac modification near the promoter of COL1A2 through the UCSC browser website (http://genome.ucsc.edu/index.html) (Figure 3(a)). Subsequently, we downloaded the ChIP-seq data of H3K27ac from the GEO database for GC tissues with corresponding paracancerous tissues (GSE144385 and GSE162213). The promoter of COL1A2 had higher H3K27ac modification in GC tissues (Figures 3(b) and  3(c)). Thus, we further employed ChIP-qPCR analysis on AGS and MKN-45 cells, and we found that the use of anti-H3K27ac antibody enriched more promoter sequences of COL1A2 (p < 0:01) (Figure 3(d)). Relative COL1A2 protein level P a r e n t a l P a r e n t a l R e s i s t a n t R e s i s t a n t 5 Analytical Cellular Pathology histone-modifying enzymes that promote COL1A2 expression, we first treated AGS and MKN-45 cells with a HAT inhibitor, and we found that the level of H3K27ac modification of the COL1A2 promoter was significantly reduced in the cells after the use of HAT inhibitor (p < 0:01), and the mRNA and protein expression of COL1A2 was similarly reduced (all p < 0:01) (Figures 4(a)-4(c)). We further observed in the UCSC website that EP300 had a binding relationship with the promoter of COL1A2 (Figure 4(d)). Relative TWIST1 protein level TWIST1 GAPDH P a r e n t a l P a r e n t a l R e s i s t a n t R e s i s t a n t

Analytical Cellular Pathology
To clarify that EP300 promoted H3K27ac modification of the COL1A2 promoter and thus COL1A2 expression, we transfected EP300 overexpression plasmids into AGS or MKN-45 cells. After that, a significant increase in the level of H3K27ac modification was observed (both p < 0:01) (Figure 4(e)). Moreover, the enhanced mRNA and protein expression of COL1A2 was observed as well (all p < 0:01) (Figures 4(f) and 4(g)). Subsequently, we further found that TWIST1 colocalized with EP300 in AGS and MKN-45 cells by double-labeled immunofluorescence (Figure 4(h)). This result suggests that during the transcriptional activation of COL1A2 by TWIST1, EP300 was recruited near the promoter of COL1A2, which in turn promotes H3K27ac modification, leading to the overexpression of COL1A2.

Overexpression of COL1A2 Promotes Drug Resistance to
Apatinib in Parental GC Cells. Based on the aforementioned speculation that COL1A2 was significantly correlated with apatinib resistance in GC cells, we overexpressed COL1A2 in parental AGS and MKN-45 cells. RT-qPCR and western blot were utilized to verify the transfection efficiency (all p < 0:01) (Figures 5(a) and 5(b)). Subsequently, we examined the IC50 value of parental GC cells using the CCK-8 assays. The repressive effect of apatinib on GC cell proliferation was significantly reduced after overexpression of COL1A2 (both p < 0:01) ( Figure 5(c)). Similarly, a significant increase in the number of colonies formed by cells after treatment with 1 μM apatinib (both p < 0:01) ( Figure 5(d)) and a significant decline in the number of apoptotic cells (all p < 0:01) (Figures 5(e) and 5(f)) were observed.

Knockdown of COL1A2 Sensitizes Drug-Resistant Cells to
Apatinib. In Figure 5, we found that overexpression of COL1A2 significantly promoted the resistance of parental GC cells to apatinib. To further verify the effect of COL1A2 on apatinib resistance, we transfected shRNAs targeting COL1A2 into resistant cells and confirmed the knockdown efficiency in cells by RT-qPCR with western blot (all p < 0:01) (Figures 6(a) and 6(b)). Subsequently, we further determined the changes in the sensitivity of drug-resistant cells to apatinib after knockdown of COL1A2 by CCK-8. The sensitivity of MKN-45/R and AGS/R cells to apatinib was significantly increased in the presence of shRNA-COL1A2 (both p < 0:01) (Figure 6(c)), and the inhibiting    Analytical Cellular Pathology effects of cell growth by 1 μM apatinib were also significantly enhanced (both p < 0:01) (Figure 6(d)). In addition, the number of cell death was elevated after treatment (all p < 0:01) (Figures 6(e) and 6(f)).

3.7.
Knockdown of TWIST1 or EP300 Reduces Resistance of Parental Cells Overexpressing COL1A2 to Apatinib. To explore the effect of TWIST1 or EP300 on apatinib in GC cells, we transfected shRNAs targeting TWIST1 or EP300 into cells overexpressing COL1A2 and verified the transfection efficiency using RT-qPCR and western blot (all p < 0:01 ) (Figures 7(a) and 7(b)). We observed a significant increase in the sensitivity of MKN-45 or AGS cells to apatinib (both p < 0:01) (Figure 7(c)). Knockdown of TWIST1 or EP300 further strengthened the effects of apatinib on GC cell growth and death (all p < 0:01) (Figures 7(d) and 7(e)).

Discussion
Apatinib has been revealed to have anticancer effects through stimulation of apoptosis, suppression of cell proliferation, and induction of the effects of conventional chemotherapy agents [16][17][18]. Moreover, it is expected to have a wider application when drug resistance is overcome [5]. Drug resistance is particularly important for patients who proceed to second-line or above chemotherapy since there are no further chemodrugs available once resistance to apatinib occurs, and antiangiogenic agent resistance is suggested to be mainly associated with activation of cell adhesionrelated or growth-related signaling pathways [19]. This study showed that COL1A2 potentiated the resistance of GC cell to apatinib in vitro. The mechanism involved in the process was that COL1A2 expression was elevated by the transcription factor TWIST1, and EP300 was recruited near the promoter of COL1A2 to increase its H3K27ac modification, leading to COL1A2 overexpression and apatinib resistance in GC cells.
In our study, we first verified the significant overexpression of COL1A2 in GC using five relevant datasets and clarified its association with the somber prognosis of GC patients. To decipher the detailed mechanism for the overexpression of COL1A2 in GC, integrated bioinformatics tools were utilized, which exhibited that TWIST1 showed a significant positive correlation with COL1A2. TWIST1 is a basic helix-loop-helix transcription factor which plays pivotal parts in manifold stages of embryonic development and considerably results in tumor initiation, primary tumor growth, and even tumor metastasis [20]. Interestingly, TWIST1 transcriptionally upregulated AKT2 in breast cancer cells, leading to increased migration, invasion, and resistance to paclitaxel [21]. Kim et al. showed using ChIP that UV irradiation reduced the local recruitment of histone H3 acetylation as well as EP300 to the binding site (-1406/-1393) in the COL1A2 promoter in human dermal fibroblasts [22]. Our ChIP-seq data also revealed the distinct H3K27ac modification in the COL1A2 promoter in GC tissues, which was validated using ChIP-qPCR. Moreover, the interaction between EP300 and TWIST1 has been attested by Hamamori et al. [23]. Therefore, we believed that the overexpression of COL1A2 in GC was caused by the coordination between EP300 and TWIST1. A similar mechanism was observed in multiple myeloma cells as well where SP1/EP300 complex promoted proliferation of cancer cells through modulating the transcription of IQGAP1 [24]. Furthermore, Asano and Trojanowska reported that Fli1 recruited HDAC1/EP300 to the COL1A2 promoter and suppressed the expression of

10
Analytical Cellular Pathology the COL1A2 through histone deacetylation in human dermal fibroblasts [25]. These discordant results may be due to the influence of HDAC1, a histone deacetylase. Later, we conducted gain-and loss-of-function assays to validate their roles in GC cell resistance to apatinib. The interaction between COL1 and cisplatin resistance has been indicated by Rekowski et al. in ovarian cancer [26]. Also, Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis revealed that COL1A1 regulated carboplatin resistance of ovarian cancer cells through the "ECMreceptor interaction" and "focal adhesion" pathways [27]. In addition, pcCOL1A2 was found to reverse the effects of

12
Analytical Cellular Pathology miR-29a-3p on GC cell proliferation and apoptosis [28]. In the present study, overexpression of COL1A2 effectively enhanced the parental GC cell resistance to apatinib, as evidenced by enhanced cell growth and repressed apoptosis. As for TWIST1, TWIST1 was upregulated in various malignances and has a positive correlation with poor prognosis [29,30]. In addition, TWIST1 increased (about 2-fold) the ability of thyroid cancer cells to migrate into collagen I matrix, whereas TWIST1 knockdown in CAL62 cells reduced cell migration into collagen I matrix [31]. EP300 (also known as p300 and KAT3B) is a large gene spanning about 90 kilobases on chromosome 22q13.2; it encodes a 2414 amino acid protein with a molecular weight of about 260 kDa [32]. EP300 knockdown abolished the cancer stem cell phenotype by reducing sphere formation capacity in vitro as well as tumor formation in a xenograft mouse model in triple-negative breast cancer [33]. More relevantly, Furlan et al. reported that EP300/CREBBP inhibitors mediate the MYC/ribosomal protein axis in enzalutamideresistant cells and might have promising therapeutic implications in prostate cancer [34].

Conclusion
Collectively, TWIST1 synergizes with EP300 to enhance COL1A2 expression. COL1A2 could induce the apatinib resistance in GC. In addition, TWIST1 or EP300 depletion could sensitize GC cells to apatinib by blocking COL1A2 expression. However, due to the lack of clinical samples, future validation is needed in both human and animal tissues.

Data Availability
All the data generated or analyzed during this study are included in this published article.

Conflicts of Interest
The authors declare no competing interests.

Authors' Contributions
Gang Yu and Wanjing Chen contributed equally to this work.