Downregulation of ALDH5A1 Promotes Tumor 1 Metastasis and Contributes to Poor Prognosis in Ovarian 2 Cancer

43 Background: Despite modern therapies, ovarian cancer (OC) remains a major clinical problem with a 44 high risk of mortality. We previously reported that low expression of ALDH5A1 could serve as an 45 indicator for predicting poor prognosis in OC. However, the function of ALDH5A1 in OC progression 46 has not been elucidated yet. 47 Methods: We firstly compared ALDH5A1 expression in metastatic tissues to primary site of OC based 48 on the Oncomine database. Then wound healing assay and Transwell assay were utilized to determine 49 the biological role of OC cells transfected with ALDH5A1 siRNA. To unravel the potential mechanism 50 of ALDH5A1 meditating metastasis of OC, the co-expression profile of ALDH5A1 in OC cell lines and 51 OC patients were generated using cBioPortal. Moreover, qRT-PCR and WB analysis were used to detect 52 the expression levels of metastasis-related genes after ALDH5A1 suppression, HPA database was used 53 to confirm the relative expression of ALDH5A1 and MMP in OC patients. In addition, KM survival plots 54 in 578 OC patients from the TCGA database were analyzed. 55 Results: We proved lower ALDH5A1 expression in metastatic tissues compared to primary site of OC, 56 and knockdown of ALDH5A1 promoted the malignant behavior of OC cells. Additionally, the co- 57 expression profile of ALDH5A1 was significantly enriched in extracellular matrix (ECM) organization 58 pathway. We further confirmed ALDH5A1 was negatively associated with MMP expression in OC, 59 indicating that ALDH5A1 was closely related to OC metastasis via ECM organization pathway. Finally, 60 KM survival plots revealed that low ALDH5A1 expression contributed to poor OC survival. 61 Conclusions: These results suggested a key role of ALDH5A1 in driving the progression of OC and 62 identified ALDH5A1 as a robust therapeutic target of OC.

The siRNA and the negative control were transfected into SKOV3 cells using lipofectamine TM 3000 115 (Thermo Fisher) according to the protocols. The efficiency of silence was determined at 48 hours after 116

transfection. 117
Cell migration and invasion assays 118 SKOV3 cells either native or transfected at the concentration of 1× 10 6 cells/well were seeded in a 119 6-well plate and incubated under 5% CO 2 at 37°C. After an overnight incubation the cells grew to 100% 120 confluence. A rectangular lesion on the monolayer cells was generated using a sterile 100ul pipette tip. 121 The debris were removed and the edge of the scratch was smoothed by washing the cells once with PBS 122 and then replaced with 2.5ml of McCoy's 5A medium, after that cells were cultured. Photographic images 123 of the lesion border were acquired using an inverted microscope. 124

SKOV3 cells invasion assay was evaluated by transwell chambers with Matrigel-coated inserts (BD 125
Biosciences) according to the manufacture's protocol. In brief, Matrigel was thawed and liquefied on ice, 7 and then 30ul of Matrigel was added to a 24-well transwell insert and solidified. 1× 10 5 cells either native 127 or transfected with siRNA were plated in the insert on top of the Matrigel coating and incubated for 10 128 minutes under 5% CO 2 at 37°C to allow the cells to settle down. The lower chamber contained McCoy's 129 5A medium with 10% FBS as the chemoattractant. After incubation for 24h under 5% CO 2 at 37°C, any 130 cells had not penetrated the membrane were removed using cotton swabs. The cells had successfully 131 migrated to the bottom surfaces of the membranes were fixed with 4% polyoxymethylene and stained 132 with 0.2% crystal violet for 10 minutes. The number of cells was counted underneath an inverted 133 microscope. 134

Database search and analysis for ALDH5A1 co-expression genes 135
Data from OC cell lines (n = 47) in the Cancer Cell Line Encyclopedia (CCLE) and from patients 136 with ovarian serous cystadenocarcinoma (n = 489) in the Cancer Genome Atlas were analyzed by 137 cBioPortal(11) online platform. The genes were considered as ALDH5A1 co-expression genes when the 138 |Spearman's correlation|>0.2 and P<0.05. The common co-expression genes of ALDH5A1 in both 139 databases were conducted for pathway enrichment analysis using the g:Profiler(12) and Metascape(13) 140 online platform. Significant GO terms with similar function were visualized as interaction networks using 141 the Metascape online platform to further determine the relationship among terms, where terms with 142 P<0.01 and similarity score >0.3 were connected by edges. 143 In addition, the co-expression analysis between ALDH5A1 and key genes in the ECM organization 144 pathway were also performed using cBioPortal. 145

Quantitative RT-PCR (qRT-PCR) and Western blotting 146
Total RNAs were extracted using TRIzol reagent (Invitrogen) according to the instructions. cDNA to nitrocellulose membranes and immunoblotted with primary antibodies against ALDH5A1, GAPHD, 159 MMP-2, MMP-3 and MMP-14. Following incubation with secondary antibodies, the protein bands were 160 visualized using a chemiluminescence reagent (Thermo Fisher). 161

Prognostic implications of ALDH5A1 in OC 162
A web-based tool PROGgeneV2(14) was used to assess the prognostic implication of ALDH5A1 163 in OC. The KM survival plots were established using ALDH5A1 mRNA expression data and overall 164 survival information of the 578 OC patients from the TCGA database. ALDH5A1 was entered in the 165 database to get the KM survival plots, Hazard ratio (HR), 95% confidence intervals and P value were 166 presented on the main plots. 167

Statistical analysis 168 9
All data were analyzed using GraphPad Prism 7.0 and were presented as mean±SD of triplicates. 169 Quantitative data were analyzed using Student's t-test between two groups. For all analyses, a P<0.05 170 was considered statistically significant and were indicated with an asterisk. 171

Down-regulation of ALDH5A1 Correlated with tumor malignant features 174
To explore whether the expression of ALDH5A1 was correlated with malignancy in human OC, we 175 firstly evaluated the expression level of ALDH5A1 in primary and in metastatic tissues of OC patients. 176 Clinical data of OC patients were obtained from the Anglesio Ovarian data in Oncomine dataset. As 177 shown in Fig.1A, the transcription level of ALDH5A1 in metastatic site of OC (n = 16) was markedly 178 downregulated compared with that in primary site (n = 74) ( ** P<0.01). 179 Next, we further understood the relationship between ALDH5A1 expression and tumor malignancy 180 in vitro. ALDH5A1 gene was knocked down by siRNA to test the possible roles in tumor aggressiveness. 181 qRT-PCR analysis confirmed that ALDH5A1 expression was successfully down-regulated in SKOV3 182 cells (Fig.1B). Then we performed a scratch-wound healing assay and a transwell assay to determine the 183 effects of ALDH5A1 on OC cell migration and invasion. After down-regulation of ALDH5A1, the 184 migratory and invasion abilities of OC cells were dramatically increased ( Fig.1C-F). 185

ALDH5A1 and ECM signaling pathways in OC 187
To unravel the potential mechanism meditating the biological functions of ALDH5A1, we extracted 188 1575 co-expression genes of ALDH5A1 from 47 OC cell lines in CCLE database (supplement table 1), 189 10 and 1220 co-expression genes of ALDH5A1 from 489 OC patients in TCGA database (supplement table  190 2) using the cBioPortal online platform. In total, 128 common co-expression genes were found to be 191 overlaped through taking the intersection of these two co-expression gene sets ( Fig.2A and supplement  192 table 3). 193 To explore the aim of identifying possible signaling pathways from the list of co-expression genes 194 of ALDH5A1 in OC, we performed functional enrichment analysis with these 128 common co-195 expression genes of ALDH5A1 obtained from CCLE and TCGA database. (REAC: R-HSA-1474244). We then performed ontology analysis again using the Metascape platform to 203 confirm these results. Fig.2C showed the top 20 putative biological processes, and the most significantly 204 enriched gene set was ECM organization pathway (GO: 0030198). This analysis also revealed 205 ALDH5A1 and correlated genes were largely related to tissue morphogenesis (GO: 0048729) and 206 skeletal system development (GO: 0001501). Fig.2D showed all these biological processes identified 207 with a significant P value were closely inter-related. All the above results indicated that ALDH5A1 and 208 correlated genes were mainly related to the ECM and influence the development of cancer. 209

ALDH5A1 was negatively associated with MMP expression in OC 210 11
Nextly, we checked the key genes participated in the ECM organization pathway which may be 211 correlated with ALDH5A1 using g:Profiler. Among these genes, a negative correlation was found 212 between ALDH5A1 and matrix metalloproteinases (MMPs). The negative correlation between 213 ALDH5A1 and MMP2 expression (R = 0.33), between ALDH5A1 and MMP3 expression (R = 0.25), 214 between ALDH5A1 and MMP14 expression (R = 0.30) were confirmed using Spearman and Pearson 215 correlation analyses (Fig.3A). 216 We confirmed that both MMP mRNA and protein expression levels were upregulated by siRNA-217 mediated ALDH5A1 knockdown in OC cells. When the expression levels of ALDH5A1 were decreased, 218 MMP2, MMP3, MMP14 mRNA expression was significantly increased in OC cells (Fig.3B). In parallel, 219 western blot results showed that the MMP2, MMP3, MMP14 protein expression were increased in 220 ALDH5A downexpressing OC cells compared with control cells (Fig.3C). 221 In addition, we analyzed an OC TMA cohort obtained from the HPA database. 222 Immunohistochemical (IHC) staining results also demonstrated the similar expression patterns of 223 ALDH5A1 and MMP in OC patients (Fig.4). 224

The poor prognosis of the patients with lower expression of ALDH5A1 in OC 225
Finally, to investigate whether ALDH5A1 is associated with OC patient prognosis, a Kaplan-Meier 226 analysis based on the TCGA ovarian adenocarcinoma data was organized by the web-based tool 227 PROGgeneV2. The results showed that the patients with lower expression of ALDH5A1 presented 228 poorer prognosis than those with higher expression in OC. The overall survival (OS) rates of OC patients 229 with ALDH5A1 high were obviously higher than those of patients with ALDH5A1 low [HR = 0.75 (0.64-230 0.88), P = 0.0005, Fig.5A]. We also explored the correlation between ALDH5A1 mRNA expression and 12 pathological grades of OC. As the sample size of patients with pathological grade I was too small (n = 232 6), we did not analyze the survival curves in this group. As shown in Fig.5B  The ECM is a dynamic structure influences tumour progression(28), which is commonly 268 deregulated and becomes disorganized in cancer. Deregulated ECM dynamics disrupt tissue polarity, 269 architecture, and integrity and promote epithelial-mesenchymal transition (EMT) and metastasis(29). 270 The MMPs are a family of zinc-dependent enzymes, and MMP-mediated ECM degradation leads to 271 disrupt the physiological barrier and cancer cell metastasis has been a guiding principle in MMP 272 research(30). Using g:Profiler, we extracted MMP2, MMP3, MMP14 from the key genes participated in 273 14 the ECM organization pathway, verified a negative correlation between ALDH5A1 and MMP expression. 274 In terms of MMP2 and MMP3, both of them were found to function as early response proteins in OC 275 metastasis(31-33). In recent years, several studies revealed that MMP14 plays a central role in 276 pericellular matrix degradation during basement membrane and interstitial tissue transmigration 277 programs(34), which stimulates a tumor-stromal signaling pathway and promotes angiogenesis and 278 tumor growth on OC cells(35). All the growing evidence verified our hypothesis that ALDH5A1 was 279 somewhat relative to the metastasis of OC. 280 Moreover, the OS rates of OC patients with ALDH5A1 high were obviously higher than those of 281 patients with ALDH5A1 low , and the high expression of ALDH5A1 in pathological grade II and III 282 patients were associated with improved OS. This indicated low expression of ALDH5A1 was a 283 significant predictor of worse clinical prognosis in OC patient. 284 285

Conclusion 286
Overall, the present study revealed that ALDH5A1 may play an important role in metastasis of OC, 287 and ALDH5A1 may be a therapeutic target of OC that is potentially effective in treating OC metastasis 288 according to the bioinformatic analyses and verification experiments. Therefore, although much remains 289 to be learned and further studies are needed to fully understand the reciprocal interactions that are 290 essential for OC metastasis, and the precise role of interaction between ALDH5A1 and metastasis of OC 291 needs to be further investigated, our findings confirmed that ALDH5A1 might be a promising molecular 292 target for OC therapeutic intervention. 293 294

Declarations 295
Ethical approval and consent to participate: Not applicable. 296 Consent for publication: Not applicable. 297 Availability of data and materials: All data generated or analyzed during this study are included in this 298 article and its additional files. 299 Competing interests: The authors declare that they have no competing interests.  Authors' contributions: XT and QHZ contributed to the conception and design of the study. CC, XW and 303 RL performed the study and drafted the article. PJ, HWC and MX conducted data acquisition, data 304 analysis and interpretation. All authors discussed the results and agreed to be accountable for all aspects 305 of the work. All authors read and approved the final manuscript. 306