ETV5 expression positively correlates with promoter methylation and predicts response for 5-FU-based adjuvant therapy response in proximal colon cancer

Discovery of markers predictive for 5-Fluorouracil (5-FU)-based adjuvant chemotherapy (adjCTX) response in patients with locally advanced stage II and III colorectal cancer (CRC) is necessary for early identification of potential responders as only 20-65% of CRC patients benefit from the treatment. PEA3 subfamily of ETS transcription factors (ETV1, ETV4, and ETV5) are upregulated in multiple cancers including colon cancers. However, the underlying epigenetic mechanism regulating their overexpression and their role in predicting therapy response in colon cancer is largely unexplored. In this study, using gene expression and methylation data from The Cancer Genome Atlas (TCGA) project, we showed that promoter DNA methylation negatively correlates with ETV4 expression (ρ= -0.17, p=5.6×10−3) and positively correlates with ETV5 expression (ρ= 0.22, p=1.43×10−4) in colon cancer tissue. Further, our analysis in 662 colon cancer patients treated with 5-FU-based-adjCTX revealed that higher ETV5 expression associated with shorter relapse-free survival (RFS) of treated patients with proximal tumors (Hazard ratio = 3.30 - 6.22, p=0.005-0.02). We also observed higher expression of signaling molecules involved in cellular proliferation (e.g. GNB5, DUSP4, FYN) in patients with high ETV5 level, suggesting that the increased cellular proliferation due to overexpression of these genes could drive the therapy resistance. The present study suggests ETV5 expression as a strong predictive biomarker for 5-FU-based adjCTX response in stage II/III CRC patients with proximal tumors.


Introduction 67
Colon rectal adenocarcinoma (CRC) is the fourth most commonly diagnosed cancer (2 million cases 68 in 2018) in the world and kills nearly 1 million people annually1 mostly due to the spread of tumor 69 cells to other secondary organs (e.g. liver) in the later stage (stage III and IV) of the disease2. 70 Therefore, successful treatment of the locally advanced early-stage (stage I, II, and III) cancer is 71 necessary in order to prevent disease progression and improve the overall survival of the patients3. 72 Usually, the early-stage patients are cured by surgical removal of the tumor only without the use of 73 chemotherapy, however, systematic use of 5-Fluorouracil (5-FU)-based-adjCTX is recommended for 74 stage II and III cases with high risk (e.g. with perineural invasion and poor histological 75 differentiation) of reoccurrence3,4. However, the use of adjCTX after surgery cures only 20% of 76 additional stage III patients over surgery alone (cures 50% of cases) and improves the chance of 10-77 year overall survival only by 10%-20% in stage II patients4. Further, it incurs considerable toxicity 78 (e.g. myelosuppression, diarrhea) and economic costs to the patients5, 6. The higher toxicity and low 79 efficacy of 5-FU-based adjCTX demand novel and reliable molecular markers that can predict the 80 treatment response in early-stage (II and III) patients and help to stratify patients with different 81 response7,8. 82 Attempts to predict response for adjuvant chemotherapy have identified molecular alterations (e.g. 83 microsatellites status9, TP53 mutations10, genetic polymorphism in MTHFR11 and DPYD12,13) as a 84 predictive marker for 5-FU-based adjCTX response. Further, recent studies exploring gene 85 expression signatures as predictive markers for treatment response in colon cancer have identified 86 ESR114, and CD815 expression as predictors for 5-FU-based adjCTX response in CRC. However, 87 none of the identified markers can successfully segregate the responders from nonresponders before 88 initiating the therapy indicating the need to search for additional novel markers predictive for adjCTX 89 implemented in the R package affy. Gene expression was summarized as the average expression 134 levels of all the probes of the genes and was used for differential and survival analysis.  Statistical analysis: All the statistical analysis has been performed using R version 3.5.3. The Non-161 parametric Wilcoxon rank test has been used to compare the gene expression between two groups. 162 The survival analysis has been performed using Cox-proportional regression as implemented in the cancerous tissue in TCGA data and observed higher expression of only ETV4 (Wilcox test 171 P=4.90x10-25) and ETV5 (Wilcox test P=5.17x10-9) in tumor suggesting their possible role in tumor 172 biology ( Figure 1A, and B). Further, we checked the role of promoter DNA methylation over 173 expression of ETV4 and ETV5 in the cancer tissue as promoter hypermethylation can shutdown gene 174 expression. Our analysis revealed that promoter DNA methylation nominally differs between cancer 175 and normal tissue for both ETV4 and ETV5 ( Figure 1C

ETV5 expression differs in CRC patients with proximal and distal tumors 191
After observing that ETV4 and ETV5 are overexpressed in colon cancerous tissue from TCGA 192 cohort, we studied their role in predicting adjCTX response in CRC using two publicly available 193 datasets. Since the location of the tumor can affect the response to chemotherapy29, first we checked 194 the expression level of ETV4 and ETV5 in stage II/III CRC patients (both treated and untreated) with 195 proximal and distal tumors. We observed a significant difference in ETV4 expression between distal 196 and proximal tumors only in discovery cohort samples (Figure 2A  In order to assess the effect of ETV4 and ETV5 expression over chemotherapy response, we used 210 Cox-regression with RFS as the treatment outcome and observed that higher ETV5 expression is a 211 strong and selective predictor for poor RFS (HR= 2.29, P=0.00178) in treated patients in the discovery 212 cohort ( Figure 3A). Further, stratified analysis revealed that ETV5 expression significantly predicts   3rd quartile has been categorized as high (red) and with expression <1st quartile has been categorized 232 as low (low). The log-rank analysis was used to test for significance between the survival curves. 233 234

Higher ETV5 level associated with genes involved in cell differentiation 235
After observing that higher ETV5 expression correlates with shorter RFS in 5-FU-based adjCTX 236 treated patients with proximal tumors. We explored the possible mechanism by which ETV5 may 237 induce poor response for therapy. A differential analysis between adjCTX treated patients with higher 238 (the 4th quartile) and lower (the 1st quartile) ETV5 expression revealed a strong difference (P < 10-239 5) in the expression of 22 genes enriched in the cell differentiation process (Suppl.   of their expression as a predictive marker for chemotherapy response in stage II and III colorectal 259 cancer patients using 2 publicly available independent colon cancer datasets. 260 Our analysis revealed higher expression ETV4 and ETV5 genes in colon cancer tissue compared to 261 normal tissue in TCGA samples ( Figure 1A and B). Further, ETV4 expression showed a negative 262 correlation with promoter methylation ( Figure 1E MEK-ERK and PI3K-AKT pathways leading to cellular proliferation and viability30. GNB5 has 284 emerged as the major target to overcome cetuximab resistance in colorectal cancer30. Furthermore, 285 we observed an expression difference of DUSP4 in patients with a high level of ETV5. DUSP4 is a 286 MAPK phosphatase and its role has been contradictory as downregulation of DUSP4 enhances cell 287 proliferation and invasiveness in colorectal carcinomas41 but inhibits growth in human colorectal cancer cell lines42. These results suggest that higher ETV5 expression may induce drug resistance by 289 upregulation of genes involved in colon cancer proliferation and growth. However, studies in larger 290 human cohort and animal model systems can fully explain the detailed mechanism for the ETV5 role 291 in 5-FU-based adjCTX resistance. 292 The current study identified ETV5 as a biomarker of 5-FU-based adjCTX response in CRC patients 293 with evidence II level as defined by Simon et al43, and revealed that higher ETV5 is associated with 294 poor response in proximal CRC tumors. These results suggest that ETV5 could be useful for the 295 identification of responders before administration 5-FU-based adjCTX when included along with 296 other already established clinicopathological markers. 297

Author contribution 300
AKG conceptualized the study, analyzed the data, and wrote the manuscript. 301

Acknowledgment 302
The author is very much thankful to Prof. Tero Aittokallio and Aleksandr Ianevski for their valuable 303 suggestions during the analysis and editing of the manuscript. 304 Suppl.