This article is mentioned in:

Introduction

Long noncoding RNAs (lncRNAs) have been recognized as important functional agents of the genome, containing intricate structural and informational capacity (1–3). Although these transcripts have little or no protein-coding capability, lncRNAs regulate gene expression through multiple mechanisms including chromatin modification and both transcriptional and post-transcriptional regulation (1–3). Chromosome 8q24, a large gene desert known for its paucity of coding regions contains the lncRNA Colon Cancer Associated Transcript 2 (CCAT2), located in close proximity to the MYC oncogene (2,4–6). LncRNA CCAT2 was initially recognized to have oncogenic effects in colorectal cancer, and has since been linked to various cancers, including ovarian cancer (2,4–12). CCAT2 harbors the rs6983267: NR_109834.1:n.662G>T SNP and has been shown to interact near the MYC promoter and regulate gene expression, mediating tumor metastasis and growth in colon, lung, breast, prostate, endometrial and gastric cancers (2,4–9).

Ovarian cancer continues to be one of the leading causes of death from gynecological cancers accounting for over 200,000 deaths per year worldwide (13). Thus, there have been multiple efforts to discover novel biomarkers and therapeutic targets, including lncRNAs, for this cancer. The rs6983267 SNP within CCAT2 has such potential due to its allele-specific features which are hypothesized to affect MYC expression (2,4–6,8–10,14,15). Ghoussaini et al were the first to associate the 8q24.21.a locus with increased risk of ovarian cancer (12). Further investigation by Huang et al found that CCAT2 expression was significantly higher in ovarian cancer tissue with high CCAT2 gene expression correlating to poor prognostic parameters and shorter overall and disease-free survival (7). Other studies revealed similar results, showing that CCAT2 expression was upregulated in ovarian cancer cells, and knock down of CCAT2 expression in vitro significantly repressed proliferation and promoted apoptosis in certain cell lines (11). Specifically, the G allele of rs6983267 was shown to significantly increase a women's risk of ovarian cancer (10).

Although there has been a growing body of evidence linking ovarian cancer risk with CCAT2 expression, there is little information about the association of the rs6983267 risk allele and clinical outcomes in ovarian cancer patients. The aim of this study is to assess the association between genotypes at the rs6983267 locus (within lncRNA CCAT2) and clinical outcomes in patients with high-grade ovarian cancer (HGSOC), including survival and response to chemotherapy.

Materials and methods

A retrospective genetic association study was conducted using genomic DNA from ninety-eight patients with HGSOC at the University of Iowa Hospitals and Clinics. The genomic DNA originated from flash frozen tumor tissues stored in the Department of Obstetrics and Gynecology Gynecologic Oncology Bank (IRB, ID#200209010 and ID#201804817) which is part of the Women's Health Tissue Repository (IRB, ID#201809807). All tissues archived in the Women's Health Tissue Repository were originally obtained from adult patients under written informed consent in accordance with the University of Iowa IRB guidelines. Genomic DNAs were purified from frozen tumor tissues using the DNeasy Blood and Tissue Kit according to the manufacturer's (Qiagen) recommendations.

Clinical data

Clinical and pathological data were collected from the electronic medical record. Clinical variables previously observed to be associated with chemo-response were included in the data collection (16). Only baseline clinical and pathological characteristics which can be obtained before starting initial chemotherapy were included.

Genotyping

The genomic region around rs6983267 was amplified via standard PCR on a BioRad T-100 thermal cycler using primers detailed in Fig. S1. The reagents used in each PCR sample included: 3 µl 10X reaction buffer with 1.5 mM MgCl2, 1 µl 10 mM dNTPs, 1 µl (10 pmole) of each primer, and 0.5 µl Taq polymerase (manufactured by New England BioLabs) with a final reaction volume of 30 µl. The thermal cycler was programmed for five minutes at 95°C for initial denaturation, followed by 35 cycles of 30 sec at 95°C for denaturation, 30 sec at 55°C for annealing, 30 sec at 72°C for extension, and seven minutes at 72°C for the final extension. The PCR amplicon was purified using the QIAGEN QIAquick PCR amplification kit. The purified amplicons were sequenced using conventional Sanger sequencing carried out on an Applied Biosystems Model 3730×l capillary sequencer in the Genome Facility at the University of Iowa Institute of Human Genomics. The results provided the genotype (GG, GT, TT) of each patient for the rs6983267 SNP.

Association with rs6983267 genotypes

Univariate logistic regression was used to explore the association between the clinical outcomes and biological variables and the rs6983267 genotypes (genotypes containing any G allele and the TT genotype). This was performed to assess advantageous characteristics in HGSOC survival. Clinical outcomes analyzed included: Age, body mass index (BMI), Charlson Co-morbidity Index, pre-operative CA-125, cancer stage, disease in the upper abdomen by imaging (other than the omentum), disease in chest by imaging, tumor grade, residual disease after surgery, removal of pelvic lymph nodes, removal of para-aortic lymph nodes, neoadjuvant chemotherapy, number of chemotherapy cycles delivered, dose dense chemotherapy, and death by disease. Biological variables included: CCAT2, MYC, MYCL, MYCN, MYCBP (MYC Binding Protein), MYCBP2 (MYC Binding Protein 2), MYCBP2-AS1 (MYC Binding Protein 2 Antisense RNA 1), MYCBPAP (MYC Binding Protein Associated Protein), MYCNUT (MYCN Upstream Transcript), MYCNOS (MYCN Opposite Strand), and MYCT1 (MYC Target 1). Variables in the univariable analysis with a p-value <0.10 were introduced in the multivariate logistic regression model. This P-value was used to create a more inclusive multivariate model (17). The multivariable logistic regression model was used to assess independent association of clinical and biological variables with the rs6983267 genotypes (genotypes containing any G allele and the TT genotype). Variables with a P-value <0.05 in the multivariate analysis were considered significant.

Survival analysis

Survival analysis for patients with the different genotypes at the rs6983267 SNP (GG, GT, TT) and MYC and CCAT2 expressions was performed using the Cox proportional hazard model. Survival analysis was performed using a log-rank test for a model with three genotypes (GG, GT and TT), and for a model comparing the homozygous TT genotype with genotypes containing the most frequent allele, G (GG or GT). Survival assessment of clinical variables (age, BMI, tumor grade, FIGO stage, pre-operative CA-125, neoadjuvant chemotherapy, residual disease after surgery, and optimal surgery) also were performed using the Cox proportional hazard ratio (HR). Clinical and biological variables associated in the univariate analysis with a P≤0.10, were introduced in a Cox Proportional hazard ratio multivariable model (17). Proportional hazards assumptions were assessed for the final survival model. Variables with a P-value <0.05 in the multivariate Cox model were considered significant.

Gene expression of the biological variables listed above was determined from previous RNA sequencing experiments using the same patients (GEO accession number GSE156699) (18,19).

Power calculation

With 98 samples, and SNP (rs6983267) frequencies of 69% for GG/GT and 31% for TT, our study had a power of 79% to find differences in survival of >30% at 5 years when comparing between genotypes, with an α error of 0.05. Statistical analysis, power calculations and graphics were performed with R statistical package and computer environment (20). R packages survival, stats, and survcomp were used for the statistical analyses (21).

Statistical analysis. A univariate logistic regression was used to assess the association between the clinical and biological variables and the rs6983267 genotypes. Variables from this analysis (P≤0.1) were introduced into a multivariate logistic regression to assess independent association of clinical and biological variables with the rs6983267 genotypes. A log-rank test was used to analyze survival for a model assessing the rs6983267 genotypes (GG, GT, and TT) and for a model assessing the homozygous TT genotype and genotypes containing the most frequent allele, G (GG and GT). Survival assessment of clinical variables was performed using the Cox Proportional hazard ratio. Significant variables from this analysis were introduced into a multivariable Cox Proportional hazard ratio model.

Results

Association with rs6983267 genotype

To assess which characteristics were associated with the rs6983267 genotype that showed advantages in HGSOC survival (TT), we performed univariate logistic regression analyses (Table I). Significant clinical and biological variables associated with the TT genotype were introduced into the multivariate logistic regression model and included: Age (OR=0.95, 95% CI 0.92–0.99, P=0.010), pre-operative CA-125 (OR=1.00, 95% CI 1.001–1.002, P=0.042), cancer stage (OR=2.49, 95% CI 1.08–6.15, P=0.040), and response to chemotherapy (OR=2.91, 95% CI 1.03–9.17, P=0.052). In the multivariate logistic regression model (Fig. 1; Table II) TT genotype was independently associated with FIGO stage and response to chemotherapy. HGSOC patients with a TT genotype at the rs6983267 locus were over five times more likely to have a higher FIGO stage (OR=5.34, 95% CI 1.50–22.62, P=0.014) and were four times more likely to respond to chemotherapy (OR=4.51, 95% CI 1.40–18.00, P=0.018) compared to individuals with a GG or GT genotype.

Figure 1. Multivariate logistic regression of association with rs6983267 TT genotype. Multivariate logistic regression demonstrated that patients with the TT genotype were associated with more advanced disease (FIGO stage) (OR 5.34, 95% CI, 1.50–22.62; P=0.014) and improved response to chemotherapy (OR 4.51; 95% CI, 1.40–18.00; P=0.018) when compared with patients with genotypes containing any G allele (GG, GT). OR, odds ratio; T, tyrosine; G, guanine; CI, confidence interval; FIGO, International Federation of Gynecology and Obstetrics.

Table I. Table of patient characteristics analyzed using univariate logistic regression to demonstrate association with the TT genotype.

Table I.

Table of patient characteristics analyzed using univariate logistic regression to demonstrate association with the TT genotype.

Variable	G-allele (GG, TG) N=68	TT genotype, N=30	OR (95% CI)	P-value
Mean age, years	63	55	0.95 (0.92–0.99)	0.010a
Mean body mass index, kg/m2	27.1	25.7	0.97 (0.89–1.04)	0.427
Charlson comorbidity index
Low	6	5
Medium	40	15	0.68 (0.20–2.29)	0.536
High	6	2	0.60 (0.07–3.90)	0.605
Mean pre-operative CA-125	1674.14	4719.79	1.00 (1.001–1.002)	0.042a
FIGO stage
Stage I–II	4	0	2.49 (1.08–6.15)	0.040a
Stage III	50	18
Stage IV	14	11
Disease in upper abdomen by imaging (other than omentum)
Large bowel	2	1	1.00 (0.42–2.50)	0.992
Spleen	0	0
Portahepatis	2	2
Mesenteric	3	0
Disease in chest by imaging	0	6	1.2×108 (8.3×10−6-NA)	0.991
Grade
1	0	0	1.22 (0.44–3.79)	0.712
2	16	6
3	48	22
Residual disease after surgery
Microscopic	13	3	0.47 (0.01-1,61)	0.269
Macroscopic	55	27
Optimal	46	16	0.55 (0.23–1.32)	0.178
Suboptimal	22	14
Removal of pelvic lymph nodes	13	2	0.30 (0.05–1.19)	0.132
Removal of para-aortic lymph nodes	8	1	0.26 (0.14–1.51)	0.212
Neoadjuvant chemotherapy	9	5	1.38 (0.39, 4.44)	0.600
Response to chemotherapy
Responders	25	19	2.91 (1.03–9.17)	0.052a
Non-responders	23	6
Number of cycles delivered
<6 cycles	12	2	1.05 (0.80–1.37)	0.686
>6 cycles	54	26
Dose dense chemotherapy	1	2	4.79	0.209
Death by disease	55	24	1.45 (0.44–105.5)	0.594
CCAT2b	1.96	2.15	1.10 (0.81–1.49)	0.552
MYCBPb	3.34	3.32	1.1×1014 (NA-4.2×10174)	0.991
MYCLb	8.83	9.02	1.08 (0.82–1.49)	0.575
MYCBP2b	12.12	12.03	0.79 (0.39–1.58)	0.509
MYCBP2-AS1b	3.40	3.44	2.08 (0.38–11.12)	0.376
MYCBPAPb	4.90	4.68	0.87 (0.61–1.23)	0.441
MYCNUTb	3.32	3.34	2.96×108 (NA-2.3×10−179)	0.991
MYCNOSb	3.51	3.47	0.84 (0.26–2.17)	0.739
MYCNb	6.91	7.58	1.22 (0.96–1.56)	0.109
MYCT1b	5.61	5.38	0.88 (0.62–1.22)	0.451
MYCb	10.63	11.01	1.27 (0.90–1.84)	0.189

a P<0.10 so introduced into multivariate analysis.

b Data are represented as mean gene expression. OR, odds ratio; CI, confidence interval; CCAT2, colon cancer-associated transcript 2; MYCBP, MYC-binding protein; MYCBP2, MYC-binding protein 2; MYCBP2-AS1, MYC-binding protein 2 antisense RNA 1; MYCBPAP, MYC-binding protein-associated protein; MYCNUT, MYCN upstream transcript; MYCNOS, MYCN opposite strand; MYCT1, MYC target 1.

Table II. Multivariate logistic regression of patients with the TT genotype compared with patients with genotypes containing any G allele.

Table II.

Multivariate logistic regression of patients with the TT genotype compared with patients with genotypes containing any G allele.

Variable	Odds ratio	95% confidence interval	P-value
Age, years	0.99	0.95–1.05	0.853
Pre-operative CA-125	1.00	0.999–1.0002	0.145
FIGO stage	5.34	1.50–22.62	0.014a
Response to chemotherapy	4.51	1.40–18.00	0.018a

a Statistically significant, P<0.05. CA-125, cancer antigen 125; FIGO, International Federation of Gynecology and Obstetrics.

Survival analysis

Analysis comparing survival among the three genotypes (GG, GT, and TT) revealed no difference in survival (P=0.20). When comparing survival curves between patients with G allele presence (GG or GT) and patients with the TT genotype, there was stronger evidence for differences in survival (P=0.10), even after applying several weighted Kaplan-Meier tests due to late-crossing survival curves (Fig. 2; Table SI). The results of univariate survival analysis for clinical and biological variables are summarized in Table III. The majority of patient deaths were attributed to HGSOC (84%, 46 out of 55).

Figure 2. Kaplan-Meier curves comparing survival rates. (A) Survival curves among genotypes GG, TG and TT did not show a significant difference in survival (P=0.20), with 20.6, 20.2 and 35.2 months of median survival, respectively. (B) Stronger evidence for differences in survival was observed when comparing survival curves between patients with any G-allele (GG, TG) and patients with the TT genotype (P=0.10), with 20.6 and 35.2 months of median survival, respectively. T, tyrosine; G, guanine.

Table III. Univariate survival analysis of the effect of variables on survival time in patients with high-grade serous ovarian cancer.

Table III.

Univariate survival analysis of the effect of variables on survival time in patients with high-grade serous ovarian cancer.

Variable (Ref.)	HR	95% confidence interval	P-value
rs6983267 (Ref: G presence)	0.665	0.40–1.10	0.100a
Age, years	1.029	1.02–1.04	<0.001a
Body mass index, kg/m2	0.997	0.98–1.02	0.698
Grade	0.821	0.61–1.10	0.191
FIGO stage	1.366	1.06–1.76	0.015a
Pre-operative CA-125	1.000	1.00–1.00	0.680
Neoadjuvant chemotherapy (ref: no)	2.249	1.50–3.38	<0.001a
Residual disease (ref: micro)	2.056	1.48–2.85	<0.001a
Optimal surgery (ref: yes)	1.586	1.26–2.00	<0.001a
Colon cancer-associated transcript 2	1.000	0.84–1.19	0.997
MYC	1.091	0.89–1.33	0.391

a P≤0.10 so introduced into multivariate analysis. HR, hazard ratio; ref, reference; FIGO, International Federation of Gynecology and Obstetrics.

Based on the univariate survival analysis, the following variables were introduced into the multivariate Cox proportional hazard ratio model: Genotype TT at the rs6983267 locus (reference: G allele) (HR=0.665, 95% CI 0.40–1.10, P=0.100), age (HR=1.029, 95% CI 1.02–1.04, P<0.001), FIGO stage (HR=1.366, 95% CI 1.06–1.76, P=0.015), neoadjuvant chemotherapy (reference: No neoadjuvant chemotherapy) (HR=2.249, 95% CI 1.50–3.38, P<0.001), residual disease (reference: Microscopic disease) (HR=2.056, 95% CI 1.48–2.85, P<0.001), and optimal surgery (reference: Yes) (HR=1.586, 95% CI 1.26–2.00, P<0.001). The multivariate survival analysis (Fig. 3; Table IV) demonstrated that patients with the TT genotype had improved survival time compared to patients with genotypes containing any G allele (HR=0.59, 95% CI 0.36–0.97, P=0.039;), even after accounting for other significant co-variates, like neoadjuvant chemotherapy (HR=3.06, 95% CI 1.64–5.69, P=0.039) and residual disease after surgery (HR=2.21, 95% CI 1.13–4.33, P=0.021).

Figure 3. Multivariate survival analysis-Cox proportional hazard ratio. Multivariate Cox proportional hazard analysis demonstrated that patients with the TT genotype were associated with improved survival time (HR 0.59; 95% CI, 0.36–0.97; P=0.039). Neoadjuvant chemotherapy (HR 3.06; 95% CI, 1.64–5.69; P<0.001) and residual disease (HR 2.21; 95% CI, 1.13–4.33; P=0.021) were associated with decreased survival time. HR, hazard ratio; T, tyrosine; CI, confidence interval.

Table IV. Multivariable Cox proportional hazard ratio analysis of the significant variables.

Table IV.

Multivariable Cox proportional hazard ratio analysis of the significant variables.

Variable (Ref.)	Hazard ratio	95% confidence interval	P-value
rs6983267 (ref: G presence)	0.59	0.36, 0.97	0.039a
Age, years	1.02	0.999, 1.04	0.057
FIGO stage	1.09	0.66, 1.80	0.747
Neoadjuvant chemotherapy (ref: no)	3.06	1.64, 5.69	<0.001a
Residual disease (ref: micro)	2.21	1.13, 4.33	0.021a
Optimal surgery (ref: yes)	0.73	0.44, 1.20	0.215

a Statistically significant, P<0.05. Ref, reference; FIGO, International Federation of Gynecology and Obstetrics.

Discussion

In this study we showed that HGSOC patients with the TT genotype at the rs6983267 locus had improved survival time when compared to HGSOC patients with any G allele (GG and GT genotypes). This is despite patients with the TT genotype having a higher FIGO stage. A possible explanation for this outcome is patients with the TT genotype responded better to initial standard chemotherapy. The mechanism by which this occurs is unclear; however, an example of this phenomenon can be seen in BRCA and homologous repair deficient (HRD) ovarian cancer patients. In epithelial ovarian cancer, the strongest known genetic risk factor is BRCA1 and BRCA2 germline mutations, which account for ~10% of cases (22,23). Yet BRCA-associated epithelial ovarian cancer patients have been shown to have greater 5-year survival when compared to sporadic mutations (22,23). This has been attributed to improved response to chemotherapy due to a process coined ‘synthetic lethality’ in which simultaneous impairment of two DNA repair pathways leads to cytotoxicity and cell death (23–25). PARP inhibitors, which block the repair of DNA single-strand breaks, have been shown to be 100–1,000 times more effective in cells deficient in BRCA1 or BRCA2 (25). This same response has been demonstrated in HGSOC patients who exhibit aberrations in other homologous recombination repair genes (24–26). Further studies to assess the mechanisms and pathophysiology specific to the role of lncRNA CCAT2 in ovarian cancer may be able to characterize other treatments or maintenance therapies that have greater efficacy based on the rs6983267 genotype.

Ghoussaini et al associated two other SNPs within 8q24 with ovarian cancer, rs10505477 and rs10808556, that may also warrant further investigation (12). SNP rs10505477 is located within long non-coding RNA Cancer Susceptibility Candidate 8 (CASC8), which is in the chromosome 8q24 locus (27). Several studies have looked at clinical outcomes in patients with gastric and lung cancer in relation to SNP rs10505477. In patients with gastric cancer undergoing cisplatin chemotherapy, GA and AA genotypes of rs10505477 were correlated with poorer overall survival, compared with the GG genotype (28). Similarly, the rs10505477 and rs6983267 polymorphisms have been shown to respond to platinum-based chemotherapy in lung cancer (29,30). In addition, several studies have suggested a strong linkage disequilibrium between the SNP rs10505477 and the SNP rs6983267 (27,31). Wu et al analyzed the association between CCAT2 and CASC8 polymorphisms, suggesting effects from both variants play a role in hepatocellular carcinoma risk (32). It may be the case that several polymorphisms contribute to the development of ovarian cancer, treatment response, and overall survival. The collective effects of several polymorphisms within 8q24 have yet to be addressed in ovarian cancer and may provide interesting insight.

There are limited studies regarding SNP rs10808556. Tong et al conducted a meta-analysis evaluating twenty-eight variants in 8q24 and their association with cancer risk. They found that rs10808556 was significantly associated with colorectal cancer risk (33). Another study assessed the relationship between rs10808556 and thyroid carcinoma risk; however, findings did not suggest an association (34). Similar to rs6983267, numerous studies involving SNPs within 8q24 focus on cancer susceptibility. Our study highlights the need to not only look at the role that SNPs play in the onset of disease but also the response to therapeutic interventions and therefore prognosis.

Studies involving 8q24 often revolve around MYC activity in relation to cancer risk-associated SNPs. While these SNPs have been associated with increased cancer risk, there have been uncertainties regarding MYC's role behind this observation. Goode et al analyzed common variants at 8q24 in relation to ovarian cancer and noted significant SNPs for ovarian cancer were located in a gene desert relatively far from the 3′ end of MYC (35). SNP rs6983267, for instance, is located 335 kb from MYC, its closest gene (5,9,36–39). It was suggested that MYC may not be the target gene for ovarian cancer or that these polymorphisms were capable of influencing MYC from a distance (35). However, in colorectal and prostate cancer tissues, evidence revealing long-range physical interaction between rs6983267 and MYC was discovered (36,40). Furthermore, risk loci within 8q24 appeared to act in a tissue-dependent manner such that the risk loci associated with prostate cancer, for instance, interacted with MYC in prostate cancer cells but not breast or colon cancer cells (39). It was concluded that rs6983267, along with other risk loci within 8q24, likely acts as a tissue-specific cis-regulatory enhancer element, leading to increased expression of MYC (36,39). It has been proposed that the mechanism behind this is similar to what is seen with the Colon Cancer Associated Transcript 1 locus (CCAT1) located 515 kb upstream from MYC within a colorectal cancer super-enhancer. CCAT1 encodes two lncRNAs: Colon Cancer Associated Transcript 1 long isoform (CCAT1-L) and Colon Cancer Associated Transcript 1 short isoform (CCAT1-S) (6). LncRNAs CCAT1-L and CCAT1-S facilitate the formation of chromatin looping, which allow for MYC interaction with its enhancers (6,37). In fact, interaction between various cancer risk variants within 8q24 and the MYC oncogene via chromatin looping has been observed in multiple studies (5,37,39). In addition, chromatin looping allows the lncRNAs to accumulate around the MYC locus and carry out their role in MYC regulation (6).

MYC expression is regulated through the binding of Wnt proteins to their receptors on the cell surface (39). This results in a signaling pathway that stabilizes β-catenin and allows it to enter the nucleus to bind to the TCF4 transcription factor (5,15,39). Wright et el assessed the rs6983267 risk allele's effect on chromatin loop formation in order to better elucidate whether increased MYC expression was a result of alterations in loop formation versus interactions with the TCF4 transcription factor. It was shown that the loop does not alter in frequency of formation or interactions in response to the rs6983267 SNP, and that the loop exists regardless of which genotype is present. This suggested that increased MYC activity is a result of increased TCF4 recruitment and not by altered loop formation (37). Additionally, affinity for the TCF4 transcription factor was found to be higher for the G allele of rs6983267 than the T allele (9,38).

Several studies have aimed to elucidate the pathophysiology of the effect of lncRNA CCAT2 on MYC expression. It has been proposed that lncRNA CCAT2, transcribed from the MYC-335 enhancer region involving the rs6983267 site, associates with TCF4 to augment its transcription activity, though the specific mechanism is unknown. Binding of CCAT2 may alter protein structure or modify the association of TCF4 and its partners within the transcription complex in an allosteric manner. This in turn leads to increased Wnt and MYC activity (6,15). Alternatively, given that the G allele appears to increase transcription of CCAT2 compared to the T allele in colorectal cancer, it has been suggested that G, the risk allele of rs6983267, changes the property of the final CCAT2 transcript, ultimately influencing its binding capacity to TCF4 (4,6).

Our study supports the notion that increased MYC expression is responsible for unregulated growth in ovarian cancer; however, our results did not show a significant genotype-dependent increase in MYC expression. In fact, direct evidence consistently linking rs6983267 alleles to level of MYC expression have not been demonstrated (5,9,36,37,39,40). Several reasons have been proposed, including differential expression of MYC among the different cell types that comprise an organ. For example, MYC may be expressed at different levels in epithelial cells, germ cells, and stromal cells within the ovary (39). Another reason for this discordance may be the inadequacies of our current technology to pick up subtle differences in MYC transcription (5). Further, this association may not be detected due to timing of risk elevation in relation to presentation of clinical disease, which occurs earlier in the disease course (5,36,40). This was also described by Wasserman et al who analyzed the cancer risk allele in prostate cancer (40). This study demonstrated that allele-specific enhancer activity may be more active early in development before tumorigenesis occurs (40). Capturing a protein level at a single time point, in other words, may not correctly reflect the gene's role in tumorigenesis. Similarly, obtaining tissue samples later in the disease course, as was the case in our study, may not accurately reflect the differences in MYC activity among different genotypes, especially since our population consisted of patients who had known ovarian cancer. In addition, the prior mentioned studies focus on allele-specific MYC expression in relation to cancer risk. Future studies assessing MYC expression after a patient has been diagnosed with cancer in regard to clinical course and response to treatment may prove to be beneficial.

Although differences in allele-specific MYC transcription remain unclear, previous studies have suggested that the T allele had a 2-fold increase in MYC transcription despite extensive evidence associating the G allele with increased cancer risk (4,6,7,10,11,15). These seemingly contradictory findings were addressed by Sotelo et al who connected MYC to a phenomenon called ‘intrinsic tumor suppression’ (15). This phenomenon, first proposed by Lowe et al, describes the tight coupling of cell proliferation and cell death. In normally functioning cells, mutations that drive cell proliferation also possess the ability to activate senescence and apoptosis (41). Oncogenic MYC, for instance, has been shown to trigger the ARF/p53 tumor suppressor pathway to induce apoptosis when levels of MYC reach a certain threshold (41,42). However, Murphy et al showed low levels of MYC failed to activate the apoptotic transcriptional pathway, allowing for tumorigenesis (42). Thus, it has been theorized that low-level uninhibited MYC is more likely to initiate oncogenesis than MYC that is overexpressed (42). Extending this theory to our findings, it may be the case that the TT genotype activates the tumor suppressor pathway to a greater extent than the GG or GT genotype.

In this study variables with a P-value of <0.10 in the univariate analysis were included in the multivariate analysis. A higher P-value than the traditional level of 0.05 was chosen in order to decrease the risk of excluding a potentially important variable (17). Several other studies have utilized similar methods. Hoshimoto et al assessed pre-operative factors associated with survival of cholangiocarcinoma, and incorporated variables with P<0.10 in the univariate analysis into the multivariate model (43). This was also utilized in a publication by Chao et al, which assessed characteristics associated with hepatocellular carcinoma survival after liver transplant. Variables that had a P-value <0.10 in the univariate analysis were included in the multivariate analysis (44). While this model allows for increased inclusivity, it also increases the risk of introducing variables that have a confounding effect on each other, resulting in high intercorrelations among non-significant variables (17).

A limitation of this study was the relatively small sample size and the retrospective nature of the design. Variables such as the Charlson Comorbidity Index, cancer stage, and location of metastatic disease contained subcategories that often had 0–2 cases for a particular genotype, limiting the data analysis. In addition, it is unknown whether the rs6983267 SNP investigated in this study resulted from a germline or somatic DNA mutation. However, given the aim of this study was to analyze the association of survival, rather than cancer risk, with the tumor genotype, our conclusions should not change based on type of DNA mutation. This study was strengthened by the data collection occurring at a single tertiary medical center, which ensured consistency with study protocol in all sample collection and analysis procedures. In addition, due to the diversity of patients treated at a large tertiary medical center, the 98 samples in this study likely represent a broad array of clinical phenotypes in ovarian cancer, although all of them shared a common ancestry (45).

To our knowledge, this is the first study to assess the association between clinical outcomes in patients with HGSOC and genotypes of the rs6983267 SNP within the lncRNA CCAT2. HGSOC patients with the TT genotype at this locus had improved survival time compared to patients with genotypes containing any G allele, despite patients with the TT genotype being diagnosed at a more advanced disease stage. Increased survival may be due to better response to initial chemotherapy by patients with the TT genotype. This study suggests individualized cancer outcomes are influenced by patient genomic variation. We know that certain patients will respond better to chemotherapy, but we are starting to untangle some of the reasons why this may happen. Further studies are needed to discern the intrinsic biological mechanisms of this observation and its potential use as target therapy.

Supplementary Material

Supporting Data

Acknowledgements

Not applicable.

Funding

The present work was supported in part by the National Institutes of Health (grant nos. R01 CA99908 and R01 CA184101), The Basic Research Fund from the Department of Obstetrics and Gynecology at the University of Iowa and by The American Association of Obstetricians and Gynecologists Foundation Bridge Funding Award.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Authors' contributions

DI, NC, JGB and ED conceived the study. ED designed the methodology. ED and JGB validated the study. JGB performed the formal analysis. DI and ED performed the investigation. JGB and ED procured the resources. ED was responsible for data curation. DI wrote the original draft manuscript. DI, NC, JGB and ED reviewed and edited the draft manuscript. JBG visualized the study. JGB and ED supervised the study, were project administrators and acquired funding. JGB and ED confirm the authenticity of all the raw data. All authors have read and approved the final manuscript.

Ethics approval and consent to participate

The study was conducted according to the guidelines of the Declaration of Helsinki, and was approved by The Institutional Review Board (or Ethics Committee) of the University of Iowa (approval nos. #200209010, #201804817, #201809807). Informed consent was obtained from all subjects involved in the study.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

References