TGFBR2 and BAX Mononucleotide Tract Mutations, Microsatellite Instability, and Prognosis in 1072 Colorectal Cancers

Background Mononucleotide tracts in the coding regions of the TGFBR2 and BAX genes are commonly mutated in microsatellite instability-high (MSI-high) colon cancers. The receptor TGFBR2 plays an important role in the TGFB1 (transforming growth factor-β, TGF-β) signaling pathway, and BAX plays a key role in apoptosis. However, a role of TGFBR2 or BAX mononucleotide mutation in colorectal cancer as a prognostic biomarker remains uncertain. Methodology/Principal Findings We utilized a database of 1072 rectal and colon cancers in two prospective cohort studies (the Nurses' Health Study and the Health Professionals Follow-up Study). Cox proportional hazards model was used to compute mortality hazard ratio (HR), adjusted for clinical, pathological and molecular features including the CpG island methylator phenotype (CIMP), LINE-1 methylation, and KRAS, BRAF and PIK3CA mutations. MSI-high was observed in 15% (162/1072) of all colorectal cancers. TGFBR2 and BAX mononucleotide mutations were detected in 74% (117/159) and 30% (48/158) of MSI-high tumors, respectively. In Kaplan-Meier analysis as well as univariate and multivariate Cox regression analyses, compared to microsatellite stable (MSS)/MSI-low cases, MSI-high cases were associated with superior colorectal cancer-specific survival [adjusted HR, 0.34; 95% confidence interval (CI), 0.20–0.57] regardless of TGFBR2 or BAX mutation status. Among MSI-high tumors, TGFBR2 mononucleotide mutation was associated with CIMP-high independent of other variables [multivariate odds ratio, 3.57; 95% CI, 1.66–7.66; p = 0.0011]. Conclusions TGFBR2 or BAX mononucleotide mutations are not associated with the patient survival outcome in MSI-high colorectal cancer. Our data do not support those mutations as prognostic biomarkers (beyond MSI) in colorectal carcinoma.


Introduction
Colorectal cancer represents a group of molecularly heterogeneous diseases with different sets of epigenetic and genetic abnormalities. High degree of microsatellite instability (MSI-high) is caused by deficiency of DNA mismatch repair system, and observed in approximately 15% of colorectal cancers. MSI testing is widely used as screening for patients with Lynch syndrome/ hereditary nonpolyposis colorectal cancer (HNPCC) [1,2,3]. In addition, MSI is generally accepted as a prognostic marker [4], and likely a predictive marker for resistance to 5-fluorouracil [5]. Since Markowitz et al. [6] discovered mutations in the coding mononucleotide repeats of TGFBR2 in MSI-high colon cancer cells, similar mutations of coding mononucleotide repeats in many other genes (including BAX, MSH3, MSH6, IGF2R and PTEN) have been found in MSI-high colorectal cancers [3,7,8,9]. Among those genes, mononucleotide coding repeats of TGFBR2 (A) 10 and BAX (G) 8 have frequent frameshift mutations resulting in the production of truncated, inactive form of the proteins [3,10]. TGFB1 (transforming growth factor-b, TGF-b) and its receptor TGFBR2 constitute a signaling pathway that regulates the transcription of many genes, and functions as a tumor suppressor [11,12,13,14] and an immune response regulator [15]. BAX generally promotes apoptosis and antagonizes the effect of BCL2 [16,17,18]. Thus, inactivation of TGFBR2 or BAX may contribute to tumor progression.
Several previous studies have examined the prognostic role of TGFBR2 or BAX mononucleotide mutations in MSI-high colorectal cancers, yielding inconclusive results due to limited statistical power in most studies [19,20,21,22,23,24,25] (Table 1). All but one previous study [23] examined the prognostic role of TGFBR2 or BAX mononucleotide mutation in less than 100 MSIhigh tumors (the number of MSI-high tumors ranging from 16 to 98) [19,20,21,22,23,24,25]. In addition, none of the previous studies [19,20,21,22,23,24,25] has comprehensively examined potential confounding effect of key molecular biomarkers in colorectal cancer, including the CpG island methylator phenotype (CIMP), and KRAS, BRAF and PIK3CA mutations. Thus, the prognostic role of TGFBR2 or BAX mononucleotide mutation in MSI-high tumors still remains uncertain. We conducted this study to test the hypothesis that TGFBR2 or BAX mononucleotide mutations in colorectal cancer were associated with altered tumor behavior (beyond MSI), utilizing a database of 1072 stage I to IV colorectal cancers in two prospective cohort studies. Our current study represents the first study which utilized a database of prospective cohort studies to test the stated hypothesis. This fact increases generalizability of our study findings. Moreover, because we concurrently assessed clinical, pathologic and tumor molecular variables such as the CpG island methylator phenotype (CIMP), LINE-1 methylation, KRAS, BRAF and PIK3CA mutations, we could evaluate the effect of TGFBR2 or BAX mutation independent of these potential confounders.

Study group
We utilized the database of two prospective cohort studies, the Nurses' Health Study (N = 121,701 women followed since 1976) and the Health Professionals Follow-up Study (N = 51,529 men followed since 1986) [26,27]. Participants have been sent biennial questionnaires to update information on potential risk factors and to identify newly diagnosed cancers in themselves and their first degree relatives. We collected paraffin-embedded tumor tissue blocks of incident colorectal cancers from hospitals throughout the U.S. where participants with colorectal cancer underwent tumor resection [26,27]. Clinical characteristics of the cases are described in Table 2 (on the left, under the column heading ''All cases''). There was no significant difference in demographic features between cases with tissue available and those without available tissue among our cohort studies [26]. A majority of cases have previously been characterized for statuses of TGFBR2, MSI, CIMP, KRAS, BRAF, PIK3CA and LINE-1 methylation [28,29,30,31]. However, none of our previous studies have analyzed the prognostic significance of mononucleotide mutation of TGFBR2 or BAX. BAX mutation has not been analyzed in any of our previous studies. Thus, this study represents a new study utilizing a resource of the existing materials and database, analogous to novel studies using well-described cell lines (e.g., SW480 cell line) or mouse models (e.g., Apc min mouse model).
Hematoxylin and eosin stained tissue sections from all colorectal cancer cases were reviewed by a pathologist (S.O.) unaware of other data. Tumor differentiation was categorized as wellmoderate vs. poor (.50% vs. #50% glandular areas). We excluded cases which were preoperatively treated. Based on the availability of adequate follow-up and tumor tissue data, 1072 stage I-IV colorectal cancer cases diagnosed up to 2004 were included in the current study ( Figure 1). Patients were observed until death or June 30 2009, whichever came first. Death of a participant was confirmed by the National Death Index. Returning questionnaire indicated informed consent from all study subjects. Informed consent was obtained from all study subjects. Tissue collection and analyses were approved by the Human Subjects Committees at Harvard School of Public Health and Brigham and Women's Hospital.

Statistical analysis
We used SAS program (Version 9.1, SAS Institute, Cary, NC) for all statistical analyses. All p values were two-sided. When we perform multiple hypothesis testing (i.e., analyses of molecular correlates and interactions), a p value for statistical significance was adjusted to p = 0.0038 ( = 0.05/13) by Bonferroni correction. The chi-square test (or Fisher's exact test) was performed for categorical variables. The t test assuming unequal variances was done to compare mean age and mean LINE-1 methylation level. For survival analysis, the Kaplan-Meier method and log-rank test were used. For analyses of colorectal cancer-specific mortality, deaths as a result of causes other than colorectal cancer were censored. To control for confounding, we used multivariate stagematched (stratified) Cox proportional hazards model to compute hazard ratio (HR) of death. To avoid residual confounding and overfitting, disease stage (I, II, III, IV, unknown) was used as a stratifying variable, utilizing the ''strata'' option in the SAS ''proc phreg'' command. The multivariate model initially  avoid overfitting. We confirmed that excluding cases with missing information in any of the covariates did not substantially alter results (data not shown). A multivariate logistic regression analysis was performed to examine an independent relationship of each covariate with TGFBR2 mutation (as an outcome variable). The multivariate model initially included a similar, but not the same set of the covariates as the initial Cox model, considering possible cause-effect relationship with TGFBR2 mutation. Specifically, disease stage and tumor differentiation were likely consequences (rather than causes) of TGFBR2 mutation. Thus, those variables were not included in the logistic regression model. A backward elimination with a threshold of p = 0.10 was used to select variables in the final model and avoid overfitting.

Multivariate analysis to assess independent relations with TGFBR2 mutation
We performed multivariate logistic regression analysis to examine whether TGFBR2 mutation was independently associated with any clinical, pathologic and other molecular variables. In MSI-high tumors, TGFBR2 mutation was independently associated with CIMP-high [multivariate odds ratio (OR), 3.57; 95% confidence interval (CI), 1.66-7.66; p = 0.0011].

Mononucleotide mutations of TGFBR2 and BAX and colorectal cancer prognosis
During adequate follow-up (11.6 years of median follow-up of censored cases), there were 505 deaths including 302 colorectal cancer-specific deaths. Among all cases, MSI-high was significantly associated with longer colorectal cancer-specific survival compared to MSS/MSI-low cancers by log-rank test (p,0.0001), univariate and multivariate Cox regression analysis (adjusted HR, 0.34; 95% CI, 0.20-0.57; p,0.0001) ( Table 3). When we separately examined TGFBR2-mutated MSI-high cases and TGFBR2-wildtype MSI-high cases, both groups showed significantly longer colorectal cancer-specific survival compared to MSS/MSI-low cases ( Figure 2, Table 3). When we separately examined BAX-mutated MSI-high cases and BAX-wildtype MSI-high cases, both groups  showed significantly longer colorectal cancer-specific survival compared to MSS/MSI-low cases ( Figure 2, Table 3). In overall mortality analyses, although somewhat attenuated, results showed similar trends ( Table 3). Among MSI-high cases, patient survival did not significantly differ by TGFBR2 or BAX mutation status.
We compared colorectal cancer specific and overall survival between TGFBR2-mutated MSI-high cases and TGFBR2-wildtype MSI-high cases (or between BAX-mutated MSI-high cases and BAX-wild type MSI-high cases). There was no significant difference between the two groups ( Table 1).

Discussion
We conducted this study to examine the prognostic significance of mononucleotide tract mutations in the coding regions of TGFBR2 or BAX in MSI-high colorectal cancers. We utilized two prospective cohort studies with a large number of clinically and molecularly well-annotated colorectal cancer cases with adequate follow-up. Our result showed that MSI-high tumors were associated with indolent tumor behavior regardless of TGFBR2 or BAX mononucleotide mutation status, independent of CIMP and other key tumor molecular biomarkers. Nonetheless, it may be of interest to examine interactions between these molecular alterations and dietary and lifestyle factors if there is a hypothesis in evolving science of molecular pathological epidemiology [42,43].
It should be noted that small studies are more prone to ''publication bias'' than large studies [44]. This phenomenon of publication bias occurs because studies with null findings have a higher likelihood of being unwritten and unpublished compared to those with significant results. Compared to small studies (e.g., studies with a sample size of ,200 cancers) with null data, large studies with null data are more likely published. As a result, large studies are less prone to publication bias than small studies. Furthermore, academic pressures might force investigators to design small studies which are easy to complete and get data for haste publications, which might contribute to bias [45,46,47]. Therefore, we should weigh more on large-scale studies when we evaluate the published literature on prognostic significance of any biomarker such as TGFBR2 or BAX mononucleotide mutation. Publishing null data in well-powered studies [44,48,49,50,51,52] are important because publishing significant results in small underpowered studies leads to publication bias.
Our data are generally consistent with some of previous studies [19,22,24,25] (Table 1). Watanabe et al. [21] used stage II and III cases that underwent adjuvant chemotherapy, and reported that, TGFBR2 mutation was associated with improved 5-year overall survival among 73 MSI-high tumors. In another study [20], among 67 MSI-high tumors, BAX mutation was associated with poor prognosis. In an underpowered study by Fernández-Peralta et al. [23], among 16 MSI-high tumors, both TGFBR2 mutation and BAX mutation were associated with better prognosis. The largest study (total N = 1427; 170 MSI-high cancers) by Samowitz et al. [22] showed no prognostic role of TGFBR2 or BAX mutations among MSI-high colorectal cancer cases, in agreement with our current study -the second largest study to date and the only study which examined other key tumor molecular biomarkers such as CIMP, LINE-1 methylation and KRAS, BRAF and PIK3CA mutations.
Studying somatic molecular changes and molecular correlates is important in cancer research towards personalized medicine [53,54,55,56]. The CpG island methylator phenotype (CIMP) has been established as an epigenomic molecular classifier of colorectal cancer [57,58,59,60,61,62,63,64,65,66,67,68,69,70,71]. In the past, Iacopetta et al. [20] showed no significant association between KRAS mutation and TGFBR2 mutation. We assessed the association between tumor molecular variables (CIMP, LINE-1, KRAS, BRAF and PIK3CA) and TGFBR2 mutation and did not find significant relation between TGFBR2 mutation and KRAS or BRAF mutation. Interestingly, we have found that, among MSI-high tumors, TGFBR2 mutation was associated with CIMP-high, independent of clinical and other molecular features. A recent study [72] has reported that genetic variants in the TGFB1 pathway related genes (MAPK1, RUNX1 and RUNX2) are associated with CIMP-high colon cancer. Further studies are needed to elucidate the exact mechanism of the relationship between CIMP and the TGFB1 pathway.
There are limitations in this study. For example, data on cancer treatment were limited. Nonetheless, it is unlikely that chemotherapy use substantially differed according to TGFBR2 or BAX mutation status in tumor, since such data were typically unavailable for treatment decision making. As another limitation, beyond cause of mortality, data on cancer recurrences were unavailable in these cohort studies. Nonetheless, given median follow-up of over 11 years for censored cases, colorectal cancerspecific survival might be a reasonable surrogate of colorectal cancer-specific outcome.
There are advantages in utilizing the database of the two prospective cohort studies, the Nurses' Health Study and the Health Professionals Follow-up Study, to examine prognostic significance of tumor biomarkers. Anthropometric measurements, family history, cancer staging, and other clinical, pathologic, and tumor molecular data were prospectively collected, blinded to patient outcome [26]. Cohort participants who developed cancer were treated at hospitals throughout the U.S., and thus more representative colorectal cancers in the U.S. population than patients in one to a few academic hospitals. There were no demographic difference between cases with tumor tissue analyzed and those without tumor tissue analyzed [26]. Finally, our rich tumor database enabled us to simultaneously assess pathologic and tumor molecular correlates and control for potential confounding by the tumor molecular features.
In conclusion, our large tumor database has shown that, compared to MSS/MSI-low cases, MSI-high colorectal cancer is associated with longer cancer-specific survival, regardless of TGFBR2 or BAX mononucleotide tract mutation status. The importance of large-scale studies cannot be overemphasized because, compared to large studies, small studies are much more prone to publication bias, which can mislead clinical practice.