GNAS1 T393C Polymorphism and Survival in Patients with Sporadic Colorectal Cancer

Colorectal cancer is one of the leading causes of cancer-related death worldwide. In Germany, there has been a stable incidence of colorectal cancer since 1980, with 51,000 newly detected cases per year and an average 5-year survival rate after tumor resection in the range of 50% (1). Currently, the standard method for predicting outcome in patients is the tumor-node-metastasis system and the International Union Against Cancer (UICC) staging system (2). Patients with UICC stage III (any T, N_1-2, M₀, 5-year survival 30-55%) generally have a worse prognosis than patients with UICC stages I to II (T_3-4, N₀, M₀, 5-year survival 60-95%). However, patients with comparable pathologic disease stage exhibit significantly different survival, especially in stage II colon cancer (3).

The majority of patients with stage II disease is not treated with adjuvant chemotherapy. However, this treatment regimen is mandatory in patients presenting with stage III disease because decreases in relapse and mortality rates by 30% to 40% were observed compared with surgery alone (4). Up to 40% of patients with stage II colorectal cancer will develop recurrent disease during their lifetime and the role of adjuvant chemotherapy in this setting is still unclear (3, 5, 6). The identification of patients with high-risk colorectal cancer could help to set up novel treatment strategies in the lymph node–negative disease (UICC I-II) and may, in principle, improve outcome (7). Therefore, it is desirable to identify molecular markers offering to identify more aggressive colonic cancer phenotypes to individually tailor therapy. In fact, research in the past several years suggested that molecular markers could help to define subgroups of patients who, after surgery, may benefit from adjuvant chemotherapy (8–10). However, lack of standardization in the analysis of immunohistochemically detected biomarkers often prevents their application as prognostic factors. Most of this research has focused on properties of the tumor tissue itself (e.g., somatic mutations or differential expression of genes or proteins), whereas a potential role of genetic host factors has rarely been investigated (9–12).

We have recently shown that genotypes of the single-nucleotide polymorphism (SNP) T393C in the gene GNAS1, encoding the ubiquitously expressed Gαs subunit of heterotrimeric G proteins, predict the clinical outcome of patients with urothelial carcinoma (13). Patients with TT genotypes showed a prolonged progression-free, metastasis-free, and overall survival compared with patients with TC and CC genotypes. We also showed that Gαs mRNA expression is increased in TT genotypes, not only in bladder cancer but also in heart and fat cell specimens (13). These findings strongly suggest that increased Gαs mRNA expression is a general phenomenon in individuals with the GNAS1 393TT genotype (13). Data from in vitro experiments indicate that increased expression of Gαs enhances apoptosis (14–16). Hence, it is tempting to hypothesize that increased Gαs expression with concomitantly enhanced apoptosis may be related to better survival in patients with TT genotype.

Thus, one aim of the present study was to accumulate further evidence that the T393C polymorphism–related altered expression of Gαs is not only associated with the prediction of outcome in bladder cancer patients but may also represent a more general feature with the capacity to predict the clinical course in other tumors, too. To this end, we investigated a potential association between genotypes of the T393C polymorphism and survival in a series of patients suffering from sporadic colorectal carcinoma. It was recently suggested that the T393C polymorphism lies within a recombination hotspot surrounded upstream and downstream within the same gene by two haplotype blocks (17). The second aim of the study was therefore to investigate possible associations of one SNP from the 5′block (T2291C, dbSNP rs6026584) and one SNP from the 3′ block (T18830C, dbSNP rs234628) with outcome from colorectal cancer.

Patients. All patients of this study were operated at a single Hospital (Department of General Surgery, Klinikum Herford, Germany) in the years 1996 to 1998 according to the recommendations of the German Society of Surgery. Mesorectal excision was not done at that time in the group of patients with rectal carcinoma of the middle and lower third of the rectum. Primarily, the files of all patients operated during this time period with cancer of the colon and rectum were retrieved and reviewed; finally, the study cohort (151 patients; for clinicopathologic data, see Table 1) consisted exclusively of patients with a meticulously complete follow-up record. The minimum follow-up period for patients still alive was 60 months.

Histopathologic diagnosis was made according to the WHO classification of tumors of the colon and rectum. Additionally, the tumors were staged and graded by standard histologic analysis according to the guidelines of International Union Against Cancer (2).

Patients with UICC stage III colorectal carcinoma received in the adjuvant setting a standardized chemotherapy with leucovorine 20 mg as an i.v. bolus on day 1 to 5 and fluorouracil 425 mg/m²/d as an i.v. bolus on day 1 to 5, repeated on day 28 for a total of 6 months [Mayo Clinic protocol (18)]. According to the recommendations of the German Surgical Oncology Working Group (CAO), patients with stage II colorectal carcinoma were not advised to any adjuvant treatment. The study was done according the Declaration of Helsinki and approved by the Ethics Committee of the University Hospital Essen.

Blood donors. The Caucasian control sample consisted of 820 healthy white individuals of either gender who were recruited at the local Department for Transfusion Medicine, University Hospital Essen. All samples were collected at random from subjects donating blood and the details of this sample have been published previously (19).

Genotyping. DNA was extracted from paraffin-embedded tumor tissue using a commercially available kit (QIAAmp, Qiagen, Hilden, Germany). Genotypes for the T393C polymorphism were determined as described (13). The T2291C polymorphism was genotyped using primers 5′-AGGCAGAATTATGCTGTTGGGA-3′ and 5′-AGATCCGTGCCTCAGTTTCCAC-3′ to amplify a 508 bp fragment followed by restriction analysis with 1 unit SspI (New England Biolabs, Beverly, MA) resulting in CC = 508 bp, CT = 508, 318, 190 bp, and TT = 318, 190 bp. For the T18830C SNP, PCR was done using primers GNAS_in11_se 5′ TTCAGGGGTTCAGCTACC 3′ and biotinylated primer GNAS_in11_as_BT 5′ GGCACAAACAAGGGGACT 3′ resulting in a 200 bp fragment. Genotypes were determined using Pyrosequencing (20) with sequencing primer GNAS_in11_seq 5′-CTTTGCGCCCCTC-3′.

Statistical analysis and presentation of data. The clinical outcome analyzed in this study was survival dependent on UICC stages and T393C genotypes. Kaplan-Meier plots and the log-rank test for trend were used to evaluate the relationship between UICC stages or T393C genotypes, and clinical outcome from the date of the primary diagnosis to the end of follow-up. Log-rank tests for T393C genotypes were done using all three genotypes unless stated otherwise. The effects of gender, UICC stages, grade, age, and T393C genotypes as prognostic factors for the clinical outcome were analyzed by stepwise multivariate Cox-regression analysis. Hazard ratios and 95% confidence intervals (95% CI) were calculated from the Cox-regression model including all factors for multivariate analysis and for the indicated factor for univariate analysis. Contingency tables and the Pearson's χ² test were used for categorical variables using T393C genotypes. Because the T393C polymorphism shows a gene-dose effect (13), linear ANOVA was used for comparison of continuous variables where appropriate. Haplotype analysis and control for deviation from the Hardy-Weinberg equilibrium were conducted with the public domain programs EH and Hardy-Weinberg equilibrium by J. Ott (http://linkage.rockefeller.edu/ott/linkutil.htm). The EH program analyzes linkage or absence of linkage between allelic markers by comparison of the observed and the predicted genotype distributions using contingency tables to test whether the observed distributions were compatible with a model of independent segregation (21). This allowed subsequent determination of estimated haplotype frequencies and estimated linkage disequilibrium coefficient, D, through the use of the 2LD program (22). D was expressed as D′, giving the value of D as a percentage of the maximum calculated value given the observed allele frequencies. Values of D′ range between −1 and +1. A ∣D′∣ value of 1 denotes complete linkage disequilibrium whereas a value of 0 denotes complete linkage equilibrium. Differences were regarded significant at P < 0.05. All statistical analyses were done using SPSS 11.0 (SPSS, Chicago, IL) or GraphPad Prism 4.0 (GraphPad Software, San Diego, CA). Continuous variables are given as means ± SD or ± SE as indicated.

Demographic characteristics and tumor grade and stage in the whole case group and by genotype are displayed in Table 1. The mean age was 68.1 years (range, 37-91 years) and mean follow-up time was 43 months (range, 1-92 months). Tumor site, grade, and UICC stage were compatible with those reported in the literature (23).

T393C genotype distribution is also displayed in Table 1. The frequency of the C allele (fC) in the patient group was 0.52 and the distribution was compatible with the Hardy-Weinberg equilibrium. To investigate whether the T393C polymorphism is predictive for colorectal cancer, we compared genotypes and allele frequencies with those from 820 healthy, white blood donors. Genotype distribution (TT, n = 182; TC, n = 403; CC, n = 235) as well as C-allele frequency (fC = 0.53) was not significantly different from that of our patient group, which argues against an association of T393C genotypes with an increased risk for colorectal cancer. Gender, tumor site, grade, lymph node status, metastases, and UICC stage were not associated with genotypes (Table 1). However, the comparison of pathologic stages and genotype distribution showed a significant relationship between genotype and pathologic stage at the time of initial diagnosis. CC genotypes seem to be associated with higher pathologic stages compared with TT genotypes (Table 1; P = 0.024).

To confirm that our sample displays the typical clinical course of patients with colorectal cancer, we calculated Kaplan-Meier curves for overall survival depending on the UICC stage. As shown in Fig. 1, overall survival was significantly dependent on pathologic stage (P < 0.0001) and survival rates were comparable with published data (23).

Clinical outcome by GNAS1 T393C genotypes. Overall survival dependent on T393C genotypes was analyzed using Kaplan-Meier survival curves (Fig. 2). Survival was not significantly associated with T393C genotypes when all UICC stages were analyzed together (Fig. 2A). However, in UICC stages I to II, survival was significantly dependent on the T393C genotype with an apparent gene-dose effect (Fig. 2B; P = 0.009). GNAS1 393C homozygous patients displayed a higher risk for death than T393 homozygous patients, with heterozygous patients being at intermediate risk [hazard ratios (HR): CC versus TT: 5.8; 95% CI, 1.4-15.9; P = 0.02; CT versus TT: 2.8; 95% CI, 0.7-7.9; P = 0.19]. Five-year survival rates were 87.8% for TT, 71.0% for TC, and 50.0% for CC (Fig. 2B). No association between survival and genotype was detected for UICC stages III to IV (Fig. 2C).

Because it has recently been suggested that associations reported for the T393C polymorphism could be confounded by SNPs from the 5′ or 3′ haplotype block (17), we chose one SNP in the 5′ block (T2291C) and one SNP in the 3′ block (T18830C) and calculated the linkage disequilibrium between these SNPs and T393C. In addition, we examined their possible associations with survival both using these SNPs separately as well as by haplotype analysis. As shown in Table 2, the T393C and T2291C are in linkage disequilibrium (χ² = 11.6; ∣D′∣ = 0.45; P < 0.01). However, Kaplan-Meier and log-rank statistics using the T2291C SNP alone as well as in haplotype combination with the T393C SNP revealed no significant associations with the clinical outcome at any UICC stage (not shown). The T393C and T18830C SNPs were also in linkage disequilibrium, which was weaker compared with the T2291C (χ² = 7.93; ∣D′∣ = 0.27; P < 0.05; Table 2). No significant association of the T18830C SNP to progression of colorectal cancer could be detected. Moreover, no additional effect or interaction was observed when the T393C and T18830C haplotype was investigated (not shown).

Multivariate analysis including age at diagnosis, UICC stage, grade, gender, and T393C genotypes revealed that the T393C polymorphism was an independent risk factor for death in UICC stages I to II (Table 3). Patients with CC genotype had an HR of 12.1 (95% CI, 2.02-72.6; P = 0.006) compared with the reference group consisting of T393 homozygous individuals. Heterozygous patients were at intermediate risk (HR, 2.51; 95% CI, 0.45-14.1; P = 0.190). The T2291C and T18830C SNPs as well as the corresponding haplotypes with the T393C SNP were not significantly associated with clinical outcome in this multivariate model (not shown).

Characterization of GNAS1 mRNA folding containing the T393C SNP. Previous studies suggest a relationship between mRNA secondary structure and gene expression (24, 25). We therefore hypothesized that the T393C substitution may result in changes of mRNA folding. Using MFOLD (26), we modeled the effect of the T393C polymorphism GNAS1 mRNA folding structures of the coding region. Partial mRNA folding structures of 393T and 393C mRNA are shown in Fig. 3. The T>C substitution causes an obvious change in the mRNA folding structures suggesting differences in Gαs mRNA stability due to this nucleotide exchange. This is consistent with experimental data showing altered mRNA expression in different T393C genotypes (13).

Although UICC stages generally correlate with outcome in a large percentage of patients, more precise prediction of the individual course would facilitate clinical decision making (e.g., about appropriateness of adjuvant therapy). Unfortunately, biological markers for colorectal cancer describing properties of the tumor itself have yielded inconclusive results (8–10). One of the aims of the present study was, therefore, to investigate whether a genetic host factor, the common T393C polymorphism in the gene GNAS1, may be predictive for survival in patients with colorectal cancer. The results of the study clearly indicate that mortality in UICC stages I to II was significantly increased in CC genotypes compared with TC or TT genotypes whereas no association with the genotype was observed in stages III to IV. Furthermore, the T393C polymorphism was shown to be an independent risk factor for clinical outcome when using UICC stages and grading as covariates with the highest risk for an unfavorable clinical course in CC genotypes (Table 3). Finally, CC genotypes presented at an advanced pathologic stage at first diagnosis (P = 0.024; see Table 1).

The T393C polymorphism is located within a recombination hotspot which was supposed to be in linkage equilibrium with two haplotype blocks, one 5′ and one 3′ of the T393C polymorphism (17). Therefore, one SNP (T2291C) of the 5′ block and one SNP (T18830C) of the 3′ block were chosen to analyze linkage disequilibrium; a weak but significant linkage disequilibrium was found between each SNP and the T393C polymorphism (Table 2). However, neither the T2291C nor the T18830C polymorphism was associated with survival of colorectal cancer patients, alone or in combination with the T393C SNP. Together with our previous observation that the T393C polymorphism is a predictive marker for clinical outcome in bladder cancer (13), the present results strongly support its potential role as a possible general genetic marker for tumor progression.

We have recently shown that the GNAS1 TT genotype is associated with increased Gαs mRNA expression in different tissues (13). From the above-mentioned observations, we hypothesized that the T393C exchange itself could have an effect on mRNA stability. Interestingly, determinants of mRNA stability have been described in the coding region of some genes (27–29). We found that the T>C substitution at position 393 changes the mRNA folding structures predicted by MFOLD (Fig. 3). Therefore, genotype-dependent differences in mRNA decay due to altered secondary structure could finally cause differences in Gαs mRNA expression (13). Functional studies will have to verify this hypothesis.

Finally, it still may be speculated that yet unidentified functional SNPs in GNAS1 are in strong linkage disequilibrium with T393C and thus exert an effect on transcription of GNAS1.

In vitro experiments suggest that increased expression of Gαs is associated with enhanced apoptosis (14–16). The second messenger, cyclic AMP, which is generated by activated Gαs, seems to play a major role in this proapoptotic process. An increased concentration of the intracellular second messenger cyclic AMP promotes apoptosis in several cell types including leukemic cells (30), ovarian cancer cells (31), and lymphoma cells (32). Increased Gαs expression in tissues of patients with TT genotypes may therefore confer an enhanced apoptosis in 393T allele carriers. Hypothetically, this mechanism may contribute to a more favorable clinical course in cancer patients. This hypothesis, although attractive, remains to be supported by additional functional studies which, however, were beyond the scope of the present report.

A major finding of the present study is that genotypes of the T393C polymorphism are associated with survival of colorectal cancer patients only in UICC stages I to II, but not in stages III to IV. Because the T393C polymorphism is expected to exert an effect on early events of intracellular signal transduction (i.e., receptor-G protein coupling and downstream signaling events), this effect is hypothetically expected only in earlier cancer stages, in which the cell cycle is still under control of the “normal” signal transduction machinery. Once cell cycle regulation, migration, and proliferation of cancer cells are “out of control” due to accumulating somatic mutations in cell cycle-controlling proteins, as expected in higher-grade tumors, alterations in the efficacy of early signal transduction may have little or no effect. Despite these considerations, the effects of the T393C polymorphism are relatively strong, as seen from the high HR associated with the CC genotype, which is comparable to that of UICC staging for all patients or tumor grade in patients with tumors UICC I to II (see Table 3).

Both the results of the present study and previous findings (13) strongly suggest a role of the T393C polymorphism in tumor progression. Nevertheless, it has to be emphasized that in both studies a limited number of patients were investigated, which became especially evident when subgroups were analyzed. Therefore, although our findings support the concept of a role of genetic host factors in tumor progression, further independent studies are necessary to confirm their validity.