Pathological analysis of resected tumor specimens is still the best established method to predict the survival of colorectal cancer patients [20
]. Nevertheless, the clinical outcome of patients bearing tumors with the same stage is very heterogeneous. To define the prognosis and to plan appropriate adjuvant treatment, additional predictors are needed. This is particularly important for patients with stage II disease who are not routinely treated with adjuvant chemotherapy, and among whom selection of bad-risk patients who would significantly benefit from adjuvant treatment would be especially helpful. Furthermore, with the introduction of more efficient but also more toxic adjuvant therapy regimens, risk stratification becomes an issue of even higher relevance.
This study strongly suggests that the gene copy status of SKI is an independent predictor of relapse-free survival and OS in early colorectal cancer. Amplification of SKI was identified as a negative prognostic marker in stage II and stage III colorectal cancer, with a risk of relapse twice as high as for tumors bearing a normal diploidy at the SKI locus. In the multivariate analysis, this association was shown to be independent of age, sex, lymph node status, tumor size, and tumor location, and can therefore be used as an independent prognostic marker to identify high-risk situations, thus influencing the decision to provide adjuvant treatment. Based on our data, patients with stage II disease and SKI-amplified tumors should be considered to have a higher risk of relapse, representing a molecularly defined subgroup that might benefit from adjuvant treatment. However, due to the low rate of SKI amplification (10.1%), the 95% confidence interval for HR is quite large and, therefore, a confirmatory investigation in a larger sample would be important to definitely establish the role of SKI amplification.
The association of SKI
amplification and unfavorable prognosis implies a causal relation to colorectal tumorigenesis. Considering, however, that it affects only 10% of the tumors, it should not be regarded as a prerequisite but as a possible lesion in the multistage colorectal tumorigenesis. This is in line with the findings of other TGF-β signaling molecules, which are frequently, but not in all cases, altered in colorectal cancer by point mutation [11
] or deletion. SMAD2, SMAD4
, and SMAD7
, for example, were found to be deleted in 66%, 64%, and 48% of the colorectal tumors, respectively [4
]. Of these, SMAD7
has been shown to be a positive prognostic factor [2
], whereas SMAD2
seem to be unrelated to a prognostic effect. This is in line with the view that colorectal cancer may harbor defects in the TGF-β signaling pathway at different levels, influencing tumor biology and thereby the patient's prognosis.
Our result of an association between SKI
amplification and worse prognosis is easily explained by several previous findings. As expected for a proto-oncogene, overexpression of SKI
has been shown to induce morphological transformation and anchorage independence [13
]. Furthermore, it has been demonstrated that overexpression of Ski abrogates not only transcription of TGF-β-responsive genes, but also TGF-β-induced growth arrest (i.e., it has a tumor-promoting potential) [15,16
]. In melanoma, SKI has been found not only in the nucleus but also in the cytoplasm, where it comes to a SKI/SMAD2/3 association that seems to prevent SMAD3 nuclear translocation in response to TGF-β [21
]. A recent report demonstrated that SKI can block TGF-β signaling by interfering with the phosphorylation of SMAD2 and SMAD3 by the activated TGF-β type I receptor. In addition, it facilitates the assembly of SMAD2/SMAD4 and SMAD3/SMAD4 complexes independent of TGF-β signaling, thereby disabling TGF-β signaling by engaging SMAD proteins in nonproductive complexes [22
]. TGF-β target gene transcription, which in early tumor stages results in cell cycle arrest, can be inhibited by all these mechanisms, thereby explaining the facilitated growth of SKI
-amplified tumors. As an interesting alternative mechanism, in human melanoma cells, SKI has been shown to be a coactivator of the Wnt/β-catenin signaling pathway [23
]. This pathway has been well characterized in colorectal cancer and activation promotes tumorigenesis. Through this mechanism, the tumor-promoting activity of SKI
amplification in colorectal cancer would be easily explained.
On the contrary, SKI
deletion did not have any prognostic relevance in our analysis. In accordance with our study, which found SKI
deletions in 47% of the samples, loss of chromosome 1p36, the chromosomal region where SKI
is located, has been reported in several studies at frequencies between 30% and 80% [24–27
]. Two studies reported a negative prognostic value of 1p36 deletion [28,29
]. Our data, however, do not support a prognostic relevance of 1p36 and specifically not of SKI
. The frequent deletions of SKI
in 47% of colorectal cancers suggest that SKI
might act as a tumor-suppressor gene. In fact, in Ski
-deficient heterozygous mice, an increased susceptibility to tumorigenesis, mainly of lymphomas and leukemias, was observed. In that study, however, no colorectal cancers were observed [30
]. This might be due to interspecies variations in tumor susceptibility, but it could also mean that SKI
deletion is not causally linked to colorectal carcinogenesis.
, the other member of the SKI
gene family, was also found to be frequently deleted (55% of the tumors) in our colorectal cancer samples. SNON
is located at 3q36, a region with a high frequency of loss in human osteosarcoma. In the mouse, SnoN
has been demonstrated to act as a tumor suppressor, but again, there were no colorectal cancers observed [31
]. This is in line with our data suggesting SNON
gene copy alterations not to be significantly associated with patients' prognosis. We have to be careful when excluding a possible prognostic value of SNON
deletion because our lack of statistical significance might be due to the relatively low power of our analyses. The statistical power for SNON
deletion (deletion versus
no change) was 0.33 in the DFS analysis and 0.29 in the OS analysis.
Our finding of a lack of significant interaction (P > .2) between treatment and gene copy status of SKI and SNON is consistent with the suggestion that the effect of chemotherapy is not mediated through genes that are regulated by the interaction of SKI or SNON with the SMAD3 and SMAD4 complex. However, due to the small sample size, the lack of significance is difficult to interpret because the statistical power to exclude an interaction is low.
In conclusion, we have identified SKI amplification as an independent prognostic factor in early-stage colorectal cancer. Our finding might become relevant in the current attempts to improve risk stratification in the decision for or against an adjuvant treatment regimen.