The main finding of our study was that CK7 expression in CRC was an independent relatively strong, negative prognostic indicator. The p-value for CK7+/CK7 − was surprisingly low (p = 0.007 and p = 0.011 adjusted on grade); lower than in low/high grade (0.062) and right/left-sided tumors (p = 0.043); however, the association between CK7 + and high-grade CRC and the association between CK7 + and right-sided tumors was not significant.
Our results differ from those reported in several previous studies over the last two decades regarding anatomic distribution and prognostic significance of CK7 and CK20 expression in CRC. Zhang et al. described enrichment of rectal carcinomas in terms of CK7 positivity compared to proximal colon tumors [19], which sharply contrasts with our finding of more frequent CK7 expression in right-sided tumors. In line with our results, Bayrak et al. found that CK20 + tumors were more frequent in the left colon and rectum, and they found an association between CK20 + status and low-grade tumors. Moreover, Bayrak et al. described a more frequent expression of CK7 in CRCs with regional lymph node metastatic involvement, without association of CK7 with grade [6]. Hernandez et al. described a more frequent occurrence of CK7 positivity in advanced stage CRCs than early-stage cancers [10]; nevertheless, as with Bayrak [6], the authors did not perform a survival analysis that included patient follow-up. In our study, CK7 + tumors had worse survival without evidence for an association with advanced stage.
Harbaum and colleagues specifically addressed the prognostic role of CK7 in CRC using a similar cohort size (350 patients with follow-ups) as we did. Although their study showed CK7 positivity in 9% of CRCs and the negative prognostic value of CK7 was borderline insignificant with p = 0.06. Moreover, Harbaum et al. found a significant association between CK7 positivity and high grade (p = 0.013); and slight association (borderline insignificant) between CK7 positivity and right-sided tumors (p = 0.07) and an MMR-deficient status (p = 0.12). The association between CK7 expression and an MMR-deficient status was not corroborated in our study; however, this is hard to explain since CK7 + tumors had a worse prognosis and MMR-deficient tumors had a better prognosis. Harbaum et al. also found CK7 positivity in tumor budding cells and explained this as a marker of dedifferentiation and invasion in CRC [9]. In comparison with Harbaum´s study, our results showed a surprisingly higher level of statistical significance (p = 0.007) despite similar cohort sizes: This might be explained by random chance given the relatively low numbers of patients with CK7 + CRCs in general. Moreover, our observations of CK7 + CRCs were irrespective of the presence of budding.
Yamagishi et al. paid specific attention to CRCs with variant histology (i.e., poorly differentiated, mucinous, signet ring, etc). The authors found a negative prognostic significance for CK20 expression in poorly differentiated carcinomas, while this association was not found in well/moderately differentiated CRCs or in CK7 [18]. In concordance with our results, the authors showed increased CK20 expression in MMR proficient tumors, which is in line with previously published studies [8, 14, 15].
CK7 expression signifying worse survival was described by Loupakis et al. in a subset of V600EBRAF mutated CRCs with metastatic disease [13], CK7 expression was found to be markedly more common in BRAF mutated MSS tumors [12]. Moreover, in BRAF mutated CRCs, there was evidence that CK20 negativity was a negative prognostic indicator [13]. Our analysis of the prognostic role of CK7 was performed retrospectively and without knowledge of BRAF status.
Droy-Dupré et al. clustered 122 CRCs describing CK20 + CK7-CDX2 + MMR + so called “crypt-like carcinoma” and minor MUC5AC + CK7+/- subtype with foveolar gastric phenotype; the latter cluster of CRCs displayed worse prognosis [20]. However, this study is not directly comparable to ours because of different way of cohort grouping based on different examined markers.
A recent study performed by Al-Maghrabi and colleagues included survival analysis of 144 cases of CRC to clarify the significance of the CK expression pattern; the study failed to find evidence of any significant relationship [5]. However, due to the overall paucity of CK7 + CRCs, statistically significant results are hard to obtain when analyzing relatively small datasets.
Concerning methodology, the best percentage cut-off points for considering CK7+/CK7 − and CK20+/CK20 − tumors, remains unresolved. Loupakis et al. considered tumors with staining in > 10% of tumor cells to be positive, just as we did with CK7 [13]. Al-Maghrabi considered CRCs staining in > 5% to be positive [5]. Yamagishi et al. used a cut-off value of 1% for CK7 and 25% for CK20. The authors argue that since normal colonic mucosa does not express CK7 and only show diffuse expression of CK 20, any CK7 expression should be considered abnormal even at 1%, and CK20 < 25% should be evaluated as abnormally downregulated [18]. The point concerning CK20 is reasonable from our point of view, but for CK7, we think that expression in < 10% of tumor cells barely represents a significant expression since we had some cases in which CK7 + staining was only present in single cells, which in the words of Harbaum et al. could be a “freak of nature” [14]. The biological significance of CK7, relative to the progression of CRC, can be seen in our results; the p-value observed in the Kaplan-Meier analysis using a cut-off value of 10% was lower (rmean survival 4.98 vs. 7.74 years, p = 0.007) than using 1% as the cut-off (rmean survival 5.79 vs. 7.74 years, p = 0.036); notwithstanding, the latter analysis contained more cases regarded as CK7+ (18 and 28 CK7 + probands, respectively). Nevertheless, like all studies using the time and cost sparing TMA technique, questions related to tumor heterogeneity cannot be ignored. When using a 1% cut-off value, there is a significant possibility that the entire tumor will have an even smaller overall percentage of CK7 positive cells.
From our point of view, the other significant results from our study, i.e., the association between MMR status and tumor location (the negative prognosis of a proximal site) as well as high grade and advanced stage are well known and have been widely discussed in the past and thus, do not require any further discussion.
An interesting observation was found when comparing the 5-year survival in CK7 negative vs. CK7 positive tumors, 65.4%, and 29.4%, respectively, while the general 5-year survival for CRC is 65% (based on a large dataset) [20]. The noticeable similarity to the 5-year survival of patients with CK7 − CRCs in our study was not so surprising since the vast majority of CRCs are CK7−. However, the 5-year survival of CK7 + CRCs strikingly resembles the 5-year survival of lung adenocarcinoma, which ranges from 22–35% [25, 26]. The same can be seen in our study when comparing the mean cancer-specific survival of 4.98 years in CRCs with > 10% CK7 + cells and 5.79 in those with > 1% CK7 + cells, respectively, whereas the mean survival in lung adenocarcinoma varies between 3.4 and 5.4 years [27, 28]. There was a surprising similarity between the survival rate in CK7 + CRC and lung adenocarcinoma, which consistently express CK7. However, for a more in-depth analysis of this interesting phenomenon, a larger data set is needed. Concerning carcinomas originating in endoderm-derived organs, the prognosis of pancreatobiliary cancer (usually CK7+) is much worse than in CK7 + CRC and lung adenocarcinoma. The Kaplan-Meier analysis in our study suggests CK7 + is a stronger negative prognostic indicator than histopathological grade; with p-values of 0.0072 for CK7 + and 0.062 for grade, which is despite many high-grade CRCs being compared to a small number of CK7 + CRCs.
All this suggests that CK7 positivity represents a particular molecular cytoskeletal phenotype in CRC with more aggressive tumor behavior; however, at the cellular level, this is difficult to interpret. There is much more that needs to be explained in the future.
In routine metastatic adenocarcinoma histology, CK7 positivity is often regarded as an argument against a lesion`s colorectal origin. Our study results suggest that this reasoning needs to be revised to acknowledge that the CK7 + is a rare feature in CRC, but one that has substantial prognostic value. Since CK7 immunohistochemistry is easy, widely used, and relatively inexpensive, we argue that it could be included in the standard histology panel to aid with the prognostic stratification of patients. However, more studies focused on the prognostic significance of CK7 in unselected cases of CRC are needed in this field.
In conclusion, our study provides interesting results indicating that cytokeratin 7 in colorectal carcinoma is an independent marker of a poor prognosis, which may be regarded as unexpected in context of several previously published studies that examined the same question.