br Although Muc is one of
Although Muc2 is one of the downstream transcriptional targets of CDX2, 42% of cases had discordant overall expres-sion of CDX2 and Muc2 with loss of 1808011-22-4 of one but not both genes. Data from The Cancer Genome Atlas (TCGA) da-tabase of colon cancer cases showed reduced Muc2 gene ex-pression in 38% of cases and reduced CDX2 expression in 20% of cases (Fig. 7A). However, there is little overlap be-tween these cases; most cases with reduced Muc2 expression showed no decrease in CDX2 expression. Separating cases based on CDX2 expression alone in the TCGA cohort did not show a difference in overall survival or progression-free survival (Fig. 7B and C). In contrast, if Muc2 gene expression status is used, a significant difference in overall survival is present and the progression-free survival shows a trend toward improved survival in cases that express Muc2 (Fig. 7D and E). In cases with loss of CDX2 and/or Muc2, there is a significant difference in both overall and progression-free survival (Fig. 7F and G). 4. Discussion
Given the need for an improved means to identify high-risk early-stage CRC patients who would benefit from chemother-apy, some studies have suggested that CDX2 status may pro-vide a means to identify these patients . However, as our study shows, CDX2 expression status is difficult to implement into clinical practice as a reliable predictive biomarker. We found that CDX2 expression is variable within a section as well as between different sections of the same tumor. The de-creased CDX2 expression is thought to be mediated by vari-able methylation of the CDX2 gene and has been associated with CIMP in CRC [19-21]. Furthermore, the lack of well-characterized mutations, deletions, or gene arrangements pre-cludes the use of sequencing or fluorescence in situ hybridiza-tion–based assays to confirm equivocal cases. Taken together, this makes CDX2 immunohistochemistry a potentially unreli-able biomarker for the prognostication of patients with CRC.
We focused our study on stage II patients who did not re-ceive adjuvant chemotherapy, as this group had the largest dif-ference in survival in the previous studies  and represents the population where there is an unclear role for chemother-apy. All cases were reviewed retrospectively, as this would be a challenging question to answer in a prospective manner given the low rate of recurrence and long follow-up times re-quired. Ideally, the question of CDX2 as a predictive marker would be addressed in a randomized prospective trial, but this is beyond the scope of this and previous studies. In our cohort, there was an increased number of right-sided colon cancers, which may have selected for an increased number of MSI-re-lated cancers. There are mixed results regarding the associa-tion between MSI status and CDX2 expression [20-26], but it seems clear that CDX2 loss is associated with the CIMP methylator phenotype [19-21] that is often seen in right-sided colon cancers. The MSI status is unknown for most of our cases, as they predated routine immunohistochemistry testing for MMR proteins at our institution. Given the fact that this question has been extensively studied in the literature with varying results and we have not identified an association be-tween CDX2 and outcome, linking CDX2 and MSI status in our cohort would likely be uninformative.
We did not identify a difference in survival based on CDX2 immunohistochemical expression status as has been reported previously . Other groups have also reported variable find-ings in terms of CDX2 expression and outcome in colon can-cer [12-15]. The reason for the discordant results is not clear. To exclude an antibody factor, we repeated immunohisto-chemistry with the same antibody used in another study  in a subset of our cases (Fig. 4). We found a similar number of CDX2-negative cases in our cohort (11% versus 9%) to the previous study . The rate of CDX2 loss in CRC has been variably reported in the literature from 4% to 29% [6,19,39-41]. The negative effect of CDX2 loss has been suggested to be limited to a specific subgroup of colon cancers with a mes-enchymal gene expression signature . One possible
explanation for the divergent results may be a larger number of cases with this mesenchymal gene signature in some patient cohorts.
Our study was powered to detect a difference based on the outcome reported in the previous 2016 study . A very recent study has addressed this question using a large cohort of pa-tients with 1157 stage II colon cancer cases and has found a statistically significant but less dramatic effect on outcome with a hazard ratio of 1.54 . Using this effect size, our study would be underpowered to detect a difference in out-come based on CDX2 status in stage II colon cancer. How-ever, in our cohort, Muc2 loss was associated with significantly reduced cancer-specific survival with a hazard ra-tio of 3.32. This suggests that other markers alone or in combi-nation with CDX2 may better predict early-stage colon cancer with a poor prognosis that may benefit from additional therapy.