Biomarker use to help predict treatment response for mCRC patients first began in 2008, when KRAS mutation was found to predict for lack of benefit to anti-EGFR agents. Presently, in mCRC, we now only have two lines of therapy for the 40% of patients that have KRAS mutation, and KRAS wild-type patients have three regimens. The main goal of this study was to find additional drug combinations that might have activity and which are traditionally not used in treatment of CRC; moreover, we aimed to identify the patients which might benefit most from these repurposed regimens.
Docetaxel has proven activity in many solid tumors including breast, prostate, gastric and lung cancers but early studies in unselected colon cancer patients were disappointing.36–39
The EORTC phase II study of docetaxel in metastatic colon cancer showed a 9% response rate, with one complete response noted, while Pazdur et al. reported 2 minor responses not reaching criteria for a partial response in another phase II study of the agent. Our data showed differential sensitivity based on CHFR
expression, as regulated by methylation, in CRC cell lines. Considering the mechanism of CHFR as an important checkpoint protein causing cellular arrest, this suggests that treatment with an anti-microtubule agent would result in increased cytotoxicity if CHFR is not available. We and others have previously reported that there is a subset of colon cancer tumors, approximately 31%, that have CHFR
Using the highly purified TCGA sample data, we confirmed this prevalence in a larger cohort of primary tumors.
It has been previously shown in gastric cancer that CHFR
methylated cell lines lead to absence of gene expression for that gene and are sensitive to microtubule inhibitors such as paclitaxel and docetaxel,27
though a retrospective study in advanced and recurrent gastric cancer did not find the same association.41
An association between taxane response and CHFR
methylation was also reported in a preclinical study of endometrial cancer where endometrial cell lines with CHFR
methylation were more sensitive to paclitaxel and this sensitivity was reversed with re-expression of CHFR
by demethylation of the gene.42
methylation predicted for sensitivity to taxanes in cervical adenocarcinoma cell lines also reversed with demethylation of the gene.24
Our study is the first report of the impact of CHFR
methylation status on chemosensitivity in CRC. Recent work has suggested a potential prognostic value for the biomarker as well, with Tanaka et al. reporting that CHFR
gene methylation is associated with disease recurrence in locally advanced colon cancer.43
Previous work has demonstrated that ectopic expression of CHFR
-null cells results in taxane resistance.22
These reports are consistent with our findings that re-expression of CHFR
using 5-azacitidine in methylated cell lines does shift the dose-response curve of docetaxel towards resistance. This is particularly meaningful as treatment with a second cytotoxic agent would be expected to show more cytotoxicity, not less. Since CHFR
is a densely methylated gene, we were not able to induce sufficient demethylation in all cell lines with our lower dose of 5-azacitidine as was seen in HCT116. In cell lines where CHFR
was not able to be re-expressed (HCT116) or is already unmethylated and expressed highly at baseline (SW480), there was no change in sensitivity to docetaxel following 5-azatidine treatment. This experiment is limited due to the non-specific demethylating activity of 5-azacitidine but lends support to the hypothesis that CHFR itself may be the cause of increased resistance to taxanes.
Our data also support the use of gemcitabine in MSI-H CRC patients, approximately 15% of the CRC population. In the original phase I study of the agent, one of two responders was a colon cancer patient,44
but the phase II trial of gemcitabine in unselected CRC patients reported minimal activity.45
Our work demonstrates that MSI status predicts sensitivity to gemcitabine. The ability of microsatellite status to predict sensitivity to chemotherapy has been controversial in regards to 5FU-based treatments.9, 11, 46, 47
Data supporting that MSI status may predict for improved prognosis has been more consistent.48
In this paper, we found that MSI cell lines were significantly more sensitive to gemcitabine than MSS cell lines. DLD1 was resistant to all agents tested. Previous studies have also reported that DLD1 is resistant to agents to which other MSI cell lines are susceptible; as DLD1 has a loss of hMSH6,33
it has a less microsatellite unstable phenotype and may be the reason why we and others did not see the same response as other MSI cell lines. The mechanism of increased sensitivity of MSI cell lines to gemcitabine in our experiments is unknown; silencing of the mismatch repair genes themselves are not likely the cause of increased gemcitabine sensitivity, as gemcitabine-related DNA repair is usually repaired through other DNA repair pathways. We did not find any correlation with gemcitabine sensitivity and gemcitabine-transporter or metabolizing protein expression. There has been difficulty elucidating the mechanism for differential sensitivity seen of MSI cell lines to other agents such as irinotecan and 5FU previously.32–34
MSI-H cell lines likely have some other molecular abnormality common between them, for which MSI-H is a surrogate biomarker, perhaps just overall genomic instability. Takahashi et al. reported that there was increased in vitro
sensitivity of a MLH1
-deficient/MSI when compared to MLH1
introduced MSI colon cancer to DNA polymerase inhibitors such as gemcitabine and hydroxyurea.12
In our study, there were some inherent challenges in assessing the contributions of the MSI vs. CHFR-methylated phenotype to drug sensitivity due to the overlap between CHFR methylation and MSI. Our work attempted to tease out these differences using candidate cell lines that express MSI or CHFR exclusively as well as those with both biomarkers, as would be found in a standard patient population. The selected cell lines were very rapid growing cell lines; this rapid growth did make post-treatment effects of treatment difficult to interpret. Still, during active treatment, there was clearly differential activity of each agent and the combination predicted by biomarker presence or absence. Moreover, after active treatment of the mice ceased (at the time of sacrifice of mock group), we noted sustained growth inhibition in the combination treated group compared to docetaxel alone for COLO205 and RKO ().
Our data support the possibility that CHFR methylation and MSI may be predictive biomarkers for docetaxel and/or gemcitabine treatment for subsets of colon cancer patients. For this work, we chose 10 cell lines that are commonly used in preclinical work for colorectal cancer, and which provided a group of cell lines with differential MSI and CHFR methylation status. We feel this was an adequate, albeit limited, sample size to test our hypothesis. The overlap between MSI and CHFR methylation as well as MSI and the hypermethylator phenotype (CIMP) creates great opportunities for combination treatment with a taxane plus gemcitabine regimen for ~15% of CRC, a regimen that would be predicted to have at least additive benefit for these CHFR-methylated, MSI patients. In addition, another 15% of patients with only CHFR methylation or MSI would be predicted to benefit from monotherapy with either a taxane or gemcitabine. A phase II trial to test this hypothesis in a biomarker-driven patient metastatic CRC population has recently been activated. More broadly, oncologic drug testing based on epigenetic biomarkers may be an important piece of the drug development paradigm to appropriately select therapy for cancer patients of all tumor types.