Exploratory biomarker analysis for treatment response in KRAS wild type metastatic colorectal cancer patients who received cetuximab plus irinotecan

Colorectal cancer (CRC) has been a significant cause of morbidity and mortality throughout the world [1]. Metastatic CRC (mCRC) is associated with a particularly poor prognosis. Despite progress in cytotoxic chemotherapy during past decade, the five year survival rate for mCRC remains below 10 % [2]. There have been recent and rapid advances in the development of agents targeted against components of receptor tyrosine kinase signaling cascades for use in cancer-therapy. A monoclonal antibody (mAb) against epidermal growth factor receptor (EGFR), referred to as cetuximab, is a product of this movement, and has been implemented to treat selected mCRC patients.

However, only 10–20 % of the selected mCRC patients benefit from anti-EGFR therapy [3‐6], highlighting a distinct need for individualized treatment. Following the discovery that mutations in KRAS are associated with resistance to anti-EGFR treatment, determination of KRAS status is now recommended in mCRC patients before starting anti-EGFR therapies. Despite the application of these selective strategies, less than half of the chosen wild type KRAS patient population benefits from anti-EGFR treatment [7‐9]. More recently, other oncogenic alterations such as mutations in BRAF and PIK3CA were identified as candidates associated with resistance to anti-EGFR therapies in wild type KRAS patients [8, 10‐13]. However, there is still a need to identify and confirm additional biomarkers that can be used to more accurately select wild type mCRC patients that will respond to anti-EGFR therapy.

Protein kinases control many cellular processes including metabolism, transcription, cell cycle progression, cytoskeletal rearrangement, cell movement, apoptosis, and differentiation [14, 15]. Therefore, protein kinases are essential targets for molecular therapy. Indeed, various protein kinase inhibitors have been shown to be effective against cancer cells. Cancers often result from the interconnectivity of complex pathways, some of which are not well understood. For this reason, we surmise that the anti-tumor activity of cetuximab may be affected by various kinase genes involved different pathways.

In order to identify additional selective biomarkers for CI indication, we genotyped wild type KRAS colorectal tumor samples from patients that received CI treatment for mutations in either BRAF or PIK3CA, and profiled comprehensively the expression of 522 kinase genes involved different pathways.

Despite exclusion of 30–40 % of patients with KRAS mutant tumors, cetuximab-based therapy fail in more than half of CRC patients harboring wild type KRAS [4, 9, 17‐19]. Recently, mutations in other downstream effectors of the EGFR pathway, such as BRAF, NRAS, and PIK3CA have also been implicated as possible predictors for anti-EGFR response [12, 18, 20, 21]. Overall, the benefit of analyzing these candidates has been difficult to determine due to low mutation frequencies for individual markers. In an attempt to identify additional biomarkers for the efficacy of cetuximab-based therapy in CRC patients with wild type KRAS, we performed targeted gene expression profiling for 522 kinase genes in addition to genotyping for BRAF and PIK3CA. Our analysis suggested that the overexpression of PSKH1, TLK2 and PHKG2 may be a considered a potential biomarker to predict the efficacy of CI in wild type KRAS CRC. Moreover, the higher expression value of PSKH1 (r = 0.462, p < 0.001) and TLK2 (r = 0.361, p = 0.005) had the significant correlation to prolonged PFS. However, among 522 kinase genes, any gene didn’t have significant prognostic value for overall survival (OS). We also conducted the multivariate analysis (Cox proportional hazard modeling) for PFS with clinical variable and expression nature of kinase genes such as PSKH1, TLK2 and PHKG2. Univariate analysis showed that prolonged PFS was significantly associated with metastatic sites of ≤ 2, and expressed of PSKH1, TLK2 and PHKG2. In multivariate analysis, only expressed of PSKH1 (hazard ratio [HR], 2.219; 95 % confidence interval [CI], 1.085 TO 4.541; p = 0.029) was significantly with prolonged PFS.

We used an nCounter assay to analyze the expression of kinase genes. The nCounter assay can profile 522 human kinase genes in one reaction tube with only 100 ng of total RNA. Previously, we compared the results of an nCounter assay with immunohistochemistry results using three biomarkers (EGFR, HER2 and MET) in a gastric cancer (GC) model [22]. Relative to immunohistochemistry findings, the NanoString-based assay sensitivities and specificities were 85.7 % and 82.8 % for EGFR, 100 % and 97.2 % for HER2, and 100 % and 100 % for MET, respectively. Hence, nCounter assays may be feasible as a relatively reliable option in the clinic, where only small amount of RNA (~100 ng) can be rendered from paraffin-embedded tissues. Based on our previous study, NanoString-based kinase assay has been used as a feasible screening tool for initially identifying patients who have GC with overexpressed drug targets in our institution. This study suggested that nCounter assays could be used efficiently as a platform for implementing personalized medicine of CRC.

PSKH1 is a 424 amino acids long ubiquitously expressed autophosphorylating humane protein serine kinase [23, 24]. PSKH1 has been known to have a useful structural and regulatory role in the maintenance of the Golgi apparatus [25]. TLK2 is a nuclear serine/threonine kinase which is a cell cycle check point regulating chromatin assembly and is inactivated in response to DNA damage [26]. Although the higher expression value for PSKH1 and TLK2 is likely to be related to prolonged PFS as well as antitumor effect for CI in CRC patients in this analysis, the molecular determinants for overexpression for PSKH1 and TLK2 are unknown for the effect of cetuximab. Further functional analysis of these kinase genes is needed to apply these markers in clinical practice.

Assessment of some mutations could improve the objective response rate; objective response rates were 24.4 % in unselected population, 36.3 % in population harboring the wild type KRAS, and 41.2 % in the KRAS, BRAF, NRAS, and PIK3CA exon 20 wild type populations [18]. However, cetuximab still fails to meet expectations in more than half of patients who were selected via genotyping of multiple genes. There are some limitations in our study. A small sample size, heterogeneous patient population, retrospective nature, and limited validated data didn’t allow us to draw any significant conclusions. Especially, our study included only 3 patients with BRAF mutation. Thus, the role of BRAF mutation as predictive marker for cetuximab needs to be validated and studied. Tissue availability was also a potential limitation of the current retrospective biomarker-analysis. It is better to verify the PSKH1, TLK2, and PHKG2 expression identified by nCounter assay in tissues using independent methods such as in situ hybridization, immune histochemistry or RT-qPCR. However, patients analyzed in this study had refractory cancer. Among these patients, it was difficult to get the tumor tissue for biomarker-analysis. Nevertheless, our attempts to whether expression of kinase genes could serve as biomarkers for the efficacy of cetuximab therapy in wild type KRAS CRC patients is worth noting.

Many crucial signaling pathways become dysregulated during cancer initiation and progression. Identifying the pathways involved and quantifying their deregulation is an important step toward understanding carcinogenesis [27‐30]. Because advanced therapies target specific pathways, pathway-level understanding is also a necessity in the development of personalized cancer treatments. Previously, we found that KRAS mutant CRC patients with low RAS signature scores were likely to benefit from treatment with cetuximab, irinotecan plus simvastatin [31]. This finding suggested that differential RAS signaling in KRAS mutant CRC patients is a useful biomarker for choosing the proper treatment regimen. Based on our analysis, CCR3, cdMac, and integrin pathways are significantly associated with the activity of cetuximab as the down-regulation of these pathways correlated with an observed increase in cetuximab efficacy, though these results must be validated. Additional, work will also need to be done to evaluate whether the CCR3, cdMac, and integrin pathways are interconnected with the EGFR signaling that was known as the key pathway for the effect of cetuximab. Many new investigational agents, such as inhibitors of MAP2K, PI3K, mTOR, EGFR, and BRAF will be extremely useful for elucidating these networks [32]. These pathway-based approaches will be used as novel algorithms for precision medicine.

In conclusion, our study suggested that expression nature of kinase genes such as PSKH1, TLK2 and PHKG2 may be informative to predict the efficacy of CI in wild type KRAS CRC. In addition to known EGFR pathway, the CCR3, cdMac, and integrin pathways may also be useful in predicting cetuximab response in these patients. Our findings warrant further independent validation. In the future, multiple-biomarker assessment will be needed in mCRC patients in order to optimize identification of patients that will benefit from cetuximab.

The result of this work demonstrated that CI did not have anti-tumor activity in BRAF mutant CRC. Additionally, expression nature of kinase genes such as PSKH1, TLK2 and PHKG2 may be informative to predict the efficacy of CI in wild type KRAS CRC.