A rapid, sensitive, reproducible and cost-effective method for mutation profiling of colon cancer and metastatic lymph nodes

Recent evidence suggests that mutation profiling can assist in the prognosis and prediction for colon cancer. KRAS, PIK3CA and BRAF mutations are frequent in tumors of the colon and have been associated with poor prognosis [1‐6]. However, these results remain controversial because other studies have shown that mutations in these genes are not prognostic. A large study, a meta analysis, of KRAS mutations, found that only KRASG12V was a bad prognostic marker; other KRAS mutations were not associated with bad prognosis [2]. Evidence has also demonstrated that KRAS mutations are potential markers for prediction because tumors with KRAS mutations are significantly associated with resistance to EGFR antibody based therapies [7‐11]. Publications have reported the same phenomenon with BRAF and PIK3CA mutations, although these observations are still not well established [12, 13]. The published study suggesting that BRAF mutant tumors were resistant to EGFR therapies was a small study [13]. Predictive value of PIK3CA mutations remains controversial in that other publications have shown that these mutations have no predictive value [14, 15]. These inconsistencies, together with two other factors, have limited the impact of mutation profiling for prognosis and prediction in standard care of colon cancer. A large sample size is required to establish that a gene mutation has a significant impact for patient prediction or prognosis. Another limitation is that until recently conducting such large studies with the standard sequencing technologies was too time consuming and too expensive to be practical for clinical studies.

Moreover, while the high frequency of KRAS, BRAF and PIK3CA mutations in colon cancer is well documented, other potentially important mutations have not been profiled with a large number of clinical samples. Whole genome sequencing of a small number of colon samples demonstrated that somatic cancer mutations consist of a few genes that occur frequently and many more mutations that occur very infrequently in many different genes [16, 17]. Mutations in these infrequently mutated genes could have a similar effect or synergize with mutations in KRAS, PIK3CA, and BRAF.

Given these considerations, our goal was to find a cost-effective and high throughput methodology that would detect frequent and infrequent cancer mutations genes in a large number of samples. Furthermore, it was essential that the methodology would work with degraded DNAs isolated from FFPET.

The mass spectrometric SNP genotyping technology based on the Sequenom platform provided an ideal choice for mutation profiling to address these several criteria. It has been shown to work with small amounts of degraded DNAs (5 ng), and the high multiplexing capacity minimizes the use of irreplaceable clinical samples. In addition, a variety of studies have demonstrated that the sensitivity of mass spectrometric methods exceeds that of traditional Sanger sequencing and is highly concordant with Sanger sequencing, Pyrosequencing, and allele-specific PCR [16, 18, 19]. Furthermore, Sequenom has recently developed the OncoCarta Panel, an oncogene panel that offers a rapid and parallel analysis of 238 simple and complex cancer mutations across 19 genes.

The OncoCarta panel includes assays for most colon cancer mutations in the clinically relevant genes, BRAF (99%), KRAS (98%), and PIK3CA (78%), and in addition contains assays for other cancer mutations in genes that intersect with the same pathways as that of KRAS, BRAF, and PIK3CA, such as AKT1, EGFR, HRAS, NRAS, MET and others. The frequency of KRAS, PIK3CA, and BRAF mutations in the National Surgical Adjuvant Breast and Bowel Project (NSABP) trial C-07 were similar to the frequencies for colon samples listed in the COSMIC data base. This observation provides evidence that the mutation data obtained with the Sequenom platform is accurate. Our results also demonstrate that a majority of colon cancer samples have aberrant PIK(3) - RAS/RAF network; similar results have been seen previously [6].

Mutations in ABL1 and MET, not previously identified in colon cancer, were identified, and 13 other genes were screened and found not to be mutated in hot spot locations. Furthermore, this study identified the most frequent colon cancer mutations from OncoCarta, providing the necessary information to reduce the number of assays from 187 to 24, creating a smaller, more specific and economical panel, requiring less DNA, and thus conserving precious clinical samples.

The Sequenom platform provides a superior technology for the screening of many hot spot mutations in cancer samples. Sanger sequencing would require amplification of at least 60 different fragments per sample, and many reactions would require optimization, thus adding considerable time and expense. Multiplexing and the use of the OncoCarta panel allowed us to skip this time consuming step. Thus, conservatively, Sanger sequencing would be 40 times more expensive, and require at least 2 times more DNA. Other sequencing technologies, which employ differential melting of mutant and wt sequences, such as HRMA, still require that the PCR product be sequenced. This would add significant cost and time to the procedure because 60% of the colon samples contained one or more mutations. In addition the Sequenom platform is more sensitive than Sanger sequencing in that it was able to detect mutations that represented only 5% of the DNA. Pyrosequencing represented a potential alternative to the Sequenom platform, but in our hands assays needed to be optimized, and the lack of multiplexing made the procedure more time consuming and demanded more DNA. The Sequenom methodology also focuses on only those nucleotides that are known to be cancer mutations and thus makes review of the sequence information considerably faster than Sanger. Next-Generation sequencing was cost prohibitive and has not been shown to work with DNAs isolated from FFPET. Thus, the Sequenom platform and the OncoCarta Panel provided the simplest, most rapid, sensitive and cost-effective method for detecting hot spot cancer mutations in degraded DNAs isolated from archival and routinely processed FFPET. The ColoCarta panel provides a more specific panel for colon cancer mutation detection and greatly reduces the amount of DNA needed for mutation profiling.

The frequencies and specific amino acid mutations detected here were similar to the COSMIC database and other publications [6]. The small variation in frequency between our data and other reports may be attributed to differences in the stage of the samples analyzed, the number of samples considered, and the sensitivity of the technology [18]. These observations, combined with the perfect match that we obtained between the expected and the detected mutations in our control cell lines, both fresh and FFPE, and the fact that mutations detected with OncoCarta and ColoCarta were identical, suggest that the technology is reliable and reproducible in DNAs isolated from FFPE samples.

In our study, the majority of tumors (60.3%) had one or more mutations in KRAS, PIK3CA, and BRAF. Mutations in these genes are likely to perturb many different and overlapping signaling pathways, including PI3K/AKT, ERK/MAPK, SAPK/JNK, NFKβ, and others. We were also able to detect other less frequent mutations that are likely to perturb the same pathways and these may cause resistance to EGFR-targeted therapies, as reported for KRAS, PIK3CA and BRAF. For example, AKT1 and NRAS are molecules that are downstream mediators of the EGFR signaling pathway, and mutations in these genes are likely to affect the response to drugs that target EGFR.

Mutations in ABL, AKT1, and MET were detected here but were not listed in COSMIC, probably due to the small number of samples analyzed. The AKT1-E17K mutation was initially identified as a SNP, rs34409589, but in a recent publication it was found to be a somatic mutation and was found in 3 of 51 colon cancers [27]. The frequency of these mutations in this small study (51 samples) was 6% and is much greater than in the C0-7 samples (0.4%). This difference in frequencies may be because the Carpten et al [27] samples were from more advanced stages than those from the C-07 trial. Moreover, they selected large tumors (>100 mg) and containing more than 60% tumor cells. No such selection was done for our study, and samples were from stages II and III exclusively. The significance of ABL1 and AKT1 mutations for patient prediction and prognosis in our study is questionable given that they each were found in only in one sample and represented only 0.4% of the cases.

To our knowledge, this is the first report of MET mutations in the primary colon cancer, but a different MET mutation (N1118Y) was found in a lung metastasis of the large intestine [28]. The MET mutations, R970C and T992I, were detected in 8 out of 239 C-07 colon cancers. These mutations correspond to MET-R988C and MET-T1010I, respectively, in the long form of MET which is the isoform referred to in the COSMIC database [29]. The R970C and T992I mutations are located in the juxtamembrane segment of the protein and were detected in lung carcinoma [30]. These mutations, when introduced into a lung cell line, increased focus formation, formation of colonies in soft agar, cell motility, and migration. These mutations also resulted in constitutive tyrosine phosphorylation on several cellular proteins including paxillin at key tyrosine residues and may account for the increased motility of cells with this mutation. Another critical amino acid in this location is a Ser 985, which, when phosphorylated, has been found to diminish MET signaling [31]. If phosphorylation at Thr residue 992 (1010) reduces signaling, then the R992I mutation would inhibit this negative feedback and may result in constitutive signaling [30].

If MET mutations confer an alternative activated signaling pathway, then these mutations could also confer resistance to anti-EGFR-based therapies or provide a new target for directed therapies. Therapeutic drugs have been developed to specifically target MET, including small molecule kinase inhibitors, anti-MET monoclonal antibodies, and inhibitors of HGF, the MET ligand. Invitro assays have demonstrated that a number of MET targeted therapies were able to prevent MET signaling, decrease cell viability, and limit cell motility and migration in vitro [32]. The small molecule ARQ 197, a kinase inhibitor, has entered phase II clinical trials so may represent a possible therapeutic strategy for some colon tumors.

To our knowledge, this is also one of the most exhaustive analyses of mutation profiling of metastatic lymph nodes and their corresponding primary colon tumors. Our analysis showed that a majority of samples were concordant (89.7%) but in a few samples mutations were detected only in the primary tumor and not in the metastatic lymph node. Also in samples with 2 co-occurring mutations, the ratio of the double mutations varied in primary and lymph node tumors. Discordance in the genetic profile between primary tumors and the metastatic lymph nodes has been observed [33]. Such data may indicate that tumor cell migration selects different cell populations from the one in the primary tumor. However, it is also possible that these mutational differences between the lymph node and the primary tumor are a result of tumor heterogeneity.

Another interesting observation in our study was that BRAF mutations were significantly correlated with poorly differentiated tumors and the prevalence of mucin; similar observations have been reported [25, 26]. These characteristics are both associated with a worse prognosis and are consistent with other reports associating BRAF mutations with a bad prognosis [3]. However, in our study we found that there were 2 metastatic lymph nodes that did not maintain the BRAF mutation present in the corresponding primary tumor, suggesting that BRAF mutations are not essential for metastatic spread to the lymph node in all tumors. Clearly, additional studies would be required to understand these apparent inconsistencies; additional lymph node samples are not currently available but could be the subject of further studies when samples become available [3].

The Sequenom platform provided a superior technology for the screening of 238 common hot spot cancer mutations in 19 genes. The frequent occurrence of KRAS, PIK3CA, and BRAF was confirmed, and mutations not detected before in colon cancer were found in MET and ABL1. Twenty-five assays from the OncoCarta were replexed to form a new panel, termed ColoCarta, which will be used to screen an additional 800 tumors from NSABP clinical trial C0-7 with the purpose of identifying prognostic or predictive markers for stage II and III colon cancer.