Specific genomic aberrations in primary colorectal cancer are associated with liver metastases

Colorectal cancer (CRC) is the second leading cause of cancer death in the Western world. Every year one million people worldwide will develop CRC[1]. The overall five-year survival is 57% and up to 50% of all patients will eventually develop metastases. Metastases are responsible for the great majority of cancer deaths, mainly metastatic liver disease. Even with surgery and modern chemotherapy most metastases are eventually fatal.

Besides lymphatic and haematogenous dissemination, CRC can spread into the intra-abdominal cavity and cause peritoneal metastases (PM).

Of all patients who die of advanced colorectal cancer, 60-70% show evidence of liver metastasis. Even with the use of targeted drugs, the overall survival in patients with non-resectable CRC liver metastases is only 2-years. In patients with resectable liver metastasis a 5-year survival of 30% can be achieved and up to 20% of the population will still be alive after 10 years. Eligibility for hepatic surgery depends on the possibility that all metastases are resectable and adequate liver preservation can be maintained. Therefore early detection of liver metastases is of utmost priority and will result in more radical surgery and thus long term survival[2].

Similar to liver metastasis, peritoneal metastasis is uniformly seen as a fatal condition. However, in the last decade survival has improved due to aggressive cytoreductive surgery in combination with hyperthermic intraperitoneal chemotherapy (HIPEC). Several predictors of outcome after HIPEC treatment, such as completeness of cytoreduction and number of affected intra-abdominal tumor regions, have been described. Patients who underwent complete cytoreduction in combination with HIPEC showed a 5-year survival of 22-49%. Early detection of peritoneal metastases will result in a better cytoreduction and less affected intra-abdominal regions[3]. In conclusion, early detection of liver and peritoneal metastasis will result in a tailored follow-up program and through this better patient outcome.

Accurate staging of CRC with clinicopathological parameters is important in predicting prognosis and guiding treatment but can currently not predict the site of metastases. Therefore, understanding of the molecular and cellular mechanisms underlying colorectal cancer formation, in particular progression to site of metastases is of utmost importance.

The development of CRC arises and develops through the adenoma-carcinoma sequence. This adenoma-carcinoma sequence is a well defined pathway of histopathological stages, each characterized by distinct mutations in oncogenes and tumor suppressor genes [4, 5]. Two molecular pathways have been well described; the microsatellite instability and the chromosomal instability pathway (reviewed by Jass et al) [4, 6, 7]. The majority of CRC (85%) are chromosomally unstable [5], characterized by allelic losses, chromosomal amplifications and translocations [8], whereas mismatch repair deficiency is the underlying cause of the remainder of CRCs.

Cardoso et al. reviewed multiple studies that have reported on the existence of chromosomal abnormalities and gene expression profiles in CRC [8]. With this approach they described several specific chromosomal loci and corresponding genes which play an important role in colorectal cancer progression. A meta-analysis by Diep et al. on 31 comparative genomic hybridization (CGH) studies, comprising a total 859 CRCs, described chromosomal alterations that occurred early in the establishment of primary CRC, as well as those that are present in the different Dukes' stages and in liver metastases[9].

So far, a number of chromosomal aberrations have been related to liver metastases in CRC [9–13]. Metha et al. studied the relation between the extent of chromosomal instability and the survival of patients with liver metastases. They showed that with a larger chromosomally unstable fraction in the liver metastases, survival for patients was better[14].

Various studies have described prognostic gene expression profiles for CRC patients although these profiles have very few genes in common [15–20]. Gene expression profiles have been described for breast cancer that predict site specific recurrence e.g., bone and lung metastases[21–23]. However, gene expression or genomic profiles in CRC that predict site specific recurrence have not been well studied.

Here, we investigated genome-wide chromosomal aberrations in defined groups of primary colorectal tumors to determine copy number signatures for site specific metastases. The understanding of the molecular and cellular mechanisms underlying colorectal cancer formation, progression to malignancy and site specific metastases are important to perform targeted follow-up and eventually develop targeted therapy in patients with CRC.

One of the challenges in CRC therapy lies in the early detection and treatment of CRC metastases. Elucidation of the molecular and cellular mechanisms of developing metastases will play an important role in future diagnostic and therapeutic interventions. A well established screening method to detect the genetic changes that underlie carcinogenesis is comparative genomic hybridization and was first introduced by Kallionemi and colleagues in 1992 [32]

In the present study we examine genome wide chromosomal aberrations in primary CRC to identify molecular markers predictive for liver metastases. The method we chose to examined genome wide chromosomal aberrations is a published and publically available method to look for recurrent copy number alterations and differential copy number alteration[31]. The KC-SMART method was applicable to our research questions and therefore used. An important principle in the analysis of this 20q amplification is that a threshold value can be (arbitrarily) set, but the logistic regression shows a continuous relationship between the aCGH measurement and the risk of a liver metastasis. Binarizing the predictive variable would cause information loss in this case. We believe that there is more information in the continuous mean log2 value of the 20q probes than just an on/off call.

To the best of our knowledge, this is the first study reporting a classifier predictive for liver metastases in primary CRC based on genomic aberrations. Our data demonstrate that primary colorectal tumors that developed liver metastases are characterized by an amplification of chromosome 20q. This amplification of chromosome 20q occurred significantly more often in the LM group compared to the other groups (PM and M0).

In this study there were significantly more left sided and rectum tumors in the LM group as compared to the PM group. So far, however, there is no evidence that peritoneal metastasis are related to the location of the primary colorectal tumor. These primary lesions are often characterized as mucinous T4 tumors which spread tumor cells into the peritoneal cavity[33]. It might be that left sided and rectum tumors could modify the penetrance of chromosome 20q and as result more frequently develop liver metastases. This could contribute to the fact that left sided and rectal tumors in our study result more frequently into liver metastases whereas right sided tumors more often into peritoneal metastases.

We created a classifier to predict liver metastases in patients with CRC. This LM-PAM classifier was able to identify patients who would develop liver metastases and resulted in an AUC of 0.86. Although the classifier was constructed by employing cross-validation to obtain unbiased error estimates, this result should be confirmed in an independent validation set. Logistic regression analysis on the mean 20q copy number ratio showed for every 0.1 log₂ 20q amplification an odds ratio of 16.2 (90% CI, 2.3-141.2) for developing liver metastases.

Amplification of chromosome 20q has been identified in several tumor types including breast, ovary, bladder, pancreas and stomach [32, 34–38]. In colorectal cancer, chromosome 20q has been related to tumor progression, liver metastases and as an indicator of worse patient survival [10–13, 16, 29, 39–42]. Knösel et al. created a progression model to identify relevant chromosomal imbalances specific for metastases but this model was not created to predict site of metastases in primary CRC [43]. Recently, Nakao et al. showed that specific copy number aberrations were linked to nodal metastases and reported a significant difference in 20q amplification in primary colorectal tumors between patients who had liver metastases at time of surgery and those who had no liver metastases. Unfortunately, they did not propose a prediction model for liver metastases [44, 45]. Diep et al. presented a genetic pathway for CRC progression based on a meta-analysis of 31 CGH studies. They identified specific chromosomal alterations linked to different stages of tumor progression and liver metastases and found that the majority of chromosomal alterations were present in both primary carcinomas and liver metastases. They showed that the number of alterations increases in the transition from primary carcinomas to liver metastases. Furthermore, they showed that the role of chromosome 20q was evident in patients with Dukes D classification [9].

Hence, across all studies, chromosome 20q is of importance in the development of liver metastases. Therefore, more understanding of the candidate genes located on chromosome 20q may guide us to understand the biological mechanisms in the development of liver metastases. So far, several genes located on 20q have been described to play an important role in tumor progression and liver metastases. For example genes such as CAS/CSE1L, NABC1, ZNF217, Aurora2 (BTAK, STK15), LIVIN, PTK6, HD54, EEF1A2, PSMA7, TPX2, AURKA and the ubiquitin-conjugating enzyme E2C (UBE2C) [10–13, 45–51].

These candidate genes located on chromosome 20q should be taken into account when examining new targeted therapeutic regimens for patients with CRC. For example, chromosome 20q amplification in CRC showed in an in vitro study response to Kinesin-5 Inhibitor. This inhibitor plays a role in the mitotic spindle function in the cell. Resistance to Kinesin was dominated by amplification of chromosome 20q. It was suggested that AURKA and the TPX2 gene located on 20q were the genes resistance for the Kinesin-5 Inhibitor[51]. Amplification of 20q could therefore be a potential target for novel antimitotic cancer therapies.

In summary, organ specific CRC metastases localization can be predicted by a LM-PAM classifier on the basis of specific genomic aberrations in the primary colorectal tumor. The validation of the LM-PAM classifier will further potentiate its role as a tool in clinical practice. As a result, patients at risk for developing liver metastases should be frequently screened with modern imaging tools and are most likely to benefit from additional chemotherapy.

We show the possibility for specific CGH profiles to predict CRC metastases target organ with 80% precision. This is the first tool to do this and as such may provide the information to guide individual treatment protocols. In daily clinical practice q-PCR or FISH probes could be used on FFPE tissue for detecting patients at risk based on a 20q amplification. We are now developing an MLPA analysis of chromosome 20q to identify specific regions involved in the development of liver metastasis.

The ability to predict liver metastases based on specific genomic aberrations in the CRC is important for understanding the biology of the tumor, to perform tailored follow-up and eventually develop targeted therapy in patients with CRC. Further research should be done by investigation specific target regions of chromosome 20q.