Mortality from colorectal cancer (CRC), the fourth most frequent cause of cancer deaths, is mainly due to metastatic liver disease. Much is known about the adenoma-carcinoma progression of CRC 123 and sporadic CRC is recognised as a heterogeneous and complex disease involving many genes and pathways 45. There has been intensive analysis of the prognostic value of molecular markers for CRC in risk assessment and disease management 67891011. Despite intense study of the metastatic process many aspects of its molecular genetic basis remain unclear. Improved understanding of the molecular events underlying metastasis is crucial for the development of new methods for early detection and treatment of colorectal cancer.

Traditionally, loss of heterozygosity (LOH) analyses were used to map regions harbouring tumour suppressor genes; this method exploits Knudson's two hit hypothesis of tumorigenesis 12 We reasoned that LOH analyses could be used to map chromosomal regions specifically disrupted in metastases, and might therefore highlight the presence of a gene(s) involved in metastasis.

Chromosome 8p is frequently lost in CRC, many studies implicate loss in later stage disease and metastases 131415, and several 8p genes have been implicated in metastasis 16171819. However, to date no strong candidate CRC metastasis suppressor has been identified showing loss of expression and/or function in a significant proportion of tumours, as compared to the frequent mutation and/or silencing of genes involved in adenoma-carcinoma progression 20. We therefore concentrated our analysis on 8p, identified a region of metastasis-specific loss, and screened genes within this region for changes at the DNA, mRNA and/or protein level associated with metastasis.

Disruption of 8p is common in many cancers and could simply indicate the relative instability of this region such that disruption is a consequence of disease rather than playing a causative role in tumour progression. In CRC 8p loss has been implicated in later stage disease and metastasis 131415 and this study again highlights 8p identifying a 2 Mb region of metastasis-specific loss at 8p21-22. In addition, there is compelling evidence for the role of several 8p genes in carcinogenesis, including a number studied here, such as the prostate tumour suppressor NKX3.1, TRAILR DR5, DBC2 and STC1. This tends to argue against 8p disruption being primarily a consequence of tumourigenesis.

In this study the most promising candidate at the mRNA level was ADAMDEC1. We observed decreased mRNA expression in CT and LM compared to CN (Figure 4b). No detectable protein was expressed in 13 LM, and in 3/13 matched CT samples ADAMDEC1 was detected which adds weight to the possibility that loss of expression is progressive through tumourigenesis and may play a role in metastasis. (Figure 5). ADAM family members are involved in cancer progression49 adding further support to the possibility that this unique family member also has a role in tumourigenesis.

We also identified a germline mutation of DR5 which may provide additional insights into the function of this TRAILR. C790T truncates DR5 (Figure 3b) resulting in a protein devoid of the death domain and a potential functional similarity to DcR2. Several studies have also identified truncating mutations of DR5 located prior to the death domain 52535455. One (another germline mutation) showed loss of the growth suppressive function of wild-type DR5 in HNSCC (head and neck squamous cell carcinoma), ovarian and CRC cell lines 53. Despite the diagnosis of synchronous liver metastases the C790T patient did not develop extra-hepatic metastases, responded well to hepatic chemotherapy (5-flurouracil (5-FU)) and SIRT (selective internal radiation) and had > 5-year survival post LM diagnosis. Inducible loss of DR5 protein expression promotes the growth of colon tumours in mice and confers resistance to 5-FU, without causing resistance to TRAIL-induced apoptosis 56. This is in contrast to the clinical observations for the C790T patient (slow tumour growth and p53-responsiveness to 5FU and SIRT).

The heterogeneous or "patchy" C790T DR5 immunostaining and lack of a wild-type DR5 in C790T LM by western analysis suggests complete loss of the wild-type DR5 in C790T LM. The clinical data indicate that the C790T DR5 may have a dominant negative effect, similar to DcR2, including retention of the ability to respond to p53-dependent therapy, and prevention of controlled proliferation, in contrast to null DR5 experimental models 56.

In light of the wealth of evidence implicating 8p in CRC progression it is perhaps surprising that no strong candidate tumour and/or metastatic suppressor has been identified. Other possibilities are that the 'candidate(s)' could be a microRNA or other non-coding RNA or that haploinsufficiency (rather than gene 'knock-out') is sufficient for tumourigenesis. Alternatively, a single gene/RNA may not be the main 'effector' but rather it could be a combinatorial effect whereby a number of genes are involved and perturbation of them all, or a subset thereof, results in tumour progression.

The 2 Mb region in this study certainly appears to be a hot-spot for genes involved in carcinogenesis, and contains 3 gene clusters, TRAILR, ADAM and NKX3.1/NKX2.6, each encoding members with a role in tumour progression. There is evidence for the clustering of co-expressed genes in eukaryotes 57, as well as increasing recognition that dynamic chromosomal architecture and genomic repositioning play an important role in gene regulation 58. An indication that clustering is of functional importance at 8p21-22 is suggested by the observation of co-regulation at the mRNA transcriptional level and co-methylation patterns for the TRAILR pairs DcR1 and DcR2 and DR4 and DR5 in neuroblastoma cell lines 59. We performed Pearsons' correlation analysis on the gene expression data for the 13 candidate genes in both CT and LM. This revealed potential transcriptional relationships between a number of genes including DR4 and DR5 (Additional file 6) which adds support to the possibility that the clustering of 'tumour' genes within this region is of functional significance.

In addition, this possibility is further supported by recent evidence demonstrating that the nuclear protein SATB1 acts as a 'genomic organiser' involved in the epigenetic remodelling of chromatin to facilitate upregulation of metastasis-associated genes and down-regulation of tumour suppressors 60. Although the genes investigated by Han et al were not clustered, 8p21-22 may be a key candidate target of such regulation, the clustering providing a further mechanism for co-ordinated control.

Metastasis might be viewed as a complex disorder in which genetic and environmental factors interact, subtle modulations of cellular activity being required to facilitate survival. We propose that 8p21-22 may not contain a CRC metasatasis suppressor(s) and that the clustering of a large number of genes in one region under co-ordinated control bears closer resemblance to a complex disease, whereby the overall combined profile of multiple genes contributes to the phenotype.

This study identified a metastatic susceptibility locus within a 2 Mb region of 8p21-22, which appears to be a "hot-spot" for genes with a role in carcinoma development and revealed ADAMDEC1 as a potential tumour suppressor. We suggest that the rich nature of this region for genes with a role in tumour development is of pathological significance such that the genes may form a cluster disruption of which favours CRC tumour progression. The possibility of relationships between the genes is supported by the presence of several gene clusters, our observation of potential transcriptional associations, and van Noesels et al observation of transcriptional co-regulation and co-methylation of the DR4/DR5 and DcR1/DcR2 TRAILR pairs 59. As understanding of the role of genomic architecture increases it may become essential to consider neighbouring genes (in genomic and/or nuclear space) to fully understand both gene function and carcinoma progression.

CRC: colorectal cancer; CN: colon normal; CT: colon primary tumour; LM: liver metastasis; LOH: loss of heterozygosity; FFPE: formalin-fixed paraffin-embedded; IHC: immunohistochemistry; MSL: metastasis specific loss.

The authors declare that they have no competing interests.

DPM–C carried out the semi-quantitative real-time PCR, contributed to the mutation analyses and study design, and drafted the manuscript. KAH participated in the IHC, study design and manuscript preparation. AW and HjS performed the IHC. TW carried out the mutation analyses. RO'C was involved in the IHC and critical reading of the manuscript. DAH and RAL performed the statistical analyses. JAR participated in IHC, study design and manuscript preparation. RSS provided clinical expertise, clinical samples and critical reading of the manuscript. SR conceived the study, participated in its design, performed LOH analyses, and helped with manuscript preparation. All authors read and approved the final manuscript.