Chromosome 20p11 gains are associated with liver-specific metastasis in patients with colorectal cancer
- Leonie J M Mekenkamp1,2,
- Josien C Haan3,
- Miriam Koopman4,
- M Elisa Vink-Börger1,
- Daniëlle Israeli3,
- Steven Teerenstra5,
- Bauke Ylstra3,
- Gerrit A Meijer3,
- Cornelis J A Punt2,6,
- Iris D Nagtegaal1
- 1Department of Pathology, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
- 2Department of Medical Oncology, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
- 3Department of Pathology, VU University Medical Center, Amsterdam, The Netherlands
- 4Department of Medical Oncology, University Medical Center Utrecht, Utrecht, The Netherlands
- 5Department of Epidemiology, Biostatistics and Health Technology Assessment, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
- 6Department of Medical Oncology, Academic Medical Centre, University of Amsterdam, Amsterdam, The Netherlands
- Correspondence to Dr Iris D Nagtegaal, Department of Pathology, Radboud University Nijmegen Medical Centre, PO Box 9101, 6500 HB Nijmegen, The Netherlands;
Contributors LJMM, JCH, MEVK and DI carried out the experiments. LJMM, JCH, MEVK, MK, DI, ST, BY and IDN analysed and interpreted data. LJMM, JCH, GAM, CJAP and IDN were responsible for the study design and wrote the paper. JCH, MK, MEVK, DI, ST, BY, GAM, CJA and IDN revised the manuscript critically. All authors were involved in editing the paper and gave final approval of the submitted and published versions.
- Revised 6 December 2011
- Accepted 21 December 2011
- Published Online First 20 January 2012
Objective Metastatic colorectal cancer (CRC) cells have a selective preference for certain target organs that cannot be explained by circulatory patterns alone. This study aimed to identify clinicopathological features and chromosomal aberrations of primary tumours associated with organ-specific CRC metastasis.
Design Clinicopathological features were investigated in patients with CRC who had exclusively hepatic (n=182) versus exclusively extrahepatic (n=139) metastases. A total of 139 primary tumours of patients with hepatic (n=85) and extrahepatic metastases (n=54) were screened for chromosomal aberrations by microarray-based comparative genomic hybridisation, and the findings were validated in an independent set of 80 primary tumours. A publicly available database was used to correlate chromosomal aberrations with gene expression. Protein expression was evaluated by immunohistochemistry on tissue microarrays.
Results Patients with hepatic metastases were significantly more often male (71% vs 53% p=0.002), more often had abnormal lactate dehydrogenase activity (37% vs 14% p<0.0001), exhibited primary tumour localisation in the colon (52% vs 40% p=0.03) and had synchronous onset of metastases (70% vs 19% p<0.0001). Primary tumours of patients with hepatic metastases were more commonly T3 tumours (79% vs 63% p=0.006) and less commonly of mucinous histology (5% vs 16% p=0.02). Gain of 20p11 was more often observed in patients with hepatic metastases (p<0.05), which was confirmed in an independent dataset (p<0.05; false discovery rate <0.05). Twelve genes mapping at 20p11 were significantly overexpressed as a consequence of 20p11 copy number gain. C20orf3 showed the strongest correlation between RNA expression and DNA copy number. This was reflected in significantly higher protein expression in patients with hepatic metastases (59%; n=325) than in those with extrahepatic metastases (41%; n=256) (p=0.01).
Conclusion C20orf3 mapping at 20p11 is associated with hepatic-specific metastasis in patients with CRC. This gene is a candidate biomarker for liver metastases and may be of clinical value in early-stage CRC.
- Colorectal cancer
- hepatic metastasis
- extrahepatic metastasis
- genomic aberration
- colorectal metastases
- colorectal cancer genes
- liver metastases
- colorectal pathology
- tumour markers
- colon carcinogenesis
- gastric cancer
- genetic instability
- colonic adenomas
- surgical oncology
- small intestine cancer
- colorectal adenomas
Significance of the study
What is already known about this subject?
Both mechanical factors and (epi)genetic features of tumour cells determine where a metastasis will develop after the arrest of tumour emboli.
Organ-specific metastasis has mainly been investigated in animal models, but also in human primary breast cancer tissues, where gene expression profiling can predict bone and lung metastases. This suggests that ‘homing’-related factors can be detected in the bulk of the primary tumour.
What are the new findings?
Gain of chromosome 20p11 is over-represented in primary colorectal cancers that preferentially metastasise to the liver.
Of the 12 of 28 genes at chromosome 20p11 with a dosage effect on expression, C20orf3 showed the strongest correlation, and protein expression was associated with hepatic-specific metastases.
How might it impact on clinical practice in the foreseeable future?
These findings help to unravel the biology behind organ-specific metastases. The possible role of this hepatic metastasis-associated gene in specific steps of the hepatic metastatic process needs to be functionally validated. Ultimately, this could result in (1) the identification of new prognostic markers that select patients with early-stage colorectal cancer who are most likely to develop liver metastases, and (2) the development of liver-specific anti-metastatic therapies for the future.
Colorectal cancer (CRC) is one of the most commonly diagnosed malignancies, and the majority of these patients die as a result of metastatic disease. The phenomenon of cancer metastasis has been extensively studied and characterised as a complex, multistep process. To produce metastatic outgrowth, tumour cells need to be proficient in all steps of the metastatic process, including invasion, embolisation, survival in the circulation, arrest in a distant capillary bed, and proliferation within the organ parenchyma. Tumour cells acquire these biological properties by accumulating (epi)genetic alterations.1 One hypothesis on the acquisition of a metastatic phenotype is that these modifications are already present in the primary tumour. This is supported by the finding that gene expression signatures of the primary tumour have been shown to predict the occurrence of metastasis in patients with breast cancer.2
Like other types of cancer, CRC shows organ preference for metastasis formation. The liver is the predominant site in ∼80% of patients with CRC. In 40–50% of these patients, extrahepatic organs are also involved in metastatic colonisation.3 Lung metastases develop in 5–15% of the patients, and metastases in the central nervous system, adrenal glands, ovaries, skeleton and skin together account for <10% of all colorectal metastases.4 Metastatic cells prefer to grow in certain organs in a way that cannot be explained by circulatory patterns alone. Mechanical entrapment combined with receptor-specific seed and soil adhesions are currently being discussed as determining factors for cancer cell arrest in target organs. Organ specificity has mainly been investigated in animal models,5 ,6 but gene expression profiling in human breast cancer tissue can predict bone and lung metastases.7 ,8 The development of DNA microarray technology, which allows genome-wide profiling, has provided new insight into the genetic basis of metastasis. However, so far, neither chromosomal aberrations nor gene expression profiling in the primary tumour have been correlated with hepatic versus extrahepatic metastasis in CRC.
CRC in many aspects, including prognosis and survival, is a heterogeneous disease. In the case of unresectable metastatic CRC, patients are treated with cytotoxic regimens (fluoropyrimidines, oxaliplatin, irinotecan) in combination with targeted therapy (vascular endothelial growth factor and epidermal growth factor receptor antibodies). There are conflicting data on the prognostic value of organ-specific metastasis in patients treated with systemic therapy. Several studies have reported the presence of liver metastases as a negative predictor,9–14 while others observed survival benefit in patients with hepatic metastases compared with patients with lung metastases.15–17 So far, no studies have been performed to evaluate the differences in patients with hepatic versus extrahepatic metastases in terms of clinicopathological features and outcome.
This study aimed to identify clinicopathological features, chromosomal aberrations and outcome associated with hepatic versus extrahepatic metastasis in patients with CRC.
Materials and methods
The patients included in this analysis participated in the CAIRO study (CKTO 2002-07, http://ClinicalTrials.gov; NCT00312000) of the Dutch Colorectal Cancer Group.18 In this multicentre phase III trial, 820 patients with metastatic CRC without previous systemic treatment for metastatic disease were randomised between sequential and combination treatment with capecitabine, irinotecan and oxaliplatin. The primary end point of the study was overall survival (OS). The written informed consent required for all patients before study entry also included translational research on tumour tissue. For the present analysis, we selected 550 eligible patients who underwent a resection of the primary tumour, for which formalin-fixed paraffin-embedded (FFPE) material of the primary tumour was available. Patients were divided according to the site of the metastases in exclusively hepatic (n=182) and exclusively extrahepatic (n=139) disease. Patients with a combination of hepatic and extrahepatic metastases (n=221), locally advanced disease (n=7) and for whom the metastatic site was unknown (n=1) were excluded from this analysis.
Clinical and histopathological parameters
The following clinical features were collected for each patient: age, gender, Eastern Cooperative Oncology Group (ECOG) performance status, serum lactate dehydrogenase (LDH), site of the primary tumour, previous adjuvant therapy, number of metastatic sites involved, metachronous (>6 months after initial diagnosis) or synchronous (≤6 months of initial diagnosis) onset of metastases, and regimen used as first-line treatment.
The TNM (tumour, node, metastases) classification was used to describe the extent of cancer spread in terms of invasion depth and lymph node stage.19 Histopathological review was carried out by two independent observers (LJMM, IDN). If the scoring was not unambiguous, the opinion of the pathologist (IDN) was final. Tumours were classified using the WHO guidelines.20 A tumour was considered to be of the mucinous type when at least 50% of the tumour volume consisted of mucin. Primary tumours were graded into well, moderately and poorly differentiated adenocarcinomas based on the part of poorest differentiation in the tumour. The mismatch repair system status was determined by immunohistochemistry and microsatellite instability analysis.21
Data analysis of clinicopathological features and outcome
Clinical and histopathological characteristics of patients with hepatic and extrahepatic metastases were compared using the Wilcoxon signed rank test or χ2 test where appropriate. OS was calculated as the interval from the date of randomisation until death from any cause or until the date of last follow-up. Progression free survival (PFS) for first-line treatment was calculated from the date of randomisation to the first observation of disease progression or death from any cause. OS and PFS curves were estimated using the Kaplan–Meier method and compared using the log-rank test. Patients were considered evaluable for response if they had completed at least three cycles of chemotherapy. Overall response was defined as partial response or complete response. Disease control was defined by stable disease with a duration of more than 4 months or partial response or complete response. Differences in response and disease control rates were analysed by a χ2 model. Multivariate analysis of OS was performed by means of a Cox proportional hazards model, including the following covariates: gender, performance status, serum LDH, site of the primary tumour, number of metastatic sites involved, T stage, N stage, classification and differentiation grade of the primary tumour. All statistical tests were two-sided, and p values of <0.05 were considered significant.
Sample selection for DNA copy number profiling
To asses DNA copy number profiles we used an array comparative genomic hybridisation (CGH) dataset of 222 primary colorectal tumours from patients with metastatic CRC who were treated within the CAIRO study (Haan et al, unpublished data). We selected the 85 and 54 patients with exclusively hepatic and exclusively extrahepatic metastases, respectively. The accuracy of the observed differences in DNA copy number profile was validated in an independent validation set of 45 and 35 primary tumours with hepatic and extrahepatic metastases, respectively. These tumours were derived from an array CGH dataset of 134 patients with metastatic CRC (Haan et al, unpublished data), who participated in the CAIRO2 study (CKTO 2005-02, http://ClinicalTrials.gov; NCT00208546).22 In this multicentre phase III trial, patients were randomly assigned to first-line treatment with capecitabine, oxaliplatin and bevacizumab, or the same schedule with the addition of cetuximab. Since the CAIRO2 study had a negative outcome, possibly because of a negative interaction between the study drugs, the array CGH profiles were determined only in patients receiving capecitabine, oxaliplatin and bevacizumab. The selection criteria for the patients used for both the training and validation set are described in detail elsewhere (Haan et al, unpublished data). Briefly, primary tumours were selected from patients who underwent a resection of the primary tumour and for whom FFPE material of the primary tumour was available. Since we used normal as well as tumour DNA from the same patient, FFPE material for both needed to be available. Stringent criteria were used to select patients based on tumour cell percentage (at least 70%), clinical variables (matched to the stratification variables in the original studies) and DNA quality (specific activity at least 16 pmol/μg).
The clinicopathological features of patients in both the learning (online supplemental table 1) and validation (online supplemental table 2) array CGH datasets are representative of the larger dataset used for clinicopathological comparisons.
DNA was isolated using an extensively validated protocol as previously described.23 For each tumour, an area was marked containing at least 70% tumour cells. Of the FFPE blocks, two to six 10 μm sections were cut, deparaffinised and microdissected. DNA was extracted using a column-based method (QIAmp microkit; Qiagen, Hilden, Germany). Matched normal mucosal DNA was used from all of these samples as a reference and was obtained from the resection margins or at least 1 cm distance from the tumour. Normality was confirmed in silico for each reference by comparing the array signals of the normal reference of another patient by across array.24 All DNA concentrations were measured on a Nanodrop ND-1000 spectrophotometer (Thermo Scientific, Wilmington, Delaware, USA).
Labelling and hybridisation was carried out as previously described.24 Briefly, 500 ng genomic DNA and matched normal DNA were labelled using a CGH labelling kit for oligo arrays (Enzo Life Sciences, Farmingdale, New York, USA) with cyanine 3-dUTP and cyanine 5-dUTP nucleotide mixture, respectively. Labelled DNA quality was tested by using the Nanodrop ND-1000 spectrophotometer (Thermo Scientific) to measure specific activity. Samples with specific activity <16 pmol/μg were considered insufficient, and these cases were replaced by comparable samples. Hybridisations were performed on the Agilent 4×180K oligonucleotide arrays (AMADID number 022522; Agilent Technologies, Palo Alto, California, USA). These arrays contain 180 880 in-situ synthesised 60-mer oligonucleotides, representing 169793 unique chromosomal locations distributed across the genome at 17 kb intervals and is enriched with 4548 additional unique oligonucleotides, located at 238 of the Cancer Census genes. The exact array design can be found in the Gene Expression Omnibus (GEO) platform GPL8687 http://www.ncbi.nlm.nih.gov/geo/. Images of the arrays were acquired using a microarray scanner G2505C (Agilent Technologies).
Preprocessing array CGH data and data analysis
Array image analysis was performed using Features Extraction software (V.10.5.1.1; Agilent Technologies). Local background was subtracted from the signal median intensities of both tumour and normal DNA. The log2 tumour to normal signal ratio was calculated and normalised against the median value of the log2 ratios of all the oligonucleotides mapped to the March 2006 human reference sequence (NCBI36/hg18) on chromosome 1-22 and X. The median absolute deviation value was calculated as a quality measure of the final array CGH data. Samples from the learning and validation set with a median absolute deviation value above 0.4 were excluded.
Wave patterns occurring in the genomic profiles were smoothed with the R package NoWaves.25 The R package CGHcall26 was used to preprocess and normalise the data. Cellularities were set to the cellularity determined by a pathologist, and median normalisation was performed. The R package DNAcopy27 segmented the log2 ratios and they were renormalised by mode normalisation. Chromosomal copy number losses and gains were identified by CGHcall, calling probabilities of 0.5 or more. The accuracy of the normalisation, segmentation and calling was verified by visual inspection. To reduce the dimension of the array CGH dataset without loss of information, regions were defined as previously described.28
For supervised analysis, a two-sample Wilcoxon test using 10.000 permutations was performed to calculate the significance of DNA copy number differences between patients with hepatic versus extrahepatic metastases.29 Two separate tests were performed to compare frequencies of gains and losses between the two groups. To account for multiple testing, a permutation-based false discovery rate (FDR) was applied to the p values.
Gene dosage effects
To identify genes on the relevant chromosomal regions that show a gene dosage effect (ie, mRNA expression levels are correlated with DNA copy number status) in CRC, a dataset was used from The Cancer Genome Atlas (TCGA) Data Portal (http://tcga-data.nci.nih.gov/tcga/) with combined mRNA and DNA copy number data. DNA copy numbers were available as segmented data, of which primary tumours with log2 ratios lower than −0.5 were designated as ‘loss’, and values higher than 0.5 were designated as ‘gain’.
A two-sample Wilcoxon test was used to compare log2 mRNA expression ratios between patients with and without DNA copy number gain. To account for multiple testing, a permutation-based FDR was applied to the p values.
Immunohistochemistry on tissue microarrays (TMAs)
From the FFPE primary tumour tissues available from patients with CRC in the CAIRO and CAIRO2 studies with hepatic and extrahepatic metastases, a 2 mm punch was taken to assemble TMAs. From each TMA, a 4 μm section was mounted on glass, deparaffinised and rehydrated. Endogenous peroxidase activity was blocked with 3% H2O2 for 30 min. After microwave antigen retrieval using 10 mM sodium citrate buffer (pH 6.0), slides were incubated with rabbit antibody to human polyclonal C20Orf3 (1:800 dilution; Sigma-Aldrich, St Louis, Missouri, USA) overnight at 4°C. Subsequently, sections were incubated with Powervision Poly-HRP anti-Ms/Rb/Ra IgG (Immunologic, Duiven, The Netherlands) and developed using PowerDAB (Immunologic). The slides were counterstained with haematoxylin and evaluated by two independent observers (LJMM, MEVB). In the case of discrepancy, a definite result was generated based on the expertise of a third investigator (IDN). All three investigators were blinded to the clinical and array CGH data. Stromal tissue served as a positive internal control. C20orf3 protein expression was negative if none of the tumour cells showed cytoplasmic/nuclear membrane staining, and positive if at least one tumour cell showed staining. The staining intensity was graded as negative (no staining), weak (light brown), moderate (brown) or strong (dark brown).
A χ2 model was used to compare protein expression between patients with hepatic versus extrahepatic metastases. Subsequently, protein expression was also correlated with the ordinal array CGH data. All statistical tests were two-sided, and differences were considered significant when p values were below 0.05.
Clinical and histopathological features associated with organ-specific metastasis
Compared with patients with extrahepatic metastases (n=139), patients with hepatic metastases (n=182) were significantly more often male (p=0.002), more often had serum LDH activity above the upper limit of normal (p<0.0001), exhibited primary tumour localisation in the colon (p=0.03), showed synchronous onset of metastases (p<0.0001), and had less commonly received previous adjuvant chemotherapy (p<0.0001). There were no significant differences between patients with hepatic versus extrahepatic metastases in age (median 65 vs 63 years, p=0.39, respectively), performance status at randomisation (both groups 97% WHO 0-1) and treatment arm (both groups 47% sequential arm) (table 1).
Primary tumours of patients with hepatic metastases more often had T3 stage (p=0.006), and mucinous histology (p=0.02) was less often observed in patients with hepatic compared with extrahepatic metastases. No significant differences between patients with hepatic versus extrahepatic metastases were observed in diameter (both median 45 mm), lymph node status, differentiation grade, and mismatch repair (MMR) status (deficient MMR in 3% vs 4%, p=0.44, respectively) (table 1).
Prognostic value of organ-specific metastases
No significant difference in median OS was observed for patients with hepatic versus extrahepatic metastases in univariate analysis (20.3 vs 19.9 months, respectively; p=0.55). The absence of a significant impact on prognosis was confirmed for hepatic versus extrahepatic metastases in multivariate analysis (HR 0.90; 95% CI 0.66 to 1.23; p=0.52). The median PFS in first-line treatment was not significantly different between patients with hepatic versus extrahepatic metastases (8.2 vs 7.0 months, respectively; p=0.46). A total of 291 patients were assessable for response in first-line treatment: 158 in the hepatic group and 133 in the extrahepatic group. The overall response rate (complete plus partial tumour response) in first-line treatment was significantly better in patients with hepatic metastases than patients with extrahepatic metastases (43% vs 27%, respectively; p=0.007). The disease control rate (complete plus partial tumour response plus stable disease) in first-line treatment was not significantly different between patients with hepatic and extrahepatic metastases (85% vs 86%, respectively; p=0.74).
Identification of a DNA copy number profile associated with organ-specific metastasis
The frequency plots of DNA copy number aberrations throughout the genome in patients with hepatic (n=85) and extrahepatic (n=54) metastases were very similar (figure 1). Despite the high concordance level between the two groups, small differences in DNA copy number profile were observed. Patients with hepatic metastases had significantly more gains at 20p11 than patients with extrahepatic metastases (p<0.05; FDR=0.88) (online supplemental table 3). Loss at 5q12 was significantly less commonly observed in patients with hepatic versus extrahepatic metastases (p<0.05; FDR=0.54) (online supplemental table 4).
To validate these differences in DNA copy number profiles, an additional independent validation set of 45 patients with hepatic metastases and 35 patients with extrahepatic metastases was selected. Significantly more gains at 20p11 in patients with hepatic versus extrahepatic metastases were confirmed (FDR<0.05) (figure 1, online supplemental table 5). Differences in copy number aberrations at 5q12 could not be validated.
Identification of differentially expressed genes at 20p11
To determine the most relevant genes at 20p11 with a potential role in hepatic-specific CRC metastasis, we used a publicly available dataset of 141 patients with CRC. For these patients, both DNA copy number and gene expression profiling of the primary tumour were available. Putative genes with a dosage effect were identified by comparing tumours with 20p11 gain with tumours without 20p11 gain. This approach revealed 12 out of 28 genes with expression levels that were significantly influenced by the occurrence of 20p11 gain, namely XRN2, NXT1, GZF1, NAPB, CSTL1, CST3, CST5, C20orf3, ACSS1, ENTPD6, PYGB, ABHD12 (online supplemental table 6). Of these 12 differentially expressed genes, C20orf3 showed the strongest correlation between copy number status and RNA expression (FDR<0.0001) (figure 2).
Confirmation of differential expression by immunohistochemistry
Using the TMAs of both CAIRO and CAIRO2 patients, C20orf3 expression could be determined in 581 patients, of which 325 had hepatic metastases and 283 had extrahepatic metastases. Examples of C20orf3 expression in TMA cores of primary adenocarcinomas are shown in figure 3. In situ confirmation of C20orf3 protein expression yielded a higher percentage of primary tumours with the presence of C20orf3 staining in patients with hepatic versus extrahepatic metastases (59% vs 41%, respectively; p=0.01) (table 2). In 199 out of the 219 patients with copy number data, copy number status was correlated with C20orf3 expression. A significant positive correlation of C20orf3 protein expression with the ordinal array CGH ratios was shown as well (p<0.0001) (table 3). Validation of other genes was hampered by the unavailability of adequate antibodies for immunohistochemistry on FFPE tissues.
In this study we characterised specific clinicopathological features and genomic aberrations of the primary tumour in patients with CRC with hepatic versus extrahepatic metastases.
In patients with hepatic metastases, the primary tumour was more often in the colon. The venous drainage of the colon is via the portal system, therefore the liver has always been regarded as the first site of haematogenous spread. The increased incidence of extrahepatic metastases in rectosigmoid and rectal carcinoma can be attributed to the direct haematogenous spread into the systemic circulation via the inferior and middle rectal veins. However, over one-third of patients with colon carcinomas develop extrahepatic metastases. This supports a substantial role for features that are inherent to the cancer cell and the micro-environment. In the present study, liver-specific metastasis was more often observed in male than female patients. This observation was also reported in a smaller cohort of patients,30 but a clear explanation is lacking. The onset of metastases is another clinical feature associated with site-specific metastasis. The majority of patients with stage IV disease present with hepatic metastases at diagnosis, and only one-sixth of the patients had extrahepatic metastases. These differences may partly be related to the diagnostic procedures, since extrahepatic metastases may be more difficult to detect. Indeed, progress has been made in diagnostic techniques, which may explain the rising incidence of synchronous pulmonary metastases from CRC.31 The third clinical feature associated with site-specific metastasis is serum LDH. Our results confirm that this variable is not only a surrogate estimate of tumour burden, but also a serological factor for hepatic metastases.32
Pathological examination of CRC resection specimens identified T stage and tumour type, which are correlated with organ-specific metastases. T1 and T2 tumours do not often metastasise, but, if distant spread occurs, they are more likely to stop outside the liver. These tumours are located in the mucosa or submucosa, which have the greatest density of lymphatic vessels.33 We hypothesise that T1 and T2 tumours metastasise via these lymphatic vessels, thereby escaping entrapment in the liver and producing outgrowth in extrahepatic organs. Mucinous adenocarcinomas are a less common histological subset of CRC, also appearing to metastasise more often in extrahepatic organs. These tumours probably possess specific traits that stimulate invasion in extrahepatic organs, but the underlying mechanism is unknown.
Several prognostic factors have been investigated in metastatic CRC, but the influence of the metastatic site as an additional predictor of outcome is highly controversial. Despite differences in clinicopathological features between patients with hepatic and extrahepatic metastases, the median OS and PFS were not significantly different between the groups. Even after correction for multiple prognostic factors, established in the same study population in a previous study,34 metastatic site is not an independent prognostic factor for survival in CRC. However, we observed a higher overall response rate to first-line chemotherapy in patients with hepatic metastases than in those with extrahepatic metastases. First, there may be a bias in response assessments, since liver metastasis may be easier to assess than extrahepatic metastases. Second, significantly more patients with extrahepatic metastases were treated with previous adjuvant chemotherapy, while patients with hepatic metastases were not. This could in theory have resulted in (partial) resistance to chemotherapy in the former group. However, the differences in overall response rates between patients with hepatic and extrahepatic metastases did not translate into differences in survival. This is in line with a previous study by our group34 in which we established that the application of previous adjuvant chemotherapy was not of prognostic value in the same patient cohort.
Genomic aberrations responsible for CRC metastasis to the liver remain speculative and poorly understood. Most of the existing data were obtained using experimentally derived mouse tumour models. In the present study, we aimed to investigate genomic aberrations that drive CRC cells to the liver directly from human samples. The comparison of primary tumours derived from patients with hepatic and extrahepatic metastases identified more gains at 20p11 in patients with hepatic metastases. In contrast with 20q, the correlation of 20p11 with liver metastasis has not previously been described. Gain of 20q is a common genomic aberration in CRC and an indicator of poor prognosis35 and metastatic potential.36 Chromosome 20q gains are more often observed in primary tumours that metastasise to the liver than in tumours that metastasise to the peritoneum and tumours without distant metastases.30 The design of our study and our patient selection are different, as we exclusively used tumours that possess metastatic potential. Therefore differences in gain of 20p11 are more likely to be attributable to organ-specific metastasis. Next, our array contained 180880 nucleotides, and an adequate number of samples was analysed, therefore we expect increased power in our analysis. However, our results only suggest a correlation between 20p11 gain in the primary tumour and liver metastases. We should keep in mind that primary tumours are highly heterogeneous in terms of both their cell populations and their ability to metastasise. Therefore it may be that the genes responsible for organ tropism are not detected in the bulk of the primary tumour. A next step could be to search directly for genes involved in organ-specific metastases by profiling metastatic samples from different secondary sites in relation to their primary tumour.
The number of genes at 20p11 is too high to really disclose those that play a role in organ-specific metastases. Not all genes mapping at gained regions are recurrently overexpressed; therefore we compared gene expression between tumours with and without 20p11 gain. Of the 12 of 28 genes at 20p11 with a dosage effect on expression, C20orf3 showed the strongest correlation, and protein expression was associated with hepatic-specific metastases. C20orf3 is a member of the lactonohydrolase super family, and the potential involvement of this protein in enzymatic processes is suggested.37 Protein expression has mainly been demonstrated in the liver, but no relation to hepatic-specific metastases has yet been reported. In vitro assays should provide evidence for a causal role for this gene in metastasis formation. In addition, confirmation of their relevance to liver-specific metastasis can be made by showing functional evidence of the pro-liver-metastatic effect in a xenograft model. It is interesting that C20orf3 maps on the opposite allele at a distance of a few kilobases from the human CMAP gene, which is correlated with liver metastases.38 However, in the publicly available dataset we used, CMAP was not overexpressed in 20p11 gained primary tumours. Expression of XRN2, NXT1, GZF1, NAPB, CSTL1, CST3, CST5, ACSS1, ENTPD6, PYG and ABHD12 was also increased in primary tumours with 20p11 gain, but we could not analyse the correlation of their protein expression with organ-specific metastases. However, if 20p11 gain influences organ-specific metastases, this could well be caused by altered expression of multiple genes rather than a single gene.
Organ-specific metastasis is a highly complex process, and the capacity to disseminate also depends on additional features other than chromosomal aberrations at chromosome 20p11. First, smaller alterations including somatic mutations are probably important, and novel technologies such as next-generation sequencing will play a pivotal role. Germline variants detectable by genome-wide association studies may also contribute to organ-specific metastasis, but have not yet been identified. Second, the barriers to infiltrating an organ depend on the architecture and specific features of the microenvironment. Endothelial adhesive interactions and certain aspects of the vasculature have been proposed to contribute to dissemination in specific organs.39 Some studies suggest that the secretion of cytokines results in a pro-metastatic microenvironment,40 ,41 but these data need further characterisation in cancer models.
In conclusion, an array CGH profile including the protein-encoding gene C20orf3 was over-represented in primary CRCs that preferentially metastasise to the liver. Although selected from a large clinical trial, it should be realised that our results are derived from a retrospective analysis. Therefore selection bias cannot be excluded, and prospective studies on this topic are warranted. Furthermore, the possible role of this liver metastasis-associated gene in specific steps of the hepatic metastatic process needs to be functionally validated. This could result in the development of (1) new prognostic markers that could help in identifying patients who are most likely to develop liver metastases and (2) liver-specific anti-metastatic therapies for the future.
Funding Dutch Colorectal Cancer Group. This funder had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests None.
Ethics approval The two randomised clinical trials were approved by the Committee on Human-Related Research Arnhem—Nijmegen and by the local institutional review boards. The written informed consent required for all patients before study entry also included translational research on tumour tissue.
Provenance and peer review Not commissioned; externally peer reviewed.