Background
Oxaliplatin, in combination with 5-Fluorouracil (5-FU, capecitabine) and additional biological agents, presents one of two chemotherapy options available for the treatment of advanced colorectal cancer (CRC). Similar to other platinum compounds, oxaliplatin exerts its cytotoxic activity by causing damage to cellular DNA in the form of helix distorting DNA-platinum adducts, such as intra- and interstrand DNA and DNA-protein crosslinks [
1,
2]. Several DNA repair systems are involved in the removal of this damage, including nucleotide excision repair (NER) [
3]. NER is initiated by recruitment of several proteins to the site of damage. ERCC1 and ERCC4 (also known as XPF) form a heterodimer with endonuclease activity, which is recruited at 5′ to the DNA lesion. Following incision by the complex, another endonuclease (ERCC5, also known as XPG) cleaves at 3′ to the lesion, allowing removal of the damage nucleotide(s). The missing fragment is replaced and ligated [
4]. Due to a central role in NER, as well as interstand crosslink repair [
5], the ERCC1-ERCC4 heterodimer has been widely studied in relation to platinum resistance.
In a landmark study, Shirota and colleagues found a link between low ERCC1 mRNA expression levels and longer survival in a stage IV CRC oxaliplatin-treated patient cohort [
6]. Numerous other studies have attempted to link ERCC1 protein levels to platinum sensitivity [
7‐
9], however these studies have relied on the use of a particular monoclonal antibody, which has recently been found to bind an unrelated protein, raising questions towards the validity of these results [
10,
11]. Interestingly, studies of gene copy number alterations involving the
ERCC1 locus at 19q13 and
ERCC4 locus at 16p13.12 are very limited in number and have not been performed in CRC [
12‐
14].
Predictive biomarkers can encompass both prognostic and predictive components, a phenomenon which may hamper the detection of a beneficial effect from treatment, unless the prognostic element has been investigated and mapped [
15]. With the aim of identifying a predictive biomarker profile for oxaliplatin sensitivity, we have constructed two novel FISH probes directed at
ERCC1 and
ERCC4 genes, as well as relevant reference probes. These were subsequently tested these in a CRC cell line metaphase panel to identify potential aberrations. Both probe combinations were subsequently tested in a chemonaive stage III CRC patient cohort to determine the presence, frequency and prognostic impact of
ERCC1/
ERCC4 gene aberrations. Based on the collected FISH data, scoring guidelines were established for future use.
Discussion
A single
ERCC1-19q13 copy number aberration was observed in the CRC cell line panel. In HT29, a total of four
ERCC1-19q13 signals were detected in this near-triploid cell line. The finding is somewhat similar to the two different karyotypic descriptions available in the NCBI and NCI’s SKY/M-FISH & CGH database. Both descriptions list HT29 as harboring a normal copy of chromosome 19 and differ with regards to whether the cell line additionally harbors either a 19q isochromosome and a derivative fusion chromosome with material from 17q fused at 19qter [
20], or a derivative 19q isochromosome with a duplication of 19q13.1-13.4 and an unbalanced translocation of 19q12-qter to chromosome 17 [
21]. In the current study it appears that the additional signals from
ERCC1-19q13 may be attributed to the formation of a 19q isochromosome fused with an additional (at least partial) duplicated 19q chromosome fragment. Taken together, these different karyotypes indicate that this region appears to be unstable in HT29. No
ERCC4 aberrations were detected.
Analysis of the 152 tumor specimens with
ERCC1-19q13/CEN-2 revealed the presence of both gene deletions and gene gains. To determine whether the two tumor specimens with an ‘
ERCC1-19q13 Deletion’ status were correctly classified, mean gene and reference signals were compared to those previously acquired from unaffected colon mucosa [
16,
17]. In the first sample, CEN-2 signal counts were in the diploid range, whereas gene signals were in the haploid range, indicative of gene deletion. In the second sample,
ERCC1-19q13 counts were in the diploid range, while those of CEN-2 were in the triploid range, suggesting that the low ratio observed for this specimen may be attributed to either loss of
ERCC1-19q13 in a triploid tumor, or chromosome 2 aneusomy. Taken together, these findings suggest that loss of
ERCC1-19q13 occurs infrequently in stage III CRC.
Copy number alterations of the
ERCC1 locus are not widely reported. In ovarian cancer,
ERCC1 appears not to frequently undergo copy number aberrations [
12], whereas it is a more common phenomenon in glioma [
13] (note: blotting-based method were applied in both studies). In non-small cell lung cancer,
ERCC1 gene copy number increases were detected by FISH in 25.5% of samples a finding comparable to the 27.0% reported in the present study, although the FISH probe design and scoring differed substantially [
14]. In this study, Vanhecke and colleagues applied an
ERCC1 gene probe in combination with a reference probe directed at 19p13 and followed
EGFR consensus scoring guidelines to classify samples as either normal, high polysomy or gene amplified [
14]. It should be noted that the use of a reference probe directed at 19p does not allow the detection of
ERCC1 gene copy increases which occur independently of the rest of 19q, i.e. gene amplification-driven copy number increases, which presents a flaw in the design of their probe.
Gain of 19q has previously been reported in colorectal cancer [
22,
23], indicating that a reference probe located on 19q would be ideal in differentiating between arm-level 19q gains and those involving a smaller chromosomal fragment, such as an amplicon. In the present study, the use of an
ERCC1-19q13/CEN-2 ratio cut-off of 1.5 in combination with CEN-2, allows detection of gene copy number increases of 50% or more relative to tumoral ploidy levels. Therefore, the assay does not distinguish between gains due to large chromosomal events, such as chromosome 19 polysomy or 19q isochromosome formation, and those due to gene amplification, which involve an amplicon. To determine whether focal amplification occurs in CRC, GISTIC analysis (Genomic Identification of Significant Targets in Cancer, [
24]) of CRC samples (128 tumor specimens and 33 cell lines) available in the public tumorscape database (broadinstitute.org/tumorscape) was performed [
25]. The results of this analysis suggest that
ERCC1 does not undergo focal amplification.
As previously mentioned, the
ERCC1-19q13 FISH probe covers several other genes, including
ERCC2,
FOSB,
PPP1R13L, MARK4 and
GPR4. ERCC2 (also known as XPD) is a 5′ → 3′ helicase involved in unwinding the double stranded DNA structure around the DNA lesion in NER prior to ERCC1-ERCC4 incision [
4]. Copy number alterations of this gene have previously been reported, but only in the form of infrequent gene loss [
12,
13]. These findings are in line with the deletion frequency observed in the present study. It should be noted that FOSB, PPP1R13L and GPR4 appear to play a role in oncogenesis [
26‐
28], but do not appear to have been investigated in relation to gene copy number alterations. Interestingly,
MARK4 appears to undergo gene amplification in glioblastoma cell lines, resulting in overexpression the MARK4L isoform and increased proliferate capacity [
29]. Due to the nature of the
ERCC1-19q13/CEN-2 FISH probe, it is unknown to what extent other 19q genes are gained in specimens harboring an
ERCC1-19q13 gain. Gain of 19q would result in increased copy number of several well known genes with involvement in cancer, such as
BAX[
30],
CEACAM1[
31,
32],
AKT2[
33] and
BCL2L12[
34].
Higher
ERCC1-19q13 copy numbers were significantly associated with longer survival in the univariate analysis, whereas higher
ERCC1-19q13/CEN-2 ratios and
ERCC1-19q13 gain produced non-significant trends (see Table
2). No relationships were observed for TTR and LR as clinical endpoints in the univariate analysis.
In the multivariate analysis, tumor localization was significantly associated with OS and TTR, where rectal cancers exhibited a poorer prognosis [
17]. This finding may be attributed to the conventional surgical techniques at the time of specimen collection, which was performed before the introduction of total mesorectal excision (TME) in Denmark. TME has since its implementation as a standard operative procedure significantly improved overall 5-year survival, with the greatest improvement observed for stage III patients [
35]. Therefore, the clinical outcome of patients in the rectal cancer subgroup is likely to differ from those receiving surgery today. It could be of future interest to investigate the relationship between
ERCC1-19q13 gain and stage III rectal cancer patient prognosis.
Multivariate analysis, performed adjusting for age and gender, revealed significant relationships between higher
ERCC1-19q13 copy numbers and
ERCC1-19q13/CEN-2 ratios and longer survival and TTR in patients with colon tumors, but not rectal tumors (see Table
3). Similarly,
ERCC1-19q13 gain was significantly associated with longer survival and exhibited a non-significant trend towards longer TTR in the colon subgroup. This non-significant finding may be attributed to the low number colon tumor specimens (17 of 81 specimens) harboring an
ERCC1-19q13 gain. Taken together, these findings suggest that
ERCC1-19q13 copy number increases occur in both colon and rectal tumors, but are only related to better prognosis in patients with colon tumors. Colon and rectal tumors are widely studied as a single entity and a landmark genome-scale analysis has revealed striking similarities between tumors from either localization, with the exception of tumors located in the right/ascending colon, which frequently exhibit microsatellite instability (MSI) [
22]. While tumors may exhibit similar genomic profiles, the prognostic impact of a given genetic alteration may differ according to tumor localization. A study of
TP53 mutations, specifically denaturing mutations, revealed significant prognostic impacts in some tumor localizations (distal colon), but not others (proximal colon and rectum) [
36], a finding with similarities to that of the present study.
After analysis of the first 50 tumor specimens,
ERCC4/CEN-16 hybridization was terminated due to a lack of aberrations, a finding similar to what was observed in the cell line metaphase panel and also supported by GISTIC analysis in the tumorscape database. In ovarian cancer,
ERCC4 mRNA expression has previously been shown to be tightly correlated with
ERCC1 mRNA expression, indicating that the mechanisms regulating the expression of these genes are linked [
37]. We therefore suggest that future explorative studies of
ERCC1 copy number alterations in other cancer types also investigate
ERCC4 copy numbers, as these may potentially play a role in
ERCC1 expression.
Observer workload can be reduced substantially by requiring fewer nuclei to be scored when determining
ERCC1-19q13 status. Wolff and colleagues [
38] suggest a minimum concordance of 0.95 as a requirement for the validation of a
HER2 assay when compared to a validated method, a guideline which was adopted for the current study. Scoring an initial 30 nuclei containing both gene and centromere signals produced concordance of 0.89, when compared to scoring all relevant nuclei. By introducing a borderline interval between 1.35 and 1.65, where an additional 30 nuclei must be scored, concordance increased to 0.99, surpassing the guidelines set forth by Wolff
et al. In the specimens scored in the present study, these updated guidelines would have reduced the amount of nuclei scored by 30.6%. We therefore suggest that future investigations with
ERCC1-19q13/CEN-2 in CRC score nuclei based upon the aforementioned updated guidelines.
Competing interests
David Hersi Smith and Sven Müller are employed at Dako A/S. Ib Jarle Christensen is an external consultant for Dako A/S. The remaining authors have no disclosures or conflicts of interest.
Authors’ contributions
DHS – collection and analysis of data, preparation of manuscript. IJC and BM – statistical analysis of data and preparation of manuscript. NFJ – prepation of metaphase specimens and manuscript. SVM – FISH probe design and preparation of manuscript. HJN - provided patient samples and preparation of manuscript. NB – preparation of manuscript and experimental design. KVN – preparation of manuscript and experimental design. All authors read and approved the final manuscript.