Introduction

In colorectal cancer (CRC) there are two classical pathways that direct tumorigenesis: microsatellite instability (MSI or MIN) and chromosomal instability (CIN). MSI results from a defective DNA mismatch repair (MMR) system and therefore characterises tumors from patients with Lynch syndrome (previously HNPCC, hereditary nonpolyposis colorectal cancer). In addition 15% of sporadic CRC displays MSI due to MLH1 promoter hypermethylation [1–3]. Tumor cells with abrogated MMR function accumulate small deletions and insertions in stretches of short repetitive DNA sequences distributed throughout the genome. These mutations lead to frameshifts within coding sequences and thus inactivation of genes, thereby contributing to tumor development and progression [4–6]. MSI carcinomas most often show a diploid or near-diploid genome [7], while up to 73% of sporadic CRC tumors show aneuploidy, the equivalent of a gross amount of CIN [8]. In sporadic microsatellite unstable (MSI-H) carcinomas the most frequent aberrations are gains of chromosome 8, 12 and 13 while chromosomal losses occurred predominantly at 15q14 [9]. In sporadic microsatellite stable (MSS) CRC, CIN is characterized by losses and amplifications of arms of, or complete, chromosomes [10–12]. In general, physical loss of chromosomes 17p and 18q, and gain at 8q, 13q, and 20 occur at early stages during the transition from adenoma to carcinoma, whereas loss of 4p is associated with transition from Dukes’ A to B–D. Chromosomal loss of 8p and gain of 7p and 17q is reported to be associated with the transition from primary carcinoma to local and distal metastases. Loss of 14q and gains of 1q, 11, 12p, and 19 are considered late events [13, 14]. Both chromosomes 5 and 17p are more often targeted by copy number neutral LOH than by copy number variations [15, 16].

Clinically, the uncertainty about the contribution of an MMR unclassified variant (MMR-UV) to the risk of developing cancer is a major problem. While carriers of a pathogenic MMR mutation are at increased risk, those with an MMR-UV could also represent rare variants without increased risk of cancer. For pathogenic MMR carriers, clinical geneticists offer pre-symptomatic testing for the detection of neoplasia at an early stage. For patients carrying an MMR-UV with unproven pathogenicity, offering pre-symptomatic testing is difficult.

Since 2001 evidence for differences between sporadic and familial MSI-H carcinomas with respect to both genotype and phenotype is accumulating. [17, 18] To expand this knowledge we determined the possible difference in genomic tumor profiles of patients with pathogenic MMR mutations, MMR-unclassified variants and of sporadic carcinomas with MLH1 promoter hypermethylation.

Material and methods

Thirty-seven formalin-fixed paraffin-embedded (FFPE) microsatellite unstable (MSI-H) colorectal tumors from 37 patients selected from the pathology archives were included in our study. Corresponding histological normal tissue from 30 of these patients and leukocyte DNA for seven patients was available. In addition, four FFPE endometrial carcinomas with corresponding normal DNA were analyzed. Thirty-one of these samples originated from patients with familial MMR deficiency; the following mutation carriers were included: 11 MLH1 (6 pathogenic, 5 UVs), 10 MSH2 (7 pathogenic, 3 UVs), 5 MSH6 (all pathogenic), and 5 PMS2 (all pathogenic) mutation carriers. One MLH1–UV carrier also showed a mono-allelic G382D mutation in MUTYH and one PMS2 carrier showed a V878A UV in MSH6 as well. A subset of these cases has been reported previously [19, 20]. The mean age at diagnosis of cancer was 49 years for the pathogenic MMR mutation carriers, and 43 years for the MMR-UV carriers. Clinical and mutation data are given in Table 1. The additional 10 samples originate from 10 patients that present with sporadic MMR deficient right sided (RST) colon carcinomas based on MLH1 promoter hypermethylation with a mean age of 76 years. The study was approved by the Medical Ethical Committee of the LUMC (protocol P01-019) and the tumors were analyzed following the guidelines described in the code for proper secondary use of human tissue established by the Dutch federation of medical sciences (http://www.federa.org/).

Table 1 Characteristics of pathogenic MMR mutation, MMR-UV carriers and patients with sporadic MSI-H carcinomas
Full size table

MSI analysis and immunohistochemistry (IHC) of the MMR genes

MSI analysis and immunohistochemical staining of the MMR proteins was performed as described by de Jong et al. [19].

DNA isolation

Normal and tumor tissue was selected by a pathologist (HM), guided with microscopy of a hematoxylin eosin-stained slide. DNA of the selected tissue was extracted from FFPE material as described [19]. The DNA was subsequently cleaned up using protein precipitation solution (Promega, Leiden, The Netherlands) and 2-propanol precipitation. Leukocyte DNA was obtained by salting out precipitation. DNA concentrations were measured using picogreen (Invitrogen-Molecular Probes, Carlsbad, CA, USA).

Hypermethylation analysis of the MLH1 promoter

The MLH1 promoter hypermethylation status of the five MLH1-UVs and the 10 sporadic MSI-H right-sided tumors were determined by hypermethylation analysis of the MLH1 promoter using a methylation-specific MLPA assay as previously described [21].

Single nucleotide polymorphism array analysis

DNA was tested using Illumina BeadArrays and the GoldenGate assay (Illumina, San Diego, CA, USA). The GoldenGate assay was carried out according to the manufacturer’s protocol with minor differences: 1 μg DNA was used as input in a multi-use activation step and subsequently dissolved in 60 μl resuspension buffer. For each sample, four SNP panels (linkage panel, LP), LP1-4, were tested together covering the genome: LP1 covers chromosomes 1–3 and 22, LP2 for chromosomes 5–9, LP3 for chromosomes 10–15 and 21, and LP4 for chromosomes 4, 16–20, X and Y. Each panel was analyzed separately on a beadarray. Due to the limited availability of archival tumor tissue some of the LPs could not be analyzed. In 13 cases one LP, and in one case two LPs could not be analyzed. Two carcinomas (cases 13 and 15,) could therefore not be analyzed for loss at MSH2 or MSH6, respectively, and two for the hypermethylated MLH1 locus (case S32 and S78, Table 1). Overall, we were able to analyze 91% of the genome in the three groups we corrected for the missing information in subsequent calculations.

We used linkage mapping panel version IV_B containing 6008 SNP markers distributed evenly over the genome with an average physical distance of 482 kb. Gene calls were extracted using GeneCall (version 6.0.7) and GTS Reports (4.0.10.0) (Illumina, San Diego, CA, USA). The software provides two quality scores: an experiment-wide gene train score (GTS) and a sample-specific gene call score (GCS).

Copy number and loss of heterozygosity (LOH)

Copy numbers were determined from the signal intensity of the individual SNPs. LOH was analyzed by comparing the genotypes from paired normal and tumor DNA. Both genomic profiles were generated with the R-package BeadArray SNP [22]. In addition, chromosome visualization of LOH was performed in Spotfire DecisionSite (Spotfire, Somerville, MA, USA) [15]. Furthermore, LOH was computed from the GCS and the GTS. LOH was called for high quality heterozygous SNPs in the normal tissue (relative gene call score (rGCS) > 0.8) that were, in the paired tumor, either homozygous or showed an rGCS/GTS ratio <0.8. In practice, regions of LOH always presented with stretches of markers showing LOH. LOH at one or two SNPs was ignored [15, 23]. Our interpretation of LOH has been verified in separate experiments with tumors using microsatellite and FISH probes (results not shown). When both physical loss and LOH were detected at a specific region, we considered the detected LOH as an additional indication of physical loss. If no copy number change was detected, LOH was interpreted as copy neutral LOH (cnLOH).

Statistics

With a one-way ANOVA F test the amount of chromosomal aberrations in the three MSI-H groups was compared. A Scheffe-post hoc test was performed between the contrasts when the 0-hypothesis was rejected.

Results

We studied genome wide copy number changes and copy neutral LOH (cnLOH) in formalin-fixed paraffin-embedded (FFPE) tumor tissue using 6 K SNP arrays. The cohort consisted of 23 MSI-H tumors of 23 Lynch syndrome patients with pathogenic mutations eight tumors of patients with unclassified variants in MLH1, or MSH2 genes. In addition, 10 sporadic MSI-H carcinomas with MLH1 promoter hypermethylation from 10 patients were analyzed (Table 1).

Lynch syndrome cases with pathogenic MMR mutations

Immunohistochemical analysis revealed that in all carcinomas of pathogenic MMR mutation carriers the protein of the mutated gene was abrogated. In 14 of these cases both proteins of the heterodimer (MLH1/PMS2 or MSH2/MSH6) were abrogated. In six of these carcinomas (from three MLH1 and three PMS2 mutation carriers) also MSH6 was not expressed, this might be due to a frameshift in the C8 repeat which is located in the coding region of MSH6 [24] (Table 1).

As expected from the literature [7, 25, 26], very few copy number aberrations were observed in the carcinomas from the carriers of pathogenic MMR mutations (Table 1). Only five of 23 (22%) MSI-H tumors presented with copy number abnormalities. Four of these cases, showed a single loss or gain of a chromosomal region. The fifth tumor presented with gain in two chromosomal regions. The chromosomes 2p, 3q, 9p, 19q and 20p were targeted in these tumors and the size of the affected segments ranged from 1 to 19 chromosomal sub-bands. Chromosome band 9p24.3 was targeted twice, in cases 7 and 17 by physical loss and gain, respectively. Physical loss of the MMR gene involved was only detected in case 10. Interestingly, in this case two different types of alterations were detected around chromosomal sub-band 2p21 harbouring the (mutated) MSH2 gene; physical loss adjacent to cnLOH. We designated this alteration as physical loss.

Also genome wide cnLOH was infrequent in these tumors (Table 1). However cnLOH around the locus of the mutated MMR gene was frequently observed in the 23 carcinomas from patients with pathogenic mutations. Five of six tumors from MLH1 mutation carriers showed cnLOH at the MLH1 locus (3p22.2) (Table 2). The extend of the LOH ranged from chromosome 3p26.3 to 3p14.1, and the smallest region of overlap (SRO) spanned 3p25.5–21.31, which encompasses MLH1. Two of the six MSH2 tumors showed cnLOH of the MSH2 locus at chromosome 2p21 (the interval of LOH ranged from 2p25.3 to 2p13.3; SRO 2p25.3–14) (Table 2). For PMS2 mutation carriers, cnLOH was seen in two of five tumors (interval of LOH, 7p22.3–11.1; SRO, 7p22.3–13) (Fig. 1, Table 2). None of the five tumors from MSH6 mutation carriers showed cnLOH at the MSH6 locus (2p16, Table 2).

Table 2 Copy neutral LOH at MMR loci and mean percentage of aberrant sub-bands in cases from pathogenic MMR mutation versus MMR-UV carriers
Full size table
Fig. 1
figure 1

LOH view of tumors from a pathogenic PMS2 mutation carrier generated with Spotfire DecisionSite (Spotfire, Somerville, MA). Heterozygous SNPs (upper diamonds in the figure) are dispersed over the chromosomes. These were analyzed in both tumor and corresponding normal DNA. For LOH, ≥3 SNPs in a specific region that are altered from heterozygote in normal to homozygote in tumor (lower diamonds) are scored as LOH. In practice, regions of LOH always presented with stretches of markers showing LOH. LOH at one or two SNPs was ignored. In this case LOH of PMS2 is seen on chromosome 7 and none of the pseudogenes on chromosome 7q are affected

Full size image

In addition, two of seven MSH2 carcinomas presented with cnLOH at 6q with SRO; 6q24.3–25.2 (cases 8 and 14, Table 1). One patient with a pathogenic PMS2 germline mutation (case 22, Table 1) presented with additional LOH of the chromosomal region 2p25.3–15 that harbours MSH2 and MSH6. In this left-sided colon carcinoma, the protein expression of MLH1, MSH6 and PMS2 was abrogated and MSH2 expression was retained. MLH1 and MSH6 germline mutation analysis were negative.

MSI-H carcinomas with unclassified variants in MMR genes

Eight carcinomas from MMR-UV carriers were tested. Five of the eight cases showed normal positive staining of MMR proteins tested. Two MLH1-UV cases showed absent staining of at least MLH1, whereas one MSH2-UV case showed only absence of MSH6 protein. The five MLH1-UVs did not show promoter hypermethylation of MLH1.

Copy number abnormalities were detected in five of eight (62%) carcinomas from MMR-UV carriers. These carcinomas were from four MLH1-UV and one MSH2-UV carriers. Four tumors displayed a single copy number abnormality and the fifth tumor displayed two copy number abnormalities. The affected segments ranged in size from 1 to 52 chromosomal sub-bands. The copy number abnormalities affected chromosome 6p, 7, 8, 9p and 17. Chromosome 9p24.3 was affected in two of these five tumors (a gain in tumor 27 and physical loss in tumor 28a, Table 1). None of the analyzed tumors from MMR-UV carriers showed physical loss at the specific MMR gene locus involved.

CnLOH at the locus of the mutated MMR gene was found to a lesser extent than in tumors from pathogenic MMR mutation carriers (Table 2). Two of the five MLH1-UV carcinomas showed cnLOH at the MLH1 locus on chromosome 3p22.2 while none of MSH2-UV carriers showed cnLOH at chromosome 2p21 (Table 2). Also genome wide cnLOH was limited. Five of the eight carcinomas showed cnLOH, ranging from one to three genomic regions at eight different chromosomes (Table 1).

Sporadic MSI-H carcinomas with MLH1 promoter hypermethylation

Genome-wide profiles of copy number abnormalities and cnLOH were determined from 10 MSI-H carcinomas with MLH1 promoter hypermethylation. Protein expression of MLH1 and PMS2 was abrogated in all 10 carcinomas as determined by immunohistochemistry. In six of the 10 (60%) sporadic MSI-H carcinomas limited copy number changes were detected. Three of these tumors exhibited one copy number abnormality and the other three displayed two changes. The affected segments ranged in size from 8 to 47 sub-bands of the genome of these carcinomas, affecting chromosome 1q, 4, 6, 8p, 9, 10 and 12. Amplification of complete chromosome 12 occurred in two cases (case S19 and S51) all additional copy number changes were unique. The locus of MLH1 showed neither physical loss nor cnLOH in eight tumors that could be tested (Table 2). CnLOH was observed in 3 of the 10 carcinomas (30%). Two tumors showed one segment of cn LOH and the other tumor displayed two segments of cnLOH, affecting chromosomes 6, 12q and 19p (Table 1).

Comparison of three groups

We compared the average number of segments with cnLOH or copy number abnormalities detected in the carcinomas of the different groups (Table 3). The fraction of aberrant segments in each group and the distribution over the chromosomes is shown in Fig. 2. This comparison shows that the carcinomas of patients carrying an UV in one of the MMR genes display more aberrations (on average 2.79, range 0–4), than the carcinomas of patients with a pathogenic MMR mutation (on average 1.44, range 0–3) and the carcinomas with MLH1 promoter hypermethylation (on average 1.32, range 0–4). The average number of aberrant segments of the three groups were compared with a one-way ANOVA test. A significant difference was found (P = 0.045) comparing the total number of segments per group, in a post hoc test (Scheffe test) no significant difference was revealed between the individual groups. The average size (chromosomal sub-bands) of the aberrant segments is larger in the tumors of patients carrying an UV in MLH1 or MSH2 (13 sub-bands) and in tumors with MLH1 promoter hypermethylation (20 sub-bands), compared to the tumors of patients with a pathogenic MMR mutation (8 sub-bands).

Table 3 Average number of chromosomal segments with cnLOH and genomic aberrations
Full size table
Fig. 2
figure 2

Fraction of chromosomal events, per chromosome arm, in MSI-H carcinomas. The shaded bars indicate the percentage of 23 carcinomas from pathogenic mutation carriers and the black bars represent the eight carcinomas from MMR-UV carriers. The grey bars indicate the MSI-H carcinomas with hypermethylation of the MLH1 promoter that exhibit events of chromosomal aberration of a chromosome. This percentage has been calculated for the respective chromosome arms

Full size image

Although subtle, the distribution of the types of chromosomal events—copy number aberrations versus cnLOH—is different in the carcinomas with MLH1 promoter hypermethylation compared to the carriers of a pathogenic mutation or an UV in one of the MMR genes. Whereas in these last two groups the majority of events comprise cnLOH, copy number aberrations are more prevalent in carcinomas with MLH1 promoter hypermethylation (Table 3). The one-way ANOVA test identified a significant difference (P = 0.027) between the number of cnLOH events in the three carcinoma groups; the Scheffe test assigned this result to the difference between the sporadic carcinomas with MLH1 promoter hypermethylation and carcinomas from MMR-UV carriers (P = 0.027). A comparison of the percentage of chromosomal gain, loss and/or cnLOH was made for the three groups. The increase of chromosomal aberrations in carcinomas from MMR-UV carriers compared to the other two groups is again evident. The chromosomes involved and the distribution of the events over the chromosomes is different between the groups. Chromosomes 6p, 9p, 10q and 12p are affected by events in all three groups although to a different level. The suggested increase in events on chromosome 3p in the tumors from MMR-UV (Fig. 2) carriers can be explained by an unequal distribution of MLH1 carriers (pathogenic 6/26 vs UVs 5/8) in the groups. Furthermore, one MLH1-UV case with cnLOH on 3p does not comprise the MLH1 locus.

Discussion

This is the first study that compares genome wide SNP array profiles of MSI-H carcinomas from MMR pathogenic mutation carriers, MMR-UV carriers and carcinomas with promoter hypermethylation of MLH1. With both comparative genomic hybridization (CGH) and SNP arrays, copy number information can be obtained however with SNP arrays also genome wide copy neutral LOH (cnLOH) can be studied which provide us with additional information. We used Illumina 6K SNP arrays on FFPE material and analyzed the data with the BeadArray SNP package [22].

Overall we did not detect extensive cnLOH in MSI-H carcinomas. Most of the cnLOH we found in carcinomas from pathogenic MMR mutation carriers, involved the MMR gene locus. Especially, for MLH1 such cnLOH was seen in tumors from pathogenic mutation carriers (in five of six tumors). In the MMR-UV cases, cnLOH at the MMR locus was less frequent. In literature a varying frequency of LOH has been described on the MLH1 and MSH2 locus in series of pathogenic MMR mutation carriers and MMR-UV carriers. LOH at chromosome 3p has been reported in 35–85% of all tumors with a germline mutation (pathogenic as well as UVs) in MLH1 [3, 27–34]. LOH at chromosome 2p has been described in 14–50% of all tumors with a germline MSH2 mutation (pathogenic as well as UVs) [3, 28, 31, 35]. We detected cnLOH of PMS2 in 40% of tumors, which to our knowledge has not been published previously. Using SNP arrays we have delineated the intervals of LOH around the affected genes. The LOH at the PMS2 locus on chromosome 7p (SRO 7p22.3–13) points at the sensitivity of the technique in view of the existence of about 14 pseudogenes of PMS2 [36–39] that were not targeted by the specific cnLOH of the PMS2 locus.

Of interest is the increased number of aberrant segments in carcinomas from MMR-UV carriers compared to pathogenic MMR mutation carriers and carcinomas with MLH1 promoter hypermethylation. Apparently, CIN is added to microsatellite instability in these MMR-UV cases during tumorigenesis. This could suggest that such additional CIN is necessary for tumorigenesis in cases with a priori weak mutator effects. Furthermore, this finding supports the observations that CIN and MIN are not mutually exclusive [9, 40–42]. With the detection of an unclassified variant in one of the MMR genes in patients that are highly suspected to be affected with Lynch syndrome, the uncertainty that a pathogenic mutation has been missed remains. We now suggest that finding a relatively increased CIN might make this less likely, as was seen in five of eight MMR-UV cases. However, the finding of MSI-H with absence of nuclear staining in cases from MMR-UV carriers does not definitively prove the pathogenicity of such UV. The five tumors from MMR-UV carriers, in which all MMR proteins tested are expressed, suggest the presence of a stable protein that is defective in MMR. It should be remembered that the staining and MSI results also depend on the nature of the second somatic hit that occurred in the tumor. Furthermore, in series of cases with specific MMR-UVs not always the same results are obtained [19].

We see that the chromosomal segment that is targeted is larger in the tumors of patients carrying an UV in MLH1 or MSH2 and in tumors with MLH1 promoter hypermethylation, compared to the tumors of patients with a pathogenic MMR mutation. Aberrations of whole chromosomes are found in, respectively, five of the eight MMR-UV carcinomas, in five of the 10 MLH1 methylated carcinomas and only in two of the 23 MMR pathogenic carcinomas. In addition, the distribution of the types of chromosomal events—copy number aberrations versus cnLOH—is slightly different in the carcinomas with MLH1 promoter hypermethylation compared to the carriers of a pathogenic mutation or an UV in one of the MMR genes. Whereas in these last two groups the majority of abnormalities concerns cnLOH (79% and 64%, respectively), copy number aberrations are the more prevalent abnormality seen in carcinomas with MLH1 promoter hypermethylation (60%). In contrast to other publications we detected equal amounts of gain and physical loss of parts or whole chromosomes in the sporadic MSI-H carcinomas [9, 42]. Trautmann et al. studied 23 sporadic MSI-H carcinomas with array CGH and identified gains on chromosomes 8, 12 and 13. We also identified gain of chromosome 12 in two out of 10 carcinomas.

Moreover, we could identify several small regions with copy number changes and cnLOH that were present in more than one MSI-H tumor with a pathogenic MMR defect on chromosomes 9p24.3 and 6q24.2–25.2 respectively. These regions might harbour genes that are important for tumorigenesis. Recent association studies identified polymorphic sequences at 8q24 as associated with an increased risk for CRC. Interestingly, chromosome 9p24 was also implicated in two of these studies pointing at a role for 9p24 in carcinogenesis [43–46].

The approach we describe here appears to be an elegant way to detect (genome wide) cnLOH in MSI-H formalin fixed paraffin embedded carcinomas. Studying LOH in these type of carcinomas was often hampered due to instability of polymorphic microsatellite markers. We also suggest that the SNP array platform, as described here and applicable to FFPE tissue, may be a crucial tool in finding the genetic cause of unexplained familial colorectal cancer, since we were able to identify distinct small regions of LOH and/or copy number alterations.