Colorectal carcinomas in MUTYH-associated polyposis display histopathological similarities to microsatellite unstable carcinomas

Background MUTYH-associated polyposis (MAP) is a recessively inherited disorder which predisposes biallelic carriers for a high risk of polyposis and colorectal carcinoma (CRC). Since about one third of the biallelic MAP patients in population based CRC series has no adenomas, this study aimed to identify specific clinicopathological characteristics of MAP CRCs and compare these with reported data on sporadic and Lynch CRCs. Methods From 44 MAP patients who developed ≥ 1 CRCs, 42 of 58 tumours were analyzed histologically and 35 immunohistochemically for p53 and beta-catenin. Cell densities of CD3, CD8, CD57, and granzyme B positive lymphocytes were determined. KRAS2, the mutation cluster region (MCR) of APC, p53, and SMAD4 were analyzed for somatic mutations. Results MAP CRCs frequently localized to the proximal colon (69%, 40/58), were mucinous in 21% (9/42), and had a conspicuous Crohn's like infiltrate reaction in 33% (13/40); all of these parameters occurred at a higher rate than reported for sporadic CRCs. Tumour infiltrating lymphocytes (TILs) were also highly prevalent in MAP CRCs. Somatic APC MCR mutations occurred in 14% (5/36) while 64% (23/36) had KRAS2 mutations (22/23 c.34G>T). G>T tranversions were found in p53 and SMAD4, although the relative frequency compared to other mutations was low. Conclusion MAP CRCs show some similarities to micro-satellite unstable cancers, with a preferential proximal location, a high rate of mucinous histotype and increased presence of TILs. These features should direct the practicing pathologist towards a MAP aetiology of CRC as an alternative for a mismatch repair deficient cause. High frequent G>T transversions in APC and KRAS2 (mutated in early tumour development) but not in P53 and SMAD4 (implicated in tumour progression) might indicate a predominant MUTYH effect in early carcinogenesis.

The MUTYH protein is a base excision repair (BER) glycosylase involved in the repair of DNA damage resulting from the oxidation of guanine nucleotides. The oxidation product of guanine, 8-oxo-7,8-dihydro-2'-deoxyguanosine (8-oxoG), readily mispairs with adenosine nucleotides during DNA replication. MUTYH acts by scanning the newly synthesized DNA strand for any mispaired adenines, either with guanines or 8-oxoG's, and excising them. A dysfunctional MUTYH protein increases the occurrence of somatic G>T transversions. For instance, somatic mutations in the APC gene in MAP tumours involve almost exclusively G>T transversions, an observation that led to the discovery of the MAP syndrome. [10] Similarly, the most prevalent KRAS2 mutation in MAP tumours is a G>T transversion at codon 12 (c.34G>T), which was reported to be present in 64% of MAP carcinomas. [11] Such mutation is infrequent in sporadic CRCs, according to published consecutive series. [12] Recently it was reported that MAP CRCs are often neardiploid (52%) and commonly contain chromosomal regions of copy neutral loss of heterozygosity (LOH) (71%). In copy-neutral LOH there is no loss of genetic material and this can arise via mitotic recombination, non-disjunction, or deletion and reduplication events. This is in contrast to sporadic colon cancer, where physical loss of genetic material is the main characteristic. [13] Another set of CRCs with deficiencies in DNA repair are the mismatch repair deficient tumours. Mismatch repair deficient CRCs with high-microsatellite instability (MSIhigh) arise in the context of the Lynch syndrome or have a sporadic origin due to somatic inactivation of hMLH1. MSI-high tumours have characteristic histological and molecular features: they are most often near-diploid and arise preferentially in the right side of the colon. They have a high prevalence of mucinous and medullary histotypes, poor differentiation, a Crohn's-like lymphocytic reaction, and a high amount of (intra-epithelial) tumour infiltrating lymphocytes (TILs). These characteristics are currently employed for diagnostic purposes and may have implications for patient treatment and prognosis. [14][15][16][17] In the present study we describe histological and molecular aspects of MAP carcinomas in a Dutch cohort, and compare these with data available in literature of consecutive series of sporadic, MSI-high (sporadic), and Lynch syndrome-derived CRCs. The aim of this study was to identify specific characteristics that would aid in the diagnosis of MAP CRCs and in differentiation from carcinomas arising from distinct genetic backgrounds.

Patients
Fifty-seven MAP families, from which clinical data were available, were studied. Informed consent was obtained according to protocols approved by the Leiden University Medical Center ethics review board (02-2004). These families include 57 index-patients with biallelic homozygous or compound heterozygous MUTYH mutations and 22 siblings with biallelic MUTYH mutations. Forty indexpatients have been described previously in lesser detail. [3] One patient was reported to have CRC stage A according to the modified Astler-Coller guidelines at age 21, but after re-evaluation appeared to have high grade dysplasia (carcinoma in situ) and was not included in this series of MAP carcinomas. To date, 56% (44/79) of the carriers have developed CRC. A total of 58 CRCs were diagnosed in these 44 MAP patients, composed of 26 males and 18 females. Hematoxylin and eosin (H&E) stained slides and tissue material from 42 and 38 CRCs, respectively, belonging to 35 MAP patients, could be retrieved from 23 pathology laboratories throughout The Netherlands.

Histological examination
Histological tumour type and grade were independently assessed by two observers (HM and MN) according to the World Health Organization classification. [18] Staging was performed according to the modified Astler-Coller (MAC) and the American Joint Committee on Cancer (AJCC) TNM staging system guidelines. Metachronous tumours were defined as new tumours arising in the colon at least six months after the initial diagnosis. [19] Semiquantitative assessment defined two subgroups, according to mucinous content of the tumour: >50% of tumour area involved (mucinous) and 10 to 50% of tumour area involved. Mucinous adenocarcinomas and signet-ring cell carcinoma by convention were considered poorly differentiated [18], although are also separately scored. A Crohn's-like reaction was assessed as grade 0, 1+, or 2+. [20] Grade 2+ was referred to as conspicuous. Leukocyte infiltration was assessed semi-quantitatively on H&Eslides as none, moderate (visible only on high magnification (×40), or marked (visible already on low magnification ×10).

Molecular analysis
Genomic DNA of paired normal colon and colorectal carcinoma tissue was isolated from formalin-fixed paraffinembedded material, as described previously. [21] The percentage of tumour cells in the areas from which the punches were taken, were in all cases above 50% and in most cases above 70%. Microsatellite instability analysis was done according to the Bethesda guidelines using the markers D2S123, D5S346, D17S250, BAT25, BAT26, and BAT40. A tumour was scored MS-stable when no marker showed instability, MSI-high when > 30% of markers showed instability, and MSI-low when only one marker (<30%) showed instability. [22] For somatic mutation analysis of APC, KRAS2, p53 and SMAD4 DNA sequence analysis was performed. Codon 12 and 13 of the KRAS2 gene and the Mutation Cluster Region (codons 1286-1513) of the APC gene were analyzed as described previously. [23] Primer pairs were designed for the coding regions and exon-intron boundaries of p53 exons 5-8 and SMAD4 exons 3-13. Primer details are available from the authors upon request. Germline DNA mutation analysis of the whole MUTYH gene was performed on lymphocytic DNA or DNA from formalin-fixed paraffin-embedded normal tissue as described previously. [3,5] Primer details are available from the authors upon request. For further details see the website of our DNA diagnostic laboratory [24]. To describe MUTYH mutations we used the most upto-date annotation, see the LOVD database. [25,26] Immunohistochemical analysis using tissue microarray To construct a tissue microarray (TMA), triplicate tissue cores were taken from tumour tissue as described previously [27]on the basis of H&E-stained slides reviewed by a pathologist (HM).) Sections were deparaffinised and endogenous peroxidases were inactivated with 0.3% H 2 O 2 in methanol solution after antigen retrieval by means of microwave oven treatment for 10 minutes in 10 mM citrate buffer pH 6.0 (p53, MLH1) or 1 mM Tris-EDTA pH 8.0 (beta-catenin, PMS2). Sections were incubated overnight at room temperature with mouse antihuman monoclonal antibodies directed against p53 (clone D0-7, 1:1000, Neomarkers, USA), MLH1 (clone G168-728, 1:50, BD Pharmigen, USA), PMS2 (clone A16-4, 1:200, BD Pharmigen) and beta-catenin (encoded by CTNNB1) (clone 14, 1: 800, BD Transduction Laboratories, USA). The following day, tissue sections were incubated with a biotinylated secondary antibody in PBS/BSA 1%. Diaminobenzidine tetrahydrochloride was used as a chromogen for the development of the staining. The slides were counterstained with hematoxylin. Immunohistochemistry (IHC) was scored for p53 nuclear staining as: 1 = none, 2 = >0<25% (mostly indicative of a functional intact p53 status), 3 = 25%-75%, or 4 = >75% (the latter two indicative of p53 dysfunction). Staining of betacatenin was graded by the following scale: 1 = membra-nous staining, 2A = membranous and some nuclear staining, 2B = membranous staining, and increased nuclear staining, 3 = strong nuclear staining, with less or no membranous staining. Normal epithelium and stromal cells provided positive internal controls.

Infiltrate analysis using fluorescent immunohistochemical staining
Fluorescent immunostaining was performed as previously described. [28] Five μm TMA sections were used in all experiments. A mixture of the antibodies ab828 (rabbit polyclonal, anti-CD3; Abcam, UK), hNK-1 (mouse monoclonal IgM, anti-CD57; Department of Pathology, LUMC, The Netherlands), and 4B11 (mouse monoclonal IgG2b, anti-CD8; Novo Castra, UK) was added to each slide. The next day slides were incubated with the appropriate combination of fluorescent antibody conjugates (IgG-Alexa Fluor 546, IgM-Alexa Fluor 488 and IgG2b-Alexa Fluor 647). Alexa Fluor conjugates were obtained from Molecular Probes (Leiden, the Netherlands). The images were captured with a confocal laser scanning microscope (LSM) (Zeiss LSM510, Zeiss, Germany). The number of each leukocyte sub-type was assessed per tumour area (TILs/mm 2 tumour epithelium) using the LSM software (Zeiss). Cells staining for CD3, and not for CD8 or other CD markers, were considered to be T-helper lymphocytes. Standard immunostaining for the associated cytotoxic molecule granzyme B (NCL-GRAN-B; clone 11F1; Novo Castra) was performed in sequential sections. To calculate the density of granzyme B positive cells we used the assessed tumour area from the fluorescent immunostaining.

Literature on histological and molecular features of CRC
We compared literature available, describing histological and molecular features, on sporadic, MSI-high and Lynch CRCs. The following terms were employed as search terms: colon carcinomas, bowel cancer, CRC, sporadic, MSI-high, Lynch, HNPCC, histological, molecular, APC, KRAS2, p53, beta-catenin, CTNNB1, SMAD4, tumour infiltrating lymphocytes, tumour infiltrating lymphocytes and intra-epithelial lymphocytes. All relevant references within articles were identified and included.

Statistical Analysis
Fisher exact test was used to estimate an association between molecular and clinical-pathological parameters. The Spearman test was used to assess correlations. All Pvalues are reported for a two-tailed test; P-values of less than 0.05 were considered to be statistically significant. A group with high and low intraepithelial leukocyte infiltration was distinguished, using the median value as cutpoint of the leukocyte-infiltration scores of all patients. When a patient had more than one carcinoma the mean number of lymphocytes in these carcinomas was used for the survival analysis. The Kaplan-Meier method was used to calculate the overall survival and the log-rank test was used for comparison of the survival curves. All statistical analyses were carried out using the SPSS software package (SPSS Inc. 12.0, USA).

Histopathology of MAP carcinomas
Forty-four MAP patients were diagnosed with 58 CRCs (see Table 1) at a mean age of 49 years (Table 2 and 3). Ten patients (23%) had metachronous or synchronous carcinomas. The majority of carcinomas (35/51, 69%, metachronous tumours not included), were right-sided, i.e. proximal to the flexura lienalis. The majority of proximal tumours were located in ascending colon or cecum (80%, 28/35) and only a minority occurred in the transverse colon or hepatic flexure (14%, 5/35). Fifty-five percent (29/53) were stage B according to MAC guidelines or T>1N0M0 according to the TNM guidelines. Forty-two colon carcinomas from 35 MAP index patients were available for further study (Table 1). Histological analysis showed poor differentiation in 26% (11/42) and moderate in 71% (30/42, Figure 1A), mucinous CRCs represented 81% of these cases (9/11, Figure 1B). No MAP carcinomas showed a solid or medullary histotype or contained signet ring cells. A Crohn's like infiltrate was present in 33% (13/40, Figure 1C). In 40% (16/40) of tumours, focal necrosis (dirty necrosis within glandular lumina) was present. Tumour infiltrating lymphocytes assessed on H&E-slides were present in 74% (31/42) and were marked in 17% (7/42, Figure 1D). In 33 tumours MSI status was analyzed and all carcinomas but one were MS-stable. The latter carcinoma showed instability of less than 30% of markers (one dinucleotide marker) designating the tumour as MSI-low. Furthermore, staining for MLH1 and PMS2 was positive for all the tumours.
No significant geno-phenotype correlations for any of the main histopathological parameters could be found.

Somatic mutation analysis and immunohistochemical staining
APC mutation analysis (Table 4) of the mutation cluster region showed somatic mutations in 5/36 (14%) carcinomas; four were MUTYH associated transversions (G>T's); two were C>T transitions, one of them occurring together with a G>A transition (patient 6). KRAS2 mutations were found in 23/36 (64%) of tumours, 22 were c.34G>T transversions. An increased nuclear and reduced membranous beta-catenin staining was found in 11% (4/35). In 57% (20/35) of MAP CRCs, p53 staining indicative of a functional p53 status (>0<25% nuclear staining) was found. Nuclear staining indicative of p53 dysfunction was found in 34% (12/35) ( Figure 1E). In 9 out of 16 carcinomas (56%) that could be analyzed, ten p53 mutations were found. One carcinoma had two mutations (patient 7, Table 4). Three mutations were G>T transversions. Except in one case (patient 7, Table 4), staining was in concordance with the combined results of the p53 staining and LOH of chromosome 17p results published previously by Middeldorp et al (Table 4). [13] When staining was indicative of a dysfunctional p53 status, a mutation as well as LOH was found (patients 2, 23, 24, and 41). In cases were a mutation is present but no LOH was identified for 17p, staining was indicative of a still intact, functional p53 (patient 5, 8, and 16). Only one case (patient 22) had a nonsense mutation in p53, explaining the absence of nuclear staining. All other p53 mutations are (probable pathogenic) amino acid substitutions and all except one have been published previously [29] (Table 4). SMAD4 mutations were present in 26% of MAP carcinomas tested (5/19, Table 4). Two tumours had G>T tranversions.

Infiltrate analysis
First, we scored the presence of TILs and Crohn's like infiltrate on standard H&E sections. Secondly, in order to establish an objective lymphocytic cell count, we also performed triple-fluorescent IHC staining for CD3 + , CD8 + , and CD57 + TILs ( Figure 1F). The median number of intraepithelial T-helper (CD3 + , CD8 -), cytotoxic T lymphocyte (CTL, CD8 + , CD57 -) and natural killer cells (NK, CD8 + / CD57 + ) found in 34 CRCs available for analyses were 20, 37, and 0 cells/mm 2 tumour, respectively. There was a significant correlation between the total number of immunofluorescently detected TILs and the amount of TILs (none, present, marked) assessed on H&E slides (Spearman's test, P = 0.002). Immunohistochemistry for granzyme B, which is expressed on activated CD8+ cytotoxic lymphocytes (CTLs) and involved in the induction of apoptosis of target cells, showed a median of 4 granzyme B + cells/mm 2 . The number of CD8 + TILs showed a significant positive correlation with the number of granzyme B + cells (Spearman's test, P = 0.009). No significant differences in survival were seen between patients with high versus low levels of TILs, although in patients with a high level of CD8 + TILs and granzyme B + cells, a tendency was observed towards a better prognosis ( Figure  2A and 2B, P = 0.15 and P = 0.2, respectively, log rank test).

Literature Review
Results are shown in Table 2 and 3; different study outcomes were aggregated in Table 2 and Table 3 shows concise conclusions of these data. Data on Lynch syndrome carcinomas only included proven mismatch repair gene mutation carriers. An extended version of Table 2 is available online, as a supplement (Additional file 1), showing outcomes of all the individual articles and including data for Lynch syndrome suspected carcinomas where no germline mutations in the mismatch repair genes were searched for or found.

Discussion
Specific histological and molecular characterization of CRCs may have implications in the diagnosis, prognosis, and treatment strategies, as demonstrated for mismatch repair deficient MSI-high CRCs. [15,16,30] In order to explore similar implications for MAP, we analyzed clinical, histological and molecular pathologic characteristics in our series of MAP CRCs and compared our findings to the literature (Table 2 and 3). Age at diagnosis of CRC, when compared to sporadic cases, was relatively young in MAP patients and comparable to that in Lynch patients (49 and 47 years). MAP CRCs showed less metastases than sporadic CRCs, but more than Lynch carcinomas (Table 2 and 3). A prominent feature is the 69% proximal sublocalization of colon carcinomas in this cohort; others reported a proximal location in 29, 43 and 46% of MAP CRCs. [4,11,31] This is still more than that reported in sporadic carcinomas (23%) but less than MSI-high or Lynch carcinomas (75% and 67% respectively, Table 2 and 3). Furthermore, this study showed that proximal MAP tumours had a preferential location in the cecum or ascending colon, as opposed to the transverse colon, which was also reported by O'Shea at al. [31] The relatively high number of observed (meta) synchronous carcinomas in MAP patients is in agreement with previous reports on MAP patients (Table 2). In contrast to sporadic carcinomas but in agreement with MSI-high carcinomas, MAP CRC's were demonstrated to be relatively often mucinous in this study ( In MSI-high tumours the caretaker function of mismatch repair enzymes is disrupted resulting in somatic mutations that are accumulated throughout the cell genome. Such abundance of mutations can result in aberrant frameshift peptides that would be presented at the cell surface, through the antigen processing pathway, to cells of the immune system. This sequence of events could explain the presence of an accentuated intra-epithelial lymphocytic infiltrate in MSI-high tumours. [14,32,33] In MAP carcinomas the disruption of the caretaker function of the BER machinery, mediated by MUTYH mutations, leads to the accumulation of G>T somatic mutations, at least early in tumourigenesis, which might evoke similar specific anti-tumour immune responses. TILs were present in a majority of MAP CRCs (74%) in this study, although marked infiltration was only detected in about one fifth of MAP carcinomas. The median number of CD8 + lymphocytes in MAP CRCs, scored with IHC, fell somewhere between that found in MSI-stable and MSI-high CRC's (indicated by absent MLH-1 expression) previously analyzed with the same procedure in our laboratory. [34] O'Shea et al. reported the same percentage of TILs present in MAP CRCs as in their controls (50%, 8/16). [31] Their controls, however, might not be representative of sporadic CRCs, as the number of TILs reported by others in sporadic CRCs is much less (24%, Table 2 and 3). In MSI-high tumours, an active immune response by the host (represented by a high number of TILs) was also associated with a better survival, especially when associated with granzyme B positivity. [33] The group of MAP patients with a high number of CD8 + and granzyme B + TILs showed a better overall survival, although not statistically significant for either parameters.
The number of APC mutations in the mutation cluster region found in MAP carcinomas (14%) is notably lower than that reported in sporadic CRCs (63%, Table 2 and 3). One possible explanation for the relatively small number of APC mutations could be that in MAP CRCs more mutations lay outside the mutation cluster region. In sporadic      CRCs, APC mutations in the MCR represent 50-77% of all APC mutations found. [35][36][37] In MAP tumours (CRCs and adenomas), a substantial proportion (~60%) of mutations were found outside the MCR. [10,11] Similarly, in MSI-high CRCs noticeably more APC mutations are found when larger regions of APC were analyzed (Table  2A). Another explanation might be that distal carcinomas, which are underrepresented in our cohort, have more APC mutations than proximal tumours as shown by Luchtenborg et al. [38] In addition, APC-MCR mutations were seen in 46% (13/28) of rectum carcinomas analyzed previously in our laboratory. [39] In agreement with data derived on MSI-high carcinomas, strong beta-catenin nuclear staining was not frequent in MAP carcinomas in this study (13% and 11%, respec-tively). In sporadic CRCs this rate is much higher (77%, Table 2 and 3).

MAP CRC histology
We found a high percentage of KRAS2 mutations in our MAP CRCs (64%), comparable to reports by Lipton et al. and Jones et al. [11,40] The vast majority (96%) were c.34G>T transversions (GGT>TGT). Intriguingly, no G>T transversions at the second nucleotide of this codon (leading to GGT> GTT) have been reported so far in MAP CRCs.
In contrast, KRAS2 mutations are found on average in 29% of sporadic CRCs and 22% of sporadic MSI-high carcinomas. Furthermore, the c.34G>T tranversion comprises just 8% of KRAS2 mutations in sporadic and none in MSI-high CRCs. [12] In Lynch carcinomas the percentage of KRAS2 mutations is around 34%, Table 2 and 3)  Nuclear staining indicative of p53 dysfunction was found in 34% in this study, which is less than found in sporadic carcinomas (57% , Table 2A). We found p53 mutations in 60% of carcinomas analyzed; this finding, along with the presence of LOH, correlated with the IHC staining results (Table 4). Three mutations were typical MUTYH G>T transversions. Lipton et al. found nuclear staining in 53% of MAP carcinomas, but p53 mutations in only 21% (of which a minority were G>T transversions), suggesting an alternative mechanism of over-expression. [11] The employment of tissue arrays for the current study could imply an underestimation of cases with both p53 and beta-catenin accumulation in the nucleus. However, we reported that 3 punches representing one tumour in a TMA correctly recapitulated the observations made on analysis of the whole slides. [27].
SMAD4 somatic mutations in this series were present in 26% (5/19, Table 4). Lipton et al. did not find any SMAD4 mutations in MAP CRCs, but did find 18q LOH at the same frequency as in sporadic CRCs. [11] Recently we reported that the chromosome 18q LOH in MAP carcinomas mainly comprises copy neutral LOH and not physical loss, as observed in sporadic CRC. [13] In the early stages of MAP tumourigenesis, a dominance of the BER defect can be concluded from the high frequency  of G>T tranversions in KRAS2 and APC. In the later stages such G>T transversions seem less prominent, as seen in SMAD4 and p53. Mitotic recombination might be a driving force in MAP carcinogenesis, based on our conclusion that the LOH in MAP carcinomas mainly comprise copy neutral LOH. Two previous studies by Colebatch and Lefevre et al. [43,44] have suggested that MAP CRCs can also develop through a MSI pathway (by inactivation of MLH1) because of the finding of MSI-high phenotype in one out of three and one out of six MAP CRCs, respectively. O'Shea et al. found MMR deficiency in 2 out of 11 tumours (18%). However, on the basis of the results of this study (an absence of MSI-high in 35 analyzed carcinomas), we conclude that the MSI pathway is not an important pathway in the development of MUTYH associated tumours.

Conclusion
MAP CRCs as a group show specific histological features that differentiate them from sporadic CRCs, and have similarities with sporadic MSI-high and Lynch syndrome colon cancers, such as a preferential proximal location, mucinous histotype, and increased presence of TILs. These TILs might suggest that defects in base excision repair, similar to mismatch repair deficiency, produce secondary aberrant proteins functioning as tumour-specific neoantigens that, in turn, induce anti-tumour immune responses. Further evidence for MAP can be assembled by the detec-tion of c.34G>T transversion in KRAS2 that takes place in early tumour development. KRAS2 analysis can be implemented as a pre-screening test that helps selecting CRC patients eligibly for germline MUTYH mutation testing. In practice, above features should direct the pathologist towards a MAP aetiology of CRC as an alternative for a mismatch repair deficient cause, especially when diagnosed at a young age and in combination with polyps and/or a recessive inheritance pattern.