Article Text

PDF

Analysis of candidate modifier loci for the severity of colonic familial adenomatous polyposis, with evidence for the importance of the N-acetyl transferases
  1. M D Crabtree1,
  2. C Fletcher2,
  3. M Churchman2,
  4. S V Hodgson4,
  5. K Neale3,
  6. R K S Phillips3,
  7. I P M Tomlinson1
  1. 1Molecular and Population Genetics Laboratory, London Institute, Cancer Research UK, London, UK, and the Academic Institute, St Mark’s Hospital, Harrow, Middlesex, UK
  2. 2Genotyping Group, Cancer Research UK, Henry Wellcome Building, Headington, Oxford, UK
  3. 3The Academic Institute, St Mark’s Hospital, Harrow, Middlesex, UK
  4. 4Department of Genetics, 8th Floor, Guy’s Tower, Guy’s Hospital, London, UK
  1. Correspondence to:
    Dr I P M Tomlinson
    Molecular and Population Genetics Laboratory, London Institute, Cancer Research UK, 44 Lincoln’s Inn Fields, London WC2A 3PX, UK; ian.tomlinsoncancer.org.uk

Abstract

Background: We have recently shown that the severity of human colonic familial adenomatous polyposis (FAP) varies in a manner consistent with the action of modifier genes. These modifier genes may harbour common alleles which increase the risk of colorectal cancer (CRC) in the general population. Analyses have suggested several common polymorphisms as risk alleles for CRC.

Methods: We determined the association between the severity of colonic FAP (151 patients) and polymorphisms in MTHFR, NAT1, NAT2, GSTM, GSTT, cyclin D1, E-cadherin, and APC. All of these loci have been suggested as influencing the risk of CRC. Colonic FAP severity was quantitated as the number of polyps per colectomy specimen, standardised for colon size. We analysed the relationship between disease severity and genotype at the polymorphic site, making allowance for the position of the germline APC mutation.

Results: We identified significant associations between more severe disease and the absence of the NAT1*10 genotype in the whole group of patients. In a subset of patients with germline mutations in the so-called “mutation cluster region”, there was an association between more severe disease and the presence of NAT2*fast alleles. In the whole patient set, a relatively strong association existed between more severe disease and possession of both the NAT1*non-10 and NAT2*fast genotypes. There was weak evidence for an association between the APCT1493C allele and more severe disease in the whole patient group. No consistent association with disease severity was found for the other polymorphisms.

Conclusion: The severity of colonic FAP may be modified by alleles at the NAT1 and/or NAT2 loci. The identity of any functional variation remains unknown as NAT1*10 appears to be non-functional and there is linkage disequilibrium between alleles at multiple sites within these loci which are adjacent on chromosome 8p22. While evidence from this study cannot be conclusive, our data suggest that NAT1 and NAT2 variants may explain an approximately twofold increase in polyp number in the FAP colon.

  • modifier genes
  • carcinogen metabolism
  • phenotypic severity
  • familial adenomatous polyposis
  • FAP, familial adenomatous polyposis
  • CRC, colorectal cancer
  • NAT, N-acetyltransferase
  • SNP, single nucleotide polymorphism
  • HNPCC, hereditary non-polyposis colon cancers
  • GST, glutathione S-transferase
  • MTHF, 5,10-methylene tetrahydrofolate
  • MTHFR, methylene tetrahydrofolate reductase
  • CCND1, cyclin D1
  • CDH1, cellular dose of E-cadherin
  • MCR, mutation cluster region
  • PCR, polymerase chain reaction
  • IQR, interquartile range

Statistics from Altmetric.com

Familial adenomatous polyposis (FAP) is a rare condition caused by a germline mutation in the APC gene.1 FAP is characterised by the development of hundreds to thousands of adenomatous colonic polyps. If untreated, one or more polyps will develop into colorectal cancer, often by the fourth decade of life. The risk of cancer is related to the number of colonic adenomas.2 The number of adenomas which an individual develops is partly a function of the position of the germline mutation, but this cannot explain the phenomenon of intrafamilial variation.34 Recently, we have shown that the pattern of intrafamilial variation in colonic FAP severity is consistent with the action of modifier genes.56 These observations support those made in animal models where modifier loci modulate the phenotypic severity of the Min mouse,78 the murine equivalent of FAP.

Modifier genes of FAP are likely to be common polymorphisms. These polymorphisms are excellent candidate low penetrance alleles for differential susceptibility to colorectal cancer (CRC) in the general population.9 CRC is known to have a heritable component10 and many plausible risk alleles have been suggested. There are various potential strategies for their identification, including not only direct analysis of association with disease in the human population, but also complementary studies such as identification of modifier genes. Candidate FAP modifiers are potentially of several types, including carcinogen metabolism genes, DNA repair loci, tumour suppressors and oncogenes, Wnt pathway genes, and many other genes with putative roles in CRC.11

Metabolism of carcinogens and other chemicals foreign to the body is governed by a complex set of enzymatic reactions. Environmental carcinogens directly or indirectly damage DNA; the process is complex and consists of three elements, bioactivation, detoxification, and chemical modification. Bioactivation and detoxification are regulated by phase 1 and phase 2 enzymatic reactions. DNA damage probably occurs by the formation of adducts which then stabilise slipped mutagenic intermediates that can occur during DNA replication. The genes examined in this study are briefly outlined below. A more comprehensive review has recently been provided by Houlston and Tomlinson.11

N-acetyltransferases (NAT1 and NAT2) are involved in phase 2 reactions which metabolise xenobiotic compounds. Two types of phenotypes are recognised for each protein—fast and slow. Slow metabolisers (for example, NAT1*14 and NAT1*15) have a Vmax approximately 50% of that of rapid acetylators.1213NAT1 and NAT2 lie close together on chromosome 8p22. The most common variant at NAT1 is termed NAT1*10, which is primarily represented by two non-coding single nucleotide polymorphisms (SNPs) (table 1). It is currently thought unlikely that the NAT1*10 alleles themselves encode functionally distinct proteins but they may be in linkage disequilibrium with fast allele functional variation.12–14 Although not all studies are consistent, the NAT1*10 polymorphism has been associated with a 1.92-fold increase in the risk of CRC1516 and with a lower age of onset of hereditary non-polyposis colon cancers (HNPCCs).17

Table 1

Polymorphisms tested

For NAT2, humans can be divided into slow and fast acetylator populations on the basis of their isoniazid phenotype. In general, the fast phenotype corresponds to the presence of the NAT2*4 allele, either as a heterozygote or homozygote. NAT2*4 is defined (table 1) by a haplotype comprising specific alleles at one silent (codon 161) and three coding polymorphisms (codons 197, 268, and 286) within NAT2.18 Slow acetylator status may be protective against the onset of colon cancer in patients with HNPCC,19 although not all studies have shown this effect.20 In a meta-analysis of low penetrance susceptibility alleles, the risk of CRC was raised in the rapid acetylator group,11 although this effect did not reach statistical significance overall.

Glutathione S-transferases (GSTs) comprise a family of phase 2 enzymes which catalyse the reaction of glutathione with organic compounds. There is substantial (sixfold) interindividual variation in the activity of these enzymes in red blood cells.21 The GSTM1 null allele is a putative cancer risk allele (table 1). It has a population frequency of approximately 0.7.2223GSTM1 null allele homozygosity has been associated with a borderline increased risk of CRC.11 Some studies have found that DNA adduct levels are higher in GSTM1 null homozygotes if they are smokers, suggesting that gene-environment interactions are important.24 The GSTT1 null allele polymorphism (table 1) is responsible for variation in metabolism of mono- and dihalomethane and other similar molecules such as alkyl halides, dichloromethane, and ethylene oxide. Monohalomethanes occur naturally, but compounds such as dichloromethane and ethylene oxide are important industrial chemicals used as methylating agents, solvents, and pesticides. Sixty per cent of the human population can conjugate the above chemicals but the remainder cannot do so as they are GSTT1 null homozygotes. In a meta-analysis, two studies showed a small non-significant adverse risk profile associated with the GSTT1 null allele.11

Folate metabolism abnormalities have been implicated in the aetiology of human neural tube defects, vascular disease, and the development of malignancy.25 Methylene tetrahydrofolate reductase (MTHFR) encodes an enzyme that converts 5,10-methylene tetrahydrofolate (MTHF) to 5-methyl tetrahydrofolate. Defective MTHFR function renders individuals liable to hyperhomocysteinuria and mimics the effects of dietary folate insufficiency. MTHFR alleles have been implicated in colorectal carcinogenesis and represent some of the strongest candidates for low penetrance susceptibility alleles.11 The two most frequent polymorphisms (table 1) are the C677T substitution that converts alanine to valine at amino acid 222 (and is associated with reduced enzyme activity) and the C1298A that converts glutamic acid to alanine at amino acid 429.26

Cyclin D1 (CCND1) is a downstream target of β-catenin. The G870A polymorphism (table 1) interferes with the splicing of CCND1 exon 4/5, probably reducing transcript levels.27 This variant has been shown to be more frequent in familial colorectal cancer cases than controls28 and may modify the age of onset of colon cancer in HNPCC.27 CyclinD1 genotype also influences tumour number in the ApcMin mouse.29 Similarly, variation in the cellular dose of E-cadherin (CDH1) has caused variation in the severity of intestinal polyposis in animal models.30 Downregulation of E-cadherin is associated with some types of intestinal, breast, bladder, and lung cancer. The CDH1 A−160C promoter polymorphism (table 1) may alter expression of E-cadherin and is therefore a plausible candidate modifier allele for FAP.31

Rare APC variants (I1307K and E1317Q) have been associated with a raised colon cancer risk but these alleles are too infrequent to account for much of the population risk variation for “sporadic” colon cancer. However, our study of intrafamilial colonic FAP severity indicated that the severity correlation was higher between sibling pairs than parent-offspring pairs.6 This raised the possibility that the ostensibly “wild-type” APC allele could be influencing disease severity. There are two relatively frequent non-silent APC polymorphisms in Western populations that might be responsible for the observation (table 1). These are APCD1822V32 and a polymorphism (rs2019720) near the untranslated 5′ part of the gene (K Heinimann, personal communication); either could conceivably alter gene expression or function. Other non-synonymous APC polymorphisms include G1493A and G1678A.

We have examined the relationship between colonic FAP severity and a variety of candidate polymorphisms in the above genes (table 1).

METHODS

A total of 151 patients from 51 families were ascertained via the St Mark’s Polyposis Registry, London, UK. DNA samples and clinical details were obtained from all subjects. This sample size provided >90% power to detect at p = 0.05 a nominal threefold increased risk of severe disease conferred by an allele at a frequency of 0.5. All patients were from pedigrees with established classical FAP—in that one or more family members had more than 100 adenomas—with autosomal dominant inheritance of disease. DNA came from one of several sources: peripheral blood; established cell line; or fixed normal tissue. Most patients had known germline APC mutations, between codons 170 and 1464.

The severity of colonic polyposis was assessed, as described previously.56 In brief, we determined the number of adenomatous polyps in the resected colectomy specimen, after correction for length and assuming that the standard colon is 100 cm long, to give the standard count (actual count×100/patient’s colon length). We did not correct for the effects of age, as we have previously shown that age is not a major source of variation in severity for patients who undergo colectomy in early adult life.56

In order to correct for effects of the germline mutation on polyp number, patients were placed into groups derived from work previously published which demonstrates that germline mutations in the APC “mutation cluster region” (MCR) tend to produce a more severe phenotype than mutations in other regions.633 In practice, we used three main groupings: MCR (codons 1250–1400); MAIN (not within MCR); and UNK (germline APC mutation unknown).

A variety of established methods were employed to genotype most of the candidate modifier polymorphisms described above, alongside both positive control samples of known genotype and negative controls. In general, only the most common polymorphisms within each gene were typed. The references supplying details of the methods used for these polymorphisms and the nature of the variation typed are given in table 1. For the APC polymorphisms, we set up assays using Pyrosequencing (Uppsala, Sweden). Polymerase chain reaction (PCR) amplification was performed as recommended by the manufacturer using one biotinylated oligonucleotide, standard reagents, and cycling conditions of: 95°C five minutes ×1; 95°C 15 seconds, 53°C 30 seconds, 72°C 25 seconds ×50; 72°C five minutes ×1. The reaction product was immobilised on streptavidin-coated Dynabeads and denatured in 0.5 M NaOH, followed by sequencing using a third oligonucleotide according to the manufacturer’s standard protocol. The following oligonucleotides were used for the original PCR (table 1): for rs2019720, GTGAACAGGGTGGCAAACAG and GGCCTAACAGAGGGAGAAAAA; for rs459552, CAGACAACAAAGATTCAAAGAAACA and TCAAAAGCAAAACTTCCTCTG; for rs41115, TTCTGCCTTCTGTAGGAATGG and TAACAATCGAATCCCCTCCA; and for rs42427, AATGCTGCAGTTCAGAGGGT and CACTCAGGCTGGATGAACAA. The sequencing oligonucleotides used were, respectively, GAAAAATAATTCCAAG AACTAACAAAGACACTGGT, GAGAAGGAGTTAGAGGAGG, and ACTTTATTACATTTTGCCAC. Haplotypes were reconstructed for the APC polymorphisms wherever possible, partly to determine whether they were present on the germline mutant or wild-type background. None of our patients reported that they were of Ashkenazi origin and APC I1307K was therefore not typed.

Associations between genotypes and standard polyp count were performed, grouping genotypes at each locus, where appropriate, according to their frequencies. All statistical computations were performed using STATA 7.0. The severity data were analysed both in the whole data set and after grouping by position of the germline APC mutation. Our approach was to use non-parametric tests because standard polyp counts were not normally distributed in the MAIN (n = 98) and UNK (n = 36) groups, although counts in the smaller MCR group (n = 18) were consistent with a normally distributed variable (table 2 and Shapiro-Wilk test, details not shown). We used a threshold of p = 0.05 to indicate statistical significance, our reasoning being that it was more important to tolerate possible type II errors than type I errors as replication of our findings in further studies would be necessary whatever our results. Candidate polymorphisms were considered initially in isolation, and then in combination, for four groups ((i) NAT1 and NAT2, (ii) MTHFR alleles, (iii) GSTM1 and GSTT1, and (iv) the APC polymorphisms) because of close linkage between polymorphic sites at these loci.

Table 2

Disease severity. Colonic severity is shown as the standard count (that is, the colectomy polyp count corrected for colon specimen size alone)

RESULTS

Details of patients’ colorectal polyposis are summarised in table 2. The mean standard count was 1519 polyps (SD 1548, median 950, interquartile range 533–1865). Patients included in this study were representative of the whole population of FAP patients from St Mark’s Hospital (details not shown). There was no evidence of association between any polymorphism genotype (including at APC) and the germline mutation group (all p>0.2, details not shown). Disease severity did not vary significantly with sex (Kruskal-Wallis test, χ21 = 0.107, p = 0.74). As expected, there was a strong tendency for disease to be more severe in the MCR group, with a median corrected polyp count of 3727 compared with 760 for the MAIN and 1107 for the UNK patients (Kruskal-Wallis test, χ22 = 30.8, p<0.001). Disease severities in the MAIN and UNK groups did not differ significantly from each other (χ21 = 0.745, p = 0.38). It is highly likely that the UNK patients did not have MCR mutations because the MCR part of the gene is a small region within a large exon for which mutation detection is relatively straightforward, whereas other parts of the gene may harbour cryptic changes, such as intronic variants and large deletions. Furthermore, FAP patients with germline mutations in the MCR tend to show LOH as the “second hit” in their colonic polyps whereas other FAP patients tend to acquire a protein truncating mutation as their “second hit”.34 An FAP modifier gene might therefore act differentially with respect to the site of the germline APC mutation. Therefore, in addition to analysis of each polymorphism with respect to disease severity in the whole patient group, analysis was also performed for the (MAIN+UNK) and MCR groups separately.

The results of testing each polymorphism for an association with disease severity in the whole patient set are shown in table 3. The NAT1*10 allele as heretozygote/homozygote was associated with less severe colonic polyposis (median 563) than compound heterozygotes or homozygotes for NAT1*non-10 alleles (median 928). This association was derived solely from the (MAIN+UNK) group of patients (Kruskal-Wallis test, χ21 = 6.61, p = 0.01), with no evidence of any association in the smaller MCR group. Of the other polymorphisms (table 3), only APCT1493C provided any evidence of an association; there was more severe disease in those carrying the C allele, an effect which just failed to reach significance when genotypes were grouped (table 3), but was nominally significant (Kruskal-Wallis test, χ22 = 6.00, p = 0.05) when all three genotypes were tested separately, owing to the presence of severe disease in the three patients with the C/C genotype.

Table 3

Associations between disease severity and genotype at each polymorphism in the entire patient set

When the (MAIN+UNK) and MCR groups were tested separately for associations between polymorphisms and disease severity, the only significant association was that of the NAT2 polymorphism in the MCR group. Carriers of one or two NAT2*fast alleles in the MCR group had a median standard polyp count of 4884 (interquartile range (IQR) 4293–6489; n = 4) compared with 1971 (IQR 1020–3198; n = 7) in those with no NAT2*fast allele (Kruskal-Wallis test, χ22 = 4.32, p = 0.038). NAT2*fast carriers did have more severe disease in the (MAIN+UNK) group also (median 900 for NAT2*fast v 687 for NAT2*slow) but this difference was not significant (Kruskal-Wallis test, χ21 = 0.654, p = 0.42). Carriers of two GSTM1*null alleles in the MCR group tended to have less severe disease (median 2564 (IQR 1600–3818); n = 9) than those with other GSTM1 genotypes (median 6489 (IQR 5000–7978); n = 2) but this was of only borderline significance (Kruskal-Wallis test, χ21 = 3.56, p = 0.059). All polymorphisms which demonstrated associations with disease in the (MAIN+UNK) group alone also did so in the whole set of patients.

We tested the effects of combinations of genotypes in different ways, depending on the loci concerned: for NAT1 and NAT2, we tested the combination of absence of NAT1*10 and possession of NAT2*fast; for MTHFR, we used the hapipf command of STATA to predict haplotypes; for GSTM1 and GSTT1, we tested the combination of null alleles at each locus; and for APC, we directly determined “wild-type” haplotypes using the pedigrees. There was no detectable association between genotype and disease severity for MTHFR, GSTM1-GSTT1, and APC (details not shown). For NAT1NAT2, however, more severe disease was associated with combined absence of the NAT1*10 genotype and presence of the NAT2*fast genotype. This association was present both in the entire patient set (median 1167 for NAT1*non-10-NAT2*fast v 687 for other genotypes; Kruskal-Wallis test, χ21 = 4.34, p = 0.037) and particularly in the MAIN/UNK group (median 1167 for NAT1*non-10; NAT2*fast v 587 for other genotypes; Kruskal-Wallis test, χ21 = 10.32, p = 0.0013).

DISCUSSION

We have found evidence to suggest that the severity of colonic polyposis in FAP is associated with variation at the NAT1 and/or NAT2 loci. The NAT1*10 allele was associated with less severe disease in the entire patient set, and the NAT2*fast type was associated with more severe disease in the subset of patients with a germline mutation in the MCR. NAT1*10 has previously been associated with an increased risk of colon cancer,15 apparently in contradistinction to our data; but NAT*10 is unlikely to represent functional variation1314 and any association with disease is likely to come from linkage disequilibrium with functional variants. There is some existing evidence to suggest that NAT2*fast alleles are functionally associated with an increased risk of bowel cancer,19 although again, linkage disequilibrium cannot be excluded. To further complicate matters, NAT1 and NAT2 lie within 0.2 Mb on chromosome 8p22 and there is evidence that at least some alleles at each locus are in moderate linkage disequilibrium with each other.35 We found that the combined absence of NAT1*non-10 and the presence of NAT2*fast were strongly associated with severe disease.

Given that variation in FAP polyp number appears to be related to tumour initiation rather than progression,5 a differential effect of the NAT phenotype might influence the early stages of tumorigenesis in the FAP colon, perhaps even at the stage of the “second hit” at APC. If this is the case, then carcinogen metabolism is likely to be important in early human growth and development. It is even possible that carcinogens from the maternal diet impact on the frequency and type of the “second hit” in utero.

The possible association of the GSTM1 non-null phenotype with more severe colonic FAP in the MCR patients appears unlikely to result from a true biological effect. The possible association is somewhat puzzling on theoretical grounds and on the basis of previous studies, from which we would predict the opposite to our findings. We note, moreover, that—unlike NAT1 and NAT2—there was no evidence for the GSTM1 severity association in the largest (MAIN) group of patients. Perhaps given that the five GSTM loci lie within 100 kb on chromosome 1p13.3, there is a true association between another GSTM allele and disease severity.

Similarly, while we have previously suggested that variation in the APC gene itself might modify the FAP phenotype, we are cautious about the finding of a possible association between the APC1493C allele and more severe disease. This variation is unlikely to have functional consequences but may be in linkage disequilibrium with other variation, for example in the APC promoter (either in cis or in trans with the germline APC mutant). Haplotype reconstruction from pedigrees did not however show that the C allele was consistently associated with the mutant or wild-type APC allele and we found no evidence for a specific APC haplotype associated with disease severity. It is possible that the potential association between APC1493C and disease severity results in part from the fact that we decided to include different members of each family as providing independent data for this study. The justification for this relies on the following: (i) balancing potential errors from the assumption of independence against the loss of power; (ii) our analysis of common polymorphisms which are expected to segregate within families; (iii) relatively small number of individuals per family in our study (median 3); and (iv) allowance for the effects of different germline APC mutations by subdividing into MAIN, UNK, and MCR groups.

Our findings on the other polymorphisms (MTHFR, GSTT1, E-cadherin, and cyclin D1) provide no convincing evidence for an effect on disease severity. In the case of the MTHFR C677T polymorphism, in particular, this is an important negative result. Studies have consistently shown a weak association of the T allele homozygote with reduced colon cancer risk (summarised by Houlston and Tomlinson11) but no effect on adenoma risk has been detected, in accord with our results.

We conclude that variation at the NAT1 and/or NAT2 loci is a possible modifier of the number of colorectal polyps in patients with FAP. The nature of the variation remains unknown. Evidence from this study is not conclusive but our data suggest that NAT1 variants may explain an approximately twofold increase in polyp number in the FAP colon (table 3). The same or different modifier genes may be involved in determining the severity of extracolonic disease, such as desmoids and upper gastrointestinal tumours. It is however evident that, even if the effects of NAT1 and/or NAT2 are confirmed by other studies, there may still be unexplained variation in FAP. The possibility remains that variation in FAP severity is a polygenic trait, its study requiring analysis of large numbers of patients at multiple polymorphic loci.

Acknowledgments

We are grateful to the patients involved in the study, to the Equipment Park, Cancer Research UK, to staff from the Polyposis Registry and Colorectal Cancer Unit, Cancer Research UK, St Mark’s Hospital, and to two referees (Tim Bishop and one anonymous). MC held a Cancer Research UK Translational Fellowship.

REFERENCES

View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.