Matrix Metalloproteinase-9 (MMP-9) polymorphisms in patients with cutaneous malignant melanoma

We recruited 1002 patients with melanoma (stages 0–IV) at Memorial Sloan-Kettering Cancer Center (MSKCC), New York, USA. The study protocol was approved by the MSKCC Institutional Review Board (IRB). Ninety-six percent of the patients approached signed an informed consent and agreed to participate in the study. Patients filled out a short self-administered questionnaire that included information on gender, race, age, family history, freckling density, hair and eye color, propensity to burn, and ability to tan after sun exposure. The information on hair color, eye color, and propensity to tan or sunburn were combined into a single variable, the "phenotypic index" [16]. This index, with minimum and maximum values of 1 and 5, represents the sum of points assigned to the following phenotypic features: hair color (1 if brown/black; 2 if light brown/blond; 3 if red/auburn); eye color (0 if brown; 1 if green/hazel/blue); and propensity to tan or sunburn (0 if tend to tan; 1 if tend to sunburn). We also obtained clinicopathological information including presence of dysplastic nevi, multiple primary tumors, stage at diagnosis and at follow-up (based on the AJCC 2002 classification), disease status, disease progression, and survival among others. The characteristics of the study group are shown in Table 1.

Blood was obtained from 17 patients and buccal cells were collected from 985 individuals. DNA from buccal cells was extracted using Puregene^® kits (Gentra Systems Inc., Minneapolis, USA), and DNA from blood was extracted with the QIAamp DNA Blood kit (QIAGEN Inc. Valencia, USA) using manufacturer's recommendations. DNA concentration was measured by spectrophotometry at 260 nm in a Spectramax Plus 384 (Molecular Devices, Sunnyvale, USA). The DNA quality was determined by the ratio A260/A280.

All known non-synonymous variations in MMP-9 available in the dbSNP database of the National Center for Biotechnology Information (NCBI) [17] were evaluated with in-silico methods to determine which ones potentially disturb the molecular structure and/or function of MMP-9. In the analysis we determined: 1) the steric distortions and hydrogen bond losses caused by the polymorphisms; 2) the effect of the residue changes on the interactions with known ligands (metal cofactors, synthetic and natural inhibitors); and 3) the effect on conserved or specificity residues of four superfamilies of MMP-9 domains (MMP N-terminal domain, catalytic domain, fibronectin type II domain, hemopexin domain) by multiple sequence analysis [18].

All genotyping was done with PCR-based methods and included: melting temperature analysis [19] coupled to the LightTyper instrument (Roche Applied Science, Indianapolis, USA), pyrosequencing [20] with the PSQ™ 96 MA or PSQ™HS 96A instruments (Biotage AB, Uppsala, Sweden), fragment size analysis [21] by an ABI PRISM^® 3100 Genetic Analyzer (Applied Biosystems, Foster City, USA), and heteroduplex analysis [22, 23] using the Wave DNA Fragment Analysis System (Transgenomic, Omaha, USA). The PCR primers and PCR conditions are listed in the Additional file 1. For the heteroduplex analysis, each sample was run unmixed and mixed with equimolar amount of wild type control. The MMP-9 (-)1562 C/T was assessed with a nested PCR because the largest fragment, although specific, was not suitable for pyrosequencing analysis. All genotyping included known internal controls and were considered for the analysis when there was 100% agreement between 2 independent laboratory members. Samples that failed were repeated once or twice, as needed.

Samples were PCR amplified and then purified with a purification kit following the manufacturer's recommendations (QIAGEN Inc., Valencia, USA). One to 10 ng of each purified sample were sequenced in the DNA Sequencing Core Facility at Memorial Sloan-Kettering Cancer Center. Samples were run in an ABI 3730-XLDNA Analyzer (Applied Biosystems, Foster City, USA). Sequencing electropherograms were read at least twice, reviewed manually and with the Mutation Surveyor software, version 2.41 (SoftGenetics LLC, State College, USA).

Two-sided Chi-Square tests, Cochran-Armitage tests for trend [24], and Fisher's exact tests were performed at each individual SNP in the MMP-9 gene to test for association with various clinical and epidemiologic factors. Multivariate analyses were conducted using logistic regression, adjusting for age, sex, phenotypic index, moles and freckles. To investigate associations between SNP and overall survival, time was measured from initial date of diagnosis with melanoma to date of death or last follow-up. Potential associations between time to recurrence, defined as a patient's first recurrence of melanoma, time to disease progression, and genotype were also examined. Survival estimates were computed using the methods of Kaplan and Meier [25] and comparisons between genotypes were made using the log-rank test. All statistical analyses were carried out using SAS version 9.1 (SAS Institute, Cary, NC). The EM algorithm [26] was used to estimate the haplotype frequencies and corresponding standard errors and confidence intervals were estimated using a jackknife process. For dichotomous traits, likelihood ratio tests were conducted to test for haplotype-trait association [27]. For continuous traits, a generalized linear models framework was utilized and score tests were constructed [28]. Omnibus test statistics over all haplotypes in addition to tests for association between each individual haplotype and the trait were performed. Exact p-values were computed using 1000 permutations. Analyses were conducted using SAS Genetics 9.1 software (SAS Institute Inc, Cary, North Carolina) and haplo.stat in R version 2.1.1 (The R Foundation for Statistical Computing, Vienna, Austria).

Therefore, we derived a synthetic structure of MMP-9 by 3D superposition of different structural fragments of MMP-9 and its close homolog MMP-2. Although this synthetic structure was not optimized for residue-residue contacts, it showed the MMP-9/TIMP binding site. Structure-based analysis of residue-residue interactions and analysis of protein domain family alignments suggested that five of the 10 non-synonymous SNPs -N127K, D165N, R279Q, P574R, and R668Q- would be appropriate candidates that might affect MMP-9 function and/or interaction with its inhibitors. None of the 5 polymorphic residues studied localizes to the MMP-9/TIMP binding interface (Figure 1), however all of these residues have specific interactions between their side-chain and surrounding residues, which are abolished as a result of an amino acid change. Our analysis is consistent with the results obtained by PolyPhen (Polymorphism Phenotyping [29]).

We genotyped one microsatellite and one SNP in the promoter of MMP-9 and five coding SNPs. The genotype and allele distributions are summarized in Tables 2 and 3, and were similar to those described in the dbSNP for European panels (HapMap-CEU, AFD EUR and PGA-European panels from dbSNP build 125) and previous published data. Hardy-Weinberg equilibrium was observed for all the polymorphisms.

Statistically significant differences (p < 0.05) in genotype frequencies were observed in eleven different melanoma phenotypic groups. However, when we adjusted for age, sex, phenotypic index, moles, and freckles only four associations remained significant. The complete set of results is available as additional data (see Additional files 2, 3 and 4).

The haplotype analysis was performed with the following polymorphisms: promoter (-)1562 C/T, promoter (-)131 (CA)_n, Q279R and R668Q. The other SNPs were excluded because of their low or null heterozygosity.

The analysis revealed that the most frequent haplotype contained all reference alleles and was (-)1562C-(CA)_{< 21}-279Q-668R with a frequency of 55.4%; followed by (-)1562C-(CA)_{≥ 21}-279R-668R (18.4%), and (-)1562T- (CA)_{≥ 21}-279R-668Q (12.6%). All other haplotypes showed a frequency less than 9%. The Omnibus test showed non-significant differences between clinical stage at diagnosis, Clark level, tumor thickness, tumor site, and phenotypic index, and the MMP-9 haplotypes (data not shown).