Matrix Metalloproteinase Single-Nucleotide Polymorphisms and Haplotypes Predict Breast Cancer Progression

Breast cancer is the most common female cancer in the Western world and constitutes ∼30% of all cancers diagnosed in women. Although breast cancer–related deaths have decreased by >20% since 1995, predicting outcome still is a major challenge. Breast cancer is a very heterogeneous disease (1), and there is a continuing need to predict those patients who are likely to have more aggressive disease and to tailor treatment accordingly. As a result, intensive research is currently under way to identify the genetic determinants that may prove to be important for both classification and prognosis prediction.

Matrix metalloproteinases (MMP) are a multifunctional family, comprising >28 endoproteinases, that are involved in the degradation of extracellular matrix and basement membrane barriers (2–4). Members of the family have been implicated in the promotion of tumor growth, angiogenesis, invasion, and metastasis (5–7). Elevated levels of MMPs have been described in many tumors wherein MMPs are associated with tumor spread and poor patient prognosis (3, 8–10). MMP gene expression is regulated primarily at the transcriptional level and in many solid tumors, particularly breast cancer; the major source of MMPs is the host microenvironment in response to tumor cell–derived signals (11). Thus, factors which influence the host response may play a role in determining tumor behavior in an individual.

Over recent years, single-nucleotide polymorphisms (SNP), constituting insertions (e.g., MMP-1, G at position −1,607 bp), deletions (e.g., MMP-3, A at position −1,171 bp), or substitutions (e.g., MMP-7, A to G substitution at −181 bp; MMP-9, C to T substitution at −1,562 bp; MMP-12, A to G substitution at −82 bp; MMP-13, A to G substitution at −77 bp) of single bases upstream of the transcriptional start site of MMP genes have been described and shown to affect transcriptional activity, leading to alterations in gene expression (12). Thus, the MMP-1 2G allele creates a binding site for the Ets transcription factor, enhancing transcriptional activity (13), whereas the MMP-3 6A allele exhibits weaker transcriptional activity than the 5A allele, possibly due to enhanced binding of a transcriptional repressor (14). The MMP-7 G allele, the MMP-9 T allele, the MMP-12 A allele, and the MMP-13 G allele all have been shown and predicted to generate higher levels of gene transcription and hence enzyme activity (15–18). These polymorphisms have been variably linked to disease susceptibility (19), tumor invasive capacity (20), and patient prognosis (21) in other cancer types.

Given the role of MMPs in breast cancer progression and the association of high levels of MMPs with poor patient prognosis, we hypothesized that individuals carrying high-expressing MMP genotypes may be more susceptible to tumor spread compared with those patients lacking these alleles. This study therefore aimed to investigate the relationship between functional SNPs in MMP-1, MMP-3, MMP-7, MMP-9, MMP-12, and MMP-13 and tumor behavior in a cohort of patients with breast cancer. We further hypothesized that the joint effects of these SNPs may exert a greater risk than the individual effect of each SNP; therefore, we analyzed the relationship between gene haplotypes and tumor behavior.

Patient samples and cohort selection. A total of 221 patients with breast cancer presenting between 1990 and 2006 to St. Bartholomew's Hospital, London (95% from 1999 onwards) were included in this study, following ethics approval from the North East London LREC. From each year of recruitment, all lymph node–positive cases and approximately equal numbers of grade-matched lymph node–negative cases were selected. The cases before 1999 had presented with recurrence during the main study period of 1999 to 2006, but the original diagnosis was between 1990 and 1998. The original histology was reviewed, and details regarding tumor grade, type, and lymph node status were recorded. All patients underwent segmental or total mastectomy plus an axillary node procedure. This was level 1 clearance before 2001, and sentinel node biopsy plus sample (minimum of four nodes) from 2001. All lymph nodes were assessed on single H&E alone, and micrometastasis was excluded from the study. Data were available for estrogen receptor and progesterone receptor status. Patient ethnicity details were retrieved from the clinical records, and for the purpose of this study, individuals were recorded as Caucasian or other. Patient outcome data were retrieved from the clinical records, with a median follow-up of 60 months (range, 2-180 months). MMP SNP genotype and haplotype were related to disease-free survival and overall survival. Only breast cancer–specific deaths were considered, with data from patients who died from other causes being censored.

DNA preparation. Nontumor DNA was isolated from formalin-fixed paraffin-embedded noninvolved lymph node. Eight serial sections (5 μm) were taken from selected lymph node blocks, placed into a 1.5-mL tube, and deparaffinized using xylene. DNA was extracted using the QIAamp DNA minikit (Qiagen) following manufacturer's instructions.

MMP PCR. PCR was carried out in a 25 μL volume containing 25 ng of genomic DNA, 1 μL of each primer (2 mmol/L stock), 2 μL of 2.5 mmol/L deoxynucleotide triphosphate mix (Invitrogen), 2.5 μL of 10× PCR buffer, MgCl₂ (1.25 μL of a 50 mmol/L stock for MMP-7, MMP-12, and MMP-13 or 2.5 μL of a 50 mmol/L stock for MMP-1 and MMP-3), 0.2 μL of Platinum Taq DNA polymerase (5 units/μL; Invitrogen), and sterile deionized H₂O up to a final volume of 25 μL.

PCR conditions were as follows: initial denaturation for 2 min at 95°C; 40 cycles of denaturing for 30 s at 94°C, annealing for 30 s at a primer-specific temperature [MMP-1, forward 5′ TCGTGAGAATGTCTTCCCATT 3′ (HEX tagged)/reverse 5′ TTATCACTTCAGCACCTTATGGT 3′; MMP-7, forward 5′ CTGAATGATACCTATGAGAGCAGT 3′/reverse 5′ GCAGGAAGCACACAATGAATT 3′; the underlined base refers to an engineered base change in the primer, which substitutes a T for an A, thus generating an EcoRI restriction site; MMP-12, forward 5′ TGCTTGAGTGATGGACTAGAT 3′/reverse 5′ TTCTAGCCTAAGTTCCTGAACTGT 3′; MMP-13, forward 5′ TCGCCACGTAAGCATGTTTACCT 3′/reverse 5′ GACAAATCATCTTCATCACC 3′ all at 55°C; MMP-3, forward 5′ GGAATTCACATCACTGCCACCACT 3′ (FAM tagged)/reverse 5′ AGTGCTAGGATTACAGACATGGGTCA 3′ at 64°C; MMP-9, forward 5′ TGGTCAACGTAGTGAAACCCCATCT 3′/reverse 5′ TCCAGGCCCAATTATCACACTTAT 3′ at 67°C] and elongation for 30 s at 72°C; then 10 min at 72°C.

Genotyping. Genotype information for each gene is shown in Table 1. The MMP-1 and MMP-3 genotypes were determined by studying PCR product length by fluorescent genotyping. The forward PCR primers for MMP-1 and MMP-3 were labeled fluorescently with HEX and FAM, respectively. Each PCR product was diluted to 40 pg/μL; 2 μL of this was then added to 8 μL of Hi-Di and Rox size standard (ratio, 50:1; Applied Biosystems). The resulting 10 μL was then loaded onto an ABI 3700 automated sequencer, and the results were analyzed using Genescan and Genotyper packages (Applied Biosystems).

The MMP-7 (EcoRI), MMP-9 (SphI), MMP-12 (HpyCH4III), and MMP-13 (BsrI) genotypes were determined by PCR-RFLP assays. The restriction enzymes used for each gene are displayed in parentheses. All enzymes were obtained from New England Biosciences, and digestion was done in a 30-μL reaction volume following the manufacturer's instructions. After digestion, the products were resolved by electrophoresis on a 4% agarose gel and stained with ethidium bromide.

To determine the genotyping false call rate a 10% sample of randomly selected patients were regenotyped for each of the six MMPs (22).

Statistical analysis. Power and sample size was calculated using QUANTO version 1.2.1 (23). For the genes examined, in the Caucasian-only population, a sample size of 116 gives 80% power and a 0.05 level of significance. Differences in patients' characteristics and pathologic factors stratified by lymph node status were evaluated using χ² test. The associations between MMP genotype frequencies and lymph node status were estimated by computing odds ratios and 95% confidence intervals (95% CI) from both univariate and multivariate logistic regression analyses with adjustment for tumor grade. The association between genotype and disease-free and overall survival was investigated using the method of Kaplan-Meier and was estimated using the log-rank test. Cox regression models were used to adjust for potential confounders. All statistical analyses were two-sided and done using Minitab release 14.20 and STATA 9.2. We used Chaplin 1.2.2 (24, 25) and HAPSTAT (26, 27) to identify haplotypes that were associated with lymph node metastasis and Thesias (28, 29) to identify haplotypes that were associated with survival.

Cohort characteristics. We analyzed MMP promoter polymorphisms in a cohort of 221 patients, comprising 128 lymph node–negative and 93 lymph node–positive patients. This cohort consisted of mixed ethnicities, with Caucasian patients comprising 65% of the study group. The remaining 35% comprised 1 Pakistani, 1 Arabic, 2 Muslim, 2 Greek, 3 Turkish, 4 Asian, 4 Jewish, 18 African, and 47 other patients for whom ethnicity was not reported. Owing to the heterogeneous nature of the group, the cohort was analyzed as two groups: Caucasian-only and mixed ethnicity, which included all patients. Patient tumors were grade matched, with the exception of the grade I tumors, which were underrepresented in the lymph node–positive cases. Patients were also matched for estrogen receptor/progesterone receptor status and age. At the time of breast cancer diagnosis, lymph node–negative patients were between ages 33 and 88 years (mean, 61 ± 12) and lymph node–positive patients were between ages 36 and 77 years (mean, 57 ± 11). No associations were observed between estrogen receptor/progesterone receptor status or age and the occurrence of lymph node metastasis. Higher grade tumors were associated with a lymph node–positive status (P < 0.001), and there was a significant relationship between lymph node–positive status and reduced survival (P < 0.001). Characteristics of the study population are summarized in Table 2.

Association between individual genotypes and tumor spread. The genotype distributions of all MMPs in the mixed ethnicity group and for the selected Caucasian group of patients are shown in Table 3. All data were adjusted for tumor grade. The presence of polymorphisms within the promoters of MMP-1 and MMP-3, resulting from the insertion/deletion of a G or A, respectively, was determined by fluorescent genotyping. The SNPs for MMP-7, MMP-9, MMP-12, and MMP-13, wherein the SNP generated or eliminated a restriction enzyme recognition site, were identified by PCR-RFLP. The application of these approaches permitted the determination of an individual's genotype for each of these loci. The average genotyping success rate for the 221 patients was >90%, with a false call rate of <1%. With the exception of MMP-13, the distribution of genotypes did not deviate from Hardy-Weinberg equilibrium, which suggests that this patient cohort did not deviate genetically from the general population of the United Kingdom. In the whole cohort, regression analyses showed that patients carrying the 2G/2G genotype for MMP-1 showed a significant 3.9-fold (95% CI, 1.7-9.4; P < 0.001) increased risk of lymph node metastasis compared with subjects carrying the 1G/1G genotype (Table 3). Subjects carrying the C/T genotype for MMP-9 had a significant 3.6-fold (95% CI, 1.0-11.1; P < 0.001) increased risk of lymph node metastasis compared with C/C carriers (Table 3). There were no patients with T/T genotype. To minimize possible confounding issues due to allele frequency variation by race, separate analyses were done on the Caucasian-only group. This strengthened the association of the MMP-9 genotype with nodal status with the C/T genotype exhibiting a significant 9.1-fold (95% CI, 1.7-48.4; P = 0.01) increased risk of nodal invasion, whereas the association between the 2G/2G MMP-1 genotype and nodal status showed a borderline-significant increased risk of metastasis (2.6-fold; 95% CI, 1.0-6.9; P < 0.06). The genotype frequencies for MMP-3, MMP-7, MMP-12, and MMP-13 were not significantly different between node-negative and node-positive patients in either the mixed ethnicity or Caucasian-only group. Separate analyses were not done on individual ethnic groups because the numbers in each group were too small.

Association between individual genotypes and overall survival and disease-free survival. Using the Cox regression model (Table 4A), only the presence of the 2G/2G genotype for MMP-1 compared with the presence of the 1G/1G genotype was predictive of a shorter patient survival time [crude hazard ratio (CHR) for overall survival, 3.1; 95% CI, 1.1-8.7; P = 0.03; Fig. 1A]. None of the other MMP polymorphisms was associated significantly with survival. After adjusting for tumor grade and lymph node status, the MMP-1 polymorphism was no longer significantly associated with overall survival (adjusted hazard ratio for overall survival, 2.4; 95% CI, 0.8-6.6; P = 0.1; Fig. 1B; Table 4A). There was no significant association between individual MMP SNPs and disease-free survival (Table 4B).

Association between MMP haplotype and tumor behavior. Because MMP-1, MMP-3, MMP-7, MMP-12, and MMP-13 genes are clustered on chromosome 11q22.3, we addressed whether there were any associations between specific haplotypes and tumor behavior. Eleven common haplotypes were identified among both node-negative and node-positive patients, with the distribution of most being similar between the two patient groups (Table 5). However, two haplotypes did exhibit a significant association with lymph node metastasis: 2G-5A-A-A-A (2% frequency in lymph node–negative patients versus 10% in lymph node–positive; P = 0.02) and 2G-5A-G-A-A (8% in lymph node–negative patients versus 20% in lymph node–positive; P = 0.02). In addition, these two haplotypes (2G-5A-A-A-A: CHR 2.5, 95% CI 1.2-5.2, P = 0.02; 2G-5A-G-A-A: CHR 4.5, 95% CI 2.3-8.8, P = 0.00001) also showed a significant association with overall survival. After adjusting for tumor grade and lymph node status, the 2G-5A-G-A-A haplotype (adjusted hazard ratio, 2.6; 95% CI, 1.3-5.0; P = 0.006) was still associated with overall survival.

The haplotype 2G-5A-G-A-A was associated with disease-free survival in the unadjusted analysis (CHR, 2.7; 95% CI, 1.08-6.70; P = 0.03) but was no longer significant after adjusting for lymph node status and grade (CHR, 1.9; 95% CI, 0.68-5.45; P = 0.21).

MMPs have been implicated in tumor progression, and high levels of expression have been related to more aggressive tumor behavior and poorer patient prognosis (20, 21). This study, therefore, investigated the association between functional SNPs in MMP genes, previously implicated in breast cancer, and breast cancer spread and patient survival. In the genotype analysis, wherein each MMP SNP was studied separately in the mixed ethnicity patient cohort, the 2G/2G genotype of MMP-1, which generates an increased level of gene expression (13), was more frequent in the lymph node–positive patients and conferred a 3.9-fold increased risk of lymph node metastasis. The C/T genotype of MMP-9 was found to confer a 3.6-fold increased occurrence of lymph node metastasis. Interestingly, there is a change in the association pattern when analysis is restricted to the Caucasian-only group, with the MMP-1 2G/2G genotype risk decreasing to 2.6-fold, whereas the MMP-9 C/T genotype confers a 9.1-fold increase in risk of metastasis in this group. Such differences between different ethnic groups may, in some part, explain the discrepancies reported between associations in the literature and the work described herein. Thus, although Przybylowska et al. (3) reported a similar association between MMP-1 2G/2G genotype and lymph node metastasis in patients with breast cancer, they showed no such association with the MMP-9 T allele. This could be a consequence of variation in the ethnicity of the patients from the two cohorts. In our study, when the Caucasian patients were subtracted from the ethnically mixed group, and the latter analyzed, there was no association between the T allele and node status when corrected for grade (P = 0.96; odds ratio, 1.05; 95% CI, 0.18-5.93). Similarly, in colorectal cancer, whereas a relationship between the MMP-1 2G/2G genotype and poor prognosis has been identified in an Italian cohort of patients with colorectal cancer (30), no such relationship was identified in a recent large Australian study (31).

Previous studies have identified links between promoter polymorphisms in MMP-1 and MMP-9 (15, 32) and increased expression of enzyme levels. In addition, increased expression for both MMP-1 and MMP-9 has been associated with lymph node metastasis (33) and poor prognosis (3, 34, 35) in breast cancer. In keeping with this, our data and others' point to the potential prognostic role of MMP-1 and MMP-9 polymorphisms in identifying patients at risk of more aggressive disease.

Having identified a link between genotype and node status, the analysis was extended to determine any association between genotype, disease-free survival, and overall survival. Previous studies have identified such links for both MMP-2 (36) and MMP-12 (37) in non–small cell lung cancer, between MMP-7 and gastric cancer (38) and between MMP-1 and colorectal cancer (39). In this study, analysis was restricted to the Caucasian-only group. No association between MMP genotype and disease-free survival was identified; however, initial analysis showed a significant association between the MMP-1 2G/2G genotype and overall survival. However, after adjustment for node status, this association was lost. This is in keeping with our hypothesis that elevated levels of MMPs promote spread of breast cancer. Thus, the high-expressing MMP-1 2G/2G genotype promotes lymph node metastasis, and prognosis is dependent on this facet of tumor behavior. Interestingly, even when adjusted for lymph node status, there remains a trend for decreased overall survival in patients with the 2G/2G genotype compared with 1G/2G or 1G/1G, which suggests that the enzyme may play a role in distant tumor dissemination. There was no association between MMP-9 T allele and overall survival despite the relationship between this allele and lymph node metastasis, which is difficult to explain, apart from acknowledging that this attests to the complexity of function of these enzymes and the process of tumor spread.

Given the likely multifaceted role of MMPs in tumor progression and the involvement of different enzymes at different stages of the disease (6, 34), it is plausible that the combined effect of multiple high-expressing MMP alleles may provide a stronger predictor of prognosis than individual SNPs. We therefore investigated disease behavior with MMP SNP haplotypes. This identified two haplotypes, 2G-5A-A-A-A and 2G-5A-G-A-A, linked to increased incidence of lymph node metastasis. The 2G-5A-A-A-A haplotype was shown to have a dominant effect (subjects with one copy of the haplotype have the same risk of lymph node metastasis as those subjects with two copies), whereas the 2G-5A-G-A-A haplotype exhibited a multiplicative effect, such that subjects with one copy of the haplotype are at an intermediate risk of lymph node metastasis, with respect to those subjects with zero or two copies. Furthermore, the 2G-5A-G-A-A haplotype was also associated with overall survival even after adjustment for lymph node status. Whereas this haplotype was associated with disease-free survival in unadjusted analysis, this relationship was lost when adjusted for lymph node status. In keeping with the genotype data, these haplotype associations are driven by the MMP-1 2G allele. The haplotypes also reflect the previously described linkage disequilibrium between the MMP-1 2G allele and the MMP-3 5A allele (40).

Polymorphisms in the promoters of MMPs have a functional role and have been shown to exhibit allelotypic effects on levels of gene transcript (15, 32). Although the lifelong effects of altered transcription are unclear, overexpression of MMPs caused by promoter polymorphisms may enhance cancer progression by virtue of their role in degradation of the ECM, thereby generating an environment that favors tumor cell migration. Modulation of the microenvironment is emerging as a critical factor in determining tumor cell behavior (41, 42), and changes in the stroma, such as those mediated through proteolytic remodeling, can have a major influence on tumor cell function (43). Thus, the altered expression of MMPs generated through functional SNPs and the consequent effect on tumor microenvironment may contribute to the patient-to-patient variability in breast cancer susceptibility and outcome. In patients with breast cancer, knowledge of the MMP SNP genotype and/or haplotype, as shown here, could contribute to identification of patients at higher risk and therefore influence treatment decisions.