Gene–Gene Interactions Between Glutathione S-Transferase M1 and Matrix Metalloproteinases 1, 9, and 12 in Chronic Obstructive Pulmonary Disease in Serbians

ABSTRACT

Chronic obstructive pulmonary disease (COPD) is a complex disorder influenced by multiple genetic and environmental factors, as well as their interactions. Since elevated oxidative stress and protease activity characterize the pathogenesis of COPD, variants of genes that can affect these processes have been commonly studied in COPD. However, interactions among genes that can influence oxidative stress and protease activity remain poorly investigated in COPD. The aim of this study was to look into the role of functional variants in matrix metalloproteinases (MMPs) 1, 9, and 12 in the occurrence and/or modulation of COPD, and to analyze their interactions with glutathione S-transferases (GSTs) M1, T1, and P1 in the pathogenesis of COPD in Serbians. The MMP1 rs1799750 G > GG, MMP9 rs3918242 C > T, and MMP12 rs2276109 A > G variants were analyzed by direct detection methods. Gene–gene interactions between variants in MMPs and GSTs were assessed using a case-control model. Our results showed association of the MMP1 GG/GG genotype with COPD (p = 0.036, OR = 2.50). Gene–gene interactions between the GSTM1 null and MMP1 GG (p = 0.028, OR = 2.99) and the GSTM1 null and MMP12 AA variants (p = 0.015, OR = 3.82) were found to significantly increase the risk of COPD occurrence. Furthermore, the MMP12 G variant was found to modify the age of COPD onset (p = 0.025, OR = 3.30), while interaction between the GSTM1 null and MMP9 T variants was found to modify the severity of disease (p = 0.019, OR = 4.83). To our best knowledge, this is the first study revealing several gene–gene interactions affecting oxidative stress and protease activity in the pathogenesis of COPD.

KEYWORDS:

• Pulmonary disease
• gene–gene interactions
• functional variants
• oxidative stress
• extracellular matrix remodeling

Introduction

Chronic obstructive pulmonary disease (COPD) is a chronic respiratory disorder characterized by chronic obstruction of lung airflow that interferes with normal breathing and is not fully reversible (1). This disease may lead to chronic respiratory failure, and it is responsible for early mortality, high death rates, and significant cost to health systems. Tobacco smoking is a major risk factor for the development of COPD. Prevalence of COPD was estimated to be around 10%, but it varies according to country, age, and sex (2).

The pathogenesis of COPD is complex and includes a number of processes (3). The major component is abnormal inflammatory response to the inhalation of toxic substances, such as active and passive tobacco smoke and urban and rural air pollution. The chronic inflammatory response against such substances leads to mucus hypersecretion, airway remodeling, and alveolar destruction. Among several processes relevant for COPD pathogenesis, extracellular matrix (ECM) destruction, and oxidative stress could be of great importance as they can cause significant tissue injury and hence contribute to the deterioration of the initial inflammatory damage. Excessive ECM destruction, as a consequence of proteinases-antiproteinases imbalance in the lung, can amplify lung inflammation and injury (4). Oxidative stress, oxidants-antioxidants imbalance, can cause additional lung tissue damage, activation of proteinases, and inactivation of antiproteinases leading to ECM destruction (3).

Glutathione S-transferases (GSTs) are cytosolic enzymes with a major role in biological detoxification processes based on their ability to catalyze the conjugation of the reduced form of glutathione to xenobiotic substrates (5). Since airways are exposed to various potentially harmful substances, the function of GSTs is of particular importance for the lung (6). Commonly studied GST molecules are involved in the deactivation of carcinogenic intermediates of polycyclic aromatic hydrocarbons (GSTM1), phase-II biotransformation of a number of chemicals, such as hydrocarbons and halogenated hydrocarbons (GSTT1) and metabolism of hydrophobic and electrophilic compounds (GSTP1). Several lines of evidence suggest that GSTM1 and GSTT1 play a role in the deactivation of reactive oxygen species that trigger the oxidative stress and are likely to be involved in cellular processes underlying COPD, while GSTP1 is the most abundant GST in the human lungs (7). In spite of their importance, GSTM1 and GSTT1 are often genetically deleted in a high percentage of the human population, and the variant Ile106Val that affects the substrate specificity of GSTP1 is also relatively frequent. Functional variants GSTM1 null, GSTT1 null, and GSTP1 Ile105Val are considered to be a risk for COPD, and thus they have been extensively studied, but results remain controversial (8).

Matrix metalloproteinases (MMPs) are zinc-dependent endopeptidases with major role in lung tissue due to their main function, simultaneous degradation of ECM by MMP family members with different substrate specificities (9). They also participate in a number of other processes important for lung homeostasis, including inflammation, mucus hypersecretion, proliferation, apoptosis, and profibrotic pathways. The expression and activity of MMP1, 9, and 12 are elevated in COPD (10–13). Overexpression of MMPs can lead to excessive ECM degradation, and subsequently to the obstructive lung diseases, including COPD (9). Functional variants MMP1 rs1799750 G > GG, MMP9 rs3918242 C > T, and MMP12 rs2276109 A > G are known to alter the gene expression (14–16). The role of these functional variants has been extensively investigated in COPD, but the results are largely inconsistent (17–24).

Since COPD is a complex disease, influenced by multiple genetic and environmental factors, identification of gene–gene interactions may contribute to a greater understanding of the mechanisms involved in the pathogenesis of COPD. Common functional variants of genes involved in oxidants-antioxidants and proteinases-antiproteinases balance have been commonly investigated in COPD, but only for the genes involved in either oxidants-antioxidants or proteinases-antiproteinases balance (20–29). However, interactions among genes that can affect these two processes remain poorly studied in COPD, even though they are related (7,30–32). The aim of this study was to analyze the influence of functional variants in MMP1, 9, and 12 on the occurrence and/or modulation of COPD in Serbian population. Also, we hypothesized influence of gene–gene interactions between genes involved in oxidants-antioxidants and proteinases-antiproteinases balance on the pathogenesis of COPD. Hence, interactions between two chosen gene families, representing these processes, were analyzed by employing functional variants of MMP1, 9, and 12, and GSTM1, T1, and P1, relying on our previously published data (25,33,34).

Material and methods

Case-control study

The patient and control subjects enrolled in this study were previously described (25). Briefly, the patient group consisted of 122 subjects with COPD diagnosis established based on medical history, physical examination, pulmonary function tests, blood gas analyses, and chest radiography, according to GOLD standard (35).

The control group encompassed 100 unrelated subjects from the same geographic area, without clinical evidence of COPD and with normal values of pulmonary function tests.

The study was approved by the local Ethics Committee and informed consent was obtained from each participant.

The data of GSTM1, T1, and P1 genotypes for controls and patients, as well as of MMP1 and 9 for controls, have been already published (25,33,34). The data were employed in this study in order to analyze gene–gene interactions among GSTM1, T1, and P1 with MMP1, 9, and 12.

Determination of genotypes

Genomic DNA was extracted from whole blood using a GFX Genomic Blood DNA Purification Kit (Amersham Biosciences).

The MMP1 rs1799750 (−1607G > GG) variant was detected using the polymerase chain reaction (PCR) followed by the conformation sensitive gel electrophoresis (CSGE), as previously published (33). Briefly, a fragment containing the MMP1 G-1607GG deletion/insertion variant was amplified using primers TGAGAAGAGGATTTCCTTTTCGTG and GGATTCCTGTTTTCTTTCTGCGTC. The PCR products of 199 bp and 200 bp corresponding to the −1607G and the −1607GG allele, respectively, were distinguished after separation by CSGE and visualization by the silver staining.

The MMP9 rs3918242 (−1562C > T) variant was distinguished using the PCR-restriction fragment length polymorphism (RFLP) method as previously described (14). In brief, a fragment encompassing the MMP9 C-1562T variant was amplified using primers GCCTGGCACATAGTAGGCCC and CTTCCTAGCCAGCCGGCATC. The PCR products digested with PaeI (Fermentas) were separated on 2% agarose gel and the −1562C (435 bp) and the −1562T (247 bp and 188 bp) alleles were visualized by ethidium bromide staining.

Detection of the MMP12 rs2276109 (−82A > G) variant was enabled using the PCR-mediated site-directed mutagenesis (PSM) with subsequent RFLP method, as previously published (26). Briefly, a fragment containing the MMP12 A-82G variant was amplified using primers GAGATAGTCAAGGGATGATATCAGC and AAGAGCTCCAGAAGCAGTGG. After digestion with PvuII (Fermentas), the PCR products were separated on 3% agarose gel and the −82A (199 bp) and the −82G (175 bp and 24 bp) alleles were visualized by ethidium bromide staining.

Genotypes were scored without knowledge of the sample phenotypes by two^{Table 1} independent observers.

Statistical analysis

Basic characteristics, clinical parameters, genotype and allele frequencies of patient, and control subjects are expressed as a percentage or mean ± standard error of mean (SEM). The Hardy–Weinberg equilibrium of the genotype distributions was assessed for each cohort using the χ²-test. The normality of distribution for continuous variables was assessed using the Shapiro–Wilk test. A comparison of categorical and continuous variables was performed using the χ²-test and the Mann–Whitney U-test, respectively. Association of genotypes and alleles with COPD occurrence, severity or early-onset, was tested using the odds ratio (OR) with the 95% confidence interval (95% CI) calculated by binary logistic regression. The outcome variable was adjusted for potential confounding factors. Gene–gene interaction was estimated by calculating effect for each risk allele and their joint effect relative to nonrisk genotype (reference category) in a case-control model using approach described by Ottman (36). Interaction is evaluated on a multiplicative scale as the departure of joint effects from multiplicative ORs (36,37). As variants relating to preestablished hypothesis were analyzed adjustment for multiple testing was not applied (38). A p value of less than 0.05 was considered significant. Statistical analysis was performed using the Statistical Package for the Social Sciences (SPSS, Version 20).

Results

The basic characteristics of COPD group and controls included: age of disease onset/age (45.0 ± 1.6 versus 50.9 ± 1.4 years), gender (70.2% versus 35.0% males), and smoking status (69.7% versus 50.0% smokers; 27.4 ± 1.7 versus 26.3 ± 2.3 pack-years) and clinical parameters included: forced expiratory volume in 1 second (FEV₁, % of predicted) (45.3 ± 2.3 versus 110.1 ± 1.7) and FEV₁/forced vital capacity (62.4 ± 2.0 versus 98.3 ± 0.9), respectively (25). The differences in age, gender, and smoking status among patients and controls were used for adjustment of the outcomes.

The distributions of the MMP1 rs1799750, MMP9 rs3918242, and MMP12 rs2276109 alleles and genotypes in COPD and control group are listed in Table 1. The observed genotype distributions have confirmed the expectations of Hardy–Weinberg equilibrium.

Table 1. Distribution of MMP1, MMP9, and MMP12 Gene Variants in COPD Patients and Controls.

Gene, reference SNP ID number	Genotype, allele	COPD, % (N = 122)	Controls, % (N = 100)	p, OR (95.5% CI)^a	p, OR (95.5% CI)^b
MMP1 rs1799750	G/G	23.8	31.0	1	1
	G/GG	46.7	51.0	0.581, 1.19 (0.63–2.25)	0.598, 1.21 (0.60–2.45)
	GG/GG	29.5	18.0	0.049, 2.14 (1.00–4.57)	0.036, 2.50 (1.06–5.91)
	GG	52.9	43.5	0.049, 1.46 (1.00–2.12)	0.037, 1.57 (1.03–2.40)
MMP9 rs3918242	C/C	67.2	72.0	1	1
	C/T	31.2	24.0	0.283, 1.39 (0.76–2.54)	0.296, 1.43 (0.73–2.82)
	T/T	1.6	4.0	0.350, 0.44 (0.08–2.47)	0.894, 0.87 (0.11–6.63)
	T	17.2	16.0	0.733, 1.09 (0.66–1.81)	0.438, 1.25 (0.71–2.23)
MMP12 rs2276109	A/A	82.0	73.0	1	1
	A/G	15.6	24.0	0.110, 0.58 (0.29–1.13)	0.094, 0.53 (0.25–1.11)
	G/G	2.4	3.0	0.705, 0.73 (0.14–3.72)	0.612, 0.60 (0.09–4.23)
	G	10.2	15.0	0.132, 0.65 (0.37–1.14)	0.100, 0.59 (0.31–1.11)

COPD, chronic obstructive pulmonary disease; MMP, matrix metalloproteinase; and SNP, single nucleotide polymorphism.

^a Crude.

^b Adjusted for age, gender, and smoking status.

A significantly higher frequency of the MMP1 rs1799750 GG variant was found in patients than in controls (52.9% versus 43.5%, p = 0.037) as shown in Table 1. Whereas the frequency of heterozygotes was similar among patients and controls (46.7% versus 51.0%, respectively), a significant increase of GG homozygotes was detected in patients compared to controls (29.5% versus 18.0%, p = 0.036). According to our results, homozygotes for the GG variant have 2.5-fold higher risk of the occurrence of COPD, as compared to homozygotes for the G variant.

Our first data of the MMP9 rs3918242 variant showed lack of association with COPD, that is confirmed by this study, conducted on enlarged group of patients (Table 1) (34).

The distribution of the MMP12 rs2276109 variant was not significantly different between the groups, but a trend of decrement of G variant carriers was noticed in the patient group as compared to controls (18.0% versus 27.0%, respectively), Table 1.

In order to analyze interactions among the genes that can influence oxidative stress and protease activity in the pathogenesis of COPD, we combined previously published data regarding GST's M1, T1, and P1 variants with MMP's 1, 9, and 12 variants reported here, as shown in Table 2 (25). For all variants the minor allele was considered as the risk allele, except for MMP12 A > G where the major allele was taken as the risk allele. Gene–gene interactions were estimated using the dominant model, with the exception of MMP12 where the AA genotype was taken as the risk genotype.

Table 2. Gene–Gene Interactions Between GSTs M1, T1, and P1 and MMPs 1, 9, and 12 in COPD.

Gene–gene interaction	Genotype	COPD, % (N = 122)	Controls, % (N = 100)	p, OR (95.5% CI)^a	p, OR (95.5% CI)^b
GSTM1-MMP1	wt-G/G	9.0	19.0	1	1
	null-G/G	14.8	13.0	0.097, 2.39 (0.85–6.70)	0.210, 2.08 (0.66–6.51)
	wt-GG/	29.5	34.0	0.178, 1.83 (0.76–4.40)	0.186, 1.96 (0.72–5.30)
	null-GG/	46.7	34.0	0.015, 2.90 (1.23–6.81)	0.028, 2.99 (1.12–7.97)
GSTT1-MMP1	wt-G/G	18.0	25.0	1	1
	null-G/G	5.8	7.0	0.834, 1.14 (0.34–3.75)	0.666, 1.35 (0.35–5.18)
	wt-GG/	59.8	51.0	0.158, 1.63 (0.83–3.20)	0.107, 1.86 (0.87–3.96)
	null-GG/	16.4	17.0	0.510, 1.34 (0.56–3.17)	0.397, 1.53 (0.57–4.14)
GSTP1-MMP1	Ile/Ile-G/G	9.0	13.0	1	1
	Val/-G/G	14.8	19.0	0.830, 1.12 (0.40–3.13)	0.683, 0.79 (0.25–2.50)
	Ile/Ile-GG/	30.3	32.0	0.511, 1.37 (0.54–3.47)	0.847, 1.11 (0.39–3.12)
	Val/-GG/	45.9	36.0	0.188, 1.84 (0.74–4.55)	0.258, 1.79 (0.65–4.93)
GSTM1-MMP9	wt-C/C	28.7	38.0	1	1
	null-C/C	38.5	34.0	0.211, 1.50 (0.79–2.84)	0.254, 1.52 (0.74–3.13)
	wt-T/	9.8	15.0	0.756, 0.87 (0.36–2.11)	0.782, 1.15 (0.42–3.12)
	null-T/	23.0	13.0	0.038, 2.34 (1.05–5.22)	0.072, 2.29 (0.93–5.63)
GSTT1-MMP9	wt-C/C	52.5	56.0	1	1
	null-C/C	14.7	16.0	0.968, 0.98 (0.46–2.11)	0.804, 1.12 (0.46–2.70)
	wt-T/	25.4	20.0	0.370, 1.36 (0.70–2.64)	0.247, 1.57 (0.73–3.35)
	null-T/	7.4	8.0	0.976, 0.98 (0.36–2.72)	0.942, 1.04 (0.32–3.40)
GSTP1-MMP9	Ile/Ile-C/C	24.6	32.0	1	1
	Val/-C/C	42.6	40.0	0.322, 1.39 (0.73–2.65)	0.771, 1.11 (0.53–2.32)
	Ile/Ile-T/	14.8	13.0	0.380, 1.48 (0.62–3.52)	0.945, 1.03 (0.39–2.71)
	Val/-T/	18.0	15.0	0.287, 1.56 (0.69–3.57)	0.134, 2.11 (0.79–5.63)
GSTM1-MMP12	wt-G/	5.7	17.0	1	1
	null-G/	12.3	10.0	0.033, 3.64 (1.11–11.97)	0.129, 2.73 (0.75–9.95)
	wt-A/A	32.8	36.0	0.049, 2.70 (1.00–7.25)	0.083, 2.66 (0.88–8.02)
	null-A/A	49.2	37.0	0.006, 3.94 (1.49–10.40)	0.015, 3.82 (1.29–11.33)
GSTT1-MMP12	wt-G/	11.5	20.0	1	1
	null-G/	6.5	7.0	0.432, 1.63 (0.48–5.55)	0.573, 1.51 (0.36–6.25)
	wt-A/A	66.4	56.0	0.062, 2.07 (0.96–4.43)	0.074, 2.17 (0.93–5.06)
	null-A/A	15.6	17.0	0.332, 1.60 (0.62–4.11)	0.241, 1.91 (0.65–5.62)
GSTP1-MMP12	Ile/Ile-G/	8.2	11.0	1	1
	Val/-G/	9.8	16.0	0.740, 0.82 (0.26–2.57)	0.435, 0.60 (0.167–2.16)
	Ile/Ile-A/A	31.2	34.0	0.677, 1.23 (0.46–3.25)	0.884, 1.08 (0.37–3.16)
	Val/-A/A	50.8	39.0	0.247, 1.75 (0.68–4.50)	0.293, 1.75 (0.62–5.00)
MMP1-MMP9	G/G-C/C	14.7	25.0	1	1
	GG-C/C	52.5	47.0	0.080, 1.89 (0.93–3.86)	0.051, 2.21 (0.99–4.922)
	G/G-T/	9.0	7.0	0.174, 2.18 (0.71–6.72)	0.111, 2.80 (0.79–9.94)
	GG-T/	23.8	21.0	0.122, 1.92 (0.84–4.38)	0.081, 2.29 (0.90–5.83)
MMP1-MMP12	G/G-G/	6.5	14.0	1	1
	GG/-G/	11.5	13.0	0.280, 1.88 (0.60–5.96)	0.141, 2.60 (0.73–9.26)
	G/G-A/A	17.2	18.0	0.192, 2.04 (0.70–5.97)	0.096, 2.74 (0.84–9.00)
	GG/-A/A	64.8	55.0	0.053, 2.51 (0.99–6.40)	0.023, 3.33 (1.18–9.45)
MMP9-MMP12	C/C-G/	10.6	22.0	1	1
	T/-G/	7.4	5.0	0.091, 3.05 (0.84–11.07)	0.051, 4.39 (0.99–19.42)
	C/C-A/A	56.6	50.0	0.032, 2.33 (1.07–5.08)	0.020, 2.81 (1.18–6.70)
	T/-A/A	25.4	23.0	0.064, 2.28 (0.95–5.46)	0.041, 2.76 (1.04–7.32)

COPD, chronic obstructive pulmonary disease; GST, glutathione S-transferase; and MMP, matrix metalloproteinase.

^a Crude.

^b Adjusted for age, gender, and smoking status.

Gene–gene interaction between GSTM1 and MMP1 was identified in COPD occurrence (Table 2). A significantly higher frequency of the double risk genotype, GSTM1 null and MMP1 rs1799750 GG, was detected in patients than in controls (46.7% versus 34.0%, p = 0.028). Interaction between the GSTM1 null and MMP1 GG variants was less than multiplicative (2.39 × 1.83 > 2.90), showing significant increase of almost threefold in the risk of COPD, as compared to the reference nonrisk genotype, GSTM1 wt and MMP1 G, which was underrepresented in patients.

Interaction between the GSTM1 and MMP9 genes was detected in COPD (Table 2). The genotype composed of two risk variants, GSTM1 null and MMP9 rs3918242 T, was markedly overrepresented in patients compared to controls (23.0% versus 13.0%, p = 0.038). Interaction between GSTM1 and MMP9 was more than multiplicative (1.50 × 0.87 < 2.34), but should be interpreted with caution considering that the significance was lost after adjusting for potential confounders (p = 0.072).

Gene–gene interaction among GSTM1 and MMP12 was identified in the occurrence of COPD (Table 2). Significantly higher frequency of the double risk genotype, GSTM1 null and MMP12 rs2276109 AA, was found in the group of patients compared to controls (49.2% versus 37.0%, p = 0.015). The double risk genotype of GSTM1 and MMP12 increased the risk factor of the occurrence of COPD 3.82-fold, as compared to the reference nonrisk genotype, GSTM1 wt and MMP12 G, which was underrepresented in COPD group. The interaction between GSTM1 and MMP12 variants was less than multiplicative (3.64 × 2.70 > 3.94).

Furthermore, we examined weather the MMP variants, as well as significant interactions between GSTM1 and MMPs can act as modifiers of COPD affecting the severity or the age of onset of the disease. The influence of variants on COPD severity was estimated among subgroups of patients with severe (FEV1 < 50) and mild/moderate (FEV1 ≥ 50) disease, while the influence of variants on the age of disease onset was estimated among subgroups of patients with COPD onset before (<40 years) and after (≥40 years) the age of 40 years, as shown in Table 3. After testing the variants MMP1 G > GG, MMP9 C > T, and MMP12 A > G no association with the severity of the disease was found. However, MMP12 A > G was found to influence the age of onset of COPD, as the G variant was significantly higher represented in patients with the disease onset before the age of 40 years (32.4% versus 12.5%, p = 0.025). Our results indicate that the presence of G variant increase the chance of early onset of COPD 3.3-fold, compared to the AA genotype, Table 3. Moreover, we tested the mean values of the age of disease onset in relation to MMP12 A > G and confirmed significantly lower mean value of the age of COPD onset in patients with G variant compared to AA genotype (37.3 ± 4.2 versus 46.7 ± 1.6, p = 0.048).

Table 3. The MMPs 1, 9, and 12 Variants in the Severity and Onset of COPD.

		FEV1, % pred				COPD onset, years
Gene	Genotype	<50, %	≥50, %	p, OR (95.5% CI)^a	p, OR (95.5% CI)^b	<40, %	≥40, %	p, OR (95.5% CI)^a	p, OR (95.5% CI)^c
		82 (67.2)	40 (32.8)			34 (27.7)	88 (72.3)
MMP1 rs1799750	G/G	23.2	25.0	1	1	14.7	27.3	1	1
	GG/	76.8	75.0	0.824, 1.10 (0.46–2.67)	0.934, 1.04 (0.42–2.59)	85.3	72.7	0.150, 2.17 (0.75–6.27)	0.057, 3.15 (0.97–10.23)
MMP9 rs3918242	C/C	63.4	75.0	1	1	64.7	68.2	1	1
	T/	36.6	25.0	0.203, 1.73 (0.74–4.03)	0.109, 2.09 (0.85–5.17)	35.3	31.8	0.714, 1.17 (0.51–2.69)	0.762, 1.15 (0.47–2.83)
MMP12 rs2276109	A/A	80.5	85.0	1	1	67.6	87.5	1	1
	G/	19.5	15.0	0.544, 1.37 (0.49–3.83)	0.531, 1.41 (0.48–4.19)	32.4	12.5	0.013, 3.35 (1.29–8.71)	0.025, 3.30 (1.16–9.40)

COPD, chronic obstructive pulmonary disease; FEV¹, forced expiratory volume in 1 second; and MMP, matrix metalloproteinase.

^a Crude.

^b Adjusted for age, gender, and smoking status.

^c Adjusted for gender and smoking status.

Gene–gene interactions identified in this study also demonstrated the influence on the severity and the age of onset of COPD, as presented in Table 4. Interaction between GSTM1 and MMP9 was identified in severe COPD. The frequency of the double risk genotype, GSTM1 null and MMP9 rs3918242 T, was significantly higher in severe COPD compared to mild/moderate disease (28.1% versus 12.5, respectively, p = 0.019). Interaction between GSTM1 and MMP9 was more than multiplicative (1.60 × 1.05 < 3.45) with 4.83-fold increase in the risk of severe COPD, as compared to the reference nonrisk genotype, GSTM1 wt and MMP9 CC. Furthermore, we analyzed interaction between GSTM1 and MMP9 by testing the mean values of FEV₁ among patients. Genotypes wt-C/C, null-C/C, wt-T/, and null-T/ had respective FEV₁ 50.7 ± 5.0, 45.9 ± 3.7, 44.6 ± 4.8, and 37.9 ± 4.0, confirming significantly lower FEV₁ (p = 0.036) in patients with GSTM1 null and MMP9 T variants compared to the reference nonrisk genotype (GSTM1 wt and MMP9 CC).

Table 4. Gene–Gene Interactions Among GSTM1 and MMPs 1, 9, and 12 in the Severity and Onset of COPD.

		FEV(1), % pred				COPD onset, years
Gene–gene interaction	Genotype	<50, %	≥50, %	p, OR (95.5% CI)^a	p, OR (95.5% CI)^b	<40, %	≥40, %	p, OR (95.5% CI)^a	p, OR (95.5% CI)^c
		82 (67.2)	40 (32.8)			34 (27.7)	88 (72.3)
GSTM1-MMP1	wt-G/G	7.3	12.5	1	1	5.9	10.2	1	1
	null-G/G	15.9	12.5	0.335, 2.17 (0.45–10.44)	0.281, 2.47 (0.48–12.79)	8.8	17.1	0.917, 0.90 (0.12–6.46)	0.833, 1.25 (0.15–10,36)
	wt-GG/	25.6	37.5	0.824, 1.17 (0.30–4.54)	0.835, 1.17 (0.27–4.96)	41.2	25.0	0.218, 2.86 (0.54–15.25)	0.084, 5.01 (0.80–31,28)
	null-GG/	51.2	37.5	0.210, 2.33 (0.62–8.78)	0.209, 2.47 (0.60–10.10)	44.1	47.7	0.571, 1.61 (0.31–8.30)	0.244, 2.88 (0.49–17,08)
GSTM1-MMP9	wt-C/C	24.4	37.5	1	1	38.2	25.0	1	1
	null-C/C	39.0	37.5	0.31, 1.60 (0.64–3.97)	0.269, 1.74 (0.65–4.61)	26.5	43.2	0.073, 0.40 (0.15–1.09)	0.132, 0.44 (0.15–1.28)
	wt-T/	8.5	12.5	0.943, 1.05 (0.28–3.96)	0.683, 1.33 (0.33–5.35)	8.8	10.2	0.447, 0.56 (0.13–2.47)	0.411, 0.51 (0.10–2.54)
	null-T/	28.1	12.5	0.039, 3.45 (1.06–11.16)	0.019, 4.83 (1.29–18.01)	26.5	21.6	0.679, 0.80 (0.28–2.29)	0.814, 0.87 (0.28–2.72)
GSTM1-MMP12	wt-G	3.7	0.0	1	1	14.7	2.3	1	1
	null-G	15.8	5.0	0.045, 8.67 (1.05–71.57)	0.038, 10.04 (1.14–88.46)	17.6	10.2	0.181, 0.27 (0.04–1.85)	0.196, 0.25 (0.03–2.03)
	wt-A/A	29.3	40.0	0.403, 2.00 (0.39–10.16)	0.377, 2.16 (0.39–1.91)	32.4	33.0	0.038, 0.15 (0.03–0.90)	0.041, 0.14 (0.02–0.92)
	null-A/A	51.2	45.0	0.163, 3.11 (0.63–15.34)	0.145, 3.54 (0.65–19.43)	35.3	54.5	0.010, 0.10 (0.02–0.58)	0.019, 0.11 (0.02–0.69)

COPD, chronic obstructive pulmonary disease; FEV¹, forced expiratory volume in 1 second; GST, glutathione S-transferase; and MMP, matrix metalloproteinase.

^a Crude.

^b Adjusted for age, gender and smoking status.

^c Adjusted for gender and smoking status.

Interaction between GSTM1 and MMP12 was found to modify the age of COPD onset and represents protective factor for early disease onset in patients with the GSTM1 null and MMP12 rs2276109 AA variants (OR = 0.11, p = 0.019), compared to the reference nonrisk genotype (GSTM1 wt and MMP12 G), Table 4. We further tested the GSTM1 and MMP12 genotype combinations wt-G/, null-G/, wt-A/A, and null-A/A with respective mean values of the age of disease onset 27.4 ± 6.9, 42.2 ± 4.8, 44.4 ± 2.9, and 48.2 ± 1.8 and found significantly higher mean age of the disease onset (p = 0.007) in patients with the GSTM1 null and MMP12 AA variants, compared to the reference genotype (GSTM1 wt and MMP12 G).

Discussion

The main finding of this study is that gene–gene interactions among GSTM1 and MMP1, 9, and 12 may be involved in the pathogenesis of COPD. Two interactions, between the GSTM1 null and MMP1 GG and between the GSTM1 null and MMP12 AA variants were found to significantly increase the risk of COPD occurrence. Additionally, the MMP1 GG variant showed a sole association with COPD. Furthermore, interaction between the GSTM1 null and MMP9 T variants, as well as MMP12 G variant were found to modify the severity and the age of onset of COPD, respectively.

Our results showed the association of the MMP1 GG/GG genotype with 2.5-fold higher risk for COPD occurrence. The G > GG variant is associated with higher promoter activity, as well as with elevated mRNA and protein levels of MMP1 (15). In the lungs of the COPD subjects, expression and activity of MMP1 is upregulated and inversely correlated with the severity of the disease (10,11). Previous studies did not identify relation between the MMP1 G > GG variant and COPD, but in a meta-analysis slight, significant increase in the risk (OR = 1.2) for COPD was noticed in Caucasians (20,21,23,24,39). However, other meta-analyses could not confirm the link of the GG variant and COPD (40,41). Additionally, our study identified gene–gene interaction between the GSTM1 null and MMP1 GG variants in the pathogenesis of COPD. The GSTM1 is thought to have an important role in the response and regulation of oxidative stress in the lungs (7). A meta-analysis reported significant association of the GSTM1 null variant with COPD in Caucasians (42). Moreover, it is suggested that GSTM1 acts through interactions with environmental or genetic factors (7,27,28,42). We previously reported the GSTM1 null variant as a sole risk of 1.8 for COPD, while in this study the MMP1 GG/GG genotype represents a sole risk of 2.5 for COPD (25). These two genetic risk factors significantly associate in the pathogenesis of COPD, augmenting the risk factor to almost threefold in carriers of the double risk genotype (Table 2). The deficiency of GSTM1 may disturb oxidants-antioxidants balance, while oxidative stress can activate MMP1 and, together with the MMP1 GG variant, intensify ECM degradation (30). In previous studies, the G > GG variant was found to have an allele specific response to cigarette smoke and diesel exhaust particles, implicating the importance of gene–environment interactions in the regulation of MMP1 expression (43,44). Until now, no interaction between GSTM1 and MMP1 in the pathogenesis of COPD was reported.

According to our previous results, as well as results in this study, the MMP9 C > T variant is not the risk factor for COPD (34). The C > T variant is characterized by greater promoter activity of the T allele and investigation of its role in the pathogenesis of COPD yielded controversial results (14). A meta-analysis of Jiang et al. found positive correlation of the T variant and susceptibility to COPD (41). However, in other meta-analyses the T variant tended to associate with COPD in Asians and showed association with COPD in comparison to smoker controls, while the TT genotype showed association with severe COPD cases and in age-matched studies (39, 40). In our study, gene–gene interaction between GSTM1 and MMP9 was found to influence COPD severity with 4.8-fold higher risk of severe COPD in patients with the double risk genotype (null and T). Interactions of the MMP9 T variant with other risk factors have been emphasized in COPD. A study of Yanchina et al. is the first that reported gene–gene interaction of the GSTM1 null and MMP9 T variants as a significant risk for COPD (45). In our study, the same interaction showed marginal significance (p = 0.072, Table 2) in the occurrence of COPD, but modifying effect (p = 0.019, Table 4) on severity of the disease. Apart from this, the study of Yanchina et al. reported markedly lower frequency of the double risk genotype for GSTM1 and MMP9, in COPD (13.9% versus 23.0%) and controls (2.6% versus 13.0%), compared to our study. While both studies, ours and the study of Yanchina et al., emphasis importance of gene–gene interactions of common functional variants in GSTM1 and MMP9 in the pathogenesis of COPD, they also showed variety among two populations, that is a feature of complex disorders, such as COPD (46). Another study, conducted in Taiwan, indicated interaction between the CYP1A1 2A/2A and MMP9 T/T variants in COPD (47). This interaction may be more specific for Taiwanese, since both variants in CYP1A1 and MMP9 have much higher frequency in Asians, than in Caucasians, and where the MMP9 T/T is a sole risk for COPD. Moreover, we have recently reported gene–environment interaction between the MMP9 T variant and cigarette smoke in the pathogenesis of COPD (34). Therefore, we can speculate that the MMP9 T variant has a low penetrance in the pathogenesis of COPD, and rather acts as a modifier of the disease or its “pathological” effect may depend on ethnicity or interaction with another environmental or genetic factor.

Our results regarding the MMP12 rs2276109 A > G variant showed no significant difference among patients and controls, Table 1. The A > G variant is associated with a lower promoter activity of MMP12 (16). Previous research has shown protective effect of the A > G variant in COPD and chronic bronchitis and a positive correlation with the lung function, while several haplotypes that include A variant have been linked to COPD severity, chronic bronchitis, or recurrent pneumonia (17–19,21). On the contrary, other studies and meta-analysis could not confirm the association of A > G variant with COPD (20,22,40). We therefore considered the AA genotype as a risk for COPD in the analysis of gene–gene interactions, based on a trend of lower frequency of the G variant in patients in our study and previously reported protective role of the G variant. We identified interaction between GSTM1 null and MMP12 AA variants, with almost fourfold higher risk for COPD in individuals with double risk genotype (Table 2). MMP12 is highly expressed in the lungs of COPD patients and its activity positively correlates with the severity of emphysema (12,13). Carriers of AA genotype may express higher levels of MMP12 that can intensify elastolysis, while the GSTM1 null variant may affect detoxification augmenting oxidative stress and further lead to the development of emphysema. Our results also indicate the modifying role of the MMP12 A > G variant on the age of onset of COPD, expressing more than threefold higher risk of early COPD onset in carriers of G variant (Table 3). Since the G variant is associated with lower expression and activity of MMP12, it is not expected to influence early COPD onset (16). In our study, the G variant was underrepresented in COPD group, while carriers of G variant were found to have COPD onset before the age of 40 years. Early-onset of COPD may lead to the earlier occurrence of severe symptoms diminishing the quality of life, but in our study the G allele was not associated with severity of COPD (Table 3). A possible explanation is that the G variant can be in linkage disequilibrium with a variant that can initiate early disease onset. Furthermore, we identified interaction between GSTM1 null and MMP12 AA with protective role for early COPD onset. Two risk variants included in the interaction with a protective effect is a rather controversial finding. As mean value of the age of COPD onset did not differ in relation to the GSTM1 variant (GSTM1 wt and null, 41.7 ± 2.8 and 47.0 ± 1.8, p = 0.260), we concluded that this interaction results from the association of the MMP12 G variant with early COPD onset, as found in this study (Table 3).

Conclusions

Our study demonstrated a sole association of the MMP1 GG/GG genotype with COPD occurrence, as well as association of MMP12 A > G variant with early COPD onset. We investigated gene–gene interactions between GST and MMP gene families using common functional genetic variants, commonly studied in COPD, but separately (20–29). Despite the emphasized role of oxidative stress and protease activity in the pathogenesis of COPD, interactions between the genes involved in these two processes have been poorly examined, with only a few studies reporting interactions in COPD (45,47). To our best knowledge, this is the first study reporting gene–gene interactions between GSTM1 and MMP1, 9, and 12 in COPD occurrence or severity. The interactions encompass two risk variants that can influence oxidative stress and protease activity in an unfavorable manner. Oxidative stress can upregulate MMP expressions via activation of redox sensitive transcription factors (e.g., NFkB and AP1) (30). Also, GSTM1 regulates a critical kinase of mitogen-activated protein kinase signaling pathway, and thus can affect activity of downstream kinases that regulate MMP expression (7,31,32). Since COPD has a complex aetiology, identification of interactions can reveal more significant risk factors for the disease allowing development of therapeutics and preventive strategies. Among various genetic variants analyzed in COPD, many have given controversial results, which is probably due to heterogeneity of COPD cases, type of control sample, small sample size, incomplete allele penetrance, ethnic differences, etc. Besides, genetic factors identified in large cohorts have suggested that still there is a large portion of “missing heritability” in COPD (48). We can speculate that unidentified gene–gene and gene–environment interactions may be a part of “missing heritability” in COPD.

Although small sample size represents a limitation of our study, we used COPD cases selected strictly according to the GOLD criteria, and subjects with asthma, malignant disease, bronchiectasis, and other severe concomitant nonpulmonary diseases were excluded from the analysis. We also had a well-defined control group that included subjects with normal pulmonary function and no evidence of lung diseases. Still, more data regarding gene–gene interactions in the field of oxidative stress and protease activity are needed. The future studies need to be conducted on larger cohorts and in different ethnic groups, and should include functional validation of interactions, in order to reveal the actors of complex interplay between multiple factors in the pathogenesis of COPD.

Declaration of interest statement

The authors declare that they have no conflict of interest.

Acknowledgment

This work was supported by grant 173008 from the Ministry of Education, Science and Technological Development of Serbia.