Introduction

Cardiovascular diseases are a global problem of modern healthcare and the second most common cause of total mortality1,2. Among cardiovascular diseases, essential hypertension (EH) is of tremendous clinical importance in terms of health, working capacity, and life expectancy3,4,5. According to some estimates, the number of patients with EH will reach 1.56 billion by 20256. More than 9.4 million death cases resulting from EH complications, such as stroke, myocardial infarction, renal failure, etc., are recorded worldwide annually7,8,9.

The development of EH is determined by complex interaction mechanisms of genetic and environmental factors5,10,11,12. The contribution of hereditary factors to EH is estimated from 25 to 75% in different populations according to family and twin studies13,14. Among the possible candidate genes for EH are matrix metalloproteinases (MMP). This is a group of enzymes with a wide range of biological functions that are responsible for the hydrolysis of the extracellular matrix (ECM) components15,16,17,18. The imbalance between synthesis and degradation of ECM caused by the change in MMP gene expression can lead to a decrease in the ability of the vascular wall to remodel and to the development of cardiovascular diseases19,20. It was found that single nucleotide polymorphisms (SNP) of the MMP genes are associated with EH and its complications in different populations: Australian (rs3025058 MMP3), Polish (rs3025058 MMP3), American (rs652438 MMP12), Swedish (rs11568818 MMP7), Brazilian (rs243865 MMP2), Serbian (rs11225395 and rs1320632 MMP8, rs1799750 MMP1) and others16,19,21,22,23,24,25,26. However, no associations of the MMP polymorphisms with EH and its complications were found in Chinese (rs3025058 MMP3, rs17577 MMP9), Indian (rs11568818 MMP7), and Mexican (rs1799750 MMP1) populations27,28,29. The observed inconsistencies prompt for further studies of the role of the MMP polymorphisms in the development of EH.

In this study, we analyzed the associations of matrix metalloproteinases and their gene–gene interactions with EH in a Caucasian population of Central Russia. We also examined the biological mechanisms of their phenotypic effects and SNPs linked to them (nonsynonymous substitutions, regulatory and cis-eQTL influences, pathways).

Results

Study participants characteristics

The patients with EH had higher rates of BMI, total cholesterol, triglycerides, low-density lipoprotein cholesterol, low high-density lipoprotein cholesterol, and a high incidence of smokers (p < 0.001) (Table 1). These factors were applied as covariates in the association analyses.

Table 1 Phenotypic characteristics of the study participants.
Full size table

Analysis of SNP association

The data about the studied SNPs is given in Table 2. All loci had MAF > 5% and showed no departure from the HWE (p > 0.05).

Table 2 The allele and genotype frequencies (%) of the studied SNPs in the EH and control groups.
Full size table

The allele G of the rs11568818 MMP7 polymorphism was associated with EH according to the additive (OR = 0.70, 95% CI 0.54–0.89, p = 0.001, pperm = 0.01, power 99.03%), dominant (OR = 0.66, 95% CI 0.46–0.97, p = 0.02, pperm = 0.03, power 94.50%) and recessive (OR = 0.58, 95% CI 0.37–0.91, p = 0.01, pperm = 0.02, power 91.54%) models (Table 3).

Table 3 Association of the MMPs genotypes with EH.
Full size table

SNP × SNP interactions

In total 12 most significant two-, three, and four-locus models of gene–gene interactions associated with EH were determined (pperm ≤ 0.02, cross-validation consistency CVC = 10/10, testing balanced accuracy 49.2–55.3%) (Tables 4 and 5). These models included eight SNPs. Loci rs1320632 MMP8 and rs11568818 MMP7 were involved in the largest number of models (12 and 10, respectively). Loci rs652438 MMP12, rs11225395 MMP8, rs1320632 MMP8, and rs11568818 MMP7 contributed to the most significant epistatic models (pperm = 0.002). The following combinations of genotypes had most significant associations with EH: rs11225395 CT × rs1320632 AG × rs11568818 CC (beta = − 0.827, p = 0.002), rs652438 AA × rs11225395 CT × rs1320632 AG × rs11568818 CT (beta = − 0.847, p = 0.003), rs17577 AG × rs1320632 AG × rs11568818 CT × rs1799750 1G/1G (beta = − 1.726, p = 0.002) (Supplementary Table 1).

Table 4 Gene–gene interactions significantly associated with the risk of EH according to MB-MDR.
Full size table
Table 5 Cross-validation statistics for best models of the gene–gene interactions in EH.
Full size table

The graph of the most significant epistatic models of the four SNPs (Fig. 1) suggests these interactions are concerted. Pronounced synergism was observed between the polymorphisms of the MMP8 gene, while an antagonistic interaction was suggested between rs652438 MMP12 and rs11568818 MMP7. The graph of interactions (Fig. 1b) shows that rs11225395 MMP8 and rs11568818 MMP7 eliminated 0.34 and 0.25% of class entropy, respectively, thereby having the largest univariate effects.

Figure 1
figure 1

High-order gene–gene interaction analysis for SNPs of matrix metalloproteinases in EH (data obtained by Multifactor Dimensionality Reduction, version 3.0.2): (a) Interaction dendrogram; (b) Interaction entropy graph.

Full size image

Functional SNP

Non-synonymous SNPs

Among the eight SNPs studied, two loci (rs17577 MMP9 and rs652438 MMP12) cause the replacement of amino acids in the encoded polypeptide and a decrease in its activity (Supplementary Table 2). Also, three non-synonymous variants (rs679620 MMP3, rs1940475 and rs3765620 MMP8) were determined among the polymorphisms linked to the studied SNPs (Supplementary Table 2).

Regulatory effects

The data on the regulatory effects of the EH-associated loci are presented in Supplementary Table 3. According to the HaploReg database (v4.1), two SNPs were located in evolutionarily conserved regions, two polymorphisms—in the promoter histone marks region, and eight SNPs—in the enhancer histone marks region in various tissues. Seven SNPs were in the hypersensitivity region to DNAse-1, three SNPs—in the protein-bound region, five SNPs—in the motifs changed region. According to the SNPinfo Web Server database, polymorphisms rs1320632 MMP8 and rs11225395 MMP8 possessed the most significant regulatory potential (0.53 and 0.20, respectively). Four SNPs were located in the regions of the transcription factor binding site (TFBS), one was in the microRNA binding region, and two were in the exonic splicing enhancer and exonic splicing silencer. SNP rs11568818 was located in the DNA regulatory motifs region: the allele G increased affinity to transcription factors Foxa (ΔLOD scores = − 5.1), PLZF (ΔLOD scores = − 1.5), Pou5f1 (ΔLOD scores = − 3.2) and reduced affinity to GR transcription factor (ΔLOD scores = 0.8).

In addition to the eight EH-associated SNPs, regulatory significance was estimated for 114 polymorphisms linked to them (Supplementary Table 4). Eight SNPs (including five non-synonymous and three synonymous substitutions) were located in exons of the studied genes, three were located in 3′-UTR and one in 5′-UTR, 86 were in introns, and 44 were in intergenic regions. Nine loci were located in evolutionarily conserved regions.

The in silico analysis of SNPs linked to the EH-associated loci suggested several polymorphisms with pronounced regulatory effects (Supplementary Table 4). For example, rs243862 (linked to rs243865 MMP2) is located in the promoter histone marks region in 21 tissues and enhancer histone marks in two tissues, the hypersensitivity region to DNAse-1 in 19 tissues, region 14 motifs changed.

Expression QTLs

According to the GTExportal database, four EH-associated SNPs had the cis-eQTL significance (p < 8 × 10–5, pFDR ≤ 0.05) and might affect the expression of five genes (MMP7, MMP27, RP11-817J15.3, SNX21, SLC12A5) in several tissues and organs (Supplementary Table 5). Six EH-associated loci were in strong LD with the SNPs affecting the expression (p < 8.5 × 10–5, FDR ≤ 0.05) of 11 genes in more than 20 tissues and organs, including those pathogenetically significant for the development of EH (whole blood, tibial artery, left ventricle of heart, etc.) (Supplementary Table 6).

Pathway analyses. The in silico analysis of the functional significance was conducted for the 7 EH-associated genes (MMP7, MMP8, MMP1, MMP2, MMP3, MMP9, MMP12) and for genes whose expression is affected by the EH-associated SNPs according to the eQTL analysis (Supplementary Tables 5 and 6).

Of the 15 genes considered, the Gene Ontology databases have the relevant information about 12 genes; the information about three genes (RP11-817J15, WTAPP1, and RPL13P2) was not available. We found evidence of enrichment for pathways involved in the metalloendopeptidase activity (PFDR = 1.37E−16), collagen catabolic process (PFDR = 4.68E−15), collagen degradation (PFDR = 8.92E−15), extracellular matrix disassembly (PFDR = 4.79E−14), activation of matrix metalloproteinases (PFDR = 2.28E−13), and 10 other pathways (FDR ≤ 0.05) (Supplementary Table 7).

The network of the intergenic interactions between the 15 EH-associated genes and other 20 genes inferred by GeneMANIA is given in Fig. 2. These interactions are realized through common protein domains (47.03%), co-expression (35.73%), pathogenetic pathways (6.11%), co-localization (2.63%), and genetic interactions (0.53%). The EH-associated genes may interact either directly or via other genes (e.g., NOP56, NOP58, MMP10, MMP21, MMP28, etc.). Among the 35 candidate genes for EH, the most significant interactions were determined for NOP56 and NOP58 (common protein domains, weight = 0.65), PRPF31 и NOP56 (common protein domains, weight = 0.41), PRPF31 и NOP58 (common protein domains, weight = 0.41), MMP7 и MMP3 (pathways, weight = 0.10).

Figure 2
figure 2

The interaction network of the candidate genes for EH obtained using the GeneMANIA online service, http://www.genemania.org/ (the genes associated with EH according to the present study are cross-shaded).

Full size image

Discussion

The present study determined significant associations of eight loci of matrix metalloproteinase genes with EH in a Caucasian population of Central Russia.

Allele G of locus rs11568818 MMP7 was associated with EH according to the dominant, additive and recessive models (OR = 0.58–0.70) and was involved in 10 of 12 two-, three-, and four-locus models of gene–gene interactions associated with EH. Marker rs11568818 was characterized by a significant regulatory effect: it was located in the region hypersensitive to DNAse-1 in 15 tissues, in the region of modified histones (H3K4me1 and H3K4me3) that marked promoter and enhancer in 11 different organs and tissues, and affected the MMP7 gene expression. The locus was located in the region of DNA that binds to the TATA-binding protein (TBP), c-FOS, c-Jun, and located in regulatory protein binding sites. According to the GeneCards database, TBP is responsible for proper RNA polymerase positioning on a promoter during transcription; regulatory proteins c-FOS and c-Jun interact with each other and control cell proliferation, differentiation, and transformation. The data about the possible contribution of rs11568818 to cardiovascular disease in different ethnic populations was somewhat inconsistent. For example, Jormsjö et al.23 showed that carriers of genotype GG rs11568818 MMP7 had an increased risk for developing cardiovascular pathology in the Swedish population, while in populations from India, Mexico, and Turkey associations of this marker with EH and its complications were not determined28,30,31. The observed inconsistencies could stem from the differences in study designs (e.g., differences in the used covariates, sample sizes, gene–gene and gene-environment interactions, etc.). In addition, the differences in the results might be associated with ethnicity-specific pathogenetic features of the emergence and course of EH32,33,34,35,36,37 or/and ethnicity-related differences in the genetic structure of the populations38,39.

According to the GeneCards database (http://www.genecards.org/), the MMP7 gene belongs to the gene cluster on chromosome 11 and encodes the enzyme of the same name, which is characterized by the absence of a conserved C-terminal hemopexin domain. Matrix metalloproteinase 7 is responsible for proteolytic cleavage of elastin, type I, III, IV, V gelatins, fibronectin, casein, proteoglycans and is involved in the regeneration processes after damage, remodeling of the extracellular matrix, and also modulates cell migration, proliferation, and apoptosis15.

We did not determine monolocus effects for rs1320632 of the MMP8 gene; however, this marker is involved in all 12 identified gene–gene interactions models associated with EH. SNP rs1320632 has significant regulatory potential—it is located in the region of DNA regulatory motifs and modulates affinity for seven transcription factors (CAC-binding-protein, Foxc1-2, GATA-known13, GCM, MAZ, PRDM1-known1, STAT-disc6). In addition, this SNP is located in the region of hypersensitivity to DNAse-1 in 10 tissues, in the region of H3K4me1 and H3K4me3, marking enhancers and promoters in six tissues. We found that rs1320632 MMP8 is strongly linked to 14 SNPs that have important regulatory significance, and this locus is associated with a level of the MMP27 gene expression in four tissues. Our results are consistent with those previously reported for the Serbian population22. The MMP8 gene encodes a proteolytic enzyme involved in the cleavage of the extracellular matrix in the proliferation and remodeling of tissues, embryonic development, as well as in pathological processes such as arthritis and metastasis. Selective proteolysis of the polypeptide leads to the formation of many active forms of the enzyme with different N-ends. MMP8 is involved in the degradation of type I-III collagen and is expressed by macrophages, while the production of MMP8 sharply increases with inflammation.

It should be noted that the current study is somewhat limited because: (a) only one ethnic population was analyzed. The well-known ethnic disparities in the prevalence of complex diseases warrant validation studies of the determined associations of the MMP genes and EH in other ethnic populations; (b) a transethnic meta-analysis of the studied MMP SNPs would help to clarify this issue, but is currently impossible due to the insufficient data available about MMP and EH; (c) the obtained results are not sufficient to construct a reliable predictive model of EH based on the eight studied SNPs using the multi-model deep learning method.

Conclusions

Thus, in this work, we found that genetic polymorphism rs11568818 MMP7 and gene–gene interactions of eight SNPs are associated with EH in a Caucasian population of Central Russia, and their phenotypic effects are realized through non-synonymous substitutions, regulatory and cis-eQTL effects, and shaed biological pathways.

Materials and methods

Study subjects

The study sample included 1405 people: 939 patients with essential hypertension and 466 controls. The participants were recruited through the cardiological and neurological departments of the St. Joasaph Belgorod Regional Clinical Hospital during 2013–2016. The following inclusion criteria were adopted: self-declared Russian descent, a birthplace in Central Russia40.

Essential hypertension was diagnosed by certified physicians in cardiology and neurology as recommended by the World Health Organization (n = 939, 100%). All study subjects had a clinical history of hypertension for more than one year. Untreated hypertensive patients had the established hypertension defined by seated systolic (SBP) and/or diastolic (DBP) blood pressure above 140 and/or 90 mm Hg, respectively, measured at least twice. All hypertensive patients had no clinical signs, symptoms, and laboratory findings suggesting secondary hypertension, and liver or/and kidney failure. The controls were recruited during regular medical examinations at the above Center. The criterion for inclusion in the control group was the level of SBP < 140 mmHg and the level of DBP < 90 mmHg, no history of metabolic syndrome, autoimmune disorders, and oncological diseases.

The level of blood pressure (BP) was determined by the auscultation method using a sphygmomanometer and according to Korotkov41. BP was measured throughout several days (at least twice). The patients had not consumed caffeine, exercise, and smoke for at least 30 min before the measurement procedure bean. The measurement was performed in the seated position of the patient after 5 min of rest. The blood pressure was measured on both arms: at least two measurements were taken with an interval of 1–2 min. A mean of at least two readings taken at least two times was used to assess individual blood pressure.

The study was carried out in accordance with the standards of Good Clinical Practice and the principles of the Helsinki Declaration. The study was approved by the Regional Ethics Committee of Belgorod State University. All participants signed informed consent before the enrolment in the study.

Data on anthropometric characteristics (height, weight, and body mass index), smoking and alcohol use were collected for each participant. Blood samples for determining total cholesterol (TC, mmol/l), triglycerides (TG, mmol/l), high-density lipoprotein cholesterol (HDL-C, mmol/l), and low-density lipoprotein cholesterol (LDL-C, mmol/l) were collected after 8-h fasting, the analysis was performed in the certified clinical diagnostic laboratory of the St. Joasaph Belgorod Regional Clinical Hospital. The baseline and clinical characteristics of the study population are given in Table 1. The control group was matched to the EH group for sex and age (p > 0.05).

SNP selection and DNA handling

DNA was extracted from whole blood by the phenol–chloroform protocol and then checked for quality (as described earlier42).

Eight single nucleotide polymorphisms (SNPs) of seven matrix metalloproteinase genes (rs1799750 MMP1, rs243865 MMP2, rs3025058 MMP3, rs11568818 MMP7, rs1320632 and rs11225395 MMP8, rs17577 MMP9, and rs652438 MMP12) were selected for the study based on the following criteria42,43: (1) SNP associations with the development of EH and/or its complications according to the results of previous studies; (2) the regulatory effect of a polymorphism (regSNP); (3) the effect of the locus on gene expression (eSNP); (4) relation to non-synonymous substitutions (nsSNP); (5) MAF > 0.05. The SNPs regulatory potential and effect on the gene expression were assessed using HaploReg (v4.1) (http://archive.broadinstitute.org/mammals/haploreg/haploreg.php), SNPinfo Web Server—SNP Function Prediction (FuncPred) (https://snpinfo.niehs.nih.gov/snpinfo/snpfunc.html), and GTExportal (http://www.gtexportal.org/).

Information on the biological role of the studied polymorphisms and their associations with cardiovascular pathology is presented in Supplementary Table 8. All SNPs appeared to have significant regulatory potential, four of them (50.00%) were eSNPs, and 2 were nsSNPs (Supplementary Table 3).

SNP genotyping

The polymorphisms were genotyped using the MALDI‐TOF mass spectrometry iPLEX platform (Agena Bioscience Inc, San Diego, CA). Genotyping of blind replicates was performed to ensure quality control. The repeatability test was performed for 5% of randomly selected samples and showed 100% reproducibility.

Statistical analysis

Correspondence of the studied loci to the Hardy–Weinberg equilibrium (HWE) was checked by the chi-square test. The loci were analyzed for associations with EH using logistic regression and according to additive (i.e., comparison of all genotypes, e.g., TT vs TC vs CC), dominant (CC/TC vs TT, where C is a minor allele), and recessive (CC vs TC/TT, where C is a minor allele) genetic models with adjustment for covariates. The following covariates were applied as quantitative variables: BMI, total cholesterol, triglycerides, high-density and low-density lipoprotein cholesterol; and while smoking status was used as qualitative variables (yes/no) (Table 1). The adaptive permutation test44 was applied to adjust the results for multiple comparisons. The significance level was set at pperm < 0.05. The calculations were performed using software PLINK v.2.050 (http://zzz.bwh.harvard.edu/plink/). Statistical power for each SNP was estimated using Quanto 1.2.4 (http://hydra.usc.edu/gxe, 2009).

The epistatic interactions were analyzed assuming two-, three-, and four-locus models. The MB-MDR (Model Based Multifactor Dimensionality Reduction)45,46 approach and respective software (v. 2.6) for the R programming environment were utilized for the computations. The significance of the gene–gene interaction models was evaluated by the permutation test44. For the permutation test, the following threshold p values (after the Bonferroni correction based on the numbers of combinations studied for eight loci) were adopted for the models of the gene–gene interactions: p < 1.8 × 10–3 (< 0.05/28) for the two-locus models, p < 8.9 × 10–4 (< 0.05/56) for the three-locus models, and p < 7.1 × 10–4 (< 0.05/70) for the four-locus models. The significance level was set at pperm < 0.05.

The cross-validation of the most significant models of intergenic interactions associated with EH was conducted by MDR (Multifactor Dimensionality Reduction)47, as implemented in the MDR software (v.3.0.2) (http://sourceforge.net/projects/mdr). The MDR method was used to assess the nature and strength (contribution to entropy) of gene–gene interactions and visualize them in graph form48.

Functional SNPs

The SNPs associated with EH and those strongly linked to them were evaluated for their functional significance (non-synonymous SNPs49, regulatory potential50,51, and eQTLs52). The loci in linkage disequilibrium (LD) (r2 ≥ 0.8) with the EH-associated ones were determined using HaploReg (v4.1) (http://archive.broadinstitute.org/mammals/haploreg/haploreg.php) and the data of the European population from the 1000 Genomes Project Phase 142,50,53,54.

Non-synonymous SNPs

Non-synonymous SNPs and their predictive potential were analyzed using the SIFT tool (https://sift.bii.a-star.edu.sg/)49.

Regulatory effects

The candidate loci for EH were analyzed in silico for their regulatory potential using HaploReg (v4.1) (http://archive.broadinstitute.org/mammals/haploreg/haploreg.php)50 and SNP Function Prediction (FuncPred) (https://snpinfo.niehs.nih.gov/snpinfo/snpfunc.html)51.

Expression QTLs

The effect of the candidate SNPs for EH on gene expression in various tissues and organs was estimated using the GTExportal data (http://www.gtexportal.org/) as of 10.12.2019 (Release V8 updated on 26/08/2019) (dbGaP Accession phs000424.v8.p2)52. The False Discovery Rate (FDR) ≤ 0.05 was applied as statistical significance threshold55.

Pathway analyses

The genes associated with EH were analyzed for functional significance in the various metabolic pathways using the Gene Ontology Portal (PANTHER Overrepresentation Test accessed on 13.04.2017; PANTHER version 12.0 accessed on 10.07.2017, http://geneontology.org)56. The adjustment for multiple comparisons was made using the FDR test. The networks of intergene interaction were inferred using GeneMANIA (version 3.5.0, accessed on 13 March 2017, http://genemania.org) and the automatic weighting for the network57.