Journal of Genetics (2019) 98:32 © Indian Academy of Sciences https://doi.org/10.1007/s12041-019-1077-2

RESEARCH ARTICLE

Matrix polymorphisms in chronic periodontitis: a case–control study in the Indian population

POULAMI MAJUMDER1∗, SUJAY GHOSH2 and SUBRATA KUMAR DEY1

1Department of Biotechnology, Centre for Genetic Studies, School of Biotechnology and Biological Sciences, Maulana Abul Kalam Azad University of Technology (Formerly West Bengal University of Technology), Kolkata 700 064, India 2Cytogenetics and Genomics Research Unit, Department of Zoology, University of Calcutta, Kolkata 700 019, India *For correspondence. E-mail: [email protected].

Received 2 April 2018; revised 25 November 2018; accepted 26 November 2018; published online 13 March 2019

Abstract. Chronic periodontitis (CP) is the common form of inflammatory oral disease. Matrix (MMPs) play a pivotal role in the progression of CP by degrading gingival tissue and its remodelling. Here, we conducted a case–control study to investigate a possible association of single-nucleotide polymorphism of MMP and their interaction with CP in the Indian population. A total of 357 DNA samples of venous blood was isolated, of which 157 were identified as CP patients and 200 were healthy individuals. Genotyping of six MMP genes (MMP1, MMP3, MMP7, MMP8, MMP12 and MMP13) was done using polymerase chain reaction following Sanger’s method of sequencing. Statistical analyses were performed by SPSS v16.0, R package (SNPassoc). Gene–gene interactions were evaluated by MDR 3.0.2. The frequency of 6A allele of MMP3 −11715A-6A gene polymorphisms (36%) and G allele of MMP8 +17G-C gene polymorphisms (34%) were higher in the CP population compared with the healthy population (19% and 24%, respectively). A significant association of T allele of MMP8 −799C-T gene promoter polymorphism was found with CP (OR = 2.95, 95%CI = 2.16 − 4.04, P < 0.0001). Genotypic frequency of MMP12 −82A-G polymorphism is associated with CP risk while its allelic distribution is not (OR = 1.32, 95%CI = 0.93 − 1.88, P = 0.129). Gene–gene interactions show the best cross validation consistency model, i.e. MMP1 −519A-G X MMP7 −181A-G X MMP8 −799C-T polymorphisms with a value of 9/10. This gene–gene interaction shows that the significant association of MMP8 −799C-T polymorphism with CP increased susceptibility.Allelic distribution of MMP8 +17G-C and MMP3 −11715A-6A polymorphisms revealed their protective role towards decreased risk of CP. MMP1 −519A-G and MMP7 −181A-G polymorphisms show combinatorial synergistic effect on CP risk.

Keywords. chronic periodontitis; matrix metalloproteinases; single-nucleotide polymorphism; genotype; allele; cross validation consistency.

Introduction and determine the pathogenesis and rate of progression of the disease. The mechanism by which an individual Chronic periodontitis (CP) is considered to be the most may develop CP is not completely understood. In most of common multifactorial complex oral disease (Sorsa et al. the cases, the presence of pathogenic subgingival microbes 2004). CP is the destructive form of periodontitis affect- alone does not result in periodontal tissue destruction ing adults, which can be explained as bacterial infection (Hajishengallis 2015). The interplay between microbial that induced inflammatory gingival tissue, which attach and host factors causes the imbalance of potential micro- and support teeth. The phenotypic characteristics mani- bial levels that intrigues the immunological as well as fest tissue destruction, progressive irreversible bone loss, host response (Molander et al. 1998; Hajishengallis 2015). gingival bleeding and tooth loss (Albander 2005). This This response is mostly regulated by the genetic factors multifactorial disease is regulated by bacterial, genetic and which can be modified by several other factors including environmental factors that affect adult individuals (Rin- demographic, behavioural, environmental and systemic takoski et al. 2010). Although CP is initiated by dental aspects (Parmar et al. 2009). With respect to genetic plaque (microbial), host factor, i.e. genetic factors sustain factors, the (ECM) metalloproteinases

1 32 Page 2 of 13 Poulami Majumder et al.

(MMPs) have an important role due to their , MMP1, MMP3, MMP7, MMP8, MMP12 and MMP13. which are involved in physiological and pathological All these MMP genes are located on 11. The processes, including remodelling and destruction of ECM aim of our study was to investigate the possible associa- (Li et al. 2012). Any disparity that occurs in MMPs is tion of MMP genes (MMP1, MMP3, MMP7, MMP8, secreted by neutrophils (Holla et al. 2004). Subjected tis- MMP12 and MMP13) with CP in the Indian popula- sue inhibitors initiate the destruction of in gum tion. In India, to the best of our knowledge, there is only tissue, leading to CP. However in case of CPs, this balance one study on the association between MMP9 gene poly- may be affected initially by microbial factors and slowly morphism and CP prevalence (Majumder et al. 2017). progressed by genetic factors like MMPs with the help of This quantitative study may be implicated in the aetiology other factors like behavioural factors (smoking, chewing of CP. tobacco) etc., which helps to enhance the risk of CP occur- rence (Parmar et al. 2009). MMP is a large family of zinc-dependent extracel- Clinical significance lular proteinases, which are responsible for the tissue remodelling and degradation of ECM, including , (i) India is a genetically diverse country. A large number elastins, gelatin, matrix glycoproteins and proteoglycans of individuals are affected by periodontitis but most of the (Weng et al. 2016). Most of the members of MMPs fam- cases are neglected, which may lead to more severe inflam- ily are secreted as its inactive pro form (Holla et al. matory and other related diseases. (ii) MMP may act as a 2012). The proteolytic activities of MMPs are precisely biomarker of CP due to its direct involvement with tissue regulated by the involvement of key factors such as micro- degradation. (iii) Risk stratification of patients with CP bial and other environmental, behavioural factors etc. remains unclear in everyday clinical treatment. (iv) MMP and MMPs are activated from their precursors (Visse gene–gene interaction may conclude the chronic inflam- and Nagase 2003). Studies suggest that MMPs comprise matory pathway events in CP patients. (v) This work may the most important pathway towards the tissue destruc- contribute towards the aetiology of CP and may impart a tion and remodelling associated with periodontal tissue theoretical basis for prevention and clinical treatment of (Letra et al. 2012). The dramatic changes in certain lev- CP. els of MMPI, MMPII, MMPIII, MMPVII, MMPVIII, MMPXII and MMPXIII have been observed in gingi- val crevicular fluid and gingival tissue of periodontitis patients (Gurkan et al. 2008). Likewise, other studies have Materials and methods also shown that transcript levels of MMPs are signifi- cantly increased in affected periodontal tissue (Malemud Participants 2006). Accordingly, it can be premised that functional polymorphisms in MMP genes may affect MMP pro- A case–control study was carried out with a total of tein expression and may predispose to chronic periodontal 357 participants, among which 157 patients were affected disease conditions. Several genotype analyses of single- with CP and the rest 200 participants were considered as nucleotide polymorphisms (SNPs) in MMP genes have periodontal healthy control (HC). All participants, who been done to investigate the association between genetic volunteered, agreed to consent in accordance with Helsinki factors and the disease occurrence. Some remarkable stud- Declaration and the Indian Council of Medical Research ies have shown the increased frequency of some common (ICMR) guidelines and were informed about the pur- MMP SNPs in patients with periodontitis (Holla et al. pose of the study; the confidentiality of the participants 2005; Keles et al. 2006; Chou et al. 2011; Loo et al. was preserved during the study. All study methods were 2011; Li et al. 2012; Majumder et al. 2017). On the con- also performed in accordance with the above-mentioned trary, some other studies have demonstrated little or no guidelines. A large number of participants was residing in association of these SNPs in MMP genes with aetiol- the eastern region of India, specifically from West Ben- ogy of CP (Holla et al. 2004; Itagaki et al. 2004; Astolfi gal, though some of participants were from north eastern et al. 2006; Luczyszyn et al. 2012). In different popula- region. All study subjects were recruited for a time period tions there are different outcomes regarding the MMP of over one year between September 2014 and January gene SNP association. The different gene polymorphisms 2016. The study was approved by the institutional ethics determines the different levels of expression at committee. All possible clinical and epidemiological data transaction levels which lead to various disease pheno- were taken for this study. All study subjects were in age types. Several common SNPs have been identified in MMP range of 21–65 years. We prepared some inclusion and genes (Li et al. 2016). We selected those genes and their exclusion criteria during the sample collection. Figure 1 polymorphisms based on their effect on the immune sys- shows the standards for reporting diagnostic accuracy tem, ECM degradation and gingival inflammation. We study (STARD) flowchart that represents the selection have selected six genes from the MMP gene family, namely process of subjects for the entire study. MMP gene polymorphisms in CP patients Page 3 of 13 32

Potentially eligible participants (n = 437) Excluded for not getting participants’ consent n= 13

Eligible participants (n = 424) Excluded (n = 21) Due to having systemic diseases like diabetes, flu, Clinical test blood pressure etc. Based on CAL, PPD, GI (n = 403)

Healthy Control Case (n = 200) (n = 199) Excluded (n = 42; 40 Inconclusive diagnosed with AgP and 2 (n = 0) were inconclusive)

Final diagnosis (Chronic Final diagnosis inflammation, accumulation of Healthy Control dental plaque) (n = 200) Chronic Periodontitis (n= 157)

Figure 1. STARD flowchart for the study population selection.

Inclusion criteria: All participants must have at least 15 history, tooth mobility and radiographs. Clinical param- remaining teeth. For CP selection, patients must pos- eters include PD, CAL, plaque index (PI) and gingival sess probing depth (PD) and clinical attachment loss index (GI). (CAL) more than 3 mm. Smoking status of the partici- pants was taken on the basis of the number of cigarettes smoked (≥10/day) for the last five years. Participants Genotyping who chewed tobacco were also identified by defining their tobacco usage ≥3 times per day for 5 years (Page A measure of 4-mL blood was taken by venipuncture and Beck 1997). The status of tea intake was divided from the arm vein of each subject and kept in ethylenedi- into three categories, namely more than four cups per aminetetraacetic acid (EDTA) vacuitner. Genomic DNA day and less than four cups per day and no intake of was isolated from each sample by Genomic DNA Mini kit tea. (DSRGT DNA Isolation kit, India) based on the instruc- tions of the protocol. DNA was purified by using sequen- tial phenol–chloroform extraction and salt–ethanol pre- cipitation methods and purity was measured by the ratio Exclusion criteria: Those who had systemic diseases, which of OD260/OD280. Extracted pure DNA was labelled and could modify the periodontal status (namely, the cerebro- stored in TE buffer at −20◦C until use. Polymerase chain vascular disease, arthrosclerosis, hypertension, coronary reaction (PCR) was performed for amplification of the fol- heart disease etc.) were excluded. lowing SNPs: MMP1 −519A-G, MMP1 −16071G-2G, The socio-demographic data of participants were MMP3 −11715A-6A, MMP7 −181A-G, MMP8 −799C- included in this study.Clinical assessment of study subjects T, MMP8 +17G-C, MMP12 −82A-G and MMP13 was as follows: CP subjects with signs of clinical inflam- −77A-G. All PCR was carried out in 50 μL containing mation consistent with local aetiological factors, GI score 0.1 μgofDNA,5μLof10× buffer (Invitrogen, Sao > ≥ ≥ μ 1, PD 4 mm, CAL 4 mm, with radiographic evidence Paulo, Brazil), 5 L of 0.5 mM MgCl2 (Invitrogen), 1 of bone loss were included in this study (Zeng et al. 2015). μL of 10 mM dNTPs (HiMedia, India), 1 μLof0.5μM HC subjects have healthy periodontium with no evidence of each primer (Sigma-Aldrich, India), 2.5 U Taq DNA of loss of connective tissue attachment or supporting bone polymerase (Invitrogen). The designed primers for each or other signs of disease activity. All CP patients were SNPs and the thermal cycling parameters for the PCR diagnosed according to their physical, medical and dental amplification of those polymorphisms of MMPs gene are 32 Page 4 of 13 Poulami Majumder et al. Cfor Cfor Cfor Cfor Cfor Cfor Cfor Cfor ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ C for 1 min); final C for 1 min); final C for 1 min); final C for 30 s); final C for 1 min); final C for 1 min); final C for 1 min); final C for 30 s); final ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ Cfor7min Cfor7min Cfor7min Cfor7min Cfor7min Cfor7min Cfor7min Cfor7min ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ C for 5 min; 30 cycles (94 C for 5 min; 30 cycles (94 C for 5 min; 30 cycles (96 C for 5 min; 35 cycles (96 C for 5 min; 27 cycles (96 C for 6 min; 30 cycles (96 C for 6 min; 30 cycles (96 C for 5 min; 27 cycles (96 ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ C for 1 min, 72 C for 1 min, 72 C for 1 min, 72 C for 2 min, 72 C for 1 min, 72 C for 1 min, 72 C for 1 min, 72 C for 1 min, 72 ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ 30 s, 60 30 s, 58 extension at 72 extension at 72 extension at 72 extension at 72 extension at 72 extension at 72 extension at 72 extension at 72 30 s, 60 30 s, 56 30 s, 63 30 s, 60 30 s, 58 30 s, 59 ) PCR condition  –3  ) Reverse primers (5  –3  AGTCCTTGCCCTTCCAGAAA CAGGCAGCTTAACAAAGGCA 1 Cycle 94 GAAGGAATTAGAGCTGCCACACCTGCCATCTTTCCCCTGTA AAGGGATTTCTCTGTGGCAATCTGGAGGATGTGGTTTGGT 1 Cycle 94 ACGGTGAGTCGCATAGCTG TCAGAATAAGTGGGACCAGGT 1 Cycle 96 1 Cycle 96 GAAAGCACTCATTTACTCCAGGA AGTGTCCATACTTACTTCACCAATGGGTAAACATGCCATCTTGA 1 Cycle 96 TCATCTTCATCACCACCACTG 1 Cycle 96 TGCTGCTTTCCTGAGTTAACC CCCCACTCTCCTTCCTTGG 1 Cycle 96 GGCCTAGTCCTTACCAGCTT CCTTGTTCCTTTTCCTCAACCA 1 Cycle 96 519A-G 11715A-6A 181A-G 799C-T 16071G-2G 82A-G 77A-G (rs494379) − (rs35068180) (rs11568818) (rs11225395) − − − (rs2276109) (rs2252070) − (rs1799750) +17C-G (rs2155052) − − Primers and PCR conditions used in this study. Table 1. GeneMMP1 SNP (rs number) Forward primers (5 MMP3 MMP7 MMP8 MMP12 MMP13 MMP gene polymorphisms in CP patients Page 5 of 13 32

Table 2. Sociodemographic and clinical characteristics in study population.

Parameters CP (n = 157) HC (n = 200) CP versus HC (P# value)

Age Age range (years) 22–69 24–65 ∗ Mean (±SD) 41.59 ± 11.12 38.41 ± 9.48 0.0038 Gender (%) ∗ Male 101 (65.33) 95 (47.5) 0.0016 Female 56 (34.67) 105 (52.5) Ref Smoking (%) ∗ Yes 96 (62.0) 41 (20.5) < 0.0001 No 61 (38.0) 159 (79.5) Ref Chewing tobacco (%) ∗ Yes 94 (62.67) 59 (49.5) < 0.0001 No 63 (37.33) 141 (50.5) Ref Drinking tea (%) ∗ ≥4 cups/day 67 (44.67) 62 (31.0) < 0.0001 ∗ <4 cups/day 73 (46.0) 77 (38.5) 0.0001 Never 17 (9.33) 61 (30.5) Ref PD (mm) (full mouth mean ± SD) 6.36 ± 1.62 0.34 ± 0.66 < 0.0001∗ CAL (mm) (full mouth mean ± SD) 8.79 ± 1.94 0.03 ± 0.21 < 0.0001∗ PI 2.95 ± 0.81 0.05 ± 0.2 < 0.0001∗ GI 2.61 ± 1.01 0.01 ± 0.08 < 0.0001∗

*Statistically significant. #Fisher’s exact test P value < 0.05. detailed in table 1. All PCR products were sequenced Results (Sanger method of sequencing) by Prism 3100 DNA Genetic Analyzer (Applied Biosystems, Carlsbad, USA). Baseline characteristics

In this study, 157 CP patients were enrolled and their mean age was 41.59 ± 11.12 years. There is no dras- Statistical analysis tic difference in mean age between patients and control population (38.41 ± 9.48). The basal characteristics and Statistical analysis was performed using SPSS v16.0 (SPSS, clinical parameters are tabulated in table 2. Among the CP Chicago, USA). Continuous variables with normal dis- population, 62% participants are regular smokers, which tribution are shown as mean ± standard deviation (SD). is significantly higher than the control group (20.5%). Conformity to Hardy–Weinberg equilibrium (HWE) was The frequency of smoking and chewing tobacco in CP determined by χ 2 analysis. All categorical nonparametric are 62% and 62.67%, respectively. Both the smokers and data were evaluated by using χ 2 test and Fisher’s exact chewing tobacco users are significantly greater than the test. One-way analysis of variance (ANOVA) was per- control (Fisher’s exact test P < 0.0001). About 44.67% CP formed to evaluate the parametric data. The χ 2 test was patients drank tea more than four cups per day while 73% also used to evaluate whether genotype and allele fre- CP patients take tea less than four cups per day. We found quencies were in HWE. All genotyping distribution and that the frequency of tea drinkers are significantly higher statistical analysis were assessed by SNPassoc v3.4.1 in than the control group (<0.0001). The mean of PD in R package. The association of each SNP with the risk patient group (6.36 ± 1.62) is significantly greater than the for CP was analysed by multivariate logistic regression control (<0.0001). Meanwhile, CAL also is significantly using the stepwise backward approach; odds ratio (OR), higher (8.79 ± 1.94) in CP group accordingly (<0.0001). and 95% confidence intervals (95%CI) were also calcu- The other clinical parameters in this study, i.e. PI, GI are lated. Differences were considered statistically significant significantly higher in CP patients (2.95±0.81, 2.61±1.01, when P value was less than 0.05. For quantitative data, the respectively) (<0.0001 and <0.0001, respectively). mean and 95%CI were calculated. Bonferroni correction was done and was used only for multiple genotypic com- parisons of the genes studied (Pc ≤ 0.006). Multifactor Associations between MMP SNPs in CP patients dimensionality reduction (MDR) 3.0.2 software was used to determine the gene–gene interactions. The ‘protective Chromatograms of the MMP gene polymorphisms stud- alleles’ and ‘risk alleles’ combinations were determined by ied are displayed in figure 2,a–h.Intable3, the genotypic the same method. distributions of models are discussed. The distribution 32 Page 6 of 13 Poulami Majumder et al.

(a)

-519A-A -519A-G -519G-G

(b)

-16071G-1G -16071G-2G -16072G-2G

(c) -11715A-5A -11715A-6A -11716A-6A

(d)

-181A-A -181A-G -181G-G

Figure 2. (contd) MMP gene polymorphisms in CP patients Page 7 of 13 32

(e) -799C-C -799C-T -799T-T

(f) +17 G-G +17 G-C +17 C-C

(g)

-82 A-A -82 A-G -82 G-G

(h)

-77 A-A -77 A-G -77 G-G

Figure 2. Chromatograms showing genotypic variants: (a) variants for MMP1 −519A-G; (b) variants for MMP1 −16071G-2G; (c) variants for MMP3 −11715A-6A; (d) variants for MMP7 −181A-G. Chromatograms showing genotypic variants: (e) variants for MMP8 −799C-T; (f) variants for MMP8 +17G-C; (g) variants for MMP12 −82A-G; (h) MMP13 −77A-G. of polymorphisms studied in both CP and control (HC) dominant, recessive and over dominant models. MMP1 population obeys the HW law (χ 2 < 3.84; P > 0.05) −519A-G genotypic frequencies are not to be found except MMP12 −82A-G in CP and MMP13 −77A in the significant. The genotypic distribution of the codominant control. There are four models described, codominant, model between AA and GG is insignificant and so are 32 Page 8 of 13 Poulami Majumder et al.

Table 3. Genotypic distribution of MMP gene variants.

Gene SNPs Genotype CP (%) HC (%) Model AOR (95%CI) P value

MMP1 −519A-G AA 68 (43.3) 91 (45.5) Codominant AA versus AG 1.05 (0.63−1.75) 0.258 AG 62 (39.5) 86 (43.0) GG 27 (17.2) 23 (11.5) AA versus GG 1.78 (0.88−3.60) Dominant AA versus AG+GG 1.20 (0.75−1.94) 0.44 Recessive AA+AG versus GG 1.74 (0.9−3.37) 0.101 Over dominant AA+GG versus AG 0.91 (0.56−1.46) 0.688 −16071G-2G 1G1G 81 (51.6) 101 (50.5) Codominant 11 versus 12 0.95 (0.57−1.58) 0.839 1G2G 58 (36.9) 73 (36.5) 2G2G 18 (11.5) 26 (13.0) 11 versus 22 0.8 (0.38−1.67) Dominant 11 versus 12+22 0.91 (0.57−1.46) 0.702 Recessive 11+12 versus 22 0.82 (0.4−1.65) 0.573 Over dominant 11+22 versus 12 1.0 (0.62−1.62) 0.997 ∗ MMP3 −11715A-6A 5A5A 72 (45.9) 134 (67.0) Codominant 55 versus 56 2.01 (1.18−3.42) < 0.0001 5A6A 56 (35.7) 56 (28.0) 6A6A 29 (18.5) 10 (5.0) 55 versus 66 6.89 (2.96−16.02) ∗ Dominant 55 versus 56+66 2.73 (1.67−4.44) < 0.0001 ∗ Recessive 55+56 versus 66 5.31 (2.35−12.0) < 0.0001 Over dominant 55+66 versus 56 1.45 (0.87−2.39) 0.015 MMP7 −181A-G AA 94 (59.9) 137 (68.5) Codominant AA versus AG 1.3 (0.77−2.21) 0.4 AG 50 (31.8) 50 (25.0) GG 13 (8.3) 13 (6.5) AA versus GG 1.64 (0.68−4.0) Dominant AA versus AG+GG 1.37 (0.84−2.23) 0.205 Recessive AA+AG versus GG 1.52 (0.63−3.66) 0.349 Over dominant AA+GG versus AG 1.24 (0.74−2.08) 0.423 ∗ MMP8 −799C-T CC 43 (27.4) 110 (55.0) Codominant CC versus CT 2.47 (1.44−4.24) < 0.0001 CT 66 (42.0) 70 (35.0) TT 48 (30.6) 20 (10.0) CC versus TT 5.46 (2.74−10.89) Dominant CC versus CT+TT 3.18 (1.93−5.22) < 0.0001∗ Recessive CC+CT versus TT 3.50 (1.86−6.56) < 0.0001∗ Over dominant CC+TT versus CT 1.44 (0.89−2.34) 0.013 +17G-C GG 76 (48.4) 122 (61.0) Codominant GG versus GC 1.57 (0.93−2.64) 0.071 GC 56 (35.7) 60 (30.0) CC 25 (15.9) 18 (9.0) GG versus CC 2.07 (0.99−4.34) Dominant GG versus GC+CC 1.70 (1.06−2.72) 0.028 Recessive GG+GC versus CC 1.75 (0.86−3.58) 0.12 Over dominant GG+CC versus GC 1.37 (0.83−2.27) 0.216 ∗ MMP12 −82A-G AA 94 (59.9) 130 (65.0) Codominant AA versus AG 1.91 (1.10−3.29) 0.002 AG 45 (28.7) 56 (28.0) GG 18 (11.5) 14 (7.0) AA versus GG 3.09 (1.62−5.91) ∗ Dominant AA versus AG+GG 2.26 (1.38−3.73) 0.001 Recessive AA+AG versus GG 2.21 (1.24−3.94) 0.01 Over dominant AA+GG versus AG 1.25 (0.77−2.04) 0.36 MMP13 −77A-G AA 120 (76.4) 163 (81.5) Codominant AA versus AG 1.54 (0.86−2.75) 0.34 AG 32 (20.4) 31 (15.5) GG 5 (3.2) 6 (3.0) AA versus GG 1.53 (0.39−6.04) Dominant AA versus AG+GG 1.54 (0.86−2.75) 0.142 Recessive AA+AG versus GG 1.41 (0.36−5.53) 0.626 Over dominant AA+GG versus AG 1.52 (0.82−2.80) 0.18

*Statistically significant after Bonferroni correction P ≤ 0.006. Clinical parameters, age, gender, smoking and chewing tobacco habits were adjusted. SNP association study was done with SNPassoc analysis in R package (Repository Cambridge University). MMP gene polymorphisms in CP patients Page 9 of 13 32

Table 4. Allele frequency of studied MMP gene variants

Allele frequency Gene SNPs Allele CP (N = 157) HC (N = 200) Chi-squarea (P value) OR (95% CI) P valueb MMP1 −519A-G A 198 (0.63) 268 (0.67) 1.21 (0.272) 1.189 (0.87−1.62) 0.303 G 116 (0.37) 132 (0.33) −16071G-2G 1G 220 (0.7) 276 (0.69) 0.09 (0.764) 0.951 (0.689−1.31) 0.806 2G 94 (0.3) 124 (0.31) ∗ MMP3 −11715A-6A 5A 201 (0.64) 324 (0.81) 26.08 (< 0.0001) 2.39 (1.7−3.36) < 0.0001 6A 113 (0.36) 76 (0.19) MMP7 −181A-G A 239 (0.76) 324 (0.81) 2.52 (0.112) 1.33 (0.93−1.91) 0.117 G 75 (0.24) 76 (0.19) ∗ MMP8 −799C-T C 150 (0.48) 292 (0.73) 47.8 (< 0.0001) 2.95 (2.16−4.04) < 0.0001 T 164 (0.52) 108 (0.27) ∗ +17G-C G 207 (0.66) 304 (0.76) 8.78 (0.003) 1.63 (1.18−2.27) 0.003 C 107 (0.34) 96 (0.24) MMP12 −82A-G A 232 (0.74) 316 (0.79) 2.58 (0.108) 1.32 (0.93−1.88) 0.129 G 82 (0.26) 84 (0.21) MMP13 −77A-G A 273 (0.87) 356 (0.89) 0.71 (0.39) 1.21 (0.771−1.912) 0.416 G 41 (0.13) 44 (0.11)

aPearson χ2 test; bFisher’s exact test. *Statistically significant at P < 0.05. the other models. MMP1 −16071G-2G polymorphisms P < 0.0001). In MMP8 +17C-G polymorphism, G allele is do not show any association with CP. The codominant the mutant one and acts as protective allele in disease pro- (5A5A versus 5A6A and 5A5A versus 6A6A), domi- gression (OR = 1.63, 95%CI = 1.18 − 2.27, P = 0.003). nant (5A5A versus 5A6A+6A6A) and recessive model G allele of MMP12 is not significant to CP occurrence (5A5A+5A6A versus 6A6A) of MMP3 −11715A-6A (OR = 1.26, 95%CI = 0.89 − 1.795, P = 0.208). The genotypic distribution is significant at P < 0.001, 0.003 G allele of MMP13 does not show statistical significance and < 0.0001. However, the over dominant models are with CP progression (OR = 1.46, 95%CI = 0.929−2.314, not significant after Bonferroni correction (P ≤ 0.006). P = 0.102). Finally,we conducted a haplotype analysis but There was no trace of significance in MMP7 −181A-G we did not find a link (D < 0.8) in MMP genes with CP polymorphism. MMP8 −799C-T polymorphisms show an except two SNPs of MMP8 which is obvious (D = 0.92). association with CP. All the models are found to be signif- This discouraged us to study the haplotype distribution icant (codominant model: CC versus CT; AOR = 2.47, further. 95%CI = 1.44 − 4.24, P < 0.0001 and CC versus TT; AOR = 5.46, 95%CI = 2.74 − 10.89, P < 0.0001. Dom- inant model: CC versus CT+TT; AOR = 3.18, 95%CI = Gene–gene interaction 1.93 − 5.22, P < 0.0001. Recessive model: CC+CT ver- sus TT; AOR = 3.50, 95%CI = 1.86 − 6.56, P < 0.0001. We analysed the best interaction model composed of Over dominant model: CC+TT versus CT; AOR = 1.44, the eight SNPs studied in MMP genes by using MDR 95%CI = 0.89 − 2.34, P < 0.0001). MMP8 +17C-G does software. We selected the best model for gene–gene inter- not show any significance with CP. MMP12 polymorphism action (table 5). The three-factor model including MMP1 shows that codominant and dominant model of genotypic −519A-G X MMP7 −181A-G X MMP8 −799C-T SNPs, distribution is significant (≤ 0.006). In table 4, allelic dis- which yielded the highest training balance accuracy of tributions are discussed. MMP1 −519A-G polymorphism 0.721, highest testing accuracy of 0.676 and with CV con- has rare alleles that means the C allele which is not signifi- sistency of 9/10. The interaction between three gene SNPs cant (OR = 1.23, 95%CI = 0.909−1.687, P = 0.18) in CP are associated with the CP risk (AOR = 6.69, 95%CI = group compared with the HC group. MMP1 −16071G-2G 3.037 − 8.352, P < 0.0001). In figure 3, entropy-based cir- polymorphism, 2G allele has no significant relation with cle graph of gene–gene interactions in CP population has disease (OR = 0.92, 95%CI = 0.669 − 1.268, P = 0.626). been illustrated. Entropy values in cells reflect indepen- In MMP3 gene 6A allele is significantly higher in CP group dent effects of indicated allelic variants whereas those in with OR = 2.39, 95%CI = 1.7 − 3.36, P < 0.0001. connecting lines represent the effect of interaction. The red MMP7 −181A-G polymorphism have the G allele which lines indicate a high degree of synergy whilst the light-green is not significant (OR = 1.42, 95%CI = 0.993 − 2.056, line indicates a redundancy (Sinitsky et al. 2017). MMP8 P = 0.062). MMP8−799C-T gene, the T allele is signifi- −799C-T SNP has independently strong effect (7.51%) on cant in CP progression (OR = 2.95, 95%CI = 2.16−4.04, CP risk while MMP1 −519A-G and MMP7 −181A-G 32 Page 10 of 13 Poulami Majumder et al. ∗ ∗ ∗ 0001 0001 0001 . . . 0 0 0 < < < tion model was the 5.1957) 8.3528) 10.9601) − − −

Figure 3. Entropy-based circle graph of gene–gene interactions in the CP population. Entropy values in cells reflect independent effects of indicated allelic variants whereas those in connect- ing lines represent the effect of interaction. The red lines reflect 71 9/10 6.6979 (4.0932 55 10/10 3.2403 (2.0208 55 9/10 5.0368 (3.0372 . . . a high degree of synergy while the light-green line indicates a redundancy. The figure was generated by MDRv.3.0.2 (Compu- tational Genetics Laboratory; Dartmouth). *The expression of polymorphisms in this figure, namely, MMP1 (−519A-G) (which is software generated) is same as mentioned in the text as MMP1 733 0 726 0 799 0 –519A-G and onwards. . . .

represent synergistic interaction effect (9.04%) towards 799C-T polymorphisms.

− CP susceptibility. Figure 4 demonstrates the risk alleles 676 0 638 0 66 0 (dark-grey cells) and protective alleles (light-grey cells) and . . . a combination of the three-factor model alleles. MMP8 × Discussion 181A-G − 721 0 6381 0 678 0 A good periodontal health needs a balance between tis- . . . 0 sue destruction , i.e. MMPs and its inhibitors. As mentioned earlier that CP is a multifactorial disease, where MMP7 Training balance Testing balance P for permutation gene–gene interactions associated with CP risk when compared with control group.

× genetic factors play an important role in disease suscep- tibility. The reason behind the study of these genes that

MMP cause periodontitis is to find out the aetiology of disease

519A-G genetically (Malemud 2006). It is an inflammatory oral −

181A-G X disease and therefore is obvious for the presence of inflam- −

05. matory response in host cells. However, it is clear that . 181A-G 0 0 MMP1

− inflammatory responses are regulated by immune < MMP7 responses. All genes act like a collaborating network

× along with their regulatory system. This study reveals MMP7 −

× the following: (i) 6A allele of MMP3 11715A-6A gene polymorphism and G allele of MMP8 +17G-C gene poly- morphisms act as protective alleles in the CP population 519A-G 799C-T519A-G 0 Characteristics of the model of − 799C-T investigated. It is found to be as protective as these mutant − − − allele frequencies and are also higher in control healthy groups. (ii) A significant association of MMP8 −799C-T Table 5. MMP8 MMP1 MMP1 *Statistically significant at AOR, adjusted OR. The best model was detected as the one with the maximum testing accuracy and maximum cross validation (CV) consistency. In this study, the best interac three-factor model including MMP8 Modelgene accuracy promoter accuracy polymorphism Sensitivity Specificity CVconsistency with AOR increased susceptibility test MMP gene polymorphisms in CP patients Page 11 of 13 32

Figure 4. Allele combinations of indicated SNPs associated with high (dark-grey cells) and low (light-grey cells) susceptibility risk in CP patients. The figure was generated by MDRv.3.0.2 (Computational Genetics Laboratory; Dartmouth). The most high-risk cells have been found in the third interaction block (six high-risk cells out of nine cells) i.e. T allele of MMP8 −99C-T with MMP1 −519A-G and MMP7 −181A-G. Hence, the effect of a rare allele of MMP8 −799C-T polymorphism caused more risk towards CP prevalence. *The expression of polymorphisms in this figure, namely, MMP1 (−519A-G) (which is software generated) is same as mentioned in the text as MMP1 −519A-G and onwards. to CP prevalence. (iii) No significant association of MMP3 gene can act as protective factor and has decreased MMP1 −519A-G, MMP1 −16071G-2G and MMP13 susceptibility to the disease. The same event happens in −77A-G gene polymorphisms with CP. (iv) The allelic case of MMP8 +17G-C polymorphism. G allele is the distribution of MMP1 −519A-G, MMP1 −16071G-2G, rare allele but found dominant in both disease and control MMP12 −82A-G and MMP13 −77A-G are nonsignifi- group, which means G allele has protective shield for CP. cantly associated with CP susceptibility. (v) Codominant MMP8 −799C-T polymorphism has direct effect on tissue and dominant models of MMP12 −82A-G polymor- degradation. The results shows a significant association of phism is associated with CP increasing risk. (vi) If the MMP8 gene polymorphism with the increasing suscepti- mutant genotypes of all gene polymorphisms are com- bility to CP. Therefore, MMP8 polymorphisms may act bined from the three-factor models, the CP risk seems as disease markers for the early diagnosis of periodonti- to increase. (vii) The best gene–gene interaction model is tis especially of CP (Chou et al. 2011). MMP13 −77A-G MMP1 −519A-G × MMP7 −181A-G × MMP8 −799C- polymorphism seems to be negatively associated with CP T polymorphisms with a CV consistency of 9/10. We progression that means the presence of variant in lower have managed to analyse this small data statistically. frequency decreases the CP severity (Rossa et al. 2010). In Actually, the promoter of the gene responds to various the CP group, the two polymorphisms of MMP8 gene are stimuli, including growth factors, cytokines, tumour pro- linked as they were showing a D value of more than 0.8, i.e. moters and oncogene products. In our study, we found 0.92. This linkage somehow reveals that if a periodontitis no association with MMP1 gene with CP susceptibil- patient carries MMP8 gene polymorphisms then there is ity. Several studies revealed that some inhibitors such as a possibility to inherit the polymorphisms together to the tissue inhibitor, COX inhibitor, even anti-inflammatory next generation. While we studied gene–gene interaction, interleukins may suppress the effect of MMP1 −519A- we found a strong effect of MMP8 −799C-T polymor- GandMMP1 −16071G-2G polymorphisms, which may phism separately but when it interacts with another gene help to decrease the severity of CP (Visse and Nagase polymorphism the entropy value reduces and the 2003; Fontana et al. 2012). 6A polymorphism in the redundancy found between MMP8 −799C-T with MMP1 promoter of the MMP3 gene is caused by a variation in the −519A-G and MMP7 −181A-G. On the contrary, we number of adenosines located at position −1171 of MMP3 found synergistic interaction (red line) between MMP1 relative to the transcription start site, resulting in one allele −519A-G and MMP7 −181A-G gene polymorphisms having five adenosines (5A) and the other allele having six (9.04%) while their separate effects bring no impact on CP adenosines (6A). It has been shown that in different studies occurrence. The dark grey cells containing the risk allele the individuals carrying the 6A allele have decreased sus- frequency and the light grey cells indicate protective allele ceptibility to this disease (Astolfi et al. 2006; Li et al. 2012). combinations (Wang et al. 2016; Sinitsky et al. 2017). It In this study we have found the frequency of 6A variants is interesting that separately polymorphisms in MMP1 more than 5A in diseased group, thus 6A variant of −519A-G and MMP7 −181A-G gene polymorphisms 32 Page 12 of 13 Poulami Majumder et al. have no effects on CP susceptibility but while combined Holla L., Jurajda M., Fassmann A., Dvorakova N., Zno- with MMP8 −799C-T they show a significant increase in jil V. and Vacha J. 2004 Genetic variations in the matrix susceptibility towards CP risk (figure 4). We therefore con- metalloproteinase-1 promoter and risk of susceptibility and/or severity of chronic periodontitis in the Czech population. J. clude that these genotype combinations as key players in Clin. Periodontol. 31, 685–690. modulating the CP risk in the Indian population. Cumu- Itagaki M., Kubota T., Tai H., Shimada Y., Morozumi T. and latively, all genes and their polymorphisms contribute to Yamazaki K. 2004 -1 and -3 gene the periodontitis progression. The extent of their expres- promoter polymorphisms in Japanese patients with periodon- sion impacts the signs of CP that results in the emergence titis. J. Clin. Periodontol. 31, 764–769. Keles G., Gunes S., Sumer A., Sumer M., Kara N., Bagci H. et of different types of periodontal destruction and oral dis- al. 2006 Association of matrix metalloproteinase-9 promoter eases. This study had limitations such as the small sample gene polymorphism with chronic periodontitis. J. Periodontol. size, and participants were from the eastern zone of India. 77, 1510–1514. Hence, further studies, in a larger ethnically diverse popu- Letra A., Silva R. M., Rylands R. J., Silveira E. M., de Souza lation are necessary to elucidate the role of MMP genes in A. P., Wendell S. K. et al. 2012 MMP3 and TIMP1 variants contribute to chronic periodontitis and may be implicated in CP risk. disease progression. J. Clin. Periodontol. 39, 707–716. This study concludes the significant association of Li G., Yue Y., Tian Y., Li J. L., Wang M., Liang H. et al. MMP8 −799C-T polymorphism with CP increased sus- 2012 Association of matrix metalloproteinase (MMP)-1, 3, 9, ceptibility while MMP8 +17G-C and MMP3 −11715A- interleukin (IL)-2, 8 and cyclooxygenase (COX)-2 gene poly- 6A mutant alleles are associated with the decreased sus- morphisms with chronic periodontitis in a Chinese population. − − Cytokine 60, 552–560. ceptibility to CP. MMP1 519A-G and MMP7 181A- Li W., Zhu W., Singh P., Ajmera D., Song J. and Ji P. 2016 Asso- G polymorphisms are showing combinatorial synergistic ciation of common variants in MMPs with periodontitis risk. effect on CP risk. Oral habits such as smoking and chewing Dis. Markers 2016, 1545974. tobacco enhance the risk to CP susceptibility. Loo W. T., Wang M., Jin L. J., Cheung M. N. and Li G. R. 2011 Association of matrix metalloproteinase (MMP-1, MMP-3 and MMP-9) and cyclooxygenase-2 gene polymorphisms and Acknowledgement their with chronic periodontitis. Arch. Oral. Biol. 56, 1081–1090. We thank Dr Vineet Nair and all participants of this study. This Luczyszyn S., De Souza C., Braosi A., Dirschnabel A., Claudino work was supported by DBT, India (102/IFD/SAN/1699/2015– M., Repeke C. et al. 2012 Analysis of the association of an 2016). MMP1 promoter polymorphism and transcript levels with chronic periodontitis and end-stage renal disease in a Brazilian population. Arch. Oral. Biol. 57, 954–963. Majumder P., Singh S. J., Nair V., Bhaumik P., Mukherjee A., Ghosh P. et al. 2017 Alliance of matrix metalloproteinase- References 9 (MMP-9) promoter gene polymorphism with chronic and aggressive periodontitis in Indian population. Meta Gene. 12, Albander J.M. 2005 Epidemiology and risk factors for periodon- 88–93. tal diseases. Dent. Clin. N. Am. 49, 517–532. Malemud C. J.2006 Matrix metalloproteinases (MMPs) in health Astolfi C., Shinohara L., da Silva R., Santos M. C., Line S. and and disease: an overview. Front. Biosci. 11, 1696–1701. de Souza A. P. 2006 Genetic polymorphisms in the MMP-1 Molander A., Reit C., Dahlén G. and Kvist T. 1998 Microbio- and MMP-3 gene may contribute to chronic periodontitis in a logical status of root-filled teeth with apical periodontitis. Int. Brazilian population. J. Clin. Periodontol. 33, 699–703. Endontic. J. 31, 1–7. Chou Y. H., Ho Y. P., Lin Y. C., Hu K. F., Yang Y. H., Ho Page R. and Beck J. D. 1997 Risk assessment for periodontal K. Y. et al. 2011 MMP-8 -799 C > T genetic polymorphism diseases. Int. Dent. J. 47, 61–87. is associated with the susceptibility to chronic and aggressive Parmar G., Sangwan P., Vashi P., Kulkarni P.and Kumar S. 2009 periodontitis in Taiwanese. J.Clin. Periodontol. 38, 1078–1084. Effect of chewing a mixture of areca nut and tobacco on peri- Fontana V., Silva P., Gerlach R. and Tanus-Santos J. 2012 odontal tissues and oral hygiene status. J. Oral. Sci. 50, 57–62. Circulating matrix metalloproteinases and their inhibitors in Rintakoski K., Kaprio J., Murtomaa H. 2010 Genetic and envi- hypertension. Clin. Chim. Acta. 413, 656–662. ronmental factors in oral health among twins. J. Dent. Res. 89, Gurkan A., Emingil G., Saygan B. H., Atilla G., Cinarcik 700–704. S., Kose T. et al. 2008 Gene polymorphisms of matrix Rossa C., Min L., Bronson P. and Kirkwood K. L. 2010 Tran- metalloproteinase-2, -9 and -12 in periodontal health and scriptional activation of MMP-13 by periodontal pathogenic severe chronic periodontitis. Arch. Oral. Biol. 53, 337–345. LPS requires p38 MAP kinase. J. Endotoxin. Res. 13, 85–93. Hajishengallis G. 2015 Periodontitis: from microbial immune Sinitsky M. Y., Minina V. I., Asanov M. A., Yuzhalin A. E., subversion to systemic inflammation. Nat. Rev. Immunol. 15, Ponasenko A. V. and Druzhinin V. G. 2017 Association of 30–44. DNA repair gene polymorphisms with genotoxic stress in Holla L. I., Fassmann A., Vasku A., Goldbergova M., Beranek underground coal miners. Mutagenesis 32, 501–509. M., Znojil V. et al. 2005 Genetic variations in the human Sorsa T., Tjäderhane L. and Salo T. 2004 Matrix metallopro- A (matrix metalloproteinase-2) promoter are not teinases (MMPs) in oral diseases. Oral. Dis. 10, 311–318. associated with susceptibility to, and severity of, chronic peri- Visse R. and Nagase H. 2003 Matrix metalloproteinases and tis- odontitis. J. Periodontol. 76, 1056–1060. sue inhibitors of metalloproteinases: structure, function, and Holla L., Hrdlickova B., Vokurka J. and Fassmanna A. 2012 biochemistry. Circ. Res. 92, 827–839. Matrix metalloproteinase 8 (MMP8) gene polymorphisms in Wang B., Xu Q., Yang H., Sun L. and Yuan Y. 2016 The asso- chronic periodontitis. Arch. Oral. Biol. 57, 188–196. ciation of six polymorphisms of five genes involved in three MMP gene polymorphisms in CP patients Page 13 of 13 32

steps of nucleotide excision repair pathways with hepatocellu- Zeng X., Zhang Y., Kwong J. S., Zhang C., Li S., lar cancer risk. Oncotarget. 7, 20357–20367. Sun F. et al. 2015 The methodological quality assess- Weng H., Yan Y., Jin Y. H., Meng X., Mo Y. and Zeng, X. 2016 ment tools for preclinical and clinical studies, system- Matrix metalloproteinase gene polymorphisms and periodon- atic review and meta-analysis, and clinical practice guide- titis susceptibility: a meta-analysis involving 6,162 individuals. line: a systematic review. J. Evid. Based. Med. 8,2– Sci. Rep. 6, 24812. 10.

Corresponding editor: Indrajit Nanda