Variants in matrix metalloproteinase‐9 gene are associated with hemorrhagic transformation in acute ischemic stroke patients with atherothrombosis, small artery disease, and cardioembolic stroke

Abstract Objective The potential effect of matrix metalloproteinase‐9 (MMP‐9) variants and these variants interactions on hemorrhagic transformation (HT) risk after ischemic stroke (IS) remain unclear. The aims of present study were to investigate the associations of six variants in MMP‐9 with HT, and these variants interactions whether related to increased HT risk. Method A total of 705 patients with IS who were admitted to the participating hospitals within 48 hr of symptom onset were consecutively enrolled between March 2014 and December 2016. HT was confirmed by brain computed tomography (CT) scan during 14 days from stroke onset. Six variants of MMP‐9 gene were measured by mass spectrometry. Interactions of gene variant–gene variant were assessed through generalized multifactor dimensionality reduction method (GMDR). Results HT occurred in 104 (14.8%) patients. There were no differences in genotypes for the six variants between patients with and without HT using univariate analysis (all p > 0.05). GMDR analysis revealed that there was a synergistic effect of gene variant–gene variant interactions between rs3918242 and rs3787268 in MMP‐9 gene. Cox regression analysis showed that high‐risk interactions of rs3918242 and rs3787268 were associated with increased risk of HT after adjusting for covariates (hazard ratio: 2.08; 95% confidence interval: 1.34–7.85; p = 0.016). Conclusion Incidence of HT is common in acute IS in Chinese population. The mechanisms leading to HT are most likely multifactorial. Two‐loci interactions of rs3918242 and rs3787268 in MMP‐9 gene may confer a higher risk for HT.


| BACKG ROU N D
Hemorrhagic transformation (HT) is the most feared complication in patient with acute ischemic stroke (IS). HT occurs in as many as 10%-40% of patients with IS (Beslow et al., 2011;Terruso et al., 2009). HT can exacerbate brain ischemic injury by promoting glial cell and neuronal death, and is associated with increased mortality and poor stroke outcomes (Khatri, Wechsler, & Broderick, 2007;Park et al., 2012).
However, the underlying mechanisms for HT are unclear. Thus, to clarify the complex pathophysiology and basic mechanisms of HT within the context of stroke are essential to better confirm IS patients in prevention and treatment of HT, and reduce its impact on IS patients.
Blood-brain barrier (BBB) destruction is one of the major causes in development of HT in IS (Jickling et al., 2014). HT is associated with increased BBB permeability (Bang et al., 2007;Lin et al., 2007).
The mechanisms of BBB breakdown in HT are not fully understood, although proteolytic degradation of neurovascular matrix and oxygen free radical damage are very important (Jickling et al., 2014;Kelly et al., 2008). Matrix metalloproteinase-9 (MMP-9) protein expression may increase after ischemia, and it plays an important role in BBB destruction (Barr et al., 2010). A number of studies have shown that high plasma concentration of MMP-9 in acute phase of IS may increase HT risk within ischemic core (Castellanos et al., 2007;Jha et al., 2014;del Zoppo et al., 2012). Use of MMP-9 inhibitors can decrease the BBB destruction and reduce the incidence of HT in experimental animals (Lu et al., 2008(Lu et al., , 2009).
MMP-9 single nucleotide polymorphisms (SNPs) encode and regulate the transcription of MMP-9 protein, and are associated with plasma MMP-9 levels (Blankenberg et al., 2003). It has been known MMP-9 SNPs are associated with carotid atherosclerosis and increased IS incidence (Lin et al., 2012;Nie, Wang, & Tang, 2014;Yuan et al., 2013). Although many studies have revealed that there is an association of plasma MMP-9 levels with HT risk (Castellanos et al., 2007;Jha et al., 2014;del Zoppo et al., 2012), the effect of MMP-9 SNPs on HT susceptibility is not clear. Zhang, Cao, Xu, Li, and Xu (2015) reported that MMP-9 variants were associated with HT of IS in the Chinese population. However, one study in a Mediterranean population did not show the associations of MMP-9 SNPs with HT (Montaner et al., 2003).
These inconsistent results revealed that the genetic etiologies for HT may be complex, and common limitation of these studies is that these trials only assessed single gene polymorphisms without consideration of interactions between gene variant-gene variant. Nowadays, it is emphasized that investigation of gene variant-gene variant interactions is necessary to elucidate genetic mechanisms for the complex diseases by the generalized multifactor dimensionality reduction method (GMDR) (Lou et al., 2007). To date, the effect of gene variant-gene variant interactions in MMP-9 gene on HT risk is unknown.
In this study, we hypothesized the interactions among gene variant-gene variant in MMP-9 gene may affect HT risk in acute IS. To test this hypothesis, we measured six variants in MMP-9 to assess these gene variant-gene variant interactions whether increased the risk of HT. This study was earnestly hoped to provide insight into the mechanism for HT and better prevention of HT.

| Study population
Between March 2014 and December 2016, we consecutively enrolled 705 acute IS patients who were admitted to People's Hospital of Deyang City and the Third Affiliated Hospital of Wenzhou Medical University within 48 hr of symptom onset, and were confirmed by magnetic resonance imaging (MRI). All patients were performed brain computed tomography (CT) scan at admission and at 10-14 days of admission. Additional CT scans were performed whenever symptom deteriorated. Brain CT angiography or magnetic resonance angiography, carotid duplex ultrasound, electrocardiogram (ECG) or 24-hr Holter ECG, echocardiogram were assessed in all patients. The inclusion criteria were as follows: (a) age ≥40 years old; (b) etiologies of IS were atherothrombosis, cardioembolism and small artery disease on the basis of classification system (Han et al., 2007). Exclusion crite- infarct; (f) malignant paraproteinemias, platelet count <100 × 10 9 /L; (g) severe liver, renal disease. The overall response rate was approximately 95% (705/741) for IS patients. This study was reviewed and approved by ethics committees of participating hospitals. The informed consent was obtained from each participant before study enrollment.
All patients were treated according to the standard guideline (Kernan et al., 2014). For each patient, National Institutes of Health Stroke Scale (NIHSS) score was performed by a member of stroke team at admission, and subsequently once a day during period of hospitalization. The various risk factors were recorded, including age, sex, hypertension, diabetes mellitus, atrial fibrillation, and blood pressure at presentation. Fasting glucose, total plasma cholesterol, triglycerides, low-density lipoprotein cholesterol were measured (Yi, Han, Zhou, Lin, & Liu, 2016).

| HT assessment
All enrolled patients were performed CT scan and MRI scan at admission, and a follow-up CT at 10-14 days from stroke onset. On the basis of ECASS (European Cooperative Acute Stroke Study) classification (Larrue & Von, 1997), HT was defined as any degree of hyperdensity within the infarction zone during 14 days from stroke onset by brain CT scan.
Blood sample (3 ml) from each patient was drawn from an arm vein. DNA was extracted from the peripheral leukocytes. The six variants of MMP-9 were measured by matrix-assisted laser desorption and ionization time-of-flight mass spectrometry method, as our previously described (Yi, Liao, Fu, Zhang, & Wang, 2015).

| Statistical analysis
According to the suggested sample requirement of gene-gene interaction (Wang & Zhao, 2003), we calculated the sample in this study.
Incidence of HT was about 10%-40% in patients with acute IS (Beslow et al., 2011;Terruso et al., 2009). According to the incidence of HT, we speculated a sample of 700 could provide 90% power to detect 10% relative risk increment of HT in patients harboring the high-risk interactions genotype, assuming 10% rate of HT in IS patients carrying the low-risk interactions genotype, if a two-sided type I error of 0.05.

Hardy-Weinberg equilibrium for each variant and genotype distributions of the six variants between patients with and without HT
was analyzed by chi-squaredtest. Discrete variables were compared by chi-squaredtest and continuous variables were compared by Student's t test between patients with and without HT. Gene-gene interaction was analyzed by GMDR method (β version 0.7, www. healt hsyst em.virgi nia.edu/inter net/addic tion-genom ics/Software), as previously described (Lou et al., 2007;Yi et al., 2015).
The incidence of HT between patients with and without high-risk interactive genotype was compared by chi-squared test. Survival function estimate of HT was calculated using Kaplan-Meier analyses between patients with and without high-risk interactive genotype. The risk factors for HT conferred by high-risk interactive variable was analyzed by Cox proportional-hazards model to adjust confounding variables, which were significantly associated with HT (p < 0.05) on univariate analysis, and were reported as hazard ratio (HR) with 95% confidence interval (CI).

| Genotype distributions in patients with and without HT
The distributions of the six variants for MMP-9 were in Hardy-Weinberg equilibrium (p > 0.05). The genotype distributions of the six variants did not differ between patients with and without HT using single-locus analytical method (p > 0.05 for each variant individually, Table 2).

| Gene variant-gene variant interactions and HT
Then, we investigated the association between the high-order interaction for the six variants and HT using the GMDR analysis. After adjusting covariates, the best interaction model for HT was rs3918242 and rs3787268, which cross-validation consistency was scored 10/10, and the sign test was 9/10 (p = 0.019,

| Different genotype combinations of rs3918242 and rs3787268 and HT risk
Furthermore, we investigated the associations of different genotype combination of rs3918242 and rs3787268 with risk of HT.
Compared with patients carrying rs3918242TT and rs3787268AA (wild-type genotype), the relative risk of the nine genotype com-

| Risk factors for HT
The risk factors for HT conferred by the combinations of rs3918242 and rs3787268 were assessed using Cox proportional hazards model. The low-risk interactive genotype of rs3918242 and rs3787268 was assigned as zero, and the high-risk interactive genotype was assigned as one. The other variables entered the model to adjust, including age, NIHSS score at admission, ischemic areas, cardioembolism, atrial fibrillation, and systolic blood pressure. The results showed that the high-risk interactive genotype of rs3918242 and rs3787268 was independent predictors for the HT risk after adjustment for the covariates (HR: 2.08; 95% CI: 1.34-7.85; p = 0.016, Table 5). Cumulative freedom for HT was lower in patients with highrisk interactive genotype than those patients with high-risk interactive genotype using Kaplan-Meier estimate (Figure 1).

| D ISCUSS I ON
The possible role of genetic variants of MMP-9 gene in HT has not been thoroughly assessed. In this study, the results showed that the six variants of MMP-9 gene were not associated with the risk of HT using single-locus analysis. GMDR analysis revealed that there was a synergistic effect of variant-variant interactions between rs3918242 and rs3787268 in MMP-9 gene, and the high-risk interactive genotype was significantly associated with increased the HT risk after adjusting for the covariates.
Extensive evidence showed MMP-9 polymorphisms were significantly associated with carotid atherosclerosis and increased IS risk (Lin et al., 2012;Nie et al., 2014;Yuan et al., 2013). However, few studies investigated whether MMP-9 polymorphisms affected HT susceptibility. Zhang et al. (2015) assessed rs3918242 (C vs. T allele) polymorphisms in MMP-9 and found that the incidence of HT was higher in patients with CC genotype compared to those patients with CT/TT genotype, or patients carrying C-allele compared to those carrying T allele. However, one study from Mediterranean population did not show the association between rs3918242C/T polymorphisms and HT (Montaner et al., 2003). Our current results exhibited that there was no association between the six variants of MMP-9 and HT risk using single-locus analysis. There potential causes for inconsistent results may be: (a) racial differences of study populations; (b) IS and HT are complex diseases that does not follow Mendelian pattern of inheritance. Gene-environment and genegene interaction may play a key role in these complex diseases (Bevan et al., 2012). It is highly likely that single variant contributes to complex diseases only through their interaction with other variant (Culverhouse, Suarez, Lin, & Reich, 2002). Therefore, singlelocus linkage analysis seems unsuitable for complex genetic etiology of HT; and (c) social differences exist among different population, this may alter environmental risk to which patients are exposed.
The noteworthy observations in this study were that there was a synergistic effect of gene variant-gene variant on the risk of HT using the GMDR methods. GMDR revealed that rs3918242 and rs3787268 in MMP-9 had a synergistic effect to increase HT risk.
The risk of HT increased by 2.08-fold in IS patients carrying high-risk interactive genotype of rs3918242 and rs3787268 compared with those carrying low-risk interactive genotype, suggesting that interactions of the two variants may play an important role in genetic etiology for HT. To the best of our knowledge, current study is the first to identify that interaction of rs3918242 and rs3787268 in MMP-9 contribute to HT risk.
The pathophysiological mechanisms of the interactions between rs3918242 and rs3787268 in MMP-9 gene effect on HT susceptibility are unclear. Breakdown of BBB is one of important causes in the development of HT (Jickling et al., 2014). Previous studies have shown that MMP-9 concentration in plasma or cerebral extracellular fluid is closely related to vascular damage, resulting from collagenase and elastase degrade extracellular matrix (Culverhouse et al., 2002).
In addition, MMP-9 plays a role in destruction and reconstruction of vascular endothelium (Hou et al., 2014), and is associated with BBB destruction and increased HT risk of acute IS (Barr et al., 2010;Castellanos et al., 2007;Jha et al., 2014;del Zoppo et al., 2012). It has been confirmed that MMP-9 inhibitor can decrease MMP-9 activation, neurovascular injury, BBB destruction, and incidence of HT (Lu et al., 2008(Lu et al., , 2009). Polymorphisms of MMP-9 encode and regulate the transcription of MMP-9 protein, and are associated with plasma MMP-9 levels (Blankenberg et al., 2003). Therefore, one possible explanation for the rs3918242 and rs3787268 interaction is that the two variants participate and regulate the transcription of MMP-9, one of the important mechanisms of HT. MMP-9 gene is located on the long arm of chromosome 20q13.12 in human genome, F I G U R E 1 Probability of Survival Free of HT. Kaplan-Maier analysis of cumulative freedom from HT associated with high-risk interactive genotype (Figure 1). HT indicates hemorrhagic transformation and is regulated at transcriptional level. rs3918242 variant in MMP-9 may inhibit protein binding, and reduce the rate of transcription and downregulate MMP-9 expression (Lin et al., 2012). Degradation of the vascular extracellular matrix by MMP-9 is one important cause for vascular remodeling and angiogenesis (Hashimoto et al., 2003), and may contribute to increase the risk of intracerebral hemorrhage (Gaetani et al., 1999). Ho et al. (2015) showed that MMP-9 rs3787268 polymorphisms were associated with intracerebral hemorrhage (ICH), and may interact with TIMP-1 (tissue inhibitors of metalloproteinases) polymorphisms or alcohol to increase ICH risk.
Thus, we reason that high-risk interaction between rs3918242 and rs3787268 could provide these patients with higher MMP-9 concentration than those patients without this particular high-risk interaction, thereby increasing the risk of HT in acute IS.
Several limitations should be noted in this study. First, although previous study revealed MMP-9 polymorphisms were associated with plasma MMP-9 levels (Blankenberg et al., 2003), plasma MMP-9 levels were not assessed in this study. Second, because of limited samples and two-center study, current results can not represent full spectrum of Chinese population. Thus, these findings should be confirmed in multicenter studies in future. Third, we genotyped multiple known variants in MMP-9 gene, some rare functional variants were not investigated in this study. Fourth, this study only investigated gene variant-gene variant interactions in six variants in MMP-9 gene.
As previously mentioned, oxygen free radical damage, inflammation reactions, MMP-2 and MMP-3 may play key roles in BBB destruction and HT (Jickling et al., 2014). We did not measure the variants in oxygen free radical relevant genes, inflammation relevant genes, and MMP-2 and MMP-3 genes. Thus, a larger set of variants must be investigated to elucidate the effect of full extent of gene-gene interaction on HT susceptibility in future studies.

| CON CLUS ION
The six variants in MMP-9 were not associated with the risk of HT by single-locus analysis. However, GMDR showed that there was a synergistic effect of gene variant-gene variant interactions between rs3918242 and rs3787268. High-risk interactive genotype of rs3918242 and rs3787268 was significantly associated with increased risk for HT in acute IS. The GMDR analysis could provide further insight into the complex genetic pathogenesis of HT.

ACK N OWLED G M ENTS
This study was supported in part by grants from the Sichuan

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.