Assessment of the association of matrix metalloproteinases with myopia, refractive error and ocular biometric measures in an Australian cohort.

Extracellular matrix proteins have been implicated in protein remodelling of the sclera in refractive error. The matrix metalloproteinases (MMPs) falling into the collagenase (MMP1, MMP8, MMP13), gelatinase (MMP2, MMP9) and stromelysin (MMP3, MMP10, MMP11) functional groups are particularly important. We wished to assess their association with myopia, refractive error and ocular biometric measures in an Australian cohort. A total of 543 unrelated individuals of Caucasian ethnicity were genotyped including 269 myopes (≤-1.0D) and 274 controls (>-1.0D). Tag single nucleotide polymorphisms (SNPs) (n = 53) were chosen to encompass these eight MMPs. Association tests were performed using linear and logistic regression analysis with age and gender as covariates. Spherical equivalent, myopia, axial length, anterior chamber depth and corneal curvature were the phenotypes of interest. Initial findings indicated that the best p values for each trait were 0.02 for myopia at rs2274755 (MMP9), 0.02 for SE at both rs3740938 (MMP8) and rs131451 (MMP11), 0.01 for axial length at rs11225395 (MMP8), 0.01 for anterior chamber depth at rs498186 (MMP1) and 0.02 at rs10488 (MMP1). However, following correction for multiple testing, none of these SNPs remained statistically significant. Our data suggests that the MMPs in the collagenase, gelatinase and stromelysin categories do not appear to be associated with myopia, refractive error or ocular biometric measures in this cohort.

Introduction

Refractive errors such as myopia are a group of common ocular disorders that result in blurred vision. The refractive status of the eye can be clinically defined using spherical equivalent (SE) measures and quantitated in dioptres (D). SE is commonly used, both in the clinic and academia, to define the overall refractive status of the eye and determine the nature and degree of refractive error [1]. The overall refractive status of the eye is influenced by a number of underlying components including ocular axial length, corneal curvature and lens thickness [2]. In particular, axial length is the most common factor associated with refractive error. Assessment of the determinants of refractive errors must not only include SE measures but also ocular biometric measures if we are to gain a better overall understanding of refractive error [3]. Unfortunately, often, such measures are lacking in many studies.

Biologically, the size and shape of the eye globe plays a key role in influencing the refractive status of the eye. In particular, an increase in ocular axial length can result in myopia and a reduction in scleral thickness [4]. This thinning of the sclera is not simply due to a passive stretching process but is the result of active remodelling of extracellular matrix components [5]. Scleral remodelling is a dynamic process resulting from an imbalance between the synthesis of extracellular matrix components such as collagens and proteoglycans and the degradation of extracellular matrix components by factors that include matrix metalloproteinases (MMPs).

MMPs are a group of zinc dependent endopeptidases that are involved in the degradation of extracellular matrix proteins. Depending on their substrate specificity, domain structure and cellular localisation, MMPs can be classified into collagenases (MMP1, MMP8, MMP13), gelatinases (MMP2, MMP9), stromelysins (MMP3, MMP10, MMP11), membrane-type (MMP14, MMP15, MMP16, MMP17, MMP24, MMP25) and matrilysins (MMP7, MMP26). Microarray based studies using healthy human donor scleral tissue have shown that multiple MMPs (MMP1–3, MMP7–17, MMP20, MMP24) are expressed suggesting that the scleral remodelling process in humans is, at least in part, driven by these MMPs [6]. Further evidence for the role of MMPs in scleral remodelling has come from animal studies.

Our understanding of the factors driving the scleral remodelling process has increased through the use of animal studies which have implicated the gelatinase, collagenase and stromelysin group of MMPs in myopia and refractive error. Animal studies typically induce myopia using form deprivation techniques that involve the application of monocular goggles that either restrict vision range or restrict clear vision. The scleral remodelling process in form deprivation myopia animal models results from an increase in collagen degradation and a decrease in collagen fibril diameter. In particular, collagen I, the major collagenous component of the mammalian sclera, has been shown to be selectively down regulated in tree shrew and chick models of form deprivation myopia [7]. This down regulation is the direct result of an increase in gelatinase A (MMP2) activity and may also be due to the activity of collagenases whose role is to cleave collagen I [8], [9]. In addition to collagen disturbances, scleral remodelling in form deprivation myopia results from an increase in proteoglycan turnover that is driven by the activities of gelatinase A and stromelysin [10]. For these reasons, MMPs are of interest in genetic association studies for myopia and refractive error.

The collagenase, gelatinase and stromelysin groups of MMPs have been partially analysed in genetic association studies for refractive error and myopia in human cohorts. In the case of refractive error, a recent study assessed, amongst others, these groups of MMPs, with the exception of MMP11, for association with refractive error in two family based cohorts, one of Ashkenazi Jewish origin and one Amish [11]. Positive associations were found in the Amish families with rs9928731 in MMP2 (p = 00026) and rs1939008 (p = 00016) located in the intergenic region between MMP1 and MMP10. These associations were not confirmed in the Ashkenazi Jewish families suggesting a potential founder effect. Thus, so far it is difficult to assess whether these findings are generalisable to non founder cohort groups in the wider population or only specific to certain groups. No other SNPs in this group of MMPs in these families showed positive associations with refractive error. In the case of high grade myopia there have been three studies, all in cohorts of Asian descent. These studies have used either a tag SNP approach in a single gene such as MMP2 [12] or MMP3 [13] or selected SNPs covering multiple MMPs such as MMP1–3 [14] and have not reported positive associations. In the case of common myopia there has been one study in a Caucasian cohort that assessed selected SNPs in MMP1, MMP3 and MMP9 and found a positive association of myopia with the rs3025058 in MMP3 (p = 0.015) and the rs17576(R279Q) in MMP9 (p = 0.026) [15].

Despite many genetic studies analysing variants in the collagenase, stromelysin and gelatinase groups of MMPs, there is still more work that needs to be undertaken in order to have a comprehensive understanding of the role these genes play in refractive error and myopia. There are two areas, in particular, that need addressing. The first is an association study looking at ocular components that contribute to myopia such as axial length, corneal curvature and anterior chamber depth. These traits are yet to be examined in relation to the gelatinase, stromelysin and collagenase groups of MMPs. The second area is a replication study of the positive associations that have been detected so far. These are in MMP2 for refractive error and in MMP3 and MMP9 for common myopia. Given this, our study ultimately aims to address these gaps in our current knowledge in order to both extend, as well as complement existing studies. The purpose of our study is to undertake a comprehensive genetic association study to assess the role of all the collagenase, gelatinase and stromelysin groups of MMPs in refractive error and myopia. An additional assessment as to the role that these MMPs might have in the endophenotypes of axial length, anterior chamber depth and corneal curvature will also be undertaken.

Materials and Methods

Subjects

Individuals were selected from the Genes in Myopia study with the recruitment process previously described [16], [17] [18]. Individuals were excluded based on four criteria (1) history of other eyes pathologies such as retinal detachment or keratoconus, (2) history of genetic disorders known to predispose to myopia, (3) anisometropia >2D difference between eyes and, (4) non-Caucasian ancestry.

Measurements for refraction (SE), axial length, corneal curvature and anterior chamber depth were taken for all individuals as described previously [16], [17]. For the current study myopia was defined as ≤−1.0D in the right eye. DNA from all consenting individuals was collected from venous blood samples [19]. Written informed consent was obtained from all individuals prior to any clinical examination, and ethics approval was provided by the Human Research and Ethics Committee of the Royal Victorian Eye and Ear Hospital, Melbourne. The study was conducted in accordance to the tenets of the Declaration of Helsinki.

SNP selection and genotyping

Tag SNPs encompassing the coding region as well as 2 kb upstream of the start codon and 2 kb downstream of the stop codon of MMP1, MMP2, MMP3, MMP8, MMP9, MMP10, MMP11, MMP13 were chosen. Methodology for choosing tag SNPs has been previously described [18]. Briefly, the Tagger section within the HaploView (version 4.2) software was used to identify tSNPs in these gene utilising a pairwise tagging approach, with the criteria of r2>0.8 and a minor allele frequency (MAF) >10% [20]. Tag SNPs were based on the CEU Hap Map population. All chosen SNPs were genotyped by the Australian Genome Research Facility (Melbourne, Australia) using Illumina® GoldenGate Genotyping assays and the Illumina iScan array scanner.

Statistical analysis

Genotyping data were assessed for deviations from Hardy Weinberg equilibrium using PLINK (version 1.04) [21]. Any SNPs not passing this test in controls (P<0.05) were excluded from the analysis. Association tests were also performed using PLINK. Association tests for refraction (SE), axial length, corneal curvature and anterior chamber depth were performed using linear regression and tests for myopia were performed using logistic regression. All association tests included age and gender as covariates and adjustments for multiple testing using the Bonferroni correction were applied. Power calculations were performed using Quanto version 1.2.4 [22].

Results

Cohort Demographics

A total of 543 individuals were included in this study including 269 with myopia and 274 controls. For this study, myopia was defined as any individual with a SE ≤−1.0D and controls as those with SE >1.0D. This definition of myopia was used to reflect that used in the Hall et al study (2009) which is currently the only other study that assessed MMPs for an association with myopia. Refraction measures for the right and left eyes were highly correlated (r2>0.99) and thus only measures from the right eye were used in our analysis.

The mean age for the overall cohort was 50.9±14.9 years (49.9±16.1 years in the controls; 51.9±13.6 in the myopes). There were 35.8% males overall (31.1% in the controls; 40.7% in the myopes) and 64.2% females (69.0% in the controls; 59.3% in the myopes). A summary of all the clinical measures is shown in Table 1.

Table 1

Clinical characteristics of the study cohort for spherical equivalnet (SE), axial length (AL), corneal curvature (CC) and anterior chamber depth (ACD).

Phenotype	All	Controls	Cases
	Mean (SD)	Mean (SD)	Mean (SD)
SE	−1.8 (3.2)	0.7 (1.2)	−4.3 (2.6)
AL	24.3 (1.5)	23.5 (1.4)	25.1 (1.3)
CC	42.8 (2.8)	43.8 (1.8)	41.9 (3.3)
ACD	3.5 (0.4)	3.5 (0.3)	3.6 (0.4)

Power calculation

Our power calculation showed that using a cohort size of 543 individuals (269 cases and 274 controls) has >80% power to detect an Odds Ratio of 1.8 assuming a minor allele frequency of 0.2 and an alpha of 0.001.

Genetic association tests and power calculation

A total of 53 SNPs were genotyped including eleven in MMP1, ten in MMP2, three in MMP3, six in MMP8, three in MMP9, ten in MMP10, five in MMP11 and five in MMP13 (Table 2). The average SNP call rate was 97.62%. Three SNPs (rs11225426 in MMP1, rs7948454 in MMP10 and rs3758854 in MMP13) were not in Hardy Weinberg Equilibrium and were therefore excluded from further analysis.The threshold for statistical significance for this study defined as P = 0.05/50 = 0.001.

Table 2

List of tagged SNPs genotyped for each matrix metalloproteinase gene.

Gene	Chromosome	Physical Location* (bp)	SNPs
MMP1	11	102,165,861–102,174,104	rs10488, rs11225426, rs1144393, rs2071232, rs3213460, rs470358, rs470504, rs470558, rs470747, rs498186, rs7125062
MMP2	16	54,070,589–54,098,103	rs1053605, rs11541998, rs11639960, rs11646643, rs1992116, rs243835, rs243840, rs243842, rs243866, rs7201
MMP3	11	102,211,738–102,219,552	rs3020919, rs522616, rs639752
MMP8	11	102,088,542–102,100,868	rs11225394, rs11225395, rs12284255, rs1320632, rs2012390, rs3740938
MMP9	20	44,070,954–44,078,606	rs17576, rs2274755, rs3918253
MMP10	11	102,146,444–102,156,554	rs12290253, rs17099562, rs17359286, rs3819099, rs4431992, rs470154, rs470171, rs486055, rs7119084, rs7948454
MMP11	22	22,445,036–22,456,502	rs131451, rs2267029, rs28382576, rs738791, rs738792
MMP13	11	102,318,935–102,331,672	rs11225490,rs17860584,rs3758854,rs478927

HapMap Data Release 28 Phase2/3, August 2010.

Association tests for myopia using logistic regression indicated the best p-value (unadjusted) was 0.04 at rs28382576 for MMP11 (Table 3). Association tests using linear regression analysis showed the best p-values (unadjusted) of 0.02 at rs3740938 in MMP8 for spherical equivalent, 0.01 at rs11225395 in MMP8 for axial length, 0.01 at rs498186 in MMP1 for anterior chamber depth and 0.02 at rs10488 in MMP1 for corneal curvature (Table 4). Following Bonferroni-correction, none of the SNPs retained statistical significance at a threshold of P<0.001.

Table 3

Tagged SNPs with the most associated p-values for each MMP gene using myopia as the trait.

SNP	Gene	Minor Allele	Cases		Controls		P*	OR	95% CI
			Freq.	HWE P	Freq.	HWE P
rs470358	MMP1	A	0.42	0.90	0.40	0.10	0.51	1.09	0.85–1.40
rs243866	MMP2	A	0.23	0.86	0.26	0.43	0.28	0.85	0.64–1.14
rs639752	MMP3	A	0.52	0.71	0.48	0.33	0.18	1.18	0.93–1.50
rs3740938	MMP8	A	0.09	1.00	0.06	1.00	0.05	1.58	1.00–2.51
rs2274755	MMP9	A	0.20	0.33	0.14	0.81	0.02	1.48	1.06–2.06
rs17099562	MMP10	A	0.06	0.58	0.04	0.38	0.20	1.44	0.83–2.52
rs28382576	MMP11	A	0.07	1.00	0.04	1.00	0.04	1.84	1.03–3.26
rs478927	MMP13	A	0.36	0.50	0.30	0.77	0.07	1.27	0.98–1.64

Unadjusted P-values with a significance threshold of 0.05/50 = 0.001.

Table 4

Tagged SNPs with the most associated p-values for each MMP gene using spherical equivalent (SE), axial length (AL), anterior chamber depth (ACD) and corneal curvature (CC) as the traits.

Gene	Trait	SNP	Minor Allele		HWE	P*	OR	95% CI
			Name	Freq.
MMP1	SE	rs470747	G	0.36	0.93	0.10	1.40	0.93–2.10
	AL	rs10488	A	0.06	0.46	0.28	0.81	0.55–1.19
	ACD	rs498186	C	0.43	0.59	0.01	0.94	0.90–0.99
	CC	rs10488	A	0.06	0.46	0.02	2.40	1.17–4.92
MMP2	SE	rs11541998	C	0.11	0.13	0.17	1.52	0.83–2.77
	AL	rs11541998	C	0.11	0.13	0.25	0.83	0.61–1.13
	ACD	rs11541998	C	0.11	0.13	0.17	1.06	0.98–1.14
	CC	rs1053605	A	0.07	0.35	0.07	1.85	0.94–3.64
MMP3	SE	rs522616	G	0.21	0.19	0.37	1.23	0.78–1.96
	AL	rs522616	G	0.21	0.19	0.35	0.90	0.72–1.12
	ACD	rs3020919	A	0.24	0.29	0.15	0.96	0.91–1.02
	CC	rs3020919	A	0.24	0.29	0.51	1.15	0.75–1.76
MMP8	SE	rs3740938	A	0.08	1.00	0.02	0.42	0.20–0.85
	AL	rs11225395	A	0.46	0.93	0.01	1.28	1.05–1.54
	ACD	rs12284255	A	0.07	0.52	0.08	0.93	0.85–1.01
	CC	rs11225395	A	0.46	0.93	0.09	0.73	0.51–1.05
MMP9	SE	rs3918253	G	0.45	0.43	0.23	0.79	0.54–1.16
	AL	rs2274755	A	0.17	0.54	0.18	1.14	0.94–1.37
	ACD	rs3918253	G	0.45	0.43	0.43	0.97	0.91–1.04
	CC	rs3918253	G	0.45	0.43	0.30	0.83	0.58–1.18
MMP10	SE	rs470154	A	0.06	0.40	0.31	0.64	0.27–1.52
	AL	rs486055	A	0.15	0.87	0.06	1.29	0.99–1.69
	ACD	rs7948454	G	0.08	0.02	0.09	1.07	0.99–1.16
	CC	rs17099562	A	0.05	0.39	0.06	2.17	0.98–4.83
MMP11	SE	rs131451	G	0.12	0.29	0.02	0.50	0.28–0.90
	AL	rs738791	A	0.49	0.93	0.15	1.15	0.95–1.40
	ACD	rs738791	A	0.49	0.93	0.10	1.04	0.99–1.09
	CC	rs738792	G	0.11	0.51	0.31	0.75	0.43–1.30
MMP13	SE	rs478927	A	0.33	0.77	0.40	0.84	0.56–1.26
	AL	rs10502009	G	0.11	1.00	0.19	0.82	0.61–1.10
	ACD	rs3758854	A	0.07	0.01	0.29	1.05	0.96–1.15
	CC	rs10502009	G	0.11	1.00	0.21	1.42	0.82–2.48

Unadjusted P-values with a significance threshold of 0.05/50 = 0.001.

Discussion

Our study has undertaken a detailed genetic analysis into the association of the gelatinase, collagenase and stromelysin groups of MMPs in myopia and refractive error. More importantly this is the first study to assess associations of these MMPs in the ocular biometric measures of axial length, anterior chamber depth and corneal curvature. To date this is the most comprehensive study into this group of MMPs that has been undertaken whereas previous studies were more limited, in that only selected variants were chosen or there was no analysis of endophenotypes. Our methodological approach used a tag SNP strategy for association testing that allowed for complete genetic coverage of all known SNPs in the coding regions, intronic regions and 2 kb upstream of the start codon and 2 kb downstream of the stop codon of these MMPs. Using this approach, we were not able to detect any statistically significant association with the phenotypes analysed. There have been two SNPs rs3025058 (MMP3) and rs17576 (R279Q; MMP9) previously associated with myopia [15]. We directly genotyped rs17576 and were not able to confirm the association with common myopia in our cohort (Tables S1 and S2). In the case of rs3025058 we did not directly genotype this SNP. This SNP is physically located with the region tagged by our SNPs but there is no genotype information available from the Hap Map reference population (CEU) to enable assessment of the LD relationships between this SNP and those that we genotyped. Hence we cannot comment on weather the SNPs we genotyped will act as proxies for this SNP and it will have to be directly genotyped in order to confirm its lack of association with common myopia in our cohort.

Although our study did not indicate association of genetic variants with refractive error for these MMPs, there may be other extracellular matrix component genes that play a role. Extracellular matrix components such as collagen type I alpha 1 (COL1A1), collagen type II alpha 1 (COL2A1), lumican (LUM), decorin (DCN) and epiphycan (EPYC or DSPG3) have been previously assessed for genetic associations with myopia. These include assessment of DSPG3 and DCN with high myopia but resulted in no association whereas assessment of COL1A1, COL2A1 and LUM showed both positive and negative associations in different studies [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33]. Of these genes only COL2A1 has been assessed for associations with common myopia with a positive result being reported [34].

Evidence for genetic variants in extracellular matrix components as playing a role in refractive error is not strong so far. Most evidence has been derived from alterations in the expression of these genes where strong evidence from animal studies has shown that extracellular matrix components such as MMP2, proteoglycans and type I collagen are differentially expressed in form deprivation myopia where vision is modified using artificial lenses or translucent occludes. [7], [8], [35], [36]. These studies in combination with the lack of genetic association reported in our work suggest that single base changes in the primary DNA sequence, at least in these genes, are not the main driving force behind myopia in this cohort. This does not imply that these genes do not play a role in myopia in general but simply relates to the state of allele associations in the current study cohort. It should also be noted that the ability to detect an association in any population is dependent on the SNPs present which may in some cases account for lack of reproducibility between studies. In addition to gene expression changes it is also possible large structural variations, rather than single nucleotide polymorphisms in the DNA may play a role in myopia. These would include deletions, duplications and more complicated genomic rearrangements, and are commonly referred to as structural or copy number variations. In support of this, evidence has emerged that show copy number variations at chromosome Xq28 play a role in X-linked myopia that is associated with other clinical features such as color vision deficiencies [37]. Structural variations have been previously implicated in many other diseases such as age related macular degeneration [38], [39].

The study design used for this study has many strengths including the choice of cohort and the tag SNP approach. The cohort was chosen to include a homogenous population of Caucasian ancestry where we collected ancestry information from at least two generations. This allowed minimization of the potential effects of population admixture. The power calculation for this study suggests that the cohort size utilised is of sufficient size to detect modest genetic effects up to an odds ratio of 1.8. Although our cohort may potentially be too small to detect very small effects, the volume of the sample required to detect these effects exceeded the capacity of the recruitment processes of this study. However, our sample size is within the range of what has been previously reported which strengthens our justification for using a cohort of this size. In addition to this the reported relatively narrow confidence intervals also suggest that the study had suffient power to detect relavant association. The tag SNP approach is also important as it allows good coverage of all the SNPs in these genes. However, the tag SNP approach also has limitations in that in this study it only covered common SNPs with a minor allele frequency of 0.1 or greater and hence rare variants may be excluded. Additionally the tag SNP approach does not cover genetic variants such as epigenetic modification, copy number variations and structural variations that may influence the development of disease. Clearly, more work needs to be undertaken for these genes in order to gain a complete understanding of all the potential genetic variations that may contribute to the development of myopia.

Our study suggests that polymorphisms in MMP genes categorised as collagenases, stromelysins and gelatinases do not play a major role in refractive error, myopia, axial length, corneal curvature and anterior chamber depth. Although there is strong evidence that these genes are involved in the sclera remodelling process that accompanies myopia, we propose that their role is not driven by single nucleotide polymorphisms but instead is influenced by other genetic changes that might include copy number changes, epigenetic changes or alterations in the regulatory elements of the genes all of which may results in changes in expression.

Results for all tagged SNPs for each MMP using myopia as the trait.

(DOCX)

Click here for additional data file.

Results for all tagged SNPs for each MMP using spherical equivalent (SE), axial length (AL), anterior chamber depth (ACD) and corneal curvature (CC) as the traits.

(DOCX)

Click here for additional data file.