Association of Polymorphisms of Tumor Necrosis Factor and Tumor Protein p53 with Primary Open-Angle Glaucoma

Abstract

Purpose.: To evaluate the variants of 10 genes for association with primary open-angle glaucoma (POAG) in a Chinese population.

Methods.: A total of 405 unrelated patients with POAG (255 high-tension glaucoma [HTG], 100 normal-tension glaucoma [NTG], and 50 juvenile-onset open-angle glaucoma [JOAG]) and 201 control subjects were recruited. Seventeen variants in 10 genes with reported association with POAG were genotyped for analysis of allele and haplotype frequencies between cases and control subjects. These genes included CDH1 (cadherin 1, type 1, E-cadherin), CDKN1A (cyclin-dependent kinase inhibitor 1A), CYP1B1 (cytochrome P450, family 1, subfamily B, polypeptide 1), GSTM1 (glutathione S-transferase mu 1), GSTT1 (glutathione S-transferase theta 1), MTHFR (5,10-methylenetetrahydrofolate reductase), NOS3 (nitric oxide synthase 3), OPA1 (optic atrophy 1), TNF (tumor necrosis factor), and TP53 (tumor protein p53).

Results.: One SNP (−308G>A; rs1800629) in TNF demonstrated a significant association with HTG (P = 0.012). The allele G frequency was higher in HTG patients than in control subjects (94.6% vs. 90.3%; OR = 1.89). One haplotype consisting of rs1799724 and rs1800629 was significantly associated with HTG (P = 0.015, corrected P = 0.045). One SNP (R72P; rs1042522) in TP53 was significantly associated with NTG (P = 0.018). The allele G frequency was higher in NTG patients than in control subjects (56.1% vs. 45.8%; OR = 1.52). The significance of these associations survived the Bonferroni correction (corrected P < 0.024). Other gene variants were not significantly associated with HTG (P > 0.063) or NTG (P > 0.13). None of the studied variants was significantly associated with JOAG (P > 0.17).

Conclusions.: The findings suggest that variants in TNF and TP53 are risk factors for POAG, whereas variants in other studied genes are not major risk factors for POAG, at least in the Chinese population.

Glaucoma is a heterogeneous group of diseases characterized by apoptosis of the retinal ganglion cells and progressive degeneration of the optic nerve. It is the second leading cause of blindness worldwide, estimated to affect more than 60 million people by 2010. ¹ Primary open-angle glaucoma (POAG) and exfoliation glaucoma (XFG) are common forms of glaucoma in most populations. Family segregation studies have shown that genetic factors make major contributions to the etiology of these diseases. ² Lysyl oxidase-like 1 (LOXL1) was recently associated with XFG as a major gene, accounting for more than 90% of XFG cases in most populations. ^3–6 However, the genetics of POAG appears to be more complex, and no major associated genes have been identified thus far. ⁷  

At least 20 genetic loci have been linked to POAG, 14 of which are designated GLC1A to GLC1N. ^7,8 Only three causative genes have been identified from these loci: myocilin (MYOC), ⁹ optineurin (OPTN), ¹⁰ and WD repeat domain 36 (WDR36). ¹¹ Mutations in these genes account for only approximately 10% of patients with POAG. ^7,12,13 More than 20 other genes have been associated with POAG, but most of them have been reported in single studies; only a few of them have been investigated in multiple studies. ^7,14 However, reports on these associations, such as those for optic atrophy 1 (OPA1), are often conflicting in the different studies, probably due to small samples or ethnic differences. ^15,16 Thus, further evaluation of these genes is warranted in larger samples and different populations. 

Retinal ganglion cell death in glaucoma has been characterized—especially that involving the apoptotic pathway. ¹⁷ Growing evidence supports the possibility that oxidative stress and mitochondrial dysfunction contribute to retinal ganglion cell death by apoptosis and lead to glaucoma. ^18,19 In addition, elevated intraocular pressure (IOP) is an important risk factor for glaucoma. Thus, an understanding of the genes that are involved in apoptosis, oxidative stress, mitochondrial dysfunction, and regulating IOP should assist in elucidation of the mechanisms of glaucoma. In the present study, we investigated 10 genes for their association with POAG in a Chinese case–control cohort containing 405 patients with POAG and 201 control subjects. Five of these genes are involved in apoptosis: CDH1 (cadherin 1, type 1, E-cadherin), ²⁰ CDKN1A (cyclin-dependent kinase inhibitor 1A), ²¹ MTHFR (5,10-methylenetetrahydrofolate reductase), ²² TNF (tumor necrosis factor), ²³ and TP53 (tumor protein p53). ²⁴ Two are involved in oxidative stress: GSTM1 (glutathione S-transferase mu 1) ²⁵ and GSTT1 (glutathione S-transferase theta 1). ²⁶ One is associated with mitochondrial dysfunction: OPA1. ¹⁵ Two are involved in regulating IOP: CYP1B1 (cytochrome P450, family 1, subfamily B, polypeptide 1) ²⁷ and NOS3 (nitric oxide synthase 3). ²⁸ Seventeen variants in these genes that have been reported to be significantly associated with POAG were analyzed in the present study. 

Methods

Patients and Control Subjects

Patients with POAG were recruited from the Eye Clinic of the Prince of Wales Hospital, Hong Kong. POAG was defined as meeting all the following criteria: exclusion of secondary causes (e.g., trauma, uveitis, steroid-induced glaucoma, or exfoliation glaucoma), Shaffer grade III or IV open iridocorneal angle on gonioscopy, characteristic optic disc damage or typical visual field loss by standard automated perimetry with the Glaucoma Hemifield Test (Humphrey Field Analyzer; Carl Zeiss Meditec, Inc., Dublin, CA). IOP was determined by applanation tonometry. Control subjects were recruited from people who attended the same clinic for conditions of senile cataract, floaters, refractive errors, or itchy eyes. They were excluded from glaucoma using the same criteria of diagnosis as was used for the POAG patients after going through the same ophthalmic examination. This study was approved by the Ethics Committee for Human Research, the Chinese University of Hong Kong. Informed consent was obtained from all study subjects after explanation of the nature and possible consequences of the study, in accordance with the tenets of the Declaration of Helsinki. 

A cohort of 405 unrelated patients with POAG and 201 unrelated control subjects without glaucoma were included in this study. The demographic and clinical features of the study subjects were summarized in Table 1. In the POAG group, 255 patients had late-onset HTG with highest IOP ranging from 22 to 67 mm Hg (mean ± SD: 29.8 ± 8.0 mm Hg), 100 patients had late-onset NTG, having highest IOP between 10 and 21 mm Hg (mean ± SD: 17.6 ± 3.0 mm Hg), and 50 patients were JOAG whose age at diagnosis was between 4 and 34 years (mean ± SD: 21.3 ± 9.2 years) and highest IOP ranged from 25 to 69 mm Hg (mean ± SD: 33.7 ± 6.1 mm Hg). Age at inclusion in the control group ranged from 50 to 90 years (mean ± SD: 69.8 ± 8.7 years), the highest IOP ranged from 6 to 21 mm Hg (mean ± SD: 15.1 ± 3.4 mm Hg), and vertical cup-disc ratio from 0.2 to 0.5 (mean ± SD: 0.38 ± 0.08). Visual fields were within normal range. They had no family history of glaucoma. 

Table 1.

 

Demographic and Clinical Features of the Study Subjects

Group	n	Gender (M/F)	Age at Diagnosis* (y)		Highest IOP (mm Hg)		Vertical Cup–Disc Ratio
			Range	Mean ± SD	Range	Mean ± SD	Range	Mean ± SD
HTG	255	167/88	35–90	62.8 ± 12.4	22–67	29.8 ± 8.0	0.7–1.0	0.82 ± 0.08
NTG	100	54/46	35–88	63.2 ± 11.5	10–21	17.6 ± 3.0	0.7–1.0	0.85 ± 0.06
JOAG	50	33/17	4–34	21.3 ± 9.2	25–69	33.7 ± 6.1	0.8–1.0	0.81 ± 0.07
Control	201	120/81	50–90	69.8 ± 8.7	6–21	15.1 ± 3.4	0.2–0.5	0.38 ± 0.08

*  For the control subjects, age at diagnosis refers to age at inclusion.

All the subjects were Han Chinese living in Hong Kong. They were recruited from the same eye clinic and had a similar ethnic background. The cases and control subjects were matched for sex, with 62.7% and 59.7% male POAG patients and control subjects, respectively (P = 0.47). Because of the age-dependence of POAG, only control subjects older than age 50 were included in the study. 

Genotyping

Genomic DNA was extracted from 200 μL of whole blood with a commercial kit (Qiamp Blood Kit; Qiagen, Hilden, Germany). Quantification of extracted DNA was performed with a spectrophotometer (model ND-1000; NanoDrop Technologies, Wilmington, DE). The deletion variants in GSTM1 and GSTT1 were determined by using the multiplex polymerase chain reaction (PCR) protocol, as previously described. ²⁹ The amplified PCR products were identified by electrophoresis in a 1.5% agarose gel with 0.5 μg/mL ethidium bromide. Individuals with homozygous deletions were deemed to have a null genotype and those with one or two copies of the relevant genes had a positive genotype. Single-nucleotide polymorphisms (SNPs) in other genes (CDH1, CDKN1A, CYP1B1, MTHFR, NOS3, OPA1, TNF, and TP53) were genotyped either by SNP genotyping assay (TaqMan; Applied Biosystems [ABI], Foster City, CA) or by direct sequencing. Oligonucleotide primers were obtained from ABI (assay by demand) and the assays were performed according to the manufacturer's instructions. For direct sequencing, products from PCR amplification were purified and sequenced using dye-termination chemistry (BigDye; ABI) and an automated genetic analyzer (model 3130; ABI). 

Statistical Analysis

Statistical analyses for the SNPs were performed using PLINK (ver. 1.06), which is a free statistical analysis toolset, designed to perform a range of basic and large-scale analyses for genome-wide association studies in a computationally efficient manner. ³⁰ Hardy-Weinberg equilibrium was assessed by using the χ² test. The minor allele frequencies of each SNP between patients with HTG, NTG, or JOAG and control subjects were compared by using Fisher's exact test. Odds ratio (OR) and 95% confidence interval (CI) were calculated by using the logistic regression method. Haplotype frequencies were estimated with the standard E-M algorithm and tested with the χ² test. The omnibus P-value was obtained from the omnibus test. Specific P-values were obtained from the haplotype-specific test. OR and 95% CI were calculated for each haplotype compared with all the other haplotypes. 

Statistical analyses for the deletions in GSTM1 and GSTT1 were performed (SAS ver. 9.1; SAS Institute, Cary, NC). Genotype frequencies of each gene between patients with HTG, NTG, or JOAG and control subjects were compared by Fisher's exact test. 

Multiple comparisons were corrected by using the Bonferroni method. Gene-wide correction was performed with all studied variants within each gene. 

Results

Characteristics and genotype counts of the 17 variants were summarized in Table 2. All SNPs followed Hardy-Weinberg equilibrium in the control group. One promoter SNP (−308G>A; rs1800629) in TNF demonstrated a significant association with HTG (P = 0.012; Table 3). The allele G frequency of rs1800629 was higher in the HTG patients than in the control subjects (94.6% vs. 90.3%; OR = 1.89; 95% CI, 1.14–3.13). One haplotype consisting of rs1799724 and rs1800629 in TNF was significantly associated with HTG (P = 0.015, Bonferroni-corrected P = 0.045; Table 4). One nonsynonymous SNP (R72P; rs1042522) in TP53 was significantly associated with NTG (P = 0.018; Table 3). The allele G frequency of rs1042522 was higher in the NTG patients than in the control subjects (56.1% vs. 45.8%; OR = 1.52; 95% CI, 1.08–2.13). These associations survived the gene-wide Bonferroni correction (corrected P < 0.024). Other gene variants were not significantly associated with HTG (P > 0.063) or NTG (P > 0.13). None of the studied gene variants was significantly associated with JOAG (P > 0.17; Table 3). 

Table 2.

 

Characteristics and Genotype Counts of the 17 Variants in 10 Genes

Gene	SNP	Chr	Position (bp)*	Location	Sequence Change	Codon Change	Minor Allele†	Genotype Count (AA/AB/BB)‡
								HTG	NTG	JOAG	Control Subjects
CDH1	rs1801026	16	67424957	3′UTR	c.2649+54C>T	NA	T	8/73/171	1/29/69	2/11/33	9/52/140
CDKN1A	rs3176352	6	36760317	Intron 2	IVS2+16C>G	NA	C	39/131/82	17/50/32	12/20/14	36/89/76
CDKN1A	rs1801270	6	38207008	Exon 2	c.93C>A	S31R	A	58/127/67	19/51/29	10/20/16	48/101/52
CYP1B1	rs1800440	2	38151643	Exon 3	c.1358A>G	N453S	G	0/1/251	0/1/98	0/0/46	0/0/201
CYP1B1	rs1056836	2	38151707	Exon 3	c.1294G>C	V432L	G	1/39/212	2/13/84	0/9/37	1/35/165
CYP1B1	rs10012	2	38155894	Exon 2	c.142C>G	R48G	G	14/79/159	3/35/61	0/16/30	9/72/120
MTHFR	rs1801133	1	11778965	Exon 5	c.665C>T	A222V	T	11/87/154	5/30/64	0/20/26	6/60/135
NOS3	rs2070744	7	150321012	Promoter	−813C>T	NA	C	6/53/193	2/16/81	2/8/36	4/40/157
NOS3	rs3918226	7	150321109	Promoter	−716C>T	NA	T	0/0/252	0/0/99	0/0/46	0/0/201
NOS3	rs1799983	7	150327044	Exon 8	c.894T>G	D298E	T	5/41/206	1/16/82	0/15/31	1/43/157
OPA1	rs166850	3	194837768	Intron 8	IVS8+4C>T	NA	T	0/35/217	1/9/89	0/4/42	1/27/173
OPA1	rs10451941	3	194837796	Intron 8	IVS8+32T>C	NA	C	26/125/101	12/49/38	9/21/16	23/104/74
TNF	rs1799724	6	31650461	Promoter	−857C>T	NA	T	5/58/189	0/19/80	0/11/35	3/39/159
TNF	rs1800629	6	31651010	Promoter	−308G>A	NA	A	1/25/226	0/12/87	0/9/37	5/29/167
TP53	rs1042522	17	7520197	Exon 4	c.215G>C	R72P	C	64/114/74	22/43/34	14/19/13	55/108/38
GSTM1	GSTM1	1	110032359	Whole gene	Positive>null	Deletion	Null	121/131	41/58	19/27	81/119
GSTT1	GSTT1	22	22706462	Whole gene	Positive>null	Deletion	Positive	123/129	42/57	20/26	94/106

 NA, not available.

*  Chromosomal position is based on NCBI Build 36.2 (National Center for Biotechnology Information, Bethesda, MD).

†  For GSTM1 and GSTT1, minor allele refers to minor genotype.

‡  A represents the minor allele and B represents the common allele. For GSTM1 and GSTT1, the count indicates minor/common genotype.

Table 3.

 

Single-Variant Association of the Studied Genes with HTG, NTG, and JOAG

Gene	SNP	Minor Allele*	Minor Allele Frequency*				P †
			HTG	NTG	JOAG	Control Subjects	HTG	NTG	JOAG
CDH1	rs1801026	T	0.177	0.157	0.163	0.174	0.92	0.59	0.80
CDKN1A	rs3176352	C	0.415	0.424	0.478	0.401	0.67	0.58	0.17
CDKN1A	rs1801270	A	0.482	0.450	0.435	0.490	0.81	0.35	0.34
CYP1B1	rs1800440	G	0.002	0.005	0	0	0.37	0.15	NA
CYP1B1	rs1056836	G	0.081	0.086	0.098	0.092	0.57	0.80	0.86
CYP1B1	rs10012	G	0.212	0.207	0.174	0.224	0.67	0.64	0.29
MTHFR	rs1801133	T	0.216	0.202	0.217	0.179	0.16	0.50	0.39
NOS3	rs2070744	C	0.129	0.101	0.130	0.119	0.67	0.50	0.77
NOS3	rs3918226	T	0	0	0	0	NA	NA	NA
NOS3	rs1799983	T	0.101	0.091	0.163	0.112	0.60	0.43	0.18
OPA1	rs166850	T	0.069	0.056	0.043	0.072	0.88	0.44	0.32
OPA1	rs10451941	C	0.351	0.369	0.424	0.373	0.49	0.92	0.37
TNF	rs1799724	T	0.135	0.096	0.120	0.112	0.30	0.55	0.84
TNF	rs1800629 ‡	A	0.054	0.061	0.098	0.097	0.012	0.13	0.98
TP53	rs1042522 §	C	0.480	0.439	0.511	0.542	0.063	0.018	0.59
GSTM1	Gene deletion	Null	0.480	0.414	0.413	0.405	0.13	0.90	1.00
GSTT1	Gene deletion	Positive	0.488	0.424	0.435	0.470	0.71	0.46	0.74

 Significant SNPs and P-values are shown in bold type.

*  For GSTM1 and GSTT1, minor allele refers to minor genotype.

†  Obtained from Fisher's exact test versus control subjects. NA, not available.

‡  The allele G frequency of rs1800629 was higher in the HTG patients (94.6%) than in control subjects (90.3%; P = 0.012; OR = 1.89, 95% CI = 1.14–3.13).

§  The allele G frequency of rs1042522 was higher in the NTG patients (56.1%) than in control subjects (45.8%; P = 0.018; OR = 1.52, 95% CI = 1.08–2.13).

Table 4.

 

Haplotype Analysis of TNF in Patients with HTG and Control Subjects

Haplotype (rs1799724-rs1800629)	Haplotype Frequency		P	OR (95% CI)
	HTG	Control
CG	0.815	0.802	0.58	1.10 (0.79–1.53)
TG	0.133	0.106	0.23	1.28 (0.86–1.93)
CA *	0.052	0.092	0.015	0.53 (0.32–0.89)
Total	1.000	1.000	0.04†

*  Haplotype CA shown in bold type was significantly associated with HTG (P = 0.015; Bonferroni-corrected P = 0.045).

†  This P value was obtained from the omnibus test, whereas the other P-values were obtained from the haplotype-specific test. OR was calculated for each haplotype compared with all the other haplotypes.

Discussion

In the present study, we investigated 17 variants in 10 genes for association with POAG in a Chinese case–control sample. Except for one SNP rs3176352 in CDKN1A, all the other gene variants have been reported to be significantly associated with POAG. However, most of them have been reported only in single studies, and reports of associations of the other variants often conflict in multiple studies. Replication of the association of these gene variants with POAG in an independent sample is thus warranted. In the present study, only one SNP (rs1800629) in TNF and one SNP (rs1042522) in TP53 were found to be significantly associated with HTG and NTG, respectively. All the other 15 variants were not significantly associated with HTG or NTG. In addition, none of the studied variants was significantly associated with JOAG, suggesting that these genes are not major risk factors for JOAG. 

Our results suggested that TNF and TP53 have significant roles in the development of POAG. Intriguingly, both TNF and p53 are important components involved in the apoptosis pathway. In experimental models, an increased TNF expression level has been demonstrated in glaucomatous eyes, ^31,32 and TNF has been reported to participate directly in the apoptosis of retinal ganglion cells. ³³ p53 is involved in the apoptosis of retinal ganglion cells through acting as a transcription factor and upregulating the expression of the proapoptotic gene BAX and downregulating the expression of the antiapoptotic gene BCL2. ³⁴ In zebrafish mutants, loss of Wdr36 function leads to an activation of the p53 stress-response pathway, suggesting that defects in p53 pathway genes may influence the impact of WDR36 variants on POAG. ³⁵  

TNF was initially reported to associate with POAG in a Chinese population (Fig. 1). The allele A frequency of the promoter SNP rs1800629 (−308G>A) was higher in patients with POAG than in control subjects (42.5% vs. 21.4%; OR = 2.72; 95% CI, 1.66–4.45). ²³ This association was replicated in an Iranian population, with a higher allele A frequency of rs1800629 in patients with POAG than in control subjects (9.3% vs. 2.5%; OR = 3.99; 95% CI, 1.71–9.33). ³⁶ However, the frequencies of allele A in both cases and control subjects in the Iranian study were much lower than those reported in the Chinese population. ^23,36 A study in a Japanese population revealed no significant association between three promoter SNPs (−308G>A, −857C>T and −863C>A) of TNF and POAG, despite a possible interaction between TNF and OPTN. ³⁷ Similarly, a study in a Caucasian population from southern Austria also did not find a significant association between two promoter SNPs (−308G>A and −238G>A) of TNF and POAG. ³⁸ In the present study, we identified a significant association between TNF and HTG. In contrast to previous studies in which the allele A of rs1800629 was more frequent in POAG patients than in control subjects, ^23,36 we found that the allele G frequency was higher in the HTG patients than in the control subjects (94.6% vs. 90.3%; OR = 1.89; 95% CI, 1.14–3.13; Table 3). Our haplotype analysis further confirmed this association (Table 4). The allele A frequency in our control subjects (9.7%) was similar to that in the HapMap data from Chinese (3.3%) but was much lower than that reported in the initial Chinese study (21.4%). ²³ In a study of lipopolysaccharide-stimulated whole-blood-cell cultures, the allele A of rs1800629 was associated with increased TNF production. ³⁹ However, investigators in another study were unable to replicate this effect of the allele A on TNF production. ⁴⁰ In the present study, the allele A of rs1800629 was found to be protective against POAG, whereas the allele G was a risk factor. A similar protective effect of the allele A has been indicated in other diseases in specific populations, such as XFG in Turks and ischemic stroke in Asians. ^41,42 It remains unclear how the G allele works as a risk factor to influence the development of POAG. 

Figure 1.

 

The gene structure of TNF and TP53 and the polymorphic sites for association with POAG published to date. Solid squares: translated coding exons; horizontal lines: introns. The polymorphisms shown inside the squares were investigated in the present study.

TP53 was originally shown to be associated with POAG in a Chinese population (Fig. 1). The allele C frequency of the nonsynonymous SNP rs1042522 (R72P) was higher in POAG patients than in control subjects (57% vs. 36%; OR = 2.39; 95% CI, 1.14–5.01). ²⁴ Two studies in Caucasian populations replicated this association and showed that specific haplotypes consisting of rs1042522 and rs17878362 (a 16-bp ins/del polymorphism in intron 3) were significantly associated with POAG. ^43,44 However, no significant association was identified between rs1042522 and POAG in several studies in Indian, Australian, Japanese, Turkish, and Brazilian populations. ^45–49 In addition, no significant association was found between rs17878362 and POAG in Indian and Australian populations. ^45,46 In contrast to the original study in the Chinese, in which the allele C of rs1042522 was more frequent in POAG patients than in control subjects, ²⁴ we found that the allele G frequency was higher in our NTG patients than in the control subjects (56.1% vs. 45.8%; OR = 1.52; 95% CI, 1.08–2.13; Table 3). Our results are consistent with those in a study by Daugherty et al., ⁴⁴ who reported a higher frequency of the allele G in POAG patients than in control subjects (80% vs. 71%) and even higher in NTG (84%) patients than in HTG (78%) patients. Note that the allele G frequency in our control subjects (45.8%) was similar to that in the HapMap data from Chinese (51.1%) subjects but was much lower than that in the original Chinese study (64%). ²⁴ SNP rs1042522 is a common sequence polymorphism, located in the proline-rich domain of p53, which is essential for the protein to fully induce apoptosis. In cell lines containing inducible alleles and in cells with endogenous p53, the G allele of rs1042522 was shown to have a 15-fold increased apoptotic ability compared with the C allele. ⁵⁰ Thus, it is possible that concurrence of the G allele with higher risk of POAG is due to increased susceptibility of retinal ganglion cells to apoptosis. On the basis of p53 involvement in the development of POAG, apoptosis would more likely have a greater role in NTG than in HTG. ⁴⁶ This hypothesis is further supported by findings of a significant association of TP53 with NTG other than HTG in our present study and the study by Daugherty et al. ⁴⁴  

In the present study, no significant association was identified for CDH1, CDKN1A, CYP1B1, GSTM1, GSTT1, MTHFR, NOS3, and OPA1, suggesting that genetic variants in these genes may not be major risk factors for the development of POAG in Chinese patients. 

Association with POAG of a 3′UTR SNP (rs1801026) in the CDH1 gene and a SNP in exon 2 (rs1801270) of CDKN1A had been reported in the Chinese. ^20,21 We did not reveal any significant association of either SNP with POAG in this study on an independent and larger Chinese sample. For rs1801026, whereas the allele T frequency was similar in POAG patients, in control subjects, it was much lower in our study (17.4%) than in the original study (52.4%). ²⁰ However, our results in control subjects are very close to statistics in the dbSNP database for Asians (16.7%). As for rs1801270, association with POAG had not been replicated in two studies of Caucasians. ^48,51  

Rare heterozygous mutations in the CYP1B1 gene have been consistently reported in POAG, especially in JOAG. ^27,52 One study of an Indian population has suggested an association between a common SNP (rs1056836) in exon 3 of CYP1B1 and POAG. ⁵³ However, our study, like many other reported studies, did not support the notion that common polymorphisms of CYP1B1 are associated with the development of POAG. ^54–57 It is noteworthy that unlike in Caucasians, the allele G of rs1800440 was very rare in our Chinese sample, at 0.5% in patients with POAG, and is completely absent in control subjects, consistent with the HapMap database (1.1% in Chinese). We think it is more likely that CYP1B1 contributes to only a small proportion of POAG with Mendelian inheritance. However, complex forms of POAG are unlikely to be due to common polymorphisms of CYP1B1. 

The GSTM1-positive genotype was first shown to be associated with POAG in an Estonian population. ²⁵ Only one study in a Turkish population support this finding, ²⁶ whereas several other studies did not find any association. ^58,59 There were even reports of a reversed association in which the null genotype was more common in POAG patients. ^60–62 The GSTT1 null genotype was associated with POAG in studies of a Turkish population ²⁶ and of an Arabian population, ⁶² but not in other Turkish or European studies. ^25,59–61 In the present study in the Chinese, we did not find any association of GSTM1 and GSTT1 with POAG. 

One SNP in exon 5 (rs1801133) of the MTHFR gene was associated with POAG in a German population. ²² However, this association with POAG has not yet been replicated in other studies, including three studies of Caucasians ^63–65 and two of Asians, ^66,67 although one study of Koreans suggested a possible association with younger NTG. ⁶⁸ In the present study, we did not find any association between this SNP and POAG in the Chinese. 

A promoter SNP (rs3918226) in the NOS3 gene associated with POAG was first reported in an Australian population. ²⁸ However, another promoter SNP (rs2070744) of NOS3 was not associated with POAG in two other studies. ^69,70 In addition, the 27-bp repeat alleles in intron 4 of NOS3 were associated with POAG in British and Pakistani populations ^70,71 but not in a Chinese population. ⁶⁹ In the present study, we did not find any association between two NOS3 SNPs (rs2070744 and rs1799983) and POAG. Of note, the minor allele T of rs3918226 in NOS3 was not found in our Chinese sample (Table 3). One recent study in Caucasians demonstrated that three SNPs of NOS3 (rs3918188, rs2070744, and rs1800779) were significantly associated with HTG among women but not among men. ⁷² The other two SNPs (rs1799983 and rs7830) did not show this sex-dependent association. In the present study, stratified analysis did not identify a significant difference between the men and the women in association of two NOS3 SNPs (rs2070744 and rs1799983) with HTG (P-heterogeneity by sex > 0.14; data not shown). 

Two SNPs (rs166850 and rs10451941) in intron 8 of the OPA1 gene were initially associated with NTG in a British population. ¹⁵ To date, this association has been replicated in two studies of Caucasians ^73,74 and in one Japanese study, ⁷⁵ but not in two other studies of Caucasians, ^16,57 two studies of Africans, ^16,76 and one study of Koreans. ⁷⁷ In the present study, we did not find any association of OPA1 with POAG in our Chinese sample. 

The pathogenesis of POAG is genetically heterogeneous and complex. Despite several genetic loci and genes that have been associated with POAG, the major genes that confer significant susceptibility remain unknown. ^7,14 In particular, the complex forms of POAG are most likely caused by interactions of multiple genes and environmental factors. To date, most association studies for POAG have investigated only single genes or single gene variants without accounting for contributions from gene–gene and gene–environment interactions. This deficit in studies of POAG-associated genes may be a main reason for the conflicting results in different association studies, in addition to ethnic differences. An ideal replication study would be performed in the same population with exactly the same sampling criteria as in the original study. Unfortunately, almost none of the association studies published have been replicated thusly. It is also noted that original studies that identified significant associations were usually performed in small samples and thus lack sufficient representation of the study populations, leading to false-positive associations. Another problem caused by small samples is lack of sufficient power to detect true associations in replication studies, leading to false-negative associations. In this regard, associations obtained from the studies using large samples are more reliable. 

Data from the HapMap project have shown obviously different allele frequencies of many genes between populations. It is therefore crucial to avoid population stratification in genetic association studies. One of the strengths of the present study is the well-matched cases and control subjects in ethnicity, and thus the observed association in our study is unlikely to be affected by population stratification. Another major strength of our study is the larger sample size in both the disease and control groups that gave our study sufficient power to detect a true association. In addition, we classified patients with POAG into three subgroups and analyzed association for HTG, NTG, and JOAG separately, which also improved the study's power by reducing the phenotype heterogeneity. 

One limitation of our study is that we did not screen the whole gene region but investigated significant variants in each gene as reported by other groups. As a result, we could not rule out the possibility of the existence of other variants in these genes that might be associated with POAG in our samples. Further comprehensive study of whole genes would be helpful to examine this issue. In addition, the associations we observed in the present study were nominally significant after the gene-wide Bonferroni correction. In genome-wide association studies, up to 500-K SNPs are tested for association and thus the Bonferroni-corrected significance level is usually set as 1 × 10⁻⁷ (0.05/500 K). However, in the present study, our null hypothesis was that the allele frequencies of all variants tested in a specific gene were not significantly different between cases and control subjects, the gene-wide Bonferroni-corrected significance level was thus set as a function of the number of variants tested in a specific gene. In addition, one critical assumption of the Bonferroni correction is that all tests performed are independent. However, this is not the case in many association studies in which markers are not independent because of linkage disequilibrium. A Bonferroni correction is thus considered to be overly conservative, especially for replication studies that have tended to obtain less significant association than the initial studies. ⁷⁸  

In summary, our findings suggest that variants in TNF and TP53 are significant risk factors for POAG. However, variants in other studied genes are not major risk factors for the development of this complex disorder, at least in the Chinese population. 

Footnotes

 Supported in part by a block grant from the University Grants Committee, Hong Kong, and by General Research Fund Grant 2140597 from the Research Grant Council, Hong Kong.

Footnotes

 Disclosure: B.J. Fan, None; K. Liu, None; D.Y. Wang, None; C.C.Y. Tham, None; P.O.S. Tam, None; D.S.C. Lam, None; C.P. Pang, None