Methylenetetrahydrofolate reductase (MTHFR) C677T gene polymorphism and diabetic nephropathy susceptibility in patients with type 2 diabetes mellitus

Abstract

Methylenetetrahydrofolate reductase (MTHFR) is a crucial enzyme that regulates nucleotide synthesis and DNA methylation. The MTHFR C677T gene polymorphism (rs1801133), a C → T transition at nucleotide 677 in exon 4, is a common gene variant of MTHFR and has been implicated in diabetic nephropathy, albeit with inconsistent results. Here, we performed a meta-analysis to assess the common effect size of this polymorphism on DN susceptibility. Case–control studies on the association of the MTHFR C677T gene polymorphism with DN risk were retrieved from databases up to August 1, 2013, and eligible studies were recruited into the meta-analysis and further analyzed. Of 132 studies, 33 were identified as suitable for this analysis. The results showed that T allele and TT genotype were distinctly associated with DN susceptibility in the overall population and Asians, and might be a risk factor in Caucasians and Africans (T allele: Overall population: p < 0.00001, Asians: p = 0.0002, Caucasians: p = 0.02, Africans: p < 0.00001; TT genotype: Overall population: p < 0.00001, Asians: p = 0.0003, Caucasians: p = 0.008, Africans: p = 0.0003). Furthermore, the analysis suggested that the CC genotype might play a protective role against DN onset in patients with type 2 diabetes for the overall population, Asians, Caucasian and Africans. However, due to the limited sample size in the African population, these results should be interpreted with care. In conclusion, the MTHFR C677T T allele or TT genotype might be a significant genetic molecular marker to determine the risk of DN in patients with type 2 diabetes and help to develop suitable disease prevention and management strategies.

• C677T
• diabetic nephropathy
• gene polymorphism
• methylenetetrahydrofolate reductase
• MTHFR
• meta-analysis

Introduction

Diabetes is a metabolic disorder caused by loss of the insulin-producing beta cells of the Langerhans islets in the pancreas (type 1) or insulin resistance (type 2) and is characterized by hyperglycemia, which produces the classical symptoms of polyuria, polydipsia and polyphagia. Furthermore, diabetes negatively affects the hemodynamics in the macro and microvasculature of various organs, and the damage inflicted to the vasculature through various degenerate mechanisms results in a significantly increased risk of developing cardiovascular disease and ischemic stroke,1 diabetic retinopathy,2 diabetic neuropathy3 and diabetic nephropathy (DN)4–6.

DN is one of the most common microvascular complications of diabetes and the primary cause of end-stage renal disease and involves nearly 20–40% of diabetic patients.4 DN affects numerous kidney cell types, including tubular and vascular epithelia, glomerular podocytes, interstitial fibroblasts, mesangial and endothelial cells,4^,5 and is associated with extensive cellular dysfunction, hypoxia, oxidative stress and extracellular matrix accumulation.6–8 Although both type 1 and type 2 diabetes are able to induce DN, the current DN epidemic has been shown to be predominantly associated with type 2 diabetes.9 However, a fundamental understanding of the interplay between degenerate biological mechanisms in humans and exogenous influences that produce the clinical and pathologic alterations observed in DN remains incomplete. Next to environmental and life-style stressors, the susceptibility to DN has an intrinsic genetic basis, which is demonstrated by familial aggregation and ethnic-specific prevalence rates, and appears to be the most prominent in African-Americans and Pima Indians.10 Recent investigations have implicated a myriad of gene polymorphisms as key players in the modulation of the pathogenesis of DN and have been distinctly associated with DN risk.11–15 However, emerging evidence also suggests that additional genetic variants and derangements,10 as well as alterations in epigenetic mechanisms, including DNA methylation variations, histone modifications and micro-RNA (miRNA) changes, are involved in diabetes and DN susceptibility.16

A candidate gene implicated to play a major role in DN susceptibility is the methylenetetrahydrofolate reductase (MTHFR) gene, located at the end of the short arm of chromosome 1 (1p36.3).17 MTHFR codes for a crucial enzyme (flavoprotein) in folate and homocysteine metabolism (Figure 1A) that catalyzes the flavin adenine dinucleotide (FAD)-dependent reduction of CH₂-H₄folate by NAD(p)H to CH₃-H₄folate and is involved in nucleotide synthesis, DNA repair and methylation.18 Folate and vitamin B12 are required in the synthesis of methionine via remethylation of homocysteine, which results in the formation of S-adenosylmethionine; a methyl donor for the maintenance of methylation patterns, which determine gene expression and DNA conformation. Because of MTHFR’s pivotal role in many aspects of normal gene function (Figure 1A), including homocysteine levels (an established risk factor for vascular disease19), it is not surprising that MTHFR gene polymorphisms may have dramatic effects on normal cellular function and that increasingly these are implicated in various diseases, including major depressive disorder, congenital heart diseases, breast cancer, leukemia and others.20–22

Figure 1. (A) Schematic overview of the involvement of 5,10-methylenetetrahydrofolate reductase (MTHFR) in folate and homocystein metabolism, nucleotide synthesis and (re)methylation events. (B) Flow chart for study recruitment into the meta-analysis of the MTHFR C677T gene polymorphism and diabetic nephropathy relationship.

The MTHFR C677T single nucleotide polymorphism (rs1801133) involves a C → T transition at nucleotide 677 in exon 4 and is a common gene variant of MTHFR that includes CC, CT and TT genotypes and C and T alleles. MTHFR C → T transition induces enzyme thermolability and is distinctly associated with diminished enzyme activity.23^,24 Even though the present epidemiologic evidence implicates the MTHFR C677T gene polymorphism in the etiology of DN in patients with type 2 diabetes mellitus, the available proof reported to date is inconclusive, predominantly due to sparseness of data and disagreements among studies. We performed this meta-analysis to investigate the relationship between the MTHFR C677T gene polymorphism and DN susceptibility in patients with type 2 diabetes mellitus, with the intention to provide a more reliable analysis of the significance of the purported association.

Materials and methods

Search strategy

Relevant studies were extracted from PubMed, Embase, Cochrane Library and CBM-disc (China Biological Medicine Database) on August 1, 2013, and association reports were identified. The retrieval strings entered into these databases were: “(methylenetetrahydrofolate reductase OR MTHFR) AND (DN OR DN OR diabetes mellitus nephropathy) AND (gene OR polymorphism OR allele OR genotype OR variant OR variation OR mutation)”. Additional reports were identified by scrutinizing the references cited in the recruited articles. If multiple publications for the same data from the same study group occurred, we only recruited the later dated paper into our final analysis.

Inclusion and exclusion criteria

Inclusion criteria

(1) A case–control study, (2) the outcome had to be DN; (3) there had to be at least two comparison groups (DN group vs. control group).

Exclusion criteria

(1) Review articles, editorials and case reports, (2) articles that did not provide detailed genotype data, (3) investigations on the association of other genes with DN, (4) investigations of the role of MTHFR related to diseases and (5) multiple publications of the same data from the same study group.

Data extraction and synthesis

The following information was extracted from each study independently by at least two investigators: first author’s surname, year of publication, ethnicity of study population, and the number of cases and controls for MTHFR genotype. Frequencies of T allele were calculated for case and control groups from the corresponding genotype distribution. The results were compared, and disagreements were resolved by discussion.

Statistical analysis

Available data were entered into Cochrane Review Manager Version 5 (Cochrane Library, UK) and analyzed. The pooled statistics were counted using the fixed effects model, but a random effects model was applied when the p value of the heterogeneity test was less than 0.1.25^,26 Results were expressed with odds ratios (OR) for dichotomous data, and additionally 95% confidence intervals (CI) were calculated. p < 0.05 was required for the overall OR to be deemed statistically significant.27–29 I² was used to test the heterogeneity among the included studies. We classified the investigations into studies for Caucasians, Asians and African populations, because the genotype frequencies and prevalence of DN were quite different between the various ethnic groups. In order to avoid excessive comparisons, the OR was calculated using three methods,30^,31 that is, method 1: allele comparison (T allele vs. C allele); method 2: comparison of TT homozygotes with the other two combinations (TT vs. TC + CC); and method 3: comparison of CC genotype with the other two combinations (CC vs. TC + TT). A chi-square test was used to determine if the reported genotype distribution of the control populations conformed to Hardy–Weinberg equilibrium (HWE; p < 0.05 was considered significant). Sensitivity analysis was performed when studies with controls not in HWE. Sensitivity analysis was also performed according to the source of the controls (healthy vs. hospital), genotyping methods (PCR-RFLP, etc.), by sequential omission of individual studies, and sample size of case or control (<100 vs. ≥100). Stata 11.0 (StataCorp LP, College Station, TX) was used to test for publication bias. The Begg adjusted rank correlation32 and the Egger regression asymmetry tests33 were used for exploring publication bias (p < 0.1 was considered significant) when the number of the included studies exceeded six.

Results

Study characteristics

The search yielded 132 studies, of which, after carefully applying the inclusion and exclusion criteria, 33 studies were identified as suitable for inclusion in the meta-analysis on the association between the MTHFR C677T gene polymorphism and DN susceptibility (Figure 1B). Eighteen reports were published in English34–51 and 15 reports were published in Chinese.52–66 Twenty-two reports were conducted in Asian subjects,42–44^,46–48^,50^,52–56 10 reports in Caucasians34–41^,45^,51 and 1 report in Africans.49 The data of interest were extracted, as summarized in Table 1. In total, the 33 investigations contained 3959 DN patients and 3915 controls.

Table 1. General characteristics of the included studies in this meta-analysis.

					Case				Control
Author, Year	Ethnicity	Country	Genotyping methods	Source of control	TT	TC	CC	Total	TT	TC	CC	Total	HWE
Neugebauer, 1998	Asian	Japan	PCR-RFLP	Healthy	12	31	24	67	17	43	86	146	0.004
Odawara, 1999	Asian	Japan	PCR-RFLP	Hospital	26	65	52	143	25	68	38	131	0.579
Fujita, 1999	Asian	Japan	PCR-RFLP	Hospital	17	57	31	105	9	39	20	68	0.142
Blüthner, 1999	Caucasian	Germany, Poland	PCR-RFLP	Healthy	23	50	74	147	15	68	67	150	0.709
Shpichinetsky, 2000	Caucasian	Israel	PCR-RFLP	Hospital	10	22	23	55	6	16	21	43	0.317
Yang, 2001	Asian	China	PCR-RFLP	Healthy	23	27	17	67	8	28	26	62	0.914
Sun, 2001	Asian	China	PCR-RFLP	Healthy	8	20	11	39	31	16	10	57	0.008
Wang, 2001	Asian	China	PCR-RFLP	Healthy	28	34	20	82	12	36	37	85	0.503
Guo, 2002	Asian	China	PCR-RFLP	Healthy	22	19	12	53	5	11	12	28	0.392
Moczulski, 2003	Caucasian	Poland	PCR-RFLP	Hospital			45	121			75	159
Xu, 2003	Asian	China	PCR-RFLP	Healthy	21	33	15	69	7	25	20	52	0.853
Mao, 2004	Asian	China	PCR-RFLP	Healthy	6	20	14	40	3	18	26	47	0.961
Chen, 2004	Asian	China	PCR-RFLP	Healthy	14	21	6	41	5	9	21	35	0.038
Sun, 2004	Asian	China	PCR-RFLP	Healthy	26	53	45	124	22	34	74	130	0
Ksiazek, 2004	Caucasian	Poland	PCR-RFLP	Healthy	29	65	77	171	16	83	71	170	0.237
Yoshioka, 2004	Asian	Japan	PCR-RFLP	Hospital	6	13	21	40	29	107	71	207	0.261
Cao, 2005	Asian	China	PCR-RFLP	Hospital	14	20	6	40	26	18	3	47	0.961
Yue, 2006	Asian	China	PCR-RFLP	Healthy	34	55	23	112	2	11	17	30	0.903
Mtiraoui, 2007	Caucasian	Tunisia	PCR-RFLP	Healthy	62	135	163	360	36	94	270	400	0
Böger, 2007	Caucasian	Germany	PCR-RFLP	Hospital	32	219	188	439	15	69	64	148	0.566
Eroglu, 2007	Caucasian	Turkey	PCR-RFLP	Healthy	1	20	26	47	7	58	63	128	0.172
Maeda, 2008	Asian	Japan	PCR-RFLP	Hospital	5	6	14	25	10	51	26	87	0.046
Wen, 2008	Asian	China	PCR-RFLP	Healthy	23	50	22	95	5	25	27	57	0.816
Lin, 2009	Asian	China	PCR-RFLP	Hospital	47	36	56	139	24	22	93	139	0
Ukinc, 2009	Caucasian	Turkey	PCR-RFLP	Hospital	0	16	6	22	0	8	22	30	0.399
Nemr, 2010	Caucasian	Lebanon, Bahrain	PCR-RFLP	Hospital	78	162	236	476	41	186	410	637	0.002
Qian, 2010	Asian	China	PCR-RFLP	Hospital	22	35	15	72	22	63	42	127	0.845
Rahimi, 2010	Caucasian	Iran	PCR-RFLP	Hospital	17	63	60	140	2	26	45	73	0.438
Movva, 2011	Asian	India	PCR-RFLP	Healthy	0	30	106	136	0	9	91	100	0.637
Dai, 2012	Asian	China	PCR-RFLP	Healthy	12	26	22	60	2	27	31	60	0.177
El-Baz, 2012	African	Egypt	PCR-RFLP	Hospital	24	46	32	102	3	19	78	100	0.189
Sun, 2013a	Asian	China	PCR-RFLP	Healthy	30	123	75	228	6	42	30	78	0.094
Sun, 2013b	Asian	China	PCR-RFLP	Healthy	14	53	35	102	8	53	43	104	0.128

Notes: PCR–RFLP: polymerase chain reaction-restriction fragment length polymorphism; HWE: Hardy–Weinberg equilibrium.

The average distribution frequency of the MTHFR C677T T allele in the DN group was 42.14 and 38.41% in the control group. Overall, the results demonstrate that the average distribution frequency of the MTHFR C677T T allele in case group was markedly higher than frequency in the control group (case/control = 1.10). In the sub-group analysis, the average distribution frequency of the MTHFR C677T T allele was 45.46% in Asian patients with DN (41.09% in controls) and 33.65% in Caucasian patients with DN (31.65% in controls). However, in Africans, a significantly lower frequency was observed in controls, that is, 12.50%, with an average distribution frequency in cases that was in the same order of magnitude as the other ethnic groups, that is, 46.08%.

Overall, these results translate into case to control ratios for the average distribution frequency of the MTHFR C677T T allele that were larger than 1 for the overall population (case/control = 1.10), Asians (case/control = 1.11), Caucasians (case/control = 1.07) and Africans (case/control = 3.69). However, the results in Africans showed a large discrepancy in the control group, which resulted in a significantly higher ratio compared with other ethnic groups and might be a consequence of the limited number of studies and low-sampling frequency. Therefore, the results in the African group should be interpreted with care. Consequently, the MTHFR C677T T allele might only be considered a risk factor for DN disease in the overall population, Asians and Caucasians.

Association of MTHFR C677T gene polymorphism with DN susceptibility

The results from this meta-analysis showed clear trends with respect to the association between MTHFR C677T genotypes and the risk of developing DN. As summarized in Table 2, the MTHFR C677T T allele (Figure 2) or TT genotype (Figure 3) was distinctly associated with DN susceptibility in the overall population (T allele: OR = 1.57, 95% CI 1.31–1.88, p < 0.00001; TT genotype: OR = 1.84, 95% CI 1.43–2.37, p < 0.00001). Conversely, evaluation of the MTHFR C677T CC genotype in the overall population (Figure 4) suggested a protective role for the MTHFR C677T CC genotype and a reduced risk of developing DN (OR = 0.60, 95% CI 0.48–0.74, p < 0.00001).

Figure 2. Association between T allele and diabetic nephropathy risk in the overall population.

Figure 3. Association of TT genotype with diabetic nephropathy risk in the overall population.

Figure 4. Association of CC genotype with diabetic nephropathy risk in the overall population.

Table 2. Meta-analysis of the association of MTHFR C677T gene polymorphism with diabetic nephropathy risk.

Genetic contrasts	Group and subgroups	Studies number	Q test p value	Model selected	OR (95% CI)	p
T versus C	Overall	32	<0.00001	Random	1.57 (1.31, 1.88)	<0.00001
	Asian	22	<0.00001	Random	1.55 (1.23, 1.95)	0.0002
	Caucasian	9	<0.00001	Random	1.41 (1.07, 1.86)	0.02
	African	1	–	Fixed	5.98 (3.62, 9.88)	<0.00001
TT versus CT + CC	Overall	32	<0.00001	Random	1.84 (1.43, 2.37)	<0.00001
	Asian	22	<0.0001	Random	1.80 (1.31, 2.47)	0.0003
	Caucasian	9	0.01	Random	1.73 (1.15, 2.60)	0.008
	African	1	–	Fixed	9.95 (2.89, 34.27)	0.0003
CC versus CT + TT	Overall	33	<0.00001	Random	0.60 (0.48, 0.74)	<0.00001
	Asian	22	<0.00001	Random	0.60 (0.44, 0.80)	0.0006
	Caucasian	10	<0.00001	Random	0.70 (0.04, 0.96)	0.03
	African	1	–	Fixed	0.13 (0.07, 0.24)	<0.00001
Sensitivity analysis according to HWE
T versus C	Overall	23	<0.00001	Random	1.57 (1.25, 1.97)	<0.0001
	Asian	15	<0.00001	Random	3.65 (1.23, 2.07)	0.0005
	Caucasian	7	0.01	Random	1.20 (0.92, 1.57)	0.18
	African	1	–	Fixed	5.98 (3.62, 9.88)	<0.00001
TT versus CT + CC	Overall	23	0.0002	Random	1.95 (1.41, 2.70)	<0.0001
	Asian	15	0.005	Random	2.03 (1.38, 2.98)	0.0003
	Caucasian	7	0.08	Random	1.40 (0.82, 2.39)	0.22
	African	1	–	Fixed	9.95 (2.89, 34.27)	0.0003
CC versus CT + TT	Overall	23	<0.00001	Random	0.61 (0.45, 0.82)	0.001
	Asian	15	<0.0001	Random	0.60 (0.43, 0.84)	0.003
	Caucasian	7	0.005	Random	0.83 (0.57, 1.21)	0.34
	African	1	–	Fixed	0.13 (0.07, 0.24)	<0.00001
Sensitivity analysis according the control source from healthy
T versus C	Overall	20	<0.00001	Random	1.81 (1.43, 2.28)	<0.00001
	Asian	15	<0.00001	Random	1.88 (1.48, 2.38)	<0.00001
	Caucasian	4	<0.0001	Random	1.20 (0.75, 1.92)	0.45
TT versus CT + CC	Overall	20	0.0007	Random	2.18 (1.57, 3.02)	<0.00001
	Asian	15	0.001	Random	2.20 (1.44, 3.38)	0.0003
	Caucasian	4	0.46	Fixed	1.87 (1.36, 2.56)	<0.0001
CC versus CT + TT	Overall	20	<0.00001	Random	0.49 (0.37, 0.66)	<0.00001
	Asian	15	0.02	Random	0.45 (0.35, 0.59)	<0.00001
	Caucasian	4	<0.00001	Fixed	0.90 (0.46, 1.77)	0.76
Sensitivity analysis according the control source from hospital
T versus C	Overall	13	<0.00001	Random	1.41 (1.02, 1.96)	0.04
	Asian	7	<0.00001	Random	1.02 (0.66, 1.58)	0.91
	Caucasian	5	0.0003	Random	1.62 (1.09, 2.43)	0.02
	African	1	–	Fixed	5.98 (3.62, 9.88)	<0.00001
TT versus CT + CC	Overall	13	<0.00001	Random	1.62 (1.05, 2.49)	0.03
	Asian	7	0.03	Random	1.31 (0.83, 2.06)	0.25
	Caucasian	5	0.002	Random	1.77 (0.73, 4.28)	0.20
	African	1	–	Fixed	9.95 (2.89, 34.27)	0.0003
CC versus CT + TT	Overall	14	<0.00001	Random	0.70 (0.47, 1.02)	0.06
	Asian	7	<0.00001	Random	1.13 (0.60, 2.14)	0.70
	Caucasian	6	0.02	Random	0.60 (0.43, 0.84)	0.003
	African	1	–	Fixed	0.13 (0.07, 0.24)	<0.00001
Sensitivity analysis according to sample size of case (≥100)
T versus C	Overall	11	<0.00001	Random	1.64 (1.23, 2.19)	0.0007
	Asian	5	0.0001	Random	1.62 (1.04, 2.52)	0.03
	Caucasian	5	<0.00001	Random	1.35 (0.96, 1.90)	0.08
	African	1	–	Fixed	5.98 (3.62, 9.88)	<0.00001
TT versus CT + CC	Overall	11	0.001	Random	1.82 (1.29, 2.57)	0.0007
	Asian	5	0.13	Fixed	1.54 (1.12, 2.12)	0.008
	Caucasian	5	0.009	Random	1.75 (1.12, 2.75)	0.01
	African	1	–	Fixed	9.95 (2.89, 34.27)	0.0003
CC versus CT + TT	Overall	12	<0.00001	Random	0.59 (0.42, 0.82)	0.002
	Asian	5	0.0005	Random	0.56 (0.32, 0.98)	0.04
	Caucasian	5	<0.00001	Random	0.75 (0.52, 1.09)	0.13
	African	1	–	Fixed	0.13 (0.07, 0.24)	<0.00001
Sensitivity analysis according to sample size of case (<100)
T versus C	Overall	21	<0.00001	Random	1.53 (1.20, 1.96)	0.0007
	Asian	17	<0.00001	Random	1.53 (1.15, 2.01)	0.003
	Caucasian	4	0.01	Random	1.55 (0.87, 2.77)	0.14
TT versus CT + CC	Overall	21	0.0001	Random	1.87 (1.28, 2.73)	0.001
	Asian	17	<0.00001	Random	1.91 (1.27, 2.87)	0.002
	Caucasian	14	0.13	Fixed	1.76 (0.85, 3.64)	0.13
CC versus CT + TT	Overall	21	<0.00001	Random	0.60 (0.43, 0.82)	0.002
	Asian	17	<0.00001	Random	0.61 (0.42, 0.88)	0.008
	Caucasian	4	0.01	Random	0.56 (0.26, 1.18)	0.13

Note: HWE: Hardy–Weinberg equilibrium.

In the subgroup analysis, the MTHFR C677T T allele (Figure 5), TT (Figure 6) and CC (Figure 7) genotypes were associated with DN susceptibility in the overall population and Asians, with similar results (T allele: OR = 1.55, 95% CI 1.23–1.95, p = 0.0002; TT genotype: OR = 1.80, 95% CI 1.31–2.47, p = 0.0003; CC genotype: OR = 0.60, 95% CI 0.44–0.80, p = 0.0006). In Caucasians, the MTHFR C677T T allele or TT genotype was associated with DN susceptibility (T allele: OR = 1.41, 95% CI 1.07–1.86, p = 0.02; TT genotype: OR = 1.73, 95% CI 1.15–2.60, p = 0.008; Table 2), but not the MTHFR C677T CC genotype (OR = 0.70, 95% CI 0.04–0.96, p = 0.03; Table 2). Therefore, this genotype might be a factor that reduces the risk of DN disease in Caucasians. Although the African sample size was relatively low, analysis of the available data showed that, equal to Caucasians, the MTHFR C677T T allele or TT genotype was associated with DN susceptibility (T allele: OR = 5.98, 95% CI 3.62–9.88, p < 0.00001; TT genotype: OR = 9.95, 95% CI 2.89–34.27, p = 0.0003; Table 2), but the CC genotype was not (OR = 0.13, 95% CI: 0.07–0.24, p < 0.00001; Table 2). Further studies are required to further clarify this trend and observations.

Figure 5. Association between T allele and diabetic nephropathy risk in Asians.

Figure 6. Association between TT genotype and diabetic nephropathy risk in Asians.

Figure 7. Association between CC genotype and diabetic nephropathy risk in Asians.

Sensitivity analysis

Sensitivity analysis examines data consistency, and if results are sensitive to restrictions on the included data to provide stronger evidence of an effect and of generalizability of the results. Therefore, the data included in this meta-analysis were subjected to further scrutiny via sensitivity analysis. Sensitivity analysis according to the genotype distribution of the control population reported for the MTHFR C677T gene polymorphism conformed to HWE, and the results were shown to be similar for the overall population, Asians and Africans compared with non-sensitivity analysis, but not for Caucasians (Table 2).

When focusing on the source of the controls (healthy vs. hospital), sensitivity analysis according to the source of the controls from healthy showed that in the overall population (T allele: OR = 1.81, 95% CI 1.43–2.28, p < 0.00001; TT genotype: OR = 2.18, 95% CI 1.57–3.02, p < 0.00001; CC genotype: OR = 0.49, 95% CI 0.37–0.66, p < 0.00001) and Asians (Asians: T allele: OR = 1.88, 95% CI 1.48–2.38, p < 0.00001; TT genotype: OR = 2.20, 95% CI 1.44–3.38, p = 0.0003; CC genotype: OR = 0.45, 95% CI 0.35–0.59, p < 0.00001), both sensitivity and non-sensitivity analyses produced similar results. This was not the case for Caucasians, because the sensitivity analysis according to the source of the controls from healthy showed that the TT genotype was distinctly associated with DN risk, but T allele and CC genotype were not (Table 2). Furthermore, in the sensitivity analysis, according to the source of the controls from hospital, we found that the results in the overall population and in Africans were consistent with the non-sensitivity analysis, whereas this was not the case for the Asian population (Table 2). In this ethnic group, only the T allele and CC genotype were associated with DN risk (Table 2). Examination of the genotype determination method resulted in similar results for both sensitivity and non-sensitivity analyses.

Sensitivity analysis for the relationship between the MTHFR C677T gene polymorphism and DN risk was also performed according to sample size of case (<100 vs. ≥100). In the meta-analysis of sample size of case ≥100, we found that the results for the overall population, Asians and Africans were similar to those from non-sensitivity analysis, but in Caucasians, the results from T allele and CC genotype were inconsistent with the non-sensitivity analysis (Table 2). Similarly, when evaluating the sample size of case <100, the results of sensitivity and non-sensitivity analyses were consistent, and a clear association of the MTHFR C677T gene polymorphism with DN risk was established in the overall population and Asians, but inconsistent results were obtained for Caucasians (Table 2). Sequential omission of individual studies showed no significant deviation between both analyses (data not shown).

Evaluation of publication bias

Publication bias was tested (Figure 8), and no publication bias was established for the MTHFR C677T gene polymorphism in the overall population (Begg p = 0.412; Egger p = 0.728), Asians (Begg p = 0.735; Egger p = 0.909) and Caucasians (Begg p = 1.000; Egger p = 0.645).

Figure 8. Begg’s funnel plots with pseudo 95% confidence limits. Evaluation of publication bias for the association of MTHFR C677T gene polymorphism with diabetic nephropathy risk in the overall population (A), Asians (B) and Caucasians (C).

Discussion

Diabetes has been associated with renal disease in a large number of patients. However, diabetes-associated renal disease is neither simply determined by the duration nor severity of diabetes and, therefore, not an inevitable complication of diabetes. Consequently, other factors, such as environmental and life-style stressors and genetic factors, must play key roles in the pathogenesis of DN. Increasing scientific evidence suggests that genotype seems to be one of the major determinants of both the incidence and severity of DN. The probability of developing DN is significantly increased in both type 1 and type 2 diabetes patients with first degree relatives suffering from DN (familial aggregation), as clearly demonstrated in Pima Indian families with successive generations of diabetes sufferers.67 Numerous studies have aimed to determine the genetic components that predispose patients with type 2 diabetes mellitus to developing DN and renal failure. However, very few genetic variants have conclusively been studied in relation to disease and population dynamics.

Alterations in the MTHFR gene have been implicated in the etiology of DN, especially because this gene code for an enzyme is intrinsically involved in nucleotide synthesis, DNA repair and DNA methylation and plays a pivotal role in folate and homocysteine metabolism (Figure 1A). Recently, Li et al.68 reported that total plasma homocysteine concentration was independently associated with the occurrence of DN in Chinese type 2 diabetes patients at an early stage of the disease. Elevated homocysteine levels, predominantly caused by gene polymorphisms in key enzymes involved in homocysteine metabolism, are distinctly considered to be a risk factor for renal disease, including DN.44^,47^,69 The C → T transition in the MTHFR C677T gene polymorphism results in an alanine to valine substitution, which causes thermolability23^,24 and negatively affects enzyme activity and homocysteine levels. MTHFR TT homozygotes and CT heterozygotes have been shown to retain 30 and 65% enzyme activity compared with the wild-type CC genotype,70 whereas T allele homozygosity equally results in reduced enzyme activity, but only mild to moderately increased homocysteine levels.71 Overall, the MTHFR C677T gene polymorphism might represent an important risk factor for DN in diabetes.

Thirty-three suitable studies were recruited into our meta-analysis. Evaluation of the MTHFR C677T gene polymorphism in the overall population showed that the average distribution frequency of the MTHFR C677T T allele was 1.1-fold larger in the case group compared with the control group. Both T allele and TT genotype were associated with DN risk, whereas the analysis showed that CC genotype may be considered a protective factor against DN. Sensitivity analysis according to HWE, the source of the controls, genotyping methods (PCR-RFLP), by sequential omission of individual studies and sample size of case or control (<100 vs. ≥100) were similar to those in the non-sensitivity analysis for the overall population. No significant differences were observed between the sensitivity and non-sensitivity for the T allele and TT genotype according to the source of the controls from hospital. Furthermore, the pooled OR for the CC genotype from the sensitivity analysis according to the source of the controls from hospital was favorable with respect to the control group; although the difference was not statistically significant. Nonetheless, our results demonstrate a significant level of robustness in the analysis outcome and indicate that in the overall population, the T allele and TT genotype may be considered risk factors for DN.

Because geographic and ethnic difference might be important factors that affect the susceptibility to DN, we further evaluated the data according to ethnicity. The analysis showed that the average frequency of the T allele in controls was 41.09% for Asians and 33.65% for Caucasians and therefore a factor ∼1.22 higher in Asians. Conversely, the T allele frequency in Africans was only 12.50% for the control group (∼3.3- and ∼2.7-fold lower than in Asians or Caucasians). This low frequency might not necessarily reflect a lower occurrence of the polymorphism in this particular ethnic group, but might be a result of the small sample size. Furthermore, the data collected originated from patients geographically restricted to northern Africa49 and by no means reflected an ethnic cross-section of the whole continent. Nonetheless, a new principle finding can equally not be ruled out, but will require further prospective studies in Africans with various ethnic and geographic backgrounds.

Subgroup analysis showed that the average frequency of the T allele in the Asian case group was 1.11-fold higher compared with the control group and, therefore, the MTHFR C677T gene polymorphism might be a risk factor for DN in Asians. No publication bias was observed for the studies involving Asians, which largely validates the obtained results for the Asian population. In Caucasians, the average frequency of the T allele in the case group was 1.07-fold higher compared with the control group, which suggested and increased susceptibility to DN in patients with type 2 diabetes. In the sensitivity analysis according to the control source from healthy and the sensitivity analysis according to sample size of case (≥100), the TT genotype was associated with DN risk, and in the sensitivity analysis according the control source from hospital, the T allele was associated with DN risk. Sensitivity and non-sensitivity analyses were consistent, but when focusing on other factors, no such consistency could be established. Although we included 10 studies for the analysis in Caucasians and no publication bias was observed, the meta-analysis results showed that it was difficult to draw robust conclusions for this ethnic group. Consequently, more studies in Caucasians with significantly larger sample sizes are required to conclusively determine the exact MTHFR variants that represent a risk factor for DN in Caucasians. In Africans, the average frequency of T allele in the case group was 3.69-fold higher than the control group. Despite investing considerable efforts in searching for suitable published studies that focused on Africans, only one study fulfilled our selection criteria and was recruited into this meta-analysis, with the consequence that the sample size was low. As a result, our estimates of the role that the MTHFR C677T gene polymorphism plays in DN susceptibility in Africans are still likely to represent an overestimation, despite equal results from both sensitivity analysis and non-sensitivity analysis. Overall, information on the prevalence of nephropathy in African diabetic populations is scarce. With obesity on the incline (an established risk factor for diabetes), particularly in urbanized areas, diabetes and DN are likely to become an increasing health burden on the African population. Further investigations on the genetic predisposition of Africans are, therefore, urgently needed. For the overall population, sensitivity analysis according to HWE, the source of the controls (healthy), genotyping methods (PCR-RFLP), by sequential omission of individual studies and sample size of case or control (<100 vs. ≥100) were similar to those in non-sensitivity analysis. However, the results for T allele, TT genotype and CC genotype from sensitivity and non-sensitivity analyses according to the source of the controls from hospital were inconsistent. Interestingly, the patients from all the control groups from hospital were type 2 diabetes patients not suffering from DN. Further evaluation showed that no significant differences in genotype distribution between DN patients with type 2 diabetes and non-DN patients with type 2 diabetes could be established.

In recent years, several meta-analyses explored a possible association between the MTHFR C677T gene polymorphism DN susceptibility. Zintzaras et al.72 included 11 studies involving subjects with DN and reported that the T allele was associated with DN risk in patients with type 2 diabetes. An evaluation by Niu et al.73 included 22 studies and showed that the MTHFR TT genotype might confer a moderately augmented risk for DN. Chang et al.74 recruited 13 studies focusing on the Chinese population in their meta-analysis and reported that the MTHFR C677T polymorphism might influence DN risk in this ethnic group. Yang et al.75 performed a meta-analysis included nine reports in Caucasian individuals with type 2 diabetes, and this meta-analysis supported that there was an association between MTHFR C677T polymorphism and DN risk. Overall, these studies reported heterogonous results. Our meta-analysis (33 studies) showed that the MTHFR T allele or TT genotype was a distinct risk factor for the DN susceptibility in patient with type 2 diabetes in the overall population and Asians, and that the CC genotype might play a protective role against DN susceptibility, confirming the results by Zintzaras72, Chang74 and Yang75 and their co-workers. In Caucasians and Africans, there are some indications that an equal association exists. However, the results should be regarded cautiously, because of possible heterogeneity in the enrolled cases, limited statistical power, variable study designs and different interventions. Although a clear genetic component is observable in DN, it is likely that multiple genetic variants, which might isolated only have nominal effects, in conjunction with exogenous triggers, determine if diabetes patients ultimately will develop DN. Nonetheless, many questions regarding the role of MTHFR alleles in various diseases still remain. Alternate and more precise strategies would be required to obtain more accurate data. First, precise estimates of MTHFR genotype frequencies from large and representative samples among well-defined populations are required. Second, both relative and absolute risks as well as attributable fractions need to be determined for each disease and genotype of interest. Third, additional genes and alleles should be identified. Fourth, since the fundamental metabolic roles of folate and homocysteine suggest that genetic variation in the regulatory genes may also play a role in the etiology of DN, more outcomes should be evaluated. Fifth, the interplay between various genetic variants per se (gene-gene) and multiple genetic variants and exogenous factors (gene-environment) should be studied in more detail, in particular gene–nutrient interactions. Finally, in addition to study design considerations, special attention should be given to sample size requirements in order to maximize statistical significance. Such strategies would allow significantly more robust meta-analyses and minimize some of the limitations associated with meta-analyses. Alternatively, an integrated approach that incorporates datasets from multiple sources, including genetics and epigenetics, transcriptomic, proteomics, metabolomics and the effect of exogenous and life-style stressors might shed more light on which genetic elements contribute to this serious complication of diabetes. Prospective multilayer studies that address the interplay between various factors will enable researchers to identify promising candidate genes and pathways, develop better risk profiles, and allow a more targeted approach to both the management of diabetes and DN, and the development of novel therapeutic interventions. Various initiatives to collate data on renal disease from multiple sources of evidence have already been established, including the Kidney and Urine Pathway Knowledge Base,76 the Human Kidney and Urine Proteome,77 Nephromine (renal gene expression profiles),78 and EuReGene (gene expression in mouse kidney).79

Declaration of interest

This study was supported by the sub-item of 985 Project Foundation of Sun Yat-Sen (The Hundred Talents Program Foundation; No. 88000-3311300) and was supported by the Guangzhou Medical Key Subject Construction Project (2013--2015). The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.