Molecular Vision 2026; 32:308-323 <http://www.molvis.org/molvis/v32/308>
Received 09 September 2025 | Accepted 22 July 2026 | Published 31 July 2026

Association and interaction of glucose transporter 3 and monocarboxylate transporter 2 gene polymorphisms in Japanese patients with normal-tension glaucoma

Makoto Nakamura,1 Akira Meguro,2 Eiichi Nomura,2 Chiaki Yamaguchi,3 Yu Yokoyama,3 Satoru Tsuda,3 Kota Sato,3 Kazuki Hashimoto,3 Masataka Sato,3 Akira Narita,4 Mika Sakurai-Yageta,4 Sotaro Mori,1 Mari Sakamoto,1 Yuko Yamada-Nakanishi,1 Toru Nakazawa,3 Nobuhisa Mizuki2

1Department of Surgery Division of Ophthalmology, Kobe University Graduate School of Medicine; 2Department of Ophthalmology, Yokohama City University; 3Department of Ophthalmology, Tohoku University Graduate School of Medicine; 4Tohoku Medical Megabank Organization, Tohoku University

Correspondence to: Makoto Nakamura, Department of Surgery, Division of Ophthalmology, Kobe University Graduate School of Medicine. 7-5-2 Kusunoki-cho, Chuo-ku, Kobe 650-0017, Japan; Phone: 81-78-382-6048; email: [email protected]

Abstract

Purpose: Animal models have shown that stress to retinal ganglion cells and the optic nerve compromises the astrocyte-to-neuron lactate shuttle (ANLS) and increases glucose transporter (GLUT) expression and intraretinal glucose concentration. The aim of this study is to investigate the risk of glaucoma development associated with genes related to ANLS and GLUT by conducting single-nucleotide polymorphism (SNP) association and gene–gene interaction analyses in Japanese patients with normal-tension glaucoma (NTG) and controls.

Methods: Comprehensive association analysis of 120 SNPs in the genes aquaporin 9 (AQP9), glucose transporter 1 (GLUT1), glucose transporter 2 (GLUT2), glucose transporter 3 (GLUT3), glucose transporter 4 (GLUT4), monocarboxylate transporter 1 (MCT1), monocarboxylate transporter 2 (MCT2), monocarboxylate transporter 4 (MCT4), lactate dehydrogenase A (LDHA), lactate dehydrogenase B (LDHB), lactate dehydrogenase C (LDHC), lactate dehydrogenase D (LDHD), lactate dehydrogenase A-like 6A (LDHAL6A), and lactate dehydrogenase A-like 6B (LDHAL6B), as well as gene–gene interaction analysis of 153 SNP pairs from 18 genes, was performed in two different populations, comprising a total of 888 patients with NTG and 3,178 normal controls, both within each population and by integrating the populations.

Results: Three SNPs, GLUT3rs7976243, MCT2rs2706299, and LDHDrs11648082, were significantly associated with NTG in the combined analysis of the two populations (Pc = 0.00044–0.030). The minor allele of rs7976243 was protective (odds ratio [OR] = 0.86), while those of the other two SNPs increased NTG risk (OR = 1.16). SNP–SNP interaction analysis identified one significant pair, GLUT3rs11610602–MCT2rs11173115, with a protective effect on NTG (Pc = 0.0060, OR = 0.43). A joint additive effect analysis of 12 SNPs selected using a p < 0.05 threshold in the meta-analysis demonstrated a significant combined effect (p = 0.0017), and the SNP sum score was associated with NTG risk (OR = 1.09 per allele, p = 2.7 × 10−7).

Conclusions: SNPs in genes of ANLS-related proteins and GLUT, as well as their interactions, may contribute to the genetic risk of NTG in the Japanese population.

Introduction

Glaucomatous optic neuropathy (GON) is a multifactorial disease involving a complex interplay of genetic factors, epigenetic modifications, and environmental influences, resulting in irreversible degeneration of retinal ganglion cells (RGCs) and their axons [1]. In particular, in normal-tension glaucoma (NTG), where intraocular pressure (IOP) is within the normal range, it is hypothesized that IOP-independent factors are involved in the onset of GON [1].

It is known that brain neurons, including RGCs, utilize not only glucose taken directly from the bloodstream via specific glucose transporters (GLUT) but also lactate produced by astrocytes through aerobic glycolysis from the glucose they absorb. This lactate is transported to neurons via monocarboxylate transporters (MCTs) [2–6]. Lactate transferred from astrocytes to RGCs is converted to pyruvate by lactate dehydrogenase, which is then metabolized through the tricarboxylic acid cycle. During this process, adenosine triphosphate, the primary energy source for cells, is generated through the coupling of oxidative phosphorylation and the electron transport chain.

The transport of lactate as an energy substrate from astrocytes to neurons is referred to as the astrocyte-to-neuron lactate shuttle (ANLS). Although such glia–neuron compartmentalized glucose metabolic processes were discovered in Caenorhabditis elegans [7] and honeybees [8] over a quarter century ago, recent studies have increasingly revealed that they also play a critical role in maintaining physiologic functions in the mammalian brain and inner retina [2–6].

GLUT and MCT belong to the solute carrier transporter (SLC) superfamily, which does not require energy for solute transport and passively transports substrates along their concentration gradients [9,10]. The GLUT family is classified as SLC2A, while MCT is classified as SLC16A. Both families are further subdivided into transporters specific to glial or vascular endothelial cells and those specific to neurons. In the case of GLUT, the former corresponds to GLUT1 and the latter to GLUT3; for MCT, the former includes MCT1 and MCT4, and the latter includes MCT2.

In RGCs, lactate transport is also mediated by aquaporin (AQP) 9, which belongs to the broader class of water channels. The AQP family includes 13 isoforms, which are generally classified into water-selective AQPs that permit only water molecules; aquaglyceroporins that allow the passage of uncharged solutes such as lactate, glycerol, and urea; and super-AQPs that can transport other substances [11,12]. AQP9 belongs to the aquaglyceroporin group, along with AQP3, AQP7, and AQP10.

Previously, we reported that in the retina, MCT1, MCT2, and MCT4 form complexes with AQP9; that MCT2 and AQP9 are expressed in RGCs; that expression of AQP9 is reduced, and these complexes are disrupted in models of optic nerve crush and experimental ocular hypertension; and that these changes are accompanied by a reduction in RGC number and function, along with increased expression of GLUT1 and GLUT3 [13,14]. Furthermore, metabolomic analysis of retinas from a mouse model of ocular hypertension revealed elevated retinal glucose levels, along with increased glutamate concentrations [15].

These observations suggest that MCT and AQP9 cooperatively participate in ANLS at the protein level in RGCs and that stress to RGCs or the optic nerve may disrupt the ANLS interaction between MCT and AQP9. As a result, this disruption may induce a state of excessive glucose dependence, at least in part, via changes in GLUT expression. These findings prompted us to consider the hypothesis that genetic factors regulating ANLS-related molecules or GLUT expression may constitute one of the determinants for the onset of GON in patients with NTG.

The aim of this study is to investigate the risk of GON development associated with genes related to the ANLS and GLUT by conducting a single-nucleotide polymorphism (SNP) association analysis and gene–gene interaction (epistasis) analysis [16] in Japanese patients with NTG and normal controls.

Methods

Ethics statement

This study was approved by the Medical Ethical Review Board of Kobe University Graduate School of Medicine (approval numbers: B212043 and B242041), Yokohama City University School of Medicine (approval numbers: A150122005 and B240100055), and Tohoku University Graduate School of Medicine (approval numbers: 2021–1-184 and 2024–1-150) and adhered to the tenets of the Declaration of Helsinki. Written informed consent was obtained from all participants.

Participants

For the discovery study, 407 patients with NTG aged 40 years or older and 1,096 normal controls aged 50 years or older were recruited from patients attending Kobe University Hospital or Yokohama City University Hospital. For the replication study, 481 patients with NTG aged 40 years or older from Tohoku University Hospital and 2,082 individuals over 50 years of age from a general population cohort collected by the Tohoku Medical Megabank Organization [17] were enrolled.

The inclusion and exclusion criteria were based on previously published reports [18–20] as follows: NTG was diagnosed in the patients based on the following stringent inclusion criteria: the existence of GON accompanied by corresponding visual field loss, normal open angle with a Shaffer grade of 2 or higher, absence of IOP exceeding 21 mm Hg upon repeated measurement using Goldmann applanation tonometry without pharmacologic intervention, and the exclusion of any other retinal, optic nerve, or cerebral pathologies. Glaucomatous optic nerve abnormality was diagnosed when the vertical cup/disc ratio of the optic nerve head was 0.7 or greater, the rim width at the superior (11 to 1 o’clock) or inferior (5 to 7 o’clock) regions was less than or equal to 10% of the disc diameter, the interocular difference in the vertical cup/disc ratio was 0.2 or more, or a nerve fiber bundle defect was present. The Humphrey visual field test was conducted using the Swedish Interactive Standard Algorithm Program 30–2 or 24–2 (Carl Zeiss Meditec, Oberkochen, Germany). In accordance with Anderson and Patella’s criteria [21], the hemifield was considered abnormal when the pattern deviation probability plot revealed a cluster of three or more nonedge contiguous points (when the 30–2 program was used) with a probability of less than 5% in the upper or lower hemifield and in one of these points with a probability of less than 1%.

Additionally, the patient groups were rigorously categorized using the following inclusion and exclusion criteria. Patients with a refractive error of less than −8.0 D (the spherical equivalent [SE]) and those with a diagnosis of NTG at age under 40 or over 60 were excluded. The following changes were made to the selection criteria based on participant age to account for the detrimental effects of aging on the RGCs [18,19]: (1) no modification if the patient was diagnosed before the age of 50, (2) mean deviation (MD) of −10.00 dB or worse in at least one eye if the disease was diagnosed when the patient was between the ages of 50 and 55, and (3) MD of −15.00 dB or worse in at least one eye if the disease was diagnosed when the patient was older than 55. Patients diagnosed before 40 years of age were excluded to reduce potential genetic heterogeneity associated with early-onset disease. Patients whose refractions had altered during diagnosis—for instance, as a result of refractive surgery or cataract surgery—were not included. When only one eye had glaucomatous visual field loss, the affected eye’s refraction was used, and the glaucomatous visual field loss was assessed. When both eyes had glaucomatous visual field loss, the more seriously affected eye’s refraction and glaucomatous visual field loss were used.

Healthy volunteers from a geographic area comparable to that of the patients with NTG served as the control participants. They had no myopia or mild myopia with refractive errors of −3.00 D or less, nor any glaucoma or systemic or local illnesses that could cause glaucoma or alterations to the optic disc.

Candidate gene association analysis

A comprehensive SNP analysis was conducted on 14 gene regions: AQP9, GLUT1, GLUT2, GLUT3, GLUT4, MCT1, MCT2, MCT4, LDHA, LDHB, LDHC, LDHD, LDHAL6A, and LDHAL6B. We selected a total of 120 tagging SNPs that collectively covered the entire region of each gene, including 5 kb upstream and downstream. SNP selection was based on the HapMap Japanese dataset (minor allele frequency ≥5%, pairwise r2 ≥ 0.8), using the LD TAG SNP selection tool available on the SNPinfo web server.

SNP genotyping for the discovery cohort was performed using the TaqMan 5′ exonuclease assay, and primer–probe sets were supplied by Thermo Fisher Scientific (Foster City, CA). The polymerase chain reaction (PCR) for each SNP was performed with a 10-μl reaction mixture that contained 1× TaqMan GTXpress Master Mix (Thermo Fisher Scientific), 1× TaqMan SNP Genotyping Assay primer/probe mix, and 3 ng genomic DNA. The PCR conditions were as follows: 95 °C for 20 s, followed by 40 cycles of denaturation at 95 °C for 3 s and annealing/extension at 60 °C for 20 s. The fluorescent probe signal was detected using the StepOnePlus Real-Time PCR System (Thermo Fisher Scientific), according to the manufacturer’s instructions.

In the replication cohort, we used genome-wide association study data generated using the Japonica Array NEO by Tohoku Medical Megabank Organization [22]. These data were imputed following the method described in a previous study [23]. Specifically, a cross-imputed reference panel was first constructed from the 3.5KJPNv2 [24] and 1000 Genomes Project phase 3 [25] datasets using IMPUTE2 [26]. Genotype imputation was then performed using IMPUTE4 [27], with the resulting merged reference panel, after prephasing with SHAPEIT2 [28]. Of the 120 SNPs selected for analysis, one SNP (rs11024665) could not be imputed and was excluded. The remaining 119 SNPs, all with an INFO score >0.8, were used for subsequent analyses.

Association analyses of SNPs were performed under an additive genetic model using logistic regression for binary traits, as implemented in SNP & Variation Suite software version 8.8.3 (Golden Helix, Bozeman, MT) and PLINK software [29]. Age and sex were included in the model as covariates. The obtained p values were corrected for multiple testing with the Bonferroni method by multiplying the nominal p values by the number of tagging SNPs (n = 120). A corrected P (Pc) value <0.05 was considered significant. Linkage disequilibrium (LD) between SNPs was assessed using SNP & Variation Suite software.

SNP–SNP interaction analysis

We also performed SNP–SNP interaction analysis using the epistasis function of PLINK. During this process, SNPs in strong LD, defined as r2 > 0.70, were subjected to LD pruning by selecting one representative SNP from each LD group, choosing the variant with the lowest meta-analysis p value within each group, resulting in the exclusion of 20 SNPs. In addition, SNPs with a minor allele frequency <5% or a Hardy–Weinberg equilibrium p value <0.05 were excluded to further ensure data quality. As a result, 19 SNPs were removed, leaving 81 after quality control and pruning. Of these 81 SNPs, 18 that showed nominal significance (p < 0.10) in the meta-analysis of SNP associations combining the discovery and replication cohorts were selected. A total of 153 SNP pairs derived from these 18 SNPs were then included in the SNP–SNP interaction analysis. Asymptotic p values for the interaction terms were computed based on logistic regression for binary traits. The obtained p values were corrected for multiple testing with the Bonferroni method by multiplying the nominal p values by the number of SNP pairs (n = 153), and a Pc value <0.05 was considered significant.

Meta-analysis

Meta-analyses combining the discovery and replication cohorts were conducted for both the SNP association results and the SNP–SNP interaction results, using the inverse-variance method under a random-effects model with the META software, which accounts for heterogeneity in effect sizes (odds ratios [ORs]) between cohorts. The obtained p values were corrected for multiple testing with the Bonferroni method by multiplying the nominal p values by the number of tagging SNPs (n = 120) or SNP pairs (n = 153). A Pc value <0.05 was considered significant.

Joint additive effect analysis

A joint additive effect analysis was performed to evaluate the combined contribution of multiple candidate SNPs. Twenty SNPs that met the selection threshold of p < 0.05 in the meta-analysis of single-variant association tests were first selected as candidate variants. These SNPs were further filtered by LD pruning (r2 > 0.70), minor allele frequency (<5%), and deviation from Hardy–Weinberg equilibrium (p < 0.05), resulting in 12 SNPs for analysis.

For the joint additive effect analysis, genotype and phenotype data from the discovery and replication cohorts were merged into a single dataset, and a cohort indicator variable was included as a covariate to account for between-cohort differences. Missing genotype data were imputed using the cohort-specific mean allele dosage for each SNP. Joint additive effects were assessed using a multivariable logistic regression model in which all 12 SNPs were simultaneously included under an additive genetic model, with adjustment for age, sex, and cohort. For each SNP, mutually adjusted ORs, 95% confidence intervals (CIs), and p values were estimated. The overall joint effect of the 12 SNPs was evaluated using a likelihood ratio test comparing the full model (covariates + all SNPs) with the covariate-only model. In addition, a SNP sum score, defined as the total number of risk alleles across the 12 SNPs, was calculated for each individual, and its association with NTG was examined using a separate logistic regression model.

All statistical analyses were conducted in R (version 4.3.1) using the glm function with a binomial family. Regression summaries and effect estimates were obtained with the broom and dplyr packages.

Functional annotation

We investigated the functional roles of the identified SNPs using HaploReg v4.2 [30] and RegulomeDB [31]. Expression quantitative trait locus data were extracted from the GTEx Portal online database, version 10, which provides expression quantitative trait locus results across 47 different human tissues for assessment of associations between an identified SNP and the target gene [32].

Results

Demographic and clinical characteristics of the study populations

The demographic and clinical characteristics of patients with NTG and controls are summarized in Table 1. In the discovery cohort, the mean ages of patients and controls were 49.3 ± 5.4 years and 63.1 ± 8.5 years, respectively, while in the replication cohort, they were 61.5 ± 12.1 years and 64.0 ± 6.8 years, respectively. A significant difference in age was observed between patients and controls in both cohorts. The NTG cases in the discovery cohort were significantly younger than those in the replication cohort. In contrast, there was no difference in age among the controls between the two cohorts. In terms of sex distribution, 43.0% of patients and 44.5% of controls in the discovery cohort were male, and 49.9% of patients and 50.1% of controls in the replication cohort were male. No significant difference in sex distribution was observed in either cohort. The proportion of females among NTG cases in the discovery cohort was significantly higher than that in the replication cohort, whereas there was no difference between the two cohorts among the controls. Age and sex were included as covariates in the association analyses (see Methods). The mean SE in the discovery cohort was −3.82 ± 2.95 D for patients and 0.28 ± 1.88 D for controls, showing a significant difference between the groups. In the replication cohort, the mean SE for patients was −4.51 ± 4.12 D. The NTG cases in the replication cohort had significantly greater myopic refractive error than those in the discovery cohort. SE data were not available for controls in the replication cohort; therefore, SE was not included as a covariate in the association analyses to ensure consistent covariate adjustment across both cohorts. The mean MDs of patients in the discovery cohort and replication cohort were −9.57 ± 7.81 dB and −14.12 ± 8.14 dB, respectively. The latter had significantly worse MD compared with the former. The mean IOPs of patients in the discovery cohort and replication cohort were 15.73 ± 2.78 mmHg and 13.13 ± 2.60 mmHg, respectively. The former was significantly higher than the latter.

To assess the potential confounding effect of SE, we performed a quantitative linear regression analysis within the discovery cohort, treating SE as a continuous outcome variable and each SNP as a predictor. None of the SNPs tested in this study showed a significant association with SE, suggesting that SE is unlikely to be a confounding factor in our genetic association analyses.

Association analysis

In this study, we conducted association analyses on 120 tagging SNPs covering 14 gene regions (AQP9, GLUT1, GLUT2, GLUT3, GLUT4, MCT1, MCT2, MCT4, LDHA, LDHB, LDHC, LDHD, LDHAL6A, and LDHAL6B). First, these 120 SNPs were analyzed in the discovery cohort, and subsequently, the same set of SNPs was examined in an independent replication cohort to validate the findings. Of these SNPs, 12 showed an association with NTG with p < 0.05 in the discovery cohort, with the strongest association observed for rs3820549 located in GLUT1 (Appendix 1). However, none of these associations remained significant after Bonferroni correction for multiple testing (Pc > 0.05). In the replication cohort, six SNPs showed an association with NTG with p < 0.05, with the strongest signal observed for rs12313154 located in GLUT3 (Appendix 1); however, none remained significant after Bonferroni correction (Pc > 0.05). Among these six SNPs, only rs7976243 in GLUT3 showed p < 0.05 in the discovery cohort, exhibited a consistent direction of effect (OR) between the two cohorts, and achieved statistical significance in the subsequent meta-analysis (Pc = 0.00044; Table 2, Appendix 1).

In addition to GLUT3rs7976243, the meta-analysis combining the discovery and replication cohorts identified two more SNPs that showed statistically significant associations with NTG: rs2706299 in MCT2 and rs11648082 in LDHD (Pc = 0.030 and 0.0084, respectively; Table 2, Appendix 1). Among these three SNPs, the minor allele of GLUT3rs7976243 was associated with a decreased risk of NTG (OR = 0.86), whereas the minor alleles of the other two SNPs were associated with an increased risk of NTG (MCT2rs2706299: OR = 1.16; LDHDrs11648082: OR = 1.16).

SNP–SNP interaction analysis

In the discovery cohort, seven SNP pairs showed associations with NTG at p < 0.05, and 12 SNP pairs did so in the replication cohort; however, none of these associations remained significant after Bonferroni correction (Appendix 2). Among them, two SNP pairs, GLUT3rs11610602-MCT2rs11173115 and AQP9rs2414535-LDHDrs755099, showed p < 0.05 in both cohorts and exhibited consistent directions of effect (ORs). Of these, the GLUT3rs11610602–MCT2rs11173115 pair reached statistical significance in the meta-analysis after multiple testing correction (Pc = 0.0060), with an odds ratio of 0.43, indicating a protective interaction between their minor alleles in NTG development (Table 3). In contrast, the AQP9rs2414535–LDHDrs755099 pair did not reach significance in the meta-analysis (Pc = 0.18; Table 3). In total, 16 SNP pairs showed p < 0.05 in the meta-analysis, but only the GLUT3rs11610602–MCT2rs11173115 pair remained significant after correction (Appendix 2).

Joint additive effect analysis

To further evaluate the combined effect of the identified variants, we performed a joint additive effect analysis of the 12 SNPs selected based on the meta-analysis of association tests with p < 0.05 (Table 4). In the SNPwise analysis, none of the SNPs showed a significant independent association after correction for multiple testing, although rs7165001 showed a nominal association (p = 0.034). These findings suggest that the individual effects of each SNP are modest and may not be independently detectable when accounting for other variants. In contrast, the likelihood ratio test demonstrated that inclusion of all 12 SNPs significantly improved the model fit compared with the base model (p = 0.0017), indicating a significant joint effect of these variants. Furthermore, the SNP sum score, defined as the total number of risk alleles across the 12 SNPs, was significantly associated with the risk of NTG (p = 2.7 × 10−7, OR = 1.09), corresponding to an approximately 9% increase in NTG odds per additional risk allele. In other words, if all 12 alleles are risk alleles, the OR would be 1.0912 = 2.81. Thus, having 12 risk alleles results in an OR 2.81 times higher compared with having zero risk alleles (OR = 1).

Functional analysis

The five SNPs identified in the association analyses and SNP–SNP interaction analyses are located within the noncoding region, including intronic and untranslated regions, which can significantly affect gene expression. Among the three SNPs identified in the association analyses, the HaploReg v4.2 database predicted that all would be located in promoter histone marks and enhancer histone marks, three in DNase hypersensitivity regions, and two in protein-binding regions (Appendix 3). The database also predicted that all of these three SNPs would alter the regulatory motif of transcription factors, suggesting a potential role for them in regulating gene expression. In addition, the RegulomeDB scores of these three SNPs were 1f, indicating that these SNPs are likely to affect transcription factor binding and the expression of a target gene (Appendix 3).

Among the two SNPs identified in the SNP–SNP interaction analyses, HaploReg predicted that GLUT3rs11610602 would be located in a DNase hypersensitivity region and a protein-binding region and would alter the regulatory motif of transcription factors. In addition, its RegulomeDB score was 1f, suggesting its potential role in regulating gene expression (Appendix 3). MCT2rs11173115 was predicted to be located within promoter histone marks, enhancer histone marks, DNase hypersensitivity, and protein-binding regions and alter the regulatory motif of transcription factors in HaploReg. Its RegulomeDB score was also 1f (Appendix 3).

The GTEx Portal database indicated that two of the three SNPs identified in the association analyses, GLUT3rs7976243 and LDHDrs11648082, affected expression levels of their target genes. The C allele of GLUT3rs7976243 was significantly associated with increased GLUT3 expression in multitissue meta-analysis (p = 8.0 × 10−15; Appendix 3). The T allele of rs11648082 in LDHD was significantly associated with increased expression of LDHD in many tissues/organs (p = 3.0 × 10−9; Appendix 3). On the other hand, although MCT2rs2706299 showed a significant increase or decrease in MCT2 expression in a limited number of tissues, the multitissue meta-analysis did not demonstrate a significant association between rs2706299 and MCT2 expression (p = 0.28; Appendix 3). Regarding heterogeneity across tissues, rs7976243 (I2 = 0.58, p = 4.0 × 10−7) and rs2706299 (I2 = 0.55, p = 5.0 × 10−6) showed moderate to high levels, indicating substantial variation in expression effects. rs11648082 showed a lower degree of heterogeneity (I2 = 0.38, p = 0.0050), suggesting relatively consistent effects across tissues.

Among the two SNPs identified in the SNP–SNP interaction analyses, GLUT3rs11610602 exhibited tissue-specific expression differences. The T allele of rs11610602 was significantly associated with decreased GLUT3 expression specifically in brain tissues (p = 4.0 × 10−7 in brain tissue meta-analysis). However, this association was not observed in the meta-analysis across all GTEx tissues (p = 0.73; Appendix 3). For MCT2rs11173115, the direction of effect on MCT2 expression varied across tissues. In some tissues, the SNP was associated with a significant increase in expression, while in others, a significant decrease was observed. Nonetheless, the overall multitissue meta-analysis did not show a significant association between rs11173115 and MCT2 expression (p = 0.84; Appendix 3). Across tissues, rs11173115 showed high heterogeneity (I2 = 0.73, p = 2.0 × 10−16), whereas rs11610602 showed minimal heterogeneity (I2 = 0.14, p = 0.21).

Discussion

Recent large-scale genome-wide association studies have been reported to explain approximately 14% of the genetic predisposition to glaucoma [33]. Nevertheless, many genetic factors remain unaccounted for as part of the so-called missing heritability [34].

In the present study, we conducted a discovery and replication study in two distinct cohorts using a comprehensive set of SNPs from 14 candidate genes involved in ANLS and GLUT, followed by a meta-analysis. Among the 120 SNPs analyzed in the association study, SNP rs7976243 in GLUT3 showed nominal significance in the additive model in both the discovery and replication cohorts. In the meta-analysis, it reached statistical significance after Bonferroni correction, with the minor C allele being associated with a reduced risk of NTG (meta-analysis: OR = 0.86; 95% CI, 0.80–0.91; Pc = 0.00044; Table 2 and Appendix 1).

Previously, we reported increased expression of GLUT1 and GLUT3, along with elevated retinal glucose levels in a mouse model of optic nerve crush [13]. Similarly, in a metabolomic analysis using a mouse model of ocular hypertension, we also observed increased glucose concentrations in the retina [15]. According to prior metabolomic studies, retinal glucose levels were approximately threefold higher 2 weeks after optic nerve crush compared to controls [35] and nearly 50-fold higher during the IOP elevation phase in DBA/2J mice, a model of spontaneous glaucoma [36].

Under physiologic conditions, glucose levels in brain parenchyma are tightly regulated to remain below half of blood glucose levels. When blood glucose rises, cerebral glucose uptake significantly decreases to maintain normal brain metabolic activity [37]. Artificially elevating glucose levels in the brain also leads to a doubling of glutamate concentrations [38]. It is well known that excess glutamate induces neuronal cell death via excitotoxicity [39]. In emergency medicine, elevated blood glucose or increased brain glucose levels following traumatic brain injury are associated with increased patient mortality [40,41]. Proposed mechanisms include tissue acidification in injured brain regions, increased glucose–sodium exchange, mitochondrial fragmentation, enhanced inflammatory responses, promotion of advanced glycation end-product formation, excessive production of reactive oxygen species, and activation of endothelial protease C.

Recently, it has been reported that patients with glaucoma showed higher levels of serum glucose compared to patients without glaucoma and that glaucoma incidence was much greater in patients with hyperglycemia than in those with a high polygenic risk score (PRS) but normoglycemia [42]. Moreover, another recent study showed that while glucose levels were positively associated with glaucoma risk, glycolytic metabolites showed a significant negative correlation [43].

Taken together, the observed risk-reducing effect of the neuron-specific GLUT3 SNP rs7976243 on NTG may suggest that Japanese individuals with a genetic predisposition for excessive glucose uptake in RGCs are more susceptible to developing NTG.

However, our functional annotation analysis in this study suggests that rs7976243 has a regulatory effect on the expression of a target gene, with the minor C allele being significantly associated with increased GLUT3 expression across multiple tissues. Nevertheless, the analysis also found moderate heterogeneity of GLUT3 expression (I2 = 0.58), particularly among brain tissues, with both upregulation and downregulation depending on the brain region. However, since the GTEx database does not include retinal tissue, the effect of rs7976243 on GLUT3 expression in the retina remains unclear. Therefore, further studies using retina-specific datasets are needed to clarify its role in NTG pathogenesis.

In the association analysis, although rs2706299 in MCT2 and rs11648082 in LDHD did not reach statistical significance in each individual cohort, all three showed a trend toward increased odds ratios for NTG onset under the additive model. In the meta-analysis, they reached statistical significance after Bonferroni correction (Table 2).

MCT2 is a neuron-specific monocarboxylate transporter known to play a critical role in the survival and functional maintenance of RGCs. In our previous study, mice receiving intravitreal injections of an MCT2 inhibitor exhibited more severe RGC damage after optic nerve crush compared to untreated controls. Harun-Or-Rashid et al. [44] reported that knockout of Mct2 in DBA/2J mice—a model of spontaneous glaucoma—resulted in significantly reduced visual evoked potentials and retinal adenosine triphosphate production, whereas overexpression of Mct2 preserved RGC density and axon count at levels comparable to controls. These meta-analysis findings may suggest that Japanese individuals genetically predisposed to lower MCT2 expression have a higher risk of developing NTG.

LDHD is a mitochondrial enzyme that converts lactate into pyruvate. Although the precise function of LDHD remains unclear, genetic variants in LDHD have been reported to cause central nervous system disorders accompanied by elevated blood D-lactate levels [45]. On the other hand, a recent study reported that high levels of lactate, pyruvate, and citrate can reduce glaucoma risk by 29%, even among individuals in the top PRS quartile, and that administration of pyruvate in drinking water suppressed optic nerve degeneration in a spontaneous glaucoma mouse model [46]. It is known that the physiologic isomer of lactate is the L-form, whereas the D-form is inactive or even exerts toxic effects on retinal neurons [47]. Taken together, it has been suggested that in a Japanese population carrying certain LDHD gene polymorphisms, the serum or local concentration of the toxic D-lactate isomer increases, resulting in enhanced oxidative stress in the electron transport chain and a potentially elevated risk of developing GON.

Next, we examined pairwise interactions among SNPs from the 14 candidate genes. As a result, individuals carrying both GLUT3 rs11610602 and MCT2rs11173115 showed nominal significance in both the discovery and replication cohorts and reached statistical significance after Bonferroni correction in the meta-analysis, indicating a reduced risk of NTG (meta-analysis: OR = 0.43; 95% CI, 0.28–0.64; Pc = 0.0060; Table 3). Although these SNPs are located within the same genes as those identified in the single-SNP association analyses (GLUT3rs7976243 and MCT2rs2706299), they are in low LD at each locus (r2 = 0.20 for GLUT3 and r2 = 0.02 for MCT2), suggesting that the observed interaction effect likely involves genetically distinct signals relative to the single-SNP findings.

Notably, neither rs11610602 nor rs11173115 reached statistical significance in the single-SNP association analyses in either cohort when considered individually. Epistasis refers to interactions between multiple, often distant, genes that influence phenotypic expression [16,48–50]. In some cases, each gene may have an additive or multiplicative effect on the trait individually, while in others, no single gene shows an effect on its own, yet a significant phenotypic effect emerges when specific combinations occur. The present findings appear to correspond to the latter case, highlighting the importance of SNP–SNP interactions that cannot be detected through single-marker effects.

Both GLUT3rs7976243 and MCT2rs2706299 are located in regions where histone modifications characteristic of promoters or enhancers have been identified, suggesting a high likelihood that they function as promoter/enhancer elements. This may explain why each SNP independently contributes significantly to disease risk without requiring interaction with other SNPs. In contrast, although MCT2rs11173115 also exhibits regulatory features, including promoter and enhancer histone marks, it did not show a significant independent effect. Instead, its association with NTG emerged through interaction with GLUT3rs11610602. These findings suggest that regulatory potential alone may not be sufficient for independent disease association and that certain variants may exert their effects in combination with other genetic factors.

In recent years, PRS, which incorporates a large number of SNPs, has been explored as a means to improve prediction of various phenotypes or diseases, including glaucoma [51,52]. However, SNPs included in the PRS are typically those that show statistically significant associations on their own, and current PRS models remain insufficient to explain the missing heritability. One possible reason is that interactions between SNPs that do not individually affect phenotypic expression may not have been taken into account [16,47,48]. Since no consensus exists regarding significance thresholds in epistasis analysis, the current study limited its discussion to the SNP pair described above. However, when we conducted interaction analysis on 153 SNP pairs derived from 18 selected SNPs, 23 SNP pairs reached nominal significance in either or both cohorts and/or in the meta-analysis (Appendix 2). These interactions showed both increased and decreased odds ratios for NTG onset, depending on the SNP pair. How to best analyze and interpret such data remains an open question.

In this study, we also performed a joint additive effect analysis to evaluate the combined contribution of the 12 SNPs selected based on the meta-analysis of association tests with p < 0.05 with PRS in mind (Table 4). Although none of the individual SNPs showed a significant independent association after mutual adjustment, the likelihood ratio test indicated a significant joint effect, and the SNP sum score was significantly associated with NTG, with risk increasing according to the number of risk alleles, with an approximately 9% increase in NTG odds per additional risk allele. These findings suggest that multiple variants with modest individual effects may collectively contribute to NTG susceptibility, consistent with a modest cumulative polygenic effect. As an example, the ORs for GLUT3rs11610602 and MCT2rs11173115 were 0.86 and 0.82 in the discovery cohort, 0.94 and 0.85 in the replication cohort, and 0.93 and 0.84 in the meta-analysis, respectively. Therefore, if a joint additive effect were present, the expected OR in the meta-analysis would be 0.93 × 0.84 = 0.78. In contrast, the interaction OR obtained from the epistasis analysis in the meta-analysis was 0.43 (Pc = 0.0060). Thus, the interaction effect of this specific pair greatly exceeded the expected additive effect. Taken together, our data suggest the involvement of both additive and interaction effects in the genetic architecture of NTG with a different power of effect depending on the combination of SNP pairs.

The limitations of this study include a relatively small sample size and the restriction to a Japanese population, which limits the generalizability of our findings. A more stringent threshold (e.g., p < 0.001) might help confirm associations in future studies with larger samples or independent cohorts. Diurnal IOP variation over 24 hours was not assessed, and central corneal thickness was not measured. For these reasons, the possibility that some patients with high-tension glaucoma were inadvertently included cannot be entirely excluded. Differences in control subject backgrounds between the discovery and replication cohorts may also explain the limited consistency between association and interaction results across the two groups. Furthermore, we did not conduct functional validation of the SNPs examined, so the interpretations discussed above remain speculative. Nevertheless, our findings are consistent with previous reports showing altered expression and interaction of ANLS-related transporters and GLUTs, as well as altered retinal glucose metabolism, in glaucoma model animals [13–15].

In conclusion, SNPs in genes of ANLS-related proteins and GLUT, as well as their interactions, may contribute to the genetic risk of NTG in the Japanese population.

Appendix 1. Association results of all 120 analyzed SNPs with normal tension glaucoma.

Appendix 2. Results of SNP–SNP interaction analysis for 153 pairs derived from 18 nominally associated SNPs.

Appendix 3. Functional annotation of significant SNPs associated with normal tension glaucoma.

Acknowledgments

This study was supported in part by JSPS KAKENHI Grant Number 24K02603 (MN, AM). We thank Mses. Sachiyo Sugimoto and Chinatsu Gocho in Tohoku Medical Megabank Organization, Tohoku University, Sendai, Japan for their technical assistance. None of the authors have any conflict of interest.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoD­erivatives 4.0 International License, which permits reusers to copy and distribute the material in any medium or format in unadapted form only, for noncommercial purposes only, and only so long as attribution is given to the creator. CC BY-NC-ND includes the following elements: BY: credit must be given to the creator. NC: Only noncommercial uses of the work are permitted. ND: No derivatives or adaptations of the work are permitted.

References

  1. Jayaram H, Kolko M, Friedman DS, Gazzard G. Glaucoma: now and beyond. Lancet. 2023; 402:1788-801. [PMID: 37742700]
  2. Pellerin L, Magistretti PJ. Glutamate uptake into astrocytes stimulates aerobic glycolysis: a mechanism coupling neuronal activity to glucose utilization. Proc Natl Acad Sci U S A. 1994; 91:10625-9. [PMID: 7938003]
  3. Magistretti PJ, Pellerin L, Rothman DL, Shulman RG. Energy on demand. Science. 1999; 283:496-7. [PMID: 9988650]
  4. Magistretti PJ, Allaman I. Lactate in the brain: from metabolic end-product to signalling molecule. Nat Rev Neurosci. 2018; 19:235-49. [PMID: 29515192]
  5. Brooks GA. The Science and Translation of Lactate Shuttle Theory. Cell Metab. 2018; 27:757-85. [PMID: 29617642]
  6. Hurley JB, Lindsay KJ, Du J. Glucose, lactate, and shuttling of metabolites in vertebrate retinas. J Neurosci Res. 2015; 93:1079-92. [PMID: 25801286]
  7. Volkenhoff A, Weiler A, Letzel M, Stehling M, Klämbt C, Schirmeier S. Glial glycolysis is essential for neuronal survival in Drosophila. Cell Metab. 2015; 22:437-47. [PMID: 26235423]
  8. Tsacopoulos M, Veuthey AL, Saravelos SG, Perrottet P, Tsoupras G. Glial cells transform glucose to alanine, which fuels the neurons in the honeybee retina. J Neurosci. 1994; 14:1339-51. [PMID: 8120629]
  9. Jha MK, Morrison BM. Glia-neuron energy metabolism in health and diseases: New insights into the role of nervous system metabolic transporters. Exp Neurol. 2018; 309:23-31. [PMID: 30044944]
  10. Zhang Y, Zhang Y, Sun K, Meng Z, Chen L. The SLC transporter in nutrient and metabolic sensing, regulation, and drug development. J Mol Cell Biol. 2019; 11:1-13. [PMID: 30239845]
  11. Amiry-Moghaddam M, Ottersen OP. The molecular basis of water transport in the brain. Nat Rev Neurosci. 2003; 4:991-1001. [PMID: 14682361]
  12. Badaut J, Ashwal S, Obenaus A. Aquaporins in cerebrovascular disease: a target for treatment of brain edema? Cerebrovasc Dis. 2011; 31:521-31. [PMID: 21487216]
  13. Mori S, Kurimoto T, Miki A, Maeda H, Kusuhara S, Nakamura M. Aqp9 gene deletion enhances retinal ganglion cell (RGC) death and dysfunction induced by optic nerve crush: Evidence that aquaporin 9 acts as an astrocyte-to-neuron lactate shuttle in concert with monocarboxylate transporters to support RGC function and survival. Mol Neurobiol. 2020; 57:4530-48. [PMID: 32748371]
  14. Murai Y, Mori S, Okuda M, Kusuhara S, Kurimoto T, Nakamura M. Effects of elevated intraocular pressure on retinal ganglion cell density and expression and interaction of retinal aquaporin 9 and monocarboxylate transporters. Ophthalmic Res. 2023; 66:1222-9. [PMID: 37647868]
  15. Arai-Okuda M, Murai Y, Maeda H, Kanamori A, Miki T, Naito T, Sugihara K, Kono M, Tanito M, Onoe H, Hirooka K, Kiuchi Y, Shinohara M, Kusuhara S, Mori S, Ueda K, Sakamoto M, Yamada-Nakanishi Y, Nakamura M. Potentially compromised systemic and local lactate metabolic balance in glaucoma, which could increase retinal glucose and glutamate concentrations. Sci Rep. 2024; 14:3683 [PMID: 38355836]
  16. Blanco-Marchite C, Sánchez-Sánchez F, López-Garrido MP, Iñigez-de-Onzoño M, López-Martínez F, López-Sánchez E, Alvarez L, Rodríguez-Calvo PP, Méndez-Hernández C, Fernández-Vega L, García-Sánchez J, Coca-Prados M, García-Feijoo J, Escribano J. WDR36 and P53 gene variants and susceptibility to primary open-angle glaucoma: analysis of gene-gene interactions. Invest Ophthalmol Vis Sci. 2011; 52:8467-78. [PMID: 21931130]
  17. Kuriyama S, Yaegashi N, Nagami F, Arai T, Kawaguchi Y, Osumi N, Sakaida M, Suzuki Y, Nakayama K, Hashizume H, Tamiya G, Kawame H, Suzuki K, Hozawa A, Nakaya N, Kikuya M, Metoki H, Tsuji I, Fuse N, Kiyomoto H, Sugawara J, Tsuboi A, Egawa S, Ito K, Chida K, Ishii T, Tomita H, Taki Y, Minegishi N, Ishii N, Yasuda J, Igarashi K, Shimizu R, Nagasaki M, Koshiba S, Kinoshita K, Ogishima S, Takai-Igarashi T, Tominaga T, Tanabe O, Ohuchi N, Shimosegawa T, Kure S, Tanaka H, Ito S, Hitomi J, Tanno K, Nakamura M, Ogasawara K, Kobayashi S, Sakata K, Satoh M, Shimizu A, Sasaki M, Endo R, Sobue K, Tohoku Medical Megabank Project Study Group T, Yamamoto M. The Tohoku Medical Megabank Project: Design and Mission. J Epidemiol. 2016; 26:493-511. [PMID: 27374138]
  18. Shibuya E, Meguro A, Ota M, Kashiwagi K, Mabuchi F, Iijima H, Kawase K, Yamamoto T, Nakamura M, Negi A, Sagara T, Nishida T, Inatani M, Tanihara H, Aihara M, Araie M, Fukuchi T, Abe H, Higashide T, Sugiyama K, Kanamoto T, Kiuchi Y, Iwase A, Ohno S, Inoko H, Mizuki N. Association of Toll-like receptor 4 gene polymorphisms with normal tension glaucoma. Invest Ophthalmol Vis Sci. 2008; 49:4453-7. [PMID: 18586872]
  19. Yasumura R, Meguro A, Ota M, Nomura E, Uemoto R, Kashiwagi K, Mabuchi F, Iijima H, Kawase K, Yamamoto T, Nakamura M, Negi A, Sagara T, Nishida T, Inatani M, Tanihara H, Aihara M, Araie M, Fukuchi T, Abe H, Higashide T, Sugiyama K, Kanamoto T, Kiuchi Y, Iwase A, Ohno S, Inoko H, Mizuki N. Investigation of the association between SLC1A3 gene polymorphisms and normal tension glaucoma. Mol Vis. 2011; 17:792-6. [PMID: 21528001]
  20. Shiga Y, Akiyama M, Nishiguchi KM, Sato K, Shimozawa N, Takahashi A, Momozawa Y, Hirata M, Matsuda K, Yamaji T, Iwasaki M, Tsugane S, Oze I, Mikami H, Naito M, Wakai K, Yoshikawa M, Miyake M, Yamashiro K, Kashiwagi K, Iwata T, Mabuchi F, Takamoto M, Ozaki M, Kawase K, Aihara M, Araie M, Yamamoto T, Kiuchi Y, Nakamura M, Ikeda Y, Sonoda KH, Ishibashi T, Nitta K, Iwase A, Shirato S, Oka Y, Satoh M, Sasaki M, Fuse N, Suzuki Y, Cheng CY, Khor CC, Baskaran M, Perera S, Aung T, Vithana EN, Cooke Bailey JN, Kang JH, Pasquale LR, Haines JL, Wiggs JL, Burdon KP, Gharahkhani P, Hewitt AW, Mackey DA, MacGregor S, Craig JE, Allingham RR, Hauser M, Ashaye A, Budenz DL, Akafo S, Williams SEI, Kamatani Y, Nakazawa T, Kubo M, Japan Glaucoma Society Omics Group (JGS-OG). NEIGHBORHOOD Consortium. Genome-wide association study identifies seven novel susceptibility loci for primary open-angle glaucoma. Hum Mol Genet. 2018; 27:1486-96. [PMID: 29452408]
  21. Anderson DR, Patella VM. Automated Static Perimetry. 2nd ed. St. Louis: Mosby; 1999. p.121–190.
  22. Sakurai-Yageta M, Kumada K, Gocho C, Makino S, Uruno A, Tadaka S, Motoike IN, Kimura M, Ito S, Otsuki A, Narita A, Kudo H, Aoki Y, Danjoh I, Yasuda J, Kawame H, Minegishi N, Koshiba S, Fuse N, Tamiya G, Yamamoto M, Kinoshita K. Japonica Array NEO with increased genome-wide coverage and abundant disease risk SNPs. J Biochem. 2021; 170:399-410. [PMID: 34131746]
  23. Hanyuda A, Raita Y, Ninomiya T, Hashimoto K, Takada N, Sato K, Inoue J, Koshiba S, Tamiya G, Narita A, Akiyama M, Omodaka K, Tsuda S, Yokoyama Y, Himori N, Yamamoto Y, Taniguchi T, Negishi K, Nakazawa T. Metabolomic Profiling of Open-Angle Glaucoma Etiologic Endotypes: Tohoku Multi-Omics Glaucoma Study. Invest Ophthalmol Vis Sci. 2024; 65:44 [PMID: 39565301]
  24. Tadaka S, Katsuoka F, Ueki M, Kojima K, Makino S, Saito S, Otsuki A, Gocho C, Sakurai-Yageta M, Danjoh I, Motoike IN, Yamaguchi-Kabata Y, Shirota M, Koshiba S, Nagasaki M, Minegishi N, Hozawa A, Kuriyama S, Shimizu A, Yasuda J, Fuse N, Tamiya G, Yamamoto M, Kinoshita K, Tohoku Medical Megabank Project Study Group. 3.5KJPNv2: an allele frequency panel of 3552 Japanese individuals including the X chromosome. Hum Genome Var. 2019; 6:28 [PMID: 31240104]
  25. 1000 Genomes Project Consortium; Auton A, Brooks LD, Durbin RM, Garrison EP, Kang HM, Korbel JO, Marchini JL, McCarthy S, McVean GA, Abecasis GR. A global reference for human genetic variation. Nature. 2015; 526:68-74.
  26. Howie BN, Donnelly P, Marchini J. A flexible and accurate genotype imputation method for the next generation of genome-wide association studies. PLoS Genet. 2009; 5e1000529 [PMID: 19543373]
  27. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, Motyer A, Vukcevic D, Delaneau O, O’Connell J, Cortes A, Welsh S, Young A, Effingham M, McVean G, Leslie S, Allen N, Donnelly P, Marchini J. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018; 562:203-9. [PMID: 30305743]
  28. Delaneau O, Zagury JF, Marchini J. Improved whole-chromosome phasing for disease and population genetic studies. Nat Methods. 2013; 10:5-6. [PMID: 23269371]
  29. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, Maller J, Sklar P, de Bakker PI, Daly MJ, Sham PC. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007; 81:559-75. [PMID: 17701901]
  30. Ward LD, Kellis M. HaploReg v4: systematic mining of putative causal variants, cell types, regulators and target genes for human complex traits and disease. Nucleic Acids Res. 2016; 44D1:D877-81. [PMID: 26657631]
  31. Boyle AP, Hong EL, Hariharan M, Cheng Y, Schaub MA, Kasowski M, Karczewski KJ, Park J, Hitz BC, Weng S, Cherry JM, Snyder M. Annotation of functional variation in personal genomes using RegulomeDB. Genome Res. 2012; 22:1790-7. [PMID: 22955989]
  32. GTEx Consortium. The Genotype-Tissue Expression (GTEx) project. Nat Genet. 2013; 45:580-5. [PMID: 23715323]
  33. Han X, Gharahkhani P, Hamel AR, Ong JS, Rentería ME, Mehta P, Dong X, Pasutto F, Hammond C, Young TL, Hysi P, Lotery AJ, Jorgenson E, Choquet H, Hauser M, Cooke Bailey JN, Nakazawa T, Akiyama M, Shiga Y, Fuller ZL, Wang X, Hewitt AW, Craig JE, Pasquale LR, Mackey DA, Wiggs JL, Khawaja AP, Segrè AV, MacGregor S, 23andMe Research Team. International Glaucoma Genetics Consortium. Large-scale multitrait genome-wide association analyses identify hundreds of glaucoma risk loci. Nat Genet. 2023; 55:1116-25. [PMID: 37386247]
  34. Maher B. Personal genomes: The case of the missing heritability. Nature. 2008; 456:18-21. [PMID: 18987709]
  35. Agudo-Barriuso M, Lahoz A, Nadal-Nicolás FM, Sobrado-Calvo P, Piquer-Gil M, Díaz-Llopis M, Vidal-Sanz M, Mullor JL. Metabolomic changes in the rat retina after optic nerve crush. Invest Ophthalmol Vis Sci. 2013; 54:4249-59. [PMID: 23696609]
  36. Harder JM, Guymer C, Wood JPM, Daskalaki E, Chidlow G, Zhang C, Balasubramanian R, Cardozo BH, Foxworth NE, Deering KE, Ouellette TB, Montgomery C, Wheelock CE, Casson RJ, Williams PA, John SWM. Disturbed glucose and pyruvate metabolism in glaucoma with neuroprotection by pyruvate or rapamycin. Proc Natl Acad Sci U S A. 2020; 117:33619-27. [PMID: 33318177]
  37. Xie J, Kittur FS, Li PA, Hung CY. Rethinking the necessity of low glucose intervention for cerebral ischemia/reperfusion injury. Neural Regen Res. 2022; 17:1397-403. [PMID: 34916409]
  38. Meierhans R, Béchir M, Ludwig S, Sommerfeld J, Brandi G, Haberthür C, Stocker R, Stover JF. Brain metabolism is significantly impaired at blood glucose below 6 mM and brain glucose below 1 mM in patients with severe traumatic brain injury. Crit Care. 2010; 14:R13 [PMID: 20141631]
  39. Sucher NJ, Lipton SA, Dreyer EB. Molecular basis of glutamate toxicity in retinal ganglion cells. Vision Res. 1997; 37:3483-93. [PMID: 9425525]
  40. Timofeev I, Carpenter KL, Nortje J, Al-Rawi PG, O’Connell MT, Czosnyka M, Smielewski P, Pickard JD, Menon DK, Kirkpatrick PJ, Gupta AK, Hutchinson PJ. Cerebral extracellular chemistry and outcome following traumatic brain injury: a microdialysis study of 223 patients. Brain. 2011; 134:484-94. [PMID: 21247930]
  41. Kurtz P, Rocha EEM. Nutrition therapy, glucose control, and brain metabolism in traumatic brain injury: A multimodal monitoring approach. Front Neurosci. 2020; 14:190 [PMID: 32265626]
  42. Jee D, Huang S, Kang S, Park S. Polygenetic-Risk Scores for A Glaucoma Risk Interact with Blood Pressure, Glucose Control, and Carbohydrate Intake. Nutrients. 2020; 12:3282 [PMID: 33114701]
  43. Zeleznik OA, Kang JH, Lasky-Su J, Eliassen AH, Frueh L, Clish CB, Rosner BA, Elze T, Hysi P, Khawaja A, Wiggs JL, Pasquale LR, UK Biobank Eye and Vision Consortium. Plasma metabolite profile for primary open-angle glaucoma in three US cohorts and the UK Biobank. Nat Commun. 2023; 14:2860 [PMID: 37208353]
  44. Harun-Or-Rashid M, Pappenhagen N, Zubricky R, Coughlin L, Jassim AH, Inman DM. MCT2 overexpression rescues metabolic vulnerability and protects retinal ganglion cells in two models of glaucoma. Neurobiol Dis. 2020; 141104944 [PMID: 32422282]
  45. Jin S, Chen X, Yang J, Ding J. Lactate dehydrogenase D is a general dehydrogenase for D-2-hydroxyacids and is associated with D-lactic acidosis. Nat Commun. 2023; 14:6638 [PMID: 37863926]
  46. Li K, Tolman N, Segrè AV, Stuart KV, Zeleznik OA, Vallabh NA, Hu K, Zebardast N, Hanyuda A, Raita Y, Montgomery C, Zhang C, Hysi PG, Do R, Khawaja AP, Wiggs JL, Kang JH, John SWM, Pasquale LR, UK Biobank Eye and Vision Consortium. Pyruvate and related energetic metabolites modulate resilience against high genetic risk for glaucoma. Elife. 2025; 14RP105576 [PMID: 40272416]
  47. Akashi A, Miki A, Kanamori A, Nakamura M. Aquaporin 9 expression is required for l-lactate to maintain retinal neuronal survival. Neurosci Lett. 2015; 589:185-90. [PMID: 25637697]
  48. Verma SS, Cooke Bailey JN, Lucas A, Bradford Y, Linneman JG, Hauser MA, Pasquale LR, Peissig PL, Brilliant MH, McCarty CA, Haines JL, Wiggs JL, Vrabec TR, Tromp G, Ritchie MD, eMERGE Network. NEIGHBOR Consortium. eMERGE Network; NEIGHBOR Consortium. Epistatic gene-based interaction analyses for glaucoma in eMERGE and NEIGHBOR Consortium. PLoS Genet. 2016; 12e1006186 [PMID: 27623284]
  49. Hu JK, Wang X, Wang P. Testing gene-gene interactions in genome wide association studies. Genet Epidemiol. 2014; 38:123-34. [PMID: 24431225]
  50. Chattopadhyay A, Lu TP. Gene-gene interaction: the curse of dimensionality. Ann Transl Med. 2019; 7:813 [PMID: 32042829]
  51. de Vries VA, Hanyuda A, Vergroesen JE, Do R, Friedman DS, Kraft P, Turman C, Luo YL, Tran JH, Liefers B, Wong SH, Lee RH, Zebardast N, Klaver CCW, Segrè AV, Pasquale LR, Wiggs JL, Kang JH, Ramdas WD. The Clinical Usefulness of a Glaucoma Polygenic Risk Score in 4 Population-Based European Ancestry Cohorts. Ophthalmology. 2025; 132:228-37. [PMID: 39128550]
  52. Akiyama M, Tamiya G, Fujiwara K, Shiga Y, Yokoyama Y, Hashimoto K, Sato M, Sato K, Narita A, Hashimoto S, Ueda E, Furuta Y, Hata J, Miyake M, Ikeda HO, Suda K, Numa S, Mori Y, Morino K, Murakami Y, Shimokawa S, Nakamura S, Yawata N, Fujisawa K, Yamana S, Mori K, Ikeda Y, Miyata K, Mori K, Ogino K, Koyanagi Y, Kamatani Y, Ninomiya T, Sonoda KH, Nakazawa T, Biobank Japan Project. Japan Glaucoma Society Omics Group. Genomic Research Committee of the Japanese Ophthalmological Society. Genetic Risk Stratification of Primary Open-Angle Glaucoma in Japanese Individuals. Ophthalmology. 2024; 131:1271-80. [PMID: 39023470]