Age-dependent regional retinal nerve fibre changes in SIX1/SIX6 polymorphism

SIX1/SIX6 polymorphism has been shown to be associated with glaucoma. Studies have also found that, in older adults, retinal nerve fibre layer (RNFL) thickness is significantly thinned with each copy of the risk allele in SIX1/SIX6. However, it is not known whether these genetic variants exert their effects in younger individuals. Comparing a healthy young adult with an older adult cohort (mean age 20 vs 63 years), both of Northern European descent, we found that there was no significant RNFL thinning in each copy of the risk alleles in SIX1/SIX6 in the eyes of younger individuals. The older cohort showed an unexpectedly thicker RNFL in the nasal sector with each copy of the risk allele for both the SIX1 (rs10483727) and SIX6 (rs33912345) variants. In the temporal sector, thinner RNFL was found with each copy of the risk allele in rs33912345 with a decrease trend observed in rs10483727. Our results suggest that SIX1/SIX6 gene variants exert their influence later in adult life.

www.nature.com/scientificreports/ in the Japanese study 22 . In rs10483727, an European cohort study reported significant global RNFL thinning with each copy of the risk allele as well as the inferior and superior sectors 23 but the study cohort contained a large portion of glaucoma cases and suspects (~ 80%).
To date, the association between RNFL thickness and SIX1/SIX6 variants have only been reported in older individuals (mean age > 50 years) [20][21][22][23] . Hence it is not known whether these genetic variants exert their effects in younger individuals. In this study, we investigate the effect of age on SIX1/SIX6 polymorphism by examining the RNFL thickness in two large adult cohorts of similar descent but with substantial age difference.

Results
A total of 2,541 individuals; (943 from younger cohort, the Raine Study Gen2 (RS GEN2) and 1598 from the older cohort, TwinsUK adult twin Registry (TwinsUK)) for whom both RNFL and genotypic data were available were included in the analysis. Demographic, genotypic and ocular characteristics are presented in Table 1. In the younger RS GEN2 cohort, the mean age was 20.0 years (range: 18.3-22.1) and 488 participants (52%) were female. In the older TwinsUK cohort, the average age was 62.6 years (range: 21.7-89.7) and the proportion of female participants (91%) was much greater than in the RS GEN2 (52%) cohort. While there were no glaucoma suspects or cases in the RS GEN2, the prevalence of self-report glaucoma from questionnaire data in TwinsUK was 1.5%. The SIX6 variant rs33912345 had C allele frequency of 39.4% in the RS GEN2 cohort. It was in strong linkage disequilibrium with the previously identified variant rs10483727 (R 2 = 0.98). The genotype frequencies of SIX6 rs33912345 among the 943 participants were 338 (35.8%) for genotype A/A, 466 (49.4%) for A/C, and 139 (14.7%) for C/C. A similar distribution of genotype frequencies was observed in the TwinsUK older cohort (Table 1). In RS GEN2, for rs10483727, the genotype frequency for C/C, C/T and T/T were 334 (35.4%), 478 (50.7%) and 131 (13.9%), respectively. The genotype distribution for both rs10483727 and rs33912345 in the TwinsUK cohort were similar to that of the RS GEN2 cohort.
Linear regression showed that, in rs33912345, the younger RS GEN2 cohort showed non-significant decrease in global and all four RNFL sectors with each copy of the C risk allele (Table 3). However, the temporal (β = − 0.73, p = 0.04) quadrant was significantly thinner with the C risk allele in the older TwinsUK cohort. In agreement with the genotype data presented in Table 2, each copy of the risk allele was associated with increase in RNFL thickness (β = + 1.02, p = 0.03). Similarly, the younger cohort showed no significant thinning with each copy of the T risk allele in rs10483727 (Table 4). In the older cohort, the nasal sector showed significant thicker RNFL

Discussion
Large GWA studies allow the identification of genetic variants that are associated with common, complex diseases of old age. However, in the case of POAG and SIX1/SIX6, it is not known when these genetic variants start exerting their effects 24 . Studies have reported that individuals with risk allele in SIX1/SIX6 are more susceptible to glaucoma 1,3,25 and have thinner RNFL [20][21][22][23] . Note that these RNFL studies were conducted in older individuals, hence it is not clear whether thinner RNFL found in the older individuals with SIX1/SIX6 variants is due to these individuals being born with thinner RNFLs or due to a faster rate of RNFL degeneration as they age. Our data suggests the latter scenario, as there was no significant decrease in RNFL thickness for both rs33912345 and rs10483727 variants in the younger cohort but we found evidence of sectorial RNFL thinning in rs33912345 in  We found, in our older cohort, each copy of the rs10483727 risk allele was associated with significantly thicker nasal RNFL. Temporal RNFL was trending thinner, albeit not significantly, with each copy of the risk allele. This finding differs to a previous study in individuals of European descent (n = 231, mean age 65.1 years), which reported that each copy of the rs10483727 risk allele was associated with significant thinning of the inferior and superior but not nasal or temporal sectors 23 . The study cohort was relatively small and skewed towards subjects with glaucoma (46 healthy individuals, 101 glaucoma suspects, 84 clinically diagnosed glaucoma), which differs to the current study. The authors did not report the proportion of T-risk in their cohort. However, given the high proportion of glaucoma cases (suspect and confirmed) and that these glaucoma cases present with a higher proportion of T risk allele, it can be assumed that the study population contains higher T risk allele proportion than our TwinsUK cohort. In addition, the RNFL thickness in the aforementioned study is thinner than our cohort for all three genotypes, most likely also due to the higher proportion of glaucoma cases. Hence these factors could combine to result in different findings between the two studies. Nevertheless, this does not explain our atypical finding of thicker nasal RNFL with each copy of the rs10483727 risk allele in the TwinsUK cohort, which we have no satisfactory explanation for but it may be partly attributed to the higher variability in nasal RNFL measurement 26 . In rs33912345, we found significantly thinner RNFL in the temporal but thicker RNFL in the nasal sector with each copy of the risk allele in the older TwinsUK cohort, which again differs from previous reports. In a Japanese population 22 (n = 2,306, mean age 57.6 years), significant thinning was found between each copy of the rs33912345 risk allele in both inferior and temporal RNFL sectors. In contrast, a Singaporean-based study 21 (n = 1,243, mean age 55.0 years) showed significant thinning with rs33912345 risk allele in both superior and inferior sectors. One possible explanation of the difference in sectoral thinning between the different studies could be the difference in allele distribution. In support of our postulation, the rs33912345 genotype distribution of the Singaporean study (A/A 4.0%, A/C 32.1%, C/C 63.9%) is vastly different when compared to the TwinsUK cohort (A/A 34.1%, A/C 48.4%, C/C 17.5%). The higher prevalence of the C risk allele in the Singaporean population may explain why global RNFL thinning was observed but not in the TwinsUK cohort. Allele distribution was not provided in the Japanese study. Further investigation is required to confirm whether difference in genotype distribution can result in difference in sectorial RNFL thinning.
A key strength of this study is the inclusion of two large adult cohorts with similar ancestry but with a clear age gap between the two groups. The extensive ocular and genetic data collected in both groups allowed us to perform similar analyses, which enabled direct comparison between the two populations. Furthermore, normative RNFL thickness values are reported in two large populations of different ages, providing reference value for future studies. The main limitation of this study is the cross-sectional design. A longitudinal study will allow us to define the age range when RNFL thickness in adults with risk alleles deviates from those with major alleles, which could shed light on glaucoma disease mechanism and advise on age for glaucoma screening.
In contrast to evidence from the older populations, we found no effects of the SIX1/SIX6 gene variants rs33912345 and rs10483727 on peripapillary RNFL thickness of young adults. However, we found evidence for sectorial thinning in the older cohort with each copy of the risk allele in an older cohort with similar ancestry, albeit with different RNFL profile compared to previous studies in other ethnicities. Our study suggests that SIX1/SIX6 gene variants exert their influence later in adult life. Further research should concentrate on validating these results as well as identifying functional pathways involving SIX1/SIX6 gene and other common POAG risk factors. Table 4. Association of rs10483727 with peripapillary retinal nerve fibre layer thickness. a β, changes in RNFL thickness (in µm) per copy of the risk allele. b Model 1: RNFL = α + age + sex + spherical equivalent + principal components + rs10483727_T. c Model 2: RNFL = α + age + sex + spherical equivalent + population structure + relatedness + rs10483727_T.

Methods
The Raine Study Gen2. The Raine Study is a multigenerational, longitudinal study conducted in Perth, Western Australia 27 . At the 20-year follow-up of the cohort, 1,344 participants were enrolled in RS GEN2, a cross-sectional study of eye diseases in young adults 28 . All participants underwent a comprehensive standardised ocular examination. This included optical coherence tomography (OCT) measurements of the peripapillary RNFL by Spectralis (Heidelberg Engineering GmbH, Heidelberg, Germany). In each eye, a circular scan (diameter 3.5 mm) centred on the optic nerve was acquired by averaging 16 consecutive frames in real time. RNFL thickness was determined in four quadrants (superior, inferior, temporal, nasal) around the optic disc automatically using a Heidelberg Eye Explorer (HEYEX) software version 6.7.12.0 as well as global mean RNFL thickness. Intraocular pressure (IOP) with an ICare TAO1i Tonometer (Icare Finland, Oy, Helsinki, Finland), central corneal thickness (CCT) with an Oculus Pentacam (Optikgerate GmbH, Wetzlar, Germany) and axial length using noncontact partial coherence interferometry (IOL Master V.5; Carl Zeiss Meditec AG, Jena, Germany) were also acquired at the same visit. Genotyping was performed on 1593 participants, including those who did not attend the vision assessments, using the Illumina Human 660W-quad BeadChip (Illumina, Inc., San Diego, CA, USA), which included the SIX6 missense variant rs39912345 and SIX1/SIX6 variant rs10483727. After standard quality control, the cleaned genotypic datasets included 1,494 individuals. Details of the quality control step have been described previously 29 .
Only subjects with both RNFL and genotyping data as well as of Northern European descent are included in the analysis.
TwinsUK cohort study. The TwinsUK Adult Twin Registry is based at St Thomas' Hospital, London. Participants were unaware of any hypotheses or proposals for specific studies; only later were they invited to have an eye examination. The St Thomas' Hospital Local Research Ethics Committee approved the study, and all the twin participants volunteered to join the TwinsUK Registry and gave informed consent to attend the hospital for phenotyping and for their data to be used for scientific research. Spherical equivalent was calculated from autorefraction data (ARM-10, Tagaki Ltd, Takaoka, Japan). RNFL was assessed with the OCT (Optovue iVue sd-OCT, Fremont, CA, USA) using a circular scan (4.9 mm diameter) centred at the optic nerve head; both global and quadrant thicknesses were determined by the Optovue algorithm 30 .
Genotyping of the TwinsUK cohort was done with a combination of Illumina HumanHap300 and Human-Hap610Q chips. Intensity data for each of the arrays were pooled separately and genotypes were called with the Illuminus32 calling algorithm, thresholding on a maximum posterior probability of 0.95. Imputation was performed using the IMPUTE 2.0 31 software package using haplotype information from the Haplotype Reference Consortium panel, pre-phased using ShapeIT 32 . Analyses were done using GEMMA 33 , which implements a mixed model that adjusts for similarity between individuals.
Only subjects with both RNFL and genotyping data as well as of Northern European descent are included in the analysis.

Statistical analysis.
Average of right and left eye values were included in the analysis for both studies, regardless of glaucoma status. In both studies, potential covariates affecting both the global-and sectorial-specific RNFL measures were evaluated using scatterplots and a correlation matrix (data not shown). An additive genetic model was constructed to determine the association between rs33912345 and RNFL thickness as a quantitative trait whereby the number of minor allele copies was coded into 0, 1 and 2 (the C allele of rs39912345) and used in a linear regression model adjusted for age, sex and significant predictors of RNFL measures. In the case of the TwinsUK cohort, relatedness was also taken into account. RNFL measurements were compared between genotype groups using a trend test. Logistic regression analyses were also performed with case/control status as an outcome (dependent variable) and rs33912345 genotype as the predictor variable. The same additive genetic modelling was performed for rs10483727 in the RS GEN2 and TwinsUK data.
The statistical analyses were performed using the statistical software R version 3.6.2 (R Foundation for Statistical Computing, https ://www.r-proje ct.org/ [in the public domain]) in the RS GEN2 and STATA14 statistical package (www.stata .com) in TwinsUK. A p value < 0.05 was considered statistically significant.
Ethics statement. This study utilised data from the Raine Study Gen2 (RS GEN2) and the TwinsUK adult twin Registry (TwinsUK). Both studies were conducted in accordance with the tenets of the Declaration of Helsinki. Individual study protocols were approved by the University of Western Australia Human Research Ethics Committee and the Guys and St. Thomas' ethics committee for the RS GEN2 and TwinsUK, respectively. Informed consent was obtained from all participants prior to the ocular examination session.