Single nucleotide polymorphisms in LOXL1 as biomarkers for progression of exfoliation glaucoma in Sweden

Exfoliation glaucoma is a common and aggressive type of glaucoma with high prevalence in Scandinavia. The aim of this study was to elucidate whether the allele frequencies of two single nucleotide polymorphisms (SNPs) located in LOXL1 were associated with the progression of exfoliation glaucoma in Swedish patients.

outflow, thus increasing intraocular pressure (IOP). Exfoliation glaucoma is an aggressive form of glaucoma, with rapid visual field deterioration, which is approximately 3-4 folds faster in exfoliation glaucoma than in primary open glaucoma (Ayala, 2020).
Although the intrinsic causes of exfoliation glaucoma have not been revealed, genetic mechanisms are known to play a role. LOXL1 is the most critical gene related to exfoliation glaucoma development (Aung et al., 2018;Liu & Allingham, 2011;Pasutto et al., 2017;Wiggs & Pasquale, 2017). Single nucleotide polymorphisms (SNPs) are variations in the human DNA that influence gene expression and can explain individual variations of the disease. SNPs have been associated with the development of different conditions and interindividual differences in the effect of drugs. Different SNPs located in LOXL1 are associated with an increased risk of developing exfoliation glaucoma (Eivers et al., 2020;Jaimes et al., 2012;Li et al., 2021;Wang et al., 2016).
As a progressive disease without a cure, the progression of glaucoma should be determined. The only way to slow down glaucoma progression is to reduce IOP. Glaucoma progresses at different rates among individuals. Since glaucoma leads to deterioration of the visual field, the best way to determine its progression is through repeated visual field tests. The European Glaucoma Society (EGS) recommends 5-6 visual fields the first 2 years after diagnosis ('European Glaucoma Society Terminology and Guidelines for Glaucoma, 5th Edition', 2021). The Swedish guidelines postulate 5-6 visual fields 3 years after glaucoma diagnosis (Heijl et al., 2012). This approach was judged to be more compatible with resource allocation to patients with glaucoma in the Swedish healthcare system.
The aim of this study was to examine the association between previously highlighted LOXL1 SNPs (i.e., LOXL1_rs2165241 and LOXL1_rs1048661) and progression of exfoliation glaucoma in a Swedish population. Due to the low frequency of LOXL1_rs3825942 SNP in patients with glaucoma and a limited sample size, it was not included in the present study.

| Patients and examinations
In this prospective non-randomised cohort study, we enrolled patients with exfoliation glaucoma at the Ophthalmology Department of the Skaraborg's Hospital, Skövde, and Sahlgrenska University Hospital, Gothenburg, from 1st January 2014 to 31st December 2017. All patients were followed up for 3 years ± 3 months. Informed consent was obtained from all patients. The study protocol was granted ethical approval by the University of Gothenburg (DN:119-12). The study was performed in accordance with the tenets of the Declaration of Helsinki.
At the recruiting visit, an ophthalmic nurse checked the visual acuity of patients with Snellen's chart and performed a visual field test. Humphrey field analysis (Carl Zeiss, Carl-Zeiss-Straße 22, 73447 Oberkochen, Germany) was performed using the software threshold 24-2. Subsequently, an ophthalmologist measured IOP with a Goldmann applanation tonometer and performed slit-lamp biomicroscopy, including gonioscopy. Pupils were then dilated with 2.5% phenylephrine and 0.5% tropicamide (Bausch & Lomb UK Ltd., 106 London Road-Kingston-upon-Thames-Surrey-KT2 6TN-England). After 20 min, the presence of exfoliation was confirmed, and the optic nerve was assessed using a 90-D lens. Subsequently, the central corneal thickness (CCT) was measured using an ultrasound device (Tomey Pachymetry; Tomey Corp, Nagoya 451-0051, Japan). The average value of seven measurements was automatically calculated. At the end of the visit, blood samples were collected. The number of medicines was registered as the number of compounds and not the number of bottles used.
Inclusion criteria were (1) a diagnosis of exfoliation glaucoma based on the criteria established by the EGS ('European Glaucoma Society Terminology and Guidelines for Glaucoma, 5th Edition', 2021), that is an untreated IOP of ≥21 mmHg, an open anterior chamber angle, glaucomatous visual field defects (at least two repeatable Humphrey 24-2 tests), glaucomatous optic nerve damage and the presence of exfoliation material; (2) at least five reliable visual field tests during the 3-year follow-up, with reliability defined as false positives ≤15%, false negatives ≤20% and/or fixation losses ≤30%; and (3) age ≤ 85 years at the recruiting visit.
Exclusion criteria were (1) a diagnosis of advanced glaucoma defined as mean deviation (MD) ≥18 dB and/or visual field index (VFI) ≤40% because of 'floor effects', in which further loss of visual field defects can no longer be detected (Nguyen et al., 2019;Wall et al., 2009); (2) a history of glaucoma surgery other than uneventful cataract surgery or selective laser trabeculoplasty (SLT); and (3) other eye diseases (central venous occlusion, retinal detachment, etc.) that could affect the visual fields during the 3-year follow-up.

| Visual field progression assessment
The visual field progression was studied with three methods. The first method was based on MD visual field. The difference in MD values from the beginning to the end of the study was calculated. Higher values indicated higher progression. The MD values were chosen because several studies use MD as an indicator of progression (Berchuck et al., 2019;Liebmann et al., 2017;Salonikiou et al., 2018). However, cataract development can also modify MD values.
The second method was based on VFI. A device calculated the VFI and performed a regression analysis to calculate the rate of progression (ROP). The machine calculated the ROP as the amount of VFI deterioration (%)/year. The ROP calculation is also referred to as a 'trend analysis'.
The third method was the guided progression analysis (GPA), which is also included in the device and performed automatically (GPA Alert) but differs from ROP. GPA is an 'event analysis', while ROP is a 'trend analysis'. The machine compares every single point before examinations. The GPA alert result options for progression are 'no', 'possible' and 'likely'. We evaluated glaucoma as 'no progression' or 'progression', the latter including both 'possible' and 'likely'.

| Genotyping
All included patients underwent venipuncture for blood sample extraction. Two ethylenediamine tetra acetic acid tubes were obtained and stored in a −70°C freezer until analysis. DNA extraction from blood samples was performed using standard procedures. Genetic analysis was performed at LGC Genomics (Hoddesdon, Herts, UK) using competitive allele-specific PCR (KASP) genotyping (How does KASP work | LGC Biosearch Technologies). The success rate of the samples was over 95% for all genotyped SNPs, and the SNPs were in Hardy-Weinberg equilibrium (based on a p-value threshold of 0.05).

| Statistical analysis
All statistical analyses were performed using SPSS version 28.0.1.0 (IBM, 1 New Orchard Road Armonk, NY 10504, USA) software. Clinical characteristics of the cohort were compared in relation to progression/no progression of glaucoma based on GPA. Categorical variables were compared with Fisher's exact test while continuous variables with Student's t-test. The SNP allele distribution between no progression and progression cases based on GPA was compared using Fisher's exact test. The distribution of the haplotypes between no progression and progression cases based on GPA was compared using Chi2 analyses within the Haploview 4.1 software (Barrett et al., 2005). The relation between the SNPs and progression/no progression in glaucoma was also tested using regression analyses (logistic or linear). In these analyses, MD and ROP, in addition to GPA, were used as outcome variables measuring progression/no progression of glaucoma. Three genetic models were used: additive, dominant, and recessive (Table 1). Regression analyses were performed both in unadjusted and adjusted models. Potential covariates were first tested for association with the outcome variable (i.e., GPA, MD and ROP) using univariate linear or logistic regression (depending on the character of the outcome variable). Associated variables were then included as covariates in multivariate regression analyses. Finally, a regression analysis was performed, including the genotypes of the two SNPs as predictors (both SNPs within the same model) and GPA/MD/ROP as outcome variables.
A power analysis based on Fisher's exact test was performed using the Power and Sample Size Calculations software (http://biost at.mc.vande rbilt.edu/Power Sampl e Size). Each SNP was analysed separately. In all cases, the type I error (α error) was set to 0.05.

| R E SU LT S
We enrolled a total of 130 patients in the present study. Most patients (n = 125) were recruited at the Department of Ophthalmology of Skaraborg's Hospital, Skövde. The average age at the recruiting appointment was 71.4 ± 6.29 years. There were 64 (49%) men and 66 (51%) women. The average IOP values at the glaucoma diagnosis and recruiting visit were 32.9 ± 6.51 and 17.6 ± 2.84 mmHg, respectively. There were 102 (78%) phakic and 28 (22%) pseudophakic patients. Exfoliation glaucoma was unilateral and bilateral in 82 (63%) and 48 (37%) patients, respectively. The mean CCT was 542.73 ± 34.04 μm. The average time interval between the glaucoma diagnosis and the recruitment appointment was 2.69 ± 1.13 years. The average number of medicines at the recruiting visit was 1.75 ± 0.72. At the beginning of the study, the average MD was −6.88 ± 4.94 dB, and the average VFI was 83.76% ± 15.32%.
The clinical characteristics of the patients were also compared according to their GPA ( Table 2). The distribution of allele frequencies based on GPA differed significantly for rs2165241 and rs1048661. In the case of LOXL1_rs2165241, the frequency of the C allele was 13% in no progression patients, while it was 38% in progression patients (Chi-square; p = 8 × 10 −7 ). In the case of LOXL1_rs1048661, the frequency of the T allele was 9% in no progression patients, while it was 32% in progression patients (Chi-square; p = 2 × 10 −6 ) ( Table 3).
Linkage disequilibrium (LD) between LOXL-1 rs_2165241 and LOXL1 rs_1048661 was r 2 = 0.8 (strong) (based on calculations using the Haploview 4.1 software). The haplotype distribution between no progression and progression patients differed significantly for T A B L E 1 Coding of the different genotypes in the regression analyses.

Genotypes Additive Dominant Recessive
Non-risk homozygote 0 0 0 Heterozygote 1 1 0 Risk homozygote 2 1 1 two of the three haplotypes. For the GT haplotype, the frequency among progression patients was 67%, while it was 85% among no progression patients (Chi-square; p = 9 × 10 −4 ). For the TC haplotype, the frequency was 28% among progression patients but only 12% among no progression patients (Chi-square; p = 0.002). For the GC haplotype, there was no significant difference in the frequency among progression (5%) and no progression (3%) patients (Chi-square; p = 0.42) ( Table 4). The relationship of LOXL1_rs2165241 and LOXL1_ rs1048661 with GPA was significant for the additive and dominant models, both in the unadjusted and adjusted analyses. In the case of LOXL1_rs2165241 and LOXL1_ rs1048661, the OR was approximately 3 in the additive and dominant unadjusted model. However, the OR increased to approximately 6 for both SNPs when the model was adjusted for IOP at diagnosis, SLT treated/ untreated during 3 years' follow-up, MD, VFI and the number of medicines at recruitment ( Table 5).
The association of LOXL1_rs2165241 and LOXL1_ rs1048661 with MD was significant for the additive and recessive models, both in the unadjusted and adjusted analyses. Similar results were found for the association of LOXL1_rs2165241 and LOXL1_rs1048661 with ROP. The 'β Coefficient' values were approximately 1 in the unadjusted additive model, while they were approximately 3 in the unadjusted recessive model. The values showed a moderate increase when covariates (Unilateral/bilateral glaucoma, SLT treated/untreated during 3 years' follow-up, MD and the number of medicines at recruitment) were added to the models. (Tables 6 and 7).
Regression analyses using both SNPs as predictors and GPA/MD/ROP as the outcome variable showed no significant associations. In the case of the logistic T A B L E 2 Clinical characteristics of the patients according to progress/no progress (GPA).

Variables
No T A B L E 3 Distribution of minor allele frequencies of SNPs in LOXL1 according to no progression/progression (GPA).  as the outcome. The significant variables were IOP at diagnosis (p = 0.02), SLT during the 3-year follow-up (p = 0.01), MD (p < 0.001), VFI (p = 0.02), and the number of medicines at recruitment (p < 0.001; Table S1). A univariate linear regression analysis was used to test possible covariates in relation to the MD difference and ROP (continuous variables) as the outcomes. The significant variables in relation to the MD difference were unilateral/bilateral glaucoma (p = 0.04), SLT during the 3-year follow-up (p < 0.0001), MD (p = 0.02) and the number of medicines at recruitment (p = 0.004). In relation to ROP, the significant variables were IOP at diagnosis (p = 0.04), SLT during the 3-year follow-up (p = 0.01), MD (p < 0.001), VFI (p = 0.004), and the number of medicines at recruitment (p = 0.005; Table S2).

Rs-ID
Regarding power calculation, based on the GPA analysis, for LOXL-rs2165241, with a sample size of 53 patients with no progression and 77 patients with progression and allele frequencies of 13.2% and 38.31%, respectively, the α error was 0.05, and the power (1β) was 87%. For LOXL-rs1048661, with the same sample size of no progression and progression and allele frequencies of 9.43% and 32.46%, respectively, the α error was 0.05, and the power was 85%.

| DI SC US SION
The present study showed an association of the SNPs LOXL1-rs2165241 and LOXL1-rs1048661 with exfoliation glaucoma progression, using the additive model of GPA, MD, and ROP. SNPs are markers for disease susceptibility (Kido et al., 2018). How the investigated SNPs can alter the function of LOXL1 is still unknown. Using an animal model, Sharma et al. reported that the SNPs LOXL1-rs1048661 and rs3825942 might alter protein function and binding of the LOXL1 protein surface with consequent alterations in protein-protein interactions. Changes in these residues might change the role of LOXL1 (Sharma et al., 2016). Genetic mechanisms underlying exfoliation and exfoliation glaucoma have been described before. According to the literature, the SNPs included in this study showed a strong association with exfoliation syndrome and/or exfoliation glaucoma (Eivers et al., 2020;Jaimes et al., 2012;Li et al., 2021;Wang et al., 2016). Other SNPs located in LOXL1 can probably alter the gene function. The intrinsic mechanisms between genetic changes and exfoliation glaucoma have not been elucidated (Schlotzer-Schrehardt, 2011). Increased production of protein material could occur in patients because of genetic alterations, thus increasing IOP and visual field damage. In our study, some patients developed progression despite an IOP reduction of approximately 60%-70% from IOP at diagnosis. Glaucoma progression in the whole cohort showed significant variations among individuals (ROP at the 3-year follow-up = 2.58%/year ±2.61%/year). Genetic mechanisms could induce glaucoma progression apart from IOP. Due to genetic mechanisms, apoptosis in the ganglion cells could be directly or indirectly induced.
The association of SNPs LOXL1_rs2165241 and LOXL1_rs1048661 with exfoliation/exfoliation glaucoma has been extensively described in Scandinavia and worldwide. Thorleifsson et al. published the first article in 2007(Thorleifsson et al., 2007. Their study included controls from Iceland and cases of glaucoma from Iceland and Sweden. Even patients in Finland have been studied, showing similar results as the Icelandic study (Lemmelä et al., 2009).
Exfoliation syndrome and glaucoma have been described worldwide. However, the prevalence of exfoliation varies among countries. For example, the prevalence is high in Scandinavia but low in China. Åström et al. reported a prevalence of 23% in the Swedish population at 63 years of age while a low prevalence of 0.55% in the Chinese population at 60 years or older Ren et al., 2015). Previous genetic studies have reported noticeable variability in the disease associations of many LOXL1 variants. The risk of developing glaucoma associated with specific allele frequencies differs with ethnicity. For rs2165241, the T allele is a risk factor in the European populations (Anastasopoulos et al., 2014;de Juan-Marcos et al., 2016;Lemmelä et al., 2009;Thorleifsson et al., 2007), while the C allele is a risk factor for the development of exfoliation in Asian populations (Wang et al., 2016). No previous article using the same design as our study was found in the literature. The closest was a study involving Turkish patients (n = 48) with exfoliation glaucoma. The study showed no association of LOXL1_rs1048661 or LOXL1_rs41435250 with severe disease. That was a cross-sectional study on the severity, and not progression, of the disease. Moreover, the severity of glaucoma was estimated as morphological changes in the optic nerve and not as visual field damage (Yilmaz et al., 2016).
The present study showed a strong LD (r 2 = 0.8) between LOXL1_rs2165241 and LOXL1_rs1048661. Linkage disequilibrium between SNPs varies among different populations studied. The reasons for variation are multiple. Common reasons mentioned for LD are recombination, genetic diversity and location of the SNPs (Conrad et al., 2006). Based on a population of European residents in Utah and using the LD-link software from the National Institute of Health (NIH) (Machiela & Chanock, 2018), the LD between LOXL1_ rs2165241 and LOXL1_rs1048661 showed to be moderate (r 2 = 0.4). The difference between this value (r 2 = 0.4) and the one seen in our study (r 2 = 0.8) is relatively big. The only information found on a comparable population is the one present in Thorleifsson et al. (2007). The authors described an LD between LOXL1_rs2165241 and LOXL1_rs1048661 of r 2 = 0.46 among Icelandic individuals. All included subjects were controls and not glaucoma patients. Patients included in the present study were all born in Sweden (apart from three born in Finland). No healthy individuals were included in the present study; all patients suffered from the same disease, exfoliation glaucoma. One possible explanation for the divergent LD seen in our study could be the relatively homogenous and highly selected sample. The strong LD in our sample raises the question if it was necessary to study both SNPs, or if it had been enough just to include one of them. A previous study showed that the relationship between LD and association was not linear. An LD of r 2 = 0.8 could explain an association effect of 60% (Nielsen et al., 2008). The authors of that study concluded that high LD between markers does not necessarily imply that the association test result of one marker is a direct substitute of the result for the other marker; eliminating one of these markers in an association study will lead to a reduction in overall power. Still, regression analyses including both SNPs in our study within the same model, showed no associations with the outcomes (i.e., GPA, MD and ROP), indicating that the two SNPs have no independent effects (i.e. they give rise to the same genetic signal). Moreover, haplotype analyses showed that none of the three main haplotypes, constructed based on allele combinations of the two SNPs, was more strongly associated with the outcomes than each SNP separately, further indicating that the two SNPs do not contribute with different effects.
The present study used three different progression models to define progression variables: MD, GPA and ROP. Unfortunately, there is still no consensus on the best model. The MD model, the oldest one to follow glaucoma progression, is affected by cataract development and surgery. However, Hoddap's classification of glaucoma, still in use, is based on MD (Hodapp et al., 1993).
The progression variables (MD, GPA and ROP) were tested, as outcome variables, in three genetic models: additive, dominant and recessive. Associations were found in all analyses using the additive model. However, the results for the additive model, including MD, were not as significant as those for the models including GPA and ROP. As pointed out above, MD is probably not the best method to assess glaucoma progression because several other causes can change MD.
Regarding logistic regression analyses involving GPA, both the unadjusted and adjusted models showed a significant association between the allele status and disease progression. For LOXL1_rs2165241 SNP, both the additive and dominant models showed an odds ratio of approximately 6 in the adjusted model. This means that the odds of disease progression increased approximately six folds per C allele (in the additive model) or when the C allele was present (in the dominant model). For LOXL1_rs1048661, the results were similar (i.e., the odds for progressive disease increased approximately 6-7 folds with the T allele). However, considering the relatively small sample size, these results must be interpreted with caution. For the models including ROP and MD, similar results in relation to GPA were found. However, in the cases of ROP and MD, the additive and recessive models showed significant results (compared to the additive and dominant models for GPA). The B-coefficient in the recessive model including ROP was approximately 3 for LOXL1_rs2165241 and LOXL1_rs1048661. This means that disease progression increased approximately 3 dB in the model including MD and 3%/year VFI in the model including ROP, when the C allele was present for LOXL1_rs2165241 and T allele for LOXL1_rs104866. In relative terms, it means a disease worsening with 10% in the MD values and 9% in the VFI values. Again, these results must be interpreted cautiously, considering the relatively small sample size.
Regarding clinical characteristics of the cohort, the participants' average age was consistent with previous studies Heijl et al., 2009). We wanted to evaluate visual field defects; therefore, patients >85 years were excluded because they could have difficulty performing visual field tests. Exfoliation glaucoma was presented more frequently as unilateral than bilateral, consistent with previous studies (Tarkkanen & Kivela, 2004).
The patients were recruited close to establishing their glaucoma diagnosis. This factor and high IOP at the diagnosis could explain the high number of patients identified with progression in the GPA analysis. With time, the disease could slow, and visual field defect progression would diminish with decreasing IOP. However, ROP at the 3-year follow-up showed that the average progression was relatively high. This is a limitation because our results are only applicable to patients with highly progressive glaucoma. On the other hand, this is a strength because the higher progression rate makes the difference between patients even more apparent, reducing the effects of a reduced sample size.
The study has certain limitations. According to Hodapp's glaucoma classification, MD of <−6, −6 to −12, and −12 dB is early, moderate and advanced glaucoma, respectively (Hodapp et al., 1993). Since all patients included in our study had early and moderate glaucoma, our results probably do not apply to patients with advanced glaucoma. As assessed by visual field defects, advanced glaucoma damage exhibit 'floor effects' prohibiting evaluation of further progression (Nguyen et al., 2019;Wall et al., 2009). Moreover, the study involved patients with parents born in Scandinavia. Since LOXL1 allele frequencies vary among countries (Li et al., 2021), our results cannot be generalised to other ethnic groups. Another limitation is the small sample size. However, the cohort's well-defined phenotype (exfoliation glaucoma) diminished bias probabilities. Furthermore, the power analysis showed sufficient power to render reliable results. No similar studies have comparable sample size as the present study. Moreover, the present study did not include morphological parameters for progression, such as optic disc measurements on optical coherence tomography (OCT). At the commencement of the study, OCT was not commonly performed for the clinical diagnosis of glaucoma. The gold standard modality for establishing glaucoma progression is repeated visual field testing ("European Glaucoma Society Terminology and Guidelines for Glaucoma, 5th Edition," 2021). Finally, OCT measurements also cause 'floor effects' in patients with advanced glaucoma, thus contributing no additional advantage (Abu et al., 2021).
In conclusion, the present study showed an association of LOXL1_rs2165241 and LOXL1_rs1048661 with visual field defect progression in Swedish patients with exfoliation glaucoma. Carrying the C allele of LOXL1_ rs2165241 or the T allele of LOXL1_rs1048661 increased the risk of developing a progressive form of exfoliation glaucoma in Swedish patients. The two SNPs seem, however, to be highly correlated with each other (i.e., in high LD). Further evidence from a larger sample is required to confirm the results.

AC K NO W L E DGE M E N T S
The study was financed by grants from the Swedish state under the ALF agreement between the Swedish government and the county councils (ALF-GBG-145921 and ALFGBG-966230), Konung Gustav and Drottning Victorias Frimurarstiftelse, Agneta Prytz-Folkes och Gösta Folkes stiftelse, Kronprinsessan Margaretas Arbetsnämnd för Synskadade and De Blindas Vänner. The sponsors and funding organisations had no role in the design or conduct of this research. None of the authors has a financial or proprietary interest in any material or method mentioned.

CON F L IC T OF I N T E R E ST
Nothing to declare.