Common Risk Variants in AHI1 Are Associated With Childhood Steroid Sensitive Nephrotic Syndrome

Introduction Steroid-sensitive nephrotic syndrome (SSNS) is the most common form of kidney disease in children worldwide. Genome-wide association studies (GWAS) have demonstrated the association of SSNS with genetic variation at HLA-DQ/DR and have identified several non-HLA loci that aid in further understanding of disease pathophysiology. We sought to identify additional genetic loci associated with SSNS in children of Sri Lankan and European ancestry. Methods We conducted a GWAS in a cohort of Sri Lankan individuals comprising 420 pediatric patients with SSNS and 2339 genetic ancestry matched controls obtained from the UK Biobank. We then performed a transethnic meta-analysis with a previously reported European cohort of 422 pediatric patients and 5642 controls. Results Our GWAS confirmed the previously reported association of SSNS with HLA-DR/DQ (rs9271602, P = 1.12 × 10−27, odds ratio [OR] = 2.75). Transethnic meta-analysis replicated these findings and identified a novel association at AHI1 (rs2746432, P = 2.79 × 10−8, OR = 1.37), which was also replicated in an independent South Asian cohort. AHI1 is implicated in ciliary protein transport and immune dysregulation, with rare variation in this gene contributing to Joubert syndrome type 3. Conclusions Common variation in AHI1 confers risk of the development of SSNS in both Sri Lankan and European populations. The association with common variation in AHI1 further supports the role of immune dysregulation in the pathogenesis of SSNS and demonstrates that variation across the allele frequency spectrum in a gene can contribute to disparate monogenic and polygenic diseases.

to disease. First, SSNS is defined by response to initial treatment with corticosteroid therapy and in patients that develop a relapsing course of disease, SSNS also responds to additional immunosuppressive medications. 8 Second, the onset of disease is typically associated with a preceding infection, suggesting that prior activation of the immune system may trigger the disease. 9 Third, antibodies directed toward nephrin, a protein in the slit diaphragm in the glomerulus, have recently been identified in patients with SSNS. 10 These clinical observations suggest that SSNS is an autoimmune disorder, implicating both genetic and environmental factors contributing to development of the disease.
GWAS have been instrumental in elucidating genetic risk factors for developing SSNS in childhood. The HLA-DR/DQ region has exhibited the strongest association with disease in European, South Asian, and Japanese populations, [3][4][5][6]11 supporting the inference from clinical observations that SSNS has an immunologic basis. Beyond the HLA region, genome-wide associations at CALHM6 and PARM1 have been identified in European children, 3 and at NPHS1 and TNFSF15 in Japanese children. 6 In the latter study, the NPHS1 and TNFSF15 loci were not replicated in a European population, suggesting that SSNS possesses different genetic architecture outside of HLA in these 2 different groups. We set out to perform a GWAS in a Sri Lankan population, followed by a European-Sri Lankan transethnic meta-analysis, to identify additional genetic loci associated with SSNS to aid in further understanding of the pathophysiology of disease.

METHODS
Abbreviated methods follow. Detailed methods may be found in the Supplementary Material.

Study Populations
Sri Lankan patients diagnosed with childhood SSNS (age of onset <18 years) were recruited into the study. Most patients were of self-reported Sri Lankan ancestry, with additional ancestrally matched patients identified by principal component analysis. All patients were diagnosed with SSNS as per the Kidney Disease: Improving Global Outcomes guidelines. 12 Patients were recruited by collaborating clinicians at their affiliated institutions, as well as from the Prednisolone in Nephrotic Syndrome (PREDNOS, EudraCT 2010-022489-29) and PREDNOS2 (EudraCT 2012-003476-39) trials (Cattran et al. 13 and para 2 of Webb et al. 14 ). Informed written consent was obtained from each participant and ethical approval was granted at each contributing institution. Ancestrally matched controls were obtained from the UK Biobank. 15 Genotyping, Quality Control, and Whole-Genome Imputation Isolation of DNA and genotyping were performed using standard procedures (see Supplementary Methods). Patients were genotyped via the Infinium Multi-Ethnic Global Array BeadChip v.A1 at University College London Genomics (Institute of Child Health, University College London, UK). UK Biobank controls had been genotyped using the Applied Biosystems UK Biobank Axiom Array. Before imputation, quality control was performed on the case and control cohorts separately ( Figure 1). Individuals were excluded by low call rate (<95%), low genotyping quality (heterozygosity rates >3 SDs AE from the mean), and relatedness (IBD # 0.1875). Single-nucleotide polymorphisms (SNPs) were excluded by >2 alleles, low call rate (<99%), low minor allele frequency (<0.01), and in the control cohort only, deviation from Hardy-Weinberg equilibrium (P < 0.01). A further filter was applied to both cohorts to remove SNPs genotyped discrepantly between the Multi-Ethnic Global Array BeadChip and Axiom arrays. These SNPs were identified by comparison of a separate group of Sri Lankan healthy control subjects genotyped on the Multi-Ethnic Global Array BeadChip and the control cohort genotyped on the Axiom array (Supplementary Figure S1). Principal component analysis was used to identify the subset of cases and controls of Sri Lankan ancestry (Supplementary Figure S2).
Whole-genome imputation was performed with minimac4 on the Michigan Imputation Server 16,17 using the 1000 Genomes Project Phase 3 as the reference panel. 18 SNPs with a dosage R 2 of <0.8 were excluded. Postimputation quality control excluded SNPs by low call rate (<99%), low minor allele frequency (<0.01) and deviation from Hardy-Weinberg equilibrium (P < 0.01) in controls. PLINK versions 1.90 and 2.00 were used for quality control analysis. 19 Genome-Wide Association Analysis GWAS was performed in SAIGE 20 with adjustment for sex and the first 3 principal components of ancestry. Using >3 principal components resulted in genomic deflation, suggesting overfitting. Conditional analysis of the lead SNPs was performed in SAIGE using the same model adjusted for sex and principal components. A genome-wide significance threshold of P < 5 Â 10 À8 was used. R v4.2.1 was used to generate Manhattan plots. Regional plots were generated using LocusZoom with 1000 Genomes Nov 2014 used as the linkage disequilibrium (LD) reference. 21

HLA FINE-MAPPING
HLA imputation was performed with minimac4 on the Michigan Imputation Server using the HLA-TAPAS (HLA-Typing at Protein for Association Studies) reference panel. 16,17,22 HLA association analysis was performed in PLINK v2.00 using a logistic regression model adjusted for sex and the first 10 principal components of ancestry. Conditional analysis of the lead HLA allele was performed using the same logistic regression model adjusted for sex and principal components. Association testing and conditional analyses were performed on the HLA 4digit classical alleles and HLA amino acid polymorphisms separately. Significance thresholds of P < 3.0 Â 10 À4 (0.05/136) and P < 2.8 Â 10 À5 (0.05/ 1778) were used to adjust for multiple comparisons with the n ¼ 136 4-digit HLA classical alleles and n ¼ 1778 HLA amino acid polymorphisms used in the analysis, respectively.
Transethnic meta-analysis Transethnic meta-analysis was performed with a previously reported GWAS of European children with SSNS. 3 Analysis was conducted using the set of overlapping markers between the 2 data sets. The inverse-variance method was used based on a fixed-effects model in META: https://mathgen.stats.ox.ac. uk/genetics_software/meta/meta.html. The genomic inflation factor (l) and population sizes of each study were corrected for in the model. Results were considered significantly heterogeneous with a Cochran Q test P < 0.10. The genome-wide significance threshold for the meta-analysis was considered for P < 5 Â 10 À8 .

Replication
Replication of the 2 novel candidate SNPs (in TMEM131L and AHI1) was assessed in an independent population that comprised 150 South Asian (including Sri Lankan) participants from the INSIGHT 23 cohort and 277 controls from the Spit for Science study. 24 South Asian genetic ancestry was determined by principal component analysis using 1000 Genomes 18 ancestry controls as reference. Association analyses were carried out under an additive model. Significance threshold for replication was considered as P < 0.05/ 2 ¼ 0.025. For replication of candidate SNPs at the HLA locus, we examined the results of previously published GWAS in SSNS. [3][4][5][6] Power Calculation The GWAS and replication study power were calculated using the Michigan Genetic Association Study power CLINICAL RESEARCH calculator 25 assuming a disease prevalence of 1:10,000. For the initial GWAS (420 cases and 2339 controls), the minimum genotype relative risk with a power of 0.8 was calculated using an additive model assuming a disease allele frequency of 0.10 in the control population and a significance level of 5 Â 10 À8 . For the replication analysis (150 cases and 277 controls), power calculation assumed the genotype relative risk and allele frequency at each locus observed in the discovery GWAS, with a significance threshold of P < 0.025.

GWAS Study Cohort
A total of 663 individuals with childhood-onset SSNS and South Asian ancestry were available for our study, and 420 Sri Lankan cases were included in the association analysis following quality control and selection for Sri Lankan ancestry ( Figure 1). The control data set was obtained from the UK Biobank from cohorts of selfreported Indian, Bangladeshi, Pakistani, Any other Asian background, and other ethnic group ancestry for an initial total of 14,398 individuals. After ancestrally matching these individuals to our cases and performing quality control, 2339 healthy individuals with genetically determined Sri Lankan ancestry were included in the association analysis, with the majority obtained from the Any other Asian background and other ethnic group cohorts. The case and control cohorts were imputed and combined to yield a total of 5,265,125 high quality SNPs for analysis.
The next strongest association was outside the HLA region at 4q31.3 in the gene, TMEM131L, previously called KIAA0922 (rs74537360, P ¼ 2.98 Â 10 À8 , OR ¼ 2.37, 95% CI 1.75-3.22) (Figure 3b). Genome-wide significance was lost after conditioning on rs74537360 (Supplementary Figure S4). A further isolated marker (rs78120384) outside of HLA reached genome-wide significance, which was on the lower border of accepted allele frequencies and was deemed a falsepositive result (Figure 3c).
The power of this GWAS exceeded 80% to detect common alleles (minor allele frequency >0.01) with genotypic relative risk >2.2 at a significance threshold   are listed along the x-axis. The level of significance is depicted along the y-axis as Àlog 10 (P). Each dot represents a variant. The red line represents the threshold of genome-wide significance (P ¼ 5 Â 10 À8 ). Three loci achieve genome-wide significance on chromosomes 3, 4, and 6. QQ-plot and lambda are displayed in the top right corner. GWAS, Genome-wide association study; SSNS, sensitive nephrotic syndrome.
of P > 5 Â 10 À8 under an additive model. The inflation factor (l) was calculated to be 1.00 suggesting no evidence of genomic inflation.
HLA Fine-Mapping Significant association with SSNS was detected in 6 classical HLA alleles, including 3 previously reported subtypes associated with SSNS in Europeans: DQB1*02:01, DQA1*01, and DQA1*02:01 3 (Table 2 3,26,27 ). The strongest association was observed in DQB1*02:01, which was a risk haplotype. The strongest protective allele was in HLA-DQA1*01. Conditional analysis on the lead HLA allele, HLA-DQB1*02:01, revealed that the only further independent signal was in HLA-B*52:01 (Table 2 , Four HLA amino acid polymorphisms were significantly associated with disease, with the strongest association observed with an Alanine residue at position 74 of the HLA-DQB1 protein, which was a risk allele. Conditional analysis revealed a further protective allele with a Leucine substitution at position À4 of the HLA-DQB1 protein ( Table 2 , 3,26,27 ).

Replication
The novel genome-wide significant signal at 6q23.3 (AHI1) was replicated in an independent South Asian population (INSIGHT cohort) (rs2746432, P ¼ 1.13 Â 10 À2 , OR ¼ 1.58), although the power to do so was only 0.466 (Table 4). This lead SNP also showed evidence of association in the Japanese cohort published by Jia et al. (rs2746432, P ¼ 1.08 Â 10 À3 ). 6 The signal at 4q31.3 (TMEM131L) was not replicated in this cohort (rs74537360, P ¼ 0.76, OR ¼ 1.09), despite power to detect this signal being 0.894.

Gene Annotation
The lead SNP at the 6q23.3 locus (rs2746432) is a protein-coding variant for AHI1 and exhibits cis-eQTLs in the GTEx 28 database in almost all of the 54 tissues tested. Notably, rs2746432 shows strong cis-eQTL effects in fibroblasts (normalized effect size (NES) 0.51, P ¼ 1.6 Â 10 À28 ), Epstein-Barr virustransformed lymphocytes (NES 0.60, P ¼ 2.6 Â 10 À9 ), and in the spleen (NES 0.41, P ¼ 6.0 Â 10 À9 ). The SSNS risk (minor) allele in rs2746432 decreased the expression of AHI1 in all cell types, indicating that in cases where the risk allele was more frequent, the expression of AHI1 is down-regulated. The lead SNP (rs2746432) is also an eQTL for the genes LINC00271 (nonproteincoding variant) with strongest effects in testis (NES 0.15, P ¼ 4.1 Â 10 À6 ) and thyroid (NES 0.14, P ¼ 9.0 Â 10 À8 ) and RP3-388E23.2 (novel transcript in noncoding gene) with strongest effects in the cerebellum (NES 0.29, P ¼ 7.7 Â 10 À6 ) and pituitary gland (NES 0.26, P ¼ 2.7 Â 10 À5 ). Gene annotation in the UCSC genome browser demonstrates that the promoter region of AHI1 and LINC00271 overlap, and that when AHI1 is turned on, LINC00271 is turned off. 29 In the Human Kidney Cell Atlas, AHI1, LINC00271, and RP3-388E23.2 did not show any significant expression in adult kidney-related tissues; however, in the fetal kidney, AHI1 expression was significant in many kidney tissues, and was highest in the proximal tubule and plasmacytoid dendritic cells (a cell type that specializes in interferon production). 30 There was no significant eQTL for rs2746432 in the Human Kidney eQTL Atlas. 31

DISCUSSION
The present Sri Lankan GWAS and transethnic metaanalyses were performed in the largest South Asian cohort to date and identified common variants in AHI1 as a new susceptibility locus for childhood SSNS. This study has also confirmed previous association findings of SSNS with HLA-DQ/DR. Furthermore, the larger  sample size enabled additional fine-mapping of the HLA locus. These findings provide new insights into our understanding of the genetic background of childhood SSNS, and further support an immunologic basis to its pathogenesis. The identification of AHI1 and its associated proteins also reveals new targets for biological inquiry and potential therapeutic development, and provides evidence that genes implicated in rare Mendelian disorders can also harbor common variants in a complex disease. Gene annotation of the lead SNP at the 6q23.3 locus (AHI1) revealed that this SNP has eQTL effects in AHI1, LINC00271, and RP3-388E23.2. Indeed, in the meta-analysis, the lead SNP demonstrated LD extending into LINC00271, and functional annotation observed the transcription start site of these 2 genes to be overlapping, suggesting the potential for coregulation of these genes. Previous studies have indeed reported association of disease traits with an LD block encompassing all 3 genes. 32,33 AHI1 encodes the protein, jouberin, which is a component of a ring-like protein complex in the transition zone at the base of cilia. 34 Together, with the other proteins that compose the complex, AHI1 acts to restrict protein diffusion between the plasma and ciliary membranes; disruption of the complex leads to reduction in cilia formation and a reduction in signaling receptors from the remaining cilia. 35 Rare biallelic mutations in AHI1 cause Joubert syndrome, a rare monogenic disorder manifesting in agenesis of the cerebellum, ataxia, hypotonia, and intellectual disabilities. 36 Interestingly, our meta-analysis revealed that common variants in AHI1 are associated with SSNS.
AHI1 has a diverse array of biological functions. It is known to be important in the kidney through its interaction with NPHP1, which encodes another protein at the basal body of cilia. Mutations in NPHP1 are associated with Joubert syndrome accompanied by renal dysfunction, accounting for the majority of cases of nephronophthisis. 37 AHI1 and NPHP1 form heterodimers and heterotetramers, and mutations in AHI1 have been shown to change this binding pattern. 36 AHI1 is also involved in immune system function. Jiang et al. found that AHI1 is highly expressed in primitive types of normal hematopoietic cells and is down-regulated during early differentiation. 38 Therefore, alterations in AHI1 expression may contribute to the development of certain types of human leukemias. Notably, a GWAS in the autoimmune disease multiple sclerosis detected a susceptibility variant in AHI1 (rs4896153) (with LD extending into LINC00271) that was subsequently shown to have strong cis-eQTL effect on overall AHI1 expression. 32 Functional studies showed that expression peaked after stimulation of human CD4þ T cells, suggesting that it may play a role in early T-cell receptor activation. AHI1 has also been  shown to be involved in actin organization, 39 and therefore the authors of this study speculated that AHI1 may play a role in the formation or stabilization of the T-cell receptor synapse as a mechanism for its association with multiple sclerosis. 32 The eQTL analysis of the lead SNP, rs2746432, showed cis-eQTL effects on AHI1 in Epstein-Barr virus-transformed lymphocytes, with the risk allele at this variant associated with decreased AHI1 expression. Thus, it is possible that decreased AHI1 expression in the lymphocytes of individuals with SSNS could lead to increased cytokine production, and/or destabilization of the T-cell receptor complex, both resulting in immune system dysregulation. The association of common variation at AHI1 with SSNS in addition to the established association of (biallelic) rare variants of AHI1 with Joubert syndrome demonstrates that variation across the allele frequency spectrum in a gene can contribute to both monogenic and polygenic disease, and that these alleles might act by different mechanisms, resulting in altogether different disorders.
The strongest association in our analysis was in the HLA region. In the Sri Lankan discovery cohort, the lead SNP, rs9271602, was in the HLA-DR/DQ region, specifically in the HLA-DQA1 and HLA-DQB1 genes. This finding was also detected in the previously published European GWAS, 3 and was therefore unsurprisingly also observed in our transethnic meta-analysis of the European and Sri Lankan cohorts, with the strongest association at rs2856665, between HLA-DQB1 and HLA-DQA2 but with LD extending to HLA-DQA1. All previous GWAS published on SSNS have found association within these genes, including populations of European, South Asian, and Japanese ancestry. [3][4][5][6]40 Fine-mapping of the HLA alleles identified HLA-DQB1*02:01, HLA-DQA1*02:01, HLA-DPB1*17:01, and HLA-B*52:01 to be associated with increased risk of SSNS. Of these, HLA-DQA1*02:01 and HLA-DQB1*02 were also found to be associated with increased risk of disease in European and South Asian studies. 3,5,40 HLA-DQA1*01 and HLA-DQB1*05 were the protective alleles associated with SSNS in our Sri Lankan discovery cohort, with the HLA-DQA1*01 allele replicating in European 3 and South Asian 40 populations. Conversely, in Japanese populations, altogether different risk and protective HLA alleles have been identified. 4,6 These findings demonstrate substantial overlap between European and South Asian populations, but not Japanese. This is most likely explained by differing allele frequencies in the different populations (see Table 2 and Supplementary  Table S2).
In the conditional analysis on the lead allele, HLA-DQB1*02:01, in our Sri Lankan discovery cohort, both the risk and protective alleles at HLA-DR/DQ disappeared, leaving only HLA-B*52:01 as the independent signal. HLA-B*52:01 (the most common subtype of allele at B*52) was the only class I HLA allele associated with disease in our discovery GWAS analysis. Although this allele was relatively rare in the Sri Lankan population (minor allele frequency 0.08), it is interesting because of its association with several other immune-mediated diseases, including ulcerative colitis 41 and Takayasu's arteritis. 42  HLA-DQB1 A74 is involved in the formation of the peptide-binding cleft, 43 and it is also in high LD with HLA-DQB1 A57 which is critical for peptide binding and recognition. 44 The strongest association outside of HLA in our transethnic meta-analysis was on chromosome 6 (rs2637681, P ¼ 5.44 Â 10 À13 , OR ¼ 0.62, 95% CI 0.54-0.70) in the gene, CALHM6. 45 This locus was associated with SSNS in the previously published European GWAS, 3 and was also reported as a potential signal in the SSNS GWAS by Debiec et al. 5 However, it was not significantly associated with disease in our Sri Lankan discovery cohort (Supplementary Table S1). Unsurprisingly, association of rs2637681 in our transethnic metaanalysis was mainly driven by the European cohort. The direction of effect was the same in both cohorts, however, which supports the relevance of this finding.
Power calculation indicated that the Sri Lankan discovery GWAS was powered to detect a signal of similar strength at this locus at P < 0.05, even considering the lower frequency of the associated allele in the Sri Lankan population (0.162 compared with 0.413 in Europeans). However, the observed signal (P ¼ 0.199, OR ¼ 0.86, 95% CI 0.69-1.08) was not as strong as this. There are several potential explanations for this, including differences in LD patterns in individuals of different ancestries, or that there is a true difference in effect size at this variant, perhaps because of differences in genetic background or environmental exposures. 46 It has been previously demonstrated that variants associated with a particular disease in one ancestral group are not always reproduced in another. 47 Furthermore, variants shared among autoimmune disorders have been shown to be protective in one disorder and risky in another. 48 This study has several limitations. First, there was limited clinical information on the individuals included in the study. Details such as age of onset or relapse pattern could have enhanced our understanding of the relationship between markers of clinical severity and number of risk alleles, although the relatively small size of the cohort would limit the power to perform this type of analysis. Second, the case and control data sets were genotyped on different platforms, which limited the number of overlapping markers. We overcame this problem by filtering for genotyping discrepancies and imputing the data sets separately, but this has greater potential for error than if the case and control data sets were genotyped on the same platform. Third, though this study provides evidence of association with alleles at TMEM131L in the Sri Lankan discovery cohort, this was not replicated in either the meta-analysis or in an independent South Asian cohort, suggesting that this association is most likely a type 1 error; it is also possible that this association represents a genetic risk factor uniquely found in the Sri Lankan (as opposed to the South Asian or European) population. Replication in an independent Sri Lankan cohort is needed.
In summary, our study showed a novel association of childhood SSNS with alleles at AHI1 and confirmed previous associations at HLA-DR/DQ. These findings further support the role of immune dysregulation in the pathophysiology of disease. The AHI1 association, in particular, suggests a link between a ciliary gene and glomerular disease and reinforces an emerging paradigm in nephrology: in genes harboring rare Mendelian variants, common alleles can increase the susceptibility of polygenic diseases. Our study also illustrates the importance of performing GWAS in larger data sets by combining populations of diverse ancestry, because by doing so, we were able to increase the power to detect a novel variant associated with SSNS.

DISCLOSURE
All the authors declared no competing interests.

ACKNOWLEDGMENTS
The authors would like to acknowledge the patients and families who participated in this research, without direct benefit to themselves. MLD

DATA AVAILABILITY STATEMENT
The data from our previously published European SSNS GWAS are available here: https://ega-archive.org/studies/ EGAS00001003607. The data underlying the discovery GWAS from this article will also be shared via the NHGRI-EBI GWAS Catalog. Accession numbers and/or DOIs will be made available after acceptance.

AUTHOR CONTRIBUTIONS
MLD designed the study, performed quality control and association analysis, interpreted results, drafted the manuscript, and approved of the manuscript as written. SG contributed to data collection, interpreted results, critically revised the manuscript, and approved of the manuscript as written. CV provided computing support, interpreted results, critically revised the manuscript, and approved of the manuscript as written. APL designed in-house computing software, critically revised the manuscript, and approved of the manuscript as written. OS-A provided computing support, interpreted results, critically revised the manuscript, and approved of the manuscript as written. SD-K contributed to data collection, critically revised the manuscript, and approved of the manuscript as written. JC analyzed validation cohort data, critically revised the manuscript, and approved of the manuscript as written. MC contributed patient data and approved of the manuscript as written. JAK contributed patient data and approved of the manuscript as written. ST contributed patient data and approved of the manuscript as written. RR contributed patient data and approved of the manuscript as written. AA contributed patient data and approved of the manuscript as written. RG contributed patient data and approved of the manuscript as written. RP contributed patient data, interpreted results, critically revised the manuscript, and approved of the manuscript as written. RK, DB, HCS, and DPG designed the study, interpreted results, critically revised the manuscript, and approved of the manuscript as written.

SUPPLEMENTARY MATERIAL
Supplementary File (PDF) Supplementary Methods Supplementary Figure S1. Genotyping discrepancy analysis in Sri Lankan controls genotyped alongside Sri Lankan SSNS cases versus UK Biobank controls. Supplementary Figure S2. Principal component analysis of case-control data set anchored by 100 Genomes controls. Supplementary Figure S3. Conditional analysis in HLA-DQ/ DR region identified in Sri Lankan discovery cohort. Supplementary Figure S4. Conditional analysis in 4q31.3 region (TMEM131L) identified in Sri Lankan discovery cohort.
Supplementary Figure S5. HLA 4-digit allele analysis in the Sri Lankan discovery cohort. Supplementary Table S1. Genome-wide significant variants associated with SSNS published in European, Japanese, and Sri Lankan populations. Supplementary Table S2. HLA alleles associated with SSNS in Sri Lankan and European populations. SUPPLEMENTARY REFERENCES