Genome-Wide Admixture Mapping of Estimated Glomerular Filtration Rate and Chronic Kidney Disease Identifies European and African Ancestry-of-Origin Loci in Hispanic and Latino Individuals in the United States

Significance Statement

Populations of Hispanic or Latino individuals have an increased risk of CKD, yet little is known about CKD genetics in these underserved groups. A genome-wide admixture mapping study of CKD traits conducted in 12,601 participants from the Hispanic Community Health Study/Study of Latinos (HCHS/SOL) identified three novel ancestry-of-origin loci on African-derived and European-derived chromosomal haplotypes—implicating novel candidate genes for kidney function in these loci. Two of these loci were validated in Black individuals, indicating potential generalizability of the loci across populations with shared ancestry. Interestingly, a genome-wide association study in HCHS/SOL failed to identify these ancestry-specific regions. These results illustrate the utility of leveraging diverse ancestries via admixture mapping for new insights into the genetics of CKD traits in studies of recently admixed populations.

Visual Abstract

Abstract

Background

Admixture mapping is a powerful approach for gene mapping of complex traits that leverages the diverse genetic ancestry in populations with recent admixture, such as Hispanic or Latino individuals in the United States. These individuals have an increased risk of CKD.

Methods

We performed genome-wide admixture mapping for both CKD and eGFR in a sample of 12,601 participants from the Hispanic Community Health Study/Study of Latinos, with validation in a sample of 8191 Black participants from the Women’s Health Initiative (WHI). We also compared the findings with those from a conventional genome-wide association study.

Results

Three novel ancestry-of-origin loci were identified on chromosomes 2, 14, and 15 for CKD and eGFR. The chromosome 2 locus comprises two European ancestry regions encompassing the FSHR and NRXN1 genes, with European ancestry at this locus associated with increased CKD risk. The chromosome 14 locus, found within the DLK1-DIO3 imprinted domain, was associated with lower eGFR and driven by European ancestry. The eGFR-associated locus on chromosome 15 included intronic variants of RYR3 and was within an African-specific genomic region associated with higher eGFR. The genome-wide association study failed to identify significant associations in these regions. We validated the chromosome 14 and 15 loci associated with eGFR in the WHI Black participants.

Conclusions

This study provides evidence of shared ancestry-specific genomic regions influencing eGFR in Hispanic or Latino individuals and Black individuals and illustrates the potential for leveraging genetic ancestry in recently admixed populations for the discovery of novel candidate loci for kidney phenotypes.

CKD affects one in seven US adults, with a disproportionate burden on Black individuals and Hispanic or Latino (HL) individuals.^1,2 Ancestry-specific genetic contributions to CKD risk have been identified in Black individuals and HLs, such as the African-derived APOL1 G1 and G2 genotypes.^3,4 Few genome-wide association studies (GWAS) or multi-ethnic GWAS meta-analysis of CKD and eGFR have included HL populations.^5–11 Because genetic risk may not be shared across populations,¹² leveraging the genetic diversity in recently admixed populations offers an opportunity to identify novel loci that have not been detected in genetic studies of homogeneous ancestry populations.

Admixture mapping is a powerful approach for genetic mapping of complex diseases in multi-ethnic populations. Recent admixture between continental ancestral populations, such as in HLs, creates long-range blocks of correlation [linkage disequilibrium (LD)] between genetic variants that have allele frequency differences in the founding populations, turning the genome of an admixed individual into a mosaic of chromosome segments of different ancestral origins (Figure 1A). The rationale and motivation for admixture mapping is based on causal variants having higher frequencies on chromosomal segments inherited from the ancestral population(s) with higher disease prevalence. Admixture mapping leverages differences in genetic ancestry to identify associations between these ancestry-specific chromosomal segments (local ancestry) at each locus and traits, which can yield both new and complementary findings compared with GWAS (association mapping) of complex diseases. Admixture mapping led to the discovery of the APOL1 genomic region for focal segmental glomerulosclerosis and hypertension-attributed ESKD in Black individuals, with subsequent identification of the G1 and G2 risk genotypes within the chromosome 22 region.³

Figure 1.

Admixture mapping and association mapping for eGFR and CKD. (A) Mosaic of chromosome segments of different ancestral origins of Hispanic and Latino populations. (B) Admixture mapping Manhattan plots for eGFR (left) and CKD (right) in HCHS/SOL. (C) GWAS Manhattan plots for eGFR (left) and CKD (right) in HCHS/SOL. Only GWAS results for the chromosomes of interest 2, 14, and 15 are represented. SNPs in the regions associated in the admixture mapping are highlighted in green. Note two significant signals on chromosome 2 unrelated to the admixture mapping findings in the GWAS for CKD driven by low frequency variants (rs146424372 and rs116137931).

HL populations are admixed with African, European, and Native American ancestries. Using an admixture mapping approach that allows for the testing of multiple ancestry effects simultaneously, our recent study identified a novel American Indian locus at chromosome 2 associated with urine albumin excretion.¹³ In addition to identifying trait-associated loci with causal variants that have large allele frequency differences across ancestries, our approach also allows for the identification of the ancestries that are driving these associations. We also recently extended our admixture mapping approach for analyses of categorical outcomes such as CKD.¹⁴ This study aims to identify and fine-map kidney loci for eGFR and CKD using admixture mapping in HL participants of the Hispanic Community Health Study/Study of Latinos (HCHS/SOL). It is worth noting that this study leverages ancestry, which represents the genetic background of the individuals, and not race and ethnicity, which are two social constructs with no inherent biologic meaning.

Methods

Study Samples and Phenotype

HCHS/SOL is a community-based cohort study that recruited 16,415 participants aged 18–76 years from four US field centers (Bronx, Chicago, Miami, and San Diego) representative of the largest US HL groups: Central Americans, Cubans, Dominicans, Mexicans, Puerto Ricans, and South Americans. The study has extensive familial relatedness, as previously discussed.¹⁵ Details about the complex sampling design and cohort selection for HCHS/SOL are described elsewhere.¹⁶ Sociodemographic, physical and lifestyle evaluations, and fasting blood and spot urine samples were collected at the baseline visit (2008–2011). Participant demographics were obtained through a questionnaire during the clinical visits, which included questions about gender and HL background group. The HCHS/SOL study was approved by the institutional review board at each participating institution and collected a written informed consent from all participants.

Participants were genotyped for more than 2.5 million single-nucleotide polymorphisms (SNPs) using a customized Illumina array, and imputation was performed using the 1000 Genomes Project phase I reference panel. The quality control procedures and details about genotyping, imputation, and local ancestry calls were previously described.^15,17,18 eGFR, estimated using the Chronic Kidney Disease Epidemiology Collaboration equation¹⁹ based on serum creatinine measured using a creatinase enzymatic method, was inverse normal transformed for validity of the normality assumptions for the statistical procedures. CKD was defined by an untransformed eGFR<60 ml/min per 1.73 m². We performed analyses for eGFR and CKD in a subset of 12,601 subjects with available genotype and phenotype data.

An independent sample of 3050 HL and 8191 Black women from the Women’s Health Initiative (WHI)²⁰ SNP Health Association Resource was used to validate our findings. Local ancestry calls for both WHI HLs and Black individuals were inferred previously.²¹

Statistical Analyses

Multiple Random and Fixed Effects

The HCHS/SOL study was developed under a complex sampling design. To account for the complex correlation structure of the data, we included three random effects: (a) genetic relatedness (or kinship), (b) household and (c) census block group (the geographic cluster of the households), which accounts for shared polygenic effects among relatives as well as shared environment effects. The statistical model also included the top five principal components that reflect the population structure in the sample, recruitment center, genetic analysis group, age, and sex as fixed effects. The genetic analysis groups correspond to the same self-identified background groups (Central American, Cuban, Dominican, Mexican, Puerto Rican, or South American), where samples were relocated to compound more genetically homogenous clusters. Details on the estimation of genetic analysis groups, principal components, and kinship coefficients were previously reported.¹⁵ For linear mixed models (LMMs), we also estimated separate residual variance components by genetic analysis group to allow for violations of the homoscedasticity assumptions.

Admixture Mapping

The admixture mapping analysis was performed using an LMM and a generalized linear mixed model association test (GMMAT)¹⁴ for eGFR and CKD outcomes, respectively, in which all African, European, and Native American ancestries were tested simultaneously. Statistical details about LMM and GMMAT are provided in Supplemental Appendix 1. Admixture mapping with LMM and GMMAT are implemented in the GENESIS R package.²² We first fit the models under the null hypothesis of no genetic ancestry effect, including the random and fixed effects described above. We then used the null models to evaluate the association between the local ancestry at each locus and the outcomes using a score test. Here, locus is defined as a long-range block of admixture-LD created by the recent admixture in HL populations, and local ancestry corresponds to the locus-specific ancestry allelic dosages (zero, one, or two copies of African, European, or Native American alleles) estimated from the genotype data. The width of the admixture mapping peak was defined by the boundaries of the admixture-LD block (or local ancestry locus) based on the physical positions. Because every SNP in a local ancestry block has the same admixture mapping signal and effect size, only one SNP per locus, referred to as the lead SNP, was included in the analyses, which resulted in a total of 15,500 loci tested via admixture mapping. Secondary admixture mapping analyses were performed by testing each ancestry against the others to identify which ancestry is driving the association at each locus. Based on previous simulation analysis, a nominal P of 5.7 × 10⁻⁵ controls the type I error at the level of 0.05,²³ and accounts for the number of independent tests performed under populational structure. Given the long-range LD structure in recent admixed populations, a smaller number of tests should be conducted, and consequently, significance thresholds applied for admixture mapping are not as stringent as in GWAS. The effect sizes for the significant ancestry blocks were estimated based on the allelic dosage of the ancestry driving the signal. To provide evidence for the locus accounting for the observed local ancestry association, we conducted conditional admixture mapping including the lead SNPs representing each associated locus as covariates in the models. Additional sensitivity analyses evaluated the robustness of our findings to the inclusion of diabetes and hypertension as covariates. Variants were annotated using Ensembl (http://uswest.ensembl.org/index.html), Variant Effect Predictor (https://uswest.ensembl.org/info/docs/tools/vep/index.html), AMP web tools (https://hugeamp.org/), and Combined Annotation Dependent Depletion score for deleteriousness (scaled C-score; cutoff ≥ 15 for potential deleteriousness). We used FORGE2 (https://forge2.altiusinstitute.org/) to identify tissue type– or cell type–specific signal of variants for overlap with epigenomic DNase I hypersensitive sites, histone mark chromatin immunoprecipitation broad peaks, and hidden Markov model chromatin states using Roadmap epigenetic data (https://www.encodeproject.org/). Known and predicted regulatory elements overlapping SNPs were also assessed using RegulomeDB.²⁴ We queried datasets of single-cell RNA sequencing for gene expression in healthy adult human kidney and in diabetic human kidney tissues using the Kidney Interactive Transcriptomics tool (http://humphreyslab.com/SingleCell/).

Association Mapping

We performed a GWAS on the HCHS/SOL imputed genotype data using the GENESIS R package.²² We first filtered imputed SNPs based on the information metric provided by IMPUTE2²⁵ (info>0.8), the ratio of observed variance of imputed dosages to the expected binomial variance (oevar>0.8), and the effective minor allele count (2p(1-p)*N*oevar, where p is the minor allele frequency and N is the sample size) >30 for eGFR and >50 for CKD. Using the null models fit for the admixture mapping, we tested the association between each SNP, using the dosage of reference allele, and the outcomes using a score test.

Admixture Mapping Validation

We performed validation studies for the HCHS/SOL admixture mapping findings in 3050 HL and 8191 Black independent samples from the WHI SNP Health Association Resource,²⁰ using the local ancestry calls previously estimated.²¹ Admixture mapping for eGFR and CKD outcomes in HL and Black samples were conducted in the regions of interest using mixed models adjusted by age and the top four principal components. Admixture mapping and coefficient estimation for the associated loci followed the methods described above^14,22 and detailed in Supplemental Appendix 1. P values were adjusted for false discovery rate.

Results

Table 1 displays the characteristics of HCHS/SOL participants including comorbidities. The mean age was 46.1 (range 18–76 years), and 59% were women. The overall CKD prevalence based on an eGFR<60 ml/min per 1.73 m² was 3.4%.

Table 1.

Descriptive characteristics of HCHS/SOL samples

Trait	Total Sample (n=12,601)
Age (yr), mean±SD	46.1±13.9
Women, n (%)	7428 (59.0)
eGFR (ml/min per 1.73 m2), mean±SD	96.5±18.9
CKD, n (%)	427 (3.4)
BMI (kg/m2), mean±SD	29.8±6.0
Diabetes, n (%)	2470 (19.6)
Hypertension, n (%)	3525 (28.0)
Genetic groups, n (%)
Central America	1381 (11.0)
Cuban	2246 (17.8)
Dominican	1172 (9.3)
Mexican	4660 (37)
Puerto Rican	2229 (17.7)
South American	913 (7.2)
Global ancestry, mean±SD
African	0.14±0.16
European	0.56±0.21
Native American	0.30±0.24

Genetic groups are derived from the original self-identified background groups to have more genetically heterogeneous samples. Global ancestry refers to global ancestries estimated by averaging the local ancestry calls across all chromosomes. BMI, body mass index.

Admixture Mapping Findings for eGFR and CKD

The admixture mapping of eGFR identified two genome-wide significant loci on chromosomes 14 and 15, whereas the admixture mapping of CKD identified two loci on chromosome 2 (Figure 1B). When testing each ancestry contribution to admixture mapping findings, the association on chromosome 14 was driven by European ancestry, whereas African ancestry was driving the association on chromosome 15 (Figure 2A). For the CKD loci, the chromosome 2 associations were driven by European ancestry (Figure 2B). Therefore, European and African ancestry chromosome segments contributed to our new findings (Table 2). Sensitivity analyses showed that these results were robust to the diabetes and hypertension adjustments (Supplemental Figure 1).

Figure 2.

Single ancestry admixture mapping analyses. (A) Mapping for eGFR and (B) mapping for CKD. In both cases African, European, and Native American ancestries were tested separately. Note that the eGFR chromosome 14 signal is driven by European ancestry and the chromosome 15 signal is driven by African ancestry. For CKD, the chromosome 2 signal is driven by European ancestry.

Table 2.

Admixture mapping results for eGFR and CKD

	HCHS/SOL Admixture Mapping Results							WHI Black individuals Admixture Mapping Validation
Trait	Chr	#SNPsa	Physical position (GRCh37/hg19)	Admixture Mapping joint Pb	Effect (95% CI)c	Ancestry backgroundd		#SNPsa	Physical position (GRCh37/hg19)	Admixture Mapping joint Pb	Effect (95% CI)c
						Ancestry	P
eGFR	14	4	101329926–101350298	4.0 x 10-5	−0.05 (−0.07 to −0.03)	EUR	1.0 x 10-5	2	101349017–101350298	0.02	−0.04 (−0.08 to −0.01)
	15	37	33801207–33884488	2.4 x 10-5	0.07 (0.04 to 0.11)	AFR	2.2 x 10-5	5	33801207– 33812915	0.04	−0.04 (−0.08 to −0.003)
								41	33821715–33884488	0.04	−0.04 (−0.08 to −0.002)
CKD	2	66	49571948–49894154	3.3 x 10-5	1.46 (1.23 to 1.73)	EUR	6.1 x 10-6	No validation
		37	49909155–50197075	5.1 x 10-5	1.45 (1.23 to 1.71)	EUR	9.1 x 10-6	No validation

Chr, chromosome; AM, admixture mapping; EUR, European; AFR, African.

^aNumber of SNPs within each locus.

^bP for the admixture mapping joint test, in which all ancestries were tested simultaneously. The significance threshold for the HCHS/SOL admixture mapping is a P<5.7 x 10^-5; for WHI admixture mapping validation, the AM joint P was adjusted for false discovery rate.

^cCoefficient (95% confidence interval) for eGFR and odds ratio (95% confidence interval) for CKD.

^dAncestry driving the signal, according to the single ancestry admixture mapping analysis, and the respective P of the statistical test.

The eGFR association on chromosome 14 was captured by four SNPs. The beta coefficient for local ancestry was −0.05 (SEM=0.01, P=9.8 x 10^-6) indicating a decrease in eGFR for each additional copy of the European allele at this locus. The eGFR association on chromosome 15 was observed in a locus with 37 SNPs, all of them within the RYR3 gene, where African ancestry conferred a protective effect against low eGFR (beta coefficient = 0.07, SEM=0.02, P=2.1 x 10^-5). The CKD association on chromosome 2 was driven by two loci with 66 and 37 SNPs, respectively. These regions include intergenic variants near the FSHR and NRXN1 genes and intronic variants of NRXN1. The odds ratio (OR) for CKD was similar across the two loci (OR, 1.46; 95% confidence interval [95% CI], 1.23 to 1.73; OR, 1.45; 95% CI, 1.23 to 1.71 for the first and second loci, respectively) which increased risk of CKD for each additional copy of the European allele (Supplemental Table 1, Table 2). Overall, our study has identified three novel candidate loci for eGFR or CKD in HL samples, which were driven by ancestry-specific backgrounds.

Conditional Admixture Mapping Analyses

To provide supporting evidence for identified loci accounting for observed admixture mapping associations, we performed conditional admixture mapping analyses including the lead SNP of each locus as covariate. Conditional analyses on one or more lead SNPs completely accounted for the associations on chromosomes 2, 14, and 15 (Supplemental Figure 2). Note that the two regions on chromosome 2 are only 15 kb apart and show similar effect, and they explained the admixture association whether tested jointly (Supplemental Figure 2D) or separately (Supplemental Figure 2, E and F). These findings support the presence of a single signal at these loci.

Association Mapping Results

To compare gene discovery from admixture mapping to discovery using genome-wide association, we performed a GWAS for eGFR and CKD, fitting similar statistical models as applied to admixture mapping to avoid bias owing to adjustments. The GWAS did not identify significantly associated loci at the chromosomes 2, 14, and 15 regions (Figure 1C). In the GWAS for eGFR, no SNP reached the genome-wide significant threshold (P<5.0 x 10^-8) on chromosomes 14 and 15. For CKD, the GWAS identified two low frequency variants (minor allele frequency = 0.3%) statistically significant on chromosome 2 (rs146424372, 25 kb downstream of SCG2, P=3.5 x 10^-8; rs116137931, intronic variant of LINC01317, P=4.1 x 10^-8) in distinct regions of the admixture mapping signals. These findings support potential gains in gene discovery when using admixture mapping that leverages local ancestry compared with GWAS in HL populations.

Validation of the Admixture Mapping Findings

To validate loci identified in the HCHS/SOL admixture mapping, we performed admixture mapping on the chromosomes 2, 14, and 15 regions of interest in independent HL samples from the WHI (n=3050). We also conducted the validation on the WHI Black individuals (n=8191) because the secondary admixture mapping analyses identified regions of European and African backgrounds. Both eGFR admixture mapping findings on chromosomes 14 and 15 were validated in WHI Black individuals (Table 2) but not in WHI HL individuals. On chromosome 14, we identified a locus with two SNPs associated with eGFR (rs6575805 and rs3825569; P=0.02), in which the European background showed a similar magnitude and direction of the effect (beta coefficient = −0.04, SEM=0.02, P=0.02) of that observed on HCHS/SOL. On chromosome 15, a significant association was captured by two loci including, respectively, five and 41 SNPs (P=0.04). Interestingly, the African background in this region conferred an opposite direction of the effect observed in HCHS/SOL, which could be explained by distinct genetic variants underlying the admixture signals in HL and Black individuals or by their different ancestry backgrounds. More discussion is provided below. The effect sizes estimated for both loci were similar (beta coefficient = −0.04, SEM=0.02, P=0.03 and P=0.04, respectively) suggesting that the transformed eGFR outcome decreases 0.04, on average, for each additional copy of an African haplotype in this region. We did not identify a significant association in the chromosome 2 admixture mapping region of interest for CKD. In summary, we validated two loci associated with eGFR but not the chromosome 2 region associated with CKD.

Discussion

This study is a gene discovery effort to map genetic loci for eGFR and CKD using chromosome ancestral background in HL populations. We compared our approach for gene discovery with the traditional GWAS. We demonstrated that leveraging diversity in genetic ancestry across the genome in the analysis via admixture mapping can help to identify novel candidate loci influencing eGFR and CKD in recently admixed HL populations. The main findings of this study were the discovery of three novel loci on chromosomes 2 (for CKD) and 14 and 15 (for eGFR), in which the associations were largely driven by European and African ancestries. These loci were not identified in our HL sample using GWAS and also have not been implicated in previous large-scale GWAS meta-analyses using homogeneous ancestry populations, despite these loci harboring genes that have been implicated in kidney function and diseases, as discussed below.

The European region on chromosome 2 associated with CKD comprises two loci spanning 322.2 kb and 287.9 kb, respectively. Variants in these loci are mostly intergenic near the FSHR and NRXN1 genes (2p16.3), with some intronic to NRXN1, a gene with low expression in different cell types in both healthy and diabetic kidney human tissue. Four variants in this region show potential deleterious effects (Combined Annotation Dependent Depletion [CADD] ≥15; rs1405959, rs17038952, rs13425236, rs1045881), and several variants overlap with epigenetic functional elements in fetal kidney tissue (Supplemental Table 1). FSHR encodes the receptor for follicle-stimulating hormone (FSH). High FSH levels were associated with the increased risk of CKD in Chinese postmenopausal women; in experimental models, FSH causes kidney dysfunction and tubulointerstitial fibrosis, and worsening kidney injury through macrophage recruitment and activation.²⁶ The NRXN1 gene encodes a single-pass type I membrane protein, which is involved in the central nervous system. Variants in this gene have been implicated in blood pressure, body mass index, and a range of cognitive traits.¹¹ The chromosome 2 region has multiple SNPs associated with lifestyle risk factors for CKD. Variants rs4521079 and rs13401858 have been associated with Western dietary patterns of processed meat in the UK Biobank.²⁷ Additionally, rs13401858 and rs971732 are in high LD with rs10179773 and rs12465974, respectively, which have been significantly associated with smoking status in large-scale European ancestry GWAS.^28,29

The European locus on chromosome 14 associated with eGFR includes four SNPs (Supplemental Table 1) spanning a 20.4 kb region within the DLK1-DIO3 imprinted domain (14q32.2), which contains paternally expressed genes (RTL1, DLK1, DIO3) and maternally expressed long and small noncoding RNAs including MEG3 (a gene expressed in endothelial cells in healthy and diabetic human kidney tissue), antisense RTL1, microRNAs, and pseudogenes. Epigenomic annotation showed that rs8008201 and rs3825569 overlap the histone mark H3K36me3, and rs11851174 is likely to affect transcription factor binding (ranking 2b in RegulomeDB²⁴), suggesting regulatory function. The imprinted region on 14q32.2 has been associated with type I diabetes susceptibility.³⁰ Furthermore, the DLK1-DIO3 domain has been implicated in placental and embryonic tissue development and metabolism.³¹ A knockout mice of Peg11/Rtl1 gene showed abnormalities of fetal capillaries leading to lethality and fetal growth retardation.³² All SNPs upstream/downstream of microRNAs are directly implicated on urological cancers and kidney fibrosis.³³ MIR433 is a profibrotic microRNA, and it has been considered as a potential therapeutic target for kidney fibrotic disease.³⁴

The 83.2 kb African locus on chromosome 15 harbors 37 variants (35 intronic, one missense [rs2077268; CADD=16.97], and one missense/synonymous [rs674155; CADD=14.02]) in the RYR3 gene (15q13.3–14). Several variants overlap DNase I hypersensitive sites and histone marks, including rs581954, a deleterious SNP (CADD=15.66) that also demonstrates evidence of affecting transcription factor binding (ranking 2a) using RegulomeDB.²⁴ Rs2077268 also has a potential deleterious effect (CADD≥15) and has shown a suggestive pharmacogenetic effect on heart failure among Black individuals and HL individuals.³⁵ RYR3 is involved in intracellular calcium ion release channels; it is highly expressed in skeletal muscles, and has varying expression in other tissues including kidney.^36–38 In kidney tissue single-cell RNA sequencing databases, RYR3 is expressed in podocytes in healthy adult human kidney and in diabetic human kidney tissue. The RYR3 gene has been associated with diabetic kidney disease (rs2596230) in both a GWAS in Black individuals and a transethnic GWAS meta-analysis, although findings did not reach genome-wide significance thresholds.³⁹ RYR3 has also been associated with plasma adiponectin levels and inflammation, and elevated adiponectin has been implicated with CKD occurrence and progression.^40,41

Of the loci identified in the admixture mapping, we provided evidence for validation of the two loci for eGFR but we did not validate the chromosome 2 loci associated with CKD. The chromosome 14 locus had a consistent direction of effect for the European ancestry region when comparing HCHS/SOL results and WHI. Interestingly, the African locus on chromosome 15 had the opposite direction of effect for HL and Black individuals, and a variety of factors may have contributed to this result. Note that admixture mapping nominates an ancestry-of-origin region associated with a phenotype, but it does not identify the underlying causal variants, which could be risk or protective alleles. There could be distinct genetic variants underlying the admixture association in the African locus in HL and Black individuals owing to different genetic architecture influencing CKD. The HL analysis also included an additional ancestral population (Native American), and so the ancestry effects are not directly comparable because the ancestral baseline is different owing to the Native American ancestry contributions to HL individuals that are not present in Black individuals.

Several factors may have contributed to the nonsignificant admixture mapping results for validation in the WHI HL sample for eGFR and in the WHI HL/Black samples for CKD. Although HL populations share African, European, and Native American backgrounds, they are highly heterogeneous in terms of proportions of ancestral populations (global ancestry proportions) and ancestry at the level of chromosomal regions (local ancestry). The WHI HL samples have a different ancestry composition to that observed in HCHS/SOL (Supplemental Figure 3). Almost half of the subjects are Mexicans (versus 37% in HCHS/SOL), and ancestry backgrounds other than Cuban, Mexican, and Puerto Rican are not distinguished and are likely small. The WHI HL sample is also 4-fold smaller than HCHS/SOL. Therefore, low statistical power, differences in genetic background, or a combination of both can explain the lack of validation of some of the regions in WHI HL individuals. For CKD validation, the sample of WHI Black individuals was larger and had a higher proportion of prevalent CKD compared with the discovery sample (6% versus 3.4%). However, the frequency of European alleles at the chromosome 2 loci was lower in WHI Black individuals (20% for CKD cases and 24% for non-CKD in WHI Black individuals versus 65% for CKD cases and 54% for non-CKD in HCHS/SOL). This suggests that the number of European alleles may not have captured the association in this region.

We compared our results with a GWAS and attempted to fine-map our regions using GWAS data. The GWAS did not identify genome-wide significant associations within the admixture mapping regions. However, we identified two low frequency variants on chromosome 2 associated with CKD, one of them in the region of SCG2 gene, which had been previously implicated with end-stage renal disease.⁴² These results illustrate the complementarity of admixture and association approaches in identifying and improving complex trait mapping in multi-ethnic populations.

There are some limitations to our findings. Our admixture mapping analysis captures both common and rare variants through local ancestry. However, the GWAS fine-mapping was largely limited to the common variants from the directly genotyped or imputed data and rare variants within the associated region were not interrogated. In addition, the available WHI HL samples for validation had a different ancestry composition compared with the HCHS/SOL samples, which could explain, in part, the nonsignificant admixture mapping results for validation in HL individuals, in addition to the small sample size, as previously mentioned.

In summary, we identified three novel loci associated with eGFR and CKD using admixture mapping in a HL sample, which were not identified by the traditional GWAS approach. The signals on chromosomes 14 and 15 were driven by European- and African-specific loci within the DLK1-DIO3 imprinted domain and RYR3 gene, respectively, and both replicated in WHI Black samples. The CKD signal on chromosome 2 was captured by a European-specific loci near the FSHR and NRXN1 genes but the findings did not replicate. Our study provides new information into genetic contributions to eGFR in HL individuals, and in particular, ancestry-specific genetic regions that are implicated for eGFR.

Disclosures

N. Franceschini reports being a scientific advisor or having membership with WHI Publication and Presentation Committee, WHI Vice-Chair of Ancillary Committee, National Heart, Lung, and Blood Institute (NHLBI) TOPMed kidney working group convener, on the editorial board of American Journal of Physiology: Renal Physiology, and on the editorial board of Contemporary Clinical Trials journal. T. Thornton reports having ownership interest in Regeneron Genetics Center; and being a scientific advisor or membership as associate editor for American Journal of Human Genetics. J. Lash reports being a scientific advisor or having membership with Kidney360. J. Cai reports being a scientific advisor or having membership with the editorial board for Lifetime Data Analysis, Statistics in Biosciences, and Journal of the Royal Statistical Society, Series B. All remaining authors have nothing to disclose. The baseline examination of HCHS/SOL was carried out as a collaborative study supported by contracts from the NHLBI to the University of North Carolina (N01-HC65233), University of Miami (N01-HC65234), Albert Einstein College of Medicine (N01-HC65235), Northwestern University (N01-HC65236), and San Diego State University (N01-HC65237). The following institutes/centers/offices contributed to the first phase of HCHS/SOL through a transfer of funds to the NHLBI: National Institute on Minority Health and Health Disparities, National Institute on Deafness and Other Communication Disorders, National Institute of Dental and Craniofacial Research (NIDCR), National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), National Institute of Neurological Disorders and Stroke, National Institutes of Health Office of Dietary Supplements. The Genetic Analysis Center at Washington University was supported by NHLBI and NIDCR contracts (HHSN268201300005C AM03 and MOD03). Genotyping efforts were supported by NHLBI HSN 26220/20054C, National Center for Advancing Translational Sciences Clinical and Translational Science Institute grant UL1TR000124, and NIDDK Diabetes Research Center grant DK063491.

Funding

This study is supported by National Institutes of Health grants DK11744, MD5012765, HL140385, and HL123677 (to N. Franceschini).

Data Sharing Statement

The Hispanic Community Health Study/Study of Latinos (HCHS/SOL) data supporting the findings of this study is openly available in the dbGap repository at https://www.ncbi.nlm.nih.gov/gap/, under the accession numbers phs000880.v1.p1 and phs000810.v1.p1.

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2021050617/-/DCSupplemental.

Supplemental Appendix 1. Details on the admixture mapping linear and logistic mixed models used for eGFR and CKD analyses.

Supplemental Figure 1. Admixture mapping for eGFR and CKD including diabetes and hypertension as covariates.

Supplemental Figure 2. Conditional admixture mapping analyses for eGFR and CKD including the lead SNP of each locus as covariate.

Supplemental Figure 3. Ancestry background of HCHS/SOL and WHI HL samples.

Supplemental Table 1. Annotation of the SNPs within ancestry-of-origin loci associated with eGFR and CKD.