Re-sequencing of the APOL1-APOL4 and MYH9 gene regions in African Americans does not identify additional risks for CKD progression

Abstract

Background

APOL1 G1 and G2 nephropathy risk variants are associated with non-diabetic end-stage kidney disease (ESKD) in African Americans (AAs) in an autosomal recessive pattern. Additional risk and protective genetic variants may be present near the APOL1 loci since earlier age ESKD is observed in some AAs with one APOL1 renal-risk variant and because the adjacent gene MYH9 is associated with nephropathy in populations lacking G1 and G2 variants.

Methods

Re-sequencing was performed across a ~275 kb region encompassing the APOL1-APOL4 and MYH9 genes in 154 AA cases with non-diabetic ESKD and 38 controls without nephropathy who were heterozygous for a single APOL1 G1 or G2 risk variant.

Results

Sequencing identified 3246 non-coding single nucleotide polymorphisms (SNPs), 55 coding SNPs, and 246 insertion/deletions (InDels). No new coding variations were identified. Eleven variants, including a rare APOL3 Gln⁵⁸Ter null variant (rs11089781), were genotyped in a replication panel of 1571 AA ESKD cases and 1334 controls. After adjusting for APOL1 G1 and G2 risk effects, these variations were not significantly associated with ESKD. In subjects with <2 APOL1 G1 and/or G2 alleles (849 cases; 1139 controls), the APOL3 null variant was nominally associated with ESKD (recessive model, OR 1.81; p=0.026); however, analysis in 807 AA cases and 634 controls from the Family Investigation of Nephropathy and Diabetes (FIND) did not replicate this association.

Conclusion

Additional common variants in the APOL1-APOL4-MYH9 region do not contribute significantly to ESKD risk beyond the APOL1 G1 and G2 alleles.

Introduction

Five common etiologies of end-stage kidney disease (ESKD) are associated with the G1 and G2 coding variants in the apolipoprotein L1 gene (APOL1) on chromosome 22q13 [1–3]. This observation has altered our understanding of the pathogenesis of non-diabetic nephropathy in populations with recent African ancestry. APOL1-associated nephropathies are inherited in an autosomal recessive pattern; risk is present in those inheriting two copies of the G1 and/or the G2 nephropathy variants.[1] Despite this inheritance pattern, individuals with a single APOL1 nephropathy variant initiate renal replacement therapy or develop nephropathy at earlier ages than those without APOL1 nephropathy variants[1;4–6]. In addition, the majority of African Americans (AAs) who inherit two APOL1 nephropathy variants will not develop ESKD [1;7;8].

These findings suggest that additional factors beyond the G1 and G2 variants modulate risk for development of focal segmental glomerulosclerosis (FSGS), HIV-associated nephropathy (HIVAN), focal global glomerulosclerosis with interstitial and vascular changes (FGGS or hypertension-attributed nephropathy), sickle cell nephropathy, and severe lupus nephritis, diseases comprising the APOL1-associated disease spectrum[1;3;4;8–12]. Variation in neighboring genes may contribute to risk, based upon the association of markers in the adjacent non-muscle myosin heavy chain 9 gene (MYH9) with ESKD in European and Asian ancestry populations. These ancestral groups generally lack APOL1 G1 and G2 nephropathy variants[13–17].

In an effort to detect novel risk and protective variants on chromosome 22q13 that impact the risk for nephropathy, Next Generation DNA sequencing (NGS) was performed on a ~275 kb region encompassing the APOL1-APOL4 and MYH9 genes using capture enrichment. The primary goal was to identify genetic variations that may contribute to ESKD risk in cases who were heterozygous for APOL1 G1 or G2 alleles, but which may be absent in subjects who were heterozygous for the APOL1 G1 or G2 alleles and lacked nephropathy.

Materials and Methods

Study Populations

One hundred and ninety-two self-described subjects of African descent born in North Carolina, South Carolina, Virginia, Georgia, or Tennessee formed the Next Generation Sequencing (NGS) discovery panel (Table 1). Cases had ESKD attributed to FSGS, FGGS, HIVAN, chronic glomerulonephritis (without a kidney biopsy), hypertension, or unknown causes and were recruited from dialysis clinics. Patients with ESKD ascribed to diabetes mellitus, urologic disease, surgical nephrectomy, IgA nephropathy, polycystic kidney disease, membranous glomerulonephritis, membranoproliferative glomerulonephritis and other etiologies of ESKD were excluded. Controls had a serum creatinine concentration <1.5 mg/dl (men) or <1.3 mg/dl (women), with a urine albumin:creatinine ratio <30 mg/g. All 192 subjects (154 cases and 38 controls) in this discovery panel were heterozygous for either the G1 or G2 APOL1 risk allele with equal proportions of each risk allele in cases and controls.

Table 1.

	Wake Forest Discovery Samples			Wake Forest Complete Sample Set*			FIND Replication
	ESKD Cases (N=154)	Controls (N=38)	Total (N=192)	ESKD Cases (N=1427)	Controls (N=1304)	Total (N=2731)	ESKD Cases (N=394)	Controls (N=660)	Total (N=1054)
Female (%)	46.8	89.5	55.2	43.7	51.7	47.5	33.0	61.8	51.0
Age (years)	55.9±13.7	45.9±12.1	54.0±14.0	53.7±14.3	49.2±12.2	51.6±13.5	47.5±11.7	46.6±11.8	47.0±11.8
Duration ESKD (years)	4.7±5.1	-	-	5.6±5.7	-	-	-	-	-
BMI (kg/m2)	27.0±6.3	32.5±8.7	28.2±7.2	27.5±7.1	29.6±7.2	28.5±7.3	-	-	-
BUN (mg/dL)	-	10.7±3.4	-	-	13.7±5.2	-	-	-	-
Serum creatinine (mg/dL)	-	0.8±0.1	-	-	1.0±0.4	-	-	-	-

^*includes the 192 samples from the Discovery set

The complete Wake Forest study population included in the full genotyping set consisted of a total of 1,571 subjects of African descent having non-diabetic ESKD and 1,342 controls without kidney disease (Table 1). Because the 192 participants included in the sequencing discovery panel were a subset derived from the complete genotyping set and constituted a significant number of heterozygotes, these samples were included in the complete analysis in order to limit possible type 1 and type 2 errors that could have occurred due to missing heterozygotes. The complete analysis set of cases met identical disease criteria as NGS study cases. All controls in the complete analysis set denied knowledge of kidney disease and met the serum creatinine concentration cut-offs employed in the NGS analysis; albeit, some subjects lacked serum creatinine concentrations and the majority lacked measures of albuminuria. Therefore, it is possible that a small number of controls had mild or undiagnosed nephropathy, an effect that would reduce power for confirming genetic associations. The Family Investigation of Nephropathy and Diabetes (FIND) [18,19] served as a second replication population, consisting of 807 non-diabetic ESKD cases and 634 non-nephropathy controls with complete phenotypic data (Table 1).

Next Generation Sequencing

For re-sequencing, a 275 kb region of Chr22:36529953-36804892 (hg19) was designated as the capture region. The custom enrichment probes were designed and synthesized using TargetSeq™ system (Life Technologies; Grand Island, New York). Repetitive and low complexity sequences were removed from the enrichment design using RepeatMasker (www.repeatmasker.org). Genomic DNA was sheared to an average of 180 bp using a Covaris S220 focused ultrasonicator and DNA libraries constructed using the SOLiD Fragment Library kit on an automated SOLiD Library Builder (Life Technologies; New York). Samples were barcoded to allow multiplex sequencing. The fragment libraries were enriched and amplified according to the TargetSeq protocol and then sequenced on an ABI 5500XL at 50–75 bp read lengths.

Bioinformatic and Statistical Analysis

Data generated on the ABI SOLiD (.xseq) were analyzed by first pass using LifeScope (Grand Island, New York) Genomic Analysis Software (v 2.5) to generate BAM files against the reference human genome (GRCh37/hg19). A BED file generated for the capture enrichment design was used to filter fragments aligning to the region Chr22:36529953-36804892. BAM files were used for secondary alignment and variant calling. Variants were called using GATK 'best practices'. Multiple-sample variant calling was done using GATK UnifiedGenotyper (v2.6) after de-duplication (PicardTools) and realignment around indels (GATK IndelRealigner).

A logistic regression model was computed to test for association between an individual single nucleotide polymorphism (SNP) and ESKD. In the NGS discovery cohort, the logistic model was adjusted for age and gender as implemented in SNPGWA (www.phs.wakehealth.edu). By design, the discovery cohort had one copy of either the APOL1 G1 or G2 risk variant. In the validation analysis, comprised of all discovery and additional replication samples, the logistic regression model was computed adjusting for age, gender, global ancestry, and APOL1 G1/G2 risk.

Genotyping

Two SNPs in APOL1, the G1 nephropathy risk allele (rs73885319; rs60910145) and an insertion/deletion polymorphism for the G2 risk allele (rs71785313), were genotyped in subjects using a custom assay designed in the Center for Genomics and Personalized Medicine Research at Wake Forest School of Medicine on the Sequenom platform (San Diego, California) [20]. Of the 270 blind duplicates that were genotyped, only 3 were found to be discordant for a 99% concordance rate. The APOL1 G1 and G2 genotype calls were visually inspected for quality control.

SNPs identified as being associated with ESKD (Supplementary Table 1) in the first pass association analysis in the 192 subject NGS discovery panel and one SNP in APOL3 (rs11089781) encoding the APOL3 null allele (Gln⁵⁸Ter) were genotyped using the Sequenom platform, as described above. Ninety blind duplicates were genotyped and 2 were discordant (98% concordance rate). Admixture estimates were computed based on the frequencies of 106 bi-allelic ancestry informative markers distributed throughout the genome. The maximum likelihood approach as coded in the package Frequentist estimation of individual ancestry proportion (FRAPPE) was used to obtain the proportion of African and European ancestry in each individual [21]. Genotype data at these markers were obtained from 44 HapMap Yoruba individuals (YRI) and 39 European American controls as anchors and provided starting values for the Expectation-Maximization algorithm in FRAPPE.

Results

Deep Sequencing Results

Table 1 contains demographic characteristics for the NGS (discovery) and the full sample of Wake Forest ESKD cases and non-nephropathy controls that were used for replication. In the NGS discovery study, our goal was to identify new genetic variants that may contribute to ESKD risk in subjects with ESKD that had inherited only a single copy of either the G1 or G2 APOL1 risk variant (recessive inheritance pattern). A matched set of subjects without nephropathy and a single copy of either the G1 or G2 APOL1 risk variants served as the comparison group. For subsequent validation by genotyping the full sample of ESKD cases and non-nephropathy controls, a mixture of genotypes representative of these populations was included. Cases with ESKD generally initiated renal replacement therapy in the late fourth or fifth decade of life, and were compared to control subjects of similar age. There were disproportionately greater numbers of females in the NGS control sample relative to NGS cases and the full case/control sample. Body mass index (BMI) was also higher in the non-nephropathy controls, likely reflecting poorer nutrition and inflammation observed with advanced nephropathy.

The mean sequencing depth across the 275 kb region was >100X for the 192 subjects in the NGS discovery panel. The median fraction of mapped reads was 94%. SNP calling using GATK resulted in 3300 SNPs of which 69.8 % were found in dbSNP. Fifty-five coding changes were identified across all five genes, none of which were novel. The transition/transversion ratio was 2.04. One hundred twenty SNPs previously genotyped were detected with Cohen's kappa 0.83, representing high agreement. Two hundred and forty six insertion-deletions (InDels) were also identified of which 60% were found in dbSNP. The APOL1 G1 risk alleles (rs73885319; rs60910145) and G2 risk allele insertion/deletion (rs71785313) were used as internal controls to validate re-sequencing results in the NGS discovery panel. APOL1 G1 and G2 genotypes were 100% concordant with previous genotyping in all 192 individuals comprising the NGS discovery sample.

Association Analysis

After adjusting for age and gender and compensating for linkage disequilibrium (LD), 10 variants from the top 25 hits in the dominant and additive models were chosen for genotyping in the complete Wake Forest (validation) sample (Supplementary Table S1). Of these 10 variants, two were found to be monomorphic in the validation set; both of these SNPs had been identified as novel polymorphisms in the deep sequencing analysis. After adjusting for the effects of the APOL1 G1 and G2 risk alleles and without multiple comparison adjustment, only the APOL3 SNP rs132672 approached statistical significance (p=0.053) in the complete validation set under a recessive model.

A second analysis focused on coding variants that had been identified in the APOL1-APOL4-MYH9 capture region. A summary of coding variants identified in each gene in the discovery panel cases and controls is shown (Table 2). No novel coding variations were identified in the APOL1-APOL4-MYH9 genes. Based on allele frequency, a single APOL3 SNP (rs11089781) encoding the APOL3 nonsense allele (Gln⁵⁸Ter) was associated with ESKD in the NGS discovery panel (p=0.02) with 30% of cases and 15% of controls having the null allele. This SNP was genotyped and found to be associated with ESKD in the full case-control validation sample after adjustment for APOL1 G1 and G2 compound risk (Table 3, recessive model, odds ratio 1.81; p=0.026). The recessive mode of inheritance was consistent with expectations for a null allele in a highly inducible gene. No other coding variants had strong evidence for association with ESKD. However, further genotyping of rs11089781 in 807 FIND non-diabetic ESKD cases and 634 FIND non-nephropathy controls did not support association (data not shown).

Table 2.

	Chr 22 position	rs ID	Reference allele	Amino Acid Change	Cases MAF	Controls MAF
APOL1	36657740	rs41297245	G	Gly112Arg	0.02	0.025
	36661330	rs2239785	G	Glu166Lys	0.27	0.32
	36661409	rs116136671	A	Asn192Ser	0.044	0.01
	36661536	rs136174	C	Ala234Ala	0.96	0.96
	36661566	rs136175	G	Met244Ile	0.96	0.96
	36661646	rs136176	G	Arg271Lys	0.96	0.96
	36661674	rs73885316	C	Asn280Lys	0.03	0.01
	36661691	rs73403889	G	Gly286Asp	0.003	-
	36661791	rs141788376	C	Arg319Arg	0.003	-
	36661796	rs150588135	G	Arg321Gln	0.003	-
	36661810	rs139590386	A	Ile326Val	0.003	-
	36661842	rs136177	G	Arg366Arg	0.05	0.04
	36661891	rs16996616	G	Asp353Asn	0.08	0.04
	36661906	rs73885319	A	Ser358Gly	0.25	0.28
	36661940	rs143845266	T	Val369Glu	-	0.01
	36662034	rs60910145	T	Ile400Met	0.25	0.28
APOL2	36623671	rs73885303	G	Ala265Thr	0.006	0.01
	36623902	rs148745051	G	Gly188Ser	0.003	-
	36623920	rs7285167	C	Arg182Cys	0.63	0.61
	36623980	rs15125817	C	Ala162Val	0.006	-
	36624231	rs118097350	C	Arg78Lys	0.01	-
APOL3	36537431	rs141045937	C	Ala342Ala	0.003	-
	36537500	rs132618	A	Gly319Gly	0.13	0.16
	36537650	rs147303125	A	Val269Val	0.003	-
	36537725	rs61731692	G	Tyr244Tyr	0.08	0.06
	36537740	rs138851629	G	Ile239Ile	0.003	-
	36537763	rs61741884	C	Val232Met	0.006	-
	36537798	rs116147257	G	Thr220Ile	0.03	0.10
	36537893	rs3827346	A	Asn188Asn	0.02	0.04
	36538053	rs6000152	G	Ala135Val	0.075	0.14
	36541536	rs141517147	G	Ala112Val	0.003	-
	36556729	rs79419411	T	Arg71Gly	0.11	0.15
	36556768	rs11089781	G	Gln58Ter	0.30	0.15
	36556823	rs132653	G	Ser39Arg	0.45	0.48
	36556882	rs148712729	T	Cys13Trp	0.003	0.01
APOL4	36587154	rs111781032	C	Thr339Asn	0.1	0.06
	36587202	rs6000172	C	Ser323Leu	0.74	0.65
	36587215	rs61730820	G	Glu319Lys	0.013	-
	36587223	rs6000173	C	Ala316Glu	0.74	0.65
	36587279	rs6000174	T	Leu297Leu	0.74	0.65
	36587310	rs76473883	G	Arg287Gln	0.013	0.04
	36587311	rs150226555	C	Arg287Trp	-	0.025
	36587346	rs61730819	G	Arg275His	0.1	0.06
	36587486	rs2227167	T	Ser228Ser	0.74	0.65
	36587511	rs2227168	G	Arg220His	0.75	0.68
	36587682	rs146308862	G	Gly163Glu	-	0.01
	36587704	rs132700	A	Met156Val	0.32	0.44
	36587886	rs78582347	G	Met94Val	0.1	0.06
	36587952	rs2227169	G	Glu72Glu	0.73	0.64
	36591380	rs2007468	T	Phe62Phe	0.87	0.78
	36591402	rs140324556	G	Ser55Thr	0.003	-
	36591475	rs77639244	C	Arg31Cys	0.1	0.06
	36598049	rs80587	G	Val12Leu	0.65	0.56
	36598058	rs132736	A	Ile9VAl	0.64	0.50
	36598081	rs6000181	T	Met1Leu	0.05	0.01

Table 3.

	N (case/controls)	MAF(case/control)	Dominant model		Additive model		Recessive model
			p-value	OR (95%CI)	p-value	OR (95%(CI)	p-value	OR (95% CI)
All Subjects	1559/1332	0.30/0.22	0.31	1.09(0.92, 1.29)	0.18	1.1(0.96,1.25)	0.2	1.25(0.89,1.76)
Subjects with 0 or 1 APOL1 G1 or G2 allele	849/1139	0.21/0.19	0.38	1.09(0.90,1.32)	0.13	1.14(0.96,1.35)	0.022	1.83(1.09,3.08)
Subjects with 0 APOL1 G1 or G2 allele	374/562	0.11/0.12	0.24	0.81(0.58,1.15)	0.40	0.87(0.64,1.20)	0.41	1.6(0.52,4.90)
Subjects with 1 APOL1 G1 or G2 allele	463/574	0.28/0.26	0.38	1.12(0.87,1.45)	0.15	1.17(0.94,1.46)	0.064	1.75(0.97,3.18)
Subjects with 2 APOL1 G1 or G2 allele	722/196	0.40/0.40	0.64	1.08(0.78,1.50)	0.79	1.03(0.82,1.30)	0.92	0.98(0.64,1.50)
Subjects with 1 G1 allele	166/219	0.26/0.21	0.09	1.46(0.95,2.25)	0.084	1.41(0.96,2.08)	0.49	1.55(0.44,5.46)
Subjects with 1 G2 allele	297/355	0.30/0.28	0.75	0.95(0.70,1.31)	0.69	1.06(0.81,1.37)	0.10	1.76(0.90,3.46)

^*P-values not adjusted for multiple comparisons

OR, odds ratio

Discussion

This study is the first to describe NGS based re-sequencing of the ~275 kb region encompassing the APOL1-4-MYH9 region. Based on our strategy of sequencing affected and unaffected subjects heterozygous for either the APOL1 G1 or G2 risk allele, our data suggest that no additional common genetic variations in this ~275 kb contribute significantly to ESKD risk beyond the APOL1 G1 and G2 alleles. Although we did identify the common null variant rs11089781 (Gln>Ter) as having a potential recessive effect on ESKD risk within our primary screening population, the inability to reproduce this association in the independent FIND cohort suggests that the initially observed association could be due to a false positive result, a cohort-specific observation, or different diagnoses in FIND participants [18,19]. Given the frequency and the potential functionality of this null variant, analysis in other African ancestry populations with nephropathy is warranted, as are tests to determine how this APOL3 null variation functions in ESKD.

The data in this study are consistent with a recent report showing that re-sequencing of all of the APOL1 exons in 1437 subject of African and European ancestry did not identify any new common or rare coding variants in APOL1 that contribute to additional ESKD risk [22]. Of the 23 coding variants identified in that study, we were able to identify 15 of the same variants. Of the remaining 8 coding variants, our study was not powered (N=192) to detect these rare variants (frequency <1% in all cases) as has been the case in other re-sequencing studies [23] We did, however, identify one additional rare variant (rs143845266) not identified in the APOL1 re-sequencing study. Because the APOL1 re-sequencing study focused only on exons, we are unable to compare our NGS sequencing data obtained in the non-coding regions (introns and promoter region) and we could not compare our InDel sequencing data since no comprehensive InDel sequencing data was reported in the APOL1 focused re-sequencing study. However, combining our analysis with this APOL1 focused study, the cumulative data support that additional common coding variations in APOL1 are highly unlikely to contribute significantly to additional nephropathy risk.

The present analysis utilized a capture bait strategy to enrich for target regions and sequencing read lengths of 50–75 bp, thus our data were not ideal for properly evaluating copy number variations or large structural variations in the APOL1-4-MYH9 region. There is evidence that the APOL1 region could be a hotspot for genetic recombination based on the mapping of crossover recruiting motifs within surrounding the APOL1 gene [24]. Additionally, there are low divergent sequence regions within APOL4 and APOL1, which could be evidence of a recombination event or a large sequence inversion involving these two genes. A recent analysis of 123 subjects with kidney disease and 255 control subjects found that in the subjects with nephropathy, there was a ~4% APOL1 duplication rate compared to a APOL1 duplication rate of only ~0.8% in the controls [25]. Thus it is possible that in a small subset of subjects heterozygous for the APOL1 G1 or G2 allele, individuals could carry two copies of a risk allele. Whether the second copy is expressed or functional isn’t known, but it does suggest that in some subjects a second copy of an APOL1 G1 or G2 allele could set up an imbalance in gene expression where the dose of the mutant copies of APOL1 protein could be higher in subjects with an APOL1 CNV.

Given that additional common genetic variants in the APOL1-4-MYH9 region that contribute significantly to ESKD risk apparently do not exist, the possibility remains that altered gene expression or epigenetics could also be in play. In the case of APOL1, at least 4 splice variants are identified in genomic databases (UCSC Genome Browser: genome.ucsc.edu). One report identified that differential splicing out of exon 4 regulates the toxicity of the G0 (non-G1 or non-G2) APOL1 allele [26]. Based on a review of APOL1 exon 4, only 4 extremely rare coding variants have been identified in this exon. Neither this study nor our detailed APOL1 re-sequencing study identified these 4 rare variants, suggesting that these rare variants are most likely novel germline variants, and that this exonic region is under conservation pressure, critical for proper function of APOL1.

In summary, the APOL1 G1 and G2 renal risk variants are important determinants of ESKD susceptibility. Based on the present analyses and other reports, these two variants appear to be responsible for the bulk of APOL1 derived risk for progressive non-diabetic nephropathy. While additional coding changes occur in the genes APOL1, APOL2, APOL3, and APOL4, none of the variants identified in this study significantly contributed to nephropathy risk. Although this doesn’t exclude the possibility that these coding variants may affect gene function, and thus be minor contributors to disease risk, the data support other uncharacterized genetic or epigenetic factors as likely responsible for additional ESKD risk residing in the chromosome 22q13 region.