Clin Exp Nephrol DOI 10.1007/s10157-014-0969-y
ORIGINAL ARTICLE
Role of the p.E66Q variant of GLA in the progression of chronic kidney disease Hirofumi Watanabe • Shin Goto • Akinori Miyashita Hiroki Maruyama • Minako Wakasugi • Akio Yokoseki • Ryozo Kuwano • Ichiei Narita
•
Received: 4 November 2013 / Accepted: 25 March 2014 Ó Japanese Society of Nephrology 2014
Abstract Background The p.E66Q variant of the a-galactosidase A gene (GLA) is frequently found during screening for Fabry disease in dialysis patients in Japan. However, recent reports suggest that the p.E66Q variant is not a diseasecausing mutation but is a risk factor for cerebral smallvessel occlusion. To evaluate the role of the p.E66Q in the progression of renal diseases, we performed a genetic association study in patients with chronic kidney disease (CKD). Methods In this study, we enrolled 1651 chronic hemodialysis and 941 non-dialysis patients who attended medical institutions in the Niigata Prefecture, Japan. The frequency of the p.E66Q allele was compared between hemodialysis and non-dialysis patients, with data from a previously published study of Japanese male newborns. In addition, we compared estimated glomerular filtration rates
H. Watanabe S. Goto I. Narita (&) Division of Clinical Nephrology and Rheumatology, Niigata University Graduate School of Medical and Dental Sciences, 1-757 Asahimachi-dori, Chuo-ku, Niigata, Niigata 951-8510, Japan e-mail:
[email protected] A. Miyashita R. Kuwano Department of Molecular Genetics, Brain Research Institute, Niigata University, Niigata, Japan H. Maruyama Department of Clinical Nephroscience, Niigata University Graduate School of Medical and Dental Sciences, Niigata, Japan M. Wakasugi A. Yokoseki Center for Inter-organ Communication Research, Niigata University Graduate School of Medical and Dental Sciences, Niigata, Japan
(eGFR) in the presence or absence of the p.E66Q variant in non-dialysis patients. Results Of the 2233 alleles in hemodialysis and 1447 alleles in non-dialysis patients, 21 and nine harbored p.E66Q, respectively. However, p.E66Q allele frequencies did not differ between the two patient groups (0.90 versus 0.62 %, P = 0.35), and no significant difference in p.E66Q allele frequency was observed between male hemodialysis patients and the general Japanese population (0.52 versus 0.63 %, P = 0.67). Moreover, eGFR did not significantly differ between non-dialysis patients with the p.E66Q variant and patients with the wild-type allele (65.5 ± 10.7 versus 62.7 ± 16.6 mL/min/1.73 m2, P = 0.69). Conclusion This study indicated that the p.E66Q variant of GLA does not affect the progression of CKD. Keywords a-Galactosidase A Fabry disease Genetic association study Hemodialysis Polymorphism SNP genotyping
Introduction Fabry disease is an X-linked lysosomal storage disorder caused by abnormalities in the a-galactosidase A gene (GLA), which leads to a deficiency in a-galactosidase A activity [1]. Absent or deficient activity of the lysosomal exoglycohydrolase, a-galactosidase A, results in progressive accumulation of globotriaosylceramide (GL-3 or Gb3) and related glycosphingolipids within lysosomes [2]. With advancing age, progressive accumulation of lysosomal GL3, particularly in the vascular endothelium, leads to renal failure, vascular disease of the heart and brain, and premature death [3].
123
Clin Exp Nephrol
The advent of an effective treatment, enzyme replacement therapy (ERT), has increased the importance of early detection of Fabry disease [4]. Diagnosing Fabry disease is also important because it permits family studies to identify other affected relatives and initiation of genetic counseling and therapeutic intervention [3]. However, disease presentation is generally heterogeneous, and the diagnosis of Fabry disease can therefore be difficult and delayed, and require confirmation by genetic analysis. Fabry disease can be caused by a variety of missense or nonsense point mutations, splicing mutations, small deletions or insertions, or large deletions [2], with more than 600 distinct mutations of GLA having been identified. The p.E66Q variant of GLA is a G to C transversion at nucleotide 196 (cDNA) of exon 2 and has been frequently found during screening for Fabry disease in chronic dialysis patients in Japan [3–7]. The activity of a-galactosidase A in patients with the p.E66Q variant has been reported to be lower than that in patients with the wild-type allele [8, 9]. On the other hand, recent pathological investigations have suggested that the p.E66Q variant is not a diseasecausing mutation [9, 10]. As mentioned earlier, the p.E66Q variant causes biochemical changes such as lower lysosome function. These findings therefore suggest that the p.E66Q variant may be a risk factor for the development and progression of some chronic diseases, even though it may not be associated with classical Fabry disease. To evaluate the role of the p.E66Q variant in the progression of renal disease, we performed a genetic association study in patients with chronic kidney disease (CKD).
Materials and methods Patients Of the 2592 participants enrolled in the study, 1651 were chronic hemodialysis patients who attended dialysis institutions in the Niigata Prefecture, Japan, and 941 were nondialysis patients who were registered with the Project in Sado for Total Health (PROST) between June 2008 and July 2012. PROST is a prospective observational cohort study that is currently being conducted at Sado General Hospital, Niigata Prefecture (http://square.umin.ac.jp/prost/). No study participants were diagnosed with Fabry disease and all participants provided informed consent. The institutional review board of Niigata University Graduate School of Medical and Dental Sciences, Niigata, Japan approved the study (approval no. 511). Genotyping of the p.E66Q variant Peripheral venous blood samples were obtained for DNA isolation, with genomic DNA being extracted using
123
standard methods. SNP genotyping of individual samples was performed using TaqMan technology. The TaqMan SNP genotyping assays were purchased from Applied Biosystems (CA, USA) and used for allelic discrimination in 384-well plates. The sequences of the probes have been reported in a previous study [8]. A 20-ng aliquot of each DNA sample was pipetted into a 384-well plate, followed by drying on a clean bench. The dried DNA was then mixed with 700 lL of 29 TaqMan Universal Master Mix II, 17.5 lL of 409 TaqMan Assay (final concentration 0.59), and 682.5 lL of nuclease-free water on a 384-well optical PCR plate. The PCR conditions were as follows: 95 °C for 10 min, followed by 48 cycles at 92 °C for 15 s, and 58 °C for 1 min. After PCR, genotyping was performed on an Applied Biosystems 7900HT Fast Real-Time PCR System using 384-well plates. The SDSv2.4 software was used for amplification detection. Genetic association study Continuous variables are presented as mean ± standard deviation (SD) or as medians and interquartile range, whereas categorical variables are presented as numbers and ratios (%). The allele frequencies of the p.E66Q variant in chronic hemodialysis and non-dialysis patients were calculated and compared using the v2 test. We also used the v2 test to compare the allele frequency in chronic hemodialysis and non-dialysis patients grouped according to sex. A previous study estimated the frequency of the p.E66Q variant in the Japanese general population by testing 5051 consecutive male neonates. Thirty-two hemizygous male neonates were identified, with the frequency of the p.E66Q allele being 0.63 % [11]. We compared the allele frequencies of hemodialysis and non-dialysis patients with that of the previously reported Japanese male population. We also compared the renal function in non-dialysis patients with or without the p.E66Q variant using the estimated glomerular filtration rate (eGFR). The eGFR values were compared using the Student’s t test. P values \0.05 were considered statistically significant.
Results Patient characteristics The baseline characteristics of chronic hemodialysis and non-dialysis patients are shown in Table 1. More than half of the patients started receiving dialysis due to chronic glomerulonephritis. No patient had been diagnosed with Fabry disease. Of the 941 non-dialysis patients, 577 (61.3 %) had an eGFR [60 mL/min/1.73 m2 and 364 (38.7 %) had an eGFR \60 mL/min/1.73 m2.
Clin Exp Nephrol Table 1 Baseline characteristics of the patients
Total (n = 1651)
Male (n = 969)
Female (n = 682)
Characteristics of the chronic hemodialysis patients Age (year)
61.6 ± 13.1
60.5 ± 12.9
63.1 ± 13.2
Age dialysis started (year)
53.8 ± 16.0
52.6 ± 16.0
55.5 ± 16.0
Duration of hemodialysis (month), median (25, 75 %)
64 (28, 137)
64 (27, 139)
64.5 (29, 133)
Cause of ESKD (%) Chronic glomerulonephritis
973 (58.9)
577 (59.5)
396 (58.1)
Diabetes mellitus Nephrosclerosis
359 (21.7) 101 (6.1)
233 (24.0) 52 (5.4)
126 (18.5) 49 (7.2)
Polycystic kidney disease
75 (4.5)
36 (3.7)
39 (5.7)
Uropathy
41 (2.5)
18 (1.9)
23 (3.4)
Rapidly progressive glomerulonephritis
14 (0.8)
10 (1.0)
4 (0.6)
Hyperuricemia
15 (0.9)
8 (0.8)
7 (1.0)
Systemic lupus erythematosus
10 (0.6)
2 (0.2)
8 (1.2)
Others
53 (3.2)
28 (2.9)
25 (3.7)
Unidentified
10 (0.6)
5 (0.5)
5 (0.7)
Total (n = 941)
Male (n = 435)
Female (n = 506)
70.1 ± 9.3
69.4 ± 9.2
70.7 ± 9.4
Height (cm)
155.5 ± 9.3
162.7 ± 6.5
149.3 ± 6.5
Weight (kg)
58.8 ± 12.1
64.0 ± 11.2
54.4 ± 11.2
Systolic blood pressure (mmHg)
136.6 ± 19.8
138.9 ± 19.5
134.7 ± 19.9
Diastolic blood pressure (mmHg)
74.7 ± 12.1
76.4 ± 12.4
73.2 ± 11.7
S-Cre (mg/dL)
0.88 ± 0.38
1.02 ± 0.42
0.75 ± 0.28
Characteristics of the non-dialysis patients Age (year)
2
eGFR (mL/min/1.73 m )
Data are mean ± standard deviation; median (interquartile range); number (ratio) ESKD end-stage kidney disease, S-Cre serum creatinine
62.7 ± 16.6
62.8 ± 17.8
62.7 ± 15.5
C90, n (%)
37 (3.9)
21 (4.8)
16 (3.2)
60–89, n (%)
540 (57.4)
241 (55.4)
299 (59.1)
45–59, n (%)
241 (25.6)
105 (24.1)
136 (26.9)
30–414, n (%)
91 (9.7)
50 (11.5)
41 (8.1)
15–29, n (%)
27 (2.9)
16 (3.7)
11 (2.2)
\15, n (%)
5 (0.5)
2 (0.5)
3 (0.6)
Genotyping assay and allele frequencies of the p.E66Q variant We identified the 20 hemodialysis patients with the p.E66Q variant (Table 2). There were a variety of causes of their end-stage kidney disease, and no clear trend was observed for age at sampling, age dialysis started, or duration of hemodialysis. Of the 20 patients, only one heterozygous female patient (patient 10) had a renal biopsy performed 9 years before starting dialysis that showed minor glomerular abnormalities and scattered tubulo-interstitial nephritis, with no accumulation of GL-3 in the renal tissues. Of the 2333 alleles identified in the chronic hemodialysis patients, 21 alleles were found to harbor the p.E66Q variant. Five alleles were hemizygotes from males,
whereas 14 heterozygotes and 1 homozygote alleles were from females. The allele frequency in the chronic hemodialysis patients was 0.90 % (95 % confidence interval 0.558–1.373). Of the 1447 alleles in the non-dialysis patients, 9 were found to harbor the p.E66Q variant (4 hemizygote alleles from males and 5 heterozygotes alleles from females). The allele frequency was 0.62 % (95 % confidence interval 0.279–1.154) (Table 3). Comparison of the p.E66Q allele frequency among subjects There was no significant difference in the p.E66Q allele frequency in the chronic hemodialysis patients compared with the non-dialysis patients (P = 0.35). In males, the
123
Clin Exp Nephrol Table 2 Characteristics of the 20 chronic hemodialysis patients with the p.E66Q variant
ESKD end-stage kidney disease, CGN chronic glomerulonephritis, DM diabetes mellitus, SLE systemic lupus erythematosus, PKD polycystic kidney disease, HU hyperuricemia, GS glomerular sclerosis, RPGN rapidly progressive glomerulonephritis, CPN chronic pyelonephritis
Patient
Sex
Allele
Age (year)
Age dialysis started (year)
1
Male
Hemizygote
36
29
83
CGN
2
Male
Hemizygote
49
22
324
CGN
3
Male
Hemizygote
55
52
43
DM
4
Male
Hemizygote
58
56
17
DM
5
Male
Hemizygote
59
50
115
CGN
6
Female
Heterozygote
33
29
44
SLE
7
Female
Heterozygote
40
33
82
CGN
8
Female
Heterozygote
47
45
20
CGN
9
Female
Heterozygote
48
47
16
CGN
10
Female
Heterozygote
49
41
96
CGN
11 12
Female Female
Heterozygote Heterozygote
53 56
49 42
53 170
PKD HU
13
Female
Heterozygote
57
49
99
CGN
14
Female
Heterozygote
59
57
31
DM
15
Female
Heterozygote
65
63
24
CGN
16
Female
Heterozygote
75
72
38
DM
17
Female
Heterozygote
78
77
9
GS
18
Female
Heterozygote
78
73
60
19
Female
Heterozygote
79
76
32
CGN
20
Female
Homozygote
60
51
112
CPN
Table 3 Result of the p.E66Q genotyping assay
Homozygote
p.E66Q heterozygote
Hemizygote
Duration of hemodialysis (month)
Wild
Cause of ESKD
RPGN
p.E66Q allele frequency
Allele frequency 95 % CI
Hemodialysis patients Male (n = 969)
–
–
5
964
0.52
Female (n = 682)
1
14
–
667
1.17
Male (n = 435)
–
–
4
431
0.92
Female (n = 506)
0
5
–
501
0.49
–
–
32
5019
0.63
0.90
0.558–1.373
0.62
0.279–1.154
Non-dialysis patients Frequency is rate (%). Control is adapted from the Japanese newborn males in [11] CI confidence interval
Control Male (n = 5051)
p.E66Q allele frequency in the chronic dialysis patients was 0.52 % and in the non-dialysis patients was 0.92 %. This difference in frequency between the two groups was not significant (P = 0.38). Similarly, we found no significant difference in the p.E66Q allele frequency for females between the chronic dialysis and non-dialysis patients (1.17 versus 0.49 %, P = 0.08). Comparisons between chronic hemodialysis patients and previously reported Japanese neonates showed no significant difference in allele frequencies in males (0.52 versus 0.63 %, P = 0.67). Similarly, no significant difference in allele frequency was observed between male non-dialysis patients and Japanese neonates (0.92 versus 0.63 %, P = 0.69).
123
0.434–0.893
Table 4 The eGFR in the non-dialysis patients by p.E66Q allele p.E66Q
Wild
P value
Male (n = 435)
64.5 ± 13.4
62.8 ± 17.8
0.57
Female (n = 509)
66.3 ± 9.7
62.6 ± 15.5
0.70
Total (n = 941)
65.5 ± 10.7
62.7 ± 16.6
0.69
2
Data are mean ± SD (mL/min/1.73 m )
Comparison of renal function in the non-dialysis patients As shown in Table 4, there was no significant difference in the eGFR of non-dialysis patients with or without the
Clin Exp Nephrol
p.E66Q variant when examined either as the whole study cohort or in males and females.
Discussion This study showed that the frequency of the p.E66Q allele in chronic hemodialysis patients was not significantly different from that measured in non-dialysis patients and the Japanese general population. This finding indicates that the variant does not affect the development and progression of renal disease. This is the first report investigating the genetic influence of the p.E66Q variant of GLA on CKD. To assess adaptation to specific treatments such as ERT, it is important to determine whether the p.E66Q variant is a disease-causing mutation or a risk factor of any disease. Initially, the p.E66Q variant was considered to be a pathogenic mutation that caused later-onset Fabry disease [3, 12, 13] because the first patient reported to harbor p.E66Q presented with classic Fabry disease due to another mutation (p.R112C) on the same allele [14]. However, some recent reports have suggested that the p.E66Q is not a disease-causing mutation but a functional polymorphism [8– 10]. Histopathological studies of biopsied skin tissues [9], myocardial tissues [10], or renal tissues [4] from patients with the p.E66Q variant have not revealed any changes specific to Fabry disease. In addition, no accumulation of GL-3 in cultured fibroblasts or increased level of plasma globotriaosylsphingosine was observed in patients with the p.E66Q variant [9]. Consistent with these reports, we observed no lesion suggesting GL-3 accumulation in the renal tissue of a patient with a heterozygous p.E66Q variant in this study. All p.E66Q positive cases including this female patient did not present clinically with a Fabry disease phenotype. These data suggested that the p.E66Q variant may be a functional polymorphism, though the possibility that the variant is involved in the development of renal variant type of Fabry disease, could not be completely excluded because only a few patients with p.E66Q were examined for pathological lesion using renal biopsy. In previous studies of Japanese dialysis patients, p.E66Q genotyping was performed after enzymatic screening [3–7, 13, 15, 16]. Because the activity of the a-galactosidase A enzyme with the p.E66Q variant is lower, but may remain in the normal range in some patients, the frequency of the p.E66Q variant was likely to have been underestimated in those studies. For example, a recent study that enzymatically screened 21170 neonates followed by DNA sequencing found only 2 boys harbored the p.E66Q [17]. However, in the present study, we showed that the actual frequency of the p.E66Q allele was considerably higher than that reported in previous reports that indicated the variant was rare and caused disease.
On the other hand, it has been proposed that the p.E66Q variant may be a risk factor for some chronic diseases. Several examples of rare and low-frequent variants with large effect sizes have been reported in complex traits [18]. In fact, we showed the frequency of the p.E66Q variant was less than 1 % in this study. Previous studies have shown that residual enzyme activity in white blood cells of subjects harboring the p.E66Q was functionally lower than those with the wild-type genotype [8–10]. Maruyama et al. [4] also described 2 patients with the p.E66Q variant with a detectable globotriaosylsphingosine level. We hypothesized that the p.E66Q variant may be a risk factor for CKD, but observed no association between the p.E66Q variant and end-stage kidney disease and/or renal function. We could not obtain evidence that the variant was a risk factor for CKD progression. However, a recent report demonstrates an association between p.E66Q and cerebral smallvessel occlusion in elderly Japanese males [11]. In the present study, we determined whether the p.E66Q variant could affect the course of CKD progression, but collected only incomplete data on cerebrovascular events in our patients. Hence, we have not tested the association between the p.E66Q variant and cerebrovascular or cardiovascular diseases in patients with CKD. There were some limitations in this study. First, we selected only consenting patients who were in stable condition, leading to a bias that may have affected the results. In addition, we used a cross-sectional design that excluded the p.E66Q variants who may have died from its effects. Therefore, prospective cohort studies are required to accurately predict the role of the p.E66Q variant in the progression of renal diseases. Moreover, because the allele frequency of the p.E66Q variant was less than 1 %, which is categorized as low-frequency variant, the present data may lack the ideal sample size and statistical power that is required to determine disease associations of this rare variant [19]. Thus, future studies require more numbers of CKD patients to demonstrate the full effects of the p.E66Q variant. In conclusion, this genetic association study indicated that the p.E66Q variant of GLA does not affect the progression of CKD. With striking recent developments of next generation sequencing, sequencing data for some patients can be clinically applied at the bedside. Thus, detailed elucidation of genotype–phenotype correlations for Fabry disease will contribute to the understanding of disease pathways and will facilitate personalized medicine. Acknowledgments We thank Ms. Hiroko Aita for technical assistance. We are indebted to the patients, nurses, medical staff, and physicians who supported this study. This work was supported by a Grant-in-Aid for Scientific Research (B) from the Ministry of Education, Science and Culture of Japan (23390223) (H.M.), and by a Grant-in-Aid for Project in Sado for Total Health (PROST) from the
123
Clin Exp Nephrol Ministry of Education, Culture, Sports, Science and Technology of Japan. Conflict of interest
None.
References 1. Zarate YA, Hopkin RJ. Fabry’s disease. Lancet. 2008; 372:1427–35. 2. Germain DP. Fabry disease. Orphanet J Rare Dis. 2010;5:30. 3. Nakao S, Kodama C, Takenaka T, Tanaka A, Yasumoto Y, Yoshida A, Kanzaki T, Enriquez AL, Eng CM, Tanaka H, Tei C, Desnick RJ. Fabry disease: detection of undiagnosed hemodialysis patients and identification of a ‘‘renal variant’’ phenotype. Kidney Int. 2003;64:801–7. 4. Maruyama H, Takata T, Tsubata Y, Tazawa R, Goto K, Tohyama J, Narita I, Yoshioka H, Ishii S. Screening of male dialysis patients for Fabry disease by plasma globotriaosylsphingosine. Clin J Am Soc Nephrol. 2013;8:629–36. 5. Kikumoto Y, Sugiyama H, Morinaga H, Inoue T, Takiue K, Kitagawa M, Saito D, Takatori Y, Kinomura M, Kitamura S, Akagi S, Sada K, Nakao K, Maeshima Y, Kitayama H, Makino H. The frequency of Fabry disease with the E66Q variant in the a-galactosidase A gene in Japanese dialysis patients: a case report and a literature review. Clin Nephrol. 2012;78:224–9. 6. Fujii H, Kono K, Goto S, Onishi T, Kawai H, Hirata K, Hattori K, Nakamura K, Endo F, Fukagawa M. Prevalence and cardiovascular features of Japanese hemodialysis patients with Fabry disease. Am J Nephrol. 2009;30:527–35. 7. Nishino T, Obata Y, Furusu A, Hirose M, Shinzato K, Hattori K, Nakamura K, Matsumoto T, Endo F, Kohno S. Identification of a novel mutation and prevalence study for Fabry disease in Japanese dialysis patients. Ren Fail. 2012;34:566–70. 8. Lee BH, Heo SH, Kim GH, Park JY, Kim WS, Kang DH, Choe KH, Kim WH, Yang SH, Yoo HW. Mutations of the GLA gene in Korean patients with Fabry disease and frequency of the E66Q allele as a functional variant in Korean newborns. J Hum Genet. 2010;55:512–7. 9. Togawa T, Tsukimura T, Kodama T, Tanaka T, Kawashima I, Saito S, Ohno K, Fukushige T, Kanekura T, Satomura A, Kang DH, Lee BH, Yoo HW, Doi K, Noiri E, Sakuraba H. Fabry disease: biochemical, pathological and structural studies of the agalactosidase A with E66Q amino acid substitution. Mol Genet Metab. 2012;105:615–20.
123
10. Kobayashi M, Ohashi T, Fukuda T, Yanagisawa T, Inomata T, Nagaoka T, Kitagawa T, Eto Y, Ida H, Kusano E. No accumulation of globotriaosylceramide in the heart of a patient with the E66Q mutation in the a-galactosidase A gene. Mol Genet Metab. 2012;107:711–5. 11. Nakamura K, Sekijima Y, Nakamura K, Hattori K, Nagamatsu K, Shimizu Y, Yazaki M, Sakurai A, Endo F, Fukushima Y, Ikeda SI. p.E66Q mutation in the GLA gene is associated with a high risk of cerebral small-vessel occlusion in elderly Japanese males. Eur J Neurol. 2014;21:49–56 12. Okumiya T, Ishii S, Takenaka T, Kase R, Kamei S, Sakuraba H, Suzuki Y. Galactose stabilizes various missense mutants of alpha-galactosidase in Fabry disease. Biochem Biophys Res Commun. 1995;214:1219–24. 13. Tanaka M, Ohashi T, Kobayashi M, Eto Y, Miyamura N, Nishida K, Araki E, Itoh K, Matsushita K, Hara M, Kuwahara K, Nakano T, Yasumoto N, Nonoguchi H, Tomita K. Identification of Fabry’s disease by the screening of alpha-galactosidase A activity in male and female hemodialysis patients. Clin Nephrol. 2005;64:281–7. 14. Ishii S, Sakuraba H, Suzuki Y. Point mutations in the upstream region of the a-galactosidase A gene exon 6 in an atypical variant of Fabry disease. Hum Genet. 1992;89:29–32. 15. Ichinose M, Nakayama M, Ohashi T, Utsunomiya Y, Kobayashi M, Eto Y. Significance of screening for Fabry disease among male dialysis patients. Clin Exp Nephrol. 2005;9:228–32. 16. Utsumi K, Kase R, Takata T, Sakuraba H, Matsui N, Saito H, Nakumura T, Kawabe M, Iino Y, Katayama Y. Fabry disease in patients receiving maintenance dialysis. Clin Exp Nephrol. 2000;4:49–51. 17. Inoue T, Hattori K, Ihara K, Ishii A, Nakamura K, Hirose S. Newborn screening for Fabry disease in Japan: prevalence and genotypes of Fabry disease in a pilot study. J Hum Genet. 2013;58:548–52. 18. Manolio TA, Collins FS, Cox NJ, Goldstein DB, Hindorff LA, Hunter DJ, McCarthy MI, Ramos EM, Cardon LR, Chakravarti A, Cho JH, Guttmacher AE, Kong A, Kruglyak L, Mardis E, Rotimi CN, Slatkin M, Valle D, Whittemore AS, Boehnke M, Clark AG, Eichler EE, Gibson G, Haines JL, Mackay TF, McCarroll SA, Visscher PM. Finding the missing heritability of complex diseases. Nature. 2009;461:747–53. 19. Raychaudhuri S. Mapping rare and common causal alleles for complex human diseases. Cell. 2011;147:57–69.