G3: Genes|Genomes|Genetics Early Online, published on July 13, 2016 as doi:10.1534/g3.116.032979
1
Title Page
2
3
From Genotype to Phenotype: Nonsense Variants in SLC13A1 are Associated with Decreased
4
Serum Sulfate and Increased Serum Aminotransferases
5
6
*Christina G. Tise, 1033 West Barre Street, Baltimore, MD 21230;
7
[email protected]
8
James A. Perry1, 685 West Baltimore Street, MSTF Room 357, Baltimore, MD 21201;
9
[email protected]
10
Leslie E. Anforth1, 135 Strummer Lane Gaithersburg, MD 20878; [email protected]
11
Mary A. Pavlovich1, 685 West Baltimore Street, MSTF Room 357, Baltimore, MD 21201;
12
[email protected]
13
Joshua D. Backman1, 660 West Redwood Street, Howard Hall 460, Baltimore, MD 21201;
14
[email protected]
15
Kathleen A. Ryan1, 685 West Baltimore Street, MSTF Room 357, Baltimore, MD 21201;
16
[email protected]
17
Joshua P. Lewis1, 660 West Redwood Street, Howard Hall 498, Baltimore, MD 21201;
18
[email protected]
19
Jeffrey R. O’Connell1, 685 West Baltimore Street, MSTF Room 300 B, Baltimore, MD 21201;
20
[email protected]
21
Laura M. Yerges-Armstrong1¶, 709 Swedeland Road, UW2230, King of Prussia, PA 19406;
22
[email protected]
1
© The Author(s) 2013. Published by the Genetics Society of America.
23
Alan R. Shuldiner1¶, 685 West Baltimore Street, MSTF Room 379, Baltimore, MD 21201;
24
[email protected]
25
26
1
27
Nutrition, and, University of Maryland School of Medicine, Baltimore, Maryland, United States of
28
America
Program for Personalized and Genomic Medicine and Division of Endocrinology, Diabetes and
29
30
* Corresponding author
31
¶
LMY and ARS are Joint Senior Authors.
32
33
34
35
36
37
38
39
40
41
42
43
44
45
2
46
Abstract
47
Using genomic applications to glean insights into human biology, we systematically
48
searched for nonsense, single nucleotide variants (SNVs) that are rare in the general population but
49
enriched in the Old Order Amish (Amish) due to founder effect. We identified two non-linked,
50
nonsense SNVs (R12X and W48X) in SLC13A1 (allele frequencies 0.29% and 0.74% in the Amish;
51
enriched 1.2-fold and 3.7-fold, compared to the outbred Caucasian population, respectively).
52
SLC13A1 encodes the apical, sodium-sulfate cotransporter (NaS1) responsible for sulfate
53
(re)absorption in the kidneys and intestine. Both SLC13A1 R12X and W48X were independently
54
associated with a 27.6% (P=2.7x10-8) and 27.3% (P=6.9x10-14) decrease in serum sulfate,
55
respectively (P=8.8x10-20 for carriers of either SLC13A1 nonsense SNV). We further performed the
56
first exome- and genome-wide association study (ExWAS/GWAS) of serum sulfate and identified a
57
missense variant (L348P) in SLC26A1, which encodes the basolateral sulfate-anion transporter
58
(Sat1), that was associated with decreased serum sulfate (P=4.4x10-12). Consistent with sulfate’s
59
role in xenobiotic detoxification and protection against acetaminophen-induced hepatotoxicity,
60
SLC13A1 nonsense SNV carriers had higher aminotransferase levels compared to non-carriers.
61
Furthermore, SLC26A1 L348P was associated with lower whole-body bone mineral density (BMD)
62
and higher serum calcium, consistent with the osteochondrodysplasia exhibited by dogs and sheep
63
with naturally-occurring, homozygous, loss-of-function mutations in Slc13a1. This study
64
demonstrates the power and translational potential of systematic identification and characterization
65
of rare, loss-of-function variants and warrants additional studies to better understand the importance
66
of sulfate in human physiology, disease, and drug toxicity.
67
68
3
69
70
Article Summary
While less genetically diverse, founder populations are more likely to have multiple
71
individuals who share the same ancestral variant. The authors take advantage of this concept by
72
systematically identifying nonsense variants enriched in a cohort of Old Order Amish individuals.
73
Upon selecting two nonsense variants in the sulfate transporter gene, SLC13A1, to study further, the
74
authors investigate the consequences of these variants on serum sulfate, a micronutrient
75
infrequently measured. In addition, the authors examine the effect of these variants on clinical
76
phenotypes, results of which are particularly relevant given sulfate’s known role in acetaminophen
77
toxicity and osteoarthritis.
78
79
80
Keywords
81
SLC13A1; serum sulfate; loss-of-function variants; Old Order Amish; GWAS/ExWAS
82
83
84
85
86
87
88
89
90
4
91
Background
92
Population isolates have been critical in elucidating the genetic basis of both classical
93
Mendelian and complex diseases. While many loss-of-function variants associated with disease are
94
rare in the general population, genetic drift in founder populations like the Old Order Amish
95
(Amish) increases the probability that many individuals will share the same rare, disease-causing
96
variant. Recent advances in genome-wide genotyping and sequencing technologies have made
97
possible the systematic identification of these variants and their subsequent investigation.
98
In this study, we systematically interrogated nonsense single nucleotide variants (SNVs)
99
from the Illumina Human Exome BeadChip for minor allele frequency (MAF) enrichment in the
100
Amish compared to outbred, European-derived populations. Using this approach, we observed an
101
enrichment of two nonsense SNVs (c.34G>A, p.R12X and c.144C>T, p.W48X) in SLC13A1, which
102
encodes an apical membrane, sodium-sulfate cotransporter (NaS1) responsible for sulfate
103
(re)absorption in the intestines and kidneys.
104
Inorganic sulfate (SO42-) is an important micronutrient vital for numerous cellular and
105
metabolic processes in human development and physiology (Dawson 2011, 2013; Markovich
106
2014). Sulfate is critical in the biotransformation of multiple compounds via sulfotransferase-
107
mediated sulfate conjugation (sulfation) (Klaassen and Boles 1997). These compounds include
108
hormones, neurotransmitters, proteoglycans and xenobiotics during phase II metabolism (Dawson
109
2011; Markovich 2014, 2011a; Council 2005; Markovich 2011b). Impaired sulfation capacity
110
substantially alters the metabolism and activities of these compounds and has been implicated in
111
several human pathologies including reduced xenobiotic clearance, skeletal dysplasia, premature
112
pubarche, autism spectrum disorder (ASD), and neurological disease (Alberti et al. 1999; Bowling
113
et al. 2013; Heafield et al. 1990; Ahmad et al. 1998; Noordam et al. 2009; Dawson 2013; Dawson
5
114
and Markovich 2005). Despite the importance of sulfate in these physiological processes, little is
115
known regarding sulfate homeostasis in humans nor is it routinely measured in clinical settings.
116
Given the potential importance of sulfate in the development of several disorders, we
117
comprehensively characterized serum sulfate in 977 Amish individuals in order to 1) determine the
118
heritability of serum sulfate, 2) examine the relationship between SLC13A1 SNVs and serum
119
sulfate, 3) identify additional novel genetic contributors through completion of the first exome-wide
120
association study (ExWAS) and genome-wide association study (GWAS) of serum sulfate, and 4)
121
investigate associations between sulfate-altering genotypes and relevant, clinical phenotypes
122
including aminotransferase levels, bone mineral density (BMD), and serum calcium.
123
124
125
Methods
126
Study Population
127
This report is based on the Amish community living in Lancaster County, PA, whom our
128
research group has been studying since 1993. This community was founded by several hundreds of
129
individuals who immigrated to Lancaster County, PA from central Europe during the early 18th
130
century, with the present day Lancaster County Amish community comprised of their descendants
131
(Lee et al. 2010). Cultural and religious beliefs have maintained the Amish as distinct from the
132
general population. Due to the availability of extensive genealogical records (Agarwala et al. 2003),
133
virtually all present-day Amish can be linked into a single, 14-generation pedigree. To date, we
134
have screened approximately 6,500 Amish adults for a variety of risk factors related to
135
cardiovascular disease (Mitchell et al. 2008), diabetes (Hsueh et al. 2000), and osteoporosis
136
(Streeten et al. 2006) as part of the Amish Complex Disease Research Program (ACDRP).
6
137
Subjects included in this report are members of the Amish community in Lancaster County,
138
PA who were at least 18 years old and have previously participated in one or more Institutional
139
Review Board (IRB)-approved studies conducted at the University of Maryland School of Medicine
140
Amish Research Clinic (ARC). Written informed consent was obtained from each participant.
141
142
143
Genotyping
Exome-wide genotyping, including four SLC13A1 SNVs, R12X (rs28364172), W48X
144
(rs138275989), N174S (rs2140516) and R237C (rs139376972), was performed on 1,725 Amish
145
subjects using the Illumina Human Exome BeadChip (Illumina Inc., San Diego, California).
146
Genotyping was performed according to the manufacturer’s instructions and genotypes were
147
assigned using Illumina GenomeStudio software (Illumina Inc., San Diego, California).
148
In subsequent follow-up investigations, genotyping of SLC13A1 nonsense SNVs, R12X and
149
W48X was performed on additional Amish subjects using TaqMan® SNP genotyping assays (Life
150
Technologies, Foster City, California). In total, SLC13A1 R12X and W48X genotypes were
151
available in 3,924 and 3,782 unique individuals, respectively. For both SNPs, the TaqMan genotype
152
concordance was greater than 99.8% in a subset of duplicate samples, and for subjects genotyped
153
using the Illumina Human Exome BeadChip and TaqMan® SNP genotyping assays, genotype
154
concordance between platforms was 100% for both SLC13A1 R12X and W48X.
155
For 917 of the 977 subjects for which serum sulfate was measured, genome-wide
156
genotyping was performed using the Affymetrix GeneChip Human Mapping 500K or 1M (version
157
6.0) arrays according to the manufacturer's instructions (Affymetrix Inc., Santa Clara, California).
158
Genotype calls were performed using BRLMM (500K array) or Birdseed version 2 (1 M array).
7
159
SNVs present on both arrays with a minor allele frequency greater than 1% were included in the
160
analyses.
161
162
163
Serum Sulfate Measurements
All known Amish carriers of the rare SLC13A1 SNVs, R12X, W48X and R237C, for whom
164
fasting, serum aliquots were available, were selected for sulfate measurement. Such subjects
165
consisted of participants from the Amish Pharmacogenomics of Anti-Platelet Intervention (PAPI)
166
Study (Shuldiner et al. 2009; Bozzi et al. 2015), the Amish Heredity and Phenotype Intervention
167
(HAPI) Heart Study (Mitchell et al. 2008), the Amish Family Calcification Study (AFCS) (Post et
168
al. 2007), the Amish Family Longevity Study (AFLS) (Mitchell et al. 2001; Sorkin et al. 2005), and
169
the Amish Wellness Study (AWS) (File S1). All other subjects consisted of randomly selected
170
participants of the Amish PAPI (Shuldiner et al. 2009; Bozzi et al. 2015) or HAPI Heart (Mitchell
171
et al. 2008) studies for whom fasting serum aliquots were available.
172
Sulfate measurements were completed on frozen, fasting serum aliquots stored at -80◦C.
173
Sulfate concentration was determined by turbidimetry according to Dodgson & Price (Dodgson and
174
Price 1962) using a Quantichrom™ Sulfate Assay Kit (BioAssay Systems, Hayward, CA). In order
175
to improve accuracy, a quadratic least squares fit was used instead of a linear fit to generate the
176
standard curve (NIST/SEMATECH e-Handbook of Statistical Methods). All standards and samples
177
were measured in duplicate. For each sample, sulfate concentration was calculated as the mean of
178
the duplicate measurements. Samples with an absolute difference greater than 20% were considered
179
discordant duplicate measurements and were not included in the analysis. Of the 1,003 samples in
180
which serum sulfate was measured, 26 subjects’ samples were excluded due to discordant duplicate
181
measurements, resulting in valid sulfate concentration measurements for 977 subjects’ samples.
8
182
183
Serum Alanine Aminotransferase (ALT), Aspartate Aminotransferase (AST), Calcium, and
184
Albumin Measurements
185
Serum ALT, AST, calcium, and albumin measurements were collected as part of the Amish
186
PAPI (Shuldiner et al. 2009; Bozzi et al. 2015) and Amish Family Longevity (Sorkin et al. 2005;
187
Mitchell et al. 2001) studies and completed on fresh, fasting serum aliquots by Quest Diagnostics
188
(Horsham, Pennsylvania) at the time of subject enrollment.
189
190
191
Bone Mineral Density (BMD) Measurements
BMD was measured as part of the Amish PAPI Study (Shuldiner et al. 2009; Bozzi et al.
192
2015) by dual-energy x-ray absorptiometry at the proximal femur, lumbar spine, one-third radius,
193
and whole-body using a Hologic 4500W (Hologic). Daily phantom measurements were obtained to
194
detect measurement shifts over time. Coefficients of variation for the repeat measures of the sites in
195
normal volunteers were between 1.5% and 0.8%.
196
197
198
Statistical analysis
The MAF for each nonsense SNV was obtained from the Total European Ancestry
199
population from 1000 Genomes (1000g-EUR) and the European American population from the
200
National Heart, Lung, and Blood Institute (NHLBI) Exome Sequencing Project (ESP) (ESP-EA).
201
For each SNV, the total allele count was assumed to be 1338 in 1000g-EUR (n=669) and 6728 in
202
ESP-EA (n=3,364). For each nonsense SNV, the MAF in the Amish was compared to the MAF in
203
1000g-EUR and ESP-EA (when available) using a chi-squared test.
9
204
Association analyses between genotypes and serum sulfate, and other phenotypic measures,
205
were conducted using a regression-based method that models variation of the trait of interest as a
206
function of measured covariates, measured genotypes, and a polygenic component that accounts for
207
phenotypic correlation due to relatedness. This method was implemented using the Mixed Models
208
Analysis for Pedigrees and Populations (MMAP) program (O'Connell). For each association
209
analysis performed, individuals with a missing covariate, genotype and/or trait of interest were
210
excluded from the analysis. All analyses of serum sulfate included gender, study, and study-age as
211
covariates. All analyses of clinical phenotypes included age and gender as covariates. All analyses
212
of clinical phenotypes obtained in Amish PAPI Study (Shuldiner et al. 2009; Bozzi et al. 2015)
213
participants containing serum sulfate as a covariate were limited to subjects for whom serum sulfate
214
was measured using a serum aliquot collected at the time of participation in that study.
215
216
217
Results
218
SLC13A1 Variants are Enriched in the Amish
219
In this Amish cohort (n=1,725), 43,459 of the >240,000 markers on the Illumina Human
220
Exome BeadChip were polymorphic, 266 of which were nonsense SNVs. For 45 of the 266
221
nonsense SNVs, no MAF was reported in 1000g-EUR nor ESP-EA. For 131 of the 266 nonsense
222
SNVs, the MAF was greater in this Amish cohort compared to the MAF in 1000g-EUR averaged
223
with the MAF in ESP-EA. Of these 131 SNVs, 67 and 103 were significantly enriched (P<0.05)
224
compared to 1000G-EUR and ESP-EA, respectively (Table S1). With an intent to unveil new
225
biology, we filtered out variants in genes with an associated Online Mendelian Inheritance in Man
226
(OMIM®) (Online Mendelian Inheritance in Man, OMIM®) phenotype, leaving us with 118
10
227
nonsense SNVs. Two nonsense SNVs (c.34G>A, p.R12X and c.144C>T, p.W48X) in SLC13A1
228
(Figure 1) were enriched 1.9-fold (0.43% vs. 0.23%) and 4.4-fold (0.87% vs. 0.20%), respectively,
229
compared to ESP-EA allele frequencies. In addition, two missense SNVs (c.521T>C, p.N174S and
230
c.877G>A, p.R237C) in SLC13A1 were also present on the Illumina Exome BeadChip and enriched
231
1.2-fold (39% vs. 32%) and 10.7-fold (1.7% vs. 0.16%), respectively, compared to ESP-EA allele
232
frequencies (Table S2).
233
Upon further genotyping of SLC13A1 R12X and W48X in additional Amish subjects
234
(nR12X=3,924; nW48X=3,782), the allele frequencies were enriched 1.2-fold (0.29% vs. 0.023) and
235
3.7-fold (0.74% vs. 0.20%), respectively (Table S2). All four SLC13A1 SNVs conformed to Hardy-
236
Weinberg expectations (Table S2); no R12X homozygotes, W48X homozygotes, or R12X/W48X
237
compound heterozygotes were observed. None of the four SNVs were in significant linkage
238
disequilibrium (LD) with one another with the exception of N174S and R237C (r2=0.02 and
239
|D’|=1.00, LOD≥2) (Figure S1).
240
241
242
SLC13A1 Variants are Associated with Serum Sulfate Levels
Given the role of SLC13A1 in sulfate (re)absorption, we measured sulfate levels in 977
243
individuals. Amish research subjects selected for sulfate measurement consisted of 481 males
244
(49.2%) and 496 females (50.8%), with a mean age of 46.0 ± 14.4 years, and a mean BMI of 27.1 ±
245
4.8 kg/m2 (Table 1). Serum sulfate was normally distributed in this population with an unadjusted
246
mean of 0.36 mM (Figure S2). Serum sulfate was associated with increasing age (β=0.0007,
247
P=4.9x10-6), but not with gender (β=-0.0049, P=0.36). The total variance in serum sulfate
248
explained by age, gender, and study was 13% (Table S3a). The residual heritability of serum sulfate
249
after adjustment for age, gender, and study was 0.40 (P=1.8x10-16).
11
250
In our primary model, we assessed associations between each of the four SLC13A1 SNVs
251
and serum sulfate, independently (Table S3a-b). Specifically, we observed strong associations
252
between R12X and W48X, and lower serum sulfate. SLC13A1 R12X was associated with a 26.7%
253
lower serum sulfate (β=-0.10 mM, P=2.9x10-7) (Figure S3) and SLC13A1 W48X was associated
254
with a 26.5% lower serum sulfate (β=-0.10 mM, P=1.3x10-12) (Figure S4). We evaluated the
255
influence of N174S and R237C on serum sulfate levels in 900 of these individuals. In contrast to
256
R12X and W48X, N174S was associated with a 3% higher serum sulfate per allele (β=0.01 mM,
257
P=1.6x10-4) (Figure 2a). No association was observed between R237C and serum sulfate (P=0.15)
258
(Figure 2b). The estimated heritability after adjustment for each of these genotypes, along with the
259
additive and total variance explained by each variant is shown in Table S3.
260
While LD data suggested the SLC13A1 R12X and W48X variants are not correlated with
261
each other, we performed conditional association analyses to evaluate whether these variants
262
represent independent association signals (Table S2a). Indeed, both R12X and W48X were
263
independently associated with a 27.6% (β=-0.10 mM, P=2.7x10-8) and 27.3% (β=-0.10 mM,
264
P=6.9x10-14) decrease in serum sulfate, respectively (Figure 2c). SLC13A1 R12X and W48X
265
together, explained approximately 31% of the additive variance and 9% of the total variance in
266
serum sulfate. The increased significance of these two SNVs, along with the increase in additive
267
and total variance explained, compared to our primary models, is consistent with these two loci
268
being independent genetic determinants of serum sulfate.
269
Finally, we assessed the association between SLC13A1 nonsense carriers and serum sulfate
270
by including carrier status of an SLC13A1 nonsense SNV (R12X or W48X) as a covariate (Table
271
S3a). We observed a 27% decrease in serum sulfate amongst SLC13A1 nonsense SNV carriers
272
compared to subjects who do not carry R12X or W48X (β=-0.10 mM, P=8.8x10-20) (Figure 2c).
12
273
The estimated residual heritability after adjustment for SLC13A1 nonsense SNV carrier status was
274
0.28 with carrier status of an SLC13A1 nonsense SNV explaining 31% of the additive variance and
275
9% of the total variance in serum sulfate.
276
277
278
SLC26A1 L348P is Associated with Serum Sulfate Levels
An ExWAS of serum sulfate, performed using genotype data from the Illumina Human
279
Exome BeadChip (n=900) (Figure S5a), demonstrated genome-wide significance (P<5.0x10-8) for
280
W48X (β=-0.08 mM, P=2.7×10-8), and suggestive-significance for R12X (β=-0.09 mM,
281
P=7.5×10-6) in SLC13A1 on chromosome 7q31.32 (Figure 3 and Table S3b). The ExWAS also
282
revealed a novel cluster of four SNVs spanning approximately 3 megabases on chromosome 4p16.3
283
demonstrating genome-wide significance (Figure 3). These SNVs were found to be in strong LD
284
with one another, i.e. pairwise r2=0.49 to 0.97 and |D’|=0.79 to 1.00, LOD≥2 (Figure S6), with
285
L348P in SLC26A1 (rs148386572; MAF=0.06) being the most significant, associated with a 12%
286
decrease in serum sulfate per allele (β=0.05 mM, P=4.4x10-12). SLC26A1 encodes the basolateral
287
membrane, sulfate-anion transporter 1 (Sat1), consistent with this locus as a true determinant of
288
serum sulfate (Figure 1), and indicative of L348P causing decreased or loss of function in the Sat1
289
protein. SLC26A1 L348P explained 3% of the additive variance and 5% of the total variance in
290
serum sulfate (Table S3b). No other genomic region revealed association signals that met or
291
exceeded genome-wide significance (Table S3).
292
A secondary ExWAS of serum sulfate, adjusting for SLC13A1 R12X and W48X, and
293
SLC26A1 L348P genotypes, was also performed (n=900) (Figure S5b). No genomic region revealed
294
association signals that met or exceeded genome-wide significance (Figure S7 and Table S4).
13
295
A GWAS of serum sulfate, performed using genotype data from the Affymetrix GeneChip
296
(n=917) (Figure S8), demonstrated genome-wide significance for a variant on chromosome 7
297
(rs17684427) and genome-wide suggestive-significance for a variant on chromosome 4 (rs362272)
298
in strong LD with SLC13A1 W48X and SLC26A1 L348P, respectively (Figure S9-S11 and Table
299
S6). A secondary GWAS of serum sulfate, adjusting for rs17684427 and rs362272 genotypes was
300
also performed (n=917) (Figure S12). No genomic region revealed association signals that met or
301
exceeded genome-wide significance (Figure S13).
302
A model including all five SNV genotypes (SLC13A1 R12X, W48X, N174S, and R237C,
303
and SLC26A1 L348P) into the model were used to further estimate residual heritability (Table S3b).
304
In this model, SLC13A1 R12X, W48X, and N174S, and SLC26A1 L348P were all independently
305
associated with decreased serum sulfate while SLC13A1 R237C was not. The estimated residual
306
heritability after adjustment for SLC13A1 R12X, W48X, N174S, and R237C, and SLC26A1 L348P
307
genotype was 0.31 with these variants, together, explaining approximately 29% of the additive
308
variance and 13% of the total variance in serum sulfate.
309
310
311
SLC13A1 Nonsense SNVs are Associated with Increased Aminotransferase Levels
Given sulfate’s role in drug metabolism, we assessed the association between SLC13A1
312
nonsense SNV carriers and levels of alanine aminotransferase (ALT) and aspartate
313
aminotransferase (AST) in 684 participants from the Amish PAPI Study (Shuldiner et al. 2009;
314
Bozzi et al. 2015) (Table 1). ALT levels were 20% higher in SLC13A1 nonsense SNV carriers
315
compared to non-carriers (β=3.7 U/L, P=0.03). Similarly, a trending association was observed
316
between SLC13A1 nonsense carriers and increased AST (β=2.1 U/L, P=0.09) (Figure 4a). No
14
317
association was observed between SLC26A1 L348P, and ALT or AST levels (P=0.92 and 0.37,
318
respectively) (Table 2a).
319
We were able to replicate these associations in an independent Amish cohort consisting of
320
264 participants from the Amish Family Longevity Study (AFLS) (Mitchell et al. 2001; Sorkin et
321
al. 2005) who were, on average, 20.5 years older than the 684 participants from the Amish PAPI
322
Study (Shuldiner et al. 2009; Bozzi et al. 2015) (Table 1). ALT and AST levels were found to be
323
42% and 55% higher, respectively, in SLC13A1 nonsense SNV carriers than in non-carriers (β=7.3
324
U/L, P=2.0x10-3 and β=10.4 U/L, P=1.3x10-7, respectively) (Figure 4b). Additionally, an
325
association between SLC26A1 L348P and a 13% increase in AST per allele was observed in this
326
cohort (β=2.4 U/L, P=6.8x10-3) (Table 2b).
327
328
SLC26A1 Sulfate-lowering SNVs are Associated with Altered Serum Calcium and BMD
329
Given the role of sulfate in cartilage and bone development, we assessed the association
330
between SLC13A1 nonsense SNV carriers and whole-body BMD. No association was observed
331
between SLC13A1 nonsense carriers and whole-body BMD (P=0.19); however, SLC26A1 L348P
332
was significantly associated with decreased whole-body BMD (β=-0.03 g/cm2, P=6.8x10-3) (Figure
333
5 and Table 2c). Upon further investigation, SLC26A1 L348P was also associated with several other
334
BMD measurements while SLC13A1 nonsense SNV carrier status was not (Table 2d and S7).
335
Furthermore, SLC26A1 L348P was significantly associated with increased total serum calcium
336
(β=0.11 mg/dL, P=1.4x10-3), and showed a nearly-significant association with albumin-adjusted
337
serum calcium (β=0.06, P=0.06) (Figure 5 and Table 2c). No association was observed between
338
SLC13A1 nonsense SNV carriers, and serum calcium or albumin-adjusted serum calcium (P=0.98
339
and 0.68, respectively) (Table 2c).
15
340
341
342
Clinical Phenotypes are Not Associated with Serum Sulfate Concentration
Serum sulfate concentration by itself was not associated with any of the aforementioned
343
clinical phenotypes with the exception of trochanter and intertrochanter BMD. Lower serum sulfate
344
was marginally associated with lower trochanter (β=0.13 g/cm2, P=0.03) and lower intertrochanter
345
BMD (β=0.18 g/cm2, P=0.04) (Table S8).
346
347
348
349
Discussion
Using genomic approaches to unveil new human biology, we systematically identified
350
nonsense SNVs that are rare in the general population but enriched in the Amish due to founder
351
effect. Two nonsense SNVs (R12X and W48X), as well as two missense SNVs (N174S and
352
R237C), in SLC13A1 were not only enriched in the Amish but likely functional, thus we
353
hypothesized that these SNVs would be associated with altered serum sulfate due to altered sulfate
354
(re)absorption and perhaps other traits for which sulfate may play a role. We further conducted the
355
first ExWAS and GWAS of serum sulfate in an attempt to identify other genetic contributors of
356
serum sulfate, a trait not measured clinically despite its important physiological role in human
357
development, disease, and drug toxicity. To the best of our knowledge, our group is the first to
358
establish that serum sulfate is heritable (h2=0.40, P=1.8x10-16). In addition, these studies have
359
identified novel associations between R12X, W48X, and N174S in SLC13A1, and L348P in
360
SLC26A1, and serum sulfate levels. Further clinical characterization of the three sulfate-lowering
361
variants implicates sulfate wasting in liver function, and perhaps BMD and calcium homeostasis.
16
362
In regards to SLC13A1 R12X, our findings are consistent with work performed by Bowling
363
et al., which demonstrated an association between R12X and increased urinary fractional excretion
364
index (FEI) of sulfate in a cohort with autism spectrum disorder (ASD) (Bowling et al. 2013).
365
Furthermore, the R12X allele has been shown by Lee et al. to abolish 100% of NaS1 sulfate
366
transport in Xenopus oocytes (Lee et al. 2006). It is likely that this complete loss of NaS1 function
367
occurs through a nonsense-mediated decay (NMD) mechanism. While SLC13A1 W48X has not
368
been described in the literature to date, the similar effect size of R12X and W48X on serum sulfate
369
and their lack of LD with one another suggest similar loss-of-functionality. Together, these two
370
SNVs alone explain approximately 31% of the additive variance and 9% of the total variance of
371
serum sulfate in the Amish. Based on previous experience, these estimates are likely to be lower in
372
outbred populations. Nevertheless, these estimates clearly validate our approach of systematically
373
identifying and studying nonsense SNVs to glean insights into human biology.
374
Unexpectedly, we observed an association between the more common SLC13A1 N174S
375
variant and increased serum sulfate, albeit an effect size of approximately 1/10th the magnitude of
376
SLC13A1 R12X or W48X, consistent with a gain-of-function. In contrast, Bowling et al. found
377
N174S to be associated with increased FEI of sulfate (Bowling et al. 2013) in a cohort with ASD
378
while Lee et al. found the N174S allele to abolish greater than 60% of NaS1 sulfate transport (Lee
379
et al. 2006). With the exception of R12X, other loss-of-function variants, such as W48X, were not
380
accounted for by Bowling et al.
381
Our ExWAS and GWAS findings support the results of our candidate gene study regarding
382
the sulfate-lowering effect of R12X and W48X in SLC13A1, while also identifying L348P in
383
SLC26A1 as an additional genetic contributor of serum sulfate. SLC26A1 encodes Sat1, the
384
basolateral membrane, sulfate-anion transporter, consistent with this locus as a true determinant of
17
385
serum sulfate. While SLC26A1 L348P has not been described in the literature to our knowledge, at
386
half the effect size of SLC13A1 R12X or W48X, each copy of the L348P allele results in an additive
387
decrease in serum sulfate. This association is suggestive of SLC26A1 L348P causing decreased or
388
loss of function in the Sat1 protein; however functional studies are needed to fully understand the
389
extent to which this occurs. It should be noted that all three of the sulfate-lowering SNVs identified
390
in this report are enriched in the Amish, with SLC13A1 W48X and SLC26A1 L348P being
391
markedly enriched. Therefore, our study is likely better powered to detect these variants compared
392
to other ExWAS or GWAS of serum sulfate in outbred populations of comparable size.
393
Lastly, our results indicate that serum sulfate is a heritable trait, and while our results
394
explain a large portion of the genetic component of serum sulfate, there is an equally large portion
395
yet to be explained. It is possible that much of this heritability can be explained by additional
396
variants in SLC13A1, SLC26A1, and/or other sulfate transporters that were missed by the
397
genotyping arrays used in this study. Such variants have the potential to be identified through other
398
genotyping arrays, exome sequencing, and/or copy number analysis, thus studies aiming to identify
399
additional genetic variants in genes associated with serum sulfate in the Amish, as well as outbred
400
populations, are warranted.
401
402
403
SLC13A1 and Aminotransferase Levels
Serum AST and ALT are biomarkers of hepatocyte integrity and thus elevations in these
404
serum aminotransferases suggest liver cell injury (Giannini et al. 2005; Gowda et al. 2009). Given
405
sulfate’s role in drug metabolism and increased sensitivity to acetaminophen-induced hepatotoxicity
406
exhibited by Slc13a1- and Slc26a1-knockout mice (Lee et al. 2006; Dawson et al. 2010; Markovich
407
2012), we hypothesized that individuals with sulfate-lowering variants would have higher liver
18
408
aminotransferase levels. The association between SLC13A1 nonsense SNV carriers and higher
409
aminotransferase levels in participants of the Amish PAPI Study (Shuldiner et al. 2009; Bozzi et al.
410
2015), support our hypothesis, with even greater effects seen in participants of the Amish Family
411
Longevity Study (Mitchell et al. 2001; Sorkin et al. 2005). The increase in aminotransferase levels
412
seen in SLC13A1 nonsense SNV carriers is consistent with increased susceptibility to drugs or
413
exogenous substrates metabolized through sulfation. This is potentially due to decreased inward
414
transport of sulfate into cells, independent of serum sulfate concentration, resulting in decreased
415
ability to detoxify xenobiotics. In contrast to the Amish PAPI Study (Shuldiner et al. 2009; Bozzi et
416
al. 2015), participants of the Amish Family Longevity Study (Mitchell et al. 2001; Sorkin et al.
417
2005) were not asked to discontinue prescription medications and/or supplements 7 days prior to
418
their initial clinic visit (File S1). We attribute the increased effect of sulfate-lowering variants in
419
participants of the Amish Family Longevity Study to routine medication/supplement use, as well as
420
increased participant age which may render hepatocytes more susceptible to damage from
421
xenobiotics. This phenomenon may partially explain the lack of an association between SLC26A1
422
L348P and elevated aminotransferase levels in the participants of the younger Amish PAPI Study
423
(Shuldiner et al. 2009; Bozzi et al. 2015), along with differential tissue expression of SLC13A1 and
424
SLC26A1, and/or L348P causing decreased, but sufficient, Sat1 function, as opposed to a complete
425
loss of function. Moreover, the associations observed between sulfate-lowering SNVs and
426
aminotransferase level in this study may be under-estimated given the relative drug-naïve lifestyle
427
practice by the Amish community. Additional studies are warranted to better understand the
428
importance of sulfate in human physiology and its potential role in disease and drug toxicity.
429
430
SLC26A1, and BMD and Serum Calcium
19
431
Sulfate’s physiological role in bone and cartilage is well known. Sulfated proteoglycans are
432
a critical component of extracellular matrices in cells throughout the body, specifically in
433
connective tissues and are required for maintaining normal bone and cartilage structure
434
(Yanagishita 1993; Fernandes et al. 1997). The essentiality of sulfate for proper bone and cartilage
435
formation is demonstrated by the spectrum of osteochondrodysplasias in humans with homozygous,
436
loss-of-function mutations in SLC26A2 (Hastbacka et al. 1996; Hastbacka et al. 1999; Online
437
Mendelian Inheritance in Man, OMIM®), as well as dog and sheep with homozygous loss-of-
438
function mutations in Slc13a1 (Neff et al. 2012; Zhao et al. 2012). We thus hypothesized that
439
individuals with sulfate-lowering SNVs would have altered BMD. We observed several significant
440
associations between SLC26A1 L348P and decreased BMD measurements which were not
441
associated with SLC13A1 nonsense carrier status, nor serum sulfate concentration. However,
442
compared to SLC26A1 L348P (allele count ~70, 6 homozygotes), there is far less power to detect
443
such associations with SLC13A1 nonsense SNVs due to the lower allele frequency (allele count=20)
444
and the lack of homozygotes. The lack of an association with SLC13A1 nonsense variants may also
445
be explained by differential tissue expression of SLC13A1 and SLC26A1.
446
The association between increased serum calcium and SLC26A1 L348P may be specific to
447
decreased Sat1 (SLC26A1) function and is seemingly independent of serum sulfate. These findings
448
are consistent with mouse models in which hyposulfatemic, Slc26a1-knockout mice exhibit
449
nephrocalcinosis and increased calcium oxalate kidney stone formation while hyposulfatemic,
450
Slc13a1-knockout mice do not exhibit these manifestations (Dawson et al. 2010; Markovich 2011b,
451
2011a, 2012). Additionally, SLC26A1 has been has been suggested to play a role in humans with
452
recurrent calcium oxalate kidney stones (Dawson et al. 2013). While the mechanism for
453
nephrocalcinosis, calcium oxalate urolithiasis, and increased serum calcium is not totally clear,
20
454
these phenotypes associate with decreased Sat1 function, suggesting they may be specific to anion
455
transport rather than sulfate transport. We hypothesize that the SLC26A1 L348P allele decreases
456
Sat1 function, decreasing sulfate transport into the blood as well as decreasing anion transport into
457
cells. As a result, serum oxalate and thus urine oxalate concentrations may increase resulting in the
458
formation of calcium oxalate crystals which either accumulate to form stones and/or increase serum
459
calcium levels. We do not have an adequate number of subjects with renal stones to address this
460
hypothesis further.
461
462
463
464
Conclusions
Sulfate is an essential micronutrient, as impaired sulfation capacity has been reported in a
465
number of human conditions including reduced xenobiotic clearance, skeletal dysplasias, premature
466
pubarche, ASD, and neurological disease (Alberti et al. 1999; Bowling et al. 2013; Heafield et al.
467
1990; Ahmad et al. 1998; Noordam et al. 2009; Dawson 2013; Dawson and Markovich 2005).
468
Despite sulfate’s known role in the biotransformation of hormones, neurotransmitters,
469
glycosaminoglycans (GAGs), cerebrosides, and xenobiotics (Dawson 2011; Markovich 2014,
470
2011a; Council 2005), sulfate is not routinely measured clinically (Dawson 2011, 2013). In this
471
study, we measured and comprehensively characterized serum sulfate in 977 Amish subjects. We
472
are the first to estimate the heritability of serum sulfate and the first to perform an ExWAS and a
473
GWAS of serum sulfate. We identified novel associations between serum sulfate and three SNVs in
474
the gene, SLC13A1, encoding the apical membrane, sodium-sulfate cotransporter, NaS1. We further
475
identified a novel association between L348P in the gene, SLC26A1, encoding the basolateral
476
membrane, sulfate-anion transporter, Sat1, and decreased serum sulfate. We also investigated
21
477
associations between sulfate-lowering SNVs, aminotransferase levels, BMD, and serum calcium,
478
results of which implicate sulfate and these SNVs in important human metabolism, physiology, and
479
disease processes.
480
In conclusion, this study demonstrates the power and translational potential of leveraging
481
genetic data in founder populations like the Old Order Amish to unveil and better understand the
482
phenotypic consequences of rare, deleterious variants enriched by genetic drift. Further, our clinical
483
findings warrant additional studies to better understand the importance of sulfate in human
484
physiology, disease, and drug toxicity.
485
486
487
488
489
490
491
Declarations
492
List of abbreviations
493
ACDRP: Amish Complex Disease Research Program
494
AFCS: Amish Family Calcification Study
495
AFLS: Amish Family Longevity Study
496
ALT: alanine aminotransferase
497
Amish: Old Order Amish
498
ARC: Amish Research Clinic
22
499
ASD: autism spectrum disorder
500
AST: aspartate aminotransferase
501
AWS: Amish Wellness Study
502
BMD: bone mineral density
503
ESP: Exome Sequencing Project
504
ExWAS: exome-wide association study
505
FEI: fractional excretion index
506
GAG(s): glycosaminoglycan(s)
507
GWAS: genome-wide association study
508
HAPI: Heredity and Phenotype Intervention
509
IRB: Institutional Review Board
510
LD: linkage disequilibrium
511
MAF: minor allele frequency
512
MMAP: Mixed Models Analysis for Pedigrees and Populations
513
NaS1: Sodium-sulfate cotransporter 1 protein
514
NHLBI: National Heart, Lung, and Blood Institute
515
NMD: nonsense-mediated decay
516
OMIM: Online Mendelian Inheritance in Man
517
PAPI: Pharmacogenomics of Anti-Platelet Intervention
518
Sat1: Sulfate-anion transporter 1 protein
519
SNV(s): single nucleotide variant(s)
520
521
Ethics approval and consent to participate
23
522
Subjects included in this report are members of the Old Order Amish community in
523
Lancaster County, PA who were at least 18 years old and have previously participated in one or
524
more Institutional Review Board (IRB)-approved studies conducted at the University of Maryland
525
School of Medicine Amish Research Clinic (ARC). Written informed consent was obtained from
526
each participant.
527
528
Availability of data and material
529
In accordance to our human subjects research protocols, we are unable to release the
530
individual-level data set. We continue to deposit aggregate-level data into the National Institutes of
531
Health (NIH) database of Genotypes and Phenotypes (dbGaP); however, given the Old Order
532
Amish are a founder population with publically available pedigree data, depositing individual-level
533
data on our subjects creates the potential for personal identification of our research subjects. In the
534
past, the NIH has provided exemptions for this reason. Furthermore, all of our analyses were
535
performed using the Mixed Models Analysis for Pedigrees and Populations (MMAP) program
536
which models variation of the trait of interest as a function of measured covariates, measured
537
genotypes, and a polygenic component that accounts for phenotypic correlation due to relatedness.
538
For this reason, outside investigators attempting to replicate our analyses will get different results if
539
the family structure of the Old Order Amish is not accounted for in their analyses. We would be
540
glad to assist any investigators who have questions or would like to perform related analyses
541
regarding this data. Authors will certainly entertain requests for individual-level data and pedigree
542
information with qualified investigators and collaborators after necessary modifications have been
543
made to our study protocols.
544
24
545
Competing interests
546
Christina G. Tise: No conflict of interest to disclose.
547
James A. Perry: No conflict of interest to disclose.
548
Leslie E. Anforth: No conflict of interest to disclose.
549
Mary A. Pavlovich: No conflict of interest to disclose.
550
Joshua D. Backman: No conflict of interest to disclose.
551
Kathleen A. Ryan: No conflict of interest to disclose.
552
Joshua P. Lewis: No conflict of interest to disclose.
553
Jeffrey R. O’Connell: No conflict of interest to disclose.
554
Laura M. Yerges-Armstrong: Dr. Yerges-Armstrong is employed by GlaxoSmithKline but her
555
contribution to this work was completed as an assistant professor at the University of Maryland
556
School of Medicine.
557
Alan R. Shuldiner: In addition to his part-time appointment at the University of Maryland School of
558
Medicine, Dr. Shuldiner is Vice President and Co-Head of the Regeneron Genetics Center, LLC, a
559
fully owned subsidiary of Regeneron Pharmaceuticals, Inc. The Regeneron Genetics Center focuses
560
on early discovery research, applying human genomics to identify novel drug targets.
561
562
Funding
563
National Institute on Aging
564
Award Number: R01 AG018728
565
Recipient: Alan R. Shuldiner
566
567
National Institute of Arthritis and Musculoskeletal and Skin Diseases
Award Number: R01 AR046838
25
568
569
Recipient: Alan R. Shuldiner
National Heart, Lung, and Blood Institute
570
Award Number: U01 HL072515
571
Recipient: Alan R. Shuldiner
572
National Institute of General Medical Sciences
573
Award Number: U01 GM074518
574
Recipient: Alan R. Shuldiner
575
National Heart, Lung, and Blood Institute
576
Award Number: U01 HL084756
577
Recipient: Jeffrey R. O’Connell
578
National Institute of Diabetes and Digestive and Kidney Diseases
579
Award Number: P30 DK072488
580
Recipient: Simeon I. Taylor (not applicable)
581
National Heart, Lung, and Blood Institute
582
Award Number: K01 HL116770
583
Recipient: Laura M. Yerges-Armstrong
584
585
Authors’ contributions
586
CGT designed the genetic filtering approach and subsequent experiments, performed sulfate
587
measurements, genotype assays, and genetic analyses, and wrote the manuscript. JAP performed
588
genetic analyses and produced figures. LEA performed sulfate measurements and genotype assays.
589
MAP retrieved clinical phenotype data and performed genetic analyses. JDB performed genetic
590
analyses. KAR retrieved clinical phenotype data and performed genetic analyses. JPL aided in
26
591
designing the genetic filtering approach and wrote the manuscript. JRO provided genotype data,
592
aided in genetic analyses, and oversaw the project. LMY provided genotype data, performed
593
genetic analyses, oversaw the project, and wrote the manuscript. ARS aided in designing the genetic
594
filtering approach, provided genotype data and clinical phenotype data, oversaw the project, and
595
wrote the manuscript. All authors read and approved the final manuscript.
596
597
598
Acknowledgements
We gratefully acknowledge and thank the Amish Research Clinic staff, as well as the Amish
599
community for their extraordinary cooperation and support, without which these studies would not
600
have been possible.
601
602
603
604
605
Tables
606
Table 1. Characteristics of Old Order Amish research participants by cohort.
607
Characteristic
Sulfate
AFLS
PAPI Cohort
(Units)
Cohort
Cohort
Number (n)
977
684
264
Male (%)
49.2
49.8
47.3
Age ± SD (years) 46.0 ± 14.4 45.0 ± 13.4 65.5 ± 10.4
BMI ± SD (kg/m2) 27.1 ± 4.8 27.1 ± 4.7 27.9 ± 5.2
Abbreviations: SD, standard deviation; BMI, body mass index; PAPI, Pharmacogenomics of Anti-
608
Platelet Intervention; ALFS, Amish Family Longevity Study.
609
610
27
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
Table 2. Associations between sulfate-lowering SNVs and clinical phenotypes.
626
a.
Trait
(PAPI)
ALT
AST
627
628
Sulfate-lowering SNV(s)
n
Het.
Hom.
SLC13A1 R12X or W48X
SLC26A1 L348P
SLC13A1 R12X or W48X
SLC26A1 L348P
684
650
684
650
20
71
20
71
0
6
0
6
SNV covariate(s)
n
Het.
Hom.
SLC13A1 R12X or W48X
SLC26A1 L348P
SLC13A1 R12X or W48X
SLC26A1 L348P
264
232
264
232
7
21
7
21
0
2
0
2
βSNV ± SE
βSNV/
(U/L)
MeanWT
3.73 ± 1.74
0.20
-0.08 ± 0.84
0.00
2.13 ± 1.26
0.11
0.53 ± 0.59
0.03
PSNV
0.03
0.92
0.09
0.37
b.
Trait
(AFLS)
ALT
AST
629
28
βSNV ± SE
βSNV/
(U/L)
MeanWT
7.25 ± 2.33
0.42
0.81 ± 1.20
0.04
10.42 ± 1.92
0.55
2.42 ± 0.89
0.13
PSNV
2.0x10-3
0.50
1.3x10-7
6.8x10-3
630
c.
n
Het.
Hom.
βSNV ± SE
βSNV/
MeanWT
PSNV
679
20
0
-0.03 ± 0.02
-0.03
0.19
645
69
6
-0.03 ± 0.01
-0.03
6.8x10-3
684
20
0
0.00 ± 0.07
0.00
0.98
650
71
6
0.11 ± 0.03
0.01
1.4x10-3
Corrected SLC13A1 R12X or W48X 684
Calcium
SLC26A1 L348P
650
(mg/dL)
20
0
-0.03 ± 0.06
0.00
0.68
71
0
0.06 ± 0.03
0.01
0.06
Trait
Sulfate-lowering SNV(s)
Whole- SLC13A1 R12X or W48X
body BMD
SLC26A1 L348P
(g/cm2)
Calcium SLC13A1 R12X or W48X
(mg/dL)
SLC26A1 L348P
631
632
633
634
635
636
637
638
639
d.
BMD
Trait
Mid-arm
1/3-arm
Total-arm
Ultradistal-arm
Femoral
neck
Intertrochanter
Total-hip
Trochanter
Sulfate-lowering SNV(s)
n
Het.
Hom.
SLC13A1 R12X or W48X
SLC26A1 L348P
SLC13A1 R12X or W48X
SLC26A1 L348P
SLC13A1 R12X or W48X
SLC26A1 L348P
SLC13A1 R12X or W48X
SLC26A1 L348P
SLC13A1 R12X or W48X
SLC26A1 L348P
SLC13A1 R12X or W48X
SLC26A1 L348P
SLC13A1 R12X or W48X
SLC26A1 L348P
SLC13A1 R12X or W48X
680
646
680
646
680
646
680
646
681
647
680
646
681
647
680
20
70
20
70
20
70
20
70
20
70
20
70
20
70
20
0
6
0
6
0
6
0
6
0
6
0
6
0
6
0
29
βSNV ± SE
βSNV/
PSNV
2
(g/cm )
MeanWT
-0.02 ± 0.01
-0.04
0.08
-0.01 ± 0.01
-0.02
0.03
-0.03 ± 0.02
-0.04
0.08
-0.02 ± 0.01
-0.02
0.03
-0.02 ± 0.01
-0.04
0.06
-0.02 ± 0.01
-0.02
0.01
-0.02 ± 0.01
-0.03
0.26
-0.01 ± 0.01
-0.03
0.03
-0.02 ± 0.03
-0.02
0.46
-0.02 ± 0.01
-0.02
0.26
-0.04 ± 0.04
-0.03
0.35
-0.05 ± 0.02
-0.04 5.1x10-3
-0.03 ± 0.03
-0.03
0.28
-0.04 ± 0.01
-0.04
0.01
-0.03 ± 0.03
-0.04
0.23
640
SLC26A1 L348P
646 70
6
-0.03 ± 0.01
-0.03
0.04
SLC13A1 R12X or W48X 679 20
0
-0.03 ± 0.03
-0.03
0.39
Total-spine
SLC26A1 L348P
645 70
6
-0.03 ± 0.02
-0.03
0.10
Adjusted for age and gender. The P-value of significant associations are highlighted in bold text.
641
Abbreviations: SE, standard error. a) Serum alanine aminotransferase (ALT) and aspartate
642
aminotransferase (AST) levels from the Amish Pharmacogenomics of Anti-Platelet Intervention
643
(PAPI) Study. b) Serum ALT and AST levels from the Amish Family Longevity Study (AFLS). c)
644
Whole-body bone mineral density (BMD) and serum calcium from the Amish PAPI Study. d)
645
Additional BMD measurements from the Amish PAPI Study. Abbreviations: Het., heterozygotes;
646
Hom., homozygotes; WT, wild type.
647
648
649
650
651
652
653
References
654
Agarwala, R., L.G. Biesecker, and A.A. Schaffer, 2003 Anabaptist genealogy database. Am J Med
655
656
Genet C Semin Med Genet 121c (1):32-37.
Ahmad, M., M. Faiyaz Ul Haque, W. Ahmad, H. Abbas, S. Haque et al., 1998 Distinct, autosomal
657
recessive form of spondyloepimetaphyseal dysplasia segregating in an inbred Pakistani
658
kindred. Am J Med Genet 78 (5):468-473.
30
659
660
661
662
663
Alberti, A., P. Pirrone, M. Elia, R.H. Waring, and C. Romano, 1999 Sulphation deficit in "lowfunctioning" autistic children: a pilot study. Biol Psychiatry 46 (3):420-424.
Bowling, F.G., H.S. Heussler, A. McWhinney, and P.A. Dawson, 2013 Plasma and urinary sulfate
determination in a cohort with autism. Biochem Genet 51 (1-2):147-153.
Bozzi, L.M., B.D. Mitchell, J.P. Lewis, K.A. Ryan, W.R. Herzog et al., 2015 The
664
Pharmacogenomics of Anti-Platelet Intervention (PAPI) Study: Variation in Platelet
665
Response to Clopidogrel and Aspirin. Curr Vasc Pharmacol.
666
Council, N.R., 2005 Sulfate, pp. 424--448 in National Research Council. Dietary Reference Intakes
667
for Water, Potassium, Sodium, Chloride, and Sulfate. The National Academies Press,
668
Washington, DC.
669
Dawson, P.A., 2011 Sulfate in fetal development. Semin Cell Dev Biol 22 (6):653-659.
670
Dawson, P.A., 2013 Role of sulphate in development. Reproduction 146 (3):R81-89.
671
Dawson, P.A., and D. Markovich, 2005 Pathogenetics of the human SLC26 transporters. Curr Med
672
Chem 12 (4):385-396.
673
Dawson, P.A., C.S. Russell, S. Lee, S.C. McLeay, J.M. van Dongen et al., 2010 Urolithiasis and
674
hepatotoxicity are linked to the anion transporter Sat1 in mice. J Clin Invest 120 (3):706-
675
712.
31
676
677
678
679
680
681
682
683
684
685
686
687
688
Dawson, P.A., P. Sim, D.W. Mudge, and D. Cowley, 2013 Human SLC26A1 gene variants: a pilot
study. ScientificWorldJournal 2013:541710.
Dodgson, K.S., and R.G. Price, 1962 A note on the determination of the ester sulphate content of
sulphated polysaccharides. Biochem J 84:106-110.
Fernandes, I., G. Hampson, X. Cahours, P. Morin, C. Coureau et al., 1997 Abnormal sulfate
metabolism in vitamin D-deficient rats. J Clin Invest 100 (9):2196-2203.
Giannini, E.G., R. Testa, and V. Savarino, 2005 Liver enzyme alteration: a guide for clinicians.
Cmaj 172 (3):367-379.
Gowda, S., P.B. Desai, V.V. Hull, A.A. Math, S.N. Vernekar et al., 2009 A review on laboratory
liver function tests. Pan Afr Med J 3:17.
Hastbacka, J., A. Kerrebrock, K. Mokkala, G. Clines, M. Lovett et al., 1999 Identification of the
Finnish founder mutation for diastrophic dysplasia (DTD). Eur J Hum Genet 7 (6):664-670.
Hastbacka, J., A. Superti-Furga, W.R. Wilcox, D.L. Rimoin, D.H. Cohn et al., 1996
689
Atelosteogenesis type II is caused by mutations in the diastrophic dysplasia sulfate-
690
transporter gene (DTDST): evidence for a phenotypic series involving three
691
chondrodysplasias. Am J Hum Genet 58 (2):255-262.
32
692
Heafield, M.T., S. Fearn, G.B. Steventon, R.H. Waring, A.C. Williams et al., 1990 Plasma cysteine
693
and sulphate levels in patients with motor neurone, Parkinson's and Alzheimer's disease.
694
Neurosci Lett 110 (1-2):216-220.
695
Hsueh, W.C., B.D. Mitchell, R. Aburomia, T. Pollin, H. Sakul et al., 2000 Diabetes in the Old
696
Order Amish: characterization and heritability analysis of the Amish Family Diabetes Study.
697
Diabetes Care 23 (5):595-601.
698
Klaassen, C.D., and J.W. Boles, 1997 Sulfation and sulfotransferases 5: the importance of 3'-
699
phosphoadenosine 5'-phosphosulfate (PAPS) in the regulation of sulfation. Faseb j 11
700
(6):404-418.
701
Lee, S., P.A. Dawson, A.K. Hewavitharana, P.N. Shaw, and D. Markovich, 2006 Disruption of
702
NaS1 sulfate transport function in mice leads to enhanced acetaminophen-induced
703
hepatotoxicity. Hepatology 43 (6):1241-1247.
704
Lee, W.J., T.I. Pollin, J.R. O'Connell, R. Agarwala, and A.A. Schaffer, 2010 PedHunter 2.0 and its
705
usage to characterize the founder structure of the Old Order Amish of Lancaster County.
706
BMC Med Genet 11:68.
707
708
Markovich, D., 2011a Physiological roles of mammalian sulfate transporters NaS1 and Sat1. Arch
Immunol Ther Exp (Warsz) 59 (2):113-116.
33
709
710
711
712
Markovich, D., 2011b Physiological roles of renal anion transporters NaS1 and Sat1. Am J Physiol
Renal Physiol 300 (6):F1267-1270.
Markovich, D., 2012 Slc13a1 and Slc26a1 KO models reveal physiological roles of anion
transporters. Physiology (Bethesda) 27 (1):7-14.
713
Markovich, D., 2014 Na+-sulfate cotransporter SLC13A1. Pflugers Arch 466 (1):131-137.
714
Mitchell, B.D., W.C. Hsueh, T.M. King, T.I. Pollin, J. Sorkin et al., 2001 Heritability of life span in
715
716
the Old Order Amish. Am J Med Genet 102 (4):346-352.
Mitchell, B.D., P.F. McArdle, H. Shen, E. Rampersaud, T.I. Pollin et al., 2008 The genetic response
717
to short-term interventions affecting cardiovascular function: rationale and design of the
718
Heredity and Phenotype Intervention (HAPI) Heart Study. Am Heart J 155 (5):823-828.
719
Neff, M.W., J.S. Beck, J.M. Koeman, E. Boguslawski, L. Kefene et al., 2012 Partial deletion of the
720
sulfate transporter SLC13A1 is associated with an osteochondrodysplasia in the Miniature
721
Poodle breed. PLoS One 7 (12):e51917.
722
NIST/SEMATECH e-Handbook of Statistical Methods.
723
Noordam, C., V. Dhir, J.C. McNelis, F. Schlereth, N.A. Hanley et al., 2009 Inactivating PAPSS2
724
mutations in a patient with premature pubarche. N Engl J Med 360 (22):2310-2318.
725
O'Connell, J.R., MMAP Documentation.
34
726
Online Mendelian Inheritance in Man, OMIM®, pp. An Online Catalog of Human Genes and
727
Genetic Disorders, edited by J.H.U. McKusick-Nathans Institute of Genetic Medicine,
728
Baltimore, MD.
729
730
731
Online Mendelian Inheritance in Man, OMIM®, pp. MIM Number: 606718, edited by B. Johns
Hopkins University, MD.
Post, W., L.F. Bielak, K.A. Ryan, Y.C. Cheng, H. Shen et al., 2007 Determinants of coronary artery
732
and aortic calcification in the Old Order Amish. Circulation 115 (6):717-724.
733
Shuldiner, A.R., J.R. O'Connell, K.P. Bliden, A. Gandhi, K. Ryan et al., 2009 Association of
734
cytochrome P450 2C19 genotype with the antiplatelet effect and clinical efficacy of
735
clopidogrel therapy. Jama 302 (8):849-857.
736
737
738
Sorkin, J., W. Post, T.I. Pollin, J.R. O'Connell, B.D. Mitchell et al., 2005 Exploring the genetics of
longevity in the Old Order Amish. Mech Ageing Dev 126 (2):347-350.
Streeten, E.A., D.J. McBride, T.I. Pollin, K. Ryan, J. Shapiro et al., 2006 Quantitative trait loci for
739
BMD identified by autosome-wide linkage scan to chromosomes 7q and 21q in men from
740
the Amish Family Osteoporosis Study. J Bone Miner Res 21 (9):1433-1442.
741
742
Yanagishita, M., 1993 Function of proteoglycans in the extracellular matrix. Acta Pathol Jpn 43
(6):283-293.
35
743
Zhao, X., S.K. Onteru, S. Piripi, K.G. Thompson, H.T. Blair et al., 2012 In a shake of a lamb's tail:
744
using genomics to unravel a cause of chondrodysplasia in Texel sheep. Anim Genet 43
745
Suppl 1:9-18.
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
36
764
765
Figure Legends
766
Figure 1. Sulfate transporters in the renal and small intestinal epithelial cell. Modified from
767
Markovich (Markovich 2012). Permission to reuse figure obtained from copyright holder.
768
769
Figure 2. Serum sulfate concentration by SLC13A1 genotype. Box and whisker plots: box
770
represents 2nd and 3rd quartiles (IQR, interquartile range); horizontal band represents median value;
771
whiskers represent 1st and 4th quartiles; ends of whiskers represent minimum and maximum values,
772
excluding outliers; open circles represents outliers (data point more than 1.5*IQR below the 1st
773
quartile or above the 3rd quartile). Association analyses between genotypes and serum sulfate were
774
conducted using a regression-based method that models variation of the trait of interest as a
775
function of measured covariates, measured genotypes, and a polygenic component that accounts for
776
phenotypic correlation due to relatedness. a) Serum sulfate by SLC13A1 N174S genotype. b) Serum
777
sulfate by SLC13A1 R237C genotype. c) Serum sulfate by SLC13A1 nonsense SNV genotype. The
778
P-value for the R12X or W48X Heterozygotes (dark green box and whisker plot) results from the
779
model including carrier status of an SLC13A1 nonsense SNV (R12X or W48X) as a covariate. The
780
P-values for the R12X Heterozygotes and the W48X Heterozygotes (light green box and whisker
781
plots) result from the conditional model including both R12X and W48X as covariates.
782
783
Figure 3. ExWAS of serum sulfate enables discovery of SLC26A1 L348P. a) Manhattan plot
784
depicting association results (plotted as –log10 P-value on y-axis) of individual single-nucleotide
785
polymorphisms distributed across the 22 autosomes (x-axis) from the serum sulfate ExWAS
786
performed using the Illumina Human Exome BeadChip platform (n=900). Red horizontal dotted
37
787
line indicates P=5.0×10−8. Blue horizontal dotted line indicates P=5.0x10-6. b) Serum sulfate by
788
SLC26A1 L348P genotype. Box and whisker plots: box represents 2nd and 3rd quartiles (IQR,
789
interquartile range); horizontal band represents median value; whiskers represent 1st and 4th
790
quartiles; ends of whiskers represent minimum and maximum values, excluding outliers; open
791
circles represents outliers (data point more than 1.5*IQR below the 1st quartile or above the 3rd
792
quartile). Association analyses between genotypes and serum sulfate were conducted using a
793
regression-based method that models variation of the trait of interest as a function of measured
794
covariates, measured genotypes, and a polygenic component that accounts for phenotypic
795
correlation due to relatedness.
796
797
Figure 4. Serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) by
798
SLC13A1 nonsense SNV carrier status. Box and whisker plots: box represents 2nd and 3rd
799
quartiles (IQR, interquartile range); horizontal band represents median value; whiskers represent 1st
800
and 4th quartiles; ends of whiskers represent minimum and maximum values, excluding outliers;
801
open circles represents outliers (data point more than 1.5*IQR below the 1st quartile or above the
802
3rd quartile). Association analyses between genotypes and serum sulfate were conducted using a
803
regression-based method that models variation of the trait of interest as a function of measured
804
covariates, measured genotypes, and a polygenic component that accounts for phenotypic
805
correlation due to relatedness. a) Amish PAPI Study subjects. b) Amish Family Longevity Study
806
(AFLS) subjects.
807
808
Figure 5. Whole-body bone mineral density (BMD) and corrected serum calcium by SLC26A1
809
L348P genotype. Box and whisker plots: box represents 2nd and 3rd quartiles (IQR, interquartile
38
810
range); horizontal band represents median value; whiskers represent 1st and 4th quartiles; ends of
811
whiskers represent minimum and maximum values, excluding outliers; open circles represents
812
outliers (data point more than 1.5*IQR below the 1st quartile or above the 3rd quartile). Association
813
analyses between genotypes and serum sulfate were conducted using a regression-based method
814
that models variation of the trait of interest as a function of measured covariates, measured
815
genotypes, and a polygenic component that accounts for phenotypic correlation due to relatedness.
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
39
831
832
Supplemental Figure Legends
833
Figure S1. Linkage disequilibrium (LD) plots depicting r2 and D’ amongst SLC13A1 SNVs
834
within the entire Amish cohort (n=1,725). The LD plots were created using Haploview. The
835
numbers in the squares on the LD plots are the r2 and D’ values (only the decimals are shown) - r2
836
and D’ values of 1 are not listed. The color code on the r2 plot follow the r2 color scheme for
837
Haploview: white=(r2=0), shades of grey=(0<r2<1), black=(r2=1). The color code on the D’ plot
838
follow the standard color scheme for Haploview: white=(|D′|<1, LOD<2), shades of
839
pink/red=(|D′|<1, LOD≥2), blue=(|D′|=1, LOD<2), bright red=(|D′|=1, LOD≥ 2). Each SNV is
840
labeled with gene name, amino acid change and position, and the reference sequence id (rs-
841
number). The distance between each SNV is indicated on the white bar above.
842
843
Figure S2. Descriptive statistics and distribution of serum sulfate in 977 Old Order Amish
844
research participants.
845
846
Figure S3. Serum sulfate by SLC13A1 R12X genotype. Box and whisker plots: box represents 2nd
847
and 3rd quartiles (IQR, interquartile range); horizontal band represents median value; whiskers
848
represent 1st and 4th quartiles; ends of whiskers represent minimum and maximum values, excluding
849
outliers; open circles represents outliers (data point more than 1.5*IQR below the 1st quartile or
850
above the 3rd quartile). Association analyses between genotypes and serum sulfate were conducted
851
using a regression-based method that models variation of the trait of interest as a function of
852
measured covariates, measured genotypes, and a polygenic component that accounts for phenotypic
853
correlation due to relatedness.
40
854
855
Figure S4. Serum sulfate by SLC13A1 W48X genotype. Box and whisker plots: box represents
856
2nd and 3rd quartiles (IQR, interquartile range); horizontal band represents median value; whiskers
857
represent 1st and 4th quartiles; ends of whiskers represent minimum and maximum values, excluding
858
outliers; open circles represents outliers (data point more than 1.5*IQR below the 1st quartile or
859
above the 3rd quartile). Association analyses between genotypes and serum sulfate were conducted
860
using a regression-based method that models variation of the trait of interest as a function of
861
measured covariates, measured genotypes, and a polygenic component that accounts for phenotypic
862
correlation due to relatedness.
863
864
Figure S5. QQ plot for the serum sulfate ExWAS performed using the Illumina Human
865
Exome BeadChip platform (n=900). a) Unadjusted ExWAS. b) ExWAS adjusted for SLC13A1
866
R12X, SLC13A1 W48X, and SLC26A1 L348P. Note about horizontal line seen at Y=3.4: There are
867
29 SNVs on chromosome 11 in perfect LD with each other and thus all have the same P value.
868
869
Figure S6. Linkage disequilibrium plots depicting r2 and D’ amongst the top serum sulfate
870
ExWAS hits on chromosome 4 from the Illumina Human Exome BeadChip platform (n=900).
871
The LD plots were created using Haploview. The numbers in the squares on the LD plots are the r2
872
and D’ values (only the decimals are shown) - r2 and D’ values of 1 are not listed. The color code
873
on the r2 plot follow the r2 color scheme for Haploview: white=(r2=0), shades of grey=(0<r2<1),
874
black=(r2=1). The color code on the D’ plot follow the standard color scheme for Haploview:
875
white=(|D′|<1, LOD<2), shades of pink/red=(|D′|<1, LOD≥2), blue=(|D′|=1, LOD<2), bright
876
red=(|D′|=1, LOD≥ 2). Each SNV is labeled with gene name, amino acid change and position, and
41
877
the reference sequence id (rs-number). The distance between each SNV is indicated on the white
878
bar above.
879
880
Figure S7. Serum sulfate ExWAS results adjusted for SLC13A1 R12X, SLC13A1 W48X, and
881
SLC26A1 L348P (n=900).. Manhattan plot depicting association results (plotted as –log10 P-value
882
on y-axis) of individual single-nucleotide polymorphisms distributed across the 22 autosomes (x-
883
axis) from the serum sulfate ExWAS performed using the Illumina Human Exome BeadChip
884
platform. Red horizontal dotted line indicates P=5.0×10−8. Blue horizontal dotted line indicates
885
P=5.0x10-6.
886
887
Figure S8. QQ plot for serum sulfate GWAS using Affymetrix GeneChip platform (n=917).
888
889
Figure S9. Serum sulfate GWAS results (n=917). Manhattan plot depicting association results
890
(plotted as –log10 P-value on y-axis) of individual single-nucleotide polymorphisms distributed
891
across the 22 autosomes (x-axis) from the serum sulfate GWAS performed using the Affymetrix
892
GeneChip platform. Red horizontal dotted line indicates P=5.0×10−8. Blue horizontal dotted line
893
indicates P=5.0x10-6. Affymetrix GeneChip does not contain SLC13A1 R12X, SLC13A1 W48X, or
894
SLC26A1 L348P.
895
896
Figure S10. Linkage disequilibrium plots depicting r2 and D’ amongst SLC26A1 L348P (from
897
the Illumina Human Exome BeadChip platform) and the top serum sulfate GWAS hits on
898
chromosome 7 from the Affymetrix GeneChip platform (n=879, subject overlap between
899
Illumina Human Exome BeadChip and Affymetrix GeneChip platforms). The LD plots were
42
900
created using Haploview. The numbers in the squares on the LD plots are the r2 and D’ values (only
901
the decimals are shown) - r2 and D’ values of 1 are not listed. The color code on the r2 plot follow
902
the r2 color scheme for Haploview: white=(r2=0), shades of grey=(0<r2<1), black=(r2=1). The color
903
code on the D’ plot follow the standard color scheme for Haploview: white=(|D′|<1, LOD<2),
904
shades of pink/red=(|D′|<1, LOD≥2), blue=(|D′|=1, LOD<2), bright red=(|D′|=1, LOD≥ 2). Each
905
SNV is labeled with gene name, amino acid change and position, and the reference sequence id (rs-
906
number). The distance between each SNV is indicated on the white bar above.
907
908
Figure S11. Linkage disequilibrium plots depicting r2 and D’ amongst SLC13A1 W48X (from
909
the Illumina Human Exome BeadChip platform) and the top serum sulfate GWAS hits on
910
chromosome 4 from the Affymetrix GeneChip platform (n=879, subject overlap between
911
Illumina Human Exome BeadChip and Affymetrix GeneChip platforms). The LD plots were
912
created using Haploview. The numbers in the squares on the LD plots are the r2 and D’ values (only
913
the decimals are shown) - r2 and D’ values of 1 are not listed. The color code on the r2 plot follow
914
the r2 color scheme for Haploview: white=(r2=0), shades of grey=(0<r2<1), black=(r2=1). The color
915
code on the D’ plot follow the standard color scheme for Haploview: white=(|D′|<1, LOD<2),
916
shades of pink/red=(|D′|<1, LOD≥2), blue=(|D′|=1, LOD<2), bright red=(|D′|=1, LOD≥ 2). Each
917
SNV is labeled with gene name, amino acid change and position, and the reference sequence id (rs-
918
number). The distance between each SNV is indicated on the white bar above.
919
920
Figure S12. QQ plot for serum sulfate GWAS adjusted for rs17684427 and rs362272 using
921
Affymetrix GeneChip platform (n=917).
922
43
923
Figure S13. Serum sulfate GWAS results adjusted for rs17684427 and rs362272 (n=917).
924
Manhattan plot depicting association results (plotted as –log10 P-value on y-axis) of individual
925
single-nucleotide polymorphisms distributed across the 22 autosomes (x-axis) from the serum
926
sulfate GWAS performed using the Affymetrix GeneChip platform. Red horizontal dotted line
927
indicates P=5.0×10−8. Blue horizontal dotted line indicates P=5.0x10-6. Affymetrix GeneChip does
928
not contain SLC13A1 R12X, SLC13A1 W48X, or SLC26A1 L348P.
44
a.
b.