Genetic Variants Associated with Circulating Parathyroid Hormone

Abstract

Parathyroid hormone (PTH) is a primary calcium regulatory hormone. Elevated serum PTH concentrations in primary and secondary hyperparathyroidism have been associated with bone disease, hypertension, and in some studies, cardiovascular mortality. Genetic causes of variation in circulating PTH concentrations are incompletely understood. We performed a genome-wide association study of serum PTH concentrations among 29,155 participants of European ancestry from 13 cohort studies (n=22,653 and n=6502 in discovery and replication analyses, respectively). We evaluated the association of single nucleotide polymorphisms (SNPs) with natural log-transformed PTH concentration adjusted for age, sex, season, study site, and principal components of ancestry. We discovered associations of SNPs from five independent regions with serum PTH concentration, including the strongest association with rs6127099 upstream of CYP24A1 (P=4.2 × 10⁻⁵³), a gene that encodes the primary catabolic enzyme for 1,25-dihydroxyvitamin D and 25-dihydroxyvitamin D. Each additional copy of the minor allele at this SNP associated with 7% higher serum PTH concentration. The other SNPs associated with serum PTH concentration included rs4074995 within RGS14 (P=6.6 × 10⁻¹⁷), rs219779 adjacent to CLDN14 (P=3.5 × 10⁻¹⁶), rs4443100 near RTDR1 (P=8.7 × 10⁻⁹), and rs73186030 near CASR (P=4.8 × 10⁻⁸). Of these five SNPs, rs6127099, rs4074995, and rs219779 replicated. Thus, common genetic variants located near genes involved in vitamin D metabolism and calcium and renal phosphate transport associated with differences in circulating PTH concentrations. Future studies could identify the causal variants at these loci, and the clinical and functional relevance of these variants should be pursued.

Parathyroid hormone (PTH) is the major calcium regulatory hormone in humans. The physiologic role of PTH is to maintain serum calcium concentration by (1) stimulating calcium release from bone, (2) enhancing renal calcium reabsorption, and (3) catalyzing the production of 1,25-dihydroxyvitamin D (1,25[OH]₂D), the active form of vitamin D, from 25-hydroxyvitamin D (25[OH]D).^1–6 Higher circulating PTH concentrations are associated with increased risk of fracture, hypertension, left ventricular hypertrophy, and in some studies with cardiovascular mortality.^7–11

PTH is regulated in response to circulating calcium concentrations by the calcium-sensing receptor, which is highly expressed in parathyroid tissue.^12–14 PTH concentration is further modulated by calcitriol, which interacts with DNA regulatory elements via the vitamin D nuclear receptor to suppress PTH, and is an established treatment for the secondary hyperparathyroidism seen in CKD.^15,16 Additional causes of secondary hyperparathyroidism include calcium deficiency, vitamin D deficiency, vitamin D resistance, hypercalciuria, hyperphosphatemia, and vitamin D nuclear receptor abnormalities.^17,18

A genetic basis for variability in PTH concentrations is suggested by rare Mendelian disorders that cause primary hyper- and hypoparathyroidism.^19–27 Further, a classic twin study estimated that 54%–65% of variation in PTH concentration (broad-sense heritability) could be due to genetic/familial factors.²⁸ Additionally, mutations in cell cycle and cell regulatory genes have been described in surgically removed parathyroid adenomas^29,30; however, the contribution of genetic changes to the initial development of nonadenomatous parathyroid disease is unclear.

To investigate the role of common genetic variants (single nucleotide polymorphisms, SNPs) on variation in PTH concentration we conducted a genome wide association study (GWAS) meta-analysis of serum PTH concentrations in 27,561 individuals from 13 cohort studies.

Results

Genome-Wide Associations for Serum PTH Concentrations in Cohorts of European Ancestry

A total of 22,653 participants from ten cohort studies were available for the discovery meta-analysis.^31–40 The majority of studies were community-based or population-based cohort studies of healthy adults from the United States and Europe. (Table 1) There was little evidence for population stratification at study level (median genomic inflation factor, λ=1.02) or meta-analysis level (λ=1.03). Inspection of the quantile-quantile plot suggested an excess of association signals beyond those expected (Supplemental Figure 1). SNPs in five independent regions (chromosomes 3, 5, 20, 21, and 22) exceeded the threshold of genome-wide significance for association with PTH concentration (Figure 1).

Table 1.

Characteristics of study participants

Variables	ARIC (Whites)	CHS (Whites)	DCCT	Indiana	MESA (Whites)	MrOS	NESDA	OGP–Talana	SHIP-1	SHIP-Trend	GOOD	LURIC	TwinsUK	Amish
Cohort information	Discovery Cohorts										Replication Cohorts
Study design	Community-based	Community-based	RCT	Community-based	Community-based	Community-based	Community (19%), general practice (54%), and secondary mental health care (27%)	Population- and family-based	Population-based	Population-based	Population-based	Hospital-based case-control study	Twins registry	Community-based
Overall sample size with PTH and genotype	8135	1782	278	994	2455	1785	2393	696	3176	977	911	2729	1982	621
Mean age, yr (SD)	57.1 (5.7)	73.8 (4.8)	32.4 (6.9)	36.4 (8.5)	62.6 (10.3)	73.7 (5.8)	42.3 (13.0)	49.2 (20.8)	54.3 (15.2)	50.1 (13.7)	18.9 (0.6)	62.9 (10.6)	48.2 (13.2)	53.5 (13.9)
Men, n (%)	3747 (46.1)	680 (30.3)	154 (55.4)	0 (0)	1164 (47.4)	1785 (100)	813 (34.0)	284 (41.3)	1526 (48.0)	432 (44.2)	911 (100)	1879 (68.9)	44 (2.22)	265 (42.7)
Season of PTH measurement
Winter, n (%)	1973 (24.3)	564 (25.1)	82 (29.5)	198 (19.9)	682 (27.8)	350 (19.6)	628 (26.2)	696 (100)	773 (24.3)	184 (18.8)	13 (1.4)	593 (21.7)	646 (32.6)	183 (29.5)
Spring, n (%)	2410 (29.6)	515 (23.0)	71 (25.5)	249 (25.1)	658 (26.8)	468 (26.2)	578 (24.2)	0 (0)	900 (28.3)	293 (30.0)	336 (36.9)	549 (20.1)	645 (32.5)	183 (29.5)
Summer, n (%)	1984 (24.4)	650 (29.0)	56 (20.1)	268 (26.9)	522 (21.3)	515 (28.9)	592 (24.7)	0 (0)	678 (21.4)	268 (27.4)	215 (23.6)	754 (27.6)	414 (20.9)	187 (30.1)
Fall, n (%)	1768 (21.7)	514 (22.9)	69 (24.8)	279 (28.1)	593 (24.2)	452 (25.3)	595 (24.9)	0 (0)	825 (26.0)	232 (23.8)	347 (38.1)	833 (30.5)	277 (14.0)	68 (11.0)
BMI, kg/m2 (SD)	27.3 (5.0)	26.5 (4.6)	25.9 (3.5)	26.3 (6.0)	27.7 (5.1)	27.5 (3.7)	25.6 (5.0)	26.3 (4.9)	28.0 (4.9)	27.3 (4.6)	22.3 (3.1)	27.5 (4.1)	26.0 (4.9)	28.0 (5.1)
eGFR (MDRD), ml/min per 1.73 m2 (SD)	87.8 (13.2)	79.6 (18.7)	106.3 (13.4)	NA	76.0 (17.1)	77.7 (16.7)	103.7 (15.7)	77.6 (21.5)	89.2 (21.0)	96.0 (18.0)	100.1 (14.5)	82.1 (18.3)	83.6 (16.2)	90.2 (19.8)
Mean Serum PTH, pg/ml (SD)	40.4 (14.6)	56.3 (28.8)	30.6 (13.3)	29.5 (13.2)	40.6 (16.8)	32.8 (12.1)	55.7 (27.8)	39.4 (25.1)	37.6 (19.0)	34.7 (20.7)	39.3 (14.7)	32.9 (17.9)	34.5 (18.1)	57.2 (19.6)

Indiana, Indiana Sisters Cohort; NESDA, Netherlands Study of Depression and Anxiety; OGP–Talana, Ogliastra Genetic Park–Talana Study; GOOD, Gothenburg Osteoporosis and Obesity Determinants; Amish, Amish Family Osteoporosis Study; MDRD, Modification of Diet in Renal Disease.

Figure 1.

Graphical summary of genome-wide association results. Manhattan plot of the strength of association of SNPs ln-transformed PTH, on the basis of association with 22,653 participants for autosomal SNPs and 17,865 for X-chromosome (discovery stage).

The SNP having the strongest association with PTH concentration was rs6127099 (P=4.2 × 10⁻⁵³; Supplemental Figure 2A, Table 2). This SNP lies 38 kbp upstream of CYP24A1 (cytochrome P450, family 24, subfamily A, polypeptide 1) on chromosome 20q13.2. Each additional copy of the rs6127099 T allele is associated with approximately 7% higher PTH concentration, after adjustment for age, sex, the first ten principal components of ancestry (where available), geographic site (where applicable), and season of PTH measurement (model 1). The CYP24A1 gene encodes the major enzyme responsible for catabolizing calcitriol to water-soluble 1,24,25-trihydroxyvitamin D and 25(OH)D to 24,25-dihydroxyvitamin D (24,25[OH]₂D) for excretion.⁴¹ Further adjustment for body mass index (BMI) and eGFR (model 2) did not alter the magnitude of the association (7% higher PTH concentration after adjustment; P=2.2 × 10⁻⁴³).

Table 2.

Associations of top single nucleotide polymorphisms with ln-transformed serum PTH concentrations

SNP	Nearest Gene	Chr	Position	PTH-Increasing Allele	Other Allele	PTH-Increasing Allele Frequencya	Model 1						Model 2
							Discovery (n=22,653)		Replication (n=6502)		Discovery + Replication (n=29,165)		Discovery (n=22,653)		Replication (n=4908b)
							βc (SEM)	P Value	β (SEM)	P Value	β (SEM)	P Value	β (SEM)	P Value	β (SEM)	P Value
rs6127099	CYP24A1	20	52,731,402	T	A	0.34	0.07 (0.003)	4.2 × 10−53	0.07 (0.011)	9.1 × 10−11	0.07 (0.003)	2.4 × 10−72	0.07 (0.005)	2.2 × 10−43	0.08 (0.012)	2.3 × 10−11
rs4074995	RGS14	5	176,797,343	G	A	0.71	0.03 (0.003)	6.6 × 10−17	0.05 (0.012)	4.3 × 10−4	0.03 (0.003)	3.3 × 10−23	0.03 (0.005)	7.7 × 10−12	0.05 (0.012)	9.3 × 10−5
rs219779	CLDN14	21	37,833,751	G	A	0.75	0.04 (0.003)	3.5 × 10−16	0.05 (0.011)	4.3 × 10−5	0.04 (0.003)	8.9 × 10−22	0.03 (0.005)	1.6 × 10−11	0.05 (0.012)	1.6 × 10−4
rs4443100	RTDR1	22	23,372,864	G	C	0.32	0.02 (0.003)	8.7 × 10−9	0.01 (0.01)	0.50	0.02 (0.003)	4.1 × 10−11	0.02 (0.005)	3.0 × 10−7	0.01 (0.012)	0.20
rs73186030	CASR	3	122,013,465	T	C	0.14	0.03 (0.003)	4.8 × 10−8	0.03 (0.01)	0.02	0.03 (0.004)	1.2 × 10−9	0.03 (0.006)	6.4 × 10−7	0.03 (0.015)	0.04

In aggregate, rs6127099, rs4074995, rs219779, rs4443100, and rs73186030 together accounted for 4.2% of circulating PTH variation. Only top SNP from each region shown. Model 1 includes age, sex, season of PTH measurement, geographic site (if applicable), and first ten principal components of ancestry (if available). Model 2 additionally adjusts for BMI, eGFR, and eGFR-squared. Top SNPs were defined as the most significant SNP (lowest P value) within a 500 kb window. Chr, chromosome.

^aAllele frequency data from 1000 Genome Phase 1 genotype data (European Super Population [EUR]).

^bResults for model 2 were not available for the Amish Family Osteoporosis Study.

^cβ-estimates are interpreted as the absolute difference in ln(PTH) per PTH-increasing allele, e.g., +0.07 is a 0.07-unit higher ln(PTH) per T allele. The exponentiated β coefficients are interpreted as the relative difference in PTH concentration per T allele, e.g., for rs6127099, every allele is associated with a e^0.07-fold, or 1.073-fold higher, or 7.3% higher PTH concentration per T allele.

rs4074995 was associated with PTH concentrations (P=6.6 × 10⁻¹⁷, Table 2) and is located on chromosome 5q35.3. This SNP is intronic within RGS14 (regulator of G-protein signaling 14).⁴² However, rs4074995 is in strong linkage disequilibrium (LD) with SNPs in the directly adjacent gene, SLC34A1 (solute carrier family 34 [type 2 sodium/phosphate cotransporter], member 1). SLC34A1 is a strong candidate gene for PTH concentration, as it encodes the kidney-specific sodium-phosphate type 2a transporter (Npt2a) responsible for phosphate reabsorption in the proximal tubule.⁴³

The third SNP robustly associated with serum PTH concentrations was rs219779 (P=3.5 × 10⁻¹⁶, Table 2), located in a transcript (AP000695.6) on chromosome 21q22.13, adjacent to CLDN14 (Claudin 14). Each copy of the rs219779 G allele was associated with approximately 4% higher PTH concentration, after full adjustment. CLDN14 is a candidate gene for serum PTH concentration, as it plays a major role in tight junction-specific obliteration of the intercellular space, through calcium-independent cell-adhesion activity.⁴⁴

rs4443100 was associated with PTH concentration and is located approximately 20 kb 3′ of RTDR1 (RSPH14, radial spoke head 14 homolog) on chromosome 22q11.23 (P=8.7 × 10⁻⁹, Table 2). This gene encodes a protein with no known function but is located in a region deleted in pediatric rhabdoid tumors of the kidney.⁴⁵

rs73186030 was also significantly associated with PTH concentration (P=4.8 × 10⁻⁸, Table 2), and is located approximately 8 kb 3′ of CASR (calcium-sensing receptor) on chromosome 3q13.33. The CaSR protein is a G protein–coupled receptor that is expressed primarily in the parathyroid gland and the ascending loop of the kidney. This receptor senses small changes in circulating calcium concentration and couples this information to intracellular signaling pathways that modify PTH secretion and renal cation handling, thus this protein plays an essential role in maintaining mineral ion homeostasis.^46,47

Analyses of the CYP24A1 region, conditioning on rs6127099, identified one additional locus which was independently associated with circulating PTH (rs35194449; PTH-increasing allele: T [frequency=0.24]; effect estimate: 4% higher PTH per additional copy of T-allele; p_c=1.8 × 10⁻¹⁰; r² between SNP and rs6127099 = 0.00). Conditional analyses of the remaining four regions did not identify any secondary association signals, indicating no additional independently-associated SNPs after conditioning on the region’s lead SNP.

In aggregate, the loci (rs6127099, rs35194449, rs4074995, rs219779, rs4443100, and rs73186030) explained 4.5% of the variance in circulating PTH.

Replication in Four Cohorts of European Ancestry

A total of 6502 individuals were available for the replication meta-analysis of model 1 results, and 4908 for model 2. Regression coefficients for each of the top five SNPs in the discovery sample were in the same direction in the replication sample (Table 2). Replication associations were considered significant at the Bonferroni-corrected P=0.01 level for three of five SNPs meta-analyzed in replication. Post hoc power calculations of SNPs rs6127099, rs4074995, rs219779, rs4443100, and rs73186030 for replication analyses with 6502 individuals revealed 99%, 77%, 96%, 36%, and 31% power, respectively, to detect the β coefficients from the discovery analyses, with a type 1 error α=0.01. Primary regression coefficients and interpretation of our results were not affected by additional adjustment for BMI and eGFR in sensitivity analyses (Table 1).

Replication in Populations of African Ancestry

In black populations, data were available for four of the five SNPs. Three of the four SNPs had β coefficients that were direction-consistent with the primary analysis and one SNP (rs4074995) was significantly (P<0.01) associated (Table 3). rs219779 and rs4443100 showed the most allelic differentiation between individuals of African and European ancestry. The frequencies of PTH-increasing alleles differed across the five Super Populations assessed in the 1000 Genomes Project (Supplemental Table 5).

Table 3.

Associations of top single nucleotide polymorphisms with ln-transformed serum PTH concentrations among individuals of black descent (n=4279)

SNP	Nearest Gene	Chr	Position	PTH-Increasing Allele	Other Allele	PTH-Increasing Allele Frequencya	β (SEM)b	P Value	Fstc
rs6127099	CYP24A1	20	52,731,402	T	A	0.21	+0.03 (0.0136)	0.0363	0.020
rs4074995	RGS14	5	176,797,343	G	A	0.92	−0.04 (0.0160)	0.0059	0.057
rs219779	CLDN14	21	37,833,751	G	A	0.69	−0.01 (0.0096)	0.5337	0.102
rs4443100	RTDR1	22	23,372,864	C	G	0.80	−0.02 (0.0106)	0.05366	0.100
rs73186030	CASR	3	122,013,465	T	C	0.01	NA	NA	0.055

Chr, chromosome; Fst, fixation index.

^aMAF data from 1000 Genome Phase 1 genotype data (African Super Population [AFR]).

^bEffect estimate adjusted for age, sex, season of PTH measurement, geographic site (if applicable), and first ten principal components of ancestry (if available) and are interpreted as the absolute difference in ln(PTH) per minor allele, e.g., +0.03 is a 0.03-unit higher ln(PTH) per minor allele. The exponentiated β coefficients are interpreted as the relative difference in PTH concentration per T allele, e.g., for rs6127099, every T allele is associated with a e^0.03-fold, or 1.03-fold higher, or 3% higher PTH concentration per minor allele.

^cFst is an estimate of genetic differentiation calculated using the variance in allele frequencies among European and African samples from the 1000 Genomes and standardized according to the mean allele frequency in the combined sample.⁷³

Expression Quantitative Trait Locus and Functional Prediction

For each of the replicated SNPs in European populations, we identified all proxy SNPs with r²>0.8 in the 1000 Genomes Pilot 1 pairwise LD data, yielding a total of 120 SNPs. We then queried each of these SNPs in the expression quantitative trait locus (eQTL) database of the Phenotype-Genotype Integrator and the Genotype-Tissue Expression project.^48–50 rs4074995 was associated with gene expression at RGS14 (strongest association in transformed fibroblast cells, P=1.0 × 10⁻²⁸), F12 in esophagus mucosa (P=1.2 × 10⁻⁷), and FGFR4 in tibial nerve tissue (P=6.8 × 10⁻⁷), whereas rs219779 was associated with expression at AP000695.4 in skeletal muscle tissue (P=9.1 × 10⁻⁶). Associations between SNPs and SLC34A1 and CLDN14 expression could not be analyzed, as no information for these genes was included in either database. Furthermore, neither kidney nor parathyroid tissue eQTL data were available to be queried.

Associations with Bone Mineral Density

Each of the three replicated SNPs was associated (P<0.01) with femoral neck bone density in the Genetic Factors of OSteroporosis (GEFOS) consortium.⁵¹ Each additional copy of the PTH-increasing allele at rs6127099 (CYP24A1), rs4074995 (RGS14/SLC34A1), and rs219779 (CLDN14) was associated with a +0.030 (SEM=0.009, P=5 × 10⁻⁴), +0.023 (SEM=0.008, P=8 × 10⁻³), and −0.022 (SEM=0.008, P=9 × 10⁻³) difference in bone mineral density (BMD) SD.

Associations with Mineral Metabolism Biomarkers

rs6127099 (CYP24A1) is significantly associated with circulating concentrations of PTH, and circulating fibroblast growth-factor 23 (FGF-23), serum phosphorus, and urine concentrations of calcium and phosphate normalized to urine creatinine (Table 4). rs4443100 (RTDR1) and rs219779 (CLDN14) are associated with PTH concentrations and with FGF-23 concentrations. rs4074995 (RGS14/SLC34A1) is also associated with FGF-23 and phosphorus concentrations, and 25(OH)D and calcium concentrations. Serum concentrations of 24,25(OH)₂D were not associated with any of the most associated SNPs identified in the discovery/replication GWAS meta-analysis (Table 4).

Table 4.

Associations of top single nucleotide polymorphisms with mineral metabolism measurements

SNP and Nearest Gene	PTH, pg/ml		FGF23, pg/ml		24,25D, ng/ml		25D, ng/ml		Serum Calcium, mg/dl		Serum Phosphorus, mg/dl		Urine Calcium-to-Creatinine Ratio, mg/mg		Urinary FePO4
	N	Mean (SD)	N	Mean (SD)	N	Mean(SD)	N	Mean (SD)	N	Mean (SD)	N	Mean (SD)	N	Mean (SD)	N	Mean (SD)
rs6127099
CYP24A1
AA	5164	38.88 (10.5)	5164	44.17 (11.3)	1244	4.54 (2.7)	5164	27.93 (6.7)	6638	9.44 (0.24)	6638	3.57 (0.3)	1244	0.102 (0.07)	1244	12.71 (5.7)
AT	3906	41.76 (11.1)	3906	42.24 (12.9)	1009	4.51 (2.7)	3906	27.17 (6.6)	5041	9.44 (0.24)	5041	3.57(0.3)	1009	0.102 (0.07)	1009	12.62 (5.3)
TT	703	44.08 (11.8)	703	40.83 (10.2)	202	4.69 (2.8)	703	27.16 (6.1)	907	9.45 (0.26)	907	3.55 (0.3)	202	0.090 (0.06)	202	11.78 (5.3)
P-for-trenda	<0.001		<0.001		0.46		0.004		0.33		0.02		0.03		0.03
rs4074995
RGS14
AA	830	37.22 (9.19)	830	40.79 (9.97)	202	4.66 (2.8)	830	28.02 (6.6)	1053	9.45 (0.25)	1053	3.52 (0.3)	202	0.105 (0.07)	202	12.44 (5.0)
GA	4173	40.08 (11.1)	4173	42.78 (12.3)	968	4.49 (2.6)	4173	27.59 (6.6)	5416	9.44 (0.24)	5416	3.56 (0.3)	968	0.101 (0.07)	968	12.70 (5.4)
GG	5587	41.23 (11.1)	5587	43.69 (11.7)	1285	4.55 (2.7)	5587	27.44 (6.6)	7109	9.44 (0.24)	7109	3.58 (0.3)	1285	0.101 (0.07)	1285	12.54 (5.7)
P-for-trend	<0.001		<0.001		0.59		0.02		<0.001		<0.001		0.46		0.81
rs219779
CLDN14
AA	645	39.12 (11.2)	645	44.46 (12.4)	168	4.53 (2.5)	645	27.63 (6.5)	828	9.45 (0.23)	828	3.58 (0.3)	168	0.087 (0.06)	168	12.09 (5.2)
AG	3823	39.91 (10.6)	3823	42.98 (11.7)	907	4.55 (2.7)	3823	27.41 (6.6)	4858	9.44 (0.24)	4858	3.57 (0.3)	907	0.104 (0.07)	907	12.56 (5.3)
GG	5878	40.89 (11.0)	5878	43.03 (12.1)	1380	4.53 (2.7)	5878	27.62 (6.6)	7647	9.45 (0.24)	7647	3.57 (0.3)	1380	0.101 (0.07)	1380	12.68 (5.7)
P-for-trend	0.001		0.004		1.00		0.99		0.65		0.39		0.02		0.19
rs4443100
RTDR1
CC	4898	39.90 (10.7)	4898	43.60 (13.1)	1188	4.48 (2.6)	4898	27.56 (6.5)	6246	9.46 (0.24)	6246	3.57 (0.3)	1188	0.099 (0.07)	1188	12.62 (5.4)
CG	4406	40.86 (11.2)	4406	42.73 (11.2)	1014	4.61 (2.7)	4406	27.73 (6.7)	5680	9.42 (0.25)	5680	3.57 (0.3)	1014	0.102 (0.07)	1014	12.52 (5.8)
GG	994	41.74 (11.3)	994	42.64 (10.1)	253	4.52 (2.8)	994	26.87 (6.5)	1340	9.43 (0.23)	1340	3.58 (0.3)	253	0.107 (0.07)	253	12.80 (5.1)
P-for-trend	<0.001		0.02		0.83		0.003		0.001		0.20		0.11		0.64
rs73186030
CASR
CC	7817	40.00 (10.9)	7817	43.05 (12.3)	1746	4.54 (2.7)	7817	27.61 (6.7)	9916	9.41 (0.24)	9916	3.59 (0.3)	1746	0.101 (0.07)	1746	12.45 (5.6)
CT	2456	41.76 (11.3)	2456	43.32 (11.4)	636	4.52 (2.5)	2456	27.37 (6.5)	3264	9.51 (0.24)	3264	3.53 (0.3)	636	0.102 (0.07)	636	12.94 (5.5)
TT	212	42.15 (10.4)	212	42.98 (9.68)	73	4.60 (2.6)	212	27.17 (5.5)	283	9.55 (0.21)	283	3.53 (0.3)	73	0.103 (0.07)	73	13.03 (4.8)
P-for-trend	0.01		0.94		0.85		0.35		<0.001		0.001		0.81		0.38

24,25D, 24,25-dihydroxyvitamin D; 25D, 25-hydroxyvitamin D; FePO₄, Fractional excretion of phosphorus.

^aP-for-trend obtained from Wald test of linear regression coefficient for number of copies of the minor allele.

Discussion

In this first reported GWAS of PTH concentration, we identified five loci, located on chromosomes 3, 5, 20, 21, and 22, that were associated with variation in PTH concentrations. Upon replication, three of the five loci were associated with PTH concentration after Bonferroni correction. The strongest association was observed for rs6127099, located 38 kbp upstream of CYP24A1, encoding the primary catabolic enzyme for 1,25(OH)₂D. Other significant associations in discovery and replication included SNPs located near genes encoding the renal type 2a sodium-phosphate cotransporter and claudin 14. Taken together, these findings provide the first human evidence linking PTH concentrations and common polymorphisms near genes involved with vitamin D and calcium mineral metabolism.

The cytochrome p450 enzyme CYP24A1 is critically important for maintaining serum 1,25(OH)₂D concentrations within a tight physiologic range and preventing vitamin D toxicity by catalyzing the conversion of 1,25(OH)₂D to 1,24,25(OH)₃D for subsequent excretion.⁴¹ Inactivating mutations of CYP24A1 cause severe infantile hypercalcemia, hypercalcemic syndrome in adults, and nephrolithiasis, with potent suppression of PTH.^52–58 We found the T allele of rs6127099 to be associated with greater serum PTH concentrations, suggesting that the associated haplotype may confer increased CYP24A1 activity, accelerated 1,25(OH)₂D catabolism, and disinhibition of PTH levels. The association of the T allele of rs6127099 with lower serum concentrations of FGF-23, which is upregulated by 1,25(OH)₂D, is also consistent with increased CYP24A1 activity and accelerated 1,25(OH)2D catabolism. CYP24A1 also converts 25(OH)D to 24,25(OH)₂D; however, we detected only small differences in serum 25(OH)D and no differences in serum 24,25(OH)₂D concentrations according to the number of rs6127099 minor alleles. This suggests that environmental factors (diet and sunlight) play a more important role than genetic variation for circulating 25(OH)D.⁵⁹ Of note, rs6127099 is in near perfect linkage (or LD) with a variant (i.e., rs1570669) which has been associated with lower circulating calcium and with BMD at the lumbar spine.⁶⁰ Differences in 1,25(OH)₂D catabolism could have important clinical implications for the therapeutic approach to parathyroid disorders; future studies are needed to test whether rs6127099 is associated with variation in the response to vitamin D treatment.

rs4074995 is located within RGS14, which encodes a G-protein signal-regulating protein, and is in LD with multiple SNPs within the gene coding for type 2a sodium phosphate cotransporter (Npt2a, SLC34A1) expressed in the proximal tubule of the kidney.⁶¹ In a previous GWAS, as in ours, rs4074995 was found to be associated with serum phosphate concentrations.⁶² Hypothetically, this variant could lead to greater ion leakage in the sodium channel, leading to reduced need for phosphaturic hormones. Not surprisingly, we did not detect differences in the spot urinary excretion of calcium or phosphate according to the number of rs4074995 minor alleles; these parameters are highly influenced by dietary intake. The functional relevance of this polymorphism to renal phosphate transport and PTH regulation awaits further elucidation.

CLDN14 encodes a tight junction protein expressed in the thick ascending limb of the kidney that inhibits calcium and magnesium reabsorption, and is regulated by the CaSR.⁴⁴ The CaSR, located at the surface of parathyroid cells, mediates their detection of small changes in blood ionized calcium concentration, and subsequent modification of PTH release.^44,63,64 Located in a CLDN14 intron, rs219779 is in LD with several other SNPs within this gene. The rs219779 allele was associated with significantly lower urinary calcium excretion and circulating PTH, consistent with enhanced CLDN14 function. The common variant rs219780, 400 bp upstream of, and in moderate LD with, rs219779 (r²=0.78) has been associated with kidney stones and BMD.⁶⁵ Further studies are needed to identify causal variants in this region that affect renal calcium handling and modulate PTH.

Perhaps surprisingly, common SNPs near CYP24A1, SLC34A1, and CLDN14 were more strongly associated with serum PTH concentrations than those located near or within CASR, the gene encoding the primary receptor involved in PTH regulation. This may have been due to differences in minor allele frequency (MAF), however, as the CASR locus had the lowest MAF of the discovered loci. Post hoc power calculations of the replication analysis of rs73186030 to detect the same β coefficient as the discovery analyses showed 31% power with a type 1 error α=0.01. Nevertheless, rs73186030, which is in LD with several SNPs within CASR, was associated with significant differences in serum PTH in combined discovery and replication analyses. Rare genetic disruptions of the CASR gene cause familial hypocalciuric hypercalcemia and autosomal dominant hypoparathyroidism with hypocalcemia, and rs1801725, a common polymorphism within CASR, was previously identified to be associated with serum calcium concentrations in a GWAS.^60,63,66 The rs73186030 identified here is in strong LD with rs1801725 (R²=0.925, D′=1.00), and was associated with differences in serum calcium and phosphate concentrations.

Strengths of our study include the large and diverse sample, the population-based settings, and the comprehensive set of common genetic variants and mineral metabolites examined. Potential limitations include a restriction to common variants only and discovery efforts in an exclusively European American sample. The top SNPs identified in this study were located near genes encoding proteins that are typically expressed in the kidneys and parathyroid tissue and are therefore unlikely to be related to gene expression in circulating white blood cells or fibroblasts.^49,50 Follow-up expression work in renal and parathyroid tissue is required and may shed additional light on the functional relevance of the identified polymorphisms.

Although the observed variants were associated with mild differences in PTH concentrations, it is possible that the effects may be more pronounced among individuals with mineral metabolism disturbances, such as those with CKD. It will be crucial to test these results among individuals with CKD, in whom the biologic implications and potential treatment options are most relevant.

In summary, we demonstrate that common genetic variants are associated with circulating PTH concentrations in adults using meta-analysis of study-specific GWAS from thirteen large population-based studies. Follow-up studies are needed to identify potential causal genetic loci in or near the identified candidate genes. Candidate genes may be explored in more comprehensive metabolic studies of PTH metabolism and in translational animal models that will shed new light on the mechanisms and clinical implications of PTH in calcium homeostasis and its treatment.

Concise Methods

Subjects

Discovery Study Populations

Ten cohorts contributed to the discovery meta-analysis, by providing study-specific genome-wide analyses of PTH concentrations, for a total of 22,653 individuals of European ancestry (Table 1). Contributing studies included the Atherosclerosis Risk in Communities Study (ARIC; number of individuals of European ancestry, n=8135),³¹ the Cardiovascular Health Study (CHS; n=1782),³² the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications (DCCT/EDIC; n=278),³³ the Indiana Sisters Study (n=994),³⁴ the Multi-Ethnic Study of Atherosclerosis (MESA; n=2446),³⁵ the Osteoporotic Fractures in Men Study (MrOS; n=1785),³⁶ the Netherlands Study of Depression and Anxiety (n=2393),³⁷ the Ogliastra Genetic Park–Talana Study (n=687),³⁹ the Study of Health in Pomerania (SHIP; n=3176),³⁸ and the SHIP-Trend Study (n=977).⁴⁰

Replication Study Populations

In a second stage, we followed up SNPs showing significant association with PTH (P<5 × 10⁻⁸) and performed meta-analysis of β-estimates and SEM across discovery and replication stages. The replication analyses encompassed 1982 subjects from the TwinsUK Study,⁶⁷ 911 subjects from the Gothenburg Osteoporosis and Obesity Determinants Study,⁶⁸ 2988 participants from the Ludwigshafen Risk and Cardiovascular Health (LURIC) study,⁶⁹ and 621 participants from the Amish Family Osteoporosis Study⁷⁰ (Table 1).

For all study populations, we excluded participants who had an eGFR<30 ml/min per 1.73 m² calculated from serum creatinine, because of the possibility that the strong influence of kidney disease may overwhelm potentially subtler influences of SNPs on circulating PTH concentrations. All participants provided written informed consent. The study was approved by the institutional ethics committee of each study site and was conducted according to Declaration of Helsinki principles.

The meta-analysis was extended to include the evaluation of the X chromosome across seven cohorts from the discovery GWAS (MESA, ARIC, the Indiana Sisters Study, DCCT/EDIC, MrOS, SHIP, and SHIP-TREND), for a total of 17,865 participants.

Genotyping and Imputation

Genome-wide commercial arrays were used by all cohorts as described in Supplemental Table 1. Genotype calling for SNPs on the X-chromosome was performed separately in men and women. Quality control methods implemented by each study are described in the Supplemental Material. Imputation to the Phase 1 or 2 1000 Genomes panel was conducted using either MACH, IMPUTE, or BIMBAM with quality control metrics as shown in Supplemental Table 2. Imputation results were summarized as an “allele dosage” (a fractional value between 0 and 2), defined as the expected number of copies of the minor allele at that SNP.

Measurement of PTH

Circulating PTH concentrations were measured using second or third generation PTH immunoassays (Supplemental Table 3). We natural-log–transformed PTH as the dependent variable in analyses to evaluate proportionate differences in serum PTH concentrations, minimizing differences in mean PTH concentrations to yield comparable interpretation of coefficients across assays.

Statistical Analyses

Individual centers performed GWAS analyses using linear regression of natural-log–transformed PTH concentrations as the dependent variable, and genotypes (SNP allelic dosage) as predictors, under an additive genetic model. The exponentiated β coefficients from these models can be interpreted as the proportionate (or fold-) difference in serum PTH concentration associated with each additional copy of the minor allele at a given SNP, holding other model covariates constant.

Covariates included in model 1 were age, sex, and season of PTH measurement. If applicable, cohorts included geographic study site, and, where available, the first ≤10 principal components of ancestry to adjust for population stratification (Supplemental Table 2). PTH concentrations are strongly associated with eGFR and BMI. To detect PTH loci independent of these pathways and to diminish associations with effects modulated through these factors, model 2 added adjustment for BMI (calculated as weight in kilograms divided by height in meters-squared) and both linear and quadratic terms for the eGFR, calculated using the creatinine-based Chronic Kidney Disease Epidemiology Collaboration equation.⁷¹ Both discovery and replication analyses used the same adjustment models.

Genomic control parameters were estimated for each cohort and appropriate genomic control correction was applied to input statistics before performing meta-analysis to correct for residual cryptic relatedness or population stratification.⁷² Study-specific genomic control (λ_GC) estimates are shown in Supplemental Table 2.

Genetic differentiation was estimated using the Weir unbiased estimator of the fixation index, calculated using the variance in allele frequencies among European and African samples from the 1000 Genomes and standardized according to the mean allele frequency in the combined sample.⁷³

Conditional analyses were performed using the Genome-wide Complex Traits Analysis tool, version 1.25.3,⁷⁴ to test whether multiple independent risk alleles existed at any of the genome-wide significant loci, using a stepwise selection procedure and summary-level statistics from the meta-analysis (combined discovery and replication, model 1).

Post hoc power analyses were performed using the Quanto software.⁷⁵

Meta-Analysis of Discovery Data

Fixed-effects inverse variance–weighted meta-analysis was conducted within each cohort. This approach takes directionality into account by aligning study results according to the same effect allele. Study-specific effect estimates and SEMs were combined using METAL.⁷⁶

Analyses of the X-chromosome were carried out separately in men and women, and the studies were meta-analyzed separately by sex using an inverse-variance model with fixed effects. The sex-specific meta-analysis results were then combined using a sample-size weighted model.⁷⁷

Variants with imputation quality <0.3 or a MAF <0.05 were excluded from each dataset before meta-analysis. The following exclusions were also applied: call rate <97%, Hardy–Weinberg equilibrium P value <10⁻⁵, duplicate error or Mendelian inconsistency, heterozygote frequency approximately 0, or SNP not found in dbSNP Build 142. SNPs were further excluded from analysis if the ratio of the variance of the allele dosage to the variance expected under Hardy–Weinberg equilibrium was <0.01. In the X-chromosome analyses, all exclusions and filters were applied to sex-stratified SNP-level data. In total, 8,020,965 autosomal SNPs and 538,222 markers on the X-chromosome were meta-analyzed in stage 1. Quantile-quantile plots and Manhattan plots were produced using R and regional association plots were created using SNAP.⁷⁸

Selection of SNPs for Replication and Combined Discovery and Replication Analysis

SNPs exhibiting statistically significant association with PTH concentration (P<5 × 10⁻⁸) were evaluated for replication in four independent cohorts. For genetic regions that contained multiple SNPs significantly associated with PTH concentration, we selected the most significant SNP (lowest P value) within a 500 kb window for replication. Four of the top five SNPs found to be significantly associated with PTH concentration in individuals of European ancestry were imputed, with the following mean ratios of observed variance of the allele dosage to the expected binomial variance at Hardy–Weinberg equilibrium: rs6127099, oevar=0.91; rs219779, oevar=1.03; rs4443100, oevar=1.00; rs73186030, oevar=1.00. The rs4074995 SNP was directly genotyped.

We performed an inverse variance–weighted fixed-effects meta-analysis across replication results using METAL⁷⁶ and obtained two-sided P values for the final effect estimates.

Proportion of Phenotypic Variance Explained

The proportion of variance (PVE) in circulating PTH levels explained by each top novel locus, jointly across all discovery cohorts, was estimated as:

where are the effect size estimate of each minor allele on the relative concentration of PTH, SEM of the effect size, sample size, and MAF for the SNP, respectively.⁷⁹

Follow-Up in African Ancestry Cohorts

The five top SNPs found to be significantly associated with PTH concentration in individuals of European ancestry were evaluated for replication among black individuals from three cohorts: ARIC (n=2464), MESA (n=1510), and CHS (n=305). Four of the SNPs were available in black participants and meta-analyzed. These analyses were performed using the same quality control and imputation exclusions, and meta-analysis was performed using METAL.⁷⁶

eQTL and Functionality Prediction

For each of the SNPs replicated in the European ancestry populations, we identified all proxy SNPs with r²>0.8 in the 1000 Genomes Pilot 1 pairwise LD data using the SNP Annotation and Proxy Search online database (SNAP, Broad). For each SNP, we queried the eQTL database of the Phenotype-Genotype Integrator (http://www.ncbi.nlm.nih.gov/gap/phegeni).

Associations with BMD

We conducted look-ups for femoral bone density in the GEFOS dataset.⁵¹ BMD is used in clinical practice for the diagnosis of osteoporosis and bone density at the femoral neck is predictive of fracture risk. BMD was measured in all cohorts at the femoral neck using dual-energy x-ray absorptiometry following standard manufacturer protocols (General Electric Lunar Corp., Madison, WI or Hologic Inc., Bedford, MA).⁵¹

Associations with Mineral Metabolism Traits

For the SNPs associated with serum PTH concentrations, we examined their association with related mineral metabolism biomarkers where available, including levels of FGF-23 (ARIC and MESA), 25(OH)D (ARIC and MESA), 24,25(OH)₂D (MESA), serum calcium and phosphate (ARIC, LURIC, and MESA), and urine calcium and phosphorus (MESA). Details of these laboratory measurements are provided in Supplemental Tables 4 and 5.

Disclosures

B.K. reports receiving consulting fees and grant support from Amgen Inc. (Thousand Oaks, CA). I.H.d.B. reports receiving research support from Abbvie (Chicago, IL), MedTronic (Minneapolis, MN), and Abbott (Chicago, IL), and consulting fees from Amgen Inc., Bayer (Leverkusen, Germany), and Janssen (Beerse, Belgium). B.M.P. serves on the Data and Safety Monitoring Board of a clinical trial of a device funded by the manufacturer (Zoll LifeCor Corp.; Pittsburgh, PA) and on the Steering Committee of the Yale Open Data Access Project funded by Johnson & Johnson (New Brunswick, NJ).