Developmental programming of growth: genetic variant in GH2 gene encoding placental growth hormone contributes to adult height determination.

INTRODUCTION: Given the physiological role of placental growth hormone (PGH) during intrauterine development and growth, genetic variation in the coding Growth hormone 2 (GH2) gene may modulate developmental programming of adult stature. Two major GH2 variants were described worldwide, determined by single polymorphism (rs2006123; c.171 + 50C > A). We sought to study whether GH2 variants may contribute to adult anthropometric measurements.
METHODS: Genotyping of GH2 SNP rs2006123 by RFLP, testing its genetic association with adult height and Body Mass Index (BMI) by linear regression analysis, and combining the results of three individual study samples in meta-analysis. STUDY SAMPLES: HYPEST (Estonia), n = 1464 (506 men/958 women), CADCZ (Czech), n = 871 (518/353); UFA (Bashkortostan), n = 954 (655/299); meta-analysis, n = 3289 (1679/1610).
RESULTS: Meta-analysis across HYPEST, CADCZ and UFA samples (n = 3289) resulted in significant association of GH2 rs2006123 with height (recessive model: AA-homozygote effect: beta (SE) = 1.26 (0.46), P = 5.90 × 10⁻³; additive model: A-allele effect: beta (SE) = 0.45 (0.18), P = 1.40 × 10⁻²). Among men (n = 1679), the association of the A-allele with taller stature remained significant after multiple-testing correction (additive effect: beta = 0.86 (0.28), P = 1.83 × 10⁻³). No association was detected with BMI. Notably, rs2006123 was in strong LD (r² ≥ 0.87) with SNPs significantly associated with height (rs2665838, rs7209435, rs11658329) and mapped near GH2 in three independent meta-analyses of GWA studies.
CONCLUSIONS: This is the first study demonstrating a link between a placental gene variant and programming of growth potential in adulthood. The detected association between PGH encoding GH2 and adult height promotes further research on the role of placental genes in prenatal programming of human metabolism.

Introduction

The human Growth hormone/Chorionic somatomammotropin (hGH/CSH) genes belong to a rapidly evolving primate-specific gene cluster spanning 48 kb at 17q22-24. In most mammals, a single GH1 gene encodes the pituitary growth hormone (GH), whereas in primates, novel placenta-specific GH-related genes have arisen through gene duplications [1]. In humans, GH/CSH cluster consists of five highly homologous (91–97%) and structurally similar genes: GH1, GH2, CSH1, CSH2 and CSHL1 (chorionic somatomammotropin-like 1) [2]. It is well known that pituitary GH (also known as somatotropin) promotes postnatal growth of skeletal and soft tissue, and acts as an important regulator of immune function, bone turnover and muscle mass. GH deficiency (1:4000–1:10,000/births) is associated with short stature [3] and common GH1 substitutions have been shown to contribute to height determination [4].

In the human placenta, the expression of four GH/CSH genes (GH2, CSH1, CSH2, CSHL1) is coordinately induced during fetal development in the syncytiotrophoblasts from 5 to 8 weeks of pregnancy [5], and their expression profile exhibits pleiotropic effects at the maternal–fetal interface in regulating of fetal growth and modulating maternal metabolism [6-9]. Two of these, CSH1 and CSH2, encode jointly an identical protein chorionic somatomammotropin (CSH, also known as placental lactogen, PL). CSHL1 was originally considered as a pseudogene, although low levels of its expression in placenta have been reported [8]. GH2 encodes placental GH (PGH), which progressively replaces maternal pituitary GH from mid-gestation onwards, peaking towards term [10,11]. Only 13 amino acid residues constitute the difference between PGH and GH. To execute its function, PGH binds to GH cell surface receptors (GHR) with similar affinity to pituitary GH [6]. Interestingly, secreted PGH is found predominantly, but not exclusively, in maternal circulation [12,13]. Maternal PGH serum levels have been positively correlated with infant birth weight [11,12,14]. Significantly lower placental expression of GH2 and reduced levels of circulating PGH have been reported in women with fetal intrauterine growth retardation/small-for-gestational-age pregnancies [8,12,14].

We hypothesized that given the physiological role of PGH during intrauterine development, the genetic variation in the GH2 gene may modulate growth in utero and in early infancy, therefore possibly affecting the developmental programming of human stature in adulthood. However, in contrast to the pituitary-expressed GH1, there are limited data on the impact of polymorphisms in the placental hGH/CSH genes on intrauterine growth and programming of the postnatal metabolism [15]. Detailed research on hGH/CSH cluster has been hindered by its complex genomic structure rich in repetitive genic and intergenic sequence fragments. Our pioneer study had revealed that the duplicated hGH/CSH genes exhibit substantial heterogeneity in diversity patterns and low linkage disequilibrium (LD) between allelic variants, driven by the interplay between active intergenic gene conversion and locus-specific selective pressures [16]. For the GH2 gene, only two major gene variants were described, determined by the allelic status of one polymorphism (rs2006123; c.171 + 50C > A) located 50 bp from the donor splice site within intron 2 (original nomenclature [16], g.943C > A). GH2 rs2006123 alleles were differentially distributed in studied populations: 92% of the Chinese Han individuals carried the ancestral C-allele, whereas the derived A-allele was enriched in African Mandenkalu (carrier frequency 95%). In European Estonians both alleles were commonly represented (C, 66%; A, 34%). As other hGH/CSH genes showed no or low intercontinental differentiation, it was suggested that the observed GH2 variation pattern might reflect regional population-specific selection.

The present study aimed to test the association between human PGH coding GH2 intron 2 polymorphism rs2006123 and anthropomorphic phenotypes (height, BMI) in three Eastern/Central European sample sets and in the subsequent meta-analysis (total sample size, n = 3289). We report significant association between GH2 rs2006123 and adult height, and show that the studied GH2 variant is in strong LD (r2 > 0.8) with top hits from three independent meta-analysis of genome-wide association studies (GWAS) for height, mapped within the hGH/CSH gene cluster (rs2665838 [17]) or to its vicinity (<250 kb: rs7209435 [18]; rs11658329 [19]).

Materials and methods

Study groups

The analyzed sample collections HYPEST (Estonians), CADCZ (Czech) and UFA (Bashkirs and Tatars from Volga-Ural region, Russia) represent populations of Eastern/Central European origin and their basic characteristics are provided in Table 1. The recruitment of the three sample sets has been carried out in compliance with the Helsinki Declaration and participants have given the written informed consent. The HYPEST study has been approved by the Ethics Committee on Human Research of University of Tartu (permissions 122/13, 22.12.2003; 137/20, 25.04.2005). The CADCZ study has been approved by the Ethics Committee of Charles University—1st Faculty of Medicine (December 1996) and the UFA study by the Independent Ethics Committee of the Institute of Biochemistry and Genetics, Ufa Scientific Centre of Russian Academy of Sciences, Ufa, Russia (permission no. 5, 25.01.2005).

Table 1

Anthropometric characteristics of the study samples.

Parametera	HYPEST			CADCZ			UFA
	All	Men	Women	All	Men	Women	All	Men	Women
No of subjects	1464	506	958	871	518	353	954	655	299
Age (y)	47.5 ± 13.4	45.3 ± 13.6	48.7 ± 13.2	50.0 ± 10.6	51.2 ± 9.8	48.1 ± 11.4	54.1 ± 17.1	50.3 ± 15.9	62.3 ± 17.0
Height (cm)	169.6 ± 9.3	178.9 ± 6.9	164.6 ± 6.1	172.4 ± 8.7	177.3 ± 6.9	165.3 ± 5.8	167.6 ± 8.2	169.1 ± 7.5	164.4 ± 8.6
BMI (kg/m2)	26.3 ± 4.8	27.0 ± 3.9	26.0 ± 5.2	27.0 ± 4.4	27.7 ± 3.8	26.1 ± 5.0	27.9 ± 3.9	27.6 ± 3.7	28.7 ± 4.4

For each parameter mean ± SD is shown; y, age in years; BMI, body mass index (kg/m2).

Originally, HYPEST subjects were recruited across Estonia, North-Eastern Europe during 2004–2007 (1966 individuals, age range 18–85 years) with the main aim to analyze genetic-epidemiological risk factors for cardiovascular disease in Estonian population [20]. In the current study, the total number of genotyped HYPEST subjects was n = 1464 (aged 47.5 ± 13.4 years) including 506 men (aged 45.3 ± 13.6 years) and 958 women (aged 48.7 ± 13.2 years). The CADCZ samples have been recruited across Czech Republic, Central Europe by the Cardiology Department of the 2nd Clinic of Internal Medicine, Faculty Hospital Královské Vinohrady in Prague with the main aim to study genetic factors relating to homocysteine metabolism in coronary artery disease and details are published elsewhere [21]. The number of CADCZ samples available for genotyping in the current study was 871 (aged 50.0 ± 10.6 years), including 518 men (aged 51.2 ± 9.8 years) and 353 women (aged 48.1 ± 11.4 years). UFA sample is comprised of subjects recruited in Bashkortostan, the Volga-Uralic region of Russia, located at the border of Eastern Europe and Asia (prof. E. Khusnutdinova and collaborators). In this study, 954 of Bashkortostan samples (n = 464 Tatars, n = 490 Bashkirs) were genotyped (aged 54.1 ± 17.1 years). Analyzed UFA samples included 655 men (aged 50.3 ± 15.9 years) and 299 women (aged 62.3 ± 17.0 years).

At the recruitment, anthropometric data (height in cm, weight in kg) of each participant of the three studies were documented. In order to consider in association testing both genders together, in addition to the direct height measurements, the transformed values were used as recommended [17]. Z-score (standard deviation score) for height was calculated as the subject's height minus the mean height in the sex appropriate subsample of each study divided by the standard deviation of that mean. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m2).

Genotyping

In order to exclude amplification of other highly homologous hGH/CSH genes, the genotyping protocol of GH2 rs2006123 SNP included long-range PCR to amplify specifically the entire GH2 gene region (5634 bp product [16]; forward primer: 5′-AGGTCTGGAAAGGAGGAGACAAAAGAG-3′, reverse primer: 5′-TTGAATTAGACTTGGGATTCTCCTGACA-3′), and allele-specific restriction fragment length polymorphism (RFLP) analysis (Supplemental Figs. S1–S2). In case the long-range PCR had resulted in insufficient product quantity for the RFLP detection, nested PCR (1855 bp product; forward primer: 5′-GCTTGTGGTCCGTGTCTGTTCC-3′, reverse primer: 5′-GGTGGCGATAGACGTTGCTGTCC-3′) was applied to further amplify the DNA fragment involving the targeted SNP. PCR conditions are described in Supplemental Text S1.

The amplified product was digested with FastDigest XapI restriction endonuclease according to manufacturer's recommendations (Thermo Fisher Scientific, Lithuania). Allele-specific RFLP products were separated by electrophoresis on a 1–2% agarose gel and 0.5xTris/Borate/EDTA buffer. The following fragments were detected in the RFLP analysis of (i) long-range PCR: 2658, 1626, 541, 484, 218, 107 bp (C-allele); 1626, 1614, 1044, 541, 484, 218, 107 bp (A-allele) (Supplemental Fig. S1A) or (ii) nested PCR: 1546 and 309 bp (C-allele); 1045, 501, and 309 bp (A-allele) (Supplemental Fig. S1B). In developmental phase of the RFLP-based genotyping assay, a subset of samples were sequenced for the quality control and to verify few ambiguous genotypes using previously published sequencing primers and conditions [16].

Data analysis

Estimation of allele frequencies, conformance to Hardy–Weinberg Equilibrium (HWE; χ2, P > 0.05) and association testing of the GH2 rs2006123 with anthropometric characteristics using linear regression under additive and recessive genetic models was implemented in PLINK v1.07 software http://pngu.mgh.harvard.edu/purcell/plink/ [22]. Additive genetic model assumes that having two copies of the minor allele has twice the effect of having a single copy of a minor allele, while recessive model assumes that only having two copies of a minor allele has an effect on phenotype. In association testing in the full sample, adjustment for age and gender were applied; analyses among men or women, and in the full sample using transformed values (Z-score for height) were adjusted only for age. Bonferroni threshold for multiple testing was estimated for the discovery analysis α = 0.05/[2 (models) × 2 (parameters: height, BMI) × 2 (gender)] = 6.25 × 10−3, and for the meta-analysis α = 0.05/[3 (studies) × 2 (models) × 2 (parameters: height, BMI) × 2 (gender)] = 2.08 × 10−3. Results were combined in a meta-analysis using the inverse-variance method under fixed-effects model using R, ver. 2.13.1 (R Development Core Team 2011, http://www.r-project.org/). Presence of heterogeneity in the meta-analysis was assessed by Cochran's Q and I2 statistics [23]. Power calculations for meta-analysis were performed in R (pwr package). Height distribution data was used to estimate f2 values assuming additive genotype effects.

Linkage disequilibrium (LD; r2) in the hGH/CSH region was calculated using 1000GENOMES:pilot_1_CEU_low_coverage_panel dataset generated by sequencing the HapMap CEPH (Centre d'Etude du Polymorphisme Humain) samples representing Northern and Western European ancestry (n = 60; http://browser.1000genomes.org/index.html). The LD calculations using the above-mentioned data included also top SNPs associated with height within or in the vicinity of hGH/CSH region in the published meta-analysis of GWAS studies (rs2665838 [17]; rs7209435 [18]; rs11658329 [19]; rs2854160 [24]). Additionally, LD between GH2 rs2006123 and SNPs within the GH1 and GH2 genes was estimated using published resequencing dataset from Ref. [16]. All LD plots were composed using the Haploview 4.2 program (www.broadinstitute/haploview/haploview) [25].

Sequence alignments were performed using the Web-based ClustalW2 program (http://www.ebi.ac.uk/Tools/msa/clustalw2/). Statistical difference between men and women in the distribution of rs2006123 genotype frequencies was assessed using χ2-test implicated in Genepop 4.2 software (http://kimura.univ-montp2.fr/∼rousset/Genepop.htm; “population differentiation” option). The testing conditions: dememorization = 10,000, batches = 1000, iterations = 10,000.

Results

GH2 rs2006123 genotype and allele frequencies

The HYPEST (n = 1464), CADCZ (n = 871), and UFA (n = 954) samples were genotyped for the GH2 rs2006123 polymorphism (intron 2; c.171 + 50C > A; original nomenclature g.943C > A [16]) using the combination of gene-specific long-range and nested PCR followed by allele-specific RFLP. In all sample sets, the distribution of rs2006123 genotypes was in accordance with HWE (P > 0.05; Table 2). Minor A-allele frequency of the GH2 rs2006123 polymorphism exhibited a gradient from Volga-Ural to Central Europe (UFA 20.4%, HYPEST 25.4%, CADCZ 30.4%; Table 2). Genotype frequency distributions among male and female subjects in all three sample sets were statistically similar (χ2-test, P > 0.27; Supplemental Table S1).

Table 2

Allele and genotype frequencies of the GH2 rs2006123.

Population		N		Allele frequenciesa		Genotype frequenciesa			HWEb
				C	A	C/C	C/A	A/A	P-value
HYPEST	Estonia	1464	%	74.6	25.4	54.9	39.4	5.7	0.129
			(N)	(2185)	(743)	(804)	(577)	(83)
CADCZ	Czech	871	%	69.6	30.4	47.9	43.5	8.6	0.471
			(N)	(1213)	(529)	(417)	(379)	(75)
UFA	Bashkortostan	954	%	79.6	20.4	64.3	30.6	5.1	0.073
			(N)	(1518)	(390)	(613)	(292)	(49)

N, number of genotyped subjects in each sample set.

Data presented as percentage with number of allele/genotype carriers in brackets.

HWE, P-value of the χ2-test for Hardy-Weinberg-Equilibrium.

GH2 rs2006123 is associated with adult height

In the discovery analysis of HYPEST samples (Estonians), GH2 rs2006123 SNP was significantly associated with male height under additive genetic model (n = 506; A-allele effect: beta (SE) = 1.52 (0.5); P = 2.54 × 10−3), whereas no association was detected in the female-only analysis (n = 958; Table 3). Although in the replication samples (UFA, Bashkortostan: n = 655 men, n = 299 women; CADCZ, Czech: n = 518 men, n = 353 women; Table 1) the association testing did not reach statistical significance, five of the six analyzed groups under the additive model and all groups under the recessive model showed consistent trend (positive beta) of association with height (Table 3).

Table 3

Association testing of the GH2 rs2006123 with measured adult height (in cm).

Study		N	Additive model			Recessive model
			Beta (SE)	95% CI	P-value	Beta (SE)	95% CI	P-value
HYPEST	All	1464	0.43 (0.26)	[−0.07, 0.94]	9.09 × 10−2	1.41 (0.67)	[0.10, 2.73]	3.52 × 10−2
	Men	506	1.52 (0.50)	[0.54, 2.50]	2.54 × 10−3*	2.07 (1.51)	[−0.89, 5.04]	1.72 × 10−1
	Women	958	−0.05 (0.29)	[−0.63, 0.52]	8.53 × 10−1	1.21 (0.73)	[−0.22, 2.63]	9.67 × 10−2
CADCZ	All	871	0.27 (0.33)	[−0.37, 0.91]	4.06 × 10−1	0.74 (0.74)	[−0.72, 2.20]	3.20 × 10−1
	Men	518	0.37 (0.45)	[−0.51, 1.25]	4.16 × 10−1	0.64 (1.03)	[−1.38, 2.66]	5.34 × 10−1
	Women	353	0.08 (0.46)	[−0.82, 0.99]	8.58 × 10−1	0.75 (1.05)	[−1.31, 2.80]	4.78 × 10−1
UFA	All	954	0.81 (0.43)	[−0.04, 1.66]	6.25 × 10−2	2.05 (1.15)	[−0.21, 4.31]	7.62 × 10−2
	Men	655	0.83 (0.50)	[−0.14, 1.80]	9.47 × 10−2	2.39 (1.32)	[−0.20, 4.97]	7.08 × 10−2
	Women	299	0.66 (0.86)	[−1.03, 2.35]	4.46 × 10−1	1.27 (2.29)	[−3.21, 5.75]	5.79 × 10−1
Meta-analysis	All	3289	0.45 (0.18)	[0.09, 0.81]	1.40 × 10−2	1.26 (0.46)	[0.36, 2.15]	5.90 × 10−3
	Men	1679	0.86 (0.28)	[0.32, 1.41]	1.83 × 10−3**	1.47 (0.71)	[0.07, 2.87]	3.93 × 10−2
	Women	1610	0.04 (0.24)	[−0.43, 0.50]	8.79 × 10−1	1.07 (0.58)	[−0.06, 2.20]	6.37 × 10−2

Linear regression (age and gender as covariates) was used to test association with height under additive and recessive genetic models. Effect is given as beta, SE. P < 0.05 is highlighted in bold and P-values lower than Bonferroni threshold for multiple testing correction in the discovery analysis (α = 6.25 × 10−3) and in the meta-analysis (α = 2.08 × 10−3) are indicated with asterisks (*) or (**), respectively. Meta-analysis of across individual studies (HYPEST, CADCZ, UFA) was performed using an inverse-variance method under fixed-effect model. N, number of subjects; SE, standard error, CI, confidence interval.

Prior to meta-analysis, the consistency of effects across study groups was estimated. For all performed meta-analysis tests, the calculated heterogeneity statistic was indicative to the consistency of studies' results (I2 = 0%). In the meta-analysis across the HYPEST, CADCZ and UFA men (n = 1679), the association of the GH2 rs2006123 A-allele with taller stature was enhanced (additive effect: beta = 0.86 (0.28), P = 1.83 × 10−3) and remained statistically significant after multiple testing correction (Table 3). In meta-analysis across women (n = 1610), the association was not statistically significant.

Based on direct height measurements, the HYPEST men carrying AA-genotype were in average 2.7 cm taller than the CC-homozygote men, and the mean height difference between AA- and CC-homozygote women was 1.6 cm. In the UFA sample, the AA- compared to CC-homozygotes were in average 2.4 cm and 1.5 cm taller among men and women, respectively; and in the CADCZ sample the estimated height difference was 1.8 cm (male) and 0.7 cm (female) (Table 4). Alternative testing of the full study sample based on the calculated Z-scores instead of raw height measurements (in cm) further supported the detected association between GH2 rs2006123 and adult height (Table 5).

Table 4

Measured height in subgroups carrying the alternative GH2 rs2006123 genotypes.

Genotype	HYPEST		CADCZ		UFA
	Men	Women	Men	Women	Men	Women
N: CC, CA, AA	281, 206, 19	523, 371, 64	255, 218, 45	162, 161, 30	413, 208, 34	200, 84, 15
CC	178.1 ± 6.8a	164.7 ± 5.8	176.9 ± 7.4	165.3 ± 5.7	168.8 ± 7.7	164.2 ± 8.8
CA	179.8 ± 7.0	164.2 ± 6.3	177.4 ± 6 .3	165.1 ± 5.7	169.3 ± 7.1	164.8 ± 8.5
AA	180.8 ± 5.0	166.3 ± 6.3	178.7 ± 6.5	166.0 ± 7.1	171.2 ± 7.8	165.7 ± 7.4

Data is provided as mean ± SD of height in centimeters (cm); N, number of subjects in each genotype group (CC, CA, AA).

Table 5

Association testing of the GH2 rs2006123 with adult height based on calculated Z-scores (Standard deviation scores).

Study	N	Additive model			Recessive model
		Beta (SE)	95% CI	P-value	Beta (SE)	95% CI	P-value
HYPEST	1464	0.066 (0.040)	[−0.012, 0.145]	1.01 × 10−1	0.238 (0.105)	[0.032, 0.444]	2.36 × 10−2
CADCZ	871	0.038 (0.050)	[−0.061, 0.136]	4.56 × 10−1	0.113 (0.115)	[−0.113, 0.338]	3.27 × 10−1
UFA	954	0.103 (0.055)	[−0.005, 0.210]	6.20 × 10−2	0.264 (0.146)	[−0.023, 0.550]	7.13 × 10−2
Meta-analysis	3289	0.067 (0.027)	[0.013, 0.120]	1.44 × 10−2	0.199 (0.069)	[0.065, 0.333]	3.64 × 10−3

Linear regression (age as covariate) was used to test association with height under additive and recessive genetic models. Effect is given as beta, SE. P < 0.05 is highlighted in bold. Meta-analysis of across individual studies (HYPEST, CADCZ, UFA) was performed using an inverse-variance method under fixed-effect model.

N, number of subjects; SE, standard error, CI, confidence interval.

No association of GH2 rs2006123 with BMI

The GH2 rs2006123 was not associated with BMI in any of the analyzed studies (HYPEST, UFA, CADCZ) either in the full sample or in sex-specific sub-samples (Supplemental Tables S2–S3).

GH2 rs2006123 is in strong LD with top hits from meta-analysis of GWA studies for height within or nearly the hGH/CSH region

Intergenic SNPs within the hGH/CSH cluster or in the vicinity have been associated with adult height in several independent meta-analysis of conducted genome-wide association studies (GWAS) among Europeans and non-Europeans (rs2665838 [17]; rs7209435 [18]; rs11658329 [19]; rs2854160 [24]). We estimated the strength of LD between the GH2 rs2006123 and the surrounding SNPs using the 1000Genomes dataset for the subjects of European origin (CEPH; n = 60). Consistent with the previous report in Estonians, Mandenkalu and Chinese Han [16], also in the 1000Genomes dataset the entire hGH/CSH gene region was characterized by weak LD (Supplemental Fig. S3). Notably, the GH2 rs2006123 showed strong LD with three previously identified SNPs associated with adult height within or near the hGH/CSH cluster: rs2665838 (r2 = 0.87, ∼7.2 kb upstream from GH2), and rs7209435, rs11658329 (r2 = 0.92, within the MAP3K3 gene ∼245 kb and ∼195 kb upstream from rs2006123, respectively) (Fig. 1). The GH2 rs2006123 was not in LD with the hGH/CSH cluster SNP rs2854160 detected in combined GWA results for height from 9 studies of individuals of African descent (r2 = 0.19, CSH1-CSHL1 intergenic region).

Fig. 1

Pairwise LD plot (296 kb region; 17q23.3) including SNPs in the GH1 and GH2 genes and top-hits from four independent meta-analysis of GWA studies of adult height mapped within or nearby the hGH/CSH genome cluster (rs2665838 [17], rs7209435 [18], rs11658329 [19], rs2854160 [24]). The GH2 rs2006123 is indicated with an oval dashed line and the GWAS top-SNPs are marked with an asterisk. SNPs in strong LD (r2 > 0.8) are boxed. The genomic context of the region is shown above the LD plot. Solid arrows indicate to the gene location and transcriptional orientation. The location of SNPs rs2665838 and rs2854160 within the zoomed-in hGH/CSH gene cluster is shown with dashed arrows. LD (r2) was calculated based on 1000GENOMES:pilot_1_CEU_low_coverage_panel dataset containing HapMap individuals with European ancestry (CEPH; n = 60). Numbers shown on the individual squares of the LD triangle represent the estimated r2 (%) values between SNP pairs, scaling from no LD (white squares, r2 = 0) to complete allelic association (black squares, r2 = 100%).

Importantly, in the 1000Genomes dataset the SNPs within or nearby the hGH/CSH region reported in four meta-analyses of GWAS data to be associated with height, completely lacked (r2 = 0) or exhibited low LD (r2 ≤ 0.3) with SNPs in GH1 encoding the pituitary growth hormone (Fig. 1, Supplemental Fig. S3). Also, in the published resequencing study of the hGH/CSH genes [16], almost no (r2 ≤ 0.09 for 14 of 18 SNPs) or low LD was estimated between the GH2 rs2006123 and SNPs in the GH1 gene (average r2 = 0.066) or its upstream region (r2 = 0.19), and other SNPs the GH2 gene (r2 = 0.025) (Supplemental Table S4).

Discussion

The current study reports the significant effect of the genetic variant (rs2006123; c.171 + 50C > A) in intron 2 of GH2 gene encoding placental growth hormone (PGH) to modulate the adult height. To our knowledge, this is the first study to demonstrate a link between a polymorphism in a gene with known expression restricted to the placenta during prenatal period and the programming of the growth potential in adulthood. Humans with the derived GH2 rs2006123 A-allele appeared to be taller and the effect was stronger among men. In meta-analysis across the three study samples of Eastern/Central European origin (n = 3289), the effect of the A-variant was detected 0.45 cm per allele (P = 1.40 × 10−2) and 1.26 cm for the AA-genotype (P = 5.90 × 10−3); whereas among men (n = 1679) the A-allele carrier effect was 0.86 cm per allele (P = 1.83 × 10−3) and 1.47 cm for AA-homozygotes (P = 3.93 × 10−2). The limitation of modest sample size possibly affected proper association testing with female height (Supplemental Fig. S4). Still, the observed effect sizes in women showed consistent non-significant trend (positive beta) of association with height under the recessive model (meta-analysis: beta (SE) = 1.07 (0.58); P = 6.37 × 10−2; Table 3). Unfortunately, further postnatal influences were not known for the study groups and thus, could not be ruled out as additional modifiers of the study outcome. Overall, the present results contribute to a larger array of genetic effects reported to be involved in human adult height determination [17-19,24].

The studied GH2 polymorphism rs2006123 (C/A; located in intron 2, 50 bp from donor splice site) is determining the two main GH2 variants in humans [16]. It shows high allelic differentiation among world populations and possible balancing selection acting on its genetic variation. Whereas in Asia the prevalence of the derived A-allele is low (≤8% in Han Chinese and Japanese; dbSNP, http://www.ncbi.nlm.nih.gov/projects/SNP/snp_ref.cgi?rs=2006123), it represents the major GH2 variant in Africa (Senegalese Mandenkalu 85%, Nigerian Yorubans 79%). In Europe, there appears to be a West-East gradient of the A-allele frequency, from 30% in Central-Europe to 20% at the Volga-Ural region bordering Europe and Asia. In human-chimpanzee comparison, GH2 is the most conserved gene in the GH/CSH cluster and so far only chromosomes with C-allele at the position c.171 + 50C > A have been described [26]. Also, all the rest of the highly homologous genes in the hGH/CSH cluster (GH1, CSH1, CSH2, CSHL1) are monomorphic for the C-nucleotide at this position (Supplemental Fig. S5). In case the C-allele is considered as ancestral, the derived A-allele associated with a taller statue, arose as a novel mutation among humans.

PGH secreted by syncytiotrophoblast functions as an insulin antagonist, controls maternal Insulin Growth Factor 1 (IGF-1) production and glucose utilization in pregnancy [6,7]. The induction of PGH by glucose deprivation provides a feedback loop to ensure a delivery of nutrients to developing fetus. Multiple studies have provided convincing evidence for a positive correlation between maternal serum PGH concentration and birth-weight [11,12,14]. In addition, it has been shown that transgenic mice over-expressing the gene encoding for PGH became larger than their normal littermates and they are at risk to develop insulin resistance in later life [27]. The observed effects of PGH speak for its influence on growth process, thus explaining the association found with the adult height, but the exact molecular mechanisms of the effect are yet to be discovered. However, based on the accumulated evidence, it has been concluded that altered metabolism of PGH in humans may lead to consequences not only in fetal growth and gestational metabolism, but also in programming the birth-weight of a newborn and long-term metabolic function [7]. Unfortunately, the datasets of the current study did not allow addressing the association of GH2 polymorphism rs2006123 with birth weight and height.

The secretion of PGH is inhibited by glucose in vitro and in vivo [28], indicating that maternal nutrition, food availability and diet may affect PGH synthesis. Different GH2 expressional variants could have had a selective advantage to guarantee the optimal birth weight for a given population environment and life-style. However, as human dietary habits and lifestyle have changed tremendously in past centuries, the genetic composition evolved to support normal growth axis even in nutrient-limited situations may contribute to the programming of accelerated growth in a nutrient-abundant environment. Notably, a recent study showed that frequencies of SNP alleles associated with increased height were systematically elevated in Northern compared with Southern Europeans mirroring the intra-European height differences and being consistent with weak selection on these genetic variants [29].

Several meta-analyses of GWA studies have mapped genetic variants contributing to height determination within (intergenic region) or in the vicinity (within 245 kb) of the hGH/CSH gene cluster. The studied GH2 intronic variant rs2006123 was in strong LD (r2 ≥ 0.87) with three of the four reported top hits mapped to this region to be associated with height (rs2665838 [17]; rs7209435 [18]; rs11658329 [19]). Thus, the current paper is essentially providing important experimental support for an in silico imputation, an essential step in meta-analyses of GWAS datasets. This observation favors the scenario that the reported GWAS associations with height may be attributable to the functionality of the GH2 rs2006123 alternative alleles, although it does not explicitly exclude that it acts as a proxy of the actual causal variant(s) within hGH/CSH cluster modulating the growth potential. Here, we speculate that alternative alleles of this SNP (c.171 + 50C > A) located in intron 2 may modulate splicing patterns and expression of GH2, encoding in total four splice variants [9]. Interestingly, one of the GH2 gene alternative mRNA transcripts, GH2-4 (20-kDa hGH-V) skips the regular acceptor site in intron 2 and uses alternative site 45 bp downstream within exon 3 [30]. Functional studies on human pituitary growth hormone (GH) encoding GH1 (homologous to GH2) have indicated that a specific secondary structure within the native human GH transcript controls the relative utilization of the two competing splice-acceptor sites with the consequent generation of two functionally distinct hormone isoforms [31]. The deleted region (identical in GH1 and GH2) involves receptor binding site 1 and secreted hormone has therefore lower activity for GH receptor and GH binding protein [32]. The profile of PGH isoforms during fetal period may not only contribute to the determination of intra-uterine growth and metabolism, but may also program the activity of the GH/IGF-1 axis for the post-natal period [33].

Increasing attention is drawn to the in utero programming of postnatal metabolism and risks for adult disease [34]. Although the majority of observations have been ascribed to maternal malnutrition, infection, exposure to environmental factors or placental pathology, there is growing evidence that developmental programming may be also modulated by genetic factors. A polymorphism in the IGF2BP2 gene was shown to interact with fetal malnutrition to program postnatal glucose metabolism [35]. The carrier status of the T-allele of rs12979860 upstream of IL28B has been suggested to contribute to subsequent innate immune development in children who develop allergic disease, as it correlates with differences in the pro-inflammatory profile during the first five years of life [36]. A recent study showed that many genes with high expression in mid-gestation placenta have also been implicated in adult complex disease, promoting the discussion on the role of placenta in developmental programming [37].

In summary, the study reports the association between PGH encoding GH2 polymorphism and adult height, and promotes discussions and further research on the role of placental genes in prenatal programming of human metabolism and growth potential.

Disclosure statement

The authors have nothing to disclose.

Funding

This project was supported by the post-doctoral grant [MJD39] to Y.T. received from researcher mobility programme MOBILITAS financed by European Social Fund. Additional support was provided by the following grants to M.L.: Wellcome Trust International Senior Research Fellowship in Biomedical Science in Central Europe [070191/Z/03/A], Estonian Science Foundation grant [ETF7471, ETF9030], Estonian Ministry of Education and Science core grant [SF0180022s12]. V.K. was receiving institutional support by research programs of the Charles University in Prague PRVOUK-P24/LF1/3 and UNCE 20401.