Chorionic Gonadotropin Beta gene variants are associated with recurrent miscarriage in two European populations

Abstract

Context

The incidence of recurrent miscarriage (RM; ≥3 consecutive pregnancy losses) is estimated as 1%-2% in fertile couples. Familial clustering of RM has suggested the contribution of a genetic component.

Objective

Low level of HCG in maternal serum during the first trimester of the pregnancy is a clinically accepted risk factor for miscarriage. We sought to study whether variation in Chorionic Gonadotropin beta subunit genes (CGB) expressed in placenta may contribute to the risk of RM.

Design

Resequencing of CGB5 and CGB8, the two most actively transcribed loci of the four HCG beta duplicate genes.

Setting

A case-control study involving two sample sets, from Estonia (n=194) and from Finland (n=185).

Patients

RM patients (n=184) and fertile controls (n=195).

Results

From 71 identified variants in CGB5 and CGB8, 48 SNPs were novel. Significant protective effect was associated with two SNPs located at identical positions in intron 2 in both CGB5 (p=0.007, OR=0.53) and CGB8 (p=0.042, OR=0.15); and with four CGB5 promoter variants (p<0.03; OR=0.54-0.58). The carriers of minor alleles had reduced risk of RM. The haplotype structure of the CGB8 promoter was consistent with balancing selection; a rare mutation in CGB8 initiator element was detected only among patients (n=3). In addition, three rare non-synonymous substitutions were identified among RM cases as possible variants increasing the risk of recurrent pregnancy loss.

Conclusions

The findings encourage studying the functional effect of the identified variants on CGB expression and HCG hormone activity to further elucidate the role of CGB variation in RM.

Introduction

Recurrent miscarriage (RM) or habitual abortion is defined as three or more consecutive pregnancy losses before 22 gestational weeks or the spontaneous abortion of an embryo/fetus weighing less than 500g. The occurrence of RM is estimated as 1%-2% of fertile couples (1, 2). Although the patients with RM undergo multiple diagnostic tests to detect parental chromosomal anomalies, maternal thrombophilic, endocrine, or immunological disorders, over 50% of the RM cases are classified as idiopathic (3). An increased prevalence of miscarriage among first-degree relatives of the women suffering from RM (4) suggests genetic contribution in recurrent pregnancy loss. Possible candidates include genes regulating the development of maternal immunotolerance and inflammatory response, coagulation, angiogenesis, vascular tone, and apoptosis. Prime candidates of the molecular causes of RM have been various trombophilic gene mutations (5-7). Convincing data has also been reported on the association between the miscarriage rates and the polymorphisms in HLA-G gene expressed on the surface of the invading cytotrophoblasts (8).

So far, major interest has focused on the physiological response of the mother to the pregnancy. Less attention has been paid to the placental proteins coded by the fetal genome with contribution from both maternal and paternal genes and their variants. One of the first proteins produced by the conceptus is human chorionic gonadotropin (HCG), also known as “the pregnancy hormone” due to its essential role in human reproduction. The main function of HCG is to delay the apoptosis of the corpus luteum during the first trimester of pregnancy. HCG has several paracrine effects in the process of implantation (9), angiogenesis and placentation (10, 11), and development of maternal immunotolerance (12). Low level and non-exponential increase of HCG in maternal serum during the first trimester of the pregnancy is a clinically accepted risk factor for miscarriage (13-15).

The hormone-specific HCG beta-subunit is expressed by syncytiotrophoblasts of placenta and is encoded by four Chorionic Gonadotropin Beta genes (CGB, CGB5, CGB7 and CGB8) located within the LHB/CGB gene cluster at 19q13.3 (Fig. 1B). Among the four HCG beta duplicate genes CGB8 and CGB5 are the most actively transcribed and contribute together 62-82% to the total pool of beta-subunit mRNA transcripts (Fig.1A, ​16​-​18). Our previous data on the HCG beta genes showed that (i) their diversity level is one of the highest reported for human genes; (ii) there is high interindividual and intergenic difference in expression and (iii) mRNA transcription level is significantly lower in cases of RM compared to normal first trimester pregnancies (18​-20). Now we have addressed the question whether particular variants in these genes may contribute to pregnancy failure. High genetic variation in the LHB/CGB region and the aim to capture both rare and common variation prompted us to choose resequencing instead of traditional genotyping. We analyzed CGB5 and CGB8 in Estonian and Finnish RM cases (n=184) and fertile women (n=195) by comparing variation and haplotype patterns between the two groups. Consistent with hypothesis of the study, we identified genetic variants in HCG beta genes either significantly increasing or reducing a subject’s risk to experience recurrent pregnancy loss.

Figure 1.

Genomic and expressional context for the design of the association study targeting HCG beta genes. (A) The contribution of each individual gene into the total mRNA transcript pool of all six CGB genes (18). (B) Schematic presentation of the LHB/CGB gene cluster with genes marked as black wide arrows in the direction of transcription on sense strand. (C) The position of long-range PCR primers (black arrows) and extent of resequenced CGB5 and CGB8 regions (short black bars). Gene exons are depicted with grey boxes. Capital letters correspond to the primer sequences listed in Supplementary Table S1. (D) The aligned consensus sequences of the 5′upstream element of LHB/CGB genes. The nucleotide positions distinctive for each HCG beta and LHB gene co-localizing with CGB5 SNPs c5EF-155, c5EF-147, c5EF-144 and c5EF-142 (Table 1) are highlighted.

All positions are given relative to mRNA transcription start site.

Subjects and Methods

Study subjects

The study was approved by the Ethics Committees of the University of Tartu, Estonia (protocols no 117/9, 16.06.03, 126/14, 26.04.2004) and the Department of Obstetrics and Gynecology, Helsinki University Central Hospital Outpatient Clinic for women with recurrent miscarriage (protocol no 298/E2/2000). Subjects were recruited and blood samples for the DNA extraction were collected at the Women’s Clinic of Tartu University Hospital and Nova Vita Clinic, Centre for Infertility Treatment and Medical Genetics, Tallinn, Estonia in 2003-2007; and in the Department of Gynaecology and Obstetrics of the Helsinki University Hospital in Finland during 2001-2004. Written informed consent was obtained from every study participant. In both participating centers patients with at least ≥3 abortions during the first trimester of pregnancy were recruited (n=184; age 18-40 yrs). As maternally and paternally derived gene variants contribute equally to the function of a fetal genome, the patient group included both, the women and their partners, who had experienced recurrent pregnancy losses. In Estonian sample collection the patient group consisted of 32 couples and 29 females with RM, and additional 3 couples with ≥3 unsuccessful in vitro fertilization procedures. In Finnish sample collection the RM group consisted of 40 couples and 5 females with RM (detailed description in ​21​, ​22). The control group (n=195) consisted of age-matched fertile women with no history of miscarriage and either at least one normal pregnancy (the Finnish subjects, n=100) or more stringently, ≥3 successful deliveries (the Estonian subjects, n= 95). The control group was designed under the assumption that fertile women with no history of spontaneous abortions are carrying gene variants supporting successful pregnancies. Their male partners were not recruited into the control group as detailed reliable information on their past reproductive history was unavailable.

All patients had a normal karyotype tested from peripheral blood lymphocyte cultures. Female patients having uterine anomalies were excluded by ultrasonography or hysterosonogram.

Amplification and resequencing of CGB5 and CGB8

DNA was extracted from peripheral blood using a protocol based on the salting-out method for DNA extraction. The CGB5 (~ 1.7 kb fragment) and CGB8 (long-range PCR ~ 8.3 kb; nested PCR ~2.5 kb fragment) genomic regions (Fig. 1C) were amplified and resequenced using previously described primers and conditions (19). The resequenced region involving CGB8 covered 2050 bp including the entire CGB8 (1474 bp), 400 bp of 5′ upstream region. The resequenced region for CGB5 (1468 bp) covered the full genic region and part of 3′downstream region (Fig. 1C). Additional primers were designed for the analysis of the 5′upstream region of the CGB5 gene (450 bp) using the Primer3 software (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi). Specificity of the PCR products was verified in three steps: (1) design of unique primer pairs for specific amplification of only one of the seven duplicated genes; (2) verification of monomorphic status of gene-specific positions used as markers for each individual gene (Supplementary Fig. S1); (3) test for Hardy-Weinberg Equilibrium (HWE) for each identified SNP. Primer sequences for PCR and resequencing are listed in Supplementary Table S1. The sequences were resolved using either ABI 3730 X1 or ABI 3730 XL DNA Analyzer (Applied Biosystems) and assembled into a contig as described (19). Polymorphisms were identified using the PolyPhred program (Version 6.02) (http://www.phrap.org/phredphrapconsed.html) (23) and confirmed by manual checking. A genetic variant was verified only if it was observed in both forward and reverse orientations. In case of indel heterozygosity the genotype of the subject was confirmed using two independent forward and two reverse primers. The nomenclature of the identified polymorphisms was based on the GenBank reference sequences: NM_033043.1 GI:15451747 for CGB5, NM_033183.2 GI:146229337 for CGB8.

Data analysis

Allele frequencies were estimated and conformance to HWE was calculated (α = 0.05). In total 8 rare SNPs in 5′upstream region of CGB5 were found to be deviating from HWE, as one individual was homozygous for minor allele of all these SNPs.

Haplotypes were inferred from unphased genotype data using the Bayesian statistical method in the program PHASE 2.1.1 (http://www.stat.washington.edu/stephens/; ​24), applying the model allowing recombination. The running parameters were: number of iterations = 1000, thinning interval = 1, burn-in = 100; the - X10 parameter was used for increasing the number of iterations of the final run of the algorithm.

Sequence diversity parameters and neutrality tests were calculated using DnaSP (ver. 4.0; http://www.ub.es/dnasp/; ​25) with the most probable phased haplotypes as an input sequence. The direct estimate of per-site heterozygosity (π) was derived from the average pairwise sequence differences, while Watterson’s θ represents an estimate of the expected per-site heterozygosity based on the number of segregating sites (S). The basis of the Tajima’s D statistic (26) is the difference between the π and θ estimates: under neutral conditions π = θ and D^T = 0. The Ewens-Watterson homozygosity test implemented in Arlequin 2.000 software (http://cmpg.unibe.ch/software/arlequin3/​; ​27) was used to test the hypothesis that haplotypes are selectively neutral. An excess of rare variants (= homozygosity excess) indicates directional selection, while an excess of intermediate frequency variants (= homozygosity deficiency) indicates balancing selection. The relationship between inferred haplotypes was analyzed with NETWORK 4.201 software (http://www.fluxus-technology.com) using the Median-Joining network algorithm (28). Haplotype networks of CGB5 and CGB8 were calculated using (i) SNPs located in genic region from the transcription initiation site until the end of the mRNA and (ii) promoter SNPs located 5′upstream of the genic region. Singleton polymorphisms were excluded from network calculations (cannot be reliably phased) performed with default parameters. The descriptive statistics of linkage disequilibrium (LD), r² was calculated for pairs of markers and summarized by Haploview software (29).

The significance of the association between the identified SNPs in CGB5 and CGB8 genes and occurrence of RM was tested using Cochran-Armitage test for trend implemented in statistical analysis package JMP® 6.0.3 with Genomics module 2.0.6 (http://www.jmp.com/software/genomics/). The same test was applied to address the interpopulation (Estonians, Finns) differentiation. Odds ratio (OR) with 95% confidence intervals (CI) were calculated to show the strength and direction of the association. In all tests, p<0.05 was considered statistically significant.

Results

Resequencing of CGB5 and CGB8

We sequenced the entire genic and 5′upstream regions of CGB5 and CGB8 genes in a sample collection consisting of Finnish and Estonian patients with recurrent miscarriage (RM) (n=184; n=85 Finns, n=99 Estonians) and fertile controls (n=195; n=100 Finns, n=95 Estonians). For every subject the entire sequenced region covered 4.3 kbp (Fig. 1B-C). In total 71 variants were identified: 29 and 19 SNPs in the genic part of CGB5 and CGB8, respectively; 18 and 3 SNPs in the 5′ upstream regions of CGB5 and CGB8, respectively; and 2 SNPs 3′downstream of CGB5 (Table 1). Among the 71 detected SNPs 48 (68%) were novel variants, previously not described in dbSNP database (http://www.ncbi.nlm.nih.gov/SNP/) and literature. Neither CGB5 nor CGB8 has been covered by the most recent version of HAPMAP (http://www.hapmap.org/; release March 2008). The diversity parameter π that describes the mean nucleotide diversity per bp differed in the genic and 5′ upstream regions (Table 2). Among fertile women the diversity of CGB5 (π=2.71×10^-3) and CGB8 (~π=2.01×10^-3) 5′ upstream regions was approximately twofold higher compared to the genic regions of CGB5 (~π=1.69×10^-3) and CGB8 (~π=9×10^-4) (Table 2).

TABLE 1.

Characteristic of SNPs identified in CGB5 and CGB8 in Estonian and Finnish sample sets

SNP codea	Position relative to ATG	Location	Alleleb major/minor Aminoacid changec	Minor allele frequency in a subsample (%)		population difference p-valued	rs numbere
				Estonian (n=194)	Finnish (n = 185)
Variants in CGB5 genomic region
c5F-447	-812	5′ up-stream	T/G	0	S (Co)	0.304
c5F-399	-764		T/C	0	0.82	0.167
c5EF-322	-687		T/C	0.52	0.82	0.667
c5EF-315	-680		T/G	S (Pa)	1.9	0.071
c5EF-314	-679		C/A	S (Pa)	1.09	0.234
c5EF-309	-674		C/T	S (Pa)	1.09	0.234
c5EF-306	-671		T/G/C	S (Pa)	2.45	0.037
c5EF-291	-656		C/T	21.65	17.66	0.162	rs4801789
c5F-204	-569		A/G	0	1.36	0.051
c5F-191	-557		T/C	0	1.36	0.051
c5EF-155	-520		G/C	10.57	9.24	0.543
c5EF-147	-512		G/del	10.57	8.7	0.373
c5EF-144	-509		T/C	10.57	10.6	0.989
c5EF-142	-507		T/A	10.57	10.6	0.989
c5EF-82	-447		G/A	1.8	1.63	0.853
c5EF-30	-395		G/C	0.77	1.36	0.429
c5E-28	-393		C/T	1.55	0	0.016
c5EF-1	-366		A/G	S (Co)	S (Co)	0.973
c5E101	-265	5′UTR	C/T	0.52	0	0.166
c5EF138	-228		A/G	12.11	6.76	0.011	rs710899
c5E157	-209		C/T	0.52(Pa)	0	0.166
c5F206	-160		C/T	0	S (Co)	0.305
c5F324	-42		G/A	0	0.54 (Co)	0.146
c5EF345	-21		G/C	22.16	16.76	0.055	rs12610392
c5F354	-12		G/del	0	1.08	0.040
c5E519	154	intron1	G/T	0.52	0	0.166
c5E525	160		A/G	0.52	0	0.166	rs35621293
c5E527	162		G/A	0.52	0	0.166
c5E529	164		G/A	S (Pa)	0	0.328
c5EF544	179		T/G	33.25	25.68	0.024	rs3956245
c5E551	186		C/T	0.77	0	0.089	rs4002422
c5F553	188		T/C	0	0.54 (Co)	0.146
c5EF580	215		G/A	21.65	17.03	0.104	rs34524624
c5F660	295		A/C	0	S (Pa)	0.305
c5EF666	301		C/T	32.73	25.41	0.027	rs35871536
c5EF789	424	exon2	G/A p.Pro24Pro	21.13	17.03	0.148	rs35133942
c5E912	547	intron2	G/C	0.77	0	0.089
c5E918	553		C/G	0.52	0	0.166	rs33933429
c5EF1038	673		C/T	11.86	10.81	0.651	rs34335161
c5EF1069	704		A/G	1.55	2.43	0.376	rs33976607
c5F1111	746		A/T	0	0.54 (Co)	0.146
c5EF1115	750		C/T	3.35	0.54	0.005	rs34935416
c5F1178	813	exon3	G/C p.Val76Leu	0	S (Pa)	0.305
c5EF1258	893		C/T p.Tyr102Tyr	4.64	0.81 (Co)	0.001	rs35756580
c5E1390	1025		C/A p.Pro146Pro	S (Co)	0	0.328
c5EF1402	1037		T/C p.Ser150Ser	S (Co)	S (Pa)	0.973
c5EF1426	1061		G/A p.Ser158Ser	S (Co)	S (Pa)	0.973
c5E1501	1136	3′ down-stream	T/C	S (Pa)	0	0.328
c5EF1660	1295		A/T	1.29	1.08	0.502
Variants in CGB8 genomic region
c8EF-287	-659	5′ up-stream	T/C	26.49	23.85	0.417	rs4801790
c8EF-186	-558		G/T	40.21	39.08	0.754	rs8102901
c8EF-4	-376		T/A	0.52 (Pa)	S (Pa)	0.627
c8EF105	-268	5′UTR	G/C	4.12	0.57 (Co)	0.003	rs34212754
c8EF108	-265		C/T	39.95	39.08	0.808	rs13345685
c8EF301	-72		T/A	5.41	6.32	0.597	rs35930240
c8EF432	60	intron1	A/C	0.52	0.57 (Co)	0.913
c8F461	89		T/C	0	S (Pa)	0.290
c8EF523	151		G/T	S (Co)	0.86 (Pa)	0.264
c8F526	154		T/G	0	S (Co)	0.290	rs2387591
c8EF541	169		G/C	39.69	39.37	0.928	rs13345575
c8F551	179		G/T	0	S (Co)	0.290
c8F558	186		T/C	0	S (Co)	0.290
c8EF673	301		T/C	S (Co)	S (Co)	0.938
c8E806	434	exon2	C/T p.Arg28Trp	S (Pa)	0	0.343
c8EF869	497		G/A p.Val49Ile	2.32	S (Pa)	0.017
c8EF925	553	intron2	G/C	2.06	3.16	0.341	rs2303050
c8EF1045	673		C/T	0.52	1.72 (Co)	0.112	rs33943298
c8EF1076	704		G/A	1.8	2.01	0.836
c8EF1122	750		T/C	1.8	2.01	0.836
c8E1237	865	exon3	C/G p.Pro93Arg	S (Pa)	0	0.343
c8E1418	1046		A/T pArg153Arg	S (Pa)	0	0.343

^aa SNP code includes gene and sample name (e.g.c5=CGB5; E=Estonians, F=Finns), location relative to mRNA start site; GenBank references: NM_033043.1 GI:15451747 for CGB5, NM_033183.2 GI:146229337 for CGB8; non-synonymous changes detected only in patients are underlined

^balleles at the coding strand

^ccoding from ATG including signal protein

^dCochran-Armitage test for trend

^eVariants originally described by Hallast et al., 2005 (19) are highlighted in bold. All variants identified in the current study have been submitted to dbSNP database (http://www.ncbi.nlm.nih.gov/SNP/), accession numbers ss105106983 - ss105107053.

S - singleton SNP; Co - only among fertile women with no miscarriages; Pa - only among RM patients

TABLE 2.

Diversity parameters and neutrality tests of the CGB5 and CGB8 genes in fertile women and patients of recurrent miscarriages (RM)

	CGB5			CGB8
	Full regiona	5′upstream regionb	Genic regionc	Full regiona	5′upstream regionb	Genic regionc
No of SNPs	46	17	29	23	3	20
Fertile women
Diversity (π)d	0.00193	0.00271	0.00169	0.00119	0.00201	0.00095
Tajima De	-1.12980	-1.16944	-0.88132	-0.35277	2.29389 *	-0.96141
P-value of Ewens-Watterson F statisticf	ns	ns	ns	ns	0.007	ns
RM Patients
Diversity (π)d	0.00172	0.00195	0.0016	0.00115	0.00199	0.0009
Tajima De	-1.23744	-1.23550	-0.91052	-0.56306	1.26879	-1.06361
P-value of Ewens-Watterson F statisticf	ns	ns	ns	ns	ns	ns

^aSNPs in 5′upstream and genic regions

^bSNPs located in the region of −435 bp up to −1 bp from the start site of mRNA sequence

^cSNPs located in mRNA seqence: +1 bp up to +1082 bp from the start site of mRNA sequence

^dEstimate of nucleotide diversity (π) per basepair from average pairwise differences among individuals

^eThe basis of the Tajima’s D statistics (D^T) is the difference between observed (π) and expected (θ) diversity estimates: under neutral conditions π = θ and D^T = 0

^fThis statistic tests the observed allele frequency spectrum with the expected allele frequency spectrum under the neutral model (Hardy-Weinberg Equilibrium).

^*p=0.0169; ns - non-significant (p>0.05)

Two thirds of the identified variants (n=41; 58%) were shared by Estonian and Finnish sample collections. In both sample collections there were 15 population-specific SNPs represented as single or low frequency variants (<2%). Majority of the shared SNPs showed no differences (p>0.05) among the two study populations. Significant difference in allele frequencies was detected for 8 out of 71 SNPs, most being rare variants (Table 1). Linkage disequilibrium (LD) between the identified SNPs in the resequenced region was nearly absent in both population samples (Fig. 2).

Figure 2.

LD structure of the resequenced CGB5-CGB8 genomic region in Estonian and Finnish sample sets. The plot has been drawn based on r² statistic and polymorphisms are ordered as located on the sense strand of genomic DNA sequence. Singleton variants have been excluded for reliable estimation of LD pattern. The direction of transcription of CGB5 and CGB8 genes is depicted with arrows. Polymorphisms with a significant association to recurrent miscarriages (Table 3) are given above the LD plot.

Four SNPs represented non-synonymous amino acid changes: CGB5 p.Val76Leu in a single Finnish RM patient, CGB8 p.Arg28Trp and CGB8 p.Pro93Arg in single Estonian patients, and CGB8 p.Val49Ile in one Finnish and two Estonian patients, and also seven Estonian fertile women (Table 1). Further experimental studies have to be conducted before drawing any conclusions about their effect on the hormone function.

CGB5 and CGB8 variants lowering the risk for recurrent pregnancy loss

A case-control study targeting the association of identified CGB5 and CGB8 genetic variants with RM was carried out separately for the Estonian (RM cases n=99; fertile women defined as controls n=95) and the Finnish subjects (cases n=85; controls n=100) as well as for the joint dataset. The comparison of single marker and haplotype distribution in the two sample sets revealed low population stratification (Table 1, Supplementary Fig. S2) facilitating the joint analysis in order to increase the statistical power of the study.

In the full case-control sample set a significant association with RM was detected for the CGB5 5′upstream polymorphisms (c5EF-155, c5EF-147, c5EF-144, c5EF-142) (p<0.03; OR=0.58 [95% CI 0.35-0.93]; Cochran-Armitage trend test) (Table 3). Analysis of the Estonian (p=0.083; OR=0.54 [95% CI 0.27-1.1]) and the Finnish (p≤0.131; OR=0.58 [95% CI 0.29-1.19]) subsamples supported trend for association in both study populations independently, but the p-values did not reach statistical significance (p>0.05) due to reduced samples sizes. The significant association with all four CGB5 promoter polymorphisms results from higher minor allele frequency (MAF) in fertile women (12.05%-13.08%) compared to RM group (7.10%-7.92%). This difference between the control group and RM cases was consistent in both study populations: 13.16% compared to 8.08 % and 11-13% compared to 5.95-7.74%, in Estonians and Finns, respectively (Table 3). On the haplotype network of CGB5 upstream region the promoter variants carrying the minor alleles of the four polymorphisms form a remote clade (H1-H2, H10-H11; Fig. 3A).

TABLE 3.

Variants in CGB5 and CGB8 genes significantly associated with recurrent miscarriage (RM). Association P-values and odds ratio (OR) with 95% confidence interval (CI) was calculated by Cochran-Armitage test for trend

	Estonians (n=194)				Finns (n=185)				All individuals (n=379)
	MAF (%)		p-value	OR (95%CI)	MAF (%)		p-value	OR (95%CI)	MAF (%)		p-value	OR (95%CI)
SNP	Fertile women n=95	RM patients n=99			Fertile women n=100	RM patients n=85			Fertile women n=195	RM patients n=184
c5EF-155	13.16	8.08	0.083	0.54 (0.27-1.1)	11.50	6.55	0.129	0.58 (0.29-1.19)	12.31	7.38	0.024	0.57 (0.35-0.93)
c5EF-147	13.16	8.08	0.083	0.54 (0.27-1.1)	11.00	5.95	0.094	0.52 (0.24-1.13)	12.05	7.10	0.018	0.54 (0.32-0.91)
c5EF-144	13.16	8.08	0.083	0.54 (0.27-1.1)	13.00	7.74	0.131	0.61 0.31-1.17)	13.08	7.92	0.023	0.58 (0.36-0.93)
c5EF-142	13.16	8.08	0.083	0.54 (0.27-1.1)	13.00	7.74	0.131	0.61 (0.31-1.17)	13.08	7.92	0.023	0.58 (0.36-0.93)
c5EF1038	14.47	9.09	0.079	0.57 (0.30-1.08)	14.00	7.06	0.036	0.48 (0.24-0.97)	14.36	8.15	0.007	0.53 (0.32-0.85)
c8EF301*	5.79	5.05	0.740	0.86 (0.35-2.12)	8.85	3.21	0.034	0.35 (0.12-0.96)	7.33	4.24	0.072	0.55 (0.29-1.06)
c8EF1045*	0.53	0.51	0.977	0.97 (0.06-15.6)	3.13	0	0.025	na	1.83	0.28	0.042	0.15 (0.02-1.03)

^*significant association (p<0.05) only in one population

MAF - minor allele frequency; na - not applicable as monomorphic among fertile women

Figure 3.

Median-Joining (MJ) networks of predicted haplotypes in the 5′ upstream region of CGB5 (A) and CGB8 (B). Singleton polymorphisms were excluded from the analysis, because of unreliable phasing. The size of each node is proportional to the haplotype frequency in the total dataset. The relative distribution of each haplotype among the recurrent miscarriage (RM) cases (black) and fertile controls (white) is indicated. Haplotype nomenclature is shown in Supplementary Table S2. (A) The carrier status of haplotypes H1-H2 and H10-H11 lowered 1.7-fold the risk of RM. (B) Haplotype H2 defined by the minor allele of a proximal promoter mutation c8EF-4 was exclusively identified in RM patients in both study populations, Estonians and Finns.

MJ networks of predicted haplotypes in the genic regions of CGB5 and CGB8 are shown in Supplementary Fig. S3.

Among the CGB5 genic SNPs a strong protective effect was detected for the minor allele of intron 2 c5EF1038 (p<0.007; OR=0.53 [95% CI 0.32-0.85]), represented with the frequency 14.36% in fertile women compared to 8.15% in the RM group (Table 3). This effect reached statistical significance in the separate analysis of the Finnish subjects (p=0.036; OR=0.48 [95% CI 0.24-0.97]) and showed a trend for association in the Estonian (p=0.079; OR=0.57 [95% CI 0.30-1.08]) subsample. No increase in protection towards RM was detected for the combination of the minor alleles of the CGB5 5′upstream and the intronic SNP (data not shown).

Notably, the association of four CGB5 SNPs with the protective effect towards pregnancy loss was sufficiently robust to remain significant even when only the female RM patients (n=109) were considered as cases (Table 4). A separate analysis of male RM patients revealed similar trends for association and protective effect sizes as compared to female RM cases, although the p-values were non-significant possibly due to smaller sample size (n=75) that reduced the statistical power. However, the inclusion of both sexes gave a stronger effect than gender-specific analysis in all but one SNP (c5EF1038; Table 3, 4) further supporting the contribution of both, maternal and paternal genes in the reproductive success.

TABLE 4.

Case-control analysis after subdividing recurrent miscarriage (RM) patients by gender. Cochran-Armitage test for trend was performed for the variants in CGB5 and CGB8 genes showing significant association with RM in the full sample set (Table 3)

	Fertile women n=195	Female RM patients n=109			Male RM patients n=75
SNP	MAF %	MAF %	p-value	OR (95%CI)	MAF %	p-value	OR (95%CI)
c5EF-155	12.31	7.34	0.058	0.59 (0.32-1.03)	7.43	0.105	0.57 (0.26-1.13)
c5EF-147	12.05	6.88	0.039	0.53 (0.28-0.98)	7.43	0.115	0.57 (0.28-1.15)
c5EF-144	13.08	7.34	0.031	0.53 (0.29-0.95)	8.78	0.171	0.64 (0.34-1.22)
c5EF-142	13.08	7.34	0.031	0.53 (0.29-0.95)	8.78	0.171	0.64 (0.34-1.22)
c5EF1038	14.36	6.88	0.006	0.45 (0.25-0.81)	10	0.182	0.66 (0.36-1.22)

MAF - minor allele frequency; OR - odds ratio; CI - confidence interval

Population-specific associations were detected in the Finnish sample collection with two rare SNPs (MAF <10%) in CGB8: c8EF301 (p=0.034) and c8EF1045 (p=0.025) (Table 3). Interestingly, the protective variant in the intron 2 of CGB8 (c8EF1045) is located at the same position within the gene as the CGB5 intronic variant (c5EF1038).

Rare CGB8 promoter variants increase the susceptibility to recurrent miscarriages

The resequenced CGB8 5′upstream region stands out with only 3 SNPs (two common and one rare) compared to the respective region for CGB5 with 18 SNPs (Table 1). The rare allele A of SNP c8EF-4 was solely represented in patients, one from Finland and two from Estonia (Cochran-Armitage trend test, p=0.071). This polymorphism is located within the AP1-like sequence overlapping the HCG beta initiator element critical for basal transcription and downstream of the Ets-2 binding site acting as a major enhancer of HCG beta gene expression (30).

We applied two neutrality tests to explore observed versus expected distribution of SNPs and haplotypes in the 5′upstream region of CGB8. Both, the Tajima’s D statistic (D^T=2.29, p<0.05; Table 2) as well as Ewens-Watterson homozygosity test (p=0.007) indicated a possible scenario of balancing selection driving the three apparently most efficient CGB8 promoter variants (H1, H3, H4) to high frequency in both populations (Fig. 3B; Supplementary Table S2: Supplementary Fig. S2). The rare variant H2 carried the minor allele of c8EF-4, identified solely in patients. Notably, the haplotype combining the minor alleles of c8EF-287 (C; MAF=25.2%) and c8EF-186 (T; MAF=39.7%) is expected to be present with the frequency of 10%, but was not observed in the current study (Fig. 3B; Supplementary Table S2).

Discussion

Here we report the first case-control study targeting the variation in HCG beta genes in association with recurrent miscarriages (RM). Most association studies on RM have so far focused on susceptibility variants of maternal genes involved in physiological adaptation to pregnancy, such as development of immunotolerance at feto-maternal interface or alterations in fibrinolytic and coagulation pathways. As these genes also contribute to complex diseases, the role of their variants in susceptibility to RM may not be specific (5, 8, 22, 31). HCG beta genes are expressed in blastocysts shortly after fertilization (20, 32) and are essential for successful implantation. Thus, a genetic variant of these genes is more likely to have an effect on pregnancy outcome. Our study focused on CGB8 and CGB5 that provide the major fraction of HCG beta mRNA transcripts and the resequencing method was chosen instead of genotyping.

The human CGB8 and CGB5 genes are located among the seven duplicate genes within the LHB/CGB gene cluster. Major complications in targeting duplicated genes in association studies are high sequence similarity (>92%), high diversity, large number of population-specific variants and low LD due to high gene conversion activity (19, 33). These characteristics make it technically challenging to select reliable tag-SNPs and establish genotyping methods capable of targeting unique SNPs in duplicated genes. In the 379 subjects we identified only 14 out of 30 (47%) SNPs in CGB5 present in a public SNP database NCBI dbSNP and 9 out of 44 (20%) in CGB8. Several of the variants not observed in our study have been predicted in silico or by using high-throughput methods and may actually be multisite or paralogous gene variants (34). Alternatively, some of these SNPs could indeed represent variants specific to other than Estonian or Finnish populations. For example, an amino acid substitution Val79Met (nomenclature based on mature protein; from ATG p.Val99Met) in CGB5 exon 3 has been reported at carrier frequency 4.2% in a random population from the Midwest of the United States (35) but it was absent in a 580 DNA samples originating from five European populations (36). In the current study, in relatively large samples sets drawn from two neighboring populations, one third of the identified variants (MAF<2%) were found in only one population, although the sample size was sufficient to identify all common variants (MAF>5%) originally described in a large mutation screening of LHB/CGB genes (Table 1; ​19). Full resequencing data collected in this study enabled to identify several rare non-synonymous and promoter variants and to conduct haplotype analysis.

Consistent with the hypothesis of the study, we identified genetic variants in HCG beta genes significantly increasing or reducing the risk of RM. A protective effect was detected for the minor alleles of two SNPs (c5EF1038 and c8EF1045) located at the identical positions in intron 2 in both CGB5 and CGB8 and for four CGB5 promoter variants (c5EF-155; c5EF-147; c5EF-144; c5EF-142). The carrier status of the minor alleles of these six SNPs reduced the risk of RM 1.7-fold in comparison to the wildtype carriers. Interestingly, the “protective” alleles of the CGB5 promoter SNPs form a motif (C-del-C-A; H2 on Fig. 3A) identical to the promoter sequence of CGB8 (Fig. 1D), which has been shown to be most actively transcribed HCG beta gene (18). The actual contribution of these sequence variants to mRNA transcription and splicing efficiency is still to be explored.

The current data suggest the CGB8 and especially its promoter region to be under stronger functional constraint compared to CGB5 in spite of high DNA sequence similarity (98-99%) between the two genes (Supplementary Fig. S1). Firstly, we detected > 2 times less polymorphisms in CGB8 genomic region (n=22) compared to CGB5 (n=49). Secondly, three rare CGB8 variants that may exhibit an effect on hormone action were present exclusively in RM patients (p.Arg28Trp, p.Pro93Arg and a c8EF-4 within proximal promoter) compared to only one such SNP in CGB5 (p.Val76Leu). Thirdly, the applied neutrality tests indicated a balancing selection in the promoter region of CGB8, but not of CGB5. Additionally, we identified only three of the four predicted major CGB8 promoter haplotypes (Fig. 3B). The haplotype combining the minor alleles of c8EF-287 (MAF 25.2 %) and c8EF-186 (MAF 39.7%) was absent in the current dataset in spite of the relatively high minor allele frequency. The discrepancy between observed (0%) and expected (10%) frequency may be explained by the localization of these SNPs within Sp1/AP-2 binding sites (37) residing in the critical region for the trophoblast-specific expression as well as cAMP-responsiveness of the HCG beta gene transcription (38, 39). Functional studies should reveal whether these sequence variants indeed possess a combinatory effect influencing the binding of the AP-2 and Sp1 transcription factors to the promoter of HCG beta and alter the transcription of genes.

One of the key factors in obtaining reliable results in a case-control study is a clearly defined study group and replication of the results in an independent dataset. We applied parallel analysis of case-control sample sets collected from two neighboring countries in order to confirm the robustness of the association across populations. As the stratification was low between these populations we also conducted a joint analysis of the two sample sets in order to raise the statistical power. Although there were minor differences in subject recruitment, the obtained results were concordant in two populations and the strength of detected associations increased in the analysis of the pooled dataset. In addition we identified two gene variants lowering the risk of RM in the Finnish dataset only possibly owing to the specific demographic history of the Finnish population (40).

In conclusion, these data from two populations provide the first evidence for the role of the variation in HCG beta genes in contributing to the susceptibility of RM. The findings encourage further studies addressing the functional effect of the identified promoter, intronic and rare protein-altering variants on HCG beta gene expression and HCG hormone activity. The diagnostic application of our findings may facilitate the improvement of early and preventive treatment of RM.