Analysis of X chromosome genomic DNA sequence copy number variation associated with premature ovarian failure (POF)

Abstract

BACKGROUND

Premature ovarian failure (POF) is a heterogeneous disease defined as amenorrhoea for >6 months before age 40, with an FSH serum level >40 mIU/ml (menopausal levels). While there is a strong genetic association with POF, familial studies have also indicated that idiopathic POF may also be genetically linked. Conventional cytogenetic analyses have identified regions of the X chromosome that are strongly associated with ovarian function, as well as several POF candidate genes. Cryptic chromosome abnormalities that have been missed might be detected by array comparative genomic hybridization.

METHODS

In this study, samples from 42 idiopathic POF patients were subjected to a complete end-to-end X/Y chromosome tiling path array to achieve a detailed copy number variation (CNV) analysis of X chromosome involvement in POF. The arrays also contained a 1 Mb autosomal tiling path as a reference control. Quantitative PCR for selected genes contained within the CNVs was used to confirm the majority of the changes detected. The expression pattern of some of these genes in human tissue RNA was examined by reverse transcription (RT)–PCR.

RESULTS

A number of CNVs were identified on both Xp and Xq, with several being shared among the POF cases. Some CNVs fall within known polymorphic CNV regions, and others span previously identified POF candidate regions and genes.

CONCLUSIONS

The new data reported in this study reveal further discrete X chromosome intervals not previously associated with the disease and therefore implicate new clusters of candidate genes. Further studies will be required to elucidate their involvement in POF.

Introduction

Premature ovarian failure (POF also known as primary ovarian insufficiency) occurs in 1% of women (Coulam et al., 1986) and is defined as amenorrhoea for >6 months before age 40, with an FSH serum level >40 mIU/ml (Conway, 2000). POF is a heterogeneous disease, which can develop as a result of a broad spectrum of pathogenic mechanisms including genetic, autoimmune and iatrogenic causes. However, in many cases the cause is unknown. At birth, the ovary has a fixed number of primordial follicles (~ 10 × 10⁶), which under normal fertile conditions steadily decline either by atresia or during ovulation (Gosden and Faddy, 1998). However, in women with POF, there is usually premature depletion of the primordial follicle pool. This is due to defects in oocyte apoptosis mechanisms leading to either a decrease in follicular formation, resulting in a reduction of oocytes formed during ovarian development, or accelerated follicle loss. In some cases, follicles are present but there is no response to hormonal stimulation (Goswami and Conway, 2005). In addition to a reduction in the duration of fertility, there are other important health issues that may be associated with POF such as an increased risk of overall mortality, cardiovascular diseases, osteoporosis and autoimmune disorders such as diabetes or problems with the thyroid or adrenals. Furthermore, the earlier the menopause, the higher the risk of morbidity and mortality. Estrogen treatment may prevent some of these outcomes but not all, suggesting that other mechanisms are involved (Shuster et al., 2009).

There is a strong genetic association with POF, as confirmed by numerous familial cases. Familial POF is reported to have an incidence of around 4% but epidemiological studies have suggested it may be as high as 30% (Goswami and Conway, 2005). Vegetti et al. (2000) have found that idiopathic POF and early menopause patients (amenorrhoea age 41–44) may in fact be due to variable expression of the same disease with both having a high rate of familial transmission of the condition (28.5 and 50%, respectively), and pedigree analyses have suggested an autosomal or an X-linked dominant sex-limited pattern of inheritance for POF. It therefore appears that a large proportion of idiopathic POF cases are also due to genetic defects. Recently, a preliminary genome-wide case/control association study with idiopathic POF patients has provided evidence of significant linkage disequilibrium between POF and a single-nucleotide polymorphism marker in the ADMTS19 gene (Knauff et al., 2009). This study also produced suggestive associations with the BDNF, CXCL12, LHR, USP9X and TAF4B genes, confirming that multiple loci contribute to this phenotype.

For cytogenetically visible aberrations, abnormalities of the X chromosome are particularly relevant and have frequently been associated with POF. Turner syndrome presents with primary amenorrhoea, streak ovaries and X chromosome monosomy. The ovarian phenotype in these individuals cannot be explained purely by pairing failure as women presenting with Turner stigmata have also shown partial X interstitial deletions of both Xp and Xq. This illustrates how monosomy for certain X-linked genes, which may escape X-inactivation, could account for the common phenotypic anomalies associated with the condition, including ovarian failure.

The critical Xq region for ovarian development and function has been identified ranging from Xq13.3 to Xq27 (Therman et al., 1990) and is often divided into two portions. The first is the POF1 region, which has been proposed to extend from Xq21 to Xqter (Davison et al., 1998). The second is the POF2 region located in the interval Xq13.3–Xq21.1. These intervals have been defined by a combination of deletions and X:autosome translocations. Sala et al. (1997) mapped the breakpoints of X:autosome translocations in 11 women to the POF2 region and a review of cytogenetic and molecular analysis of balanced X-autosome translocations by (Persani et al., 2009) found breakpoints were commonly found between Xq13 and Xq21, whereas interstitial deletions were preferentially located between Xq23 and Xq27. These studies have suggested several candidate genes, although very few discrete mutations have actually been detected in these loci (summarized in Knauff et al., 2009). Some of the X-linked candidate genes interrupted by translocations and therefore possible candidates for POF are the following: XPNPEP2 (Xq25), POF1B (Xq21.2), DACH2 (Xq21.3), CHM (Xq21.2) and DIAPH2 (Xq22). FMR1 (Xq27) is also deemed a candidate in view of the significant association of CGG repeat expansion with POF (Bodega et al., 2006). Studies have also indicated the proximal Xp region to play a key role in ovarian function (Zinn et al., 1998; Ogata et al., 2001; Lachlan et al., 2006). Some of the key potential POF3 candidate genes suggested by structural abnormalities in Xp include ZFX (Xp22.1–21.3) and BMP15 (Xp11.2).

A few studies have reported investigation of POF patients with chromosome aberrations using array technology. Han et al. (2006) used array comparative genomic hybridization (aCGH) to reveal that their patient with POF was the carrier of an unbalanced de novo translocation between the X chromosome and chromosome 11, resulting in partial monosomy Xq and partial trisomy 11p. This patient also had features consistent with Beckwith-Wiedemann syndrome. A similar analysis of a patient with a balanced translocation [t(X;15)] was performed by Bertini et al. (2010). In 2007, Tachdjian et al. (2008) used aCGH to refine the breakpoints of three POF patients presenting with Xq deletions. In one case, an Xp duplication involving the SHOX gene was also detected although it had not been visible cytogenetically. A whole genome analysis using a 0.7 Mb resolution interval array further identified eight statistically significant CNVs, one of which was on the X chromosome (Aboura et al., 2009).

In the current study, 42 idiopathic cytogenetically normal POF patients were analysed by aCGH for CNV using a complete end-to-end X/Y chromosome tiling path. The arrays also contained a 1 Mb autosomal tiling path as a reference control. As a large proportion of idiopathic POF is likely to be due to genetic causes, the aim of this study was to detect cryptic CNV of the X chromosome which were not visible cytogenetically. No autosomal abnormalities were detected with the 1-Mb array. A number of CNV were identified by the complete X chromosome tiling path on both Xp and Xq. Some CNV fall within known polymorphic copy number variation regions and others span previously identified POF candidate intervals and genes. The new data reported in this study reveal further discrete intervals on the X not previously associated with the disease, and therefore implicate new clusters of X-linked candidate genes.

Materials and Methods

Patient samples and clinical assessment

Our current cohorts of idiopathic POF patients have come from two clinical centres: from Dr Gerard Conway, University College London (UCL) Hospitals, London (2 cases) and Dr Anna Murray, Wessex Regional Genetics Laboratory, Salisbury District Hospital, Wiltshire (40 cases). Patients from UCL were all <40 years of age, had an FSH serum >40 mlU/ml on two occasions and had no previous medical history to explain POF. Ethical approval had been obtained for this study from the local Research Ethics Committee at University College London. Patients from the Wessex Regional Genetic Laboratory had been referred following diagnosis of POF on the basis of the normally accepted criteria (see above). All patients provided consent to be tested for genetic causes of their ovarian failure, under ethical approval from Salisbury Research Ethics Committee, ref SA 32/2000. All patients were tested by conventional cytogenetics and additionally the Wessex cases were tested for FMR1 expansion mutations. None of the patients showed any abnormalities of the X chromosome. Six were cases of primary and 36 were cases of secondary amenorrhoea. Genomic DNA was extracted from peripheral blood lymphocytes using standard procedures.

Array CGH

The description of the X and Y tiling path arrays and the procedures for labelling genomic DNA, for hybridizing the arrays, for capturing the data and for subsequent data analysis to reveal CNVs were as detailed in Karcanias et al. (2007). Briefly, the chromosomes X and Y clone sets (derived from the Golden Path tile set) were obtained from the Sanger Institute, Cambridge. The X chromosome tiling path consists of 1708 clones (1083 BACs, 517 PACs, 86 cosmids and 22 fosmids), providing 94.4% coverage. The Y chromosome tiling path (195 BACs, 4 cosmids and 2 fosmids) covers 92.6% of the euchromatic region, excluding the pseudoautosomal region 2. The array also included (for the normalization of copy number changes) 3000 BAC clones covering all autosomes at ~1-Mb intervals and 6 Drosophila BAC clones for background measurement. In order to detect CNVs, 400 ng of test and reference DNA labelled with Cy5-dCTP and Cy3-dCTP, respectively, were hybridized to the arrays and washing was carried out according to the published protocol (McCabe et al., 2006). The reference DNA was from a single female. The arrays were scanned using a GenePix 4100A personal scanner (Axon Instruments, Union City, CA, USA). The scanned images were quantified using BlueFuse software (BlueGnome, Cambridge, UK). Analysis and normalization were carried out as published elsewhere (McCabe et al., 2006).

RT–PCR

Human tissue total RNA samples were purchased from Ambion (Austin, TX, USA) and from each RNA sample, cDNA was prepared using an AMV Reverse Transciption Kit (A3500: Promega, Madison, WI, USA) according to the manufacturer’s instructions. Primer pairs were designed from the exonic sequence of candidate genes (AIFM, BCORL1, XPNPEP2, ZFX, H2BFM, H2BFWT, RBMX2, USP9X, USP27X, UTP14A, CENPI) and used in a hot-start PCR (see Table I for primer sequences). PCR was performed on 10 ng cDNA in 10 μl reaction volumes using HotStarTaq DNA polymerase (Qiagen, Hilden, Germany). Reactions were carried out on an Applied Biosystems thermocycler according to the following conditions: 95°C for a 15 min hot start, then 95°C for 30 s (denaturation), 55°C for 30 s (annealing), 72°C for 1 min (extension) for 40 cycles and finally incubation at 72°C for 10 min. PCR products were electrophoresed on a 3% agarose gel that was subsequently stained with ethidium bromide and visualized under UV irradiation.

Table I.

Primer sequences used for RT–PCR and qPCR.

Gene	Primer	Sequence	Q-PCR/ RT–PCR
AIFM1	AIFM1ex1-F	GTCGTGCGTGAGAGGAAAG	both
	AIFM1ex1-R	GCACCAGCTTCTGCTTCAA	both
H2BFWT	H2BFWTEX-F	TGTCTGGTCGTGCCATCTAA	both
	H2BFWTEX-R	CCTCTGCTTGCTCTGCTTCT	both
USP27X	USP27XEX1-F	TCTACCAGTGCTTCGTGTGG	both
	USP27XEX1-R	CAGGAAAGGCAAGAGTGGAG	both
UTP14A	UTP14AEX6-F	GCAGATGGGCCGAATCCCTGG	both
	UTP14AEX6-R	TGGGCCTCAAGCTCTGGCAGT	both
BCORL1	BCORL1-F	TTTTCTCTCTCCCCCAATCC	Q-PCR
	BCORL1-R	GCAGGGAAAAGAAGGAACAG	Q-PCR
CENPI	CENPI-F	TTTAGGGCAAGGACTTTCTGAG	Q-PCR
	CENPI-R	CTCCACAGAGACAGGGTGTG	Q-PCR
CT45A4	CT45A4EX2-F	CCAGCCAATTGGATTCTCAG	Q-PCR
	CT45A4EX2-R	GTAACGTTTCCTCCCACAGG	Q-PCR
FAM123B	FAM123B-F	GCTGGAACAGCTGTGTAACG	Q-PCR
	FAM123B-R	TTGAGTCCGCAGAAAGGAG	Q-PCR
H2BFM	H2BFM-F	TTCCGTAATCGTGTTTCGTG	Q-PCR
	H2BFM-R	ATCCCACCATCCAGTCGTAG	Q-PCR
HDAC6	HDAC6-F	AGGGAGAAGGCCTGAGAGAG	Q-PCR
	HDAC6-R	GGTATGTGAGGGGGCTAGTG	Q-PCR
HPRT1	HPRT1-F	CCTGGGAAAAGAGGACTGC	Q-PCR
	HPRT1-R	CATCATTCCCGAATCTGC	Q-PCR
MFN2	MFN2 -F	TTTGCCAGCATTTACTCAGC	Q-PCR
	MFN2 -R	AAACAGGGCTGAACGAGAAG	Q-PCR
PNPLA4	PNPLA4ex2-F	TGGATTTCTGGGCATTTACC	Q-PCR
	PNPLA4ex2-R	ACAGAAGCAACCAACGATCC	Q-PCR
POF1B	POF1B	GAAAGGCAATCTTCAACCTG	Q-PCR
	POF1B	CAGCCAAGAGATGAATTTGG	Q-PCR
RP3.378P9.2	RP3.378P9.2-F	TGACTTCAGGGCACTTTTGA	Q-PCR
	RP3.378P9.2-R	CAGGACTGATGATTCCAGCA	Q-PCR
SAGE1	SAGE1-F	TCACGATATCCAGGAGGAGG	Q-PCR
	SAGE1-R	TGGGTGGCATACAATGTCCT	Q-PCR
TSPAN7	TSPAN7-F	TTTCTCTCCCTTCCCCTACC	Q-PCR
	TSPAN7-R	GTGGGGTCAGGAATAACTGG	Q-PCR
USP9X	USP9X-F	AGCGTGTCTGTGTGTTTTGG	Q-PCR
	USP9X-R	CGGAGACTCCATCCTCCTAC	Q-PCR
VCX	VCX-F	CCATGTAGGTCAGGCTGGTC	Q-PCR
	VCX-R	CGGAGGGCTATATGAAGACG	Q-PCR
XPNPEP2	XPNPEP2-F	CAAGGCCAAGTGGAGAAGAG	Q-PCR
	XPNPEP2-R	GGGCACCTAGTTGAGGACAG	Q-PCR
ZC3H12B	ZC3H12BEX-F	CCAAGGAAGAGAAGCAGCAG	Q-PCR
	ZC3H12BEX-R	ATCCCTAGGTTGGCAGCTCT	Q-PCR
ZFX	ZFX-F	CGCTATGTAAATATCGGTGAGG	Q-PCR
	ZFX-R	GGCTGTAGCGGAATTTTCTC	Q-PCR
BCORL1	BCORL1EX4-F	TGAGCAAACAGGTTGACTGC	RT–PCR
	BCORL1EX4-R	AGGGACTCCAGTTCCCAGAG	RT–PCR
CENPI	CENPIEX1-F	AGTGATTCTCCTGCCTCAGC	RT–PCR
	CENPIEX1-R	CTGGCCAATATGAGGAAACC	RT–PCR
H2BFM	H2BFMEX1-F	TATCTAATGGCCGCTGCTTC	RT–PCR
	H2BFMEX1-R	CTTCTGCTTCTGGGCCTTC	RT–PCR
RBMX2	RBMX2EX6-F	GCACAGCAGCAAGAACTCAG	RT–PCR
	RBMX2EX6-R	TGGGTTTCTCCTTCTTCAGC	RT–PCR
USP9X	USP9XEX1-F	GGAGCGAGCTACTTCAAAGC	RT–PCR
	USP9XEX1-R	GTTTCTCCTGCGTCACCTC	RT–PCR
XPNPEP2	XPNPEP2EX10-F	CTCCAGTTGCACAGGCCCCA	RT–PCR
	XPNPEP2EX10-R	GCTGGTCCCAATCCAGATCCTCACA	RT–PCR
ZFX	ZFXEX6-F	ACATTGCACAGTCCCAGATG	RT–PCR
	ZFXEX6-R	TGTCCAACATGTCCAACGTC	RT–PCR

Q-PCR

Primer pairs were designed to amplify products from genes present within each CNV, to confirm the aCGH copy number changes, using quantitative PCR (Q-PCR) (see Table I for primer sequences). Where possible, candidate genes were chosen. Q-PCR was carried out on 10 ng of patient DNA using the QuantiTect SYBR Green PCR kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. Reactions were run on a BioRad iCycler (BioRad, Bath, UK) as previously described (Quilter et al., 2008). Genes were tested in triplicate and normalized to single copy control genes HPRT1 (on the X chromosome) and MFN (representing an autosome). Quantification of the copy number was carried out using the comparative C_t method (2^−ΔΔC_t) (Livak and Schmittgen, 2001). C_t values from patients were normalized to the control genes and the ΔΔC_t was calculated by finding the difference between each sample ΔC_t and that of the reference normal female control (46,XX), to give an estimated copy number of each gene, for each patient.

Results

Copy number variation analysis of POF patients

Of the 42 patients tested on the X/Y tiling path and 1-Mb array, 20 were found to have 15 different CNVs affecting the X chromosome. No CNVs were detected by the 1 Mb interval array covering the autosomes. This may be because of the lower resolution coverage across the autosomes. All patient DNA samples were tested against the same normal female reference DNA and a CNV had to involve a minimum of three consecutive BAC clones within the X chromosome tile path. Table II summarizes the estimated size of the CNV intervals. This is based on taking the base-pair position on the X chromosome representing the first base of the BAC clone at the start of the CNV and the last base of the BAC clone at the end of the CNV. These base-pair positions are derived from the Human Genome 19 (hg19) build. As a complete X tiling path is being used with an average BAC clone size of 130–140 Kb (with some clones having smaller inserts and some overlap), this has allowed detection of intervals as small as 100 Kb. The database of genomic variants which is derived from phenotypically normal individuals from control populations (http://projects.tcag.ca/variation/?source=hg18) (Redon et al., 2006) was used to determine the locations of commonly occurring X chromosome polymorphisms that are not obviously associated with any phenotype. These are the same control populations used by Aboura et al. (2009) to determine which CNVs were significantly associated with POF. This database defines structural variation as genomic alterations that involve segments of DNA that are larger than 1 Kb. Any CNV polymorphism on the X chromosome described in the database and detected by the X tiling path in the POF patient cohort was excluded from consideration. Fig. 1A and B illustrate example profiles of an amplification and deletion and Fig. 1C summarizes the CNV found among the cohort of 42 patients after removing those documented by Redon et al. (2006). The majority of these CNV were found on Xq and in some cases more than one CNV was detected in an individual. These findings highlight new interstitial deletion and amplification intervals not previously associated with POF on the X chromosome. The log2 ratio values <1 or >−1 may reflect mosaicism for the X chromosome carrying the CNV.

Table II.

Genes covered by the CNV intervals as determined by array CGH using BAC clones.

Patient ID	CNV	Genomic locus	hg19 left breakpoint/ bp	hg19 right breakpoint/ bp	Size of del/amp bp	Genes covered by BAC clones assessed to be involved in CNV
135	Gain	Xp22.31–Xp22.31	7446551	8300790	854239	VCX, PNPLA4, MIRN651, VCX2
163	Gain	Xp22.31–Xp22.31	6274144	8330326	2056182	VCX3, LOC392425, HDHD1A, STS, VCX, PNPLA4, MIRN651, VCX2
77	Loss	Xp22.11–Xp21.3	23730647	24956038	1225391	ACOT9(? in region), SAT1(? in region), APOO, CXorf58, EIF2S3, ZFX, PDK3, PCYT1B, POLA1
134	Gain	Xp11.4–Xp11.4	38273879	38918655	644776	OCT (just), TSPAN7, MID1IP1
142	Loss	Xp11.4–Xp11.4	38949164	41032683	2083519	BCOR, IMPDH1P2, ATP6AP2, CXorf38, MED14, MKRNP5, DPRXP6, USP9X
152	Loss	Xp11.4–Xp11.4	38949164	41032683	2083519	BCOR, IMPDH1P2, ATP6AP2, CXorf38, MED14, MKRNP5, DPRXP6, USP9X
239	Loss	Xp11.23–Xp11.22	48288148	49575768	1287620	Gene rich incl HDAC6 and USP27X
213	Gain	Xq11.23–Xp11.22	62959494	64089712	1130218	ARHGEF9, FAM123B, ASB12, MTMR8
218	Gain	Xq12–Xq12	64258047	64988641	730594	ZC3H12B, LAS1L
1294	Gain	Xq13.3–Xq21.33	72662706	99004256	26341550	SATL, FAM121A, ZNF711, POF1B
152	Loss	Xq22.1–Xq22.2	99499401	103736785	4237384	includes PCDH19—ESX1 incl CENP1- and H2BFM, H2BFWT
1215	Gain	Xq22.2–Xq22.2	103043842	103379972	336130	H2BFM, H2BFWT, MCART6, CXorf39, TYBN
142	Loss	Xq22.1–Xq22.2	99667366	103670508	4003142	Just excludes PCDH19, TNMD—ESX1 many genes incl CENPI and H2BFM, H2BFWT
147	Loss	Xq22.2–Xq22.2	103188399	103610253	421854	AL03448.16, H2BFM, H2BFWT, MCART, Z82254.1, CXorf39, ESX1
135	Loss	Xq22.2–Xq22.2	103230288	103328709	98421	H2BFM, H2BFWT
87	Loss	Xq23–Xq24	116332256	116953378	621122	No genes except RP3-378P9.2-001, RP3-378P9.1-001
142	Loss	Xq25–Xq25	128469287	129577503	1108216	SMARCA1, OCRL, AL022162.1, ALPN, XPNPEP2, SASH3, ZDHHC9, UTP14A, BCORL1, ELF4, AIFM1, RAB33A, ZNF280C, SLC25A14, GPR119, RBMX2
152	Loss	Xq25–Xq25	128469287	129577503	1108216	SMARCA1, OCRL, AL022162.1, ALPN, XPNPEP2, SASH3, ZDHHC9, UTP14A, BCORL1, ELF4, AIFM1, RAB33A, ZNF280C, SLC25A14, GPR119, RBMX2
219	Loss	Xq25–Xq25	128684218	129345805	661587	OCRL, AL022162.1, ALPN, XPNPEP2, SASH3, ZDHHC9, UTP14A, BCORL1, ELF4, AIFM1, RAB33A, ZNF280C
106	Loss	Xq25–Xq25	128692536	129577503	884967	OCRL, AL022162.1, ALPN, XPNPEP2, SASH3, ZDHHC9, UTP14A, BCORL1, ELF4, AIFM1, RAB33A, ZNF280C, SLC25A14, GPR119, RBMX2
160	Loss	Xq25–Xq25	128818767	129775968	957201	XPNPEP2, SASH3, ZDHHC9, UTP14A, BCORL1, ELF4, AIFM1, RAB33A, ZNF280C, SLC25A14, GPR119, RBMX2, FAM45B, ENDX2
208	Gain	Xq26.3–Xq26.3	134683316	135047206	363890	CT45A1, CT45A3, CT45A4, SAGE1

PCR primers developed for genes in bold italics have confirmed gains or losses by qPCR.

Figure 1.

Summary of the copy number variations (CNV) detected among 42 POF patients. Genomic DNA from peripheral blood lymphocytes from POF patents was hybridized to complete X and Y chromosomes and 1 Mb autosomal BAC tiling path arrays using a normal female DNA sample as a common reference. (A) and (B) present example X chromosome profiles for a CNV amplification (A) and a deletion (B). To be scored as a CNV, a change had to include three or more consecutive BAC clones. The start and end clones are indicated. The inset provides an enlarged view of the clones involved in the CNV. (C) provides a summary of the location on the X chromosome of CNV detected among this cohort of patients after removal of common CNV variants found in the database of genomic variants.

Confirmation of CNV by qPCR

In order to verify the CNV shown in Fig. 1, qPCR was used as an alternative means of measuring copy number changes. Selected qPCR primers were developed for reporter genes deemed to lie in deleted or amplified regions. Fig. 2A illustrates the qPCR analysis for the BCORL1 gene mapping in Xq25 and demonstrates its hemizygous deletion in a series of patients, when normalized to control gene HPRT1. Fig. 2B illustrates the qPCR analysis for four genes (ZFX, TSPAN7, VCX and H2BFM) on four selected patients, relative to HPRT1 showing both amplifications (134, 163) and deletions (77, 147). Table II summarizes the qPCR analysis for each region, where the loci used for the analysis are marked in bold. The majority of the changes detected by aCGH were confirmed by the qPCR analysis. Of the 15 patients in whom qPCR was used to confirm the CNV, only patient 142 suffers from primary amenorrhoea.

Figure 2.

qPCR confirmation of the copy number variations. qPCR analysis was performed on genomic DNA using markers from each CNV interval identified among the cohort of patients. PCR primers were developed for at least one marker locus per interval. (A) illustrates the analysis using primers designed for the BCORL1 genes across a series of patients including a 45, X individual and a normal female normalized to the control gene HPRT1. (B) illustrates the analysis using primers for four genes (ZFX, TSPAN7, VCX and H2BFM) on four selected patients relative to HPRT1 and highlights both sequence amplification (134, 163) and deletion (77, 147) of patients (solid shading) compared with the normal female control (hatched shading).

Expression pattern in human tissue RNA of selected candidate genes

RT–PCR was used to establish the expression pattern of 12 selected candidate genes in a range of RNA preparations from 16 different tissues. Transcripts for genes H2BFM and H2BFWT could not be detected in any of the RNA samples used in this screen, but these genes are reported as transcribed specifically in the testis and we have confirmed this finding (data not shown). However, these genes may be expressed in primordial follicles at an earlier stage of development and therefore may not be detectable in adult tissues or ovary. The remaining genes tested are all expressed in the ovary and are transcribed in a wide range of tissues. Fig. 3 shows the analysis for the genes XPNPEP2 and ZFX and Table III summarizes the data set.

Figure 3.

RT–PCR analysis of selected genes in tissue RNA from a range of 17 human tissues. The figure illustrates the expression pattern of the XPNPEP2 and ZFX genes as determined by semi-quantitative RT–PCR. A, skeletal muscle; B, thyroid; C, spleen; D, heart-fetal; E, adipose tissue; F, brain; G, colon; H, thymus; I, heart-adult; J, Esophagus; K, placenta; L, small intestine; M, bladder; N, lung; O, trachea; P, ovary; Q, blank control.

Table III.

Summary of RT–PCR analysis for genes found in CNV intervals.

Genes
cDNA	AIFM	BCORLI	XPNPEP2	ZFX	H2BFM	H2BFWT	RBMX2	USP9X	USP27X	UTP14A	CENPI
Skeletal muscle	++	−	−	+	−	−	++	++	−	++	+
Thyroid	++	++	++	+	−	−	++	++	++	++	+++
Spleen	++	++	−	+	−	−	++	++	++	++	++
Heart	++	++	+++	+	−	−	++	++	++	++	+++
Adipose	++	+	+++	+	−	−	++	++	++	++	+++
Brain	++	++	−	+	−	−	++	++	++	++	+++
Colon	++	++	+++	+	−	−	++	++	++	++	+++
Thymus	++	++	++	+	−	−	++	++	++	++	+++
Heart-adult	++	+	++	+	−	−	++	++	++	++	++
Esophagus	++	++	++	+	−	−	++	++	++	++	+++
Placenta	++	+	−	+	−	−	++	++	++	++	+++
Small intestine	++	++	+++	++	−	−	++	++	++	++	+++
Bladder	++	+	++	+	−	−	++	++	++	++	+++
Lung	++	++	++	++	−	−	++	++	++	++	++
Trachea	++	+	+++	+	−	−	++	++	++	++	+++
Ovary	++	++	++	++	−	−	++	++	++	++	++

The number of + indicates relative expression level. Genes H2BFM and H2BFWT have been reported to be expressed in the testis. We have confirmed this (data not shown) and thus they appear to be testis specific.

Discussion

This is the first large study using a complete X tiling path array to focus CGH analysis of the X chromosome found in POF patients. The new data reported in this study reveal several discrete X chromosome intervals not previously associated with POF and therefore implicate new clusters of X-linked candidate genes. Autosomal CNVs were not detected in the present study and this probably reflects the lower resolving power of the 1 Mb autosomal array. With a 0.7 Mb whole genome array, Aboura et al. (2009) were able to detect eight significant CNVs, seven on autosomes and one on the X (in Xq28, distal to all the CNV detected by our complete X tiling path after we have removed all the common CNV polymorphisms).

Previous publications have suggested that translocations and structural rearrangements involving chromosome X (X:autosome translocations, large terminal deletions and chromosomal rearrangements) may cause POF by disrupting normal pairing in meiosis as a consequence of alterations to chromatin structure (Schlessinger et al., 2002). It has also been suggested that these structural abnormalities may exert an epigenetic effect influencing the expression of X-linked or autosomal ovary-expressed genes. For example, Rizzolio et al. (2007) have noted that most breakpoints in X-autosome translocations in the POF2 interval fall outside gene coding regions. They have suggested that the observed effect on expression of ovary and ooctye autosomal and X-linked genes flanking the translocation breakpoints may arise as a consequence of long range effects on promoter activity. More recently, the same group has shown that heterochromatin rearrangements of Xq12–q21 down-regulate oocyte-expressed genes during oocyte and follicle maturation, further suggesting that epigenetic mechanisms may contribute to X-linked POF (Rizzolio et al., 2009). If disruption of pairing is the causative agent, this implies that individual X-linked genes are not necessarily implicated in specific genetic pathways leading to the POF phenotype. Where structural change has altered the expression of key genes, these will not necessarily be included in regions of deletion or amplification or be disrupted by rearrangement and thus may be difficult to identify unequivocally. As discussed below, the higher resolution approach of aCGH has revealed much more subtle X-linked CNVs associated with POF, implying the direct involvement of genes mapping to the X.

The analysis of the 42 cytogenetically normal idiopathic cases of POF reported in this study has failed to reveal any gross abnormalities of the X chromosome or autosomes. On the contrary, the study has identified a number of new discrete X chromosome deletions and amplifications (confirmed by testing patients with at least one locus per interval using qPCR) of more limited size. The complete X chromosome tiling path has reported a high frequency of 15 novel CNVs in this POF patient cohort (20/42 patients, 48%). Despite removal of all CNV polymorphisms found in the database of genomic variants present in normal individuals, it is possible (but unlikely) that some of these new CNVs represent as yet undiscovered neutral variants. However, this higher frequency of potential POF loci on the X is consistent with selection during sex chromosome evolution leading to the accumulation of reproductive loci on the X and Y chromosomes. Some of these lie between the POF 1 and POF 2 intervals that were described previously by the work of others and some in proximal Xp and Xq regions (see Fig. 1 delineating the broad locations of the intervals). These findings are important as they provide an association between the POF phenotype and discrete perturbations of the X chromosomes that are unlikely to impair pairing or have epigenetic effects on the expression of surrounding genes as has been hypothesized to be the case for translocation patients. These intervals implicate new clusters of plausible X-linked candidate genes and sequences (see Table I) potentially implicated in genetic pathways leading to POF; this could include those genes where expression (but not dosage) can be altered by chromosome rearrangement. Table IV highlights those that are the most plausible candidates for further consideration and Table III summarizes the expression patterns of some of these genes.

Table IV.

Summary of candidate genes that may potentially be implicated in the POF phenotype.

Gene name	Gene locus	Gene function	X-inactivation status/Y homologue	Potential involvement in POF
XNPEP2	Xq25	N-terminal imido bond-specific hydrolase. Partial escape from X-inactivation	E	Breakpoints disrupting gene have been reported in a number of POF patients (Mumm et al., 2001; Rizzolio et al., 2007)
UTP14A	Xq25	Encodes protein (UTP14) involved in18S rRNA synthesis	E	Ubiquitously expressed in many tissues including ovary (Rohozinski et al., 2006)
CENPI	Xq22.1	Involved in the gonadal tissue response to FSH and assembly of the kinetichore	NE	Has been shown to have critical role in chromosome segregation (Okada et al., 2006; Cheeseman et al., 2008), with deletions potentially leading to cell death. Involved in the gonadal tissue response to FSH. This gene is also a potential candidate for human X-linked disorders of gonadal development and gametogenesis
PCDH19	Xq13.3	Cell adhesion		None currently deduced other than location in POF2 region
VCX	Xp22	Encodes small, highly charged (putatively nuclear) protein of unknown function	NE, Y	Expressed in male germ cells only (Lahn et al., 2000). Duplications are observed in 2 POF patients and not observed on X chromosomes of any other patients
STS	Xp22.32	Catalyses the conversion of sulphated steroid precursors to estrogens during pregnancy. Mutations are known to cause XLI. Escapes X-inactivation	E, Y	First reported duplication of this gene. Over-production of enzyme may lead to over-production of estrogen with potentially suppressive effects on LH and FSH production, potentially affecting oocyte production or follicle maturation
ZFX	Xp21.3	Encodes a zinc-finger containing protein and functions as transcription factor [46]. Escapes X-inactivation	E, Y	Mice with null-mutations at Zfx locus have fewer oocytes and lower reproductive life-span (Page et al., 1987; Mumm et al., 2001; Luoh et al., 1997). ZFX is located within the less common POF3 region
BCORL1	Xq25-q26.1	Repressor activity through an association with histone deacetylase (HDAC) [40]	I, Y	Deletion of BCORL1 may potentially lead to insufficient repression of apoptosis resulting in atresia of ovarian follicles
USP9X	Xp11.4	Ubiquitin specific protease with ubiquitous expression pattern. Escapes X-inactivation [31, 32]	E, Y	Maps to partial X deletion region associated with Turner syndrome (Jones et al., 1996) and has shown suggestive association with POF (Knauff et al., 2009) Expressed in a stage and cell-specfic manner in murine oogenesis (Noma et al., 2002) and has a role in chromosome segregation (Sargent et al., 1999)
TSPAN7		Member of the tetraspanin family of cell-surface glycoproteins that mediate signal transduction events involved in regulation of cell development, activation, growth and motility. This gene may have a role in control of neurite outgrowth and complexes with integrins. Associated with X-linked mental retardation and neuropsychiatric diseases	NE	Expressed in the ovary
POF1B	Xq21	Unknown function but binds to actin and has some homology to myosin heavy chain. Escapes X-inactivation	E	Expressed in ovary during early embryonic development (Bione et al., 2004). Interrupted by X:1 translocation in POF patient (Riva et al., 1996). Mutated in POF patient (Lacombe et al., 2006) where mutation affects actin binding. May play a role in meiotic chromosome pairing and apoptosis
AIFM1		Apoptosis gene—degeneration of terminally differentiated cerebellar and retinal neurons	NE	Flavoprotein essential for nuclear disassembly in apoptotic cells found in the mitochondrial intermembrane space (Erhart et al., 2005) Effects chromosome condensation and fragmentation in the nucleus and releases the apoptogenic proteins cytochrome c and caspase-9. May be involved in accelerated oocyte atresia

E, gene known to escape partially or completely from X-inactivation; I, gene known to undergo X-inactivation; Y, gene has Y homologue; NE, No evidence of X-inactivation status. All these genes are expressed in the ovary.

It is interesting to note that a number of amplifications have been identified in some of the POF patients. This association of X chromosome amplification with the phenotype is of interest in the light of early reports associating triple X syndrome with POF (Villanueva and Rebar, 1983). It is also of note that three of the loci (POF1B, STS and PNPLA4) covered by amplification are known to escape X-inactivation. Thus, any amplification of these loci would increase the number of active copies. It will be important to verify the X-inactivation status of other genes included in the amplifications.

A collation of promising candidate genes encompassed by the CNV reported in this study is presented in Table IV with comments on their potential function(s). Several translocations disrupting the XNPEP2 have been reported in POF patients (Prueitt et al., 2000), but it is not clear how this gene is relevant to ovarian function. In terms of the potential function(s) of other genes, it is clear that a subgroup may impact on some or all of three cellular processes that are likely to be important for normal ovarian function.

First, a number of these genes influence chromosome pairing and segregation. POF1B (expressed in the ovary during early embryonic development) may play a role in meiotic chromosome pairing and apoptosis and has been considered by others previously on the basis of evidence that the gene is disrupted in POF patients (Bione et al., 2004). Mutations in POF1B have been shown to affect actin binding (Lacombe et al., 2006) and this may be important for successful passage through meiosis. CENPI has been shown to have a critical role in chromosome segregation as it is involved in directing kinetochore assembly (Okada et al., 2006; Cheeseman et al., 2008), with deletions potentially leading to cell death. The observation of a CNV involving the USP9X gene is intriguing. We had originally identified and described the USP9X and USP9Y genes in mouse and human and the involvement of USP9Y in deletions of the human Y chromosome covering the AZFa male infertility deletion interval (Jones et al., 1996; Brown et al., 1998; Sargent et al., 1999); thus USP9X may also play a role in ovarian function. This is further strengthened by the inclusion of USP9X in partial X deletions associated with the gonadal aspects of the Turner phenotype (Boucher et al., 2001). It is interesting that patient 142 has loss of this region and suffers from primary amenorrhoea. More recent studies have demonstrated that the gene is expressed in a stage- and cell-specific manner during murine oogenesis (Noma et al., 2002) and that it has an important role in regulating chromosome alignment and segregation (Vong et al., 2005). Both USP9X and ZFX escape X-inactivation suggesting that (in common with Turner syndrome), it may be a hallmark of genes associated with POF that expression is required in diploid dose.

Second, some of these genes can influence hormone levels and tissue responses to hormones that may impact on efficient oocyte development and maturation. STS plays a central role in the synthesis of steroid hormones. Amplification of the gene with subsequent excess of the enzyme may lead to over-production of estrogen with potentially suppressive effects on LH and FSH production, thus potentially affecting oocyte production or follicle maturation. In addition to a role in centromere formation, CENPI is involved in gonadal tissue response to FSH (Slegtenhorst-Eegdeman et al., 1995; Roberts et al., 1996) and is thus a potential candidate for human X-linked disorders of gonadal development and gametogenesis.

Third, two of the genes in Table IV are involved in apoptotic responses. AIFM1 is a flavoprotein found in the mitochondrial intermembrane space and is essential for nuclear disassembly in apoptotic cells (Erhart et al., 2005). The protein effects chromosome condensation and fragmentation in the nucleus and releases the apoptogenic proteins cytochrome c and caspase-9. The second gene, BCORL1, is a corepressor of BCL6 (Pagan et al., 2007) which is known to repress the expression of BCL2 an anti-apoptotic gene (Saito et al., 2009). Thus perturbations of both these genes may contribute to insufficient repression of apoptosis resulting in atresia of ovarian follicles.

As a consequence of the greater resolution that aCGH can provide, the data in this paper suggest that there are multiple discrete intervals on the human X chromosome that are associated with POF. Furthermore, both amplification and deletion are associated with the phenotype and the regions covered by these CNVs contain several candidate genes with appropriate functions. These data provide the basis for more detailed investigations of the contribution of individual genes to the genetic aetiology of POF.