A Single-Nucleotide Polymorphism in the EAP1 Gene Is Associated with Amenorrhea/Oligomenorrhea in Nonhuman Primates

Abstract

Current evidence suggests that the acquisition of female reproductive capacity and the maintenance of mature reproductive function are related processes transcriptionally regulated by gene networks operating within the neuroendocrine brain. One of these genes, termed enhanced at puberty 1 (EAP1), encodes an upstream regulator of these processes. Selective inhibition of EAP1 expression in discrete regions of the rat and nonhuman primate (NHP) hypothalamus, via targeted delivery of RNA interference, either disrupts (rats) or abolishes (monkeys) reproductive cycles. The striking loss of menstrual cyclicity resulting from knocking down hypothalamic EAP1 expression suggests that diminished EAP1 function may contribute to disorders of the menstrual cycle of neuroendocrine origin. Here we show that a single-nucleotide polymorphism in the 5′-flanking region of EAP1 gene is associated with increased incidence of amenorrhea/oligomenorrhea in NHP. In the presence of the risk allele, binding of the transcription factor mothers against decapentaplegic homolog 3 (SMAD3) to its recognition site contained within the polymorphic sequence in the monkey EAP1 promoter is reduced. The risk allele also diminishes the increase in EAP1 promoter activity elicited by TGFβ1, a peptide that activates a SMAD3/4-mediated signaling pathway to regulate gene transcription. These findings indicate that common genetic variation in the EAP1 locus increases the susceptibility of NHP to loss/disruption of menstrual cyclicity. They also raise the possibility that polymorphisms in EAP1 may increase the risk of functional hypothalamic amenorrhea in humans.

Using a combination of DNA arrays and small interfering RNA (siRNA) technology, we recently identified enhanced at puberty 1 (EAP1), a gene of heretofore unknown function, as a novel component of the transcriptional machinery that regulates female reproductive function (1). We found that EAP1 transactivates genes involved in facilitating gonadotropin output (e.g. GnRH) but suppresses the expression of genes inhibitory to GnRH neuronal function (such as the preproenkephalin gene). Because of this dual activity, a function of EAP1 in the neuroendocrine brain might be to coordinate the activity of cellular networks controlling GnRH secretion. Our results also showed that siRNA-mediated reduction of Eap1 expression in the preoptic area of female rats disrupts estrous cyclicity and reduces plasma gonadotropin levels, suggesting that the loss of Eap1 in a specific region of the neuroendocrine brain leads to altered GnRH release (1). In the companion paper in this issue of Endocrinology (2), we extend these studies to nonhuman primates (NHP) and show that knocking down EAP1 mRNA in the medial basal hypothalamus of the rhesus monkey disrupts menstrual cyclicity.

These observations strongly suggest that EAP1 contributes to controlling the rhythmic manifestation of the estrous cycle in rodents and the menstrual cycle of NHP. They also raise the possibility that a diminished EAP1 function may contribute to disorders of the menstrual cycle of neuroendocrine origin. Functional hypothalamic amenorrhea (FHA) is one such disorder; FHA is characterized by the absence of menstrual cycles, low estrogen levels, and low or normal levels of gonadotropins (3, 4). FHA affects 3% of women of reproductive age and accounts for up to 30% of all cases of amenorrhea (4, 5). The factors responsible for the disease are largely unknown, but the contribution of a genetic component to its etiology in at least a fraction of affected individuals is now clear (6).

Because FHA affects women of reproductive age and does not manifest itself as a permanent condition, it would not appear that it is primarily caused by incapacitating gene mutations. In fact, all mutations recently described to be associated with hypothalamic amenorrhea are heterozygous (6). It is likely that FHA results from interactions between environmental factors and predisposing genetic factors. Common DNA sequence variation, including single-nucleotide polymorphisms (SNP), copy number variations, block substitutions, insertions/deletions, and inversions appear to be uniquely poised to provide susceptibility to disease. Among the different forms of common DNA sequence variation, SNP have proven instrumental in identifying quantitative trait loci in humans that otherwise remained elusive. A relevant example is provided by recent genome-wide association studies demonstrating an association of a sequence variation in LIN28B (a gene encoding an RNA-binding protein) with early menarche (7–10). These studies also found a significant association of sequence polymorphisms in more than 30 other genes with the age at menarche (8, 9, 11). Additional examples include the association of SNP with coronary artery disease, rheumatoid arthritis, type 1 and type 2 diabetes (12), and the association of SNP in the CFH (complement factor H) (13–15) and the HTRA1 (high-temperature requirement A) loci (16, 17) with macular degeneration.

Whether genetic variations in EAP1 sequence affects the functional capabilities of the EAP1 gene (18, 19) and the involvement of EAP1 in regulating menstrual cyclicity is not known. Here we show that a SNP located in the 5′-flanking region of the EAP1 gene is associated with amenorrhea/oligomenorrhea (A/O) in NHP. We also show that this SNP compromises the ability of mothers against decapentaplegic homolog 3/4 (SMAD3/4), a transcriptional activator complex, to interact with the EAP1 promoter and that it significantly reduces the ability of TGFβ1, which acts via SMAD3/4, to activate EAP1 transcription.

Materials and Methods

Animals

In the wild and in captive animals housed outdoors, menstrual cyclicity of rhesus macaques is either erratic or absent during the summer. Animals that are housed indoors, however, may display menstrual cyclicity throughout the year. This is probably because indoor animals do not have the visual cues to indicate a season and therefore continue to cycle. These are the animals we targeted for this study. Animal use was duly approved by the Institutional Animal Care and Use Committee of the Oregon National Primate Research Center (ONPRC).

At the ONPRC, observational menses data are recorded for historical record keeping for every female housed indoors. For this study, menses data were obtained from January 1, 2004, through September 1, 2007, for females that were 5–19 yr old at the time of this query (October 25, 2007). Females were selected from a cohort that had data for the entire time frame, and the average number of cycles per month was calculated based on a 45-month time frame. Because moving animals from one location (individual cage, group housing) to another can affect the incidence and/or regularity of menses, the average number of moves was calculated, and females were chosen for both groups that had less than an average number of moves. Females with an average of one cycle per month more than 90% of the time frame studied (45 months), and had an average of fewer than 3.4 moves, were considered to be normal cyclers (NC) and therefore were included in the control cohort (NC group; n = 30). Females with an average of one cycle per month less than 60% of the time and had an average of fewer than 3.4 moves were considered to be amenorrheic or oligomenorrheic (A/O group; n = 24). All females either had zero or one birth during the time frame analyzed. We also added group-housed animals (n = 11) to the case cohort because there were insufficient qualifying animals residing in caged housing. Because menses data are not gathered for females in group housing, we assumed that a group-housed breeding-age female (10–14 yr of age and co-housed with adult male animals) was A/O if she had zero or one infant born during the 45-month period analyzed.

DNA samples

Three milliliters of blood were drawn from each animal into an EDTA Vacutainer (BD Biosciences, Franklin Lakes, NJ), following standard procedures implemented by the Division of Animal Resources of the ONPRC. DNA was extracted from the white blood cells using the Magnisil DNA Isolation Kit (Promega, Madison WI), which yields 100–200 μg DNA per sample.

Sequencing

We sequenced the rhesus monkey EAP1 gene (NCBI accession no. Q2MJS2) including 2 kb of the 5′-flanking region (that contains the promoter), the 5′- and 3′-untranslated regions, and all 2388 bp of the coding sequence. Because EAP1 is intronless, we surveyed its entire genomic sequence by generating 12 overlapping PCR products. The primer sets used (Eurofins MWG Operon, Huntsville, AL) are shown in Table 1. The PCR amplification conditions included a denaturing step at 95 C for 5 min, followed by 31 cycles of denaturing (95 C for 30 sec), annealing (at the temperatures shown in Table 1) for 1 min, and extension (72 C for 1 min). The PCR products were sequenced on both strands by Agencourt Biosciences (Beverly, MA), using the amplification primers for each product. The sequence trace files and a preliminary list of potential variants were provided.

Table 1.

Primers used for SNP analysis

Amplicon no.	Amplicon start	Amplicon end	Forward primera	Reverse primera	Annealing temperature (°C)
1	−1914	−1222	acaaattgaggctgggtgag	caaggcggtgatttcttttg	58
2	−1448	−749	ctatctgctggccagactcc	gccaaacaggtgagtggttt	58
3	−407	+265	atacgaaattcacggcgaaa	cacgagagcagaggagagga	58
4	+62	+711	gcctccggaacgtagagtaa	ctcttcgtccccacttccta	55
5	+621	+1321	gagcttccaacccaagtctg	aacgtggttgagctgttgct	55
6	+1169	+1884	cgtcggggtcaagacagt	gacgaagaggtcgaagacga	55
7	+1665	+2331	gaggtccccagcttactgtg	gagccgtgcttcttttcgta	58
8	+2252	+2968	taccaggactgcatgaagga	attcaggtccccattacgtg	55
9	+2740	+3427	acctcattctaaccggacca	ttctacaccttgggggtgag	55
10	+3235	+3943	gaatgtaccttgggccttca	ccctcacaaacttgcaacaa	56
11	+3814	+4525	aactgccccaagtgttgaag	tgtcaagtcaccagctgcat	56
12	+4326	+5025	ccctcggtaccaggctattt	atttgccttcccctaccact	58

^aPrimers amplifying regions of EAP1 mRNA are shown in bold. Other primers are those amplifying the 5′- and 3′-flanking regions.

SNP analysis

A secondary assessment of DNA variants was generated in-house using Polyphred software (http://depts.washington.edu/uwc4c/express-licenses/assets/polyphred/) (20) and by visual validation of the identified genotypes. Custom Perl scripts were used to extract the verified genotype data and convert it into the format required for statistical comparisons within and between defined case (A/O)/normal control (NC) groups. A total of 18 SNP markers genotyped in 54 animals were identified (from 30 controls and 24 cases). Failed SNP (genotyping rates <50%), SNP with minor allele frequency (MAF) of 1% or lower, and those out of Hardy-Weinberg equilibrium were excluded from the analysis. Four variants were resequenced, and analysis was done on a subset of the previous animals (29 controls and 22 cases). Therefore, a total of 12 SNP were analyzed for association with A/O. Sequences containing these newly detected SNP have been placed into the NCBI dbSNP (http://www.ncbi.nlm.nih.gov/SNP/).

Association between each SNP and A/O was tested using a likelihood-ratio statistic to compare the allele frequencies under a null hypothesis that the SNP is in linkage equilibrium with the trait and the alternative hypothesis that the SNP is in some level of linkage disequilibrium with the putative disease locus (21). A genetic model-free analysis using an estimated prevalence of A/O of 10% was used (21).

Functional promoter assays

Because SNP HA2494, which was found to be associated with A/O, is located within a consensus binding site for SMAD3, we cloned a 1491-bp fragment from the 5′-flanking region of the rhesus monkey EAP1 gene comprising nucleotides −1262 to +229 [transcription start site (TSS) = +1], and including SNP HA2494 allele G, into the luciferase reporter plasmid pGL-2 Basic (Promega). To generate a reporter plasmid carrying allele C, we performed site-directed mutagenesis using the Change-IT multiple-mutation site-directed mutagenesis kit (USB Corp., Cleveland, OH) and the mutagenesis primer 5′-CCTGGGCGACAGAGCCAGACTCCGTCTCAAA-3′, following the manufacturer's instructions.

To determine the effect of SNP HA2494 on EAP1 promoter activity, both versions of the promoter (EAP1-C or EAP1-G) were examined for transcriptional activity in NIH3T3 cells. The cells (180,000 cells per well) were seeded onto 12-well plates in DMEM containing10% donor calf serum. Twenty-four hours later, each of the two reporter plasmids (EAP1-C pGL2 or EAP1-G pGL2) were transiently transfected for 5 h using Lipofectamine LTX and Plus reagent (Invitrogen, Carlsbad, CA) at a ratio of 1 μg DNA to 1 μl Plus reagent to 1.5 μl Lipofectamine LTX in DMEM with 3% donor calf serum. The next day, the cells were exposed to 2 ng/ml TGFβ1 (Cell Signaling, Danvers, MA) in DMEM with 0.3% donor calf serum. Twenty-four hours later, they were harvested and assayed for luciferase activity using the Luc-Screen system (ABI, Bedford, MA). Transfection efficiency was normalized by cotransfecting the plasmid CMV-Sport-β-gal (Invitrogen) at 100 ng/ml and determining β-galactosidase activity using the Tropix Galacto reaction (Applied Biosystems, Foster City, CA) as reported earlier (1).

Electrophoretic mobility shift assays

Double-stranded oligodeoxynucleotides were generated by annealing the primer 5′-GCGACAGAGCCAGACTCCGTCTCA-3′ with its reversed complement primer (for oligo C) and the primer 5′-GCGACAGAGCGAGACTCCGTCTCA-3′ with its reversed complement primer (for oligo G; the nucleotide corresponding to SNP HA2494 is shown in bold). Annealing was performed in saline-sodium citrate buffer [3 mm sodium citrate (pH 7), 30 mm NaCl] (Sigma Chemical Co., St. Louis, MO) at 95 C for 5 min and then cooled slowly to room temperature for several hours. Absorbance was determined at 260 nm using the Epoch multi-volume spectrophotometer system (BioTek, Winooski, VT).

The double-stranded 24-mer DNA oligos were diluted in binding buffer (10 mm Tris-HCl, 150 mm KCl, 0.1 mm dithiothreitol, and 0.1 mm EDTA) to a final amount of 50 ng per tube in a 10-μl reaction. Binding assay was performed using increasing amounts of recombinant human SMAD3 (Prospec, Ness Ziona, Israel) for 30 min at room temperature. Reaction was terminated by adding 2 μl of 6× EMSA gel-loading solution (Molecular Probes EMSA kit; Invitrogen) and loaded onto an 8% nondenaturing polyacrylamide mini-gel (Invitrogen). The gel was run at 150 V for 60 min in prechilled 0.5× TBE [44.5 mm Tris base, 44.5 mm boric acid, 1 mm EDTA (pH 8.5)] buffer. The gel was then stained with SYBR Green EMSA stain (Invitrogen) at 1:10,000 dilution in 0.5× TBE buffer for 20 min at room temperature, and washed twice with deionized H₂O for 10 sec to remove excess stain. The stained gel was visualized using a GelDoc XR+ (Bio-Rad, Hercules, CA) instrument. The OD of each band was determined using AlphaEaseFC software (Alpha Innotech, Santa Clara, CA).

Statistics

Association between each SNP and A/O was analyzed as described above. The differences in luciferase activity observed in promoter assays were analyzed by the Student's t test when comparing two groups or one-way ANOVA followed by the Student-Newman-Keuls multiple-comparison test for unequal replications when comparing several groups. Data are expressed as mean ± sem. The null hypothesis was rejected at the 0.05 level for all analyses.

Results

A subpopulation of female NHP fail to cycle regularly for extended periods of time

A common observation at National Primate Research Centers is that some females of reproductive age either fail to display regular menstrual cycles or do not show menstrual cycles at all for extended periods of time. Because amenorrhea and/or oligomenorrhea severely compromise fertility, this disorder significantly impacts breeding and research programs requiring normally cycling females. The NC control group of animals used in this study had nearly 12 cycles per year during the 4-yr period studied, whereas the A/O case group had half as many cycles per year (P = 1.77 × 10⁻⁹) (Fig. 1).

Fig. 1.

Menstrual cyclicity in two cohorts of female rhesus monkeys individually housed at the ONPRC. Each bar depicts the mean number of cycles per year. Vertical lines are sem.

An association exists between a SNP located in the EAP1 5′-flanking region and A/O in NHP

We first analyzed the genotype data to identify SNP loci that were not adequate for statistical analysis. To do this, we determined the existence of marker heterozygosity and missing data for all SNP (Table 2). Three SNP had genotyping rates below 50% (HA552, HA4611, and HA6226). Two SNP had MAF of 1% or less (HA1156 and HA1188). Two SNP were out of Hardy-Weinberg equilibrium (HA552 and HA5410 at P = 0.0102 and P = 0.0009, respectively). These six SNP were excluded from the analysis (Table 3), leaving 12 SNP analyzed for association with A/O.

Table 2.

Marker genotype statistics

Marker	No. of genotypesa	Percent genotyped	Heterozygosityb	MAF
HA140	50	92.60	52.00	0.5
HA1156	53	98.10	1.90	0.01
HA2229	49	92.60	51.00	0.49
HA2494c	48	96.10	18.40	0.07
HA2581	40	78.40	12.50	0.06
HA3589	5	9.30	0.00	NA
HA4450	42	77.80	33.30	0.17
HA4611	25	46.30	16.00	0.08
HA552	26	48.10	11.50	0.17
HA5410	44	81.50	22.70	0.36
HA6145	32	62.70	31.10	0.23
HA6226	11	20.40	27.30	0.14
HA8111	46	85.20	50.00	0.32
HA9247	50	92.60	6.00	0.03
HA1188	45	83.30	2.20	0.01
HA11590	51	94.40	49.00	0.34
HA1278	50	92.60	50.00	0.33
HA12243	51	94.40	17.60	0.11
HA12346	50	92.60	50.00	0.35

^aNumber of genotypes for each SNP.

^bPercentage of animals heterozygous for that SNP.

^cThese results were confirmed by resequencing a subpopulation of animals (29 controls and 22 cases).

Table 3.

SNP markers excluded from analysis

Excluded marker	Reason for exclusion
	Out of HWE	<50% genotyped	MAF ≤1%
HA552	Y	Y
HA5410	Y
HA4611		Y
HA6226		Y
HA1156			Y
HA1188			Y

HWE, Hardy-Weinberg equilibrium; Y, yes.

Of the 12 SNP analyzed, a suggestive association was found between SNP HA2494 (C/G) and A/O (unadjusted P < 0.0042, Table 4). HA2494 is located 957 nucleotides upstream from the TSS of EAP1. All A/O cases are homozygous at this location, whereas 14.58% of the NC controls are heterozygous. The HA2494 is independent from the other markers tested because it is not in linkage disequilibrium with the other SNP (r² = 0–0.154, Fig. 2). These results demonstrate a significant association between the A/O phenotype and the presence of a SNP located in the monkey EAP1 promoter, with animals homozygous for the G allele showing increased risk for A/O over that of heterozygotes (Table 4).

Table 4.

Association between SNP markers and trait

Marker	Test for associationa				Fitted penetrancesb
	Reference allelee	LOD	P value	Frequencyc	A/A	A/B	B/B
HA140	4	0.06	0.87	0.5	0.111	0.089	0.111
HA2229d	4	0.05	0.89	0.486	0.122	0.094	0.091
HA2494d	3	2.38	0.0042	0.855	0.137	0	0
HA2581d	3	0.12	0.75	0.949	0.094	0.162	0
HA6145d	3	0.06	0.8	0.778	0.106	0.083	0.142
HA4450	3	0.54	0.29	0.798	0.129	0.056	0
HA8111	4	0.12	0.75	0.665	0.12	0.082	0.092
HA9247	3	0.13	0.73	0.979	0.095	0.21	0
HA11590	3	0.35	0.44	0.632	0.133	0.072	0.102
HA1278	3	0.17	0.67	0.658	0.117	0.079	0.115
HA12243	4	0.46	0.35	0.922	0.088	0.143	0.777
HA12346	2	0.26	0.55	0.622	0.131	0.077	0.092

^aLOD score and P values for the association of each SNP to the A/O trait.

^bBased on a disease prevalence of 10% and no genetic model, three penetrances are estimated: 1) one for individuals who are homozygous for this causal allele, 2) another for heterozygotes, and 3) a penetrance for individuals who carry no copies of the allele.

^cMarker allele frequencies of the reference allele in the population are estimated using maximum likelihood. The assumption of the model is that one of the marker alleles directly influences disease susceptibility.

^dThese results were confirmed by resequencing a subpopulation of animals (29 controls and 22 cases).

^eReference allele: 2 = cytosine, 3 = guanine and 4 = thymine.

Fig. 2.

Pairwise linkage disequilibrium among all SNP used in the association analysis. The correlation (r²) between each pair of SNP measures how often the two SNP are found together in the population. A low correlation indicates that the pair of SNP segregates separately in that population.

SNP HA2494 is associated with reduced EAP1 transcriptional activation in response to TGFβ1 stimulation

TRANSFAC (https://portal.biobase-international.com) analysis of the sequence encompassing SNP HA2494 revealed that the C/G variation may affect the theoretical binding affinity of several transcription factors, including nuclear factor-1, E2F transcription factor 4, POU class 2 homeobox 2, androgen receptor, glucocorticoid receptor, GATA binding protein, and SMAD3/4 (Fig. 3A). However, with the exception of SMAD3/4, all of the other factors have the theoretical capability of recognizing consensus sequences frequently present (>20 times) within 1.5 kb of the EAP1 flanking region. In contrast, SMAD3/4 is recognized by only two sequences, one located more than 1500 bp upstream from the EAP1 TSS and another located between nucleotides −961 and −952 from the TSS. This region contains the C/G SNP (located at nucleotide −957). Whereas the C/C genotype results in a sequence predicted to allow SMAD3/4 binding, the G/G genotype results in theoretical loss of this recognition site. To determine whether SNP HA2494 affects basal EAP1 transcriptional activity, we performed promoter assays by transfecting NIH 3T3 cells with a 1491-bp construct containing either the C or G allele of HA2494. These assays showed that a promoter carrying the G allele, predominant in A/O animals, had slightly lower activity than the promoter carrying the C allele, mostly found in NC monkeys (Fig. 3B). Because TGFβ regulates gene transcription by activating a SMAD3-mediated signaling pathway (22), we determined whether the substitution of G for C in this site affects the ability of TGFβ1 to trans-regulate the EAP1 promoter. EAP1-G promoter activity increased significantly (P < 0.001) less in response to TGFβ1 (2 ng/ml) than EAP1-C (Fig. 3B), suggesting that the G allele of SNP HA2494 compromises the ability of TGFβ1 to trans-activate the EAP1 promoter.

Fig. 3.

SNP HA2494 regulates EAP1 transcriptional activity by differentially modulating EAP1 promoter response to TGFβ1. Panel A, SNP HA2494 (C/G) is located 957 bp upstream from the predicted TSS of the rhesus monkey EAP1 gene. TRANSFAC analysis shows that this SNP is the first nucleotide (underlined) in the core sequence (CAGAC) for the binding of SMAD proteins. Bold letters denote the similarities between the SMAD3/4 consensus sequence and the sequence flanking SNP HA2494. Panel B, Two constructs containing the region −1262 to +229 of the EAP1 5′-flanking region, each carrying a different version of SNP HA2494, were generated to assess the effect of SNP HA2494 on EAP1 transcriptional activity. Luciferase activity is shown as percentage of the activity displayed by the pGL2 empty vector, the luciferase reporter plasmid in which the EAP1 regulatory region was cloned. Although both versions of the promoter had similar basal activities, TGFβ1-dependent trans-activation of the promoter carrying the G allele was significantly reduced compared with the promoter carrying the C allele. Bars are means (n = 12 per group), and vertical lines are sem; ***, P < 0.001 (one-way ANOVA followed by the Student-Newman-Keuls multiple-comparison test). Panel C, EMSA performed with double-stranded oligomer probes containing the putative SMAD3/4 binding core motif shown in A. Two 24-mer DNA oligomers named oligo C and oligo G were prepared, each carrying a different allele of SNP HA2494. Left panel, Delayed migration of double-stranded oligodeoxynucleotides by increasing amounts (1, 2, and 4 μg) of human recombinant SMAD3 in the presence of a constant amount (50 ng per reaction) of DNA (arrow B, protein-bound DNA; arrow F, free DNA); right panel, relative binding activities calculated as percentage of the highest concentration of SMAD3 (4 μg) bound to oligo C.

SNP HA2494 functionally compromises SMAD3 recruitment to its proximal binding site in the EAP1 5′-flanking region

We next used EMSA to determine whether the SNP HA2494 alleles alter the recruitment of SMAD3 to its proximal binding site (−961 and −952) bp in the EAP1 promoter. SMAD3 binding to oligodeoxynucleotide 5′-GCGACAGAGCCAGACTCCGTCTCA-3′ was unambiguous and dependent on the concentration of SMAD3 protein used (Fig. 3C). This oligonucleotide carries the C allele of SNP HA2494 (underlined) and contains the canonical core SMAD3/4 binding site (denoted in bold). However, SMAD3 binding to an oligodeoxynucleotide (5′-GCGACAGAGCGAGACTCCGTCTCA-3′) containing the G allele of SNP HA2494 (underlined) was reduced, and this loss was observed at all concentrations of SMAD3 used (Fig. 3C). These results indicate that SNP HA2494 compromises the recruitment of SMAD3 to its proximal binding site in the EAP1 promoter.

Discussion

The present results indicate that a SNP in the promoter region of EAP1 is associated with an increased incidence of A/O in rhesus monkeys. The SNP maps to nucleotide −957 upstream from the gene's TSS. We observed that animals with disrupted or absent menstrual cyclicity are more likely to carry the G/G allele genotype than animals with normal cycles, in which the C/G genotype was predominant. Our results also show that the G/G polymorphism is associated with diminished response of the EAP1 promoter to TGFβ1. This deficiency appears to be related to a lower SMAD3 association to a binding site located in the region of the EAP1 promoter bearing the G allele.

One key signaling pathway used by members of the TGFβ superfamily to affect gene expression is the receptor-activated translocation of phosphorylated SMAD forms to the cell nucleus (22). After TGFβ1-mediated receptor activation, SMAD3 is phosphorylated and forms a heteromeric complex with the common mediator SMAD4 (23, 24). Activated SMAD oligomers translocate to the nucleus where they bind to Smad binding elements containing the CAGAC sequence (23, 25, 26) located in the promoter region of target genes and associate with other transcription factors to regulate gene expression (22–24).

A reduced TGFβ1 effectiveness to enhance EAP1 expression may be physiologically significant for the neuroendocrine control of GnRH secretion, as TGFβ1 and other family members have been implicated in the hypothalamic regulation of GnRH release. For instance, activin stimulates GnRH release from a neuronal GnRH cell line (27) and from hypothalamic explants (28). Studies using the GT1-7 GnRH neuronal cell line, and isolated astrocytes in culture showed that TGFβ1 is released by astrocytes and stimulates GnRH release directly by activating specific receptors located on these neurons (29, 30). A significant fraction of GnRH neurons in the intact rat hypothalamus express the type I serine-threonine kinase receptors that would allow the neurons to respond to TGFβ1 or activin in vivo (31). In addition to stimulating GnRH release, TGFβ1 transiently increases GnRH mRNA levels in GT1-7 cells (29, 32). However, TGFβ1 does not stimulate GnRH release from median eminence nerve terminals disconnected from the GnRH cell bodies (33), suggesting that a significant fraction of its stimulatory effect on GnRH release is exerted on GnRH neuronal perikarya. The absence of immunoreactive TGFβ receptors I and II in GnRH axons of the median eminence, in contrast to their presence in GnRH cell bodies (34), is consistent with this concept. An involvement of astrocytic TGFβ1 in the hypothalamic mechanism underlying the preovulatory surge of gonadotropins has been suggested by the presence of estradiol receptors in hypothalamic astrocytes (30, 35) and the ability of estradiol to stimulate the production of TGFβ1 from them (30). It thus appears that TGFβ1 produced by hypothalamic astrocytes is one of the glial factors involved in facilitating GnRH release and that it does so at least in part via activation of specific receptors, presumably located on GnRH neuronal perikarya.

Mice carrying a targeted deletion of the Smad3 gene exhibit a marked reduction in reproductive capacity (36). However, their growth is also severely retarded, and the mutants develop colorectal adenocarcinomas, dying between 4–6 months of age. Conditional deletion of the Smad3 gene from specific neuronal subsets or astrocytes using Cre-loxP technology will help in defining the role of the TGFβ1-SMAD3 pathway in the hypothalamic control of reproduction.

Although a role for TGFβ1 in regulating GnRH secretion appears well established, the mechanism by which a reduced EAP1 production may affect menstrual cyclicity is unclear. Like some other unique transcriptional regulators that exhibit either trans-activating or repressive activities, depending on the target promoter and the partners involved (37, 38), EAP1 appears to have dual transcriptional activity (1). It trans-activates the GnRH promoter and represses the proenkephalin promoter in GnRH-secreting GT1-7 cells (1), but, surprisingly, it also represses the KISS1 promoter in nonneuronal 293MSR cells and GT1-7 cells (39). This latter observation needs to be considered with caution because in vitro analyses of promoter activity using cell lines may not faithfully reflect the situation in vivo. The cell lines used to examine the transcriptional activity of the KISS1 promoter may not express the same complement of transcriptional regulators expressed in kisspeptin neurons in situ. Resolution of this issue awaits development of a cell line derived from immortalized kisspeptin neurons.

Because the present results and those of studies involving RNA interference-mediated reduction in hypothalamic EAP1 expression (1, 2) show that loss of EAP1 function disrupts, instead of enhancing, reproductive cyclicity, the possibility needs to be considered that a main function of EAP1 in the reproductive hypothalamus is to repress downstream genes involved in the inhibitory control of the GnRH neuronal network. The identity of such repressors is unknown, but some of them may be genes encoding transsynaptic inhibitors, such as the proenkephalin gene (1) and, perhaps more importantly, genes involved in transcriptional silencing, such as the Polycomb silencing complex, which has been recently implicated in the epigenetic repression of female puberty and reproductive cyclicity (40). Supporting the idea that EAP1 mostly behaves as a transcriptional repressor comes from the recent demonstration (41) that EAP1 is a key component of a repressive complex that includes two other related proteins, interfering regulatory factor-2 binding protein 2 (IRF2BP2, also known as DIF-1) and IRF2BP1. In cancer cells, the complex is repressed by an upstream regulator known as nuclear receptor interacting factor 3 and targets a downstream proapoptotic gene for repression, resulting in diminished cell death (41). Although these observations were made in breast cancer cells, we also observed that all components of this repressive complex, including the upstream repressor nuclear receptor interacting factor 3, are coexpressed in hypothalamic neurons including the preoptic area and arcuate nucleus (41). It is thus likely that the DIF-1/IRF2BP1/EAP1 complex is physiologically involved in controlling specific neuroendocrine functions such as those underlying the control of reproductive function. Additional studies are needed to assess the validity of this hypothesis.

The present findings considered in conjunction with the striking loss of reproductive cyclicity resulting from the siRNA-mediated knockdown of EAP1 in the hypothalamus of rhesus monkeys (2) suggest that EAP1 plays an important role in controlling the rhythmic manifestation of the menstrual cycle in higher primates, including humans. They also raise the possibility that a defect in hypothalamic EAP1 function may contribute to disorders of the menstrual cycle of hypothalamic origin. FHA is one such disorder. It is primarily due to alterations in the central neuroendocrine mechanism controlling pulsatile GnRH secretion (3, 4). As in EAP1 knockdown monkeys, the hypothalamus in FHA fails to generate a normal pattern of LH secretion and to respond to estrogen-positive feedback with a GnRH-dependent LH surge (42–44).

These considerations suggest that sequence variations in the EAP1 gene may increase the susceptibility of NHP and humans to develop a FHA-like phenotype. A recent study demonstrated the existence of heterozygous mutations in seven of 55 patients with hypothalamic amenorrhea, leading to the conclusion that such mutations, although not sufficient to cause hypothalamic hypogonadism by themselves, may increase the susceptibility to adverse conditions leading to FHA (6). In both humans and NHP, FHA can develop as a consequence of stress, whether metabolic, exercise-related, or psychosocial (45–47). Intriguingly, there are individual differences in the sensitivity of the hypothalamic-pituitary-gonadal axis to stress with some women showing marked suppression, whereas others are relatively stress resistant (45, 47–50). These differences appear to be functionally determined, at least in part, by alterations in the transsynaptic control of GnRH secretion, as indicated by the studies of Bethea, Cameron, and colleagues (51–55) who showed that stress-sensitive monkeys display abnormalities in serotonergic, GABAergic, and CRH neurotransmission, as compared with stress-resistant animals.

Association studies have been increasingly used to identify genes underlying complex human diseases. The most comprehensive approach employed by a majority of these studies is the genome-wide association analysis of large populations of subjects. Prominent examples of studies using this approach are those describing genetic associations to coronary artery disease, rheumatoid arthritis, type 1 and type 2 diabetes (12, 56, 57), and macular degeneration (13–17). Although genome-wide association analyses such as these represent a powerful tool to identify previously unknown loci for complex diseases, targeted association studies aimed at recognizing the contribution of a candidate gene, such as the present study, have also been successful (16, 17).

The association study we report here is limited with respect to power, because we do not have a large cohort of NHP available for study. However, earlier human studies using a similar approach, i.e. the use of small populations of subjects for the targeted association analysis of a single gene instead of a genome-wide analysis of multiple unknown genes, have been successful (16, 17). Because of the small sample size, it is possible to obtain spurious positive associations or an erroneous absence of association. We have addressed the former concern by performing extensive in vitro experiments that, in our view, establish the existence of functional consequences of this polymorphic site in the EAP1 promoter. We cannot, however, rule out the possibility of false-negative outcomes, because identification of such outcomes requires a much larger study, currently unfeasible. The lack of sequence variation in the body of EAP1 that could be associated with alterations in EAP1 function was indeed surprising. The multiple domains of the EAP1 gene would appear to make this gene particularly vulnerable to alterations in functional capacity caused by sequence polymorphisms. EAP1 contains several PEST domains (58) that target the mRNA for rapid degradation; a RING finger motif required for EAP1 transcriptional activity (1); a zinc-finger domain required for interactions with partners of a repressor complex (41); a polyglutamine tract, the expansion of which is central to the pathogenesis of Huntington disease and related disorders (59, 60); and a polyalanine tract known to be associated with the molecular pathogenesis of gene transcription (61).

Although to date, few candidate gene studies have been reported in rhesus macaques, the results thus far obtained show that some gene variants identified in humans are also present in these animals. For example, a conserved 21-bp insertion/deletion in the serotonin transporter gene promoter (l/s rh5-HTTLPR), first associated with increased risk for alcohol dependence in humans (62–64), is also associated with increased alcohol consumption in female rhesus monkeys (65). In a study of the rhesus tryptophan hydroxlyase-2 (TPH2) gene, encoding a protein involved in serotonin biosynthesis, naturally occurring rhesus variants were identified that mimicked the phenotypic effects of human TPH2 variants (66).

Taken all together, the present results show that a SNP in the EAP1 promoter is associated with an increase incidence of A/O in NHP and a decreased EAP1 promoter activity in response to a physiologically relevant stimulus. They also raise the intriguing possibility that common genetic variation in the EAP1 gene may be associated with FHA in humans.