Transcriptional Profiling of Equine Endometrium During the Time of Maternal Recognition of Pregnancy¹

Establishment and maintenance of pregnancy are critically dependent on embryo-maternal communication during the preimplantation period. To gain new insights into this complex process in the horse, transcriptional profiling of Day 13.5 pregnant and cyclic endometrial tissue samples was carried out using custom-designed microarrays. Selected array data were validated using quantitative RT-PCR, and proteins of interest were localized using immunohistochemistry. One hundred and six transcripts were up-regulated, whereas 47 transcripts showed lower expression levels in pregnant mares, that is, were down-regulated in pregnant mares. Half of the genes with known or inferred function are classically regulated by estrogens. Elevated transcript levels were found for genes involved in cell-cell signaling, heat shock response, and secretory proteins, among others. Solute carrier family 36 (proton/amino acid symporter), member 2, SLC36A2, was one of the most highly up-regulated genes, potentially reflecting the nutritional needs of the rapidly developing embryo. Among the genes showing lower expression in pregnant mares, estrogen receptor 1 was of particular interest because of its potential involvement in the initiation of luteolysis in cyclic mares. We hypothesize that either conceptus' estrogens or luteinizing hormone of uterine origin is involved in the observed down-regulation of estrogen receptor 1. Several of the genes identified in the current study are known to play a role in early pregnancy in species other than the horse. Thus, products of these commonly expressed genes likely contain universal activities for controlling endometrial receptivity to the conceptus, whereas other factors play unique roles within specific species in ensuring ongoing corpus luteum function. This is the first systematic study of endometrial transcriptome changes in response to the presence of an embryo during maternal recognition of pregnancy and an important step toward deciphering the embryo-maternal dialogue in equids.

Introduction

Early conceptus development, implantation, and maintenance of a pregnancy are critically dependent on a precisely orchestrated conceptus-maternal interaction that prompts continued progestagenic support, thereby ensuring a receptive uterine environment that allows the conceptus to thrive. In most mammals, the conceptus signals its presence through biochemical processes that interrupt estrous cyclicity and maintain primary CL function. The events controlling these processes are one component of the phenomenon often termed “maternal recognition of pregnancy” (MRP) [1]. These processes are well understood in domestic ruminants and other ungulates such as the pig. IFNT has been identified as a conceptus-derived paracrine factor exhibiting antiluteolytic properties in ruminants, while conceptus-derived estrogens represent the primary pregnancy recognition signal in pigs [2, 3].

The horse is one of the few domestic species in which the conceptus-derived signals used for pregnancy recognition have not been identified. Equids appear to be distinct from ruminants and pigs in the signal or signals being used for maternal recognition of pregnancy. Horses exhibit some unusual features during early pregnancy that likely contribute to pregnancy recognition. For example, the spherical equine conceptus migrates continuously throughout the uterine lumen between at least Day 9 and Day 16 after ovulation. Restriction of conceptus movement results in luteolysis with subsequent failure of pregnancy maintenance [4]. Equine conceptuses also produce substantial amounts of estrogens and prostaglandins [5–7]; however, their precise roles during early pregnancy remain unclear. Experiments attempting to prove that embryo-derived estrogens are responsible for extension of corpus luteum function have been inconclusive [8, 9]. Endometrial prostaglandin F2 alpha secretion is reduced in pregnant mares, and there is accumulating evidence that disruption of oxytocin receptor function contributes to abrogation of the luteolytic cascade during pregnancy in the horse [10–12], though the underlying mechanism leading to this alteration is not known. The mRNA for PTGS2 (previously called PGHS2), the rate-limiting enzyme in prostaglandin production, is substantially lower in endometrium from pregnant mares on Day 14 after ovulation [13]. Moreover, recent work done by Ealy and coworkers showed that exposing endometrial explants to conceptus secretions decreases PTGS2 transcript abundance [14], and the authors therefore hypothesize that PTGS2 is a target for the antiluteolytic signal produced by the equine conceptus.

Microarray analysis of both endometrial and embryonic tissues has been applied in various species to gain further insight into the molecular events underlying embryo-maternal interactions during MRP [8, 15–17]. The objective of the present investigation was to profile transcriptional changes in endometrial tissues from Day 13.5 pregnant and cyclic mares using custom-designed microarrays. Array data were validated using quantitative RT-PCR, and proteins of interest were localized using immunohistochemistry.

Materials and Methods

Animals and Tissue Collection

All animal procedures were completed in accordance with and with the approval of the institutional animal care and use committee at the University of Kentucky and the University of Florida. All mares were subjected to a breeding soundness examination, and only those mares having no apparent detectable abnormalities of the reproductive tract were used in the study. The estrous cycle of four mares was shortened by the administration of 10 mg of a prostaglandin analogue, dinoprost (Lutalyse; Pfizer, New York, NY). After induction of luteolysis, the animals were examined daily for follicular development, development of uterine edema, as well as cervical and uterine tone. In the absence of a distinct corpus luteum, the presence of a follicle at least 35 mm in diameter, pronounced uterine edema, and diminished uterine and cervical tone, each mare received 2500 IU of human chorionic gonadotropin (hCG; Chorulon; Intervet, Millsboro, DE) as an ovulation-inducing agent and was simultaneously inseminated. For insemination, a minimum of 500 × 10⁶ progressively motile sperm were used. Semen was collected the day of insemination from a Thoroughbred stallion with previously proven fertility. The collected semen was extended with E-Z Mixin (Animal Reproduction Systems, Chino, CA) to a final concentration of 50 × 10⁶/ml and supplemented with ticarcillin. Mares were examined twice daily by ultrasonography for signs of ovulation. If ovulation had not occurred within 48 h after insemination, another artificial insemination was performed. All mares were examined for accumulation of free uterine fluid on Days 1 and 2 after ovulation.

Endometrial tissue samples were recovered on Day 13.5 of pregnancy. The day of pregnancy was determined based on the results of repeated ultrasonic examinations of the genital tract and the knowledge that most ovulations occur 36 h after administration of hCG. The first pregnancy examination was performed on Day 12.5, and an increase in diameter of the vesicle occurring from Day 12.5 to Day 13.5 served as an indicator of embryonic viability. Each embryo showed the expected growth rate of 3 mm from Day 12.5 to Day 13.5. For recovery of tissue samples, the position of the embryo within the uterus was determined by ultrasound, and endometrial biopsies were retrieved transcervically from the non-embryo-bearing uterine horn. A total of three endometrial biopsies were obtained and snap-frozen in liquid nitrogen. Each mare received 10 mg dinoprost after sample collection and was allowed to ovulate spontaneously during the following estrus. During diestrus of the consecutive cycle, the same procedure, that is, induction of luteolysis, induction of ovulation, insemination, and recovery of tissue samples on Day 13.5, was repeated. During this cycle, insemination was carried out as a sham insemination using semen extender only. A blood sample for progesterone analysis was obtained prior to each sample collection.

Isolation of RNA

Total cellular RNA was isolated from endometrial samples using Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's recommendation. The volume of RNA was then adjusted to 150 μl and precipitated using 150 μl of isopropanol and 15 μl of 3 M sodium acetate. RNA was analyzed for quality by determining the RNA integrity number (RIN) using a Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA) and was quantified via spectrophotometry using a NanoDrop ND-1000 (Agilent Technologies). Samples with a RIN of 8.0 or greater, a 260/280 ratio of 1.95 or greater, and a 260/230 ratio of 2.0 or greater were used for analysis.

Construction of Custom Microarray

Construction of normalized cDNA libraries

Equal masses of total RNA from four individual samples were mixed to create a pooled sample with a final concentration of 1 μg/μl. A cDNA library was constructed from 1 μl (1 000 ng) of this material using the SMART kit from Clontech (Mountain Views, CA). Briefly, the RNA was used as template to synthesize cDNA, which then was amplified using an Advantage 2 PCR kit (Clontech) for a total of 21 cycles as determined from an optimization experiment. PCR products were purified (QIAquick PCR Purification kit, Qiagen, Germantown, MD) and eluted in 50 μl deionized water. This product (20 μl) was divided into two tubes and used as a template for a second PCR for 16 cycles. The products were pooled, purified using QIAquick PCR Purification kit, eluted in deionized water, and then reconstituted to a final concentration of 100 ng/ml. The generated cDNA library was normalized using the Trimmer kit from Evrogen (Evrogen Joint Stock Company, Moscow, Russia) following the provided protocols, starting with 1200 ng of cDNA. Normalization efficiency was determined by assessing the rate of actin, beta amplification, a gene expected to be expressed at high levels in nonnormalized cDNA pools. Amplifications were performed for 15 cycles and visualized on a 1.0% agarose gel. PCR reactions performed on normalized cDNA resulted in weaker band intensity than those reactions performed on nonnormalized cDNA (data not shown). This is consistent with the premise that the normalization process was effective.

Sequencing of the Normalized Libraries

DNA sequencing was performed on a 454 Life Sciences GS-FLX pyrosequencer by the University of Florida's Interdisciplinary Center for Biotechnology Research (UF-ICBR). Sequencing was performed as described in the supplementary material and methods in Margulies et al. [18] with slight modifications as specified by 454 Life Sciences. Briefly, high-molecular-mass DNA amplified using the rolling circle amplification reaction was sheared by nebulization to a size range of 300–800 base pairs. DNA fragment ends were repaired and phosphorylated using T4 DNA polymerase and T4 polynucleotide kinase. Adaptor oligonucleotides “A” and “B,” supplied with the 454 Life Sciences sequencing reagent kit, were ligated to the DNA fragments using T4 DNA ligase. Purified DNA fragments were hybridized to DNA capture beads and clonally amplified.

Bioinformatics and Array Construction

DNA sequence data from the titration and production runs were combined in a single assembly using version 1.1.03.24 of the GS FLX Newbler sequence assembly software. Paracel Transcript Assembler was then used to combine 454 contigs and reads with genes predicted from the horse EquCab2 assembly, where a series of sequence cleaning, chimera identification, clustering, and assembly steps was performed.

All contigs and singlets were annotated using a BLAST search against NCBI NR and NT databases. The e-value threshold was set at 1e-4. For each query sequence, the top 100 BLAST hits were obtained, if available, and stored in BlastQuest, a SQL database developed by the UF-ICBR that facilitates similarity-based sequence annotation with gene ontology information. NCBI Gene database was used to map horse transcripts to homologs from human, mouse rat, dog, chimpanzee, or other mammals. For sequences with no BLAST hit passing 1e-4 threshold, ESTScan [19] was used to predict CDS (coding sequence) region. NR/NT BLAST and ESTScan results were combined to determine the sequence orientation. The final set of target sequence included forward strands of annotated transcripts and both strands for unannotated transcripts.

The “GE Probe Design” tool on eArray (Agilent Technologies) was used to design the 60-mer probes. For each target sequence, one best probe was selected based on melting temperature, base composition, and cross-hybridization potential. These probes were printed in a 4 × 44 K format with a random layout.

Array Hybridization

Fluorescent cRNA was generated using the Quick Amp Labeling Kit (Agilent Technologies). In brief, total RNA (5 μg) was reverse transcribed to cDNA, during which a T7 sequence was introduced. T7 RNA polymerase-driven RNA synthesis was used for the preparation and labeling of RNA with Cy3 or Cy5 dye. The fluorescent cRNA probes were purified using Qiagen RNeasy Mini kit (Qiagen), and an equal amount (825 ng) of Cy3 and Cy5 labeled cRNA probes were hybridized on the custom array. Only samples that showed a specific activity equal or greater than 8.0 pmol Cy3 or Cy5 per microgram of cRNA were used for array hybridization. Two of the four samples obtained from pregnant and nonpregnant mares were each labeled with Cy5 and Cy3, respectively. Pregnant and nonpregnant samples from each of the four mares included in the study were paired for the hybridization. Hybridization was carried out overnight at 65°C for 17 h, and hybridized slides were washed using the Gene Expression Wash Buffer kit followed by Stabilization and Drying solution (Agilent Technologies). Genepix 4100A scanner was used for image acquisition (Molecular Devices Corporation, Sunnyvale, CA).

Analysis of Array Data

GenePixPro 6.0 microarray image analysis software was used for computation of feature intensities. Mean feature intensity corrected for the local median intensity background was used for further analysis and is hereafter referred to as raw data. Raw data were imported into JMP Genomics 3.0 (Cary, NC) and log base 2 transformed. Quality of raw data was assessed using Distribution Analysis and Correlation and Principal Component Analysis functions in JMP Genomics. MA plots were generated to visualize intensity-dependent ratio of raw data. Raw data were then normalized using locally weighted linear regression. Mixed model analysis was performed to identify significantly differentially expressed genes whereby pregnancy status was included as a fixed effect and array as random effect. An F-test on least-square means was used to estimate the significance of difference for each gene in each comparison where P < 0.01 was considered to be statistically different. Only genes showing a difference in signal intensity of at least 1.3 between pregnant and nonpregnant in all four animals were considered to be differentially expressed. Of the transcripts identified as differentially expressed between pregnant and cyclic animals, only those with a signal-to-noise ratio of equal to or greater than 3.0 were kept for further analysis. The coefficient of variation of the expression ratios between pregnant and nonpregnant animals was calculated.

Based on the data for human orthologous genes, the Database for Annotation, Visualization and Integrated Discovery (DAVID) was used to categorize the genes regarding their molecular function or the biological processes in which they are involved, respectively. Resulting data were supplemented with additional information from Entrez Gene (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene) and from the literature (http://www.ncbi.nlm.nih.gov/PubMed). Existing literature was reviewed for evidence regarding regulation of differentially expressed transcripts by estrogen. This search was not limited to a certain species or tissue type, and evidence arising from both whole animal and cell culture experiments was included. Evidence for regulation of differentially expressed transcripts by steroid hormones other than estrogen, that is, progesterone, was not taken into consideration for the purpose of this study.

Real-Time RT-PCR

Total RNA was isolated as described previously. RNA samples (5 μg/reaction) were treated with RNase-free DNase I (Ambion, Woodward Austin, TX) for 15 min at 37°C, heat denatured (75° for 10 min), then reverse transcribed using High Capacity cDNA Reverse Transcription Kit and random hexamers (Applied Biosystems, Foster City, CA). Primers specific for the selected transcripts were designed using Jellyfish 3.3.1 (Field Scientific LLC, Lewisburg, PA). Amplification efficiencies were similar between all primer sets used as determined using 1:4 serial dilutions of standard endometrial RNA samples. Real-time PCR was completed using SYBR Green PCR Master Mix (Applied Biosystems) with the following cycling conditions: 95°C for 10 min, 40 cycles of 95°C for 15 sec and 59°C for 1 min, and 55–95°C for dissociation. Each PCR was performed in triplicate. All reactions were automatically pipetted using the epMotion Automated Pipetting Systems (Eppendorf, Westbury, NY). Specificity of amplification was monitored by including non-reverse-transcribed RNA reactions for each sample and by completing a dissociation analysis at the end of each real-time run to verify the amplification of a single product. Changes in gene expression were calculated by mean threshold cycle (C_T) and then normalized for the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) to generate delta (Δ) C_T values. GAPDH has been shown to be stably expressed across endometrial samples obtained from a wide range of reproductive statuses (unpublished observation). Changes in relative abundance of specific transcripts were examined by calculating the fold effect using the delta delta (ΔΔ) C_T method [20].

Immunohistochemistry

Endometrial tissue samples obtained from pregnant and nonpregnant mares were fixed by immersion in 10% formalin over 24 h. After the 24-h fixation period, samples were dehydrated and embedded in paraffin wax. The paraffin blocks were then cut into 4-μm sections, and the layers were placed onto poly-l-lysine-coated glass slides. Sections were deparaffinized in xylene and rehydrated through a decreasing alcohol gradient. Antigen retrieval was accomplished through microwave irradiation of the sections in 10 mM sodium citrate buffer, pH 6.0. Sections were then allowed to cool to room temperature, followed by washing with PBST. The following primary antibodies were chosen based on their potential biological implication and used in conjunction with ImmunoCruz staining systems (rabbit and goat; Santa Cruz Biotechnology Inc., Santa Cruz, CA) to localize protein expression within the endometrial sections: ESR1 (Santa Cruz, sc-542), HSPB1 (Santa Cruz, sc-1048), ISG15 (Santa Cruz, sc-50366), SLC36A2 (Santa Cruz, sc-130224), and LHB (Leica Microsystems, Bannockburn, IL, HCGP-U). In brief, endogenous peroxidase was quenched followed by washing with PBST. Nonspecific antigen sites were blocked by incubating the sections with serum block for 20 min. After blocking, sections were incubated with the primary antibody (diluted 1:50) for 2 h at room temperature and then washed with PBST. Incubation with biotinylated secondary antibody was carried out for 30 min at room temperature, followed by 30-min incubation with horseradish peroxidase (HRP)-streptavidin complex. Slides were then washed with PBST and incubated with HRP substrate for 5 min at room temperature. Counterstaining was performed using hematoxylin. Sections were dehydrated and mounted with a cover slip. As negative controls, the primary antibody was substituted with serum block solution or normal serum from the species the primary antibody was raised in.

Results

454 Pyrosequencing

Two separate sequencing runs were performed from the pregnant and nonpregnant normalized cDNA libraries. The initial titration run, which is performed at a smaller scale, is used to identify the approximate density of beads on the plate, resulted in 656 122 vector-trimmed bases for the pregnant library and 1 020 337 vector-trimmed bases for the nonpregnant library. The production run of the pregnant library resulted in 37 553 462 vector-trimmed bases, whereas the production run of the nonpregnant library resulted in 22 254 857 vector-trimmed bases. Assembly of sequences from both libraries resulted in 33 881 contiguous sequences.

Bioinformatics and Array Design

The array format used was the 4 × 44 k Agilent microarray. For each array, 45 220 total features included 1 417 Agilent control probes and 43 803 user probes. To occupy the 43 803 user features, the four probe groups were printed as specified in Table 1.

Table 1.

Occupation of the 43 803 available user features on the 4 × 44 k Agilent microarray.

Probe group	No. of probes printed
Control probes	250
Annotated	30 443
Annotated_minus	2134
Nonannotated_pregnant	2233
Nonannotated_non_pregnant	650
Nonannotated_genome	26
Nonannotated_long	8067

Detection of Differentially Expressed Genes in Endometrial Tissue Samples of Pregnant and Nonpregnant Animals

Custom oligonucleotide microarrays were used to analyze the transcriptional response of the endometrium to a conceptus. Endometrial tissue samples were obtained from mares at Day 13.5 of gestation and Day 13.5 of the estrous cycle (n = 4 mares/pregnancy status). Peripheral progesterone concentrations did not differ between pregnant and nonpregnant mares (7.25 ± 1.5 and 6.95 ± 1.3 ng/ml, respectively). Total RNA was isolated, fluorescently labeled, and analyzed after array hybridization. A total of 153 transcripts were identified as differentially expressed between pregnant and cyclic animals. One hundred and six transcripts showed higher expression levels in tissue samples obtained from pregnant mares with the fold change ranging from 1.30 to 15.46 (referred to as up-regulated genes). Forty-seven genes showed lower expression levels in tissue samples obtained from pregnant mares with the fold change ranging from 1.42 to 4.54 (referred to as down-regulated genes). The mean coefficient of variation of expression ratio between pregnant and nonpregnant mares was 41%, ranging from 2% to 136%.

Sixty-one (58%) of the up-regulated transcripts correspond to genes with known or inferred function. The biological processes for most of the up-regulated genes are depicted in Figure 1. The majority of genes are involved in either one of the biological processes—transport, signal transduction, proteolysis, or ion homeostasis—or are secreted proteins. DAVID analysis resulted in the following significantly enriched gene ontologies: signal sequence, secreted, cell-cell signaling, heat shock protein, and tetratricopeptide repeat.

Fig. 1.

Classification of genes up-regulated at Day 13.5 of pregnancy into functional categories. Gene ontology processes are ranked by the number of genes affected, and those significantly enriched are marked with an asterisk. Genes may appear multiple times within these hierarchies so that the number of genes provided in this figure is relative to a certain biological process, not absolute.

Twenty-seven (56%) of the down-regulated transcripts correspond to genes with known or inferred function The biological processes for most of the down-regulated genes are involved in are depicted in Figure 2. The majority of genes are either involved in one of the biological processes: protein binding, response to stimulus, cell communication, or cell motility. DAVID analysis resulted in the following significantly enriched gene ontologies: protein binding, protein metabolic process, and lysosome.

Fig. 2.

Classification of genes down-regulated at Day 13.5 of pregnancy into functional categories. Gene ontology processes are ranked by the number of genes affected, and those significantly enriched are marked with an asterisk. Genes may appear multiple times within these hierarchies so that the number of genes provided in this figure is relative to a certain biological process, not absolute.

Twenty-two of the up-regulated transcripts are regulated by estrogens (Table 2), which corresponds to 27% of all up-regulated genes, or 36% of up-regulated genes with known or inferred function. Of the down-regulated transcripts, 15 are regulated by estrogens (Table 2), corresponding to 31% of all down-regulated genes or to 55% of down-regulated genes with known or inferred function, respectively.

Table 2.

Transcripts showing differential mRNA expression levels in equine endometrium between pregnant and cyclic mares 13.5 days after ovulation.^a

Equine gene/cDNA or homologue	Gene symbol	GenBank accession no.	E2	Putative function	FC	CV (%)	P value
Homo sapiens cDNA FLJ11245 fis, clone PLACE1008629	–	AK002107		–	22.9	64	0.003
Equus caballus insulin-like growth factor binding protein 1	IGFBP1	XM_001496203	x	Regulation of cell growth	19.2	89	0.007
Homo sapiens proton/amino acid transporter 2 (PAT2)	SLC36A2	AY162214		–	15.5	14	0.001
Equus caballus similar to LOC155006 protein (LOC100066723)	–	XM_001496914		–	14.5	136	0.01
Equus caballus DNA, microsatellite TKY2496	–	AB216439		–	12.9	81	0.002
Equus caballus DNA, microsatellite TKY504	–	AB103722		–	11.2	77	0.002
Equus caballus similar to heat shock 22kDa protein 8 (HSPB8)	HSPB8	XM_001490413	x	Response to stress	10.6	41	0.002
Equus caballus subtracted library fragment 75	–	AY246839		–	10.0	70	0.007
Equus caballus similar to ATPase, H+ transporting, lysosomal V0 subunit a4	ATP6V0A4	XM_001499288		Ion/proton transport	5.7	82	0.008
Homo sapiens chromosome 15, clone CTD-3221M10	–	AC103967		–	5.4	72	0.002
Equus caballus DNA, microsatellite TKY2163	–	AB216106		–	5.4	74	0.006
Equus caballus similar to fibroblast growth factor 9 (glia-activating factor)	FGF9	XM_001489697	x	Growth factor activity	5.3	60	0.005
Homo sapiens GM2 ganglioside activator (GM2A)	GM2A	NM_000405	x	Lipid transport; oligosaccharide catabolism	5.0	68	0.01
Equus caballus similar to F-box and leucine-rich repeat protein 7	FBXL7	XM_001499794	x	Ubiquitin cycle	4.4	131	0.009
Rhesus Macaque BAC CH250-204A18 complete sequence	–	AC198676		–	4.3	51	0.007
Equus caballus genomic BAC clone CH241-440E11, containing KIT	–	AM420315		–	4.1	89	0.01
Homo sapiens chromosome 5 clone CTC-224D3	–	AC008385		–	3.8	52	0.002
Equus caballus DNA, microsatellite TKY3263	–	AB217206		–	3.8	83	0.009
Equus caballus coagulation factor C homolog, cochlin (Limulus polyphemus)	COCH	XM_001489788		–	3.6	56	0.007
Equus caballus similar to solute carrier family 4, sodium borate transporter-like, member 11	SLC4A11	XM_001496863	x	Cellular cation homeostasis	3.6	49	0.008
Equus caballus similar to stanniocalcin	STC1	XM_001493195	x	Cell-cell signaling; calcium ion homeostasis	3.5	117	0.007
Homo sapiens neuropeptide FF receptor 1	NPFFR1	NM_022146		G-protein coupled receptor protein signaling pathway	3.5	21	0.007
Homo sapiens 12 BAC RP11-591N1	–	AC009522		–	3.3	56	0.008
Canis familiaris hypothetical protein LOC610484	–	XM_847996		–	3.3	38	0.006
Homo sapiens olfactory receptor, family 4, subfamily C, member 16	OR4C16	NM_001004701		–	3.2	76	0.001
Equus caballus colony stimulating factor 3 receptor (granulocyte)	CSF3R	XM_001499027		Cell adhesion; signal transduction	3.1	66	0.01
Homo sapiens chromosome 15, clone RP11-186H10	–	AC113190		–	3.0	33	0.008
Homo sapiens chromosome 5 clone RP11-428C6	–	AC093283		–	3.0	65	0.005
Homo sapiens 3q BAC RP11-82C9	–	AC078985		–	3.0	67	0.001
Homo sapiens inositol 1,4,5-triphosphate receptor, type 1	ITPR1	NM_001099952	x	Calcium ion transport; signal transduction	3.0	35	0.008
Homo sapiens family with sequence similarity 101, member A	FAM101A	NM_181709		Transport/receptor activity	3.0	66	0.004
Human DNA sequence from clone RP3-477H23 on chromosome 22q12.1–12.2	–	AL033538		–	3.0	63	0.01
Human DNA sequence from clone RP6-64P14 on chromosome Xq24–25, complete sequence	–	AL109800		Glutamate signaling pathway; ion transport	2.9	70	0.01
Equus caballus similar to kin of irregular chiasm 2 splice	KIRREL2	XM_001493050		Cell adhesion	2.9	62	0.01
Felis catus clone BAC 160a17 major histocompatiblity complex classical class II region	–	AY152836		–	2.9	67	0.01
Homo sapiens synaptogyrin 3	SYNGR3	NM_004209	x	–	2.9	58	0.008
Equus caballus similar to IL-22 receptor, alpha 1	IL22RA1	XM_001501288		Receptor activity	2.9	60	0.009
Homo sapiens chromosome 5 clone CTC-483L24	–	AC008533		–	2.9	49	0.01
Pan troglodytes BAC clone CH251-700P20 from chromosome 7	–	AC184876		–	2.9	14	0.001
Equus caballus DNA cross-link repair 1B (PSO2 homolog, S. cerevisiae)	DCLRE1B	XM_001495742		DNA repair	2.8	54	0.008
Homo sapiens BAC clone RP11-574K22 from 2	–	AC016776		–	2.8	51	0.01
Equus caballus similar to tumor-associated calcium signal transducer 2	TACSTD2	XM_001497584	x	Signal transduction; cell proliferation	2.8	70	0.006
Equus caballus similar to protocadherin beta 1	PCDHB1	XM_001502250		Homophilic cell adhesion	2.8	46	0.019
Homo sapiens poly(A) binding protein interacting protein 2B	PAIP2B	NM_020459		Negative regulation of translational initiation	2.7	48	0.007
Equus caballus similar to angiopoietin-like 2	ANGPTL2	XM_001501670		Signal transduction	2.7	61	0.004
Equus caballus similar to makorin ring finger protein 1	MKRN1	XM_001496338	x	–	2.7	59	0.008
Bos taurus BAC CH240-161G19 (Children's Hospital Oakland Research Institute Bovine BAC Library (male))	–	AC150517		–	2.7	60	0.01
Homo sapiens ubiquitin specific peptidase 54	USP54	NM_152586		Ubiquitin-dependent protein catabolic process	2.7	59	0.008
Human chromosome 14 DNA sequence BAC C-2510G9 of library CalTech-D from chromosome 14	–	AL133238		–	2.6	56	0.008
Mus musculus BAC clone RP23-306C13 from chromosome 13	–	AC169128		–	2.6	22	0.01
Homo sapiens clone TA40 untranslated mRNA	–	AF126109		–	2.6	60	0.01
Equus caballus Rho GTPase activating protein 30	ARHGAP30	XM_001503866		Signal transduction	2.6	39	0.001
Homo sapiens heat shock 27kDa protein family, member 7 (cardiovascular)	HSPB7	NM_014424	x	Response to stress	2.6	35	0.01
Equus caballus zinc finger protein 646	ZNF646	XM_001500835		Nucleic acid/zinc ion binding	2.6	65	0.01
Equus caballus similar to solute carrier family 44 member 5	SLC44A5	XM_001497544		–	2.6	67	0.01
Homo sapiens Chromosome 17 BAC GS1-531I17	–	AC004805		–	2.5	58	0.009
Homo sapiens chromosome 5 clone CTB-78F1	–	AC008708		–	2.5	54	0.01
Homo sapiens BAC clone RP11-492P11 from 4	–	AC109518		–	2.4	41	0.009
Equus caballus similar to DEAQ box RNA-dependent ATPase 1	DQX1	XM_001498853		Nucleic acid binding	2.4	45	0.006
Equus caballus clone CH241-280K8	–	AC124156		–	2.4	24	0.008
Equus caballus similar to signal peptide, CUB domain, EGF-like 1	SCUBE1	XM_001500812		Protein heterooligomerization	2.4	21	0.005
Equus caballus similar to putative serine protease 23	PRSS23	XM_001489895		Proteolysis	2.4	16	0.002
Pan troglodytes BAC clone CH251-391M23 from chromosome 7	–	AC187377		–	2.3	63	0.009
Equus caballus similar to S100 calcium binding protein A2	S100A2	XM_001494667	x	Calcium ion binding	2.3	45	0.01
Mouse DNA sequence from clone RP23-399H3 on chromosome 4	–	AL627259		–	2.3	39	0.009
Equus caballus similar to solute carrier family 46 member 2	SLC46A2	XM_001490426		Transport	2.2	74	0.01
Equus caballus similar to G protein-coupled bile acid receptor 1	GPBAR1	XM_001491146		Signal transduction	2.2	38	0.007
Equus caballus similar to fibroblast growth factor-binding protein	FGFBP	XM_001498741	x	Growth factor binding	2.1	21	0.007
Equus caballus FIC domain containing	FICD	XM_001501073		Protein binding	2.1	24	0.008
Equus caballus similar to crystallin Alpha B	CRYAB	XM_001501779	x	Protein tyrosine kinase signaling; anti-apoptosis	2.0	30	0.008
Equus caballus similar to zona pellucida glycoprotein 2 (sperm receptor)	ZP2	XM_001494769		Binding of sperm to zona pellucida	2.0	7	0.004
Human chromosome 14 DNA sequence BAC R-132J14 of library RPCI-11 from chromosome 14 of Homo sapiens (Human)	–	AL138499		–	2.0	45	0.010
Equus caballus N-myc downstream regulated 1	NDRG1	XM_001498825	x	Mast cell activation	2.0	23	0.01
Equus caballus similar to vanin 3	VNN3	XM_001504346		–	1.9	12	0.005
Equus caballus similar to transmembrane 7 superfamily member 3	TM7SF3	XM_001502786		–	1.9	30	0.01
Equus caballus similar to lectin, mannose-binding 2-like	LMAN2L	XM_001492041		Protein folding	1.9	52	0.007
Rhesus Macaque BAC CH250-419N13 complete sequence	–	AC192013		–	1.9	20	0.00033
Equus caballus similar to interferon-induced protein with tetratricopeptide repeats 1	IFIT1	XM_001501389		Binding	1.8	45	0.01
Equus caballus similar to galactosidase, beta 1	GLB1	XM_001490147	x	Carbohydrate metabolism	1.8	32	0.01
Equus caballus clone CH241-67M21	–	AC124908		–	1.8	20	0.005
Equus caballus similar to chromosome 6 open reading frame 81	C6orf81	XM_001499224		Binding	1.7	36	0.01
Equus caballus regulatory factor X, 4 (influences HLA class II expression)	RFX4	XM_001499033	x	Regulation of transcription, DNA-dependent	1.7	13	0.007
Equus caballus similar to dopey family member 2	DOPEY2	XM_001916299		Golgi to endosome transport	1.7	8	0.001
Equus caballus similar to interferon stimulated gene 15	ISG15	XM_001496658		Cell-cell signaling; protein modification process	1.7	50	0.008
Methanococcus vannielii SB, complete genome				-	1.7	9	0.008
Equus caballus similar to interferon, alpha-inducible protein 6	IFI6	XM_001500542		Immune response; anti-apoptosis	1.6	44	0.01
Equus caballus similar to membrane-spanning 4-domains, subfamily A, member 8B	MS4A8B	XM_001493583		Signal transduction	1.6	18	0.009
Homo sapiens chromosome 1 clone RP11-430H15	–	AC093577		–	1.6	15	0.007
Equus caballus similar to interferon regulatory factor 7	IRF7	XM_001494157		Transcription	1.6	51	0.006
Equus caballus S100 calcium binding protein A6	S100A6	AF083065	x	Cell-cell signaling; fibroblast proliferation	1.6	16	0.004
Equus caballus clone CH241-91H15	–	AC124909		–	1.5	21	0.009
Equus caballus similar to Ras association (RalGDS/AF-6) domain family (N-terminal) member 8	RASSF8	XM_001502654		Signal transduction	1.5	19	0.004
Equus caballus kallikrein 1 (glandular kallikrein precursor, KLK1E2)	KLK1	NM_001081893	x	Proteolysis	1.5	3	0.002
Equus caballus similar to programmed cell death 1	PDCD1	XM_001497864	x	Signal transducer activity	1.5	14	0.005
Equus caballus similar to gap junctional protein, beta 5, 31.1kDa	GJB5	XM_001503718	x	Cell communication	1.4	6	0.001
Equus caballus luteinizing hormone beta polypeptide	LHB	XM_001490342		Hormone activity	1.4	5	0.003
Equus caballus similar to WD repeat domain 63	WDR63	XM_001495268		Microtubule-based movement	1.4	11	0.006
Equus caballus similar to syndecan 4	SDC4	XM_001500606		Cytoskeletal protein binding	1.4	2	0.002
Equus caballus TIMP metallopeptidase inhibitor 1	TIMP1	U95039	x	Metallopeptidase inhibitor	1.4	16	0.006
Homo sapiens tubulin polymerization-promoting protein family member 3	TPPP3	NM_015964		Microtubule polymerization	1.4	24	0.005
PREDICTED: Equus caballus ATP-binding cassette, sub-family G (WHITE), member 2	ABCG2	XM_001496332	x	Transport; ATPase activity	1.3	15	0.006
Campylobacter hominis ATCC BAA-381	–	NC_009713		–	1.3	7	0.01
Equus caballus discs, large (Drosophila) homolog-associated protein 3	DLGAP3	XM_001503709		Cell-cell signaling	1.3	18	0.001
Equus caballus similar to chymotrypsin C (caldecrin)	CTRC	XM_001489099		Proteolysis	1.3	19	0.001
Equus caballus solute carrier family 37 (glycerol-3-phosphate transporter) member 1	SLC37A1	XM_001492192		Carbohydrate transport	1.3	3	0.002
Equus caballus similar to chromosome 2 open reading frame 70 (Homo sapiens)	C2orf70	XM_001501068		–	1.3	25	0.002
Homo sapiens chromosome 3 clone RP11-16E8, complete sequence	–	AC096914		–	0.7	12	0.01
Equus caballus similar to tetraspanin 12	TSPAN12	XM_001502043		–	0.7	54	0.01
Equus caballus similar to fucosyltransferase 8 (alpha (1,6) fucosyltransferase)	FUT8	XM_001499525	x	TGF beta and integrin mediated signaling pathway	0.7	10	0.009
Equus caballus similar to family with sequence similarity 165, member B	FAM165B	XM_001496974		–	0.7	40	0.01
Homo sapiens 12 BAC RP11-424C20 (Roswell Park Cancer Institute Human BAC Library)	–	AC112777		–	0.7	27	0.008
Homo sapiens chromosome 18, clone RP11-57G22	–	AC012123		–	0.7	24	0.009
Equus caballus similar to myeloid/lymphoid or mixed lineage-leukemia (trithorax homolog, Drosophila); translocation to 3	MLLT3	XM_001917557		Transcription; intracellular signaling cascade	0.7	2	0.01
Homo sapiens BAC clone RP11-73K9 from 4	–	AC109347		–	0.7	37	0.009
Equus caballus similar to UDP-N-acetyl-alpha-D-galactosamine:polypeptide N-acetylgalactosaminyltransferase-like 4	GALNTL4	XM_001501162		Calcium binding, transferase activity	0.7	8	0.008
Homo sapiens BAC clone RP11-132H23 from 2	–	AC074090		–	0.7	16	0.006
Equus caballus similar to membrane-spanning 4-domains, subfamily A, member 4	MS4A4A	XM_001493295		Signal transduction	0.7	6	0.001
Equus caballus similar to epithelial stromal interaction 1 (breast)	EPSTI1	XM_001915526		-	0.7	13	0.003
Homo sapiens secretogranin III	SCG3	NM_013243	x	-	0.7	26	0.002
Canis familiaris DENN/MADD domain containing 4A	DENND4A	XM_847605	x	Transcription	0.7	11	0.01
Homo sapiens chromosome 18, clone RP11-57G22	–	AC012123		–	0.7	4	0.003
Pig DNA sequence from clone PigE-259M5 on chromosome 7, complete sequence	–	CR956428		–	0.6	25	0.003
Homo sapiens 3′-phosphoadenosine 5′-phosphosulfate synthase 2	PAPSS2	NM_001015880		Nucleic acid metabolic process	0.6	46	0.01
Equus caballus nuclear receptor subfamily 2, group C, member 2	NR2C2	XM_001491050	x	Cell division; transcription	0.6	16	0.004
Equus caballus estrogen receptor 1	ESR1	NM_001081772	x	Estrogen receptor signaling	0.6	28	0.01
Rhesus Macaque BAC CH250-348K7 () complete sequence	–	AC191835		–	0.6	42	0.01
Equus caballus similar to connective tissue growth factor	CTGF	XM_001503316	x	Cell adhesion; regulation of cell growth; angiogenesis	0.6	33	0.010
Equus caballus similar to lymphatic vessel endothelial hyaluronic acid receptor 1	LYVE1	XM_001504920		Cell adhesion; transport; receptor	0.6	8	0.001
Zebrafish DNA sequence from clone DKEY-270G12 in linkage group 6	–	CR628393		–	0.6	4	0.001
Equus caballus monoamine oxidase A	MAOA	NM_001081832	x	Electron transport; dopamine catabolic process	0.6	28	0.001
Human chromosome 14 DNA sequence BAC R-998D10 of library RPCI-11 from chromosome 14	–	AL161668		–	0.6	10	0.008
Glycine tomentella clone gtd1-31b20	–			–	0.6	36	0.004
Equus caballus similar to low density lipoprotein-related protein 2	LRP2	XM_001494269	x	Calcium binding; receptor	0.6	42	0.005
Homo sapiens solute carrier family 31 (copper transporter), member 1	SLC31A1	NM_001859	x	Ion transport	0.6	34	0.004
Equus caballus similar to KIAA1797	KIAA1797	XM_001496721		–	0.6	58	0.008
Equus caballus similar to plasminogen	PLG	XM_001500502	x	Neg. regulation of cell proliferation	0.5	50	0.009
Mustela vison clone IX6N2 genomic sequence	–	DQ271579		Response to stimulus	0.5	26	0.004
Homo sapiens chromosome 18, clone RP11-27G24	–	AC016229		–	0.5	35	0.005
PREDICTED: Equus caballus similar to reticulocalbin 2, EF-hand calcium binding domain	RCN2	XM_001493149		Calcium binding	0.5	35	0.005
Equus caballus cysteine-rich secretory protein 3	CRISP3	NM_001081874	x	–	0.5	40	0.009
Homo sapiens gap junction protein, alpha 1, 43kDa	GJA1	NM_000165	x	Signal transducer activity	0.5	57	0.010
Equus caballus similar to ectonucleotide pyrophosphatase/phosphodiesterase 3	ENPP3	XM_001503291		Endonuclease activity	0.5	39	0.005
Pan troglodytes BAC clone CH251-457N21 from chromosome 7	–	AC192766		–	0.5	33	0.006
SLC39A9 solute carrier family 39 (zinc transporter), member 9	SLC39A9	DQ786762		Metal ion transporter	0.5	76	0.005
Equus caballus similar to cholecystokinins	CCK	XM_001501653	x	Signal transduction	0.5	22	0.003
Bos taurus BAC CH240-85N4 (Children's Hospital Oakland Research Institute Bovine BAC Library (male))	–	AC151117		–	0.4	6	0.003
Equus caballus similar to secretogranin III	SCG3	XM_001501619		–	0.4	33	0.006
Equus caballus similar to kynurenine 3-monooxygenase (kynurenine 3-hydroxylase)	KMO	XM_001492701		Electron carrier activity	0.4	52	0.005
Zebrafish DNA sequence from clone DKEY-51M13 in linkage group 5 Contains the 3′ end of a novel gene similar to vertebrate fibrinogen C domain containing 1	FIBCD1	CT573108		–	0.4	58	0.01
Equus caballus similar to protein kinase domain containing, cytoplasmic homolog (mouse)	PKDCC	XM_001499180		Protein phosphorylation	0.4	26	0.009
Equus caballus similar to cathepsin L	CTSL1	XM_001495745		Proteolysis	0.3	52	0.006
Equus caballus fibroblast growth factor 7	FGF7	XM_001502116	x	Cell proliferation; signal transduction	0.3	58	0.005
Equus caballus similar to apolipoprotein A-I	APOA1	XM_001502469		Protein phosphorylation; lipid transport	0.2	15	0.01

^a

 E2: regulated by estrogens; FC: fold-change (expression in pregnant mares relative to nonpregnant mares); CV: coefficient of variation.

Real-Time RT-PCR

To verify selected expression data obtained from the microarray analysis, real-time RT-PCR was performed on a subset of genes that showed a significant up- or down-regulation between pregnant and nonpregnant animals. Four up-regulated and four down-regulated genes were included in the analysis (Fig. 3; results for SLC36A2 are omitted because of large fold change). Difference in expression data obtained by microarray analysis were confirmed using this approach. The ratios obtained by real-time RT-PCR were, in most cases, slightly higher and, in one case (SLC36A2), substantially higher compared to those obtained by array hybridization. The discrepancy in fold change for SLC36A2 (mean ratio microarray: 15.46 versus mean ratio qPCR: 380.7) highlights the tendency of microarrays to show suppressed changes in expression due to their narrower dynamic range relative to real-time RT-PCR, a phenomenon most pronounced for highly expressed genes such as SLC36A2. Five transcripts (PTGS1, PGR, PGFS, OXTR, and PLA2G4A) that showed no significant difference in expression level in the microarray analysis were also analyzed by real-time RT-PCR. No statistically significant difference in mRNA levels were detected.

Fig. 3.

Validation of the microarray results obtained by real-time RT-PCR of the expression of a subset of up- and down-regulated transcripts. The y-axis shows the amplitude of relative expression determined by microarray (gray bars) or real-time RT-PCR (black bars). The error bar represents the SEM.

Immunohistochemical Localization of Selected Gene Products

Immunohistochemistry was performed using four selected up-regulated genes (ISG15, HSPB1, LHB, and SLC36A2) and one down-regulated gene (ESR1) to localize expression of the gene product within the endometrium (Fig. 4). No staining could be observed when the primary antibody was omitted and replaced with serum block solution or normal serum obtained from the species the primary antibody was raised in. In samples obtained from pregnant mares ISG15 localized to the superficial epithelium and endometrial stroma, while only slight staining for glandular epithelial cells could be observed. In cyclic mares, staining for ISG15 was considerably weaker and scattered throughout the endometrial stroma. LHB exclusively localized to superficial epithelial cells in pregnant mares, while the endometrium of nonpregnant mares showed weak to no staining for LHB. HSPB1 expression was observed in superficial epithelial cells and scattered throughout the endometrial stroma in pregnant mares, and staining intensity was noticeably weaker in samples obtained from cyclic mares. SLC36A2 localized to the apical membrane of superficial epithelial cells in pregnant mares. In nonpregnant mares, very weak to no staining for SLC36A2 was observed. In cyclic mares, ESR1 was most prominently expressed in superficial and glandular epithelial cells. In pregnant mares staining for ESR1 could be observed in superficial epithelial cells, while staining of the glandular epithelium was diminished compared to cyclic mares.

Fig. 4.

Immunohistochemistry. Endometrial tissue samples were obtained from mares at Day 13.5 of pregnancy (preg) or the luteal phase (NP) for the detection of ESR1, LHB, ISG15, HSPB1, and SLC36A2.

Discussion

The objective of the present study was to identify changes in gene expression of the equine endometrium due to the presence of a conceptus during the time period of maternal recognition of pregnancy. Endometrial tissue samples collected from pregnant and cyclic mares 13.5 days after ovulation were subjected to holistic gene expression analysis, that is, microarray analysis. When this work began, no equine specific microarray was available commercially so that we constructed a whole genome microarray based on publicly available sequences supplemented with expression data obtained by high-throughput sequencing. This approach is considered advantageous over using cross-species microarray hybridization.

Pregnant and nonpregnant samples from each of the four mares included in the experiment were paired for array hybridization so that each animal served as its own control. Using this experimental approach, we can potentially eliminate genetic variability as a factor that could alter the gene expression analysis results. We detected 106 up-regulated and 47 down-regulated genes in the endometrium of pregnant versus control animals. The main criterion for differential expression was a difference in signal intensity of 1.3-fold or more in all four mares included in the study, a signal-to-noise ratio of individual spots equal to or greater than 3.0, and a P-value ≤0.01. There is no general consensus on the statistical analysis of microarray data. Historically, an arbitrary fold-change cutoff of two has been used as criterion for selection of differentially expressed genes. In light of recent improvements in microarray technology (e.g., in-house spotted cDNA arrays versus computer-guided synthesized oligonucleotide arrays), this cutoff value has been lowered to as low as 1.3. In general, there is an inherent problem with this selection criterion, as genes that are expressed at low levels have a greater inherent error (i.e., variation) in their measured levels. These genes will then tend to numerically meet a fold change of two even if the gene is not truly differentially expressed, thereby producing false positives. In contrast, highly expressed genes, while having less error in their measured levels, may not meet an arbitrary 2-fold cutoff even when they are truly differentially expressed, thereby leading to an increased number of false negatives. The analysis of a composite tissue such as the endometrium adds to the complexity of microarray analysis, as the tissue is composed of various cell types, such as luminal and glandular epithelium and stromal cells.

In the current study, a fold-change cutoff of 1.3 was chosen for the following reasons: 1) endometrial tissue samples were collected early on in maternal recognition of pregnancy before the onset of structural luteolysis, and hence only subtle changes in gene expression were expected; 2) the raw data before any normalization procedures proved to be of good quality such that variations in gene expression levels due to technical influence are considered to be low; and 3) the moderate variation of expression ratios between pregnant and nonpregnant animals (mean coefficient of variation 41%) indicates consistent mechanisms of gene regulation. Quantitative PCR on a subset of selected genes confirmed expression data obtained by microarray supporting the results. The protein expression of five differentially expressed transcripts was assessed via immunohistochemistry. Results support mRNA expression data; that is, genes with higher mRNA expression in pregnant mares showed more intense staining pregnant than cyclic mares and vice versa. Existing literature was reviewed across species and tissues for evidence regarding regulation of differentially expressed transcripts by estrogen. Reviewing existing literature, for 45% of all genes identified within this study with known or inferred function, evidence for regulation by estrogen was found. Given the large quantities of estrogen synthesized by the equine conceptus, this finding supports the validity of the data.

Up-Regulated Genes

Not surprisingly, it appears that a majority of the up-regulated genes identified herein are involved in endometrial remodeling in response to a conceptus. Extracellular matrix turnover, stimulation of stromal cell proliferation, and stimulation of angiogenesis take place. Genes involved in cell-cell signaling could promote a synchronized uterine and embryonic development. Furthermore, appropriate delivery of nutrients to the preattachment conceptus may be ensured through the gene products thought to be involved in nutrient transport, such as SLC36A2, the transcript displaying one of the largest fold changes. GM2 ganglioside activator (GM2A) also is of particular interest. It was up-regulated 5.02-fold. GM2A is a major protein expressed by the equine trophoblast up until Day 18.5 of pregnancy [21], while its expression by uterine tissue has not been reported thus far. It has been suggested that GM2A has a lipid carrier role in the conceptus. We hypothesize that part of the GM2A protein detected in the early conceptuses is of uterine origin as has been described for P19 lipocalin (uterocalin) [22].

Heat Shock Proteins

Three heat shock proteins, namely, HSPB7, CRYAB, and HSPB8, all of which are known to be regulated by estrogens, were expressed at greater levels in pregnant mares. Heat shock proteins in general have been hypothesized to be part of the molecular repertoire that make the endometrium receptive to the conceptus during the window of implantation in humans [23], and we suggest that these proteins play a pivotal role in preparing the endometrium for conceptus fixation in the horse.

Putative Secretory Proteins

Secreted proteins make up a substantial portion of the gene products up-regulated in pregnant mares. The rate of production of secretory proteins during pregnancy profoundly affects pregnancy outcome in other species. Sheep that fail to form uterine glands do not maintain conceptuses beyond the period of maternal recognition of pregnancy [24]. Similarly, mares suffering from endometrosis, which is characterized by destruction of uterine glands and stromal fibrosis, have a lower reproductive efficiency. This outcome is attributed, in part, to disrupted protein secretion [25].

Another up-regulated gene of particular interest is LHB, which encodes the beta subunit of luteinizing hormone (LH) and choriogonadotropin [26]. Endometrial production and secretion of hCG during the progesterone-dominated phase of the menstrual cycle has been demonstrated in women [27] and is thought to modulate the uterine environment in preparation for implantation through exerting local effects such as promoting differentiation of stromal cells into decidual cells, suppressing cellular apoptosis, stimulating angiogenesis, and exerting immune-modulatory effects [27]. It is possible that the observed up-regulation of LHB leads to increased release of bioactive LH into the uterine environment during early pregnancy in the horse where it could exert paracrine effects preparing the uterus for conceptus implantation.

It also was interesting to see that the expression of tissue inhibitor of metalloproteinase 1 (TIMP1) was higher in the pregnant group. TIMPs regulate extracellular matrix turnover and tissue remodeling [28]. Temporal and spatially regulated expression of matrix metalloproteinases and their inhibitors are thought to be essential for endometrial tissue remodeling during implantation, and mice deficient in TIMP1 exhibit lower pregnancy rates [29]. We hypothesize that TIMP1 is one of the factors contributing to endometrial remodeling observed in equine pregnancy.

Insulin-like growth factors (IGFs) and their binding partners have received widespread attention for their role as regulators of embryonic and fetal development. Uterine expression of IGFBP1, which showed an 19.2-fold higher expression level during pregnancy in the current study, has been described in a number of species. In the ovine uterus, IGFBP1 mRNA expression is significantly up-regulated by Days 16–17 of pregnancy [30]. IGF1 expression has been described by both the conceptus and the endometrium in the horse [31]. As hypothesized in sheep, IGFBP1 might play role in trafficking IGFs between the endometrium and the uterine lumen in equine pregnancy supporting endometrial and conceptus growth.

The protein encoded by FGF9 is a member of the fibroblast growth factor (FGF) family. FGFs exert a variety of mitogenic and cell survival activities [32]. In human uterine tissue, FGF9 expression is enhanced by 17β-estradiol and induces proliferation of endometrial stromal cells [33]. During Day 14 of equine pregnancy, proliferation of all endometrial layers is present as observed by Ki-67 immunohistochemical staining (L.A. Silva, personal communication), implicating a potential role for FGF9 or others mitogens in endometrial remodeling.

Cell-Cell Signaling

Cell-cell signaling refers to any process that mediates the transfer of information from one cell to another. Given the importance of a synchronized uterine and embryonic development, the authors propose that up-regulation of genes belonging to this functional Category are pivotal to successful establishment of pregnancy. Stanniocalcin 1 (STC1) is one of the up-regulated genes involved in cell-cell signaling, and pregnancy-specific uterine expression of STC1 has been described in mice [34], sheep [35], and pigs [36]. In pigs, STC1 has been identified as luminal epithelial marker for implantation. STC1 expression in the equine endometrium has not yet been reported, but the large number of species this gene can be in found during pregnancy implies a species-independent role in pregnancy.

Transport

Solute carrier family 36 (proton/amino acid symporter), member 2, SLC36A2, is the transcript exhibiting the largest fold change in the current study (15.46-fold). To our knowledge, its expression in the uterus has not been described. SLC36A2, also known as PAT2, mediates the transport of amino and fatty acids [37]. Amino acid transport is critical to early embryo development. In the mouse, even a brief 5-min exposure of fertilized oocytes to medium without amino acids is detrimental to embryonic development [38]. The marked up-regulation of this proton/amino acid symporter is hypothesized to reflect the increased need of the rapidly developing equine conceptus for nutrients.

Down-Regulated Genes

Among the down-regulated genes, estrogen receptor 1 was of particular interest because of its potential involvement in the initiation of luteolysis in cyclic mares. Down-regulated transcripts indicate that invasion of the conceptus is counteracted, which is in accordance with the prolonged preimplantation period of the equine embryo.

Microarray analysis revealed endometrial samples collected from pregnant mares to have 1.6-fold lower mRNA levels for estrogen receptor 1 (ESR1) than samples collected from nonpregnant mares. This observation could be confirmed using real-time RT-PCR. Estrogens and their receptors play a central role in luteal regression in ruminants via the up-regulation of endometrial oxytocin receptors during the expected time of luteolysis, when oxytocin initiates pulsatile release of PGF_2alpha through binding to its receptor [39]. In sheep, IFNT inhibits cyclic up-regulation of the estrogen receptor during early gestation, thereby preventing up-regulation of oxytocin receptors, an event central to inhibition of luteolysis [40]. Oxytocin plays a central role in luteal regression in mares, and a reduced ability of the endometrium to secrete prostaglandin in response to oxytocin can be observed during early pregnancy. In accordance with this decreased oxytocin responsiveness, oxytocin receptors are reduced in both number and their affinity in pregnant mares [10]. Lower expression of ESR1 in pregnant mares has been reported at Day 15 of pregnancy [41]. This is the first description of estrogen receptor down-regulation as early as Day 13.5 of equine pregnancy. The authors hypothesize that a similar relationship between decreased expression of ESR1 and subsequent decreased expression of oxytocin receptor exists in the mare as described in ruminants. Estrogen can autoregulate the expression of its own receptor, both positively and negatively depending on the cell type examined [42–44]. For endometrial tissue, both decreased [44] and increased [45] expression of ESR1 after estrogen treatment had been described depending on the species examined. In face of the large quantities of estrogen synthesized by the equine conceptus and the observation that estrogen negatively regulates the expression of its own receptor in various tissues including uterine tissue, one might hypothesize that conceptus-derived estrogens are the most likely cause of the observed down-regulation of ESR1 in pregnant mares.

Counteraction of Trophoblast Invasion

The equine conceptus is unusual in exhibiting a prolonged preattachment phase of intrauterine development with implantation initiated around Day 40 of pregnancy [46]. The horse is further considered a noninvasive species regarding endometrium-trophoblast interaction, so that invasion of the trophoblast, except for the chorionic girdle cells, is prevented. In the current study, a 3.0-fold down-regulation of cathepsin L and a 1.8-fold down-regulation of plasminogen activator is suggestive of such a counteraction of trophoblast invasion. Cathpesin L is a lysosomal proteinase involved in cellular protein catabolism, primarily collagen and elastin. Cathepsins, in general, are implicated in implantation in mice [47], and a pregnancy-specific up-regulation of cathepsin L mRNA also exists in the ovid [48]. Plasminogen is an inactive enzyme precursor that is converted to the active enzyme plasmin through cleavage by plasminogen activator. Plasmin can degrade a variety of proteins, possibly including components of the extracellular matrix [49]. Pig blastocysts secrete plasminogen activator during early pregnancy, whereas the endometrium secretes plasminogen inhibitor. The concomitant secretion of plasminogen inhibitor is thought to prevent a proteolytic cascade initiated by the blastocysts plasminogen activator [50]. The down-regulation of plasminogen precursor observed in the equine endometrium during early pregnancy in the current study could be interpreted similarly. To prevent erosion of the endometrium, the protease precursor plasminogen is down-regulated to reduce substrate for plasminogen activator potentially secreted by the equine conceptus.

Miscellaneous Down-Regulated Genes

Connective tissue growth factor, CTGF, acts to promote fibroblast proliferation, and migration, adhesion, along with extracellular matrix formation and remodeling [51]. In corpus luteum tissue, there is evidence that hCG negatively regulates CTGF through paracrine signaling [52]. As discussed previously, there may be an increased release of bioactive LH into the uterine environment during early pregnancy in the horse, potentially accounting for the observed down-regulation of CTGF.

A 1.75-fold down-regulation of Gap junction protein, alpha 1, 43 kDa, GJA1, was observed in the present study. Gap junctions mediate intercellular communication and are hypothesized to participate in preparation of the uterus for implantation. In sheep, GJA1 expression of endometrial stromal cells is enhanced during implantation [53]. In mice, in which Gja1 expression is markedly increased in stromal cells surrounding the implanting embryo, conditional deletion of Gja1 results in arrest of embryo growth and early pregnancy loss [54]. In the current study, abundance of GJA1 was lower during early pregnancy than during the corresponding day of the estrous cycle, which differs from expression levels specific to other species. The observed down-regulation of GJA1 might reflect the delayed implantation of the equine conceptus.

To conclude, the results give insight into the changes occurring at the mRNA level during the intense embryo-maternal dialogue at Day 13.5 of gestation. To our knowledge, this is the first report of its kind in the equine species. Several of the genes identified in the current study are known to play a role in early pregnancy in species other than the horse. We thus hypothesize that a subset of genes crucial to endometrial receptivity is conserved across species, representing a common pattern during early pregnancy, whereas each species has a distinct mechanism ensuring ongoing corpus luteum function. The identification of several new genes that may be important for endometrial preparation with respect to successful outcome of pregnancy provides many new starting points for more detailed investigations. Future studies will investigate earlier stages of pregnancy using a similar approach to provide a comprehensive view of the dynamic transcriptome changes underlying maternal pregnancy recognition.

Author notes

¹Supported by the Grayson-Jockey Club Research Foundation.