Microarray analysis of the primate luteal transcriptome during chorionic gonadotrophin administration simulating early pregnancy

Abstract

To explore chorionic gonadotrophin (CG)-regulated gene expression in the primate corpus luteum (CL), adult female rhesus macaques were treated in a model of simulated early pregnancy (SEP). Total RNA was isolated from individual CL after specific intervals of exposure (1, 3, 6 and 9 days) to recombinant hCG in vivo and hybridized to Affymetrix™ GeneChip Rhesus Macaque Genome Arrays. The mRNA levels of 1192 transcripts changed ≥2-fold [one-way ANOVA, false discovery rate (FDR) correction; P< 0.05] during SEP when compared with Day 10 untreated controls. Real-time PCR validation indicated that 15 of 17 genes matched in expression pattern between PCR and microarray. Protein levels of three genes identified as CG-sensitive, CYP19A1 (aromatase), PGRMC1 (progestin-binding protein) and STAR (steroidogenic acute regulatory protein) were quantified by western blot analysis. To further analyze global changes in gene expression induced by CG exposure, luteal gene expression was compared between SEP (rescued) and regressing CL, utilizing previously banked GeneChip data from the luteal phase of the menstrual cycle. Expression patterns and mRNA levels were analyzed between time-matched intervals. Transcripts for 7677 mRNAs differed in expression patterns ≥2-fold (one-way ANOVA, FDR correction; P< 0.05) between the hCG-exposed (SEP) CL and regressing CL. Regressed CL (at menses) were most unlike all other CL. Pathway analysis of significantly affected transcripts was performed; the pathway most impacted by CG exposure was steroid biosynthesis. Further comparisons of the genome-wide changes in luteal gene expression during CG rescue and luteolysis in the natural menstrual cycle should identify additional key regulatory pathways promoting primate fertility.

Introduction

During fertile cycles, timely regression of luteal structure–function in many primates is prevented by the LH-like molecule chorionic gonadotrophin (CG) secreted by the primate conceptus beginning at implantation (Jarvela et al., 2008). CG exposure stimulates additional progesterone (P), estrogen (E) and relaxin secretion by the corpus luteum (CL) (Ghosh et al., 1997). However, CG rescue of the primate CL is also transient, prolonging luteal P secretion in rhesus macaques for ∼2 additional weeks until the luteal–placental shift, after which the placenta then produces the P needed to sustain pregnancy to term (Atkinson et al., 1975).

To investigate the effects of CG on luteal tissue, a non-human primate model simulating early pregnancy was developed by treating female rhesus monkeys with increasing dosages of hCG mimicking the rise detected during early pregnancy (Wilks and Noble, 1983; Ottobre et al., 1984; Duffy et al., 1996). The rhesus macaque model of simulated early pregnancy (SEP) results in a similar pattern of systemic E, P and relaxin secretion to that in early pregnancy (Duffy et al., 1996). It is also a useful model to examine the role of CG-R signaling in the primate CL used by primatologists to obtain tissue without terminating early pregnancy (Duffy and Stouffer, 1997). It was determined that CL during the mid-late luteal phase are most responsive to CG stimulation (Wilks and Noble, 1983). Ottobre et al. (1984) also noted that available (but not total) luteal-binding sites for CG decrease during prolonged exposure to CG. It was later determined that the receptor (LHCGR) was uncoupled from adenylate cyclase signaling after prolonged exposure to CG (VandeVoort et al., 1988), which precedes the observed decrease in luteal P response to CG. Similar responses to exogenous treatment with escalating dosages of hCG are also noted in women (Illingworth et al., 1990).

The effects of CG-stimulated LHCGR signaling and action in the primate CL have been investigated in limited studies mostly focused on single genes and gene products (Duffy and Stouffer, 1997; Sanders and Stouffer, 1997; Sugino et al., 2000; Duncan et al., 2005). However, recent advances in microarray technology and the development of the Affymetrix™ GeneChip^® Rhesus Macaque Genome Array allow investigation of global changes in gene expression in this clinically relevant primate model (Bogan et al., 2008b, 2009, Bishop et al., 2009a, 2011b; Priyanka et al., 2009; Xu et al., 2011). Building on past experiments investigating regulation of the luteal transcriptome by LH during the menstrual cycle (Bishop et al., 2009a, 2011b), the primary goal of the current investigation was to identify changes in gene expression induced by CG in macaque CL during SEP using the rhesus gene array (Duffy and Stouffer, 1997). Changes in gene expression are also compared between CL rescued during SEP and time-matched CL undergoing regression in non-gravid cycles (Bogan et al., 2008b).

Materials and Methods

Experiments were performed at the Oregon National Primate Research Center (ONPRC), West Campus of Oregon Health & Science University (OHSU). All protocols were approved by the ONPRC/OHSU Institutional Animal Care and Use Committee (IACUC), in accordance with the National Institutes of Health (NIH) Guidelines for Care and Use of Laboratory Animals. Adult, female rhesus macaques (Macaca mulatta, n= 20) exhibiting regular menstrual cycles were selected for this study and were under the direct care of the ONPRC Division of Animal Resources (DAR) as described (Sitzmann et al., 2010). All surgical procedures (aseptic laparotomy under anesthesia) were performed as previously described (Duffy et al., 2000) by ONPRC DAR Surgical Staff.

Simulated early pregnancy

The SEP protocol was performed as described previously (Duffy and Stouffer, 1997). Briefly, serum E and P levels were monitored daily beginning 6 days after the onset of menstrual bleeding. The first day of the luteal phase (Day 1; ovulation) was defined as the first day of low E (<100 pg/ml) following the mid-cycle E surge, with a coincident rise in serum P above 0.2 ng/ml (Bishop et al., 2009b). SEP treatment began on Day 9 of the luteal phase (Duffy and Stouffer, 1997). Recombinant hCG (Novarel™, Ferring Pharmaceuticals Inc., Parsippany, NJ, USA) was administered in increasing dosages (15, 30, 45, 90, 180, 360, 720, 1440 and 2880 IU) twice daily by i.m. injection. Individual CL were collected on Days 10, 12, 15 and 18 of the luteal phase, representing 1, 3, 6 and 9 days of hCG treatment (n= 4 CL/day). Additionally, CL were collected from untreated monkeys on Day 10 of the luteal phase to serve as baseline controls. All CL were dissected from ovarian tissue, snap-frozen in liquid nitrogen and stored at −80°C until TRIzol^® extraction of total RNA and protein (Invitrogen, Carlsbad, CA, USA) according to manufacturer's protocols.

Microarray analyses of mRNA from SEP CL

Before microarray hybridization, RNA was further purified using an RNAeasy kit (Qiagen, Valencia, CA, USA). Total RNA integrity was assessed by the RNA 6000 LabChip using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Aliquots of RNA from individual CL were hybridized to Affymetrix™ GeneChip^® Rhesus Macaque Genome Arrays by the OHSU Affymetrix Microarray Core as reported by Bishop et al. (2009a) and Bogan et al. (2008b). For the complete hybridization protocol, see http://www.ohsu.edu/xd/research/research-cores/gmsr/services/affymetrix-microarray-core/upload/AMC-3-IVT-Expression-Assay-Methods.pdf. Processed image (.CEL) files from each array are deposited in NCBI GEO data sets (http://www.ncbi.nlm.nih.gov/gds, Series GSE25335). These files were uploaded into the web-based GeneSifter^® software (Geospiza, Inc., Seattle, WA, USA) for analysis. Probeset hybridization data were log₂-transformed using the Robust Multi-array Average (RMA) algorithm (Noriega et al., 2009). Probesets on the Rhesus GeneChip^® represent individual rhesus macaque gene (mRNA) transcripts (see Noriega et al., 2009, for further details on the Rhesus GeneChip^®). Significant changes in mRNA transcript levels were defined as a ≥2-fold change after false discovery rate (FDR) correction (Benjamini and Hochberg, P< 0.05). All mRNA transcripts were interrogated by one-way ANOVA for every SEP time-point versus Day 10 controls. Pairwise comparisons (Welch's t-test) were performed between Day 10 control CL versus Day 10 SEP, also Day 10 versus Day 12, Day 12 versus Day 15 and Day 15 versus Day 18 of SEP.

Real-time PCR validation of select gene products

Several gene products (mRNAs; n= 17) were chosen for real-time PCR validation. Aliquots of RNA from SEP CL were rendered DNA-free by treatment with TURBO™ DNase and reverse-transcribed using a RETROscript^® kit (both by Applied Biosystems/Ambion, Inc., Austin, TX, USA) to generate cDNA. Real-time PCR for various gene products (Supplementary data, Table SI: Probes and primers) was performed with TaqMan^® MGB probes (Applied Biosystems/Life Technologies Inc., Foster City, CA, USA) as described by Bogan and Hennebold (2010). Quantified PCR data were analyzed by one-way ANOVA and log₂-transformed if needed to correct for heterogeneity of variances (SAS^® version 9.2, SAS Institute Inc., Cary, NC, USA).

Additional validation of mRNA for RLN1, RLN3 and the receptor, LGR7, was performed. Primers specific to the rhesus macaque sequences of RLN3 and LGR7 were designed by use of the Primer-BLAST program (http://www.ncbi.nlm.nih.gov/tools/primer-blast/; Table I). Primers designed as described for real-time PCR gene expression analysis were utilized for validation of relaxin 1 (RLN1). PCR on amplified cDNA was performed with GoTaq^® Green Master Mix (Promega Corp., Madison, WI, USA) according to the manufacturer's protocols (primer annealing temperatures RL1: 53°C, 30 cycles; RLN3: 55°C, 40 cycles; LGR7: 55°C, 40 cycles). Products were subjected to agarose gel electrophoresis and bands visualized with SYBR^® Safe DNA gel stain (Invitrogen/Life Technologies Inc., Carlsbad, CA, USA). To confirm amplification of the correct rhesus macaque LGR7 sequence, the band representing amplified LGR7 was excised from the gel, and cDNA was purified by use of QIAquick^© PCR purification kit (Qiagen Inc.). This was then sequenced to confirm amplification of LGR7 by use of a 96-capillary ABI3700XL (Applied Biosystems), and resulting sequences were analyzed by Sequence Scanner v 1.0 software (Applied Biosystems) and compared with rhesus macaque sequences in GenBank by nucleotide BLAST (blastn) alignment (http://blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=blastn&BLAST_PROGRAMS=megaBlast&PAGE_TYPE=BlastSearch&SHOW_DEFAULTS=on&LINK_LOC=blasthome).

Table I.

Pairwise analyses of Rhesus GeneChip^© Microarray Data.*

Simulated early pregnancy (SEP)
Day 10	292↑a	Day 10 + hCG
	127↓
Day 10 + hCG	71↑	Day 12 + hCG
	191↓
Day 12 + hCG	10↑	Day 15 + hCG
	21↓
Day 15 + hCG	25↑	Day 18 + hCG
	30↓
Rescued (SEP) versus regressing CL
Day 10	292↑	Day 10 + hCG
	127↓
Day 12	1543↑	Day 12 + hCG
	440↓
Days 14–16	2078↑	Day 15 + hCG
	452↓
Days 18–19	3980↑	Day 18 + hCG
	1581↓

^aArrows indicate up ↑ or down ↓ regulation of mRNA.

*Analyses represent pre-planned comparisons only. Significant changes determined by Welch's t-test with Benjamini and Hochberg FDR correction P> 0.05, expression changes between groups ≥2-fold.

Correlation between mRNA and protein expression

To compare the patterns of mRNA expression to those of their respective proteins, CYP19A1 (aromatase), STAR (steroidogenic acute regulatory protein) and PGRMC1 (progestin-binding protein) were analyzed from protein isolated from individual CL by western blotting as reported by Bishop et al. (2009a) (Supplementary data, Table SII: Antibodies). Proteins were transferred onto PVDF membranes by iBlot^® Dry Blotting System (iBlot^® device and iBlot^® Transfer Stack, Invitrogen) and bands of interest were visualized using the WesternDot™ 625 Goat Anti-Rabbit Western Blot Kit (Invitrogen) according to the manufacturer's protocols. Bands were quantified by ImageJ Software (Abramoff et al., 2004). Protein expression levels were quantified by calculating the area under the curve (AUC) of peak band intensity and dividing by loading control (GAPDH) AUC, after eliminating background signal (ImageJ website: http://rsb.info.nih.gov/ij/docs/menus/analyze.html#gels). All data were analyzed by one-way ANOVA (SAS^® version 9.2).

Comparison of mRNA levels in CL of SEP with those in time-matched CL from the late luteal phase of the non-fecund menstrual cycle

Microarray data from macaque CL collected at specific stages during the luteal phase of natural menstrual cycles (Bogan et al., 2008b), also generated at ONPRC using Affymetrix™ GeneChip^® Rhesus Macaque Genome Arrays, were downloaded from NCBI GEO data sets (http://www.ncbi.nlm.nih.gov/gds, Series GSE10367). These arrays included probeset hybridization data that are time-matched to the interval of SEP in non-fecund cycles, i.e. CL from Days 10 to 12 (mid-late), 14 to 16 (late) or 18 to 19 (very-late) of the luteal phase (n= 4 CL/luteal day). Processed .CEL files were uploaded into GeneSifter with SEP .CEL files. RMA-transformed transcripts from regressing CL (luteal days 10–19; Group 1) were compared with those from SEP luteal rescue (Days 10–18 + hCG; Group 2). Previous microarray comparisons indicated that probeset data from Day 10 to 12 CL in these experiments are virtually indistinguishable from control day 12 rhesus CL data from Bishop et al. (2009a) (unpublished data). Therefore, untreated day 10 SEP control CL were added to Group 1 and Day 10–12 CL from Series GSE10367 (Bogan et al., 2008b) denoted as Day 12 only for clarity.

Levels of mRNA (probeset) transcripts were interrogated by one-way ANOVA to identify significant changes in expression patterns between these two groups (≥2-fold, Benjamini and Hochberg FDR correction, P< 0.05). Pairwise comparisons between time-matched rescued (SEP) and regressing CL were performed (Welch's t-test with FDR correction) to identify significant (≥2-fold) changes in mRNA transcript levels induced by hCG treatment.

Results

Changes in mRNA levels in CL during SEP

Patterns of steroid hormones (P and E) presented in Fig. 1A and B demonstrate an acute (24-h) increase in P and E after the onset of hCG exposure on luteal day 9. By the first luteal collection on Day 10, rescued CL (SEP) were associated with increased serum P and E levels compared with those from controls. By the last CL collection on Day 15, E levels remained elevated, but a decline in serum P to pretreatment levels was observed as described previously (Duffy et al., 1996).

Figure 1.

(A and B) Serum P (A) and estradiol (B) levels during SEP (Day 10, n= 16; Day 12, n= 12; Day 15, n= 8; and Day 18, n= 4) and untreated control (n= 4) females. Treatment interval for hCG is indicated by black bar at the bottom of graphs. P levels transiently and estradiol levels gradually rise during SEP treatment (one-way ANOVA repeated-measures P< 0.001); means with different lowercase letters differ (P< 0.05), within group over time. P levels between SEP and control groups differ on luteal day 10 and estradiol levels are significantly different on luteal day 9, as depicted by means with different upper case letters (P< 0.05). Arrows indicate days of CL collection.

GeneSifter analysis of all hCG-treated CL compared with Day 10 untreated CL identified 1192 mRNA transcripts whose levels changed ≥2-fold during the SEP protocol (one-way ANOVA, Benjamini and Hochberg FDR correction, P< 0.05; Supplementary data, File S1). A heat map of these changes is presented in Fig. 2A. Five major patterns were identified based on changes to observed mRNA expression on this heat map. GeneSifter k-medoids clustering was then performed for five patterns of gene expression to determine the number of mRNAs in each pattern (Fig. 2B) and examples of these patterns are noted in Fig. 2A: transcripts abruptly decreasing expression after 1 (Pattern 4) or 3 days (Pattern 2) of hCG treatment; those gradually increasing (Pattern 3) or decreasing (Pattern 1) throughout hCG treatment; and several transiently increasing (Pattern 5) following 1 day of CG exposure. Hierarchical clustering of treatment groups indicated that Days 10 and 10 + hCG are very dissimilar and show diverged gene expression from the other treatment groups (Fig. 2C).

Figure 2.

(A) Heat map of all significant changes in mRNA transcript levels in macaque CL during SEP. Lanes: 1, Day 10 control; 2, Day 10 + hCG; 3, Day 12 + hCG; 4, Day 15 + hCG; 5, Day 18 + hCG (hCG treatment designated by grey bar). Bold lines and numbers indicate borders between different expression patterns of transcripts (see text for detailed explanation). Legend indicates color representation of increases (+1 log₂) and decreases (−1 log₂) in expression. (B) K-medoids cluster analysis based on observed five patterns of expression and total numbers of probesets (mRNAs) that correspond to each cluster. Numbers on x-axis correspond to lanes denoted in (A). The mean silhouette width corresponds to error associated with each cluster. (C) Hierarchical clustering of all SEP treatment groups. Relationship is depicted not only by branching but also by the length of branches as indicated by the arrow below the dendrogram.

Pairwise analyses (Table I; Supplementary data, Files S2–S5) revealed many mRNA transcripts that display rapid responses to CG exposure. The majority of significant (≥2-fold) changes in mRNA levels during SEP occur between Day 10 versus Day 10 + hCG and Day 10 + hCG verses Day 12 + hCG. Fewer significant changes in mRNA levels are observed after additional CG exposure, e.g. Day 12 versus Day 15 and Day 15 versus Day 18 SEP, although Fig. 2A and B suggests that the effects of CG are sustained for many mRNAs throughout SEP.

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of significantly impacted mRNA transcripts identified biological themes associated with CG exposure during SEP (Table II). The most affected KEGG pathway is steroid biosynthesis (Table III), and many other affected KEGG pathways are related to steroid hormone production. Most of the transcripts in the steroid biosynthesis pathway are up-regulated by CG exposure, either acutely or chronically.

Table II.

KEGG pathway analyses of luteal genes changing in expression during SEP.^a

KEGG pathways all SEP	List	Gene set	z-score
Steroid biosynthesis	10	15	10.96
Biosynthesis of alkaloids derived from terpenoid and polyketide	14	71	5.73
Biosynthesis of terpenoids and steroids	11	66	4.37
γ-Hexachlorocyclohexane degradation	5	19	4.27
Nicotinate and nicotinamide metabolism	5	22	3.82
Biosynthesis of plant hormones	11	83	3.48
p53 signaling pathway	9	63	3.4
Cell cycle	13	114	3.17
Androgen and estrogen metabolism	5	31	2.85
ECM–receptor interaction	9	77	2.71
PPAR signaling pathway	6	57	1.92
Metabolism of xenobiotics by cytochrome P450	6	58	1.87
Cell adhesion molecules (CAMs)	11	132	1.78
Metabolic pathways	55	903	1.69
Complement and coagulation cascades	5	54	1.44
MAPK signaling pathway	8	234	−1.15
Calcium signaling pathway	5	168	−1.24
Pathways in cancer	10	297	−1.35
Regulation of actin cytoskeleton	5	187	−1.5
Cytokine–cytokine receptor interaction	6	222	−1.62

^aEdited by #transcripts in list ≥5 and z-score >1.

Table III.

Steroid biosynthesis.

Acute effect of CGa	Chronica	Gene title	Gene ID
None	Down	24-dehydrocholesterol reductase	DHCR24
Up	Up	7-dehydrocholesterol reductase	DHCR7
Up	Up	Emopamil-binding protein (sterol isomerase)	EBP
Up	Up	Farnesyl-diphosphate farnesyltransferase 1	FDFT1
Up	Up	Hydroxysteroid (17-beta) dehydrogenase 7	HSD17B7
Up	Up	Lanosterol synthase (2,3-oxidosqualene-lanosterol cyclase)	LSS
None	Down	Similar to 24-dehydrocholesterol reductase precursor	LOC717022
Up	Up	Squalene epoxidase	SQLEb
Up	Up	Sterol-C4-methyl oxidase-like	SC4MOL
Up	Up	Sterol-C5-desaturase (ERG3 delta-5-desaturase homolog, S. cerevisiae)-like	SC5DL

^aAcute: Day 10, Chronic: Days 12–18.

^b2/3 Affymetrix probesets for SQLE showed an acute effect of CG.

Of the 17 genes chosen for validation of mRNA changes, the expression patterns for 15 transcripts (probesets) were similar in the microarray and real-time PCR assays (Supplementary data, Fig. S1). Some of the validated genes exhibiting acute changes in gene expression (Day 10 + hCG) are depicted in Fig. 3A and B (HSD3β2 and HSD11β1). Others showed chronic/delayed effects of hCG (≥Day 12 + hCG) (e.g. CYP11A1 and HSD11β2; Fig. 3C and D). RLN1 demonstrated the most robust change in mRNA levels of all validated genes (Fig. 4A). This is the major isoform of RLN present in the rhesus CL, as confirmed by reverse transcription (RT)–PCR (Fig. 4B). The mRNA for the relaxin receptor RXFP1 (LGR7) was also detected by RT–PCR and confirmed by sequencing in rhesus CL (Fig. 4B, sequencing data not shown). However, microarray analysis indicated that mRNA levels for both RLN3 and RXFP1 did not significantly vary with CG treatment (data not shown).

Figure 3.

Examples of validated gene expression in macaque CL during SEP. For clarity, only statistical differences associated with real-time PCR values are depicted (means with different letters differ P< 0.05). Hydroxysteroid dehydrogenases, HSD3β2 (A) and HSD11β1 (B), both demonstrate rapid increases in mRNA levels, compared with controls (D10) after 1 day of hCG exposure, which then decline to or below control levels by 9 days of hCG exposure. Aromatase CYP11A1 (C) and HSD11β2 (D) demonstrate delayed (≥Day 12 + hCG) responses to hCG treatment during SEP. Note that statistically significant changes in mRNA levels occur on or after 3 days of hCG treatment. For all of these genes, the pattern of expression between real-time PCR and microarray are similar.

Figure 4.

(A) Changes in transcript levels for RLN1 in macaque CL during SEP (one-way ANOVA P< 0.001). Significant differences in gene expression were detected after 3 days of hCG treatment (Day 12). (B) RT–PCR detected mRNAs for rhesus RLN1, RLN3 and the relaxin receptor RXFP1 (LGR7). Band for RXFP1 (LGR7) was confirmed by sequencing analysis (data not shown).

Comparison of mRNA and protein levels for selected gene products

Analysis of STAR gene products demonstrated that mRNA levels did not increase until 3 days after CG (Day 12 + hCG), whereas protein levels were elevated after 1 day of CG (Day 10 + hCG; Fig. 5A). Conversely, CYP19A1 (aromatase) mRNA levels increased after 1 day of hCG treatment (Fig. 5B, Day 10 + hCG), but protein levels did not increase until 3 days of treatment. A transient rise in mRNA and protein levels was detected on Day 10 + hCG for the progestin-binding protein PGRMC1, followed by a significant decrease between days 10 + hCG and 12 + hCG (Fig. 5C). Thereafter, mRNA levels increased on Days 15 + hCG and 18 + hCG, while protein levels remain suppressed and comparable to Day 10 control CL.

Figure 5.

Comparison between mRNA and protein expression patterns for steroidogenic acute regulatory protein (STAR) (A), aromatase (CYP19A1) (B) and progestin-binding protein (PGRMC1) (C). Significant differences are depicted for real-time PCR data (different lower case letters, P< 0.05) and protein data from western blot analyses normalized to GAPDH controls (different upper case letters, P< 0.05).

Comparison of mRNA expression levels in time-matched CL during SEP versus regressing CL in the non-fecund cycle

Analyses of mRNA expression were compared between CL during SEP and CL obtained during a non-fecund cycle to assess CG effects on stage-matched tissue. Serum P levels (ng/ml) during CL regression in the late luteal phase of non-fecund cycles averaged 4.4 ± 0.6 (Day 12), 5.1 ± 1.5 (Days 14–16) and 0.2 ± 0.1 (Days 18 and 19; Bogan et al., unpublished data), well below those of time-matched stages in hCG-treated cycles (Fig. 1A). Patterns of 7677 mRNA transcripts differed significantly by ≥2-fold between time-matched CL from the mid-late to very-late luteal phase and SEP (Supplementary data, File S6). A heat map of these data is depicted in Fig. 6A. Comparing patterns of luteal gene expression between regressing and rescued (SEP) CL, two primary effects of CG are apparent: mRNA levels that decrease during luteolysis are maintained by SEP protocols similar to those observed in functional CL (Fig. 6A, Pattern 1); conversely, mRNA transcripts whose levels increase during luteolysis are typically suppressed by hCG treatment (Fig. 6A, Pattern 2). Clustering of samples (Fig. 6B) revealed that Day 10 CL (in the presence and absence of hCG) are again somewhat similar. After Day 10, hCG-treated CL are unlike those of time-matched controls. Regressed CL (Days 18 and 19) differ from all other samples. Pairwise analyses (Table I; Supplementary data, Files S7–S9) demonstrate increasing divergence of mRNA levels for a number of genes between CL under the influence of tonic LH (regressing CL) versus CG (SEP CL) as the luteal phase progresses.

Figure 6.

(A) Heat map of all significant differences in mRNA transcript levels in time-matched CL during regression in the menstrual cycle and SEP. Groups are indicated by brackets. Lanes: Day 10, Day 12, Days 14–16 and Days 18–19 of the luteal phase; Day 10 + hCG, Day 12 + hCG, Day 15 + hCG and Day 18 + hCG in SEP. Two different patterns of gene expression were observed; a bar denotes the border between patterns (see text for explanation). (B) Hierarchical clustering of all treatment groups of regressing and rescued CL. Relationship is depicted not only by branching but also by the length of branches as indicated by the arrow below the dendrogram.

KEGG pathway analysis of affected mRNA transcripts (Table IV) indicates that genes in the steroid biosynthesis category were most likely to be differentially expressed between rescued (SEP) and regressing CL. Expression levels of many mRNA transcripts in this pathway (Table V) were down-regulated during regression, but are either up-regulated or maintained at high levels by hCG exposure. Additional impacted pathways involved in regulation of steroid hormone expression include biosynthesis of plant hormones, androgen and E metabolism, and biosynthesis of terpenoids and steroids. Other pathways implicate changes to genes associated with the immune system (i.e. natural killer cell-mediated cytotoxicity, graft-versus-host disease and asthma) in CL during SEP. Other significant pathways re-enforce the importance of cell death and apoptosis pathways (i.e. Parkinson's disease), as key components of primate luteolysis, and therefore suppression/disruption of these processes, are possibly critical for successful luteal rescue (Dickinson et al., 2008; Peluffo et al., 2009).

Table IV.

KEGG pathways of luteal genes differing in expression between SEP and regression.^a

KEGG pathway rescue versus regression	List	Gene set	z-score	List/gene set (%)
Steroid biosynthesis	11	15	3.93	73
Lysosome	45	102	3.7	44
Amino sugar and nucleotide sugar metabolism	19	34	3.65	56
Valine, leucine and isoleucine degradation	21	39	3.63	54
Fc gamma R-mediated phagocytosis	35	79	3.28	44
Biosynthesis of plant hormones	35	83	2.93	42
γ-Hexachlorocyclohexane degradation	11	19	2.92	58
Terpenoid backbone biosynthesis	9	15	2.78	60
Biosynthesis of alkaloids derived from terpenoid and polyketide	30	71	2.72	42
Focal adhesion	67	184	2.64	36
Pathogenic Escherichia coli infection—EHEC	17	37	2.46	46
Natural killer cell-mediated cytotoxicity	45	119	2.45	38
p53 signaling pathway	26	63	2.39	41
Biosynthesis of terpenoids and steroids	27	66	2.38	41
Type I diabetes mellitus	20	46	2.37	43
Allograft rejection	18	41	2.3	44
Jak-STAT signaling pathway	47	128	2.26	37
Gap junction	29	74	2.19	39
Autoimmune thyroid disease	20	48	2.14	42
PPAR signaling pathway	23	57	2.11	40
Graft-versus-host disease	17	40	2.07	43
Asthma	13	29	2.04	45
Aminoacyl-tRNA biosynthesis	6	41	−1.9	15
Nucleotide excision repair	6	41	−1.9	15
ABC transporters	4	34	−2.11	12
Pyrimidine metabolism	13	83	−2.51	16
Cardiac muscle contraction	9	65	−2.55	14
Proteasome	5	46	−2.59	11
Alzheimer's disease	29	163	−2.93	18
Oxidative phosphorylation	17	117	−3.27	15
Huntington's disease	28	171	−3.43	16
Parkinson's disease	16	119	−3.57	13
Neuroactive ligand–receptor interaction	36	232	−4.33	16

^az-score ≥1.9, with listed genes >10% of total gene set.

Table V.

Steroid biosynthesis.

Regression	SEP rescuea	Gene title
Down	Up	7-dehydrocholesterol reductase
Down	Up	Squalene epoxidase
Down	Up	Farnesyl-diphosphate farnesyltransferase 1
Down	Maintain	NAD(P) dependent steroid dehydrogenase-like
Down	Maintain	Emopamil-binding protein (sterol isomerase)
Down	Maintain	Sterol-C5-desaturase (ERG3 delta-5-desaturase homolog, S. cerevisiae)-like
Down	Maintain	Hydroxysteroid (17-β) dehydrogenase 7
Down	Up	Farnesyl-diphosphate farnesyltransferase 1
Down	Maintain	Transmembrane 7 superfamily member 2
Down	Up	Squalene epoxidase
Down	Maintain	Sterol-C4-methyl oxidase-like
Down	Maintain	Similar to 24-dehydrocholesterol reductase precursor
Down	Maintain	24-dehydrocholesterol reductase

^aMaintain = levels similar to Day 10 CL.

Discussion

Analysis of gene transcripts in the macaque CL during SEP indicates that rising hCG levels have early rapid effects on mRNA levels of several gene cohorts between 1 and 3 days of hCG treatment (Days 10 and 12 + hCG groups; Table I). The rapid increase closely parallels the pattern of P secretion in rescued CL (Fig. 1A). It is tempting to consider the rapid, transient changes in gene expression by CG exposure as critical for maintaining and enhancing P secretion. The most impacted KEGG pathway, steroid biosynthesis, demonstrates that the most significantly expressed steroid hormone-associated genes are rapidly up-regulated by CG exposure (Table III). The transcript for the major enzyme catalyzing P biosynthesis, HSD3β2, is rapidly up-regulated by CG (Fig. 3A). It is important to note that not all rhesus macaque transcripts are fully annotated to KEGG pathways (Bishop et al., 2011b). The macaque probeset for HSD3β2 is not included in the KEGG gene list under steroid biosynthesis due to poor annotation of this transcript. Similar rapid up-regulation of HSD3β2 by CG treatment is reported in macaque luteinized granulosa cells within 24 h of exposure (Puttabyatappa et al., 2010).

However, many gene transcripts that rapidly respond to hCG treatment (patterns 4 and 5, Fig. 2A), especially those whose levels increase (Pattern 5), exhibit a transient response and return to or below pretreatment levels by 3–5 days of CG exposure. Notable examples include hydroxysteroid dehydrogenase HSD3β2 mRNA (Fig. 3A) and the progestin-binding protein PGRMC1 (Fig. 5C). Previously reported declines in local LHCGR availability [unoccupied receptors (Ottobre et al., 1984)] or receptor coupling to adenlyate cyclase (VandeVoort et al., 1988) occur within the macaque CL of SEP by 3–6 days of CG exposure. The results from this study also revealed declining LHCGR mRNA levels after Day 12 during CG exposure (data not shown). Thus, reductions in LHCGR mRNA levels, protein expression and receptor signaling may be a contributing factor to the decline in luteal CG sensitivity, which may account for altered mRNA expression observed for some gene cohorts. This likely includes genes responsible for P synthesis, as P levels plateau by 3 days of CG exposure and decline by 6 days of CG exposure (Day 15 + hCG; Fig. 1A), following the transient rise in HDS3β2 mRNA expression. Alternately, transient expression of local ligand–receptor signaling pathways may indirectly influence the expression of gene activities. For example, the protein for the putative membrane receptor for P, PGRMC1, transiently increases in the CL after 1 day of CG exposure (Fig. 5C) and thereafter returns to pretreatment levels. Based on reports of trophic actions for P in the primate CL, including PGRMC1-mediated prevention of cellular apoptosis (Peluso et al., 2009), this mechanism may transiently promote luteal structure–function during early pregnancy.

Other genes displayed delayed and sustained (e.g. Patterns 1, 2 and 3; Fig. 2A) patterns of expression levels after 3 or more days of exposure, e.g. CYP11A1 and HSD11β2 (Fig. 3C and D), during the time of decreasing P response to CG. Changes in expression levels for these genes may be due to loss of LH/CG actions following reduced LHCGR signaling or indirect effects of CG modulation of other local luteotropic or luteolytic factors besides P. Estradiol levels continue to climb with increasing dosages of hCG (Fig. 1B), which could indicate a further role for E during SEP. Protein expression for STAR and CYP19A1 (aromatase, Fig. 5A and B), both important for E production, is maintained at high levels during SEP in contrast to reported decreases in expression for both of these genes during spontaneous and induced luteolysis (Bishop et al., 2009a; Bogan et al., 2009). Estradiol could be acting to regulate CG-induced sustained gene expression in SEP CL via the classical E receptor ESR2 (ERβ), which was detected at high levels in luteal protein extracts collected during the mid-late luteal phase in rhesus monkeys (Duffy et al., 2000). Microarray analysis of ESR2 levels in SEP CL indicates that the mRNA is present, but changes in transcript levels between SEP and regressing CL are not significant by this analysis. Expression of ESR2 is also detected in human CL during luteal rescue, with some evidence for regulation by CG in vitro (van den Driesche et al., 2008).

Steroids are not the only potential local factors regulated by exposure to CG. Locally produced prostaglandin E2 has emerged as another factor important for primate luteal function (Bogan et al., 2008a). Analyzing the SEP microarray data (Supplementary data, File S1) for dynamically expressed probesets involved in prostaglandin synthesis and signaling reveals a pattern similar to those reported in rhesus macaque early to mid-luteal CL during the luteal phase (Bogan et al., 2008b). A changing balance between luteotropic prostaglandin signaling versus luteolytic signaling is observed during SEP; with early transient increases in expression for HPGD (hydroxyprostaglandin dehydrogenase), PTGER3 (prostaglandin E receptor) and PTGES (prostaglandin E synthase) in response to 1 day of hCG exposure (data not shown). Expression of mRNA for the COX-1 (cyclooxygenase) gene increases on Days 6 and 9 of hCG exposure, while expression of mRNA for the PTGFR (prostaglandin F receptor) is unchanged, albeit present at relatively high levels (data not shown). This transition resembles a change similar to prostaglandins present during the late luteal phase near regression (Bogan et al., 2008a). Prostaglandin signaling may contribute to acute regulation of the luteal transcriptome by CG exposure during SEP.

One gene demonstrating a robust delayed response to CG exposure is relaxin (RLN1, Fig. 4A). Previous studies on rhesus monkeys indicated that unlike the rapid increase in serum levels of P following CG exposure, relaxin levels gradually increase during SEP and plateau 4 days after the final hCG injection (Duffy et al., 1996). In vitro culture of human granulosa cells exposed to constant levels of CG mimicking SEP respond by increasing relaxin secretion only after ∼5 days of treatment, and levels peak 10 days after initial exposure (Stewart and VandeVoort, 1997). While there is evidence for an endocrine role for relaxin in macaque blastocyst development (VandeVoort et al., 2011), a paracrine role for relaxin in the ovary is also possible. The current evidence for expression of LGR7, as well as its ligand RLN1 in rhesus CL, confirms previous reports (Bogan et al., 2008b; Maseelall et al., 2009b). Since E promotes LGR7 expression in human uterine cells (Maseelall et al., 2009a), it is possible that steroids influence the RLN1–LGR7 pathway. Local relaxin signaling in the primate CL remains an underappreciated pathway that should be investigated further.

Comparison of gene expression patterns between rescued and regressed CL demonstrated that the primate CL responds to CG in the presence of tonic LH secretion during the late luteal phase with increasing divergence of mRNA transcripts when compared with CL of non-fecund cycles. Down-regulation (Pattern 1, Fig. 6A) of the majority of mRNA transcripts was reported previously during spontaneous luteolysis of rhesus CL (Bogan et al., 2009), although a number of mRNAs displayed increased expression by late luteal phase (Pattern 2). During SEP, CG treatment prevented most of the changes in luteal gene expression observed by CL undergoing luteolytic processes. Early prevention of these luteolytic processes by exposure to CG on Day 9 is critical for successful primate luteal rescue, as evidenced by studies in the rhesus macaque showing diminished responses to CG administration beginning on and after Day 14 of the luteal phase (Wilks and Noble, 1983). While these time-matched CL are very different on Day 18 of the luteal phase, the rescued CL will ultimately regress during pregnancy in the presence of continued CG (Treloar et al., 1972). Further mining of the differences between conceptive and non-fertile cycles, especially in the first 3 days of CG exposure, could reveal other important luteotropic processes critical for sustaining primate pregnancy.

Complete microarray databases now exist for the rhesus macaque ovary from the ovulatory luteinizing follicle (Xu et al., 2011, GEO data sets GSE22776) and CL spanning its lifespan throughout the non-gravid luteal phase (Bogan et al., 2008b, GEO data sets GSE10367), particularly the time period of spontaneous regression during the natural menstrual cycle (Bogan et al., 2009, GEO data sets GSE12807) and in SEP (current data, GEO data sets GSE25335). All include biological replicates (n= 4–6 samples/time point) of ovarian tissue biopsies (ovulating follicle or CL) and are hybridized to Affymetrix™ GeneChip^® Rhesus Macaque Genome Arrays at the same facility using the same labeling protocol. These are publically available in the NCBI Gene Expression Ominbus (GEO) and will allow investigators to explore changes in gene expression based on the cycle stage, CL regression and CL rescue in SEP. In addition, data sets are available on the response to ablation/replacement of LH and steroids at the mid-late non-gravid luteal phase (Bishop et al., 2009a, GEO data sets GSE12281). GEO data sets available from cDNA of bonnet monkey CL hybridized to the rhesus array (Priyanka et al., 2009, GEO data sets GSE7827 and GSE8371) were successfully compared with rhesus ovarian array data to identify common themes of LH-sensitive gene regulation in the primate CL (Bishop et al., 2011b). The bonnet monkey CL array data were also mined to identify progestin-related luteal gene transcripts (Suresh et al., 2011). The impact of steroid ablation during rhesus macaque SEP luteal rescue is currently under investigation (Bishop et al., 2011a, GEO data sets GSE29363). These primate ovarian microarray databases can be further mined based on the Affymetrix™ probeset identifier, Pathway/Ontology, or gene identifier for mRNAs represented on Rhesus GeneChip Array, to consider novel gene pathways controlling primate female fertility and to investigate effects of CL luteinization/regression/rescue on known ovarian gene pathways. The rhesus macaque genome is still undergoing annotation, and once completed, it will allow for full utilization of these resources (Noriega et al., 2009; Bishop et al., 2011b).

In summary, exposure of the primate CL to CG during early pregnancy greatly impacts the activity of the luteal transcriptome when compared with the CL of the non-fecund cycle. Further mining of impacted gene pathways can lead to identification of novel luteotropic factors and new treatments for infertility during early pregnancy.

Authors' roles

C.V.B.: experimental conception and design, data acquisition, analysis and interpretation, manuscript preparation and final approval. S.S.: data acquisition and interpretation, manuscript revision and final approval. L.X.: data acquisition and interpretation, manuscript revision and final approval. J.D.H.: experimental conception and design, data interpretation, manuscript revision and final approval. R.L.S.: experimental conception and design, data interpretation, manuscript revision and final approval.

Supplementary data

Supplementary data are available at http://molehr.oxfordjournals.org/.

Funding

R01 HD20869, RR00163, T32-HD07133 (C.V.B.), China Scholarship Council (CSC; L.X.).

Conflict of interest

None declared.