Global analysis of genes regulated by HOXA10 in decidualization reveals a role in cell proliferation

Abstract

Homeobox (HOX) A10 is essential for fertility as demonstrated in transgenic mice, specifically affecting implantation and decidualization. Its role in human decidualization, however, remains unknown. In this study, we used gene silencing followed by microarray analysis to decipher the role of HOXA10 during decidualization of human endometrial stromal cells (HESCs). HOXA10 was knocked down using siRNA oligonucleotide transfection and cells were treated with estradiol, medroxyprogesterone acetate and dibutyryl cAMP (H + cAMP) to induce decidualization. Genes significantly regulated were identified using the Affymetrix microarray chip. With this method, 2361 transcripts were significantly altered by 1.5-fold or higher (P < 0.05) with H + cAMP treatment only. Of these genes, 258 were significantly up-regulated by HOXA10 knockdown and 236 transcripts were significantly down-regulated by more than 1.5-fold, totaling 494 genes that were regulated by HOXA10 during decidualization. Data analysis using the Ingenuity System revealed that many of the genes regulated by HOXA10 knockdown during H + cAMP treatment were associated with cell cycle. Real-time PCR was used to confirm that HOXA10 knockdown decreased expression of the cell cycle genes CDC2 and CCNB2. In addition, a higher percentage of cells were arrested in the G2/M phase. Next, we observed that cell proliferation as measured by BrdU incorporation was decreased upon HOXA10 knockdown and H + cAMP treatment. Apoptosis, on the other hand, as measured by Annexin V staining was not influenced by siHOXA10 in decidualizing cells. Together, these data demonstrate that during decidualization of HESC, HOXA10 is actively involved in promoting cell proliferation through the regulation of hundreds of genes.

Introduction

The human endometrium undergoes extensive remodeling in response to the rising levels of estrogen and progesterone during the menstrual cycle. During the secretory phase, decidualization is initiated in the stroma and progresses if there is a pregnancy and develops into the decidua of pregnancy. Decidualization is best described as the transformation of the endometrial stroma into a dense cellular matrix known as the decidua (Brosens and Gellersen, 2006). Specifically the endometrial stromal cells differentiate to a decidual cell which differs morphologically and biochemically, expressing and secreting factors that are favorable for an implanting embryo. Over the years, the decidualization process has been studied extensively and the use of transgenic mice has not only demonstrated its necessity for pregnancy but has made it possible to identify critical genes involved in this process. The genes important for decidualization identified thus far include progesterone receptor A (Lydon et al., 1995), homeobox (HOX) A10 (Satokata et al., 1995), cyclooxygenase 2 (Lim et al., 1997), leukemia inhibitory factor (Stewart et al., 1992; Stewart and Cullinan 1997), interleukin 11 receptor (Robb et al., 1998), Hoxa11 (Gendron et al., 1997), FK506 binding protein 4 (Tranguch et al., 2007), CCAAT/enhancer binding protein beta (Mantena et al., 2006), Indian Hedgehog (Lee et al., 2006), steroid receptor coactivator-1 and steroid receptor coactivator-2 and -1 (Mukherjee et al., 2006).

HOXA10 is a member of the homeobox gene family known to be involved in the genetic control of development, particularly in the specification of the body plan, pattern formation, the determination of cell fate and several other basic developmental processes (reviewed in Gehring, 1987). HOXA10 is expressed during embryonic development (Krumlauf, 1994) and continues to have a regulatory role in the adult female reproductive tract. Expression of HOXA10 in the uterus is regulated in response to estrogen and progesterone, and its levels increase dramatically during the mid-secretory phase of the menstrual cycle when progesterone levels are high (Taylor et al., 1998). In addition, levels of HOXA10 expression are diminished in the endometria of women with endometriosis, polycystic ovary syndrome (Cermik et al., 2003) and idiopathic infertility (Gui et al., 1999), indicating that HOXA10 may be essential for fertility, as is the case in mice. HOXA10 acts as a transcription factor containing a unique homeodomain that comprised a 61 amino acid residue polypeptide which binds to DNA. Genes such as cyclin-dependent kinase inhibitor 1A (Bromleigh and Freedman, 2000), genes of the Wnt pathway (Ferrell et al., 2005), integrin, beta 3 (Daftary et al., 2002), empty spiracles homolog 2 (Troy et al., 2003) and FK506 binding protein 4 (Daikoku et al., 2005) are among some that are regulated by HOXA10. In the context of decidualization, we recently demonstrated that HOXA10 attenuates the expression of insulin-like growth factor binding protein (IGFBP1) in decidualizing stromal cells (Kim et al., 2007). Thus, it is evident that HOXA10 regulates a number of genes depending on the tissue and hormonal milieu; however, its role in human endometrial function remains unclear.

In this study, we performed a genome-wide microarray analysis in order to identify the genes that are regulated by HOXA10 during decidualization and by doing so, elucidated a physiological function for HOXA10. Specifically, HOXA10 was silenced in endometrial stromal cells using siRNA oligonucleotides and the cells were then treated with the hormones, estrogen and medroxyprogesterone acetate (MPA) and dibutyryl cAMP (H + cAMP) to induce decidualization. Microarray analysis revealed that a significant number of the HOXA10-dependent decidua-specific genes were involved in cell cycle regulation. Confirmatory studies demonstrated that HOXA10 does indeed promote cell proliferation during decidualization.

Materials and Methods

Cell culture

Human endometrial tissue was obtained from premenopausal women undergoing hysterectomies for fibroids and with no clinically documented abnormalities of the endometrium. Menstrual cycle phase was not recorded since our previous studies showed that stromal cells responded similarly to hormones whether they came from proliferative or secretory phase tissues. This study was approved by the Human Subject Committee at Northwestern University, in accordance with U.S. Department of Health regulations. Human endometrial stromal cells (HESCs) were isolated as previously described (Kim et al., 2003). Cells were grown until 80% confluence and subsequently remained untreated or treated with 36 nM 17b-estradiol, 1 µM MPA and 0.1 mM dibutyryl-cAMP (Sigma, St Louis, MO, USA) for 48 h. We will refer to this hormonal treatment as H + cAMP.

Small interfering RNA and transient transfection

Transfections of small interfering RNA (siRNA) oligonucleotides were carried out using X-tremeGENE siRNA Transfection Reagent (Roche Applied Sciences, Indianapolis, IN, USA) according to the manufacturer’s protocol. In knockdown experiments, HESCs were transiently transfected with 50 nM of the HOXA10 siRNA oligonucleotides (Dharmacon, ON-TARGETplus SMARTpool L-006336-00), or siRNA specific to the firefly luciferase protein (Dharmacon, Lafayette, CO, USA) acting as a control and referred to as siControl (Kim et al., 2007). After siRNA transfection, cells treated with H + cAMP for 48 h in media containing 2% charcoal-stripped fetal bovine serum (Kim et al., 2007). Since it has been demonstrated that transfection of siRNA oligonucleotides into cells can also activate the innate antiviral response in some cell types and promote the expression of the cytokine interferons, we determined whether interferon response was activated in our system. Interferon stimulates the cellular stress response by inducing the expression of numerous interferon-stimulated genes (ISGs). A kit for measuring changes in ISGs by real-time PCR was used according to the manufacturer’s protocol (Systems Biosciences, Mountain View, CA, USA). We found no activation of ISGs when siHOXA10 was transfected into HESCs.

Western blot analysis

HESCs were lysed with RIPA buffer (150 mM NaCl, 1% IGEPAL, 0.5% sodium deoxycholate, 0.1% SDS and 50 mMTris, pH 8.0) with protease inhibitors (Sigma) to recuperate whole-cell proteins. Protein content was measured using the Micro BCA protein assay kit (Pierce, Rockford, IL, USA). Proteins were run on a precast 7.5% acrylamide gel (Bio-Rad, Hercules, CA, USA) and transferred onto PVDF membrane. Membranes were blocked with 5% milk in Tris-buffered saline with 0.1% Tween 20 and then incubated with primary antibody to HOXA10 (N-20, Santa Cruz) followed by incubation with secondary peroxidase-conjugated donkey anti-goat. Protein complexes were detected with a chemiluminescent detection kit (Amersham Biosciences, Piscataway, NJ, USA).

Microarray analysis and statistical analysis

Three separate primary HESC cultures derived from three different patients were used for microarray analysis. Each culture consisted of three experimental conditions, hence, a total of nine microarray chips were used. The experimental conditions were (i) transfected with the siControl oligonucleotides, (ii) transfected with siControl oligonucleotides and treated with H + cAMP for 48 h and (iii) transfected with an siRNA targeting HOXA10 with subsequent treatment H + cAMP for 48 h. All RNA samples were processed at the Microarray Core Facility in the Center for Genetic Medicine at Northwestern University (Chicago, IL, USA). The quality of total RNA was evaluated using Bioanalyzer 2100 (Agilent Technologies Inc., Santa Clara, CA, USA). Each RNA sample (1.5 µg), with 260/280 and 28S/18S ratio of greater than 1.8, was used to make double-stranded cDNA and labeled cRNA following the One-Cycle Target Labeling Assay from Affymetrix (Affymetrix Inc., Santa Clara, CA, USA). The size distribution and fragmentation quality of the biotin-labeled cRNA were checked by the Bioanalyzer 2100. The fragmented labeled cRNA was hybridized to the Human U133-Plus 2.0 Array (Affymetrix Inc.) for 18 h. The chips were then scanned and the data extracted using the Affymetrix GeneChip Operating Software (Affymetrix Inc.). Gene expression levels were quantified using the RMA algorithm with the quantile normalization built in. Data analysis was conducted using the Bioconductor limma package (Dudoit et al., 2003; Wettenhall and Smyth, 2004). To find statistically consistent genes with differential expression, we utilized a linear model with empirical Bayesian correction. The candidate gene list was selected by fold changes larger than 1.5-fold and P < 0.05 (false discovery rate adjusted). To interpret the biological significance of the candidate genes, the Ingenuity Pathways Analysis system was used (IPA, Ingenuity Systems, www.ingenuity.com). Specifically, the Functional Analysis tool identified the biological functions and/or diseases that were most significant to the dataset. The 494 genes that were regulated by HOXA10 during decidualization were analyzed and these genes from the dataset with the fold change cutoff of 1.5-fold were associated with biological functions and/or diseases in the Ingenuity Pathways Knowledge Base were considered for the analysis. The Fischer’s exact test was used to calculate a P-value determining the probability that each biological function and/or disease assigned to the dataset is due to chance alone. Threshold is at 1.25 = −log (P = 0.05).

Real-time PCR

Cells were lysed using TriReagent (Sigma) and total RNA was extracted using the manufacturer’s protocol. Total RNA was reverse transcribed in a final volume of 20 µl. Real-time PCR analysis was performed using either SYBR green fluorescence for HOXA10 (Kim et al., 2007) or Taqman for CDC2 and CCNB2. Each real-time PCR consisted of 1 µl RT product, 10 µl SYBR Green or Taqman PCR Master Mix (PE Applied Biosystems, Foster City, CA, USA), and 500 nM forward and reverse primer pairs. All reactions were carried out on an ABI PRISM 7700 Sequence Detection System (Applied Biosystems) for 40 cycles (95°C for 15 s, 60°C for 1 min) after 10 min incubation at 95°C. The fold change in expression of each gene was calculated using the ΔΔC_t method (Livak and Schmittgen, 2001), with the ribosomal protein, large, P0 (RPLP0) mRNA or TBP as an internal control.

Proliferation and cell cycle analysis

Cells were grown in 60 mm plates until 50% confluency. Cells were transfected with siHOXA10 or siControl oligonucleotides for 24 h. Cells were trypsinized and plated into a 96-well plate followed by treatment with H + cAMP for 48 h. We used two methods to determine cell proliferation and viability. First, the Quick Cell Proliferation Assay Kit (#K301, BioVision, Mountain View, CA, USA) was used to measure viable cells after treatments. Ten microliters of WST-1/Electro Coupling Solution solution was added per well and incubated at 37°C. Samples were read on the Synergy HT from Bio-Tek with the KC4 3.4 software at 440 nm to determine cell proliferation and viability. Second, BrdU incorporation immunoassay was used to specifically measure cell proliferation. Cells were treated as above and BrdU was added during the last 24 h of H + cAMP treatment. During this time, BrdU is incorporated in place of thymidine into the DNA of cycling cells. BrdU was then detected by an immunoassay according to the manufacturer’s protocol (Roche Applied Sciences) and the reaction product quantified by measuring the absorbance using a scanning multi-well spectrophotometer.

For cell cycle analysis, cells were trypsinized after HOXA10 knockdown and H + cAMP treatment and fixed with 75% ethanol for 2 h. Cells were resuspended in 1 mL of propidium iodide (PI) staining solution containing 50 µg/mL PI (Sigma), 2 mg of RNase A (Invitrogen, Eugene, OR, USA) and 0.1% Triton-X (Fisher) made in PBS. Samples were incubated for 20 min at 37°C and analyzed for G0/G1, S and G2/M fractions on a Coulter EPICS-XL flow cytometer (Beckman–Coulter, Fullerton, CA, USA).

Apoptosis

Cells were grown in 60 mm plates until 50% confluency. Cells were transfected with siHOXA10 or siControl oligonucleotides followed by treatment with H + cAMP for 48 h. Cells were trypsinized and resuspended in annexin-binding buffer (10 mM Hepes, 140 mM NaCl and 2.5 mM CaCl₂, pH 7.4) to a concentration of ~1 × 10⁶ cells/mL. Annexin V, Alexa Fluor 647 conjugate (Invitrogen) and DAPI (Invitrogen) were added to each cell solution and samples were analyzed using the CyAn flow cytometer (Dako, Fort Collins, CO, USA) for early and late apoptosis.

Statistical analysis

For Fig 1, Fig 3 and Fig 4, data were analyzed by a one-way ANOVA followed by the Tukey post hoc test.

Figure 1. HOXA10 knockdown in decidualizing HESCs.

HESC were transfected with siRNA to HOXA10 (siHOXA10) or to the luciferase gene as a control (siCTRL). Cells were then treated with H + cAMP for 48 h. HOXA10 mRNA (A) and protein (B) were measured by real-time PCR or western blot to verify knockdown of HOXA10. Data are expressed as fold changes compared with the siCTRL and no treatment and presented as the mean ± SEM of three experiments for (A) or representative of three experiments for (B). Statistical differences are noted as ‘a’. P < 0.05.

Figure 3. HOXA10-dependent decidua-specific genes associated with cell cycle.

(A) The top network identified by IPA included genes associated with cell cycle, cellular assembly and organization, and cancer. CDC and CCNB2 are highlighted with a box. Upregulated genes upon HOXA10 knockdown are represented as gray shapes marked with a dotted circle. Down-regulated genes are represented by the other shapes in gray. The biological relationship between two nodes is represented as either a solid line (direct interaction) or dotted line (indirect interaction). (B) Primary cultures were transfected with or without HOXA10 siRNA and treated with H + cAMP for 48 h. Levels of CCNB2 and CDC2 transcripts were measured by real-time PCR. The data were normalized to TBP transcript levels and expressed as fold change compared with the expression level of untreated cells. Data are expressed as fold changes compared with the siCTRL and no treatment and presented as the mean ± SEM of triplicate measurements. Statistical differences are noted as ‘a or b’. P < 0.01.

Figure 4. HOXA10 knockdown attenuates cell proliferation in differentiating HESCs.

(A) HESC cells transfected with HOXA10 siRNA (dotted line) or control siRNA (solid line) were plated in 96-well plates overnight and then treated with H + cAMP. Cell viability was measured 0, 24, 48 and 72 h later. The results are the relative fold change in viable cells compared with untreated cells at time 0. (B) HESCs transfected with control or HOXA10 siRNA and treated with H + cAMP for 48 h were analyzed by flow cytometry. The percentage of cells in each phase of the cell cycle (G0/G1, S and G2/M) is indicated. Histograms are representative of four independent experiments. (C) HESCs were treated as in (A) and cell proliferation was measured using BrdU incorporation for 24 h. The results are shown as fold change of cell proliferation compared with siCTRL and corrected for initial cell number. Data are the mean + SEM of three independent experiments. P < 0.05. (D) HESCs were treated as in (B). Annexin V/DAPI staining was used to identify early apoptotic cells. Data are the mean + SEM of three independent experiments. Statistical differences are noted as ‘a’. P < 0.05.

Results

Identification of HOXA10 target genes in decidualizing HESCs

HESCs decidualize in culture when treated with estrogen, progestin and a cAMP analog. During this time, numerous genes are regulated in a time-dependent manner as previously shown by others (Brar et al., 2001; Tierney et al., 2003). In order to elucidate the role of HOXA10 during decidualization, HOXA10 was first silenced in HESCs using siRNA and subsequently treated with H + cAMP for 48 h. This technique efficiently knocked down HOXA10 mRNA and protein expression (Fig. 1).

Following the knockdown of HOXA10 and H + cAMP treatment, the RNA was subjected to microarray analysis using the Human U133 Affymetrix chip which provides over 47 000 transcripts. Three treatment groups for each of the HESCs from three different patients were subjected to analysis. The three treatment groups were (i) cells transfected with the siControl oligonucleotides, targeting the luciferase gene (ii) cells transfected with siControl and treated with H + cAMP for 48 h and (iii) cells transfected with a siRNA targeting HOXA10 with subsequent treatment H + cAMP for 48 h. The genes were first compared in groups 1 versus 2 in order to identify genes significantly regulated by H þ cAMP treatment. These genes were then compared with group 3 in order to identify genes regulated by siHOXA10 knockdown. First, we found 2361 transcripts significantly altered by 1.5-fold or higher (P < 0.05) with H + cAMP treatment only. All data are available for download in Supplementary Table S1. The 50 most highly induced and repressed genes are shown in Supplementary Table S2. Many of these genes have previously been reported to be decidua-specific (Brar et al., 2001; Tierney et al., 2003; Takano et al., 2007). Of the 2361 decidua-specific genes, 258 were significantly up-regulated by HOXA10 knockdown and 236 transcripts were significantly down-regulated by more than 1.5-fold, totaling 494 genes that were regulated by HOXA10 during decidualization (Supplementary Table S3). Table I and Table II list the top 50 decidua-specific genes that are up or down regulated by HOXA10 knockdown.

Table I.

Top 50 genes induced after HOXA10 knockdown in decidualizing cells.

Gene symbol	Gene name	H + cAMP (fold change)	HOXA10 siRNA (fold change)
ADAM30	ADAM metallopeptidase domain 30	−7.31	20.74
SVIL	Supervillin	−6.60	10.10
ASH1L	ash1 (absent, small, or homeotic)-like (Drosophila)	−4.39	9.93
CAPZA2	Capping protein (actin filament) muscle Z-line, alpha 2	−3.19	9.90
MULK	Multiple substrate lipid kinase	−15.30	9.67
SERPINB3	Serpin peptidase inhibitor, clade B (ovalbumin), member 3	−2.41	8.98
SMN2	Survival of motor neuron 2, centromeric	−4.31	7.94
TIE1	Tyrosine kinase with immunoglobulin-like and EGF-like domains 1	−2.99	7.82
PMFBP1	Polyamine modulated factor 1 binding protein 1	−1.91	7.71
ANKMY1	Ankyrin repeat and MYND domain containing 1	−4.42	7.27
RBM12B	RNA binding motif protein 12B	−3.83	7.15
FRAS1	Fraser syndrome 1	−3.56	7.02
FAT	FAT tumor suppressor homolog 1 (Drosophila)	−4.09	6.96
SEC14L3	SEC14-like 3 (S. cerevisiae)	−2.73	6.84
CCDC28A	Coiled-coil domain containing 28A	−3.94	6.53
SLC26A8	Solute carrier family 26, member 8	−3.48	6.52
KCTD4	Potassium channel tetramerisation domain containing 4	−4.93	6.35
VN1R2	Vomeronasal 1 receptor 2	−3.98	6.30
	UDP-N-acetyl-α-d-galactosamine:polypeptide N-
GALNT5	Acetylgalactosaminyltransferase 5 (GalNAc-T5)	−2.37	6.25
RAB40C	RAB40C, member RAS oncogene family	−4.67	5.94
RCE1	RCE1 homolog, prenyl protein peptidase (S. cerevisiae)	−5.28	5.80
MARCH1	Membrane-associated ring finger (C3HC4) 1	−3.14	5.79
CYP19A1	Cytochrome P450, family 19, subfamily A, polypeptide 1	−4.44	5.78
DNASE1	Deoxyribonuclease I	−4.99	5.78
CLSTN2	Calsyntenin 2	−4.77	5.78
RNASE7	Ribonuclease, RNase A family, 7	−6.78	5.77
ANKRD15	Ankyrin repeat domain 15	−5.36	5.76
SOST	Sclerosteosis	−6.10	5.75
MS4A6A	Membrane-spanning 4-domains, subfamily A, member 6A	−4.03	5.53
C9	Complement component 9	−7.43	5.48
RBM20	RNA binding motif protein 20	−3.24	5.47
TSHR	Thyroid stimulating hormone receptor	−3.79	5.45
ZBTB37	Zinc finger and BTB domain containing 37	−2.20	5.44
RFC3	Replication factor C (activator 1) 3, 38 kDa	−5.27	5.09
ZNF226	Zinc finger protein 226	−6.37	5.09
LRRC61	Leucine rich repeat containing 61	−2.83	5.07
ITGB1	Integrin, beta 1 (fibronectin receptor, beta polypeptide, antigen CD29 includes MDF2, MSK12)	−4.82	5.01
BMPR2	Bone morphogenetic protein receptor, type II (serine/threonine kinase)	−3.54	4.99
HAP1	Huntingtin-associated protein 1 (neuroan 1)	−3.26	4.97
ZFP42	Zinc finger protein 42	−4.00	4.94
MTMR9	Myotubularin related protein 9	−3.22	4.93
PKP2	Plakophilin 2	−4.45	4.91
RAPGEF5	Rap guanine nucleotide exchange factor (GEF) 5	−4.60	4.87
SORBS2	Sorbin and SH3 domain containing 2	−2.77	4.84
PPP1R16B	Protein phosphatase 1, regulatory (inhibitor) subunit 16B	−6.13	4.74
GPC5	Glypican 5	−6.24	4.70
NEB	Nebulin	−4.18	4.69
ATP8A1	ATPase, aminophospholipid transporter (APLT), Class I, type 8A, member 1	−3.57	4.67
PHGDHL1	Phosphoglycerate dehydrogenase-like 1	−6.55	4.66
	UDP-N-acetyl-α-d-galactosamine:polypeptide N-
GALNT8	Acetylgalactosaminyltransferase 8 (GalNAc-T8)	−5.59	4.65
FBXO31	F-box protein 31	−2.82	4.58

Table II.

Top 50 genes repressed after HOXA10 knockdown in decidualizing cells.

Gene symbol	Gene name	H + cAMP (fold change)	HOXA10 siRNA (fold change)
KRTAP4–10	Keratin associated protein 4–10	3.92	−8.99
PNLIPRP2	Pancreatic lipase-related protein	2 4.12	−8.00
CCDC38	Coiled-coil domain containing 38	8.66	−7.84
ORM1	Orosomucoid 1	7.63	−7.67
ANKRD30B	Ankyrin repeat domain 30B	6.65	−7.38
RECK	Reversion-inducing-cysteine-rich protein with kazal motifs	10.37	−7.27
JARID2	Jumonji, AT rich interactive domain 2	4.52	−6.51
STK35	Serine/threonine kinase 35	3.07	−6.02
PAK2	p21 (CDKN1A)-activated kinase 2	3.84	−5.81
IL17F	Interleukin 17F	3.60	−5.73
GRIN3B	Glutamate receptor, ionotropic, N-methyl-d-aspartate 3B	5.21	−5.68
SLC28A1	Solute carrier family 28 (sodium-coupled nucleoside transporter), member 1	4.38	−5.67
SCML4	Sex comb on midleg-like 4 (Drosophila)	4.59	−5.65
VPS18	Vacuolar protein sorting protein 18	5.09	−5.64
PRKCE	Protein kinase C, epsilon	9.63	−5.58
WNT6	Wingless-type MMTV integration site family, member 6	3.11	−5.36
EFCAB1	EF-hand calcium binding domain 1	5.32	−5.33
SERPINB7	Serpin peptidase inhibitor, clade B (ovalbumin), member 7	−3.27	−5.32
RFXDC1	Regulatory factor X domain containing 1	2.41	−5.15
LPHN3	Latrophilin 3	3.67	−5.14
SH3GLP2	SH3-domain GRB2-like pseudogene 2	5.18	−5.13
MITF	Microphthalmia-associated transcription factor	3.03	−5.10
FILIP1	Filamin A interacting protein 1	11.75	−4.98
ZFY	Zinc finger protein, Y-linked	2.79	−4.88
CKMT2	Creatine kinase, mitochondrial 2 (sarcomeric)	4.15	−4.82
NUDT7	Nudix (nucleoside diphosphate linked moiety X)-type motif 7	2.90	−4.77
GPC5	Glypican 5	4.65	−4.70
VGLL1	Vestigial-like 1 (Drosophila)	7.10	−4.68
EVI5L	Ecotropic viral integration site 5-like	2.40	−4.66
TRERF1	Transcriptional regulating factor 1	3.86	−4.62
ZNF292	Zinc finger protein 292	2.22	−4.52
MIDN	Midnolin	3.75	−4.50
EP400	E1A binding protein p400	5.56	−4.44
PTOV1	Prostate tumor overexpressed gene 1	5.22	−4.39
P2RX2	Purinergic receptor P2X, ligand-gated ion channel, 2	5.73	−4.38
TBC1D10A	TBC1 domain family, member 10A	3.26	−4.36
ZNF160	Zinc finger protein 160	3.17	−4.34
HNRPM	Heterogeneous nuclear ribonucleoprotein M	3.88	−4.34
BAGE2	B melanoma antigen family, member 2	6.01	−4.32
	B double prime 1, subunit of RNA polymerase III transcription initiation factor
BDP1	IIIB	4.97	−4.28
KLHL1AS	Kelch-like 1 antisense (Drosophila)	2.48	−4.27
SLC24A1	Solute carrier family 24 (sodium/potassium/calcium exchanger), member 1	2.82	−4.27
ATXN3L	ataxin 3-like	3.11	−4.24
ACSBG1	Acyl-CoA synthetase bubblegum family member 1	3.34	−4.21
GRIP2	Glutamate receptor interacting protein 2	2.56	−4.18
AAK1	AP2 associated kinase 1	2.28	−4.17
TSGA10	Testis specific, 10	4.68	−4.15
KRTAP19-1	Keratin associated protein 19–1	3.08	−4.07
ZNF595	Zinc finger protein 595	2.10	−4.05
EIF4E3	Eukaryotic translation initiation factor 4E member 3	3.43	−4.01
SLC41A2	Solute carrier family 41, member 2	1.52	−4.01

Classification of HOXA10 dependent decidua-specific genes

Genes identified in the microarray as being HOXA10-dependent and decidua-specific were categorized into functions using IPA. The Functional Analysis tool identified the biological functions in the Ingenuity Pathways Knowledge Base that were most significant to our dataset. When data were classified according to Molecular and Cellular Functions (Fig. 2A), the most highly significant function were genes associated with cell cycle. In Fig. 2B, the data is presented as a pie chart to demonstrate the number of significant genes falling into each function. The highest number of genes (104) from our dataset was found in the category of Cellular Growth and Proliferation.

Figure 2. Ingenuity Pathway Analysis of functional categories for the 494 HOXA10-dependent decidua-specific genes.

Genes from our data set were classified into categories pertaining to Molecular and Cellular Function. (A) Significance refers to the –log (P-value), which is obtained by the Ingenuity program using the right-tailed Fisher’s exact test. Threshold is at 1.25 = −log (P = 0.05). (B) The number of genes falling into each category is illustrated.

In light of this analysis, we used IPA to generate networks from our dataset. This network generation function overlaid the HOXA10-dependent decidua-specific genes onto a global molecular network that was developed from information contained in the Ingenuity Pathways Knowledge Base. Using algorithms, the networks were generated based on their connectivity to the different genes. The top network with the highest number of genes included from our dataset was genes associated with cell cycle, cellular assembly and organization, and cancer (Fig. 3A). The molecular relationships between the genes or gene products are shown by each line which is supported by at least one reference from the literature, textbook or from canonical information stored in the Ingenuity Pathways Knowledge Base. References for the generation of network 1 are listed in Supplementary Table S4. Up-regulated genes upon HOXA10 knockdown are represented as gray shapes marked with a dotted circle. Down-regulated genes are represented by the other shapes in gray. In this network, 3 biological relationships existed with CCNB2 and 12 biological relationships existed with CDC2 both directly and indirectly (Fig. 3A). We chose these two genes as they appeared to modify numerous other genes, to confirm the microarray data using real-time PCR. In support of the microarray data, treatment with H + cAMP decreased expression of CDC2 and CCNB2 and knockdown of HOXA10 followed by H + cAMP treatment further decreased expression of these two genes.

HOXA10 and cell proliferation and apoptosis

To determine whether the alterations of expression of cell cycle genes as identified by the microarray did indeed influence cell physiology, a cell viability assay was performed. While treatment with H + cAMP increased the number of viable cells over a 72 h period, HOXA10 knockdown resulted in a decrease in the number of viable cells (Fig. 4A). Cell cycle analysis revealed that while cells in G0/G1 phase decreased slightly, cells in the G2/M phase increased with HOXA10 knockdown (Fig. 4B). These data suggest that the cells have arrested in the G2/M phase. Since the decrease in viable cells after HOXA10 silencing could be due to a decreased rate of proliferation or an increased rate of cell death or both, we then performed a BrdU assay to specifically measure proliferation of the cells. As shown in Fig. 4C, proliferation of H + cAMP treated cells with HOXA10 knockdown was lower than that of its corresponding control. Next, we assessed whether cell death was altered with HOXA10 knockdown using the Annexin V assay to measure apoptosis. As shown in Fig. 4D, the percentage of decidualizing cells undergoing early apoptosis did not significantly differ between the control and HOXA10 silenced cells. No significant difference was observed in the percentage of cells at late apoptosis either although there was a trend for a higher, albeit modest percentage of dying or dead cells (early and late apoptosis) upon HOXA10 knockdown (data not shown). However, these data did not reach statistical significance. In summary, our data demonstrate that silencing HOXA10 in HESC reduces proliferation but does not significantly affect apoptosis.

Discussion

Human decidualization is a complex process that involves interactions of numerous genes and pathways. Other microarray studies have identified the genes that are most prominently regulated during decidualization and have provided insight to types of transformations that occur (Brar et al., 2001; Tierney et al., 2003; Takano et al., 2007). While identification of a function for a particular gene during this process in human cells has been a challenge, techniques involving gene silencing in combination with microarray analysis with the use of bioinformatics tools have made it feasible to define a role for a particular gene of interest during decidualization. We have recently reported a study defining the role of FOXO1 during decidualization of stromal cells using such an approach (Takano et al., 2007). Similarly, we used the same approach here to identify genes that are regulated by HOXA10 during decidualization of HESCs as well as to elucidate physiological functions for HOXA10 during this process.

As with other homeobox genes, HOXA10 is a regulator of embryonic morphogenesis and differentiation (McGinnis and Krumlauf, 1992) and is essential in determining body pattern along the anterior–posterior axis. HOXA10 is necessary for the development and differentitation of the reproductive tract (Izpisua-Belmonte et al., 1991). In adult female mice that are HOXA10-deficient, implantation is severely compromised and defective decidualization leads to recurrent pregnancy loss and infertility (Benson et al., 1996; Lim et al., 1999). Furthermore, blocking HOXA10 expression by administering HOXA10 antisense oligonucleotides into the uterine lumen results in decreased litter size (Bagot et al., 2000). From these studies, it is apparent that HOXA10 is necessary for implantation, however how it does this remains unknown.

Studies have demonstrated that HOXA10 is an estrogen and progesterone responsive gene in the mouse, primate and human endometria (Taylor et al., 1998; Gui et al., 1999; Lim et al., 1999; Yao et al., 2003; Godbole et al., 2007). In mice, it has been shown that progesterone stimulates stromal cell proliferation (Huet-Hudson et al., 1989) and that HOXA10 deficient mice exhibit a stromal cell proliferation defect (Lim et al., 1999; Yao et al., 2003). In agreement with its role in proliferation, the expression of two cyclin-dependent kinase inhibitor genes, p57 and p15, were also increased in the hoxa10 mutant mice (Yao et al., 2003). We demonstrate here for the first time that HOXA10 promotes proliferation of HESC. The top gene network formulated by the Ingenuity program was cell cycle, cellular assembly and organization, and cancer (Fig. 3A). In this network, down-regulation of cell cycle regulatory genes such as cyclins E, B1, B2, CDC2, MCM4, E2F7, FOXM1 and AURKB are indicative of attenuation of cell proliferation. Our cell cycle analysis suggests that there is a higher percentage of cells arrested in the G2/M phase. CDC2 and CCNB2 have been reported to be checkpoint control molecules (Tan et al., 2007) and thus the decrease in expression of these genes would affect cell cycle progression. Coincidentally, we recently reported that these two genes were also regulated by FOXO1 during decidualization (Takano et al., 2007) but in the opposite way. Silencing FOXO1 resulted in an increase of CDC2 and CCNB2 in the decidualizing cells. The opposing effects of FOXO1 and HOXA10 on these two genes may be essential to control the decidualization process to prevent premature differentiation and apoptosis.

In any gene knockdown experiment, it is not possible to determine whether the genes are direct or indirect targets of the protein of interest. Thus, whether CCNB2 or CDC2 are direct targets of HOXA10 remain to be determined. It is apparent from the network schematic (Fig. 3A) that multiple interactions between the genes/gene products exist and that influencing one gene by HOXA10 knockdown can influence many others. FOXM1, which is a gene that is decreased after HOXA10 knockdown, mediates or regulates a number of cell cycle associated events, including DNA damage/checkpoint signaling (Tan et al., 2007), cell cycle protein degradation by Skp2 ubiquitination pathway (Liu et al., 2006), chromosome segregation and mitosis (Wonsey and Follettie, 2005; Schuller et al., 2007), and cell proliferation (Zhao et al., 2006). The deletion of FoxM1 specifically in endothelial cells results in a decreased expression of cyclins, Cdc2 and Cdc25C, increased p27(Kip1) expression and decreased Cdk activities (Zhao et al., 2006). Thus, it would be interesting to determine whether a master regulator of cell cycle events such as FOXM1 is a direct target of HOXA10.

From the existing literature, it has been generally assumed that HOXA10 is involved in endometrial differentiation. Recently, we have demonstrated that HOXA10 attenuates the expression of the decidual marker, IGFBP1 (Kim et al., 2007), in decidualizing stromal cells. We have shown here the evidence that HOXA10 promotes proliferation of decidualizing stromal cells. Thus by promoting proliferation, the differentiation process could be attenuated and cells would then be allowed to decidualize over an extended period of time which is one characteristic of decidualization. Full decidualization of HESC requires many days of hormone treatment. It is important to note that the endometrium comprised both glandular epithelial cells as well as stromal cells and differentiation in response to estrogen and progesterone occurs in both cell types (Noyes et al., 1975; Corbeil et al., 1985; Evans et al., 1990). As with other endometrial genes, the role of HOXA10 may differ depending on cell type. In the HOXA10 null mice, ovariectomized females given estrogen and progesterone exhibited an impairment of stromal cell proliferation while epithelial cell proliferation was unaffected (Lim et al., 1999; Yao et al., 2003), demonstrating a potential dichotomous role of HOXA10 in the endometrium. The role of HOXA10 in human endometrial epithelial cells is unclear. Studies have demonstrated that the expression of HOXA10 in endometrial adenocarcinoma, which arises from the glandular epithelium, is either increased (Lane et al., 2004) or decreased (Yoshida et al., 2006) compared with normal endometrium. Thus, it remains uncertain as to the role of HOXA10 in endometrial adenocarcinoma or in normal epithelial cell differentation. Given its role in promoting proliferation in stromal cells, it would be interesting to determine the expression profiles of HOXA10 in endometrial stromal cell sarcomas.

In the mouse, progesterone promotes proliferation of uterine stromal cells which precedes decidualization. It has been shown that HOXA10 mediates progesterone stimulated proliferation during implantation in the mouse (Yao et al., 2003). In the human endometrium, proliferation has received less recognition as a component of decidualization. Furthermore, it remains unclear whether progesterone mediates proliferation of the stromal cells during the secretory phase of the menstrual cycle. One study by Dahmoun et al., (1999) showed that proliferation in the stroma does occur during the last 3 days of the secretory phase, as demonstrated by Ki67 staining and that this paralleled with increased progesterone receptor staining. Whether this proliferation is directly regulated by progesterone is unclear. Also, whether this proliferation precedes decidualization is also unknown, although decidualization in the human endometrium begins in a subset of stromal cells surrounding the terminal spiral arteries of the superficial endometrial layer around Day 23 of a 28-day cycle (Wynn, 1974). Our data suggests that HOXA10 is likely to be involved in the proliferation of stromal cells and whether this holds true in the in vivo situation remains to be studied.

HOXA10 plays a key role in myeloid and B-lymphoid progenitor cell differentiation (Thorsteinsdottir et al., 1997a). In leukemia, uncontrolled proliferating hematopoietic cells do not differentiate and these cells continue to express HOXA10 (Lawrence et al., 1995). Mature peripheral blood leukocytes express low levels of HOXA10 (Lawrence et al., 1995). Finally, it has been shown that overexpression of HOXA10 can lead to acute myeloid leukemia (Thorsteinsdottir et al., 1997b), supporting the association of HOXA10 with proliferation events. Thus, the studies elucidating the mechanisms of HOX proteins in hematopoietic differentiation will provide clues to deciphering the mechanisms by which HOXA10 promotes proliferation during decidualization.

We present in this study the evidence that HOXA10 plays an active role in the process of decidualization. With our genome-wide based microarray approach, we were able to decipher one role of HOXA10 during decidualization which is to promote proliferation through the regulation of hundreds of genes. While it is evident that HOXA10 can regulate numerous physiological processes, we have used bioinformatics tools to focus on just one of these processes, proliferation, during decidualization of HESCs. These novel data give new insight in understanding the consequence of aberrant expression of HOXA10 in uterine pathologies, such as endometriosis, endometrial cancer and uterine leiomyomas.