| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Departments of Development and Molecular Biology and Obstetrics and Gynecology and Women’s Health, Center for the Study of Reproductive Biology and Women’s Health, Albert Einstein College of Medicine, Bronx, New York 10461
Address all correspondence and requests for reprints to: Dr. Jeffrey W. Pollard, Department of Developmental and Molecular Biology, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, New York 10461. E-mail: pollard{at}aecom.yu.edu.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Across species, the uterine luminal epithelium (LE) is the initial site of embryo attachment, whereas the glandular epithelium (GE) is thought to be the principal source of uterine secretions that are required for the establishment and maintenance of pregnancy (1). The absence of GE and reduced LE in the ovine uterine gland ewe knockout model resulted in a reduced conceptus survival, supporting a fundamental role for GE and their secretions during early pregnancy (2). The precise nature of uterine secretions is not well characterized, but that of LE and GE appear to differ biochemically (3, 4).
There is strong evidence other than differences in biochemical secretions to suggest that under receptive conditions (defined in the mouse as the 24-h period in which the LE will allow attachment of the implanting blastocyst), the LE and GE function as specialized units within the endometrium. The LE has more examples of genes that are solely expressed at periimplantation stages [such as Areg (amphiregulin), Hdc (histidine decarboxylase), Irg1 (immunoresponsive gene 1), Hbegf (heparin-binding EGF-like growth factor), and Calb1 (calbindin 28K)] than known in the underlying GE [Lif (leukemia inhibitory factor), Il6st (IL-6 signal transducer), and Calca (calcitonin-related polypeptide-
)] (5, 6). It seems that under receptive conditions, although both epithelial cell types can uniquely express certain genes, many others, also proven to be important for implantation, appear to be coordinately expressed [e.g. Cdh1 (cadherin 1), Tro (trophinin), Ihh (Indian hedgehog), Ptgs1 (prostaglandin-endoperoxide synthase 1 or COX-1)] in both compartments (5, 6). This does not dismiss the importance of genes down-regulated in either or both epithelial cell types during the receptive window such as Muc1 (mucin 1, transmembrane) (7).
In adult rodents, blastocyst implantation defects arise from the loss of expression of two uterine GE-specific proteins, leukemia-inhibitory factor (8) and calcitonin-related polypeptide- (9). WNT7A expression is restricted to the uterine LE of adult mice (10), and knockout studies show that these mutants are devoid of glands (11) and exhibit an infertile phenotype (12). These mouse models demonstrate the intricate relationship between epithelial cell types that is important for the development of a receptive endometrium. However, the extent of this relationship and the molecular signature that defines them has never been explored largely because of our inability to isolate them.
The use of modern molecular technologies such as cDNA array, laser-capture microdissection (LCM), and improved RNA amplification methods has enabled us for the first time to investigate the relationship between LE and GE that is conducive to a receptive environment. The evidence gathered in this study is sufficient to conclude that, at least at the level of transcripts, the LE and GE are specialized epithelial cell compartments. Although these epithelial cell types exhibit cooperative roles during the period of blastocyst implantation, their distinct molecular signatures undoubtedly also play a role in governing the outcome of early pregnancy.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Laser capture microdissection
Cryosections 8 µm thick of unfixed uterine tissues were placed in 90% ethanol stored within the –20 C cryochamber and fixed for no longer than 2 min. Sections were washed in distilled water, rinsed in absolute ethanol, and counterstained in alcoholic eosin for 10 sec. Rapid washing in absolute ethanol was followed by three subsequent dehydration steps, 2 min each, in absolute ethanol. Within a 60-min time frame of sectioning, approximately five to 15 sections of endometrium were used to initially capture LE cells followed by GE cells using the Arcturus Pix cell II instrument (Arcturus Engineering Inc., Mountain View, CA) as previously described (13).
RNA extraction and cDNA probe preparation
Total RNA was extracted from LCM-captured epithelium using RNeasy Mini Kits (QIAGEN, Valencia, CA) for array and RNeasy Micro Kits (QIAGEN) for quantitative real-time RT-PCR (QR-PCR) according to the manufacturer with amended steps for ethanol precipitation as previously described (13). The yields obtained ranged from approximately 10–150 ng, with the upper range targeted for array analysis. The quality of the RNA was determined using the nano-biosizing assay (Agilent Bioanalyzer; Agilent Technologies, Palo Alto, CA). To remove potential genomic DNA contamination, LCM PCR samples were treated with amplification-grade DNase 1 on a column (QIAGEN) or after resuspension of the pellet (Invitrogen, Carlsbad, CA) for array samples as previously described (13).
The RNA pellet obtained from LCM samples was resuspended into 11 µl of RNase/DNase-free water, and a single round of linear amplification was performed by the in vitro transcription T7 promoter method as outlined by the manufacturer’s protocol (Ambion’s Message Amp TM Kit; Ambion, Austin, TX). Amplified RNA was resuspended into 10 µl RNase/DNase-free water and the yield and quality established by a combination of spectrometry and Agilent Bioanalyzer. Probe cDNA was synthesized from approximately 3.5 µg/µl of amplified RNA and labeled with Cy3 or Cy5 fluorescent dyes (GE and LE, respectively).
Microarrays
Our main goal was to investigate the importance of differential gene expression patterns of LE and GE at periimplantation stages (receptive group). However, array experiments were designed to compare the differential expression of LE vs. GE from receptive and nonreceptive groups. This provided further insight as to the likelihood that receptive epithelial gene candidates were hormonally regulated. For each group, LCM samples of LE and GE were collected from one to two animals for hybridization of each array and repeated three to four times.
These samples were simultaneously hybridized to glass 28K cDNA gene chip microarray slides from the Albert Einstein College of Medicine’s Array Facility (http://microarray1k.aecom.yu.edu/). Probe preparation and hybridization were performed according to the facility’s protocol (http://microarray1k.aecom.yu.edu/) as previously described (14).
After hybridization, the arrays were scanned by Axon Gene Pix 4000A scanner (Axon Instruments, Inc., Foster City, CA). The two fluorescent probes were scanned separately (Cy3 at 532 nm excitation and Cy5 at 635 nm excitation) and the images superimposed using Gene Pix Pro 3.0 software (Axon Instruments). The data were transferred into an MS Excel worksheet and the Cy5/Cy3 ratio for each spot calculated after background subtraction.
Gene expression values were quantified by the two-based log ratio of red channel intensity vs. green channel intensity, followed by LOWESS normalization (R-Aroma version 0.85) and MAD scaling to remove the intensity-dependent dye bias and print-tip variation. We restricted our analysis to genes for which the mean fluorescent hybridization signal intensity divided by the median background intensity was at least 1.5.
We used one-class significance analysis of microarrays (SAM, version 1.2), based on Tusher et al. (15), to identify differentially expressed genes between LE and GE in the receptive and nonreceptive groups. Comparison of the gene output from one-class analysis of both groups was performed to determine periimplantation-specific genes. SAM assigns a score using a permutation test (based on a modified t test) for each gene of a data set. In addition, SAM allows control of the false discovery rate (FDR) by setting a threshold (
) to the difference between the actual test result and the result from repeated permutations of the tested groups. A q value (percent) was assigned to each detectable gene (i) in the array and represents the modified P value, measuring the FDR at which a gene is called significant. As the relative difference of each gene (di) increases (di > 0), the q value decreases.
To determine gene identity, gene bank accession numbers were used in searches of three independent databases: 1) http://genome-www5.stanford.edu/cgi-bin/source//sourceBatchSearch, 2) http://geneontology.org, and 3) http://microarray1k.aecom.yu.edu. DNA sequencing of all the targets used in PCR and in situ validation protocols was performed and revealed more than 95% homology with the published sequences. Using the information gathered from such databases as well as independent research, significant genes in the output from SAM were categorized into common gene function groups for each epithelial cell type.
QR-PCR
To determine the validity of significant gene expression array profiles reported by SAM, quantitative RT-PCR was performed in real time (QR-PCR). To establish the QR-PCR assay, a mass screening of 30 genes was first performed [QR-PCR screen (SPCR)] that included positive and negative controls. The positive control genes known to show preferential expression in the uterine LE or GE under receptive conditions were Irg1/Calb1 and Il6st, respectively. Genes that were shown by array to have similar expression levels in the epithelial cell types such as cathepsin L (Ctsl) and cyclin F (Ccnf) were used as negative controls (data not shown). LE and GE LCM samples were pooled from receptive uteri of three animals for each cell type. Hence, one sample of cDNA was used to test mRNA expression levels of all genes incorporated in the SPCR (sample size n = 1). The cDNA was made from nonamplified RNA as described.
Several gene targets in the receptive group that showed relatively high SAM scores were selected for additional mRNA analysis using the established QR-PCR assay (QPCR). This differed from the SPCR in that LCM samples were not pooled. LE and GE LCM samples were collected from receptive uteri of four independent animals for statistical analysis, LE and GE were analyzed as paired samples for each animal (sample size n = 4).
The RNA pellet was resuspended in 10 µl RNase/DNase-free water. First-strand cDNA was reverse transcribed employing a Superscript II RNase kit (90 min at 43 C) (Invitrogen). PCR was performed in real-time using the Prism GeneAmp 5700 (Applied Biosystems, Foster City, CA) to allow amplicon quantification according to the manufacturer’s instructions. The 10-µl PCR mixture consisted of 1x SYBR green PCR master mix (Applied Biosystems), cDNA template, forward and reverse primers (ranging from 0.5–1 µM), and 2 µl cDNA. The PCR cycle conditions were one cycle of 50 C for 2 min, one cycle of 95 C for 10 min, and then 40 cycles at 95 C for 15 sec followed by 60 C for 1 min. Primer pairs specific for published cDNA sequences (obtained from http://www.ncbi.nlm.nih.gov/) were designed using Primer3 software (http://frodo.wi.mit.edu/primer3/primer3_code.html). Details of the forward and reverse oligonucleotide primers that were synthesized by Invitrogen appear in Table 1
. The mRNA abundances were determined by normalization of the data to the expression levels of 18S rRNA (Universal 18S Primer Pair; Ambion) as well as hypoxanthine-guanine phosphoribosyltransferase (HPRT). In preliminary experiments, serial dilutions of cDNA were analyzed to confirm a linear relationship between cDNA content and quantity of product across the amplification range. As a result, the cDNA was diluted 1:8 for gene of interest and 1:16 for the housekeeping gene. Paired LE and GE samples from all four animals from both experimental groups were analyzed on the same plate and run in triplicate. The omission of cDNA as well as water-only samples served as negative controls. Reaction products were analyzed by dissociation curve profile and by electrophoresis in 2% agarose gel containing 0.5 mg/ml ethidium bromide and visualized over a UV light box.
|
CT method (16) that calculates 95% confidence intervals. The cycle threshold (CT) indicates the fractional cycle at which the amplified target reaches its threshold. The CT was determined from the exponential phase of the PCR by the SDS 2.2 software. The CT value for a sample is calculated by subtracting the CT of each gene of interest from that of the housekeeping gene. The CT of a gene of interest is calculated by subtracting the CT of the GE sample from the CT of the LE sample. An overall CT across all of the GE/LE pairs for each gene of interest can then be generated. Thus, 2–CT represents the fold change in gene expression of the LE over the GE in the receptive endometrial group.
In situ hybridization
Sense and antisense probes were generated from linearized plasmids that contained SP6 or T3 and T7 RNA polymerase sites. The incorporation of digoxigenin (DIG)-labeled nucleotides was performed with the DIG RNA labeling kit (Roche Applied Science, Indianapolis, IN). Uteri were collected and frozen as outlined above from animals at d 1–5 of pregnancy. In situ hybridization was performed using 14-µm uterine cryosections mounted on Plus slides (Fisher Scientific, Hampton, NH). PBS (pH 7.4) was used for all washing steps before probe application.
Sections were heated at 50 C for 2 min, air dried, and incubated briefly in chloroform to remove endogenous lipid. After fixation (4% paraformaldehyde, pH 7.4) and acetylation (0.5% acetic anhydride in 0.1 M triethanolamine, pH 8.0, and 0.4% concentrated HCl), each tissue was covered with 100 µl of the hybridization cocktail and a Hybrid plastic coverslip (Sigma) and incubated in a humidified chamber at 58 C for 2 h. The hybridization cocktail consisted of 2x standard saline citrate (SSC) (20x SCC contains 3 M NaCl, 300 mM Na3C6H5O7), 50% formamide, 5x Denhardt’s solution (Sigma), and 1 mg/ml salmon sperm DNA (Invitrogen). Incubations in hybridization buffer that contained approximately 0.5 µg/ml of denatured RNA (65 C for 10 min) antisense and sense probes followed for 16 h at 58 C as above.
Slides were washed twice in 2x SSC for 30 min, at 58 C, and then at 37 C. Sections were treated with RNase A (20 µg/ml in 2x SCC; Sigma) at 37 C for 30 min, washed once in 2x SSC for 10 min at 37 C, and washed twice in 0.2x SCC at 58 C for 30 min. After washing with buffer 1 [0.1 M Tris (pH 7.6), and 0.15 M NaCl] for 2 min at room temperature (RT), the slides were incubated in buffer 2 [0.5% blocking reagent (Roche Molecular Biochemicals Co., Indianapolis, IN) in buffer 1] at RT for 10 min. Immunological detection of the hybridized probe was carried out in a humidified chamber with anti-DIG antibody conjugated with alkaline phosphatase (1:5000; polyclonal, Fab fragments from sheep; Roche) diluted in blocking buffer. Slides were washed twice in buffer 1 for 15 min at RT and equilibrated in buffer 3 [0.1 M Tris (pH 9.5), 0.15 M NaCl, 50 mM MgCl2] for 5 min. For color development, slides were incubated overnight at RT in a nitroblue tetrazolium salt/5-bromo-4-chloro-3-indolyl phosphate tablet solution as prepared by the manufacturer (Roche) with levamisole added (200 µg/ml; MP Biomedicals Inc., Levine, CA) to block endogenous alkaline phosphatase activity. The dark purple reaction was stopped with Tris-EDTA solution (0.1 M Tris-HCl, 0.01 M EDTA, pH 8.0). Sections were briefly counterstained with 0.1% (wt/vol) methyl green, rinsed, and left to air dry before coverslipping.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
threshold of 0.41 were differentially expressed at significant levels just before blastocyst implantation. Of these, 153 and 118 had significantly greater expression in the LE and GE, respectively. Figure 1A shows the SAM plot for this comparison.
|
Comparison of LE vs. GE differential gene array patterns in both receptive and nonreceptive groups revealed that of 553 candidates, only one gene, Srrm1, was duplicated (showing greater expression in GE). Hence, the endometrial LE and GE populations showed diverse molecular signatures during early pregnancy and in mice of depleted hormone status. Therefore, the 270 genes showing differential LE vs. GE patterns in the receptive group were a unique gene subset to that of the nonreceptive endometrium and were likely to be regulated by ovarian steroids and related to blastocyst implantation. For this paper, only the genes generated from the receptive group will be the focus of further discussion.
From 271 genes, those of unknown function (these were mostly expressed sequence tags or Riken cDNA clones) were best represented in receptive epithelia (Fig. 1B
), comprising at least one quarter of the total number of genes differentially expressed in LE and GE populations (25.5 and 33.1%, respectively). Thus, the remaining 193 genes of known function that ranged in the level of differential expression from approximately 1.5- to 7-fold (Tables 2 and 3) were categorized into groups separately for the 114 genes related to LE (Table 2) and 79 genes related to GE (Table 3). A summary of the percentages of gene function groups represented in LE and GE cell types is shown in Fig. 1B. Because neither gene name nor function could be attributed to epithelial targets in the unknown category, these are not listed in Tables 2 and 3. Genes of functional groups that contained two or fewer members were pooled into a mixed-function group designated as other and comprised 9.8 and 17.8% of the LE and GE populations, respectively. The genes that could be assigned functions in the LE and GE showed great diversity with up to 14 different families represented.
|
|
|
|
). DNA-binding and transport-associated genes were also found in both LE and GE but to a lesser (<9%) and more varied extent (Fig. 1B). LE had approximately two times as many genes related to DNA binding and half the number of transport-related genes compared with GE (Fig. 1B). Immune-related genes and those involved in signal transduction comprised 7% or less of the total population of genes found in both LE and GE (Fig. 1B).
Unique to the LE population of receptive endometrium in increasing abundance were lipid-, metal ion binding-, carbohydrate-, and protein biosynthesis-related gene families comprising 5.9, 4.6, 4.6, and 2.0%, respectively (Fig. 1B). Conversely, metabolic related genes (4.2%) as well as a small population of mRNA processing genes (2.6%) were only found in GE (Fig. 1B).
Validation by QR-PCR
Validation of mRNA expression of receptive epithelial gene targets was chosen to be performed by PCR in real time owing to the enhanced sensitivity required when using small yields generated from laser-capture samples (17, 18). The QR-PCR assay was established in a preliminary screen (SPCR) using 30 epithelial targets from both LE and GE array lists. SPCR demonstrated that 22 genes correlated 100% with the array data showing significantly greater mRNA expression in LE than GE at periimplantation stages (Fig. 2A). Positive controls used for the SPCR included genes whose expression has been previously shown to be enriched in the LE during the receptive window: Irg1 (19), Calb1 (20), and Hdc (21).
|
).
The 100% correlation of array and SPCR data for 30 implantation-related epithelial candidates, including known positives for specific cell types, demonstrated the accuracy of this assay for determining the reproducibility of differential expression of array gene candidates from laser-captured samples. For the SPCR assay, both 18S rRNA and HPRT were used as reference genes for detecting quantitative mRNA expression in different epithelial cell types. As shown in Fig. 2B, both housekeeping genes produced very similar results for the mean difference in gene expression of 30 candidates with R2 and slope values of 0.9856 and 0.9589, respectively. Hence there was no significant difference in the mRNA expression in either epithelial cell types as determined by Wilcoxin sign rank test (P = 0.2948). Given the low yield of LCM epithelial samples, we used 18S rRNA for future quantitative mRNA assays, because it is more abundant than HPRT.
Expression of selected epithelial targets
A number of genes of interest that also showed higher ranked levels of fold expression in either epithelial cell type were selected for additional analysis. Quantitative mRNA assessment using greater sample numbers was performed using the established QR-PCR assay (QPCR; Fig. 2C). Furthermore, the temporal and spatial expression of novel implantation-related epithelial targets was examined across early pregnant stages using in situ hybridization.
Luminal epithelium
Lipid-related genes.
The lipid-related fraction of genes enriched in the LE of uteri from the receptive group was of interest to us given the existing knowledge and known progesterone regulation of uterine lipid content (23). More abundant lipid deposition in LE than GE as detected by Oil Red O staining (see supplemental data on The Endocrine Society’s Journals Online web site at http://endo.endojournals.org) presented technical difficulties for in situ hybridization. Background DIG-alkaline phosphatase staining produced by lipid-enriched epithelial sites was eliminated by chloroform pretreatment before fixation and hybridization steps as described in the supplemental data.
In the lipid-related genes, the mRNA expression of Etnk1 was significantly elevated by 3-fold in the LE of receptive endometrium. The hybridization pattern for Etnk1 appeared to be implantation specific (Fig. 3). Hence, there was minimal mRNA expression in the LE on d 2 of pregnancy, which reached peak levels on d 4 of pregnancy and then was reduced after implantation on d 5 (Fig. 3). The decidual cell reaction and GE peripheral to the implantation site on d 5 of pregnancy showed weak expression. The GE had minimal or no Etnk1 expression across pregnancy (Fig. 3). Uniform and weak hybridization staining in both LE and GE of immature (4 and 6 wk) and nonpregnant uteri (data not shown) confirmed that Etnk1 is maximally expressed in the LE compartment during the implantation window. The preferential expression of Etnk1 in the LE rather than GE just before implantation was confirmed by QR-PCR as shown in Fig. 2C.
|
).
Cytoskeletal/structural and cell adhesion.
Under receptive conditions, Sprr2a and Tspan8, were among the top bracket of cytoskeletal/structural and cell adhesion-related genes with the greatest levels of expression in the LE than GE. The greater than 4-fold mRNA expression in the LE than GE as shown by array just before blastocyst attachment was supported by QR-PCR findings for both genes (Fig. 2, A and C). Although there was greater mRNA expression of Sprr2a in LE than GE on d 4 of pregnancy as shown by QR-PCR (Fig. 2, A and C) and in situ hybridization findings, the expression was weak compared with the intense staining seen on d 1 of pregnancy (data not shown). Many other whole-uteri array mouse studies have also found that this gene, a structural component of the epidermal cornified cell envelope, is estrogen regulated (24, 25, 26), is primarily located in epithelia, and is implantation specific (27).
Defense responses.
Fyb, Wfdc2, and Hdc were among the immune-related genes whose expression was significantly up-regulated under receptive conditions by at least 3.5-fold more in the LE than GE. QR-PCR confirmed that all three genes were indeed preferentially expressed in the LE surface of P4E2-injected endometria of delayed mice (Fig. 2, A and C).
In the uterus, Hdc is a major source of histamine, released from mast cells of the uterine stroma (28), and is also capable of inducing an edematous response (29). Hdc expression has been shown previously to be more abundant in the LE on d 4–5 of pregnancy in the mouse (21). Thus, this gene serves as a good positive control for our data set as shown by QR-PCR (Fig. 2A).
Fyb, also known as SLP-76-associated protein of 130 kDa (SLAP-130) or adhesion and degranulation-promoting adapter protein (ADAP), is thought to be exclusively expressed in mononuclear cells of the hematopoietic lineage with preferential expression in T lymphocytes and monocytes (30). We confirmed by hybridization the preferential mRNA expression of Fyb in immune cells of the stroma, immediately underlying the basement membrane of the LE on d 4 of pregnancy (data not shown). These Fyb-positive immune cells, given their location, were obviously cocaptured with LE at the time of sample collection.
LE expression of the natural antimicrobial named Wfdc2 was maximal on d 3 of pregnancy, weakened slightly on d 4, and was further reduced on d 5 with isolated staining in the decidualized region of the endometrium as shown by in situ hybridization (Fig. 4A).
|
Enzyme.
Of all the genes significantly expressed by the array, Gsto1 and Cstd, members of the enzyme-related gene fraction, showed the greatest increase in mRNA expression (5- to 7-fold) in the LE rather than GE just before implantation. QR-PCR analysis of LE and GE isolated by LCM 8 h after P4E2 injection of delayed-implantation mice supported the preferential expression of these genes in the LE compartment (Fig. 2, A and C) as well as others such as Nudt4 (Fig. 2A).
The sudden and dramatic appearance of Gsto1 mRNA expression in LE on d 3–4 of pregnancy (Fig. 4B) with an absence of staining by d 5 as shown by in situ hybridization confirmed that this gene was specific to the implantation window in mice. Not previously described in the endometrium, Gsto1 is an unusual isoform of a member of the glutathione S transferase (GST) family, which functions as an antioxidant. The GE showed complete absence of expression for Gsto1 at all stages of pregnancy examined (Fig. 4B).
Growth factor and signal transduction.
In the category of growth factor and signal transduction, Sbk1 showed 3- to 4-fold significantly greater expression in the LE than GE just before implantation. Hybridization staining showed that on d 2 of pregnancy, the level of Sbk1 expression was similar in LE and GE but that there was a robust increase in the LE surface expression on d 3–4 with a sharp decline on d 5 of pregnancy (data not shown).
Other groups.
Under receptive conditions, calcium- and zinc-binding-related genes, Calb1/S100g and Car2, respectively, were elevated approximately three to six times higher in the LE than GE. The robust increase in and preferential mRNA expression of these genes in LE was confirmed by QR-PCR (Fig. 2A). Calbindin 28K is a well-established epithelial marker of endometrial receptivity in mice and primates (20).
Cob1 has recently been identified as an important gene in axial patterning (32). Its mRNA expression was confirmed to be significantly more expressed in LE than GE by QR-PCR (Fig. 2A) as well as in tissue sections on d 4–5 of pregnancy (data not shown). In addition, Cln5, a lysosomal-related gene, was found to have preferential expression in LE during the implantation window (data not shown), which was also supported by QR-PCR findings (Fig. 2A). Eef1g, involved in protein biosynthesis, was determined to be 4-fold preferentially expressed in LE of endometrium from P4E2-induced delayed-implantation mice as confirmed by QR-PCR (Fig. 2A).
Of the genes of mixed transport function, Fxyd4 had three times the level of expression in the LE than GE at periimplantation as supported by QR-PCR (Fig. 2A). LE and GE mRNA expression of Fxyd4 was minimal on d 2 of pregnancy but was significantly elevated in LE on d 3–4 with decreased expression on d 5 of pregnancy (data not shown). FXYD4, also known as corticosteroid hormone-induced factor (CHIF), belongs to the FXYD family of single-membrane-span proteins that have been shown to interact with NaK-ATP and have a general ion transport function (33). Despite having an expression pattern specific to the implantation window, Chif null mice are fertile (34).
Glandular epithelia
Defense response.
Rearranged mRNA from the Ig heavy chain (V region; mRNA IgH) and Lyzs were the most prevalent immune-related genes expressed up to 6-fold higher in isolated GE than LE populations obtained from receptive endometrium. The more abundant expression of these two genes in GE at periimplantation was confirmed by QR-PCR (Fig. 2, A and C) and in situ hybridization (only Lyzs shown in Fig. 5). The spatiotemporal mRNA expression pattern of Lyzs across pregnancy was particularly striking (Fig. 5). Weak staining of GE and underlying stroma was seen on d 2 of pregnancy. By d 3, there was punctate expression of Lyzs mRNA in individual cells of GE as well as in stromal cells underlying LE (Fig. 5). Late on d 4 of pregnancy, only punctuate staining of GE remained, which by d 5 was reduced in expression with weak but uniform staining of the decidual zone also (Fig. 5).
|
In addition, Arg2 mRNA was found to have 6-fold greater expression in LE than GE by array as supported by QR-PCR (Fig. 2A). Arg2 encodes an enzyme that has been recently assigned an immune-related role (36) that facilitates periimplantation events.
Enzyme.
With 4-fold greater expression in GE than LE of receptive endometrium, which was also supported by QR-PCR (Fig. 2, A and C), Sult1d1 stood out among the varied list of enzyme-related genes as a potential marker for implantation. Its differential expression pattern was one of the few examples of genes we studied that was significant only when delayed mice were treated with the combination of P4E2 and not P4 alone (data not shown). Indeed, its weak expression on d 2 of pregnancy but marked expression on d 3–4 characterized it as one of few known gland-specific markers of implantation (Fig. 6A). LE showed consistently weak Sult1d1 expression across pregnancy until d 5 when peripheral GE and decidual cells were also moderately stained (Fig. 6A).
|
). However, differential epithelial expression patterns were apparent only on d 4–5 of pregnancy as uniform, moderate staining was seen during earlier stages of pregnancy.
Other.
Of varying function, Slc23a2, Sh3tc2, and Fnbp1 are examples of other genes that showed 3- to 4-fold greater expression in GE than LE at periimplantation stages. QR-PCR confirmed the preferential GE expression of the first two of these genes (Fig. 2A) as did tissue-specific hybridization staining patterns for D430044G18Rik and Fnbp1 on d 4–5 of pregnancy (data not shown).
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Although we have largely focused on implantation, our array analysis also revealed that LE and GE have distinct molecular signatures irrespective of the uterine hormone milieu. The 282 genes that showed differential expression under hormone-depleted status (nonreceptive group) were not further investigated in this paper, but the result itself suggests that the molecular relationship between endometrial epithelial subtypes is important not just at periimplantation stages but across the mouse reproductive cycle.
Whether cDNA or oligonucleotide derived, all of the arrays performed to date in the adult mouse have used whole uterine samples in an attempt to find ovarian steroid hormone-regulated targets (25, 26, 27, 37, 38, 39, 40). Hence, our array study is unique in that we have focused on individual endometrial epithelial cell types to evaluate the extent of their relationship in inducing receptive uterine conditions. Given the huge variation in experimental design and data analysis of previous uterine array studies, it is difficult to cross-compare the results to find repeated patterns of true molecular markers of uterine responsiveness. Indeed, the implantation-specific expression patterns of certain epithelial cell types of a receptive endometrium may be diluted by the use of mixed-cell populations. For example, Wfdc2 (38) was a gene candidate found to be decreased in whole uteri from ovariectomized P4-treated mice. However, by in situ hybridization, we found its expression to peak on d 3–4 of pregnancy in the LE.
There is a consistency, however, in the published array data that suggests that less than 6% of all the mouse transcriptome (usually from 1.5–6%) represented on any given chip are regulated by ovarian steroids in the uterus. This is uniform across the many types of targets sought, whether estrogen (26, 37), progesterone (27, 38), or implantation specific (27, 39). This is also the scenario in the literature of human uterine array studies, except endometrial (not whole uterine) samples have been used, where once again, roughly 1–6% of the genome tested seem to be hormonally regulated (41, 42, 43, 44).
In this study, we reported that of 19,202 sequences expressed in the uterus of receptive groups, approximately 1.5% showed a significant level of differential expression. This is notable, given that other mouse (27) and human (42, 44) array studies of mixed-cell populations using whole endometrium or uteri taken from the receptive period have reported less than 2.5% hormonal gene regulation of the genome tested. The fact that we found 1.5% from hybridization of only two epithelial cell types is a testament to the diverse molecular nature of the LE and GE of a receptive uterus.
There is some degree of variation in the timing of implantation with natural matings. We used the delayed-implantation mouse model to induce uterine receptivity because this enabled us to control the timing, so that we could examine the endometrium during a defined 2- to 4-h preblastocyst attachment period. We found that the majority of genes examined were expressed in both GE and LE. However, significant differences in the molecular phenotype between epithelial subtypes were found, and this may be key to understanding uterine infertility and disease.
Under receptive conditions, we found more examples of genes solely expressed in the LE than GE. Furthermore, our cDNA array slides were not printed with Calca, and the Lif cDNA clone consisted of a hybrid sequence (our unpublished data), which made validation of GE-specific targets of receptivity difficult. However, other differentially expressed epithelial markers of uterine receptivity were found. For example, the significantly greater abundance of Ilst6 (22) mRNA in GE as well as the preferential LE expression of Hdc (21) and Calb1 (20) together with other implantation-specific genes, Ihh (45) and Tro (46), in both compartments were findings that supported the use of the delayed model to obtain receptive endometrium and validated the novel gene expression patterns that we have described.
In each epithelial cell type, there were distinct gene expression signatures. The increased presence of lipid bodies within the endometrial epithelium during early pregnancy has long been established (47, 48). These are spherical organelles composed of a core of neutral lipids such as triacylglycerols and cholesterol esters that often contain specific populations of proteins bound to their surface (49) like adipose differentiation-related protein. This is the first report of Adfp mRNA within the endometrium where we found it more abundantly expressed on the uterine LE surface than in GE at implantation. The exact biological function of adipose differentiation-related protein has not been determined but is believed to be in the intracellular mobilization and storage of neutral lipids (50).
There are also other lipid-metabolizing enzymes such as Etnk1 that are specifically expressed in the LE compartment. Etnk1 (ATP:ethanolamine O-phosphotransferase, EC 2.7.1.82) is the first enzyme in the CDP ethanolamine pathway, which is primarily important for the biosynthesis of phosphatidylethanolamine, an abundant phospholipid in eukaryotic cell membranes (51). Etnk1 activities in rodents were isolated from liver many years ago (52), but little is known regarding its expression in the adult mouse. The induction of Etnk1 primarily in LE during the implantation window could be associated with increased phosphatidylethanolamine synthesis. Overexpression of Drosophila Etnk1 in a NIH 3T3 cell line (53) protected these cells from apoptotic cell death. Peak levels of Etnk1 in LE on d 4 followed by a demise in expression on d 5 of pregnancy, is consistent with pre- and postapoptotic cell events in this cell type that allow the implanting embryo to invade. It is proposed that the effects of phosphorylated ethanolamines on cell survival may be mediated by alterations in the composition of the phospholipids membrane or by a direct block of a key apoptotic enzyme (54).
Another gene that appears to be involved in lipid metabolism is Irg1, which has already been shown to be highly up-regulated by P4 in the LE compartment at the implantation window (19, 55), a result confirmed in this study. There are other genes not found to be differentially expressed in our experiments that appear to play essential functions at implantation, such as the lipooxygenases that convert arachidonic acid to leukotriene A4 (55). The important role that lipid mediators such as prostaglandins and endocannabinoids are thought to be play in the molecular dialogue between embryonic and endometrial surfaces at implantation has been recently emphasized (56).
This study, together with others in both mouse and human uteri, has revealed a large number of immunologically relevant proteins expressed by the uterine epithelium. Previously, we have shown that two epithelial abundant cytokines, colony-stimulating factor 1 (57) and chemokine (C-C motif) ligand 26-like or eotaxin (CCL11) (58), are hormonally regulated in the endometrium where they effect macrophage recruitment. Indeed, the uterine epithelium seems to be a rich source of hematopoietic cytokines and growth factors throughout pregnancy, including the implantation phase (59, 60, 61). These factors are likely to regulate the innate immune responses to pathogen challenges that may occur during mating, and they may also be involved in wounding responses to the invading blastocyst.
In addition to these growth factors, there are also many proteins that play a role in mucosal immunity through their bactericidal functions (62). Among these, the whey acidic protein (WAP) motif protein, WAP four-disulfide core domain 2, identified in the current study together with other natural antimicrobial peptides, such as ß-defensins and secretory leuckocyte protease inhibitor (SLPI), are thought to be key mediators of the innate immune defense critical for successful implantation and pregnancy. The WAP domains are usually small secretory proteins that exhibit a variety of functions, including those that affect growth and differentiation (63). In fact, they are the major whey proteins in mouse milk (64). Little is known about the expression pattern and function of WAP four-disulfide core domain 2 in rodents, especially in reproduction. However, in recent array studies of primate endometrium, the expression of Wfdc2 was significantly up-regulated during the secretory phase of women (65) and monkeys (66), which coincided with the period of receptivity. Although Yanaihara et al. (65) localized peak expression of Wfdc2 to the epithelium, they did not distinguish between subepithelial populations. Whole-uterine array studies mining for P4-regulated genes in mice (38) failed to identify this gene as we have in isolated LE samples of receptive endometrium.
Lysozyme (LYZS), a bacteriolytic enzyme also known as muramidase, is typically expressed in hematopoietic cells of the macrophage and granulocyte lineage. In mammals, LYZS is also an important component of innate immunity against common pathogens at mucosal surfaces where it is confined to specialized epithelial cells including the serous (but not mucinous) glands (67) and Paneth cells (68) of the respiratory and gastrointestinal tract (terminal ileum), respectively. Although uterine Lyzs expression has been shown to vary across the mouse reproductive cycle (26), this is the first report to demonstrate epithelial-specific mRNA expression during the receptive window. The punctuate staining of select GE cells at periimplantation stages described here was striking and similar to LYZS protein expression of gut epithelium described in human patients with ulcerative colitis and Crohn’s disease (69). Changes in the cyclic and early pregnancy expression of endometrial Lyzs likely parallels the changing populations of macrophages with varying hormonal milieu (57).
The preferential expression of cytoskeletal and structural genes such as Sprr2a and Tspan8 on the LE surface at periimplantation stages highlights the specialized molecular phenotype that this structure adopts to induce receptivity. Associated with barrier function for protection against water loss (70) and infection (71), the down-regulated LE expression of Sprr2a mRNA leading up to d 5 of pregnancy in mice likely facilitates blastocyst implantation.
In the present study, we found Sult1d1 to be almost exclusively expressed in the uterine GE during the implantation period of mice. This enzyme belongs to a family of cytosolic sulfotransferases that mediate sulfation and play a critical role in the homeostasis, regulation, and detoxification of biologically active endogenous and environmental chemicals (72). Northern blot analysis of a wide spectrum of different tissues revealed that cDNA expression of the original mouse Sult1d1 clone (GB: AA244730) was limited to whole kidney and uterus (73, 74). This enzyme showed strong activities catalyzing the sulfation of several prostaglandins, confirming the first association of eicosanoids with sulfotransferase activity in mammals (75). The positive role of eicosanoids in implantation events in the uterus has already been discussed (56).
This discussion has highlighted just a few examples of the differentially expressed endometrial genes within the epithelia and during implantation. We have not addressed whether the expression of these genes is regulated to the same extent at the translational level, because commercial availability of antibodies to relatively unknown genes is limited. However, we provide significant evidence that under receptive conditions, epithelial-specific genes of the mouse transcriptome can be dually or independently expressed in LE and GE populations. The molecular relationship between these epithelial subtypes is dynamic, the consequences of which are not fully understood, at least in terms of fertility. In future studies of the endometrium, it is imperative that their identity as specialized epithelial cell types be acknowledged, so that they are not studied as one type of epithelia but as separate entities.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
This research was supported by grants from the National Institutes of Health (R01 CA 89617 to J.W.P.) and to the Albert Einstein Cancer Center (P30 13330). J.W.P. is the Sheldon and Betty E. Feinberg Senior Faculty Scholar in Cancer Research.
A.L.N. and J.W.P. have nothing to declare.
First Published Online April 20, 2006
Abbreviations: CT, Cycle threshold; DIG, digoxigenin; E2, estradiol-17ß; GE, glandular epithelium; HPRT, hypoxanthine-guanine phosphoribosyltransferase; LCM, laser-capture microdissection; LE, luminal epithelium; LYZS, lysozyme; P4, progesterone; QPCR, the established QR-PCR assay; QR-PCR, quantitative real-time RT-PCR, RT, room temperature; SAM, significance analysis of microarrays; SPCR, QR-PCR screen.
Received December 29, 2005.
Accepted for publication April 11, 2006.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CT method. Methods 25:402–408[CrossRef][Medline]
