Global Gene Expression Analysis to Identify Molecular Markers of Uterine Receptivity and Embryo Implantation

implantation, series of heterogeneous cell types of the uterus blastocyst. Although a of interactions during implantation, genetic evidence suggests that only a small number of them are critical to this process. To obtain a global view and identify novel pathways of implantation, we used a dual screening strategy to analyze the expression of nearly 10,000 mouse genes by microarray analysis. Comparison of implantation and interimplantation sites by a conservative statistical approach revealed 36 upregulated genes and 27 downregulated genes at the implantation site. We also compared the uterine gene expression profile of progesterone-treated, delayed implanting mice to that of mice in which delayed implantation was terminated by estrogen. The results show upregulation of 128 genes and downregulation of 101 genes after termination of the delayed implantation. A combined analysis of these experiments showed specific upregulation of 27 genes both at the implantation site and during uterine activation, representing a broad diversity of molecular functions. In contrast, the majority of genes that were decreased in the combined analysis were related to host immunity or the immune response, suggesting the importance of these genes in regulating the uterine environment for the implanting blastocyst. Collectively, we identified genes with recognized roles in implantation, genes with potential roles in this process and genes whose functions have yet to be defined in this event. The identification of unique genetic markers for the onset of implantation signifies that genome-wide analysis coupled with functional assays is a promising approach to resolve the molecular pathways required for successful implantation.


Summary
Infertility and spontaneous pregnancy losses are an enduring problem to women's health.
The establishment of pregnancy depends on successful implantation, where a complex series of interactions occur between the heterogeneous cell types of the uterus and blastocyst. Although a number of genes are implicated in embryo-uterine interactions during implantation, genetic evidence suggests that only a small number of them are critical to this process. To obtain a global view and identify novel pathways of implantation, we used a dual screening strategy to analyze the expression of nearly 10,000 mouse genes by microarray analysis. Comparison of implantation and interimplantation sites by a conservative statistical approach revealed 36 upregulated genes and 27 downregulated genes at the implantation site. We also compared the uterine gene expression profile of progesterone-treated, delayed implanting mice to that of mice in which delayed implantation was terminated by estrogen. The results show upregulation of 128 genes and downregulation of 101 genes after termination of the delayed implantation. A combined analysis of these experiments showed specific upregulation of 27 genes both at the implantation site and during uterine activation, representing a broad diversity of molecular functions. In contrast, the majority of genes that were decreased in the combined analysis were related to host immunity or the immune response, suggesting the importance of these genes in regulating the uterine environment for the implanting blastocyst. Collectively, we identified genes with recognized roles in implantation, genes with potential roles in this process and genes whose functions have yet to be defined in this event. The identification of unique genetic markers for the onset of implantation signifies that genome-wide analysis coupled with functional assays is a promising approach to resolve the molecular pathways required for successful implantation.
Introduction regulatory steps is necessary to further understand the biologic basis for the establishment of pregnancy or the underlying causes of pregnancy failures. In this respect, two recent reports highlight gene expression profiling in the post-implantation period (15,16). However to our knowledge, no such analysis at the onset of implantation has been reported. To address this issue, we employed two complementary strategies using murine GeneChip Expression Arrays (Affymetrix, Santa Clara, CA) to determine global gene expression profiles during implantation in mice. The first approach compared RNAs from implantation and interimplantation sites to identify genes that are specifically up-or downregulated at the implantation site. A second analysis compared RNA from P 4 -primed pregnant uteri with delayed implantation with that of P 4 -primed uteri after estrogen activation. Several genes with known expression status at the implantation site were detected. In addition, cell-specific expression patterns in the implantation and interimplantation site were observed for four candidate genes, confirming the validity of this approach. Mice with delayed implantation expressed a large number of genes associated with immunity or immune responses. The suppression of these genes at the implantation site also suggests that this site is immunologically privileged during early pregnancy and that modulation of the immune response is an active process during implantation. There were 81 genes whose expression was affected in both analyses. These results suggest that pan-genomic gene expression profiles are a promising approach for the identification of markers of uterine receptivity during implantation, and that multiple screening strategies yield a distinct set of candidate genes that appear to be critical in early pregnancy. by guest on July 9, 2020 http://www.jbc.org/ Downloaded from

Animals
All experiments were conducted in accordance with NIH standards for the care and use of animals. Mice were killed by cervical dislocation or were anesthetized for survival surgery with avertin. Adult virgin CD-1 female mice were mated with fertile males of the same strain to induce pregnancy (day 1 = vaginal plug).
Increased stromal vascular permeability at the site of initial contact of the blastocyst with the uterine luminal epithelium is the first visible sign of the implantation process (2300-2400 h on day 4) and can be monitored by an intravenous injection of a blue dye (12,13,17). For the first analysis, implantation and interimplantation sites were divided by sharp dissection at 2300-2400h on day 4 (n=12 mice). Uterine segments included uterine myometrium, stroma and epithelium. Implantation sites also included blastocysts. Because there are normal variations in the timing of implantation, only those uteri with uniformly distinct blue bands were included, while regions with embryo crowding were discarded (Figure 1).
In the second analysis, P 4 -primed uteri of mice with delayed implantation were compared to P 4 -treated uteri after estrogen activation. To induce delayed implantation, pregnant females were ovariectomized on the morning of day 4 of pregnancy (0900 h) and given daily subcutaneous injections of P 4 (2 mg/mouse in 0.1 ml sesame oil) from days 5-7 (13,14). To terminate delayed implantation and induce blastocyst activation, a single subcutaneous injection of estradiol-17β (25 ng/mouse in 0.1 ml oil) was given to one group of animals at the same time as P 4 injection on the third day of delay (day 7). Whole uteri (n=6) were collected from each of these two groups 12 h after the last injection of steroids.

Sample Preparation
Uterine tissues were flash frozen at the time of dissection and stored at -80C. Specimens of implantation and interimplantation regions, and of delayed and activated uteri were separately pooled and total RNA was extracted in Trizol reagent (Life Technologies, Gaithersburg, MD) according to the manufacturer's recommendations. An additional RNA cleanup step was performed using the Qiagen (Chatsworth, CA) RNeasy total RNA isolation kit. Total RNA (10 µg) from each group was used to generate cDNA using the Superscript Choice System (Life Technologies). First-strand synthesis was performed using a T7-(dT) 24  and all preparations met Affymetrix s recommended criteria for use on their expression arrays.

GeneChip Hybridization and Statistical Analysis
Each cRNA (15 µg) preparation was used to inoculate Murine U74A GeneChip Expression Arrays (Affymetrix, Inc.) and the hybridization, staining, scan and analysis were conducted per recommended protocols. An Affymetrix software filter was applied to mask transcripts with incorrect orientation in the public databases. Although numerous ESTs were differentially expressed, only those transcripts with known identities are reported herein. Three replicate hybridizations were performed using each of the four pooled RNA samples (implantation, interimplantation, delayed, activated) to establish the reproducibility of our 7 results. Alterations in RNA transcript levels were analyzed using the Affymetrix Analysis Suite 4.0 software. Differences in levels of fluorescent intensity, which represents levels of hybridization, between the 25 base pair oligonucleotides and their mismatches, were analyzed by multiple decision matrices to determine the Presence or Absence of gene expression and to derive an Average Difference score representing the relative level of gene expression.
Background and noise corrections account for nonspecific binding and minor variations in hybridization conditions. Values for the mean and standard deviation of the three replicate Average Difference scores were calculated for each gene on the GeneChip. Comparison between groups was performed by Student s t-test (P<0.05 considered significant). The Fold Change in expression between groups was calculated from the mean Average Difference scores.
A second approach was used to verify the identification of differentially expressed genes.
Affymetrix algorithms produced statistical decisions for an Increase or Decrease in expression when comparing the hybridization results of any two samples. Thus, paired comparisons of three replicate hybridizations resulted in 9 possible outcomes for a single analysis (e.g., implantation vs interimplantation). Transcripts with statistically significant expression (by t-test) that were also differentially expressed in 4 or more of the nine pair-wise comparisons (Increase or Decrease) were considered candidates for further evaluation. Others have used this counting approach using somewhat higher cut-off points (18). However at higher threshold levels, we noted that a number of genes, known to be expressed at the implantation site, were excluded.

Gene expression is altered at the onset of natural implantation (implantation vs interimplantation sites)
Hybridization intensity to the arrays was uniform for housekeeping genes such as GAPDH, cyclophilin, and a large number of ribosomal proteins, indicating that the expression data in the array hybridization experiments are consistent with other standards for studying gene expression.
The relative levels of gene expression at the implantation and interimplantation sites were first compared by plotting the Average Difference values for one individual array hybridization experiment against another and determining the Presence or Absence of gene expression for the entire array (Figure 2). Genes that were considered Absent or Marginally expressed were widely dispersed (black points), whereas genes that were declared Present in both hybridizations were tightly grouped (red points). An increase of more than 2-fold in the Average Difference score indicated genes with significantly higher expression at the implantation site (above the curve) or at the interimplantation region (below the curve) (Figure 2). These results show that a vast majority of the genes on the chip have similar expression patterns (Present, Absent or Marginal) and their relevance to implantation is questionable. A small number of genes were Present in both (red points) or Present in one but Marginal or Absent in the other (blue points) that also had greater than a two-fold difference in the level of expression. These genes were considered as potential candidates for further evaluation. Surprisingly, a symmetrical distribution of these candidate genes was observed, suggesting that an equal number of genes are upregulated in the implantation and interimplantation regions.
Genes with statistically significant differences in expression at implantation versus interimplantation sites were identified by comparison of replicate hybridizations and by a statistical decision for an Increase or a Decrease in gene expression. By t-test alone, there were 293 upregulated and 370 downregulated genes at the implantation site. A second statistical approach was performed to provide an additional objective analysis. We used a threshold value of 4 of the 9 possible outcomes, which resulted in 49 upregulated and 60 downregulated genes and included several known genes that are expressed during implantation or are considered biologically relevant to the implantation process. A combination of t-test and counting approaches identified 36 genes that were upregulated at the implantation site and 27 genes that were downregulated (Tables 1, 2).
Differentially expressed genes were categorized based on the best available information regarding their biologic functions. Genes with multiple functions were assigned to a single category. Many genes that are known to be associated with the implantation process fall into categories similar to the genes we detected with increased expression at the implantation site, including growth factors/cytokines and their receptors, transcription factors, genes encoding structural proteins, or genes associated with cell proliferation. We also observed upregulation of a group of calcium-related genes, including Bip, Sik similar protein (Sik-SP), and calcineurinand calcyclin-related proteins. Genes encoding Bip and Sik-SP showed highly localized expression at the implantation site, confirming our array results (Figure 3). These genes are of interest since calcium is an essential modulator of enzyme functions and signal transduction, and there is evidence that genes involved in calcium regulation play an important role during implantation (21)(22)(23)(24)(25).
Previous efforts to identify novel genes during the periimplantation period have primarily focused on the implantation site, with less attention paid to genes that are expressed at the interimplantation region. Genes with increased expression at the interimplantation site may act to guide the blastocyst to specific sites for implantation or be important for embryo spacing.
Aberrant expression of these genes may be as detrimental to implantation as the loss of genes that are expressed at the implantation site. In our gene array experiments, we observed a 5-fold decrease in levels of Cyp1b1 expression at the implantation site. Cyp1b1 is a member of the cytochrome P450 system that converts primary estrogens to their active metabolites, catecholestrogens, which are important for blastocyst activation (20). In situ hybridization showed that Cyp1b1 is restricted to the subepithelial stroma of the interimplantation region ( Figure 3), in agreement with the decreased expression levels seen in the array experiments. A similar expression pattern was noted for the PGE 2 receptor subtype, EP 4 , which showed a nearly 2-fold decrease by gene array. The overall diversity of genes with this pattern of expression suggests that further investigation of interimplantation-specific genes is warranted.

Gene expression is altered in an experimental model of implantation (delayed versus induced implantation)
To identify genes that are upregulated at the time of blastocyst activation for implantation, RNAs from P 4 -primed delayed implanting uteri and P 4 -primed uteri after estrogen treatment were obtained for comparative analysis. Scatterplot analysis of genes considered Absent or Present showed that overall gene expression patterns in delayed and activated uteri were closely correlated (Figure 2). The tight distribution of overall expression is similar to that seen in the implantation versus interimplantation analysis. The overall similarity of the two different scatterplots is not surprising, since the ultimate outcome, implantation, is similar in both models and the experiments were designed to be complementary and provide a dual approach to identify novel implantation-specific genes.
Statistical analysis by t-test alone identified 409 upregulated and 550 downregulated genes in comparisons of delayed versus activated uteri. However, our combined statistical approach reduced the number of candidate genes to 128 upregulated and 101 downregulated transcripts after termination of the delayed implantation by estrogen (Supplement, Tables 1, 2).
Mice with delayed implantation expressed a large number of genes associated with host immunity or the immune response (n = 48) compared to the uteri of mice after estrogen activation (n = 3) (Supplement , Table 1). Furthermore, nearly 50% of the genes with significant expression during delayed implantation have some immune-related function (48/101).
Specific roles for these genes during implantation have not been described.
There were striking differences in the functional categories of delayed versus activated genes. In general, more DNA processing, cell cycle-associated genes and a larger number of enzymes were observed after initiating the process of implantation (Supplement, Table 2). This shift in gene diversity suggests that embryonic activation and uterine preparation for the onset of implantation is mediated by a subset of genes that requires estrogen. In this respect, the gap junction proteins, connexin 26 and connexin 43, are implicated in implantation and their regulation is influenced by steroid hormones (26). In our array experiments, the expression of connexin 26 increased over 8-fold, while the expression of connexin 43 showed a greater than 2.5-fold increase after estrogen activation. The PGE 2 receptor subtype EP 2 , is the only other gene that was detected in our delayed versus activated gene array whose expression is also known to be induced at the site of the implanting blastocyst after termination of delayed implantation (27). Overall, our results show that delayed implantation is a valuable model to dissect the molecular aspects of implantation, and compare the distribution of genes that are expressed during dormancy or active implantation.

Comparison of gene expression in natural and delayed implanting models of pregnancy
Although implantation is the eventual outcome of the both models, it is unclear whether molecular mechanisms underlying natural implantation and induced implantation are similar. To determine whether genes upregulated in the delayed uterus after estrogen activation are similar to the group of genes with increased expression at the implantation site, the results of both hybridization analyses were combined to highlight genes that were differentially expressed in both comparisons. The intersection of these sets revealed 244 genes with significantly altered expression (t-test only) in both models (Figure 4). Considering only those transcripts with a greater than 2-fold difference in Average Difference scores, there were 54 genes that had significant expression at the interimplantation site and during progesterone-primed implantation delay ( Table 3). By similar criteria, we also observed 27 genes that had increased expression at the implantation site and after estrogen activation ( Table 4). Among these, only connexins 26 and -43, amphiregulin and nexin-1 are associated with implantation (26,28,29). The importance of the remaining genes in implantation awaits further investigation. Nonetheless, these results show that a dual screening strategy identified a small number of candidate genes that are likely to have significant roles in implantation.

Discussion
The survival of any species depends on stable mechanisms for reproduction. Thus, it is assumed that essential mechanisms for embryo implantation must be supported by redundant pathways to ensure the conception of new offspring. This predicts that a large number of genes that are important for implantation remain to be identified. Previous approaches to investigate implantation have generally relied on the analysis of individual candidate genes or gene families.
We used DNA microarray technology to screen a large cross-section of the murine genome to identify novel implantation-specific genes. Our present investigation has identified genes with recognized roles in implantation, genes with potential roles in this process and genes whose functions have yet to be defined. In addition, a small number of genes showed significantly altered expression during both natural and induced implantation.
The process of implantation involves cell-cell interactions between the blastocyst and uterus, cell-type specific proliferation and differentiation of the uterus, and immunological responses of the mother to the semi-allogenic embryo. Our data show a broad diversity of genes that are modulated during implantation. A recent report described a microarray-based approach to identify genes in the uterus during the post-implantation period (15). They observed 192 genes with increased expression and 207 genes with decreased expression levels. Similar to our results, genes typically showed 1.5 to 3-fold induction at the implantation site. Surprisingly, there are very few genes that were mutually identified in both studies. However, they compared uterine gene expression profiles on the evening of day 4 to those on day 6 of pregnancy. With this approach, both implantation and interimplantation regions would be included in a single sample so that no distinction could be made for gene localization around the implanting blastocyst. Our in situ hybridization results show the importance of differentiating these two sites. In addition, uterine horns were flushed in this study (15), and the uteri were split and the luminal surface was scraped to remove conceptuses. Physical disruption of the uterine epithelium would likely result in different gene expression profiles. We designed our experiments to include the implanting blastocyst in both analyses. The presence of blastocysts in our first (implantation vs interimplantation) and second (delayed vs activated) analyses serves to strengthen our approach by including the embryonic genome and profiling the expression of embryonic factors that may be significant to implantation. Moreover, we analyzed intact uterine horns to preserve an undisturbed relationship between the uterine myometria, stroma, and epithelium and its intimate contact with the blastocyst. In contrast, Yoshioka et al. (15), focused on two distinct time points in pregnancy, resulting in the identification of genes with differential expression between implantation and decidualization, rather than implantation site-specific genes.
We also identified genes that are differentially expressed in delayed versus estrogenactivated uteri. There are previous reports that genes encoding the EGF-like growth factors, cytokines and other inflammatory mediators, extracellular matrix proteins, cell cycle molecules and immunoregulatory proteins are expressed at the site of estrogen-induced implantation in a pattern similar to their expression in natural implantation (17,27,30,(31)(32)(33)(34)(35)(36)(37)(38). In this respect, our microarray results are consistent with the previously described expression of several steroid hormone-sensitive and implantation-specific genes, including Cyp1b1, connexin 26 and connexin 43, Sik-SP, the prostaglandin receptor EP 2 , and histidine decarboxylase (20,21,26,27,39). However, we failed to detect the anticipated changes in the expression of cyclooxygenase-2, perlecan, trophinin, HB-EGF, LIF and a number of other hormoneresponsive, implantation-associated genes at the implantation site. This is perhaps due to their highly restricted expression around the implanting blastocyst, resulting in the dilution of implantation-specific RNAs in a large pool of RNAs derived from other uterine cell types.
Indeed, we have previously observed that changes in the expression of cyclooxygenase-2 and HB-EGF and several other implantation-specific genes could not be detected by Northern hybridization, but showed discrete upregulation at the implantation site as observed by in situ hybridization (17,(30)(31)(32).
A significant shift was noted in the diversity of genes expressed in delayed implantation uteri compared to estrogen-activated uteri. In particular, mice with delayed implantation expressed a large number of genes associated with immunity and/or the immune response. The suppression of these genes at normal implantation sites and after estrogen-activation ( Table 3) suggests that the implantation site is immunologically protected. Although large numbers of maternal natural killer cells are recruited to the uterine deciduum 48 hours after embryo attachment (40), leukocytes and other bone marrow-derived cells migrate away from the site of blastocyst attachment at an earlier time during the onset of implantation (41,42). Reduced expression of numerous immune-related genes at the implantation site suggests that immunomodulatory cells, even if present, remain quiescent with the onset of implantation. Thus, reduced expression of these genes is an important finding, since it is still unclear how the embryo escapes maternal immunological responses during pregnancy (43). The mechanism of downregulation of these genes at the implantation site is unknown, but elaboration of immunosuppressive signals from active blastocysts cannot be ruled out (44,45). Our data suggest that modulation of the immune response is an active process prior to or during blastocyst implantation.
There were 81 genes with differential expression at the implantation site during both natural and induced implantation, suggesting their importance for implantation. Connexin 26 and -43 are gap junction proteins that are influenced by ovarian steroids, accumulate in the stroma around the implantation site, and are upregulated during experimentally induced decidualization (26). Their expression was significantly increased at the implantation site and in the uterus after estrogen activation. These observations coincide with a growing body of evidence for the role of structural genes in the establishment of pregnancy (46). Amphiregulin and nexin-1 were the only other implantation-associated genes that had increased expression during both natural and induced implantation ( Table 4). Amphiregulin is a member of the EGFfamily of growth factors that becomes intensely localized to the uterine luminal epithelium surrounding the blastocyst at the onset of implantation (28). Nexin-1 is a serine protease inhibitor that regulates processing of plasmin, thrombin, urokinase, plasminogen activators and other proteases, and is upregulated during implantation (29). Tight regulation of proteases and their inhibitors is considered an important aspect of embryo-uterine interaction at the site of implantation (47). Collectively, our results suggest that other genes identified in the composite analysis are physiologically relevant to the implantation process.
The remaining genes identified by the combined screening approach do not yet have any recognized roles in implantation. However, many of the genes identified with increased expression in both analyses ( Table 4) are associated with cell proliferation (spermidine synthase, PCNA), cell cycle regulation (cyclin E2), inflammation or tumor biology (ribonucleotide reductase M2, NM23), and DNA replication, synthesis, and/or repair (ribonucleotide reductase M2, CDC46 and Mcm homologs, topoisomerase, PCNA, FEN-1). Since the process of implantation is considered a proinflammatory response and involves uterine proliferation, differentiation and apoptosis, these genes could well be very important for various aspects of this process. It is interesting to note that PCNA can directly bind to FEN-1 to stimulate its nuclease activity during base excision repair of damaged DNA (48), and that ribonucleotide reductase is the rate-limiting enzyme that provides the essential deoxynucleotides for new DNA synthesis and DNA repair. All three of these genes showed upregulated expression during implantation and may act in a concerted fashion for remodeling of the uterine epithelium and stroma during blastocyst attachment and invasion. However, the diverse functions for each of these genes preclude any speculation as to their specific significance to implantation.
In summary, we employed a global gene expression strategy to identify novel genes in the implantation process. Our results show an equal number of genes that are upregulated or downregulated at the implantation site and a small number of candidate genes that have significant changes in their expression during natural and estrogen-induced implantation. A better understanding of the molecular mechanisms of embryo-uterine interactions during implantation will provide insight into the high rate of spontaneous pregnancy losses. Attempts to resolve these complex signaling networks will likely benefit from a genome-wide approach coupled with functional assays, and facilitate new methods to address infertility and contraceptive challenges in women's health.   Table 2. Genes with significantly decreased expression at the implantation site.
RNAs from implantation and interimplantation sites were obtained for GeneChip hybridization. The fold change in gene expression was determined by comparison of mean Average Difference scores. Differentially expressed genes were identified by two statistical methods. By definition, genes with decreased expression at the implantation site also have increased expression in the interimplantation region (see Figure 3). *, genes with known expression during the periimplantation period.  Table 3.

Fold change of genes that showed decreased expression at the implantation site and after initiation of estrogen-induced implantation.
Differentially expressed genes in the first (implantation vs interimplantation) and second (activated vs delayed) analyses were compared to identify common patterns of gene regulation. Genes with significant alterations in expression level (t-test only) and at least a 2-fold change in one of the analyses are shown. Negative fold change values indicate decreased expression relative to the baseline condition (interimplantation sites or delayed uteri). By definition, these genes also have increased expression at the interimplantation site and during P 4 -induced delayed implantation. *, genes with known expression during the peri-implantation period.  Table 4

. Fold change of genes that showed increased expression at the implantation site and after initiation of estrogen-induced implantation.
Differentially expressed genes in the first (implantation vs interimplantation) and second (activated vs delayed) analyses were compared to identify common patterns of gene regulation. Genes with significant alterations in expression level (t-test only) and at least a 2-fold change in one of the analyses are shown. *, genes with known expression during the peri-implantation period.           Table 1. Genes with significantly decreased expression during estrogeninduced termination of delayed implantation. RNAs from P 4 -primed delayed implanting uteri and delayed implanting uteri after estrogen treatment were obtained for GeneChip hybridization. The fold change in gene expression was determined by comparison of mean Average Difference scores. Differentially expressed genes were identified by two statistical methods. Negative fold change values indicate decreased expression relative to estrogen-activated uteri. By definition, these genes also have higher expression during delayed implantation. *, genes with known expression during the periimplantation period.