Gene induction and categorical reprogramming during in vitro human endometrial fibroblast decidualization

ANOOP K. BRAR1, STUART HANDWERGER1, CHERIE A. KESSLER1 and BRUCE J. ARONOW2

1 Departments of Endocrinology
2 Molecular and Developmental Biology, Children’s Hospital Research Foundation and Department of Pediatrics, University of Cincinnati College of Medicine, Cincinnati, Ohio 45229


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Gene induction and categorical reprogramming during in vitro human endometrial fibroblast decidualization. Physiol Genomics 7: 135–148, 2001. First published September 21, 2001; 10.1152/physiolgenomics.00061.2001.—Human decidual fibroblasts undergo a differentiative commitment to the acquisition of endocrine, metabolic, and structural cell functions in a process known as decidualization. Decidualization is critical for embryo implantation and placental function. We characterized gene expression pattern kinetics during decidual fibroblast differentiation by microarray analysis. Of 6,918 genes analyzed, 121 genes were induced by more than twofold, 110 were downregulated, and 50 showed biphasic behavior. Dynamically regulated genes were could be fit into nine K-means algorithm-based kinetic pattern groups, and by biologic classification, into five categories: cell and tissue function, cell and tissue structure, regulation of gene expression, expressed sequence tag (EST), and "function unknown." Reprogramming of genes within specific functional groups and gene families was a prominent feature that consisted of simultaneous induction and downregulation of a set of genes with related function. We previously observed a conceptually similar process during fetal trophoblast differentiation, in which the same phenomena applied to different genes. Of the 569 dynamically regulated genes regulated by either model, only 81 of these were in common. These results suggest that reprogramming of gene expression within focused functional categories represents a fundamental aspect of cellular differentiation.

decidual fibroblast differentiation; gene regulation; pregnancy; microarray


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
DURING THE LUTEAL PHASE of the menstrual cycle in which the uterus is prepared for a potential embryo implantation, endometrial stromal cells differentiate into decidual cells by the process known as decidualization (14, 30). If implantation and pregnancy ensues, then the decidual cells become the predominant cell type of the uterine lining. Decidualization is critical for successful uterine implantation of the blastocyst and abnormalities of decidualization likely are a common cause of spontaneous abortion and infertility. However, the molecular basis of decidualization has not been defined at the genomic level to permit identification of potential molecular abnormalities responsible for unsuccessful decidualization and implantation. A better understanding of the molecular events during decidualization will identify the molecules critical for the process and will delineate target genes acted upon by transcriptional factors and signaling mechanisms.

Most knowledge about the regulation of human decidualization has come from studies using in vitro model systems. Decidualization can be induced in primary cultures of endometrial stromal cells by incubation of the cells with progesterone in combination with estradiol or relaxin or with high levels of cAMP in the absence of exogenous hormone (22, 27, 41). The treated endometrial cells acquire the characteristic morphologic phenotype of decidual cells and express decidual cell-specific markers such as prolactin and insulin-like growth factor binding protein-1 (IGFBP-1). Decidualization can also be induced in primary cultures of human decidual fibroblasts (36) and in a permanent cell line of human endometrial fibroblasts (St-2 cells) (8). Treatment of primary cultures of decidual fibroblasts and St-2 cells with cAMP induces prolactin gene expression and decidualization, but treatment with progesterone, alone or in combination with estradiol, is without effect. Treatment of the cells with progesterone, however, potentiates the induction of decidualization in response to cAMP. Another model system, the N5 endometrial cell line, the cells of which have phenotypic features of primary cultures of decidualized human endometrial stromal cells and secrete low levels of prolactin and IGFBP-1 constitutively, is also decidualized further in response to cAMP (6).

Studies using these model systems have identified several genes induced during decidualization but have so far provided little information about those that are repressed or activated with differing kinetics during decidualization. To identify new genes that are dynamically regulated during decidualization and their expression patterns, we performed sequential DNA microarray analyses of human decidual fibroblasts decidualized in vitro by treatment with progesterone, estradiol, and cAMP using the Incyte human GEM-V microarray that contains 6,918 genes. In addition, we compared the expression profiles of the dynamically regulated genes during decidualization to those regulated during differentiation of cytotrophoblast cells. Our findings suggest that decidualization, like cytotrophoblast differentiation, is characterized by "categorical reprogramming" in which there is strongly diminished levels of genes within the same category as many of the major induced genes. However, differentiation of the two different lineages is highly divergent, with the same genes exhibiting nearly opposite regulation in the two lineages. The results show that decidualization and cytotrophoblast differentiation exhibit markedly different patterns of expression and demonstrate the extent to which differentiation programs can be intensely cell-type specific.


    MATERIALS AND METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

Preparation and culture of decidual fibroblasts.
Decidual fibroblasts were prepared as previously described (36). Permission to obtain term human placentas from uncomplicated pregnancies was approved by the Institutional Review Boards at the University of Cincinnati College of Medicine and the Children’s Hospital Medical Center, Cincinnati, OH. Briefly, the decidua parietalis tissue was dissected from the fetal membranes within 1–2 h after delivery, then treated with collagenase, and the cells were dispersed on plastic plates (107 cells/25 cm2 surface area) and cultured in RPMI 1640 medium (containing 50 mM HEPES, 2% NaHCO3, 25 U/ml penicillin G, 25 g/ml streptomycin, and 5 g/ml amphotericin B, pH 7.2) supplemented with 10% FBS. After 24 h, the nonadherent cells (which included most of the terminally differentiated decidual cells and bone marrow-derived cells) were freed by agitation and removed by medium exchange. Adherent cells were grown to confluence, split with trypsinization, and replated. By passage 3, essentially all cells were proliferating decidual fibroblasts with features typical of previous studies (36) in which cells appeared homogeneous by light microscopy, flow cytometry, and immunocytochemical analyses. Decidual fibroblasts do not express cell surface antigens typical of bone marrow-derived cells (CD45, CD14, and HLA-DR) or the cytoskeleton protein cytokeratin, but do express HLA-A, -B, -C, which are not present on trophoblast cells, and vimentin, which is a typical marker for fibroblasts (mesenchymal cells) and which is not present on epithelial cells.

The cells were plated in six-well tissue culture plates (2.5 x 106 cells/well; Falcon Plastics, Becton-Dickinson, Franklin Lakes, NJ) in FBS-supplemented medium. After allowing the cells to attach to the plates overnight, decidualization was induced by treatment with medroxyprogesterone acetate (1 µM), estradiol (10 nM) and dibutyryl-cAMP (50 µM). The medium was changed at 2, 4, 6, 9 and 12 days, and the experiment was terminated at 15 days. Cells from triplicate wells were collected at the time of each medium change, pooled, and stored at -70°C until the RNA was isolated.

RNA preparation.
Poly(A)+ RNA from cultured decidual cells was obtained using the Oligotex mRNA isolation kit (Qiagen, Valencia, CA) according to manufacturer’s instructions following initial total RNA isolation (11). The mRNA was purified by two passes through oligo d(T) columns.

RT-PCR analysis.
The relative amounts of prolactin and IGFBP-1 mRNA levels during decidualization of the decidual fibroblasts were determined by semiquantitative RT-PCR as described previously (39). For each assay sample, the amounts of prolactin and IGFBP-1 mRNAs were normalized to the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA content of the sample.

Northern blot analysis.
Verification of microarray results was accomplished by Northern blot analyses of RNA samples from an independent experiment in which decidual fibroblast cells were decidualized by a protocol identical to that used in the microarray experiment. The RNA samples from this experiment were analyzed for mRNA levels by Northern blot analyses using cDNA or DNA probes obtained from Incyte Genomics (St. Louis, MO): IGFBP-1 (Incyte clone 1448718), decorin (Incyte clone 3820761), epidermal growth factor (EGF)-containing fibulin-like extracellular protein-1 (Incyte clone 1798209), somatostatin (Incyte clone 2494617), tissue inhibitor of metalloproteinase 3 (TIMP3, Incyte clone 1998369), and endometrial bleeding associated factor (EBAF, Incyte clone 1865634). Equal loading and membrane transfer of RNA was verified by methylene blue staining of ribosomal bands. Quantitation of the signal on membranes was performed using a PhosphorImager screen (Molecular Dynamics, Sunnyvale, CA). The relative induction of the mRNAs determined by Northern blot analyses were then compared with that determined by microarray analyses.

DNA microarray.
DNA microarray analyses were performed using Cy3- and Cy5-labeled probes prepared from poly(A)+ mRNA from the fibroblast cells after 2, 4, 6, 9, 12 and 15 days of treatment. Microarray hybridizations were performed by Incyte Genomics (Palo Alto, CA) using the human GEM-V microarray. The reference "time 0" mRNA (Cy3) was prepared from untreated cells.

Primary data were examined using Incyte GEMTools software and GeneSpring software (Silicon Genetics, Redwood City, CA). Defective cDNA spots (irregular geometry, scratched, or <40% area compared with average) were eliminated from the data set. Data sets were subjected to normalization within each microarray experiment such that individual Cy5 spot intensities were multiplied by the ratio of the median intensity of the Cy3 channel to that of the Cy5 channel for all genes over the entire microarray (R = K x Cy5/Cy3). Each microarray contained 192 control genes present as nonmammalian single gene "spikes" or "complex targets." The complex targets consisted of probe sets that contain a pool of cellular genes expressed in most cell types. In addition, each experimental mRNA sample was augmented with incremental amounts of nonmammalian gene RNA (2x, 4x, 16x, etc.) to permit assessment of the dynamic range attained within each microarray. Little variation was observed across the microarray series with respect to the 192 control genes (not shown), providing support for interarray comparisons of temporally regulated genes.

We assembled a pool of genes that exhibited significant expression change during the cell culture period using a multistep procedure. Regulated genes were selected that exhibited a twofold increase or 50% decrease in their expression relative to the time 0 reference sample based upon the average gene expression value observed between any two adjacent time points. This maneuver filters out a significant noise component in the individual arrays but reduces the sensitivity to moderately regulated genes that are unique to a single time point. The dynamically regulated genes were clustered according to expression pattern dynamics by subjecting log2-transformed data series to K-means (17) and hierarchical tree clustering algorithms (13) as implemented in the GeneSpring program (Silicon Genetics). The hierarchical tree analysis was performed using a minimum distance value of 0.001, separation ratio of 0.5, and the standard correlation distance definition algorithm. The hierarchical tree structure was used to suggest approximate numbers of behavior sets for K-means clusters, and this was empirically confirmed by observing relatively consistent behaviors within the resulting sets. Too few groups resulted in sets composed of heterogeneously behaving genes, and too many groups led to adjacent sets that exhibited indistinguishable behavior.

Time-averaged microarray series.
Initial K-means analysis revealed that unique sets were formed on the basis of small behavior differences between subsequent time points that may have been attributable to technical variation or noise that was unrelated to biologic variation. For example, a gene that had a saw-tooth induction kinetic over the 0, 2, 4, 6, 9, 12, and 15 day time points would be grouped into a different set than one which had less of a saw-tooth component to its induction. We opted to decrease the noise contribution of the single microarray data point by using a time-averaged approach in which the average fold change of each gene at days 3, 5, 7.5, 10.5, and 13.5 was interpolated by averaging the two data points on either side. Although this decreases noise, it also decreases the ability to discern kinetic patterns that are confined to just one of the time points. However, based on the patterns of the vast majority of genes, the advantages of this approach appeared to greatly outweigh this potential disadvantage. A similar treatment was applied to the data set obtained from differentiating cytotrophoblast cells (2) to generate time-averaged values. This allowed for a comparison of regulatory behavior of genes between the decidual and trophoblast cells during differentiation.

Genes were classified into biological function groups by adaptation of the gene ontology classification we used previously in a study of human cytotrophoblast differentiation (2). To view all gene expression data, cluster groups, and the functional categories of the dynamically regulated genes, please refer to the Supplementary Material1 for this article, published online at the Physiological Genomics web site.


    RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Decidual fibroblast cells used in the DNA microarray studies responded to cAMP, progesterone, and estradiol exposure with a progressive change in morphology and induction of prolactin and IGFBP-1 gene expression as previously reported (Fig. 1) (36), modeling the in vivo maturation of a gravid uterine decidual cell. Immunocytochemistry revealed that all cells stained for vimentin. Both prolactin and IGFBP-1 mRNA levels were undetectable prior to treatment as measured by semiquantitative RT-PCR. IGFBP-1 mRNA levels increased during the first 2 days of exposure to cAMP and steroids and reached peak levels on day 9. Prolactin mRNA was first detectable at day 2 and then increased progressively to reach a peak at day 12.



View larger version (19K):
[in this window]
[in a new window]
 
Fig. 1. Validation of differentiation during in vitro culture of decidual fibroblasts. RT-PCR analysis of prolactin and insulin-like growth factor binding protein-1 (IGFBP-1) during decidualization of decidual fibroblasts. The relative levels of mRNA were determined at time 0 and 2, 4, 6, 9, 12, and 15 days from the same mRNA samples used in cDNA microarray analyses. RT-PCR mRNA measurements were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA in the same samples.

 
To analyze gene regulatory kinetics and ascertain the identities of the complement of genes dynamically regulated during decidualization, we used the Incyte human GEM-V microarray to examine gene induction and repression in hormone-treated cells relative to untreated decidual fibroblasts. mRNAs from a series of cell culture time points were isolated following the induction of decidualization and labeled with Cy5 using reverse transcriptase. mRNA derived from untreated cells was labeled with Cy3, and hybridizations were set up for days 2, 4, 6, 9, 12, and 15. We observed that 281 of the 6,918 genes on the microarray exhibited twofold or greater induction or 50% or more repression (Fig. 2). Of these, 127 genes were induced, and 154 were repressed. Over the culture period, the total numbers of regulated genes increased progressively with 103 of the upregulated genes and 110 of the downregulated genes exhibiting more than twofold changes at two or more time points, respectively. At day 2, 34 genes were induced twofold, and 49 genes declined by more than 50%. However, at day 15, there were 70 genes that were induced twofold, and 134 were repressed by more than 50%.



View larger version (41K):
[in this window]
[in a new window]
 
Fig. 2. Scattergrams of the primary microarray data used to generate expression profiles. Each data point represents Cy3 and Cy5 signal intensities for each of the 6,918 spots on the microarray. The y-axes indicate Cy5 intensities that correspond to the mRNAs purified from different days of culture. For each microarray, a balance coefficient was derived based on the overall average of median intensity values for Cy3 vs. Cy5. Thus Cy5 signal intensity values multiplied by the balance coefficient caused the median gene on the microarray to be unchanged in its expression between the sample time and time 0. The upper and lower lines indicate the positions where genes are twofold up or down relative to day 0. The vast majority of genes were unchanged in their expression compared with the time 0 reference control. The balance coefficients for the individual arrays were 1.52, 1.23, 0.84, 1.34, 190, and 1.67 for days 2, 4, 6, 9, 12, and 15, respectively.

 
The most induced annotated genes detected by microarray analysis included IGFBP-1, EBAF [a member of the transforming growth factor-ß (TGF-ß) gene family], somatostatin , forkhead box O1A, decorin, spermidine/spermine N1-acetyltransferase, EGF-containing fibulin-like extracellular matrix protein 1, IGFBP-4, TIMP3, solute carrier family 16 (monocarboxylic acid transporters), member 6, and IGFBP-3. The prolactin gene, which was also markedly induced, was not present on the microarray. The annotated genes whose mRNAs exhibited the largest repression include cysteine-rich angiogenic inducer 61; collagen, type I, {alpha}1; transgelin; regulator of G protein signaling 4; fibronectin 1; hexabrachion (tenascin C, cytotactin); and IGFBP-5. A complete list of the 100 most regulated genes, along with the accession numbers and fold change from day 0 level of expression is shown in Table 1.


View this table:
[in this window]
[in a new window]
 
Table 1. The 100 most strongly upregulated or downregulated genes during decidual fibroblast differentiation

 
The reliability of microarray quantitative data was independently corroborated by the use of Northern blot analysis of the mRNAs used in the microarray experiments as well as by replicate analyses using additional cell and mRNA preparations. As shown in Fig. 3, the expression of decorin, fibulin, EBAF and TIMP3 mRNAs based on Northern analysis followed the same induction profiles suggested by the microarray data. Similar results were obtained for IGFBP-1 (data not shown). All patterns shown by Northern blots agreed with the patterns indicated by microarray analyses.



View larger version (18K):
[in this window]
[in a new window]
 
Fig. 3. The kinetics of gene induction during decidualization as determined by microarray and Northern blot analyses. mRNA levels from representative genes were determined by microarray analysis and Northern blot analyses from the same mRNA samples. The decidual fibroblasts were exposed for 12 days to progesterone, estradiol, and cAMP as described in MATERIALS AND METHODS. Similar patterns for fibulin, decorin, endometrial bleeding associated factor (EBAF), and tissue inhibitor of metalloproteinase 3 (TIMP3) mRNA levels by Northern blot analyses were observed in two other experiments. Northern blot analysis values were in relation to the signal observed for GAPDH mRNA.

 
To categorize better the patterns of changes in gene expression associated with decidualization, we subjected the log-transformed Cy5/Cy3 expression ratios of the dataset to two forms of mathematical clustering using hierarchical tree and K-means algorithms. Application of this approach to the 281 strongly regulated genes allowed detection of coordinately regulated gene groups within the pool of dynamically regulated genes without a dilution effect generated by the inclusion of genes whose expression did not change significantly during differentiation. As expected, the hierarchical tree structure (Fig. 4) revealed a major division between induced and repressed genes, with the principal variations attributable to the length of delay prior to induction or repression. The results obtained using each sampled time point indicated that most of the 281 genes exhibited induction or repression over multiple time points.



View larger version (34K):
[in this window]
[in a new window]
 
Fig. 4. Cluster analysis of the 281 genes regulated by more than twofold during decidual fibroblast differentiation using the hierarchical-tree algorithm applied to the log of the ratio of each gene’s expression. The color code for the signal strength in the classification scheme is shown in the box at bottom right in which induced genes are indicated by shades of red; repressed genes are indicated by shades of green.

 
Initial K-means analyses revealed excessive splitting of some sets based on small variations between adjacent time points. We therefore used a time-averaged approach to smooth out technical vs. biologic variances. Figure 5 shows K-means clustering performed using data from the time-averaged values (MATERIALS AND METHODS). The 281 dynamically regulated genes were separated into 9 different K-means cluster groups composed of induced, biphasic, and repressed regulatory behaviors. The time-averaged approach thus provided an excellent smoothing function to the kinetic patterns and allowed for the formation of much more consistent gene clusters that shared similar kinetic induction and repression patterns. The individual gene traces followed simple kinetic patterns without marked fluctuations between adjacent time points, suggesting that a relatively low noise component. This was particularly true for genes that were expressed above the 50th percentile of signal intensity in both Cy3 and Cy5 channels (data not shown). Of the 281 regulated genes, 121 genes (patterns 1, 2, and 3) were induced, 50 genes (patterns 4, 8 and 9) exhibited biphasic behavior; and 110 genes (patterns 5, 6 and 7) were downregulated during decidualization (Table 2).



View larger version (41K):
[in this window]
[in a new window]
 
Fig. 5. K-means cluster analysis of the 281 genes dynamically regulated during decidualization of decidual fibroblasts as applied to the log of the expression ratio values for each gene’s expression. The patterns show early middle and delayed kinetics and a small degree of biphasic behavior.

 

View this table:
[in this window]
[in a new window]
 
Table 2. Categorical distribution of the 281 genes regulated by more than twofold during decidual fibroblast differentiation

 
Following literature review of the 281 dynamically regulated genes, we divided them into similar functional categories as used previously to describe genes implicated through DNA microarray studies of human cytotrophoblast cell differentiation (2). These include cell regulation (77 genes), cell and tissue function (74 genes), cell and tissue structure (79 genes), unknown function (6 genes), and expressed sequence tags (ESTs) (45 clones). Table 2 shows the number of dynamically regulated genes in each functional category. Table 3 shows the most dynamically regulated induced and repressed genes (4-fold changes) in the categories of cell regulation, cell and tissue function, and cell and tissue structure at different stages of differentiation.


View this table:
[in this window]
[in a new window]
 
Table 3. Categorical reprogramming scheme for decidual fibroblast differentiation

 
As shown in Table 2, single functional categories were frequently divided into groups of strongly induced and strongly repressed genes. This was particularly true within the categories of inflammatory response and polypeptide hormone in which 6 vs. 2 and 11 vs. 5 genes, respectively, were induced and repressed during decidualization. Within the categories of cytoskeletal organization and transport, 8 vs. 19 and 2 vs. 7 genes were induced and repressed.

Divergent behaviors not only occurred in the same functional class but also in members of the same gene family. As shown in Fig. 6, family members of the integrin, collagen, and laminin gene families varied widely in their regulation during decidualization. In each gene family, one or two members were strongly induced while several others were strongly downregulated. Taken together, these findings support the hypothesis that efficient execution of some biologic processes, such as the modification of adhesion or tissue structure, is best accomplished by both induction and repression of individual genes within the category. Other gene categories illustrated the same theme of coordinate induction and repression, including metabolism and transcriptional and translational regulation.



View larger version (29K):
[in this window]
[in a new window]
 
Fig. 6. Different individual genes within the integrin, collagen, laminin, and IGF/IGFBP gene families show reciprocal regulation during decidual fibroblast differentiation.

 
Divergent behavior was also observed for members of the IGF/IGFBP family (Fig. 6). Although IGFBP-1 was the most induced gene during decidualization, IGFBP-5 was the most repressed gene. Insulin-like growth factor II (IGF-II) mRNA levels increased gradually during decidualization beginning on day 2. The microarrays used for this study did not contain the cDNA for IGF-I.

In an earlier study (2), we investigated the expression profiles of genes dynamically regulated during the differentiation of human cytotrophoblast cells to syncytiotrophoblast cells using the same microarrays used in the present study. We observed 397 genes that were regulated using the same criteria as applied to this study. To better appreciate those aspects of the decidual fibroblast differentiation program that were specific vs. those that might be related to in vitro cell culture per se or are representative of general differentiation processes that were independent of the cell type of origin, we sought to compare the relative behavior of the genes identified in each model. To make these comparisons, we combined the two experiments into a single series and compared the relative behavior of the pooled 569 genes dynamically regulated in either experiment to hierarchical tree cluster analysis. Most of the strongly upregulated and downregulated genes were unique to a single model (Fig. 7), with only 81 genes meeting selection criteria for both. Using the K-means algorithm, we were able to divide these 81 commonly regulated genes into 7 clusters that consisted of genes that exhibited a variety of induction and repression kinetics across the two cell types, with 52 of these exhibiting similar regulation and 29 completely opposite (data can be found in the Supplementary Material published online at the Physiological Genomics web site). Most of the induced genes that were in common to both cell types were related to metabolism, whereas most of the repressed genes were related to cytoskeletal organization and cell adhesion.



View larger version (46K):
[in this window]
[in a new window]
 
Fig. 7. Cell type-specific gene induction and repression in decidual fibroblasts and trophoblasts. Hierarchical-tree cluster analysis of the 569 genes regulated during the differentiation of either cell type using Pearson correlation of log-transformed expression ratios relative to the time 0 gene expression for each cell type, respectively. The color bar at top left indicates the ratio values.

 

    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
In the present study, we have identified 281 genes that are dynamically regulated in decidual fibroblasts during in vitro decidualization induced by progesterone, estradiol, and cAMP. Some of the regulated genes were previously noted by ourselves and others to undergo regulation in endometrial stromal cells decidualized in vitro with progesterone and estradiol, relaxin, or cAMP (5, 7, 20, 41). Although most of the genes regulated during decidualization of the decidual fibroblasts were not previously known to be induced or repressed during decidualization, almost all were in functional categories likely to play critical roles in decidualization-associated processes, such as extracellular organization.

In an earlier report on genes changed during decidualization, Popovici and coworkers (34) performed DNA microarray analyses on human endometrial stromal cells after 10 days of exposure to progesterone or 2 days of exposure to cAMP. Of 588 genes screened, marked regulation was noted for cytokines, growth factors, nuclear transcription factors, members of the cyclin family, and mediators of the cAMP signal transduction pathway. Changes were also noted for the insulin receptor, some neurotransmitter receptors, neuromodulators, follicle-stimulating hormone receptor, inhibin/activin-ßA subunit, inhibin-{alpha}, and tumor necrosis growth factor-related apoptosis-inducing ligand (TRAIL). The total number of regulated genes was not indicated, and only two downregulated genes were identified (monocyte chemotactic protein-1 receptor and Ca2+ calmodulin-dependent kinase type IV catalytic subunit). Unfortunately, the genes represented on Clontech human Atlas cDNA expression array used by these investigators do not overlap extensively with those on the Incyte human GEM-V microarray used in the present study, thus precluding a useful cross-comparison.

A major difference between our study and earlier investigations, including that of Popovici et al. (34), is that we examined gene expression at multiple intervals during the decidualization process. We observed several different inductive and repressive kinetic patterns associated with decidualization, and we were able to correlate the functions of genes to their expression patterns during decidualization. The patterns consisted of variable delays prior to induction or repression as well as several patterns with biphasic kinetics.

Earlier studies of endometrial stromal cell differentiation, usually performed by immunochemical analyses of tissue samples, showed marked changes in proteins involved in extracellular organization, cytoskeletal organization, and cell adhesion. Cadherin-11 (10), laminin (21), integrin-{alpha}4 (fibronectin receptor), integrin-ß1, and integrin-{alpha}5ß3 (28), and desmin and vimentin intermediate elements (9) were shown to increase during decidualization. Furthermore, the ratio of type III to type I collagen decreased, and the ratio of type V to type I collagen significantly increased (23) during the decidualization process. Our microarray studies also indicate striking changes in extracellular matrix protein genes during decidualization of decidual fibroblasts. There was early induction of the extracellular matrix proteins decorin and fibromodulin and early repression of collagen type X {alpha}1, collagen type XI {alpha}1, collagen type I {alpha}1, collagen type XVI {alpha}1, collagen type IV {alpha}1, laminin-{alpha}4, integrin-{alpha}4, vimentin, glypican-1, and fibronectin-1. There was also induction of the proteolytic enzymes matrix metalloproteinases-2 and -11 and TIMP3 and repression of plasminogen activator inhibitor type I. Taken together, these findings indicate that striking reprogramming of genes that determine extracellular matrix and tissue strength and integrity occurs during decidualization and suggest that these changes are a prominent feature of decidualization. Interestingly, previous studies using in vitro models showed no changes in integrin expression by immunofluorescence and flow cytometry in cultured endometrial stromal cells decidualized by treatment with progesterone and estrogen (18). In contrast, the microarray data presented here demonstrated changes in integrin expression during decidual fibroblast cell differentiation in vitro. Of note was the induction of integrin-{alpha}4 (fibronectin receptor) expression, which is known to peak during mid-luteal phase and is downregulated in the endometrium of infertile women during implantation (35).

One of the earliest induced genes was the transcription factor forkhead O1A, which was originally isolated from a human alveolar rhabdomyosarcoma (38). The function of this member of the forkhead family member is unknown, but other members of the family have been implicated in the development of the lung and other tissues (19, 42). ID-3, a member of the inhibitors of DNA binding family, was also markedly induced. This binding protein acts as a negative helix-loop-helix protein that modifies the action of other HLX transcription factors (37). Changes in the expression of ID family members have been noted in several cell types during differentiation (3, 24), including placenta, but the role of the protein in decidualization is unknown.

Dramatic changes were observed in many members of the IGFBP family. IGFBP-1, which is widely recognized as a marker of decidualization in endometrial stromal cells (27), was the most induced gene in the decidual fibroblast cells undergoing differentiation, whereas IGFBP-5 was the most repressed gene. IGFBP-3 and IGFBP-4 were also induced. Decidual IGFBP-1 has been demonstrated to exert several different actions at the maternal-fetal interface and within the endometrium (15). IGFBP-1, which usually exists as a phosphorylated, high-affinity species, sequesters IGFs, thereby inhibiting their local action (15).

Our results support the hypothesis that there is a highly specific control network affecting IGF action at the maternal-fetal interface. Paracrine modulation of phosphorylated IGFBP-1, involving dephosphorylation and proteolysis by placental alkaline phosphatase and proteases that results in increased IGF bioavailability has been proposed (15). IGFBP-1 stimulates human extravillous trophoblast cell migration by signaling through integrin-{alpha}5ß1 and the mitogen-activated protein kinase pathway (16). This supports a role for decidual IGFBP-1 in trophoblast invasion as well as during the menstrual cycle since secretory endometrium has significantly more IGFBP-1 than proliferative endometrium (12). IGFBP-5, the major binding protein in bone tissue, inhibits IGF binding to its receptor (25). The binding of IGFBP-5 to extracellular matrix leads to a decrease in its affinity for IGF-I and an increase in IGF-I bioavailability (32). Taken together, these findings strongly suggest that the differential expression of the IGFBPs during decidualization markedly affect the biological actions of IGFs.

There was also early induction of genes involved in the mitogen-activated protein (MAP) kinase signaling pathway (MAP kinase kinase kinase 5 and SH3-domain binding protein 5) as well as the TGF-ß1 family member genes EBAF and inhibin-ßA. Induced later were several other genes involved in the phosphoinositide signal transduction pathway (phosphatidic acid phosphatase 2b and phospholipase D1). At present, the roles of these signaling pathways in human decidualization are unknown.

In addition to prolactin and IGFBP-1, other genes induced in endometrial stromal cells during decidualization include c-Src kinase (29), ubiquitin cross-reactive protein (4), superoxide dismutase (40), fatty acid synthase (33), 11ß-hydroxysteroid dehydrogenase (1), and endothelin B (26). In contrast, the mRNA expression of CD63 (a transmembrane 4 superfamily member), which has been reported to be downregulated during progesterone-induced decidualization of human endometrial stromal cells (31), was unchanged in our analysis of decidualization.

The mechanisms for gene activation or mRNA repression during decidualization remain to be identified. Mechanisms could include both transcriptional and posttranscriptional activation of gene expression systems already in place within the untreated decidual cell, or through the synthesis of new gene products that impact on gene expression. The different kinetic patterns may represent the occurrence of multiple regulatory mechanisms. Repression is particularly intriguing since specific machinery for selective mRNA decay or accelerated turnover has not been described in the decidual cell.

Comparison of genes activated and repressed in the differentiation of trophoblasts and decidual fibroblasts demonstrated only limited overlap in their identities, with only 81 out of the 569 in common. Of the overlapping genes, many were reciprocally regulated. However, the great majority of the 569 fell into gene function categories that were of vital importance to both cell types, with extensive representation of genes related to cell and tissue structure and cell-cell interaction. Of particular interest is the fact that in both models there was strong downregulation of key members of these categories simultaneous with the induction of other members of the same categories. This suggests that "categorical reprogramming" is fundamental for the differentiation of both cell types but that the two cells differ in the genes that are induced and degraded during differentiation and that the two different lineages are engaged in very distinct programs. We believe that cross-comparative analyses will be a powerful approach to identifying and distinguishing gene expression patterns that play general and specific roles in biologic processes.

In summary, we have shown that decidualization comprises a highly dynamic gene program that significantly affects the mRNA levels of 281 of 6,918 individual genes queried. The genes regulated during this process differ markedly from those observed in cytotrophoblast differentiation, demonstrating the specificity of the two programs. Several distinct kinetic patterns were observed involving both induction and strong repression. We hypothesize that decidualization, like trophoblast differentiation, requires both activation and repression of a substantial number of genes and that distinct classes of regulated genes are responsible for decidualization. The first is the induction of gene products responsible for cell functions that were not necessary prior to differentiation but that play a role for differentiated cell functions. Examples would include hormone production, unique metabolic processes, or mediators of differentiation per se. The other class represents genes induced that are to replace existing gene products with those that cause the cell to switch structure and function. To accomplish this, we envision that cells must eliminate mRNAs of gene products that compete or interfere with specific biochemical or cell biological functions of the induced genes. Thus our hypothesis is that the successful accomplishment of cellular differentiation requires both the induction of the effectors of the differentiated cell as well as dynamic reprogramming of genes within functional pathways that are critical for both precursor and product cell lineages. The fact that trophoblast cells and decidual fibroblast cells both manifest the phenomena of categorical reprogramming, but do so with largely nonoverlapping genes, suggests that this may represent a fundamental aspect of cellular differentiation.

It remains to be determined the extent to which these patterns and individual gene behaviors are polymorphic in human populations. Some variance may be of little consequence; however, regulatory variance or allelism within some of these decidualization-associated genes may be associated with decreased fertility, increased risk of endometriosis, or other pathological processes.


    ACKNOWLEDGMENTS
 
We thank Brian Richardson for many suggestions and Cathy Ebert for assistance with mRNA quality assurance.

This work was supported by National Institutes of Health Grant HD-15201. This work was also supported by an infrastructure grant to the University of Cincinnati College of Medicine from the Howard Hughes Medical Institute.

Portions of this work were presented at the 83rd Annual Meeting of the Endocrine Society, in Denver, CO, July 2001.


    FOOTNOTES
 
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).

Address for reprint requests and other correspondence: A. K. Brar, Dept. of Endocrinology, Children’s Hospital Medical Center, 3333 Burnet Ave., Cincinnati, OH 45229-2029 (E-mail: Anoop.Brar{at}).

1 Supplementary Material to this article is available online at http://physiolgenomics.physiology.org/cgi/content/full/7/2/ 135/DC1. Back


    REFERENCES
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

  1. Arcuri F, Monder C, Lockwood CJ, and Schatz F. Expression of 11ß-hydroxysteroid dehydrogenase during decidualization of human endometrial stromal cells. Endocrinology 137: 595–600, 1996.[Abstract]
  2. Aronow BJ, Richardson BD, and Handwerger S. Microarray analysis of trophoblast differentiation: gene expression reprogramming in key gene function categories. Physiol Genomics 6: 105–116, 2001.[Abstract/Free Full Text]
  3. Atherton GT, Travers H, Deed R, and Norton JD. Regulation of cell differentiation in C2C12 myoblasts by the Id3 helix-loop-helix protein. Cell Growth Differ 7: 1059–1066, 1996.[Abstract]
  4. Bebington C, Doherty FJ, and Fleming SD. Ubiquitin cross-reactive protein gene expression is increased in decidualized endometrial stromal cells at the initiation of pregnancy. Mol Hum Reprod 5: 966–972, 1999.[Abstract/Free Full Text]
  5. Bell SC, Jackson JA, Ashmore J, Zhu HH, and Tseng L. Regulation of insulin-like growth factor-binding protein-1 synthesis and secretion by progestin and relaxin in long term cultures of human endometrial stromal cells. J Clin Endocrinol Metab 72: 1014–1024, 1991.[Abstract]
  6. Brar A, Kanda Y, Kessler C, Cedars M, and Handwerger S. N5 endometrial stromal cell line: a model system to study decidual prolactin gene expression. In Vitro Cell Dev Biol 35: 150–154, 1999.[ISI]
  7. Brar AK, Frank GR, Kessler CA, Cedars MI, and Handwerger S. Progesterone-dependent decidualization of the human endometrium is mediated by cAMP. Endocrine 6: 301–307, 1997.[ISI][Medline]
  8. Brosens JJ, Takeda S, Acevedo CH, Lewis MP, Kirby PL, Symes EK, Krausz T, Purohit A, Gellersen B, and White JO. Human endometrial fibroblasts immortalized by simian virus 40 large T antigen differentiate in response to a decidualization stimulus. Endocrinology 137: 2225–2231, 1996.[Abstract]
  9. Can A, Tekelioglu M, and Baltaci A. Expression of desmin and vimentin intermediate filaments in human decidual cells during first trimester pregnancy. Placenta 16: 261–175, 1995.[ISI][Medline]
  10. Chen GT, Getsios S, and MacCalman CD. Cadherin-11 is a hormonally regulated cellular marker of decidualization in human endometrial stromal cells. Mol Reprod Dev 52: 158–165, 1999.[ISI][Medline]
  11. Chomczynski P and Sacchi N. Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction. Anal Biochem 162: 156–159, 1987.[ISI][Medline]
  12. Corleta H, Capp E, and Strowitzki T. Cycle modulation of insulin-like growth factor-binding protein 1 in human endometrium. Braz J Med Biol Res 33: 1387–1391, 2000.[ISI][Medline]
  13. Eisen MB, Spellman PT, Brown PO, and Botstein D. Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci USA 95: 14863–14868, 1998.[Abstract/Free Full Text]
  14. Ferenczy A and Bergeron C. Histology of the human endometrium: from birth to senescence. Ann NY Acad Sci 622: 6–27, 1991.[Abstract]
  15. Gibson JM, Aplin JD, White A, and Westwood M. Regulation of IGF bioavailability in pregnancy. Mol Hum Reprod 7: 79–87, 2001.[Abstract/Free Full Text]
  16. Gleeson LM, Chakraborty C, McKinnon T, and Lala PK. Insulin-like growth factor-binding protein 1 stimulates human trophoblast migration by signaling through {alpha}5ß1 integrin via mitogen-activated protein kinase pathway. J Clin Endocrinol Metab 86: 2484–2493, 2001.[Abstract/Free Full Text]
  17. Gordon AD. Classification. Boca Raton, FL: Chapman & Hall/CRC, 1999, p. 256.
  18. Grosskinsky CM, Yowell CW, Sun J, Parise LV, and Lessey BA. Modulation of integrin expression in endometrial stromal cells in vitro. J Clin Endocrinol Metab 81: 2047–2054, 1996.[Abstract]
  19. Hellqvist M, Mahlapuu M, Samuelsson L, Enerback S, and Carlsson P. Differential activation of lung-specific genes by two forkhead proteins, FREAC-1 and FREAC-2. J Biol Chem 271: 4482–4490, 1996.[Abstract/Free Full Text]
  20. Higuchi T, Kanzaki H, Nakayama H, Fujimoto M, Hatayama H, Kojima K, Iwai M, Mori T, and Fujita J. Induction of tissue inhibitor of metalloproteinase 3 gene expression during in vitro decidualization of human endometrial stromal cells. Endocrinology 136: 4973–4981, 1995.[Abstract]
  21. Holinka CF. Growth and hormonal responsiveness of human endometrial stromal cells in culture. Hum Cell 1: 207–217, 1988.[Medline]
  22. Irwin JC, Kirk D, King RJ, Quigley MM, and Gwatkin RB. Hormonal regulation of human endometrial stromal cells in culture: an in vitro model for decidualization. Fertil Steril 52: 761–768, 1989.[ISI][Medline]
  23. Iwahashi M, Muragaki Y, Ooshima A, Yamoto M, and Nakano R. Alterations in distribution and composition of the extracellular matrix during decidualization of the human endometrium. J Reprod Fertil 108: 147–155, 1996.[Abstract]
  24. Janatpour MJ, McMaster MT, Genbacev O, Zhou Y, Dong J, Cross JC, Israel MA, and Fisher SJ. Id-2 regulates critical aspects of human cytotrophoblast differentiation, invasion and migration. Development 127: 549–558, 2000.[Abstract/Free Full Text]
  25. Kalus W, Zweckstetter M, Renner C, Sanchez Y, Georgescu J, Grol M, Demuth D, Schumacher R, Dony C, Lang K, and Holak TA. Structure of the IGF-binding domain of the insulin-like growth factor- binding protein-5 (IGFBP-5): implications for IGF and IGF-I receptor interactions. EMBO J 17: 6558–6572, 1998.[Abstract/Free Full Text]
  26. Kohnen G, Campbell S, Irvine GA, Church HJ, MacLachlan F, Titterington M, and Cameron IT. Endothelin receptor expression in human decidua. Mol Hum Reprod 4: 185–193, 1998.[Abstract]
  27. Lane B, Oxberry W, Mazella J, and Tseng L. Decidualization of human endometrial stromal cells in vitro: effects of progestin and relaxin on the ultrastructure and production of decidual secretory proteins. Hum Reprod 9: 259–266, 1994.[Abstract]
  28. Lessey BA, Castelbaum AJ, Buck CA, Lei Y, Yowell CW, and Sun J. Further characterization of endometrial integrins during the menstrual cycle and in pregnancy. Fertil Steril 62: 497–506, 1994.[ISI][Medline]
  29. Maruyama T, Yoshimura Y, Yodoi J, and Sabe H. Activation of c-Src kinase is associated with in vitro decidualization of human endometrial stromal cells. Endocrinology 140: 2632–2636, 1999.[Abstract/Free Full Text]
  30. Noyes RW, Hertig AT, and Rock J. Dating the endometrial biopsy. Am J Obstet Gynecol 122: 262–263, 1975.[Medline]
  31. Okada H, Sanezumi M, Nakajima T, Okada S, Yasuda K, and Kanzaki H. Rapid down-regulation of CD63 transcription by progesterone in human endometrial stromal cells. Mol Hum Reprod 5: 554–558, 1999.[Abstract/Free Full Text]
  32. Parker A, Rees C, Clarke J, Busby WH Jr, and Clemmons DR. Binding of insulin-like growth factor (IGF)-binding protein-5 to smooth- muscle cell extracellular matrix is a major determinant of the cellular response to IGF-I. Mol Biol Cell 9: 2383–2392, 1998.[Abstract/Free Full Text]
  33. Pizer ES, Kurman RJ, Pasternack GR, and Kuhajda FP. Expression of fatty acid synthase is closely linked to proliferation and stromal decidualization in cycling endometrium. Int J Gynecol Pathol 16: 45–51, 1997.[ISI][Medline]
  34. Popovici RM, Kao LC, and Giudice LC. Discovery of new inducible genes in in vitro decidualized human endometrial stromal cells using microarray technology. Endocrinology 141: 3510–3513, 2000.[Abstract/Free Full Text]
  35. Reddy KV and Meherji PK. Integrin cell adhesion molecules in endometrium of fertile and infertile women throughout menstrual cycle. Indian J Exp Biol 37: 323–331, 1999.[Medline]
  36. Richards RG, Brar AK, Frank GR, Hartman SM, and Jikihara H. Fibroblast cells from term human decidua closely resemble endometrial stromal cells: induction of prolactin and insulin-like growth factor binding protein-1 expression. Biol Reprod 52: 609–615, 1995.[Abstract]
  37. Riechmann V and Sablitzky F. Mutually exclusive expression of two dominant-negative helix-loop-helix (dnHLH) genes, Id4 and Id3, in the developing brain of the mouse suggests distinct regulatory roles of these dnHLH proteins during cellular proliferation and differentiation of the nervous system. Cell Growth Differ 6: 837–843, 1995.[Abstract]
  38. Shapiro DN, Sublett JE, Li B, Downing JR, and Naeve CW. Fusion of PAX3 to a member of the forkhead family of transcription factors in human alveolar rhabdomyosarcoma. Cancer Res 53: 5108–5112, 1993.[Abstract]
  39. Stephanou A and Handwerger S. Retinoic acid and thyroid hormone regulate placental lactogen expression in human trophoblast cells. Endocrinology 136: 933–938, 1995.[Abstract]
  40. Sugino N, Kashida S, Takiguchi S, Nakamura Y, and Kato H. Induction of superoxide dismutase by decidualization in human endometrial stromal cells. Mol Hum Reprod 6: 178–184, 2000.[Abstract/Free Full Text]
  41. Tang B, Guller S, and Gurpide E. Cyclic adenosine 3',5'-monophosphate induces prolactin expression in stromal cells isolated from human proliferative endometrium. Endocrinology 133: 2197–2203, 1993.[Abstract]
  42. Warburton D, Schwarz M, Tefft D, Flores-Delgado G, Anderson KD, and Cardoso WV. The molecular basis of lung morphogenesis. Mech Dev 92: 55–81, 2000.[ISI][Medline]