| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Animal Science,3
Department of Health and Medical Informatics,4
Department of Computer Science,5
Molecular Biology Program,6 University of Missouri-Columbia, Columbia, Missouri 65211
Monsanto Company,7 St. Louis, Missouri 63163
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
early development, embryo, gene expression, gene regulation, in vitro fertilization, pig
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Application of many embryo-related biotechnologies requires that the embryos be cultured or produced in vitro. These technologies include embryo transfer, fertilization with sperm that are sorted based on the presence of an X or Y chromosome, cloning by nuclear transfer, embryo bisection, transgenic animal production, etc. [4, 5]. Embryo production in vitro includes oocyte maturation, fertilization, and culture as far as the blastocyst stage of development. Unfortunately, in vitro-produced embryos are generally considered to be less developmentally competent than their in vivo-produced counterparts [6, 7].
Identification of key genes involved in specific developmental processes requires an understanding of the global patterns of gene expression in that tissue at that specific time. The best way to identify those genes is to sequence cDNAs from the tissue of interest. To that end, we have embarked on an expressed sequent tag (EST) project to identify previously characterized, as well as novel, genes in the reproductive tissues and early embryos of pigs [8– 10] (http://genome.rnet.missouri.edu/Swine/). Without construction of these libraries from the tissues of interest, many of the tissue-specific novel genes would not have been identified. For example, Whitworth et al. [10] identified 1114 of the 2489 clusters to be unique submissions to GenBank. Indeed, even in the mouse, where extensive sequencing of cDNA libraries has resulted in large numbers of clones, a focused effort on the two-cell stage has identified over 4000 different cDNAs, with over half of them having not previously been studied [11].
Construction of microarrays based on EST sequences and subsequent transcriptional profiling experiments have revealed the involvement of newly discovered genes as well as provided a better understanding of the genetic pathways that are involved with embryo development in sea urchins [12, 13] and mice [14–16]. Here we evaluate global changes in message abundance in the germinal vesicle- (pgvo), four-cell- (p4civv) and blastocyst (pblivv)-stage embryos in pigs and compare these with in vitro-produced embryos at the (p4civp) and blastocyst (pblivp) stages. Improvements in in vitro conditions might be made based on the changes in transcript level that were identified between pgvo, p4civv, and pblivv.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Germinal vesicle oocytes and in vivo- and in vitro-derived four-cell-stage and blastocyst-stage embryos were collected as described previously [10]. Animals were treated according to institutional animal care and use guidelines. Cumulus cells were removed from oocytes and the zona pellucidae were removed from the embryos. Oocytes and embryos were then washed in diethyl pyrocarbonate-treated PBS with 0.1% polyvinyl alcohol and snap frozen in 1 µl of Superase RNase inhibitor (Ambion, Austin, TX). Oocytes and embryos were stored at –80°C until poly(A) RNA extraction.
Poly(A) RNA Extraction
Poly(A) RNA from pools of 100 oocytes or 34–50 embryos was extracted by using a micropoly(A) purist kit (Ambion) according to the manufacturer's instructions with minor modifications to the protocol. The modified elution buffer consisted of 10 mM Tris (pH 7.5), 1 mM EDTA, and 0.5% SDS heated to 65°C. The second round of poly(A) purification was also removed from the protocol. Due to the low amount of mRNA, the mRNA concentration, integrity, and purity were not verified.
Amplification and Aminoallyl Labeling of Poly(A) RNA
Fifty percent of the poly(A) RNA from the extraction was amplified by using the Ovation nanosample RNA amplification system (NuGEN Technologies, Inc., San Carlos, CA) following the manufacturer's instructions. The Ovation kit utilizes the Ribo-SPIA process to linearly amplify limited amounts mRNA in a three-step process resulting in microgram quantities of double-stranded aminoallyl (aa)-labeled cDNA (acDNA). In a preliminary study, Nugen Ribo-SPIA amplification kit was tested by diluting total RNA from pregnant pig endometrium (Day 12 of estrus) to 4.5 ng/µl. Diluted RNA (4.5 ng total) was amplified by using the Ribo-SPIA kit, labeled with cy5 and hybridized with reference labeled with cy3. Undiluted RNA from the same sample (15 µg) was processed by standard protocol and also hybridized to 20-K microarray. This experiment was repeated twice as described for the embryo samples below, resulting in four slides. Amplified and unamplified results were loaded into Genespring, and ANOVA (P < 0.05) with Benjamini and Hochberg multiple testing correction was performed. This tested resulted in zero significantly different genes between amplified and unamplified microarrays. Other groups have shown similar correlations (R2 values ranging from 0.82 and 0.97) between Ribo-SPIA-amplified RNA and unamplified RNA using microarrays [17]. Three pools of embryos (p4civp, p4civv, pblivp, pblivv) and oocytes (pgvo) were each analyzed twice. Such a design resulted in two technical replications of the three biological replications, for a total of six measurements for each stage.
Reference RNA
A reference RNA sample was created in house by isolating total RNA from a large representation of nonreproductive and reproductive tissues across different stages of development by using RNA Stat 60 protocol (Tel-Test, Friendswood, TX). Nonreproductive tissues were collected from a 160-day-old gilt and include heart, kidney, liver, pituitary, hypothalamus, skeletal muscle, small intestine, and spleen and consisted of approximately 50% of the reference RNA population. The other 50% of the reference RNA population was isolated from reproductive tissues, including corpus luteum, follicle, oviduct, endometrium, placenta, and fetus. RNA quality was assessed for each tissue by agarose gel electrophoresis and purity was measured by 260:280 ratios by spectrophotometry. RNA was then pooled from each tissue to create the reference RNA. The reference RNA (15 µg) was indirectly labeled with amino allyl dUTP by reverse transcription with Superscript II (Invitrogen, Carlsbad, CA) and 1 µl oligo dT primer (50 µM) as described by Hegde et al. [18]. The protocol was modified to include an equal ratio of 3-amino allyl dUTP and dTTP and a lower concentration of oligo dT primer.
Labeling and Hybridization
The acDNA from each stage (pgvo, p4civp, p4civv, pblivp, and pblivv) and amino allyl-labeled cDNA (aacDNA) from the reference sample were purified by using a PCR purification kit (Qiagen, Valencia, CA) and dried by a Centravap vacuum centrifugation system (Labconco, Kansas City, MO). Samples were resuspended in 0.1 M sodium carbonate buffer, pH 9.0. Oocyte and embryo acDNA (2 µg) and aacDNA from reference RNA were labeled with cy5 and cy3 monoreactive dyes (Amersham Biosciences, Piscataway, NJ), respectively, for 1 h at room temperature. Labeling was quenched by the addition of 100 mM sodium acetate. Samples were then purified (Qiagen), dried by Centravap (Labconco), and resuspended in 26 µl hybridization buffer (48% formamide, 4.8x saline-sodium citrate [SSC], 0.1% SDS, and 20 µg poly A in water). Samples were then denatured at 95°C for 3 min and allowed to reach room temperature before they were applied to the microarrays under 22 x 40 mm Lifter Slips (Erie Scientific, Portsmouth, NH). Spectophotometry measurements were performed after each purification so that equal amounts of acDNA and reference aacDNA could be labeled and labeling efficiency could be calculated.
Microarrays were hybridized for 16 h at 42°C and washed at room temperature in a graded series of SSC solutions for 4 min each (wash 1 = 2x SSC/0.1% SDS, wash 2 = 0.1x SSC/0.1% SDS, wash 3 = 0.1x SSC). Microarrays were then rinsed in water and dried by centrifugation at 1500 rpm for 5 min.
cDNA Microarray Preparation, Printing, and Postprocessing
A pig reproductive tissue-specific 19 968 spot cDNA microarray was created from University of Missouri-Columbia EST projects. A general description of the libraries and the ESTs that were selected for spotting on the array can be found at the following website: http://genome.rnet.missouri.edu/Swine/. The unigene set consisted of sequences from 27 cDNA libraries from tissues including follicle, corpus luteum, embryo, oocyte, oviduct, endometrium, conceptus, and fetus across different stages of the estrous cycle [8–10, 19]. The total census on the array includes 4108 different ESTs that have no match in GenBank, plus an additional 10 862 different ESTs that have useful annotation. Thus, we have a total of 14 970 potentially different genes represented. We refer to this array as our 20-K array because there are about 20 000 spots on the array (a number of control spots are included, and several clones are represented two or more times). This unigene set (Pig Reproductive Tissue Unigene 1: PRU-1) was PCR amplified from bacterial glycerol stocks in a 50-µl reaction containing 0.25 mM dNTP and 200 ng M13 forward and reverse primers. The follicle/corpus luteum/embryo/oocyte clones were amplified by using Klentaq (2.5 units; Ab Peptides, St. Louis, MO). The remaining clones were amplified by using Biolase DNA polymerase (0.85 units; Bioline, Randolph, MA). PCR conditions included a 95°C initial denaturation for 3 min followed by 30 cycles of 95°C denaturation (30 sec), 55°C annealing (30 sec), and 72°C extension (3 min). PCR success and approximate concentration for each clone was verified by agarose gel electrophoresis.
PCR reactions were purified by using MultiScreen-PCR Plates (Millipore, Billerica, MA) according to manufacturer's instructions. The purified PCR reactions were dried in a Centravap vacuum centrifugation system and resuspended in 10 µl 3x SSC. The 96-well PCR plates were reracked into 384-well plates (MJ Research, Inc., Waltham, MA) by the University of Missouri-Columbia DNA Core Facility, and the 384-well plates were stored at –80°C until printing onto slides.
Gold Seal glass microscope slides (Fisher Scientific, Hampton, NH) were coated with 0.02% poly-L-lysine (Sigma, St. Louis, MO) in 0.05x PBS. Coated slides were incubated at room temperature under desiccation for 2–3 wk before printing [20]. Slides were examined for imperfections and only high-quality slides were used in the printing process. The cDNAs were printed onto coated slides by using a pick and place robot in an environment with 35–45% humidity at room temperature [20]. Slides were rehydrated over a 37°C humidified chamber and snap dried on a 140°C heat block. Slides were then crosslinked at 120 mJ/cm2 for 20 sec (Spectrolinker; Spectronics Corp., Westbury, NY) and postprocessed with 0.018% succinic anhydride and 0.043 M sodium borate in 1-methyl-2-pyrrolidinone [20]. All chemicals were purchased from Sigma-Aldrich (St. Louis, MO). Microarrays were stored at room temperature under desiccation until hybridization.
Unigene Set Annotation
Sequences passing initial quality assessment in our sequence analysis pipeline were clustered based on a 0.95 sequence homology criteria to produce a project unigene set. The dominant member of each cluster was chosen to be the longest sequence in each cluster. The annotations tied to each cluster are based on a multilevel strategy that involves multiple searches of several reference databases. The databases were searched in the order Human Reference Sequences (HumRefSeq), Vertebrate Reference Sequences (VertRefSeq), Unigene, and GenBank EST (GBEST), with subsearches in the Human Unigene RefSeq (UGHumRefSeq), Vertebrate Unigene RefSeq (UGVertRefSeq), and Non-Replicated (NR) databases. The resulting BLAST score from each database search, in turn, was compared with a threshold score. If the score was greater than the threshold, the corresponding annotation from the best hit was used as the annotation. A threshold value (score) of 100 was used when searching HumReFSeq and VertRefSeq. When searching the Unigene sequences, a threshold value of 100 (200 for Sus Scrofa hits) was used. If the search of the GBEST database resulted in a BLAST score greater than 100, then the top scoring sequence found in GBEST was searched against the UGHumRefSeq, UGVertRefSeq, and NR databases in turn. A resulting score greater than 200 in these searches selected the annotation from the corresponding hit. If none of these criteria were met, the sequence was declared to be unique. The resulting annotations were placed into a table listing the sequence name (dominant member), database, BLAST score, E-value, and annotation tied to each member of the project unigene set. This table was then used as a guide for further investigation. A simple diagram of the search order and criteria is shown in Figure 1.
|
Scanning of Microarrays and Acquisition of Results
After hybridization, microarrays were scanned on a Genepix 4000B scanner (Axon Instruments, Union City, CA). This scanner scans at two wavelengths (532 nm for cy3 and 635 nm for cy5) simultaneously by using a dual laser. Spot quality was assessed by using Genepix Pro 4.0 (Axon Instruments). Bad spots or poor-quality spots (smeared or saturated) were removed manually and raw results files (*.gpr) were then loaded into Genespring 6.2.1 (Silicon Genetics, Redwood, CA).
Microarray Analysis by Genespring 6.2.1
Three comparisons were made by using Genespring 6.2.1. The first comparison analyzed changes in mRNA differences throughout early development. Three experiments were set up in Genespring (pgvo versus [-] p4civv, p4civv-pblivv, and pgvo-pblivv). Two additional experiments were set up in Genespring: p4civp-p4civv and pblivp-pblivv. After the results files for each comparison were loaded into Genespring, a per-spot and per-chip Lowess normalization was performed on each experiment. ANOVA was performed on each comparison by using a parametric test with variances not assumed equal (Welch t-test) and a P-value cut-off of P < 0.05. Additionally, the Benjamini and Hochberg multiple correction test with a 5% false discovery rate was performed on the pgvo-p4civv, p4civv-pblivv, and pgvo-pblivv experiments.
Microarray Analysis by Expression Analysis Systemic Explorer
Expression Analysis Systemic Explorer (EASE) version 2.0 is a software program that helps interpret gene lists from microarray results by identifying the most prominent gene ontology groups represented in the array data [21] (http://david.niaid.nih.gov/david/ease.htm). GenBank accession numbers from all significantly upregulated cDNAs from each experiment were loaded into the EASE software program and analyzed to find overrepresented categories. EASE then generated an annotation table for each gene list and provided an EASE score to identify the most significant biological themes.
k-Means Clustering
Genespring provides a k-means clustering function that divides genes into similar groups based on their expression patterns. K-means clustering (standard correlation) was performed on an experiment that combined pgvo-p4civv-pblivv to identify clusters of gene expression patterns across the first week of development.
Condition Trees
Similarities in gene expression between each microarray in the experiment can be assessed by using the condition tree function in Genespring. Condition trees (standard correlation) were created for the pgvo-p4civv-pblivv, p4civv-p4civp, and pblivv-pblivp experiments.
Validation of Microarray Results by Real-Time PCR
Microarray results were validated by using real-time PCR (RT-PCR). Seven candidate cDNAs were selected for pgvo-p4civv, p4civv-pblivv, and pgvo-pblivv experiments. Four candidate cDNAs were selected for the p4civp-p4civv and five candidate cDNAs were selected for the pblivp-pblivv. Clone names, annotation, and GenBank accession numbers as well as the GenBank accession number for the annotation are listed in Table 2. Primers for each candidate were designed by Primer Express (Applied Biosystems, Foster City, CA) by using the company default settings. The first optimal primer pair was chosen for each candidate cDNA. Primers were ordered from Integrated DNA technologies (Skokie, IL). Amplified cDNA from each sample was pooled from each biological replicate and diluted to 5 ng/µl. Serial dilutions were then performed to final concentrations of 0.5 ng/µl, 0.05 ng/µl, and 0.005 ng/µl for each pooled sample. RT-PCR was performed on the 0.05-ng/µl dilution by using the QuantiTect SYBR Green PCR Kit by following standard protocol. Amplifications were performed in triplicate on each plate and each plate was replicated three times, resulting in nine threshold cycle (CT) measurements/gene. The cDNAs were amplified by using the ABI Prism 7500 (Applied Biosystems) and relative quantification was analyzed by using the 7500 sequence detection system software (Applied Biosystems). A housekeeping gene (p4mm3-015-H07, YWHAG) was chosen. Preliminary experiments demonstrated that YWHAG expression was identical in all samples when assayed by both microarrays and RT-PCR. Threshold cycle (CT) for the target gene was subtracted from the CT for YWHAG to obtain the change (
) in CT (CT). The relative amount of each mRNA was calculated by assuming an amplification efficiency of two and using the equation 
. Differences were determined by importing 
values into VasserStats (http://faculty.vassar.edu/lowry/anova1u.html) and performing ANOVA.
|
Quality Control of Spot Identity
To confirm that lanes were called correctly, clones from the front (-a01 to -a06) and end (-h01 to -h12) of the 288 unigene plate stocks were grown and sequenced. These 576 samples were compared back to the database to confirm a match to the original data. There were 57 sequences that failed to pass the quality test so were not carried forward to the BLAST program. These failures likely resulted from loss of sample during storage. Of the remaining 519 sequences compared with the database, there were 17 in the failed category. Eleven of these did, in fact, hit the proper sequence in the database but gave a warning due to low scores. Thus, only 6 of the 519 failed to match their properly annotated clone (an error rate of 1.16%).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
After agarose gel electorphoresis, RNA samples with clear 18S and 28S bands with no visual evidence of degradation were pooled to create the reference sample. The 260:280 ratio for the pooled reference sample was 2.0. The reference sample produced a raw intensity of >50 (wavelength 532) on 18 358 of the 19 128 good spots on the microarray. This resulted in a reference signal on 96.0% of the microarray; therefore, all spots were included in the analysis.
Aminoallyl cDNA Purity and Labeling Efficiency for Unamplified Reference and Amplified Samples
Spectrophotometry results showed a 260:280 ratio for aacDNA (n = 30) from reference and acDNA (n = 30) 1.9 and 1.8, respectively. Labeling efficiencies were not significantly different between amplified and unamplified cDNAs (P = 0.447). Again, P-values were calculated on all stages as a group (Table 1).
|
Microarray Expression Results from pgvo-p4civv, p4civv-pblivv, and pgvo-pblivv Comparisons
ANOVA (P < 0.05) with the Benjamini and Hochberg false discovery rate multiple testing correction was performed in Genespring on 19 128 spots. Spots with insufficient data for comparison (n = 480) and control spots (3x SSC, vector, poly A, housekeeping genes, and plant cDNA) (n = 360) were removed from analysis by Genespring filtering procedures. In the pgvo-p4civv, p4civv-pblivv, and pgvo-pblivv comparisons, 3214, 1989, and 4528 cDNAs were found to be significantly differentially expressed, respectively. A list of all genes and their expression levels and P-values can be found on our website (http://genome.rnet.missouri.edu/Swine/Publications/). Condition tree analysis showed that the pgvo, p4civv, and pblivv clustered separately but that pgvo and the p4civv are more similar to each other than to the pblivv (Fig. 2). The k-means clustering (Fig. 3) shows that there are a large number of messages that change in level at each stage of development and that the differences detected are not simply due to changes at one particular stage or another.
|
|
Microarray Expression Results from p4civp-p4civv and pblivp-pblivv Comparisons
The Benjamini and Hochberg false discovery rate multiple testing correction was also performed after ANOVA (P < 0.05) in Genespring on the p4civp-p4civv and pblivp-pblivv comparisons and resulted in 0 and 10 differentially expressed cDNAs, respectively. The 10 gene products that were more abundant in the in vivo-produced blastocysts as compared with the in vitro-produced blastocysts include ATPase, Ca++ transporting, plasma membrane 1 (pd3end3-018-E01), CDC28 protein kinase regulatory subunit 2 (pfeto1-021-b09), SET translocation (myeloid leukemia-associated) (pfeto1-002-b11), RAS-GTPase-activating protein SH3-domain-ginding protein (MI-P-E4-abp-c-C06), high-mobility group box 1 (p2mm1-001-g09), and 5 gene products without annotation (p2mm2-003-c07, pd10en3-005-F06, p2mm2-003-e05, pd3end3-012-B11, peov1-007-G06).
Overall, this suggests that in vivo- and in vitro-produced embryos in our culture system have quite similar expression patterns relative to comparisons between the pgvo, p4civv, and pblivv. The multiple testing correction function was then removed from our analysis. The ANOVA from above without the Benjamini and Hochberg test resulted in 1409 and 1696 differentially expressed cDNAs in the p4civp-p4civv and pblivp-pblivv comparisons, respectively. A condition tree clustering of the relative expression of cDNAs of the in vitro- and in vivo-produced four-cell stage embryos is shown in Figure 4. In contrast, condition tree clustering of the blastocyst-stage embryos resulted in a mixing of some of the in vitro- and in vivo-produced embryos (Fig. 4). Thus, those pools of in vitro-produced embryos that make it to the blastocyst stage appear to be very similar to in vivo-produced embryos. A list of all cDNAs and their expression levels and P-values can be found on our website (http://genome.rnet.missouri.edu/Swine/Publications/).
|
RT-PCR Verification of Microarray Results
Seven genes (ZP3, ZP4, DNMT1, PTMA, DLD, RPL23a, and ACTB) were selected to verify the pgvo-p4civv, p4civv-pblivv, and pgvo-pblivv comparisons by RT-PCR. The expression of all seven genes followed the same pattern whether evaluated by microarray or RT-PCR. One gene, DNMT1, was a slight exception to this rule, as a significant difference could not be detected (P = 0.561) by microarray between pgvo and p4civv, while RT-PCR did detect a difference (P = 0.004). All other genes showed the same (P < 0.05) relative expression pattern by both microarray and RT-PCR analysis in the above three comparisons (Table 3), resulting in an overall success rate of 95.2% (1 minor inconsistency out of 21 comparisons: three stages by seven messages).
|
Four genes (DSTN, PAIP1, UBE4B, and NASP) were selected to verify the p4civp-p4civv comparison by RT-PCR. All four genes showed the same relative expression pattern by both microarray and RT-PCR analysis (Table 4).
|
Five genes (HMGB1, ATP5A1, LSM2, DNAJB6, and ALCAM) were selected to verify pblivp-pblivv comparison by RT-PCR. HMGB1, LSM2, DNAJB6, and ALCAM showed the same relative expression pattern by both microarray and RT-PCR analysis (Table 5). APT5J showed an opposite relative expression pattern by microarray RT-PCR analysis. Overall success rate of RT-PCR and microarray verification of the in vitro (ivp) and in vivo (ivv) comparisons was 88.9% (one major inconsistency out of nine comparisons).
|
Biological Pathways
The simplified gene ontology in Genespring was used to categorize significantly differentially expressed genes into three functions, biological processes, cellular components, and molecular function (Table 6). It should be noted that not all ESTs on the microarray have been annotated; therefore, the function cannot be assigned for all significantly different genes. Similarly, the comparison was made between in vitro- and in vivo-produced embryos, but because only ANOVA was performed on this date before entering into EASE, this table is included only on our web site.
|
By using the EASE scores and program, as was done previously in the mouse [16], we were able to quickly identify categories that either changed significantly during development or were different between in vitro- and in vivo-produced embryos. The first comparison was between pgvo and p4civv. Gene categories that were higher in the pgvo include cell adhesion receptor activity, mitotic cell cycle, primary active transporter activity, and M-phase specific microtubule process (see Table 7 for the top eight most significant different gene ontology (GO) scores; the complete list is posted on the web site (http://genome.rnet.missouri.edu/Swine/Publications/). In contrast, the gene categories that were higher for the p4civv include microtubule, nucleus, binding, development, insulin-like growth factor binding. When comparing the p4civv with the pblivv, those higher in the p4civv include transferase activity, transferring phosphorus-containing groups, phosphotransferase activity, alcohol group as acceptor, kinase activity, transferase activity. and protein kinase activity. Those pathways that were higher in the pblivv include cytosolic ribosome, hydrogen ion transporter activity, monovalent inorganic cation transporter activity, structural constituent of ribosome, and cation transporter activity.
|
A similar comparison between in vitro- and in vivo-produced embryos also yielded many differences. In vitro-produced four-cell-stage embryos have higher message levels of gene expression for the following categories: oncogenesis, cell adhesion receptor activity, membrane, and hydrolase activity, acting on ester bonds (see web site for the complete list [http://genome.rnet.missouri.edu/Swine/Publications/]). In contrast, the gene categories that were higher in the p4civv include reproduction, calcium ion binding, small ribosomal subunit, pathogenesis, and sexual reproduction. There were even more significant differences between the in vitro- and in vivo-produced blastocyst-stage embryos. Those gene categories that were more highly represented in the pblivp include plasma membrane, protein-tyrosine-phosphatase activity, protein binding, cytoskeleton, and nonmembrane spanning protein tyrosine phosphatase activity (web site for the complete list [http://genome.rnet.missouri.edu/Swine/Publications/]). In contrast, those gene categories that were higher in the pblivv include nucleolus, small nucleolar ribonucleoprotein complex, RNA binding, RNA processing, RNA metabolism, ribonucleoprotein complex, and mRNA binding.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The second theme of this article was to evaluate the pattern of gene expression in in vitro- versus in vivo-produced embryos. Only a handful of genes had significantly different levels of gene expression, as determined by the Benjamini and Hochberg multiple correction test, thus illustrating the degree of similarity between the in vitro- and in vivo-produced embryos. The one example of a difference between the microarray and RT-PCR was for ATP5A1. This difference between the ANOVA and the RT-PCR may have been a result of a misidentified clone or cross hybridization between gene family members. The condition trees showed a clear distinction between the in vitro- and in vivo-produced four-cell stage (Fig. 4). In contrast, there was a mixing of the in vitro- and in vivo-produced blastocyst-stage embryos (Fig. 4). Thus, there may have been a selective removal of the poor-quality embryos by the time they reached the blastocyst stage. These results also suggest that our in vitro-produced blastocyst-stage embryos are of a quality that is almost transcriptionally indistinguishable from in vivo-produced blastocyst-stage embryos, at least in regard to global transcription patterns. Differences in transcriptional profile due to their being in vitro- or in vivo-produced may be a result of aberrant expression of a limited number of genes. It should be noted that the comparison was between equal numbers of morphologically similar embryos, not between equal cell numbers, as the in vitro-produced blastocysts likely had few cells at the same morphological stage as compared with the in vivo-produced blastocyst-stage embryos. A more dramatic difference in gene expression between in vitro- and in vivo-cultured mouse embryos was reported by Rinaudo and Schultz [25].
To confirm that the differences that were observed on the arrays were truly indicative of changes in mRNA abundance, a number of transcripts were selected for quantitation via RT-PCR. First, a standard was selected based on its uniform abundance. This standard and the individual genes selected for RT-PCR are discussed below.
Tyrosine 3-Monooxygenase/Tryptophan 5-Monooxygenase Activation Protein, Gamma Polypeptide
The tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, gamma polypeptide (YWHAG) protein plays a role in signal transduction (by interaction with RAF1 [26]), mitosis, and cellular proliferation, and its expression is induced by growth factors and is phosphorylated by protein kinase C [27]. We found it (p4mm3-015-H07) to be uniformly high throughout development in the pig. In our oocyte/embryo EST project, this clone was found only once in the germinal vesicle library [10]. In the mouse, YWHAG message was highest during the two-cell stage [15, 24]. Because the message level of YWHAG was similar across different stages of embryo development, we used it as a control to standardize our other RT-PCR message levels.
Germinal Vesicle, Four-Cell Stage, and Blastocyst-Stage Comparisons
A number of genes were selected for comparing between pgvo, p4civv, and pblivv. They included ZP3 and ZP4, which are the major constituents of the muco-protein layer that surrounds the early embryo (ZP4 is thought to have a function similar to zona pellucida 1, i.e., that of providing structural support to the zona pellucida). In the mouse, this family of proteins is specific to the developing oocyte in the ovary and is not detected in any other tissue [28]. Consistent with other reports, we have found very high levels of both ZP3 (pgvo4-005-B01) and 4 (pnatal4-012-D12) in the germinal vesicle oocyte, with a continuing decrease observed at the four-cell stage and the blastocyst stage [10, 14, 15, 24].
The next gene selected for analysis was DNA (cytosine-5-)-methyltransferase, as it is partially responsible for regulation of transcription and genomic imprinting [29, 30]. In the mouse, there appears to be at least two DNMT1's, a somatic form (DNMT1s) and an oocyte-specific form (DNMT1o) [31]. The overall level of DNA methylation drops from the zygote to the blastocyst stage even though there is a high level of DNMT1, because the protein for DNMT1o only transiently enters the nucleus during the eight-cell stage [31, 32]. Based on the mouse observations, the DNMT1 detected in pig (pgvo3-001-C11) embryogenesis may, in fact, be DNMT1o [33]. Similar to other studies [10, 15, 24], we found the level of DNMT1 by the microarrays to be very high in the oocyte and four-cell stage and then decreases dramatically by the blastocyst stage. However, RT-PCR showed a continued high level during the four-cell stage. It should be noted that the microarray data yielded a 10-fold difference between some of the data points at the four-cell stage.
Prothymosin alpha is a proliferation-related nuclear protein [34] that may inhibit apoptosis [35], as it is involved with chromatin remodeling [36, 37]. Prothymosin alpha also inhibits cytochrome c, thus abolishing its antioxidant functions [38]; an important observation considering that oxidative stress is often observed in embryo culture. In the present study, PTMA message level was found to be higher at the four-cell stage as compared with either the oocyte or blastocyst stage. Similarly, Whitworth et al. [10] gave it a (p4civv1-015-B05) human clone annotation and found it only in the four-cell stage library. In the mouse, PTMA increased after the two-cell stage [15, 24].
Dihydrolipoamide dehydrogenase precursor is the E3 component of pyruvate dehydrogenase complex, the alpha-ketoglutarate dehydrogenase complex, and the branched-chain alpha-keto acid dehydrongenase complex [39]. Mutations in humans cause maple syrup Urine disease type 3 [40]. Dihydrolipoamide dehydrogenase precursor is also an ubiquinone reductase and thus is involved with protecting against lipid and protein peroxidation [41, 42], an important observation because embryos created in vitro appear to be oxidatively stressed [43]. This clone (pd6end2-005-D03) was not detected by Whitworth et al. [10] but found here to increase from the oocyte to four-cell to blastocyst stage. In direct contrast, in the mouse, relatively high messages were found in the oocyte, zygote, and two-cell stage but lower levels by the eight-cell stage [15, 24].
Ribosomal protein L23a is one of about 80 different proteins that comprise the mammalian ribosome and influences protein secretion and folding [44]. The message for RPL23a (pd12-14end-005-F10) was found to increase at the blastocyst stage and was found only in the blastocyst-stage libraries and not in the four-cell-stage or germinal vesicle-stage libraries created by Whitworth et al. [10]. In the mouse, RPL23 increased at the four-cell stage and beyond as compared with earlier stages [15, 24].
Actin, a cytoskeletal protein, is often chosen as a housekeeping gene, as it is involved with basic cell shape and assumed to be present at similar levels in all cells. However, recent studies have questioned this assumption [45, 46]. Indeed, here we have shown that, on a per embryo basis, ACTB message (p8mm1-002-D03) levels increase significantly (over 20-fold; P < 0.003) from the four-cell stage to the blastocyst stage at a time when total cell volume is decreasing. Our previous article, which simply evaluated the relative proportion of ESTs, found two clusters for ACTB. Neither cluster had any members from the germinal vesicle- or four-cell-stage libraries, but both had members from the blastocyst-stage libraries [10]. Similarly in the mouse, Wang et al. [24] showed an 8.4-fold increase in ACTB message levels from the germinal vesicle stage to the blastocyst. All articles examining embryo development that use ß-actin as a control should probably be reviewed with caution.
In Vitro- Versus In Vivo-Produced Embryos
Four genes were selected for RT-PCR to confirm the array data at the four-cell stage. Destrin (DSTN) is a mammalian protein that rapidly depolymerizes F-actin in a stoichiometric and pH-independent manner [47]. Destrin clones (pfeto1-012-G04) were detected mainly in in vitro-produced blastocysts by Whitworth et al. [10]. Here we found that DSTN levels were higher in the in vitro-produced four-cell stage embryos as compared with the in vivo-produced embryos.
The EST named pgvo1-003-H04 is similar to the message for poly(A) binding protein interacting protein 1 isoform 1 (PAIP1). While translation is regulated by the binding of the eIF4 complex to the mRNA, translation initiation is further regulated by the poly(A)-binding protein and its interaction with the eIF4 complex. This circular interaction facilitates poly(A) shortening [48] and translation. We found that PAIP1 mRNA level was elevated in the in vitro-produced four-cell-stage embryos as compared with the in vivo-produced embryos.
The proteasome degrades proteins marked with multiubiquitin chains. The process of ubiquitination uses an ubiquitin-activating enzyme, a ubiquitin-conjugating enzyme, a substrate-specific ubiquitin-protein ligase, and an additional conjugation factor [49]. The message level for UBE4B (pnatal4-011-G02) was higher in the in vivo-produced four-cell-stage embryos as compared with the in vitro-produced four-cell-stage embryo. It also increased significantly from the four-cell stage to the blastocyst stage.
Nuclear autoantigenic sperm protein isoform 2 was originally thought to be an acidic testis and sperm-specific protein containing two histone-binding domains [50]. The message level for NASP (pnatal4-004-E05) was higher in the in vivo-produced four-cell-stage embryos as compared with the in vitro-produced four-cell-stage embryo. It also increased significantly from the germinal vesicle stage to the four-cell stage. Mouse two-cell-stage embryos cultured in nonblocking conditions had higher levels of NASP than did those cultured in conditions that cause the embryo to stop development, or block, at the two-cell stage [51]. Thus, our results are consistent with higher levels being associated with higher developmental competence.
Five genes were selected for RT-PCR to confirm the microarray data. The high mobility group 1 (peov3-011-E12) proteins are nonhistone chromosomal proteins that localize to the nucleus and bind to specific DNA structures that are bent or kinked [52]. We found that HMGB1 was higher in blastocyst-stage embryos that were produced in vivo as compared with the in vitro-produced blastocysts. This might be an indirect indicator of more DNA (nuclei) in the in vivo-produced embryos as compared with the less developmentally competent in vitro-produced embryos.
ATP synthase is a complex composed of over 16 different proteins, 2 of which are encoded by the mitochondrial genome. ATP synthase, H+ transporting, mitochondrial F1 (ATP5A1: p8mm4-014-D06) is a nuclear encoded gene whose protein contributes to the overall function of the ATP synthase [53, 54]. We found that ATP5A1 was higher in the in vivo-produced blastocysts as compared with the in vitro-produced blastocysts, suggesting a higher metabolic rate in these more developmentally competent embryos.
The small nuclear ribonuclear proteins (pd12fol-010-H12) are responsible for the processing of pre-mRNA before it leaves the nucleus and enters the cytoplasm [55]. The B and D core protein of the snRNP does not localize to the nucleus between germinal vesicle breakdown and the late four-cell stage in the pig [56]. The localization to the nucleus of the Y12 antigen during the four-cell stage is 
-amanitin sensitive. When nuclei that have the Y12 antigen readily detectable are transferred to an enucleated oocyte and the oocyte is activated, then the nuclei lose their reactivity to the Y12 antigen, indirectly indicating that there is no mRNA processing and hence no mRNA synthesis. Similar to the HMGB1, we found a higher level of LSM2 message in the in vivo-produced embryos as compared with the in vivo-produced embryos, and this may be indicative of either more nuclei or a more transcriptionally active genome associated with more developmentally competent embryos.
The DNAj/HSP40 (pfeto1-019-A08) family of proteins stimulate ATPase activity and thus regulate molecular chaperone activity [57, 58]. A large number (31) of ESTs for DNAj subunits are on our array. We found DNAJB6 message to be higher in the in vitro-produced blastocyst-stage embryos as compared with the in vivo-produced blastocyst-stage embryos.
Activated leukocyte cell adhesion molecule (p4mm3-009-F11) is a member of the immunoglobulin superfamily of adhesion molecules expressed by a variety of hematopoietic cells, melanomas, carcinomas, and neuronal cells [59] and is usually restricted to subsets of cells involved in dynamic growth [60]. The message for ALCAM was higher in the in vitro-produced blastocyst-stage embryos as compared with the in vivo-produced blastocyst-stage embryos.
Other Genes of Interest
A few other genes of interest that have a pattern of detection similar to the mouse include Y-box proteins (MSY2) that are involved with transcriptional and translational control [61] of genes and zygote arrest 1 (ZAR1) (both decrease from oocyte to the blastocyst stage; MSY2 [24], ZAR1 [15]). Pregnancy-associated glycoprotein 2 message was previously shown to increase from the oocyte to four-cell stage [62], and this result was confirmed by the microarrays in the present article.
Comparisons with bovine are also interesting. A recent paper by Pennetier et al. [63] suggested that ZAR1 and growth differentiation factor decreased in expression from the oocyte to blastocyst stage. We show a significant decrease in transcript levels for both genes from the oocyte to four-cell stage and a further decline in ZAR1 from the four-cell stage to the blastocyst stage, while there were no differences between the in vitro- and in vivo-produced embryos. Similarly, peroxiredoxins (PRDX) 2–6 were evaluated in in vitro-produced bovine embryos [64]. We have PRDX 2–5 on our array. Similar to the bovine, we found that PRDX2 decreased at the maternal to zygotic transition and that PRDX3 decreased at the maternal to zygotic transition with an increase at the blastocyst stage. We found no significant change in PRDX4 in the pig, while Leyens et al. [64] could not detect it at any preimplantation stage. Finally, while no difference was found between the preimplantation stages of bovine embryo development for PDRX5, we found an increase at the blastocyst stage.
With a normalization of the cell stage to the onset of major amounts of transcription, clearly there is a great deal of similarity between the transcriptional patterns in mice and in pigs. However, an interesting difference that was detected was the decrease in message abundance of leukemia inhibitory factor receptor (LIFR). In the mouse, LIFR tended to increase to the blastocyst stage [15]. In the pig, we found a significant decrease from the germinal vesicle and four-cell stage to the blastocyst stage. Such an observation might explain why it has been so difficult to isolate embryonic stem cells from pig embryos.
Confirmation of Message Level
While we report here that there are many genes for which patterns are confirmed by similar expression in other species or by our previous report on the relative abundance of message within libraries generated from various stages of embryos, there is likely some bias in the system. Some bias showed up in our previous libraries, as all of the candidates that were reported previously [10] are not at the top of the list in this article. In addition, there is likely some bias introduced during the amplification performed in the present article. Fortunately, this bias should be the same across the different stages of development. There were a total of 39 comparisons of message levels evaluated by RT-PCR, and only the pattern of one was grossly inconsistent with the data generated by the microarray.
EASE Analysis
The EASE program identifies pathways of genes that appear to be up- or downregulated. The pathways that were identified to be upregulated in the four-cell stage are somewhat similar to with those seen in the mouse, i.e., pathways associated with nuclear structure and function [16]. Similarly, in the blastocyst-stage embryos, there were pathways associated with ribosomal function and ion transport that are up regulated [16]. When a similar comparison in the mouse was made between in vivo-produced embryos and in vitro-produced embryos (cultured in Whitten medium or KSOM/AA), only 114 and 29 genes, respectively, changed in message level [25]. Both in vitro groups had pathways that represented protein synthesis, cell proliferation, and transporter function being downregulated. Our results in the pig showed changes in different categories, i.e., decreases in plasma membrane, protein-tyrosine-phosphatase activity, and increases in nucleolus, small nucleolar ribonucleoprotein complex, and RNA binding and processing.
Application
Identification of individual genes that change in expression level during development is really only the first step. There are many different applications that can be envisioned to extend from the data generated to date. One thing that can be done is to look for genes that encode secreted molecules that have higher levels in the in vivo-produced embryos as compared with the in vitro-produced embryos. Such candidates that are twofold higher in the in vivo-produced blastocysts as compared with the in vitro-produced blastocysts include cortistatin (pnatal4-021-C07), bikunin (pd12-14end-003-H01), interferon gamma (MI-P-E4-abo-e-E09), and insulin-like growth factor binding protein 7 (pd12-14-end-006-D02). Assays are already developed for, or could be developed for, most of these genes. Developmentally competent embryos might be better identified by use of assays such as these.
Another is to look for putative receptors that increase as development proceeds from the oocyte to blastocyst stage, as the ligand for these receptors may be candidates to add to culture media to improve development in vitro (Table 8). One of the pathways represented in the EASE results and one of the receptors represented in Table 8 is the receptor for epidermal growth factor. Previous studies have shown that addition of epidermal growth factor to pig embryo culture medium enhances the cellular division of the embryos [65]. Conversely, the ligand for receptors that decrease in abundance might be used for the addition to culture media of one-cell- or two-cell-stage embryos (Table 9).
|
|
Summary
These initial studies into gene expression patterns during pig embryogenesis have provided a wealth of data that will be hopefully aid in 1) understanding embryogenesis in the pig, 2) identifying markers that can be used to noninvasively identify embryos of high developmental competence, and 3) provide candidate ligands that can be added to defined systems of embryo production to increase the percentage of and quality of blastocyst-stage embryos. Clearly, this article represents a starting point in describing gene expression in pig embryogenesis.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
2 Correspondence: R.S. Prather, E125D ASRC, 920 East Campus Dr., University of Missouri-Columbia, Columbia, MO 65211. FAX: 573 884 7827; PratherR{at}Missouri.Edu
Received: 22 November 2004.
First decision: 6 December 2004.
Accepted: 7 February 2005.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
