| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Infigen, Inc.,5 DeForest, Wisconsin 53532
Department of Biomedical Sciences,6 Cornell University, Ithaca, New York 14853-6401
Department of Biology,7 University of Pennsylvania, Philadelphia, Pennsylvania 19104-6018
Virginia Bioinformatics Institute and Department of Statistics,8 Virginia Tech, Blacksburg, Virginia 24061
NuPotential, LLC,9 Baton Rouge, Louisiana 70808-4124
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
-glucosidase), two lysosomal-related genes (prosaposin and lysosomal-associated membrane protein 2), and a gene associated with stress responses (heat shock protein 27) along with major histocompatibility complex class I, nidogen 2, a putative transport protein, heterogeneous nuclear ribonuclear protein K, mitochondrial 16S rRNA, and ES1 (a zebrafish orthologue of unknown function). The three remaining genes are novel. To our knowledge, this is the first report comparing individual embryos produced by NT and IVF using cDNA microarray technology for any species, and it uses a rigorous experimental design that emphasizes statistical significance to identify differentially expressed genes between NT and IVF embryos in cattle.
developmental biology, early development, embryo, trophoblast
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Published reports also indicate that the majority of embryos produced by nuclear transfer (NT) are compromised, because they are unable to develop past the blastocyst or early postimplantation stages [13, 14]. The basis for these failures is unknown, but a popular hypothesis is that inefficient reprogramming of the donor nucleus results in inappropriate expression of genes required for embryonic development. For example, using cloned mouse embryos that express enhanced green fluorescent protein driven by an Oct4 promoter, stochastic-like expression patterns were observed. For example, some blastomeres expressed both the transgene and endogenous Oct4 gene, whereas other blastomeres expressed one gene but not the other [15]. Oct 4 is a transcription factor essential for preimplantation development [16].
An expression-based reprogramming assay using preimplantation embryos could be used to test systematically hypotheses aimed at improving NT-enabled reprogramming, especially for animals with long gestations. For example, identification of gene classifiers that distinguish embryo types (e.g., in vitro fertilization [IVF] or NT) potentially represents a diagnostic tool to assess hypothesis-driven perturbations. Normalizing expression levels of NT embryo gene classifiers may indicate improvement with regard to developmental competence and reprogramming efficiency. Moreover, identification of differentially expressed genes may suggest methods of intervention. For example, genes overexpressed in cloned embryos may be caused by persistent expression from the donor nucleus. Down-regulation of these genes in donor cells before NT can potentially be achieved using technologies such as small interfering RNA (siRNA). Alternatively, exposure of donor nuclei to chromatin remodeling factors or other exogenous molecules could enhance reprogramming efficiency. At present, evaluating the efficacy of such treatments requires costly embryo transfers when using animals with long gestations. Development of embryo-based predictive models could minimize the tremendous costs associated with such transfers and provide a rationale for further testing in vivo. Moreover, gene expression analysis of individual embryos using cDNA microarrays should yield new insights regarding the regulatory mechanisms involved in NT-induced reprogramming and embryonic development.
Microarrays are a powerful discovery tool that will allow these models to be realized. In the present study, we developed a method to identify differentially expressed genes between multiple, individual embryos using cDNA microarrays. These embryos represented two distinct classes (NT and IVF), which differ greatly in average developmental competence (
2% and 50%, respectively). We used an interwoven-loop experimental design, and the data were analyzed using linear mixed model methodology (LMMM) with restricted maximum likelihood (REML) estimation of the variance components [17] implemented in the software package ASREML [18]. To our knowledge, this is the first report describing the use of these methodologies to identify differentially expressed genes between NT- and IVF-produced bovine embryos.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Oocytes aspirated from ovaries obtained from cows at slaughter were matured overnight in maturation medium (medium 199; Biowhittaker, Walkersville, MD) supplemented with LH (10 IU/ml; Sigma, St. Louis, MO), estradiol (1 mg/ml; Sigma), and 10% fetal bovine serum (Hyclone, Logan, UT) at 38.5°C in a humidified 5% CO2 incubator. Matured oocytes were inseminated by combining frozen-thawed sperm as described previously [19]. The IVF embryos were scored for blastocyst development on Day 7 (day of insemination = Day 0) and graded according to International Embryo Transfer Society (IETS) guidelines.
Cell Culture
Bovine donor cells were derived from fetal body tissue (outer part of the body minus the head and viscera) of a 58-day-old bovine fetus (obtained from a slaughterhouse) as previously described [2] and cultured in -minimal essential media (MEM) (Gibco-BRL, Carlsbad, CA) supplemented with 10% fetal bovine serum and 0.1 mM 2-mercaptoethanol (Gibco-BRL). Confluent culture dishes were passaged in 1x trypsin-EDTA (Gibco-BRL) at least once before use in NT. These cells were the source for all NT embryos used in the analysis described.
Nuclear Transfer
Bovine NT was performed and embryos cultured as described previously [2] using matured oocytes. The NT embryos were scored for blastocyst development on Day 7 (day of NT = Day 0). All NT embryos were produced from a single donor cell line.
RNA Purification and Amplification
The RNA was isolated from individual Day 7, stage 7 (blastocyst), IETS grade 1 or 2, NT- and IVF-matured embryos using the Picopure RNA isolation kit (Arcturus, Mountain View, CA). Linear amplification of total cellular RNA was performed using the RiboAmp RNA Amplification kit (Arcturus) followed by two additional rounds and according to the manufacturer's protocol. Amplified RNA (cRNA) quality and concentrations were determined by gel electrophoresis and A260/280 readings, respectively. On average, 25.7 ± 3.3 µg of cRNA were produced for each embryo after two rounds of amplification and 159.6 ± 5.5 µg after round three for each embryo used in the microarray study (data are expressed as mean ± SEM throughout). Similar results were obtained with IVF embryos.
cDNA Library Construction
A cDNA library was custom-made (Stratagene, La Jolla, CA) from bovine cultured genital ridge cells. Briefly, bovine genital ridges were isolated from a 55-day-old slaughterhouse bovine fetus. Cells were grown in -MEM (Gibco-BRL) supplemented with 10% fetal bovine serum and 0.1 mM 2-mercaptoethanol. The cDNA was synthesized from RNA isolated from 80 x 106 cultured genital ridge cells and ligated into the lambda arms of the Uni-ZAP vector (Stratagene). In vivo mass excision of the pBluescript phagemid from the Uni-ZAP XR vector was performed to generate a cDNA library following the manufacturer's protocols. The library was plated onto LB-ampicillin (100 µg/ml; Fisher Scientific, Pittsburgh, PA) agar plates (Bio101, Carlsbad, CA) and incubated overnight at 37°C. Individual colonies were picked from the agar plates using the BioPick robot (BioRobotics, Cambridge, UK), transferred into a 384-well plate (Nunc, Rochester, NY) containing LB-ampicillin (100 µg/ml) media (Bio101), and grown overnight with shaking at 37°C. Following overnight growth, glycerol (Fisher Scientific) was added to each well (10% final concentration) using the BioPick robot, and the cells were processed according to Hedge et al. [20]. Briefly, from the overnight culture suspension, 10 µl were transferred into a 96-well polymerase chain reaction (PCR) plate (MJ Research, Waltham, MA) using the Biomek 2000 workstation (Beckman, Fullerton, CA). Each well contained 90 µl of sterile water (Sigma). The plates were incubated at 95°C for 10 min to lyse the cells and release the plasmid clones. Plates were centrifuged for 5 min at 670 x g to pellet the cell debris. The supernatant (80 µl) was transferred into a fresh, 96-well PCR plate. Alternatively, 5 µl of overnight culture were spotted onto a Clonesaver (Whatman, Clifton, NJ) card. Punches of 1.2 mm were placed in a 96-well PCR plate and washed twice with water. Insert cDNA samples were amplified by PCR using the flanking vector-specific primers T7 and T3. Each 50-µl reaction contained 2 µl of lysed culture supernatant, 1x AmpliTaq Reaction buffer, 1.5 mM MgCl2, 1.0 mM of each primer, 0.2 mM of each dNTP, and 2.0 U of AmpliTaq DNA Polymerase (Perkin-Elmer, Wellesley, MA). Thermal cycling conditions were as follows: 3 min at 94°C; 30 cycles of 1 min at 94°C, 1 min at 56°C, and 1 min at 72°C; and a final extension of 3 min at 72°C on a PTC-225 Tetrad (MJ Research). Following PCR amplification of the clone inserts, the PCR products were electrophoresed in 1.0% Tris-borate-EDTA agarose gels to confirm amplification and purified using the Multiscreen PCR filter (Millipore, Billerica, MA) using the Biomek 2000.
Expressed Sequence Tag Sequencing
Sequencing of cloned cDNA inserts from the cultured genital ridge cDNA library was performed using the ABI Prism BigDye Terminator Cycle Sequencing V2.0 Ready Reaction kit (Applied Biosystems, Foster City, CA) following the manufacturer's protocol and using the supplied reagents. Sequencing reactions were electrophoresed and analyzed using a 377 DNA Sequencer (Applied Biosystems). The resulting data were trimmed of vector sequence using Factura V2.2 (Applied Biosystems) and analyzed by BLAST (Basic Local Alignment Search Tool; National Center for Biotechnology Information [NCBI], Bethesda, MD). Sequence data, samples, and containers were managed and/or analyzed using SQL*LIMS, SQL*GT, and Biomerge Software (Applied Biosystems).
Sequence Data Analysis
Sequence data for all clones were analyzed by BLASTN and BLASTX from the NCBI. A cutoff value of P = 10–10 was the general criterion for scoring a sequence as having a significant match. Subsequently, each sequence was categorized functionally. Interclone comparisons to determine redundancy and frequency of duplication was performed using Biomerge software (Applied Biosystems).
cDNA Probe Production and Labeling
Amplified RNA (cRNA) from individual NT and IVF embryos was converted into aminoallyl-labeled cDNA probes using slight modifications of protocols described by Hedge et al. [20] and DeRisi (http://www.microarrays.org; http://cmgm.stanford.edu/pbrown/protocols/amino-allyl.htm). Briefly, 6 µg of random hexamers were added to 5 µg of cRNA; final volume was 18 µl. The reaction mixture was incubated at 70°C for 10 min, chilled on ice for 3 min, and then placed at room temperature. Reverse transcription with incorporation of aminoallyl-dUTP (Sigma) was carried out according to the method described by DeRisi. The reaction products were purified according to the method described by Hedge et al. [20]. Coupling of Cye dyes was carried out as described by DeRisi, and purification was carried out with the following modification: Before the final purification with QiaQuick PCR Purification Kit (Qiagen, Valencia, CA), 35 µl of 3 M sodium acetate (pH 5.2; Sigma) were added to each reaction to lower the pH. Following QiaQuick purification, the probe was speed-vacuum centrifuged to 20 µl, and 1 µl of Human Cot-1 DNA (20 µg/µl; Invitrogen, Carlsbad, CA) was added to the probe.
Array Design and Microarray Printing
Microarrays were prepared by printing purified, amplified PCR products in 50% dimethyl sulfoxide (DMSO; Sigma) onto GAPS II aminosilane-coated glass microscope slides (Corning, Corning, NY). Briefly, 5 µl of each purified PCR product was transferred into a Genetix 384-well plate, and 5 µl of DMSO (Sigma) were added to each well. A 4 x 12, solid-pin tool was used to print the cDNAs onto the slide surface by the MicroGrid TAS arrayer (BioRobotics) according to the manufacturer's protocol. Each cDNA was arrayed in duplicate in a 10 x 11 block pattern representing 5280 spots (or features) or 2640 cDNAs on the chip. The print spots were 200 µm in diameter and 400 µm apart. Human ß-actin was printed as a positive hybridization control and herring sperm DNA as a negative hybridization control. The printed arrays were baked at 80°C for 4 h.
Hybridization
Hybridization and posthybridization washes were carried out according to the method described by Hedge et al. [20] with the following modifications: The arrayed slides were placed into a Corning GAPS II slide container and incubated in warmed, filtered prehybridization buffer containing 5x SSC (1x SSC: 0.15 M sodium chloride and 0.015 M sodium citrate; USB, Cleveland, OH), 0.1% SDS (USB), and 10 mg/ml of BSA (Roche Diagnostics, Basel, Switzerland). Following prehybridization, slides were washed in room-temperature Milli-Q water, dipped in isopropanol (Fisher Scientific), and dried with high-pressure CO2 gas. Hybridization buffer contained 50% formamide (Fisher Scientific), 10x SSC, and 0.4% SDS and was warmed to 42°C before use. Following posthybridization washes, slides were dried under CO2 pressured gas.
Image Analysis and Data Normalization
Slides were scanned using a GenePix 4000A scanner (Axon Instruments, Foster City, CA). Features were analyzed using GenePix Pro Version 3.0 software (Axon Instruments) and raw data exported into Excel (Microsoft, Redmond, WA). Two raw-intensity measurements were used: Foreground mean minus background median, and foreground mean not adjusted for background. The RI plots [21, 22], scatter plots of log(R/G) versus log(RG), with R denoting the red raw intensity and G the green raw intensity, were performed for these data and all 100 arrays. The data were log transformed, and we performed local regression (lowess and spatial lowess) to recenter the data around the fitted curves as proposed by Yang et al. [21].
Experimental Design and LMMM Analysis
The experimental design was an interwoven loop using 100 arrays in which 10 NT embryos were compared to 10 IVF embryos (Fig. 1); gene expression data were obtained by comparing 10 individual stage 7, Day 7, IETS grade 1 or 2 bovine blastocyst embryos produced by NT to 10 individual stage 7, Day 7, IETS grade 1 or 2 blastocyst embryos produced by IVF. For the first 20 arrays (the loop portion), each NT embryo was paired with two IVF embryos. For the interwoven portion, each NT embryo was paired with the remaining IVF embryos.
|
The data were analyzed with LMMM using variance-components estimates obtained by REML. The software packages used were SAS Proc Mixed [23] and ASREML [18, 22]. Data analysis was performed using suitable transformation and normalization of the raw data (intensities) before LMMM analysis with lowess and spatial lowess on median background- and nonbackground-adjusted data. We verified that these different data adjustments essentially produced the same lists of genes identified as differentially expressed. The data were analyzed with the following linear mixed model:
where y is the log-transformed and lowess-adjusted intensity (the average of the two spots per array for each gene), µ is the overall mean, Ai is the random effect of array i (i = 1, ..., 100), Dj is the fixed effect of dye j (j = 1, 2), AxDij is the random effect of interaction between array and dye, Tk is the fixed effect of embryo type (NT: k = 1; IVF: k = 2), El(Tk) is the random effect of embryo l (l = 1, ..., 10) within type, Gm is the fixed effect of gene m (m = 1, ..., 2640), T(Gkm) is the fixed effect of interaction between type of embryo and gene, E(T)xGklm is the random effect of interaction between embryo within type and gene, AxGim is the random effect of interaction between array and gene, and eijklm is a residual.
The variance-covariance structure of these types of data is not trivial to specify. Its specification was dictated to a large extent by the software available for analysis. We first analyzed the data using Proc Mixed of SAS [23] following the two-step approach described by Wolfinger et al. [24], because SAS was not capable of analyzing the data on all genes jointly. With this analysis, only one gene was significantly differentially expressed. We therefore reanalyzed the data with another software package, ASREML [18, 22], which was capable of processing the data on all genes jointly. For the global random factors array and embryo, dispersion assumptions were Ai iid N(0, 
) and El(Tk) iid N(0, 
), whereas iid denotes independently and identically distributed and N denotes normal distribution, assuming different variances among embryos for the two types (NT and IVF). For the gene-specific factors and for the two-step analysis with SAS, dispersion assumptions were AxGim iid N(0, 
), E(T)xGklm iid N(0, 
), and eijklm iid N(0, 
), with separate variance components for each gene. For the joint analysis, AxGim iid N(0, 
), E(T)xGklm iid N(0, 
), and eijklm iid N(0, 
), where all variance components except for the residual variance were assumed to be constant across all genes but the residual variance was allowed to differ between nine groups of genes (genes were grouped according to their variability of expression; details not shown), with g(m) denoting the group (m = 1, ..., 9) of gene m (m = 1, ..., 2640). Grouping was performed because of a limit on the number of different variance components allowed by ASREML and to obtain more stable estimates of the residual variances, which are used in the tests for differential expression.
To determine which genes were differentially expressed between NT and IVF embryos, the following t-statistic (tm) was computed for each gene m as
Note that previous authors (see, e.g., [25, 26]) have used a different contrast, which is based only on the interaction effects—that is, identical to the above equation with the T terms omitted. However, differential expression of gene m in types 1 and 2 implies that the difference between the average expression value of gene m in type 1 and the average expression value of gene m in type 2 is nonzero, and this difference is represented by the numerator of the above equation. If the difference among the T main effects (T1 – T2) is negligible because of most genes not being differentially expressed and/or half the differentially expressed genes being up- and the other half down-regulated, then little difference should be found between the two contrasts. Moreover, concerns exist regarding the T main effects not cleanly representing average mRNA expression levels in different cell populations but, rather, being confounded with technology effects unique to individual samples (sample extraction and amplification) and hybridizations (array x dye interactions). In experiments with biological replication, the sample effects are captured in the biological replicate effects (E(T)), and replication across arrays allowed us to separate out the array x dye interactions as well so that the type (T) effects should be clean. (For a recent discussion of these contrasts, see Black and Doerge [26]). In the identification of differentially expressed genes, P values were adjusted conservatively for multiple testing with the Bonferroni and step-down Bonferroni methods. We also implemented an alternative adjustment for multiple testing by using the q method described by Storey and Tibshirani [27], which is based on the false-discovery rate [28] and is less stringent.
Real Time TaqMan Analysis
For each target gene, a sensitive real-time TaqMan PCR [29–31] was optimized to quantify transcripts for vimentin, keratins 8 and 19, major histocompatibilty complex (MHC)-I, heat shock protein 27 (Hsp27), and nidogen 2. Briefly, two primers and an internal, fluorescent-labeled TaqMan probe (5' end, reporter dye FAM [6-carboxyflourescein]; 3' end, quencher dye TAMRA [6-carboxytetramethylrhodamine]) were designed using Primer Express software (Applied Biosystems) (Table 1). As an endogenous control, a TaqMan PCR system was used targeting bovine glyceraldehyde-3-phosphate dehydrogenase (Gapdh) as described elsewhere [29–31]. The length of the TaqMan PCR products were held very short (between 73 and 170 base pairs [bp]) to enable high amplification efficiencies. For the bovine-specific target genes, exon-exon boundaries were deduced from human sequences. Either forward or reverse primers (or both for vimentin) were placed over a boundary to prevent the amplification of genomic DNA background in the total RNA.
|
The cDNA was synthesized directly from RNA isolated from individual embryos as described by Bertolini et al. [32] or from 1 µg of amplified RNA using 100 U of SuperScript III (Invitrogen), 300 ng of random hexadeoxyribonucleotide (pd(N)6) primers (random hexamer primer), 10 U of RNaseOut (RNase inhibitor), and 1 mM dNTPs (all Life Technologies) in a final volume of 40 µl. The reverse-transcription reaction proceeded for 50 min at 50°C. After addition of 60 µl of water, the reaction was terminated by heating for 5 min to 95°C and cooling on ice.
Each PCR reaction contained 400 nM of each primer, 80 nM of the TaqMan probe, and commercially available PCR mastermix (TaqMan Universal PCR Mastermix; Applied Biosystems) containing 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 5 mM MgCl2, 2.5 mM dNTPs, 0.625 U of AmpliTaq Gold DNA polymerase per reaction, 0.25 U of AmpErase UNG per reaction, and 5 µl of the diluted cDNA sample in a final volume of 25 µl. The samples were placed in 96-well plates and amplified in an automated fluorometer (ABI PRISM 7700 Sequence Detection System; Applied Biosystems). Amplification conditions were 2 min at 50°C, 10 min at 95°C, and then 40 cycles of 15 sec at 95°C and 60 sec at 60°C.
Relative Quantification of Target Gene Abundance
Final quantification was done using the comparative CT method [29– 31] and is reported as relative transcription or the n-fold difference relative to a calibrator cDNA (i.e., lowest target gene transcription). In brief, the signal of the endogenous control Gapdh was used to normalize the target gene signals of each sample. The 
CT for gene transcription was calibrated against control groups. The relative linear amount of target molecules relative to the calibrator was calculated by 
. Therefore, all target gene transcription is expressed as an n-fold difference relative to the calibrator. Statistical analysis to assess significantly different grouping was performed using the Wilcoxon rank-sum test.
Immunocytochemistry
The IVF and NT embryos were scored for blastocyst development on Day 7 (day of insemination = Day 0; day of NT = Day 0). The embryos were mounted onto polylysine-coated coverslips by spinning at 1500 x g for 10 min. Embryos were fixed in PBS containing 4% paraformaldehyde (Fisher Scientific). For vimentin detection, the fixed embryos were washed in PBS, blocked in PF buffer (2% BSA [Fisher Scientific], 0.1% Triton X-100 [Fisher Scientific], and 0.02% SDS in Ca2+/Mg2+-free PBS [Fisher Scientific]) for 10 min and incubated with a mouse monoclonal anti-vimentin antibody immunoglobulin (Ig) M (antibody 7752, V1-01, 1:200 dilution; Abcam, Cambridge, U.K.) in PF buffer for 1 h. The coverslips were washed in PF buffer and incubated with a rabbit polyclonal secondary antibody to mouse IgG + IgM + IgA (fluorescein isothiocyanate [FITC]; antibody 8517, 1:400 dilution in PF buffer; Abcam) for 1 h. The coverslips were washed in PF, stained with Hoechst 33342 (Sigma), washed in PBS, and mounted in Fluoromount (Electron Microscopy Sciences, Hatfield, PA). Microscopy was performed with a Axiovert S100 Zeiss microscope (Carl Zeiss, Inc., New York, NY). Images were captured with a Zeiss Axiocam camera.
For MHC-I staining, whole NT and IVF embryos were mounted on glass slides, fixed in acetone, and then shipped to the Department of Biomedical Sciences of Cornell University for MHC-I labeling [33]. Briefly, whole embryos were blocked with normal goat serum and then incubated for 2 h at 37°C with IL A19 mouse anti-bovine MHC-I primary antibody diluted 1:2000. Incubation with a FITC-conjugated anti-mouse secondary antibody (1:50; Vector Laboratories) followed. Nonimmune ascites (Sigma) was used as a negative control.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
To initiate development of an expressed sequence tag database relevant to NT-induced reprogramming, we constructed a cDNA library from a cultured genital ridge donor cell line that had been used successfully to produce cloned cattle. The library contained approximately 106 independent primary clones. The average insert size of the cultured genital ridge cDNA library was 1.0 kilobase (kb), with insert sizes ranging from 0.5 to 2.2 kb. Approximately 50 000 clones from the cultured genital ridge library were randomly picked and sequenced. The average sequence length was 800 bp, with ranges from approximately 0.2 to 1.1 bp. Internal redundancy was determined by comparing sequence data with each other using BioMerge software. Clones having significant BLAST scores were classified into 21 functional categories. Sequences lacking a significant BLAST score were categorized as "unknown," and those in which no match was identified were classified as "novel." These data provided the basis to select 2640 unique sequences to serve as the basis of our first-generation reprogramming microarray for cattle. The distribution of genes present on our microarray with respect to functional category is shown in Figure 2. Minimal redundancy was further verified by hybridizing the array to a ß-actin probe. Less than 1% of the spots had detectable signal after normalization and background subtraction (data not shown), indicating that the amplification procedure did not result in overexpression of abundant transcripts.
|
Reproducibility of Linear Amplification
The reproducibility of the amplification procedure was assessed by several methods. We first assessed the reliability of amplification by dividing isolated RNA obtained from an individual, IVF-produced embryo into equal aliquots, performing parallel rounds of amplification and computing a correlation coefficient after hybridization. Relatively high ratios were observed after both round 2 (average r = 0.98) and round 3 (average r = 0.97) round of amplification. In a separate experiment, amplification was performed in duplicate from the same starting RNA, and the number of detected features versus nondetected features was determined between rounds 2 and 3. In total, 2515 genes had detectable hybridization signals when probed with cRNA from round 2. The same genes also were detected using cRNA from round 3. When probed with cRNA from round 2, 125 features were not detected, and when probed with cRNA from round 3, the same features were not detected. No discrepancies were observed between detected and nondetected genes between amplifications in round 2 and in round 3. Finally, the percentage of features having twofold or greater differences after normalization between round 2 and round 3 was determined. Sixty-one features (1.1%) were identified that met the criteria. In only one example was this discrepancy manifested in both duplicated features, and that instance likely represented an artifact introduced by the amplification protocol. The duplicate spots in the remaining 59 samples had less than twofold differences when compared to the corresponding doublets between hybridization. Thus, the technical reproducibility and reliability do not appear to be compromised significantly between two and three rounds of amplification for this set of genes. Though systematic errors can be introduced, appropriate experimental design and analyses should minimize their consequences.
LMMM Analysis
The data were first analyzed by applying the two-step LMMM analysis described by Wolfinger et al. [24] with SAS [23]. This approach identified only one gene as significantly differentially expressed between NT and IVF embryos. Therefore, a joint analysis of the data on all genes was performed using the ASREML software [18, 22]. This analysis found the same gene plus another 17 genes; that is, a total of 18 genes were identified as being differentially expressed using the Bonferroni correction for multiple testing. Table 2 lists the 15 known genes identified in the analysis. To obtain this list, several datasets were analyzed: one containing background-corrected data that were only log transformed, another containing background-corrected and lowess-adjusted data, another as the previous one but without background correction, and another containing background-corrected and spatial lowess-adjusted data. These analyses validated each other by producing the same set of genes.
|
Based on the q-value method, 87 additional genes were found to be differentially expressed, for a total of 105 genes, when a q-value cutoff of 0.05 was used. This results in a false-discovery rate of 5% among the significant genes; that is, 5 or 6 of the 105 genes are expected to be false positives. The difference in the set of identified genes between the joint analysis with ASREML and the two-step, or gene-specific, analysis with SAS was caused by the use of different contrasts (those in (5) or those with the T main effects omitted, respectively, with the main effects contrast between NT and IVF, T2 – T1, being significant in this analysis), by the difference in degrees of freedom between gene-specific and joint analyses, and by the use of joint versus gene-specific estimates of the error variance and other variance components. The REML estimates of the variance components computed with ASREML are listed in Table 3. The variance components associated with array (A, AxG) were by far the largest, whereas the sum of the two embryo-related variances (E(T), E(T)xG) was approximately equal to the residual or error variance on average across both types, but the variance among NT embryos was much smaller than the variance among the IVF embryos both globally and at the gene-specific level. The variance caused by labeling (AxD) appears to be quite small.
|
Validation of Microarray Data
In concordance with our data, aberrant expression of MHC-I had been previously reported in trophoblast cells of first-trimester cloned cattle [33] and was an indirect validation of our results. To substantiate further the microarray results, the expression levels of six differentially expressed genes (keratins 19 and 8, vimentin, hsp27, nidogen 2, and MHC-I) were confirmed by real-time TaqMan PCR. For example, cytokeratin 19 was identified as being down-regulated in NT embryos relative to IVF embryos. The average n-fold difference relative to a calibrator was 31.45 ± 8.57 in the 10 IVF embryos used for the microarray analysis, compared to 4.18 ± 1.17 for the NT embryos (Fig. 3A, white bars, and Table 4). Similarly, the up-regulation of nidogen 2 expression was also confirmed by TaqMan analysis (IVF, 11.55 ± 3.65; NT, 40.52 ± 12.16) (Fig. 3E, white bars, and Table 4). Wilcoxon rank-sum tests were performed to assess the probability that the IVF and NT embryo classes were different. Vimentin, cytokeratins 8 and 19, nidogen 2, and MHC-I grouped significantly different between NT and IVF classes (P < 0.05) and, thus, were consistent with LMMM data indicating differential expression between NT and IVF embryos. In contrast, Hsp27 did not group statistically different; however, average values were consistent with LMMM data indicating down-regulation in NT embryos.
|
|
TaqMan data also indicated variability of transcript levels between individual embryos of the same class. For example, vimentin expression ranged from 0.00 to 12 503.12 in seven of the IVF embryos. Only two NT embryos had detectable levels of vimentin mRNA. In contrast, nidogen 2 had a broader range of values (1.82–127.12) and was detected in all 10 NT embryos examined. Seven of the 10 IVF embryos had measurable levels of nidogen 2 mRNA but with a narrower range (0.00–35.02). These variations may reflect both temporal differences introduced by in vitro procedures and imperfect gene regulation induced by the NT process. In contrast, Gapdh was detected in all 20 embryos used in the microarray analysis, and very little variability was detected between embryos within a class. These data highlight the necessity for replication and analysis of statistical significance based on variance. Furthermore, the analysis of pooled embryo samples likely would mask the variability observed between individual embryos, resulting in high rates of false positives.
We further validated these genes by real-time TaqMan analysis using a new set of 10 IVF and 10 NT embryos but without linear amplification. The analyses of individual embryos using real-time TaqMan PCR without linear amplification has been described previously by Bertolini et al. [32]. Results were consistent with microarray data indicating differential expression between classes and all six genes, including Hsp27, grouped distinctly (P < 0.05). Variability was observed between individual embryos within a given class.
Though Gapdh was detected in all 20 embryos used in the analysis, we observed that vimentin values were not determined in 8 of the 10 original NT embryos used in the microarray experiment when reassessed using real-time TaqMan. Similarly, 3 of the 10 IVF samples lacked values. To address these observations further, we analyzed vimentin protein levels in individual Day 7, stage 7, IETS grade 1 or 2 NT and IVF embryos using immunocytochemistry. In general, vimentin was detected in the majority of IVF-produced embryos (see Fig. 4A for a representative example). Vimentin protein in NT embryos appeared to be reduced when compared to IVF-produced embryos (see Fig. 4C for a representative example). The majority of NT embryos examined (9 of 10) had apparent diminished levels of vimentin protein based on visual observation when compared to IVF-produced embryos (7 of 10 displayed a pronounced fluorescence). A small fraction of the NT embryos had detectable levels of vimentin protein, and some IVF embryos had apparently reduced vimentin as assessed visually by immunofluorescence. These patterns were consistent with microarray and TaqMan data indicating generally lower levels of vimentin gene expression.
|
We also examined by immunocytochemistry 24 NT and 13 IVF embryos for MHC-I protein expression, because NT-generated embryos have been reported to express aberrant MHC-I protein, which normally is suppressed in first-trimester cattle embryos [33]. A strong signal was observed in 19 of the 24 NT embryos examined (see Fig. 5C for representative example) but was less apparent, or absent, in the 13 IVF-generated embryos (see Fig. 5A for a representative example).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Interestingly, three genes encoding intermediate-filament (IF) proteins (vimentin and cytokeratins 8 and 19) are underexpressed in NT embryos. Cytokeratin 8 and vimentin expression are required for posthatching development of bovine embryos [35], which in turn is required for appropriate establishment and development of the chorioallantoic placenta. Cytokeratin 8 is detected in Day 14 trophoblast cells, indicating epithelial differentiation, but it is absent in less differentiated epiblast cells [35]. In contrast, vimentin is detected in these latter cells and associated with their recruitment to form the primitive streak as well as the definitive endoderm and mesoderm [35]. Aberrant expression of these genes as early as Day 7 likely will interfere with downstream differentiation. Furthermore, appropriate establishment and development of the chorioallantoic placenta depends on proper formation of germ layers (endoderm, ectoderm, and mesoderm) during gastrulation. Misregulation of IF gene expression likely indicates the future placental abnormalities often observed in cloned cattle.
Heat shock protein 27, an ATP-independent chaperone that promotes the refolding of denatured proteins and protects against apoptosis triggered by growth factor deprivation, hydrogen peroxide, anticancer drugs, and other stimuli [36], also is underexpressed in NT embryos. We have noted that the incidence of apoptosis in Day 7, stage 7 bovine embryos produced by NT is approximately 75%, which is significantly higher than that in either IVF (50%) or artificially inseminated embryos (22%) of the same stage (unpublished results), and similar observations have been reported by others [37–39]. It is tempting to speculate that the reduced level of Hsp27 expression in the NT embryos is somehow linked with the increased incidence of apoptosis in these embryos. In addition, Hsp27 binds to soluble IF proteins [40] and IFs. Binding is not stress dependent, but stress can facilitate this association. The down-regulation of the expression of three IFs and Hsp27 may contribute to the poor development of NT-generated embryos because of a generalized perturbation in the IF system.
Bovine embryos have a detectable hypoblast by Day 11, and further development is thought to require an extracellular matrix. The extracellular matrix is believed to connect the hypoblast, inner cell mass, and trophoblast [35] of Day 11 bovine embryos, and the localization of laminin and collagen IV between the hypoblast and trophoblast [41] supports this view. Nidogens 1 and 2 are implicated in establishing basement membrane assembly by connecting the major networks formed by laminins and collagen IV [42]. In nidogen 1 or nidogen 2 knock-out mice, neither protein is essential for basement membrane formation or maintenance, whereas preliminary data indicate that double knock-outs are lethal [43]. In cattle, neither gene has been analyzed closely during development. We found that nidogen 2 is up-regulated in NT embryos when compared to IVF controls, and this could compromise the establishment of basement membranes in relation to the hypoblast, epiblast, and trophectoderm.
Cattle MHC-I haplotypes carry several class I genes, and the number of expressed genes varies between haplotypes (C. Davies, personal communication). Consistent with previous findings [33], the MHC-I genes are also up-regulated in NT embryos when compared to IVF controls. Trophoblast MHC-I expression normally is suppressed during the first trimester of bovine gestation, and MHC I overexpression in the cloned embryos could induce a maternal lymphocytic response that is detrimental to continued development. Consistent with this hypothesis is our observation of a general reduction in placentome numbers in spontaneously aborted cloned fetuses (unpublished data), a phenotype that would be consistent with compromised placental development. It will be interesting to test the hypothesis that knock-down of MHC-I expression in donor cell lines using the siRNA technology will minimize MHC-I expression in NT-generated embryos and reduce abnormal fetomaternal interactions and fetal loss.
In summary, we have identified 18 genes aberrantly expressed in Day 7, NT-generated embryos when compared to IVF controls using cDNA microarray technology to measure large-scale gene expression and LMMM for statistical analysis of these array data. Several of these genes have been implicated in previous reports as being associated with later stages of embryogenesis and placental development. Normalizing the expression patterns of these genes may improve full-term survivability of cloned cattle. Further development of reliable, comprehensive datasets should provide a strong foundation to discover the key mechanisms required for reprogramming and to develop models that can be used to test hypothesis-driven, methodology-based alterations.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
2 Correspondence: Kenneth J. Eilertsen, Pennington Biomedical Research Center, 6400 Perkins Road, Baton Rouge, LA 70808-4124. FAX: 225 763 0273; eilertkj{at}pbrc.edu
3 Current Address: Department of Anatomy and Cell Biology, College of Medicine, University of Iowa, Iowa City, IA 52242
4 Current Address: Pennington Biomedical Research Center, Stem Cell Biology Group, 6400 Perkins Road, Baton Rouge, LA 70808-4124
Received: 25 May 2004.
First decision: 18 June 2004.
Accepted: 21 September 2004.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
B-crystallin. J Cell Sci 1999 112:2099-2112[Abstract]This article has been cited by other articles:
|
|
 |
|
  R. E. Everts, P. Chavatte-Palmer, A. Razzak, I. Hue, C. A. Green, R. Oliveira, X. Vignon, S. L. Rodriguez-Zas, X. C. Tian, X. Yang, et al. Aberrant gene expression patterns in placentomes are associated with phenotypically normal and abnormal cattle cloned by somatic cell nuclear transfer Physiol Genomics, October 8, 2008; 33(1): 65 - 77. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. M. Giraldo, D. A. Hylan, C. B. Ballard, M. N. Purpera, T. D. Vaught, J. W. Lynn, R. A. Godke, and K. R. Bondioli Effect of Epigenetic Modifications of Donor Somatic Cells on the Subsequent Chromatin Remodeling of Cloned Bovine Embryos Biol Reprod, May 1, 2008; 78(5): 832 - 840. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Jincho, Y. Sotomaru, M. Kawahara, Y. Ono, H. Ogawa, Y. Obata, and T. Kono Identification of Genes Aberrantly Expressed in Mouse Embryonic Stem Cell-Cloned Blastocysts Biol Reprod, April 1, 2008; 78(4): 568 - 576. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 V. Duranthon, A. J Watson, and P. Lonergan Preimplantation embryo programming: transcription, epigenetics, and culture environment Reproduction, February 1, 2008; 135(2): 141 - 150. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. K Berg, C. Li, G. Asher, D. N Wells, and B. Oback Red Deer Cloned from Antler Stem Cells and Their Differentiated Progeny Biol Reprod, September 1, 2007; 77(3): 384 - 394. [Abstract] [Full Text] [PDF]
|
|
|
