| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* Department of Animal Science and
and
Department of Statistics, Iowa State University, Ames 50011;
and
School of Animal Science and Veterinary Medicine, Huazhong Agricultural University, Wuhan, 430070, P. R. China; and
and
Department of Animal Science, Michigan State University, East Lansing 48824
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
Key Words: Complementary Deoxyribonucleic Acid • Gene Expression • Pigs • Skeletal Muscle
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
). In the pig, muscle fibers form in two stages during gestation, and differences between genetically obese and lean pigs for DNA content and muscle weight become apparent by 80 d of gestation (Swatland et al., 1994). This is approximately 10 d after the initiation of secondary fiber formation (Swatland et al., 1973). Detecting the gene expression differences between developmental ages or between different pigs with different muscle growth phenotypes may provide candidate genes that can be further investigated for association with meat production and quality traits. For example, myostatin gene was found to have a higher expression level in longissimus muscle in lower-birth-weight piglets than in normal-weight piglets (Ji et al., 1998). However, previous studies have only investigated one or two genes at a time, and the use of array technology to quantify the pattern of gene expression of many genes in parallel during porcine skeletal muscle development has not been published. Thus, techniques that can provide large-scale gene expression data will be powerful tools for determining potential associations between gene expression and phenotype outcomes. To monitor the gene expression patterns in different developmental ages of porcine skeletal muscle, a macroarray was developed and used to examine differential gene expression between pre- and postnatal muscle samples. The reproducibility of this low-cost and sensitive technique was also evaluated, and macroarray results were confirmed for several differentially expressed genes using quantitative RNA blot analysis. |
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
Muscle Samples
Muscle samples from four ages (75- and 105-d fetal, 1- and 7-wk postnatal) were used in the initial screen. The 75- and 105-d fetal muscle samples consisted of a mixture of hind limb muscles (including gluteus, semitendinosus, and semimembranosus) because of the small size of individual muscle types. Furthermore, tissues from multiple fetuses were pooled at each of these ages. For each of the 1- and 7-wk-old samples, semitendinosus muscle tissue from a single pig in each age was used. For the second macroarray experiment, hind limb muscle samples from two independent pools of 75-d fetal pigs and semitendinosus muscle samples from two animals at 1 wk postnatal were used to do the analysis (all were different biological samples than those used in the initial screen). All the pigs were Yorkshire x Landrace crossbreeds and were from the same herd.
Ribonucleic Acid Extraction, Labeling and Hybridization of Target to Membranes
Total RNA was extracted using Trizol reagent (Gibco, BRL, Rockville, Maryland) according to the manufacturer’s instructions. First strand cDNA targets were made from 30 to 50 µg of total RNA using 32P-dCTP and Superscript II reverse transcriptase (Gibco) and oligo-dT18 primers (Schummer et al., 1997). Duplicate membranes were hybridized with targets from each developmental age. Equivalent amounts of purified target were added to each membrane. Prehybridization was conducted at 68°C for 6 to 8 h, and hybridization was conducted for 18 to 20 h in 6x SSC, 1% SDS, 10 mM Na-phosphate buffer (pH 6.7), 10x Denhardt’s, 200 µg/mL of denatured salmon sperm DNA, and 100 mM sodium pyrophosphate (Na-POP). Membranes were washed twice with 2x SSC, 0.1% SDS for 5 min at 68°C, twice with 0.2x SSC, 0.1% SDS for 20 min at 68°C, and then washed a final time in 0.2x SSC, 0.5% SDS, and 0.1% Na-POP for 20 min at 68°C.
Quantification and Analysis of Hybridization Signals
Membranes were exposed to a phosphor imager screen (Molecular Dynamics, Sunnyvale, CA) after hybridization and washing, and then scanned by using a Storm 860 machine (Molecular Dynamics). Images were quantified by ImaGene 4.1 software (BioDiscovery, Los Angeles, CA).
Normalization and Identification of Differentially Expressed Genes
In order to compare hybridization signal intensities across membranes, a normalized signal intensity for each spot was obtained by subtracting the average of the log median of the 10 values (0.125, 0.25, 0.5, 1, and 2 ng in duplicate) of complex controls on a particular membrane from the log median intensity associated with each spot on the membrane. This complex control consisted of double-stranded cDNA obtained by reverse transcription of a total RNA sample derived from combining equal quantities of skeletal muscle, liver, lung, and spleen RNA. These normalized signal intensities are estimates of the log of the ratio of each spot’s signal intensity relative to the complex control for each membrane.
Twofold comparison and the mixed linear model analyses were used to judge if a gene was differentially expressed between muscle samples. We use yijklm to denote the normalized signal intensity corresponding to age i, animal j, membrane k, EST l, and spot m (i = 1,2; j = 1,2; k = 1,2; l = 1, . . ., 238; m = 1,2). The fold change for EST l between ages 75-d and 1-wk was estimated as exp(
), where 
denotes the average normalized signal intensity of the eight spots corresponding to age i and EST l (averaged over animals, duplicate membranes, and duplicate spots). EST with fold changes larger than two between the two ages were considered to be potentially differentially expressed between ages, subject to verification by the linear model analysis described below.
The model can be represented in symbolic form as follows:
 |
where µ denotes the average normalized signal intensity; ai denotes the effect of age i; sj(i) denotes the random effect of muscle sample j nested within age i; mk(ij) denotes the random effect of the duplicate membrane k nested within age i and muscle sample j; gl denotes the effect of the lth EST; (ag)il denotes the interaction effect of age i and the EST l; (sg)jl(i) denotes the random interaction effect of the lth EST with sample j nested within age i; (mg)kl(ij) denotes the random interaction effect of the lth EST with membrane k nested within age i and sample j; and em(ijkl) denotes the random effect of spot m of age i; muscle sample j, membrane k, and EST l (i = 1,2; j = 1,2; k = 1,2; l = 1, . . ., 238, m = 1,2). All random effects were assumed to be independent, normally distributed, mean-zero random variables with constant variance within each random effect category.
For each EST, the mean normalized signal intensities for the two ages were tested for equality using a t-test. The appropriate error mean square for the denominator of the t-test involves a linear combination of the mean square for samples nested within ages and the mean square for EST x sample interaction. The denominator degrees of freedom for each t-test were approximated using the method of Satterthwaite (1946), and the Holm (1979) method was used to adjust the P-values from the t-tests to obtain approximate strong control of the overall type I error rate. EST whose mean normalized signal intensities differed significantly between the two ages (Holm-adjusted P-value less than or equal to 0.01) were judged to be differentially expressed across ages.
RNA Blot Confirmation
Several genes were selected for RNA blot hybridizations to confirm the differential expression found in the macroarray analysis. These genes were selected for several reasons: first; to further investigate specific expression patterns; second; to compare expression patterns in the pig to those in other species for specific known genes; third, to confirm differential expression for novel porcine EST (genes). We also selected a novel cDNA-M218 for RNA blot study since this gene showed an expression pattern very close to both the twofold difference (1.94) and the P < 0.01 statistical difference criteria (0.0114). More importantly, M218 seems to be upregulated in postnatal muscle, one of only two genes to show this pattern in the macroarray analysis.
We first chose seven developmental time points, 45–, 60-, 75-, and 105-d fetal muscle samples and 1-, 3-, and 7-wk postnatal muscle samples, which included the four samples used in the first macroarray, to survey gene expression across these ages. Second, RNA from the two 75-d fetal and 1-wk postnatal pig muscle samples in the second macroarray analyses were used in additional RNA blot analyses to collect data from three biological samples to confirm the macroarray results. RNA blot analysis was performed according to Northern Max kit protocols (Ambion, Austin, Texas). Hybridization signals were visualized by phosphor imaging and scanning with the Storm 860 phosphor imager was performed to quantify results. After hybridization, blots were stripped in 0.1% SDS and re-probed with additional genes. All blots were also probed with ß-actin. Signal intensities from the 75-d and 1-wk muscle samples for each gene were normalized to the optical density values for 28S RNA band and Student’s t-test was used to analyze whether there was a significant difference in expression level between those two stages.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
Identification of Differential Gene Expression
All EST with a clear signal were re-arrayed after the first screen of four muscle ages (n = 238). These EST represent 193 different genes. The RNA samples from two additional biological replicates from 75-d and 1-wk were labeled and hybridized to membranes printed using these re-arrayed EST. Typical autoradiography patterns and scatter plot analyses within and between ages are shown in Figures 1A to 1C. Figures 1B and 1C show that the correlation coefficient of normalized signal intensity in within-age comparison is higher (r = 0.95 in 75-d samples, r = 0.86 in 1-wk samples) than the correlation coefficient of normalized signal intensity in across-age comparisons (r = 0.53).
|
), only three EST (1.2%) were found to have an estimated fold change in excess of two between the two 75-d fetal samples, whereas only five EST (2.1%) had a twofold change between the two 1-wk samples. A twofold expression change was set to indicate that a gene was likely to be differentially expressed between these ages. We found 37 EST that represent 28 genes that exhibited a twofold difference in expression level between 75-d and 1-wk ages. None of these genes had greater than twofold differences in the within-ages analysis. To further test whether these genes were differentially expressed, a linear mixed model analysis was performed. In this statistical analysis, 38 EST, which include all 28 genes identified above, had Holm-adjusted P-values less than 0.01, which indicated statistically significant expression differences. Of those 28 genes identified by both analyses of the re-arrayed EST, 23 also showed more than twofold differential expression in the original screening experiment. The estimated signal intensity relative to control within each age 
and Holm’s adjusted P-value from the t-test between ages for those 28 genes are listed in Table 1.
|
and 3 with Table 1). For example, as expected for a downregulated gene (Table 1), elongation factor 1-a had a higher level of expression at fetal age (75-d) than at postnatal age (1-wk) (Figures 2A and 3A). A similar pattern was seen for an unknown EST, E3-aaf-f-11 (Figures 2A and 3A). In addition, an unknown clone (M218), which had less than a twofold difference between ages based on the macroarray results (1-wk/75-d = 1.94, P = 0.0114, Table 1), showed significant different (P < 0.05) expression in RNA blot analysis of about sixfold (compare Figure 2B and Figure 3A). Results from RNA blot analysis using a ß-actin probe (Figure 2C), was as expected (Tuggle and Schmitz, 1994), with two transcripts (ß-actin and 
-actin) evident in 45-, 60–, and 75-d fetal muscle RNA, and only one transcript (-actin) observed in 105-d fetal, and 1-, 3-, and 7-wk postnatal muscle RNA. Finally, confirmation of macroarray results for two intermediate filament protein genes is described in the next section.
|
|
). This result is consistent with a report that indicated vimentin protein decreases relative to the muscle-specific IF protein desmin during maturation of muscle (Bilak et al., 1987). To confirm and extend the macroarray results, expression changes of vimentin and desmin were also determined by quantitative RNA blot hybridization (Figures 2D and 3). Results showed vimentin had a relatively higher expression level in 75-d fetal muscle than 1-wk postnatal muscle. The desmin gene (Tuggle et al., 1999) did not show significant changes in expression between the two ages in macroarray analysis. This result was also confirmed by RNA blot analysis (Figures 2D and 3). |
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
; Rast et al., 2000; Kelly et al., 2000). In this study, a handheld array tool was used to make cDNA macroarrays for differential gene expression analysis with radioactively labeled targets. The advantages of this approach, when compared to the widely used robotic tools, are that the macroarray is inexpensive to make, the analysis is sensitive in detecting low levels of transcripts and requires less input RNA than microarrays which use fluorescence detection, and the hybridization can be performed by using standard molecular biology laboratory equipment. High-quality, robust images with relatively low background were produced using the arrays made in this study, and data obtained were validated by additional analyses. These results demonstrate that this technique can be used to monitor the expression of hundreds or even thousands of genes in laboratories with economical, standard materials and equipment.
Definition of the source of variation in macroarray experiments is very important to avoid false positive results in macroarray analysis. We found that results from the 75-d fetal tissue samples had the largest variation when nonhomogenized portions from the 75-d hind limb muscle were used (data not shown). However, when muscle samples at any age were homogenized and used to prepare RNA, much of the variation was eliminated (data not shown); all subsequent work described used this approach. Another source of variation was in target preparation (e.g., independent RNA isolation and target incubation introduced higher variation than when one target preparation was split to two membranes for hybridization, although this kind of variation was much smaller than sample processing in fetal ages). From the experience of this study, we suggest that duplicate sampling of homogenized tissues for target preparation in duplicate macroarray hybridizations can produce sufficiently consistent results. Using this method, we found that the expression level difference between two different animals within each age for any one gene was low, which indicates that the differential expression of those genes is mainly caused by biological differences rather than animal variability (Figure 1).
To further confirm the results, several genes were selected for RNA blot hybridization. The RNA blot expression patterns observed for these genes were completely consistent with the macroarray results, in that each pattern predicted by macroarray results was seen across an expanded series of time points and quantitatively confirmed in analysis of additional animals between 75-d and 1-wk ages.
The linear model analysis for the experiment identified genes differentially expressed between 75-d and 1-wk ages by comparing variation between ages to a measure of variation within ages obtained from the two independent tissue samples used for each age. The methods of statistical analysis that we have employed are related to methods for microarray analysis proposed by Kerr et al. (2000) and Wolfinger et al. (2001). Kerr et al. (2000) used linear model analysis to identify differential expression in cDNA microarrays. In contrast to the models we used, Kerr et al. treated all effects as fixed. For the data sets considered here, this strategy would lead to a serious underestimation of standard errors for estimates of differential expression.
Wolfinger et al. (2001) emphasized the importance of identifying and accounting for random effects in microarray analyses. They illustrated how the mixed procedure of SAS (SAS Inst., Inc., Cary, NC) can be used in two steps to first normalize data and then test for differential expression, while properly accounting for random effects. The method they described requires two calls to the mixed procedure. The residuals from the first call were used as response variables in the second step of the analysis. Part of the reason for using two separate calls to the mixed procedure was that analysis with only one call to the mixed procedure could be too computationally demanding when thousands of genes are analyzed simultaneously. We were able to fit our mixed model using only a single call to the mixed procedure of SAS because we analyzed several hundred rather than several thousand genes at a time.
Both Kerr et al. (2000) and Wolfinger et al. (2001) identified genes differentially expressed among groups by testing for group x gene interaction. Implicit in such tests is a normalization based on average expression over all EST on an array. This type of normalization can be suitable if the average expression of all EST is expected to remain constant across all comparison groups. This is often a reasonable assumption if the genes spotted on the array can be considered a large sample that is generally representative of the population of all genes. The experiments that we have conducted involve a relatively small number of EST, many of which were selected because of their potential for involvement in muscle development. It is possible that a large proportion of the spotted EST could be upregulated in one age relative to another, in which case normalization by an array average could partially or completely mask the effects of interest. Thus control spots on each array were used for normalization prior to mixed linear model analysis. Furthermore, rather than testing for age x gene interaction, we directly test for differences among the mean normalized intensities of an EST across ages. In the notation used to describe the mixed linear model for the second macroarray experiment, the test for age x gene interaction for EST k is a test of H0:(ag)1l = (ag)2l against HA:(ag)1l 
(ag)2l, whereas the test for differential expression that we employed is a test of H0: a1 + (ag)1l = a2 + (ag)2l against HA: a1 + (ag)1l a2 + (ag)2l. Like Wolfinger et al. (2000), we relied on the assumption of normality for our random effects. In addition to the assumption of normality, we have made simplifying assumptions about the variance-covariance structure of the normalized intensities. Like Wolfinger et al. (2000), we used random effects to allow for a nontrivial correlation structure among all observations, but, unlike Wolfinger et al., we do not assume that each gene has its own distinct variance. Doing so would yield too few degrees of freedom for reliable estimates of gene-specific variance. Instead, we pooled information about variability from all genes to obtain more stable standard errors for our estimates of differential expression. This same strategy was used by Kerr et al. (2000).
Similar expression patterns for some known genes in this study have been demonstrated by other studies. For example, elongation factor 1- showed downregulation from 75-d fetal to 1-wk postnatal in our study. Study of the rat also found that elongation factor 1- has a very high expression level in earlier developmental ages (Lee et al., 1993). This gene belongs to a 20-member gene family, 18 of which are retropseudogenes. Only elongation factor 1- and S1 are truly expressed, and there is an age-dependent developmental switch from elongation factor 1- to S1 in brain, skeletal muscle, and heart (Khalyfa et al., 1999).
Our study also showed that a number of ribosomal protein (RP) genes, which are involved in protein synthesis, were downregulated during muscle development. An increased expression level in the hypothalamus and brown adipose tissue for the RPL3 gene was found in mouse lines with high fat deposition, which were selected for heat loss (Allan et al., 2000). Many results have shown most RP mRNA are lower in differentiated cells than in undifferentiated cells: RPL10, RPS16, RPL32, and RPL18 in heart (Agrawal et al., 1987; Kirby et al., 1995); acid ribosomal phosphoprotein P1 in intestinal (Maheshwari et al., 1993); and RPL20 in leukaemic cells (Goldstone et al., 1993). There are 15 ribosomal protein genes in our study; 12 of them were downregulated (Table 1), whereas two of them did not change their expression level (RPL3, RPL23, and acidic ribosomal phosphoprotein P1) (http://www.ans.iastate.edu/graduate/pig_muscle_macroarray.xls). Among those down regulated, a lower level of expression has also been found for RPS20 and RPS3a during neural cell differentiation (Bevort et al., 2000). Further, RPL10a, first identified as CsA-19, was shown to be downregulated in the thymus by cyclosporin A (CsA) treatment (Fisicaro et al., 1995). Our finding that expression of this CsA-responsive gene was lower in 1-wk postnatal muscle is interesting since CsA can also inhibit myoblast differentiation by blocking the expression of myogenic genes in mouse skeletal muscle cell culture (Friday et al., 2000). Our study is the first report of differential expression of these RP genes in porcine muscle development.
RNA alternative splicing is an important event in development and differentiation. Previous studies have shown that splicing factor arginine and serine-rich proteins (SFRS) are involved in splicing in other systems (Mayeda and Krainer, 1992; Caceres et al., 1994) but not in muscle (van den Bosch et al., 1996). In this study, we found that a new splicing factor SFRS12 was downregulated from 75-d fetal to 1-wk postanatal (E2-ab-f-06, Table 1). The human SFRS12 cDNA was recently deposited to Genbank (December 2002, accession No. AF 459094, Zhang et al., 2002). The functional role of this gene needs to be further investigated and our result is the first report of the differential expression for this gene in any species.
One unknown muscle clone, M218, showed an increased abundance in postnatal muscle in macroarray and RNA blot analysis. The transcript size of M218 is approximately 3 kb (Figure 2B). In unpublished work, we have determined a 1.5-kb 3'-end sequence for this gene (S.-H. Zhao and C. K. Tuggle, data not shown). When BLAST analysis to the human genome and dbEST using this sequence data were performed, no significant matches were found. Similar negative results were seen with mouse genome and EST database searches. However, the M218 sequence matches three pig EST and one bovine EST. These results may indicate that M218 represents a new gene in pig and cow, which has a relationship with muscle growth.
Some EST (40) on the macroarray were derived from differential display PCR (dd-PCR) experiments that compared RNA samples derived from either pig skeletal muscle at 60 d of gestation, 105 d of gestation, and 7 wk postnatal, or RNA derived from whole pig fetuses at 21, 35, and 45 d of gestation (C. W. Ernst, unpublished data). Our results showing that none of them exhibited differential expression may be due to the fact that the ages we tested in this study were different from the differential display PCR comparisons or that the dd-PCR were false positives.
There are two major intermediate filament genes expressed in skeletal muscle, desmin, and vimentin. Previous studies showed that more vimentin protein is present in less differentiated myocytes, whereas desmin is the major intermediate filament protein in differentiated myocytes (Bilak et al., 1987; Yang et al., 1996; Sejersen et al., 1993). In this study, both the macroarray and RNA blot results showed that the expression level of vimentin mRNA drops in skeletal muscle from 75 d to 1 wk postnatal (Figures 2D and 3). In the macroarray analyses, the desmin gene did not show differential expression between 75-d fetal and 1-wk muscle, which agreed with the RNA blot results. The change in the vimentin/desmin mRNA ratio observed by RNA blot analysis in this study is similar to a previous report that measured expression at the protein level, where desmin protein was shown to proportionally increase in differentiated muscle cells compared to Vimentin protein (Bilak et al., 1987). Our multitime point analyses were based on an n = 1 study; thus, more animals in each time point need to be tested to confirm that the decrease of vimentin protein results from a decrease in transcription or an increase in degradation of vimentin mRNA in future study. Studies in human and rat also reported a decrease of vimentin mRNA during the development of myofibers from myogenic precursor cells (Vaittinen et al., 1999; Kryszke et al., 2001).
Between the two ages, 28 genes (14.5%) showed a change in expression (Table 1). This percentage is relatively high since many transcriptional profiling experiments report 0.5 to 1% of the genes analyzed differentially expressed (Chen et al., 1998; Tanaka et al., 2000; Campbell et al., 2001). This high percentage of differentially expressed genes may be due to the fact that many of the EST were from d-20 and -45 embryo cDNA libraries. Most differentially expressed genes were embryonic EST that had a higher expression level in the earlier developmental age (75 d fetal) than in the later age (1-wk postnatal). These results indicate that expression of these genes is down regulated during muscle development. The majority of the genes examined in this study exhibited no statistically significant change in expression level across ages, which suggests these genes may not change their steady-state mRNA level during muscle development. In addition, GAPDH expression showed a greater than twofold difference between the two ages. Interestingly, Hsiao et al. (2001) also found that GAPDH was within the 15 most variable genes among 451 housekeeping genes across normal human tissues.
|
Implications |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
|
Footnotes |
|---|
2 Correspondence: 2255 Kildee Hall (phone: 515-294-4252; fax: 515-294-2401; E-mail: cktuggle{at}iastate.edu).
Received for publication February 13, 2003. Accepted for publication May 7, 2003.
|
Literature Cited |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
This article has been cited by other articles:
|
|
 |
|
  R. Davoli and S. Braglia Molecular approaches in pig breeding to improve meat quality Brief Funct Genomic Proteomic, January 21, 2008; (2008) elm036v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Ponsuksili, E. Murani, C. Walz, M. Schwerin, and K. Wimmers Pre- and postnatal hepatic gene expression profiles of two pig breeds differing in body composition: insight into pathways of metabolic regulation Physiol Genomics, May 11, 2007; 29(3): 267 - 279. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. S. Lin and C. W. Hsu Differentially transcribed genes in skeletal muscle of Duroc and Taoyuan pigs J Anim Sci, September 1, 2005; 83(9): 2075 - 2086. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. A. Byrne, Y. H. Wang, S. A. Lehnert, G. S. Harper, S. M. McWilliam, H. L. Bruce, and A. Reverter Gene expression profiling of muscle tissue in Brahman steers during nutritional restriction J Anim Sci, January 1, 2005; 83(1): 1 - 12. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
