| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
ANIMAL GENETICS |
* Cooperative Research Centre for Cattle and Beef Quality, CSIRO Livestock Industries, Queensland Bioscience Precinct, St. Lucia, Queensland 4067, Australia; and
and
Food Science Australia, Tingalpa DC, Queensland 4173, Australia
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionSummaryLiterature Cited |
|---|
Key Words: Beef • Complementary DNA • Gene Expression • Nutrition
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSummaryLiterature Cited |
|---|
). Despite an increased understanding of the role of these components, the contribution of nutrition and growth path to overall beef toughness is not well understood. There is a shortage of gene expression studies to characterize the changes associated with muscle remodeling in response to diet quality. Recent studies (Lee and Hossner, 2002; Lee et al., 2002) have shown that nutrient challenge and diet supplementation in ruminants can promote differential expression of tissue specific genes.
In this study, we compare gene expression in LM from Brahman steers fed high-, medium-, and low-quality diets using complementary DNA (cDNA) microarrays. Reverter et al. (2003a) presented aspects of the statistical analyses of this experiment, along with a preliminary list of 27 differentially expressed (DEX) elements. Two major issues were identified that warranted further research. First, the reference design did not allow for dye bias correction. Second, the list of candidate genes was neither annotated nor validated with quantitative real-time PCR (Q-PCR). Therefore, the objectives of this study were 1) to expand on the research of Reverter et al. (2003a) by adding six hybridizations, which would account for dye bias, and 2) to sequence, validate, annotate, and comprehensively profile the resulting list of DEX genes. The statistical approach is based on mixed-model ANOVA techniques and model-based clustering of the BLUP for the gene x diet interaction effect.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSummaryLiterature Cited |
|---|
. In short, 11 Brahman-cross steers weighing an average 302 ± 10 kg were randomly allotted to three diets. The ad libitum lucerne (alfalfa) hay diet (HIGH; n = 3 animals; namely H1, H2, and H3) was designed to allow the cattle to continue to grow at 0.8 to 1.0 kg/d and was the control diet. The restricted lucerne hay diet (MED; n = 4 animals; namely M1, M2, M3, and M4) was used to produce a moderate weight gain of approximately 0.3 kg/d. The ad libitum low-quality grass-based hay diet (LOW; n = 4 animals; namely L1, L2, L3 and L4) was included to produce a weight loss of 0.3 kg/d. After 27 d on the experimental diets, feed was removed from the steers the day before they were scheduled to be biopsied from the LM. For biopsy, steers were isolated and immobilized and sedated. A 1- to 2-g muscle biopsy was removed using a scalpel. All animal procedures were carried out in accordance with the Commonwealth Scientific and Industrial Research Organization (CSIRO) animal ethics guidelines.
Previous work on restricted nutrition in Bos indicus cattle (Hunter and Magner 1990a; Allingham et al., 1998) allowed us to manage the nutrition of the experimental groups to achieve three different growth rates. In the 30 d before the muscle biopsies, the experimental animals grew at rates indicative of the nutritional treatments. Animals in the HIGH group had ADG in the order of 1.0 kg/d, the MED group gained approximately 0.6 kg/d, and the LOW group lost weight at approximately 0.5 kg/d. The effects of restricted nutrition on whole-body metabolism have been published previously (Hunter and Magner, 1990b). In the 9 d before the biopsies, the animals were studied intensively in metabolic crates. Stress associated with the crates led to transient decreases in feed intake and consequent minor live weight loss due to decreased gut fill (HIGH = 5% live weight; MED = 3% live weight; LOW = 3% live weight). Animals were allowed to recover from this stress for 48 h after exiting the metabolic crates, and were fed the experimental diets during this time. Finally, animals were fasted in the 18 h before the biopsies to minimize regurgitation during anesthesia. Fasting the steers for less than 24 h would not have a significant effect on their blood glucose or live weight (Spanheimer et al., 1991) so changes to gene expression apart from those due to the experimental diets were not anticipated.
The process for RNA extraction and purification was performed as described in Reverter et al. (2003a). Anti-sense RNA (aRNA) amplification was performed using the MessageAmp aRNA Kit (Ambion, Austin, TX). Direct and indirect labeling procedures were performed as described in Lehnert et al. (2004).
We constructed a custom cDNA microarray from two cattle cDNA libraries derived from LM and s.c. fat tissue of a 24 mo-old grass-fed Angus steer (Lehnert et al., 2004). In short, 9,600 elements were printed in duplicate onto glass slides. The array consisted of 9,222 cattle probes, which comprised 7,291 anonymous cDNA from the bovine skeletal muscle and s.c. fat cDNA libraries and 1,915 bovine expressed sequence tags (EST) selected from the muscle and fat libraries, CSIRO cattle skin library (Wang et al., 2001), and the Meat Animal Research Center 1-4BOV libraries (Smith et al., 2001). In addition, 16 EST-derived oligonucleotides and the Lucidea microarray scorecard V1.1 (Amersham Biosciences, Piscataway, NJ) were printed on the array. Each array contained 19,200 cells arranged in 48 blocks of 20 rows x 20 columns each. Hence, EST were duplicated at least once on each array.
Microarray Experimental Design
Practical considerations included in the development of our design included transitivity (Townsend, 2003) and extendibility (Kerr, 2003). In this context, transitivity is defined by the ability to compare any two samples of interest directly from the same hybridization as well as indirectly through transitive association across two or more hybridizations. The concept of extendibility implies that additional samples can be added to a given design in a sensible way.
In the present study, the main objective of the gene expression experiment was to identify genes that were DEX among the HIGH, MED, and LOW groups. To this end, and considering issues of transitivity and extendibility, we developed an experimental design with two separate components. The first component involved the reference (REFE) design described in Reverter et al. (2003a), where the animals in HIGH were considered the reference sample and the total RNA from these animals was pooled and labeled with fluorescent green dye using direct incorporation of Cye3-dUTP (Amersham Biosciences, Piscataway, NJ) during the reverse-transcribed cDNA synthesis step. Consequently, RNA from animals in MED and LOW was labeled with fluorescent red dye using Cye5-dUTP as above but not pooled. The biopsy of animal L4 yielded insufficient RNA; therefore, to maintain the balance of the REFE design, two hybridizations were performed for animal L3. A total of eight hybridizations were carried out in this REFE component of the entire design.
The second component of the design involved an all-pairs (ALLP) comparison and was developed to address the possibility of dye bias and to increase technical replication. The ALLP component was performed using pooled amplified aRNA. The pairwise comparison involved pooling individual animal RNA samples within the dietary treatment groups HIGH, MED, and LOW and performing the pairwise hybridizations. Equal amounts of total RNA from individual samples within each group were pooled, and amplified aRNA was generated and used as the template for indirect red and green dye labeling of the target for microarray experimentation. For each comparison, a "dye-swap" experiment was performed. In a dye-swap, a replicate experiment is performed in which the target that was previously labeled with red is labeled with green dye and vice versa. Data from six pairwise comparison slides were generated in the ALLP component of the experimental design. Figure 1 shows a pictorial representation of the experimental design in which the two components (REFE and ALLP) are clearly distinguishable.
|
Data Acquisition Criteria
The GenePix 4000 optical scanner and the image analysis software GenePixPro 3.0, both from Axon Instruments Inc. (Union City, CA) were used to quantify the expression intensities. This software provides a distinct quality reading for bad quality spots that were subsequently excluded from the analyses. A single criterion for data acquisition was used based on those signals yielding a mean to median correlation >0.85 and measured by the ratio of the smallest of the median and mean by the largest (Tran et al., 2002). This editing criterion was applied for signals whose foreground intensity was greater than the background intensity but separately for the red and green channels. In total, there were 300,936 intensity records on 7,347 genes out of the original 9,274. Intensity records were background corrected and base-2 log-transformed. They averaged 9.5 with a SD of 2.2 and ranged from 0 to 16.
Data Analysis
The 14 slides resulting from the REFE and the ALLP components of the design were incorporated into a single analysis by fitting an ANOVA linear model to achieve normalization and estimate differential expression of genes across the three dietary treatments. The model was similar to the one suggested by Kerr and Churchill (2001), except that the effects of gene and the interactions between gene and the remaining main effects were fitted as random. Mixed models can accommodate more general covariance structure and provide shrinkage of estimated effects that can decrease bias arising when any type of data selection is applied. The optimality of mixed models to assess gene significance from cDNA microarray expression data was previously reported by Wolfinger et al. (2001) and more recently by Reverter et al. (2004b). The mixed model used in this study was as follows:
 | [1] |
where yijkgtr is the measured intensity (background corrected and based-2 logarithmic transformed) from component i (i = 1, 2, for REFE and ALLP, respectively), array j (j = 1 to 14), dye channel k (k = 1, 2, for red and green, respectively), dietary treatment t (t = 1, 2, 3, for HIGH, MED, and LOW, respectively), and repetition r on gene g.
The terms C, A, D, T, and AD were fitted as fixed effects and account for all effects that are not gene specific, have no biological significance, and the fitting of which aims at normalizing the data by accounting for systematic nongenetic effects. The random gene effect Gg contains the average level of gene expression (averaged over the other factors). The random gene x array interaction effect in (AG)jg models the effects for each spot and is crucial to the model as it serves to account for the well-known spot-to-spot variability inherent in spotted microarray data. The inclusion of this effect allows us to extract appropriate information about the treatment effects and obviates the need to form ratios (Wolfinger et al., 2001). The random gene x dye interaction effect in (DG)kg models the gene-specific dye effects that occur when subsets of genes exhibit higher fluorescent signal when labeled with one dye or the other, regardless of the treatment (Kerr et al., 2002).
The effect of interest was the random interaction between genes and diet treatments, (TG)tg because it captured differences from overall averages that were attributable to a specific combination of diet treatment t and gene g. Finally, 
ijkgtr is the random error term.
We assumed that Gg, (AG)jg, (DG)kg, (TG)tg, and ijkgtr were independent with zero mean and variance between genes (
), between genes within arrays (
2jg), between genes within dye channel (2kg), between genes within dietary treatment (2tg), and within genes (2
;), respectively. Variance components for gene effects and their interactions were assumed the same for all genes. Given that some RNA samples were from individuals and some were from pools, the assumption of independence and identical distribution could be compromised.
The model in Eq. [1] was regarded as the "Full" model, and three variations ("reduced" models 1, 2, and 3) were explored by ignoring either or both random interactions in (AG)jg and (DG)kg. Criteria for model selection include the likelihood ratio test (Stram and Lee, 1994), the Akaike information criterion (Akaike, 1969) and the Bayesian information criterion (Schwartz, 1978). Finally, the selected model was further scrutinized by examining the fitted residuals to check for modeling assumptions. To identify possible systematic patterns, we examined residuals separately for each components of the design and for each dye channel.
In all models, we obtained REML estimates of variance components and BLUP solutions using VCE software (Groeneveld and García-Cortés, 1998). The difference in gene expression for gene g (g = 1 to 7,347) under the HIGH dietary treatment compared with LOW is estimated by
 | [2] |
Large positive (negative) dg values are likely to belong to genes whose expression is down (up)regulated due to nutritional restriction. In addition, and to compare results with those reported by Reverter et al. (2003a), two additional measurements of (possible) differential gene expression will be explored and defined by
 | [3] |
and
 | [4] |
In this notation, d' g and d'' g are meant to address the contrast between HIGH and MED, and between MED and LOW, respectively.
Similar to the approach of Moser et al. (2004) and Reverter et al. (2004b), our goal here is to apply model-based cluster analysis to the preprocessed gene expression levels in dg, and see which genes will have relative levels far away from the majority. We identified DEX genes using BAYESMIX software (Reverter et al., 2003b) for Bayesian model-based clustering using mixtures of normal distributions. In the present study, mixture models with up to five components (or clusters) were contemplated.
Nucleotide Sequencing
Because most of the cDNA clones on the custom bovine microarray were anonymous at the initiation of the experiment, we sequenced and annotated the resulting DEX elements to determine their identity. Direct 5 (T3) sequence tags were generated on cDNA phage templates amplified by PCR using M13 universal forward and reverse primers (Wang et al., 2001). Sequence annotation was compiled by conducting basic local alignment search tool (Altschul et al., 1990) searches against the nonredundant and human reference sequence data-sets (GenBank; Benson et al., 2002), and against the interactive bovine single nucleotide polymorphism database (Hawken et al., 2004). To interpret the observations of differential expression, we used annotations of gene function derived from the Kyoto Encyclopedia of Genes and Genomes (http://www.kegg.com), and the Gene Ontology Consortium (Ashburner et al., 2000; http://www.geneontology.org).
Real-Time Q-PCR Validation
We selected five DEX genes for real-time Q-PCR verification. To determine the relative expression levels, we used four housekeeping genes, including glyceralde-hyde-3-phosphate dehydrogenase (GAPDH), ß-actin, 18SrRNA, and acidic ribosomal protein. ß-Actin and GAPDH were identified as suitable housekeeping genes according to the method of Wilson et al. (2003). These two genes exhibited low DEX and the lowest SD (a 95% confidence interval that was among the narrowest yet encompassing zero) using a large number of readings (137 and 193, respectively) across the 14 microarrays.
Total RNA was reverse transcribed into cDNA using random primers and SuperscriptII (Invitrogen, Carlsbad, CA). Reactions were run on an ABI PRISM 7900HT sequence detector (Applied Biosystems, Foster City, CA) using cycling parameters defined by the manufacturer. Each assay included (in triplicate) four housekeeping genes, a standard curve of five serial dilution points of a standard cDNA, a no-template control, and 1:10 dilution of each test cDNA. All PCR efficiencies were greater than 95%. Sequence detection software (Applied Biosystems) results were processed through QGENE (Muller et al., 2000) for further analysis.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSummaryLiterature Cited |
|---|
presents REML estimates of variance components for the Full model in Eq. [1] as well as for the three restricted models. Criteria for model selection (likelihood ratio test, Akaike information criterion, and Bayesian information criterion) revealed Eq. [1] as the model of choice. However, the estimates of remained unchanged across models and accounted for 79.2, 81.0, 79.5, and 81.1% of the total variation with the Full model, and Reduced models 1, 2, and 3, respectively. Similarly, the total variation accounted by (TG) was 3.6, 3.0, 3.7, and 3.0% for the same four models. These percentages provide an indication of the expected proportion of DEX genes and serve to control the false discovery rate (FDR), a major concern in gene expression (Van den Oord and Sullivan, 2003; Meuwissen and Goddard, 2004) studies. The reasoning behind using these percents to control the FDR can be found in Efron (2004), where FDR is defined to be fdr(z) 
f0(z)/(f(z), where f(z) = p0 f0(z) + p1f1(z) is the estimated density from the observed ensemble of z-values, f0(z) is the distribution of the z-values under the null hypothesis (i.e., nonsignificant or non-DEX in our context), and f1(z) is the distribution under the alternative hypothesis (i.e., DEX in our context).
|
shows plots of residuals vs. fitted values by dye channel and component of the design for the data analyzed on the base-2 log scale using the Full model in Eq. [1] and shows some modest heteroscedasticity but no other clear pattern. In spite of their power to assess model assumptions, residuals are not often examined and/or reported (one exception being Kerr et al., 2002). Systematic patterns as those reported by Kerr et al. (2002) were not found in the present study.
|
, and 
follows a t-distribution with N –2 degrees of freedom, where N = 7,347 the number of paired elements in r. Hence, the previously mentioned estimates were statistically different from zero and from each other (P < 0.01). In agreement with Reverter et al. (2003a), these results indicate that the biological distance, at the gene expression level, between HIGH and MED was much shorter than that between MED and LOW. The comparison of values in dg, d'g , and d''g from the present study with the t-statistics in Reverter et al. (2003a) (N = 4,747 genes) addressing the same contrasts revealed correlations of 0.814, 0.800, and 0.674 for HIGH vs. LOW, HIGH vs. MED, and MED vs. LOW contrasts, respectively. Only dg was considered for the task of identifying DEX genes because it had the highest variation, as well as a positive correlation with both d'g and d''g. The fitting of mixtures of normal distributions to the values in dg revealed the following three-component (or cluster) model as the one of best fit:
 |
where N(a,b) denotes a normal distribution with mean a and variance b. Figure 3 illustrates the empirical density function for dg, as well as the posterior probability (decision functions) of each value belonging to each cluster. An element in dg was classified to a cluster if its posterior probability was the largest. Note that there is no y-scale for the empirical density function. This scale corresponds to probability one for the decision function, and the empirical density is drawn to proportionality.
|
. The CSIRO sequence number of the five elements reported by Reverter et al. (2003a) that were not identified in the present study as being DEX included CCL013304, CCL013714, CCL014034, CCL013382, and CCL013742. For these elements and from the present study, measures of differential expression in dg were –1.004, –1.126, –0.884, 0.821, and 1.018, respectively. The first three elements correspond to clones that showed up-regulation due to nutritional restriction in Reverter et al. (2003a), yet their P-value (estimated from the observed proportion of more extreme differences in BLUP for gene x high-quality diet minus gene x low-quality diet solutions) was >0.05.
Applying the method of Reverter et al. (2004a) to the distribution of the 7,347 total and 156 DEX elements from this study, yielded an estimate of the sensitivity equal to 468 transcripts per million. This value contrasts with the estimate of sensitivity of 100 transcripts per million reported by Reverter et al. (2004b) for this same study after applying a multivariate mixed model to the analysis of the present study together with two other microarray studies performed using the same microarray slide.
After annotation analyses, the 57 upregulated elements were found to represent 29 unique genes, whereas the 99 downregulated elements represented 28 unique genes (Tables 2 and 3, for up- and downregulated genes, respectively). This finding reflects the inherent redundancy in the non-normalized cDNA libraries used to construct the microarray. The finding that some of the DEX genes identified in this study were represented by more than one array element adds empirical confidence to the results. Finally, there were a number of clones identified that could not be directly sequenced from PCR products. These included three clones from the upregulated group and nine clones from the down-regulated group. Hence, annotations could not be performed for these clones.
|
|
Genes Upregulated in Response to Nutritional Restriction
Putative gene identities and associated functions could be assigned to 26 of the 29 unique upregulated genes (Table 2
). The remaining three genes matched human sequences of unknown function, including DKFZp451G182, MGC:24616, and MGC:1136 (last three rows of Table 2). Upregulated genes were found to belong to functional classes involved in protein turnover (12 out of 26), cytoskeletal structure (5 out of 26), and metabolic homeostasis (3 out of 26). Additionally, genes involved in calcium signaling (calmodulin), signal transduction (TPT1), cellular trafficking (GABARAP), and polyamine synthesis (OAZ1) were identified.
Genes Downregulated in Response to Nutritional Restriction
Putative gene identities and associated functions could be assigned to 19 of the 28 unique downregulated genes (Table 3). The majority of these transcripts were associated with extracellular matrix structure (five of 29), as well as cytoskeletal structure (3 of 29; first eight rows of Table 3). The remaining eight genes matched sequences of unknown function from human (LOC51673 HSPC148, KIAA1194, DKFZp434G227), bovine (RP42-341K3, RP42-513g13, 245354 MARC 2BOV), and mouse (RP23-11P22).
Real Time Q-PCR Validation
To validate differential expression shown on the microarray, we quantified the expression of two highly expressed genes, including troponin and osteonectin(GenBank accession numbers CF614574 and CF614034, respectively) and three unknown transcripts (GenBank accession numbers CF615337, CF615329, and CF614934) by Q-PCR in HIGH vs. LOW samples. Table 4 presents the sequences of anti-sense primers used for synthesis of these five DEX genes for Q-PCR analyses. The relative quantification of candidate gene expression was performed by normalizing samples against four housekeeping genes. Similar expression profiles for all the controls were observed across the samples, adding credibility to the data. Data relative to GAPDH are presented in Table 5 as mean normalized expression values. Results from Q-PCR corroborate the differential expression results obtained using the microarray for four of the five genes tested. The data for troponin and the two unknown transcripts, CF615329 and CF615337, showed upregulation of gene expression by 4.4-, 5.7-, and 1.2-fold, respectively, and were in close agreement with the microarray results (Table 2). Results from Q-PCR for osteonectin showed a downregulation of approximately 2.5-fold, which also was similar to the microarray analysis (Table 3). Finally, a 4.3-fold upregulation of gene CF614934 between the two treatments was observed from the Q-PCR analysis, which contradicts the microarray results. This contradictory result was attributed to the very low average signal of 26 captured for this gene (Table 3). Microarray data are noisy, particularly at the low end of the spectrum and characterizing such noise is an area of intense and continuous development (Tu et al., 2002). Thirteen upregulated genes and 18 downregulated genes had average intensities lower than 784, the signal intensity of CF615329 and the lowest intensity at which fold differences were verified by Q-PCR. Genes exhibiting low-signal fold change will need to be validated independently of the microarray.
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSummaryLiterature Cited |
|---|
) and porcine(Zhao et al., 2003), which may impact on meat quality. However, large-scale gene expression studies to characterize changes associated with muscle remodeling in response to diet quality are uncommon. Investigation of these genes may provide mechanistic insights into the regulation of cellular development of carcass tissues and protein turnover in cattle. The present study used a custom microarray to uncover a set of genes that were DEX between steers losing weight due to restricted nutrition and adequately nourished or fully replete steers. The resulting DEX genes can be broadly grouped into the following categories based on the functions of the proteins they encode: proteins involved in protein turnover, cytoskeletal proteins, proteins involved in intermediary metabolism, and extracellular matrix proteins.
Genes for Proteins Involved in Protein Turnover
Caloric restriction in mammals tends to elevate protein turnover (Weindruch et al., 2001). Upregulated gene expression that is broadly consistent with this conclusion includes PSMB4, SUG1, ATP synthase, EF1
, cold shock domain protein A, and Dp-2.
Among the downregulated genes was ubiquitin-specific protease 16. This gene has been implicated in mitotic regulation (Cai et al., 1999). Downregulated expression of this gene may therefore have downstream effects other than regulation of global proteolytic processes.
Genes for Muscle-Specific Cytoskeletal Proteins
Along with muscle atrophy, caloric restriction may also induce cytoskeletal remodeling. On the synthetic side, the genes for cytoskeletal components, desmin, sarcosin, troponin-C, myosin light chain 2, titin-cap (T-cap or telethonin), and the regulatory protein CSRP3 were all upregulated (Table 2). Oxidative muscle fibers are spared preferentially during nutritional sarcopenia (e.g., White et al., 2000). Upregulation of expression of these genes may not imply increased incorporation of these proteins into muscle, but rather relative loss of glycolytic fibers.
In contrast to other cytoskeletal genes, expression of the -actin and prefoldin genes was downregulated in this study. Also, genes for the regulatory transcripts phosphatidylinositol 4-kinase, the actin-associated 17ß estradiol dehyrogenase, and the LIM mineralization protein 2 showed decreased levels of expression in response to nutritional restriction. These data argue against coordinate regulation of gene expression of all cytoskeletal protein genes.
Genes for Proteins Involved in Intermediary Metabolism
Caloric restriction may increase metabolism of amino acids to glucose via gluconeogenesis. Gene expression levels consistent with this hypothesis are argininosuccinate synthetase and calmodulin 1. Moreover, caloric restriction may induce global changes in energy metabolism as has been shown in the muscle of rhesus monkeys (Kayo et al., 2001). Gene expression levels consistent with this include cytochrome c oxidase subunits I and II.
Caloric restriction may require muscle to neutralize or excrete greater quantities of toxic waste products and metabolites (Sreekumar et al., 2002). Gene expression patterns consistent with this would be glutathione S-transferase (GSTM3) and ferritin. If animals have been adapted to limited nutrition, such responses might not be expected (Kayo et al., 2001; Weindruch et al., 2001), presumably due to a lowered metabolic rate, and a lower steady state of toxic metabolic by-products.
Genes for Extracellular Matrix Proteins
Connective tissue turnover in muscle may be down-regulated by caloric restriction. Gene expression levels consistent with this hypothesis include collagens 1 and 2, procollagen type III, osteonectin, and fibronectin. The observation that fibronectin expression is downregulated along with the collagens is interesting, considering the proposed role of the hcKrox transcriptional factor, which has been suggested to have a broad regulatory role in extracellular matrix expression (Widom et al., 2001). The synthesis of collagens (I and III as well as others) in connective tissues and specifically by mesenchymal cells is known to be downregulated by nutritional restriction (Laurent, 1987; Krupsky et al., 1997). Although lung and cartilage are known to be exquisitely responsive to nutritional restriction, skeletal muscle and vascular smooth muscle cells have also been found to respond (Laurent, 1987; Spanheimer et al., 1991).
|
Summary |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSummaryLiterature Cited |
|---|
The second trend is one of modulated protein turnover, with some processes being upregulated and others downregulated, probably reflecting the specific remodeling events that lead to a muscle that is losing myofibrillar mass more rapidly than connective tissue mass, and yet maintaining its cellularity.
Finally, metabolic homeostasis mechanisms are implicated with the modulation of genes involved in altered AA nitrogen excretion and glycogen metabolism, a downregulation of energy metabolism, as well as an increase in transcripts involved in metabolic detoxification.
On top of these clear trends, the expression of several unexpected genes was modulated during the nutritional stress experienced by the animals in the LOW group. Among these were transcriptional activators and other genes with possible regulatory roles, as well as several genes of unknown function. Further characterization of these genes may reveal new insights into the regulation of tissue remodeling and protein turnover in livestock animals.
|
Footnotes |
|---|
2 Correspondence: 306 Carmody Rd. (phone: +61-7-3214-2392; fax: +61-7-3214-2900; e-mail: Tony.Reverter-Gomez{at}csiro.au).
Received for publication March 8, 2004. Accepted for publication October 13, 2004.
|
Literature Cited |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSummaryLiterature Cited |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. Ollier, C. Robert-Granie, L. Bernard, Y. Chilliard, and C. Leroux Mammary Transcriptome Analysis of Food-Deprived Lactating Goats Highlights Genes Involved in Milk Secretion and Programmed Cell Death J. Nutr., March 1, 2007; 137(3): 560 - 567. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. P. Mane, C. Vasquez-Robinet, A. A. Sioson, L. S. Heath, and R. Grene Early PLD{alpha}-mediated events in response to progressive drought stress in Arabidopsis: a transcriptome analysis J. Exp. Bot., January 1, 2007; 58(2): 241 - 252. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Salem, P. B. Kenney, C. E. Rexroad 3rd, and J. Yao Microarray gene expression analysis in atrophying rainbow trout muscle: a unique nonmammalian muscle degradation model Physiol Genomics, December 13, 2006; 28(1): 33 - 45. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Reverter, N. J. Hudson, Y. Wang, S.-H. Tan, W. Barris, K. A. Byrne, S. M. McWilliam, C. D. K. Bottema, A. Kister, P. L. Greenwood, et al. A gene coexpression network for bovine skeletal muscle inferred from microarray data Physiol Genomics, December 13, 2006; 28(1): 76 - 83. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. A. Lehnert, K. A. Byrne, A. Reverter, G. S. Nattrass, P. L. Greenwood, Y. H. Wang, N. J. Hudson, and G. S. Harper Gene expression profiling of bovine skeletal muscle in response to and during recovery from chronic and severe undernutrition J Anim Sci, December 1, 2006; 84(12): 3239 - 3250. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. M. Reecy, D. M. Spurlock, and C. H. Stahl Gene expression profiling: Insights into skeletal muscle growth and development J Anim Sci, April 1, 2006; 84(13_suppl): E150 - E. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Reverter, W. Barris, S. McWilliam, K. A. Byrne, Y. H. Wang, S. H. Tan, N. Hudson, and B. P. Dalrymple Validation of alternative methods of data normalization in gene co-expression studies Bioinformatics, April 1, 2005; 21(7): 1112 - 1120. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||
= Cluster 1; + = Cluster 2; 
= Cluster 3).