RWD, ABO, and SKK conceived and designed the experiments. JMZ, RS, HV, EC, KMM, KGB, and ABO performed the experiments. JMZ, RS, HV, EC, KMM, KGB, and ABO analyzed the data. HV, RR, and RWD contributed reagents/materials/analysis tools. JMZ, RR, ABO, and SKK wrote the paper.
The authors have declared that no competing interests exist.
We analyzed expression of 81 normal muscle samples from humans of varying ages, and have identified a molecular profile for aging consisting of 250 age-regulated genes. This molecular profile correlates not only with chronological age but also with a measure of physiological age. We compared the transcriptional profile of muscle aging to previous transcriptional profiles of aging in the kidney and the brain, and found a common signature for aging in these diverse human tissues. The common aging signature consists of six genetic pathways; four pathways increase expression with age (genes in the extracellular matrix, genes involved in cell growth, genes encoding factors involved in complement activation, and genes encoding components of the cytosolic ribosome), while two pathways decrease expression with age (genes involved in chloride transport and genes encoding subunits of the mitochondrial electron transport chain). We also compared transcriptional profiles of aging in humans to those of the mouse and fly, and found that the electron transport chain pathway decreases expression with age in all three organisms, suggesting that this may be a public marker for aging across species.
Aging is a complex phenomenon characterized by the decay of biological function over time, eventually leading to death. High-throughput methods for examining changes in the expression of genes, such as DNA microarrays, have been successful in elucidating some of the genome-wide changes that occur with age in several human tissues. The authors profiled gene expression changes in the muscles of 81 individuals with ages spanning eight decades. They found 250 genes and 3 genetic pathways that displayed altered levels of expression in the elderly. The transcriptional profile of age-regulated genes was able to discern elderly patients with severe muscle aging from those that retained high levels of muscle function; that is, the gene expression profiles reflected physiological as well as chronological age. In order to find genetic changes that might affect most or all tissues during aging, the authors compared genome-wide profiles of aging in the muscle to those in the kidney and the brain, and found a common signature for aging shared among these three tissues consisting of six genetic pathways. One of these aging pathways (the electron transport chain pathway) is age regulated not only in humans but also in two model organisms (mice and flies), providing insights about shared age-related changes in animals with vastly different lifespans.
Aging is marked by the gradual decline of a multitude of physiological functions leading to an increasing probability of death. Some aging-related changes affect one's appearance, such as wrinkled skin, whereas others affect organ function, such as decreased kidney filtration rate and decreased muscular strength. At the molecular level, we are just beginning to assemble protein and gene expression changes that can be used as markers for aging. Rather than search for molecular aging markers by focusing on only one gene or pathway at a time, an attractive approach is to screen all genetic pathways in parallel for age-related changes by using full-genome oligonucleotide chips to search for gene expression changes in the elderly. A genome-wide transcriptional profile of aging may identify molecular markers of the aging process, and would provide insight into the molecular mechanisms that ultimately limit human lifespan.
Molecular markers of aging must reflect physiological function rather than simple chronological age because individuals age at different rates [
In this paper, we have performed a genome wide analysis of gene expression changes in the human skeletal muscle. As age increases, skeletal muscle degenerates, loses mass, loses total aerobic capacity, and becomes markedly weaker [
The extent to which age regulation of genetic pathways is specific to a particular tissue or common across many tissues is unknown. Age regulation of gene expression between the cortex and medulla regions of the human kidney was found to be highly correlated [
Another key issue is whether there are genetic pathways that are commonly age regulated in different species with vastly different lifespans, such as human, mouse, fly, and worm. Transcriptional profiles of aging have been performed on both skeletal muscle and brains in the mouse [
In this work, we present a transcriptional expression profile of 81 human skeletal muscle samples as a function of age. The symporter activity, sialyltransferase activity, and chloride transport pathways all decrease expression with age in human muscle. The age-regulated genes were found to be markers of physiological age, not just chronological age. By comparing our results on aging in muscle to previous transcriptional profiles of aging in the kidney and the brain, we found a common signature for aging across different human tissues consisting of six genetic pathways that showed common patterns of age regulation in all three tissues. Finally, by comparing the signature for aging in humans to transcriptional profiles of aging in mice, flies, and worms, we found that expression of the electron transport chain decreases with age in humans, mice, and flies, constituting a public signature for aging across species with extremely different lifespans.
In order to study the effects of aging in human muscle, we obtained 81 samples of human skeletal muscle from individuals spanning 16 to 89 y of age (
Patients Recruited by Age Group
We used a multiple regression technique on each gene to determine how its expression changes with age, as had been done previously for age regulation in the kidney (Materials and Methods) [
To identify individual genes showing strong age regulation, we examined the slope with respect to age for each gene, and identified 250 genes in which the slope was significantly positive or negative (
Rows correspond to individual genes, arranged in order from greatest increase in expression with age at top to greatest decrease in expression with age at bottom. Columns represent individual patients, from youngest at left to oldest at right. Ages of certain individuals are marked for reference. Scale represents log2 expression level (Exp). Genes discussed in the text are marked for reference. A navigable version of this figure showing identities of specific genes can be found at
We considered the possibility that some of the 250 genes might not be age regulated per se, but rather might appear to be age regulated because they are associated with a pathological condition that increases with age. For example, the incidence of diabetes is known to increase with age in the general human population [
With the exception of hypothyroidism, none of the medical factors showed a strong association with age, and so it is unlikely that these confounding factors would cause genes to appear to be age regulated (
We used two methods to test whether any of the factors affected the slope of gene expression with respect to age of the 250 age-regulated genes. First, we used a multiple regression model that included a fourth term representing the medical factor (such as hypothyroidism) in addition to age, sex, and anatomy. We then compared the aging coefficient using this new model with the one from the original model that did not include the term. If any of the 250 genes were regulated by the medical factor and not by age per se, we would expect marked differences in the aging coefficients generated by the two multiple regression models. None of the fourteen medical factors, including hypothyroidism, had a significant effect on age regulation (
In summary, we have generated a global profile of changes in gene expression during aging in human muscle (
The genetic functions of many of the 250 genes shown in
In addition to searching for age regulation one gene at a time, we also screened known genetic pathways for those showing an overall change with age. With this approach, age regulation for every gene in a pathway is combined to determine whether there is an overall regulation of the entire pathway. Screening for coordinated age regulation of genetic pathways increases the sensitivity of our analysis, as the combined effects of small regulation of many genes in a pathway can be significant. For example, in a previous study of type 2 diabetes, screening genetic pathways for changes in expression provided key insights that were not possible from analyzing genes individually [
We developed a variant of gene set enrichment analysis (GSEA) to determine whether a genetic pathway shows evidence for age regulation [
Our version of the GSEA algorithm scores a gene set according to how the genes in it show coordinated increase (or decrease) on average in response to increasing age. The increase is measured by a van der Waerden statistic. To judge whether a specific van der Waerden statistic is significant, we used bootstrap resampling. Each bootstrap sample was drawn by resampling the arrays and keeping the gene expression measurements linked with the age, sex, and anatomy variables. The 624 van der Waerden scores for the gene groups were recomputed for each of the 1,000 bootstrap samples. Six gene sets were found to have statistically significant van der Waerden scores (
Rows represent the symporter activity, sialyltransferase activity, and chloride transport gene sets. Columns correspond to individual genes within a given gene set. Scale represents the slope of the change in log2 expression level with age (
Age-Regulated Gene Sets in Muscle
Symporter genes (63 genes) and chloride transporters (35 genes) are necessary for transporting solutes during muscle contraction [
Some people age slowly and remain strong and fit in their 70s, whereas others age rapidly, becoming frail and susceptible to age-related disease. We wanted to determine whether the expression profile for the 250 aging-regulated genes correlated with physiological in addition to chronological aging. For example, patient V17 was 41 y old but expressed his age-regulated genes similarly to patients who were 10 to 20 y older, and we would like to determine whether this patient had poor muscle physiology for his age (
(A) Cross-section of histologically unremarkable deltoid muscle from a 48-y-old woman demonstrating relatively equivalent sizes of types I and II muscle fibers. Arrows denote fibers types as distinguished by enzyme histochemistry (cryosection, 200×, myosin ATPase at pH 9.4).
(B) Cross-section of deltoid muscle from an 88-y-old woman demonstrating selective atrophy of type II muscle fibers that stain darkly by ATPase enzyme histochemistry (cryosection, 200×, myosin ATPase at pH 9.4).
(C) Histograms showing a correlation between muscle physiology and gene expression for age-regulated genes. Top panel: for each of the 250 age-regulated genes, we calculated the partial correlation coefficients between the type II/type I muscle fiber diameter ratio and gene expression excluding age variation (
(D) Histogram showing the likelihood of finding 92 genes with |
A simple correlation of gene expression with muscle type ratio would not be sufficient for our purposes. Such a correlation could arise simply because the gene expression and muscle type ratio are both correlated with age. Accordingly, we employed partial correlations of gene expression with muscle type ratios after adjusting for the effect of chronological age. To do this, we regressed type II/type I muscle fiber diameter ratio on age, regressed gene expression on age, and finally correlated residuals from both regressions to obtain partial correlation coefficients. The partial correlations for the 250 age-related genes are shown in
If a gene correlates with muscle diameter ratio only because both it and muscle diameter are correlated with age, then the partial correlation described above should be close to zero. We found that a large number of the genes in our list had a statistically significant relationship with type II/type I ratio after adjusting for age. However, many of the genes not on our list were also related to type II/type I ratio adjusted for age. We were able to show that genes with large partial correlations were significantly overrepresented in our list of 250 age-regulated genes. We counted 92 of 250 age-related genes for which the (absolute) partial correlation was more than 0.2 (
Our result indicating that the 250 age-regulated genes are enriched for genes regulated by type II/type I muscle fiber diameter ratio is valid even when we use other selection thresholds for muscle physiology (i.e., other than the absolute of
In summary, these statistical tests show that the set of age-regulated genes are markers of the relative level of muscle function, even among patients that are similar in age. Our findings are further supported by two additional statistical tests described in Materials and Methods (
Some aspects of aging affect only specific tissues; examples include progressive weakness of muscle, declining synaptic function in the brain, or decreased filtration rate in the kidney. Other aspects of aging occur in all cells regardless of their tissue type, such as the accumulation of oxidative damage from the mitochondria, DNA damage, and protein damage. Our genome-wide search for gene expression changes during aging would include both types of expression changes, and it would be interesting to discern which expression changes are muscle specific and which are common to all tissues. Expression profiles that are common to aging in all tissues would provide insight into the core mechanisms that underlie cellular aging. Therefore, we compared the DNA chip expression data from our studies on muscle aging to previous DNA chip expression studies on aging in the brain and the kidney. Rodwell et al. have characterized gene expression changes with age in the cortex and the medulla of the kidney from 74 patients, and Lu et al. have examined gene expression changes in the frontal cortex of the brain from 30 patients [
Our initial attempt to compare transcriptional changes between tissues relied on a Venn analysis, in which we directly compared the overlap in the lists of the age-regulated genes from the three tissues. Next, we searched for a common aging signature by comparing the Pearson correlation of age regulation between two tissues. Both of these straightforward methods showed only borderline statistical evidence for similarities in aging between the three tissues (Materials and Methods), but neither is expected to be powerful. Ultimately, we compared tissues using a grouped gene analysis. Grouping genes can be more powerful if there are small but consistent effects in each of a number of genes. Furthermore, the specific biological processes associated with each genetic pathway provide insights into mechanisms of aging. We used the modified GSEA described above to analyze previously published data on age regulation in the kidney and the brain [
Age Regulation of Gene Sets in Three Human Tissues
Increased overall expression of the extracellular matrix gene set (152 genes) with advancing age may contribute to widespread fibrosis in the elderly (
Shown are expression data from sets of extracellular matrix genes, cell growth genes, complement activation genes, cytosolic ribosomal genes, chloride transport genes, and electron transport chain genes. Rows are human tissues (M, muscle; K, kidney; B, brain). Columns correspond to individual genes in each gene set. Scale represents the slope of the change in log2 expression level with age
The cell growth gene set (29 genes) includes genes coding for growth factors, such as
Although complement activation genes (22 genes) are induced in muscle, the kidney, and the brain, they are expressed primarily in liver [
Cytosolic ribosomal genes include 85 genes that show a general increase in expression with age in all three tissues. This result is interesting because the rate of protein synthesis is known to decrease in old age [
The chloride transport pathway is composed of 35 genes that show an overall decrease in expression with age in all three tissues. Ion transport of many types is important not only in the contraction of muscle [
The mitochondrial electron transport chain was found to show an overall decrease in expression with age. This group contains 95 genes, including genes associated with the NADH dehydrogenase family (complex I), succinate-coenzyme Q reductase (complex II), ubiquinone-cytochrome c reductase (complex III), cytochrome c oxidase (complex IV), H+-ATP synthase (complex V), and the uncoupling proteins. The finding that expression of genes involved in the electron transport chain decreases in old age supports the mitochondrial free-radical theory of aging [
The above results show that there is common age regulation for these six genetic pathways in the kidney, muscle, and the brain. Next, we determined that there was little statistical evidence for the correlation of age regulation of individual genes in a pathway in one tissue with their age regulation in another tissue (Materials and Methods). Thus, it is unclear whether or not the same genes or different genes within a pathway show age regulation between different tissues. For example, certain genes in the electron transport pathway might be age regulated in the kidney, whereas other electron transport genes might be age regulated in the muscle.
Having identified genetic pathways that are commonly age regulated in different human tissues, we next determined whether their age regulation is specific for humans (private) or whether these groups are also age regulated in other species (public). Genetic pathways that are age regulated in different species would be of particular interest because they would identify mechanisms that are inextricably related to aging, even in animals that have vastly different lifespans.
We compared age regulation in humans to previously published studies of age regulation in
We first identified orthologs of human genes in each of the other three species. Next, we determined the change in expression with respect to age for each gene in each species, using multiple regression techniques similar to the ones used for our studies of aging in human muscle (Material and Methods). We took the six gene sets shown to be aging-regulated in diverse human tissues, and then asked whether they also showed age regulation in any of the other three species. We analyzed the expression of each of the gene sets using modified GSEA to determine whether they showed an overall bias in expression with age in each species. Extracellular matrix genes, cell growth genes, complement activation genes, cytosolic ribosomal genes, and chloride transport genes did not show age regulation in other species.
The electron transport chain genes showed a consistent overall decrease in expression with age in humans, mice, and
Rows represent either human tissues or model organisms. Columns correspond to individual human genes and homologs to human genes defined by reciprocal best BLAST hits in other species. Scale represents the normalized slope of the change in log2 expression level with age (
Age Regulation of the Electron Transport Chain in Three Species
In this study, we have generated a high-resolution transcriptional profile of aging in the human muscle. Welle et al. have previously used DNA chips to profile expression changes during aging for the human muscle [
People age at different rates, especially with regard to muscular aging. Some remain fit and strong, whereas other become frail and weak when they are old. The transcriptional profile for aging in this study reflects the physiological age of the subjects, as measured by muscle diameter ratio, after making an adjustment for their chronological ages. Previous work on age regulation in the kidney also identified molecular markers that could predict the physiological age of the kidney [
Our results provide the some of the first evidence for a common signature of changes of gene expression in different human tissues. Specifically, we found similar patterns of age regulation for six biological pathways in the muscle, the kidney, and the brain. Previous studies found similar patterns of aging between different parts of the same tissue, but not between entirely different organs (i.e., age regulation was found to be similar between the cortex and medulla of the kidney [
Except for the complement activation gene set, the pathways that show common age regulation in diverse tissues also function in all cells. Changes in expression of these pathways in old age may lead to degeneration of not only core cellular functions (such as ion transport and energy production) but also to degeneration of tissue-specific functions (such as kidney filtration and synaptic signaling) that rely on housekeeping pathways. By identifying a common aging signature across tissues, we can now focus on aging pathways that are general instead of tissue-specific. The common aging signature reflects the age of diverse organs, whereas genes that are age regulated in just one tissue reflect the age of that tissue. Finally, treatments or therapies that alter expression of the four common age-regulated pathways might be expected to affect diverse tissues instead of a specific tissue, and may therefore have an overall effect on longevity.
Although some patterns of aging are similar between different human tissues, much of aging is tissue-specific. Decreases in expression of the sialyltransferases and symporter genes are changes specific to muscle, and do not appear to occur in either the kidney or the brain.
Nearly all of the age regulation that we found is specific to humans, and does not seem to occur in old mice, flies, or worms. Thus, much of age regulation in humans is species-specific (private) rather than universal for all animals (public). This result emphasizes the importance of studying aging in humans rather than model organisms with short lifespans in order to understand how people grow old.
Nevertheless, we did find one pathway that was age regulated in humans, mice, and flies. The electron transport chain gene pathway decreases expression with age in all three species. Previous studies found little or no similarity in age regulation between humans and mice [
In mammals, direct genetic tests of the functional relevance of reduced expression of the electron transport chain pathway on lifespan have not yet been reported. However, in
What types of upstream events might cause a decrease in expression of the electron transport chain pathway with age? Other mitochondrial pathways, such as the mitochondrial ribosome, do not show age regulation similar to the electron transport chain pathway. One potential cause of decreased expression of the electron transport chain pathway is that metabolism may slow in old age, resulting in reduced expression of the energy producing machinery of the cell. Another possibility is that oxidative damage to the proteins in the electron transport chain in the mitochondria may lead to reduced expression of the corresponding genes in the nucleus. The electron transport chain creates free radicals in the process of generating energy that would preferentially damage protein components of the electron transport chain [
It seems unlikely that common age regulation of the electron transport chain pathway is directly due to evolutionary conservation. Events in old age are unlikely to have a significant effect on fitness of a population because old animals (such as 3-y-old mice and 80-y-old people) are a small fraction of natural populations (except in recent human history). It could be that the electron transport chain is regulated during aging as an indirect consequence of regulation during development (antagonistic pleiotropy) [
It is interesting that the level of age regulation of the electron transport chain is nearly the same in each species, whereas lifespan varies greatly. Compared to humans, mice age 20- to 30-fold and flies age 400-fold more rapidly. Thus, the kinetics of the changes in gene expression for the electron transport chain genes precisely matches the difference in lifespan between species. This suggests that decreased expression of the electron transport chain pathway with age may be particularly informative as a marker of physiological aging.
The muscle samples were obtained from patient biopsies collected either during surgery or in an outpatient procedure, and the medical conditions associated with each biopsy are listed in
Frozen muscle samples were weighed (50–100 mg), cut into small pieces on dry ice, and then placed in 1 ml of TRIzol Reagent (Invitrogen, Carlsbad, California, United States). The tissue was homogenized using a PowerGen700 homogenizer (Fisher Scientific, Pittsburgh, Pennsylvania, United States), and the total RNA was isolated according to the TRIzol Reagent protocol.
A standard protocol designed by Affymetrix (Santa Clara, California, United States) for their HG-U133 2.0 Plus high-density oligonucleotide arrays was slightly modified by the Stanford Genome Technology Center (Stanford, California, United States), and all samples were processed in their facility (see Protocol S1). Eight micrograms of total RNA was used to synthesize cRNA for each sample, and 15 μg of cRNA was hybridized to each DNA chip. The samples were processed in random order with respect to age.
We used the DChip program [
Cross-sections of muscle cryosections were photographed at 200×, and the pictures were either measured digitally (diagnostic muscle biopsy samples, ATPase preparations) or printed (abdominal muscle samples, combined SDH-cytochrome
To determine the change in expression with age, we used a multiple regression model in which the change in expression with age takes into account the possibility that expression levels might differ in men versus women, or in abdominal muscle versus peripheral muscle. Specifically, we used the following multiple regression model:
where
For human brain, mouse kidney, and
For human kidney, we used the multiple regression model:
In
The reviewers suggested two additional methods to show that the age-regulated genes could serve as markers for physiological age. First, we showed that genes regulated by muscle physiology can also predict chronological age. We found genes that were significantly regulated by type II/type I muscle fiber diameter ratio using the multiple regression model:
Here,
Second, we repeated our age analysis taking into consideration the effect of type II/type I muscle fiber diameter ratio on age regulation. To do this, we used a four-term multiple regression model that includes terms for both age and type II/type I ratio:
Using
We used a permutation analysis to simulate the number of genes that would pass our cutoff by chance (
To examine whether pathological or pharmaceutical factors were confounding the analysis of age regulation in muscle, we performed unsupervised, average-linkage hierarchical clustering of the 81 muscle samples using the Cluster software [
GSEA [
The van der Waerden test conforms more closely to our interpretation of what it means for a group
To compute the van der Waerden test, we first find the rank r(j) for every gene j ∈
where
It is better to use resampling methods instead of the N
The original GSEA [
We adopted instead a bootstrap approach. We resampled the data and recomputed enrichment scores, obtaining a histogram roughly centered over the observed enrichment score. If the null value (zero) is far outside the resampled histogram, then the enrichment score is statistically significant. The bootstrap approach also preserves correlations among genes as well as correlations between genes and covariates.
The primary motivation for bootstrapping is the presence of covariates in our problems. Consider for example data with age, sex, and expression variables. If we permute the ages with respect to the expression data and repeat the regression, we have to decide whether the sex variable should be attached to the ages or to the expressions in the random permutation. Attaching sex to the age variables will leave us with simulated data sets in which females express Y chromosome genes as much as males. Because of such artifacts, this is not a suitable null distribution. Attaching a covariate to the expression variables is also problematic. Suppose that one of the covariates is somewhat correlated with age. The effect will be to increase the variance of the originally sampled age coefficient. In permutation samples where the covariate is attached to the expression data, it is resampled independently of age. Such independence reduces the variance of the age coefficient in the permutation data. The consequence is that the permutation-based histogram of age coefficients is then too narrow and false discoveries will result.
In the bootstrap approach we generated 1,000 sample datasets. In each sample dataset we mimicked the sampling process that gave rise to the data by resampling 81 subjects from the population of 81 subjects. The resampling keeps age, expression, and all covariates of any given subject together. Bootstrap sampling mimics the random process that generated the data.
We remark that both bootstrap and permutation sampling of the van der Waerden scores gave rise to Z scores that were nearly normally distributed, but not necessarily N(0,1) (unpublished data). In permutation sampling, the histogram of enrichment scores tended to have means near zero, but several groups had variances larger than 1.0. In bootstrap sampling, the variances often differed from 1 and the means were usually between zero and the original enrichment score.
The most direct way to compare aging in muscle, the kidney, and the brain is via a Venn analysis: we find which genes attain a stringent significance level for each tissue and judge whether the overlap is statistically significant according to a hypergeometric distribution. We did a pairwise comparison between each tissue to find genes that are aging-regulated in both sets. There are six aging-regulated genes in both the muscle and the kidney (
A more sensitive comparison can be based on correlating the age coefficients of genes in two tissues. We selected all genes that are age regulated in either of two tissues, plotted the age coefficient of each gene in one tissue versus that gene's coefficient in the other tissue, and computed the Pearson correlation
Because the genes are correlated we cannot use textbook formulas to judge the statistical significance of these Pearson scores. To get a
We also ran a bootstrap test of the tissue comparisons. In this test we resampled the microarray data with replacement 1,000 times. Each time we recomputed the correlations between age coefficients for genes in the kidney and muscle. In 1,000 trials we saw 39 in which the sample correlation was less than or equal to zero. After converting to a two-tailed test, this corresponds to a
To test for the correlation of gene ranks between tissues within those gene sets found to be commonly age regulated in the human, we used a two-tailed Spearman correlation method to first calculate a correlation coefficient for every pairwise combination of tissues (i.e., muscle–kidney, kidney–brain, muscle–brain) for that age-regulated gene set (e.g., extracellular matrix genes). In order to test for the significance of the calculated correlations, we used a permutation-based Monte Carlo method, randomizing the ranks for each gene and tissue in the gene set and recalculating Spearman correlations 1,000 times. We found that most of the correlations between tissues were not significant (
Each row denotes a medical or pharmaceutical factor. Age of patients is shown on the
(253 KB TIF)
(A) Coronary artery disease was included as an additional term in
(B–L) Similar to (A) for 11 other medical factors. (B) Coronary artery disease. (C) Colorectal cancer. (D) End-stage renal disease. (E) Hyperlipidemia. (F) Hypertension. (G) Hypothyroidism. (H) Pancreatic cancer. (I) Prostate cancer. (J) Radiotherapy. (K) Statins. (L) Villous adenoma.
(319 KB TIF)
Samples are clustered on the basis of 250 age-regulated genes in muscle, shown by the top dendrogram. Columns are individual muscle samples, marked by age of the patient. Top seven rows correspond to the expression of the first seven age-regulated genes. The diagram shows anatomical, medical, and pharmaceutical factors for each patient. Each row corresponds to one medical or pharmaceutical factor.
(1.2 MB TIF)
(2.1 MB XLS)
(59 KB XLS)
(63 KB XLS)
(45 KB XLS)
(16 KB XLS)
(27 KB XLS)
(32 KB XLS)
(14 KB XLS)
(101 KB XLS)
(982 KB XLS)
(15 KB XLS)
(15 KB XLS)
We are grateful to Drs. Jeffrey Norton, Joseph Presti, Andrew Shelton, and Mark Welton for their assistance in obtaining human muscle samples, and to Marilyn Masek for her expertise in muscle morphometry.
gene set enrichment analysis