Transcriptional profiling identifies extensive down regulation of ...

Articles in PresS. Physiol Genomics (July 29, 2003). 10.1152/physiolgenomics.00040.2003

Transcriptional profiling identifies extensive down regulation of extracellular matrix gene expression in sarcopenic rat soleus muscle J. Scott Pattison1, Lillian C. Folk 2, Richard W. Madsen 3, Thomas E. Childs1, and Frank W. Booth1 1

Departments of Biomedical Sciences and of Pharmacology and Physiology, and the Dalton Cardiovascular Institute University of Missouri at Columbia, Columbia, MO 65211; 2Department of Veterinary Pathobiology, University of Missouri at Columbia, Columbia, MO 65211, and 3 Department of Statistics, University of Missouri at Columbia, Columbia, MO 65211. Address for correspondence: Frank W. Booth, Ph.D. University of Missouri Department of Biomedical Sciences E102 Vet. Med. Bldg. 1600 E. Rollins Columbia MO 65211 Phone: 573 882 6652 email: [email protected]

Running head: Gene expression changes in growing and sarcopenic muscle

Copyright (c) 2003 by the American Physiological Society.

FINAL ACCEPTED VERSION PG-40-2003R1

2

ABSTRACT The direction of change in skeletal muscle mass differs between young and old individuals, growing in young animals and atrophying in old animals. The purpose of the experiment was to develop a statistically conservative list of genes whose expression differed significantly between young growing and old atrophying (sarcopenic) skeletal muscles, which may be contributing to physical frailty. Gene expression levels of >24,000 transcripts were determined in soleus muscle samples from young (3-4 months) and old (30-31 months) rats. Age-related differences were determined using a Student’s t-test ( of 0.05) with a Bonferroniadjustment, which yielded 682 probe sets that differed significantly between young (n=25) and old (n=20) animals. Of 347 total decreases in aged/sarcopenic muscle relative to young muscles, 199 were functionally identified; the major theme being that 24% had a biological role in the extracellular matrix and cell adhesion. Three themes were observed from 213 of the 335 total increases in sarcopenic muscles whose functions were documented in databases: 1) 14% are involved in immune response, 2) 9% play a role in proteolysis, ubiquitin-dependent degradation, and proteasome components, 3) 7% act in stress/antioxidant responses. A total of 270 differentially expressed genes and ESTs had unknown/unclear functions. By decreasing the sample sizes of young and old animals from 25x20 to 15x15, 10x10, and 5x5 observations; 682, 331, 73, and 3 statistically different mRNAs were observed, respectively. Use of large sample size and a Bonferroni multiple testing adjustment in combination yielded increased statistical power, while protecting against false positives. Finally, multiple mRNAs that differ between young growing and old, sarcopenic muscles were identified and may highlight new candidate mechanisms that regulate skeletal muscle mass during sarcopenia.

FINAL ACCEPTED VERSION PG-40-2003R1 Keywords: microarray, aged, atrophy, mRNA, statistics

3


4

The contribution of skeletal muscle strength and mass to health is under-recognized, where losses in association with advanced age result in an increased incidence of death (32). The peak mass of male skeletal muscle occurs by the age of 25 yrs in humans (30). Thereafter, approximately 10% of muscle mass is lost by the age of 50 yrs and another 30% is lost by 80 yrs of age (30). Thus, skeletal muscle transitions from a growth to an adult “steady state” to an atrophy phase with increasing age in both humans and rats (31). Because the molecular causes of sarcopenic skeletal muscle are poorly defined, the current experiment was designed to tease out contrasting mRNA levels between growing muscle in young rats and aged-atrophying (sarcopenic) muscle. The rationale for this study was to screen for gene targets in order to develop scientifically-based strategies to induce growth in sarcopenic muscle. The genes identified with differential expression for muscle growth that are present during normal adolescent muscle growth, but missing in old sarcopenic muscles, or vice versa, i.e., inhibitory factors that may be present in sarcopenia, but low in growth, were targeted as candidates for regulating muscle mass. The experimental strategy also included the decision to focus on a large sample size foregoing an adult group whose muscle mass is in a steady state. Greater importance was placed on a large sample size to obtain a larger number of significant differences with fewer false positives. Thus, the aim of the design was to identify factors promoting growth, rather than maturation, as potentially effective clinical interventions are needed when muscle mass decreases to the level associated with increased mortality. Use of microarrays allows a global, unbiased determination of mRNA expression and could provide an insight into the status of gene expression in skeletal muscle growth and sacropenia. Previous studies using Affymetrix microarrays to compare young and old skeletal muscles reported 113, 70, and 449 differentially


5

expressed mRNAs with independent observations of 3, 8, and 3, respectively (24,27,40). However, none of these reports provided valid statistical analyses to identify significant differences nor did they estimate the percentage of false positives in their lists of differentially expressed transcripts. Thus for the current experiment, it was reasoned that the use of larger sample sizes would allow greater statistical power so that conservative statistical adjustments could be made to minimize the presence of false positives.

The purpose of the current

experiment was to develop a statistically conservative list of genes whose expression differed significantly between young growing and old-sarcopenic muscles, which may be contributing to physical frailty. The hypothesis of the current study was that a subpopulation of growth factor mRNAs would be downregulated in old skeletal muscle, associated with old muscle no longer growing.

Materials and Methods Animals. Fischer 344 x Brown Norway Fl male rats obtained from the National Institute on Aging (Harlan Labs, Indianapolis, IN) and were sacrificed at the ages of 3-4-months (young, N = 25) and 30-31 months (old, N = 20). These ages were selected on the basis of previous data (8). They received regular rat chow and water ad libitum, were housed 2-3 per cage, and maintained on a 12:12 hour light-dark cycle. Prior to muscle extraction, animals were anesthetized with an intraperitoneal injection of a cocktail containing ketamine (49 mg/mL), xylazine (6.2 mg/mL), and acepromazine (2.0 mg/mL) at a concentration of 0.123 mL/100mg body weight. Soleus muscles were excised, weighed, snap-frozen in liquid nitrogen, and subsequently powdered using a mortar and pestle cooled by liquid nitrogen. Both soleus


6

muscles from a single rat formed one observation, where muscle RNA from a single animal was applied to an individual array. The University of Missouri Animal Care and Use Committee approved animal protocols. Sample processing for microarray analysis. Total RNA was isolated from an aliquot of muscle powder that was put directly into a TRIzol solution (Invitrogen) and homogenized on ice using a Polytron homogenizer (Kinematica) on setting 7 for three pulses 15 seconds each. The total RNA was further purified using RNeasy columns (Qiagen). Methods for sample preparation are described in detail in the Affymetrix Expression Analysis Technical Manual (Santa Clara, CA) and are briefly described next. Ten micrograms of purified total RNA was put into the cDNA synthesis reactions with a T7-(dT)24 primer (100 pmol/uL). First and second strand cDNA synthesis were carried out using components of the Superscript Choice kit (Invitrogen) with all incubations done in a Mastercycler Gradient thermocycler (Eppendorf). The amount of resulting double stranded cDNA was quantified using a PicoGreen kit (Molecular Probes). One microgram of cDNA was added to the in vitro transcription reaction utilizing biotinylated nucleotides provided in the BioArray High Yield RNA Transcript Labeling Kits (Enzo Diagnostics, Inc.). The resulting cRNA was further purified using RNeasy columns (Qiagen). The purified biotinylated cRNA was then fragmented and subsequently hybridized to Affymetrix rat genome U34A, B, and C arrays and analyzed by fluorescent intensity scanning according to Affymetrix protocols (Affymetrix Expression Analysis Technical Manual). The hybridization and scanning of the arrays was performed in the University of Missouri DNA Core Facility (Columbia, MO). U34 micoarrays. The U34 arrays were created in 1998. At that time the array set was


7

estimated to contain ~7000 full-length/annotated genes based on Unigene Build 34, which were all located on the A chip. Similarly, the remaining probe sets assayed for >17,000 ESTs, predominantly located on the B and C arrays. However, since that time many advances have been made to allow a considerable number of the ESTs to be identified. The human genome has been completed. The mouse genome is now available as a draft that is 95% complete. The rat genome project has begun and is projected to be finished before the year’s end. Now in 2002, many of the ESTs previously unassociated with known genes now are sufficiently homologous, such that “as of Unigene build 99 over 28,000 well substantiated genes exist” (Affymetrix technical datasheet rat230). Thus, many of the ESTs available on the B and C arrays can now be substantiated as ‘true genes’ based on their significant homologies to the known genes identified in the mouse and human genomes. The U34 array set assays ~24,000 genes and ESTs, represented by 26,379 probe sets. Thus, some mRNAs are assayed by multiple probe sets. Also, some ESTs have since been identified as portions of full length/known genes, where the ESTs are known to be part of same Unigene cluster as the full-length mRNAs causing multiple probe sets to assay the abundance of a single mRNA. GeneChip® analysis. Each probe set consisted of sixteen perfectly matched (complementary) 25-mers, corresponding to different regions along the length of a transcript. Likewise, sixteen mismatched pairs (containing a single mutated base) that do not perfectly complement a mRNA’s sequence were used as a measure of non-specific background binding. The Microarray Suite 5.0 software (Affymetrix) was employed which uses a one-sided Wilcoxon’s signed rank test to calculate a p-value reflecting the significance of differences between the perfectly matched and mismatched probe pairs, based on their fluorescent


8

intensities. The resulting p-values were used as a qualitative assessment of the ability to detect a given transcript, where p-values 0.06 were called “Absent”. Only probe sets that were “Present or Marginal” in 60% of the samples for an experimental group were analyzed statistically, with 11,165 probe sets sufficiently detected in at least one of the two experimental groups: i.e., young or old soleus muscles. ~20% of the 682 significantly different probe sets contained probe sets with more than one “Absent” call within the 20-25 muscle samples composing an experimental group (Supplemental Table 1). Approximately 4%-9% of the 682 probe sets were consistently called absent in only one of the two experimental groups, implying turning “on” or “off” of mRNA expression. Microarray Suite 5.0 software (Affymetrix) uses statistically-based algorithms to determine transcript abundance based on fluorescent intensities (termed “signal”). The “signal” for each probe set was calculated as the one-step biweight estimate of the combined differences of all of the probe pairs in the probe set. We used the calculated signal value for all subsequent statistical analyses. The fold changes of 347 probe sets that increased in the older group, relative to the young were calculated as the mean old signal intensity/mean young signal intensity. The fold changes of the 335 probe sets that decreased in sarcopenic muscle were determined as the mean young signal intensity/mean old signal intensity. A fold change of a certain magnitude can be converted to a percentage decrease by [(Y-O)/ Y]x100%. Microarray data analyses have been criticized as being “quite elusive about measurement reproducibility” (9). However, Bakay et al. (2) have reported that experimental error among Affymetrix microarrays is not a significant source of unwanted variability in expression profiling experiments (r2 = 0.979). In our hands, duplicate arrays also had small, unwanted inter-chip


9

variability (r2 = 0.981). The samples included in the current analyses were not run as replicates (i.e., assayed only once by single array). In addition, a recent publication of a single human patient, where RNA was prepared from two distinct breast tumors, and placed on duplicate U95A GeneChips (four chips total) found a very low degree of experimental variability between microarrays (r2 = 0.995), and between the two tumors (r2 = 0.987) (38). Statistical methods. The Shapiro-Wilk test for normality was employed, to determine the normality distribution of the current data set. Deviations from normality were observed about 15% of the time, suggesting it was more appropriate to use parametric statistics. Thus, an unequal variance Student’s t-test ( = 0.05) was employed to compare the signal values of young and old soleus groups. Furthermore, a Bonferroni adjustment was applied to correct for the multiple Student’s t-tests performed on 11,165 probe sets that had been detected as present, i.e., having sufficient hybridization as identified by the Affymetrix Microarray Suite 5.0 software. Power analysis was done to determine what power could be obtained with sample sizes of 20. As rats were sacrificed in groups of five over a 40-day period, a one-way ANOVA was performed to determine if any one group differed from the other groups. Groups of young and old rats were analyzed separately. A recursive analysis was performed on 20 data sets of 5x5, 10x10, and 15x15 that had been randomly selected from the 25x20 sample. An additional multiple-testing adjustment called the False Discovery Rate (FDR) was employed to control the chance of making a Type I error. The FDR is a simple, sequential Bonferroni-type procedure that has been proven to control the false discovery rate for independent-test statistics with a substantial gain in power over the Bonferroni technique at the expense of increasing acceptance false positives (4). Thus, to maximize the number of significant findings, a false discovery rate adjusted p