Gene expression profiling of gastrocnemius of “minimuscle” mice

Abstract

Few studies have investigated heterogeneity of selection response in replicate lines subjected to equivalent selection. We developed four replicate lines of mice based on high levels of voluntary wheel running (high runner or HR lines) while also maintaining four nonselected control lines. This led to the unexpected discovery of the HR minimuscle (HR_mini) phenotype, recognized by a 50% reduction in hindlimb muscle mass, which became fixed in 1 of the four HR selected lines. Here, we report genome-wide expression profiling describing transcriptome differences between HR_normal and HR_mini medial gastrocnemius. Consistent with the known reduction of type IIB fibers in HR_mini, Myh4 gene expression was −8.82-fold less (P = 0.0001) in HR_mini, which was closely associated with differences in the “calcium signaling” canonical pathway, including structural genes (e.g., Mef2c, twofold greater in HR_mini, P = 0.0003) and myogenic factors (e.g., Myog, 3.8-fold greater in HR_mini, P = 0.0026) associated with slow-type myofibers. The gene that determines the HR_mini phenotype is known to reside in a 2.6335-Mb interval on mouse chromosome 11 and 7 genes (Myh10, Chrnb1, Acadvl, Senp3, Gabarap, Eif5a, and Clec10a) from this region were differentially expressed. Verification by real-time PCR confirmed 1.5-fold greater (P < 0.05) expression of very long chain acyl-CoA dehydrogenase (Acadvl) in HR_mini. Ten other genes associated with fatty acid metabolism were also upregulated in HR_mini, suggesting differences in the ability to metabolize fatty acids in HR_normal and HR_mini muscles. This work provides a resource for understanding differences in muscle phenotypes in populations exhibiting high running capacity.

artificial selection experiments allow the study of evolution in real time and may reveal genetic correlations with the trait under selection. For example, selection of mice for enhanced protein content produced a hypermuscular phenotype (42) that subsequently was found to be caused by a deletion in myostatin (40). We (37) initiated artificial selection for increased wheel-running behavior to investigate the evolution of locomotor behavior and physiological capacities for exercise in outbred house mice (Mus domesticus). After 10 generations of selection, individuals from the four replicate high-runner (HR) lines ran, on average, 70% more revolutions/day, and exhibited significantly greater (12%) maximal oxygen uptake (V̇o_2max) compared with mice from four nonselected control lines (38). The differential in daily wheel running eventually plateaued at approximately +200% compared with control lines, and at generation 49 the endurance capacity of HR lines during forced treadmill exercise was found to be 35% greater than control lines based on total distance run prior to exhaustion (30). Moreover, the endurance running capacity of HR mice (30) was significantly greater than reported in studies using direct genetic manipulations or pharmacological approaches to alter endurance performance (e.g., Refs. 17, 32, 43).

These comparisons indicate the efficacy of artificial selection as a tool to alter complex, polygenic performance traits (12), but relatively few studies have addressed the issue of heterogeneity in the response to selection (13). That is, when replicate lines are subjected to a selection regimen their response may be dissimilar particularly regarding correlated traits. Our HR mice were developed as four replicate lines each exhibiting elevated levels of voluntary running capacity, while four nonselected lines were maintained as controls. This approach led to the unexpected discovery of the HR minimuscle (HR_mini) phenotype, originally recognized by a 50% reduction in triceps surae muscle mass (11), as well as whole limb muscle mass (21) compared with HR normal-sized (HR_normal) mice. Caused by a Mendelian recessive allele, the HR_mini phenotype increased in frequency in two of the four selected lines, demonstrating that the selection regimen has favored the phenotype (11), and the HR_mini allele has become fixed in one selected line and remains polymorphic in another (Ref. 39 and unpublished results).

Total hind limb muscle mass in the HR_mini phenotype is approximately half that of control lines, but despite the smaller muscle mass, there is no statistical difference in total number of wheel revolutions completed per day between HR_mini and HR_normal mice (21). The enhanced whole animal aerobic exercise capacity (35, 38) of HR_mini mice is in part underpinned by greater muscle capillarity (45), doubled mass-specific content of some metabolic enzymes (14), and pronounced loss of myofibers expressing myosin heavy chain (MyHC) IIb (15). The selective reduction of IIb myofibers, which have relatively large cross-sectional areas, may be largely responsible for the diminutive size of HR_mini muscles (14, 15) and likely underpins the observed differences in muscle performance. For example, medial gastrocnemius of HR_mini animals produces approximately half the mass-specific isotonic power of HR_normal, and fatigue resistance is significantly enhanced (39).

Diversity in response to selection is useful for testing hypotheses about mechanisms of correlated evolution, “trade-offs,” and “constraints.” Importantly, genetic correlations can differ from expectations based on first-principles knowledge of physiological mechanisms. The gene that determines the HR_mini phenotype is known to behave as a Mendelian recessive and to reside in a 2.6335-Mb interval on mouse chromosome 11 (MMU11) (20). This region harbors ∼100 expressed or predicted genes, many with known roles in muscle development and function, and direct sequencing of this region is currently in progress. Microarray analysis of RNA transcripts is a powerful tool for determining potential associations between gene expression and phenotypic outcomes. We used microarray of HR_mini and HR_normal to gain a better understanding of the differences and similarities in these two alternate solutions to selection on high running capacity. In addition, differentially expressed genes from the area showing linkage were prioritized to provide insight to the potential causal gene.

METHODS

Animal model.

Experimental procedures were approved by the University of California Riverside Institutional Animal Care and Use Committee.

Male mice were used from an animal model of four closed lines selectively bred for high voluntary wheel-running behavior (HR) and four control lines bred without intentional selection, described in Swallow et al. (37). Animals used in the current work were sampled from generation 37 at a mean age of 84 days (range = 79–86). To determine muscle masses, we sampled HR line #3 (lab designation), which is fixed for the minimuscle phenotype (all individuals express this phenotype) and is here termed HR_mini, as well as two other HR lines (line #7 and #8) in which the minimuscle phenotype has never been observed (here, termed HR_normal) (11). Gastrocnemius, plantaris, and soleus from both left and right legs were dissected, and masses (wet weight) of HR_normal and HR_mini muscles were compared by analysis of covariance using body mass as the covariate. For microarray analysis, we used gastrocnemius muscles (including both lateral and medial portions) from the mice described above for HR_mini line #3 and HR_normal line #8 (n = 6, in each group). To obtain sufficient RNA, gastrocnemius samples of HR_mini were pooled from left and right limbs from each animal, whereas HR_normal gastrocnemius was sampled from either left or right limb only.

Microarray analysis.

Total RNA was extracted from whole gastrocnemius muscle (>100 mg) homogenized in TRIzol reagent (Life Technologies, Gaithersburg, MD) following the manufacturer's protocol. The RNA was purified using RNeasy purification columns (Qiagen, Germantown, MD), and then RNA integrity, purity and quantity were assessed by Bioanalyzer (Agilent Technologies, Santa Carla, CA) and Nanodrop (NanoDrop Technologies, Baltimore, MD) spectrometry. The expression profiling procedure was performed following the standard Affymetrix protocol. Briefly, total RNA from each sample (5 μg) was used for cDNA synthesis followed by cRNA synthesis. Biotin-labeled and fragmented cRNAs were hybridized to the Mouse Genome 430 2.0 Array (Affymetrix, Santa Clara, CA) followed by washing, staining and image scanning per manufacturer's protocol. All profiles have been made publically accessible via the National Center for Biotechnology Information Gene Expression Omnibus (GSE41910) at http://wwwncbi.nlm.nih.gov/geo/.

Absolute analyses of the array images were performed using Affymetrix MAS 5.0. Satistically significant expression changes between HR_mini and HR_normal mice were determined using GeneSpring version 7. The expression level of each probe set was imported in to GeneSpring for probe-set filtering (10% present calls), and statistical analysis was performed using Welch t-test (corrected by Benjamini and Hochberg multiple testing with 5% false positive rate). Genes that showed significant changes (P < 0.05 after correction) were retained for functional enrichment analyses.

Functional enrichment analyses of differentially (P < 0.05, corrected) expressed genes was performed by two different methods. First, clustering of genes in to functional groups and the statistical significance of overrepresentation among the groups was calculated with the Database for Annotation, Visualization and Integrated Discovery [DAVID; http://david.abcc.ncifcrf.gov/home.jsp, (22)]. Genes were clustered according to their association with pathways of the Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.genome.jp/kegg/), and the statistical significance of clusters was estimated by modified Fisher exact P value. Second, differentially expressed genes were uploaded to Ingenuity Pathways Analysis to identify overrepresented canonical pathways from the Ingenuity Knowledge Base. To examine associations that are considered to occur specifically in skeletal muscles, we narrowed the Ingenuity Knowledge Base database to interactions reported in skeletal muscle only.

Quantitative real-time RT-PCR.

Quantitative real-time RT-PCR was used to verify the difference in acyl-coenzyme A dehydrogenase, very long chain (Acadvl) expression highlighted by microarray analysis. Gastrocnemius muscles (n = 6, in each group) were sampled from HR_normal and HR_mini mice (mean age 108 days, range 100–114 days) from generation 62 of selection. Total RNA (1 μg) isolated from gastrocnemius muscles were treated with DNase I before cDNA synthesis. cDNA were synthesized using superscript II RT (200 U/μl) with oligo-DT primers (1,000 pmol/μl). The primers used for Acadvl were 5′ GCGGACGGCGCCATTGACCTCTATG-3′ (forward) and 5′-ATCCGAGTTGCAGCCTCAATGCA-3′ (reverse). Ribosomal RNA 18S was used as the internal control. Primers used for 18S rRNA were 5′-TAGCCTTCGCCATCACTGCCATTA-3′ (forward) and 5′-AACCTGGCTGTACTTCCCATCCTT-3′ (reverse). The ABI 7900HT system (Applied Biosystem) was used for mRNA quantification as previously described (6). Briefly, SYBR Green PCR Master Mix (Applied Biosystems) was used. Samples were amplified in triplicate with cycling conditions of 50°C for 2 min, 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. All primers were tested for nonspecific amplicons and primer dimers by visualizing PCR products on 2% agarose gels prior to performing qPCR and disassociation curve after the reactions. Delta delta CT method was used to determine the expression changes of the Acadvl and an independent t-test performed to determine statistical significance of the differences.

RESULTS

Consistent with our previous findings, gastrocnemius mass was significantly less in HR_mini. Muscle mass was related (P = 0.0151) with body mass (Fig. 1), but no interactions (P > 0.07) were found between body mass and experiment group (HR_mini and HR_normal) for each of the muscles investigated. After adjusting for a mean body mass of 31.14 g, we found the mass of gastrocnemius (50 mg) in HR_mini to be 37.5% of the mass (133 mg) in HR_normal (P < 0.0001). In plantaris, the effect of body mass was also positive and significant (P = 0.0409), yet plantaris mass was significantly (P = 0.0084) less in HR_mini. After body mass adjustment, HR_mini plantaris (12.2 mg) was 84.9% of the mass of HR_normal (14.4 mg). Body mass was also significantly (P = 0.0039) related with soleus mass but in contrast to the fast-twitch muscles the mass of the slow-twitch soleus was significantly (P < 0.0001) greater in HR_mini. That is, after adjusting for body mass, soleus mass (11.1 mg) in HR_mini was 142% greater than HR_normal (7.8 mg).

Fig. 1.

Muscle mass in relation to body mass in high-runner normal-sized (HR_normal) and HR minimuscle (HR_mini) mice. Muscle mass (g) of gastrocnemius (top), plantaris (middle), and soleus (bottom) relative to body mass (g) of line #3 HR_mini (■), line #7 HR_normal (○), and line #8 HR_normal (●) mice. Data represent n = 6 mice in each group.

To identify genes differentially expressed between the two HR mouse lines, expression profiling was performed on six animals from line 8 (HR_normal) and six animals from line 3 (HR_mini). Of 45,307 probe sets on the Affymetrix mouse genome 430_2.0 chip, 56% of them (25,570 probe sets) detected target expression in at least one of the 12 arrays. These probe sets were further analyzed, and 2,650 transcripts were identified to be significantly different between HR_mini and HR_normal (P < 0.05 after multiple testing correction). Among them, 1,241 transcripts were downregulated in HR_mini, and 1,409 were upregulated. HR_mini exhibited particularly high expression levels (>20-fold greater expression) of keratin 8 and ankyrin repeat domain 1, and low (>20-fold lesser expression) levels of transmembrane protein 232, HtrA serine peptidase 4, serine/threonine kinase 33, and carbonic anhydrase 3 compared with HR_normal mice.

Functional annotation was conducted in DAVID to link genes with KEGG pathways and the top-ranking (P < 0.05) KEGG pathways are shown in Table 1. The most significant association was the KEGG pathway “focal adhesion,” which included 21 genes significantly less expressed in HR_mini, such as IGF-1, glycogen synthase kinase 3 beta, and mitogen-activated protein kinase 8; and 29 genes exhibiting greater expression in HR_mini, such as integrins, epidermal growth factor, and vascular endothelia growth factor. Other prominent associations included “regulation of actin cytoskeleton” and “ubiquitin-mediated proteolysis.” Expression of two-thirds of the genes (32 of 48) linked to regulation of actin cytoskeleton was greater in HR_mini, whereas the majority of genes (22 of 34) associated with the KEGG pathway ubiquitin-mediated proteolysis were significantly less expressed in HR_mini. Some associations, such as “hypertrophic/dilated cardiomyopathy” (Table 1), arise due to expression of specific isoforms of myofibrillar proteins, which are upregulated in failing hearts but are part of the normal expression pattern in the context of skeletal muscle. Similarly, associations to “pathways in cancer,” “glioma,” and “renal cell carcinoma” in Table 1 arise due to enrichment of genes associated with growth factor signaling pathways. Evidence exists to link the majority of genes listed within such pathways with the observed differences in muscle phenotype. However, because current ontology databases contain a proportionally greater emphasis on pathophysiological processes, particularly oncological research, this clustering of growth factor-related genes is misinterpreted as tumorigenesis. Therefore, to minimize such erroneous associations the data were further assessed using Ingenuity Pathway Analysis with the Ingenuity Knowledge Base restricted to skeletal muscle only. Under this constraint, significant (P < 0.05) enrichment was reported for 12 canonical pathways, and a cluster of four pathways (calcium signaling, glycolysis/gluconeogenesis, protein kinase A, and cAMP-mediated signaling) exhibited highly significant (P < 0.001) associations with the experiment dataset (Table 2). The validity of the microarray data is demonstrated by comparing known phenotypic differences between HR_mini and HR_normal with the most highly significant (P < 0.01) genes that were up- or downregulated in the four top-ranked canonical pathways.

Table 1.

Top-ranking KEGG pathways reported by DAVID

Name	Genes, n	BH Corrected P Value
Focal adhesion	50	0.000019
Regulation of actin cytoskeleton	48	0.00099
Ubiqutin-mediated proteolysis	34	0.0012
Hypertrophic cardiomyopathy	32	0.0062
Pathways in cancer	59	0.0098
Pyruvate metabolism	14	0.012
Glycolysis/gluconeogenesis	19	0.012
Dilated cardiomyopathy	22	0.03
Glioma	17	0.035
Renal cell carcinoma	18	0.033
Ribosome	21	0.035
MAPK signaling pathway	47	0.037

Top-ranking KEGG pathways reported by the DAVID functional annotation tool applied to all genes exhibiting significant (P < 0.05 corrected) differences in expression between normal high-runner (HR_normal) and minimuscle high-runner (HR_mini) mice.

Table 2.

Skeletal muscle genes associated with the top-ranking IPA canonical pathways

			Fold Difference
Pathway	Gene	Description	(HRmini)	Campbell
Calcium signaling
	Myl4	Myosin, light chain 4, alkali; atrial, embryonic	9.78
	Myl3	Myosin, light chain 3, alkali; ventricular, skeletal, slow	2.82
	Tpm2	Tropomyosin 2 (beta)	2.30
	Tnni1	Troponin I type 1 (skeletal, slow)	2.23
	Ppkar1a	Protein kinase, cAMP-dependent, regulatory, type I, alpha	2.05
	Mef2c	Myocyte enhancer factor 2C	2.05
	Myh7	Myosin, heavy chain 7, cardiac muscle, beta	1.97
	Tnnt1	Troponin T type 1 (skeletal, slow)	1.93
	Camk2d	Calcium/calmodulin-dependent protein kinase II delta	1.88
	Tpm3	Tropomyosin 3	1.67
	Myl9	Myosin, light chain 9, regulatory	1.62
	Myh4	Myosin, heavy chain 4, skeletal muscle	−8.82	−10.0
	Calm1	Calmodulin 1 (phosphorylase kinase, delta)	−3.08	−2.3
	Camk2a	Calcium/calmodulin-dependent protein kinase II alpha	−2.20
	Calm3	Calmodulin 3 (phosphorylase kinase, delta)	−2.06
	Calm2	Calmodulin 2 (phosphorylase kinase, delta)	−1.85	−1.6
	Tnni2	Troponin I type 2 (skeletal, fast)	−1.79
	Trdn	Triadin	−1.57
	Ppp3r1	Protein phosphatase 3 (formerly 2B), regulatory subunit B, alpha isoform	−1.53
	Asph	Aspartate beta-hydroxylase	−1.51
	Ppp3cb	Protein phosphatase 3, catalytic subunit, beta isozyme	−1.50	−6
	Atf4	Activating transcription factor 4 (tax-responsive enhancer element B67)	−1.35
	Tnnt3	Troponin T type 3 (skeletal, fast)	−1.22
	Ppkar2a	Protein kinase, cAMP-dependent, regulatory, type II, alpha	−1.21
Glycolysis/Gluconeogenesis
	Ldhb	Lactate dehydrogenase B	1.91	17.0
	Aldh1a1	Aldehyde dehydrogenase 1 family, member A1	1.83
	Aldh2	Aldehyde dehydrogenase 2 family (mitochondrial)	1.78
	Gpi	Glucose phosphate isomerase	1.67	−1.9
	Dlat	Dihydrolipoamide S-acetyltransferase	1.22
	Bpgm	2,3-bisphosphoglycerate mutase	−2.77
	Acyp1	Acylphosphatase 1, erythrocyte (common) type	−2.60
	Pgm1	Phosphoglucomutase 1	−1.97
	Gapdh	Glyceraldehyde-3-phosphate dehydrogenase	−1.31	−12
	Pgk1	Phosphoglycerate kinase 1	−1.30
	Pgam2	Phosphoglycerate mutase 2 (muscle)	−1.24
	Tpi1	Triosephosphate isomerase 1	−1.22	−2.3
Protein kinase A signaling
	Myl4	Myosin, light chain 4, alkali; atrial, embryonic	9.78
	Flnc	Filamin C, gamma	3.05
	Myl3	Myosin, light chain 3, alkali; ventricular, skeletal, slow	2.82
	Gys1	Glycogen synthase 1 (muscle)	2.57
	Tnni1	Troponin I type 1 (skeletal, slow)	2.23
	Prkar1a*	Protein kinase, cAMP-dependent, regulatory, type I, alpha	2.05
	Camk2d*	Calcium/calmodulin-dependent protein kinase II delta	1.88
	Pde8a*	Phosphodiesterase 8A	1.66
	Myl9	Myosin, light chain 9, regulatory	1.62
	Akap1	A kinase (PRKA) anchor protein 1	1.59
	H3f3c	H3 histone, family 3C	1.44
	Akap6	A kinase (PRKA) anchor protein 6	1.40
	Akap9	A kinase (PRKA) anchor protein (yotiao) 9	1.38
	Hist1 h1c	Histone cluster 1, H1c	1.34
	Add1	Adducin 1 (alpha)	1.27
	Myh4	Myosin, heavy chain 4, skeletal muscle	−8.82	−10
	Calm1	Calmodulin 1 (phosphorylase kinase, delta)	−3.08	−2.3
	Pde4b	Phosphodiesterase 4B, cAMP-specific (phosphodiesterase E4)	−2.79
	Camk2a	Calcium/calmodulin-dependent protein kinase II alpha	−2.20
	Calm3	Calmodulin 3 (phosphorylase kinase, delta)	−2.06
	Phkb	Phosphorylase kinase, beta	−1.92
	Calm2	Calmodulin 2 (phosphorylase kinase, delta)	−1.85	−1.6
	Tnni2	Troponin I type 2 (skeletal, fast)	−1.79
	Ppp1r14c	Protein phosphatase 1, regulatory (inhibitor) subunit 14C	−1.68
	Anapc5	Anaphase promoting complex subunit 5	−1.61
	Ppp3r1	Protein phosphatase 3 (formerly 2B), regulatory subunit B, alpha isoform	−1.53
	Ppp3cb	Protein phosphatase 3, catalytic subunit, beta isozyme	−1.50	−6
	Ywhae	Tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein	−1.47
	Gsk3b	Glycogen synthase kinase 3 beta	−1.41
	Atf4	Activating transcription factor 4 (tax-responsive enhancer element B67)	−1.35
	Gnas	GNAS complex locus	−1.27
	Mylpf	Myosin light chain, phosphorylatable, fast skeletal muscle	−1.26
	Ppkar2a	Protein kinase, cAMP-dependent, regulatory, type II, alpha	−1.21
cAMP-mediated signaling
	Rgs2	Regulator of G-protein signaling 2, 24 kDa	2.06
	Prkar1a	Protein kinase, cAMP-dependent, regulatory, type I, alpha	2.05
	Camk2d	Calcium/calmodulin-dependent protein kinase II delta	1.88
	Pde8a	Phosphodiesterase 8A	1.66
	Akap1	A kinase (PRKA) anchor protein 1	1.59
	Akap6	A kinase (PRKA) anchor protein 6	1.40
	Akap9	A kinase (PRKA) anchor protein (yotiao) 9	1.38
	Calm1	Calmodulin 1 (phosphorylase kinase, delta)	−3.08	−2.3
	Pde4b	Phosphodiesterase 4B, cAMP-specific (phosphodiesterase E4)	−2.79
	Camk2a	Calcium/calmodulin-dependent protein kinase II alpha	−2.20
	Calm3	Calmodulin 3 (phosphorylase kinase, delta)	−2.06
	Calm2	Calmodulin 2 (phosphorylase kinase, delta)	−1.85	−1.6
	Ppp3r1	Protein phosphatase 3 (formerly 2B), regulatory subunit B, alpha isoform	−1.53
	Ppp3cb	Protein phosphatase 3, catalytic subunit, beta isozyme	−1.50	−6
	Atf4	Activating transcription factor 4 (tax-responsive enhancer element B67)	−1.35
	Gnas	GNAS complex locus	−1.27
	Prkar2a	Protein kinase, cAMP-dependent, regulatory, type II, alpha	−1.21

Fold difference (HR_mini) represents statistically significant (P < 0.05 after correction for multiple testing) differences in expression in HR_mini compared with HR_normal gastrocnemius reported in the current work. Fold difference (Campbell) represents significant differences in expression in soleus compared with white gastrocnemius of ICR mice reported in Campbell et al. (5).

Of the 24 genes associated with “calcium signaling” 11 encode myofibrillar proteins. HR_mini are characterized by selective loss of type IIB fibers (15), and we found lesser expression of type IIb MyHC (Myh4; −8.82-fold, P = 0.0001), and attendant differences in phosphorylatable fast myosin light chain (Mylpf; −1.26-fold, P = 0.006) and fast isoforms of troponin I (Tnni2; −1.79-fold, P = 0.0003) and troponin T (Tnnt3; −1.22-fold, P = 0.004) in HR_mini. Conversely, expression of type I MyHC (Myh7; 1.97-fold, P = 0.002) and myofibril protein isoforms associated with slow-twitch muscle, including β-tropomyosin (Tpm2; 2.3-fold, P = 0.0001), slow α-tropomyosin (Tpm3; 1.67-fold, P = 0.0006), slow troponin I (Tnni1; 2.23-fold, P = 0.005), slow troponin T (Tnnt1; 1.93-fold, P = 0.0005), and ventricular/slow skeletal myosin light chain (Myl3; 2.82-fold, P = 0.006) was greater in HR_mini (Table 2).

The majority of nonmyofibrillar genes listed in calcium signaling were also associated with the protein kinase A canonical pathway. In addition, the calcium signaling network (Fig. 2) also included triadin (Trdn) and asparate β-hydroxylase (Asph), which are implicated in sarcoplasmic reticulum calcium handling, and myocyte enhancer factor 2C (Mef2c), which is associated with the expression of slow-twitch myofibril genes in adult skeletal muscle.

Fig. 2.

Differential expression of genes associated with calcium signaling. Prominent genes of the Ingenuity Pathway Analysis network “calcium signaling” are shown relative to their cellular location. The expression of genes colored red was significantly greater in HR_mini, whereas the expression of genes colored green was greater in HR_normal (P < 0.05, after correction for multiple testing).

Differential expression of genes associated with “glycolysis/gluconeogenesis” may also link with the low proportion of fast glycolytic fibers in HR_mini. Expression of the heart isoform of lactate dehydrogenase (Ldhb), which is more prominent in slow-oxidative muscle, was greater in HR_mini, but there were mixed differences in the expression of enzymes of the glycolytic pathway; phosphoglucomutase (−1.50-fold, P = 0.0006), triosephosphate isomerase (−1.22-fold, P = 0.004), and phosphoglycerate kinase (−1.30-fold, P = 0.006) were less expressed in HR_mini, whereas expression of hexokinase (1.61-fold, P = 0.001), glucose phosphate isomerase (1.67-fold, P = 0.001), and enolase (1.56-fold, P = 0.001) was greater in HR_mini.

Previous work mapped the region of the HR_mini gene to a 2.635-Mb interval on MMU11, which harbors ∼100 expressed or predicted genes (20); seven genes (Table 3) from this region were differentially expressed between HR_normal and HR_mini. Among these, very long-chain specific acyl-CoA dehydrogenase (Acadvl) expression exhibited the greatest statistical difference (P = 0.002) and was 1.60-fold greater in HR_mini. Furthermore, functional annotation of genes enriched in HR_mini highlighted the KEGG pathway “fatty acid metabolism,” which contained 11 genes (Table 4). To validate differential expression of Acadvl, we performed real-time quantitative RT-PCR and showed that the expression level of Acadvl was 1.5-fold (P < 0.05) greater in HR_mini compared with HR_normal mice (Fig. 3).

Table 3.

Differentially expressed genes from HR_mini region mapped to MU11

Gene	Affymetrix ID	Description	Fold Difference	P Value
Myh10	1452740_at	Myosin heavy chain 10, nonmuscle	1.92	0.043
Chrnb1	1420682_at	Cholinergic receptor, nicotinic, beta polypeptide 1 (muscle)	1.76	0.0108
Acadv1	1424184_at	Acyl-Coenzyme A dehydrogenase, very long chain	1.60	0.00242
Senp3	1448866_at	SUMO/sentrin specific protease 3	1.27	0.0343
Gabarap	1416937_at	Gamma-aminobutyric acid receptor-associated protein	1.15	0.0425
Eif5a	1451470_s_at	Eukaryotic translation initiation factor 5A	−1.27	0.0254
Clec10a	1419605_at	Macrophage galactose N-acetyl-galactosamine-specific lectin	−3.12	0.0205

Fold difference represents differences in expression in HR_mini compared with HR_normal. P values are from GeneSpring microarray analysis corrected by Benjamini and Hochberg multiple testing with 5% false positive rate.

Table 4.

Differentially expressed genes associated with KEGG pathway “fatty acid metabolism”

Gene	Affymetrix ID	Description	Delta	P Value
ACAA1A	1416947_s_at	Aacetyl-Coenzyme A acyltransferase 1A	1.57	0.0299
ACAA2	1428145_at, 1428146_s_at	Acetyl-Coenzyme A acyltransferase 2 (mitochondrial 3-oxoacyl-Coenzyme A thiolase)	1.68	0.0376
ACSL1	1423883_at, 1422526_at	Acyl-CoA synthetase long-chain family member 1	1.42	0.0146
ACADL	1448987_at	Acyl-Coenzyme A dehydrogenase, long-chain	1.84	0.00306
ACADS	1460216_at	Acyl-Coenzyme A dehydrogenase, short chain	1.67	0.01
ACADVL	1424184_at	Acyl-Coenzyme A dehydrogenase, very long chain	1.60	0.00242
ALDH2	1448143_at	Aldehyde dehydrogenase 2, mitochondrial	1.78	0.000667
ALDH7A1	1460167_at	Aldehyde dehydrogenase family 7, member A1	1.76	0.000674
HADH	1455972_x_at, 1460184_at	Hydroxyacyl-Coenzyme A dehydrogenase	1.41	0.0385
HADHA	1452173_at	hydroxyacyl-Coenzyme A dehydrogenase/3-ketoacyl-Coenzyme A-thiolase/enoyl-Coenzyme A hydratase (trifunctional protein), alpha subunit	1.74	0.00383
PECI	1431012_a_at	Peroxisomal delta3, delta2-enoyl-Coenzyme A isomerase	1.41	0.0472

Delta represents fold difference in expression in HR_mini compared with HR_normal. P values from GeneSpring microarray analysis corrected by Benjamini and Hochberg multiple testing with 5% false positive rate.

Fig. 3.

Acadvl expression in HR_normal and HR_mini muscle. Relative mRNA expression of Acadvl in HR_normal (line #8) and HR_mini (line #3) gastrocnemius muscle. Data are presented as means ± SD (n = 6, in each group). *Statistical significance (P < 0.05) of the difference in expression was determined by 2-tailed t-test for independent samples.

DISCUSSION

The current work provides comprehensive descriptions of two alternative muscle transcriptome “solutions” to a single type of selection. These models of artificial selection involve naturally occurring genetic variants (mutant alleles) rather than chemical or genetic manipulations that are often associated with deleterious correlated traits. HR_mini and HR_normal each exhibit enhanced physical capacity and as such might also occur, and be viable, in natural populations of house mice. In contrast, man-made genetic manipulations often induce phenotypes that are unlikely to be viable in nature. Despite their similar running capacities, gastrocnemius of adult HR_normal and HR_mini exhibited broad differences in gene expression. Genetic similarities in these two independent lines may encompass genes that are either essential or entirely unrelated to exercise capacity. Whereas gene networks exhibiting significant differences between HR_normal and HR_mini highlight aspects of genomic “flexibility” that have resulted in alternative genotypes capable of accomplishing the same task at the biochemical/ physiological level.

Previous work (20) mapped the region of the HR_mini gene to a 2.635-Mb interval on MMU11 that harbors ∼100 expressed or predicted genes. We discovered seven genes (Table 3) from this region that were differentially expressed between HR_normal and HR_mini. Published literature exists to potentially implicate each of these genes in the origin of the HR_mini phenotype. For example, nonmuscle myosin (Myh10) is associated with neurite growth cone motility (4) and hepatic mitochondrial DNA number (34). Small ubiquitin-like modifier (SUMO) demodification, mediated by Senp3, is specifically associated with regulation of ribosome biogenesis (16). Gabarap is a known atrophy gene (24) and is upregulated in cisplatin-induced atrophy of C2C12 myotubes (9). Eukaryotic initiation factor 5a (Eif5a) undergoes a unique posttranslational modification (hypusination) that correlates with embryonic myogenic differentiation, and inhibition of EIF5A hypusination impairs C2C12 differentiation (33). Clec10a (MGL1) regulates inflammatory processes, and MGL1^−/− mice are protected from glucose intolerance and insulin resistance associated with diet-induced obesity due to diminished adipose tissue macrophage response (44).

Two genes, Chrnb1 and Acadvl, from the 2.635-Mb interval on MMU11 (20) that exhibited the most robust statistical differences between HR_mini and HR_normal (Table 3) also link closely with the outcomes of the functional enrichment analyses. Chrnb1 encodes the beta subunit of the muscle acetylcholine receptor, and defects in this receptor are associated with fetal akinesia deformation sequence disorders (31). The greater expression of Chrnb1 in HR_mini gastrocnemius could be linked to the differences observed in Ca²⁺ signaling (Fig. 2). Acadvl encodes a mitochondrial enzyme that performs the first step of beta-oxidation of long-chain and very long-chain fatty acids. Acadvl deficiency is associated with mild cardiac hypertrophy (8) and altered muscle metabolism that protects Acadvl-deficient mice from high fat diet-induced obesity (46). HR_mini also exhibit cardiac hypertrophy (11, 41), but the expression of Acadvl was significantly greater in HR_mini than HR_normal muscle. Therefore, the difference in expression of Acadvl in HR_mini may be associated with the coincident changes in mitochondrial fatty acid metabolism (Table 4).

The concentration of mitochondrial enzymes, including citrate synthase, carnitine palmitoyltransferase 1, and cytochrome c oxidase, is approximately twofold greater in HR_mini than HR_normal gastrocnemius (14, 21). However, the size of triceps surae muscles in HR_mini is approximately half that of HR_normal, so, at the whole animal level, these factors counterbalance each other, and HR_normal and HR_mini muscles afford similar total energetic capacities for exercise (21). Nevertheless, differences in substrate utilization may exist, and whole animal fatty acid oxidation during submaximal exercise is significantly greater in HR_mini, which is associated with greater β-hydroxacyl-CoA dehydrogenase activity (41). The current data extend these findings and show the expression of 11 genes (including Hadh) encoding enzymes of the KEGG fatty acid metabolism pathway (Table 4) is greater in HR_mini gastrocnemius. In addition, genes associated with the transport of fatty acids to within muscle [solute carrier family 27 (fatty acid transporter member 1), Slc27a1; 1.84-fold, P = 0.0012] and mitochondria (mitochondrial carnitine/acylcarnitine translocase, Cpt1a; 1.76-fold, P = 0.0062) were also expressed at greater levels in HR_mini. These findings suggest HR_mini skeletal muscle has a greater capacity for fatty acid utilization, which may be consistent with the greater retroperitoneal fat pad mass of HR_mini animals (29).

The current genome-wide profiling data reveals the differences in the expression of fatty acid metabolism genes occur in the absence of widespread differences in regulators of mitochondrial biogenesis and genes encoding other mitochondrial enzymes. That is, few components of electron transport chain holoenzymes (e.g., cytochrome c oxidase, subunit VIIIa; 1.5-fold greater in HR_mini, P < 0.05) were differentially expressed in HR_mini, and the KEGG “oxidative phosphorylation” pathway was not significantly enriched in the current dataset. Peroxisome proliferator-activated receptor alpha (PPARα) is a key transcription factor associated with the expression of enzymes involved in fatty acid utilization (10), and Templeman et al. (41) reports muscle mRNA expression of PPARα, but not sirtuin 1, PPARβ/δ, or PPARγ coactivator-1α (PGC-1α), is significantly greater in HR_mini gastrocnemius. Six of the differentially expressed genes in the KEGG fatty acid metabolism pathway contain PPAR elements in their promoter regions, but in the current work Ppara expression was not statistically greater (1.8-fold, P = 0.41) in HR_mini. The promoter region of Acadvl contains regulatory elements for activating enhancer binding protein 2 alpha (AP-2α), AP-2α isoform 2, and AP-2α isoform 3 only. Ppara, and genes differentially expressed in HR_mini (Table 4) also have regulatory elements in their promoter regions for AP-2 isoforms, but these also were not differentially expressed. Therefore, the differences between HR_mini and HR_normal appear to be due to novel mechanisms discrete from current mainstream hypotheses regarding transcriptional regulation of muscle fatty acid metabolism.

Approximately 10% of the total number of genes detected in the current work were differentially expressed between HR_mini and HR_normal muscle. This proportion closely resembles analysis (5) comparing white gastrocnemius (100% IIb fibers) and soleus (70% type I, 30% type IIa fibers) from ICR mice, which have similar differences in type IIB myofiber content to HR_normal and HR_mini gastrocnemius (note that ICR mice formed the base population for the selection experiment that produced the HR_normal and HR_mini mice). Campbell et al. (5) report differences between white gastrocnemius and soleus in the expression of ∼12% of detectable genes including eight (Myh4, Calm1, Calm2, Ppp3cb, Ldhb, Gpi, Gapdh, and Tpi1; Table 2) that are also differentially expressed between gastrocnemius of HR_mini and HR_normal. The differences in Myh4, Calm1, Calm2, and Tpi1 expression reported here are similar to those described by Campbell et al. (5) (Table 2), whereas the magnitude of difference in Ppp3cb, Ldhb, and Gapdh expression is relatively less than expected, and HR_mini exhibit elevated expression of Gpi, which contrasts with the expected lesser expression of Gpi in slow- compared with fast-twitch muscle.

MyHC IIb is the predominant MyHC isoform expressed in mouse gastrocnemius muscle (18), and homogenous knockout of MyHC IIb results in a significant decrease in the average total number of fibers in hind limb muscles, partial compensatory hypertrophy of IId fibers (1), and a shift to greater expression of IIx/d fibers, consistent with the nearest-neighbor rule (27). Allen et al. (1) comment “We did not see any discernible difference in locomotor behavior between wild-type and IIb null mice. The pattern of gait and posture was unchanged, and there were not any overt differences in the quantity of cage locomotion (data not shown).” Harrison et al. (19), on the other hand, report lower voluntary wheel running in IIb knockout mice and signs of pathology, including fiber degeneration/regeneration. Similar to MyHC IIb knockout mice, HR_mini gastrocnemius contains small fibers in the superficial areas of the muscle (15), which normally would be exclusively occupied by IIb fibers. Due to signs of pathology and expression of embMyHC in some regenerating fibers, it is thought that the small muscle phenotype of IIb null mice is caused by atrophy and degeneration of IIb fibers (1). Interestingly, myofibrillar genes upregulated in HR_mini include embryonic isoforms of MyHC and myosin light chain along with cardiac isoforms of MyHC and troponin T, which, typically, are only expressed in regenerating adult skeletal muscle (2, 3, 28).

Microarray analysis also detected differences in myogenic regulatory factors, including myogenin (3.83-fold, P = 0.0026), alongside robust (21.35-fold greater in HR_mini) difference in the expression of cardiac ankyrin repeat domain 1 (CARP), which is a member of the muscle ankyrin repeat transcription factor family (25). Previously we (6) reported elevated expression of CARP in skeletal muscle after eccentric exercise. However, more recent evidence (26) demonstrates CARP expression is also upregulated during muscle atrophy, including denervation-induced atrophy and various models of muscular dystrophy. Interestingly, overexpression of CARP in adult skeletal muscle does not significantly alter muscle size (26). Therefore, CARP may represent a molecular marker of muscle transformation rather than specific regulator of either atrophic or hypertrophic processes. Other prominent features of the microarray data can also be linked to differences in muscle size associated with catabolic and anabolic hormone signaling (36). Keratin 8 (24.9-fold upregulated in HR_mini) and Keratin 18 (6.4-fold upregulated in HR_mini) are major components in intermediate filaments of the cytoskeleton and modulate tumor necrosis factor-α signaling in simple epithelial cells (23), whereas the ATP-independent protease serine peptidase 4 (HtrA), which regulates the availability of insulin-like growth factors (IGF) by cleaving IGF-binding proteins (7), was 40.5-fold less expressed in HR_mini.

In summary, microarray analysis of adult HR_normal and HR_mini gastrocnemius revealed the significant decrease in type IIb MyHC content (Myh4 expression) reported in previous studies is associated with attendant differences in the expression of numerous “fast” and “slow” isoforms of myofibrillar genes. The co-occurrence of prominent differences in genes of the calcium, protein kinase A, and cAMP pathways implicate differences in calcium signaling (Fig. 2) and activation of the “slow gene program” via nuclear factor of activated T-cells, cytoplasmic calcineurin-dependent 2 (Nfatc2), and myocyte-specific enhancer factor 2c (Mef2c) as possible contributing mechanisms. However, this cross-sectional study is not able to determine whether these differences in gene expression represent causes or consequences of the minimuscle phenotype.

Genes exhibiting the greatest differences in expression between HR_normal and HR_mini point to differences in anabolic/catabolic signaling and indicate the processes of muscle remodeling are ongoing in adult HR_mini gastrocnemius, which is further supported by the significant association with the KEGG focal adhesion pathway. In addition, the microarray data extend recent findings regarding the enhanced utilization of fatty acids by HR_mini mice (41) and suggest this does not correspond with prototypical transcriptional regulation of muscle fatty acid metabolism. This finding may be related to greater expression of Acadvl, which resides in the 2.6 Mb interval of MMU11 highlighted by linkage mapping as the causal region of the HR_mini phenotype.

GRANTS

Supported by National Science Foundation Grant IOS-1121273 to T. Garland. Y-W. Chen was partially supported by National Institutes of Health (NIH) Grants 1RO1AR-052027 and 1R24HD-050846. S. N. Pandey was supported by NIH Grant 1RO1AR-052027.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: J.G.B., T.G.J., and Y.-W.C. conception and design of research; J.G.B., T.H.M., S.N.P., G.B.-M., and M.F.M. performed experiments; J.G.B., T.H.M., S.N.P., G.B.-M., M.F.M., A.M.B., and Y.-W.C. analyzed data; J.G.B., T.H.M., S.N.P., G.B.-M., M.F.M., A.M.B., T.G.J., and Y.-W.C. interpreted results of experiments; J.G.B., T.H.M., S.N.P., G.B.-M., M.F.M., T.G.J., and Y.-W.C. prepared figures; J.G.B., A.M.B., T.G.J., and Y.-W.C. drafted manuscript; J.G.B., T.H.M., A.M.B., T.G.J., and Y.-W.C. edited and revised manuscript; J.G.B., T.H.M., S.N.P., G.B.-M., M.F.M., A.M.B., T.G.J., and Y.-W.C. approved final version of manuscript.