The extraocular muscles (EOM) are anatomically and physiologically distinct from other skeletal muscles. EOM are preferentially affected in mitochondrial myopathies, but spared in Duchenne’s muscular dystrophy. The anatomical and pathophysiological properties of EOM have been attributed to their unique molecular makeup: an allotype. We used expression profiling to define molecular features of the EOM allotype. We found 346 differentially expressed genes in rat EOM compared with tibialis anterior, based on a twofold difference cutoff. Genes required for efficient, fatigue-resistant, oxidative metabolism were increased in EOM, whereas genes for glycogen metabolism were decreased. EOM also showed increased expression of genes related to structural components of EOM such as vessels, nerves, mitochondria, and neuromuscular junctions. Additionally, genes related to specialized functional roles of EOM such as the embryonic and EOM-specific myosin heavy chains and genes for muscle growth, development, and/or regeneration were increased. The EOM expression profile was validated using biochemical, structural, and molecular methods. Characterization of the EOM expression profile begins to define gene transcription patterns associated with the unique anatomical, metabolic, and pathophysiological properties of EOM.

  • extraocular muscle
  • allotype
  • expression profile
  • transcriptome
  • energy metabolism

the extraocular muscles (EOM) are a group of highly specialized skeletal muscles that are required to locate and precisely track objects by the visual system. EOM rapidly and accurately provide a variety of voluntary, saccadic, and reflex eye movements. The functional requirements for EOM are wide ranging, including both slow vestibulo-ocular and optokinetic eye reflexes, as well as rapid saccadic eye movements required for reorientation of the visual system to new targets. As such, the EOM are active for extended periods of time in a fatigue-resistant manner (3, 13, 21, 42). Although the nature and extent of functional diversity of the EOM are becoming increasingly clear, the mechanisms or adaptations used by the EOM to achieve this functional diversity remain largely unknown (15, 35, 42).

Muscles achieve functional diversity through adaptive expression of various metabolic, contractile, and structural proteins that enable the performance of specialized tasks. For example, the soleus muscle contracts slowly and can sustain large loads for long periods of time in its fatigue-resistant role as a postural (antigravity) muscle. Small intrinsic hand muscles are much more easily fatigued, but provide precise and rapid movements. Functional diversity of muscle groups has been studied in terms of differences in contractile (fast vs. slow twitch) and metabolic (oxidative vs. glycolytic) properties. Muscle fibers are classified as slow (type I) and fast (type II) with subclasses within type II (a, b, and x) to reflect metabolic subtypes (11, 13, 21). At the molecular level, differences in expression of myosin heavy chain (MyHC) correlate well with physiological properties such as the velocity of contraction (Vmax) and fatigue resistance for most muscle groups (39, 43, 46), EOM being an important exception. More recently, structural proteins interacting with the MyHC, such as the myosin light chains, have also been shown to contribute to contractile properties such as Vmax (7). Additionally, MyHC immunohistochemical staining is widely used to classify (fiber-type) muscle studied in a laboratory and diagnostic setting (11, 40). However, it is clear that the currently available information defines only a small part of the functional and molecular diversity of different muscle groups and types.

The EOM differ from other skeletal muscles at multiple levels (35). Embryologically, EOM develop from the prechordal mesoderm, whereas the connective tissue of the orbit is a neural crest derivative (45). The EOM are innervated by cranial nerves rather than by motor neurons originating from the spinal cord. Unlike other mammalian skeletal muscle, the motor units of EOM are extremely small. Some myofibers in the EOM are multiply innervated, and the nerves are capable of making en grappe synapses along the fiber, an organization that is normally not found in any adult mammalian muscle (31). The EOM can be stimulated to generate individual twitch contractions at extremely high frequencies (∼400 Hz). This is in contrast to limb muscle, where stimulation at similar frequencies causes individual twitch contractions to fuse together to generate a tetanus (1, 3). The MyHC expression pattern in EOM is significantly different from that seen in limb muscle and includes expression of an EOM-specific MyHC as well as a continued expression of the embryonic MyHC in adult tissue (6, 18, 34, 41, 47). Additionally, the same myofiber can co-express multiple MyHC isoforms. The EOM phenotype is sufficiently different from other skeletal muscle that it has been referred to as the EOM “allotype” (24).

From a disease perspective, EOM are preferentially involved in specific diseases, such as myasthenia gravis, some mitochondrial myopathies, and oculopharyngeal muscular dystrophy (11). In contrast, the EOM are spared in Duchenne’s muscular dystrophy (DMD), despite the severe damage seen in other skeletal muscles in this disease (19, 22). The sparing occurs across species and is evident at a histological level in dystrophin-deficient humans, dogs, and mice (22). Although there is some selectivity of muscle involvement in early stages of DMD (proximal vs. distal musculature), this selectivity is usually lost in the course of DMD; however, the EOM remain functionally and anatomically spared even until death of patients with DMD (22).

It is generally accepted that the EOM are inherently different in terms of their molecular makeup and that the complement of genes expressed in EOM contributes to their unique pathophysiological phenotype. However, progress in defining the EOM phenotype has been slow due to their location, lack of biopsy material, and small size. Thus currently relatively few differentially expressed genes (e.g., EOM-specific MyHC, eom 1–4 genes, certain ryanodine receptor isoforms) have been identified in EOM compared with other skeletal muscle (2, 30, 47). Here, we describe expression profiling of EOM using Affymetrix (GeneChips) U34A high-density oligonucleotide microarrays (∼8,000 known rat genes), with verification of expression changes by immunohistochemistry, light microscopy, electron microscopy (EM), and biochemical and molecular methods.


RNA isolation.

RNA was extracted from muscles of adult Wistar rats that were dissected shortly after death by CO2 inhalation. All the rectus EOM were dissected from 10 rats (300 ± 25 g), rapidly frozen in liquid N2, and split into four pools (each pool ∼100 mg) for preparation of RNA. Four independent RNA preparations were made for EOM, and two individual preparations were made for tibialis anterior (TA). (The TA profiles are based on unpublished observations of Chen YW, Nader GA, Baar KR, Fedele MJ, Hoffman EP, and Esser KA.) Total RNA was extracted using the TRIzol reagent (GIBCO-BRL, Life Technologies, Gaithersburg, MD) and purified using the RNeasy kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. Briefly, frozen muscle tissue was pulverized in liquid nitrogen using a mortar and pestle and transferred into TRIzol reagent. The solution was homogenized by five aspirations through a 19-gauge needle followed by five aspirations through a 25-gauge needle. RNA was purified by phenol/chloroform extraction, precipitated using isopropyl alcohol and resuspended in diethyl pyrocarbonate (DEPC)-treated water. This RNA solution was repurified using the RNeasy kit and eluted with 40 μl of DEPC-treated water. An aliquot was used for quantification and verification of quality (260/280 ratio between 1.9–2.1) using a GeneQuant Pro spectrophotometer (Amersham Pharmacia Biotech, Uppsala, Sweden).

Affymetrix GeneChip expression profiling.

Affymetrix rat U34A series oligonucleotide-based arrays (GeneChips) were screened using guidelines provided by the manufacturer (Affymetrix, Santa Clara, CA) and optimized for expression profiling of skeletal muscle (8). Briefly, 5 μg of RNA was reverse transcribed and double-stranded cDNA was synthesized using the SuperScript Choice system (GIBCO-BRL) with a hybrid oligo-dT primer containing the T7 RNA polymerase promoter (Genset, San Diego, CA). The double-stranded cDNA was purified using phenol/chloroform extraction and precipitated with ethanol. This cDNA was resuspended in water and served as a template for an in vitro RNA transcription reaction using the ENZO BioArray High Yield RNA transcript (biotin) labeling kit (Enzo Diagnostics, Farmingdale, NY). An aliquot was quantified by spectrophotometry to calculate RNA yield and to ensure quality of in vitro transcription (in vitro transcribed RNA yield greater than 5× input). Biotin-labeled cRNA transcripts were purified using a RNeasy kit (Qiagen) and fragmented to about 200-bp size by incubating in 200 mM Tris acetate, pH 8.2, 500 mM KOAc, and 150 mM MgOAc at 94°C for 35 min prior to hybridization. Then, 15 μg of biotin-labeled cRNA samples were used for hybridization to U34A GeneChips for 16 h on an Affymetrix fluidics station 400. Subsequently, the GeneChips were washed and stained on the fluidics station to detect hybridized cRNA using phycoerythrin-conjugated streptavidin. The signal was amplified by a second round of staining using biotin-labeled anti-streptavidin antibodies, followed by staining with phycoerythrin-conjugated streptavidin. Fluorescence was read using a Hewlett-Packard model G2500A gene array scanner. Fluorescence data were analyzed using GeneChip software (Version 4). In brief, the intensity of one probe cell containing an oligonucleotide complementary to a specific sequence or perfect match (PM) was compared with the intensity of the adjacent probe cell which contains one mismatch (MM) in the center of the oligonucleotide. This analysis was repeated subsequently using multiple, different probe pairs for each gene (typically, the chips have 16–20 probe pairs per gene). The expression level of each gene was calculated based on a comparison of hybridization of the PM vs. MM signal. Iterative comparisons of different GeneChips that had been processed in parallel were conducted using four independent EOM and two TA hybridizations leading to eight data sets from six microarrays. Differential expression of genes with consistent fold changes (>2-fold) in EOM and TA was detected by comparing these eight independent data sets using previously described statistical methods that result in efficient detection of the most significant gene expression changes (8). Briefly, the determination of whether a gene was differentially expressed in EOM compared with TA depended on the following parameters; absolute (abs) call, fold change (FC), and difference (diff) call. In turn, the data for these parameters were derived using a decision matrix, which calculated analysis metrics (positive fraction, positive/negative ratio, and log average ratio) and was dependent on the probe array’s hybridization intensities. A probe set was defined as “increased” if it satisfied all of the following; abs call = P (present), FC ≥ 2, diff call = I (increased) or MI (marginally increased). A probe set was defined as “decreased” if it fulfilled all of the following; abs call = P (present), A (absent), or M (marginal) provided baseline ≠ A (absent), fold change ≤ 2, and diff call = D (decreased) or MD (marginally decreased). Probe sets that survived these criteria in all eight iterative comparisons were considered to have significant expression changes. The average values of these pair-wise comparisons are used throughout this study.

Semi-quantitative RT-PCR.

Five micrograms of total RNA from EOM and TA was reverse transcribed into cDNA using the Superscript Choice system (GIBCO-BRL). cDNA samples were purified using the GeneClean III kit (Bio 101, Carlsbad, CA). Ten percent (vol/vol) of purified cDNA was used as template for semi-quantitative PCR. Primers used for EOM-specific MyHC fragment were EOMMyHCF (5′-Gcgaattcttggatgttgagtatgcg-3′) and EOMMyHCR (5′ cgggatccagcangagctggangag-3′). For sm22 the primers were SM22F (5′-CACTGGGCAAAGATGACTGCAC-3′) and SM22R (5′ -CAGAAAAAGATGTAGAGGCAGGGTC-3′). For mbp, we used MBPF (5′ -AACACGCTGGAGATGCCATGTG-3′) and MBPR (5′ -CCGTGACCACCCTAAAGTGAGAAG-3′), and for nicotinic acetylcholine receptor (nAChR) α-subunit we used AChRAF (5′-CTCGGAGTCTTTATGCTGGTGTG-3′) and AChRAR (5′-GCGAAAGGATTCAAGTAGACAGGC-3′). As an internal control for efficiency of RT and quantification, we simultaneously amplified rat glyceraldehyde-3-phosphate dehydrogenase (GAPDH) using the primers RGAPDHF (5′-CCATGGAGAAGGCTGGGG-3′) and RGAPDHR (5′ -CAAAGTTGTCATGGATGACC-3′).

PCR was performed and aliquots removed at 22, 24, 26, 28, and 30 cycles to determine the linear amplification range of each gene. PCR products were resolved on 2% agarose gels and visualized using ethidium bromide staining on the Typhoon 8600 fluorescence imager (Amersham Pharmacia Biotech, Uppsala, Sweden). Quantification was performed on aliquots that lay within the linear amplification range for each reaction and normalized against GAPDH by using ImageQuant 1.1 software (Amersham) for the Mac OS. The entire experiment was repeated, including independent cDNA preparation, PCR, and quantification steps.

Histology, histochemistry, and EM.

Rats were killed using CO2 inhalation. The TA and EOM were dissected rapidly, then placed on thin cardboard discs using OCT in a fixed orientation to ensure identification of anatomical landmarks such as orbital and global surfaces as well as anterior and posterior ends of the muscles. Tissues were flash-frozen in liquid nitrogen-cooled isopentane and stored at − 80°C. Serial frozen sections (7–12 μm thickness) were cut with a Microm HM 500 cryostat (Zeiss, Oberkochen, Germany) and lifted using Superfrost Plus electrostatically charged slides (Menzel-Glaeser, Braunschweig, Germany). Tissue sections were subjected to gentle fixation using 100% ice-cold methanol for 5 min and stored in airtight containers at −80°C. Sections were thawed, preincubated in vehicle (1× PBS, 10% fetal bovine serum, 0.1% Triton X-100) at 4°C for 45 min, then incubated with primary antibody in vehicle at 4°C overnight [anti-EOM-specific MyHC, 1:200 (41); anti-laminin Sigma L-9393 at 1:1,000]. Sections were washed three times (for 10 min each time) in 1× PBS and incubated with Cy3-labeled (1:300) (Jackson Immunolabs, West Grove, PA) or Alexa 488-labeled (1:1,000) (Molecular Probes, Eugene, OR) secondary antibodies. Rhodamine-conjugated bungarotoxin (Molecular Probes) was applied on sections for 15 min at room temperature (1:250); the DNA-binding dye Hoechst 33258 (Molecular Probes) was applied for 5 min at a concentration of 1 μg/ml. After additional three washes in 1× PBS, sections were mounted in Gelvatol and visualized using epifluorescence illumination on an Olympus BX 51 microscope. Pictures were taken using a Magnafire CCD camera. Serial sections were processed for hematoxylin and eosin staining for histology. Digital images were prepared with Adobe Photoshop 6.0 (Adobe Systems, San Jose, CA).

For EM and other structural studies, animals were killed, and blood was removed by perfusion with a rinsing solution (15 mM EGTA in 15 mM phosphate buffer). Animals were fixed by whole body perfusion with 6% glutaraldehyde in phosphate buffer. Muscles were dissected and subjected to immersion fixation with 6% glutaraldehyde in 0.1 M cacodylate buffer for 1 h at room temperature. Tissues were postfixed with 0.8% potassium ferricyanide, 2% OsO4 in 0.1 M cacodylate buffer for 2 h, and stained en bloc in saturated uranyl acetate for 1 h. Specimens were dehydrated in ethanol and embedded in Epon. Ultrathin sections were stained with uranyl acetate and lead citrate and examined in a Philips E-410 electron microscope. Transected mitochondrial area was measured on electron micrographs of ultrathin sections from randomly chosen portions of the muscle. All fibers appearing within the grid holes were measured. The grid holes used for photographing were at approximately equal distances from each other. To avoid observer bias, a total of about 100 fibers were measured. Vascularization was determined using Nomarski optics on light microscopy of semithin sections of whole EOM and TA. For both parameters, the midportion of each rectus muscle of one eye and an undefined part of the TA was used. Reference areas were defined as myofibrillar occupied sarcoplasm. Images were digitized, montaged, and analyzed using the public domain NIH Image (version 1.63) image processing program.

Glycogen content measurement.

Two independent methods were used to measure glycogen content in frozen sections and whole muscle. Glycogen content of sections of EOM and TA was estimated using periodic acid-Schiff (PAS) reaction. Serial sections were pretreated with α-amylase enzyme to control for specificity of the glycogen staining procedure. To quantify glycogen content, samples were dissolved in KOH, and glycogen was precipitated with Na2SO4/methanol. The precipitate was treated with amyloglucosidase and glucose was measured using the Sigma 315-100 kit (Sigma Chemical, St. Louis, MO). A standard curve was generated by processing rabbit liver glycogen (Fluka Chemical, Milwaukee, WI) in parallel.


Expression profiling of EOM and TA muscle.

To identify the expression profile of EOM compared with limb muscle, we screened the Affymetrix U34A high-density oligonucleotide microarray (GeneChip) with RNA extracted from rat EOM and TA. The small size of the EOM precludes performing this type of global expression analysis on a single EOM; hence, we used pools of EOM to generate four independent samples. The pooling process also decreased the probability of intrasample variability. To verify that intrasample variability did not obscure differences between EOM and TA, as well as to determine the fold increase that we would consider as significant, we analyzed the expression profiles of the four EOM and two TA samples. Scatter graphs of expression levels of all probe sets represented on the microarray compared either among the two EOM samples (Fig. 1A) or two TA samples (Fig. 1B) show much less scatter (R2 = 0.93434 and R2 = 0.9641, respectively), compared with the scatter between EOM and TA (Fig. 1C; R2 = 0.7922).

Fig. 1.

Expression profiling data sets of extraocular muscles (EOM) and tibialis anterior (TA). The three scatter graphs represent probe sets with axes showing relative expression levels of each gene on a log scale. The graphs compare the expression profiling data sets for the following independent samples: EOM1 vs. EOM2 (A), TA1 vs. TA2 (B), and TA1 vs. EOM1 (C). The concordance (or lack of scatter) represent the experimental variability, which is minimal in A and B. C: comparison of the expression profile of limb and EOM, demonstrating considerably greater scatter. Solid lines indicate twofold difference cutoff. Slopes and correlation coefficients (R2) for the scatter graphs are as follows: y = 1.0042x + 5.8607 (A), R2 = 0.93434; y = 0.8764x + 84.748 (B), R2 = 0.9641; and y = 0.9616x + 81.014 (C), R2 = 0.7922.

To analyze and interpret the expression profile data, we performed iterative comparisons of the data sets that had been generated by expression profiling, as previously described (8). Multiple iterative comparisons (8 comparisons) of data sets were performed to control for intrasample variability (e.g., in cRNA hybridization) and to perform a stringent determination of the most significant and reproducible expression changes. Shown in Fig. 2A are eight iterative comparisons of difference calls of the independent RNA preparations and hybridizations (4 EOM preparations and 2 TA preparations). The number of surviving increase and decrease calls (or differentially expressed genes) declines in each sequential iterative comparison. Additionally, the number of increase or decrease calls is well stabilized by the eighth iterative comparison, suggesting that more iterative comparisons would be unlikely to achieve greater stringency. This method revealed 242 upregulated and 104 downregulated genes in EOM compared with TA, based on a twofold difference cutoff. These genes represent 2.75% and 1.18% of the total probe sets that were screened. Figure 2B shows a scattergram displaying the fold changes in gene expression of those genes differentially expressed between EOM vs. TA after imposing a ±2-fold cutoff on the data (i.e., deleting all genes from the graph that were not differentially expressed). Detection of embryonic and EOM-specific MyHC genes that have previously been shown to be differentially expressed in EOM (47) is indicated in Fig. 2B, demonstrating the specificity and sensitivity of our approach. The data are available as Supplementary Material1, published online at the Physiological Genomics web site. These data are also available at the Children’s National Medical Center Microarray web site (http://microarray.cnmcresearch.org/ microarray.html).

Fig. 2.

Stringent determination of differentially expressed genes in EOM vs. TA. Multiple iterative comparisons were performed between the difference calls of the EOM and TA expression profile data sets. A: effect of iterative comparisons on the eight data sets. The number of difference calls decreases in the bar graph and is stabilized by the eighth iterative comparison. The numbers of increase and decrease calls in first to eighth iterative comparison are as follows. First iterative comparison: 863 increase and 460 decrease calls. Second iterative comparison: 508 increase and 305 decrease calls. Third iterative comparison: 449 increase and 254 decrease calls. Fourth iterative comparison: 394 increase and 241 decrease calls. Fifth iterative comparison: 367 increase and 199 decrease calls. Sixth iterative comparison: 334 increase and 188 decrease calls. Seventh iterative comparison: 327 increase and 185 decrease calls. Eighth iterative comparison: 319 increase and 181 decrease calls. B: a scatter graph of the expression profile after the sequential, iterative comparisons between EOM and TA on a log scale. Individual points on the graph represent genes, which met the twofold difference cutoff (i.e., differentially expressed genes). Genes lying off the diagonal show the greatest differences in expression between EOM and TA as exemplified by the probe set detecting the gene for EOM-MyHC, which is upregulated in EOM. Based on the probe set design, cross-hybridization for embryonic, EOM-specific, and MyHC 1 is expected; all these chains have previously been shown to be coexpressed in EOM (47). The highly differential expression of EOM-specific MyHC suggested by this probe set was validated by RT-PCR (see Fig. 4). MyHC, myosin heavy chain.

Characterization of genes that are differentially expressed in EOM.

Of the 242 upregulated genes in EOM, 174 were previously described genes (representing 72% of upregulated genes; 1.98% of all probe sets screened) and 68 were expressed sequence tags (ESTs) (representing 28% of upregulated genes; 0.77% of all probe sets screened). In case of the 104 genes that were found to be downregulated in EOM, 73 were previously described genes (representing 70% of downregulated genes; 0.83% of all probe sets screened) while 31 were ESTs (representing 30% of downregulated genes; 0.35% of all probe sets screened). The EST set is predicted to contain a number of novel genes, however, these are yet to be fully cloned and characterized. Nonetheless, these genes help to accurately and comprehensively define the EOM expression profile and are available as Supplementary Material, published online at the Physiological Genomics web site.

We tabulated and sorted the known, differentially expressed genes into various clusters (exemplified in Table 1). Genes related to energy metabolism (14%) and genes encoding membrane proteins (14%) are the largest functional groups of all differentially expressed genes. Other clusters include genes of intracellular signaling (9.6%) and structural elements (7.2%), genes involved in maintenance of intracellular homeostasis (6.8%), and genes related to growth, development, and regeneration (6%). This data is consistent with the hypothesis that EOM has a unique molecular makeup or allotype (24) and is significantly different than limb muscle.

View this table:
Table 1.

Expression profile of EOM: functional clusters of selected genes

Validation of expression profiling: structural elements.

The well-studied, distinct anatomical features of EOM (e.g., higher vascular, nerve and mitochondrial content, larger number of synapses) offer an internal control for expression profiling experiments. Thus the results of a comprehensive expression profiling analysis would be predicted to demonstrate differences in genes that correlate with the anatomical features of EOM. The montages of EOM and TA in Fig. 3 show the well-recognizable small fiber size and orbital and global layers of EOM. The higher vascular and nerve content are also clearly visible and accurately reflect the expression of genes expressed in vessels (e.g., vascular α-actinin, RDC-1 protein, SM22, and angiotensin receptor) and in nerves (e.g., myelin basic protein, Schwann cell peripheral myelin, MAP2). The mRNA for the well-characterized EOM-specific MyHC (47) was found to be upregulated in EOM compared with limb muscle (Table 1). Consistent with this expression profile and previous reports, labeling with EOM-specific MyHC antibodies (25, 41) was restricted to the EOM (Fig. 4, A and B). Laminin immunohistochemistry shown in Fig. 4, C and D, demonstrates the greater intramuscular nerve content in EOM, which is consistent with the abundant mRNA of genes expressed in nerves. Double labeling using antibodies against laminin (green) and rhodamine-labeled bungarotoxin (red) demonstrates increased number of neuromuscular junctions (NMJ) in EOM compared with TA (Fig. 4, C and D). The increased synaptic content of EOM (31) is reflected by the increased expression of nAChR α- and δ-subunit genes in EOM (Table 1). Additionally, we verified the increased expression of these important structural components of EOM at the mRNA level by semi-quantitative RT-PCR analysis of EOM-MyHC, myelin basic protein, and nAChR α-subunit genes (Fig. 4, bottom). To validate the high expression of genes for mitochondrial enzymes (e.g., cytochrome C oxidase, stearoyl-CoA desaturase) we performed EM to estimate the mitochondrial content of rat EOM and TA (Fig. 5, A and B). Overall, the expression profile of genes expressed in EOM correlates well with what is known about EOM structure, suggesting that our expression profiling was sensitive and comprehensive.

Fig. 3.

Structural differences between EOM and TA. Semithin sections of EOM (left) and TA (right) were visualized and photographed using Nomarski interference contrast light microscopy. Photomontages of EOM and TA taken at low power (top: scale bar 200 μm) demonstrate the laminar organization, small fiber diameter, and larger intramuscular neural and vascular content of EOM compared with TA. Bottom: photomicrographs at high power (scale bar 20 μm) to demonstrate and quantify intramuscular vascular elements as exemplified by arrowheads (for sake of clarity, only some intramuscular vascular elements have been marked). The vascular density was about fivefold greater in EOM compared with TA.

Fig. 4.

Validation of structural differences between EOM and TA. Frozen sections (7–12 μm thickness) from flash-frozen EOM (left: A, C, and E) and TA (right: B, D, and F) were subjected to gentle fixation, processed for immunohistochemistry, visualized using epifluorescence illumination, and photographed. Significant labeling using antibodies specific to EOM-MyHC was noted in EOM (A), but not in TA (B). Increased intramuscular nerve content was noted using antibodies for the basal lamina constituent, laminin (green) in EOM (C) compared with TA (D). Double labeling using antibodies for laminin (green) and rhodamine-labeled bungarotoxin for nicotinic acetylcholine receptor (nAChR) (red) demonstrate the increased neural content as well as increased number of neuromuscular junctions (NMJs) in EOM (E) compared with TA (F). Serial sections were processed for hematoxylin and eosin staining for histology. Scale bars: A and B = 200 μm; C and D = 100 μm; E and F = 100 μm. Bottom: validation of relative increases of mRNA levels in EOM for genes associated with these structural components by semi-quantitative RT-PCR analysis. Myelin basic protein showed around eightfold increased expression (26 cycles), whereas the nAChR α-subunit gene showed around fivefold increased expression (26 cycles). The relative increase in the expression of EOM-MyHC is clearly evident, but fold increase could not be quantified, since at 30 cycles no product was detected in PCR from limb muscle. The experiments were performed in duplicate.

Fig. 5.

Validation of increased mitochondrial content of rat EOM by electron microscopy (EM). EOM and TA were processed, embedded, and analyzed by EM for mitochondrial content. A: a representative micrograph of EOM, demonstrating the increased mitochondrial content compared with TA (B). Mitochondria in the top third of the micrograph have been digitally filled in for ease of demonstration. The mitochondrial volume density was around threefold greater in EOM compared with TA. Scale bar = 2 μm.

Validation of expression profiling: metabolic pathways.

Numerous genes predominantly responsible for energy metabolism were differentially expressed between EOM and TA (Table 1). Taken together, the differential expression of the genes for phosphofructokinase C (PFK-C), lactate dehydrogenase-B (LDH-B), COX VIII-L, phosphoglucomutase, phosphorylase kinase, and muscle fructose-1,6-bisphosphatase suggested preferred utilization of blood glucose rather than glycogen-based metabolism in EOM relative to TA. To test the prediction that EOM had decreased glycogen content based on the expression profile, we analyzed the glycogen content in EOM and TA. Figure 6, A and B, shows PAS staining of frozen sections made from EOM and TA. The lack of PAS staining in EOM reflected the decreased glycogen content in EOM relative to TA. To ensure specificity of glycogen detection by PAS staining, serial sections were pretreated with α-amylase (data not shown). To verify the relative glycogen content in EOM as well as accurately measure the glycogen content of EOM and TA, we used a biochemical assay to measure the glycogen content from four independent EOM and TA samples. As shown in Fig. 6, bottom, the glycogen content of EOM was significantly lower compared with TA, providing further validation for the expression profile results (Table 1) and our predictions regarding EOM fuel utilization (see model in Fig. 7).

Fig. 6.

Differences in glycogen content in EOM and TA. Frozen sections (7–12 μm thickness) from EOM and TA were subjected to gentle fixation, semi-quantitative periodic acid-Schiff (PAS) reaction, and photographed to estimate glycogen content and distribution. Decreased glycogen content of EOM (top left) compared with TA (top right) is demonstrated as reflected by the reduced staining in EOM. To accurately quantify glycogen content, individual pools of EOM and TA were subjected to biochemical analysis for glycogen using the Can and Exton assay. As shown in the bar graph (bottom), the EOM glycogen content was 375 ng/mg tissue (open column), whereas TA contained 1,105 ng/mg tissue (gray column) (n = 4; *P < 0.05 by Student’s t-test; error bars = SE).

Fig. 7.

Model for energy metabolism in EOM and TA. In this model we propose the divergence of glucose utilization in EOM (A) vs. TA muscle (B). The EOM expression profile suggests a reduced glycogen metabolism and reliance on blood glucose-based, oxidative phosphorylation. Transcription of phosphoglucomutase (P-glucomutase), the key enzyme of glycogenesis, is reduced in EOM. Additionally, the activating enzyme for glycogen phosphorylase (G-phosphorylase) is downregulated in EOM, which suggests decreased activity of this rate-limiting enzyme in glycogen catabolism. Histochemical representation of glycogen reflects the decreased glycogen content that is predicted based on these differences in enzyme expression and/or activity. Increased expression of Phosphofructokinase-C (PFK-C) mRNA suggests efficient glucose utilization in EOM, since the PFK isoforms are the rate-limiting enzymes in glucose catabolism towards pyruvate. Increased levels of lactate dehydrogenase B (LDH-B) mRNA suggest that oxidative pathway is favored in EOM, rather than production of lactate. Additionally, TA muscle shows higher expression of pyruvate dehydrogenase kinase 4 (PDK4) mRNA. Since PDK4 is a potent inhibitor of pyruvate dehydrogenase complex (PDHC) activity, the increased expression would be predicted to result in lower PDHC activity in TA vs. EOM. PDHC catalyzes the conversion of pyruvate into acetyl-CoA, which then enters the citric acid cycle in the mitochondria and is used to generate ATP via oxidative phosphorylation. Overall, the changes would be predicted to result in high rates of aerobic, blood glucose-based metabolism in EOM. Blue color indicates decreased expression and/or activity of metabolic enzymes in EOM relative to TA. Red indicates increased expression and/or activity of metabolic enzymes in EOM relative to TA.


We used microarrays to define the expression profile of rat EOM and limb muscle to test the hypothesis that the molecular makeup of EOM is inherently different from limb muscle. Using differential display (DD) on rat EOM vs. gastrocnemius limb muscle, we had previously succeeded in defining 14 differentially expressed genes and cloned 4 novel genes (eom 1–4) expressed in EOM (30). However, the DD study was not comprehensive, since only about 100 clones were analyzed, compared with about 8,000 probe sets analyzed using microarrays. Nevertheless, it did provide clues toward aspects of the EOM structure that have also been expanded and validated by the current study, such as an increased expression of embryonic MyHC, increased mitochondrial content and decreased glycogen-based metabolism, compared with limb muscle. More recently, Porter et al. (36) used DNA microarrays to expression profile mouse EOM vs. a mixture of gastrocnemius and soleus muscle, representing limb muscle. In common with our study, they noted an upregulation of genes associated with nerve, vessels, NMJ, and glucose-based energy metabolism in EOM. The high degree of concordance toward previous studies and validation by structural, molecular, and biochemical means suggests that our expression profiling of EOM is a comprehensive study and provides meaningful mechanistic insights toward the functioning of this unique muscle.

Expression profiling: structural elements.

The expression profile obtained in this study was concordant with the known structural features of EOM. Increased expression levels of EOM-specific and embryonic MyHC mRNA in adult rat EOM are consistent with previous investigations that have studied these molecules at both the RNA and protein level (6, 41, 47). The differential upregulation of the EOM-specific and embryonic MyHC genes, as well as myosin regulatory light chain (RLC) in EOM may contribute to the unique mechanical (contractile) properties of this muscle group. The upregulation of genes expressed in vessels in the EOM expression profile provides mechanistic clues for the high vascular content in EOM (32, 48). Increased expression of genes such as myelin basic protein, Schwann cell peripheral myelin, as well as nAChR α- and δ -subunits correlate with the presence of large, multiply branched, myelinated nerves with rapid conduction velocities and abundant NMJs (3). Previous studies that analyzed nAChR are equivocal regarding the expression of the gamma (fetal) subunit of the nAChR in adult EOM (16, 20). We found no evidence for overall upregulation of the fetal subunit in EOM.

Expression profiling: features of efficient blood glucose-based, aerobic metabolism.

The expression profile suggests that EOM predominantly use aerobic metabolism of carbohydrates with reliance on blood glucose as an energy source. Support for this model comes from the decreased expression of phosphoglucomutase (a key glycogen synthesizing enzyme), phosphorylase kinase (a key regulator of glycogen breakdown or glycogenolysis), as well as a reduced glycogen content in EOM. Interestingly, deficiency of these enzymes causes glycogenosis type VIII and McArdle’s syndrome (glycogenosis type V), respectively. These diseases are associated with painful, exercise-induced cramps and severe exercise intolerance in voluntary skeletal muscles, but do not affect EOM (11). Our expression profiling suggests reasons why EOM are spared in these disorders: unlike skeletal muscles, EOM 1) synthesize only minimal amounts of glycogen and 2) do not depend on glycogen as a primary energy source, but rather oxidize blood-borne glucose to meet the energy requirements.

The EOM expression profile suggests efficient glucose utilization via the Krebs cycle and conversion to energy via oxidative phosphorylation. PFK-C mRNA was increased in EOM. Typical skeletal muscle expresses M rather than the C isoform of PFK (12). Increased expression of PFK-C is predicted to enhance efficiency of glucose utilization in EOM, since PFK is the rate-limiting enzyme in glucose catabolism toward pyruvate. Additionally, fructose-1,6-bisphosphatase mRNA (which catalyzes the opposite reaction to PFK) was decreased in EOM. Thus the EOM expression profile is consistent with a high level of unidirectional, constitutive, glucose utilization, which enables the EOM to function for long periods of time.

The expression profile suggests preferential utilization of pyruvate as anaplerotic substrate for the Krebs cycle. LDH-B mRNA was increased in EOM. In contrast to other LDH isozymes, the heart-specific isoform LDH-B oxidizes lactate into pyruvate, rather then reducing pyruvate into lactate (44). Thus increased expression of LDH-B in EOM supports oxidative utilization of pyruvate. Furthermore, expression of pyruvate dehydrogenase kinase 4 (PDK4), an inhibitor of pyruvate dehydrogenase complex (PDHC) activity, was decreased 9.95-fold. The predicted increased PDHC activity also favors oxidative phosphorylation. Consistent with models for enhanced oxidative energy metabolism in EOM, we noted an increase in mitochondrial density (Fig. 4) and mitochondrial gene expression (Table 1) in EOM. The dependence of EOM on mitochondrial-based oxidative phosphorylation offers an explanation for the early involvement of EOM in mitochondrial myopathies. The blood glucose-based energy metabolism (Table 1), along with high vascular and mitochondrial content, suggests the mechanistic basis of the overall metabolic adaptations used by the EOM to undertake fatigue resistant, rapid (∼400 Hz) contractions for relatively long periods of time, as proposed in the model (Fig. 7). These adaptations are in contrast to what is usually noted in fast skeletal muscles in mammals, but strikingly similar to those noted in the wing muscles of hummingbirds that can contract at 80 Hz for significant time periods (14).

Antioxidant capacity.

The high mitochondrial content and oxidative phosphorylation in EOM would be predicted to lead to increased levels of reactive oxygen species (ROS), which could cause significant cellular damage. Our microarray screening revealed that the EOM express high levels of the chemoprotective enzymes glutathione-S-transferase (GST) and UDP-glucuronosyltransferase (UDP-GT) mRNAs, presumably as an adaptation to the increased ROS that would be generated in EOM during oxidative phosphorylation. The increased levels of GST and UDP-GT mRNAs are consistent with previous biochemical reports suggesting that EOM had elevated antioxidant enzyme activity compared with limb muscle (38). We also found a decrease in silencer factor B (sfB), which is a repressor of GST transcription (17). A downregulation of the transcriptional repressor sfB would be predicted to increase GST transcription and steady-state mRNA levels. Thus the expression of GST may be regulated, in part, by a sfB-mediated feedback loop at the promoter level of the GST gene in the EOM.

Calcium homeostasis.

The expression profile indicates that EOM are likely to have a greater capacity to maintain intracellular Ca2+ homeostasis. EOM demonstrate an increased expression of the phospholamban (PLN) gene, which regulates the activity of sarcoplasmic reticulum Ca2+-ATPase (26). Interestingly, decreased PLN levels have been implicated in the abnormal Ca2+ flux seen in cardiac hypertrophy (33). We also found increased expression of genes encoding Ca2+-binding proteins in EOM (e.g., S-100 alpha, S-100 beta, p9Ka protein). These results are consistent with a previous study that demonstrated the superior capability of EOM to maintain Ca2+ homeostasis when challenged with pharmacological drugs that elevate intracellular calcium levels (22). These data reinforce the hypothesis that EOM maintain intracellular Ca2+ homeostasis better than limb muscles (22).

Growth, development, and regeneration.

EOM displayed an increased expression of genes that are typically associated with growth, development, and regeneration (e.g., embryonic MyHC, development-related protein). Additionally, we observed increased expression of genes for the fibroblast growth factor receptor 1 (FGFR1), insulin-like growth factor I (IGF-I), its binding proteins IGFBP-5 and IGFBP-6, as well as the satellite cell marker vascular cell adhesion molecule 1 (VCAM1). While FGF is thought to act as a competence factor and to induce quiescent satellite cells to enter the G1 phase of the cell cycle, IGF-I is a candidate progression factor and is thought to facilitate activated satellite cells to traverse the remainder of the cell cycle (11). Additionally, a number of transcription factors (e.g., Pitx2) were found to be increased in EOM, suggesting that transcriptional regulation may play a part in controlling overall gene expression in EOM during growth and development. Support for this comes from the demonstration that Pitx2 deficiency is associated with dysgenesis of EOM in Pitx2 knock out mice (23).

Disease perspective.

EOM are enigmatically spared in DMD. The exact mechanism of muscle damage in DMD is not fully understood, however, a cascade of events involving damage/ loss of sarcolemmal integrity and linkage with the extracellular matrix, a rise of intracellular Ca2+, followed by activation of calcium-activated proteases and muscle necrosis is widely accepted (4, 5, 49). Indeed a recent expression profiling study of DMD muscle supported the existence of these pathological cascades and defined a DMD-specific expression profile (8).

The hypotheses proposing why EOM escape damage, converge on the notion that EOM group-specific properties may help prevent or compensate against the deleterious consequences of dystrophin deficiency. Overexpression of the dystrophin-related protein, utrophin (which can functionally substitute for dystrophin), has been proposed as a compensatory mechanism (27). Increased utrophin was reported in some, but not all, studies performed in normal and dystrophin-deficient mouse EOM (22, 27, 37). No increase of utrophin was noted in EOM from normal or dystrophin-deficient dogs or EOM of control humans or human DMD patients (22). The current study found no increase of utrophin mRNA in rat EOM either, making it highly unlikely that utrophin upregulation plays a role in sparing of EOM in DMD. The high antioxidant capacity of EOM (38) and/or the greater capacity of EOM to maintain intracellular calcium homeostasis (22) have been suggested to protect the EOM from dystrophin-deficiency-induced necrosis. Our expression profile supports these hypotheses. The increased expression of genes associated with muscle growth, development, and regeneration in the adult rat EOM expression profile provides an alternative hypothesis that EOM may compensate for the damage in DMD by efficient regeneration/myogenesis. EOM are certainly capable of regenerating efficiently in toxin-induced myonecrosis models in mammals (9, 28), however, it is unclear whether this mechanism plays any role in adult EOM or in the sparing of EOM in dystrophin deficiency. We believe that it is important to analyze regeneration and myogenesis in EOM to help test this intriguing possibility (29).

In conclusion, we used a variety of molecular and structural methods to identify, validate, and help define the molecular determinants of the unique EOM allotype. We have also proposed a model for the glucose-based, aerobic metabolism in EOM, based on the expression profile and biochemical data presented in this study. We believe that our definition of the EOM-specific expression profile will help provide a more mechanistic understanding of the distinct metabolic and pathophysiological properties of EOM.


We are grateful to the family of “WP,” a DMD patient, for donating his body to medical research. WP’s autopsy revealed an enigmatic sparing of EOM (22) and stimulated our thinking for the current study.

We thank Dr. Clara Franzini-Armstrong (Univ. of Pennsylvania) for help in performing the EM. We thank Dr. Thomas Voit (Univ. of Essen), Dr. Michael Kaufmann (Univ. of Witten/Herdecke), and Drs. Carsten Bönnemann and Richard Stone (Univ. of Pennsylvania) for guidance and helpful discussions. We thank Dr. Yi-Wen Chen and Karyn Esser for generating the expression profiling data on normal rat TA muscle. We thank Dr. Yi-Wen Chen and Karyn Esser for generating and sharing the normal rat TA muscle expression profiling data (Chen YW, Nader GA, Baar KR, Fedele MJ, Hoffman EP, and Esser KA, unpublished observations). Rehannah Borup (Children’s Hospital Medical Centre) and Manoj Sinha (Enjbiz.com) for help in bioinformatic analysis, and Thomas Krag (Univ. of Pennsylvania) for help in dissections of EOM.

This work was supported in part by grants to T. S. Khurana from the Dutch Duchenne Parents Project (The Netherlands), the Muscular Dystrophy Association (USA), Association Française contre Les Myopathies (France). E. Felder was supported in part by the National Heart, Lung, and Blood Institute Grant HL-48093 to Dr. C. Franzini-Armstrong. M. D. Fischer was supported in part by the Henry R. Viets fellowship from the Myasthenia Gravis Foundation (USA).


  • Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).

    Address for reprint requests and other correspondence: T. S. Khurana, Dept. of Physiology and Pennsylvania Muscle Institute, Richards A601, Univ. of Pennsylvania School of Medicine, 3700 Hamilton Walk, Philadelphia, PA 19104-6085 (E-mail: tsk{at}mail.med.upenn.edu).


  • 1 Supplementary Material to this article is available online at http://physiolgenomics.physiology.org/cgi/content/full/9/2/71/DC1.


View Abstract