Physiological Genomics

Layer-specific differences of gene expression in extraocular muscles identified by laser-capture microscopy

Murat T. Budak, Sasha Bogdanovich, Martin H. J. Wiesen, Olga Lozynska, Tejvir S. Khurana, Neal A. Rubinstein


In mammals, separate muscles are typically specialized as a whole to provide distinct functional roles leading to well-recognized adaptations. This is exemplified in the lower limb by the slow, fatigue-resistant soleus, which provides a postural role vs. the fast, fatiguable tibialis anterior (TA), which provides rapid movements. A unique characteristic of extraocular muscles (EOMs) is their compartmentalization into two distinct layers, the orbital layer (OL) and global layer (GL), presumably to subserve diverse functions within the same muscle. However, molecular evidence of this diversity has been limited. We used laser-capture microscopy coupled with microarray-based expression profiling to identify molecular differences between the OL and GL of rat EOMs. We found that 210 genes were differentially regulated between these layers at a twofold expression cutoff. Differences in genes related to metabolic pathways and related to structural elements of muscle and nerve formed the largest functional clusters. Layer-specific differential expression was validated at both mRNA and protein level for MYH3, MYH6, and ACTN3. The expected layer-specific differences among genes encoding vascular elements were not evident by profiling; morphometric analysis demonstrated that the differences exist, but at a magnitude below the cutoff level established by our statistical methods. Comparison of these results with previous results comparing whole EOMs and TA suggest evolutionary mechanisms may play a role in achieving functional distinctions between OL and GL.

  • eye muscle
  • orbital layer
  • global layer
  • transcriptome
  • expression profiling
  • microarrays
  • GeneChips

the extraocular muscles (EOM) are specialized skeletal muscles required by the visual system to locate and track objects. Since these muscles are adapted to their roles in the control of eye movement, they exhibit fundamental differences from other skeletal muscles in terms of innervation, neuromuscular junctions, and mechanical properties (7, 37). Moreover, while limb muscles arise from the somites, the EOM develop from condensations of paraxial mesoderm. Hoh and Hughes (16) have suggested that muscle groups with a high degree of functional specialization would have patterns of gene expression that are distinct from those in other skeletal muscles, and the EOMs are one of these distinct muscle “allotypes”. We and others (13, 20, 31, 38) have begun to define the EOM allotype at a molecular level and have shown that EOMs have a unique pattern of gene expression profile compared with leg muscles.

One of the most striking anatomical features of all six EOMs is a distinct laminar organization, consisting of a thin outer orbital layer (OL) and a thicker inner global layer (GL) separated by a well-defined gap. Demonstrated first over 50 years ago, the laminar structure and intervening gap formed the basis of one of the earliest schemes for EOM classification (18). Although several fiber types exist within each of these layers, the fiber types of the OL are distinct from those of the GL. The OL and GL fiber types also differ in fiber size, electromyography (EMG) characteristics, vascular content, metabolic activity, and response to botulinum toxin treatment for strabismus (8, 22, 25, 37, 39, 46). Moreover, none of the fiber types in either the OL or GL are strictly homologous to those in other skeletal muscles. A number of researchers have developed systems of fiber type classification for the EOMs based on a combination of metabolic and innervation characteristics (25, 45).

The consistency across species of this compartmentalization of the EOMs into layers suggests unique functions for each layer; unfortunately, these unique functions have not yet been identified. It has been suggested that these specializations help EOM provide a functionally diverse variety of voluntary, saccadic, and reflex eye movements, with great fidelity for extended periods of time. This is in contrast to what is seen in the limbs where separate muscles are adapted to provide distinct roles. In the limb, for example, the soleus contracts slowly and can sustain large loads for long periods of time in its role as a postural (antigravity) muscle, whereas the gastrocnemius, with the same insertion as the soleus, contracts rapidly to provide fast-twitch, forceful, but easily fatiguable movements.

It would not be surprising, then, if the OL and GL also served different specialized functional processes. Different functions, however, have only been hypothesized. One theory ties the unique properties of orbital fibers to the “active pulley hypothesis,” in which the orbital fibers attach to the pulley and alter the direction of action of the GL fibers. This would have a significant effect on the rotational axis of the globe (10, 11, 21). OL fibers are recruited early during eye movements, have sustained activity throughout much of the oculomotor range, and thus are thought to be instrumental in maintaining ocular alignment (43). Consistent with this interpretation, studies suggest that, even though both orbital and global fibers are paralyzed by botulinum toxin, the OL fibers probably are the prime target of successful toxin treatment of strabismus (44, 46).

Because of the unique functional anatomy of the EOM (10, 11, 30), layer-specific differences of gene expression would be predicted to have wide-ranging consequences for eye movements in general and binocular alignment in particular. However, the small size of these muscles and difficulty in separating the layers due to its complex anatomy, in particular the contour of the EOM, raise significant obstacles for performing systemic gene expression studies using traditional microdissection methods. In this paper, we utilized laser-capture microscopy (LCM) and linear RNA amplification to unambiguously and selectively dissect out the OL and GL layers of rat EOMs and study differences in their patterns of gene expression at the level of the transcriptome. We performed expression profiling of rat OL and GL using Affymetrix (GeneChips) RAE230A high-density oligonucleotide microarrays (∼16,000 probe sets). Verification of expression changes was undertaken at RNA, protein, and structural level using a variety of cellular and molecular methods. Identification of layer-specific differences at the level of the transcriptome should help provide mechanistic clues for functional differences between the OL and GL of the EOM.


Tissue preparation and LCM.

Two adult Sprague-Dawley rats were killed using CO2 asphyxiation. Under sterile conditions, orbits were exenterated, and rectus EOM were quickly dissected out, placed in cryomolds, covered with OCT tissue embedding medium (Tissue-Tek; Sakura Finetek, Tokyo, Japan), and snap-frozen in 2-methylbutane chilled using liquid nitrogen. The EOM were then sectioned transversely into 7- to 10-μm sections using a Microm HM 500 cryostat (Zeiss, Oberkochen, Germany), mounted on SuperFrost Plus electrostatically charged slides (Fisher Scientific, Pittsburgh, PA), and immediately frozen on a dry ice. The sections were stored at −80°C until use. Every fifth slide was stained with hematoxylin and eosin to monitor overall tissue morphology. Sections taken from a region of EOM 1.3–2.3 mm anterior to the optic nerve ending were used for LCM. HistoGene LCM frozen section staining reagents were used for tissue dehydration and staining as described by the manufacturer (Arcturus Engineering, Mountain View, CA). Briefly, sections were fixed in 75% ethanol, hydrated in RNase-free water, and stained with HistoGene staining solution. Sections were washed in water and dehydrated sequentially in 75%, 95%, and 100% ethanol followed by 100% xylene. Sections were allowed to air dry and were stored in a desiccator at room temperature until utilized for LCM; care was taken to ensure that LCM was completed within 2 h of being placed in the desiccator.

The PixCell II LCM System (Arcturus Engineering) was used for LCM. A cap and thermoplastic film was placed above the frozen section. Areas of interest for microdissection were visualized using standard light microscopy; a single extremely brief pulse of laser beam (“zap”) was used to melt the thin layer of thermoplastic film coating overlying a frozen section of tissue. To obtain sufficient amounts of material the procedure was repeated with the laser focused on different areas. Typically 3,000 zaps were performed per preparation (typically from 3–4 slides containing ∼9–12 sections). The melted thermoplastic membrane and small region of tissue directly underneath (typically 10 μm2/zap) adhering to it were removed by means of lifting an overlying plastic cap. All animal experiments were approved by the University of Pennsylvania, Institutional Animal Care and Use Committee.

RNA isolation.

Eight independent total RNA preparations were made from rectus muscles of two rats using the PicoPure RNA isolation kit (Arcturus Engineering). Briefly, total RNA was extracted from the captured cells by incubating LCM caps in extraction buffer for 30 min at 42°C. RNA was purified using preconditioned MiraCol purification columns. Eluted RNA was directly used for linear RNA amplification.

Linear RNA amplification.

Linear T7-based RNA amplification (12, 36) was carried out by using the RiboAmp OA kit as suggested by manufacturer (Arcturus Engineering). One microliter (200 ng/μl) of poly-deoxyinosinic-deoxycytidylic acid (Sigma-Aldrich, St Louis, MO) was added as a nucleic acid carrier to the RNA samples. Briefly, total RNA was incubated with hybrid primers containing oligo dT/T7 RNA polymerase binding site containing primers, RNA reverse-transcribed into double-stranded cDNA, and purified using MiraCol columns. Next, in vitro transcription was performed using T7 RNA polymerase, and the amplified antisense RNA (aRNA) purified using MiraCol columns. An aliquot was used for quantification and verification of quality (260/280 ratio between 1.8–2.1) using a GeneQuant Pro spectrophotometer (Amersham Pharmacia Biotech, Uppsala, Sweden). Total aRNA yields from ∼3,000 zaps were typically 300–900 ng.

GeneChip-based expression profiling.

aRNA, 150 ng, from each independent RNA preparation was reverse transcribed into cDNA. This double-stranded cDNA was used as template for a second round of linear amplification with biotin labeling using ENZO BioArray High Yield RNA Transcript biotin labeling kit (Affymetrix, Santa Clara, CA) to yield biotin-labeled RNA that had been subjected to two rounds of linear amplification (aaRNA). Yields were typically 15–30 μg with 260/280 ratios between 1.9–2.1. Amplification of control RNA sample performed in parallel suggested that ∼20-fold first-round amplification and greater than 100-fold second-round amplification was achieved. Biotin-labeled aaRNA was fragmented to ∼200-bp size by incubating in fragmentation buffer (200 mM Tris acetate, pH 8.2, 500 mM KOAc, and 150 mM MgOAc) for 35 min at 94°C prior to hybridization.

Ten micrograms of fragmented, biotin-labeled aaRNA was hybridized to Affymetrix RAE230A GeneChips for 16 h on an Affymetrix 400 fluidics station (Affymetrix). GeneChips were labeled with phycoerythrin-conjugated streptavidin, signal amplified using biotin-labeled anti-streptavidin antibodies, followed by staining with phycoerythrin-conjugated streptavidin. Fluorescence was read using an Agilent model G2500A GeneArray scanner and data were analyzed using Affymetrix (v. 5), GeneSpring (v. 6.1), and S+ software. For all GeneChips used in this study, the 3′/5′ ratios for GAPDH were between 5.4 and 12.8, and those for β-actin were between 7.7 and 10.7. Averaged technical replicates of rat-1 OL, rat-2 OL, rat-1 GL, and rat-2 GL were used for normalization by S+ArrayAnalyzer software v. 1.1.4 (Insightful). Differential expression of genes in OL vs. GL was ascertained by the method of Benjamini-Hochberg in these four normalized data sets using local pooled error (LPE) (S+ArrayAnalyzer software) set at a 0.01% false discovery rate (FDR), as described previously (2, 17). This set of differentially expressed genes was then filtered using a >2-fold change cutoff to further increase stringency of the screen. Functional clusters were assigned by searching each gene (individually) in NetAffx ( and Entrez PubMed ( databases and the Database for Annotation, Visualization and Integrated Discovery (DAVID, Additionally, each gene was also searched (by annotation) in Online Mendelian Inheritance in Man database (OMIM, and the Rat Genome Database (

RT-PCR validation.

Independent biological validation of expression profiling at the RNA level was performed for 6 genes (MYH3, MYH6, MYH8, TNNT2, ACTN3, and EVT1) utilizing new RNA preparations that were not used for profiling experiments. These genes were validated using both using both semiquantitative reverse transcription-polymerase chain reaction (RT-PCR) and real-time RT-PCR. Six micrograms of biotin-labeled aaRNA from each layer was reverse transcribed into cDNA and purified using the RiboAmp OA kit and suspended in 40-μl volume. We used 2.5% (vol/vol) of purified cDNA (corresponding to 150 ng RNA) as template for RT-PCR. The primers used for ACTN3 were R-Actn3-F (5′-GGAATGGGATGATGGAACCTG-3′) and R-Actn3-R (5′-TGCTCTGAGGGACAGTGGAATC-3′). For TNNT2, the primers were R-Tnnt2-F (5′-GAACAGGAGGAAGGCTGAAGATG-3′) and R-Tnnt2-R (5′-GTTTTGGAGACTTTCTGGTTGTCG-3′). For MYH8, the primers were rat 5′PFast (5′-CGGGATCCGNGAGGTTCACACTAAA-3′) and Rat 3′PHNEO (5′-GCGAATTCTATTCAGCTTTAACAGGA-3′). For MYH6, the primers were Rat 5′PSlow (5′-CGGGATCCAGCANGAGCTGGANGAG-3′) and Rat 3′PAlpha (5′-GCGAATTCTGTCTGGCGCTCATGTTT-3′). For MYH3, the primers were Rat 5′PSlow (5′-CGGGATCCAGCANGAGCTGGANGAG-3′) and Rat 3′PEMB (5′-GCGAATTCAANNTTTATTGCATGTG-3′). For EVT1, the primers were R-Evt1-F (5′-CAGAACACGGCATTTGGCAC-3′) and R-Evt1-R (5′-CCATTTCCCTCCAGAGTGAAGG-3′). Additionally, as an internal control for efficiency of RT and normalization for quantification, we simultaneously amplified rat glyceraldehyde-3-phosphate dehydrogenase (GAPDH) using the primers RGAPDHF (5′-CCATGGAGAAGGCTGGGG-3′) and RGAPDHR (5′-CAAAGTTGTCATGGATGACC-3′).

Semiquantitative RT-PCR was performed as previously described (13). Briefly, aliquots were removed to determine the linear amplification range of each gene. PCR products were resolved on 2% agarose gels and visualized using SYBR Green I (Molecular Probes, Eugene, OR) gel staining a Typhoon 8600 fluorescence imager (Molecular Dynamics). Quantification was performed on aliquots that lay within the linear amplification range for each reaction and normalized against GAPDH by using ImageQuant 5.0 software (Amersham Biosciences). For independent validation, real-time RT-PCR was performed on an ABI Prism 7900HT machine (Applied Biosystems). Data was analyzed using comparative ΔΔCt relative quantitation method (23).


Adult rats were killed using CO2 asphyxiation. For immunohistochemistry and histochemistry, whole eyeballs with EOMs were dissected rapidly, then placed with the corneal surface facing down on a thin cork disc using OCT. Tissues were frozen in liquid nitrogen-cooled 2-methylbutane and stored at −80°C. Frozen sections, 8–12 μm thick, were cut with a Microm HM 500 cryostat (Zeiss) and collected on SuperFrost Plus electrostatically charged slides (Fisher Scientific). Tissue sections were fixed in ice-cold 100% methanol for 10 min, air-dried, and stored in air-tight containers at −80°C until use, for a period not exceeding 2 wk. Sections were thawed and preincubated in 10% FBS, 0.1% Triton X-100, 1× PBS for 1 h at room temperature. Monoclonal anti-mouse antibodies for myosin heavy chain-3 (14) and myosin heavy chain-6 (35) were labeled using Zenon One Alexa Fluor 546 mouse IgG1 labeling kit (Molecular Probes). Sections were incubated with primary labeled antibodies (MYH3 antibody diluted 1:200 and MYH6, diluted 1:100 in PBS) in a humidified chamber for 45 min at room temperature, washed three times in 1× PBS for 5 min each, then visualized (MYH antibodies). ACTN3 immunolabeling was performed by incubating sections with a rabbit ACTN3 antibody (1) diluted 1:500 in PBS for 1 h, washed and incubated with 1:200 dilution of Alexa 488 goat anti-rabbit secondary antibody (Molecular Probes), washed, and visualized. Glycogen content of OL and GL was estimated using periodic acid-Schiff (PAS) reaction (Sigma). After washes, sections were mounted with aqueous mounting medium, visualized on an Olympus BX51 microscope and photographed using a MagnaFire CCD camera.

For light and electron microscopy, adult rats were killed, and left common carotid was cannulated with 0.8-mm diameter tubing and perfused with 0.1 M Ca2+/Mg2+-free PBS for 1 min to relax musculature, then fixed using 6% glutaraldehyde in 0.1 M cacodylate buffer for 30 min at a flow rate of 2.5 ml/min. The left orbits of three different animals were exenterated, and the globes with all EOMs still attached were removed en bloc from the orbit. The Haderian gland was carefully dissected away from the four rectus EOMs. The recti were postfixed in 2% OsO4 for 1 h and stained overnight in saturated uranyl acetate. Individual EOM were dehydrated, infiltrated, and divided transversely into three regions, i.e., proximal, middle, and distal with respect to the scleral insertion site. Each of these regions was embedded separately in Epon. Semithin sections were cut from each region using a glass knife-equipped Leica Ultracut UCT microtome, and photographs were taken using a MagnaFire CCD camera mounted on an Olympus BX51 light microscope. Photomontages of the entire rectus muscle cross section were made digitally using Adobe Photoshop v. 7.0. Morphometric analysis was performed using Scion Image 4.02 software ( Vascular content of each of these regions was determined by counting all capillaries and fibers in randomly chosen 0.04-mm2 areas within each layer on each given slide. Thirty such 0.04-mm2 areas from 10 different muscles were used for the analysis. A total of 7,468 fibers (3,994 OL and 3,474 GL) were counted and their areas measured. A total of 5,680 capillaries (3,384 OL and 2,296 GL) were counted. Capillary identification was confirmed by cutting ultrathin sections using a diamond knife-equipped Leica Ultracut UCT microtome, from an area adjacent to region analyzed by light microscopy, and observing them on a Philips 410 electron microscope.


Isolation of material from OL and GL using LCM.

The small size of rat EOM (∼10 mm long) and even smaller dimensions of individual layers (∼500 μm) precludes isolation of layer-specific material for transcriptome analysis by traditional dissection techniques. Therefore, we used LCM to obtain layer-specific material from the OL and GL of rat rectus EOMs for this study. The principle of LCM is that an extremely brief pulse of laser beam (“zap”) melts a thin layer of thermoplastic film coating overlying a frozen section of tissue. The melted thermoplastic membrane adheres tightly to the small region of tissue directly underneath (typically 10 μm2), which can then be easily be removed by means of lifting the overlying plastic cap. To obtain sufficient amounts of material the procedure is repeated with the laser focused on another area. As illustrated in Fig. 1A, the overall laminar organization of the rat rectus EOM was clearly visible upon HistoGene staining. Figure 1B shows the OL after it had been subjected to multiple zaps and OL-specific material was combined and captured in the cap. Figure 1C shows the total postcapture, delaminated material adherent to the underside of the capturing cap. Akin to a jigsaw puzzle, the overall morphology of the collage of captured material can be still seen to retain the overall shape and size of the OL from where it was captured, despite the fact that the material shown in Fig. 1 was captured by making ∼300 individual zaps.

Fig. 1.

Visualization of orbital layer (OL) and laser-capture microscopy (LCM) of OL material. Cryosections of extraocular muscle (EOM) were stained with HistoGene, processed for LCM, and visualized using light microscopy at low power. A: clearly demarcated OL and global layer (GL), prior to LCM. B: the same section visualized post-LCM, showing removal of OL material with the GL material being unperturbed. C: the total, post-LCM, delaminated material adherent to the underside of the capturing cap. Scale bar = 50 μm.

Expression profiling of OL and GL of rat rectus EOM.

To define the expression profile of OL vs. GL of rat rectus EOM, we screened Affymetrix RAE230A GeneChips with RNA extracted from rat OL and GL. Using all the rectus muscles of two separate rats, four independent RNA preparations were made for each layer-specific microdissection and used for eight independent screening experiments yielding eight individual data sets (4 OL and 4 GL data sets). We performed hierarchical clustering for all eight individual data sets to determine the overall similarities within, and differences between, the OL profiles and the GL profiles. In these experiments normalized raw data was condition-clustered (using OL and GL as conditions) to demonstrate the overall similarity measurements of different genes within each condition (i.e., OL and GL) and differences between the conditions (i.e., OL vs. GL). Analysis of branch lengths for the OL and GL subtrees suggested that the four OL samples were closely related to each other, as were the four GL samples; importantly, the four OL samples and four GL samples were distinct from one another in a group-specific manner (Fig. 2). The profiling data was further analyzed using standard bioinformatic and statistical approaches using GeneSpring and S+ statistical software. This method revealed 103 upregulated and 113 downregulated transcripts in OL compared with GL, based on a twofold difference cutoff at an FDR of 0.01%. The heat map representation in Fig. 2 and scatter graph in Fig. 3 show the expression levels of all the transcripts that were differentially expressed between the OL and GL using these criteria. These genes represent 0.65% and 0.71% of the total probe sets that were screened (15,866). Detailed analysis of the 216 probe sets revealed presence of 6 transcripts that could be readily determined to emanate from the same genetic loci, thus reducing the actual number of genes being encoded by the transcripts to 210. Of the 102 genes found upregulated in OL, 47 were well-characterized genes (representing 46% of all upregulated genes; 0.30% of all probe sets screened) and 55 encoded genes of unknown function (representing 54% of upregulated genes; 0.35% of all probe sets screened). In case of the 108 genes that were found to be downregulated in OL, 45 were previously described genes (representing 41.7% of downregulated genes; 0.28% of all probe sets screened) while 63 encoded genes of unknown function (representing 58.3% of downregulated genes; 0.4% of all probe sets screened). The set of genes of unknown function are not discussed further in this study. Nevertheless, since this set of genes helps to accurately and comprehensively define the OL and GL expression profile, the set is also listed as supplemental data (Supplemental Table S1). All supplemental data tables are available at the Physiological Genomics web site.1 Additionally all primary data has been deposited in the Gene Expression Omnibus database (; accession number GSE1374).

Fig. 2.

Differential expression of transcripts in OL and GL GeneChips. Graphical representation of all (216) transcripts that were differentially expressed in OL and GL. As can be seen here, the four OL and four GL GeneChip data cluster into two distinct groups based on correlation of gene expression patterns. The branch lengths for OL and GL subtrees seen at the top are based on normalized raw data of all transcripts and quantitatively demonstrate that the four OL samples are closely related to each other, as are the four GL samples. Each horizontal colored bar represents one probe set, and the color of the bar determines the degree of expression (red = upregulated genes; blue = downregulated genes).

Fig. 3.

Scatter graph of differentially expressed genes in OL vs. GL, on a log scale. Each individual point on the scatter graph represents a probe set that met the twofold differential expression cut off used in this study. Genes lying furthest off the diagonal exhibit greatest expression differences between the OL and GL. The MYH3 and TNNT2 genes can be noted as expressed at higher levels in OL vs. GL, whereas ACTN3 is expressed at significantly higher levels in GL vs. OL.

We sorted known, differentially expressed genes into various functional groups, examples of which are presented in Table 1. (Additionally, we have provided a list for all the genes sorted into different functional groups in Supplemental Table S2). Genes related to metabolic pathways 28.3% (26 of 92) and structural elements for muscle and nerve 21.7% (20 of 92) were the largest functional groups of all differentially expressed genes. Other groups included genes encoding intracellular signaling 13% (12/92), chemoprotection and calcium homeostasis 11.9% (11/92), growth and/or regeneration 11.9% (11/92), channels and membrane 8.8% (8/92), and immune response 4.4% (4/92). Differential expression of genes encoding vascular elements was not detected, which was somewhat surprising given that differences in OL and GL vascularity are known to exist in humans (7). Additionally, we cross-compared (or homology mapped) the overall profile of genes differentially regulated in the OL vs. GL with those that had previously been identified by us (13) as being differentially regulated in EOM vs. leg muscle (Table 2). The homology list revealed a high number of concordant changes between differentially regulated genes in these sets of data, suggesting that the divergence between OL and GL accentuates the divergence previously noted previously between EOM and TA. Thus 8 of 9 genes that were upregulated in EOM vs. TA were also upregulated in OL vs. GL. Additionally, 15 of 17 genes that were downregulated in EOM vs. TA were also downregulated in OL vs. GL. Taken together, these data are consistent with the OL and GL possessing their own unique molecular makeup, significantly different from each other.

View this table:
Table 1.

Functional clustering of selected genes differentially regulated in OL vs. GL

View this table:
Table 2.

Cross-comparison of EOM vs. tibialis anterior with OL vs. GL gene tests

Validation of expression profiling.

We used a number of independent methods to validate important functional components and predictions revealed from expression profiling of individual layers of EOM. Expression levels of six genes representing different important structural groups [TNNT2, ACTN3, MYH3 (embryonic), MYH6 (α-cardiac), MYH8 (perinatal), EVT1] were verified at the mRNA levels by semiquantitative and real-time RT-PCR (Table 3 and Fig. 4). Furthermore, a number of predictions based on the expression profile were also verified at the protein and metabolic level using immunohistochemistry (Fig. 5) and PAS histochemical analysis, respectively. The prediction of differential glycogen content between the two layers based on profiling (Table 1) was validated using PAS staining (Fig. 6A). Detailed analysis of the metabolic cluster suggested a model for differential glucose-based oxidative metabolism in the layers (Fig. 6B). Since differential expression of genes encoding vascular elements between OL and GL was not detected in our profiling experiments, we undertook morphometric analysis to quantify the vascular density of the two layers. Light microscopy-based morphometric analysis demonstrated that rat OL is indeed ∼1.5-fold more vascular than GL, when comparing the number of capillaries per square millimeter in each layer (Fig. 7); increased OL vascular content has previously been noted in structural studies performed in humans (34). The low magnitude of difference in vascular density in rat EOM offers an explanation for why the vascular cluster was not detected in the profiling. Overall, the expression profile of EOM defined in this study correlates well with what is known about EOM structure and function, suggesting that our expression profiling was sensitive and comprehensive.

Fig. 4.

Validation of differential expression of TNNT2 and ACTN3 in OL and GL by SYBR green real-time RT-PCR. Genes were amplified using cDNA from independent OL and GL RNA preparations and analyzed on a sequence detection system (ABI 7900 HT). The x-axis indicates the number of PCR cycles, and the y-axis shows the amount of fluorescence emitted by SYBR green dye-labeled PCR products. Left: TNNT2 upregulation in OL. Right: upregulation of the ACTN3 gene product in GL. (List of all genes validated by real-time RT-PCR is provided in Table 2.)

Fig. 5.

Validation of layer-specific differential expression of MYH3, MYH6, and ACTN3 at the mRNA level by semiquantitative RT-PCR (top) and at the protein level by immunohistochemistry (bottom). Top: genes were amplified using cDNA made from independent OL and GL RNA preparations and analyzed using agarose gel electrophoresis. A: MYH3 showed around 10-fold increase in OL expression compared with GL (32 cycles). B: MYH6 was increased around 2.1-fold in OL vs. GL (32 cycles). C: mRNA for ACTN3 showed around 17.6-fold increase GL vs. OL (30 cycles). Experiments were performed in duplicate and also independently confirmed by real-time RT-PCR (Table 2). Bottom: 10-μm EOM frozen sections were subjected to gentle fixation, processed for immunohistochemistry, visualized using epifluorescence, and photographed using a MagnaFire CCD camera. A′: MYH3 antibodies strongly labeled myofibers in the OL rather than in the GL. B′: MYH6 showed greater labeling in myofibers in the OL rather than in the GL. C′: ACTN3 antibodies showed stronger labeling in GL myofibers compared with those in OL. Scale bars = 100 μm.

Fig. 6.

Metabolic differences in OL vs. GL. EOM frozen sections (10 μm) were subject to gentle fixation semiquantitative periodic acid-Schiff (PAS) staining, then photographed using a MagnaFire CCD camera to estimate glycogen distribution and content. A: PAS staining shows decreased glycogen content in GL compared with OL. Both OL and GL had much less glycogen than did limb muscle (data not shown). Scale bar = 100 μm. B: model for energy metabolism in GL. Increased expression of glycogen phosphorylase and phosphoglucomutase 1 genes suggest more efficient glycogen utilization by GL than by OL. Increased expression of pyruvate kinase activator (protein phosphatase 3) and pyruvate dehydrogenase activator (pyruvate dehydrogenase phosphatase) support efficient catabolism of glucose toward formation of pyruvate and entry to the TCA cycle for energy generation. OL fibers would be predicted to rely more upon blood-borne glucose than glycogen for metabolic needs. Red arrows indicate increased expression of metabolic enzymes in GL vs. OL.

Fig. 7.

Difference in vascular density of OL vs. GL. Morphometric analysis of 2-μm sections OL and GL was performed using light microscopy to measure capillary density. Vascular morphology was confirmed using electron microscopy to observe 60-nm thin sections taken from adjacent regions of the same block. Vascular content was determined by counting all capillaries and fibers in three randomly selected regions (each of 0.04 mm2 area) in OL and GL regions on each given slide. Thirty such areas from 10 different muscles were used for the analysis of each layer. A total of 7,468 fibers (3,994 OL and 3,474 GL) were scored, and their areas were measured. A total of 5,680 capillaries (3,384 OL and 2,296 GL) were counted. The average OL myofiber area (167.9 ± 22.4 μm2) was significantly less than average GL myofiber area (315 ± 34.0 μm2), as expected. Representative regions visualized by light microscopy of OL (A) and GL (B) are shown at top and show increased vascularity of OL. Thin sections made from adjacent regions visualized using electron microscopy are shown at bottom (A′ and B′). Orientation markers are visible on a myofiber (asterisk) and capillaries (arrowhead). Scale bar is 20 μm for light microscopy and 5 μm for electron microscopy. C: OL contains greater number of capillaries per myofiber compared with GL (OL = 0.89 ± 018, GL = 0.70 ± 0.17; P < 0.05). D: OL contains greater number of capillaries per mm2 compared with GL (OL = 2,820.0 ± 388.04, GL = 1,913.33 ± 388.60; P < 0.001).

View this table:
Table 3.

Differentially expressed genes verified by real-time RT-PCR


In the current study we used oligonucleotide microarrays (Affymetrix GeneChips) to analyze the transcriptome and define expression profiles associated with the two lamina or layers (OL and GL) that constitute a single EOM. We found 210 genes to be differentially expressed in OL compared with GL and define the OL and GL expression profiles. Verification of changes was undertaken at RNA, protein, metabolic, and structural level using a variety of cellular and molecular methods. The use of multiple, independent strategies to validate predictions of the OL and GL profile afford a high degree of confidence for this study. Taken together, our results support the hypothesis that fundamental differences in the overall patterns of gene expression exist between the OL and GL; our definition of these differences should undoubtedly help achieve a better understanding of EOM pathophysiology.

Recent advances in imaging, computer-aided anatomical three-dimensional reconstruction, and rigorous mathematical modeling have all contributed to the concept that the OL and GL are functionally distinct entities, despite being part of the same EOM. However, addressing the molecular mechanisms underlying EOM specialization (as well as EOM-layer specialization) has been hampered by biological and technical constraints. Due to the advent and advances in genetic technologies related to transcriptome analysis, the last three years have seen a number of studies that begin to define the rodent EOM transcriptome (13, 20, 31, 38). In the present study we used LCM to unambiguously sample the two layers of rodent EOM (Fig. 1) and use the material to begin understanding the molecular mechanisms underlying the remarkable functional specializations noted in the individual OL and GL lamina forming a single rectus EOM (Figs. 2 and 3, Table 1).

Laminar organization and pulley function occupy central roles in current thinking about EOM movements and pathophysiology. Anatomically, all the recti (and both their constituent layers) originate at the annulus of Zinn at the apex of the bony orbit. While the GL fibers insert into the sclera and rotate the globe, it is currently believed that a majority of fibers forming the OL do not insert onto the sclera of the globe; rather, they insert directly onto a specialized condensation of fascia. This structure, contiguous with Tenon’s fascia, has been postulated to serve as a functional “pulley.” Due to this unique aspect of EOM functional anatomy, layer-specific functional differences and/or pulley dysfunction are thought to have wide-ranging consequences for eye movements in general and binocular alignment in particular.

Defects in well coordinated functioning of these muscles and the “pulleys” can result in disorders of eye movement and/or binocular alignment (10). The EOM are also frequently targets of surgical manipulation or botulinum toxin injections for the management of strabismus (41). In the surgical management of strabismus, the EOM are typically repositioned, to adjust the EOM tension and achieve binocular alignment in the presence of defective efferent motor signals. More recently, it has been proposed that layer-specific laser ablation (using a laser fiber positioned within an EMG electrode) would allow more precise and specific therapies for strabismus and limit postoperative complications (42). A number of hypotheses have been proposed regarding the functional roles of the pulleys (10, 15). The “active pulley” hypothesis predicts that changes in location of the pulley due to contraction/relaxation of one of the layers (the OL) could have a significant effect on the rotational axis (10). Although clearly subserving different functional roles would require a differential molecular complement in these layers, little was known about the nature of molecules expressed in each of these layers that would bestow layer-specific roles.

We addressed the molecular mechanisms underlying the laminar (layer) specializations in this study using Affymetrix GeneChips to define the unique expression profiles of the OL and GL of rat EOMs. Differential expression of 210 genes support the hypothesis that there are unique patterns of gene expressions in these two layers. In general, the expression profile obtained in this study was also found to be concordant with known structural features of EOMs in rodents. Concordancy was also noted when comparing our layer-specific profile of rat rectus EOM with that of monkey EOM described in a study (19) that appeared while this paper was in submission (see Supplemental Table S3 for a comparison of expression levels for genes found in common in both studies). With respect to the myosin heavy chain isoforms, we had previously demonstrated, using immunohistochemistry, that the embryonic myosin heavy chain (MYH3) was found almost exclusively in the OL. Staining for this isoform did not occur in the GL. This current study confirmed those findings, with the MYH3 isoform showing a greater than fourfold increase in the orbital vs. the global region. Similarly, the α-cardiac myosin heavy chain isoform (MYH6) was found in much greater abundance in orbital than in global fibers. On the other hand, although slow twitch myosin isoform MYH7 was previously seen throughout both the global and the orbital regions by immunohistochemistry, expression profiling showed a preponderance of this isoform in the global region rather than the orbital region, an almost 17-fold difference. This suggests that not all the slow fibers recognized by our anti slow-twitch myosin antibody may contain the slow twitch myosin. In fact, it has been suggested that the slow fibers in the OL of mammals contain a slow tonic myosin rather than the slow twitch myosin (MYH7); our profiling data may support that conclusion (6). This is further suggested by our observation (N. A. Rubinstein, unpublished) that all the slow fibers in both the orbital and global regions of rat EOMs react with a variety of antibodies to slow twitch and to slow tonic myosins. Antigenic similarity, then, appears to make it difficult at this time to distinguish between slow twitch and slow tonic myosins. The profiling data has also helped us extend and clarify some of our previous findings. Previously, we were unable to make a definitive conclusion regarding expression of perinatal myosin heavy chain (MYH8) using monoclonal antibodies (40). Here, we find that the mRNA encoding the MYH8 is threefold more abundant in the orbital than in the global region.

Although we expected a concordance of slow and fast isoforms of the various contractile proteins within each layer, this did not prove to be the case; indeed our findings demonstrate the complex nature of the different layers of EOM. For example, in the GL, there was a preponderance of both the slow troponin I (TNNI1) and fast tropomyosin-α (TPM1). Most intriguing was the finding of a 10-fold increase in ACTN3 in the GL. Although ACTN3 is known to be expressed in type 2B fast fibers in mouse skeletal muscle, the lack of ACTN3 in OL is surprising given that these fibers are capable of extremely rapid contractions as well as considerable force generation. In humans, a naturally existing null mutation of ACTN3 occurs in ∼20% of the population due to homozygosity for a premature stop codon polymorphism at amino acid 577 of the ACTN3 gene that normally encodes an arginine (R577X) at this position (32). Interestingly, a strong allelic association has also been noted between the R577X genotype and elite athletic performance. A higher than expected frequency of the 577RR allele has been noted in elite female sprint athletes, while the frequency of the 577XX genotype is increased among endurance athletes (24, 48). The lack of expression of ACTN3 in OL may be indicative of an evolutionary adaptation for the specialized function of the OL related to pulley function of providing forceful contractions at rapid contraction speeds. Layer-specific transcriptional control mechanisms may regulate expression of ACTN3 in GL. Detailed analysis of the ACTN3 gene promoter motifs, as well as transcription factors expressed differentially between the layers, may offer mechanistic clues for this intriguing observation.

In conclusion, we used a variety of molecular and cell biological methods to help define the molecular determinants of OL and GL specialization of rat EOM. The expression profiling was concordant with some of the known structural features of these layers such as OL-specific expression of MYH3. Layer-specific differences in expression of structural genes related to force generation identified here begin to define the level of combinatorial diversity underlying contractile specializations allowing functionally diverse layer-specific functions. Differential expression of ACTN3 exemplifies the combinatorial diversity used to achieve the right balance of contraction speed and endurance by these layers of EOM. The studies support the existence of layer-specific differences in basic properties such as regenerative potential (22, 26, 27) and vascularity (7, 34). Interestingly, these properties are also known to differ between EOM and limb muscle (13, 26, 28, 29) suggesting the possibility that the divergence of OL from GL reflects divergence of EOM from limb muscle in murine species. It is interesting to note that in billfish (capable of diving to depths where it is both dark and cold) the OL of the superior rectus does not serve a primary function related to eye movements per se. Rather, its role to provide thermogenesis (as a specialized “heater organ”) for the brain to prevent cold-induced neuronal dysfunction during deep sea diving (3). The heater organ is thought to be an evolutionarily adapted structure; the adaptations including differences in the expression levels and/or isoforms of molecules in the muscle (5, 33, 47) as well as structural specializations of the sarcoplasmic reticulum itself (4). It is intriguing to speculate that the compartmentalization of EOM layers has resulted from divergent evolution of these layers among different species. Future studies (in different species) are needed to test our hypothesis that the molecular divergence of mammalian OL and GL noted in this study has evolutionary implications. We believe that our definition of expression profiles of OL and GL should help provide a more mechanistic understanding of the distinct metabolic and pathophysiological properties of EOM in general and these layers in particular.


This work was supported in part by grants from the Muscular Dystrophy Association (to S. Bogdanovich), from the Dutch Duchenne Parents Project (to T. S. Khurana), as well as National Institutes of Health Grants AR-48871, AR-41696, and EY-013862 (to T. S. Khurana), Grant EY-11779 (to N. A. Rubinstein), and Vision Research Center Core Grant EY-01583.


We thank Drs. Eric Hoffman and Javad Nazarian (Children’s National Medical Center), Dr. Kathryn North (The Children’s Hospital at Westmead), Don Baldwin (Microarray Core, Univ. of Pennsylvania), and John Tobias (Bioinformatics Core, Univ. of Pennsylvania), for guidance and helpful advice. We thank Dr. Clara Franzini-Armstrong for advice, guidance, and kind use of electron microscope imaging facilities. We thank Dr. Alan Beggs (Boston Children’s Hospital and Harvard Medical School) for kind gift of ACTN3 antibodies.


  • 1 The Supplemental Material (Supplemental Tables S1–S3) for this article is available online at

  • Article published online before print. See web site for date of publication (

    Address for reprint requests and other correspondence: T. S. Khurana, Dept. of Physiology and Pennsylvania Muscle Institute, A601 Richards Bldg., Univ. of Pennsylvania School of Medicine, 3700 Hamilton Walk, Philadelphia, PA 19104 (E-mail: tsk{at}; and N. A. Rubinstein, Dept. of Cell and Developmental Biology and Pennsylvania Muscle Institute, BRBII/III-11th Floor, Univ. of Pennsylvania School of Medicine, Curie Boulevard, Philadelphia, PA 19104 (E-mail: nrubinst{at}



  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 10.
  10. 11.
  11. 12.
  12. 13.
  13. 14.
  14. 15.
  15. 16.
  16. 17.
  17. 18.
  18. 19.
  19. 20.
  20. 21.
  21. 22.
  22. 23.
  23. 24.
  24. 25.
  25. 26.
  26. 27.
  27. 28.
  28. 29.
  29. 30.
  30. 31.
  31. 32.
  32. 33.
  33. 34.
  34. 35.
  35. 36.
  36. 37.
  37. 38.
  38. 39.
  39. 40.
  40. 41.
  41. 42.
  42. 43.
  43. 44.
  44. 45.
  45. 46.
  46. 47.
  47. 48.
View Abstract