Generating mice carrying the Pitx2 ^neo and Pitx2 ⁻ gene-targeted alleles has been described previously. ³⁹ For these experiments, each allele had been backcrossed (N =7) onto the inbred C57BL/6J background, making each animal essentially identical genetically except at the Pitx2 locus. All breeding was performed at the University of Michigan. Genotyping by polymerase chain reaction was performed as previously described. ³⁹ Animals were housed and handled in accordance with NIH guidelines for animal care. All procedures involving mice were approved by the University of Michigan Committee on the Use and Care of Animals (UCUCA). All experiments were conducted in accordance with the principles and procedures established by the National Institutes of Health (NIH) and the Association for Assessment of Laboratory Animal Care (AALAC), and in the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Three types of timed-pregnant matings were performed to produce embryos representing all levels of Pitx2 expression within the allelic series described herein. Breedings between Pitx2 ^+/− male and female mice were used to generate Pitx2 ^+/+, Pitx2 ^+/−, and Pitx2 ^−/− embryos. Likewise, Pitx2 ^+/neo mice were bred to produce Pitx2 ^+/+, Pitx2 ^+/neo , and Pitx2 ^neo/neo embryos. Pitx2 ^neo/− compound-heterozygotes were produced by breeding Pitx2 ^+/neo males to Pitx2 ^+/− females. The morning after mating was designated as embryonic day (E)0.5. Pregnant mice were killed by cervical dislocation at either E12.5 or E14.5 and embryonic tissue was either processed for histology or stored in RNA stabilization reagent (RNAlater; Ambion, Austin, TX) at −20°C for subsequent RNA isolation. DNA was isolated from extraembryonic membranes and genotyping was performed by polymerase chain reaction. ³⁹  

Embryos intended for histology were fixed in 4% paraformaldehyde in PBS for 2 to 4 hours at room temperature. Fixed embryos were washed three times in PBS, dehydrated through graded ethanol, washed three times in methyl salicylate, and embedded in paraffin (Paraplast Plus; Fisher Scientific, Pittsburgh, PA). Sections were cut at 7 μm and mounted on slides (Superfrost Plus; Fisher Scientific) using poly-L lysine solution (Sigma-Aldrich, St. Louis, MO) diluted 1:10 in DEPC-treated water. Dewaxing and rehydration was performed according to standard methods. 

Immunohistochemistry was performed with a mouse monoclonal antibody to developmental myosin heavy-chain (Vector Laboratories, Burlingame, CA). Antigen retrieval was performed by boiling in 10 mM citrate buffer (pH 6.0), for 10 or 20 minutes, followed by rinsing in water for 10 minutes to cool. Endogenous peroxidase activity was quenched for 10 minutes in 3% hydrogen peroxide in water, followed by a 30-minute block in signal enhancer (ImageiT; Invitrogen, Carlsbad, CA). Antibody blocking was performed with a kit (Mouse On Mouse [MOM]; Vector Laboratories) according to the manufacturer’s recommendations. Primary antibody was applied at a 1:5 dilution in MOM diluent (Vector Laboratories) and incubated overnight at 4°C. Biotin-labeled anti-mouse IgG supplied in the kit was used as a secondary antibody. A tyramide signal-amplification (TSA) kit (Invitrogen) was used to visualize sites of myosin heavy-chain expression. Photography was performed on a fluorescence microscope (Eclipse 800; Nikon, Tokyo, Japan) using matched aperture and exposure settings to hold signal detection invariant between genotypes. 

Eye primordia were microdissected from E12.5 Pitx2 ^+/+, Pitx2 ^+/−, and Pitx2 ^−/− embryos. Total RNA was isolated from the eyes of individual embryos (RNAqueous Micro Kits; Ambion). cDNA and biotin-labeled cRNA was generated from 1 μg of RNA from each embryo (MessageAMP kits; Ambion). The in vitro–labeling reactions were extended overnight to increase yield. Labeled cRNA was hybridized to mouse gene microarrays (Mouse Genome 430 ver. 2.0 GeneChip; Affymetrix, Santa Clara, CA) and processed by the standard protocol. Gene expression profiles were generated from four animals from each of the three Pitx2 genotypes using a total of 12 arrays. Initial data preparation was then performed with a methodology that performed background correction, quantile normalization, and summarization of expression scores (Microarray Suite ver. 5.0. Robust Multichip Average [RMA]; Affymetrix). ⁴⁰  

A two-stage, direct-screening procedure based on the Benjamini and Yekutieli ⁴¹ reconstruction of the false-discovery rate confidence interval (FDR-CI) was used to assign probabilities to the x-fold changes of gene responses. ^{42 43 44} This method is distinct from approaches based on the conventional t-test, because it allows the experimenter to control statistical significance and biological significance in determining positive differential responses. ⁴² In the first stage, a set of genes with putative expression changes was identified with an FDR-α test. In the second stage, this set of genes was further screened to establish biological and statistical significance. A minimum x-fold change (fcmin) of 1.5-fold was established as the threshold for establishing biological significance and all probesets with P < 1 were reported. For hierarchical analysis, the top 200 FDR-CI constrained gene profiles were standardized to have mean of 0 and SD of 1 across all groups and were clustered using hierarchical clustering implemented in Cluster and TreeView. ⁴⁵ Samples were grouped by expression profiles rather than genotype and Euclidean distance was chosen for clustering as the measure of expression profile similarity. Muscle-specific array hits were identified by automated and manual inspection of the National Center for Biotechnology Information (NCBI, Bethesda, MD) Entrez Gene records. Original.cel data files will be deposited with the Gene Expression Omnibus (www.ncbi.nih.gov/geo/). 

Real-time RT-PCR was performed (TaqMan Gene Expression Arrays) on custom-designed microfluidics cards (Affymetrix). E12.5 Pitx2 ^+/+ , Pitx2 ^+/−, and Pitx2 ^−/− embryos were obtained from Pitx2−heterozygous matings. The eyes and surrounding periocular mesenchyme were microdissected and stored (RNAlater; Ambion). Total RNA was extracted from homogenized tissue (RNAqueous-micro kit; Ambion). cDNA was reverse transcribed from 1 to 5 μg total RNA by (SuperScript III with random primers; Invitrogen-Life Technologies, Gaithersrburg, MD) according to the manufacturer’s protocol. cDNA was diluted as 1 μg total RNA/100 μL final volume (10 ng/μL). Gene expression assays (TaqMan; Applied Biosystems, Inc., Foster City, CA) were performed with master mix (RealTime Ready; Q.BIOgene, Carlsbad, CA) in a thermal cycler (iCycler; Bio-Rad, Hercules, CA) according to the manufacturer’s protocol. Relative changes (x-fold) compared with wild type were calculated by using the 2^−DΔCt method and normalized to each of four independent “housekeeping” genes (Hprt, Arl10c, Hsd17b12, and 18S rRNA) whose expression has been independently confirmed as unaffected by the Pitx2 genotype. Standard error was computed from four samples of each genotype. Relative expression levels obtained after normalization to Hprt are depicted in Figure 4 , but normalization to the other three housekeeping gene yielded analogous results. 

To evaluate the effects of Pitx2 gene dose in the developing eye, we used Pitx2 ⁺, Pitx2 ^neo , and Pitx2 ⁻, to generate an allelic series of mice expressing graded levels of Pitx2 expression (Fig. 1A) . ³⁹ By definition, each Pitx2 ⁺ allele contributes 50% of PITX2 protein expression in wild-type mice. Conversely, the Pitx2 ⁻ allele is a complete null allele and expresses no functional PITX2 protein. Pitx2 ⁻ homozygotes die by E15.5 of cardiovascular defects. Pitx2 ^neo is a hypomorphic (reduced-function) allele, because the PGK-neo cassette between exons 3 and 4 interferes with normal splicing of the Pitx2 transcript, leading to reduced levels of Pitx2 mRNA and accordingly reduced protein expression. Pitx2 ^neo homozygotes also die of cardiovascular defects but not until postnatal day 1, confirming that this is a milder genetic lesion. ³⁹ By interbreeding these three alleles, we were able to generate an allelic series with estimated protein expression levels ranging from 100% (Pitx2 ^+/+) down to 0% (Pitx2 ^−/−; Fig. 1B ). 

Embryos were harvested at E14.5 and examined by standard histologic techniques to determine whether eye structures were affected by Pitx2 gene dose. Changes in the extraocular muscle were the most readily apparent. As expected, Pitx2 ^+/+ embryos had discrete, well-formed, and easily identified extraocular muscle condensations (Fig 2A) , whereas Pitx2 ^−/− embryos lacked extraocular muscles altogether (data not shown). ³⁹ Consistent with our hypothesis, extraocular muscle development in embryos with intermediate Pitx2 genotypes (+/neo, +/−, neo/neo, and neo/−) had phenotypes that paralleled Pitx2 gene dose closely. The superior oblique muscle was uniformly absent in all Pitx2 ^+/− embryos examined (n = 5) and therefore appeared to be the most sensitive to Pitx2 gene dose. The inferior oblique muscle was missing in some but not all Pitx2 ^+/− embryos (Fig. 2C) . Both oblique muscles were consistently absent in Pitx2 ^neo/neo embryos (Fig. 2D) . Evidence of all four rectus muscles was still present in Pitx2 ^+/− and Pitx2 ^neo/neo embryos, but these muscles became increasingly smaller and more disorganized as Pitx2 gene dose was reduced (Figs. 2C 2D 2E) . The most common defect observed was an abnormal continuity between the lateral and inferior rectus muscles. These muscles were consistently distinct in wild-type embryos (Fig. 2A)The abnormalities were observed in both Pitx2 ^+/− (Fig. 2C)and Pitx2 ^neo/neo (Fig. 2D)embryos, but were more pronounced in Pitx2 ^neo/neo embryos. Muscle condensations are undetectable in Pitx2 ^neo/− embryos (Fig. 2E) , analogous to what is observed in Pitx2 ^−/− embryos (Ref. ³⁹ and data not shown). By examining sections taken from older embryos, we have established that these phenotypes do not result simply from delayed growth of extraocular muscles. Taken together, these results suggest that morphologic development of extraocular muscles corresponds strongly with Pitx2 gene dose. 

To be certain that extraocular muscles were really reduced or missing, rather than simply unrecognizable due to unexpected morphologic changes, we performed immunofluorescence to detect developmental myosin heavy-chain (dMyh) protein (Figs. 2A′ 2B′ 2C′ 2D′ 2E′) , which was expressed in all differentiated muscle fibers. All slides were stained at the same time, and sections were photographed with matched camera settings, to compare relative levels of protein expression between genotypes. Our results confirmed our initial conclusions from H&E-stained sections, that overall morphogenesis of extraocular muscles correspond closely with Pitx2 gene dose and that normal development of the superior and inferior oblique muscles requires somewhat higher levels of Pitx2 function than normal development of the four rectus muscles. Furthermore, we observed a marked reduction in dMyh staining intensity beginning at the Pitx2 ^+/− level, which was even more pronounced in Pitx2 ^neo/neo embryos. Although immunofluorescence is not a quantitative assay as performed, these results provide the first experimental evidence that muscle-specific gene expression levels may also correlate with Pitx2 gene dose. 

To gain further insight into the mechanisms underlying Pitx2 gene dose effects on extraocular muscle development, we initiated a comparison of global gene expression in eye primordia isolated from E12.5 Pitx2 ^+/+, Pitx2 ^+/−, and Pitx2 ^−/− embryos (n = 4 per genotype). These genotypes correspond to 100%, 50%, and 0% of the normal PITX2 function levels, respectively. RMA, a robust and well-established methodology, was used to quantify signal intensity and normalize signals. ⁴⁰ Normalized data were then analyzed using the robust two-step FDR-CI method to identify statistically significant (P < 1) differentially expressed genes. ^{41 42} This approach is now well-established. ⁴³ Hierarchical clustering of the top 200 differentially expressed genes revealed striking patterns. Three distinct patterns of gene expression were observed based on their response to Pitx2 gene dose (Fig. 3) . Expression of genes in set A decreased in parallel with Pitx2 gene dose, indicating that these genes are dose sensitive and activation correlates positively with PITX2 protein levels. Expression of genes in set B was reduced in eyes of Pitx2 ^−/− but not Pitx2 ^+/− embryos, indicating that these genes are not affected by Pitx2 gene dose. Finally, expression of genes in set C increased with decreasing Pitx2 gene dose, indicating that these genes are dose sensitive, but that activation correlates negatively with increasing PITX2 protein levels. Because whole eye primordia were used as the RNA source for these experiments, differentially expressed genes represent both myogenic and nonmyogenic functions. We next evaluated each gene in sets A, B, and C to identify those that could be confirmed as muscle-specific based on previously published results or publicly available expression data. In all, 50 muscle-specific, dosage-sensitive genes were identified (Table 1) . Consistent with the decrease of extraocular muscle morphogenesis in response to Pitx2 gene dose, expression of all the muscle-specific genes decreased in parallel with Pitx2 gene dose (Fig. 3 ; set A). No muscle-specific genes were identified in set C, whose expression increases with decreasing Pitx2 gene dose. 

To determine the overall validity of our differential gene expression data, we selected 15 muscle-specific genes representing a range of change (x-fold) and statistical confidence levels to re-evaluate by quantitative real-time PCR (TaqMan; Applied Biosystems). All 15 genes were confirmed as sensitive to Pitx2 gene dose. In all cases, the reported change and confidence interval were equal to or greater than those predicted by the microarray experiment (Fig. 4) . Taken together, we conclude that our gene-expression profiles accurately identified the muscle-specific genes that are sensitive to Pitx2 gene dose. 

To gain insight into potential underlying mechanisms that might account for the dependence of extraocular muscle development on Pitx2 gene dose, we analyzed the biological functions of the muscle-specific dose-dependent genes. As would be predicted, genes with a variety of biological functions were represented (Fig. 5) . Prominent members of the list are Myf5, Myog, Myod1, Smyd1, Msc, and Csrp3, all of which encode transcription factors or nuclear-associated proteins that have been established as essential for skeletal and/or cardiac muscle differentiation and function ^{46 47 48 49 50} (Fig. 5 , Table 1 ). Numerous genes encoding components of the muscle structural–contractile apparatus or proteins required for muscle-specific metabolism were also represented (Table 1 , Fig. 5 ). 

The possibility that alterations in PITX2 protein function levels play a significant role in development and disease was first indicated by the heterozygous gain- or loss-of-function mutations in human PITX2, which are a significant cause of Axenfeld-Rieger Syndrome. ^{32 33 34 35} Subsequently, the effects of altered PITX2 function on the expression of downstream target genes have been established in cell culture, and the central roles of Pitx2 gene dose on pituitary gland, craniofacial, and heart development were demonstrated using gene-targeted mice. ^{36 37} We now show that morphogenesis and gene expression in extraocular muscle are also sensitive to alterations in Pitx2 gene dose. Our results suggest important mechanisms regulating extraocular muscle development, which is poorly understood. More generally, these findings are significant, because abnormal development of the ocular anterior segment and a high risk for glaucoma are seminal features of Axenfeld-Rieger syndrome, and extraocular muscle represents the first opportunity to determine the effects of Pitx2 gene dose on an ocular tissue. 

There are seven extraocular muscles in the mouse (Fig. 2F) : the superior and inferior oblique muscles, the four rectus muscles, and the retractor bulbi muscle. Humans lack the retractor bulbi muscle, and so we focused our attention on understanding the effects of Pitx2 gene dose on the other six. We demonstrate that morphogenesis of extraocular muscles was highly sensitive to Pitx2 gene dose effects. The superior and inferior oblique muscles were consistently more sensitive than the four rectus muscles. This is strikingly similar to the function of Pitx2 in the developing pituitary gland, where there is a strong correlation between overall morphogenesis of Rathke’s pouch, the early pituitary primordium, and the level of Pitx2 gene dose. ³⁷ In addition, the five neuroendocrine cell lineages of the mature anterior pituitary gland are differentially sensitive to the effects of Pitx2 gene dose, with gonadotropes being the most sensitive and lactotropes and corticotropes the least sensitive. ³⁷ Similar to the pituitary, the extraocular muscles are composed of a mixture of different fiber types, distinguished in mature muscles by their diameter, innervation pattern, oxidative potential (number and size of mitochondria, pigmentation, extent of vascularization) and ability to transmit an action potential (extent of sarcoplasmic reticulum). ^{51 52} We predict that, as seen in the pituitary, the different fiber types may be differentially sensitive to Pitx2 dosage and that the overall reductions in extraocular muscle size seen in our mutants may be related to preferential loss of specific fiber types. Unfortunately, most mice in our allelic series do not survive beyond birth, and suitable molecular markers for distinguishing between the different muscle lineages or their precursors at early time points have not yet been identified. 

Previously established functions of PITX2 in other organ systems suggest molecular mechanism(s) that may underlie the effects of Pitx2 gene dose on the development of extraocular muscle. Decreased proliferation of muscle precursors may be a contributing factor, since Pitx2 has been implicated as playing a role in cellular proliferation in skeletal and cardiac muscles. ⁵³ However, we find no evidence of proliferative changes in the extraocular muscle primordia of our allelic series (data not shown). Alternatively, increasing levels of apoptosis could also explain the progressive reduction of extraocular muscle size in our mutants. Apoptosis has been implicated in the loss of pituitary cells in Pitx2 mutant mice. ⁵⁴ However, we find no evidence of increased apoptosis in the mutant extraocular muscle primordia by TUNEL assay (data not shown). Therefore, it seems likely that precursors fated to become extraocular muscle cells are present in animals of all Pitx2 genotypes but fail to initiate or sustain their muscle differentiation program efficiently in the case of animals with reduced PITX2 function and never initiate the program in animals with no PITX2 function. Some muscle-specific genes affected by Pitx2 gene dose may be direct targets of PITX2 regulation, as suggested by the observation that a subset contain ≥1 predictes PITX2 binding sites within 2 kb of the transcription-initiation site (data not shown). Others may be indirect PITX2 targets or simply markers of muscle the muscle phenotype. Identification of the regulatory elements required for endogenous expression of each gene and testing of these elements for responsiveness to PITX2 will be necessary, to distinguish between these possibilities. 

An important insight into the underlying mechanism(s) is offered by our gene expression analysis, which demonstrates that the genes for Myf5, Myog, and Myod1 all require PITX2 for their expression and are sensitive to Pitx2 gene dose effects. These genes encode bHLH class transcription factors that are well established individually and in combination for specification and differentiation of skeletal and other muscle types. ^{55 56 57 58} Expression of these genes in the primordia of extraocular muscles has been noted previously, but the functional significance of this expression has not been clear. ^{26 55} We propose that expression of these muscle-regulatory genes is required singly, or in combination, for specification and/or differentiation of extraocular muscles, similar to what has been demonstrated previously in other muscle groups. Future testing of this hypothesis requires careful analysis of the effects of individual and combinatorial knockouts of these genes on extraocular muscle development. 

Among the key differences between extraocular muscles and other muscle groups is that Pax3 and other very early key regulators of trunk myogenesis that are essential for activation of the muscle-regulatory genes are not expressed in the extraocular muscles. ^{21 22} Pax7, which is expressed, has been proposed as a functional replacement for Pax3 in extraocular and other head muscles. ²² However, there is to date no genetic evidence to support this hypothesis. Our results provide compelling genetic evidence that PITX2 is essential to initiate muscle differentiation in extraocular muscles through a mechanism that includes activation of a set of muscle-regulatory genes. Although our results establish that Pitx2 is genetically required for activation of Myf5, Myog, and Myod1, it is not possible to determine from our current data whether these genes are direct or indirect downstream targets of PITX2. However, it is interesting to note that the cis-acting regulatory elements required to recapitulate expression from these genes in transgenic mice have been identified, and each contains one or more potential PITX2 binding sites (Refs. ^{55 59 60} and data not shown). Therefore, it is feasible that each of these key regulatory genes are direct targets of PITX2 in extraocular muscles. Future testing of this hypothesis will require introduction of each relevant transgene into Pitx2 mutant mice, as well as definitive proof of the PITX2-binding sites in cell culture and biochemical assays. 

Although our current results are exciting because they identify the first model system for understanding the role of Pitx2 gene dose effects in an ocular tissue, defects in extraocular muscle development and function are unlikely to account for the anterior segment changes and glaucoma observed in patients with Axenfeld-Rieger syndrome. However, we hypothesize that other ocular structures, in addition to extraocular muscles, are sensitive to Pitx2 gene dose, and our current results establish that our Pitx2 allelic series combined with gene expression analysis are likely to be a powerful approach for testing this hypothesis. 

 

The authors thank Joseph Washburn for providing expert technical assistance with the microarrays; and Amanda Evans, Peter Hitchcock, and Anand Swaroop for critical reading of the manuscript.