Nkx2-5 and Sarcospan genetically interact in the development of the muscular ventricular septum of the heart

The muscular ventricular septum separates the flow of oxygenated and de-oxygenated blood in air-breathing vertebrates. Defects within it, termed muscular ventricular septal defects (VSDs), are common, yet less is known about how they arise than rarer heart defects. Mutations of the cardiac transcription factor NKX2-5 cause cardiac malformations, including muscular VSDs. We describe here a genetic interaction between Nkx2-5 and Sarcospan (Sspn) that affects the risk of muscular VSD in mice. Sspn encodes a protein in the dystrophin-glycoprotein complex. Sspn knockout (SspnKO) mice do not have heart defects, but Nkx2-5+/−/SspnKO mutants have a higher incidence of muscular VSD than Nkx2-5+/− mice. Myofibers in the ventricular septum follow a stereotypical pattern that is disrupted around a muscular VSD. Subendocardial myofibers normally run in parallel along the left ventricular outflow tract, but in the Nkx2-5+/−/SspnKO mutant they commonly deviate into the septum even in the absence of a muscular VSD. Thus, Nkx2-5 and Sspn act in a pathway that affects the alignment of myofibers during the development of the ventricular septum. The malalignment may be a consequence of a defect in the coalescence of trabeculae into the developing ventricular septum, which has been hypothesized to be the mechanistic basis of muscular VSDs.

If the conditional knockout mutants above represent extreme examples of how muscular VSDs arise, then subtler signs of abnormal trabecular coalescence may be associated with milder mutations of cardiac developmental genes that cause muscular VSD. Among the genes, mutations of NKX2-5, a cardiac transcription factor, cause pleiotropic congenital heart defects 13,14 . The anatomic phenotype is determined in part by modifier genes that influence the susceptibility of specific developmental pathways to Nkx2-5 mutation 14,15 . One modifier locus on chromosome 6 for muscular VSDs was previously mapped in a cross between the C57BL/6N and FVB/N inbred strains 15 . The paucity of knowledge regarding how a muscular VSD arises motivated us to evaluate sarcospan, a gene in the locus.
Sarcospan (SSPN) is a transmembrane protein in the dystrophin-glycoprotein complex. The complex links the extracellular matrix to the membrane cytoskeleton of myocytes. Sspn is specifically expressed in the heart and skeletal muscle 16,17 . Mutations of dystrophin-glycoprotein complex genes cause muscular dystrophy. Mutations are not usually associated with congenital heart disease, although an alpha-dystrobrevin mutation in one family was associated with VSD, hypoplastic left heart, and left ventricular noncompaction 18 . C57BL/6N, which bears the risk allele at the chromosome 6 locus, carries a missense mutation of a highly conserved glutamine at position 213 of SSPN. The amino acid is in the cytoplasmic C-terminal domain (Fig. 1a). Sspn knockout (Sspn KO ) mice have normal skeletal muscle function. The null mutant is obtained at the expected Mendelian frequency 19 , so Sspn deficiency alone would not be expected to cause serious congenital heart defects.

Results
Identification of Sarcospan as a candidate modifier gene for muscular VSD in Nkx2-5 +/− mice. We sequenced the Sspn alleles carried by the parental strains in our mapping cross. We confirmed that FVB/N carries the conserved glutamine (Q) and C57BL/6N carries a leucine (L) at position 213 ( Fig. 1b and Supplementary Fig. 1). The function of the conserved glutamine at position 213, which aligns to position 240 in humans, is unknown, but a survey of human and other vertebrate species suggests that the amino acid was strongly selected at the divergence of lobe-and ray-finned fishes. All 65,777 humans in the 1000 Genomes Project 20 , NHLBI GO Exome Sequencing Project 21 , and the Exome Aggregation Consortium 22 and nearly all Sarcopterygii species, e.g., coelacanth, lungfish and tetrapods, carry a glutamine at the position mutated in C57BL/6 ( Fig. 1b and Supplementary Fig. 1). The equivalent Q240L mutation in human SSPN has a scaled CADD score of 12.4, placing it in the top 10% of potential variants ranked by predicted deleteriousness 23 . Actinopterygii, i.e., ray-finned fishes, all carry a tryptophan (W) at the same position ( Fig. 1b and Supplementary Fig. 1). Q240W has a scaled CADD score of 25, placing it in the top 1% of predicted deleterious variants 23 .
Some chromatin immunoprecipitation data suggest that NKX2-5 binds to potential Sspn regulatory elements 24,25 , but we found no evidence for a regulatory polymorphism that affects expression. Sspn mRNA levels are the same in the hearts of C57BL/6N, FVB/N and C57BL/6N X FVB/N F1 hybrid mice. Wild-type and Nkx2-5 +/− mice also express Sspn at the same level (Fig. 1c).
The Sspn null mutation increases the incidence of muscular VSD in Nkx2-5 +/− mice. We next investigated whether Sspn genetically interacts with Nkx2-5 in the development of the muscular ventricular septum. Double heterozygotes, i.e., Nkx2-5 +/− /Sspn +/− , were crossed to each other or Sspn +/− single mutants. The Nkx2-5 null mutation is embryonic lethal 26,27 , but all other combinations of Nkx2-5 and Sspn genotypes were recovered at the expected Mendelian frequency in newborn mouse pups ( Supplementary Fig. 2). The co-occurrence of Nkx2-5 +/− and a heterozygous or homozygous Sspn knockout mutation does not affect embryonic survival.
The null mutation of Sspn (Sspn KO ) does not cause congenital heart defects in Nkx2-5 wild-type (Nkx2-5 WT ) mice (Fig. 2). The incidences of each type of septal defect among the Sspn wild-type (Sspn WT ) and Sspn +/− mice in either the Nkx2-5 +/− or Nkx2-5 WT groups were the same, so their results were combined for statistical analyses.
As expected, Nkx2-5 +/− mice have an increased incidence of atrial septal defects (ASDs), membranous VSDs and muscular VSDs compared to the Nkx2-5 WT . The Sspn KO genotype does not affect the incidences of ASD or membranous VSD in the Nkx2-5 +/− mice, but Nkx2-5 +/− /Sspn KO double mutants have almost twice the incidence of muscular VSD compared to Nkx2-5 +/− single mutants (Fig. 2). The results place Sspn in a genetic pathway that leads from Nkx2-5 mutation to a defect in the muscular ventricular septum.
Incidentally, a low incidence of muscular VSD was observed in Nkx2-5 WT hearts. In a previous study of Nkx2-5 WT mice in the C57BL/6N strain, which is also the predominant background of the mice in this study, we did not detect the < 5% incidence probably because of a smaller sample size 14 . The prior study did note, however, a correlation between muscular VSD incidence in Nkx2-5 +/− mice and the fraction of C57BL/6N genome in the genetic background of inbred strain crosses with FVB/N and A/J. C57BL/6N may carry genetic polymorphisms that increase the propensity toward muscular VSD in Nkx2-5 WT hearts as well.
The Sspn null mutation disrupts the organization of myocytes in the ventricular septum of Nkx2-5 +/− mice. The inspection of thousands of hearts in several projects led us to notice that the myofibers around a muscular VSD veer from their normal trajectories. For example, in a heart with an intact ventricular septum the subendocardial fibers on the left side of the septum tend to run in parallel from the apex to the base (Fig. 3a). In a heart with a muscular VSD, the left-sided, subendocardial fibers veer toward the right in sections that are near the defect (Fig. 3b). Fibers that surround the defect appear even more malaligned (Fig. 3c). The malalignment could either be a simple consequence of myofiber diversion around a defect or a sign of flawed trabecular coalescence, which could be the basis of a muscular VSD. To distinguish the two possibilities, we cut 1 μ m, thin sections of plastic-embedded hearts. The method preserves the myofiber architecture better than paraffin embedding, which is more amenable to sectioning large number of hearts quickly.
The orientation of myofibers in the normal heart follows a stereotypical pattern 28 . In the left ventricular free wall, subepicardial fibers run with a major component parallel to the long axis from apex to base, while fibers in the middle layer run circumferentially around the ventricle. The left ventricular free wall in the four genotypic groups -wild type, Nkx2-5 +/− and Sspn KO single mutants, and Nkx2-5 +/− /Sspn KO double mutant -showed this pattern ( Supplementary Fig. 3).
In the ventricular septum, the subendocardial fibers run in parallel along the left ventricular outflow tract. Subendocardial fibers on the right ventricular side do not have as consistent a pattern, perhaps owing to the shape of the right ventricle relative to the plane of section. Fibers in the middle layer of the septum run circumferentially. This pattern was observed in the intact ventricular septum of every wildtype and Nkx2-5 +/− and Sspn KO single mutant heart (Fig. 4). In contrast, each of the Nkx2-5 +/− /Sspn KO double mutants examined had a subset of  subendocardial fibers on the left side of the ventricular septum that deviated toward the right, sometimes forming whorls, even though there was no VSD (Fig. 4 and Supplementary Fig. 4). The incidence of abnormal myofiber organization in the Nkx2-5 +/− /Sspn KO double mutant is significantly greater than in the other three genotypic groups (P < 0.001, two-tailed Fisher exact test). An abnormal course of myofibers in the ventricular septum is not simply the consequence of a muscular defect. Collectively, the results indicate that Nkx2-5 and Sspn operate together in a pathway that affects the alignment of myofibers in the ventricular septum.

Discussion
Dozens of genetic mutations are known to cause muscular VSDs, but how the heart defect forms is poorly understood 29,30 . A novel genetic interaction between Nkx2-5 and Sspn offers insight into the mechanism. NKX2-5 regulates diverse cardiac developmental pathways 13 . Polymorphisms of modifier genes in these pathways influence their susceptibility to Nkx2-5 haploinsufficiency and the expression of the mutant phenotype 14 . Sspn is a gene that resides in a chromosome 6 locus that modifies the risk of muscular VSD in Nkx2-5 +/− mice 15 . The present data indicate that Sspn is an Nkx2-5 modifier gene, although we cannot state definitively whether it is the gene that underlies the chromosome 6 locus. Complete loss of Sspn function does not cause heart defects but does increase the incidence of muscular VSD in combination with an Nkx2-5 mutation. In addition, myofibers in the ventricular septum of Nkx2-5 +/− /Sspn KO double mutants commonly deviate from their normal trajectories. Myofibers similarly deviate around muscular VSDs, as noticed by us and another group 31 . We propose that the malaligned septal myofibers are a sign of a defect in trabecular coalescence that causes muscular VSDs.
The development of the muscular ventricular septum begins with the coalescence of trabeculations in the interventricular groove of the looping heart tube 6,7 . Lineage analyses in the mouse demonstrate that left and right ventricular myocytes form their respective sides of the septum. A sharp, left-right boundary delineates their contributions to the septum 32 . Myofibers that run from left to the right side of the ventricular septum appear to violate this boundary. Certain defects of trabecular development could disrupt their coalescence into the ventricular septum, thus causing the malalignment of myofibers or a muscular VSD. For example, cardiomyocyte-specific deletion of Scrib, a planar cell polarity gene, causes the stunting of trabeculae. Scrib localizes the GTPase signaling protein Rac1 to the cell membrane. Their interaction plays a critical role in the organization of trabecular myocytes, the apposition of myocytes in the ventricular septum, and the regulation of junctional complex formation at the cell membrane 12 . A quantitative defect in these processes could cause a trabecula to become malaligned if it has no neighboring trabeculae to support its normal coalescence. A muscular VSD could result when malaligned trabeculae do not come into apposition.
Interestingly, there may be subtle, quantitative differences in trabecular developmental processes even between wild-type animals. Fractal analyses indicate that the morphologic pattern of the trabecular network is more complex in the inbred C57BL/6 strain compared to the outbred NIMR:Parkes 8 . This complex, but quantifiable trabecular phenotype may be related to the low incidence of muscular VSDs in Nkx2-5 WT mice from a predominantly C57BL/6N background. Fractal analyses of trabeculae, as described by Captur et al. 8,33 , in mutants prone to muscular VSD could help to illuminate patterns of trabecular coalescence during cardiac development and the functional significance of myofiber malalignment.
The molecular and cellular bases of the genetic interaction between Nkx2-5 and Sspn remain to be defined. The hypertrabeculation phenotypes associated with human and mouse mutations indicate that NKX2-5 plays a role in trabecular development 34,35 . The present results indicate that loss of Sspn exacerbates a subtle defect associated with Nkx2-5 haploinsufficiency. In light of what is known about dystrophin-associated complex genes in skeletal muscle and other organisms, one can consider potential physical or biological functions of Sspn in ventricular septal development. The dystrophin-glycoprotein complex is best known for its role in maintaining the integrity of the muscle cell membrane and transducing force 36 . Mutations that cause muscular dystrophy are not associated with congenital heart defects, so a purely mechanical function of Sspn in ventricular septal development seems less likely. The dystrophin-glycoprotein complex also plays a role in signaling. SSPN promotes Akt signaling during muscle regeneration, and the dystrophin-glycoprotein complex modulates integrin signaling [37][38][39] . Elegant studies in Drosophila clearly demonstrate the functions of the dystrophin-associated complex in regulating planar cell polarity 40,41 , BMP and Notch signaling 42,43 . All these pathways are all involved in the formation of the ventricular myocardium 11 , and NKX2-5 has been shown to regulate BMP and Notch signaling 44,45 .
The function of amino acid Q213 and the effect of the Q213L mutation on Sspn function are unknown. The cytoplasmic C-terminal domain of SSPN, which includes Q213, contributes to protein-protein interactions that stabilize the dystrophin-glycoprotein complex 46 . The amino acid could function in this capacity or in a different protein-protein interaction. Interestingly, the conserved glutamine arose in the Sarcopterygii lineage in association with the onset of the development of a ventricular septum. The African lungfish, an early Sarcopterygii species, has a rudimentary ventricular septum 47 , whereas ray-finned fishes do not. Sspn is not essential for the formation of a ventricular septum, but the evolution of the ventricular septum undoubtedly involved the selection of polymorphisms present in the lobe-finned fishes.
Much remains to be learned regarding how mutations of genes like Nkx2-5 and their interactions with modifier genes cause muscular VSDs. Although muscular VSDs are usually benign, the investigation of their pathogenesis should illuminate mechanisms pertinent to congenital heart disease, cardiomyopathy and cardiac tissue engineering.

Methods
Mouse strains. Sspn knockout mice were purchased from the Jackson Laboratory. They were predominantly in a C57BL/6 background. The Sspn knockout was made in 129S1/Sv x 129X1/SvJ F1 ES cells and backcrossed to C57BL/6 for five generations prior to arrival at Jackson. The Sspn wild-type and knockout alleles were genotyped as described 19 . Nkx2-5 +/− mice were maintained in the C57BL/6N background and genotyped as previously described 14,27 . The C57BL/6N and FVB/N inbred strains were purchased from Charles River. The experiments were approved by the animal studies committee at Washington University School of Medicine and performed in accordance with their guidelines. The reference sequences of SSPN from multiple species were downloaded from Ensembl and NCBI on April 15, 2015. Sequences that lacked the C-terminus or the expected stop codon or had large gaps (≥ 10 amino acids) were excluded from multi-species alignment analysis. The amino acid sequences of 48 species were aligned in Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo/) 48 . The physiochemical conservation scores at each amino acid position was calculated and visualized in Jalview (http://www.jalview.org/) 49 . The transmembrane protein structure of mouse SSPN was predicted with the TMHMM Server v. 2.0 (http://www.cbs.dtu.dk/services/ TMHMM/) and drawn with TMRPres2D (http://bioinformatics.biol.uoa.gr/TMRPres2D/) 50,51 .

Sarcospan sequence analyses. SNPs among inbred mouse strains in
The deleteriousness of an amino acid substitution was estimated on the human SSPN sequence by the Combined Annotation-Dependent Depletion (CADD) method 23 . Scaled CADD scores for codon substitutions at human Q240 of SSPN were obtained from the CADD website http://cadd.gs.washington.edu/home. Quantitative RT-PCR was performed using SYBR GreenER qPCR SuperMix (Invitrogen) on an Mx3005 P qPCR system (Stratagene). Sspn expression levels were normalized to Gapdh.

Quantification of
Phenotyping of congenital heart defects. Nkx2-5 +/− mice were crossed to Sspn +/− or Sspn KO mice to generate double heterozygote F1 offspring, i.e., Nkx2-5 +/− /Sspn +/− , as well as the other genotypes. Nkx2-5 +/− /Sspn +/− mice were then crossed to other double heterozygotes or Sspn +/− single mutants. The F2 offspring were collected within hours of birth. Their hearts were dissected, fixed in 10% formalin and processed for histology. The hearts were sectioned completely at 6 μ m thickness and stained with hematoxylin and eosin. All the sections were examined for congenital heart defects without knowledge of genotype, as previously described 14 . Assessment of cardiac myofiber orientation. Newborn mouse hearts were fixed in modified Karnovsky's fixative (2% paraformaldehyde, 2.5% glutaraldehyde, 0.1 M cacodylate buffer, pH 7.4) and osmium tetroxide. The hearts were embedded in resin. Serial sections through the entire heart were cut at 1 μ m thickness in the frontal plane and stained with toluidine blue. Sections were imaged with a Nanozoomer 2.0-HT digital slide scanner (Hamamatsu) at 20x magnification.
The orientation of myofibers was analyzed using ImageJ (http://fiji.sc) and an ImageJ plugin, OrientationJ (http://bigwww.epfl.ch/demo/orientation/) 52 . OrientationJ imposes a square grid on an image. The dominant orientation of pixels within each square is depicted as a vector. The analyses were performed by two individuals blinded to genotype.
Statistical analyses. The incidences of a heart defect in two groups were compared in a 2 × 2 contingency table. Statistical significance was assessed by a chi-squared test.
The incidences of abnormal myofiber orientation were compared in a 2 × 4 contingency table. Statistical significance was assessed by a two-tailed, Fisher exact test.