Novel pathogenic variants in filamin C identified in pediatric restrictive cardiomyopathy

2.1 Ascertainment of familial RCM and ethical considerations

Patients underwent clinical evaluation by a pediatric geneticist and pediatric cardiologist at Cincinnati Children's Hospital Medical Center (CCHMC) or Children's Hospital of Wisconsin. Patients and family members enrolled in a research study approved by the Institutional Review Boards (IRB) at CCHMC or Indiana University. Written informed consent for participation in this study, as well as publication of clinical data of the affected individuals was obtained from the families. All the methods applied in this study conformed to the Declaration of Helsinki (1964) of the World Medical Association concerning human material/data and experimentation. Genomic DNA of six individuals from Family 1 (Figure 1a) and five individuals from Family 2 (Figure 2a) was extracted from whole peripheral blood leukocytes following a standard DNA extraction protocol (Promega, Madison, WI).

2.2 Exome sequencing

For Family 1, exome sequencing (ES) was performed on a research basis. Genomic DNA (2 μg) from proband IV-2 (Figure 1a) was used for enrichment of human exonic sequences with the NimbleGen SeqCap EZ Human Exome v2.0 Library (2.1 million DNA probes; Roche NimbleGen, Inc., Madison, WI). Sequencing was performed with 50 bp paired-end reads using Illumina (San Diego, CA) GAII (v2 Chemistry) to a mean depth of approximately 56×. All sequence reads were mapped to the reference human genome (UCSC hg 19) using the Illumina Pipeline software version 1.5 featuring a gapped aligner (ELAND v2). Variant identification was performed using GATK (McKenna et al., 2010). Amino acid changes were identified by comparison to the UCSC RefSeq database track. A local CCHMC realignment tool was used to minimize the errors in SNP calling due to indels. Exome data were initially screened for more than 300 genes known for CM or related phenotypes. Our filtering strategy removed previously reported variants in databases such as dbSNP (Sherry et al., 2001) and NHLBI Exome Sequencing Project (ESP) (https://evs.gs.washington.edu/EVS/), in addition to non-coding variants. Remaining variants were tested for segregation with disease in the family, and further subject to pathogenicity prediction programs including: Polyphen-2, SIFT/PROVEAN, MutationTaster, and HOPE (Adzhubei et al., 2010; Choi, Sims, Murphy, Miller, & Chan, 2012; Schwarz, Cooper, Schuelke, & Seelow, 2014; Venselaar, Te Beek, Kuipers, Hekkelman, & Vriend, 2010).

For Family 2, DNA samples were sent to the Advanced Genomics Laboratory at the Medical College of Wisconsin (MCW) for clinical ES of the parent and proband trio. In brief, after enriching for coding exons with solution based hybridization, an Illumina HiSeq 2500 was used for sequencing. Analysis of depth of coverage across exomes was performed with the GapMine software (MCW) with a minimum average depth of 85×. Sequencing variants were interpreted with the Carpe Novo software package (MCW). For this clinical pipeline, all identified variants were assessed for potential clinical significance. In addition to filtering variants based on quality and allele frequency, Carpe Novo uses multiple algorithms and databases to assess variants including bioinformatics predictions on protein function and structure (Polyphen-2 and SIFT), evolutionary conservation at the affected residues, changes to known splice site donors or acceptors, and presence or absence of the variant in aggregate databases such as dbSNP, HGMD, or OMIM (Stenson et al., 2003; Online Mendelian Inheritance in Man, OMIM^®, https://omim.org/).

2.3 Variant screening and validation

Primers (Supp. Table S1) were designed covering exonic regions containing potential candidate variants (FLNC, MYPN, and PKP2 in Family 1 and FLNC and SYNE2 in Family 2). PCR products were sequenced on an ABI (Foster City, CA) 3730XL Genetic Analyzer using BigDye Terminator (ABI). Sequence analysis was performed via FinchTV 1.4.0 (https://www.geospiza.com). All positive findings were confirmed in a separate experiment using the original genomic DNA sample as template for new amplification and bi-directional sequencing reactions. The FLNC p.Pro2298Leu and FLNC p.Tyr2563Cys variants (Refseq NM_001458.4) were determined to be pathogenic according to the American College of Medical Genetics and Genomics (ACMG) 2015 guidelines for variant interpretation (Richards et al., 2015) based on data presented here (see section Results). FLNC variants were submitted to ClinVar and have been assigned the following ClinVar accession numbers: SCV000787752 (p.Pro22298Leu) and SCV000787753 (p.Tyr2563Cys).

2.4 In silico protein homology modeling and viewing

SWISS-MODEL (Biasini et al., 2014) modeled the Ig domains 20–21 of FLNC to depict amino acid position 2298. The amino acid sequence of these domains was queried, and the template model most closely matching FLNC sequence was selected. This model was based on the known crystal structure of FLNA's Ig 20–21 domains (PDB 4P3W). A known crystal structure for FLNC Ig23 (PDB 2NQC) served as model to view amino acid position 2563. YASARA (Krieger & Vriend, 2014) was used to view both protein structures.

2.5 Cell culture, plasmid transfection, and immunocytochemistry

A human FLNC clone was purchased from Origene (Rockville, MD, catalog number RG212462) in a mammalian expression vector pCMV-AC-GFP. Mutagenesis was performed using primers designed (Supp. Table S2) for both the Q5 Site-Directed Mutagenesis Kit (New England Biolabs, Ipswich, MA) and the Infusion HD cloning kit (Clontech, Mountain View, CA). Validations of full-length FLNC pCMV-AC-GFP plasmids with wild-type and mutant transcripts were carried out using primers listed in Supp. Table S3. Both a standard calcium-phosphate method and Neon electroporation (Invitrogen, Waltham, MA) were used for transfections. For calcium-phosphate transfections, C2C12 cells (American Type Culture Collection, Manassas, VA; CRL-1772) were cultured with 2 ml Dulbecco's modified Eagle medium (DMEM) containing 10% fetal bovine serum (Hyclone, Pittsburgh, PA) and 1% penicillin-streptomyocyin (Gibco, Waltham, MA) in a 2 well (4.0 cm²) Lab-Tek chamber slide (Nunc, Rochester, NY) to reach 60% confluence on the day prior to transfection. To transfect, 5 μg of pCMV-AC-GFP FLNC wild-type or mutant plasmids were combined with 8 μl of 2 M CaCl₂. This mix was added dropwise to 62.5 μl 2X HEPES buffered saline. After 30 min of precipitation, the combined solution was added to the culture media and incubated overnight. After incubation, the transfection media was replaced with 2 ml fresh media. For Neon electroporation, 100,000 C2C12 cells and 0.75 μg of plasmid DNA were used in each transfection using electroporation conditions of 1,650 V, 10 ms pulse width, and pulse number 3. Cells were plated into two well chamber slides with 2 ml of the above described DMEM.

Forty-eight to 72 hr post transfection, cells were prepared for immunocytochemistry. All steps were carried out at room temperature. Cells were washed with phosphate-buffered saline (PBS) three times for 15 min, fixed with 4% paraformaldehyde for 30 min followed by repeated PBS washes, and then permeabilized with 1% Triton-X100 (Fisher Scientific, Waltham, MA) for 30 min. Following permeabilization, blocking was carried out with 5% non-fat dried milk and 1% goat serum (Invitrogen) in 0.1% Tween-20 in PBS for 60 min. The following primary antibodies were incubated with cells for 3 hr: desmin (polyclonal rabbit α-mouse 1:200; Abcam, Cambridge, MA; ab8592) and FLNC (polyclonal goat α-mouse 1:500; Santa Cruz Biotechnology, Dallas, TX; sc-48496). Cells were washed with PBS and then incubated for 60 minutes in the following secondary antibodies, all at 1:100 dilutions: goat α-rabbit Alexa Fluor 568 (ThermoFisher, Waltham, MA); donkey α-goat Alexa Fluor 594 (ThermoFisher); and F-actin was detected with the Phalloidin-conjugated Alexa Fluor 594 (ThermoFisher). The slides were then rinsed in PBS buffer for 15 min, mounted with Vectashield containing DAPI for nuclei counterstain (Vector Laboratories, Burlingame, CA), and examined under a fluorescence (Nikon Eclipse E400, Melville, NY) or confocal microscope (Carl Zeiss Apotome, Thornwood, NY).

2.6 Pathology

Explanted heart tissue was fixed in 10% neutral buffered formalin for 24 hr at room temperature before embedding in paraffin (FFPE), sectioning, and staining with hematoxylin and eosin (H&E) or trichrome. Additionally, a portion of the ventricle was fixed in 3% glutaraldehyde in cacodylate buffer, and sectioned to one cubic millimeter for processing and imaging by transmission electron microscopy.

3.1 Phenotypic evaluation

Family 1 showed a typical autosomal dominant mode of inheritance (Figure 1a), and all affected members presented with isolated RCM diagnosed at early ages. The family's race is African American. The proband was given a diagnosis of RCM at age 8, and upon echocardiogram was found to have an indexed left atrial volume of 72 ml/m², right atrial enlargement, and left ventricular diastolic dysfunction (Figure 1b). Cardiac catheterization confirmed the diagnosis of RCM (Rindler, Hinton, Salomonis, & Ware, 2017). He underwent transplant at the age of 11 years. The patient's sister was followed by echocardiography due to the family history of RCM and first showed evidence of biatrial enlargement at the age of 4 years. She underwent cardiac catheterization and was given a diagnosis of RCM at the age of 5 years and was listed for transplant which occurred at the age of 7 years. The paternal aunt was diagnosed with RCM at the age of 3 years and the paternal half-sister was diagnosed under the age of 10 years. All of the affected individuals presented disease symptoms without evidence of myopathy or other extra-cardiac findings. Clinical genetic testing for 10 genes (MYBPC3, MYH7, ACTC1, DES, MYL2, MYL3, TNNI3, TNNT2, TPM1, and EMD) was performed as part of the clinical diagnostic evaluation and no sequence variants were reported by the clinical testing laboratory.

The proband in Family 2 is a 1-year-old Caucasian female who presented with acute onset of unilateral weakness and was found to have a left sided thromboembolic middle cerebral artery stroke. Evaluation for etiology revealed classic phenotypic changes consistent with RCM including normal ventricular dimension and systolic function with massively dilated atria (Figure 2b). Of note, the child had an echocardiogram in the perinatal period that was normal aside from a small muscular ventral septal defect with restrictive left to right shunt. Due to elevated pulmonary vascular resistance, she was listed for heart transplant shortly after diagnosis and successfully transplanted months later. Family echocardiographic screening revealed that her twin sister had similar phenotypic findings of RCM while her brother, mother, and father were all unaffected (Figure 2a). The twin sister progressed over months to require heart transplant listing and was successfully transplanted as well. Of note, each twin also presented with extra-cardiac phenotypes at the time of RCM diagnosis, including unspecified developmental delay, hypotonia, dysmorphic facial features, and clasped thumbs. Further, they each demonstrated kidney dysfunction of unclear etiology by three years of age, post heart transplant.

3.2 Pathology

Explanted heart tissue from the proband in Family 1 was processed for both histology and transmission electron microscopy (EM). H&E staining revealed some myocyte hypertrophy, nuclear size changes, and occasional central nuclei (Figure 3a), whereas trichrome staining at lower magnification showed some differences in fiber sizes and lacy interstitial fibrosis (Figure 3b). EM of the explanted tissue showed essentially normal ultrastructure; there were no lipid or glycogen deposits. Relevant for FLNC-related myopathies, there were no inclusions or depositions observed. Mitochondrial organization and sarcomere structure (Figure 3c,d) appeared normal and regular with no disarray of the type frequently seen in patients with MYH7 or MYBPC3 variants (Wessels et al., 2015; Witjas-Paalberends et al., 2013). The proband and twin sister from Family 2 also had explanted heart tissue processed for EM. Both proband (Figure 3e) and sister (Figure 3f) also demonstrate regular sarcomeric structure, with no evidence of myofibril degeneration or signs of irregular mitochondria organization, aggregates, or depositions.

3.3 Exome analysis

After variant filtering, we identified 3 candidate variants in the proband in Family 1. These included variants in MYPN (p.Ser1282Arg), PKP2 (p.Val798Ile), and FLNC (p.Pro2298Leu). These sequence changes were tested in additional family members and only the FLNC missense variant p.Pro2298Leu (c.6893C>T; NM_001458.4) cosegregated with the RCM phenotype in the family, supporting a dominant mode of inheritance (Figure 1c). The p.Pro2298Leu variant is located in exon 41 of FLNC and falls in the immunoglobulin-like (Ig) 20 (a.a. 2244–2306) domain of the protein. This variant is not present in the Exome Aggregation Consortium (ExAC) (Lek et al., 2016). Additionally, the variant is located outside the region of homology that transcribed FLNC shares with the identified FLNC-like pseudogene (Odgerel, van der Ven, Fürst, & Goldfarb, 2010).

Clinical ES of the trio and filtering strategies were used to identify potential disease causing variants in the proband of Family 2. Ultimately, three variants in two genes, FLNC and SYNE2 (a gene associated with autosomal dominant Emery-Dreifuss muscular dystrophy), were filtered as being potentially clinically important. Each SYNE2 variant (p.Val1688Ile and p.Lys4931Arg) was inherited from a healthy parent (Supp. Figure S1). Additionally, these variants were predicted to be benign by pathogenicity prediction, were found in variant databases such as ExAC and dbSNP, and neither twin presented classical signs of Emery-Dreifuss by two years of age. The identified FLNC variant p.Tyr2563Cys (c.7688A>G; NM_001458.4) was present in both twins and absent in unaffected parents and sibling (Figure 2c) indicating it was de novo. It was not found in ExAC or dbSNP, is located in Ig domain 23, and was confirmed to be present in the functional FLNC gene and not its pseudogene.

3.4 In silico pathogenicity prediction

Both the autosomal dominant p.Pro2298Leu and de novo p.Tyr2563Cys variants in FLNC were unanimously predicted to be pathogenic using bioinformatic programs Polyphen-2, SIFT/PROVEAN, and Mutation Taster (Supp. Table S4). Further, the program HOPE (Venselaar et al., 2010) was used to discover and analyze potential changes to protein structure resulting from each variant (Supp. Table S4). Proline at position 2298 of FLNC, as well as its respective triplet codon (CCC) in the gene, are evolutionary conserved across species suggesting an important role of this residue in protein function (Figure 1d). Prolines are known to be very rigid and therefore induce a special backbone conformation which might be required at this position. All filamins consist of an N-terminus actin-binding domain followed by 24 antiparallel β-sheet repeats (called filamin or Ig-like domains), the 24^th of which is responsible for filamin homo-dimerization at the C terminus (Stossel et al., 2001). These 24 Ig filamin repeats of the FLNC protein primarily form a secondary structure of beta strands, as seen in a model of Ig domains 20–21 (Figure 4a) (Sjekloća et al., 2007). The side chain of proline will typically disrupt β-strands due to its unique ring structure (Li, Goto, Williams, & Deber, 1996), and indeed proline 2298 falls in a loop between two such strands (Figure 4b). The change to leucine at this position could disturb this special conformation and lead to protein aggregates. β-rich proteins like FLNC are prone to aggregation, and a leucine substitution could have a higher propensity to form hydrogen bonds between other nearby β-strands which could potentially disrupt protein folding and lead to aggregate formation (Hecht, 1994; Venselaar et al., 2010).

Tyrosine at position 2563 is also evolutionarily conserved across species (Figure 2d), and is located at the end of a β-strand (Figure 4c,d). The mutant cysteine residue is less hydrophobic and smaller than the wild-type tyrosine, and due to its size is predicted to cause an empty space in the protein. The difference in hydrophobicity is also predicted to lead to a loss of hydrogen bonds at the core, which could disturb proper protein folding (Venselaar et al., 2010). In addition to changes to hydrophobicity, the specific tyrosine to cysteine substitution may also be contributing to p.Tyr2563Cys protein aggregation (shown in results below). Substitution of tyrosine and cysteine at key resides of the human α-synuclein protein has been shown to cause accelerated protein aggregation in vitro, likely due to the increased ability for cysteine residues to cross-link and form stable dimers, which could be more prone to polymerization and aggregation (Zhou & Freed, 2004).

3.5 In vitro analysis of FLNC mutants

Previous studies of filaminopathies have shown evidence of mutant FLNC intracellular protein aggregation in both immuno-histological sections of patient muscle and in in vitro transfection of mutant FLNC constructs (Kley et al., 2007; Valdés-Mas et al., 2014; Vorgerd et al., 2005). These aggregates often contain additional Z-disk associated proteins such as desmin, a hallmark of myofibrillar myopathy. In order to determine whether the p.Pro2298Leu and p.Tyr2563Cys variants would also lead to intracellular protein aggregation, full-length wild-type and mutant FLNC-GFP constructs were transiently expressed in C2C12 mouse myoblasts.

Cells expressing the wild-type FLNC construct exhibited diffuse homogeneous expression of the protein (Figure 5b,n), whereas the p.Pro2298Leu construct yielded aggregates (Figure 5f,r), often near the nucleus. Additionally, the number of cells expressing p.Pro2298Leu mutant protein appeared considerably lower than those expressing the wild-type FLNC construct. Cells expressing the mutant p.Tyr2563Cys construct also demonstrated areas of FLNC aggregation (Figure 5j,v), though these aggregates were punctate and varied in size from small to intermediate. Unlike the results seen with the p.Pro2298Leu construct, the diffuse FLNC expression was not lost in these cells; rather, the p.Tyr2563C aggregates appeared to be randomly distributed throughout otherwise normal protein expression. The pattern of aggregation observed for our p.Tyr2563Cys construct most closely mirrors what has been observed in cells transfected with FLNC variants causing distal myopathy (Duff et al., 2011). Additionally, unlike FLNC variants associated with myofibrillar myopathy, we did not observe the presence of desmin aggregation with mutant FLNC aggregates (Figure 5s,w) or endogenous Flnc co-localizing to exogenous FLNC aggregates (Figure 5g,k).

Given the importance of FLNC as an actin-binding protein, and in vitro data suggesting that FLNC variants may cause actin-aggregation (Kley et al., 2012), we transiently transfected p.Pro2298Leu and p.Tyr2563Cys constructs into C2C12 cells and stained for F-actin. As previously seen, the wild-type FLNC construct showed diffuse cellular expression (Figure 6a–d). The p.Pro2298Leu mutant again resulted in intracellular FLNC aggregates, but these cells also had actin aggregates which appear to co-localize with mutant protein (Figure 6e–h). Taken together, these results suggest the p.Pro2298Leu variant results in an unstable, mutant protein which is prone to aggregation and may contribute to the overall weakening of myofibers as evidenced by actin aggregation. Cells transfected with the p.Tyr2563Cys construct also contained aggregates of both FLNC and F-actin (Figure 6i–l), which again appeared more punctate, and varied in size. Overall, these data suggest the p.Tyr2563Cys is a potentially damaging variant, as it results in aggregates containing both FLNC and F-actin.