Disease-associated mutations in the actin-binding domain of filamin B cause cytoplasmic focal accumulations correlating with disease severity


  • Raymond Dalgleish


Dominant missense mutations in FLNB, encoding the actin-cross linking protein filamin B (FLNB), cause a broad range of skeletal dysplasias with varying severity by an unknown mechanism. Here these FLNB mutations are shown to cluster in exons encoding the actin-binding domain (ABD) and filamin repeats surrounding the flexible hinge 1 region of the FLNB rod domain. Despite being positioned in domains that bind actin, it is unknown if these mutations perturb cytoskeletal structure. Expression of several full-length FLNB constructs containing ABD mutations resulted in the appearance of actin-containing cytoplasmic focal accumulations of the substituted protein to a degree that was correlated with the severity of the associated phenotypes. In contrast, study of mutations leading to substitutions in the FLNB rod domain that result in the same phenotypes as ABD mutations demonstrated that with only one exception disease-associated substitutions, surrounding hinge 1 demonstrated no tendency to form actin-filamin foci. The exception, a substitution in filamin repeat 6, lies within a region previously implicated in filamin-actin binding. These data are consistent with mutations in the ABD conferring enhanced actin-binding activity but suggest that substitutions affecting repeats near the flexible hinge region of FLNB precipitate the same phenotypes through a different mechanism. Hum Mutat 33:665–673, 2012. © 2012 Wiley Periodicals, Inc.


Filamin proteins interact with the cytoskeleton by binding to filamentous actin (F-actin), and have the potential, through dimerisation, to crosslink actin fibrils Cunningham et al., 1992; Hartwig and Stossel, 1975; Niederman et al., 1983. In humans and mice, there are three filamin genes (FLNA; MIM# 300017, FLNB; MIM# 603381, and FLNC; MIM# 102565) with distinct expression patterns, but producing highly homologous proteins. In all three proteins, there is an amino terminal actin-binding domain (ABD), followed by 15 immunoglobulin-like filamin repeats (also referred to as rod domain 1), an hinge region (termed hinge 1), then eight more filamin repeats (rod domain 2), a second hinge region, and a final filamin repeat that mediates dimerisation.

Mutations in FLNB, the gene encoding filamin B (FLNB), cause a group of skeletal dysplasias, an observation that is consonant with expression of this protein in epiphyseal growth plate chondrocytes Krakow et al., 2004. Two broad groups of FLNB-related skeletal conditions can be defined on the basis of their clinical presentation and genetic etiology. Homozygosity or compound heterozygosity for null alleles results in the recessive condition spondylocarpotarsal syndrome (SCT; MIM# 272460), which features fusion of the vertebral, carpal, and tarsal bones. The phenotype observed for Flnb-null mice is consistent with the clinical features of SCT Farrington-Rock et al., 2008; Lu et al., 2007; Zheng et al., 2007; Zhou et al., 2007. A range of autosomal dominant diseases is caused by missense mutations or small in-frame deletions or insertions in FLNB. These phenotypes include atelosteogenesis I and III (AOI; MIM# 108720 and AOIII; MIM# 108721) and boomerang dysplasia (BD; MIM# 112310) that are severe disorders in which bones are either undermodeled or have failed to initiate ossification Bicknell et al., 2005; Farrington-Rock et al., 2006. Whereas AOIII is survivable in a minority of individuals, AOI and BD invariably present with in utero lethality. The least severe phenotype is Larsen syndrome (LS; MIM# 150250), which features joint dislocations and malformations of the cervical spine as well as supernumerary carpal and tarsal ossification centers Bicknell et al., 2007.

A group of skeletal disorders with some phenotypic similarities to the FLNB group of disorders has also been linked to clustered missense mutations in the paralogous gene FLNA, which encodes the protein filamin A Robertson et al., 2003. Some of these mutations occur at amino acids homologous to those causing the LS-AO-BD spectrum of conditions in FLNB. Similar to FLNB, these diseases are clinically distinct from the disease resulting from loss-of-function mutations in FLNA Robertson, 2005.

How some missense mutations in FLNA and FLNB lead to skeletal disease is uncertain. The ability to crosslink F-actin indicates a structural role in the cytoskeleton, but interactions detected with dozens of other proteins including integrins, transmembrane receptors, and transcription factors also suggest a role in scaffolding mediators of multiple other cellular processes Nakamura et al., 2011. Latterly, a role for filamins in mechanosensation has also been established with evidence accumulating that filamin A has the ability to convert biomechanical stress sensed via integrins into a chemical signal mediated by small GTPases Ehrlicher et al., 2011. The prominent skeletal phenotypic manifestations of the dominantly inherited LS-AO-BD disorders also raise the possibility that disruption of the mechanosensory properties of filamins may relate to the pathogenesis of these developmental conditions Mammoto and Ingber, 2010. Filamin-integrin interactions can be modulated by extrinsic force Ehrlicher et al., 2011; Kiema et al., 2006b; Lad et al., 2007b and deletion of hinge 1, the structure that confers flexibility to the filamin rod domain, thereby altering the mechanosensory properties of these proteins Gardel et al., 2006. It is conceivable therefore that filamin-actin binding may be physiologically modulated to achieve the same end and that gain-of-function missense mutations in FLNA Clark et al., 2009 and FLNB Sawyer et al., 2009 may impact on cellular mechanotransduction by conferring an increase in actin-filamin avidity. Relevant to these many cellular roles, actin-bound filamin transitions between many cellular compartments including the cortical cytoskeleton, actin stress fibers, and a newly recognized perinuclear structure called the actin cap Gay et al., 2011. Although some modulators of actin-filamin binding have been identified Nakamura et al., 2005, a comprehensive understanding of the in vivo regulation of actin-filamin transitions is still lacking.

The ABD of filamins comprises two calponin homology subdomains (CH1 and CH2) arranged in tandem. The CH2 is proposed to regulate F-actin binding mediated by CH1 Lorenzi and Gimona, 2008, a model supported by the observation that deletion of CH2 from ABD constructs leads to focal accumulation of these proteins bound to actin, Lorenzi and Gimona, 2008. Significantly, missense mutations in the ABD of FLNB that lead to the LS-AO-BD spectrum of conditions occur only in CH2 and not CH1 Bicknell et al., 2007; Farrington-Rock et al., 2006; Krakow et al., 2004. These mutations confer an increased avidity to the actin-filamin binding interaction Sawyer et al., 2009 providing strong evidence for dysregulation of actin-filamin interactions as the key mechanism underlying these conditions.

Here the cytoskeletal consequences of dominant missense mutations in FLNB that lead to the LS-AO-BD spectrum of conditions is studied by cataloguing the distribution of mutations that lead to these disorders over the gene and examining the effect of expressing full length GFP-labeled mutation-containing FLNB on its intracellular distribution and actin cytoskeletal architecture. We conclude that the nonrandom clustering of LS-AO-BD disease-causing mutations is indicative of at least two pathogenic mechanisms of generating these disease phenotypes; one of which relates to enhanced actin binding and bundling and the other possibly dysregulating the function of hinge 1.

Methods and Materials

Human Subjects and Mutation Detection

All subjects were ascertained by physician-initiated referral and consented to participate under an institutional protocol approved by the Otago Ethics Committee. The clinical and radiographic phenotypes were reviewed by one of us (SR) to ensure the diagnoses were correctly assigned according to previously published criteria Bicknell et al., 2007; Farrington-Rock et al., 2006. The exon and intron–exon boundaries of FLNB were amplified by polymerase chain reaction (PCR) using previously published primers Krakow et al., 2004 and subjected to denaturing high performance liquid chromatography (DHPLC) using the WAVE platform (Transgenomic, Omaha, NE) or using direct sequencing on an ABI3100 capillary sequencer as previously described Bicknell et al., 2007. Nucleotide numbering reflects cDNA numbering with +1 corresponding to the A of the ATG translation initiation codon in the reference sequence NM_001457.3. For protein numbering, the initiation codon is denoted codon 1. Mutations are referred to in the text by the inferred amino acid substitution.

Plasmid Constructs

The full-length FLNB cDNA clones used in these studies (pCMV-FLNB and pCMV-FLNB-EGFP) were assembled by PCR from M2 melanoma cell line cDNA and pCI-FLNB-EGFP van der Flier et al., 2002, in the expression vector pCR3.1(-) (Invitrogen, Carlsbad, CA). The clones are concordant with reference sequence NM_001457.3. PCR mutagenesis was performed using forward and reverse oligonucleotide primers containing the single-nucleotide alteration, paired with 5′ and 3′ external primers. Products were subcloned into pCMV-FLNB or pCMV-FLNB-EGFP using the restriction enzymes XbaI (in pCR3.1[-]) and BamHI (NT 2175). Primers and further details are available on request.

For mutations in repeat 6 (p.Gly751Arg), repeat 14 (p.Ser1602Pro), and repeat 15 (p.Pro1699Ser), subcloning was performed using the BamHI (NT 2176) and SacII (NT 6195) sites. The presence of expected mutations was confirmed by sequencing over the subcloned regions, and two or more independent clones for each mutant were compared in transfection experiments. All PCR was carried out with Pwo polymerase (Roche Diagnostics GmbH, Mannheim, Germany).


HEK293 cells were grown in DMEM + 10% FCS. For transfection, cells were seeded in 24-well plates at 20,000 cells per well on gelatin-coated coverslips. Transfection was performed the following day with 200 ng of DNA and 0.6 μl of Fugene (Roche) per well. Cells were analyzed at 48 hr post-transfection. The DsRed-monomer-actin construct (Clontech, Mountain View, CA) was used in cotransfections at 20 ng per well cotransfected with 80 ng of FLNB expression construct per well. M2 cells were grown in alpha MEM + 8% NCS and 2% FCS. For transfection, cells were seeded in 24-well plates at 50,000 cells per well on gelatin-coated coverslips. Antibiotic-free media was used for pretransfection plating. Transfection was carried out either immediately with the cells in suspension, or the following day with 400 ng of DNA and 1 μl of Lipofectamine 2000 (Invitrogen) per well. Cells were used at 48 hr post-transfection. DNA used in transfections was prepared using either maxiprep kits (Qiagen, Hilden, Germany) or alkaline lysis minipreps further purified with Qiagen PCR purification columns.

Immunodetection and Imaging

Cells on coverslips were fixed with 1% paraformaldehyde in PBS for 5 min and then rinsed once in PBS. For imaging of EGFP constructs, cells were incubated with 10μg/ml 4',6-diamidino-2-phenylindole (DAPI) for 10 min. Coverslips were rinsed in 0.1× PBS and mounted inverted on slides with Glycergel (Dako, Glostrup, Denmark). For immunocytochemistry, cells were permeabilized with PBS containing 0.2% triton X-100 and 5% bovine serum albumin (BSA) and rinsed once in PBS. Primary antibodies were applied in PBS + 1% BSA at the following concentrations: Rabbit polyclonal anti-FLNB (AB9276; Chemicon, Temecula, CA) at 1:400 dilution; mouse monoclonal anti-FLNA (MAB1678, Chemicon) at 1:1,000 dilution.

To visualize the actin cytoskeleton, Alexafluor 350-phalloidin (Molecular Probes, Eugene, OR) was applied to permeabilized BSA-blocked cells at a concentration of 1 unit in 250 μl of PBS + 1% BSA for 30 min. Cells were incubated 30 min and then rinsed twice for 3 min with PBS. Secondary antisera (goat anti-rabbit IgG-Alexafluor594 conjugate or goat antimouse IgG Alexafluor488 conjugate, (Molecular Probes)) were applied at 1:400 dilution in PBS +1% BSA for 30 min followed with two rinses of PBS, 3 min each. Cells were stained with DAPI and mounted as described above. Analysis and image capture was performed on a Zeiss Axioplan microscope fitted with a 1.39 megapixel CCD camera (Diagnostic Instruments, Sterling Heights, MI). Images were captured and processed with Spot 4.6 (Diagnostic Instruments) and Adobe Photoshop 10.0 software (Adobe Systems Incorporated, San Jose, CA).

Western blots were carried out with rabbit anti-GFP antisera (Invitrogen) at 1:100 dilution.


Distribution of Mutations Underlying the LS-AO-BD Spectrum of Disorders

Genomic DNA samples obtained from 67 patients with a diagnosis of LS, AOI/III, or BD had a causative mutation found by sequencing of all exons and intron–exon boundaries of FLNB. The mutations found in these individuals together with those previously reported to underlie these phenotypes Bicknell et al., 2005; Bicknell et al., 2007; Farrington-Rock et al., 2006; Krakow et al., 2004 are reported in Table 1. Criteria for determination of pathogenicity for mutations included segregation with the phenotype, a demonstration that the causative variant had arisen de novo in the context of sporadic disease in the offspring of otherwise healthy parents and alteration of evolutionarily conserved residues in the parent protein as previously described (ibid).

Table 1. Distribution of FLNB Mutations among 67 Patients Diagnosed with LS, AOI/III, or BD
DiagnosisMutationProteinNOriginProtein domainReference
  1. a

    Origin: “Inherited” if any case of parental transmission is confirmed; “De novo” if any parents of patients are confirmed noncarriers, and no inherited examples are known; “Unknown” if parental DNA has not been examined. Mutation numbering relates to the reference sequence NM_001457.3 with the +1 position referring to the A of the initator methionine codon.

Boomerang dysplasiac.512T>Gp.(Leu171Arg)1De novoCH2Bicknell et al., [2005]
Boomerang dysplasiac.605T>Cp.(Met202Thr)1UnknownCH2 
Boomerang dysplasiac.703T>Cp.(Ser235Pro)1UnknownCH2Bicknell et al., [2005]
Atelosteogenesis Ic.442T>Ap.(Trp148Arg)1UnknownCH2Farrington-Rock et al., [2006]
Atelosteogenesis Ic.447_458delp.(Gln150_Ile153del)1De novoCH2 
Atelosteogenesis Ic.495G>Tp.(Trp165Cys)1De novoCH2 
Atelosteogenesis Ic.502G>Ap.(Gly168Ser)2De novoCH2 
Atelosteogenesis Ic.503G>Tp.(Gly168Val)1De novoCH2 
Atelosteogenesis Ic.512T>Ap.(Leu171Gln)1UnknownCH2Farrington-Rock et al., [2006]
Atelosteogenesis Ic.518C>Tp.(Ala173Val)1UnknownCH2Krakow et al., [2004]
Atelosteogenesis Ic.542G>Tp.(Gly181Val)1UnknownCH2Farrington-Rock et al., [2006]
Atelosteogenesis Ic.549C>Gp.(Cys183Trp)1UnknownCH2Farrington-Rock et al., [2006]
Atelosteogenesis Ic.562T>Cp.(Ser188Pro)1De novoCH2Krakow et al., [2004]
Atelosteogenesis Ic.565T>Gp.(Trp189Gly)1UnknownCH2 
Atelosteogenesis Ic.565T>Cp.(Trp189Arg)1De novoCH2 
Atelosteogenesis Ic.604A>Gp.(Met202Val)2UnknownCH2Krakow et al., [2004]
Atelosteogenesis Ic.608A>Cp.(Gln203Pro)1UnknownCH2Farrington-Rock et al., [2006]
Atelosteogenesis Ic.613G>Ap.(Ala205Thr)1De novoCH2 
Atelosteogenesis Ic.4737_4738insC; 4746_4758delp.(Tyr1580Leu;Ile1581His; Asp1583_Gly1586del)1UnknownRepeat 14Farrington-Rock et al., [2006]
Atelosteogenesis Ic.4747_4749delp.(Asp1583del)1UnknownRepeat 14Farrington-Rock et al., [2006]
Atelosteogenesis Ic.4804T>Cp.(Ser1602Pro)1UnknownRepeat 14Farrington-Rock et al., [2006]
Atelosteogenesis Ic.5095C>Tp.(Pro1699Ser)1UnknownRepeat 15 
Atelosteogenesis IIIc.500A>Tp.(Asp167Val)1De novoCH2 
Atelosteogenesis IIIc.502G>Ap.(Gly168Ser)4Inherited (Mosaic parent)CH2Farrington-Rock et al. [2006]
Atelosteogenesis IIIc.502G>Tp.(Gly168Cys)1UnknownCH2 
Atelosteogenesis IIIc.602C>Tp.(Asp201Val)1UnknownCH2Farrington-Rock et al., [2006]
Atelosteogenesis IIIc.604A>Gp.(Met202Val)1De novoCH2Krakow et al., [2004]
Atelosteogenesis IIIc.629G>Tp.(Gly210Val)1UnknownCH2Farrington-Rock et al., [2006]
Atelosteogenesis IIIc.1087G>Ap.(Gly363Arg)1De novoRepeat 2 
Atelosteogenesis IIIc.2055+1G>Ap.(Q685_686ins9)1UnknownRepeat 5 
Atelosteogenesis IIIc.2251G>Cp.(Gly751Arg)1De novoRepeat 6Krakow et al., [2004]
Atelosteogenesis IIIc.4835G>Ap.(Gly1612Asp)1UnknownRepeat 15Farrington-Rock et al., [2006]
Atelosteogenesis IIIc.4927G>Cp.(Ala1643Pro)1UnknownRepeat 15Farrington-Rock et al., [2006]
Atelosteogenesis IIIc.5071G>Ap.(Gly1691Ser)3UnknownRepeat 15Farrington-Rock et al., [2006]
Atelosteogenesis IIIc.5074G>Ap.(Gly1692Ser)1De novoRepeat 15 
Larsen syndromec.482T>Gp.(Phe161Cys)3InheritedCH2Krakow et al., [2004]
Larsen syndromec.488A>Cp.(Gln163Pro)1UnknownCH2 
Larsen syndromec.501C>Ap.(Asp167Glu)1De novoCH2 
Larsen syndromec.500A>Gp.(Asp167Gly)1De novoCH2 
Larsen syndromec.502G>Ap.(Gly168Ser)2InheritedCH2Bicknell et al., [2007]
Larsen syndromec.572C>Tp.(Pro191Leu)1De novoCH2 
Larsen syndromec.590A>Tp.(Asn197Ile)1UnknownCH2 
Larsen syndromec.622T>Cp.(Trp208Arg)1De novoCH2 
Larsen syndromec.661A>Tp.(Ile221Phe)1UnknownCH2 
Larsen syndromec.679G>Ap.(Glu227Lys)15InheritedCH2Krakow et al., [2004]
Larsen syndromec.685T>Cp.(Ser229Pro)1UnknownCH2 
Larsen syndromec.700C>Gp.(Leu234Val)1De novoCH2Bicknell et al., [2007]
Larsen syndromec.1081G>Ap.(Gly361Ser)1De novoRepeat 2Bicknell et al., [2007]
Larsen syndromec.1081G>Tp.(Gly361Cys)1UnknownRepeat 2 
Larsen syndromec.1082G>Ap.(Gly361Asp)2UnknownRepeat 2 
Larsen syndromec.1088G>Ap.(Gly363Glu)1De novoRepeat 2Bicknell et al., [2007]
Larsen syndromec.4292T>Gp.(Leu1431Arg)1De novoRepeat 13Bicknell et al., [2007]
Larsen syndromec.4580T>Ap.(Leu1527His)1UnknownRepeat 14 
Larsen syndromec.4580T>Cp.(Leu1527Pro)1De novoRepeat 14 
Larsen syndromec.4621G>Cp.(Ala1541Pro)1De novoRepeat 14 
Larsen syndromec.4625T>Cp.(Ile1542Thr)2UnknownRepeat 14 
Larsen syndromec.4711_4713delp.(Asn1571del)1De novoRepeat 14Krakow et al., [2004]
Larsen syndromec.4725_4736delp.(Ala1577_Tyr1580del)1UnknownRepeat 14 
Larsen syndromec.4756G>Ap.(Gly1586Arg)2De novoRepeat 14Krakow et al., [2004]
Larsen syndromec.4775T>Ap.(Val1592Asp)2InheritedRepeat 14Bicknell et al., [2007]
Larsen syndromec.4781A>Cp.(Tyr1594Ser)1De novoRepeat 14 
Larsen syndromec.4795A>Tp.(Ile1599Phe)1UnknownRepeat 14 
Larsen syndromec.4805C>Ap.(Ser1602Tyr)1UnknownRepeat 14 
Larsen syndromec.4808C>Tp.(Pro1603Leu)1De novoRepeat 14 
Larsen syndromec.4936G>Ap.(Gly1646Ser)1UnknownRepeat 15 
Larsen syndromec.5023_5025delp.(Phe1675del)1UnknownRepeat 15 
Larsen syndromec.5071G>Ap.(Gly1691Ser)9UnknownRepeat 15Krakow et al., [2004]
Larsen syndromec.5071G>Tp.(Gly1691Cys)1UnknownRepeat 15 
Larsen syndromec.5072G>Ap.(Gly1691Asp)1De novoRepeat 15 
Larsen syndromec.5500G>Ap.(Gly1834Arg)2De novoRepeat 17Bicknell et al., [2007]
Larsen syndromec.5706C>Ap.(Ser1902Arg)1De novoRepeat 17 
Larsen syndromec.7290_7313delp.(Cys2431_Tyr2438del)1UnknownRepeat 23 

As previously noted for these conditions, all mutations (n = 72) were either missense mutations or small in-frame deletions that are predicted to preserve the reading frame and production of full-length protein. Figure 1A depicts the distribution of these mutations mapped onto the corresponding domains of the FLNB protein. Mutations leading to these disorders are distributed in a highly nonrandom fashion with clusters of mutations found in exons encoding the CH2 portion of the ABD and repeats 14 to 17, flanking hinge 1. Mutations leading to phenotypes spanning the full range of severity were found represented in both mutation clusters, with the exception of BD, which is caused solely by substitutions within CH2.

Figure 1.

Dominant mutations in FLNB leading to the LS-AO-BD spectrum of skeletal dysplasias. A: Schematic of FLNB indicating the sites and frequencies of mutations underlying the LS-AO-BD spectrum. Column height denotes frequency within the associated domain or subdomain (actual numbers are given above the columns), while column color denotes disease association as follows: LS (black), AOI (green), AOIII (blue), BD (red). B: Distribution of mutations in FLNB CH2 subdomain. Disease associations are color-coded as in A. Recognized α-helices are underlined and identified with a letter (the prime is added to distinguish CH2 helices from helices in CH1).

Detailed analysis of the distribution of mutations in the CH2 subdomain demonstrated that although some mutations were observed recurrently, substitutions with pathogenic outcomes were distributed widely over the entirety of CH2 (Fig. 1B). A similar analysis of mutations in FLNA that lead to a spectrum of skeletal dysplasias called the otopalatodigital syndrome spectrum disorders showed a similar distribution of mutations within the CH2 subdomain Robertson, 2005. This observation indicates that the pathogenic effect conferred by amino acid substitutions with the CH2 subdomain of these proteins can be conferred by a broad range of mutations affecting this localized region of the molecule and suggests a pathophysiological mechanism analogous to the functional inactivation of the CH2 subdomain in the absence of significant structural disruption Sawyer et al., 2009. The actin-binding activity within the ABD is known to reside with CH1 although the presence of a CH2 subdomain in tandem with this subdomain appears to modulate actin-binding activity Lorenzi and Gimona, 2008. Consequently, the effect of these naturally arising substitutions was examined by transfection and expression of constructs bearing a selection of these mutations in HEK293 and in M2 cells, a cell line that lacks endogenous FLNA expression.

BD Mutation p.Ser235Pro Results in Cytoplasmic Focal Accumulations

When full-length FLNB is expressed in HEK293 cells, it is extensively distributed within the cytoplasm (Fig. 2A, panels a and d). This distribution has the appearance of a fine meshwork, consistent with binding to the actin cytoskeleton. This is in contrast to endogenous FLNB in untransfected cells (Supp. Fig. S1), which has a less even distribution featuring distinct accumulations in perinuclear areas. The ectopic FLNB distribution resembles that of endogenous FLNA (Fig. 2A, panels b and e and Supp. Fig. S1), filling the cytoplasm and filopodial extensions. The ectopic FLNB colocalizes with, rather than displaces, FLNA.

Figure 2.

Detection of FLNB protein in transfected cells and effects of mutations on cellular localization. A: HEK293 cell line transfected with pCMV-FLNB (a-f) or pCMV-FLNBSer235Pro (g-o) analyzed for FLNB and FLNA expression by immunofluorescence. Images are of FLNB signal (red; a, d, g, j, m; gamma adjustment 2), FLNA signal (green; b, e, h, k, n; gamma adjustment 2), and both signals digitally merged with DAPI (blue) (c, f, i, l, o). Scale bars 20 μM. B: HEK293 cell line transfected with pCMV-FLNB-EGFP (a, b) or pCMV-FLNBSer235Pro-EGFP (c, d) analyzed for EGFP expression (green; gamma adjustment 2) merged with DAPI (blue). Scale bar 20 μM. C: M2 melanoma cell line transfected with pCMV-FLNB (a, b) or pCMV-FLNBSer235Pro (c, d) analyzed for FLNB expression by immunofluorescence (red; gamma adjustment 2), digitally merged with DAPI (blue). Scale bar 20 μM.

When transfected FLNB contained the mutation p.Ser235Pro, (causative of the most severe FLNB-related disease, BD), the distribution of FLNB became concentrated in globular accumulations of varying sizes (panel g and j), or in some cases, networks of thickened fibrils (panel m). These globular foci also contain FLNA (panels h, k, n and merged in i, l, and o).

Similar results were seen with FLNB-EGFP fusion proteins (Fig. 2B). FLNB-EGFP occupied the entire cytoplasm (panels a and b) whereas FLNBSer235Pro-EGFP was largely restricted to globular foci and thickened fibrils (c and d). A rare example of punctate distribution for FLNB-EGFP is seen in panel b.

Since these experiments and other work Sheen et al., 2002 have suggested colocalization and possible heterodimerisation of FLNA and FLNB, it was necessary to determine whether the presence of FLNA was required for the formation of foci. To answer this question M2, a human melanoma cell line that lacks FLNA expression, was transfected with the same constructs. In these experiments, the disparate patterns of FLNB and FLNBSer235Pro localization still occurred (Fig. 2C, panels a and b compared to panels c and d). This indicates that the foci do not occur because of FLNA displacing mutant FLNB through competition for actin binding, nor is FLNA required as a heterodimerisation partner.

FLNBSer235Pro Foci Contain Actin

The p.Ser235Pro substitution occurs within the CH2 portion of the ABD of FLNB. It is conceivable that any mutation altering the structure of the ABD may prevent the association of FLNB with the actin cytoskeleton. To address this possibility, changes in the distribution of actin in the presence of mutant FLNB were investigated. In cells cotransfected, with wild-type or mutant FLNB and DsRed-tagged actin (Fig. 3A), it can be seen that actin accumulates at the foci. Similar results are seen in cells transfected with wild-type or mutant FLNB and stained with phalloidin-Alexafluor conjugate (Fig. 3B). This observation is consistent with the hypothesis that the p.Ser235Pro mutation does not abrogate actin binding. Cytoskeletal actin was also detected outside of the foci in transfected cells, indicating that the actin cytoskeleton was not entirely disrupted. Similar results were observed in experiments that utilized M2 cells (Supp. Fig. S2), showing that FLNA was not required for inclusion of actin in the foci through heterodimer formation Sheen et al., 2002. FLNB-EGFP wild-type and mutant expression constructs were also examined for evidence of degradation by transfection and Western blot (Fig. 3C and Supp. Fig. S3). All constructs were found to express at similar levels, and no degradation products were observed, even on longer exposure. This indicates that the foci are comprised of intact FLNB.

Figure 3.

Colocalization of actin and FLNBSer235Pro. A: HEK293 cell line transfected with pCMV-FLNB or pCMV-FLNBSer235Pro and cotransfected with DsRed-actin. Scale bar 20 μM. B: HEK293 cell line transfected with pCMV-FLNB or pCMV-FLNBSer235Pro and stained with Alexafluor 350-phalloidin. Scale bar 20 μM. C. Western blot of pCMV-FLNB-EGFP constructs (wild-type or bearing indicated mutations) probed with anti-GFP antisera and an anti-β-tubulin control.

The Extent of Focal Accumulation Correlates with Disease Severity

To examine whether the focal accumulation of FLNB might occur in other mutant forms of FLNB, other CH2 mutations were introduced into the FLNB-EGFP fusion protein construct. Mutations resulting in substitutions p.Leu171Arg (BD), p.Tyr148Arg (AOI), p.Met202Val (AOI and AOIII), p.Glu227Lys (LS), and p.Phe161Cys (LS) were all found to cause focal accumulation of the fusion protein in HEK293 cells (Fig. 4). Significantly, there was variability in the extent of focal accumulation. Similar to p.Ser235Pro, the p.Leu171Arg mutation left almost no cytoplasmic signal outside of the globular foci (Fig. 4, panel b). Both p.Met202Val (panel c) and p.Tyr148Arg (panel d) while producing foci in the vast majority of transfected cells also displayed more general cytoplasmic localization. The distribution of FLNB-EGFP constructs containing p.Glu227Lys and p.Phe161Cys substitutions frequently resembled that of cells transfected with wild-type fusion protein, or showed only minor focal accumulations (panels e and f). Thickened fibrils were also observed in some p.Glu227Lys transfectants (Supp. Fig. S4A, panel j). Similar results were obtained with M2 cells (Supp. Fig. S4B).

Figure 4.

Cellular localization of EGFP-tagged FLNB constructs bearing mutations causative of the BD-AO-LS disorders. HEK293 cell line transfected with pCMV-FLNB-EGFP fusion constructs carrying CH2 mutations and analyzed for EGFP expression (green; gamma adjustment 2) digitally merged with DAPI (blue). Wild-type FLNB (a) is compared to FLNB with substitutions p.Leu171Arg (b), p.Met202Val (c), p.Tyr148Arg (d), p.Glu227Lys (e), p.Phe161Cys (f), p.Gly751Arg (g), p.Ser1602Pro (h), and p.Pro1699Ser (i). Scale bar 20 μM.

Disease-Associated FLNB Mutations in Repeats 14 and 15 do not Lead to Formation of Filamin-Actin Foci

Mutations outside the ABD also result in identical disease phenotypes as those caused by mutations within the CH2 region (Fig. 1 and Table 1). The mutations are mostly clustered with the repeats flanking hinge 1 but are also occasionally seen at residues remote from this region (Fig. 1A). To determine if the focal accumulation phenomenon was restricted to mutations leading to CH2 substitutions, three amino acid substitutions from repeat 6 (p.Gly751Arg; AOIII), repeat 14 (p.Ser1602Pro; AOIII), and repeat 15 (p.Pro1699Ser; severe AOI) were introduced into the FLNB-EGFP fusion construct and transfected into HEK293 cells (Fig. 4, panels gi). While the p.Gly751Arg mutant showed good evidence of foci formation (panel g), equivalent to other AOI/III mutations, the p.Ser1602Pro and p.Pro1699Ser mutants did not differ markedly from the wild-type fusion protein distribution (panels h and i, respectively). Similar results were seen utilizing M2 cells to rule out any confounding influence that FLNA expression might have upon these observations (Supp. Fig. S4B).


Several lines of evidence indicate that the pathogenic mechanism underlying the autosomal dominant spectrum of disorders due to mutations in FLNB that includes BD, AOI/III, and LS is distinct from the recessively inherited disorder SCT. First, the diseases caused by recurrent and clustered amino acid substitutions within the CH2 subdomain and first rod domain of FLNB are quite distinct phenotypically from SCT syndrome that results from truncating mutations leading to loss of expression of this protein Farrington-Rock et al., 2008. Additionally, heterozygous carriers for SCT mutations exhibit minimal or no skeletal manifestations, indicating that the BD-AO-LS spectrum is unlikely to be due to haploinsufficiency at the FLNB locus. Third, recent data indicate that mutations within the ABD leading to the BD-AO-LS spectrum of conditions lead to increased F-actin avidity in the absence of structural destabilization of this domain Sawyer et al., 2009 as do analogous mutations in FLNA Clark et al., 2009 and FLNC Duff et al., 2011. Together these observations indicate a gain-of-function is conferred by mutations leading to BD-AO-LS but equally there remains little mechanistic understanding of the cellular consequences of the increase in actin-binding capability conferred by these alleles.

The data presented here link several lines of evidence to suggest that mutations that lead to enhanced actin binding do so by disabling CH2 regulatory function and that the result is abnormal transitioning of FLNB between networks and bundles of actin fibers. Although previously in vitro evidence demonstrated that the CH2 subdomain negatively regulates actin avidity and binding, a demonstration that this affects full-length filamin function and what the downstream consequences were on cell remain unexplored. In transient transfection experiments, BD-causing mutations in FLNB (p.Ser235Pro and p.Lys171Arg) led to nearly complete localization of the mutant proteins within actin containing cytoplasmic focal accumulations. These foci are not observed with overexpression of the wild-type protein and are unlikely to represent nonspecific aggregation of substituted protein, since they colocalize with actin and FLNA. Equally, the observed foci are unlikely to represent aggregation of degraded protein, since Western analysis of transfected cells revealed little evidence for this and filamin A would be unlikely to colocalize with mutant protein that had been directed to the proteosome. Recently, FLNA has been shown to regulate the formation of an analogous actin-rich perinuclear body, termed the actin cap Gay et al., 2011, a transient structure that relates to regulated alterations in nuclear location and shape.

Mutations associated with AOI/III and LS also cause a similar discontinuity of FLNB distribution within transfected cells, but the proportion of FLNB-specific signal associated with the foci is progressively reduced, correlating with both the reduction in severity of the associated disease phenotypes albeit still with an elevated avidity for actin Sawyer et al., 2009. Both p.Tyr148Arg and p.Met202Val have increased avidity for F-actin Sawyer et al., 2009 as do p.Glu254Lys in FLNA which is homologous with p.Glu227Lys in FLNB Clark et al., 2009.

These results demonstrate marked similarities to experiments with isolated CH1 and CH2 subdomains from α-actinin-1 and filamin A that determined that a single CH1 subdomain from either molecule was necessary and sufficient for actin binding in vitro and in vivo Lorenzi and Gimona, 2008 and that the absence of a matched CH2 subdomain adjacent to a CH1 subdomain was associated with reduced actin-filament turnover and aberrant bundling of actin fibrils. Therefore our data are consistent with a proposed model of CH1 acting as the primary F-actin-binding site and CH2 having a modulatory function.

Additionally, the data presented here also suggest that the aberrant filamin-actin binding imposed by LS-AO-BD mutations has an effect on actin cytoskeletal architecture. Similar actin focal accumulations have recently been demonstrated using a similar transient transfection approach for a gain-of-function mutation in the CH2 domain of FLNC Duff et al., 2011, suggesting that the pathogenesis of this condition, a myopathy, is similar to that operating to cause the FLNA- and FLNB-related disorders. Outside the filamin gene family, a similar mechanism has been invoked for gain-of-function mutations in another actin-binding protein, α-actinin-4, which have also been demonstrated to increase F-actin binding Weins et al., 2007.

The appearance of focal accumulations is also seen with a mutation outside the ABD in repeat 6 (p.Gly751Arg). It has been postulated that the immunoglobulin-like repeats between CH2 and the first hinge region (repeats 1 to 15) align with actin filaments and contribute significantly to actin binding Nakamura et al., 2007. It is therefore plausible that this mutation also increases F-actin binding in vivo.

Sections of epiphyseal growth plates from AOI/III patients contain regions of acellularity in the proliferating zone, with occasional multinucleated giant cells in the reserve zone Sillence et al., 1982, but no observation of focal protein accumulations as have been demonstrated here employing overexpression of EGFP-labeled gene constructs. Although focused reexamination of growth plate tissue from individuals with LS-AO-BD disorders would be useful in the light of our new results, it is unlikely that anomalies like this will be identified since examination of muscle from an individual with an analogous FLNC mutation did not demonstrate the focal accumulations seen in transient transfection experiments similar to those performed here Duff et al., 2011. It is most likely that the filamin-actin foci noted here are visible because of overexpression in the transfection system employed, and may therefore not be causative of disease. Despite this, we propose that they are still reflective of a biophysical alteration in the properties of filamin that result in a maldistribution of the protein intracellularly.

In contrast to the results obtained for mutations relating directly to filamin-actin interactions, the two mutations associated with AOI/III in repeats 14 and 15 did not result in the appearance of actin-filamin containing focal accumulations. Like CH2, the repeat 14 and 15 region is a second region within which substitutions that lead to AOI/III and LS phenotypes are clustered (Fig. 1) and lies directly adjacent to hinge 1, a structure that confers flexibility to FLNB and is central to its mechanosensory properties Gardel et al., 2006. Deletion of hinge 1 in FLNA and FLNB has demonstrated that it is this structure that confers the nonlinear viscoelastic properties that are critical to cellular mechanoprotection conferred by filamin Gardel et al., 2006. Although data are lacking to directly implicate the repeat 14 and 15 mutations described here with hinge 1 function, their identical phenotypic consequences compared to ABD mutations that alter filamin transitioning within the cell through alterations in filamin-actin avidity suggest a shared impact on the cellular response to shear forces.

How this maldistribution of FLNB leads to the pathogenesis of FLNB-related filaminopathies is uncertain. Clearly, mutations affecting actin avidity and those around hinge 1 lead to final common phenotype that results in diminished ossification and joint dislocations. Although the mechanosensory properties of FLNB are dependent on both of these functions of the molecule, there is little understanding of how mechanotransduction influences morphogenesis in these tissues. Certainly, the position of these foci would argue against dysregulation of second messenger signaling from the cortical cytoskeleton and analogous mutations in FLNA do not demonstrate an abnormal migratory or cell adhesion phenotype Clark et al., 2009. FLNA has been proposed as an important interface between extracellular matrix (ECM) and cytoskeletal actin, acting via C-terminal associations with β-integrins Kiema et al., 2006a; Lad et al., 2007a. FLNB may play a similar but more specialized role in the chondrocyte, a cell type embedded within a dense ECM. Filamins confer stress-strain characteristics to actin gels in vitro similar to those observed for intact cells Gardel et al., 2006 and furthermore extrinsic shear force, transduced via integrins, triggers filamin to change its preferred C-terminal interaction partners Ehrlicher et al., 2011. Increases in actin affinity, and changes to hinge flexibility, may therefore alter the gain of signals generated through integrin-filamin interaction that in turn may influence growth plate chondrocytes and osteoblasts. A critical test of this hypothesis would be the influence of hinge region mutations on FLNB flexibility, a possibility that should be addressed in future experiments.


The authors thank participating individuals and their families. SR is supported by Heath Research Council of New Zealand and Curekids, New Zealand.