Structural and Functional Evaluation of C. elegans Filamins FLN-1 and FLN-2

Filamins are long, flexible, multi-domain proteins composed of an N-terminal actin-binding domain (ABD) followed by multiple immunoglobulin-like repeats (IgFLN). They function to organize and maintain the actin cytoskeleton, to provide scaffolds for signaling components, and to act as mechanical force sensors. In this study, we used transcript sequencing and homology modeling to characterize the gene and protein structures of the C. elegans filamin orthologs fln-1 and fln-2. Our results reveal that C. elegans FLN-1 is well conserved at the sequence level to vertebrate filamins, particularly in the ABD and several key IgFLN repeats. Both FLN-1 and the more divergent FLN-2 colocalize with actin in vivo. FLN-2 is poorly conserved, with at least 23 IgFLN repeats interrupted by large regions that appear to be nematode-specific. Our results indicate that many of the key features of vertebrate filamins are preserved in C. elegans FLN-1 and FLN-2, and suggest the nematode may be a very useful model system for further study of filamin function.


Introduction
Filamins are long, flexible, multi-domain proteins composed of an N-terminal actin-binding domain (ABD) followed by multiple immunoglobulin-like repeats (IgFLN). The best-characterized filamins are Dictyostelium discoideum filamin (ddFLN) and human filamins (hsFLNA/B/C). Dictyostelium filamin has an ABD followed by six IgFLN repeats, whereas the human orthologs have 24 IgFLN repeats arranged into two rod domains separated by a flexible hinge. FLNA, FLNB, and FLNC are more than 70% identical at the amino acid sequence level and have overlapping expression patterns. Although FLNA and FLNB are ubiquitously expressed, FLNC is found primarily in cardiac and striated muscle [1]. Filamins are involved in diverse cellular processes including anchoring, organizing and maintaining the actin cytoskeleton, providing a scaffold for signaling components, and acting as molecular sensors for mechanical forces [1]. Due to the pleiotropic functions of filamins in humans, mutations cause a wide variety of developmental defects in the skeleton, brain, heart, and smooth muscle [2].
Although no complete structure of a filamin molecule is available, biochemical and structural studies have provided important insights into the function of filamins [3,4,5]. The beststudied role of filamin is in the organization of actin filaments into branched three-dimensional networks [1]. Filamin binds F-actin using the N-terminal ABD, although some IgFLN repeats and hinge regions may also contribute to actin binding [6]. The filamin ABD consists of two calponin homology (CH) domains that are well conserved among filamins and other actin binding proteins, such as alpha-actinin, spectrin, and fimbrin [7]. In filamin, the primary actin-binding site is hydrophobic and is located in the first CH domain (CH1) [8,9,10]. The second CH domain (CH2) has a lower affinity for actin, but is required for a fully functional ABD [10,11]. Although CH2 is less conserved across filamins than CH1, disease-related mutations suggest that CH2 may regulate the actin-binding activity of CH1 [12]. For example, gain-of-function mutations in the CH2 domain of FLNA lead to developmental disorders of the skeleton by increasing filamin affinity for F-actin, which perturbs actin cytoskeleton dynamics [13].
Individual IgFLN repeats are ,96 amino acids in length and are comprised of seven b-strands (A-G) arranged into two bsheets, which together form a b-sandwich. Filamins are predicted to interact with more than fifty different proteins, many of which interact with the CD strands of the IgFLN domains [14]. The majority of these interactions involve IgFLN domains in the second rod domain (IgFLN16-24). For example, filamin binds transmembrane proteins such as integrins [15], transmembrane receptors [16], and many signaling proteins, including the Rhofamily of GTPases [17,18]. The cytoplasmic tail of b7 integrin binds to the CD face of FLNA IgFLN21 [5], which links the actin network physically with the extracellular matrix (ECM). FLNA IgFLN24 binds RhoA, Rac1 and Cdc42, all of which regulate actin dynamics. In addition, the final repeat also mediates dimerization of filamins [6,19,20]. FLNB has also been shown to serve as a scaffold for signaling pathway components, for example, the Rac1, MEKK1, MKK4, and JNK cascade in interferoninduced apoptosis [1,21,22].
We are using the nematode C. elegans as a model system to study the conserved functions of filamin in vivo. C. elegans provides many advantages for the study of cytoskeletal regulation, including a translucent body, good visualization tools, and freely available genomic resources [23]. The C. elegans genome encodes two filamin homologs, fln-1 and fln-2. In addition to filamins, many other cytoskeletal regulators are well conserved, including integrin [24], talin [25], vinculin [26], and the Rac GTPases [27]. We have recently shown that FLN-1 is required for maintenance of actin and for proper function of the spermatheca, a somatic tissue of the nematode reproductive system that undergoes dramatic shape changes during the ovulation and fertilization of oocytes [28]. The second C. elegans filamin ortholog, FLN-2, has not been studied, but large-scale RNAi studies suggest it may play a role in molting [29,30]. In this study, we characterize the gene structures and expression patterns of fln-1 and fln-2, use homology modeling to determine the structure of the conserved IgFLN domains in these molecules, predict the conservation of function between C. elegans and other filamins, and demonstrate that the ABD of FLN-1 and FLN-2 can co-localize with actin in vivo.

Results
The C. elegans genome encodes two filamin genes, fln-1 and fln-2 In order to identify filamin-like genes in C. elegans, we performed a BLAST search of the C. elegans genome (WormBase release WS190) using human FLNA. At the time, this search identified a cluster of predicted open reading frames (ORF) on chromosome IV (Y66H1B.2, Y66H1B.5, and Y66H1B.3; Figure 1A) and another on chromosome X (C23F12.1 and C23F12.2; Figure 1B). These ORFs lack one or more of the characteristic features of filamin, the ABD or IgFLN repeats, but taken together appeared to encode proteins with a domain structure similar to other filamins. To determine which transcripts are produced from these loci, we performed RT-PCR on total RNA extracted from mixedstage wild-type (N2) nematodes, and sequenced the resulting cDNAs ( Figure S1). Sequencing of cDNAs derived from the fln-1 and fln-2 loci allowed us to identify three fln-1 isoforms and four fln-2 isoforms (Figures 1 and S1, Table 1). Sequences spanning Y66H1B.2, Y66H1B.3 and Y66H1B.5 were readily detected, suggesting these three ORFs are transcribed together and represent a single filamin locus. Similarly, C23F12.2 and C23F12.1 were found to comprise a second filamin locus. The names fln-1 and fln-2 have been assigned to the Y66H1B and C23F12 filamin loci, respectively, and the amended gene models are available in WormBase.
To identify the 59 ends of the transcripts we took advantage of SL1 trans-splicing in nematodes. The majority of transcripts in C. elegans are trans-spliced to a 22-nucleotide splice leader sequence, SL1 [31]. The SL1 sequence is commonly used in C. elegans to identify the 59 ends of transcripts [32]. We performed RT-PCR using SL1 and gene-specific primers, and sequenced the resulting amplicons. Y66H1B.3, Y66H1B.2, C23F12.1, and two independent C23F12.2 transcripts, are trans-spliced to SL1, which suggests that each of these ORFs can be transcribed independently (Figures 1 and S1). The predicted ORF Y66H1B.5 is not trans-spliced to SL1, and is probably only found as part of fulllength fln-1.
fln-1a, b, and c are the three major fln-1 transcripts, with fln-1a representing a full-length filamin that includes an actin-binding domain and 20 Ig-like filamin repeats ( Figures 1A and S1A). We are able to readily detect the full-length (6.7 kb) transcript from wild-type nematodes using RT-PCR. The structure of the FLN-1A isoform is similar to that that predicted by Heikkinen, et al. [33]. The fln-1b and fln-1c transcripts are the result of alternative splicing and an alternate transcriptional start site, respectively. fln-1b includes an exon that is not a part of the full-length fln-1a ( Figure 1A and S1A). We also detected rare transcripts spanning Y66H1B.3 and Y66H1B.2, but did not include Y66H1B.5, some of which contained stop codons at the beginning of Y66H1B.2 due to alternative splicing ( Figure S1). Analysis of GFP reporter constructs ( Figure 1A) suggests fln-1a, and likely fln-1b, are predominantly expressed in the somatic gonad, including the spermatheca, sheath, and uterus ( Figure 2A). We have shown previously that a full length, functional FLN-1A::GFP fusion protein is expressed and co-localizes with actin filaments in these tissues [28]. In contrast, the fln-1c construct is expressed in the body wall muscle, vulval muscle and hypodermis ( Figure 1E,F).
fln-2a represents the longest predicted transcript from the fln-2 locus based on cDNA sequencing ( Figure 1B and S1B); however, we were unable to detect the full-length fln-2a transcript by PCR, likely because of the transcript size (.11 kb). fln-2b is the result of alternative splicing in the final exon of C23F12.2, which introduces a stop codon, while fln-2c and d are the result of alternative transcriptional starts ( Figure 1B). In addition to the fln-2 SL1 sites we identified, the modENCODE consortium has identified numerous SL1 sites within C23F12.2 [34]. This suggests that fln-2 may have many transcriptional start sites. These additional SL1 sites would produce transcripts similar to fln-2c and d. Based on RNA-seq expression data from modENCODE it seems likely that fln-2a and b are only expressed during embryogenesis, while fln-2c and d are expressed throughout the life stages. In addition, five exons near the 39 end of fln-2, colored light gray in Figure 1B, are usually excluded, suggesting that most transcripts are lacking these five exons. Analysis of a series of GFP reporter constructs ( Figure 1B) suggests the various isoforms of fln-2 are expressed broadly, with strong expression in the hypodermis, pharynx, intestine, anal depressor muscle and distal tip cell ( Figure 2C-D, G-J). The modENCODE data suggests that the FLN-2C isoform might be the most strongly expressed version of FLN-2. To determine the expression pattern and subcellular localization of FLN-2C, we constructed a translational GFP fusion. FLN-2C::GFP localizes in small puncta at the cell membranes of vulval ( Figure 2G) and hypodermal cells ( Figure 2H,I) and to the intestinal lumen ( Figure 2I). Lacking a mutant allele of fln-2 with an overt phenotype, we are unable to determine the functional significance of these expression patterns.
To visualize evolutionary relationships between the C. elegans filamins and well-studied filamins from humans, Drosophila and Dictyostelium, we constructed a phylogenetic tree using the neighbor joining method [35]. Human filamins, Drosophila filamin cheerio, C. elegans filamin FLN-1, and Dictyostelium filamin cluster together ( Figure 1E). There is insufficient homology to indicate to which human filamin FLN-1 is most similar. In contrast, FLN-2 is a less well-conserved filamin, distant even from the divergent Drosophila ortholog jitterbug.

Domain identification of C. elegans filamins
In order to identify the actin-binding and IgFLN domains in FLN-1 and FLN-2, we used the sequence feature scan in SWISS-MODEL Workspace [36], the Simple Modular Architecture Research Tool (SMART) [37,38], and searched the protein sequence for conserved SPF motifs near the ends of IgFLN repeats [39]. (See figure S2 for an alignment of FLN-1 IgFLN repeats). The full-length form, fln-1a, is predicted to encode a 2255 amino acid protein composed of an N-terminal ABD, followed by 20 IgFLN repeats ( Figure 1C, Table 1). The fln-1b transcript encodes the ABD and 8 IgFLN repeats, two of which are alternatively spliced and not found in fln-1a. Finally, the fln-1c transcript is predicted to encode only the final 9 IgFLN repeats ( Table 1). The full-length isoform, fln-2a, encodes a 3747 amino acid protein composed of an N-terminal ABD followed by at least 23 IgFLN domains ( Figure 1D, Table 2). The FLN-2B protein consists of an ABD followed by a 300 amino acid low-homology region and two IgFLN repeats. FLN-2C and FLN-2D proteins contain IgFLN repeats, but do not have an ABD ( Figure 1, Table 2).

Homology modeling of IgFLN repeats in FLN-1 and FLN-2
We performed homology modeling of the C. elegans filamins using the SWISS-MODEL Workspace [36,40]. The human filamin actin-binding domain and many of the IgFLN repeats have structures available in the Protein Data Bank (PDB). We identified templates, and used ClustalW [41], as well as manual adjustments, to refine template alignment to the target sequence before modeling. We used stringent criteria for homology modeling, retaining only those models with a sequence identity greater than 30% and a QMEAN score greater than 0.500 [42]. For FLN-1, 13 of the 21 domains met these criteria (Table 3, Figure 3), while 11 of the 24 FLN-2 domains possessed sufficient homology for modeling (Table 4, Figure 4). The QMEAN analysis indicates that experimentally determined IgFLN structures for FLN-1 and FLN-2 are statistically likely to be similar to these predicted models [42].
In full-length FLN-1, 20 IgFLN repeats follow the N-terminal ABD ( Figure 1C, Figure 3). We were able to model three IgFLN domains from the first rod domain, and all of the second rod    IgFLN domains ( Figure 3, Table 3). Importantly, each IgFLN repeat in the second rod domain maps to a specific IgFLN repeat in human FLNA ( Figure 3, Table 3), suggesting that the domains are not just interchangeable modules, but may mediate specific and conserved interactions. In order to predict how the IgFLN domains in C. elegans filamins might interact with each other, we used multi-domain structures 2K7P (FLNA 16-17), 2K7Q (FLNA [18][19], and 2J3S (FLNA 19-21) as scaffolds to arrange our domain models ( Figure 3). The resulting models suggest that the IgFLN repeats in the FLN-1 second rod domain would adopt an arrangement similar to those found in human FLNA.
FLN-2 is composed of an N-terminal ABD followed by at least 23 IgFLN domains, interspersed with long regions for which no structures were available ( Figures 1D and 4). It is possible the low homology regions in FLN-2 encode additional IgFLN repeats. This lower overall homology resulted in fewer high-quality IgFLN domain models, and the mapping of five FLN-2 IgFLN repeats to a single PDB model 2J3S ( Figure 4, Table 4). Despite these caveats, the model quality estimation parameters for the FLN-2 IgFLN repeat models are similar to those of the FLN-1 IgFLN repeats (Table 3), suggesting FLN-2 is likely to be composed primarily of filamin-like repeats. We were unable to perform

Homology modeling of the FLN-1 and FLN-2 actin binding domains
We identified actin-binding domains (ABDs) near the N-termini of FLN-1 and FLN-2 (Figures 3, 4, and 5). Alignment of the ABDs of FLN-1, FLN-2, and human FLNA indicates the filamin ABDs are composed of well-conserved CH domains ( Figure 5D). The CH1 and CH2 domains of FLN-1 and FLN-2 were modeled using the FLNB ABD structure (PDB ID 2WA5) [43]. The resulting FLN-1 and FLN-2 ABD models are very similar to the FLNB structure ( Figure 5A-C). Both the degree of homology and statistical analysis of the models suggest the structure and function of the C. elegans ABDs are likely to be extremely well conserved (Tables 2 and 3). Interestingly, the FLN-2 ABD contains a third, more divergent, CH domain. No PDB structures were available with sufficient homology for modeling the FLN-2 CH3 domain. In addition, the linker region between CH1 and CH2 is absent in FLN-2, suggesting that interaction of the CH1 and the CH2 domain, and hence the actin binding activity of FLN-2 may be regulated differently ( Figure 5C) [10].

FLN-1 and FLN-2 actin binding domains colocalize with actin in vivo
Because of the significant homology between the C. elegans and human filamin ABDs, and the results of our homology modeling, we predicted that FLN-1 and FLN-2 would colocalize with actin in vivo. We tagged the FLN-1 and FLN-2 ABD with green fluorescent protein (GFP), generated transgenic nematodes by microinjection, and used fluorescence microscopy to determine the co-localization of the GFP fusions with actin filaments in body-wall muscle cells. FLN-1 and FLN-2 ABD localization was compared to animals expressing a VAB-10/spectraplakin ABD GFP fusion protein. VAB-10, like filamin, has an ABD consisting of tandem CH domains and has been used to label actin filaments in C. elegans [44]. F-actin was visualized in the transgenic nematodes by staining with Texas red-conjugated phalloidin. We observed colocalizaiton of the FLN-1 and FLN-2 ABDs with F-actin ( Figure 6E-J) in a manner indistinguishable from the VAB-10  fusion protein ( Figure 6B-D). To determine if the third CH domain in the FLN-2 ABD is necessary for localization to actin, we generated transgenic animals in which the two most highly conserved domains (CH1 and CH2) of FLN-2 were fused to GFP. The localization pattern ( Figure 6 K-M) was indistinguishable from the other ABD constructs, indicating the third CH domain is not required for localization. These data suggest that the CH1 and CH2 domains of C. elegans filamins can direct the localization of GFP to actin filaments, likely through direct binding to actin.

IgFLN repeats are conserved in FLN-1
IgFLN domains are composed of two beta-sheets that combine to form a beta sandwich. The first sheet contains four beta strands (A, B, E, and D) and the second is composed of three strands (C, F, and G). Human IgFLN repeats have been classified into four distinct groups: A, B, C, and D [14]. Using sequence alignments and phylogenetic trees, we identified six group A repeats (4, 8, 13, 15, 17 and 19), five group B repeats (12, 14, 16, 18, and 20), and one group D repeat (6) ( Figure 7A, S2, and S3). Group C and D repeats are located in the first rod of filamin and have a lower conservation compared to group A and B repeats. Importantly, not all repeats in the first rod domain are poorly conserved; namely repeats 4 and 8, which belong to group A.
Group A repeats have a conserved ligand-binding motif on the CD face and are responsible for most of the known filamin interactions [14]. Although other Group A repeats bind integrin with lower affinity, the hsFLNA IgFLN21 is the primary site of integrin binding [4,5]. The overall structure, position in the second rod domain, and key integrin binding residues present in FLN-1 IgFLN17 very closely resemble FLNA IgFLN21, suggesting that the nematode filamin may also bind integrin in this region ( Figure 7B-C). In contrast, while our modeling studies indicate FLN-2 IgFLN12 will fold similarly to FLNA IgFLN21, substitutions exist in several key residues on the CD face that would make integrin binding unlikely ( Figure 7B-C) [4,5].
In vitro studies of Dictyostelium and human filamins indicate that the C-terminal Group B repeat mediates homo-and hetero-dimerization between filamin monomers [9]. Our homology modeling results predict that the C-terminal IgFLN repeat of FLN-1 (IgFLN20) is structurally similar to that of the final human filamin repeat IgFLN24 (PDB structure 2EED), including the hydrophobic dimerization signature [45,46] (Figure 8A-B). In contrast, the final repeat of FLN-2 does not show specific structural homology to the dimerization domain, or possess the residues predicted to mediate dimerization ( Figure 4, Table 4). These modeling results suggest FLN-1 and FLN-2 may differ in their ability to form dimers. In the future, it will be important to determine if FLN-1 and FLN-2 can form hetero-or homodimers in vivo.

Discussion
In this study, we use transcript sequencing and bioinformatics to identify and describe the gene structure of the C. elegans filamin orthologs, encoded by fln-1 and fln-2. In addition, we use comparative homology modeling to predict the structures of the filamin repeats and the actin-binding domains. Taken together, our results suggest that the C. elegans filamin FLN-1 contains all features of the human filamins, including a well conserved actinbinding domain and 20 Ig-like filamin (IgFLN) domains separated into two rod regions. In contrast, FLN-2 is more divergent with an unusual ABD composed of three CH domains, and at least 23 Iglike repeats interspersed with large regions of low homology. These results will help inform and validate future genetic and biochemical studies of filamin function, particularly those using C. elegans as a model system. The gene structures of the two C. elegans filamins are quite divergent from each other. fln-1 is smaller and is comprised of comparatively few large exons ( Figure 1A), while fln-2 has many small exons and is considerably longer ( Figure 1B). Exons in fln-1 appear to be modular units that contain one or more complete IgFLN repeats (Figure 3). Due to the smaller exon length, domains in fln-2 are usually encoded by multiple exons (Figure 4). In both fln-1 and fln-2, a few small exons appear to be alternatively spliced, but all are in frame and are not predicted to have much effect on the overall structure of the protein. It is likely that we have not detected all splice variants of fln-1 and fln-2. C. elegans filamins appear to be extensively regulated through alternative splicing, and alternative transcriptional starts. Human filamins are also alternatively spliced, with consequences for ligand binding in the second rod domain [47]. These events are not known to produce large N-and C-terminal truncations like in C. elegans and D. melanogaster. Determining the function of these alternative filamin isoforms will be an interesting area for future research.
FLN-1 is structurally very similar to vertebrate filamins, particularly in the actin-binding domain and the second rod region. FLN-1 and Drosophila filamin cheerio both contain 20 IgFLN domains [33,[48][49][50]. The second rod region and the dimerization domain appear to be extremely well conserved between the worm, fly, and human filamins. In each case, the second rod domain consists of 8 IgFLN domains in the same order. The first rod domains of FLN-1 and cheerio lack four repeats, and have a lower overall conservation. It is possible that FLN-1 and cheerio represent ancestral filamins, which gave rise to vertebrate filamins through gene duplication events and modular expansion of the first rod domain [12]. We predict FLN-1 is likely to fold, interact and function in a manner similar to that of human filamins.
FLN-2 is a divergent form of filamin with low homology to vertebrate filamins. The closest homolog of FLN-2 outside of nematodes is the Drosophila filamin jitterbug. jitterbug shares many of the features of FLN-2, including an unusual actin-binding domain composed of three CH domains, and poorly conserved filamin repeats with long intervening sequences. jitterbug mutations cause seizures in affected flies, suggesting that jitterbug may have a role in the development or functioning of the nervous system [48,49,50]. Additionally, jitterbug is required for the mechanical response of notum tendon cells [51]. In the nematode, previously published transcript profiling data suggests full-length fln-2a is only expressed during embryogenesis, and the predominant larval form may be fln-2e, which lacks the ABD [34]. FLN-2 has not been investigated, but large-scale RNAi studies suggest it may play a role in molting [29,30]. Consistent with this, we find FLN-2C localizes in punctate structures at the membrane of hypodermal cells, where it might play a structural or attachment role.

FLN-1 and FLN-2 colocalize with F-actin in vivo
Filamins are an important class of actin crosslinking proteins that contribute to the stability of the actin cytoskeleton. We identified the actin binding domains of FLN-1 and FLN-2, and showed that they are highly conserved and structurally similar to mammalian filamins, particularly in areas critical for actin binding ( Figure 5A-C). Importantly, we demonstrated that these putative ABDs co-localize with F-actin in vivo. This work is consistent with previous results which demonstrate that both FLN-1 and cheerio are required for integrity of the actin cytoskeleton in the reproductive system [28,52,53,54].
Human disorders such as otopalatodigital syndrome (OPD) [55], periventricular heterotopia (PVH) [56], and boomerang dysplasia (BD) [57] can be caused by mutations in the filamin ABD, which modulates the affinity of filamin for F-actin. Therefore, a greater understanding of how the actin binding activity of filamin is regulated could be clinically relevant.
Although biochemical experiments have shown in vitro binding of the filamin ABD by calmodulin [10] and phosphatidylinositol 4,5bisphosphate (PIP 2 ) [58], further studies are needed to determine how actin binding is regulated physiologically. Because the structural features of the ABD are preserved in C. elegans, the nematode may be a tractable model system to study this feature of filamin function.   Homology modeling suggests FLN-1 may form dimers Dimerization has long been considered critical for the actincrosslinking function of filamin. Biochemical studies have identified the most C-terminal IgFLN domain as the primary dimerization domain [6,9,20,45,59,60,61]. Our homology modeling suggests that FLN-1 may form dimers in an orientation akin to vertebrate filamins, rather than in the manner of ddFLN, which lacks the hydrophobic dimerization interface [45]. FLN-1 dimerization could be important for its role in actin crosslinking in the nematode somatic gonad, especially in the uterus and spermatheca, where fluorescence microcopy reveals what appear to be V-shaped filamin clusters co-localized with branching actin filaments [28].

Force sensation in the second rod domain of FLN-1
Filamin molecules can function as mechanical force sensors and have been shown to engage in distinct interactions depending on conformational state. The arrangement of filamin repeats has functional relevance and provides insight into how filamin may respond to mechanical stimuli [4,33]. For example, studies using atomic force microscopy have demonstrated that the IgFLN repeats can withstand different amounts of force before unfolding [62,63], and that this unfolding can interrupt or reveal binding sites. In particular, in the FLNA group IgFLN19-21, the Nterminal b-strand A of IgFLN20 is extended away from the rest of the sandwich, and participates structurally as a part of IgFLN21, which is nestled between IgFLN19 and IgFLN21 [4]. In this orientation, the N-terminus of IgFLN20 occupies the site where integrin preferentially binds. When this auto-inhibition is disrupted, the binding site is revealed, integrin binding affinity increases, and the length of the filamin molecule increases [4]. Structural data suggest that other IgFLN repeats in Rod 2 also interact with one another and may have similar effects on ligand binding and the cellular response to mechanical stress [33]. Our analysis suggests that the second rod domain of FLN-1 is well conserved, and that the domains are physically arranged as in the human filamins. The significant sequence and structural conservation suggests that FLN-1 may also exhibit differential interactions and auto-inhibition in response to changing mechanical conditions. Ongoing studies are testing the hypothesis that FLN-1 functions as a stretch-sensitive signaling scaffold in vivo [28].

Materials and Methods
Gene structure and sequencing of fln-1 and fln-2 Wildtype N2 nematodes were obtained from the Caenorhabditis Genetics Center (Minneapolis, MN). Total RNA was extracted from mixed-stage N2 animal populations as described previously [25]. cDNA was prepared using SuperScript Reverse Transcriptase (Invitrogen; Carlsbad, CA, USA) and random hexamers as primers. cDNA sequencing was accomplished by amplifying regions of fln-1 and fln-2 by PCR, and subsequently cloning the amplicons into pGEM T-vector (Promega; Madison, WI, USA). Vector-specific primers (SP6 and T7) were used to sequence the inserts. Vector and poor quality sequences were trimmed, and the sequences were aligned to C. elegans genomic sequences using CLC bio Main Workbench version 5.2 (CLC bio; Cambridge, MA, USA). Tables S1A and S1B provide complete lists of primers used for sequencing fln-1 and fln-2 respectively. The amended C. elegans filamin gene names and structures have been submitted to WormBase (http://www.wormbase.org/) and are available as of the WS210 release [64]. Previous gene structure data is available in the referential WormBase WS205 release (http://ws205. wormbase.org/) and earlier releases.
fln-1 and fln-2 expression Sequences 59 of each filamin transcript were amplified by PCR using N2 genomic DNA as a template. Each fragment was flanked by engineered HindIII and BamHI restriction sites to facilitate cloning into the GFP expression vector pPD95.77 (Fire Vector Kit). Plasmids were isolated from E. coli and used for microinjection at an approximate concentration of 100 mg/mL. Transgenic strains were created by standard germline transformation [69] of wild type animals to create strains UN0824 xbEx0824[fln-1c::gfp; ttx-3::rfp] and UN0816 xbEx0816[fln-2a,b::gfp]. Strain UN1066 xbEx0816[fln-2c::gfp] was created by co-injection of fln-2c amplified from N2 genomic DNA using the high-fidelity Phusion polymerase (NEB, Ipswich, MA) and GFP and unc-54 39 UTR sequences from pPD95.77. Strain UN0811 xbEx0811[fln-1::gfp] is previously described [28]. All primer sequences and cloning details available upon request. For visualization, transgenic animals were mounted on 1.5% agarose pads and imaged using differential interference contrast and epifluorescence microscopy with a Nikon Eclipse 80i microscope equipped with a SPOT RT3 CCD camera (Diagnostic Instruments; Sterling Heights, MI, USA).

Phylogenetic trees
Filamin sequences from C. elegans [ACY40014.  [41]. A phylogenetic tree was constructed using the Neighbor Joining algorithm [35] and the BLOSUM62 matrix [65]. The phylogenetic tree of FLNA and FLN-1 repeats was constructed using CLC bio Main Workbench version 5.2.

Domain identification
The assembled full-length sequences encoding FLN-1 and FLN-2 were used to identify the number and location of structural domains within the protein. IprScan, a PERL-based InterProScan utility [66], integrated into the SWISS-MODEL Workspace [36] was used to predict domains and functional features in the sequence. We also used SMART (Simple Modular Architecture Research Tool) [37,38] to search for evolutionarily conserved protein domains, and scanned the sequences for SPF motifs that are often found near the end of IgFLN repeats [39].

Homology modeling
For homology modeling of FLN-1 and FLN-2, the predicted amino acid sequences were divided into individual domains for alignment and template identification. We started with the Template Identification Tool incorporated into the SWISS-MODEL Workspace (http://swissmodel.expasy.org/), which uses BLAST to identify homologs in the SWISS-MODEL Template Library (SMTL) [40], and selects the best template structure based on sequence similarity, experimental quality, and secondary structure predictions. Domain sequences with less than 50% sequence identity to their selected template were realigned outside of the SWISS-MODEL Workspace using ClustalW [41] and visually inspected to ensure proper alignment prior to modeling. Modeling was accomplished using the SWISS-MODEL Workspace Alignment Mode, which is recommended for target and template sequences with 50-30% sequence identity [67]. Targettemplate pairs with less than 29% sequence identity were discarded. The resulting models and their templates were assessed globally for quality using the QMEAN server (http://swissmodel. expasy.org/qmean/) [42], and homology models scoring less than 0.500 were discarded. Molecular graphics images were produced using the UCSF Chimera package from the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (supported by NIH P41 RR001081) [68].
To determine if the predicted ABD of FLN-1 and FLN-2 colocalize with actin in vivo, transgenic worms expressing fluorescent fusion proteins were fixed with 4% formaldehyde in PBS, placed on poly-L-lysine coated slides, and fixed in cold methanol and acetone. Fixed worms were stained with 0.4 U/mL of Texas Red-X phalloidin (Invitrogen) overnight at 4uC. The slides were washed, mounted, and viewed on a Nikon Eclipse 80i epifluorescence microscope. Images were documented with a SPOT RT3 CCD camera and SPOT Advanced software (Diagnostic Instruments; Sterling Heights, MI, USA.