Filamin actin-binding and titin-binding fulfill distinct functions in Z-disc cohesion

Many proteins contribute to the contractile properties of muscles, most notably myosin thick filaments, which are anchored at the M-line, and actin thin filaments, which are anchored at the Z-discs that border each sarcomere. In humans, mutations in the actin-binding protein Filamin-C result in myopathies, but the underlying molecular function is not well understood. Here we show using Drosophila indirect flight muscle that the filamin ortholog Cheerio in conjunction with the giant elastic protein titin plays a crucial role in keeping thin filaments stably anchored at the Z-disc. We identify the filamin domains required for interaction with the titin ortholog Sallimus, and we demonstrate a genetic interaction of filamin with titin and actin. Filamin mutants disrupting the actin- or the titin-binding domain display distinct phenotypes, with Z-discs breaking up in parallel or perpendicularly to the myofibril, respectively. Thus, Z-discs require filamin to withstand the strong contractile forces acting on them.


Introduction
Arguably, the most complex actin-related cellular structure is the sarcomere, the basic contractile unit of muscle cells. The sarcomere consists of antiparallel actin thin filaments and myosin thick filaments. The thin filaments are anchored to a big protein complex termed the Z-disc at for stabilizing the position of thin filaments perpendicular to the myofibril axis. Our data provide an explanation for the function of filamin in muscle and a framework for understanding some human FLNc myopathies [29].

Results
Cher isoforms are components of the Z-disc The cher gene spans a 34 kb region and produces at least 10 different transcript isoforms whose encoded proteins can be divided according to FlyBase into 4 molecular size groups: the small isoforms, here referred to as CherA and B (90-100 kD); a medium-sized isoform, CherC (150 kD), and CherD, containing all the big isoforms (240-260 kD) (Fig 1A) [30].
To assess the localization of Cher in the IFM we used 4 different protein trap lines that introduce a 30 kD Venus Flag-tagged artificial exon directly into the cher gene [31,32]. Due to their specific location inside the cher gene, cher CPTI1399 tags all protein isoforms, cher CPTI847 and cher GFSTF tag CherC and CherD isoforms; while cher CPTI1403 tags only CherD isoforms ( Fig 1A and S1D Fig). cher CPTI847 and cher CPTI1403 proteins are not fully functional, because homozygotes cause female infertility. cher CPTI1399 and cher GFSTF are homozygously fertile.
We first evaluated by immunoblotting the presence of Cher isoforms in the adult thorax using cher CPTI1399 , because this line should tag most predicted isoforms of cher. All Cher isoform groups were detected using this assay ( Fig 1B). To test the specificity of the detected bands we knocked down cher expression specifically in muscles using an RNAi directed against all isoforms under the control of the Mef2 expression pattern using the Gal4-UAS system: Mef2>cher-JF RNAi. Consistently, Flag-positive bands were no longer detected indicating that all bands correspond to Cher isoforms ( Fig 1B). We then analyzed the localization of these protein traps in the IFM using heterozygotes. While direct Venus fluorescence was barely detectable, anti-Flag staining was detected at the Z-disc, colocalizing with the peak of the actin signal ( Fig 1C and 1D, S1A Fig). We also used line scans to better assess their localization; all Cher-GFP trap lines show identical localization profiles peaking at the Z-disc, indicating that all filamin isoforms colocalize at the Z-disc (S1C Fig).
All three Cher protein traps localize to the Z-disc, suggesting that the small CherA isoforms contain the Z-disc localization information. To further test this, we expressed the smallest CherA isoform fused to GFP in IFM. Consistently, CherA-GFP also localizes to the Z-disc (S1B Fig). Thus, all Cher isoforms are components of the Z-disc and CherA, containing the last 8 Ig-like domains is sufficient for Z-disc localization.

Cher depletion results in sarcomeric defects
There are three available RNAi lines directed against all cher isoforms. We first assessed the efficiency of these lines by testing their ability to render flies flightless. Two of these lines, cher JF02077 and cher KK107451 , in combination with the muscle-specific driver Mef2-Gal4 led to a completely penetrant flightless phenotype. We then evaluated the IFM sarcomeric defects underlying the flying defects by first quantifying sarcomere numbers visible in confocal images. The number of recognizable sarcomeres is reduced by half in individual RNAi lines cher KK107451 and cher JF02077 , and is even further reduced in the double knock-down (Fig 1E and  S1E and S1F Fig), indicating strong defects in myofibril assembly or maintenance. A similar yet less severe sarcomere reduction phenotype is observed in other cher mutants (Fig 1E). We proceeded to analyze the sarcomeres of cher JF02077 flies in more detail.
In control flies, individual IFM sarcomeres are highly stereotyped regular structures with a perfectly defined Z-disc, revealed by the highest intensity peak of actin staining colocalizing with the Z-disc marker Kettin (Fig 1F). In Mef2>cher JF02077 flies the stereotypical sarcomere  Q1415sd . The cher Δ5 deleted region is denoted by a black line (B) Immunoblot from cher CPTI1399 thoraces incubated with anti-Flag antibody reveals 4 Cher isoforms, specifically depleted in Mef2-Gal4>UAS-cher-JF RNAi. (C, D) IFM confocal images from heterozygous Flag-tagged Cher traps stained with anti-Flag antibody to visualize the tagged isoforms. In all images, asterisks mark Z-discs of a selected myofibril immediately above the asterisks. (C) All Cher isoforms, revealed by the cher CPTI1403 protein trap, localize at the Z-disc. (D) Long Cher isoforms tagged by cher CPTI1399 localize to the Z-disc. (E) Quantification of recognizable sarcomeres in control, Cher-depleted, and cher mutant IFM. Statistical significance assessed by one-way ANOVA with post hoc Tukey: n.s. = not significant, ** = P 0.01, **** = P 0.0001. (F, G) Confocal images from control and Cher-depleted IFM, stained with anti-Kettin antibody to visualize Z-discs in red and phalloidin to visualize actin thin filaments in green. (F) Regular sarcomeric structure in Mef2-Gal4 control flies, (G) Depletion of Cher in Mef2-Gal4, UAS-cher-JF flies results in severe sarcomeric disorganization. Arrowheads indicate reduced actin staining at the Z-disc. Scale bars: 5 μm. https://doi.org/10.1371/journal.pgen.1006880.g001 Filamin function in muscles pattern is often lost, resulting in shredded myofibrils. We also observed almost normal sarcomeres, but with less actin at the Z-disc, which we call a widened Z-disc phenotype (arrowheads in Fig 1G). Infrequently, sarcomeres also exhibit very high actin accumulation at the Z-disc (middle asterisk in Fig 1G). Similar defects were seen in Mef2>cher KK107451 flies, confirming the specificity of the phenotype (S1E Fig). When both RNAi lines are combined, Mef2>cher JF02077 cher KK107451 flies display a complete loss of sarcomeric structure, where actin staining is diffuse and continuous and Kettin staining is lost (S1F Fig). To better analyze the IFM phenotype upon Cher depletion, we analyzed the sarcomere phenotype at a higher resolution using transmission electron microscopy (TEM). Control sarcomeres show an electron-dense structure corresponding to the Z-disc with a lighter surrounding area, together corresponding to the I-band. The highly ordered parallel actin thin filaments are readily seen, except in the center of the sarcomere, the H-zone, where actin is excluded (Fig 2A). In Mef2>cher JF02077 sarcomeres three distinct phenotypes were observed: 1) a widened Z-disc, 2) a smaller or fractured Z-disc, and 3) actin incorporation into the H-zone (numbers in Fig 2).

CherD stably attaches actin thin filaments at the Z-disc
The CherD isoforms correspond to the big Cher isoforms, the only isoforms containing the CH actin-binding domain (ABD). Given that Cher depletion results in thin filament disorganization mostly at the Z-disc, we hypothesized that CherD, through the ABD, might serve as a secondary attachment of thin filaments to the Z-disc. To test this, we used the cher Δ5 mutant a wild type sarcomere. Z stands for the Zdisc; H stands for the H-zone and is indicated by orange brackets; (B, C) Mef2-Gal4, UAS-cher-JF sarcomeres show widened Z-discs (green box, shown enlarged on the right), smaller than normal Z-discs (red box, shown enlarged on the right) and actin accumulation at the H-zone (orange brackets). Light green bracket denotes the width of the Z-disc. Numbers correspond to (1) a widened Z-disc, (2) a smaller or fractured Z-disc, and (3) actin incorporation into the H-zone. Scale bars: 500 nm. https://doi.org/10.1371/journal.pgen.1006880.g002 Filamin function in muscles bearing a 2.4 kb deletion uncovering the CherD transcription start site thus reducing CherD protein levels while leaving the other isoforms intact [20,24] (Fig 1A and S2A Fig). TEM of cher Δ5 mutant sarcomeres revealed three phenotypes: 1) a widened Z-disc resulting from a splitting or opening of the Z-disc (Fig 3A and 3A'), 2) actin accumulation at the H-zone (blue arrow in Fig 3A and 3B) and 3) a smaller or fractured Z-disc ( Fig 3A and 3A'). These phenotypes are similar to the spectrum of phenotypes observed in Mef2>cher JF02077 sarcomeres, but of lower penetrance (see Fig 1E), suggesting that C-terminal filamin isoforms can rescue some aspects of filamin function at the Z-disc.
We then used confocal microscopy to analyze other cher mutants that disrupt the CherD isoforms. First, we evaluated cher Δ5 mutant IFM stained for Kettin and actin. In contrast to a control staining (Fig 1F), cher Δ5 IFM display a widened Z-disc phenotype, evident by the lack of actin staining ( Fig 3C). Actin accumulation at the H-zone was also occasionally observed by confocal microscopy (S3B Fig, compare to S3A Fig). Both phenotypes agree with our TEM data. As the widened Z-disc phenotype was easier to distinguish by confocal microscopy we used it as a readout for analyzing other cher mutants.
We first tested one of the original cher alleles to be isolated, cher 1 , that has been shown to greatly reduce CherD protein levels without affecting CherA/B [18,25]. Confocal microscopy images of cher 1 mutant IFM stained for Kettin and actin confirmed the widened Zdisc phenotype, suggesting that the large CherD isoforms account for this phenotype ( Fig 3D).
CherD isoforms are the only isoforms containing an actin-binding domain begging the question of whether actin binding alone is responsible for the widened Z-disc phenotype. To test this hypothesis, we used the cher CPTI1403 protein trap that inserts a Venus-Flag tag into the first CH domain, leaving a Cher protein with severely compromised actin binding [22,31]  In cher CPTI1403 mutants widened Z-discs were evident, together with a widening of Kettin staining and loss of actin staining. CherD CPTI1403 staining at the Z-disc also appeared wider than normal ( Fig 3E). As for Mef2>cher JF02077 , sarcomeres with very high actin accumulations at the Z-disc were also observed occasionally ( Fig 3E). These data together support the notion that the ABD only found in CherD isoforms is required for keeping actin thin filaments stably attached at the Z-disc.

Removal of one copy of Actin88F enhances Cher phenotype
The Actin88F gene produces the main actin protein in the IFM and its expression is also largely restricted to the IFM. Due to these two characteristics, Act88F homozygous null mutants result in viable flightless flies with complete absence of actin in the IFM [33,34]. Act88F null heterozygous mutants display a sensitive dominant flightless phenotype suitable for assessing genetic interactions.
In Act88F KM88 heterozygotes IFM sarcomeres appear relatively normal except for the occasional splitting of myofibrils ( Fig 4A). In contrast, in Act88F KM88 cher Δ5 transheterozygotes sarcomeres are damaged; frayed sarcomeres and widened I-bands are regularly observed ( Fig  4B). To quantify the genetic interaction between Act88F and cher, we counted the number of recognizable sarcomeres compared to areas where myofibrils are too frayed to distinguish individual sarcomeres ( Fig 4C). Consistently, cher Δ5 , cher 1 and cher 1403 enhance the Act88F dominant phenotype, that is, reduce the number of recognizable sarcomeres, whereas cher Q1415sd , a C-terminal filamin mutant, does not genetically interact with Act88F. The specific genetic interaction of the cher ABD mutants with actin further confirms the role of the filamin ABD in tethering actin thin filaments to the Z-disc.  The depletion of all Cher isoforms leads to severe sarcomere disintegration ( Fig 1E). However, cher 1 and cher Δ5 mutants which completely abolish the CherD isoform, only show milder Filamin function in muscles sarcomere defects, suggesting that the remaining isoforms can rescue some aspects of filamin function. To test this hypothesis, we analyzed two mutants that introduce early stop codons at molecularly defined positions: cher Q1415sd introduces a premature stop codon after Ig 15 leaving the last 7 Ig domains untranslated and cher Q1042x introduces a stop codon after Ig 11. Due to the complexity of cher alternative splicing and internal transcription start sites, cher Q1042x TEM images of cher Q1415sd revealed 2 phenotypes in common with Mef2>cher JF02077 : 1) a smaller or fractured Z-disc, and 2) actin invasion of the H-zone (Fig 5A and 5A'). TEM images of cher Q1042x revealed a single phenotype: Z-disc fragments located at the H-zone (Fig 5B and  5B'). The widened Z-disc phenotype was not observed in any of these mutants, suggesting that the N-terminal region before the premature stop codon in cher Q1415sd is able to rescue the widened Z-disc phenotype. To confirm this result we quantified the number of widened Z-disc sarcomeres in different cher mutants using confocal microscopy. Indeed, widened Z-discs were observed with statistical significance only in the ABD mutants cher Δ5 and cher 1403 , but not in cher Q1042x or cher Q1415sd (Fig 5C). Thus, ABD and C-terminal filamin mutants have distinct phenotypes.

All Cher isoforms are required for sustained flight
We noticed that cher mutants were not entirely flightless, in contrast to Mef2>cher JF02077 knock-down flies. Consistent with our previous data, this observation argues that some remaining Cher function is present in our cher mutants. To quantify the flight ability of cher mutants we used an infrared tachometer to monitor the wing beat frequency of tethered flies ( Fig 5D). While control animals can fly at a constant speed, around 200 Hz for several minutes without interruption, cher mutants are not able to sustain flight for more than a couple of seconds (Fig 5E), demonstrating the physiological relevance of all mutants.

Cher and Sls interact in Z-disc cohesion
Resting elasticity of the sarcomere in striated muscle is determined by the giant modular protein titin. In vertebrates, titin reaches from the Z-disc across half the sarcomere up to the Mline. In invertebrates, the function of titin is split between Sls and Projectin. While Projectin associates with the thick filaments in the A-band, Sls spans from the Z-disc to the edge of the A-band [35-37]. Sls isoforms vary highly in size, from 2000 kD to 350 kD, the smallest being Zormin. The main isoform in the IFM is the 500 kD isoform called Kettin [36].
We previously noticed that Kettin staining is affected in some Cher-depleted sarcomeres as well as in Act88F cher Δ5 transheterozygotes (Figs 1G, 3C-3E and 4B). These results suggest that Kettin may be involved in Cher function at the IFM.
We first analyzed sls ZCL2144 , which bears a GFP exon trap, predominantly labels Kettin in IFM and localizes to the Z-disc. It has no muscle phenotype heterozygously ( Fig 6E), but shows IFM defects homozygously [38]. We first recombined sls ZCL2144 with Df(3R)Exel6176, a deficiency uncovering the cher locus. This transheterozygous combination led to mild sarcomere defects consisting of actin accumulation at the H-zone, evident by the lack of separation of individual sarcomeres in actin stainings (Fig 6A and 6G). These actin-related phenotypes suggested that Cher together with Sls/Kettin might be involved in keeping actin anchored at the Z-disc. To further test this, we analyzed the homozygous mutant conditions of our cher mutants. Without the addition of sls ZCL2144 , H-zone actin accumulation is not visible by confocal microscopy in cher Q1415sd or cher Q1042x mutants and present only at a low frequency in cher Δ5 mutants (S3 Fig). Interestingly, cher Q1042x in combination with sls ZCL2144 resulted in a highly significant enrichment of actin at the H-zone (Fig 6B and 6G). cher Δ5 and cher Q1415sd in combination with sls ZCL2144 also led to strong H-zone actin accumulation (Fig 6C, 6D and 6G). In many sarcomeres, the H-zone now looks like the Z-disc, if only the actin staining is considered. We therefore confirmed the location of the Z-disc by using Zasp52 as a second Z-disc marker, showing that Sls/Kettin remains at the Z-disc ( Fig 6F). Our data indicate that Sls/ Kettin genetically interacts with Cher and together with Cher mediates actin anchorage to the Z-disc.
To extend our genetic interaction analysis we tested other sls alleles. First, we asked whether removal of one copy of sls would affect the Mef2>cher JF02077 phenotype. To do so we first analyzed the IFM from sls 1 and sls j1D7 heterozygotes. Both sls 1 and sls j1D7 appear normal and can fly (Fig 7A and 7B). Yet, when these alleles were combined with Mef2>cher JF02077 , most myofibrils fray and sarcomeres disappear (Fig 7C and 7D). This genetic interaction is highly Filamin function in muscles significant for both sls 1 and sls j1D7 (Fig 7E). We also measured the lethality in Mef2>cher JF02077 alone or in combination with different sls alleles. Consistently, all sls alleles significantly enhance the lethality associated with depletion of Cher ( Fig 7F). Lastly, we tried a subtler approach, measuring the interaction using our infrared laser tachometer in transheterozygous flies. The cher mutant heterozygotes as well as sls j1D7 heterozygotes can sustain flight for more than 30 seconds (Fig 7G-7J). Yet, in sls j1D7 cher Δ5 and sls j1D7 cher Q1415sd transheterozygotes flying ability was severely affected, with flies unable to sustain flight for more than a couple of seconds (Fig 7K and 7L). Surprisingly, sls j1D7 cher Q1042x could normally sustain flight (Fig 7M). Heterozygotes and transheterozygotes showed no obvious sarcomere defects, when analyzed by confocal microscopy (S4 Fig). Thus, cher and sls strongly genetically interact across multiple allelic combinations and enhance different cher phenotypes, suggesting that the interaction of titin and filamin is of crucial importance to sarcomere structure. Filamin function in muscles The C-term Cher Ig domains bind the Sls I-band region As Cher and Sls/Kettin colocalize at the Z-disc in the IFM and our genetic interaction data indicate they function jointly, we decided to assess the physical interaction of these proteins. As a starting point, we used the cher CPTI1399 (Flag-Venus tag) allele in combination with sls ZCL2144 (His/EGFP-tag), which allows us to immunoprecipitate Sls as bait protein using Nibeads and detect all interacting Cher isoforms using Flag antibody. In this approach, all Cher isoforms were enriched in Sls-containing beads, compared to the control beads (Fig 8A). This Filamin function in muscles suggested that all Cher isoforms are in a complex with Sls. To test if Kettin, the most common Sls isoform in the IFM, is responsible for Cher binding, we immunoprecipitated cher CPTI1399 with anti-Flag affinity beads and tested for Kettin presence. Again, Kettin was enriched in the Cher-containing beads, suggesting that Cher can bind Kettin in the IFM (Fig 8B).
To independently confirm this interaction, we analyzed the localization of filamin in a titin mutant. In sls ZCL2144 heterozygotes bearing a Flag-tagged allele of Cher (cher 1399 /+) sarcomeres appear normal with both Sls:GFP and Cher:Flag colocalizing at the Z-disc (Fig 8C). However, in sls ZCL2144 homozygotes Sls is reduced at the Z-disc and accumulates into large ectopic protein aggregates (Fig 8D). Intriguingly, Cher-Flag is no longer observed at the Z-disc, but is instead recruited to the large ectopic Sls aggregates (Fig 8D). These results support the notion that Sls binds and recruits Cher into the Z-disc.
We then wondered if we could further narrow our protein-protein binding analysis. Both Sls and Cher are huge proteins not amenable to standard protein purification protocols. However, recent advances in computational approaches to study protein-protein interactions have shown that evolutionarily persistent protein complexes tend to leave a covariation signature [39]. Since protein complexes can be tightly bound, mutations in one component may be compensated with mutations in other components. This evolutionary covariation signature can be thought of as the correlated changes that appear in the coding sequence of two proteins [39][40][41][42]. Furthermore, the strength of the correlation tends to be higher at contact areas [43].
Therefore, we relied on finding the peak of positive evolutionary covariation between cher and sls as a proxy for their potential contact sites (see Materials and methods). We first divided the coding regions of both genes in alignment blocks and then used those blocks to obtain partial covariation signatures. We divided cher and sls in 4 and 250 blocks respectively and calculated the covariation between all the possible combinations. A highly-correlated area was found, comprising 2 cher blocks, encoding Ig 8-14 and Ig 15-22 and the N-terminus of Sls corresponding to the Kettin isoform (Fig 9A and 9B). These results suggest the C-terminal half of Cher binds to Kettin.
Next, we wanted to know which of the two identified Cher areas mediates the interaction with Sls/Kettin. We therefore expressed and purified two regions of Cher matching the highly coevolving blocks, Cher Ig 9-11 and Cher Ig 19-22 and used them to test for binding against Sls protein purified from thorax extracts. Cher Ig 19-22, but not Cher Ig 9-11 interacts with Sls-GFP ( Fig 9C). Thus, the computational results agree with the biochemical data, indicating an interaction of Kettin with the C-terminal Ig 19-22 domains of Cher.

Discussion
The organization of actin thin filaments is central to the function of the sarcomere and therefore of muscles. The Drosophila IFM provides great insight to our current understanding of sarcomere protein assembly and function. Here we investigated the function of Cher, the Drosophila filamin homolog.
Filamin was the first actin filament crosslinking protein identified in nonmuscle cells [4]. Filamins act by crosslinking and stabilizing a meshwork of actin filaments [8]. To date, 17 FLNc myopathy-causing variants distributed all along FLNc have been found [44]. The phenotypes of these mutations fall in several different classes, and it is often unclear how the phenotype is caused at a cell biological level [29,45]. Recently, a FLNc mutant in zebrafish was generated revealing a mild muscle fiber phenotype [46], confirming the role of filamin in vertebrate muscles.
Even though human muscles express only one FLNc isoform of around 290 kD, most Cher isoforms are present in the IFM as shown by using different protein trap lines. Like FLNc in vertebrates all Cher isoforms localize to the Z-disc [46,47]. This can now be explained by the titin interaction domain, which we mapped to Cher Ig 19-22, and which is present in all isoforms. Finally, we reveal that Cher depletion leads to three distinct phenotypes: 1) a widened Z-disc, 2) a smaller or fractured Z-disc, and 3) actin incorporation into the H-zone. Consistent with filamin's known role as an actin-binding protein, these phenotypes are directly linked to the positioning of actin thin filaments.
Distinct phenotypes can be observed in cher mutants affecting different isoforms and the most severe phenotype can be obtained by removing all isoforms using RNAi. We propose that the phenotypic differences are due to Cher truncations retaining partial functions. The best example is the cher Q1042x mutant that creates a truncated form consisting of ABD plus the first 11 Ig domains and leaving the small CherA/B isoforms unaffected: it splits Cher in two halves. Interestingly, cher Q1042x mutants present a weaker phenotype than cher Δ5 and cher Q1415sd . Retaining function in truncated filamins is not unknown. Previous studies have reported that a truncated Cher consisting of the ABD plus the first 6 Ig domains, but without Filamin function in muscles the dimerization domain, is able to rescue most filamin functions in egg chamber formation [19]. We took advantage of these hypomorphic truncations to analyze in detail the separate functions of the N-terminal ABD and the C-teminal Ig domain regions.
We propose that filamin/Cher is located at the Z-disc where it binds actin thin filaments through the N-terminal ABD and titin/Kettin through its C-terminal Ig domains. Cher is present as a homodimer held together by its most C-terminal Ig domain. This configuration allows filamin to bridge titin/Kettin with actin thin filaments from opposing sarcomeres, maintaining Z-disc cohesion both in parallel and perpendicularly to the sarcomere (Fig 10A and 10B).

Filamin-titin binding
The giant elastic protein titin is a 1 μm long flexible filament that spans half the sarcomere. In Drosophila the titin homolog Sls spans from the Z-disc to the edge of the A-band, providing elasticity between the thick filaments and the Z-disc, as a molecular spring [37,48]. Elasticity In cher mutants that disrupt actin binding, perpendicular stability is partially maintained but anchorage to the Z-disc is lost. Thus, an actin-free and widened Z-disc is often observed. (D) In mutants which compromise Sls binding and dimerization, the Z-disc ruptures perpendicular to the myofibril, causing Z-disc material and thin filaments to detach. In all mutants except cher Q1042x , individual thin filaments can invade the H-zone owing to myosin power strokes and improper anchorage of thin filaments at the Z-disc. https://doi.org/10.1371/journal.pgen.1006880.g010 Filamin function in muscles comes initially from the extension of the PEVK region, with the lowest mechanical stability, followed by the N2B region and finally the unfolding of the Ig and Fibronectin domains [49,50].
Titin is believed to function as a massive protein scaffold. Consistently, at least 24 direct human titin ligands have been found, with 8 of them in the Z-disc/I-band region including actin, ɑ-actinin, nebulin and FLNc [51]. Filamin is likewise a large scaffold protein. We show that the link previously described biochemically between filamin and titin is conserved in invertebrate muscles and required for the stability of the Z-disc.
First, using four different assays we show that cher and sls display a strong genetic interaction, suggesting Sls and Cher have a common function in muscles. We then tested for a physical interaction. His-tagged Sls can precipitate all Cher isoforms and inversely, Flag-tagged Cher co-immunoprecipitates Kettin, the most common IFM isoform of Sls. Common to all Cher isoforms is the last C-terminal region containing the last 8 Ig domains. Experiments with bacterially purified Cher Ig domains demonstrate the requirement of the last four Ig domains of Cher for this interaction. Thus, Sls-Cher binding happens near the dimerization domain of Cher. A similar in vitro binding has been shown between the Zis1-2 region of human titin to the Ig 20-24 domains of both FLNa and FLNc [52]. While we cannot rule out an indirect binding, because Sls could only be purified from thorax extracts, our results add that filamin-titin binding is crucial for sarcomere stability and indicates that this interaction is conserved.
Lastly, we provide direct evidence for the function of Cher-Sls binding. Removing the last C-terminal 8 Ig domains of Cher with the cher Q1415sd mutant, thereby removing the Sls-binding site, results in a smaller and fractured Z-disc, and actin incorporation into the H-zone. A smaller Z-disc is a very representative phenotype. Apart from cher, it has only been reported for sls [38].
In summary, Cher binds the I-band region of Sls/Kettin via C-terminal Ig domains; cher and sls genetically interact; and removal of the Sls/Kettin-binding region in Cher leads to a slslike phenotype resulting in Z-discs that are smaller or fractured along the perpendicular axis ( Fig 10D). Our results therefore strongly suggest that filamin and titin are part of a complex mediating Z-disc stability.

Filamin-actin binding
Maintaining actin thin filaments aligned and anchored is the central function of the Z-disc protein complex. Thus, many Z-disc proteins directly bind actin. Not surprisingly, filamin, a well-studied actin-binding protein localizes at the Z-disc where it binds actin through a conserved N-terminal ABD, composed of two CH domains.
We now show that filamin-actin association is critical for Z-disc cohesion, by showing a widened Z-disc phenotype for different cher mutants that specifically affect the isoform containing the ABD or by specifically disrupting the ABD. Further, we show a genetic interaction between these cher alleles and Act88F, confirming a functional filamin-actin link in muscles.
The widened Z-disc phenotype is unique compared to other IFM phenotypes described, in agreement with the proposed specific filamin-actin function in the Z-disc. The Z-disc widens and an actin-free area appears in place of the Z-disc (Fig 10C). However, the adjacent sarcomeres are still well organized and the phenotype is also unevenly distributed across myofibrils. This suggests that filamin is not required for the initial assembly of myofibrils, but rather for maintaining sarcomere structure during repetitive contractile load. In line with this notion, the Z-disc widens parallel to the vector of sarcomere contraction. A somewhat similar widened Z-disc phenotype is seen upon stretching isolated myofibrils using a piezoelectric micromotor [37]. Further, a filamin mutation in medaka, which leads to myofibril degeneration, can be rescued by inhibiting muscle contraction [53], and filamin has been shown to localize to sarcomeric microlesions formed upon strong contraction and mediate repair [54]. This again supports the proposal that stretching forces caused by muscle contractions produce the Z-disc widening phenotype in CherD mutants. This phenotype is specific to mutations of the filamin ABD, while maintaining Z-disc cohesion perpendicular to the sarcomere is likely mediated by both filamin-actin and filamin-titin binding.
Z-discs with higher levels of actin at the Z-disc were sometimes seen upon disruption of Cher ABD-containing isoforms. One explanation for this could be the better accessibility of epitopes for binding of phalloidin in the widened Z-disc sarcomeres. It is well documented that antibodies cannot penetrate IFM myofibrils well, e.g. the myosin antibody stains only at the A/I-junction and the M-line [55]. Tearing and widening of Z-discs may occasionally expose additional epitopes on thin filaments leading to the appearance of higher levels of actin.
Finally, we show that both mutating the ABD in cher Δ5 or the Sls-binding/dimerization domains in cher Q1415sd leads to actin filaments invading the H-zone. We propose that in both cases individual thin filaments are no longer stably anchored at the Z-disc, and can therefore be occasionally moved by myosin power strokes into the H-zone. Cher provides thin filaments with the necessary anchorage and elastic support to remain attached to the Z-disc. Finally, as both cher Δ5 and cher Q1415sd mutants share this phenotype and because the addition of a mutated copy of sls greatly enhances this phenotype, we propose that both titin-binding and actin-binding are required for keeping thin filaments anchored (Fig 10C  and 10D).
Importantly, we show a direct structural role for filamin in addition to the signalling and mechanosensing role ascribed to filamin in muscles so far [7]. We have shown for the first time that filamin crosslinks parallel actin filaments with the widened Z-disc phenotype, in addition to the previously demonstrated perpendicular crosslinking in nonmuscle cells [9]. Related to the Z-disc structure, our data show that Z-discs require much more actin crosslinking than just by ɑ-actinin to withstand the strong contractile forces acting on them.

Fly stocks and genetics
Unless specified, all crosses were done at 25˚C. The following fly stocks were obtained from the Bloomington Drosophila stock center: Mef2-Gal4, cher CPTI1399 , cher CPTI847 and cher CPTI1403 [31]; cher MI07480-GFSTF.0 , an insertion that integrates a Flag and an EGFP tag [32]; the RNAi transgene cher JF02077 ; the sls ZCL2144 allele (Sls-GFP) is a protein trap that incorporates a His-Tag and a EGFP sequence into all annotated sls transcripts [38]; sls 1 (D-Titin 14 ) is an EMS amorph mutant and sls j1D7 is a P{lacW} insertion into an exon encoding part of the PEVK-2 domain that fails to complement sls 1 and sls-uncovering deficiencies [56]; the deficiency line Df(3R)Exel6176 uncovers the entire cher locus and was used in combination with all cher alleles. The RNAi transgene cher KK107451 was obtained from the Vienna Drosophila RNAi Center. The cher Δ5 mutant is a deletion that removes the CherD transcription start site, was made by imprecise excision of a P-element; cher 1 , cher Q1415sd and cher Q1042X are EMS mutants; cher Q1415sd and cher Q1042X introduce a stop codon in positions 1415 and 1042, respectively; all were a kind gift from Lynn Cooley. Cher Protein Trap lines cher CPTI1399 , cher CPTI847 and cher CPTI1403 were obtained from the Drosophila Genomics and Genetic Resources at Kyoto Institute of Technology [31]. The UAS-CherA-GFP stock was a kind gift from Sven Huelsmann [27].

Confocal microscopy
All images were acquired using a 63x 1.4 NA HC Plan Apochromat oil objective on a Leica SP8 confocal microscope. Properly stained muscles were manually selected and aligned so that fibers run left to right. Once a muscle fiber was selected and aligned, random areas were imaged at 9x further magnification corresponding to 420 μm 2 at an image resolution of 1024 x 1024 pixels. All quantifications were done at the same magnification and resolution to assure homogeneity.
We used 488 nm-20 mW, 552 nm-20 mW, and 638 nm-30 mW lasers. Emitted light was detected with PMT and HyD detectors. Laser power was typically set between 1% and 3%, the pinhole was set to 1 airy unit and the gain was set between 700 V and 900 V (PMT) or between 10 V and 100 V (HyD). Smart offset was kept at 0%. Scanning speed was set to 400 Hz. Comparable settings were used for all image acquisitions.

Image quantifications
Missing sarcomeres were estimated as a decrease in the total number of distinguishable sarcomeres per image.
Actin accumulation was obtained by measuring phalloidin-TRITC staining grayscale values in the H-zone divided by the same measurement in the zone between the H-zone and the Iband, where actin is normally present. To ensure homogeneity all measurements were done using a 0.9 x 0.9 μm region of interest.
Cher intensity at the Z-disc was measured using the ImageJ plot profile tool line, which displays the intensities of pixels along a line. The X-axis represents distance (μm). To better compare pixel intensities from different flies, the pixel intensities were first normalized for each individual image. Average profiles for each Cher trap line were computed from 10 images and plotted with RStudio.

Lethality quantifications
Embryos bearing the correct genotype were selected for the absence of GFP fluorescence (TM3, twist-Gal4, UAS-GFP, Sb) and incubated at a controlled temperature. The resulting adults were counted and the lethality ratio was calculated.

Statistical analysis
Statistical significance in all figures was assessed using one-way ANOVA followed by post-hoc Tukey tests (GraphPad Prism 7) and plotted as a box plot with notches representing 95% confidence intervals using the boxplot{graphics} function in R software.
For Sls-GFP purification, 100 adult fly thoraces were homogenized in lysis buffer. Protein extracts were then incubated with prewashed anti-GFP magnetic beads (ChromoTek) for 3 hours at 4˚C. After incubation, the beads were washed twice with wash buffer (20 mM Tris-HCl pH 8, 250 mM NaCl, 5% glycerol, 0.2% Triton X-100). Sls-containing beads were then incubated with bacterially expressed proteins. Following incubation, the beads were again washed twice with wash buffer and bound proteins were eluted by boiling in 2x SDS sample buffer.

Homology modeling
CherD protein structure was calculated through the RaptorX protein structure prediction server using Cher-PA as query and FLNA, B or C as templates. The best resulting structure was analyzed and colored using Chimera UCSC software. To generate CherD 847 and CherD 1403 structures, a predicted protein sequence based on Cher-PA and Venus-GFP was generated according to the protein trap insertion site. CherD 847 introduces a Venus trap in between Ig 11 and Ig 12, while in CherD 1403 , the Venus trap is inserted into the first CH domain. The resulting sequences were modelled using RaptorX [59].

Evolutionary rate covariation
Protein covariation was calculated as previously described [40][41][42]. Briefly, exon-coding regions for each gene were obtained through the Table Browser at the UCSC genome browser (genome.ucsc.edu). BED files containing Augustus exon predictions for each gene from the R5/dm3 genome assembly were sent to the Galaxy website (www.usegalaxy.org). Alignments from 12 Drosophila species were directly obtained from Galaxy-stored MultiZ alignments in MAF format. MAF files were converted to Fasta format and imported to R software using read.dna{ape}. Pairwise distances were calculated for each species pair using dist.dna{ape} and transformed into their relative distances [60]. Finally, the Pearson correlation coefficient was used to compare relative distances using cor{stats} and plotted using corrplot{corrplot}.

Wing beat frequency assay
The wing beat frequency of a fly was determined using an optical tachometer as previously described [58]. In this study, 5-7-day old flies were first glued on pipette tips, followed by measurement of wing beat frequency for 5 minutes using an infrared laser tachometer (Model UT372, Uni-Trend Technology). A 30-second continuous flight window was selected from the 5-minute flight record.