Cbl-Associated Protein CAP contributes to correct formation and robust function of the Drosophila heart tube

The formation of a tube-like structure is a basic step in the making of functional hearts in vertebrates and invertebrates and therefore, its understanding provides important information on heart development and function. In Drosophila, the cardiac tube originates from two bilateral rows of dorsally migrating cells. On meeting at the dorsal midline, coordinated changes in cell shape and adhesive properties transform the two sheets of cells into a linear tube. ECM and transmembrane proteins linked to the cytoskeleton play an important role during these dynamic processes. Here we characterize the requirement of Cbl-Associated Protein (CAP) in Drosophila heart formation. In embryos, CAP is expressed in late migrating cardioblasts and is located preferentially at their luminal and abluminal periphery. CAP mutations result in irregular cardioblast alignment and imprecisely controlled cardioblast numbers. Furthermore, CAP mutant embryos show a strongly reduced heart lumen and an aberrant shape of lumen forming cardioblasts. Analysis of double heterozygous animals reveals a genetic interaction of CAP with Integrin- and Talin-encoding genes. In post-embryonic stages, CAP closely colocalizes with Integrin near Z-bands and at cell–cell contact sites. CAP mutants exhibit a reduced contractility in larval hearts and show a locally disrupted morphology, which correlates with a reduced pumping efficiency. Our observations imply a function of CAP in linking Integrin signaling with the actin cytoskeleton. As a modulator of the cytoskeleton, CAP is involved in the establishment of proper cell shapes during cardioblast alignment and cardiac lumen formation in the Drosophila embryo. Furthermore, CAP is required for correct heart function throughout development.


Introduction
Significant insights in heart development and function at the molecular-genetic level were obtained by study of model organisms. The morphogenetic origin, structure and function of  [55] was provided by A. Paululat (University Osnabrück, Germany). For tissue-specific RNAi knockdown, the driver lines tinC+D-GAL4 and twi+how-GAL4, were generated by combining tinCΔ4-GAL4 [56] with tinD-GAL4 (=TMEN6-GAL4 #52D1; from J. Weiss, OHSU, Portland, USA; [57]) and 2xPE-twi-GAL4 [58] with how 24B -GAL4 [59], respectively. The CAP 42b and the CAP 49e mutant fly strains and UAS-CAP-C [51] were kindly provided by A. Kolodkin (Johns Hopkins University, Baltimore, USA). All experiments were carried out at 25˚C except the RNAi knockdown experiments that were carried out at 29˚C.

Whole-mount in situ hybridizations
Whole-mount in situ hybridizations were performed as described by Lécuyer et al. [60]. Diglabeled RNA sense and antisense probes were generated using the oligonucleotides CAP-fwd: ATGCCCAATAACCGCAATC and CAP-rev: TTTGGACTCGGTTCATTTTC. For probe detection Alkaline phophatase-coupled anti-Dig antibody (Roche Molecular Biochemicals) was used. Images were taken using Zeiss Axiophot microscope system with Nomarski optics.

Drosophila CAP antiserum
To generate a Myc-His fusion protein, the DNA encoding the short CAP isoform CAP-PE/N with three SH3 domains (NP_001137639), was cloned into the pMH vector [61]. Purified Myc-His-CAP-PE/N was injected into rabbits for polyclonal antibody production (BioGenes GmbH, Berlin). The CAP antisera were affinity purified using the recombinant CAP-E/N protein.

Immunoblotting
Protein extracts were isolated from 10 adult flies of the desired genotype. Shock frozen animals were homogenized in 50 μl 2× SDS sample buffer and boiled at 90˚C for 10 min. Insoluble debris were removed by short centrifugation. For SDS-PAGE 10 μl of the resulting extracts per lane were used. The blot was blocked for 30 minutes with 10% dry milk powder (w/v) in phosphate buffered saline (PBS) containing 0.1% Triton X100 (PBT). Primary antibodies were diluted in 5% BSA in PBT and incubated overnight at 4˚C. Antibodies used were rabbit anti-CAP (1:2.000, this study) and mouse anti-H2A (undiluted cell culture supernatant, monoclonal antibody Bx65; M. Frasch, H. Saumweber, unpublished). Alkaline phosphatase-coupled secondary antibodies (Dianova, Germany) were diluted 1:10.000 and phosphatase activity was visualized by colorimetric NBT/BCIP reaction.

Immunohistochemistry and microscopy
Embryos collected from grape juice plates were dechorionated in 50% DanKlorix (Colgate-Palmolive), fixed for 30 minutes in a 1:1 (v/v) mixture of 4% formaldehyde in PBS and n-heptane and devitellinized by shaking in a 1:1 (v/v) mixture of methanol and n-heptane. Dorsal vessels from adult flies were dissected in 1xPBS and fixed for 1h with 3.7% formaldehyde in 1xPBS.
Staining of fixed embryos was as described by Knirr et al. [62]. If necessary Tyramide Signal Amplification (TSA, as indicated below) was performed using biotinylated secondary antibodies in combination with the Vectastain ABC Kit (Vector Laboratories) and fluorescent Tyramide reagent (PerkinElmer Inc.). Preparation of adult heart still inside an open abdomen was as described by Vogler and Ocorr [63]. Fixation and staining of such hearts still in their abdomina was according to Molina and Cripps [22] with the following modifications: Fixation in 3.7% formaldehyde was for 30-60 min at 4˚C, blocking was for 30 min in PBT containing 5% BSA. Following overnight reaction with primary antibodies at 4˚C, samples were washed 3x 15 min in PBT and binding of secondary antibody was for 2 h at room temperature. F-actin was visualized by staining fixed tissue using TRITC-coupled Phalloidin (Sigma, 0.3 μg/ml) in PBS for 1h at room temperature. Following three 15 min washes in PBT abdominal preparations were mounted in 85% glycerol, 4% propylgallate. The following primary antibodies were used: monoclonal antibodies for βPS1-Integrin/CF.6G11 (1:10), Rhea/Talin E16B (1:20, TSA) and Slit/C555.6D (1:10) were obtained from DSHB-NIH, University of Iowa, anti-GFP 3E6 (1:1000) was from Invitrogen. In addition the following polyclonal antibodies were used: rabbit anti-CAP (1:1000, this study), guinea pig anti-Doc2+3 (1:300, TSA; [64]) and rabbit anti-Mef2 (1:750; kindly provided by H.T.Nguyen, University of Erlangen-Nürnberg). Secondary antibodies were obtained from Invitrogen and used in a dilution of 1:1000. Data were taken using a Deltavision deconvolution microscope system (Applied precision, USA) and processed using softworx and imageJ software. The significance of changes in the number of Mef2+ or Doc+ cardioblasts was evaluated with a 2-sided t-test with equal variance. Differences with p-scores<0.05 were considered as significant (p<0.05 ( � ); p<0.01 ( �� ); p<0.001 ( ��� ); S1 File).

Histological sections and electron microscopy
Late stage embryos of the desired genotype were prepared for TEM analysis essentially as described [65]. Following washing with 0.1 M cacodylate buffer, tissue was stained with 1% osmium tetroxide for 1 hour, dehydrated in a graded ethanol series and propylene oxide and embedded in Poly/BedR 812 (Polysciences, Inc., Eppelheim, Germany). Semi-thin cross sections of each 1 μm were made on a Leica Ultracut S ultramicrotome. Sections were stained with toluidine blue for 3 minutes at 70˚C. Image acquisition was on a Zeiss Axiophot microscope system. Ultrathin sections were contrasted with uranyl acetate and lead citrate and examined with a FEI Morgagni electron microscope. Digital images were taken with a Morada CCD camera and the iTEM software (Olympus Soft Imaging Solutions GmbH, Münster, Germany).

Live-imaging of larval heart beat
Live imaging was performed using w 1118 control or CAP 49e mutant L3 larvae carrying the handC3.1-GFP reporter line [55]. Larvae were anesthetized for 3 minutes with Flynap 1 (Carolina Biological Supply Company) according to manufacturer´s instructions and immobilized on double-faced scotch tape on microscopic slides. Following 10 minutes recovery, larvae were imaged immediately on a Deltavision microscope system. Data were recorded as single optical sections with 10 frames per sec (fps) and processed using softworx software (Applied precision, USA). Data were further processed as avi files using imageJ software. From stills of these files the minimal (DCN min ) and maximal (DCN max ) distance of CMC nuclei close to the inner valve cells were measured for each 10 cycles of heart beat per animal. Average and standard deviation was determined for the control (n = 3) and the CAP 49e mutant group of animals (n = 15). From these values percent of apparent fractional shortening (AFS) was determined according to: [(DCN max -DCN min )/DCN max ] x 100%. Following data recording the larvae were carefully removed from the tape and checked for vitality. Data were graphically displayed and significance of the data was evaluated by 2-sided t-test for unequal variances (type 3).

Dye angiography
For dye injections staged pharate adults (<90 h APF) were used. The preparation of flies, the injection process and the movie acquisition were described by Drechsler et al. [66]. Briefly, pharate adults were glued on a glass slide and the operculum was removed to allow recording of the dye accumulation. 1 μg/μl Uranin, a fluorescent tracer solution, was microinjected laterally into the abdomen of the animal. Dye accumulation in the head region was recorded for 180 sec by fluorescence stereomicroscopy with a digital camera at 10 frames/sec. Pixel intensities were measured within the region of interest (R.O.I.; see red outline in Fig 8b) in the fly head using ImageJ-Fiji software. The average from three consecutive frames (0.3 sec) was determined for each animal for a time interval of 180 sec. Background was removed from the data of each animal individually by subtracting the 0.3 sec interval value with the minimal value determined from all time points. The resulting values were graphically displayed for single animals and for the average of each genotype. The p-scores were determined by a twosided t-test with unequal variance (type 3). Differences with p-scores<0.05 were considered as significant (p<0.05 ( � ); p<0.01 ( �� ); p<0.001 ( ��� ); S1 File).

Drosophila CAP is expressed in the dorsal vessel and is located at the periphery of cardioblasts
In a screen for new genes involved in Drosophila heart morphogenesis we found CAP to be expressed in the dorsal vessel (DV). Since CAP is conserved in evolution, and since the zebrafish and mouse CAP orthologues are highly expressed in the heart ( [47]; S. Abdelilah-Seyfried, University Potsdam, unpublished), we decided to further explore the heart specific function of CAP in Drosophila.
The CAP gene spans 38 kbp on chromosome 2R (46F9-47A1) encoding 23 alternate spliced transcripts that encode 20 unique polypeptides (Fig 1a; [67,68]). To detect the CAP protein, we generated an antiserum against the short CAP isoform CAP-E/N, which consists of the three highly conserved SH3 domains (Fig 1a, Inset). According to the gene model this antiserum should detect all annotated CAP isoforms except CAP-PP (Fig 1a). On Western blots with cell extracts of adult flies (Fig 1b; for entire blot see S4 Fig) our CAP antiserum reacts with several (~17) polypeptides (Fig 1b, w 1118 lane 4). This number corresponds to most of the 20 predicted CAP isoforms, as was reported previously for a similar CAP antiserum [51]. CAP 42b and CAP 49e mutant cell extracts show a weak nonspecific immunoreaction with CAP antiserum (Fig 1b, lanes 6 and 7). Extracts of animals with da-GAL4-induced UAS-CAP RNAi (see below) show a >90% reduction in signal intensity (Fig 1b, CAP RNAi lane 2 compare to. w 1118 , lane 4). Together, these results show that our antiserum is specific for CAP.
Next, we studied CAP expression in the Drosophila dorsal vessel (Fig 2, S1 Fig). By whole mount in situ hybridization using an N-terminal probe we find strong CAP expression in the DV of the Drosophila embryo (Fig 2a). The first expression of CAP is detected in embryos at stage 11 in the garland cell primordium (S1a Fig). Starting with stage 13/14 CAP expression is detected in the heart progenitors and the visceral mesoderm (S1b Fig). Thereafter CAP expression in the heart increases up to stage 16 (S1c-S1e Fig). Until stage 16/17 CAP expression in the heart is restricted to cardioblasts (CBs).
In embryos we find CAP protein in the scolopale cells, a component of the chordotonal organ (Fig 2b, open arrow), the muscle attachment sites (MAS ; Fig 2b, arrowheads) and in the DV, in agreement with data published previously [51]. CAP immunostaining of CBs is initially weak but increases during heart development. In stage 16, CAP is present in all CBs of the DV (Fig 2c), but not in the pericardial cells (Fig 2d). CAP is distributed in a granular pattern at the CB periphery. It is enriched at the luminal and abluminal side but is also present in between ipsilateral cells. On the luminal side CAP forms a rather continuous layer compared to the granular staining on the abluminal side. This difference is more obvious in the heart region. Locally, we observe abluminal accumulations of granular CAP staining near the ostial cells (Fig 2c, open arrows). In Z-sections, we see the same principal features of luminal and abluminal staining. Of note, besides the luminal location, CAP staining often reaches into the dorsal contact zones of paired contralateral cells (Fig 2e). In the larval and adult heart CAP is localized at cell contacts of contralateral cardioblasts and in the circular myofibrils of cardiomyocytes (Fig 2f-2h

CAP mutants display abnormal cardioblast cell number and alignment
To investigate the role of CAP during heart morphogenesis we used the molecularly characterized P-element excision mutants CAP 49e and CAP 42b . Both mutants carry C-terminal deletions   Fig 1a; inset). Western blot analysis and immunostaining indicates the absence of most CAP protein isoforms in these two alleles (Fig 1b). However, as mentioned, our antiserum is directed to the CAP C-terminus and therefore we cannot exclude the presence of isoform CAP-PP in these mutants (see Discussion). With this caveat in mind we studied the morphology of the DV of CAP mutants using anti-Mef2 antibodies, which stain all CB nuclei in combination with anti-Doc2+3 antibodies, which stain the ostial cell nuclei only (Fig 3).
In w 1118 controls or heterozygous CAP mutants the CBs line up in two rows that pair exactly with their contralateral partners at the dorsal midline (Fig 3a). A segmental organization of the DV becomes apparent by molecular markers that is exactly matched between opposite CB rows. Wildtype type hemisegments contain two Doc/Svp-positive cells (stained for Mef2 and Doc) and four Tinman-positive cells (stained by anti-Mef2 only; Fig 3a). When homozygous, both CAP mutants show a DV phenotype characterized by irregularities in the alignment of CBs, blister formation, occlusion of CBs between the adjacent rows of cells and gain or loss of CBs. Instead of two contralateral CBs, often three cells are meeting at the dorsal midline, with one or more central CBs separating peripheral CBs (Fig 3b and 3c). In addition, CAP mutant DVs frequently contain additional CBs in several hemisegments and may have 6 to 8 Tinmanpositive CBs per hemisegment (Fig 3b and 3c; Table 1). In a t-test of increased numbers of Mef2+ cells compared to w 1118 the p-score for CAP 42b is p = 7.00E-09 ( ��� ) and for CAP 49e p = 2.98E-05 ( ��� ; S1 File). Occasionally we observed an increased number of Doc/Svp-positive CBs as well (Fig 3b and 3c). More rarely we also found hemisegments with less than 4 Tinman-positive CBs (arrows S3a Fig). The DV phenotype is already expressed during CB migration (S3a Fig).
The DV phenotype was observed in 63% of CAP 49e and 64% of CAP 42b embryos (Table 1). However, it is variable in expression. In the CAP mutant embryos with recognizable aberrations, the phenotype ranges from embryos in which only one segment is affected to clearly abnormal hearts which feature more irregularities. We did not further subdivide recognizable phenotypes according to their strength. On the other end, some of the subtle defects may not have been recognized and thus classified as wildtype-like (see Discussion). Both CAP mutations heterozygous over Df(2R)BSC281 show a similar DV phenotype (deviations in alignment and cell number) as the homozygous CAP mutants with a slightly reduced penetrance (Fig 3d  and 3e; Table 1). However, in a t-test for the number of Mef positive CBs we see no significant difference between the CAP/Df(2R)BSC281 heterozygotes and the corresponding homozygous CAP mutants (p-scores for CAP 42b /Df(2R)BSC281 are p = 0.1759 and for CAP 49e /Df(2R) BSC281 p = 0.6491; S1 File). Homozygous Df(2R)BSC281 animals survive as stage 17 embryos and can unambiguously be identified by the absence of CAP or Nidogen (Ndg) staining (Ndg, is also removed by Df(2R)BSC281). 72% of Df(2R)BSC281 homozygotes (n = 18) show a DV alignment phenotype (Table 1). Probably the reduced penetrance in this case is caused by maternal effects of heterozygous female parents, however, this was not tested. To further corroborate the CAP DV phenotype we depleted CAP by RNAi knockdown. The UAS-CAP RNAi construct induced by the ubiquitous driver da-GAL4 strongly reduced CAP expression in adults (Fig 1b, Table 1); bar 10 μm. https://doi.org/10.1371/journal.pone.0233719.g003

PLOS ONE
Drosophila CAP contributes to correct heart formation and function described for CAP mutants. Following induction by tinC+D-GAL4 41% of the embryos show this phenotype. Induction by twi+how-Gal-4 produces a DV phenotype in 64% of randomly tested embryos. However, the number of Mef2+ CBs is closer to w 1118 (S3 Fig; Table 1). The phenotypic effects of CAP 49e and CAP 42b are partially (~50%) suppressed by overexpressing the CAP-PC isoform via the CB -specific driver tinCΔ4-Gal-4 (Fig 3f, Table 1). Under rescue conditions (CAP 49e ;UAS-CAP-C/tinCΔ4Gal4) the number of Mef2+ CBs is closer to the w 1118 than with CAP 49e (p-score CAP 49e ;UAS-CAP-C/tinCΔ4Gal4 p = 0.0956 and CAP 49e p = 2.9807E-05; S1 File).

CAP is required for heart lumen formation
Next, we asked whether a normal heart lumen is formed in CAP mutants (Fig 4). To compare the ultrastructure of the CBs in normal and mutant DVs in terms of size, shape and lumen formation, we analyzed wild type and CAP mutant embryos on serial semi-thin cross sections by light microscopy (Fig 4a) and ultrathin sections by TEM microscopy (Fig 4b). In w 1118 control animals the paired contralateral CBs are sealed at the dorsal and ventral junctional region and the intercellular space remaining between these connections establishes a DV lumen (Fig 4a,  arrow). In contrast, about 60% of CAP mutant embryos reveal a defect in lumen formation. In both CAP 49e and CAP 42b mutants, the CBs meet each other normally at the dorsal midline, but form a slit-like lumen or several split lumina, both much smaller than in the control (Fig 4b;  arrows). Serial semi-thin cross sections show that the lumen phenotype is consistent along the length of the aorta and the heart proper region (Fig 4a and 4b). The lumen defects are not exclusively connected to misplaced CBs since they also occur in areas with just two contralateral CBs (see Fig 4e). As evident in high resolution TEM, the typical crescent like shape of the CBs is distorted in CAP mutant animals. Especially the curved luminal domain seen in the controls seems to be disturbed. The cells have a more half moon-like appearance and enclose a slit-like lumen, as is emphasized in the corresponding cartoons (compare Fig 4c to 4d and 4e). Among the factors that affect CB shape and lumen formation are components of the basal membrane as well as molecules that guide CB polarization or organize membrane domains and their attachment to the cytoskeleton. According to our TEM data, CAP appears to be largely dispensable for basal membrane formation since an apparently normal basal membrane ensheaths the lumen of the DV in CAP mutants (Fig 4f-4h).

CAP is required for fine tuning the localization of Slit and Integrin at the luminal domain
To find out whether CB polarization or the distribution of typical luminal marker proteins depends on CAP, we first analyzed the distribution of Slit and βPS1-Integrin in w 1118 controls and CAP 49e mutants (Fig 5). In controls the CAP protein is localized to the luminal, abluminal and lateral membranes (Fig 5a and 5g) and the luminal marker Slit is localized predominantly at the luminal domain of the CBs (Fig 5b). Luminal Slit is co-localized with CAP protein (Fig  5c). In CAP 49e mutants no CAP protein can be detected (Fig 5d). Although Slit is still polarized and detected at the luminal domain of CAP 49e CBs, its distribution is less distinct than in the control. As a consequence the boundaries towards the lumen become ill defined (Fig 5e and  5f). In Z-sections, the lumen, if visible at all, appears to be much smaller than in the controls (Fig 5e´and 5f´). βPS1-Integrin in the control localizes to the luminal and abluminal membrane domains of CBs and is co-localized with CAP at these sites (Fig 5h and 5i). CAP 49e mutants still display a polarized βPS1-Integrin localization, but similar to Slit, its distribution looks ill defined and the boundaries of the lumen are less well defined in comparison to the controls (Fig 5k and 5l). Co-staining of Slit and anti-βPS1-Integrin in the two CAP mutants with anti-Mef2 shows that the disruption of Slit and Integrin localization is stronger when associated with misaligned CBs. In such regions, Slit spreads into the basolateral space between peripheral CBs and around the central CBs as granular material that occasionally forms extra minilumina (Fig 5m and 5n). Similar spreading and minilumen formation associated with central CBs is also observed for ßPS1-Integrin (Fig 5o and 5p). This irregular organization often spreads along the heart tube from the misaligned CBs into immediate adjacent regions (Fig 5n and 5p).

CAP genetically interacts with the integrin signaling components scab and rhea during heart development
A recent study on CAP function implicates that CAP is acting downstream of Integrin signaling at muscle attachment sites and interacts with Vinculin [51]. Due to the co-localization of CAP and βPS1-Integrin in the Drosophila DV we asked whether CAP genetically interacts with Integrin signaling components during heart development (Fig 6). Previous studies

PLOS ONE
reported that heterozygous scab and rhea mutants show normal DV development [18]. Heterozygous CAP 49e mutants also have a normal DV morphology (Fig 6a). Interestingly, double heterozygous animals, with CAP 49e either combined with scab 2 or rhea 1 show defects similar to those seen in CAP homozygous mutants (Fig 6). 48% of randomly selected scab 2 /CAP 49e double heterozygous embryos show a DV phenotype (Fig 6b; Table 1). In the case of rhea 1 / CAP 49e embryos 47% show a DV phenotype (Fig 6c; Table 1). This demonstrates that CAP genetically interacts with scab and rhea during heart development. As for Integrins, this observation is corroborated by our finding that Rhea/Talin localization is disrupted in a homozygous CAP mutant background. In control animals, Talin is localized at the luminal and abluminal periphery of CBs, demarcating a clearly defined cytoplasmic space (Fig 6d and 6e).
In CAP 49e mutant animals the luminal Talin staining is similarly ill defined as was observed for Slit and ßPS1-Integrin in CAP mutants (Fig 6f and 6g). On the abluminal domain Talin distribution is also altered and less distinct.

CAP mutant larvae and adults show an impaired heart function
We argued, that the defects we have seen in CAP mutant embryos would affect the heart function later on. Therefore, we assayed the heart function in two different ways. First we checked for the heart contractility by performing live imaging of L3 larvae carrying the cardiac handC3.1-GFP reporter. Using this reporter, cycles of heart contraction and expansion can be observed and recorded in living larvae by the movement of CMC nuclei. We used the region immediately posterior to the internal valve to record the contraction cycles of the posterior heart as shown for the w 1118 control (Fig 7a, S1 Movie). By measuring the distance of nuclei of contralateral CMCs from stills of the movies, the heart diameter at maximal extension, DCN max (Fig 7a, top) and minimal extension DCN min (Fig 7a, bottom) was determined and used to calculate the percentage of apparent fractional shortening (AFS% = [(DCN max - DCN min )/DCN max ] x 100%; for details see Materials and methods). In control animals (n = 3) we observe an average fractional shortening of 48.6 ± 8% (AFS ranging from 39.2 to 54.6%). By contrast, in CAP 49e mutant larvae (n = 15) the CMC extension is significantly reduced (Fig  7b; S2 Movie), resulting in an AFS of 20.7 ± 18.3% for these animals (Fig 7c; ranging from 0.3 to 54.3%). Data analysis by t-test demonstrates that the difference between the controls and CAP mutants is statistically significant (p = 0.0044 ( �� ); S1 File). Most of the mutant larvae display a clear reduction of the AFS with five out of 15 CAP 49e larvae featuring less than 5% of the average of the controls and eight larvae feature intermediate AFS phenotypes (see S1 File for details). Three out of 15 CAP 49e larvae (20%) exhibit an AFS in the range of the w 1118 controls consistent with the incomplete penetrance of the phenotype in the DV of embryos. Our second assay examined the adult heart function after the heart remodeling at metamorphosis. For this assay we injected a fluorescent dye as a tracer into the abdomen of late pharate adults (Fig 8a; [66]). Due to active pumping by the heart, the dye accumulates with time in the head region of the flies, which can be observed as an increase of the fluorescence emission in a defined region of interest (ROI), as seen for a w 1118 control in Fig 8b and 8c. By integration at defined time intervals and calculating the average pixel intensity, the accumulation of the tracer in the ROI is quantitated and displayed graphically for single flies as seen for three control flies (Fig 8e). The average and standard deviation for the group are plotted separately in Fig 8d. The same data is displayed for CAP 42b and CAP 49e mutants (Fig 8f-8i). Compared to the control at 180 sec, with an average pixel intensity of 61.3 (n = 3), the transport efficiency of CAP 42b (average pixel intensity 19.8; n = 7) and CAP 49e (average pixel intensity 18.8; n = 6) is strongly reduced. The differences between the w 1118 control and the CAP mutants are statistically significant (p = 0.0072 for CAP 42b ( �� ) and p = 0.0039 ( �� ) for CAP 49e ; see S1 File). However, considering single animals we observe differences in transport capability between tested animals. For instance, animals CAP 42b #2 and #4 had almost lost activity, whereas CAP 42b #3, #5, and #7 showed an intermediate phenotype, and CAP 42b #6 behaved like wild type (Fig 8g). A similar variation in phenotypic strength is also seen with CAP 49e (Fig 1i). This observation is in agreement with the variability of phenotypes observed in embryonic and larval CAP mutant hearts.

CAP mutants show local aberrations in the morphology of the adult heart tube
Normally CAP is colocalized with ßPS1-Integrin near Z-bands/costameres and cell contacts between contralateral cardiomyocytes (S2m- S2o Fig). Therefore we wondered whether the observed impaired heart function in CAP mutants results from changes in the adult heart structure, as a consequence of the disruption of Integrin adhesion complex (IAC; Fig 9). In contrast to the w 1118 control, where the contralateral cardiomyocytes (CMCs) form a continuous spiral pattern of myofibrils stained for F-actin by phalloidin (Fig 9a) and a smooth line of cell attachment sites stained by Integrin (Fig 9b, open arrow), CAP mutant hearts often exhibit local structural aberrations with clusters of actin and Integrin elements (Fig 9c and 9d, white arrows). Optical sections of the w 1118 control at higher magnification (Fig 9e-9g) reveal the regular arrangement CMCs with spiral myofibrils (Fig 9e and 9f) that line up at cell contact sites (Fig 9e and 9g) and are continued by matching myofibrils in the contralateral cell. A smoothly stained line of ßPS1-Integrin demarcates these cell contact sites between contralateral CMCs. Integrin is also found at the costameres of circular and longitudinal myofibrils (Fig  9g). In the CAP 42b mutant (Fig 9h-9m) we do not observe obvious general defects in the Zband architecture and costamere formation of myofibrils of CMCs. However, locally we detect aberrant, misaligned CMCs with a disrupted organization of βPS1-Integrin at cell contact sites

PLOS ONE
Drosophila CAP contributes to correct heart formation and function (Fig 9j and 9m). In contrast to the controls, ßPS1-Integrin is found in such cells in clusters of increased Integrin accumulation (white arrowheads in Fig 9m), interrupted by sections that appear devoid of the protein. Interestingly, amorphous actin accumulations (Fig 9i and 9l) often form near such integrin clusters (yellow arrowheads in Fig 9l). Furthermore, CAP 42b mutants feature aberrant CMCs in which the myofibrils fail to align in register and often do not attach at cell contact sites (Fig 9i and 9l). Instead they bend back into the cytoplasm or run for a while parallel to the long axis of the heart (white arrows in Fig 9l), or multiple myofibrils or amorphous actin fibers initiate from ßPS1-Integrin clusters in the mutant (yellow arrowheads in 9l). Together, the Integrin mislocalization correlates with a disturbed architecture of the circular myofibrils and a discontinuity in their arrangement across contralateral CMCs (Fig 9k). These local defects at cell contact sites could break the structural continuity between contralateral CMCs that is important to transmit coordinated contraction.

Discussion
As mentioned in the introduction, CAP has been reported to function in a variety of cell types. The data presented in our paper provide evidence that CAP plays a role in the robustness of Drosophila heart development and function. CAP mutant embryos show a DV phenotype characterized by local abnormalities in the alignment and numbers of CBs (often manifesting in luminal location of one or several CBs following closure of the dorsal vessel) as well as defects in lumen formation. Taken together with the protein localization data, this implies that CAP functions in processes near or within the membrane of cardiac cells. We suggest that the most important part of cardiac CAP function is related to the Integrin Adhesion Complex (IAC) for the following arguments. Apart from our observation that CAP mutants do not display breaks in the continuity of adjacent CBs, the DV phenotype is similar to that of homozygous slit, integrin, and talin mutants, double heterozygous mutant combinations among themselves or combinations with mutations in other Integrin-signaling components ( [18,27], and this work). This includes "clumps" [27] which we interpret as sites of misaligned CBs often forced into the DV lumen (see Fig 1H in [18]). In stage 15-16 embryos CAP colocalizes with ßPS1-Integrin (this work , Fig 5g-5i) and overlaps with Talin at the CB periphery [38]. Furthermore, we observed aberrant DV phenotypes, and hence a genetic interaction, in embryos double heterozygous for either CAP 49e and the αPS3-Integrin-encoding scb 2 or CAP 49e and the Talin allele rhea 1 . The interaction with Scab may be cell non-autonomous. CAP/Vinculin interactions were documented in vitro [44] and more recently genetically in Drosophila muscle attachment sites in vivo [51]. We propose that CAP contributes to Integrinactin cytoskeletal interaction to stably link the chain of CBs by cell adhesions and by adhesion to a common extracellular matrix (ECM). Local weakening or improper regulation of adhesiveness, possibly through a failure to transmit adhesion related forces or to coordinate organization of membrane microdomains, could result in mispositioning of migrating CBs relative to their neighbors.
The effect of the two CAP alleles used is not fully penetrant and is expressed as a series of intermediate states that differ in strength between individuals. Both alleles are small deletions removing two of the three SH3 domains common to all CAP isoforms except CAP-PP (Fig 1a  inset; [51]). Bharadwaj and colleagues [51] suggest that both alleles likely result in a complete loss of CAP protein (amorphic type). We favor the idea that CAP 42b and CAP 49e are of nonamorphic type, consistent with the phenotypic effect observed. Although the deletions in both alleles affect the C-terminus of most isoforms, except CAP-PP, they leave their N-terminal portions intact, which, like CAP-PP, might have residual activity. In addition to CAP-PP, Cterminally truncated isoforms could be variably expressed in the CAP42b and CAP49e alleles and mediate a partial CAP function leading to variable mild, dosage-sensitive effects. A mild, dosage-sensitive antimorphic character would also explain why the rather subtle morphological phenotypes are not enhanced in hemizygous embryos with the CAP-deleting deficiency Df (2L)BSC281. Specific antisera are required to detect such putative truncated protein isoforms. Unfortunately, all available CAP antisera are directed to the conserved CAP C-terminus that is affected by the CAP excision mutants.
Currently, we can only speculate what causes the observed increase of CB numbers in CAP mutants. CB number has been shown to be positively and negatively regulated by the antagonistic actions of receptor tyrosine kinase (RTK/MAPK) and Notch signaling, respectively [69][70][71][72][73]. It is conceivable that CAP could influence signaling efficiency either through its proposed impact on cell adhesion or by interfering with the activity or distribution of certain signaling components (e.g. via its SH3 domains). Interestingly, the vertebrate CAP-homolog CAP/Ponsin negatively regulates MAPK signaling [44] by binding to components of the MAPK-pathway [74]. Alternatively, Integrin dependent adhesion, mediated by cell-cell or cellextracellular matrix contacts, could serve as a mechanism to control cell division as has been reported in many mammalian cell types (reviewed in [75]). Changing the stability of the IAC by CAP mutation may interfere with such controls and trigger signals to initiate unscheduled cell division. This aspect of the cardiac CAP phenotype needs further investigation, although such investigations may be complicated by the subtleness of the phenotype and the complexity of CB number-determining signaling processes.
In the process of tube formation the CBs change their cell shape and cell surface properties. This requires a dramatic reorganization of their ECM, adhesion molecules and cytoskeletal organization [17]. CAP plays an important role in these processes since CAP mutants result in strong defects in lumen formation. CAP is most prominent at the luminal periphery of CBs where it colocalizes with ßPS1-Integrin, Talin and Slit, which are all essential for lumen formation. Integrin is the crucial player since it counteracts luminal spreading of Discs-large to the luminal domain and supports luminal location of Robo/Slit and the formation of a non-adhesive luminal ECM [18]. In CAP mutants the polarized Integrin distribution is maintained but the luminal domain of CBs stained for Integrin looks ill-defined and more expanded than in controls. This is similar to the aberrant broad distribution of ßPS1-Integrin at muscle attachment sites of somatic muscles in CAP mutant animals [51]. Interestingly, Talin is affected in a similar way. In contrast, CAP localization is not dramatically changed in zygotic rhea 79 mutations; the observed changes were interpreted as secondary to the general perturbations in heart morphology by the rhea 79 mutation [38]. In summary, this suggests a role for CAP in fine-tuning integrin-dependent adhesion that does not affect CB polarization and is upstream or independent of Talin-mediated functions at the luminal domain. Slit localization, though perturbed, is still observed at its original domains and an apparently normal basal membrane lining is still formed between contralateral CBs. Therefore, the adhesive properties of the luminal ECM might be unchanged, suggesting, that CAP mutants fail to establish a lumen not because of changes in adhesiveness, but owing to defects in the cytoskeletal architecture relevant for the formation of crescent shaped CBs. Due to defective Integrin assembly or its coupling to the cytoskeleton, CAP mutant CBs may be impaired in forming the functional connections between focal adhesions and the actin cytoskeleton that are required for the establishment of crescent cell shape. This is consistent with the observed role of CAP for the induction of stress fibers and focal adhesions in various mammalian cell types, and fits well with the proposed role for CAP and other mammalian SORBS family proteins in stiffness-sensing and force transmission [47,76].
In the transition of the late embryo to the 1 st instar larva and during pupation the dorsal vessel undergoes terminal differentiation to form the contractile larval and adult heart, respectively. This includes the establishment of robust links between the cytoskeleton and myofibrils and the cell contacts between contralateral CMCs. These contacts are structurally and functionally similar to intercalated discs in the vertebrate heart, which sense and transmit tension across the heart muscle [77]. Integrin and Talin play an important role in establishment and lifelong maintenance of CMC adhesion and their loss of function results in a severe retraction of myofibrils, losses of contacts between ipsilateral and contralateral myocytes and severely impaired function of the adult heart [6,7]. CAP staining closely correlates with the Integrin localization at contact sites of contralateral CMCs and near Z-bands. Integrin is thought to contribute to the costamere structures that link the Z-bands to the ECM (reviewed in [35]), supported by proteins like ZASP that bind to and organize Z-bands and their Integrin connection [78]. The strong CAP-ßPS1-Integrin colocalization near Z-bands and in circular myofibrils suggests association of CAP with costameres, consistent with earlier observations made by Zhang and coworkers for the CAP homolog in mouse CMCs [48]. However, this has to be established by using more protein marker combinations at higher resolution.
Our physiological tests of CAP mutant larvae and adults uncovered animals with a reduced heart function. Apparent fractional shortening in larvae and pumping efficiency in adults were strongly affected in CAP mutants, although both parameters showed a similar incomplete penetrance as was observed for morphological defects in embryos. The physiological defects in heart function correlate with a locally defective morphology of adult CAP mutant hearts. A similar phenotype was found following Talin RNAi, when the knock down was limited to 2 nd larval stage [7]. Also, hand-GAL4 driven knock down of Integrin-linked kinase ILK showed a similar phenotype [6]. Our data indicate that CAP is required for the maintenance and correct assembly of IACs at contact sites of contralateral CMCs and their connections to myofibrils. This function is in correspondence with its suggested role of CAP at muscle attachment sites of somatic muscles [51].
Overexpression of IAC components in Drosophila hearts results in premature aging and the age dependent increase in cardiac arrhythmias that could be prevented by reducing the gene dosage in heterozygous ilk and mys animals or by moderate cardiac specific knockdown of other IAC components [6]. Elimination of Talin1, one of two Talin genes in mice, also reduced the hypertrophic responses to increased cardiac load [79]. However, strong downregulation of Talin and other major components of the IAC cause serious defects, similar to dilated cardiomyopathy [6,7]. Compared to other IAC proteins CAP depletion has rather mild effects on Drosophila heart structure and function. This and the incomplete penetrance of the phenotype classify CAP as a conserved modifier of IAC function that may provide a useful tool in the therapies of age dependent heart diseases.  Table 1). Line 32 displays the average for the CB numbers of the given genotype, followed by corresponding the standard deviation (S) in line 33 and the variance (S) in line 34. Line 35 lists the p-scores obtained by a 2-sided t-test with equal variance (type 2) in relation to w1118; line 36 the same in relation to CAP 42b , line 37 in relation to CAP 49e and line 38 in relation to CAP RNAi control without GAL4-driver (compare Table 1). 2. Evaluation of heart contraction cycle by live imaging. 2.1. lists the maximal (max) and minimal (min) distance of CMC nuclei close to the inner valve cells for hand-GFP;w 1118 controls measured for each 10 cycles of heart beat per animal. The length determined for DCN max (max) and DCN min (min) are listed next to the avi file name and time point of the records in sec. Average and standard deviation for DCN max and DCN min were calculated. 2.2. Same is displayed for hand-GFP; CAP49e; 2.3. Evaluation of the data. Listed to the left are the values of AFS[%] calculated from the average DCN max and DCN min for each hand-GFP;CAP 49e and hand-GFP; w 1118 larva according to: [(DCN max -DCN min )/DCN max ] x 100%. From this the average and standard deviation for the whole group was determined for each genotype, listed in the table and used to calculate the variance. Then the p-scores were calculated using a 2-sided t-test with unequal variance (type 3). 3. Evaluation of heart transport activity of pharate adults determined by dye angiography. 3.1 Transport rate as average pixel intensity in ROI (w 1118 ) lists the pixel values determined for three w 1118 adults at different time points (for details see Materials and methods section) along with their average and standard deviation. 3.2 shows the same for 7 homozygous CAP 42b animals and 3.3 the same for 6 CAP 49e animals. These single fly values were graphically displayed in Fig 8e, 8g and 8i. 3.4 lists the average and the standard deviation of pixel intensities for each genotype that were graphically displayed in Fig 8d, 8f and 8h. From these data the p-scores to w 1118 and significance were evaluated by a 2-sided t-test with unequal variance. (PDF) 48. Zhang