Arginyltransferase regulates alpha cardiac actin function, myofibril formation and contractility during heart development

Protein arginylation is a poorly understood post-translational modification mediated by the arginyltransferase Ate1(Balzi et al., 1990), which transfers arginine (Arg) from tRNA onto proteins(Kaji et al., 1963; Soffer, 1968; Takao and Samejima, 1999). Ate1 is encoded by a single gene that is highly conserved in evolution from yeast to human, and is expressed as one or more isoforms produced by alternative splicing in different mouse tissues(Kwon et al., 1999; Rai and Kashina, 2005; Rai et al., 2006). No other arginyl transferases have been identified in eukaryotic species to date. In yeast, the Ate1 gene is not essential for cell viability, whereas Ate1 knockout in mice results in embryonic lethality and severe defects in cardiovascular development and angiogenesis(Kwon et al., 2002). Multiple proteins are arginylated in vivo(Bongiovanni et al., 1999; Decca et al., 2006a; Decca et al., 2006b; Eriste et al., 2005; Fissolo et al., 2000; Hallak and Bongiovanni, 1997; Kopitz et al., 1990; Lee et al., 2005), including 43 different proteins isolated from various mouse cells and tissues(Wong et al., 2007).

Ate1 knockout mice die between E12.5 and E17, with large hemorrhages, impairment of embryonic angiogenesis, and severe heart defects that include underdeveloped myocardium, septation defects [ventricular and atrial septal defects (VSD and ASD)], and non-separation of the aorta and pulmonary artery [persistent truncus arteriosus (PTA)](Kwon et al., 2002). The underlying molecular mechanisms and cell lineage(s) responsible for these defects are unknown. It has been hypothesized that some, or all, of these defects might be due to impaired embryonic cell migration during heart formation. VSD, ASD and PTA defects are often observed in mice with knockout of genes that affect the migration of cells of the neural crest lineage(Gitler et al., 2002) and cell adhesion (Conti et al., 2004; George et al., 1997; George et al., 1993; Tullio et al., 1997). Cells derived from Ate1 knockout embryos display severe defects in lamella formation that result in impairment of directional migration along the substrate, linked to arginylation of beta actin, a ubiquitously expressed non-muscle actin isoform that plays a crucial role in lamella formation and non-muscle cell locomotion (Karakozova et al., 2006). However, whether the heart defects seen in Ate1 knockout embryos are linked to impairments in cell migration during development, or whether additional cell-autonomous changes in the cells composing the heart contribute to the Ate1 knockout phenotype,remains to be addressed.

It has recently been found that a number of proteins in embryonic and adult mouse tissues are arginylated (Wong et al., 2007). Among these, a prominent arginylation substrate is alpha cardiac actin (also known as cardiac alpha actin, Actc1), the major component of the myofibrils in the cardiac muscle, the arginylation of which is likely to affect myofibril development and function. It is likely that impairment of this arginylation in the Ate1 knockout mouse would lead to defects in myofibril development that would be evident upon comparison of the hearts of wild-type and Ate1 knockout embryos, shedding light on the molecular role of arginylation in heart development and myofibril function.

To address the question of whether Ate1 knockout results in cell-autonomous changes in cardiac myocytes and whether arginylation plays a role in myofibril development and function, we characterized embryonic actin by gel fractionation and mass spectrometry and found that during development,alpha cardiac actin exists in a highly arginylated state and contains a total of four arginylated sites, including two that were previously unknown. The four added arginines are likely to act in conjunction within the folded actin monomer to modulate actin polymerization and co-assembly with other myofibril proteins. Analysis of myofibril development at early stages of heart development, as well as between E12.5 and E14.5 when the phenotypic changes in Ate1 knockout embryos become obvious, showed that lack of arginylation results in delayed myofibril formation and various structural defects in the myofibrils that suggest impairment in cardiac contractility. Studies of cardiac myocytes in culture support these conclusions and suggest that arginylation plays a key role in the functioning of cardiac myocytes. These results demonstrate cell-autonomous changes in cardiac myocytes that develop in response to Ate1 knockout and suggest a key role of actin arginylation in the development and function of cardiac muscle in vivo.

Mice with deletion of exons 2-4 of Ate1 were newly rederived using the targeting vector and strategy described by Kwon et al.(Kwon et al., 2002). PCR genotyping of mice and embryos used for heart and myocyte derivation was performed using a mixture of the three primers 5′-CTGTTCCACATACACTTCATTCTCAG-3′,5′-GGTGCAAGTTCCTGTCTATT-3′ and 5′-AATTGGAGGGGGATAGATAAGA-3′ to detect products of 603 bp(wild-type allele) and 417 bp (knockout allele).

For two-dimensional gel analysis, whole E12.5 embryonic hearts were washed in PBS, flash frozen in liquid nitrogen, supplemented with 50 μl of 2× SDS sample buffer [5% SDS, 5% β-mercaptoethanol, 10% glycerol,60 mM Tris (pH 6.8) (Burgess-Cassler et al., 1989)], homogenized by grinding and pipetting and boiled for 10 minutes for the subsequent electrophoretic fractionation. Complete dissolution of proteins was confirmed by centrifugation of the samples at 13,000 g for 15 minutes and visual confirmation that no pellet was present in the tube. Two-dimensional gel electrophoresis was performed by Kendrick Laboratories(www.kendricklabs.com)with SDS-boiled samples using carrier ampholines with isoelectrofocusing tube gels that enable isoelectrofocusing of samples prepared with SDS buffer as described (Anderson and Anderson,1978). Spots corresponding to individual actin isoforms were excised from dried Coomassie-stained gels and analyzed by mass spectrometry as described (Karakozova et al.,2006; Wong et al.,2007) for protein identification and mapping of the arginylation sites.

For the analysis of the protein composition of the myofibrils (see Fig. S2 in the supplementary material), myofibril isolation was performed as previously described (Meng et al.,1996). In brief, individual E12.5 embryonic hearts were homogenized in 20 volumes (per weight) of buffer I (39 mM sodium borate, 35 mM KCl, 5 mM EGTA, 1 mM DTT, pH 7.1) and centrifuged at 1500 gfor 12 minutes. The pellets were resuspended in 20 volumes of buffer II (39 mM sodium borate, 25 mM KCl, 1 mM DTT, pH 7.1) and centrifuged at 1500 g for 12 minutes. The pellets were further re-extracted for 30 minutes with Triton X-100 buffer (39 mM sodium borate, 25 mM KCl, 1 mM DTT, 1%Triton X-100, pH 7.1) and centrifuged at 1500 g for 12 minutes, followed by a wash with 20 volumes of suspension buffer (10 mM Tris,100 mM KCl, 1 mM DTT, pH 7.1) and centrifugation at 1500 g for 30 minutes. The final pellets containing myofibrils were resuspended in 15μl of 1× sample buffer and boiled for 15 minutes before loading on an SDS-PAGE gel.

Gel scanning and densitometry for determination of the percentages of arginylated actin and the myosin to actin ratio in the myofibril preparations were performed on inverted black and white images of gels, as shown in Fig. 1A, using the `gray level'quantification in the Metamorph imaging software (Molecular Devices).

Whole mouse embryos at E9.5 and hearts excised from E12.5 and E14.5 mouse embryos were washed in PBS and fixed in 2.5% glutaraldehyde and 2%paraformaldehyde in buffer C (0.1 M sodium cacodylate, pH 7.4) overnight at 4°C, followed by two 10-minute washes in buffer C and post-fixation in 2%osmium tetroxide in buffer C. For staining, fixed embryos or hearts were washed twice for 10 minutes each in buffer C, once for 10 minutes in distilled water, incubated 1 hour at room temperature in a 2% aqueous solution of uranyl acetate and then washed twice for 10 minutes each in distilled water. For embedding, stained embryos or hearts were dehydrated by incubation for 10 minutes each in 50%, 70%, 80%, 90% and 100% ethanol, followed by two 5-minute incubations in propylene oxide (PO), overnight incubation in 1:1 PO:Epon(Poly/Bed 812, Polysciences), and then 1 day in 100% Epon. Epon-embedded embryos or hearts were kept for 2 days at 60°C for Epon polymerization,sectioned, stained with 1% uranyl acetate in 50% methanol and with a 2% (w/v)solution of bismuth subnitrite at 1:50 dilution, and then overlaid onto Formvar-coated grids for electron microscopy. Four embryos at E9.5 and four hearts at E12.5 (two wild-type and two knockout each) and six hearts at E14.5(three wild-type and three knockout) were used for the measurements and observations shown in Figs 2, 3, 4 and in Fig. S3 in the supplementary material.

For cardiac myocyte derivation, whole hearts from E12.5 embryos were washed with warm PBS, placed into 200 μl of prewarmed 1% trypsin in PBS and incubated at 37°C for ∼5 minutes. After digestion, heart tissue was gently pipetted up and down until the large aggregates were broken apart, the cell suspension was diluted with 1 ml of DMEM:F10 culture medium supplemented with 10% FBS, centrifuged for 5 minutes at ∼100 g,resuspended in ∼2 ml of the same medium, and plated onto 3.5-cm collagen-coated glass-bottom dishes (Matek). Cells were incubated overnight in culture for attachment and spreading.

For myocyte beat measurements, time-lapse videos of individual myocytes or islands were obtained over 1 minute at half-second intervals (120 frames)using phase-contrast 10×, 20× or 40× objectives and an Orca AG digital camera (Hamamatsu). In each video, beats of individual cells or isolated islands were measured by selecting a 10×10 pixel region in the area of the cell where the gray value (image intensity) displayed obvious changes during each contraction (as shown in Fig. 5C), and total gray value over time in this region was measured using Metamorph. Time intervals between the centers of gray value peaks, which were determined automatically after smoothing the baseline and scaling up the peaks, were used for frequency diagrams and for calculation of means and ratios as shown in Fig. 5. For measurement of calcium changes during beats, the medium in the dishes with cultured myocytes was supplemented with 2.5 μM cell-permeable Fluo-4 fluorescent calcium indicator dye (Invitrogen), followed by a 30-minute incubation for dye loading. Individual cells and islands were then placed into fresh medium and imaged by phase-contrast time-lapse over 25 seconds, acquired at four frames per second, followed by imaging of the same field in the fluorescent channel for the same time interval. Calcium change periodicity was measured by changes in light intensity (gray levels) in a 10×10 pixel region located in the center of a beating cell, and intervals between peaks were determined and compared with those obtained during phase-contrast imaging of the same cells for the correlation plot shown in Fig. 5D.

We recently found that actin is arginylated in vivo and that this arginylation regulates its key properties and functions in the actin cytoskeleton (Karakozova et al.,2006; Wong et al.,2007). One of the two actin isoforms shown to undergo arginylation is alpha cardiac actin, a key heart protein and the major component of the cardiac myofibrils. To address the question of whether alpha cardiac actin is arginylated during embryogenesis and to estimate the percentage of actin arginylated in vivo, we fractionated whole mouse hearts derived from E12.5 wild-type and Ate1 knockout mouse embryos by 2D gel electrophoresis under a pH gradient that allows separation of actin isoforms(Fig. 1A), and identified the actin isoform composition of the individual spots by mass spectrometry. Actin isoforms (α, β, γ) are over 99% identical to each other, with small variations in amino acid composition that lead to shifts in isoelectric point (Otey et al., 1987; Otey et al., 1986; Rubenstein and Spudich, 1977; Vandekerckhove and Weber,1978) that can be seen when fractionation is performed under a shallow pH gradient (pH range 4-8); alpha actin is the most acidic and appears as the leftmost spot(s) on such gels.

Whereas adult hearts express predominantly alpha cardiac actin, a greater variety of actin isoforms are expressed simultaneously during embryogenesis(Hayward and Schwartz, 1986; Otey et al., 1988; Sassoon et al., 1988),resulting in multiple spots in the appropriate pH range of the gel in addition to the major, alpha actin spot (Fig. 1A, top panel). Comparison of the actin spot composition of wild-type and Ate1^-/- E12.5 hearts revealed a striking change in the position of the major isoform spot (α, Fig. 1A), which was significantly shifted to the more acidic end of the range (left), consistent with the loss of several positive charges. Since Arg is highly positively charged, and addition of Arg to proteins is expected to result in a shift in the isoelectric point towards the basic range, the gel shift observed between wild-type and Ate1 knockout hearts is consistent with the addition of several Arg residues to the wild-type actin molecule. To confirm this, we analyzed the major actin spots, excised from the gels similar to those shown in Fig. 1A, by mass spectrometry, using a high accuracy LTQ-Orbitrap MS instrument and a database search algorithm to look for the addition of an extra Arg to the N-terminus of the peptides (Wong et al.,2007). Although the sequence coverage of gel-excised spots was significantly lower than for samples analyzed in solution, analysis of actin spots excised from several different gels resulted in the identification of two previously unknown arginylation sites, one at isoleucine 87 (in pale pink on the diagram in Fig. 1B; this site will be described in detail elsewhere), and one at glycine 152 (for the mass spectrum of the corresponding peptide, see Fig. S1 in the supplementary material). Remarkably, although both sites are located far from the N-terminus and are expected to require a preceding proteolytic cleavage to become accessible for arginylation, they were found in actin excised from an intact 2D gel spot (43 kDa), suggesting that the molecule remained together during SDS-PAGE. The chemistry behind this reaction and the processing that leads to intrachain cleavage and arginylation without destroying the integrity of the molecule will be described elsewhere.

Mapping of all four of the identified arginylated sites onto the actin 3D structure [PDB identifier 1J6Z (Otterbein et al., 2001)] revealed that all these sites are accessible on the surface of the folded subunit, but some sites might not be accessible after polymerization (the two sites on the top), suggesting that modification of these sites could occur after subunit folding but before incorporation into the polymer. The sites are located pairwise on the actin surface, with two arginylated residues directly facing each other in the folded structure(Fig. 1B, in pink within the blue chain of the actin backbone). Insertion of Arg into these positions would be predicted to affect the molecular structure of the actin monomer, its polymerization properties and interaction with other proteins in the myofibril.

Estimation by gel densitometry of the arginylation-dependent gel shifts of the alpha cardiac actin shown in Fig. 1A suggests that as much as 40% of the total embryonic heart actin, and 50% of the alpha actin, is arginylated in vivo. Thus, alpha cardiac actin in embryonic hearts exists in a highly arginylated state and Ate1 knockout results in the abolishment of actin arginylation, which is expected to result in significant structural and functional changes in the actin molecule.

To test whether arginylation knockout results in defects in myofibril structure and/or assembly at the onset and during the progression of the phenotypic changes in Ate1^-/- embryos, we analyzed sections of fixed hearts derived from E12.5 and E14.5 wild-type and knockout littermate embryos by electron microscopy. These stages were chosen because at E12.5, Ate1^-/- embryos appear to be phenotypically normal such that only the most significant early myofibril defects, if any, are expected to be seen, whereas at E14.5, most knockout embryos look grossly abnormal and some of them start to die, so any myofibril defects in response to the Ate1 knockout would be highly prominent at this point.

Examination of the heart structure at lower magnification (×250, not shown) confirmed that Ate1 knockout hearts, consistent with previously reported observations, had thin walls owing to the reduced size of the compact zone of the myocardium (Kwon et al., 2002). Higher magnification views of the large areas of the heart (×2500, Fig. 2)revealed significant structural changes in Ate1 knockout hearts compared with their wild-type counterparts. In wild-type hearts, the development of highly organized, continuous myofibrils progressed robustly from E12.5 to E14.5, whereas in the Ate1 knockout, this development appeared to be delayed, resulting in scarce, disorganized myofibrils, the continuity of which could not be traced through multiple myocytes(Fig. 2, right). The prominence and severity of these defects varied between individual embryos and across the heart areas analyzed, but all the hearts used in this analysis displayed similar trends. In the most severe cases, as observed in hearts taken from E14.5 embryos with prominent phenotypic changes [reduced size, paleness and hemorrhages as described by Kwon et al.(Kwon et al., 2002)], the myofibril structure and connections between the myocytes were disrupted,suggesting that such hearts would be unlikely to function normally. Overall,the observed changes suggested that Ate1 knockout resulted in the delayed development and the disorganization of the myofibrils and disruption of their continuity between interconnected cells, all of which are predicted to affect contractility within the heart muscle.

To address the question of whether such global changes in myofibril layout are accompanied by ultrastructural changes in the myofibrils, we first compared the protein composition of the myofibrils isolated from wild-type and Ate1 knockout hearts at E12.5. Although no consistent changes were detected, the preparations from knockout hearts were somewhat more variable and the ratio of myosin to actin, as quantified by gel densitometry, was slightly lower than in the wild type (see Fig. S2 in the supplementary material), suggesting that at this stage the myofibrils in the knockout hearts might be more structurally unstable than in the control. This instability might be due to specific structural changes in the myofibrils themselves, or to the general onset of heart failure resulting from Ate1knockout.

We next analyzed higher magnification images of the sarcomeres and intercalated disks in wild-type and mutant hearts (Figs 3 and 4). Several structural features of the myofibrils and intercalated disks were evaluated quantitatively,including sarcomere length and Z-band thickness (as measures of sarcomere structure, Fig. 3), the angle of myofibrils and their component filaments at intercalated disks (a measure of myofibril continuity throughout the heart), and cell-cell distance at intercalated disks (a measure of heart muscle integrity)(Fig. 4). In addition, the general appearance of the myofibrils was evaluated. Only the fully developed,structurally distinct myofibrils were evaluated in both conditions, so the analysis presented in Figs 3and 4 does not reflect any measure of the rate of myofibril development, which was difficult to evaluate owing to possible variation in the area of the heart muscle that was used for sectioning in each heart.

Several prominent defects were observed in the myofibrils with the progression of Ate1 knockout-related phenotypic changes. At E12.5,most of the sarcomeres and Z-bands in fully developed myofibrils appeared normal, whereas at E14.5 the hearts with more severe defects showed sarcomere collapse (Fig. 3), with a progressive decrease in the length of individual sarcomeres and diffusion of the Z-bands in a manner that suggested disconnection of the Z-bands from the myofibrils. Some of the myofibrils became wavy and uneven, with prominent loss of Z-bands and fraying of the filaments out of the sarcomeres(Fig. 3D, bottom image). A variety of other structural abnormalities were observed, including myofibril branching at Z-bands, asymmetric sarcomeres with a different density of filaments on the two sides of the same sarcomere(Fig. 3E), and patchy or missing Z-bands (not shown; also observed in the wild type, but more often seen in the knockout).

Defects were also seen at intercalated disks, which serve as the sites of connection of the myofibrils between neighboring myocytes and ensure myofibril continuity throughout the heart. First, in most of the observed cases,including E12.5 hearts in which other ultrastructural features were apparently normal, the filaments within a myofibril approached the intercalated disk at different angles, resulting in tapered (E12.5 and E14.5), frayed and/or disoriented (E14.5) or disconnected (i.e. missing myofibrils on one side of the intercalated disk, E14.5) morphology(Fig. 4E, top three images)that suggested structural and functional discontinuity of the myofibrils between myocytes. In more severe cases, disruption of the intercalated disks themselves was observed (Fig. 4E, bottom image). All these changes suggest contractility defects and disconnection between individual myocytes in the myocardium.

To further address the question of whether the myofibril defects in Ate1 knockout hearts are accompanied by a delay in myofibril development, we analyzed myofibril ultrastructure in the hearts of Ate1 knockout embryos at earlier stages (E9.5). At these stages, no gross phenotypic changes in Ate1 knockout embryos have been reported,despite the fact that Ate1 expression is at its peak(Kwon et al., 2002) and therefore Ate1-related developmental defects should already be initiated. Consistent with the earlier studies (Kwon et al., 2002), no gross morphological changes were seen in E9.5 Ate1 knockout hearts as compared with the control. Both types of heart contained prominent myofibrils as well as disarrayed myofibril-like structures, indicating that at this stage the development of the heart muscle is still ongoing. Comparison of the `least developed' and the `most developed'myofibrils seen in both types of embryo at this stage (see Fig. S3 in the supplementary material) suggested that in Ate1 knockout mice,myofibrils were thinner than in the wild type and contained a smaller amount of prominent Z-bands. To quantify this defect, we measured the length of the Z-bands (a measure of the myofibril thickness and therefore of the extent of myofibril development) in both types of heart at E9.5 and E14.5 and compared the distributions of Z-band lengths in the wild type and knockout at these stages. Although the length of the Z-bands increased considerably at later developmental stages, they were consistently shorter in the Ate1knockout than in the wild type. This effect was especially prominent at E9.5,when the peak Z-band lengths were ∼300 nm in the wild type and ∼200 nm in the Ate1 knockout (see Fig. S3 in the supplementary material). These results suggest that in addition to the ultrastructural changes in the myofibrils seen at later stages, myofibril development is also delayed in Ate1 knockout embryos.

Thus, Ate1 knockout results in delayed myofibril development,changes in sarcomere structure and in a discontinuity of the myofibrils throughout the heart.

To address the question of whether the observed myofibril defects result in changes in the contractility of the cardiac myocytes, we isolated myocytes from E12.5 wild-type and Ate1 knockout embryos and observed their spontaneous contractility in a tissue culture dish. In both cultures, two types of cells were present: flat, well-spread polarized fibroblasts, and less well-spread, mostly hexagonally shaped myocytes that often formed small islands with two or more cells clustering together. Most of the cells with myocyte morphology exhibited spontaneous contractility during the first 2-3 days in culture (the time interval used for the analysis). Fewer myocytes appeared to be beating in the knockout cultures, but this effect was not quantified because of the lack of criteria for a formal quantitative distinction between myocytes and non-myocyte cells in culture. After a few days in culture, most of the Ate1 knockout myocytes ceased to beat,whereas in the control cultures many more cells were still beating (suggesting that the wild-type cells are `sturdier' in these culture conditions and/or more capable of continuous beating), but because by that time the culture dish was usually overgrown with fibroblasts and cells of the myocyte morphology were difficult to identify, this effect was not quantified.

To measure the beating rates of individual myocytes and myocyte islands in culture, we made time-lapse movies of beating myocytes taken at half-second intervals over the course of 1 minute in both cultures during the first 2 days after isolation, and measured the light intensity changes in phase-contrast time-lapse images over time (for the normalized plots of gray level changes over time for individual wild-type and knockout cells or islands, see Figs S4 and S5 in the supplementary material). As this relatively low sampling rate might have precluded us from measuring the actual beating rate at higher frequencies [>60 beats per minute (bpm)], we also manually calculated the number of cells beating at high frequency, and the number of cells displaying visibly irregular beating patterns, i.e. variable intervals and strength between individual beats (Fig. 5 and see Figs S4 and S5 in the supplementary material).

In both cultures, there was a wide range in the beating frequency of individual cells and islands and in the regularity of beats (for wild-type and knockout beating curves, see Figs S4 and S5 in the supplementary material). However, whereas in the wild-type cultures many cells were beating regularly and at a fairly slow rate (3-20 bpm, Fig. 5 and see Fig. S4 in the supplementary material), in the knockout cultures a considerable number of cells beat at high frequency (>20 bpm, Fig. 5 and see Fig. S5 in the supplementary material), elevating the mean bpm by a statistically significant number (Fig. 5A). The range in bpm observed in individual cells or islands displayed a somewhat wider distribution in the knockout cultures, but the difference was not statistically significant (not shown). However, the percentage of cells beating at high frequency and of cells with visible irregularities in their beating curves, from manual counting of the curves shown in Figs S4-S6 (see Figs S4-S6 in the supplementary material), was significantly higher in the Ate1 knockout cells than in the wild type(Fig. 5A). In addition, by visual observation, knockout cells and islands appeared to contract more locally, with variations throughout the cell areas, suggesting local`twitching' rather than a cell-wide contraction, consistent with the defects seen in myofibril continuity at the ultrastructural level. Although the effect was not quantified, these observations suggest that the increased occurrence of high frequency beats in the knockout cultures could reflect an abnormal,fibrillating behavior of the cells.

To confirm that the physical beating frequency measured in our experiments correlates with the changes in intracellular Ca²⁺, we measured calcium waves (using Fluo-4 Ca²⁺ vital dye) and physical beats(using light intensity measurements of phase-contrast images) in the same cells and/or islands and plotted the correlation between the two measurements(Fig. 5D, see Fig. S6 in the supplementary material for the original curves). We found that in most wild-type and knockout cells, beats and calcium waves showed a good correlation, suggesting that the changes in the beating frequency of the knockout cells, despite their origin in structural defects in the myofibrils,were still tied into calcium-dependent regulation. To minimize the UV exposure for live cells in culture, these measurements were performed at a higher sampling rate (four frames per second over 25 seconds), resulting in an elevated average bpm for both cell types (not shown) and in an elevated ratio of beating frequency between wild type and knockout, as compared with those measured in the previous set of experiments(Fig. 5A and Fig. S6 in the supplementary material). This difference suggested that the earlier measurements, taken at a lower sampling rate, might not have reflected the real beating frequencies of the cultured myocytes. The rates detected in our studies are similar to those reported by other researchers studying myocyte beating in culture, but because mouse hearts in vivo beat at a much higher frequency [over 300 bpm (Mai et al.,2004)], it is possible that sampling rates in time-lapse imaging of cultured cells preclude accurate measurement of the beating rate. For this reason, actual bpm values were not used in our analysis, and only the bpm ratios between wild-type and knockout cultures are shown.

Thus, in agreement with our ultrastructural data, we find that Ate1 knockout results in changes in the contractility and beating patterns of cardiac myocytes, suggesting corresponding changes in cardiac contractility in vivo.

Our study demonstrates that knockout of the arginyltransferase Ate1 leads to cell-autonomous defects in cardiac myocytes that develop during embryogenesis, affecting myofibril development and structure, myocardium integrity and myocyte contractility. Although it was known that Ate1knockout results in significant defects in heart development and in a reduced number of cells composing the heart walls, this is the first demonstration that some of the Ate1-regulated defects are directly associated with defects in the cells that compose the myocardium and are responsible for heart contractility.

We have previously shown that two actin isoforms, beta and alpha cardiac,are arginylated in vivo (Karakozova et al., 2006; Wong et al.,2007), and both isoforms are present in the developing heart. In this study, we identify two new sites of arginylation on alpha cardiac actin and provide a possible functional correlation between alpha actin arginylation and myofibril development and function. Our data suggest that an extremely high percentage of cardiac actin (∼50%) is arginylated, supporting the proposal that actin arginylation is crucial for myofibril integrity. It has been shown previously that substitution of alpha cardiac actin with the gamma enteric muscle isoform (presumed to be non-arginylated) leads to altered mechanochemical properties of myofibrils isolated by detergent extraction of the heart muscle (Martin et al.,2002). Although gamma enteric smooth muscle actin could,hypothetically, become partially arginylated during this substitution, this finding is, overall, consistent with our data indicating that the actin arginylation state and the N-terminal actin sequence are important for myofibril function. Furthermore, our data suggest a direct role of arginylation in the regulation of actin function during myofibril assembly and heart development.

Two of the four arginylated sites were identified in actin polypeptides of intact size isolated from an SDS-PAGE gel, suggesting that these modifications exist within the intact, folded actin molecule that maintains its integrity during SDS-PAGE, and that they do not result in protein degradation or disassembly. The chemistry and enzymology of this reaction remain to be studied. The position of the arginylation sites in the actin 3D structure(Fig. 1B) suggests that arginylation of one, or more, of these sites would significantly alter actin polymerization properties and its ability to associate with other subunits during polymerization. Since the arginylated (Ate1-positive) state of actin is the wild-type state, it is logical to assume that the absence of arginylation would result in the reduction of the ability of actin monomers to polymerize and/or interact with other actin-binding proteins, probably by the abolishment of Arg residues from key positions within the molecule.

Our findings that myofibril development is delayed from as early as E9.5 suggest that phenotypic changes in the Ate1 knockout begin much earlier than previously reported. It is conceivable that underdeveloped myofibrils, forced to contract under duress during growth and development of the embryo, become more and more abnormal as development progresses. Such strain-dependent progression would also explain the variability of the myofibril defects in the Ate1 knockout. Indeed, thicker, better developed myofibrils can endure longer, whereas thinner myofibrils disintegrate faster, leading to a variety of phenotypes both in the whole embryos and in individual myofibrils.

The timing of the myofibril defects seen in our study is consistent with the timing of other phenotypic changes in Ate1 knockout mice that become apparent after E12.5, closer to E14.5. Whereas some of the defects described in the current work (disruption of intercalated disks) might constitute secondary defects resulting from the loss of heart integrity,possibly preceding death, others (myofibril disorganization, sarcomere collapse and Z-band diffusion) are likely to be primary defects due to the lack of arginylation of the major structural components of the myofibrils. Since a prominent peak of Ate1 expression is observed at E8-9, after which Ate1 expression declines, and as the phenotypic changes take another 3-4 days to develop (Kwon et al.,2002), it is conceivable that E8-9 mark the time in development when a substantial amount of myofibril proteins are arginylated and stocked up for further use. Actin is a good candidate for this type of regulation because it is a highly abundant, highly stable protein and a major structural component of the myofibrils that, as we have shown, is arginylated in embryogenesis to a significant extent. However, other proteins might also undergo similar regulation.

Among the current models of myofibril assembly during development, the favored model suggests their formation from premyofibrils by association of cytoplasmic actin filaments into minisarcomeres that later align into nascent myofibrils and further mature into myofibrils(LoRusso et al., 1997; Rhee et al., 1994; Sanger et al., 2000; Sanger et al., 2005). Lack of arginylation of actin and other myofibril components could conceivably affect one or more stages in this process, by slowing the polymerization of the initial actin filaments, enforcing their abnormal aggregation instead of normal association, or affecting their interaction with other key proteins that form the core of the myofibril structure (myosin II, capZ, alpha actinin,troponins, tropomyosin and others). In addition to alpha cardiac actin, our previous work showed that talin (which participates in the membrane association of premyofibril complexes), as well as spectrin and filamin (which are generally responsible for the assembly of actin-containing structures),are also targets for arginylation in vivo(Wong et al., 2007). We believe that a combination of these factors leads to a delay in myofibril assembly (Fig. 6), resulting in the formation of fewer myofibrils in the Ate1 knockout hearts than in the control at the same embryonic stage. Since these myofibrils receive the same contractility signals as in the wild-type heart (as follows from our data, Fig. 5D), underdeveloped myofibrils under stress are more likely to develop structural abnormalities,eventually leading to their collapse and disintegration.

Our studies predict that cell-autonomous regulation of cardiac myocytes by arginylation is essential for myofibril stability, heart integrity and for the mechanical continuity of contraction throughout the myocardium. Changes in these processes constitute the underlying causes of congenital heart abnormalities and heart disease in humans. The study of their regulation by arginylation is an emerging field that promises to uncover new molecular mechanisms of heart development and function in normal physiology and disease.

We thank Neelima Shah and Ray Meade for heart sectioning and assistance with electron microscopy; Dr Roberto Dominguez for help with actin structure analysis and the preparation of Fig. 1B; Jon Johansen and the staff of Kendrick Laboratories for 2D gel electrophoresis; Dr Bruce Freedman for helpful suggestions on calcium imaging;Drs Howard Holtzer, Clara Franzini-Armstrong, John Murray, members of the Pennsylvania Muscle Institute and A.K.'s lab for helpful discussions; and Drs Sougata Saha and Satoshi Kurosaka for critical reading of the manuscript. This work was supported by NIH grant 5R01HL084419 and awards from WW Smith Charitable Trust and Philip Morris Research Management Group to A.K. and NIH P41 RR011823 to J.R.Y.