Native cardiac thin filaments from zebrafish

The zebrafish thin filaments in this study were isolated directly from the adult cardiac muscle, thus the native filaments were subject to any naturally occurring post translational modifications. Additionally, no cross-linking or modifications were used to enhance the stability of the regulatory proteins or flexible domains. Cardiac muscle from 14, 6-month-old wild type zebrafish (Fig. 1A) were pooled to obtain sufficient tissue for the isolation. The use of the VPP (Fig. 1B) allowed for the identification of filaments decorated with troponin and tropomyosin (Fig. 1C). However, the percentage of decorated filaments actin was ~ 10%, as discussed above: the dissociation of the thin filaments upon isolation is a known issue (Yamada et al. 2020). Protocols to isolate zebrafish myosin filaments were adapted (Gonzalez-Sola et al. 2014) and manual selection of decorated thin filaments carried out using previously established techniques (Paul et al. 2010; Yang et al. 2014). The level of decoration of negatively stained thin filaments can be seen in Fig. 1. Particles were centred on the troponin complex and with only ~ 10,000 particles we reconstructed a 480Å long segment of the thin filament (EMD-15,901). The Fourier shell correlation was calculated from two independent half maps to obtain an estimated overall resolution of 15.4Å for the final map (Fig. S1). As expected, the more rigid structure of the actin backbone is better resolved than the flexible linker regions of troponin, this is shown in Fig S2 where the reconstruction has been filtered to the local resolution. To avoid over interpretation of the density attributed to the regulatory proteins in Figs. 2, 3 and 4 the map was filtered to 15Å. A comparison of the human reconstituted reconstruction filtered to 15Å resolution is shown in Fig. S3.

Fig. 1

Cryo-EM of zebrafish cardiac thin filaments. A: An excised adult Zebrafish heart, letters (in yellow) show V-Ventricle, A-Atrium B-bulbus arteriosus. B: A typical negative stain EM micrograph of isolated thin filaments, visible troponin complexes indicated with white arrows. C: Cryo-EM image illustrating the periodic binding of troponin on the thin filament (green circles) at a spacing of ~ 385Å D: 2D class averages from cryo-EM data with pairs of troponin labelling the two strands of actin (green circles); tropomyosin strands are also visible (top and bottom images)

Fig. 2

3D reconstruction of native zebrafish cardiac thin filaments and comparison to human reconstituted thin filament structure. A-D: Surface rendered protein density map, highlighting the thin filament constituent proteins, segmented and colour coded as follows; Tn1 green, Tn2 blue, tropomyosin pink and actin purple. A & B: views of the complete map oriented to illustrate the two distinct paths taken by individual troponin molecules on each side of the thin filament as they span the two tropomyosin strands. A & B are related by 180o rotation about the central axis of the thin filament. The different paths of troponin are apparent; Tn1 is located higher on the filament than Tn2, however, the TnT linker peptide path of Tn1 to tropomyosin is longer than that of Tn2. More density is recovered for Tn1 despite its linker peptide following the longer path. C & D: Close up views of the troponin core domain illustrating its rotated ‘L’ shape which is consistent on both sides. Views as in A & B but rotated by 60o about the central axis. Segmented regions calculated using ChimeraX. The pointed (-) and barbed end (+) of the actin filament are indicated. E–J: The zebrafish (yellow) and Yamada (purple) high Ca2+ state reconstructions are superposed. E & F: Extra protein density in Tn1 and Tn2 core regions is visible in the zebrafish map. G & J: Tropomyosin and actin densities are similar. H: Strong similarity in TnT T1 domains interacting with tropomyosin overlap region. I: Extra TnT linker-region density in the zebrafish map not present in other thin filament maps

Fig. 3

Zebrafish high Ca2+ thin filament model. The full-length actin (grey), TnC (residues 2-161, zebrafish 2-161) green, TnI (41–166, zebrafish 10–135) yellow, TnT (99–272, zebrafish 101–207) orange, full length tropomyosin red, atomic models docked into the electron density. A & B: 180° rotations, C & D: close up of the central region with 180° rotations

Fig. 4

Conformation of troponin core domain and tropomyosin positions. The high and low Ca2+ models 6KN8 (green) and 6KN7 (blue) were docked into our thin filament reconstruction and compared to our zebrafish model (yellow). A & B: the core domains of Tn1 and Tn2. Regions of empty density are indicated with *. B: ɑ (45°) & (35°) the angle the TnI helix 1 makes with the horizontal in the two states C & D: the tropomyosin overlap region where TnT1 domain is located of Tn1 and Tn2 respectively. E,F, & G: Pairwise comparison of the position of tropomyosin from the three models

The regulatory proteins troponin and tropomyosin are readily identified in our reconstruction, troponin lies over two adjacent actin subunits on the genetic left-handed actin helix, leading to a 27.5 Å axial rise between the globular core regions and the characteristic troponin stagger (Fig. 2A-B). Crucially, the density linking the globular core to the opposite tropomyosin strand is recognised as the extended tail of troponin based on the assignment of troponin components made in the human reconstituted and native porcine structures (Yamada et al. 2020) (Risi et al. 2021). The stagger of the globular core domain combined with the ability of troponin to bridge the actin and tropomyosin strands results in a fully asymmetric macromolecular complex. Both troponin ternary complexes have the same rotated ‘L’ shape; with the long arm of the ‘L’ nestled between actin subunits. The general ‘L’ shape is similar to that seen in our previous negative stain structures (Paul et al. 2010, 2017). More density is recovered for the upper troponin complex (Tn1) in the zebrafish map, with additional density extending from the long arm of the ‘L’ into the TnT linker peptide region. This additional density enhances the overall asymmetry of our reconstruction in the troponin region. This difference which we observe between Tn1 and Tn2 may arise from stabilisation of Tn1 flexible linker peptide due to local interactions in this position which may not occur in the Tn2 linker peptide in the equivalent region on the other side of the thin filament. Conformational differences between tropomyosin core domains on different actin strands have also been reported and attributed to different Ca2+ states, but without differences in this linker region (Risi et al. 2021).

The TnT linker peptide crosses from the core domain on one strand and connects to the T1 domain that runs down the tropomyosin strand, towards the barbed end of the filament. Due to the stagger of the troponin core regions each linker peptide takes a path of different length. Explicitly, the path of Tn1, the upper of the two complexes, crosses to the lower of the two T1 domain-tropomyosin overlap regions, resulting in a longer path than for Tn2. We have recovered density for the T1 domain-tropomyosin overlap regions of both Tn1 and Tn2 (Fig. 2B & D). The variation in the length of the path of TnT seen in this map is a feature of the human reconstituted and native porcine structures (Yamada et al. 2020; Risi et al. 2021) and is the subject of recent study employing time-resolved fluorescence imaging & computational modelling (Deranek et al. 2022).

The zebrafish and both the Yamada high and low Ca2+ human thin filament maps are shown filtered to 15Å in Fig. S2. The zebrafish map clearly resembles the high Ca2+ human thin filament map more closely than the low Ca2+ structure. The angle at which the long arm of the L-shaped core domain emerges from the filament was consistent with the high Ca2+ map; this is even more apparent when the two structures are directly superposed (Fig. 2I-J).

A full (length) homology model of the high Ca2+ zebrafish thin filament was calculated in SWISS-MODEL (see methods) using zebrafish protein sequences. The final model was composed of actin, tropomyosin, TnC (residues 2-161), TnI (residues 41–166) and TnT (residues 10–135). The fit in map utility in Chimera gave a good initial alignment of the entire complex as a rigid body within the density. An improved manual rigid body fit of structured elements was performed. The manual fit of each troponin complex was independent as the full thin filament is asymmetric and no symmetry operators could be used. The refined fit was good (51,170 of 60,788 atoms inside contour (threshold − 0.875)) with a correlation score of 0.8, for comparison the Yamada high Ca2+ model in the zebrafish electron density gave a correlation score of 0.79.

The N-terminal lobe of TnC is an area of weak density in the zebrafish map, with the atomic model lying outside the electron density; despite this the similarity of the zebrafish map to the Yamada map and the overall quality of fit into the density establishes the agreement with the human thin filament structure.

Tropomyosin and troponin in a high Ca2+ state

The azimuthal position of tropomyosin typically indicates the activation state of the thin filament. The binding of Ca2+ to TnC is known to precipitate the movement of tropomyosin across the actin subunit revealing first the weak then the strong myosin binding sites. To that end we inspected the azimuthal position of tropomyosin to determine the activation state of our thin filament structure. Whilst it was not possible to resolve the individual strands of tropomyosin, the density was sufficient to track the path of the coiled-coil. The full-length tropomyosin molecule was docked into the density as a rigid body. The position of tropomyosin was found not to block weak myosin binding sites and in good agreement with the 6KN8 Yamada high Ca2+ state. Chopping tropomyosin into shorter segments would provide the ability to interrogate the molecule’s flexibility further (Rynkiewicz et al. 2022; Paul et al. 2017) as would higher resolution reconstructions of these flexible regions. The differential movement of tropomyosin that we reported in 2017 could have been due, in part, to low resolution density contributions from TnT, which is now known to cross strands. Through these docking experiments we were able to establish that the thin filament in our map corresponded to a high Ca2+ state. This result was unexpected due to the isolation protocol being carried out in a relaxed state, potential reasons for this are discussed below.

Further rounds of 3D classification were carried out on the data to determine whether there was an underlying mixed population of Ca2+ states. The data was sorted into two 3D classes and then four; in both cases the resultant 3D reconstructions returned a consistent tropomyosin position and location of the IT arm (Fig S5 & S6). This analysis gave no indication of a population of relaxed filaments within the data. However, minor populations of single sided troponin filaments were observed, this heterogeneity in the data will have reduced the signal to noise and the resolution, of the troponin density.

The fit of the different models into the globular core domain were critically assessed. The IT arm, which corresponds to the long arm of the L shaped motif of the troponin complex, and which is made up of a α-helical coiled-coil structure formed from part of cTnI and cTnT, makes an acute angle with the horizontal in both models, sloping down away from TnC (Fig. 4B). By comparing the angle of helix 1 in TnI in both human models, we observe that this angle is greater in a low Ca2+ state (6KN7) ~ 45° than in the high Ca2+ state (6KN8) which intersects the horizontal at a shallower angle of ~ 35° (Fig. 4A & B). The same helix in the zebrafish model runs parallel to that in the high Ca2+ state model (6KN8). Whilst the angle of the zebrafish IT arm closely aligns to that of the human model at high Ca2+, it is known that in cardiac muscle it can be mobile and as such not necessarily an indicator of activation state (Sevrieva et al. 2014). If the Tn-I switch peptide had been resolved to a higher resolution a more definitive determination of the regulatory state could have been made; by considering interaction with TnC in the closed state and with tropomyosin and actin in the blocked state.

Both globular lobes of TnC are more closely associated with the actin filament in the human high Ca2+ model (Yamada et al. 2020); this was also true of the zebrafish map density. The N-terminal domain of TnC in the zebrafish map, particularly in TN2 (Fig. 3A &C), is a region of reduced density, and neither of the Yamada models fit well in this region. However, the high Ca2+ pdb is contained within the molecular envelope of the zebrafish map to a greater extent. Thus, both the position of tropomyosin and the conformation of troponin in the zebrafish thin filament match the high Ca2+ conformation of the human thin filament.

Similarities and differences between the zebrafish and human thin filament structures

The pairwise alignments of the contractile proteins show a substantial level of sequence identity between the species (Fig S7). This is with the exception of the human N-terminal extension of TnI, which is missing in zebrafish. This region has not been resolved in the human structures and is thought to be flexible. The zebrafish map has a region of unassigned density at the base of the core adjacent to the C-lobe of TnC (Fig. 4A & B indicated with *). The proximity of this density to tropomyosin and its shape resembles the molecular dynamics modelling of the N-terminal extension of cardiac TnI (Pavadai et al. 2022). Whilst density for this region would be expected in human maps, zebrafish have no equivalent region. A BLAST search was carried out of the whole zebrafish genome to establish whether there may be an additional protein that served a similar physiological role, but no candidates were found. In humans the phosphorylation of serine 23 and 24, located in the N-terminal extension, reduces Ca2+ sensitivity which corresponds to an increased rate of relaxation and heart rate during normal response to cardiac stressors like exercise. It is conceivable that zebrafish do not need such an adaptation.

Zebrafish Troponin T is 16 amino acids shorter than human, with a small number of deletions including a 6 amino acid truncation of the C-terminus. It is likely that the small 2–4 amino acid deletions dispersed along the protein are uninterpretable in the context of our map. The extra density resolved in the zebrafish map for the linker region of TnT1 that has no equivalent density in the human maps corresponds to 161–208 in the sequence. This region has a 63% sequence identity with human, (compared to 68% for full length TnT) the slightly reduced identity the sequence does not provide us with any indication of the origin of this density. It may be indicative of a stabilising local interaction or as these are native filament preparations potentially the contribution of additional proteins. The skeletal muscle protein nebulin was recently observed to bind to the TnT linker region (Wang et al. 2022).

Other differences in the zebrafish reconstruction compared to the human is in the N-lobe of TnC, with reduced density found in the zebrafish map. There is no elucidation to the reason for this on the sequence level, with the identity between these two proteins being 90.1%, no deletions and the same length in both species. It is likely the reduced density is in part due to a less well resolved mobile region.

The architecture and orientation of the troponin core domains in human (Yamada et al. 2020) and zebrafish are consistent and reversed from the interpretation made in previous studies of negatively stained thin filaments (Paul et al. 2017; Yang et al. 2014). The different interpretation made in these studies most likely reflects the lower resolution characteristic of negative stain as well as the difference between earlier predictions of the overall architecture of the troponin complex and the unexpected arrangement identified in the cryo-EM analysis. All recently published cryo-EM maps of the thin filament have the core domain oriented in the same way demonstrating not only the step change in resolution of the thin filament given by cryo-EM but also the structural similarities across human, porcine and murine orthologues. Of note is that the orientation of the IT arm is very close to that described in an in-situ polarised fluorescence study of skeletal muscle (Knowles et al. 2012); however more flexibility was then described in the IT arm of cardiac muscle by the same lab (Sevrieva et al. 2014).

We report that both ternary troponin complexes in our reconstruction appear to be in the same state, with a very similar arrangement of main helices of TnI, C & T within the core domain. We attribute the slight differences in the N-terminal lobe of TnC to weak signal and low resolution in this area. No evidence of a differential behaviour of troponin within the core domain ‘pair’ was seen here; this was also the case as in the human thin filaments. However, this may not be expected unless prepared in a more physiological range of Ca2+ levels (Risi et al. 2021) and not resolved without a significant improvement in the resolution of the troponin complex.

The tropomyosin overlap region and its interaction with TnT is readily identified in our zebrafish map as in a similar manner to the human and porcine structures indicating this region’s importance and functional significance across species. The two distinct paths of TnT crossing the actin strands and giving rise to links between the two tropomyosin strands were also seen here. The mechanistic significance/implications of this asymmetric arrangement are yet to be fully understood. Zebrafish tropomyosin has a very high sequence identity (92%) with human cardiac, due to the limited resolution in our map no differences were resolved.

Zebrafish actin has, as expected an extremely high sequence identity (98%) to human. Our data contained large numbers of undecorated actin filaments providing the opportunity to determine the structure of zebrafish cardiac actin. Without the imposition of helical symmetry, we calculated an actin structure of equal length to our thin filament map to a resolution of 3.89Å, (EMDB 17120) (Fig. 5.). This significantly higher resolution provides the ability to assess the equivalence of our docked zebrafish model (PDB 8ORD) to human F-actin structure (6KN8) with more precision. A TM score (Zhang and Skolnick 2004) of 0.97 and a RMSD of 1.19 confirms the expected structural similarity of the two proteins (Fig S11).

Fig. 5

Cryo-EM reconstructions of the zebrafish thin filament and actin filament. A: The complete asymmetric unit of the thin filament composed of 14 actin subunits, 2 tropomyosin and troponin complexes. B: An equivalent region of the actin filament composed of 14 actin subunits. C: Detailed region of the actin filament map (yellow) with cartoon representation of the fitted coordinates (blue). D: Detailed region of an individual actin subunit protein density with side chains coordinates represented as sticks

Through a detailed comparison of the zebrafish and human reconstituted thin filament structures we can confirm that the contractile apparatus is broadly structurally equivalent. This study has revealed the zebrafish structure in a single conformation corresponding to a high Ca2+ state, and consequently we cannot report on the structural changes that underpin regulation. However, the implications of the structure being conserved across species is significant, indicating that functional domains and mechanisms are the same.