Building the 3D structure of the Lmod2s1/αTM1a_1-14Zip complex

The 3D structure of the Lmod2s1/αTM1a_1-14Zip complex represents the binding interface between Lmod2 and tropomyosin.

Lmod2s1 and αTM1a_1-14Zip (amino acid sequences shown in S1 Fig) were chosen as peptides representing Lmod2 TpmBS1 and the tropomyosin Lmod2-binding site, respectively [27]. The αTM1a_1-14Zip peptide forms a stable coiled-coil dimer under most of the conditions used in this study, and we will henceforth use the name αTM1a_1-14Zip to refer to the dimeric molecule. The stoichiometric ratios are listed hereafter with respect to the molar amount of dimer.

The obtained model of the Lmod2s1/αTM1a_1-14Zip complex is shown in Fig 1. In the complex, two α-helices of Lmod2s1 bind to the N-terminal part of the coiled-coil αTM1a_1-14Zip, thus resulting in a 4-helix bundle. The two-residue loop Leu25-Ser26 connecting two α-helices of Lmod2s1 is positioned above the two N-termini of αTM1a_1-14Zip, with the side chain of Leu25 being inserted between the αTM1a_1-14Zip chains. The N-terminal residues Ser2-Ile16 of Lmod2s1 are disordered (flexible) and interact with αTM1a_1-14Zip only transiently. Inside the disordered region, residues Lys12-Glu14 form a transient α-helical turn.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Ribbon representation of the 3D structure of the Lmod2s1/αTM1a_1-14Zip complex (PDB ID 6UT2).

N- and C-termini of Lmod2s1 and αTM1a_1-14Zip are marked. The ribbon of Lmod2s1 is shown in cyan. The N-terminal Gly residue and the first 14 residues of Tpm1.1 in αTM1a_1-14Zip are colored in black, and the C-terminal GCN4 sequence is colored in brown. The side chain of Leu25 in the link motif connecting two Lmod2 helices is shown in cyan and labeled. The residues of a transiently forming loop Lys12-Glu14 in Lmod2s1 are also labeled. 3D, three-dimensional; Lmod, leiomodin; PDB, Protein Data Bank.

https://doi.org/10.1371/journal.pbio.3000848.g001

The structure of the complex is characterized by several ionic and hydrophobic interactions between Lmod2s1 and αTM1a_1-14Zip. Sequence-specific ionic interactions between tropomyosin and Lmod2 chains are between side chains of Asp19 (Lmod2s1) and Lys5 (αTM1a_1-14Zip), Glu36 (Lmod2s1) and Lys5 (αTM1a_1-14Zip), Glu38 (Lmod2s1) and Lys7 (αTM1a_1-14Zip), and Glu41 (Lmod2s1) and Lys12 (αTM1a_1-14Zip). In members of the Tmod family, Asp19 is a highly conserved residue, and positions corresponding to 36, 38, and 41 of Lmod2 have a clear preference for negatively charged Glu or Asp residues (S2 Fig). Important residues contributing to the hydrophobic “cap” cluster are Leu22, Leu25, Leu30, Leu33, and Leu37 of Lmod2s1 and Met1/Ile4/Met8 from two chains of αTM1a_1-14Zip. Residues Leu22, Leu25, and Leu33 are highly conserved in Lmod/Tmod homologs, suggesting that the geometry of a Leu side chain in these positions is critical for favorable interactions. Leu30 and Leu37 allow for a Met substitution, which is a very similar residue in hydrophobicity and size. Consistently, Leu30 was shown to be a physiologically significant residue for Lmod2, and the replacement of Leu30 in Lmod2 with a negatively charged Glu disrupted the tropomyosin-dependent ability of an Lmod2 fragment (residues 1–201) to decrease the pointed-end polymerization of actin. In addition, the replacement also disrupted Lmod2 assembly at the pointed end of the thin filaments in cardiomyocytes [23].

The 3D structure is in agreement with spectral changes in ¹⁵N-Lmod2s1 caused by chemical/sequence alterations in the model tropomyosin peptide

The ¹⁵N-Lmod2s1 resonance peak shifts caused by alterations in the N- and C-terminal parts of the model tropomyosin peptide are consistent with the Lmod2s1/αTM1a_1-14Zip topology.

To enable labeling of the tropomyosin peptide for the NMR experiments, native acetylation of the tropomyosin N-terminus was mimicked by adding Gly at the N-terminus of αTM1a_1-14Zip (see Materials and methods, section “Peptide design and production” for more details). We compared ¹⁵N-HSQC spectra of ¹⁵N-Lmod2s1 in complex with either αTM1a_1-14Zip or Ac-αTM1a_1-14Zip (S6 Fig). Ac-αTM1a_1-14Zip differed from αTM1a_1-14Zip in that its N-terminus did not contain the mimetic Gly residue, and it was N-acetylated. Since the substitution of an acetyl group with a Gly residue has very little effect on structure and binding properties of the tropomyosin peptide [35], a comparison of ¹⁵N-Lmod2s1 spectra in respective complexes identified residues of Lmod2s1 located close to the tropomyosin N-terminus.

The amide cross-peaks that shifted considerably and did not partially overlap with peaks in the original spectrum upon the replacement of αTM1a_1-14Zip with Ac-αTM1a_1-14Zip are shown on the Lmod2s1 sequence and the 3D structure in Fig 2A and 2C. These residues tend to be closer to the middle of the peptide sequence, and they are mainly located in the first α-helix and the first half of the second α-helix (Fig 2A). In accordance with the 3D structure of the complex, mapping of the most affected residues onto the 3D structure demonstrated that larger spectral changes were observed for residues located in the proximity of the αTM1a_1-14Zip N-terminus (Fig 2C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Mapping of ¹⁵N-Lmod2s1 resonance peak shifts caused by chemical/sequence alterations in the model Tpm peptide onto the Lmod2s1 sequence and the 3D structure of the Lmod2s1/αTM1a_1-14Zip complex.

(A) Shown on a magenta background are ¹⁵N-Lmod2s1 residues with the most affected ¹⁵N-HSQC cross-peaks upon the replacement of the N-terminal Gly residue in αTM1a_1-14Zip with an acetyl group. (B) Shown on an orange background are ¹⁵N-Lmod2s1 residues with the most affected ¹⁵N-HSQC cross-peaks upon the replacement of αTM1a_1-14Zip (containing 14 N-terminal residues of Tpm) with αTM1a_1-28Zip (containing 28 N-terminal residues of Tpm). (C) Mapping of the most affected regions of ¹⁵N-Lmod2s1 (shown in magenta) upon replacement of the N-terminal Gly residue in αTM1a_1-14Zip with an acetyl group onto the 3D model of the complex. (D) Mapping of the most affected residues of ¹⁵N-Lmod2s1 (shown in orange) upon replacement of αTM1a_1-14Zip with αTM1a_1-28Zip onto the 3D model of the complex. 3D, three-dimensional; HSQC, heteronuclear single-quantum coherence; Lmod, leiomodin; Tpm, tropomyosin.

https://doi.org/10.1371/journal.pbio.3000848.g002

To map onto Lmod2s1 the spectral effects of changes in the C-terminus of the tropomyosin peptide, we used a previously reported comparison between the ¹⁵N-HSQC spectra of ¹⁵N-Lmod2s1 in complexes with either αTM1a_1-14Zip or αTM1a_1-28Zip [27]. The mapping of the considerably shifted cross-peaks onto the Lmod2s1 sequence and the 3D structure of the complex is shown in Fig 2B and 2D. Again, in agreement with our 3D structure, the most affected residues are located at the N-terminus of Lmod2s1—which was identified as mobile and potentially forming transient interactions with αTM1a_1-14Zip—and in the C-terminal half of the second Lmod2s1 α-helix. The most affected residues in the second helix are either Leu residues contributing to the hydrophobic cluster (Leu33 and Leu37) or their neighbors (Glu36, Asp39). This long-range effect of tropomyosin sequence extension on the second helix of Lmod2s1 can be explained by an impaired hydrophobic core in αTM1a_1-28Zip, which is less stable than αTM1a_1-14Zip [36, 37].

The structure of the Lmod2s1/αTM1a_1-14Zip complex reflects a complete binding interface between Lmod2 and tropomyosin.

To test if the inclusion of 28 tropomyosin amino acids into the model peptide [27] can induce the formation and binding of an extra helical structure within the N-terminal dynamic residues of Lmod2s1, we titrated ¹⁵N-Lmod2s1 with αTM1a_1-28Zip. NMR peak attenuation due to the exchange between bound and unbound ¹⁵N-Lmod2s1 was followed as a function of sub-stoichiometric αTM1a_1-28Zip concentration (S7 Fig). In this type of NMR titration experiments, the peak attenuation increases with the increase in chemical shift difference between bound and unbound states [38]. Larger peak attenuations would occur for Lmod2s1 residues if they participated in formation of a helix and/or multiple nontransient direct contacts. Fig 3 illustrates changes in ¹⁵N-Lmod2s1 spectra upon binding with either αTM1a_1-14Zip or αTM1a_1-28Zip when they were mapped onto the Lmod2s1 amino acid sequence.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Schematic comparison of the effects of αTM1a_1-14Zip and αTM1a_1-28Zip on the ¹⁵N-HSQC spectrum of ¹⁵N-labeled Lmod2s1 mapped onto the Lmod2s1 amino acid sequence.

Compound chemical shift changes (Δδ) upon Lmod2s1/αTM1a_1-14Zip complex formation were calculated as Δδ = [(Δδ_H)² + (Δδ_N/6.5)²]^1/2 [79], where δ_H and δ_N are chemical shifts of ¹H/¹⁵N nuclei of backbone amides. Peak attenuation upon the titration of Lmod2s1 with αTM1a_1-28Zip (data for 7:1 Lmod2s1: αTM1a_1-28Zip molar ratio were used) was calculated as a relative intensity of a ¹H-¹⁵N cross-peak in comparison with the intensity of the corresponding peak in the absence of αTM1a_1-28Zip. Each set of data was divided into four groups by calculating quartile values and displayed as”no color,” green, yellow, and red in the order of increasing spectral effects. Since cross-peaks from pairs of Lmod2s1 residues L10/L37 and I16/L21 overlap (see Supplementary Materials in [27]), the effect of αTM1a_1-28Zip titration on their individual resonance peaks could not be determined. Hence, the residues L10, I16, L21, and L37 are labeled in the bottom sequence as gray. Lmod, leiomodin; HSQC, heteronuclear single-quantum coherence.

https://doi.org/10.1371/journal.pbio.3000848.g003

Upon the titration of Lmod2s1 with αTM1a_1-28Zip, major peak attenuations were observed for the region Glu18-Ile40, whereas peak attenuations in the N-terminal part of Lmod2s1 were on average considerably smaller. Similarly, upon binding to the shorter αTM1a_1-14Zip, the region Glu18-Ile40 experiences the largest chemical shift changes upon binding, whereas the chemical shifts of the Lmod2s1 N-terminus are less affected (S8 Fig). Residues Glu18-Ile40 correspond to the well-structured part of Lmod2s1 in complex with the shorter αTM1a_1-14Zip. These similarities in spectral changes upon binding to either shorter αTM1a_1-14Zip or longer αTM1a_1-28Zip demonstrate that no additional secondary structure is formed in Lmod2s1 if more tropomyosin amino acid residues are included in the model peptides.

Weaker spectral effects in the N-terminal part of the Lmod2s1 peptide were also found to display some similarities for these two complexes. Specifically, within the N-terminal region, cross-peaks corresponding to residues Ser11-Lys12 were more affected by αTM1a_1-28Zip titration than cross-peaks corresponding to their neighbors. Consistently, if compared with other residues from the Ser2-Ile16 region, Lys12-Glu14 displayed larger chemical shifts upon binding to αTM1a_1-14Zip. Interestingly, the region Lys12-Glu14 was found to form an α-helical turn capable of transient interactions with αTM1a_1-14Zip in MDSs.

Taking these results together, we can conclude that αTM1a_1-14Zip provides the major part of the binding interface. Indeed, upon increasing the number of tropomyosin residues from 14 to 21, the binding affinity increased 2-fold at 30°C [27], which roughly corresponds to a free energy change of −0.4 kcal/mol. The relatively small change in Lmod2s1 binding energetics likely arose from a small number of transient interactions between the flexible Ser2-Ile16 region of Lmod2s1 and the longer tropomyosin peptide [27].

Effects of reducing Lmod2 affinity for tropomyosin on Tmod1 assembly and the thin filament length in cardiomyocytes

Decrease of Lmod2 binding to tropomyosin at the pointed end perturbs thin filament elongation and increases Tmod1 assembly.

The presence of TpmBS1 in Lmod2 is crucial for its proper localization to the pointed ends in cardiomyocytes [40]. To study how the L25G mutation affects the function of Lmod2 in cells, rat neonatal cardiomyocytes were transfected with green fluorescent protein (GFP), GFP-Lmod2, or GFP-Lmod2[L25G], and thin filament lengths were measured (Fig 4). GFP-Lmod2 expression increased thin filament lengths compared with the expression of GFP alone (GFP-Lmod2, 0.81 ± 0.01 μm; GFP, 0.77 ± 0.01 μm; Fig 4B), which was consistent with previous reports [5, 23, 41]. However, the introduction of the L25G mutation perturbed the ability of Lmod2 to elongate thin filaments (GFP-Lmod2[L25G], 0.79 ± 0.01 μm, p = 0.0177, compared with GFP-Lmod2, Fig 4B). The difference between thin filament lengths in the presence of GFP (control) and GFP-Lmod2[L25G] was not statistically significant (p = 0.4690). Additionally, GFP-Lmod2[L25G] was not detected (i.e., it did not assemble properly) at the pointed ends of the thin filaments, unlike GFP-Lmod2 (Fig 4A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. GFP-Lmod2[L25G] is unable to elongate thin filaments compared with GFP-Lmod2 WT.

(A) Representative images of rat neonatal cardiomyocytes expressing GFP, GFP-Lmod2 WT, or GFP-Lmod2[L25G]. GFP-Lmod2[L25G] was unable to localize to the pointed ends, near the M-line (“M”) of sarcomeres, unlike GFP-Lmod2 WT. Cells were stained with phalloidin to mark F-actin and α-actinin to indicate the Z-disc (“Z”). Turquoise lines show a gap in F-actin staining across the M-line. Scale bar = 1 μm. (B) Thin filament lengths from rat neonatal cardiomyocytes transfected with GFP, GFP-Lmod2 WT, or GFP-Lmod2[L25G] (S1 Data). Statistical data are shown as mean ± SEM, n = 49–78 total measurements from 10–20 cells per culture, three cultures; *p = 0.0177, **p = 0.0013, one-way ANOVA. F-actin, filamentous actin; GFP, green fluorescent protein; Lmod, leiomodin; WT, wild type.

https://doi.org/10.1371/journal.pbio.3000848.g004

We have previously shown that expression of Lmod2 leads to displacement of Tmod1 from the pointed ends in cardiomyocytes [21]. Here, we examined the effect of GFP-Lmod2[L25G] on thin filament pointed-end localization of endogenous Tmod1 (Fig 5). About 90% ± 4% of control GFP-expressing cells showed consistent Tmod1 assembly to the pointed ends of thin filaments, as expected (Fig 5B). GFP-Lmod2 expression decreased the percentage of cells displaying consistent Tmod1 assembly (GFP-Lmod2, 41% ± 10%, p = 0.0011). The assembly of Tmod1 in cells expressing the GFP-Lmod2[L25G] mutant was noticeably higher (GFP-Lmod2[L25G], 67% ± 2%, p = 0.0427) than in cells expressing WT GFP-Lmod2. The difference between the assembly of Tmod1 in cells expressing GFP-Lmod2[L25G] and those expressing GFP only (control) was not statistically significant (p = 0.079). Therefore, the Lmod2[L25G] mutant might not be able to displace Tmod1 from pointed ends, or, at least, it might be less effective than WT GFP-Lmod2. The cellular results demonstrate that the L25G mutation significantly perturbed the subcellular localization and function of Lmod2 in cardiac muscle cells, and this agrees with the structural importance of this residue for TpmBS1 and the critical role of TpmBS1 in Lmod2 function at the pointed end. Collectively, these data support the competition mechanism.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. GFP-Lmod2[L25G] does not significantly displace Tmod1 from thin filament pointed ends.

(A) Representative images showing Tmod1 assembly in rat neonatal cardiomyocytes expressing either GFP, GFP-Lmod2 WT, or GFP-Lmod2[L25G]. Cells were stained for Tmod1 and α-actinin (red) to mark Z. Scale bar = 1 μm. (B) Percentage of cells having consistent Tmod1 assembly at thin filament pointed ends (S2 Data). Mean ± SEM, n = 124–158, total number of cells measured, four cultures (*p = 0.0427, **p = 0.0011, one-way ANOVA). GFP, green fluorescent protein; Lmod, leiomodin; M, M-line; Tmod, tropomodulin; WT, wild type; Z, Z-disc.

https://doi.org/10.1371/journal.pbio.3000848.g005

A model of Lmod 3D assembly and function at the pointed end

The structure of the Lmod2s1/αTM1a_1-14Zip complex bears a resemblance to the head-to-tail tropomyosin overlap region.

We compared the 3D structures of the Lmod2s1/αTM1a_1-14Zip complex with the overlap complex formed by tropomyosin N- and C-termini [34]. Positioning of the two Lmod2s1 helices with respect to αTM1a_1-14Zip helices resembled that of the C-terminal fragment in the overlap complex (Fig 6), which means that TpmBS1 of Lmod2 can only bind the tropomyosin N-terminus if the latter is free and not a part of the overlap complex (at the pointed end). Comparison of the 3D structures also revealed sequence similarities between the tropomyosin C-terminus and the second helix of Lmod2s1. Among these similar residues, hydrophobic residues in these two complexes are responsible for maintaining important core contacts with the N-terminus of tropomyosin.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Comparison of the head-to-tail Tpm overlap junction (PDB ID 2G9J, left) with the Lmod2s1/αTM1a_1-14Zip complex (right).

The sequence alignment of the structurally homologous parts of the C-terminal helix of Lmod2s1 and C-terminus of Tpm1.1 is shown at the bottom of the figure. Similar residues are marked in green on the sequences and their side chains are displayed on the 3D structures. 3D, three-dimensional; Lmod, leiomodin; PDB, Protein Data Bank; Tpm, tropomyosin.

https://doi.org/10.1371/journal.pbio.3000848.g006

A 3D reconstruction of Lmod2 assembly at the pointed end of the thin filament.

Using our newly obtained structure and currently available structural information on Lmod2/Tmod1 interactions with actin [19, 23, 28, 42, 43], we created a model for Lmod2 assembly at the pointed end (Fig 7) using as a starting point an atomic model for F-actin in complex with one tropomyosin cable [44]. The interactions of ABS1 and ABS2 with the pointed end were modeled from the structure of the complex of homologous Tmod1 ABS1 with actin (Protein Data Bank [PDB] ID 4PKG, [19]) and from the structure of the complex of Lmod2 ABS2 with actin (PDB ID 5WFN, [42, 43]), respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. A model of Lmod2 assembly at the pointed end.

The view on the left shows the Lmod2 assembly at the pointed end of the thin filament including the entire tropomyosin molecule bound to F-actin (S3 Data). The view on the right shows the magnified boxed part of the assembly. Letters A and B label two pointed-end actin molecules (colored in tan and gray, respectively, to facilitate interpretation of the figure). Tropomyosin is shown in cyan, actin (with the exception of the pointed-end actin protomer B in the right panel) in tan, and TpmBS1 of Lmod2 in green. The Lmod2 ABS1h fragment shared with Tmod1 (residues Pro60-Leu86 in Lmod2) is shown in blue (modeled from the complex of homologous Tmod1 ABS1 with actin, PDB ID 4PKG, [19]). The fragment includes the conservative amphiphatic helix (residues Arg66-Lys79 in Lmod2). Within the linker connecting TpmBS1 and ABS1h (residues Pro42-Thr59), residues Asn45-Arg51, which were found critical for Lmod2 ABS1 binding to actin [23], are shown in red-orange. The remaining linker residues are shown in orange. ABS2 (LRR) is shown in magenta (modeled from the complex of Lmod2 LRR with actin [PDB ID 5WFN, [42, 43]]). The linker connecting ABS1h and ABS2 (residues Gly87-Asn195) is in the back and shown in brown. ABS, actin-binding site; F-actin, filamentous actin; Lmod, leiomodin; LRR, leucine-rich repeat; PDB, Protein Data Bank; TpmBS1, tropomyosin-binding site.

https://doi.org/10.1371/journal.pbio.3000848.g007

Although the existence of ABS1 in Lmod2 was questioned [25], we later demonstrated by NMR [23] and atomic force microscopy [28] that Lmod2 can interact with actin through this site. Currently, only the 3D structure of Tmod1 ABS1 in complex with actin is available [19]. Lmod2 and Tmod1 have ABS1 regions that may overlap only partially [23, 25]. However, within the segment of the Tmod1 sequence with the known 3D structure, Lmod2 and Tmod1 share a homologous amphipathic helix (residues Arg66-Lys79 in Lmod2) and several homologous residues in extended conformation flanking this helix [10]. We used these shared residues (Pro60-Leu86 of Lmod2) to model an approximate localization and structure of Lmod2 ABS1 on actin (Fig 7). Unlike in Tmod1, several residues of Lmod2 (residues Asn45-Arg51) located before the shared ABS1 region, are critical for Lmod2 ABS1 binding and might also interact with actin [23]. Since the full-size ABS1 of Lmod2 might include these additional residues, for the purposes of further discussion, we will henceforth call the shared homologous residues Pro60-Leu86 of Lmod2 as “ABS1h.”

To complete the docking of Lmod2 to the pointed end, we connected the binding sites with polypeptide linkers (residues Pro42-Thr59 to connect TpmBS1 and ABS1h, and residues Gly87-Asn195 to connect ABS1h and ABS2). The structures of these linkers and their mode of interaction with the pointed end are not known. Accordingly, to check if the linkers were sufficiently long to support our proposed 3D structure of the assembly, we tested if the linkers form secondary structure and measured their contour lengths. For that, the linkers were created in silico using UCSF Chimera as fully extended polypeptides. After 50-ns MDS, both linkers form dynamic α-helices alternating with disordered regions (S11A Fig). Contour lengths of the Pro42-Thr59 and Gly87-Asn195 linkers with the secondary structure elements were estimated as 59.5 Å and 306 Å, respectively, providing enough length to connect the binding sites at the pointed end without steric clashes with the thin filament. Secondary structure predictions made by three popular server-side predictors—Jpred4, PsiPred, and PredictProtein—suggest locations of helical regions that partially overlap with the helical regions formed in the course of the MDSs (S11B Fig). Importantly, all of the predictors suggested that a smaller number of residues were involved in the formation of α-helices, therefore indicating that the contour lengths of the linkers might be even longer. The linkers were connected using UCSF Chimera with their respective flanking TpmBS1, ABS1h, and ABS2 in such a way as to not change the integrity of secondary structure elements formed during the simulations. The disordered regions were considered conformationally variable and were changed to adapt to the shape of the thin filament. The final assembly is shown in Fig 7.

The reconstructed Lmod2/pointed-end assembly demonstrates that our model of the Lmod/tropomyosin binding interface is consistent with data on the interactions of Lmod with actin [19, 23, 28, 42, 43]. It also shows that Lmod2 and Tmod1 share a large binding surface at the pointed end, which is formed by TpmBS1, ABS1, and ABS2.

A model for the leaky cap created by Lmod2 binding at the pointed end of the thin filament.

The structural data provide a compelling model for the role of Lmod2 at the pointed end. In Fig 8, we propose steps of actin polymerization at the pointed end in the presence of bound Lmod2. As the first step of polymerization, an actin molecule (actin 1) attaches to the actin molecule B (Fig 8B). The linker connecting ABS1h and ABS2 is sufficiently long to create no clashes at the first step. For the second molecule of actin (actin 2) to attach to actin molecule A, the Lmod2 amphipathic helix (residues Arg66-Lys79) has to dissociate (Fig 8C); otherwise, it would interfere with the newly attached actin molecule 2. Interstrand interactions between actin 1 and actin 2 facilitate this displacement. Here, the helix acts as a “swinging gate” allowing further actin polymerization. Lmod2 residues Asn45-Arg51, critical for Lmod2 ABS1 binding to actin [23], do not obstruct the attachment of actin 2 to the pointed end, and therefore they may serve as a “hinge” inside the ABS1 region allowing the amphipathic helix to move away (and potentially unfold) from the site of intrastrand actin-actin interaction.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. A model for the leaky cap created by Lmod2 binding at the pointed end of the thin filament.

Top figures show side views, and bottom figures display views from the top. (A) The initial position of Lmod2 assembled at the pointed end and corresponding to the structure shown in Fig 7. ABS1 is in a “closed gate” conformation with the blue helix of ABS1 obstructing the attachment of an actin molecules to actin A. (B) Step 1, an actin molecule 1 (shown in purple) attaches to actin B. Actin molecule 1 does not interfere with the Lmod2 binding, and the ABS1 of Lmod2 remains in the “closed gate” conformation. (C) An actin molecule 2 (shown in turquoise) attaches to actin A. Interstrand interactions between actin molecules 1 and 2 facilitate displacement of Lmod2 ABS1 helix, and ABS1 adopts an “open gate” conformation (the blue helix moves away from the point of intrachain actin-actin interaction interface). This allows further actin polymerization until new tropomyosin N-termini are available for Lmod or Tmod attachment. ABS, actin-binding site; Lmod, leiomodin; Tmod, tropomodulin.

https://doi.org/10.1371/journal.pbio.3000848.g008

According to the model we propose, once two actin molecules attach to the pointed end/Lmod2 assembly, actin polymerization continues unrestricted until seven consecutive actin monomers on each side of the filament are added. Afterwards, two tropomyosin molecules bind to the seven monomers cooperatively on each side, and one of them displaces TpmBS1 of Lmod2.

The tropomyosin N-termini at the new pointed end of the filament become available for either Tmod or Lmod binding. Note that polymerization of actin at the pointed end in the presence of Lmod2 may not result in immediate removal of Lmod from the filament; it still can be bound through other actin-binding sites. This observation provides an explanation for why, unlike Tmod1, Lmod2 is found not only at the pointed end but also along the filament near the pointed end [21, 40].

Conclusions

Members of the Tmod family are major regulators of actin filament lengths and dynamics in both muscle and nonmuscle cells [8, 9, 45]. The binding of Tmod/Lmod to the thin filament occurs via several synergistic sites, each being either actin- or tropomyosin-specific [9, 46]. To date, only partial structural information is available on the mode of binding between Tmod/Lmod and actin filaments, and only with respect to actin-specific binding [19, 42, 43]. Understanding the mode of tropomyosin-specific binding, either via TpmBS1 or via TpmBS2, has remained elusive, until this report.

This study provides—for the first time, to our knowledge—a 3D model for the TpmBS1/Tpm1.1 molecular interface. This mode of TpmBS1 binding is only possible at the pointed end of actin filaments where the N-terminus of tropomyosin does not interact with the C-terminus of an adjacent tropomyosin molecule. TpmBS1 is the only tropomyosin-binding site that is universally present in all the members of the Tmod family. A high degree of sequence homology in TpmBS1 between all Lmod/Tmod isoforms suggests that this site plays an important role in Tmod and Lmod tropomyosin-specific recognition at actin filament pointed ends.

Importantly, the sequence comparison between different Tmod and Lmod proteins is in good agreement with the 3D model of the complex. S2 Fig shows TpmBS1 sequences of Lmod2 from different species and TpmBS1 sequences of Lmod/Tmod isoforms. Most of the residues that form the TpmBS1/Tpm1.1 binding interface are either conserved (highlighted in gray) or are conservative substitutions of similar volume and charge/hydrophobicity (highlighted in green). Occasional nonconservative substitutions at the binding interface (highlighted in yellow) may be useful for isoform-specific tuning of binding affinity [26, 36, 46]. The majority of sequence variability is observed for residues that are not in contact with Tpm1.1 (substitutions are highlighted in cyan). The formation of the N-terminal hydrophobic “cap” cluster is consistent with the report that tropomyosin peptides with positively charged unacetylated N-terminus poorly bind Tmod [35].

The debate about the regulatory role of Lmod2 at the pointed end has been centered on the interaction of Lmod2 with tropomyosin via TpmBS1 and with actin via ABS1 [9, 10, 46]. Our structural findings imply that in striated muscles, TpmBS1 of either Tmod or Lmod must bind to the same tropomyosin protomer at the pointed end and that this binding is mutually exclusive. A reconstruction of the Lmod2 assembly at the pointed end demonstrates that there is a considerable overlap between Lmod2 and Tmod1 binding surfaces, meaning that Tmod and Lmod compete for binding and might be involved in a process of constant dynamic exchange. It was shown that there is a continuous incorporation of actin monomers at the pointed end in cardiac myocytes [47]. If thin filaments disassemble at the barbed end, their proper length could be maintained via slow and regulated actin polymerization at the pointed end when Lmod2 substitutes Tmod1. Our data suggest that Lmod2 binding at the pointed end may play an important role in actin exchange in the process of thin filament maintenance.

Similar to Tmod1, Lmod2 should prevent depolymerization from the pointed end. However, unlike in Tmod1, the ABS1 site of Lmod2 occluding the intrastrand actin-actin binding site has the flexibility to move laterally (Fig 8). This gives Lmod the ability to allow further actin polymerization turning it into a leaky cap. Interaction of the amphipathic helix in Tmod1 ABS1 with actin is stronger than that for Lmod2 [23, 25]. Moreover, ABS1 binding in Tmod1 is additionally reinforced by the second tropomyosin-binding site (TpmBS2). In this respect, Tmod’s TpmBS2 acts like a latch, which fixes the helix in its position over actin molecule A, not allowing the helix to move, as opposed to the “swinging gate” helix in Lmod2.

Ethics statement

Experiments utilizing rats were performed under the approval by The Institutional Animal Care and Use Committee at the University of Arizona (protocol # 08–017), which conforms to all applicable federal and institutional policies, procedures, and regulations, including the PHS Policy on Humane Care and Use of Laboratory Animals, USDA regulations (9 CFR Parts 1, 2, 3), the Federal Animal Welfare Act (7 USC 2131 et. Seq.), the Guide for the Care and Use of Laboratory Animals, and all relevant institutional regulations and policies regarding animal care and use at the University of Arizona.

Peptide design and production

The choice of Lmod2s1 and αTM1a_1-14Zip amino acid sequences was described in [27]. Briefly, αTM1a_1-14Zip was designed to contain 14 N-terminal residues of Tpm1.1, followed by a GCN4 leucine zipper sequence [48] to promote formation of a coiled coil in the tropomyosin fragment. Fusing GCN4 leucine zipper sequences with tropomyosin fragments is a common approach in structural studies of tropomyosins and their complexes [34, 49]. To enable the peptide expression and uniform isotope labeling in Escherichia coli, the αTM1a_1-14Zip peptide also included an N-terminal Gly residue. The substitution of an acetyl group with a Gly residue was demonstrated to create a good mimetic for both structural and functional studies of tropomyosin N-terminal fragments and their complexes [34, 35, 50].

Synthetic peptides were synthesized in Tufts University Core Facility (Boston, MA). ¹⁵N- and ¹⁵N/¹³C-labeled human Lmod2s1 peptides were expressed as described previously [27]. ¹⁵N- and ¹⁵N/¹³C-labeled Tpm1.1 peptides were expressed following essentially the same protocol as the expression of labeled Lmod2s1. The expression minimal medium contained 50 mM Na₂HPO₄, 50 mM KH₂PO₄ (adjusted to pH 7.0 with HCl), 30 mM NaCl, 3.7 g/L ¹⁵N-ammonium sulfate, 4 g/L nonlabeled or ¹³C-glucose, 2 mM MgSO₄, 1X BME vitamins (Sigma-Aldrich, St. Louis, MO), and trace elements (20 μM FeCl₃, 4 μM CaCl₂, 2 μM ZnSO₄, 0.4 μM CoCl₂, 0.4 μM CuSO₄, 0.4 μM NiCl₂, and 0.4 μM H₃BO₃), which are known to promote the growth of BL21(DE3) E. coli cells [51]. The first rounds of purification of recombinant human Lmod2s1 and human Tpm1.1 peptides were described in [27]. When necessary, the peptides were additionally purified by reversed-phase HPLC on an Xbridge C18 5 μm 4.6 × 250-mm column (Waters, Milford, MA) to eliminate minor impurities. In the second round of purification, Lmod2s1 was purified in a linear 1%/minute ammonium carbonate/acetonitrile gradient, whereas Tpm1.1 peptides were purified in a linear 1%/minute 100 mM sodium phosphate (pH 11.2)/acetonitrile gradient. The peptides were acidified by addition of 10 μL 6 M HCl per each mL of collected peaks immediately after purification and desalted on the same HPLC column or on a Sep-Pak C18 cartridge (Waters) with 0.1% trifluoroacetic acid (TFA). The desalted peptides were eluted with a 1/1 (v/v) 0.1% TFA/acetonitrile mixture and lyophilized. Peptide concentrations were determined by measuring their difference spectra at 294 nm in 6 M guanidine-HCl between pH 7.0 and 12.5 using the extinction coefficient of 2,357 cm⁻¹ per 1 M tyrosine [27].

The L25G mutation was introduced into Lmod2s1 by amplifying a DNA vector expressing WT Lmod2s1 [27] via PCR using partially overlapping complementary primers incorporating the mutation. The primer sequences were forward primer: 5′-CGAGCGGCAGCGCTGAAGAACTGAAGGAACTGGAACGC-3′, reverse primer: 5′-AGCGCTGCCGCTCGCCAGCAGTTCATCTTCATCAATAGA-3′, where underlined nucleotides encode for Gly instead of Leu25. The PCR reaction was performed with PfuTurbo DNA polymerase (Agilent Technologies, Santa Clara, CA), and the PCR product was used to transform DH5α E. coli cells (Life Technologies, Carlsbad, CA). Presence of the L25G mutation and absence of nonspecific mutations were confirmed by Sanger sequencing (Genewiz, South Plainfield, NJ). Expression of the mutant protein was performed in BL21(DE3) E. coli cells (Life Technologies, Carlsbad, CA) grown overnight in ZYP medium [51]. Purification was performed as described for WT Lmod2s1 [27].

Plasmid construction and site-directed mutagenesis for Lmod2 and Lmod2[L25G] expression in cardiomyocytes

Full-length mouse Lmod2 (mLmod2) was cloned from cDNA generated from mouse hearts. The coding sequence was inserted into the pEGFP-C2 vector (Clontech, Mountain View, CA) using XhoI and HindIII restriction sites. The L25G mutation was introduced into pEGFP-C2-mLmod2 by PCR using Phusion High-Fidelity DNA Polymerase (New England Biolabs, Ipswich, MA) and two partially overlapping complementary oligonucleotides with the following sequences: 5′-TGGCCTCCGGCTCACCTGAAGAGCTGAAGGAGCTTGAGAG-3′ and 5′-CAGGTGAGCCGGAGGCCAGAAGTTCATCCTCATCAATGGATTCA-3′. Substituted nucleotides are underlined. Generated PCR product with the mutation was treated with DpnI (New England Biolabs) to digest the remaining WT template and used to transform DH5α E. coli cells (Life Technologies, Carlsbad, CA). Kanamycin resistance clones were analyzed for presence of the mutation. Plasmids were purified using ZR Plasmid Miniprep Classic kit (Zymo Research, Irvine, CA) according to manufacturer’s recommendations. DNA Sanger sequencing to confirm successful mutagenesis was performed by Eton Bioscience (San Diego, CA). Synthesis of the oligonucleotide primers was performed by Sigma-Aldrich (St. Louis, MO).

NMR sample preparation

NMR samples were prepared by reconstituting and mixing lyophilized Lmod2s1 and Tpm1.1 peptides in 50 mM sodium phosphate buffer (pH 6.5), 10% D₂O, 0.2 mM EDTA, 0.1% sodium azide, and 2X Pierce EDTA-free protease inhibitor cocktail. The sample pH was adjusted by the addition of small volumes of 1 M HCl or NaOH. The pH was measured with the Mettler Toledo (Columbus, OH) U402-M3-S7/200 microelectrode. In each NMR sample, one of the two peptides in the Lmod2s1/Tpm1.1 peptide complex was labeled whereas the other one was unlabeled.

The concentration of the labeled peptide was approximately 0.5 mM. Initially, 1.5-fold stoichiometric excess of the unlabeled peptide was used. However, we found that many cross-peaks in the ¹⁵N-HSQC spectra were broad (S12A and S12B Fig). Under these stoichiometric ratio conditions we estimated the apparent rotational correlation time τ_c of the ¹⁵N/¹³C-αTM1a_1-14Zip/Lmod2s1 complex to be roughly 17 ns [52], which exceeded more than 2-fold a typical τ_c (approximately 7.2 ns) for a rigid and approximately spherical protein of this size [53, 54]. We found that further increase of the concentration of the unlabeled peptide led to narrowing of the resonance lines (S12C Fig). Given that the dissociation constant (K_d) of the complex is in low μM range [27], this observation indicated that chemical exchange between bound and traces (approximately 1%) of unbound labeled peptides contributed significantly to the T2 relaxation [55]. To decrease broadening of the resonance lines that was observed as a result of the dissociation/association exchange process, most 2D and 3D NMR spectra were recorded with the unlabeled peptide being in approximately 3 times stoichiometric excess unless otherwise stated. An attempt to create higher stoichiometric excess led to severe line broadening followed by protein precipitation. Therefore, our ability to reduce the rate of T2 relaxation in the sample was limited by the solubility of the peptide components.

NMR spectra collection

The NMR spectra were recorded at 25°C on a Varian VNMRS 600 MHz spectrometer (Agilent Technologies, Santa Clara, CA) equipped with a 5-mm triple-resonance probe or on a Varian Inova 500-MHz spectrometer equipped with a 5-mm Nalorac Z-Spec HCNP triple-resonance probe. The complete data set included 2D ¹⁵N-HSQC and ¹³C-HSQC and 3D HNCO, HNCA, HN(CO)CA, HNCACB, CBCA(CO)NH, HBHA(CO)NH, HCCH-TOCSY, and ¹⁵N- and ¹³C-edited NOESY [56]. The NMR spectra were processed using NMRPipe [57] and Felix (Felix NMR, San Diego, CA). The 2D ¹⁵N-HSQC titration spectra were processed and analyzed with MestReNova (Mestrelab Research, S.L., Santiago de Compostela, Spain). NMRViewJ (One Moon Scientific) was used for 3D NMR spectra visualization and peak assignment [58] (BMRB ID 30681).

NMR assignment and analysis

Almost all the backbone atom resonances were assigned. However, as a result of the association/dissociation exchange process in the complex, resonance peaks for a number of residues were broadened beyond detection in ¹³C-correlated 3D spectra (S13 and S14 Figs), and only partial assignment of side-chain resonances was achieved. Although not being the primary reason for the incomplete assignment, significant side-chain resonance overlap caused by disproportionally high occurrence of Leu, Glu, and Lys residues in the sequences exacerbated the problem. Because of the incomplete resonance assignment, the generation of a 3D structure for the Lmod2s1/αTM1a_1-14Zip complex on the basis of NOE contacts [59] was not feasible. Therefore, to obtain the structure, we utilized the dependence of ¹⁵N, ¹³C, and ¹H chemical shift values on the local protein backbone conformation [31] to determine the backbone dihedral angles. Protein backbone torsion angle and RCI-S² values [60] were predicted on the basis of ¹⁵N, ¹³C', ¹³Cα, ¹³Cβ, ¹Hα, and ¹H^N chemical shifts by TALOS+ [61] using NMRViewJ-generated TALOS+ input files. TALOS+ predicts dihedral angles empirically on the basis of known NMR and X-ray data from 200 proteins.

Most of the well-structured elements in the complex represented α-helices. In the loop, connecting the first and second helices of Lmod2s1, Ser26 was in a β conformation, but the conformation of Leu25 was not assigned unambiguously. Out of the 10 best database matches made by TALOS+ for Leu25, nine were in the β region of the Ramachandran plot, and one remaining match had a positive φ value corresponding to an α_L (left-hand helical) conformation. Motifs ββ and α_Lβ are the most common two-residue motifs connecting two α-helices [39]. Leu and Ser show high propensities for the first and second positions in ββ motifs, respectively, whereas the α_Lβ conformation is favored by Gly and Asp. Leu is not likely to be found as the first residue in the α_Lβ motif [39], and although it was not restrained in MDSs, it was initially set to a β conformation. Calculated backbone dihedral angles (φ,ψ) of Asp2-Leu28 in αTM1a_1-14Zip were found in the α_R region (approximately −60°,−45°) of the Ramachandran plot, and the residues were found to be well-ordered. The RCI-S² values for these residues were found to be within the range of 0.79–0.90 (S1 Table).

To validate the use of TALOS+ for our specific complex, we tested how well TALOS+ identifies torsional angles for the same αTM1a_1-14Zip peptide but when it was in a complex with TM9a_251-284 (PDB ID 2G9J). Chemical shifts and 3D NMR structure for the complex were determined by Greenfield and colleagues [34]. On the basis of chemical shifts, TALOS+ generated a αTM1a_1-14Zip secondary structure similar to that in the Lmod2s1/αTM1a_1-14Zip complex, with a continuous Asp2-Leu28 α-helix. Residues Val29-Arg32 were not classified as part of the helix, and they displayed an increased dynamic behavior (disorder) toward the C-terminus. Val29 was determined to be in a α-helical conformation, Gly30 was determined to be in an α_L conformation, whereas Glu31 was mapped to the β region. In agreement with TALOS+, backbone relaxation parameters (¹H-¹⁵N heteronuclear NOE and R1 and R2 ¹⁵N relaxation rates) determined in [34] demonstrated an increased flexibility of the C-terminal residues Gly30-Arg32. However, the results of structure calculation made by Greenfield and colleagues determined both Gly30 and Glu31 to be in the α-helix [34]. The differences in Gly30-Glu31 conformations are likely to arise from their intrinsic dynamic behavior. Overall, for the known tropomyosin peptide structure, TALOS+ produced results in a good agreement with the reported NMR data.

RDC values

To obtain RDC values, S3 NMR spectra [62] of the sample were recorded in 5% alkyl-PEG bicelles (n-octanol/C8E5) [63]. First, we recorded S3 NMR spectra of ¹⁵N-labeled Lmod2s1 in the presence of excess unlabeled αTM1a_1-14Zip. By comparing ¹H-¹⁵N cross-peak splitting in oriented and isotropic (without bicelles) environments, we determined ¹H-¹⁵N RDC values for the ¹⁵N-labeled Lmod2s1 in the complex. However, we were unsuccessful in obtaining RDC values for the ¹⁵N/¹³C-labeled sample of αTM1a_1-14Zip in the presence of excess unlabeled Lmod2s1, because oriented n-octanol/C8E5 bicelles would not form for this sample. We explained this n-octanol/C8E5 behavior by its interaction with unbound Lmod2s1, which, judging by its longer retention time in the reversed-phase HPLC profile, is more hydrophobic than αTM1a_1-14Zip and can potentially interfere with the oriented phase formation. An RDC uncertainty of 1–2 Hz was estimated from comparing RDC values determined by fitting resonance peaks in NMRViewJ and NMRFAM-SPARKY [64].

NMR titration

For NMR titration of ¹⁵N-Lmod2s1 with αTM1a_1-28Zip, the Lmod2s1 peptide solution (0.3 mM) was prepared in 50 mM sodium phosphate buffer (pH 6.5), 10% D₂O, 0.2 mM EDTA, 0.1% sodium azide, and 2X Pierce EDTA-free protease inhibitor cocktail. The pH of 70 μM αTM1a_1-28Zip solution in water was adjusted to 6.5; aliquots were taken and lyophilized to dryness. The ¹⁵N-Lmod2s1 sample was added to dry αTM1a_1-28Zip to create molar ratios Lmod2s1:αTM1a_1-28Zip equal to 40:1, 30:1, 20:1, 15:1, 10:1, and 7:1. After reconstitution of αTM1a_1-28Zip, ¹⁵N-HSQC spectra were taken on the Varian Inova 500 MHz spectrometer at 25°C.

MDSs

Polypeptide structure editing and visualization was performed in UCSF Chimera [65]. Initial backbone dihedral angles were set to values predicted by TALOS+. The complexes were neutralized with Na⁺ ions and placed in a cuboid box of TIP3P water molecules [66] with a minimal distance of 10 Å from the protein complex to the edge of the box. Protonation states of histidine residues were set to neutral. The MDSs were performed using AMBER18 [67] with the ff14SB force field [68] and periodic boundary conditions. The cutoff value for nonbonded interactions was 8.0 Å (default). The pressure was set to 1 bar using a Berendsen barostat with 1-ps relaxation time (default). Energy minimization was performed with the steepest descent method for 2,500 cycles followed by the conjugate gradient method for 2,500 cycles. The temperature in MDS runs was controlled by a Langevin thermostat with a 3-ps⁻¹ collision frequency. Hydrogens were constrained using SHAKE. Backbone dihedral angles φ and ψ of ordered residues were restrained by values predicted by TALOS+ using dihedral potential constants rk2 = rk3 = 32.0 kcal × mol⁻¹ × rad⁻¹. Residues with restrained φ and ψ were Glu18-Ser24 and Ser26-Asp39 of Lmod2s1 and Asp2-Gly30 of αTM1a_1-14Zip. Dihedral angles of Leu25 were left unrestrained.

Depending on the starting point, initial 40-ns MDS runs at 298 K produced a few different modes of the crisscross Lmod2s1/αTM1a_1-14Zip binding, which could be generally grouped into structures with either the short (Glu18-Ser24) or the long (Ala27-Glu38) Lmod2s1 α-helix being roughly parallel to the tropomyosin coiled coil. These structures also differed in the conformation of the linker residue Leu25, which adopted either a β or an α_L conformation. To converge these distinct structures, we employed a simulated annealing algorithm whereby the preliminary structures were first heated from 300 K to 360 K and then cooled back in 10-degree increments. The heating was stopped at 360 K because above 360 K, the complex lost most of its secondary structure as well as ionic and nonpolar interactions. The complex was equilibrated at each temperature for 10 ns before the temperature was changed. The time step for simulated annealing was set to 1.0 fs. The equilibration time of 10 ns was chosen based on in silico tests that showed that characteristic times of α_L-to-β and β-to-α_R transitions of Leu25 were in the range of 1–2 ns.

To characterize the convergence quantitatively, we chose two structures—one from each group of the crisscross Lmod2s1/αTM1a_1-14Zip binding modes obtained in initial simulations—as starting structures in the simulated annealing protocol. We called the starting structure with the long (Ala27-Glu38) helix parallel to the tropomyosin coiled coil “model 1,” whereas the structure with the short (Glu18-Ser24) helix parallel to the tropomyosin coiled coil was called “model 2.” When backbone atoms N, Cα, and C' of well-structured residues Asp17-Asp39 (Lmod2s1) and Asp2-Leu28 (αTM1a_1-14Zip) in the two “models” were matched, RMSD between the same atoms was 3.77 Å. For the same matched structures, RMSD between the backbone atoms N, Cα, and C' of Lmod2s1 was 5.18 Å, thus reflecting the large difference between positions of Lmod2s1 in the two “models” with respect to the tropomyosin coiled coil. Seven replicas of simulated annealing were performed for each of the two starting structures. For each “model,” an ensemble of seven structures was produced. Two structures in each ensemble dissociated during the heating steps, lost the majority of interstrand contacts, and failed to associate back during the cooling steps; they were consequently excluded from analysis. The remaining five structures in each ensemble were averaged and the comparison between the two average structures gave the global RMSD (calculated as above) equal to 1.36 Å and the RMSD for Lmod2s1 equal to 2.14 Å. A pairwise global RMSD value calculated for a cluster including all 10 structures from both ensembles was 2.34 ± 0.09 Å (mean ± SEM), and the respective pairwise RMSD for Lmod2s1 was 3.63 ± 0.14 Å (S2 Table). Visual inspection of all the 10 annealed structures confirmed that their topology was similar in that in all of them the long Lmod2s1 helix (Ala27-Glu38) was approximately parallel to the tropomyosin coiled coil, whereas the linker residue Leu25 adopted a β conformation and was inserted between the strands of αTM1a_1-14Zip.

One of the annealed structures was used as a starting point for twenty 400-ns MDS production runs. The temperature was set to 298 K, and the time step was set to 1 fs. Ten equilibrated complexes were subjected to energy minimization to eliminate bond length and angle distortions caused by thermal motions (PDB ID 6UT2). Global pairwise RMSD calculated for backbone atoms N, Cα, and C' of well-structured residues Asp17-Asp39 (Lmod2s1) and Asp2-Leu28 (αTM1a_1-14Zip) in the 10 structures was 1.76 ± 0.07 Å (mean ± SEM). The pairwise RMSD statistical data are attached to supporting information (S3 Table).

Discrimination between the side-by-side and crisscross packing of the helices in the complex was performed by comparing the evolution of two metrics along MDS trajectories. In the simulations, the run time was 80 ns, the temperature was set to 300 K, volume to constant, and the Langevin thermostat collision frequency was 1 ps⁻¹. Other MDS conditions were as above. The first metric was “rolling RMSD,” which was calculated between two structures of the complex, one current and another one observed 1 ns before the current. To calculate the rolling RMSD, the two complexes were first matched over the backbone atoms N, Cα, and C' of residues Asp17-Asp39 (Lmod2s1) and Asp2-Leu28 (αTM1a_1-14Zip). After matching, the rolling RMSD value was calculated between the coordinates of the backbone atoms N, Cα, and C' of Asp17-Asp39 in Lmod2s1. The second metric was the RMSD for backbone dihedral (φ,ψ) angles between the current structure and the TALOS+ prediction (calculated for residues Asp17-Asp39 in Lmod2s1).

Simulation parameters for two linkers connecting TpmBS1 to ABS1h and ABS1h to ABS2 were the same as those for Lmod2s1/αTM1a_1-14Zip, except for those associated with temperature and pressure control. The temperature was controlled by Berendsen thermostat with a 1-ps coupling constant, and the volume was constant. The simulations for the linkers were run for 50 ns.

Model validation by RDC

RDCs observed in anisotropic samples report information on internuclei vector orientations, and they are often used for 3D structure validation and refinement [30, 69]. We used RDC values obtained from the S3 spectra in oriented medium for the validation of the Lmod2s1/αTM1a_1-14Zip structure (S15 Fig). Consistent with the RCI-S² values determined by TALOS+, the RDC values of Phe4-Ser11 were small and did not exceed approximately 2 Hz, which is typical of dynamic disoriented residues [30]. Some backbone stiffening of Lys12-Ile16 correlated with the increase in their RDC values which range from approximately −6 to 3 Hz. The experimental RDC values for ordered residues Glu18-Glu38, which corresponded to the two Lmod2s1 helices and the connecting loop, had a good correlation with theoretical RDC values obtained by REDCAT [70] from the MDS-generated structure in Fig 1 (S16 Fig). Specifically, the RMSD was approximately 2.3 Hz, which is only slightly higher than the experimental uncertainty of 1–2 Hz, and the quality Q-factor was approximately 19%. The realistic experimental minimum of Q-factor is approximately 10%, with the best structures giving Q-factor values in the range of 10%–15%, whereas 20%–25% roughly correspond to a 1.8-Å X-ray structure [71].

CD

CD measurements were performed using an Aviv model 420 CD spectropolarimeter (Lakewood, NJ). Spectra were recorded from 260 nm to 195 nm in 10 mM sodium phosphate buffer (pH 7.0), and 100 mM NaCl, at 0°C. Melting curves were measured at 222 nm from 0°C to 60°C with a temperature step of 0.2°C, a 0.15-°C deadband, a 0.3-min equilibration time, and 5-s averaging time. Concentrations of Lmod2s1 (WT), Lmod2s1 (L25G), and αTM1a_1-14Zip were 10 μM.

Lmod2 assembly at the pointed end

To build the model of Lmod2 assembly at the pointed end, an atomic model for the tropomyosin cable with F-actin was used as a starting point [44]. To create a model of the pointed end available for Lmod binding, the tropomyosin molecule providing the C-terminal “tail”in the head-to-tail tropomyosin overlap complex was removed. The part of the F-actin filament that was in contact with the removed tropomyosin protomer was also removed. In the resulting complex (named Tpm-Factin), two actin molecules at the pointed end were labeled as A and B. To dock TpmBS1, ABS1h, and ABS2 (LRR) domains of Lmod2 to Tpm-Factin, first the Lmod2s1/αTM1a_1-14Zip complex was added to Tpm-Factin via superimposing αTM1a_1-14Zip and the N-terminus of tropomyosin using the MatchMaker tool in UCSF Chimera. The Lmod2 LRR domain was docked by superimposing actin molecule B from Tpm-Factin and the complex between LRR from Lmod2 and G-actin (PDB ID 5WFN [42], calculated using raw X-ray data obtained by Chen and colleagues [43]). As an approximation for Lmod2 ABS1h docking, we superimposed the complex of G-actin with homologous Tmod1 ABS1 (PDB ID 4PKG, [19]) with actin A of Tpm-Factin.

Two polypeptide chains connecting TpmBS1 with ABS1h and ABS1h with ABS2 of Lmod2 were subjected to 50-ns MDS runs to identify potential formation of secondary structure elements and incorporated into the model. The result of the simulation was compared with secondary structure predictions made by three server-side secondary structure predictors, Jpred4 [72], PsiPred [73], and PredictProtein [74]. The contour lengths (L) of the linkers including α-helices and random coil residues were estimated using the formula L = 4 Å × (number of random coil residues) + 1.5Å × (number of α-helical residues) [75]. Redundant actin molecules (from Tpm-Factin), αTM1a_1-14Zip, nonhomologous part of Tmod1 ABS1, and gelsolin (from 4PKG) were removed. To create the final view of the assembly, we performed energy minimization using Amber18. Complexes forming at steps 1 and 2 of the putative thin filament elongation process were docked and minimized in a similar fashion.

Isolation, transfection, and immunofluorescence microscopy of neonatal cardiomyocytes

Rat neonatal cardiomyocytes were isolated and plated on 35-mm tissue-culture dishes containing 12-mm round glass coverslips as described in [5]. Within 12 hours of plating, cells were transfected with 1 μg of either GFP, GFP-Lmod2 WT, or GFP-Lmod2 [L25G] DNA using Lipofectamine 3000 reagent (Thermo Fisher Scientific, Waltham, MA). Forty-eight hours after transfection, cells were washed with phosphate-buffered saline (PBS) and incubated in relaxing buffer (10 mM 3-[N-morpholino] propane sulfonic acid [pH 7.4], 150 mM KCl, 5 mM MgCl₂, 1 mM EGTA, and 4 mM ATP) for 15 min and fixed with 2% paraformaldehyde in relaxing buffer for 15 min at room temperature. Following fixation, cells were permeabilized with 0.2% Triton X-100/PBS for 20 min at room temperature and then blocked with 2% bovine serum albumin (BSA) plus 1% normal donkey serum/PBS for 1 h at room temperature. Cells were then incubated overnight at 4°C with rabbit polyclonal anti-Tmod1 (2 μg/mL) [76] and mouse monoclonal anti-α-actinin (1:200) (EA-53; Sigma) primary antibodies. Following incubation, cells were washed with PBS for 3 × 5 min and incubated for 1.5 h at room temperature with secondary antibodies/PBS, which included Alexa Fluor 405–conjugated goat anti-mouse IgG (1:200) (Thermo Fisher Scientific), Alexa Fluor 647–conjugated donkey anti-rabbit IgG (1:300) (Jackson ImmunoResearch Laboratories, West Grove, PA), and Texas Red-X Phalloidin (Thermo Fisher Scientific). Cells were then washed with PBS for 3 × 5 min and mounted onto slides with Aqua Poly/Mount (Polysciences, Warrington, PA). Images were captured using a Nikon Eclipse Ti microscope with a 100 × NA 1.5 objective, and a digital complementary metal oxide semiconductor (CMOS) camera (ORCA-flash4.0; Hamamatsu Photonics, Shizuoka Prefecture, Japan). 3D deconvolution was performed using NIS offline deconvolution software (Nikon Corporation, Tokyo, Japan), and images were processed using Photoshop CC (Adobe, San Jose, CA). Thin filament lengths and sarcomere lengths were measured using the DDecon plugin for Image J [77, 78]. Thin filament lengths were analyzed between a sarcomere length range of 1.85–2.40 μm. Sarcomere lengths were not different between the groups within this range. For determining thin filament pointed-end assembly of Tmod1, positively transfected cardiomyocytes were classified as either “consistent” well-defined striated Tmod1 staining with little cytoplasmic background or “inconsistent,” partial striated Tmod1 staining with high levels of cytoplasmic background.

Statistics

All statistical analyses were performed in Prism 7.70 (Graphpad Software, San Diego, CA). Multiple groups were compared with one-way ANOVA with Tukey’s post hoc test. Differences with p < 0.05 were considered statistically significant.