Obscurin is a semi‐flexible molecule in solution

Abstract

Obscurin, a giant modular cytoskeletal protein, is comprised mostly of tandem immunoglobulin‐like (Ig‐like) domains. This architecture allows obscurin to connect distal targets within the cell. The linkers connecting the Ig domains are usually short (3–4 residues). The physical effect arising from these short linkers is not known; such linkers may lead to a stiff elongated molecule or, conversely, may lead to a more compact and dynamic structure. In an effort to better understand how linkers affect obscurin flexibility, and to better understand the physical underpinnings of this flexibility, here we study the structure and dynamics of four representative sets of dual obscurin Ig domains using experimental and computational techniques. We find in all cases tested that tandem obscurin Ig domains interact at the poles of each domain and tend to stay relatively extended in solution. NMR, SAXS, and MD simulations reveal that while tandem domains are elongated, they also bend and flex significantly. By applying this behavior to a simplified model, it becomes apparent obscurin can link targets more than 200 nm away. However, as targets get further apart, obscurin begins acting as a spring and requires progressively more energy to further elongate.

Short abstract

PDB Code(s): 6MG9

Introduction

Most cells in the body are subjected to motion, ranging from muscle cells contracting and relaxing to epithelial cells conforming to body movement.1 Yet cells also must be physically strong to maintain homeostasis and normal architecture amidst this strain.2 Giant cytoskeletal proteins are long, chain‐like molecules that connect distal cellular regions and have the capacity to bend and stretch.3, 4, 5, 6 Thus, these proteins provide a potential mechanism to assist the cell in its capability to be both flexible and strong.

The most well‐known giant cytoskeletal protein is titin. This protein spans from the Z‐disk to the M‐band in myocytes and is mostly composed of hundreds of consecutive, individually folded Ig domains.5 Through a domain unraveling mechanism, titin acts as a molecular spring, resisting stretch force longitudinally as the muscle cell overextends.7, 8, 9, 10

Obscurin, another giant cytoskeletal protein, has a similar architecture to titin.5, 11 This protein can be found in at least 20 different forms, ranging from 20 kDa to 970 kDa.12 At its longest, the N‐terminal two‐thirds of the protein is composed of over 60 tandem Ig and Fibronectin (FnIII)‐like domains connected to their neighbors via short linkers.13 The C‐terminus contains multiple signaling domains (i.e., PH, RhoGEF, IQ)13, 14 and either an ankyrin binding region (in obscurin A isoforms)15 or kinase domains (in obscurin B isoforms).16 Obscurin's multiple functions are closely linked to its complex subcellular localization.17, 18 In skeletal muscles, the ankyrin binding region of Obscurin A binds to small Ankyrin 1 (sAnk1) at the sarcoplasmic reticulum.15, 19, 20, 21 Ablation of this interaction reduces sAnk1 levels, which in turn leads to aberrant Ca²⁺ homeostasis.22, 23, 24 Likewise, obscurin interacts with Ankyrin‐B in the costamere. When this interaction is disrupted, skeletal muscles experience increased exercise‐induced damage due to the improper assembly of the dystrophin complex.25 Obscurin B binds to and phosphorylates N‐cadherin at the intercalated disk in cardiomyocytes, suggesting that it may modulate muscle cell adhesion.16 Complementing these membrane‐associated interactions, obscurin binds to the sarcomeric contractile apparatus in several locations.5 The 58th and 59th obscurin Ig‐like domains form a complex with the titin ZIg9 domain at the Z‐disk during development, suggesting that obscurin plays a role in myofibrillogenesis.13, 23, 26, 27, 28 Additionally, the N‐terminus of obscurin interacts with titin, slow myosin binding protein C, and myomesin at the M‐band, contributing to the M‐band lattice assembly, structure, and strength.29, 30, 31, 32 Thus, obscurin forms the only known connection between the muscle contractile apparatus and the surrounding membrane structures.5, 29, 33, 34 Clinically, obscurin is linked to breast and colorectal cancers, and obscurin knockdown cells undergo epithelial‐to‐mesenchymal transition.35, 36, 37 In muscle, specific obscurin mutations that alter target protein binding are causally linked to hypertrophic cardiomyopathy, dilated and restricted cardiomyopathy, and muscular dystrophy.33, 38, 39, 40, 41, 42

In order to better understand both how obscurin exists in solution and responds to stretch, here we study a series of representative tandem obscurin Ig domains using structural biology and computation techniques. We find that these dual domain constructs are predominantly extended in solution, yet the domains are also moderately mobile relative to each other. This finding led to the question of how these domains could be extended (suggesting a framework to maintain this conformation) and also flexible (suggesting that there is not a significant framework present). MD simulations suggest that transient non‐covalent bonds between mobile regions in neighboring domains are largely responsible for these dual domains being extended yet dynamic.

Results

Implicit in the observation that obscurin links various cellular targets to each other is the fact that the protein must act as a tether. While obscurin–target interactions in muscle are increasingly well documented, the conformation and dynamics of the obscurin region between these anchor points (the tether) are less understood.5, 16, 22, 29, 30, 38 Here, we investigate obscurin's conformation in solution. In an effort to more easily collect high‐resolution information about this protein, we utilized a reductive approach and studied a series of representative obscurin dual‐domain systems. The linkers between obscurin domains can be broadly divided into short linkers containing proline residues (48% of all obscurin linkers), short linkers with no proline residues (22%), and long linkers (>6 residues) (30%) (Table S1). Previous studies, plus basic biochemistry knowledge, suggest that the proline‐containing linkers may be more rigid, and long linkers are almost certainly more flexible.43 Here, we study two constructs with proline‐containing linkers and two constructs with proline‐absent linkers to better understand the mobility these short linkers confer on the obscurin molecule as a whole.

Multiple solution structures of individual obscurin Ig‐like domains are already in the Protein Data Base (PDB; Table S2). Included in this set of structures are many that connect to neighboring domains via short proline‐containing linkers (i.e., Ig34, Ig35, and Ig36 in full‐length obscurin).13 However, only two published structures – Ig58 and Ig59 – are connected with a non‐proline linker.38, 39 Therefore, in order to generate a more robust data set for studying domain/domain motion, we first solved the solution structure of Ig57, a domain that connects to Ig58 via a non‐proline linker. The heteronuclear single quantum coherence (HSQC) spectrum of Ig57 is well‐dispersed, and every backbone peak was subsequently sequence‐specifically assigned [Fig. 1(A)]. The resulting solution structure is of high quality, with more than 10 distance restraints per residue and no violations greater than 0.40 Å (Table S3 and Fig. S1). The 20 best structures overlay well with each other, with a backbone root‐mean‐square deviation (RMSD) of residues in the Ig‐like fold being 0.681 ± 0.061 Å. The best structure, judged by having the lowest RMSD, shows Ig57 arranged into a typical Ig‐like fold, with its two beta sheets arranged into a beta sandwich‐like fold [Fig. 1(B)].

Figure 1.

Solution structure of Ig57 (Residues 4252–4336 of human obscurin). (A) Fully assigned HSQC of Ig57. (B) Cartoon of the best Ig 57 structure. Further structural analysis can be found in Figure S1 and Table S3.

Next, we constructed a series of dual Ig domains. Ig34/35 and Ig35/36 have short proline‐containing linkers, and Ig57/58 and Ig58/59 have short proline‐absent linkers [Fig. 2(A)]. All of the domains, individually, are fully assigned using multidimensional heteronuclear Nuclear Magnetic Resonance Spectroscopy (NMR). For each dual domain system, the resulting HSQC is almost exactly the sum of the individual domain HSQCs overlaid on top of each other [Fig. 2(B)]. This indicates the individual domains do not significantly interact with their neighbor, except at the extreme poles where the linker connects the two domains [Fig. 2(C)]. In addition, there was no evidence of peak splitting in any of the HSQC spectra, indicating these tandem domains are either in a single conformation, or else are in fast exchange between several different conformations. In all cases, the linker residues between two domains were exchange‐broadened out and could not be assigned, regardless of temperature (37°C, 25°C, and 10°C). Additionally, no nuclear Overhauser effect (NOE) correlations were observed between tandem domains or between domains and their adjoining linkers, supporting the notion that these domains are dynamic relative to each other, and that all short linkers, regardless of composition, experience significant intermediate‐timescale (μs–ms) motions.

Figure 2.

NMR dual domain construct analysis. (A) Each dual domain construct labeled with the linker sequence and PDB ID. (B) HSQC overlay of Ig58 (red), Ig59 (gold), and Ig58/59 (blue). (C) Chemical shift perturbation maps of each dual domain system of obscurin. Residues colored red indicates a significant HN‐N chemical shift change (>2× st.dev.) in the dual domain systems compared to the single domains. Note that the chemical shift changes are randomly distributed around the models, which suggests that the domains do not significantly interact with each other.

Due to the paucity of inter‐domain NOE correlations, we cannot use traditional NMR methods to determine the conformation of these dual domains in solution. Therefore, we attempted two orthogonal techniques to better understand the solution structures of these tandem domain systems: small angle X‐ray scattering (SAXS) and residual dipolar couplings (RDC). Guinier plots of SAXS data [Figs. 3(A), S2, and S3] show all dual domain systems are extended in solution and have similar R _g values regardless of linker composition [Fig. 3(B)]. We next fit our SAXS data to an ensemble of tandem domain models, each with different domain/domain angles [Fig. 3(C)]. In all constructs, the models that best fit the experimental data were almost fully extended, in agreement with our Guinier analysis. However, the residuals of our best fits were non‐random in most cases. Therefore, we re‐fit the data using a two‐state model: one extended conformation and another compact [Figs. 3(D) and S4]. These two‐state models showed a better fit with the data, suggesting that all constructs are usually, but not exclusively, extended. Further fitting of more complex models yielded progressively better fits. As additional evidence of this apparent flexibility, RDC data on two of the tandem constructs (Ig35/36 and Ig58/59) show while the individual domains fit the data well, the data cannot be forced to fit any single dual domain model [Fig. S2(B, C)]. In sum, for every dual construct we tested, we conclude tandem Ig domains are relatively extended but can also exist in multiple conformations.

Figure 3.

SAXS analysis of a representative obscurin dual domain construct. (A) Guinier plots of two concentrations of Ig34/35: 1 mg/mL (black) and 3 mg/mL (red). (B) Dimensions of each dual domain system, calculated from the R _g values (R_g = √(3*Guinier slope and dimension = R _g*2). (C) The Ig34/35 best fit model (orange) compared to other models (gray), calculated via MultiFoXS.57 (D) Comparison of one state model fit (orange) and the two state model fit (blue), to Ig 34/35 1 mg/mL experimental SAXS data (circles).

The finding that every tandem dual domain system is both extended yet flexible seems paradoxical. To address the problem of how these systems can simultaneously have this kind of structure and flexibility, we require high‐resolution information of the various domain/domain interfaces. However, no NOE measurements exist between any of these regions. Additionally, this apparent domain/domain flexibility precludes X‐ray crystallography analysis due to the potential of significant crystal packing artifacts. Therefore, we turned to molecular dynamic simulations (MD) in an attempt to find possible domain/domain or domain/linker interactions. All subsequent tandem domain models were first equilibrated for >50 ns, and the angle between the domains in solution was then measured over an additional 50 ns in triplicate [Fig. 4(A)]. In all simulations, each of the dual domain systems maintained a relatively extended structure on average, but the inter‐domain angle varied widely, with a maximum change of orientation ~50–70°. A global examination of these simulations suggests that these extended conformations are the result of steric hindrance between the domains; neighboring domains with short linkers clash into each other if the angle between them is less than ~120°. These elongated but dynamic dual domain systems persist at least into the microsecond regime (Fig. S5), and are in excellent agreement with our experimental data.

Figure 4.

Molecular dynamics analysis of dual domain constructs. (A) Domain angle versus time graph. (B) Snapshot of the Ig34/35 domain/domain interface, showing likely interactions between the domains and the linker.

Closer examination of these MD trials showed that in all simulations, multiple residues at the domain poles participate in long‐lived, stabilizing interactions with moieties in the linkers [Fig. 4(B)]. Once these interactions form, they usually persist for the duration of the simulation and are largely independent of domain/domain bending. To study these interactions in more depth, we next performed steered molecular dynamics simulations (SMD) on these systems, where the domain termini were moved apart at a constant velocity of 1 Å/ns. By elongating the dual‐domain systems, SMD gives a more controlled setting to study how these putative inter‐domain and domain/linker interactions respond to bend and stretch. SMD also simulates a physiologically reasonable timescale of stretch,44, 45 and thus gives us insight into how obscurin may respond to stretch in the cell.

When a slightly bent dual domain system is stretched, the domains first straighten, yet many of the inter‐domain and domain/linker interactions remain intact [Fig. 5(A,B)]. This is accompanied by the addition of either no or very little work to the system [Fig. 5(C)]. Only after the domains completely straighten does the linker begin to extend and these interactions begin to break [Fig. 5(D)]. Thus, these non‐covalent interactions, originating on linker regions or on loops within the Ig domains, are both long‐lived and flexible. The existence of such flexible interactions explains how dual domain systems can simultaneously be extended and dynamic. Despite each construct having a different composition, all four sets of dual domain systems displayed this same behavior (Fig. S6). As the domains are stretched further, increasingly more work must be added to the system until individual domains unravel. This kind of work‐stretch profile occurs in all model constructs and is reminiscent of other well‐studied multi‐Ig‐domain systems.9, 46 These domain‐rupturing events present an oft‐used cellular mechanism through which obscurin can resist large stretch forces.10

Figure 5.

Representative steered molecular dynamics analysis. (A) Domain angle versus time graph of Ig34/35. (B) Distance between the functional groups in one Ig34/35 simulation. In these measurements, distances of ~5–6 Å denote the distance of a hydrogen bond in this trace. (C) Work versus time graph of Ig34/35. (D) Hierarchical model of obscurin extending with increasing stretch. Domains first straighten, followed by linker straightening, followed by domain unraveling.

Discussion

The N‐terminal majority of obscurin is composed of unique Ig‐like and FnIII‐like modular domains. Of the approximately 60 linkers that connect these domains, around 70% are 3–4 residues in length. Here, we study four representative short linkers. Dual‐domain systems with proline‐containing linkers and dual‐domain systems with proline‐absent linkers are equivalently flexible in solution. Domain/domain orientation tends to be around 160 ± 20°: almost fully extended. MD studies suggest these multiple orientations are of near‐equivalent energies, and thus experimental high‐resolution techniques are inadequate for studying this type of multi‐domain dynamic system. Through extensive MD simulations and analyses we find, in all constructs, short linkers facilitate specific domain/linker and domain/domain interactions. These interactions occur predominantly on loops and other disordered regions of the protein, and can tolerate both moderate compression and stretch. While the exact bonds that form are inherently unique at every interface, each construct we have studied exhibits multiple examples of these interactions. The overarching conclusion is while short linkers facilitate such interactions, the regions containing these bonds are sufficiently flexible to allow significant domain motion. However there is a limit to this flexibility; when the domains bend excessively, the surfaces begin to bump into each other thus resisting further bending. Thus, the existence of short linkers may be a mechanism in multi‐domain proteins to avoid unwanted domain/domain clamshell formation. Conversely, when two extended domains are pulled apart, interdomain bonds break well before the domains themselves rupture.

Previous structural studies of a similar system in titin concluded that short linkers, similar to those present in obscurin, lead to an extended conformation of Ig domains, and this conformation is maintained through a series of domain/domain and domain/linker non‐covalent interactions.43 However, computational studies on these same systems suggest that consecutive domains are flexible relative to each other.7 Thus, the idea presented in this study that short linkers in obscurin facilitate domain/domain and domain/linker interactions and these interactions can tolerate domain motion, reconciles longstanding discrepancies between experimental and computational work on the molecular flexibility of titin.7, 43

From the data gathered here, we created a simple model of how obscurin behaves in solution (Fig. 6). In this model, we assumed the obscurin molecule is unhindered between the beginning and the end of its tandem Ig region (i.e., it participates in no target binding in the middle of the molecule), the Ig region consists of 60 domains, each domain is 4 nm in length, and a two‐domain system bends a maximum of 45° away from 180°. With these inputs, one can create a random walk trajectory [for example, see Fig. 6(A)]. Figure 6(B) shows a distribution curve of the distance between the termini of this model and suggests they will be, on an average, around 76 nm apart from each other in solution. Of note, the input values can be altered, resulting in minor changes in the average termini distance (Fig. S7). In this model, it is worth noting that the distance between termini range from 0 nm to ~239 nm. Given these constraints, and given the work that others have done on similar proteins, a reasonable model of this system is a worm‐like chain model.43, 47 Thus, with knowledge of the persistence length and contour length, we can calculate the entropic energy required to completely extend obscurin (to 239 nm), and we find this force is small: only around 28 J/mol. Further separation of the termini, up to around 270 nm (or around 5 Å per linker), requires the flexible non‐covalent bonds to break in order to fully extend each linker region. From our SMD measurements, this extension is associated with 1–10 kJ/mol of work per linker. This extension range is likely where obscurin behaves as a physiologically relevant molecular spring.47 Extension past 270 nm begins unraveling individual Ig‐like domains, and requires a significant amount of work, likely in a manner reminiscent of how titin resists overextension.10 Thus, if obscurin links two distal targets at each termini, it will behave as a slack rope as long as those targets are less than 240 nm from each other. As the targets separate further, obscurin begins behaving as a spring, progressively resisting more force as the objects are moved farther apart from each other. This model presents obvious control points to tune such a system; adding additional anchor points to obscurin through interactions with domains in the middle of the protein, will correspondingly reduce the chain length and create a stiffer spring. Our model is overly simplistic; obscurin contains several regions of longer linkers (Fig. S1), and some tandem domains may more strongly interact with each other. Additionally, parts of the obscurin C‐terminus are non‐modular and other parts contain signaling domains, which our model does not take into account. Further research in these other obscurin regions will lead to a more refined model and should provide more detailed insights into how obscurin behaves in the context of the myocyte.

Figure 6.

Simplified model of obscurin dynamics in solution. (A) Five examples of random‐walk simulations, where the i + 1 domain is allowed to bend between 0° and 45° in any direction relative to the i domain. This model is of 60 domains. (B) The end‐to‐end distance distribution curve of 100,000 simulations, showing that the obscurin N and C termini are most often roughly 76 nm apart from each other, given the inputs specified.

Conclusions

Here we show obscurin tandem Ig‐domains adopt an elongated orientation in solution. Despite staying moderately extended, the domains have a range of flexibility. This physical characteristic is brought about through the soft interface between neighboring Ig domains, and the interactions this interface creates. These interactions are postulated to help prevent self‐association with neighboring domains. As a consequence of this elongated‐yet‐dynamic structure, obscurin does not significantly resist stretching force until the inter‐domain linkers, and eventually the Ig domains themselves, begin to unravel. This hierarchical stretching profile allows for a simple model of obscurin flexibility.

Materials and Methods

Protein isolation

All chemicals were ACS grade or higher and were purchased from Fisher Scientific, unless otherwise specified. Recombinant ¹⁵N, ¹⁵N–¹³C, and unlabeled protein were purified after overexpression in Escherichia coli (BL21[DE3]) using pET24a vector system (Novagen, San Diego, CA). All constructs were induced at 37°C with 100 μM IPTG at an OD₆₀₀ = 0.6 and grown for additional 4 h at 37°C. Cells were sonicated and centrifuged in a small amount of buffer containing 50 mM phosphate pH 8.0, 300 mM NaCl, 10 mM imidazole and 1 mM PMSF. The resulting cleared supernatant was passed over Ni‐NTA His‐bind Resin (Novagen). The column was washed extensively with 50 mM phosphate pH 8.0, 300 mM NaCl, 10 mM imidazole buffer, and eluted with the same buffer plus 500 mM imidazole. Fractions containing the protein were then concentrated in 5000 Da MWCO concentrators (Corning SpinX, Tewksburg, MA) and applied to a Sephadex G75 (Sigma, St. Louis, MO) size exclusion chromatography column in 50 mM Tris pH 7.5, 20 mM NaCl, 0.35 mM NaN₃ (G75 buffer). Pure protein, as determined by SDS‐PAGE, was once again concentrated in a 5000 Da MWCO concentrator.

Nuclear magnetic resonance spectroscopy

All data for NMR experiments were collected on a 600 MHz Bruker Avance II spectrometer equipped with a TXI room temperature 5 mm probe with z‐axis pulse field gradient coils. NMR samples were either collected at 10°C (for Ig 57) or 10–37°C (for all other samples) in 20 mM Tris pH 7.5, 20 mM NaCl, 0.35 mM NaN₃, and 0.3–1.0 mM protein with 10% D₂O. For Ig57, we collected a 2D HSQC and standard ¹⁵N‐edited triple resonance experiments including HNCACB, CBCA(CO)NH, HNCO, HN(CA)CO, C(CO)NH, HCCCONH, ¹⁵N‐edited TOCSY, ¹⁵N‐edited NOESY, and ¹³C‐edited NOESY, in as previously described.39, 48 For other constructs, we collected 3D HNCACB, CBCA(CO)NH, HNCO, and HN(CA)CO data along with 2D HSQCs. Most experiments were collected with 128, 64, and 1024 points in the T1, T2, and T3 dimensions, respectively. NMR data were processed with NMRPipe,49 extended in the indirect dimension via linear prediction, and the resulting spectra were analyzed via Sparky.50 In all samples, all visible HSQC backbone shifts were assigned. Chemical shifts for the obscurin Ig57 domain have been deposited in BMRB under the accession number 30514. Ig34, Ig35, and Ig36 chemical shift assignments were kindly provided by Dr Ayako Nomura (Riken Structural Biology Laboratory, Japan).

Structure calculation

Interproton distance constraints were derived from 3D NOESY experiments (¹⁵N‐edited and ¹³C‐edited 3D NOESY) as described previously.38 Dihedral constraints ψ ± 20° and φ ± 15° for α‐helix and ψ ± 40° and φ ± 40° for β‐sheet were included based on TALOS+ and the chemical shift index of ¹Hα and ¹³Cα atoms.51, 52 Structural calculations were performed as described in References [38, 39]. Out of 200 structures, the final 20 were selected based on lowest Q‐values and lowest RMSD from the average, and were of high quality based on the statistical criteria listed in Table S3. The overall backbone RMSD of ordered heavy atoms is 0.609 Å. The coordinates of the human obscurin Ig57 structure have been deposited in the Protein Data Bank 6MG9.

Residual dipolar coupling

Anisotropic IPAP experiments for RDC determination were performed using the same conditions as for the HSQC with the exception of using a stretched polyacrylamide gel.48, 53 The gel was prepared using 4% acrylamide, and soaked with buffer prior to soaking with protein. RDC values were calculated using PALES software.54

Small angle X‐ray scattering

Different concentrations (1.0, 3.0, and 5.0 mg/mL) of various obscurin samples were prepared in the NMR buffer. SAXS data were collected at the 12‐id‐B beamlines of the Advanced Photon Source (Lemont, IL) as previously described.55 Guinier plots were created using Origin, and the radii of gyration of the protein constructs were calculated with the Guinier approximation.56 MultiFoXS was used to analyze the fit of SAXS and RDC data together, as well as to back‐calculate the conformation that best fit the SAXS data.57

Molecular dynamics

All MD simulations were performed using the YASARA 12.4.1 software package, the Amber 03 force field, and explicit solvent (with 150 mM NaCl) in a box that extended 5 Å beyond the length of the extended construct at 37°C, and described in reference 39. All simulations were run for at least 50 ns in triplicate.

All steered molecular dynamics simulations were performed using the PMEMD module of the Amber 12 MD software package, using AMBERff12SB force field and in explicit solvent.44, 58, 59, 60 For equilibrium simulations, a constant temperature of 300 K was imposed using a Langevin thermostat with a collision frequency of 1 ps⁻¹. A constant velocity of 1.0 Å/ns (0.1 m/s) was used in order to simulate biologically relevant pulling forces.61 The SMD spring constant (rk2) was set to 0.2 and the temperature used was 310.0 K. Analysis was visualized using Gnuplot.58

Obscurin modeling

A mathematical model for obscurin was created using a 4 nm rod for each domain and nine degrees of freedom between each domain (135°, 180°, 225° in the x‐, y‐, and z‐directions, along with diagonals). Rods are connected at random in one of the nine degrees of freedom. The total distance calculated is measure from the first rod to the final rod. The model was implemented using MATLAB.62 The WLC formula

$$F\approx \frac{k_{B}T}{L_p}\left(\frac{1}{4(1-r/{L_{c)}}^2}-\frac{1}{4}+\frac{r}{L_c}\right)$$

was used, where k _B is the Boltzman constant, T is the temperature in Kelvin, L _P is the persistence length calculated in MATLAB, r is the distance between the N and C termini of our model, and L _c is the fully extended chain (the contour length).63

Statement for broader audience

The N‐terminus of obscurin is comprised of dozens of Ig‐like domains arranged consecutively. These tandem domains give obscurin a rope‐like or chain‐like architecture, permitting the protein to connect distal targets within cells. Here, we characterize obscurin's behavior in solution. We find that tandem domains are almost always elongated, yet mobile relative to each other. We propose this happens due to flexible domain/linker interactions between neighboring domains, which both prevent self‐association and contribute to passive stretch resistance.

Supporting information

Table S1 The linker between each human obscurin Ig domain

Table S2: PDB accession numbers of solution structure human obscurin Ig‐like domains (from CAC44768)

Table S3: NMR‐derived restraints and statistics of 20 NMR structures of wild‐type Ig57¹

Figure S1: NMR examples of Ig57 experiments. (A) NOESY/TOCSY overlay of A87 in Ig57. (B) Backbone walking experiments used to assign Ig57. (C) Secondary structure diagram of Ig57 with visible NOE correlations.

Figure S2: Guinier and RDC analysis. (A) Zoom in of Guinier regions of Ig34/35 with labeled R _g at two concentrations: 1 mg/mL (black) and 3 mg/mL (red). B) RDC experimental data on Ig58/59. (C) Using PALES, The fit of each individual domain and the dual domain of Ig58/59 to experimental RDC data (REF).

Figure S3: Guinier plot and Guinier region zoom in for each dual domain system at two concentrations. (A) Ig35/36 at 1 mg/mL (black) and 3 mg/mL (red). (B) Ig57/58 at 1.2 mg/mL (black) and 1.4 mg/mL (red). (C) Ig58/59 at 2.5 mg/mL (black) and 3.0 mg/mL (red).

Figure S4: Comparison of MultiFoXS one state and two state fits to experimental SAXS data of each dual domain system at two concentrations. (A) Ig34/35 at 3 mg/mL. (B) Ig35/36 at 3 mg/mL. (C) Ig35/36 at 5 mg/mL. (D) Ig57/58 at 1.2 mg/mL. (E) Ig57/58 at 1.4 mg/mL. (F) Ig58/59 at 2.5 mg/mL. (G) Ig58/59 at 3 mg/mL.

Figure S5: Angle between the domains of Ig58/59 calculated from MD over a period of ~800 ns.

Figure S6: SMD simulations for each dual domain system. (A) Domain angle versus time graph of Ig34/35 (top), work versus time graph of Ig 34/35 (middle), and distance between residues of likely interactions versus time of Ig 34/35 (bottom) for three different SMD simulations. (B) SMD data on Ig35/36, following the same organization as (A). (C) SMD data for Ig57/58. (D) SMD data for Ig58/59.

Figure S7: Modeled average length of obscurin. (A) The average obscurin length as a function of domain angle (in degrees). The number of links = (number of domains −1) for each model. (B) The average obscurin length as a function of link number. The plots represent what angle the linkers can bend. These models were generated in the XYZ dimensions.