Single Molecule Force Spectroscopy on Titin Implicates Immunoglobulin Domain Stability as a Cardiac Disease Mechanism*

Background: A mutation found in titin has been linked to arrhythmogenic cardiomyopathy (AC). Results: The mutation increases Ig10 instability and susceptibility to degradation. Conclusion: The mutation compromises the local structure of Ig10 and has a significant effect on Ig10 unfolding dynamics. Significance: Titin is the first sarcomeric protein to be implicated in AC pathology; a novel titin-based disease mechanism is suggested. Titin plays crucial roles in sarcomere organization and cardiac elasticity by acting as an intrasarcomeric molecular spring. A mutation in the tenth Ig-like domain of titin's spring region is associated with arrhythmogenic cardiomyopathy, a disease characterized by ventricular arrhythmias leading to cardiac arrest and sudden death. Titin is the first sarcomeric protein linked to arrhythmogenic cardiomyopathy. To characterize the disease mechanism, we have used atomic force microscopy to directly measure the effects that the disease-linked point mutation (T16I) has on the mechanical and kinetic stability of Ig10 at the single molecule level. The mutation decreases the force needed to unfold Ig10 and increases its rate of unfolding 4-fold. We also found that T16I Ig10 is more prone to degradation, presumably due to compromised local protein structure. Overall, the disease-linked mutation weakens the structural integrity of titin's Ig10 domain and suggests an Ig domain disease mechanism.

Titin plays crucial roles in sarcomere organization and cardiac elasticity by acting as an intrasarcomeric molecular spring. A mutation in the tenth Ig-like domain of titin's spring region is associated with arrhythmogenic cardiomyopathy, a disease characterized by ventricular arrhythmias leading to cardiac arrest and sudden death. Titin is the first sarcomeric protein linked to arrhythmogenic cardiomyopathy. To characterize the disease mechanism, we have used atomic force microscopy to directly measure the effects that the disease-linked point mutation (T16I) has on the mechanical and kinetic stability of Ig10 at the single molecule level. The mutation decreases the force needed to unfold Ig10 and increases its rate of unfolding 4-fold. We also found that T16I Ig10 is more prone to degradation, presumably due to compromised local protein structure. Overall, the disease-linked mutation weakens the structural integrity of titin's Ig10 domain and suggests an Ig domain disease mechanism.
Titin is a giant filamentous protein that spans the entire half sarcomere from the Z-disk to the M-band and is responsible for passive elasticity and the structural integrity of cardiac muscle. Titin is also involved in various cell signaling pathways (1), and mutations in titin have been implicated in numerous diseases, including dilated and hypertrophic cardiomyopathies (2,3). Recently, a mutation in titin was found in a family affected with arrhythmogenic cardiomyopathy (AC) 2 (4). AC is a primary heart muscle disorder characterized by breakdown of healthy cardiac myocytes and replacement with fibrofatty tissue and has been linked with various desmosomal mutations (5)(6)(7)(8)(9). AC is associated with severe ventricular arrhythmias and is the leading cause of sudden cardiac death in people under 35 years old (10).
Titin is connected near the Z-disk to actin-based thin filaments and in the A-band to myosin-based thick filaments and also contains an elastic region that is stretched during diastole when the heart ventricles fill with blood. The coupling of titin extension and ventricular stiffness has been extensively studied, and changes in titin-based stiffness accompany diastolic dysfunction (11)(12)(13). Titin's extensible I-band region is not bound to actin or myosin and contains three distinct springlike elements (the PEVK element, N2B element, and tandem immunoglobulin (Ig)-like domains (14)) that bear force during sarcomere stretch. The recently discovered AC-linked titin mutation (4) is found in the tenth Ig domain (Ig10) of titin's I-band region. It is the first I-band Ig mutation to be linked to cardiac disease. Ig domains have a stable ␤-barrel conformation (15) and are serially connected by short peptide linkers (16,17). When sarcomeres stretch during diastole, the tandem Ig segments in titin's I-band are the first spring element to become taut (18), although they are thought to remain folded under physiological forces (Ͻ5 pN/molecule) and may act as a force buffer under extreme loads.
This study investigates if the AC-linked point mutation affects the structure and kinetics of Ig10 using atomic force microscopy (AFM) and biochemical assays. With AFM, global unfolding of a protein can be monitored, and direct measurements of refolding rates and force-dependent unfolding rates can be made, both of which are crucial for understanding protein dynamics. Biochemical assays allow the study of protein degradation under various proteolytic conditions. The diseaselinked mutation is a C3 T nucleotide transition found in exon 37 of the titin gene (4). Exon 37 encodes for the tenth Ig domain in the proximal tandem Ig segment of the I-band, and the mutation changes the native threonine residue into isoleucine. This mutation is referred to as T16I (threonine is the 16th residue from the N terminus of Ig10). From homology modeling and secondary structure prediction (19,20), Ig10 is expected to form an Ig-I ␤-barrel structure (15), with the native Thr residue located between the AЈ and B ␤-strands (Fig. 1). Previous single molecule force spectroscopy and molecular dynamics studies have shown that residues in the AЈB loop are crucial for determining the structural stability of Ig domains via non-covalent interactions with residues in the G ␤-strand (21)(22), with AЈ mutations either increasing or decreasing the force needed to unfold the domain (23). Although many studies have investigated Ig domain unfolding (23)(24)(25)(26), this is the first time a naturally occurring Ig mutation has been studied using AFM and the first time an Ig domain in titin's elastic I-band region has been linked to cardiac dysfunction.
We hypothesize that the T16I mutation alters the conformation of Ig10 and increases its propensity to exist in a non-native, unfolded state. T16I Ig10 may then be susceptible to protease cleavage, which would completely abolish titin's force generation mechanism and would probably lead to pathological remodeling (27). A critical first step toward elucidating the relationship between mutated Ig10 and arrhythmogenic cardiomyopathies is determination of the functional effect of the T16I mutation at the single molecule level.

EXPERIMENTAL PROCEDURES
Protein Engineering-Two Ig10 pentamers (five identical Ig10 domains linked in series) were created: Ig10 wild type and Ig10 mutant. The mutant 5-mer contains the single AC-linked amino acid mutation (T16I) in all five Ig domains. To best mimic the native rotational constraints on Ig10, the linker sequence between Ig10 domains in the WT and T16I pentamers was chosen such that each Ig10 domain was flanked by the native residue at each terminus (Pro at the C terminus and Glu at the N terminus). This resulted in an Ig10(PE(Ig10)) 4 construct. We also expressed a naturally occurring fragment containing seven Ig domains (Ig7-13) and a mutant version of this heteropolyprotein in which the only difference is the T16I mutation in Ig10. Synthetic DNA coding for five human Ig10 domains connected in series and Ig7-13 was purchased from Geneart (Regensburg, Germany), inserted into vectors, and expressed in Escherichia coli. All constructs were inserted into pETM11 vectors except for Ig7-13 WT, which was inserted in FIGURE 1. A, schematic of a cardiac sarcomere. A single titin molecule spans from the Z-disk to the M-band and contains a large elastic I-band region consisting of tandem Ig domains, the N2B element, and PEVK sequence. B, schematic of part of titin's I-band region. The AC-linked point mutation is in the tenth proximal Ig domain. Note that the 15 proximal and 22 distal Ig domains of the N2B titin isoform are not all shown. C, Ig10 tertiary structure predicted from homology modeling (19). D, alignment of predicted Ig10 secondary structure (20) with solved titin Ig domains shows that the T16I mutation resides in the AЈB peptide loop (40). Residues highlighted in blue have ␤-strand secondary structure (predicted in Ig10). pET9D for technical reasons. Protein expression and purification were performed using standard methods (28).
Single Molecule Force Spectroscopy-The mechanical and kinetic properties of purified Ig10 5-mers were probed with an MFP3D atomic force microscope from Asylum Research (Santa Barbara, CA). Freshly thawed recombinant protein in buffer (25 mM BES, 2.5 mM EGTA, 1.5 mM MgCl 2 , 1.25 mM NaATP, 165 mM KCl, pH 7.0) was spotted on gold-coated microscope slides at ϳ5-10 g/ml and incubated at 4°C for at least 1 h. The Ig10 5-mers were engineered with two C-terminal cysteines to allow thiol-gold bonds to anchor the protein to the slide surface. Few functional groups besides thiols bind strongly to gold (29), so exploiting the thiol-gold bond encourages the protein's short C-terminal extension to attach to the slide surface while discouraging nonspecific binding between the globular Ig domains and gold slide. Unbound molecules were removed from the slide through pipette rinses with buffer. Surface protein density was kept low to minimize protein-protein interactions and the probability of multiple molecules attaching to the atomic force microscope tip simultaneously.
With one end of the polyprotein attached to the slide surface, a piezo-controlled silicon nitride cantilever tip (MLCT, Veeco Probes, Plainview, NY) was driven toward the protein-coated surface from above and nonspecifically adsorbed protein. The end-to-end distance of the Ig10 5-mer was constrained between the slide and the cantilever tip, and the molecule was stretched along this axis as the cantilever retracted from the slide surface at 1000 nm/s. This constant velocity stretching generates a force versus displacement curve that characterizes the elastic and mechanical properties of the tethered protein. The displacement of the cantilever base was determined directly via an integrated linear voltage differential transformer. To determine the force exerted on the molecule as the tip was pulled away from the surface, Hooke's law (F ϭ Ϫkx) was used, where F(x) is the force needed to extend the molecule a given distance, x is the distance away from equilibrium the cantilever is bent, and k is the spring constant of the cantilever. Cantilever stiffness was established by measuring its mean thermally driven vertical displacement (x) and then applying the equipartition theorem k͗x 2 ͘ ϭ k B T, where k B is Boltzmann's constant and T is absolute temperature. Stiffness is typically 20 -25 pN/nm. To accurately measure the force-displacement relationship of the Ig pentamer, the end-to-end length of the tethered molecule was corrected for cantilever bending.
As a tethered Ig pentamer is stretched, force develops in the molecule until a single Ig10 unfolding event occurs. When an Ig domain unravels from a native, compact state to a denatured, extended state, the contour length of the entire 5-mer is increased by ϳ28 nm (the diameter of a folded Ig domain is ϳ5 nm, and the backbone contour length of the unfolded Ig10 polypeptide is 87 amino acids ϫ 0.38 nm/amino acid ϭ 33 nm). This sudden increase in contour length following a high force unfolding event results in an identifiable "force peak." This process of polyprotein extension and unfolding continues until all serially linked Ig domains have unfolded or until the protein is displaced from the cantilever tip. Because the thiol-gold bond approaches covalent strengths (ϳ44 kcal/mol) (29), the molecule is much more likely to be displaced from the cantilever tip than from the slide surface. These successive unfolding events generate a characteristic sawtooth pattern; this "molecular fingerprint" is used to determine when a single Ig10 5-mer is being stretched. If more than five low force peaks are present or if consecutive force peaks are not separated by ϳ28 nm periodicity, the tip has attached to multiple pentamers at once; these data are not analyzed.
Stretch-relaxation-restretch was also implemented to directly measure the refolding rate of WT and T16I Ig10. A standard force-extension trace was overridden by a secondary trigger that initialized a user-defined drive wave that controls piezoelectric position. For the stretch-relaxationrestretch protocol, the number of unfolded domains that were allowed to refold was limited by the molecular extension during the low force hold. For example, in Fig. 6, the molecule was relaxed to the middle of the first Ig10 unfolding peak at ϳ25 nm extension (the first blue peak is adhesion between the cantilever tip and slide surface). Although there are four unfolding peaks in the initial extension, only three domains can possibly refold during the hold time (which is at an extension of ϳ25-30 nm) because the fourth refolding event would require ϳ50 pN of bend in the cantilever tip (i.e. the force needed to extend four folded domains to 25-30 nm during the initial stretch). The energy required to bend a cantilever tip (k ϭ 20 pN/nm) by this amount is 63 pN⅐nm, which is 15 times larger than k B T; therefore, the fourth unfolded Ig domain is prohibited from refolding.
Force clamp AFM protocols were also performed to measure force-dependent unfolding. The force clamp protocol uses a feedback mechanism to maintain tension in a tethered 5-mer following Ig unfolding. For each unfolding event, we measured the time interval ⌬t i between the start of the force clamp and the unfolding event. It takes tens of milliseconds for the feedback to reestablish the force clamp, and this time was subtracted from the raw time data to correct for the amount of time that the 5-mer is not held at the clamp force. For two-step unfolding, ⌬t i was taken from the start of the force clamp to the time of the second unfolding transition. Because actual hold forces varied slightly between tethered molecules, we analyzed data in 5 pN bins (i.e. data from clamps between 53 and 57 pN were analyzed together and considered as 55 pN; root mean square force noise is ϳ7 pN).
Removing Unfolding Force Peak Dependence-Although the Ig10 pentamers contain identical domains, average unfolding force increases with peak number (Table 1). To increase the robustness of fits between simulated and experimental data, we removed this peak dependence and thus reduced the spread in our data. Order statistics theory (30) was implemented to generate a parent distribution of unfolding force (describing a single isolated domain) from data gathered from unfolding of Ig10 pentamers, where ⌿ represents the parent cumulative distribution function (cdf) and is the cumulative distribution function of the lth unfolding force peaks of a homopolyprotein of length n.
Monte Carlo Simulations-Monte Carlo methods were used to simulate AFM force-extension traces. Starting from five seri-ally linked Ig domains with a defined unfolding rate at zero force (␣ 0 ) and distance along the reaction coordinate from the folded state to the transition state (⌬x u ), the molecule was stretched at 1000 nm/s (mimicking AFM experiments). At each time point, the force acting on the 5-mer was determined from force balance between cantilever bend (set at 22 pN/nm) and the force required to stretch the polypeptide (the combination of unfolded and folded domains) using the wormlike chain equation (31). The probability of unfolding is p ϭ j⅐␣(F)⅐⌬t, where j is the number of folded domains, ␣(F) ϭ ␣ 0 exp(F⅐⌬x u / k B T) is the force-dependent unfolding rate constant, and ⌬t is the time step. A contour length increase of 28 nm was imposed for each unfolding event, and the persistence length of the polypeptide chain was set to 0.4 nm.
Unfolding Rate Error Estimation-Bootstrap sampling was used to determine uncertainty in the zero force unfolding rate for WT and T16I Ig10. 10,000 bootstrap samples were generated by resampling with replacement from our AFM data. Resampling was performed for each unfolding force peak (i.e. peak i values of a bootstrap data set were taken from peak i experimental forces), and an average parent cdf (with peak dependence removed) was generated for each bootstrap sample. This cdf was then compared with Monte Carlo simulation cdfs to determine the unfolding rate that best describes the bootstrap cdf. The S.D. of the 10,000 bootstrap unfolding rates is reported as the error in unfolding rate for WT and T16I Ig10. The distance to the transition barrier (⌬x u ) was fixed at 0.3 nm for bootstrap sampling.
Maximum Likelihood Estimation-All data acquired from the AFM refolding protocol were analyzed together using maximum likelihood estimation (MLE) to determine the refolding rate of WT and T16I Ig10. Although we have evidence for two-step refolding, this process was observed too rarely to generate a data set large enough for analysis. Therefore, partial refolding was considered a 0.5 refolding event for MLE. Domain refolding was modeled as a Bernoulli trial with a success probability of 1 Ϫ e Ϫ␤⅐t . The likelihood function we want to maximize is as follows, where ␤ is the refolding rate, x and t are the measured observables, and the likelihood is the product of all possible refolding events. The vectors x and t are length N (number of domains able to refold), with x i ϭ 1 if refolding occurs and 0 otherwise, and t i equal to the user-defined hold time. To determine uncertainty in estimated refolding rate, we generated 10,000 bootstrap samples from our experimental data. We then performed MLE on each of the new data sets (to obtain a best fit ␤Ј for each bootstrap sample) and took the S.D. value of all 10,000 ␤Ј to determine the error in our refolding rate estimate for our experimental data set.
Proteolysis Assays and Mass Spectrometry-Purified trypsin from bovine pancreas was purchased from Sigma. Freshly dissolved trypsin (50 mM Tris, pH 7.6) was incubated with Ig10 5-mer at 37°C until the reaction was stopped with reducing buffer. Proteins were separated with SDS-PAGE and Coomassie-stained. Orbitrap Velos (Thermo Fisher) with an Advion Nanomate (Ithaca, NY) source operated in data-dependent MS/MS mode was used for determining N-terminal peptides of protein degradation products. High resolution mass measurement of peptides was performed by matrix-assisted laser desorption/ionization (MALDI). The MALDI experiments were performed on a Bruker Ultraflex III MALDI TOF-TOF instrument using sinapinic acid as the matrix.
Physiological protease degradation assays were performed on Ig7-13 fragments (with or without T16I mutation in Ig10). Fresh mouse hearts were blended in 800 l of 1ϫ PBS (either pH 5 or 8) ϩ 1.5 mM CaCl 2 . The resulting pH of the heart extract was ϳ6.3 and ϳ7.4, as determined by a micro-pH electrode (Lazar Research Laboratories, Los Angeles, CA). The heart extract was then added to WT and mutant Ig7-13. This mixture was incubated at 37°C, and a portion of the mixture was removed at various time points and spun down at 13,000 rpm for 2 min. The resulting supernatants were then solubilized and separated using SDS-PAGE.
In Vivo Simulation-Titin-based passive tension (in mN/mm 2 ) as a function of sarcomere length (SL) was determined from muscle mechanics data (32), and the force experienced per titin molecule was calculated from thick filament density and a stoichiometry of six titin molecules per half thick filament (33). The working SL in mice is between 1.8 and 2.2 m (34), and we assumed a sinusoidal relationship between SL and time, with SL varying between 1.8 and 2.2 m at 1 Hz (10 Hz showed similar results). From this, we estimated Ͻ5 pN of force acting on each titin molecule at a diastolic SL of 2.2 m. Force-dependent unfolding was taken as ␣(F) ϭ ␣ 0 exp(F⅐⌬x u / k B T) because the force-dependent unfolding rates we measured experimentally were for force levels much greater than the physiological range.

RESULTS
Ig10 Unfolding-To determine how the T16I mutation affects the mechanical and kinetic stability of Ig10, mutant and wild type Ig10 5-mers were stretched and unfolded at a pulling speed of 1000 nm/s. Representative force-extension traces for WT and T16I Ig10 are shown in Fig. 2B. Each of the first five force peaks represents the force needed to unfold an Ig10 domain. Ig10 unfolding results in a contour length increase, which removes the tension from the system and results in the large force drop immediately following a force peak. The last force peak is due to the fully unfolded Ig10 5-mer polypeptide displacing from the cantilever tip. Although each of the first five force peaks is due to unfolding of a single Ig10 domain, unfolding force is peak-dependent, with average unfolding force increasing with peak number (Table 1).
It should also be noted that AFM experiments on WT and mutant Ig7-13 proteins did not show different unfolding forces (data not shown). However, this result was expected because Ig domain unfolding is a stochastic process, and the average unfolding force for each peak in the sawtooth pattern probably contains the unfolding forces of Ig7, Ig8, Ig9, etc., thus masking any real effects.
Ig domain unfolding rate can be described as a force-dependent process by ␣(F) ϭ ␣ 0 exp(F⅐⌬x u /k B T), where ␣ 0 is the unfolding rate constant at zero force, F is the external applied force, ⌬x u is the unfolding distance along the reaction coordi-nate, and k B T is thermal energy at room temperature. A Monte Carlo approach was used to simulate AFM unfolding experiments over hundreds of (⌬x u , ␣ 0 ) pairs. Comparing the experimental unfolding force distribution with the simulated force distribution allows estimation of ␣ 0 and ⌬x u for WT and T16I Ig10. Because unfolding force is peak-dependent, the cdfs of unfolding forces are also peak-dependent, but this peak dependence was removed according to order statistics theory (see "Experimental Procedures"). As a result, the cdf describing the first unfolding event in a collection of force-extension curves now overlaps with the cdf describing the last unfolding event (which, by definition, is the parent cdf; Fig. 3). The transformed cdfs (one for each peak) were then averaged to generate a master parent cdf for either WT or T16I Ig10. This method is preferred because data from all unfolding force peaks are analyzed simultaneously, increasing robustness of fits. Also, it is incorrect to pool unfolding forces from different unfolding force peaks because each peak unfolding force distribution describes a different process (e.g. the force needed to unfold one domain when five folded domains are connected in series is different, on average, from the force needed to unfold one domain when 10 folded domains are connected in series). The same transformations were used on Monte Carlo simulations to generate a master parent cdf for each (⌬x u , ␣ 0 ) pair. The theoretical and experimental cdfs were compared empirically using the Kolmogorov-Smirnov test (35) to find the (⌬x u , ␣ 0 ) pair used in the Monte Carlo simulations that generated the master parent cdf most consistent with the experimental cdf for the WT and T16I Ig10 5-mer simulations. WT Ig10 was best described by ⌬x u ϭ 0.300 Ϯ 0.025 nm and ␣ 0 ϭ 0.004 Ϯ 0.0005 s Ϫ1 , and T16I Ig10 was best described by ⌬x u ϭ 0.300 Ϯ 0.025 nm and ␣ 0 ϭ 0.018 Ϯ 0.0029 s Ϫ1 (Ϯbootstrap S.D.). This analysis shows that the distance to the transition barrier along the molecular end-to-end length reaction coordinate is unchanged by the AC-linked mutation, whereas the unfolding rate at zero force is increased ϳ4-fold.
The previous method gives the unfolding rate at zero force, but we were also interested in force-dependent unfolding rates. Therefore, we probed the effect of the disease-linked mutation on the ability of Ig10 to resist unfolding at different external force levels. Using AFM feedback, we were able to hold tethered proteins at constant force to directly determine force-dependent unfolding rates. For this protocol, we stretched a tethered protein until a set force level developed and then held the protein at that force. This force is held until an Ig domain unfolds, an event that immediately increases the contour length of the molecule and removes the tension from the system. After the Ig rupture, however, the cantilever tip quickly moves further away from the surface until the set force level is reestablished. The molecule is then held at that force until another Ig domain unfolds and the process continues. This method complements standard AFM data by producing information in the time domain (e.g. hold times prior to Ig unfolding) instead of the force domain. Sample force clamp data for WT Ig10 are shown in Fig. 4A. Plotting extension versus time results in a step pattern that has constant extension increases (⌬Ext) following Ig unfolding; this pattern also acts as the molecular fingerprint of the Ig10 5-mer. Clamp forces were between 50 and 80 pN, and molecular hold time was 5 s. Force clamp and hold time values were chosen to minimize drift effects and increase the chances of seeing unfolding events prior to the molecule detaching from the cantilever tip. Fig. 4B shows the ⌬Ext histograms for WT and T16I Ig10 for all force clamp experiments. Each histogram clearly shows a well defined peak distribution in the addition to a small collection of low ⌬Ext data points. The Gaussian fit to the main peak is centered at 23.8 nm for the WT distribution and 23.0 for T16I. Although the contour length of the polypeptide increases by ϳ28 nm following domain unfolding, the extension increase is less because of polymer coiling. From the wormlike chain equation, a contour length gain of 28 nm would only result in an extension increase of ϳ22 nm at a hold force of 60 pN (average clamp, 62.1  pN for WT and 60.6 pN for T16I), which is similar to the Gaussian fit peaks in Fig. 4B. The short extension increases that lie outside of the Gaussian fit represent two-step unfolding (i.e. a short extension increase was always partnered with a second, complementary short increase, and these two extensions summed to the length of a singular unfolding event (i.e. a 7-nm increase immediately followed by a 16-nm increase instead of a single 23-nm increase)). It is worth noting that two-step unfold-ing was seen 10% of the time in WT Ig10 (23 of 221 unfolded domains) but only 2% of the time in the T16I mutant (5 of 204). Because the tethered polyproteins are under constant force in the AFM force clamp protocol, we were able to determine the force-dependent unfolding rates of WT and T16I Ig10. Fig. 5 shows the empirical cumulative distribution functions of the times at which WT and T16I Ig10 unfold while under 65 pN of external force (step pattern), with the best fit to 1 Ϫ e Ϫ␣(F)⅐t FIGURE 3. cdfs of unfolding forces for WT and T16I Ig10 5-mers. For each force value, the y axis value indicates the percentage of domains that unfolded at or below that force. The blue traces are the cdfs of peaks 1-4. The red traces are the cdfs of peaks 1-4 after transforming to the parent cdf using order statistics (30). The cdf of peak 5 is also in red (because we are stretching a 5-mer, by definition the cdf of the last unfolding peak forces is the parent cdf). The cyan traces are the average of the five parent cdf (red) traces. These averaged parent cdfs were used to compare experimental data with simulated data (after simulated data were similarly processed). This analysis reduces the spread in unfolding force data and allows all peaks to be analyzed together, which increases the sample size and improves fitting accuracy. FIGURE 4. AFM force clamp protocol. A, tethered proteins were held at constant force for 5 s. After an unfolding event (arrows), the cantilever quickly retracted from the surface until the tension level was restored, resulting in a stepwise pattern. The extension difference (⌬Ext) between steps represents the extension increase following an unfolding event. The hold time prior to a domain unfolding is measured from when the residual tension is initially reached (start of the purple trace). For example, the second Ig domain takes t ϭ ⌬t 1 ϩ ⌬t 2 to unfold. B, histograms of extension increases. The histograms represent the compilation of all force clamp data. The red Gaussian trace is a best fit to the histogram. The smooth distribution is centered at 23.3 nm for WT Ig10 and 23.0 nm for T16I. The data that do not fit the Gaussian distribution represent the individual segments of two-step unfolding events. Two-step unfolding is 5 times more prevalent in WT Ig10.

Ig10 Mutation and Domain Instability
overlaid (R 2 ϳ0.99). Note that the collection of unfolding times is well described by a single exponential, which strongly suggests that each domain in the 5-mer unfolds independently. The unfolding rate at 65 pN was found to be 1.135 s Ϫ1 for WT Ig10 and 6.865 s Ϫ1 for T16I Ig10, a 5-fold increase. The inset of Fig. 5 shows force-dependent unfolding rates at five clamp forces.
Ig10 Refolding-The unfolding rate is important to determine the susceptibility of a protein to exist in a mechanically denatured state, but the refolding rate is equally important. To determine how the disease-linked mutation affects the refolding rate of Ig10, we implemented a refolding protocol using AFM. Tethered molecules were stretched until most domains had unfolded, a user-defined drive wave was executed that relaxed the tethered molecule and allowed domain refolding events to occur for a set amount of time, and then the molecule was restretched (Fig. 6). From the force-extension trace, we can determine how many domains refolded by counting the number of force peaks during the reextension. This method allows for direct determination of the refolding rate. In the example shown, note that the first unfolding event of the second stretch is an abbreviated force peak, which indicates that a partial refolding event took place when the molecule was relaxed. This is consistent with our force clamp data and previous work (21,36) that shows that unfolding (and possibly refolding) can be a two-step process. Partial refolding events were seen infrequently but were more prevalent in the WT proteins. Combining data for all dwell times (0 -3 s), 8% of unfolded WT domains partially refolded (17 of 221), whereas only 2% of all unfolded T16I domains partially refolded (3 of 158) (recall that 10 and 2% of WT and T16I, respectively, showed two-step unfolding). Refolding probability as a function of low force dwell time is shown in Table 2. To analyze data from all dwell times together, we employed MLE to determine the effective refolding rate for WT and T16I Ig10. Assuming that domains refold independently, the probability of a domain refolding under no external force is 1 Ϫ e Ϫ␤⅐t , where ␤ is the refolding rate at zero force.
Using MLE with all of the refolding data results in ␤ ϭ 0.70 Ϯ 0.17 s Ϫ1 for WT Ig10 and ␤ ϭ 0.56 Ϯ 0.17 s Ϫ1 for T16I Ig10 (Ϯbootstrap S.D.). This suggests that the T16I mutation does not alter the refolding rate of Ig10.
Ig10 Degradation-The change in Ig10 folding kinetics is unlikely to have a relevant physiological effect in terms of titinbased passive tension. When considering the entire elastic I-band of titin's N2B isoform (the primary cardiac isoform in mice and humans (37,38)), the unfolding of a single Ig domain would change tension by Ͻ1% between SLs of 2.0 and 2.3 m. Instead, we hypothesize that unfolded Ig10 is prone to proteolysis, which would completely abolish titin's force generation mechanism and possibly lead to accelerated titin breakdown. To test if the T16I mutation affects the structural integrity of Ig10 in a way that leads to increased Ig10 proteolysis, we carried out trypsin digestion assays. Trypsin preferentially cleaves at the carboxyl side of lysine and arginine, and cleavage patterns can indicate the degree to which amino acids are structurally constrained or solvent-accessible and prone to protease cleavage.
Using a 1:30 (w/w) trypsin/Ig10 5-mer ratio in digestion buffer (50 mM Tris⅐HCl, pH 7.6), we tested the susceptibility of WT and T16I Ig10 to proteolysis. Incubation time was varied, and all reactions were performed at 37°C. Results are shown in Fig. 7. Although the only difference between WT and T16I Ig10 5-mers is the single point mutation in each Ig domain, the digestion assays clearly show that the mutant protein is more susceptible to trypsin cleavage, which implies that the mutation alters the Ig structure in such a way that trypsin can more easily perform peptide cleavage. To investigate whether the mutation is the direct cause of proteolysis, we performed MS on digested T16I Ig10. We excised the largest degradation product (arrowhead in Fig. 7A) from the gel, performed in-gel digestion with chymotrypsin, and identified N-terminal peptides. This analysis identified the N-terminal peptide to be DIPTTENLY (Fig.  7C). To be more confident that peptides at the N-terminal side of DIPTTENLY are not present in the degradation product (e.g. a peptide may be present but not seen by MS), we used a high charge/mass MS/MS selection method on the degradation product as well as the full-length (undegraded) protein. This technique clearly identified the presence of peptides containing the His tag at the N terminus of the full-length protein but zero peptides for the His tag in the degradation product. This strengthens our initial findings that the DIPTTENLY peptide found is indeed the N terminus of the primary degradation band. Next, we performed a high resolution molecular weight measurement of the Ig10 digestion products using MALDI. We found that the molecular mass of the largest degradation product is 44,380 Da. Knowing the N-terminal sequence of this protein band and its molecular mass, we were able to determine the C-terminal sequence (i.e. we could determine the cleavage site on the Ig10 5-mer that results in this digestion product). Starting from the N-terminal sequence found by Orbitrap and continuing toward the C terminus of the T16I 5-mer, the closest molecular mass match ends with EVPEIK in the fifth serially linked Ig10 (Fig. 7C). This lysine residue (Lys-17) is therefore predicted to be the last residue in the digestion product, which means that the mutant Ig10 5-mer is cleaved one residue away FIGURE 5. Force-dependent unfolding rates. From the collection of unfolding times acquired at a given residual tension, we are able to fit the cumulative distribution function to 1 Ϫ e Ϫ␣(F)⅐t , where ␣(F) is the unfolding rate at a given force. The empirical cdfs shown were generated from force clamp values between 63 and 67 pN (average ϳ65 pN). The smooth fits yield ␣(F ϭ 65) ϭ 1.135 s Ϫ1 for WT Ig10 and ␣(F ϭ 65) ϭ 6.865 s Ϫ1 for T16I Ig10. Inset, force-dependent unfolding rates for WT and T16I Ig10. Error bars, S.E. FEBRUARY 22, 2013 • VOLUME 288 • NUMBER 8 from where the T16I mutation is located. This shows that the disease-linked mutation alters the structural integrity of Ig10 in such a way that susceptibility to proteolysis is increased and that proteolysis occurs where T16I disrupts the Ig10 structure. These data complement our AFM data, which show that T16I Ig10 unfolds at lower force and has a higher unfolding rate compared with wild type Ig10.

Ig10 Mutation and Domain Instability
Although the trypsin degradation experiment and subsequent MS analysis allowed us to determine that T16I affects the structure of Ig10 near the mutation, we also wanted to see if T16I increased Ig10 degradation in the presence of physiologically relevant proteases. To do this, we incubated Ig7-13 (with and without T16I in Ig10) with heart extracts from mice to expose WT and T16I Ig10 to all of the proteases naturally found in the heart while being flanked by native Ig domains (Ig7-9 at the N terminus and Ig11-13 at the C terminus). At various time points, a portion of the Ig/heart extract mixture was removed and centrifuged, and the supernatant was solubilized. This pro-tocol was performed at low pH to mimic intracellular pH after ischemia (pH ϳ6.5 in rats (39)) and also at (relatively) high pH. The solubilized supernatants were separated by SDS-PAGE and Coomassie-stained (Fig. 8A). By measuring the optical density of the Ig7-13 band as a function of degradation time, we could determine if the T16I mutation affects the rate of titin degradation in the presence of cardiac proteases. We quantified the effect of T16I on Ig7-13 degradation by determining the half-life of full-length Ig7-13 in the heart extract mixture (Fig.  8B). The half-life of WT Ig7-13 was 21.2 h at pH 7.38 and 25.2 h at pH 6.33. The time it took for half of the mutant Ig7-13 to degrade was drastically less, about 6 h at both pH values. This indicates that the T16I mutation in Ig10 significantly increases the degradation rate of the Ig tandem in the presence of physiological proteases.

DISCUSSION
A recently discovered titin mutation linked with AC (4) is found in the tenth Ig domain of titin's proximal tandem Ig segment. We investigated how this single point mutation changes the mechanical and kinetic stability of Ig10. The AC-linked C39453T single nucleotide polymorphism in the titin gene (4) changes the native threonine (polar, hydrophilic) residue into isoleucine (nonpolar, hydrophobic). This change in hydrophobicity, in addition to the other biochemical changes associated with amino acid mutation, may play a role in disrupting Ig10 structure. The native threonine is the 16th amino acid in Ig10 and is located between the AЈ and B ␤-strands (Fig. 1) as determined by Clustal alignment with solved Ig structures (40) and secondary structure prediction of the Ig10 primary sequence (20). The residues in the AЈB loop of Ig domains are important FIGURE 6. AFM refolding protocol. A tethered molecule was unfolded and then relaxed to allow for domain refolding. Restretching the molecule and identification of overlapping force peaks allows determination of how many unfolded domains refolded while the molecule was relaxed. The blue trace is the initial stretch. After extension and force triggers were reached, the cantilever tip was driven back toward the surface (low force portion of the orange trace, from 110 to 25 nm) to release tension in the molecule. After holding for a set period of time, the molecule was restretched fully (orange trace from 25 to 140 nm).

TABLE 2 Refolding percentages as a function of hold time
The number of domains that could have refolded is given in parentheses. As expected, more domains refold during longer holds. Because the data are relatively noisy, maximum likelihood estimation was used to analyze all refolding data simultaneously to extract the refolding rate at zero force. Two partial refolding events were considered equivalent to one complete refolding event. NA, not applicable.

Ig10 Mutation and Domain Instability
for defining and maintaining mechanical stability (21,23,36), and this is supported by the fact that Thr-16 is highly conserved across species, from humans to zebrafish (4). We directly tested whether the T16I mutation alters the mechanical stability and folding kinetics of Ig10 using AFM.
Force-extension curves generated from stretching recombinant Ig10 5-mers show that the T16I mutation causes Ig10 to unfold from its native, globular state into an extended, random coil polypeptide at a lower external force. The decrease in mechanical stability of Ig10 is more relevant, considering that native Ig10 is a relatively unstable domain to begin with. Previous work has shown that other Ig domains in titin's I-band unfold at forces much higher than WT Ig10 (e.g. Ig91-98 unfolds at ϳ225 pN at 1000 nm/s pulling velocity (26); Ig27, ϳ200 pN at 600 -800 nm/s (23)) and have a much lower unfolding rate (e.g. ␣ 0 ϭ 6.1 ϫ 10 Ϫ5 s Ϫ1 for Ig91-98, ␣ 0 ϭ 1.0 ϫ 10 Ϫ5 s Ϫ1 for Ig65-70 (26); ␣ 0 ϭ 2.8 ϫ 10 Ϫ5 s Ϫ1 for Ig28 (41)). This means that, even prior to mutation, Ig10 is more likely to occupy a non-native, extended conformation. The mutation further increases the instability of Ig10 relative to other domains, which supports the hypothesis that the mutation may be physiologically relevant. For example, if the mutation simply lowered the propensity of Ig10 for unfolding, but T16I Ig10 behaved similarly to other Ig domains, then the mutation would probably not have a significant deleterious effect.
The force clamp data showed that two-step unfolding is five times more frequent in WT Ig10, which was somewhat surprising and deserving of comment. It has been shown that the AЈ ␤-strand residues of Ig domains strongly influence their ability to withstand mechanical stress via hydrogen bond networks with the G ␤-strand (21,23), and the location of the T16I mutation suggests disruption of this important native contact network. These terminal ␤-strands may also locally unfold and extend prior to complete molecular denaturing (36), resulting in short extension increases before the hydrophobic core falls apart. Our data suggest that T16I weakens the firm AЈG con- The only difference between WT and T16I 5-mer is the Thr 3 Ile mutation in each Ig domain. Mass spectrometry found that the primary degradation product of T16I begins with DIPTTENLY in the N-terminal linker sequence and ends at Lys-17 in the last Ig domain. FIGURE 8. A, degradation assay with recombinant Ig7-13 (with and without T16I in Ig10) and heart extract. Ig7-13 is exposed to all of the proteases naturally found in the heart. The mutant Ig7-13 protein has a larger N-terminal linker sequence than its wild type counterpart and has mobility similar to that of a heart extract protein. These overlapping bands were fit with a double Gaussian to isolate the intensity due to full-length Ig7-13 mutant. This experiment was performed at pH ϳ6.3 and ϳ7.4, but no pH dependence was found. B, Ig7-13 with the T16I Ig10 domain degrades much more rapidly than WT Ig7-13 at both pH values, which indicates increased degradation rate. pH values used were 6.33 Ϯ 0.06 (S.E.) and 7.38 Ϯ 0.03 (S.E.); n ϭ 3 for each bar. ***, p Ͻ 0.001 between WT and mutant using a two-way analysis of variance and Bonferroni post hoc test. FEBRUARY 22, 2013 • VOLUME 288 • NUMBER 8 tacts, such that ␤-strands sliding past each other may not always be a distinct structural transition. Instead, external force does not build up in the AЈG strand (e.g. strand stiffening (42)) but instead is transmitted more directly to the stable protein core. Subsequently, a force peak does not develop when the AЈG ␤-strands separate, and only a one-step unfolding event is recorded.

Ig10 Mutation and Domain Instability
Although arrhythmogenic cardiomyopathy is associated with a broad phenotypic spectrum (for a recent review, see Ref. 43), the most prominent pathological change is myocyte loss, which is often accompanied by fibrofatty replacement or inflammation. The disease progression is poorly understood, although it is clear that myocyte degradation is a hallmark feature (44). Our hypothesis is that the AC-linked mutation decreases the force needed to unfold Ig10 (or, similarly, increases the Ig10 unfolding rate) and increases proteolytic susceptibility. Protease cleavage of sarcomeric proteins has been linked to cardiac pathology previously, including myosin light chain (45) and myosin-binding protein-C (46). Peptide bond breakage anywhere within titin's elastic I-band would abolish titin's force generation mechanism and would probably lead to severe cardiac dysfunction. It is unclear how the cell responds to I-band cleavage, but accelerated titin turnover and/or apoptosis are possible. This myocyte loss may be the first step of AC development; therefore, we wanted to test if the T16I mutation changes the degradation proclivity of Ig10.
The trypsin digestion assays show that mutant Ig10 is more prone to proteolysis, and mass spectrometry confirmed that the cleavage site is in the AЈB loop where the mutation is located. This result is noteworthy because it shows that T16I Ig10 is more prone to degradation even under zero external force. Regardless, proteolysis would occur more readily if the protein exists in a fully unfolded state. To this point, it is worthwhile to determine to what degree the T16I mutation increases the amount of time that Ig10 is globally unfolded. From our force clamp and refolding AFM data, we determined the global unfolding and refolding rate of WT Ig10 to be ␣ 0 ϭ 0.004 s Ϫ1 and ␤ 0 ϭ 0.70 s Ϫ1 , respectively. These parameters for T16I Ig10 are ␣ 0 ϭ 0.018 s Ϫ1 and ␤ 0 ϭ 0.56 s Ϫ1 . In order to estimate the fraction of Ig10 domains unfolded in vivo, we performed computer simulations (see "Experimental Procedures"). Our simulations predict that 1.8% of WT Ig10 domains and 9.5% of T16I Ig10 domains are unfolded at any time (i.e. the T16I mutation increases the probability that Ig10 is unfolded in a beating ventricle by a factor of 5). In addition, identical analysis using the unfolding and refolding rates of other titin Ig domains (23,26) predicts that Յ0.1% of Ig27, Ig91-98, or Ig65-70 is unfolded at any time, a 100-fold decrease compared with T16I Ig10 (Fig. 9).
We also determined the free energy of unfolding for WT and T16I Ig10. To do this, we generated recombinant WT and T16I Ig10 monomers and measured the amount of protein that is unfolded under no force using pulse proteolysis (47). Briefly, Ig10 monomers were prepared in varying concentrations of denaturant (urea) and allowed to reach conformational equilibrium. Then excess protease (thermolysin) was added to digest the fraction of Ig10 monomers that were unfolded at the given urea concentration. This method quantifies the percentage of WT and T16I Ig10 monomers that are unfolded under different urea concentrations, and it also allows estimation of the Gibbs free energy of unfolding. Fig. 10A shows the fraction of WT Ig10 monomers that remain folded as a function of urea concentration as determined by pulse proteolysis. The fraction of folded protein at urea concentrations in the transition zone (where folded and unfolded populations are both substantial) was extrapolated to determine the folded fraction at 0 M urea (Fig. 10B), which gives the free energy of unfolding in water and is related to the relative population of unfolded and folded WT Ig10 using Equation 4, where [U] and [F] represent the concentration of unfolded and folded protein, respectively. We estimated that the free energy of unfolding for WT Ig10 is 2.59 Ϯ 0.4 kcal/mol, which corresponds to ϳ1% of WT monomers being unfolded at any given time at room temperature (consistent with our simulations (Fig. 9)). This exact method did not work for the T16I Ig10 monomer, however, because the mutant monomer was strongly degraded by thermolysin even in the absence of any urea. Nonetheless, we were able to determine the amount of T16I Ig10 that was degraded at 0 M urea and estimated the free energy of unfolding for T16I Ig10 to be 0.156 Ϯ 0.2 kcal/mol, which corresponds to 40 -50% of T16I monomers being FIGURE 9. In vivo unfolding simulations. The force experienced by titin's elastic I-band region in the beating heart was simulated and used to estimate the percentage of Ig domains that are unfolded in vivo. WT Ig10 is unfolded an order of magnitude more often than other Ig domains, which shows that Ig10 is less stable than other Ig domains even without the T16I mutation. T16I Ig10 is unfolded 2 orders of magnitude more than other Ig domains. Note the breaks in the y axis.

Ig10 Mutation and Domain Instability
unfolded at any given time. This result is consistent with previous NMR work that showed that T16I Ig10 co-exists as both a well structured and less structured protein in solution (4). The fraction of unfolded T16I monomers estimated by pulse proteolysis is more than that in our in vivo simulations, and this is probably because our simulations predict global unfolding, whereas the pulse proteolysis protocol does not discriminate between local and global unfolding (i.e. a flexible peptide loop may be cleaved even if the protein core is still intact). Nonetheless, these data once again show that the AC-linked point mutation greatly affects the conformational stability of Ig10 and increases its proteolytic susceptibility. We repeated this pulse proteolysis protocol with our Ig7-13 constructs and found that the T16I mutation significantly decreased the ability of Ig7-13 to remain folded under chemical denaturation. The urea concentration at which half of Ig7-13 was degraded was 4.89 Ϯ 0.08 M (S.E.) for WT Ig7-13 and 4.54 Ϯ 0.06 M (S.E.) for mutant Ig7-13 (p ϭ 0.008; data not shown). This suggests that Ig10 unfolds more easily due to the T16I mutation even when interdomain Ig contacts are conserved. The trypsin digestion assays show that T16I Ig10 is more prone to degradation, but it was important to find if Ig10 structure is compromised at the location of the T16I mutation. From MS analysis, we found the N-terminal sequence of the primary Ig10 5-mer degradation product (arrowhead in Fig. 7A) to be in the N-terminal linker sequence that was added to the Ig10 5-mer after synthetic Ig10 DNA was inserted into the pETM11 vector (which also added a His tag for protein purification). We also used MALDI to determine the molecular mass of all Ig10 digestion products and found that the primary degradation product is 44,380 Da. Starting from the N-terminal sequence found by MS, the molecular mass measurement matches the T16I 5-mer primary sequence at the lysine residue immediately following the T16I residue in the fifth Ig10 domain. The peptide D . . . K (demarcated by arrows in Fig. 7C) that results from tryptic cleavage has a predicted mass of 44,307 Da. This predicted mass is slightly less than the measured value, but MS showed that various residues within Ig10 can be modified (mostly oxidation, ϩ16 Da). Using MALDI, we found the molecular masses of the full-length 5-mer and three of the four most prominent degradation bands to be 54,171, 44,380, 34,425, and 24,087 Da. The molecular mass difference between these bands is 9791, 9955, and 10,338 Da, respectively. The predicted molecular mass of the peptide sequence between the Lys-17 (the C-terminal cleavage site in the primary digestion band) residues of adjacent serially linked Ig10 is 10,018 Da, which suggests that each of these protein bands is formed from cleavage of the full T16I 5-mer at the location of the mutation in one of the tandem Ig10 domains, resulting in one Ig10 molecular mass difference between bands. One of the four most prominent degradation bands has a molecular mass of 27,105 Da, which is not ϳ10 kDa different from the other degradation products. This band may be due to additional protein degradation following cleavage at Lys-17 because any peptide bond break probably compromises the entire ␤-barrel structure. The observed digestion band pattern is much more pronounced in the mutant Ig10 5-mer (Fig. 7A), which shows that increased proteolysis results from the AC-linked mutation, and MS showed that the cleavage site is within the amino acid loop that harbors T16I.
To provide evidence for physiologically relevant degradation, we studied if the disease-linked mutation affects Ig10 degradation when Ig10 is flanked by native Ig domains and when physiological proteases are present. We incubated our Ig7-13 construct with blended mouse heart to expose Ig7-13 to all of the proteases in the heart. Ig7-13 was degraded much more rapidly in the mutant construct (Fig.  8B), presumably due to the T16I mutation in Ig10. Similar to our pulse proteolysis data, this result shows that T16I increases titin degradation even in the absence of force. More work is needed to determine the proteolytic pathway(s) responsible for this degradation.
With the recent discovery of the T16I Ig10 mutation found in patients with arrhythmogenic cardiomyopathy overlap syndromes, we set out to determine the effects that the mutation had on the single molecule dynamics of Ig10. We found that the mutation lowered the force needed to unfold Ig10, increased its unfolding rate, and increased its susceptibility to proteolysis. These single molecule studies were a necessary first step in the ongoing investigation of the connection between titin and AC. Future studies will focus on the effect of the T16I mutation on titin breakdown in the myocardium and pathological changes of diseased hearts. Elucidating the mechanistic pathway that connects a single mutation in Ig10 to cardiac remodeling, ventricular arrhythmias, and sudden cardiac death is crucial for improving diagnosis, developing treatments, and preventing fatality.