Different Molecular Mechanics Displayed by Titin s C onstitutively and Differentially Expressed Tandem Ig Segments

Journal of Structural Biology 137, 248 258 (2002) doi:10.1006/jsbi.2002.4458 Different Molecular Mechanics Displayed by Titin s C onstitutively and Differentially Expressed Tandem Ig Segments Kaori Watanabe,* Claudia Muhle-Goll,, Miklós S. Z. Kellermayer,*, Siegfried Labeit, and Henk Granzier*,1 *VCAPP, Washington State University, Pullman, Washington 99164-6520; European Molecular Biology Laboratory, Heidelberg, Germany; Anesthesiology/Intensive Operative Medicine, U. Hospital Mannheim, Mannheim, 68135, Germany; and Biophysics, University of Pécs Medical School, Pécs H-7624, Hungary Received December 5, 2001, and in revised form February 11, 2002 Titin is a giant elastic protein responsible for passive force generated by the stretched striated-muscle sarcomere. Passive force develops in titin s extensible region which consists of the PEVK segment in series with tandemly arranged immunoglobulin (Ig)-like domains. Here we studied the mechanics of tandem Ig segments from the differentially spliced (I65 70) and constitutive (I91 98) regions by using an atomic force microscope specialized for stretching single molecules. The mechanical stability of I65 70 domains was found to be different from that of I91 98 domains. In the range of stretch rates studied (0.05 1.00 m/s) lower average domain unfolding forces for I65 70 were associated with a weaker stretch-rate dependence of the unfolding force, suggesting that the differences in the mechanical stabilities of the segments derive from differences in the zero force unfolding rate (K 0 u ) and the characteristic distance (location of the barrier) along the unfolding reaction coordinate ( X u ). No effect of calcium was found on unfolding forces and persistence length of unfolded domains. To explore the structural basis of the differences in mechanical stabilities of the two fragment types, we compared the amino acid sequence of I65 70 domains with that of I91 98 domains and by using homology modeling analyzed how sequence variations may affect folding free energies. Simulations suggest that differences in domain stability are unlikely to be caused by variation in the number of hydrogen bonds between the force-bearing -strands at the domain s N- and C-termini. Rather, they may be due to differences in hydrophobic contacts and strand orientations. 2002 Elsevier Science (USA) 1 To whom correspondence should be addressed. Fax: (509) 335-4650. E-mail: granzier@wsunix.wsu.edu. INTRODUCTION Titin is the third myofilament type of vertebrate striated muscle, with a single molecule spanning the half sarcomere from the Z-line to the M-line (for recent reviews with original citations, see Wang, 1996; Maruyama, 1997; Gregorio et al., 1999; Trinick and Tskhovrebova, 1999). Passive tension is generated by titin s extensible region, found in the I- band of the sarcomere and comprises serially linked and mechanically distinct spring elements. One of the spring elements is formed by a region rich in proline (P), glutamate (E), valine (V), and lysine residues (K), referred to as the PEVK element (Labeit and Kolmerer, 1995). The PEVK structure is not well established, but may consist of polyproline II helices interspersed with random coil structures (Ma et al., 2001). The PEVK functions as a highly flexible entropic spring that determines the passive force levels at physiological sarcomere lengths (Linke et al., 1998; Trombitas et al., 1998). The PEVK segment is flanked by spring elements of different mechanical character: the tandem immunoglobulin (Ig) segments (a proximal segment toward the Z-line and a distal segment toward the A-band). These springs consist of serially linked Ig-like domains, which are modules with a 7-strand -barrel structure (Improta et al., 1996). The tandem Ig segments extend at low forces (for a review with original citations, see Linke and Granzier, 1998), presumably via unbending of domain linkers, and their function may be that of a molecular leash. Titin is encoded by a single gene from which mrnas of up to 100 kb are transcribed (Freiburg et al., 2000; Bang et al., 2001). Differential splicing cascades adjust the number and length of the expressed spring elements (Freiburg et al., 2000). As a result, the tandem Ig segments consist of a subset of constitutively expressed domains (the N-terminal 15 1047-8477/02 $35.00 2002 Elsevier Science (USA) All rights reserved. 248

TITIN MECHANICS IN TANDEM Ig SEGMENTS 249 Ig domains of the proximal tandem Ig segment plus all of the domains of the distal segment) and a differentially spliced region that contains 0 to 53 domains (0 in N2B cardiac and 53 in soleus muscle titin). It is controversial whether unfolding of Ig domains contributes significantly to titin s extensibility under physiological conditions. When a continuous stretch is imposed on slack sarcomeres, tandem Ig segments extend rapidly to a near constant length. Further extension is halted despite the generation of high levels of passive force (in this regime PEVK extension dominates). Because the constant length of the tandem Ig in highly stretched sarcomeres can be accommodated by folded Ig domains, it has been suggested that Ig unfolding may not take place under physiological conditions (Trombitas et al., 1998). However, this conclusion is at odds with single-molecule work that has revealed that Ig domain unfolding is a probabilistic event that, given enough time, may occur even at physiological levels of force (Kellermayer et al., 1997; Rief et al., 1997; Tskhovrebova et al., 1997). In a recent report (Minajeva et al., 2001) it was suggested that a limited number of Ig domains may unfold in the stretched sarcomeres, leading to passive-force relaxation. Further work is required to more completely understand the role of Ig domains in muscle physiology. Here we studied the mechanical behavior of recombinant tandem Ig fragments from the differentially spliced (I65 70) 2 and constitutive (I91 98) regions by using an atomic force microscope (AFM) specialized for stretching of single molecules. Singlemolecule techniques have been successfully employed to study mechanically driven protein unfolding. Engineered polyproteins containing multiple repeats of the same Ig domain have been elegantly used to study the mechanical stability of the particular domain (for a review with original citations, see Carrion-Vazquez et al., 2000). The mechanical behavior of domains in such polyproteins can be different from that when the domains are coupled to neighboring domains of the titin molecule (Li et al., 2000b). Therefore, to mimic the in situ arrangement, we used in our study multidomain fragments as found in titin. Our findings indicate that the mechanical stability of domains in I65 70 is different from those in I91 98: within the range of stretch rates explored (0.05 1.00 m/s), lower mean unfolding forces and a lower stretch-rate dependence of unfolding force were observed for I65 70. Monte 2 We use the domain numbering system of Bang et al. (2001) throughout this paper, but make references to the previous nomenclature of Labeit and Kolmerer (1995) wherever necessary. Carlo simulations suggested that differences in both the characteristic distance (location of barrier) along the unfolding reaction coordinate ( X u ) and the rate constant of unfolding at zero force (K u 0 ) underlie the different mechanical stabilities of the fragments. To explore the structural basis of the mechanical differences, we compared the amino acid sequence of the I65 70 domains with that of I91 98 domains and using homology modeling studied how sequence variations may affect domain-folding energies. Finally, we also determined the persistence length of unfolded Ig domains and studied whether unfolding forces are dependent on calcium concentration. Protein Fragments METHODS Recombinant titin fragments (Fig. 1) were cloned into modified pet vectors, expressed as soluble proteins in BL21[DE3]pLysS cells, and purified on Ni-NTA columns with a His-Bind purification kit (Novagen, Madison, WI). Purified proteins were dialyzed into AB buffer (25 mm imidazole-hcl, ph 7.4, 0.2 M KCI, 4 mm MgCl 2, 1 mm EGTA, 0.01% NaN 3, 5 mm DTT), quick frozen, and stored at 80 C until later use. To characterize the resolution of our instrument under single-molecule mechanical experimental conditions we stretched spectrin which is known to display lowforce structural transitions (Rief et al., 1999). Spectrin extracted from human erythrocytes was purchased from Sigma Chemical Co. (St. Louis, MO). In order to split the heterodimer, spectrin was incubated for an hour at ph 11 in 20 mm Tris buffer. (For more detailed methods, see Rief et al., 1999.) The proteins were then used for stretching experiments. Mechanical Manipulation of Titin Fragments Molecules were mechanically stretched with a Molecular Force Probe (MFP, Asylum Research, Santa Barbara, CA), an atomic force microscope specialized for stretching molecules. The MFP was mounted on a custom-built, low-profile inverted light microscope constructed of damped mechanical posts (Thorlabs) attached to an antivibration table (TMC). Protein fragments (at a concentration of 200 g/ml for Ig fragments and 150 g/ml for spectrin) were allowed to bind to a precleaned glass microscope slide for 20 min in case of Ig fragments and for 10 min in case of spectrin. Unbound molecules were washed away with AB assay buffer. In experiments at high calcium, slides were washed multiple times with AB buffer containing 2 mm total calcium (resulting in a free calcium concentration of 1 mm or pca 3.0). Molecules were stretched in AB solution by first pressing the cantilever (gold-coated microlevers, ThermoMicroscopes, MSCT AUHW) against the titin-coated surface and then pulling the cantilever away with a predetermined rate. Force vs displacement curves were collected in repeated stretch and release cycles. The displacement of the cantilever base was measured by using an integrated linear voltage differential transformer (LVDT; Asylum Research, Santa Barbara, CA). Force (F) was obtained from cantilever bending ( d) as F K d, where K is cantilever stiffness. Stiffness was obtained for each unloaded cantilever by measuring its mean thermally driven vertical bending and applying the equipartition theorem; K d 2 k B T, where k B is Boltzmann s constant and T is absolute temperature. Typical cantilever stiffness was 50 pn/nm. Force vs displacement curves were corrected for several factors to obtain force vs molecular end-to-end length: (1) The zerolength, zero-force datapoint was obtained from the force response

250 WATANABE ET AL. that corresponded to the cantilever tip reaching (or departing from) the substrate surface. (2) Forces were corrected for baseline slope obtained from the force response of the displaced but unloaded cantilever. (3) The end-to-end length (z) of the tethered molecule was calculated by correcting the cantilever base displacement (s) with cantilever bending as z s F/K. As a result of cantilever bending during force development, imposing a constant velocity stretch on the cantilever base will result in a fluctuating velocity (due to sawtooth-like force trace) of the cantilever tip. In experiments in which we studied the velocity dependence of unfolding force (as in Fig. 3D) plotted velocities are the average velocities of the cantilever tip during the rising phase of the force peaks. Analysis of Force Data Force data were compared with the worm-like chain (WLC) equation: FA k B T z L 1 4 1 z/l 2 1 4 (Bustamante et al., 1994). The WLC model describes the molecule as a deformable continuum of persistence length A (measure of bending rigidity) and contour length L. To deduce the rate constant of unfolding at zero external force (K 0 u ) and the location of unfolding barrier, or the width of the unfolding potential along the unfolding reaction coordinate( X u ), we modeled the stretch-rate dependence of unfolding force with a Monte Carlo simulation based on previously used simulation algorithms (Kellermayer et al., 1997). Briefly, the mechanical behavior was simulated by superimposing domain unfolding/refolding kinetics on the worm-like chain force-extension curve. To calculate WLC force we considered the molecule as a serially linked chain containing segments with folded domains that each make a 4 nm contribution to the contour length and unfolded domains that add to the contour length 28.7 nm for domains in I91 98 and 31.9 for I65 70 domains (see Results). The persistence length of unfolded domains was set at 0.6 nm (see below). The force-dependent unfolding rate constant is given by K u 0 e Ea FXu /kbt ; 0 is the attempt frequency set by Brownian dynamics of the chain (10 8 s 1 ), E a is the unfolding activation energy. The width of the unfolding potential ( X u ) and K u at zero force (K 0 u ) were user adjustable. From repeated Monte Carlo simulation trials we computed the average unfolding force at different stretch rates. By fitting the calculated stretch-rate dependence to that measured (as in Fig. 3D), we obtained estimates of K 0 u and X u (for further details, see Kellermayer et al., 1997). Structural Modeling The sequence of skeletal muscle titin was derived from the EMBL data library Accession Number AC X90569. Multiple sequence alignments were obtained with ClustalW (Thompson et al., 1994) and corrected by visual inspection. Residues were highlighted according to their degree of conservation as in the default option of ClustalX (Thompson et al., 1997). When we searched for differences in the sequences that could affect domain stability we mainly focused on residues that might alter the structure and not on residues pointing to the outside of the domain. We employed homology modeling for structure prediction. Six known Ig structures of the I set (Harpaz and Chothia, 1994) served as templates. Three Ig domains were from titin: II ([Protein Data Bank (PDB) ID 1G1C], M5 [PDB ID 1NCT], and I27 (I91 in the new nomenclature) [PDB ID 1TIT]; two Ig domains were from twitchin (Caenorhabditis elegans): Ig 18 [PDB ID 1WIT] and the Ig domain attached to the kinase [PDB ID 1KOA] (1) from twitchin; and telokin [PDB ID 1FHG]. The modeling procedure was carried out with the MODELLER package (Sali and Blundell, 1993) as previously described (Muhle-Goll et al., 2001). Briefly, folding energies were computed with the program Fold-X, which uses a free energy function described by three terms: a statistic atomic contact potential of desolvation, that estimates the effective energy difference between atom-atom and atomsolvent contacts, a constant term for each buried hydrogen bond, and a local entropy term (for further details, see Guerois and Serrano, 2000). To analyze the impact of specific side chains on domain stability, we constructed structural models in which residues were replaced and compared their free energies of folding relative to those of the native structure. RESULTS To understand the functional significance of differential expression of Ig domains in titin s extensible region we studied the mechanical characteristics of the 6-domain Ig fragment I65 70 and the 8-domain I91 98 fragment, which are part of the differentially and constitutively expressed tandem Ig regions, respectively (Fig. 1). To explore the mechanical behavior and unfolding of the fragments we used an atomic force microscope (AFM) specialized for stretching single molecules (Fig. 2A). The sensitivity and stability of the instrument were tested by stretching spectrin, which is known to exhibit unfolding forces just detectable with custom-designed AFM technology (Rief et al., 1999). An example of a stretch-release curve of spectrin is shown in Fig. 2B. For comparison, a stretch-release curve for the I91 98 segment is shown in Fig 2C. Both curves display sawtooth-like force peaks, indicating that repetitive structural transitions occur during stretch. Each sudden force drop represents the unfolding of an -helical spectrin domain (Fig. 2B) or a -barrel Ig domain (Fig. 2C). The rising traces of the force peaks are well fitted with the worm-like chain model (broken thin lines), giving persistence lengths of 0.5 nm for I91 98 (see also below) and 0.8 nm for spectrin. At a stretch rate of 0.5 m/s the mean unfolding force of spectrin domains was 25 pn and that of I91 98 was 225 pn (Fig. 2D), consistent with the reported low mechanical stability of spectrin (Rief et al., 1999) and high stability of Ig domains (Li et al., 2000b). Examples of I91 98 and I65 70 stretch curves (stretch rate 0.5 m/s) are shown in Figs. 3A and B. Both fragment types display well-resolved force peaks that are typically regularly spaced (see also below) with a periodicity of 30 nm. Since the number of amino acid residues in the domains of the differentially expressed Ig segment is greater than those of the constitutive segment (Witt et al., 1998), it is expected that the contour-length gain accompanying domain unfolding events is larger in case of I65 70 than in I91 98. The contour length of par-

TITIN MECHANICS IN TANDEM Ig SEGMENTS 251 FIG. 1. Top: location of poly-ig fragments I65 70 and I91 98 in the I-band sequence of human soleus titin (domain numbering based on Freiburg et al., 2000). Bottom: electron micrograph of human soleus titin labeled with anti-titin antibodies that demarcate proximal and distal tandem Ig segment (for details, see Trombitas et al., 1998). I65 I70 is part of differentially expressed titin whereas I91 98 is constitutively expressed. tially unfolded titin segments was calculated from the WLC fit to the uprising half of the force sawtooth (see broken black lines in Figs. 3A and 3B). The contour-length gain accompanying domain unfolding events is shown in Fig. 3C. The average contourlength gain in I65 70 is 31.9 nm, whereas that in I91 98 is 28.7 nm. The observation is consistent with the greater number of residues per domain in I65 70 than in I91 98 (94 vs 89 residues) and normalizing the contour-length gain by the average number of residues per domain results in a contourlength gain per residue that is similar for the two fragment types (see also Discussion). The stretch-rate dependence of unfolding force is shown in Fig. 3D. In the range of stretch rates used in this work (50 1000 nm/s), the mean domain unfolding forces are much higher in I91 98 than in I65 70, indicating that under the employed experimental conditions the I91 98 domains are mechanically more stable than the I65 70 domains. The domain unfolding forces in both segments increased linearly with the logarithm of stretch rate, but the stretch-rate dependence of unfolding force was much stronger in I91 98 than in I65 70 (Fig. 3D). To deduce the rate constant of unfolding at zero external force (K u 0 ) and the location of unfolding barrier, or the width of the unfolding potential along the unfolding reaction coordinate ( X u ), we modeled the stretch-rate dependence of unfolding force with a Monte Carlo simulation based on previously used simulation algorithms (see Methods). Simulation results (inset of Fig. 3D) indicate K u 0 6.1 10 5 and 1.0 10 5 s 1, and X u 0.25 and 0.35 nm for I91 98 and I65 70 domains, respectively. The results for the I91 98 fragment are similar to those obtained for native titin (K u 0 : 3.3 10 5 s 1, and X u 0.3 nm (Rief et al., 1998)) and intermediate between those of I91 and I92 (or I27 and I28) linked in a heteropolyprotein (K u 0 : 6 10 4 and 4.0 10 6 s 1 ; X u 0.25 and 0.25 nm for I91 and I92, respectively (Li et al., 2000b)). The larger X u in I65 70 indicates that the unfolding energy landscapes in I65 70 and I91 98 are different: the location of the domain unfolding barrier in I65 70 domains is farther along the reaction coordinate. As noted also by others (Rief et al., 1998; Li et al., 2000b), stretching the I91 98 fragment typically resulted in a steady rise in the level of the force peaks, reflecting the hierarchical unfolding of domains with increasing stabilities. Figure 4 compares the level of the force peak as a function of the peak number. We omitted from our analysis the first and last peak of the stretch curve (the first peak was often irregular and the force leading up to this peak often deviated from WLC behavior, whereas the last peak reflects the force required to dissociate the molecule from its attachment point). Our analysis indicates that the range of domain unfolding forces is greater in I91 98 than in I65 70 (Fig. 4). We also determined the persistence length of unfolded segments by analyzing the release curve of the stretch-release force data. Because force-extension curves of unfolded Ig domains deviate at high force from pure entropic elasticity and contain enthalpic contributions due to bond angle deformations (Rief et al., 1998), we fitted release forces of less than 100 pn with a WLC equation. The data were segregated into two groups: (a) data containing stretch curves with regularly spaced force sawteeth (as in Fig. 5A), and (b) data containing stretch

252 WATANABE ET AL. FIG. 2. (A) Schematic of force-measuring AFM. A molecule is attached between the tip of a cantilever and a glass microscope slide. Extension of the molecule by retraction of the piezoelectric positioner generates force in the molecule that deflects the cantilever. Cantilever deflection is measured from which the force generated in the molecule can be calculated following cantilever stiffness calibration. (B and C) Examples of force-extension curves obtained for spectrin (B) and I91 I98 (C). Stretch is the lighter trace (at end of stretch molecule in C detached). The stretch curve has a sawtooth-like appearance, with each sudden force drop reflecting a structural transition (unfolding) in the molecule. Rising phase of each sawtooth can be fitted (broken black curves) with the worm-like chain (WLC) equation. (D) Distributions of unfolding forces of spectrin and I91 98 (stretch rate 0.5 m/s) obtained by stretching a large number of molecules and determining for each stretch curve the average peak height. Forces required to unfold spectrin domains are much lower than those required to unfold Ig domains. curves with irregularly spaced force sawteeth (as in Fig. 5B). (Note that results shown in Figs. 1 4 were all obtained from curves with regularly spaced force peaks.) We found that the persistence-length distribution from data with regularly spaced force sawteeth had in both fragment types a major peak at 0.6 nm. The mean persistence lengths were 0.58 0.02 and 0.56 0.02 nm for I91 98 and I65 70, respectively. The distribution obtained from curves with irregular force peaks was shifted to shorter lengths and had a major peak at 0.3 0.4 nm. The mean values were 0.42 0.02 for both fragment types. It has been suggested that calcium ions bind to Ig domains (Lu et al., 1998). Furthermore, there are reports in the muscle literature that calcium influences the magnitude of passive force (see, for example, Bagni et al., 1994). We therefore determined the effect of calcium concentration on the average domain unfolding force in I91 98 and the persistence length of the unfolded I91 98 chain. No significant differences were found between pca ( log[ca 2 ]) 9.0 and 3.0, a range that covers the physiological calcium range (Table I). This suggests that the results of our study apply to passive as well as activate muscle, conditions with high and low pca, respectively. To explore the structural basis of differences between the mechanical stabilities of I65 70 and I91 98 domains, we studied how sequence variations in the fragments may influence folding energy. Schulten and colleagues (Lu et al., 1998; Lu and

TITIN MECHANICS IN TANDEM Ig SEGMENTS 253 FIG. 3. (A and B) Examples of force-extension curve measured during stretch of I91 98 (A) and I65 70 (B). Stretch rate 0.5 m/s. (C) Contour-length gain vs the sequential number of unfolding events in I91 98 (solid line) and I65 70 (broken line). The slope of I65 70 is significantly (P 0.05) larger than that of I91 98 (31.9 vs 28.7 nm). (D) Dependence of mean unfolding force (F) on stretch rate (v) for Ig 91 98 (closed circles) and Ig 65 70 (open circles). Unfolding forces of I65 70 are 30 50 pn lower than those of I91 98. Slopes and offset of the linear regression lines of two fragments are different. Results were simulated by using a Monte Carlo simulation and modeling unfolding/refolding as a two-state process with one folded and one unfolded state (cf. Kellermayer et al., 1997). The obtained unfolding rates at zero force (K 0 u ) and the width of the unfolding potential ( X u ) are shown in the inset. Schulten, 2000) performed steered molecular dynamics studies of I91 (I27; see footnote 2). They proposed that the initial event during the mechanical unfolding of Ig domains is the breakage of hydrogen bonds connecting strands A and B, followed by the rupture of hydrogen bonds connecting strands A and G. Based on this proposal, the lower average unfolding force in I65 70 domains (Fig. 3D) might result from a less extensive hydrogen-bond network in these domains. We tested this hypothesis by comparing structural models of the Ig domains comprising the I65 70 and I91 98 segments. For each Ig domain we constructed a structural model based on its sequence homology to six known structures of Ig domains of the I set (see Methods). When we compared the hydrogen bond network of the models, it became apparent that differentially expressed Ig domains (I65 70) contain two additional hydrogen bonds between strand A and B, due to the insertion of an extra residue at position 31. Using an engineered polyprotein of I91 Marszalek et al. (1999) provided

254 WATANABE ET AL. FIG. 4. Unfolding force of domains gradually increases during stretch. We measured the height of the individual force peaks (except for the first and last peak: see inset) at stretch velocities of 1000, 500, 200, and 100 nm/s for I91 98 (open circles and broken lines) and I65 70 (closed circles and solid lines). Linear regression lines of data (R 2 0.95) had larger slopes for I91 98 than for I65 70. evidence that the hydrogen bonds between A and B strands rupture prior to those between A and G strands, giving rise to a low-force hump ( 100 pn) during stretch of the polyprotein. Assuming that this conclusion is valid for other domains as well, the hump is predicted to appear at a higher force in I65 70 domains than in I91 98 domains. Rupturing of the A and B strands is considered to gives rise to an unfolding intermediate whereas rupturing hydrogen bonds between A and G strands is thought to underlie the main unfolding event that leads to the sudden force drop associated with the force sawtooth (Marszalek et al., 1999; Carrion- Vazquez et al., 2000). However, our models did not reveal differences in the number and length of hydrogen bonds between A and G strands of I65 70 and I91 98 domains. Thus, differences in hydrogen bond network do not appear to provide a ready explanation for the different mechanical behavior of I65 70 and I91 98 domains. It is possible that the mechanical stability of Ig domains depends not only on the number of hydrogen bonds but also on other types of noncovalent interactions. Therefore, we explored the effects of additional sequence differences between the domains. Multiple sequence alignments of the domains revealed that differences in the sequences of I65 70 and I91 98 occur mainly in the boxed areas indicated by arrows in Fig. 6, and we tested the role of these residues in domain stability. Structural models were constructed in which the residues of one Ig domain group were replaced by those of the other. For example, P2 of I65 was substituted for I2 derived from the proximal Ig sequences and vice versa. We computed the free energies of folding of the modified domains by using the program Fold-X (Guerois and Serrano, 2000) and compared them with those of the native domains. The analysis revealed that residues in boxes 1 3 may contribute to differences in the stability of I65 70 and I91 98. Replacing proline 2 in I65 70 domains with isoleucine (residue 2 of I91) decreased the free energy of folding of I65 I70 domains, indicating that the domains were now more stable. When the reverse mutation was introduced in I91 98 domains (Ile2 replaced with proline), the free energy of folding was increased. Since in the structure of I91 the side chain of Ile2 points into the interior of the domain, its replacement by a proline might weaken hydrophobic contacts of the N-terminus. A stabilizing effect was also found when residue 7 in I65 70 domains (box 2 in Fig. 6) was exchanged by proline (found in 7 of the I91 98 domains), albeit to a lesser extent than the above discussed box 1 replacements. Proline at position 7 influences the direction of the main chain and may strengthen thereby the hydrogen bonds of strands A and B. Box 3 residues are located on strand A. They participate both via their main chain in the hydrogen bonds network of strands A and G and via their side chains in the interactions of the hydrophobic core. In I91 98 positions 11 and 13 both tend to be occupied by large hydrophobic residues (11 out of 16 residues are valines). By contrast, in I65 70 domains there are many polar (thr, ser) or charged residues (glu) in positions 11 and 13. Calculating the free energies of folding of modified I65 70 domains indicated a stabilizing effect (reduction in free energy of folding) of having valine at both position 11 and position 13, whereas the reverse replacement destabilized I91 98. The destabilizing effect of box 3 residues in I65 70 could be explained by the fact that polar (thr, ser) or charged residues (glu) are present in the hydrophobic core or by that less hydrophobic interaction surface (ala) is provided, reducing the interaction surface. Box 4 residues enlarge the BC loop and add an additional hydrogen bond to I65 70 domains (see also above). Box 5 residues had no significant effect on the computed free energy of folding. Replacing the cysteines at the position of the classical cysteine bridge (see Fig. 6, boxes marked C) with hydrophobic residues had an impact on the computed stability of most domains. Residues like

TITIN MECHANICS IN TANDEM Ig SEGMENTS 255 FIG. 5. Examples of stretch-release curves of I91 98 with WLC curve fitted to release (black line). (A) Example of a stretch curve with regularly spaced ( 30 nm) force peaks. (B) Example of curve with irregularly spaced force peaks. The shown WLC fits had a persistence length of 0.60 nm (A) and 0.28 nm (B). See text for details. isoleucine, valine, and phenylalanine make several hydrophobic contacts in the interior of the domain. Because cysteines are small and polar, they may allow fewer and less stable hydrophobic contacts (we assume that covalent S S bridges are absent in the reducing environment encountered in muscle). Since five of the six Ig domains in I65 70 contain a cysteine pair (I70 has a threonine-cysteine pair), but only three of the eight Ig domains in I91 98, the high number of cysteine pairs in I65 70 domains could contribute to their lower stability. DISCUSSION TABLE I Effect of Calcium Concentration on Unfolding Force and Persistence Length of Unfolded Polypeptide (Stretch Rate 0.5 m/sec) Unfolding force Persistence length 234 4 pn (n 86) 0.58 0.02 nm (n 19) pca 3.0 223 4 pn (n 111) 0.58 0.04 nm (n 16) Note. Shown values are mean SE. Unfolding forces and persistence lengths in the presence of calcium do not significantly differ from the control. The extensible region of skeletal muscle titins consists of two main elements, the tandem Ig and PEVK segments. The PEVK segment functions as a flexible entropic spring with a persistence length that is likely to be 2 nm or less (Kellermayer et al., 1997; Linke et al., 1998; Trombitas et al., 1998; Li et al., 2001). Tandem Ig segments, on the other hand, function as entropic springs with a much longer persistence length. Based on immunoelectron microscopy of titin in soleus muscle, this persistence length has been estimated at 15 nm (Trombitas et al., 1998), a value consistent with light-scattering studies of titin in solution (Higuchi et al., 1993), and the recent analysis of the flexibility of isolated titin molecules by electron microscopy (Tskhovrebova and Trinick, 2001). Thus, the extensible region of skeletal muscle titin consists of two types of serially linked entropic spring: the flexible PEVK and the 10-fold stiffer tandem Ig segments. Due to differential splicing, both the PEVK and the tandem Ig elements vary in length in different muscle tissues (Freiburg et al., 2000). To better understand the functional significance of differential expression of Ig domains, we mechanically characterized Ig domains, at the single molecule level, and compared domains that are part of the constitutive region with those of the differentially expressed region. At stretch rates between 0.05 and 1 m/s the average unfolding force was lower in the differentially expressed region. This finding is consistent with results of Rief et al. (1998) who determined at a stretch rate of 0.5 m/s that unfolding force of a differentially expressed fragment was lower than that of a constitutive fragment. We used a wide range of stretch rates and this revealed that the stretch-rate dependence is weaker in the differentially expressed segment (Fig. 3D). The weaker stretch-rate dependence of unfolding force in I65 70 derives from a longer characteristic length along the unfolding reaction coordinate, X u ( X u determines the slope of the lines in Fig. 3D and K u 0 their y-offset, see also Carrion-Vazquez et al., 1999). As a result of

256 WATANABE ET AL. FIG. 6. Alignment of the human Ig repeat sequences I65 to I70 (upper panel) from the differentially expressed proximal tandem Ig segment and I91 98 (lower panel) from the constitutive distal tandem Ig segment. The sequences are from the previously described human titin sequences (Labeit and Kolmerer, 1995; EMBL data library Accession Number AC X90569). Residues are highlighted according to their degree of conservation as in the default option of ClustalX (Thompson et al., 1997). The color code is by conserved property in more than 60% of the residues in one column: blue, hydrophobic; red, positive; magenta, negative; green, hydrophilic; cyan, tyrosine or histidine; orange, glycine. Proline residues are colored in yellow, and cysteine residue in pink. Blue boxes I5 indicate main regions where differences occur in residues that might alter the structure. (We did not focus on residues that point to the outside of the domain as it is unlikely that they influence domain stability.) Boxes indicated by C highlight residues potentially involved in cysteine bridge formation. (Note domain numbering based on Freiburg et al. (2000). I65 70 and I91 98 correspond to I54 59 and I27 34 of Labeit and Kolmerer (1995). See text for additional details.) the different stretch-rate dependencies of unfolding forces, the linear regression lines of the two fragments (Fig. 3D) have a crossover point at a velocity of 0.3 nm/s. Thus, our results predict that at stretch rates lower than 0.3 nm/s, I65 70 domains require on average higher forces for unfolding than I91 98. This crossover behavior of I65 70 and I91 98 is similar to that of unfolding of wildtype and mutated Ig domains (Li et al., 2000a) and it further highlights that differences in mechanical behavior cannot be fully captured at a single stretch rate. To understand the structural basis for the different mechanical behavior of the domains we studied how sequence variations may impact the hydrogen bond patch between A and G strands, a region in titin s Ig domains that is viewed as critical for unfolding (Lu et al., 1998). Because all these bonds have to be ruptured simultaneously, they are considered determinants of the force level at which the domains unfold (Carrion-Vazquez et al., 2000). In a recent mutational analysis by Li et al. (2000a) valines at positions 11, 13, and 15 of the A strand of I91 were substituted with prolines, to disrupt hydrogen bonds formed with the G strand. Unfolding of mutant proteins occurred at forces lower than those in wild-type I91, and the stretch-rate dependence of unfolding force indicated that the rate constant for unfolding at zero force (K u 0 ) was reduced whereas the width of the unfolding potential ( X u ) was increased. Because these differences between mutant and wild-type proteins are similar to those observed by us for I65 70 and I91 98, we expected fewer hydrogen bonds between A and G strands of I65 70 domains. However, our analysis did not support this prediction, as we were unable to detect differences in the number of A G hydrogen bonds in any of the domains. Thus, the unfolding force may not just be a function of the number of hydrogen bonds between the A and the G strand, but other types of noncovalent interactions are likely to contribute as well. This conclusion is consistent with the mechanical results of a mutated I91 domain in which tyrosine at position 9 was replaced by proline (Li et al., 2000a). Even though this mutation eliminates a hydrogen bond between A and G strands, it actually increases the unfolding force. Our analysis revealed several areas with differences in structural residues that may impact noncovalent chain interactions. All involve residues of strand A or A or neighboring residues. Most notably, I65 70 domains have a proline at position 2 that is absent in I91 98 and that may negatively impact domain stability, via weakening of hydrophobic contacts between the domain s N terminus and the core. (A proline at position 2 is also present in 6 out of the 7 domains of a differentially expressed fragment with low unfolding forces (Rief et al. 1998).) Another area of

TITIN MECHANICS IN TANDEM Ig SEGMENTS 257 importance may be residue 7 which in most of the I91 98 domains (7 out of 8) is a proline, but in none of the I65 70 domains (only 2 out the total 53 differentially expressed domains contain a proline at position 7). Our analysis indicates that proline at position 7 may strengthen the hydrogen bonds of strand A by allowing a more optimal strand orientation. Residues of strand A that may influence stability are residues 11 and 13. Both of these residues are in I91 98 hydrophobic residues, whereas in I65 70 one of them is typically polar. Interestingly, this allows for an additional interpretation of the results of the mutational analysis by Li et al. (2000a). The observed destabilization of I91 domains, where valine 11 or valine 13 was mutated to proline, could be partially due to a weakening of hydrophobic contacts. In summary, our findings suggest that differences in structural residues of I65 70 and I91 98 domains impact hydrophobic contacts and angles between strands of their force bearing C- and N-termini, and this may contribute to differences in mechanical behavior of I65 70 and I91 98 domains. The average number of residues per Ig domain is 94 in I65 70 and 89 in I91 98 (Labeit and Kolmerer, 1995). The 5 extra residues in I65 70 are located near the B C and E F regions of these domains (Fig. 6). These residues are part of the hidden amino acids that are mechanically isolated by the hydrogen bonds that connect the A G strands, but that become exposed to force when unfolding takes place. (The remaining domain residues are force bearing and make up the interdomain linker sequence and the N- and C-terminal -strands, see also Carrion- Vazquez et al., 1999.) Thus, the additional 5 residues of I65 70 domains add to the unfolding-induced contour-length gain and will contribute to the larger contour-length gain in I65 70 than in I91 98 (31.9 vs 28.7 nm, see Fig. 3C). However, the 5 extra residues can only explain a 2 nm gain difference (an amino acid contributes 0.38 nm to the contour length of an unfolded polypeptide) whereas 3.2 nm was measured (31.9 28.7). The unexplained 1.2 nm suggests that folded I65 70 domains have a larger fraction of their residues (not counting the extra 5 residues) hidden. Fitting the release curves of stretch-release cycles with the WLC equation revealed that the persistence length of unfolded domains is 0.6 nm. This value is slightly higher than that reported in other studies ( 0.4 nm; see, for example, Rief et al., 1998) and may be explained with the 100 pn force limit imposed on the data, to limit enthalpic contributions at high forces. We obtained the same persistence length for I65 70 and I91 98, which is expected considering the similar amino acid composition of the two fragments. In some of the stretch-release curves, irregular force peaks were seen during stretch (Fig. 5B) and the distribution of the persistence lengths of their release curves had a peak at a value about half of that of curves with regularly spaced force peaks. Our interpretation for this finding is that the irregular curves are derived mainly from doublets (two molecules connected in parallel), since such arrangement would double the force of the release curve and half the persistence length. The effect of differential expression of Ig domains on passive force developed by skeletal muscle can be evaluated by considering the molecular mechanisms of passive-force generation. A passive force model has emerged in which the tandem Ig segments and the PEVK segment behave as serially linked entropic springs. In short sarcomeres, these springs are in a contracted state and possess high configurational entropy. Upon sarcomere extension the springs straighten, their configurational entropy is decreased, and an entropic force is generated, which constitutes passive muscle force. This passive force will increase with the fractional extension of the segment (end-to-end length divided by the contour length; see Methods, Eq. (1)). An increase in the contour length of the tandem Ig segment, via splicing in of Ig domains, for example, reduces, for a given sarcomere length, the fractional extension and thereby passive force. If the muscle operates under conditions where passive forces are high (fast stretch rates), our single molecule results suggests that Ig domains from the differentially spliced region (I65 70 domains) rather than those of the constitutive region (I91 98) are more likely to unfold. Since the unfolding-induced contour-length gain in the differentially spliced domains is greater than that in the constitutive region, the reduction in passive force as a result of Ig domain unfolding will be greater in the differentially spliced domains. Thus, splicing in of Ig domains such as those found in the I65 70 segment may provide a mechanism to lower passive force by selective domain unfolding and to prevent domain unfolding in titin s constitutive region. We gratefully acknowledge Dr. Robert Yamasaki for purifying I91 98, and the generous financial support by the Deutsche Forschungsge-meinschaft (MU 1606/1-1 to C.M.G.; La668/6-2 to S.L.) and the NIH (HL61497 and HL62881 to H.G.), and by the Hungarian Science Foundation (OTKA F025353) TO M.S.Z.K. M.S.Z.K. is a Howard Hughes Medical Institute International Research Scholar. REFERENCES Bagni, M. A., Cecchi, G., Colomo, F., and Garzella, P. (1994) Development of stiffness precedes cross-bridge attachment during the early tension rise in single frog muscle fibres. J. Physiol. 481, 273 278. Bang, M. L., Centner, T., Fornoff, F., Geach, A. J., Gotthardt, M.,

258 WATANABE ET AL. McNabb, M., Witt, C. C., Labeit, D., Gregorio, C. C., and Granzier, H., et al. (2001) The complete gene sequence of titin, expression of an unusual approximately 700-kDa titin isoform, and its interaction with obscurin identify a novel Z-line to I-band linking system. Circ. Res. 89, 1065 1072. Bustamante, C., Marko, J. F., Siggia, E. D., and Smith, S. (1994) Entropic elasticity of lambda-phage DNA. Science 265, 1599 1600. Carrion-Vazquez, M., Oberhauser, A. F., Fisher, T. E., Marszalek, P. E., Li, H., and Fernandez, J. M. (2000) Mechanical design of proteins studied by single-molecule force spectroscopy and protein engineering. Prog. Biophys. Mol. Biol. 74, 63 91. Carrion-Vazquez, M., Oberhauser, A. F., Fowler, S. B., Marszalek, P. E., Broedel, S. E., Clarke, J., and Fernandez, J. M. (1999) Mechanical and chemical unfolding of a single protein: A comparison. Proc. Natl. Acad. Sci. USA 96, 3694 3699. Freiburg, A., Trombitas, K., Hell, W., Cazorla, O., Fougerousse, F., Centner, T., Kolmerer, B., Witt, C., Beckmann, J. S., Gregorio, C. C., et al. (2000) Series of exon-skipping events in the elastic spring region of titin as the structural basis for myofibrillar elastic diversity. Circ. Res. 86, 1114 1121. Gautel, M., and Goulding, D. (1996) A molecular map of titin/ connectin elasticity reveals two different mechanisms acting in series. FEBS Lett 385, 11 14. Gregorio, C. C., Granzier, H., Sorimachi, H., and Labeit, S. (1999) Muscle assembly: A titanic achievement? Curr. Opin. Cell. Biol. 11, 18 25. Guerois, R., and Serrano, L. (2000) The SH3-fold family: Experimental evidence and prediction of variations in the folding pathways. J. Mol. Biol. 304, 967 982. Harpaz, Y., and Chothia, C. (1994) Many of the immunoglobulin superfamily domains in cell adhesion molecules and surface receptors belong to a new structural set which is close to that containing variable domains. J. Mol. Biol. 238, 528 539. Higuchi, H., Nakauchi, Y., Maruyama, K., and Fujime, S. (1993) Characterization of beta-connectin (titin 2) from striated muscle by dynamic light scattering. Biophys. J. 65, 1906 1915. Improta, S., Politou, A. S., and Pastore, A. (1996) Immunoglobulin-like modules from titin I-band: Extensible components of muscle elasticity. Structure 4, 323 337. Kellermayer, M. S., Smith, S. B., Granzier, H. L., and Bustamante, C. (1997) Folding-unfolding transitions in single titin molecules characterized with laser tweezers. Science 276, 1112 1116. Labeit, S., and Kolmerer, B. (1995) Titins: Giant proteins in charge of muscle ultrastructure and elasticity. Science 270, 293 296. Li, H., Carrion-Vazquez, M., Oberhauser, A. F., Marszalek, P. E., and Fernandez, J. M. (2000a) Point mutations alter the mechanical stability of immunoglobulin modules. Nature Struct. Biol. 7, 1117 1120. Li, H., Oberhauser, A. F., Fowler, S. B., Clarke, J., and Fernandez, J. M. (2000b) Atomic force microscopy reveals the mechanical design of a modular protein. Proc. Natl. Acad. Sci. USA 97, 6527 6531. Li, H., Oberhauser, A. F., Redick, S. D., Carrion-Vazquez, M., Erickson, H. P., and Fernandez, J. M. (2001) Multiple conformations of PEVK proteins detected by single-molecule techniques. Proc. Natl. Acad. Sci. USA 98, 10682 10686. Linke, W. A., and Granzier, H. (1998) A spring tale: New facts on titin elasticity. Biophys. J. 75, 2613 2614. Linke, W. A., Ivemeyer, M., Mundel, P., Stockmeier, M. R., and Kolmerer, B. (1998) Nature of PEVK-titin elasticity in skeletal muscle. Proc. Natl. Acad. Sci. USA 95, 8052 8057. Linke, W. A., Ivemeyer, M., Olivieri, N., Kolmerer, B., Ruegg, J. C., and Labeit, S. (1996) Towards a molecular understanding of the elasticity of titin. J. Mol. Biol. 261, 62 71. Lu, H., Isralewitz, B., Krammer, A., Vogel, V., and Schulten, K. (1998) Unfolding of titin immunoglobulin domains by steered molecular dynamics simulation. Biophys. J. 75, 662 671. Lu, H., and Schulten, K. (2000) The key event in force-induced unfolding of titin s immunoglobulin domains. Biophys J. 79, 51 65. Ma, K., Kan, L., and Wang, K. (2001) Polyproline II helix is a key structural motif of the elastic PEVK segment of titin. Biochemistry 40, 3427 3438. Marszalek, P. E., Lu, H., Li, H., Carrion-Vazquez, M., Oberhauser, A. F., Schulten, K., and Fernandez, J. M. (1999) Mechanical unfolding intermediates in titin modules. Nature 402, 100 103. Maruyama, K. (1997) Connectin/titin, giant elastic protein of muscle. FASEB J. 11, 341 345. Minajeva, A., Kulke, M., Fernandez, J. M., and Linke, W. A. (2001) Unfolding of titin domains explains the viscoelastic behavior of skeletal myofibrils. Biophys. J. 80, 1442 1451. Muhle-Goll, C., Habeck, M., Cazorla, O., Nilges, M., Labeit, S., and Granzier, H. (2001) Structural and functional studies of titin s fibronectin type III modules reveal conserved surface patterns, binding to myosin S1- A possible role in the Frank- Starling mechanism of the heart. J. Mol. Biol. 313, 431 447. Rief, M., Gautel, M., Oesterhelt, F., Fernandez, J. M., and Gaub, H. E. (1997) Reversible unfolding of individual titin immunoglobulin domains by AFM. Science 276, 1109 1112. Rief, M., Gautel, M., Schemmel, A., and Gaub, H. E. (1998) The mechanical stability of immunoglobulin and fibronectin III domains in the muscle protein titin measured by atomic force microscopy. Biophys. J. 75, 3008 3014. Rief, M., Pascual, J., Saraste, M., and Gaub, H. E. (1999) Single molecule force spectroscopy of spectrin repeats: Low unfolding forces in helix bundles. J. Mol. Biol. 286, 553 561. Sali, A., and Blundell, T. L. (1993) Comparative protein modelling by satisfaction of spatial restraints. J. Mol. Biol. 234, 779 815. Thompson, J. D., Gibson, T. J., Plewniak, F., Jeanmougin, F., and Higgins, D. G. (1997) The CLUSTAL_X windows interface: Flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. 25, 4876 4882. Thompson, J. D., Higgins, D. G., and Gibson, T. J. (1994) CLUSTAL W: Improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 22, 4673 4680. Trinick, J., and Tskhovrebova, L. (1999) Titin: A molecular control freak. Trends Cell Biol. 9, 377 380. Trombitas, K., Greaser, M., Labeit, S., Jin, J. P., Kellermayer, M., Helmes, M., and Granzier, H. (1998) Titin extensibility in situ: Entropic elasticity of permanently folded and permanently unfolded molecular segments. J. Cell Biol. 140, 853 859. Tskhovrebova, L., and Trinick, J. (2001) Flexibility and extensibility in the titin molecule: Analysis of electron microscope data. J. Mol. Biol. 310, 755 771. Tskhovrebova, L., Trinick, J., Sleep, J. A., and Simmons, R. M. (1997) Elasticity and unfolding of single molecules of the giant muscle protein titin. Nature 387, 308 312. Wang, K. (1996) Titin/connectin and nebulin: Giant protein rulers of muscle structure and function. Adv. Biophys. 33, 123 134. Witt, C. C., Olivieri, N., Centner, T., Kolmerer, B., Millevoi, S., Morell, J., Labeit, D., Labeit, S., Jockusch, H., and Pastore, A. (1998) A survey of the primary structure and the interspecies conservation of I-band titin s elastic elements in vertebrates. J. Struct. Biol. 122, 206 215.