Titin stiffness modifies the force-generating region of muscle sarcomeres

The contractile units of striated muscle, the sarcomeres, comprise the thick (myosin) and thin (actin) filaments mediating active contraction and the titin filaments determining “passive” elasticity. We hypothesized that titin may be more active in muscle contraction by directly modulating thick-filament properties. We used single-myofibril mechanical measurements and atomic force microscopy of individual sarcomeres to quantify the effects of sarcomere strain and titin spring length on both the inter-filament lattice spacing and the lateral stiffness of the actin-myosin overlap zone (A-band). We found that strain reduced the lattice spacing similarly in sarcomeres with stiff (rabbit psoas) or compliant titin (rabbit diaphragm), but increased A-band lateral stiffness much more in psoas than in diaphragm. The strain-induced alterations in A-band stiffness that occur independently of lattice spacing effects may be due to titin stiffness-sensing by A-band proteins. This mechanosensitivity could play a role in the physiologically important phenomenon of length-dependent activation of striated muscle.


Results and Discussion
Titin isoform size and passive stress in uncompressed vs. dextran-compressed myofibres. Two different rabbit muscle types were used for the experiments, one expressing a relatively small titin isoform (psoas; M W ~ 3.3 MDa), the other a large titin isoform (diaphragm; M W ~ 3.7 MDa) (Fig. 1b). The size difference arises due to alternative splicing of modules in the titin spring segment (Fig. 1a) 15 . By performing force and sarcomere length (SL) measurements of skinned myofibres in relaxing buffer we found a 2-3 times higher passive tension in psoas vs. diaphragm sarcomeres (Fig. 1c). However, the myofilament lattice becomes expanded with skinning, compared to the in vivo situation, which could affect titin-based stiffness. Therefore, we also measured the passive stress-strain relationships in the presence of an osmotic compressing agent, dextran T-500, which at high concentrations (15%) was shown to modestly increase the passive tension of skinned cardiac muscle strips 16 . We supplemented the relaxing buffer with 5% dextran (lattice compression, 4.7 kPa), because myofilament calcium sensitivity is stable within the 2-6% concentration range 17 and because a dextran concentration of 5% has frequently been used by others to normalize the lattice spacing of skinned skeletal myofibres and to mimic the physiological situation [18][19][20] . We found that 5% dextran did not significantly alter the passive stress-strain curve of either fibre type (Fig. 1c).

Strain causes similar A-band compression in psoas and diaphragm sarcomeres.
Strain is expected to decrease the width of a muscle fibre 8 . We wanted to know whether sarcomere strain also reduces the width of the A-band in the isolated myofibril, as predicted, e.g., from the presence of titin's radial force component (Fig. 1a). Because fibre width and myosin lattice spacing are linearly related in skinned rabbit psoas fibres 19 , we assumed that the relationship between A-band diameter and lattice spacing is also linear. We measured the A-band diameter of non-activated single rabbit psoas or diaphragm myofibrils during stepwise stretching from slack SL (2.1-2.3 μm; strain, 1.0) up to ~145% slack SL (3.0-3.2 μm) (Fig. 2a), the latter of which is at the high end of the physiological SL range in rodent muscles 21 . For sarcomeres in relaxing buffer lacking dextran, the A-band diameter (mean value at slack SL, 1.10 μm for psoas and 0.95 μm for diaphragm, with no significant difference; Fig. 2b, inset) successively decreased with increasing strain, reaching ~79% and ~85% of the initial value at the highest strain level, in psoas and diaphragm, respectively (Fig. 2b). The mean values were well fit by linear regressions and agreed with those obtained by others for the effect of SL-increase on the width of skinned myofibres 5,20,22 . We observed a slightly larger decrease in psoas vs. diaphragm A-bands.
Upon addition of 5% dextran to the relaxing buffer in order to normalize the lattice, the A-bands became thinner by 15-20% at slack SL (Fig. 2a, bottom; Fig. 2b). This finding is consistent with previous measurements on permeabilized myofibres or myofibrils showing similarly reduced widths in response to osmotic compression by these low dextran concentrations 19,20,23,24 . In our hands, the dextran-effect was somewhat larger in psoas than in diaphragm sarcomeres. Stretch of the dextran-compressed myofibrils further reduced the A-band diameter. However, the negative slopes of the fit curves to the means were less steep than before osmotic compression (Fig. 2b). Importantly, the proportional decrease in A-band diameter at comparable strain levels was now identical in the two myofibril types (Fig. 2c). Elsewhere, skinned rat skeletal and cardiac muscle fibres, which differed greatly in titin-based stiffness, were studied by X-ray diffraction for stretch-dependent changes in the myosin lattice spacing 4 . The relative reduction in lattice spacing that was found with strain was the same in the different muscle types, while the absolute values differed. Another study reported a decrease in inter-filament lattice spacing with titin-based spring force, for skinned rabbit psoas fibres; however, the effect was much reduced following osmotic compression by dextran 11 . We conclude that the strain-induced reduction in A-band diameter or lattice spacing does not depend much on the stiffness of the titin spring, in particular so in sarcomeres in which the lattice is modestly compressed (as under physiological conditions).

The lateral thick-to-thick filament spacing in individual A-bands deduced from AFM images.
Since the A-band diameter is only an indirect measure of inter-filament lattice spacing, we wanted to obtain information about the thick-to-thick filament spacing in our single myofibrils. The aim here was to compare the lattice spacing in the single myofibrils, which have been removed from the crowded cellular environment, to that in muscle cells/bundles as determined previously by X-ray diffraction. To this end, we employed the AFM following an approach developed for insect myofibrils 25 . We chose rabbit psoas for these measurements, since values of myosin lattice spacing are readily available for this muscle type from X-ray diffraction experiments. In AFM tapping mode we acquired high-resolution height images of non-stretched myofibrils firmly adhering to a cover glass and bathed in relaxing buffer (Fig. 3a). The sarcomeric Z-disk, I-and A-bands could readily be distinguished and filamentous structures running axially along the A-band were clearly resolved. The peak-to-peak lateral distances between these elevated structures were analyzed on profile plots perpendicular to the myofibril axis (Fig. 3a). In the example shown, the observed inter-filament distances were 120 nm, 81 nm, and 48 nm. These values agree well with those expected for the thick-to-thick-filament spacings measurable on the myofibrillar surface (Fig. 3b). The diversity in spacings is due to the intersection of different planes of the thick-to-thick filament lattice at the surface of the myofibril 25 . Calculation of the spacing along these planes is based on a 47-nm primary spacing, A P , calculated as A P = 2/√ 3 · d 1,0 , where d 1,0 is the interfilament distance (Fig. 3b). We used a d 1,0 value of 40.3 nm, which is from X-ray diffraction experiments and applies to relaxed rabbit psoas muscle at ambient temperature and SL ~ 2.2 μm 26 . A d 1,0 value of ~40 nm has consistently been found for non-stretched, skinned rabbit psoas fibres under comparable buffer conditions 19,23,27,28 . Taken together, it is very likely that the filamentous structures running axially along the A-bands on our AFM images are the myosin thick filaments.
For a more comprehensive analysis of the myosin pattern in our AFM images we applied an algorithm that detected the elevated longitudinal structures in each line scan and followed them in 50-100 line scans/image to cover half of an A-band (Fig. 3c, top). Using this algorithm, we measured all nearest inter-filament distances in a given half-A-band (Fig. 3c, bottom). From the data shown in the histogram we inferred the primary spacing A P and a d 1,0 value of 43.2 ± 7.2 nm (mean ± SD). We applied this approach to 3 high-quality images of different psoas A-bands at 2.2-2.3 μm SL in relaxing buffer and pooled the values in a histogram (Fig. 3d, top). We found distinct peaks, among others, for inter-filament distances of approximately 47 nm (the expected primary spacing), 82 nm, 125 nm and 171 nm. From these data we generated a binned histogram for the distribution of calculated d 1,0 values, which showed a mean d 1,0 of 40.9 ± 4.5 nm (Fig. 3d, bottom). Thus, the myosin spacing in our single isolated psoas myofibrils is indistinguishable, within experimental error, from that reported in X-ray diffraction studies for relaxed (skinned, non-stretched) psoas muscle cells. These results suggest that the normal inter-filament lattice spacing is maintained in single isolated myofibrils dissected out of their cellular environment.
Lateral stiffness maps of sarcomeres by AFM force mapping. We reasoned that increasing the stiffness of titin in sarcomeres could affect the A-band region, (i) because the reduction in lattice spacing due to titin's radial force component promotes the repulsive forces between the charged actin and myosin filament surfaces, (ii) and/or because the strained titin directly affects the conformation or structural arrangement of proteins in the overlap zone of thick and thin filaments. We speculated that these putative titin stiffness-dependent alterations could increase the lateral stiffness of the A-band. To measure lateral stiffness we used AFM force mapping, which monitors cantilever deflection in response to sample indentation 25,[29][30][31] . Force maps consisting of 2,500 (50 × 50) force-indentation curves/image were generated on individual sarcomeres bathed in relaxing buffer, and occasionally also in "rigor" buffer lacking ATP (to maximize actin-myosin interactions). Representative force maps of psoas sarcomeres demonstrated much higher lateral stiffness under rigor compared to relaxing conditions (Fig. 4a,b), consistent with earlier findings 32 . We could distinguish regions of different stiffness along the sarcomere. In rigor, the A-band was the stiffest, followed by the M-band and the Z-disk (Fig. 4a). In relaxing buffer, the sarcomere was generally softer than in rigor and the Z-disk was the stiffest, followed by the M-and A-bands (Fig. 4b). For a convenient quantification of lateral stiffness at a given indentation we calculated the pointwise apparent modulus, E app 33 , from the force-indentation curves (Fig. 4c). Under relaxing conditions the E app for A-bands decreased with indentation and reached a quasi-plateau beginning at 60-80 nm indentation depth; only the value at the plateau was used for further analysis. We prepared single myofibrils expressing stiff (psoas) or compliant titin (diaphragm) and measured E app both at slack SL and following a ~45% stretch (SL, 3.1-3.2 μm), with or without 5% dextran. The stretch was accomplished under a phase-contrast microscope, the ends of the stretched myofibril were glued to a cover glass, and the sample (in relaxing buffer) on the cover glass was transferred to the AFM for force mapping (Fig. S1a). Force maps of stretched sarcomeres in the absence of dextran, as well as those of dextran-treated sarcomeres at slack SL, showed lower indentability reflecting higher A-band lateral stiffness compared to slack sarcomeres in the absence of dextran (Fig. 4d).

Titin stiffness is an important determinant of A-band lateral stiffness.
In the absence of dextran, the mean E app of relaxed psoas and diaphragm A-bands at slack SL was identical, ~3 kPa (Fig. 4e). Addition of 5% dextran increased E app by a factor of ~3 in both myofibril types. While an increase in myofibril lateral stiffness with dextran treatment was reported before 32 , our results show that modest compression of the myofilament lattice increases A-band lateral stiffness independently of titin spring length, at least in unstretched sarcomeres. In contrast, a 45% stretch affected E app of psoas and diaphragm A-bands differentially. Without dextran, E app significantly increased to 7.95 ± 0.33 kPa (mean ± SEM) in stretched psoas but only to 4.46 ± 0.53 kPa in stretched diaphragm (Fig. 4e). With dextran present in the relaxing buffer, E app increased to 17.51 ± 1.27 kPa in stretched psoas and significantly less, to 11.35 ± 1.11 kPa, in stretched diaphragm. Interestingly, the lateral stiffness of the Z-disk region, measured for comparison, increased with osmotic lattice compression and with sarcomere strain, but was indifferent to myofibril type (Fig. S1b). Possibly, the known differences in Z-disk structure (and strength?) between fast (psoas) and slow (diaphragm) muscle 34 offset any effect of titin stiffness on Z-disk lateral stiffness. We conclude that the effects of osmotic compression and titin stiffness increase on lateral stiffness are cumulative (although not necessarily in terms of the arithmetical sum). A certain proportion of the lateral stiffness increase at the A-band should simply be mediated by the reduction in lattice spacing, which probably increases the electrostatic repulsion between the myofilaments. However, the stiffness of titin per se is important too, since it affects A-band lateral stiffness (Fig. 4e), but is not a critical factor in the strain-induced compression of the inter-filament lattice (Fig. 2c). To explain this scenario, we consider a direct effect of titin stiffness on proteins within the actin-myosin overlap zone. Several lines of evidence suggest that various A-band proteins could "sense" titin stiffness. Among these proteins are the myosin heads, which bind to A-band titin modules 35 and show altered ordering with increasing SL in relaxed muscle 36 , although a positioning closer to actin could not be confirmed in a new study 37 . Furthermore, the stiffening of titin strains the myosin filament by a small degree 11 and this modification could translate into altered A-band lateral stiffness. Other candidate proteins are the troponin subunits and tropomyosin, which both regulate the activation of the thin filament and are important for LDA 2 . Because the titin spring binds to sarcomeric actin 38,39 and tropomyosin 40 , stretch of titin could in theory affect the structural arrangement of these thin filament proteins, perhaps with long-range effects on their conformation in the A-band region. Notably, a special type of LDA, the stretch activation of insect indirect flight muscle, is mediated by a troponin isoform that forms bridges with myosin subunits 41,42 . If similar connections were present in mammalian sarcomeres, and if they were somehow engaged by the stretching of the titin filaments, one could envision such troponin bridges to mediate the effect of titin stiffness increase on lateral A-band stiffness. An additional candidate is myosin-binding protein-C (MyBP-C), which binds to myosin, titin, and actin 43 . Conceivably, increased titin stiffness from sarcomere strain could impact the positioning of MyBP-C between the actin and myosin filaments and thus, affect lateral A-band stiffness.
In summary, our work provides evidence for a direct effect of titin stiffness on A-band lateral stiffness, which appears under sarcomere strain. This effect is largely independent of changes in inter-filament spacing. The titin contribution to A-band lateral stiffness is additive to the contribution from ionic interactions between actin and myosin filaments probed by osmotic compression. Taken together with previous findings, including latest data from time-resolved small-angle X-ray diffraction measurements on cardiac muscle 37 , it appears that the sarcomere can sense titin stiffness via structural re-arrangements or conformational changes of proteins within the actin-myosin overlap zone. These effects may provide, in part, the molecular basis for LDA. Our study thus corroborates other recent evidence in favour of an active involvement of titin in muscle contraction 44,45 . As titin is increasingly recognized as an important (cardio)myopathy gene/protein, our findings also have implications for a better understanding of the pathophysiology of heart and muscle diseases. (e) Pointwise apparent modulus (at 60-80 nm indentation depth) of psoas and diaphragm A-bands in relaxing buffer at slack SL (n = 12 − 13 sarcomeres); ~145% stretch (n = 11 sarcomeres); slack SL, 5% dextran (n = 7 sarcomeres); and ~145% stretch, 5% dextran (n = 6 − 7 sarcomeres). mean ± SEM; *p < 0.05; **p < 0.01; ***p < 0.001; ANOVA/Bonferroni-adjusted t-test.

Methods
Myofibril stretch experiments and A-band diameter measurements. Myofibrils were prepared from rabbit psoas and diaphragm muscle as described 15 . Briefly, muscle strips were dissected, tied to glass rods and used up fresh or stored at − 20 °C in rigor buffer (75 mM KCl, 10 mM Tris, 2 mM MgCl 2 , 2 mM EGTA, pH 7.1; 40 μg/mL protease inhibitor leupeptin) containing 50% glycerol. Experimental procedures were performed in accordance with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (8 th Edition, 2011). All experimental protocols were approved by the Institutional Animal Care and Use Committee of Ruhr University Bochum. To obtain myofibrils, small muscle pieces were minced, exposed for ~5 h to rigor buffer containing 0.5% Triton X-100 for skinning, and homogenized in rigor buffer. A drop of the myofibril suspension was placed on a cover slip under a Zeiss Axiovert 135 microscope. A single myofibril was selected and glued at the ends to the tip of a fine glass micropipette controlled by a micromanipulator (MHW 103, Narishige). Surplus myofibrils were removed by several washes with rigor buffer or relaxing buffer (7.8 mM ATP, 10 mM phosphocreatine, 20 mM imidazole, 4 mM EGTA, 12 mM Mg-propionate, 97.6 mM K-propionate, pH 7.0, 40 μg/ml leupeptin, 1 mM DTT; ionic strength,180 mM). In a subset of experiments, we added 5% dextran T-500 (w/v) to the relaxing buffer as an osmotic agent. All experiments were performed at room temperature. Using a custom-made setup for isolated myofibril mechanics 46,47 , a single myofibril was stretched stepwise in relaxing buffer with or without dextran. At each stretch state the myofibril image was recorded using an 885-EMCCD camera (Andor Technology), the SL was determined, and the A-band diameter of at least five sarcomeres/myofibril was measured (ImageJ software; NIH, Bethesda) by calculating the full-width-at-half-maximum (FWHM) signal on intensity profiles perpendicular to the myofibril axis.
Passive force measurements on single myofibres. Psoas or diaphragm muscle fibres were dissected from Triton-skinned muscle bundles. Dissection was done in relaxing buffer and care was taken to avoid excessive stretching of the fibres. Single fibres were suspended in a muscle mechanics workstation (Scientific Instruments, Heidelberg) and force-extension measurements were performed in relaxing buffer at room temperature, as described 15 . Fibres were stretched from slack SL (2.0-2.2 μm) in 5 steps of ~0.2 μm/sarcomere (completed in 1 s), while force was recorded (sampling rate, 100 Hz) and SL measured by laser diffraction during a 1-min hold period after each step. Following the last stretch-hold, fibres were released back to slack SL to test for possible shifts of baseline force. Three identical stretch-release protocols were performed on the same fibre (with a 5-min pause in-between) and the mean forces calculated. The force value at the end of each hold period (quasi steady-state force) was used to determine force per cross-sectional area (stress). Fibre cross-sectional area was estimated from the diameter (assuming a circular shape) measured at slack SL under a binocular microscope using a μm-grid. Following a series of measurements in relaxing buffer, 5% dextran T-500 was added to the solution and another series of measurements performed on the same fibre.
Titin gels. Titin isoforms were separated as described 48 . Briefly, tissue samples from the muscles used for mechanical measurements were solubilized in 50 mM Tris-sodium dodecyl sulfate (SDS) buffer (pH 6.8) containing 8 μg/mL leupeptin (Peptin Institute, Japan) and phosphatase inhibitor cocktail (PIC [P2880], 10 μL/mL; Sigma). Samples were heated for 3 minutes and centrifuged. Psoas, diaphragm, and a mix of both muscles were loaded at low (15 μg) and high (30 μg) concentrations and separated by 1.8% SDS polyacrylamide gel electrophoresis. Gels were run at 5 mA constant current for 16 hours. Titin bands were visualized by Coomassie-staining and digitized using LAS-4000 Image Reader (Fujifilm). AFM imaging of myofibrils and topographical analysis of putative myosin filaments. Glass cover slips were extensively washed and incubated in 5 M KOH for 30 minutes. Then the slides were sonicated, washed with ethanol and air dried. Psoas myofibrils in relaxing buffer were placed on a cover slip and height images of well-adhering single myofibrils or doublets were acquired (scan rate, 1 Hz; 256 pixels/scan) using an integrated AFM/inverted optical microscope (MFP-3D-BIO, Asylum Research). Rectangular, gold-coated silicon nitride cantilevers (Biolever BL-RC150VB-C1 (A lever) from Olympus; length, 60 μm; spring constant, 0.03 N/m; resonant frequency, 37 kHz (in air)) with a V-shaped tip (apex diameter, 60 nm; height, 7 μm) were used. Tip geometry resembles a hollow pyramid sliced in half vertically, with a sharpened apex, allowing for exact positioning on the myofibril. To minimize tip/sample interaction, tapping mode was used. Setpoint amplitude and integral gain were carefully adjusted for each experiment, following published protocols 33 . Line scans recorded perpendicular to the myofibril axis (in the A-band region) were searched for clear peaks representing elevated structures on the myofibrillar surface, presumably indicating the myosin filaments. The peak-to-peak distances were determined and collectively used to infer the primary myosin spacing (A p ) and the inter-filament distance, d 1,0 , using d 1,0 = A P · √ 3/2, as described 25 . Furthermore, an algorithm was used that detected the peaks in each line scan, followed them in 50-100 line scans/A-band image, and calculated all nearest peak-to-peak (= inter-filament) distances.
AFM force mapping and analysis of lateral stiffness. Myofibrils adhering to the cover slip were analyzed for their lateral stiffness in rigor buffer and in relaxing buffer with or without dextran. We studied relaxed myofibrils either at slack SL or following a controlled stretch to ~145% slack, performed under a Zeiss Axiovert 145 inverted microscope. The stretched myofibril was glued at the ends to the cover glass using silicone adhesive (1:1 mixture (v/v) of Dow Corning 3145 RTV and 3140 RTV) and the glass needle tips to which the myofibril was attached were broken by pressing the tips onto the glass surface. The stretched myofibril immersed in relaxing buffer at any time was then transferred together with the cover slip to the AFM. For force mapping, we first measured the resonant frequency (~8 kHz) and the spring constant (~0.03 N/m) of the Biolever BL-RC150VB-C1 (A lever) in relaxing buffer, followed by a short scan in contact mode to ensure an intact and clean cantilever tip. After Scientific RepoRts | 6:24492 | DOI: 10.1038/srep24492 selecting a region of interest containing a whole sarcomere, 2,500 (50 × 50) individual force curves were recorded over this region to generate a force (stiffness) map. Individual force curves were analyzed as follows 25 . The contact point was identified using the second derivative of the filtered deflection-extension curve and the origin of the curve was set to this point. The indentation δ was computed by subtracting the deflection d of the cantilever from the piezo movement z after contacting the myofibril (δ = z − d), to obtain force-indentation curves. Since the myofibrils have nonlinear transversal elastic moduli, we calculated the pointwise apparent modulus, E app , to obtain depth-dependent changes in the lateral stiffness of the sarcomere, using F = 2π × E app × Ø(δ), where F is the indentation force and Ø(δ) expresses the geometry of the cantilever tip 33 . For Ø(δ) we used the equation for a parabolic indenter, Ø(δ) = 4/3π × √ (R C δ 3 ), with R C being the radius of curvature of the tip (~50 nm). E app is a measure of lateral stiffness and can be related to the Young's modulus E via E = E app × 2(1 − ν 2 )], where ν is the Poisson ratio. (For biological samples, ν is usually set to 0.5, assuming a constant volume.) For the analysis we used the E app values at 60-100 nm indentation, because E app approached a plateau at this indentation depth. E app was determined in the actin-myosin overlap region of a given sarcomere (= A-band region) and also in the Z-disk region. The A-band, M-band, and Z-disk regions were detected by their different lateral stiffness and by their different surface topology measured on AFM height images acquired in tapping mode following the force mapping experiment. Only those sarcomeres were included in the analysis where the Z-disk stiffness could readily be distinguished from the lateral stiffness of the remainder of the sarcomere. In relaxed sarcomeres at slack length, the Z-disk was usually the stiffest sarcomeric region.
Statistics. Results were depicted as mean ± SEM, unless indicated otherwise. Significance was tested by Bonferroni-adjusted t-test in conjunction with ANOVA. Three levels of significance were used: *p < 0.05; **p < 0.01; and ***p < 0.001.