A small proportion of Talin molecules transmit forces to achieve muscle attachment in vivo

Cells in a developing organism are subjected to particular mechanical forces, which shape tissues and instruct cell fate decisions. How these forces are sensed and transmitted at the molecular level is thus an important question, which has mainly been investigated in cultured cells in vitro. Here, we elucidate how mechanical forces are transmitted in an intact organism. We studied Drosophila muscle attachment sites, which experience high mechanical forces during development and require integrin-mediated adhesion for stable attachment to tendons. Hence, we quantified molecular forces across the essential integrin-binding protein Talin, which links integrin to the actin cytoskeleton. Generating flies expressing three FRET-based Talin tension sensors reporting different force levels between 1 and 11 pN enabled us to quantify physiologically-relevant, molecular forces. By measuring primary Drosophila muscle cells, we demonstrate that Drosophila Talin experiences mechanical forces in cell culture that are similar to those previously reported for Talin in mammalian cell lines. However, in vivo force measurements at developing flight muscle attachment sites revealed that average forces across Talin are comparatively low and decrease even further while attachments mature and tissue-level tension increases. Concomitantly, Talin concentration at attachment sites increases five-fold as quantified by fluorescence correlation spectroscopy, suggesting that only few Talin molecules are mechanically engaged at any given time. We therefore propose that high tissue forces are shared amongst a large excess of adhesion molecules of which less than 15% are experiencing detectable forces at the same time. Our findings define an important new concept of how cells can adapt to changes in tissue mechanics to prevent mechanical failure in vivo.


Introduction 24
The shape of multicellular organisms critically depends on the presence of mechanical forces, 25 during development [1,2]. Forces not only generate form and flows within tissues [3,4] but 26 can also control cell fate decisions [5,6] or trigger mitosis [7]. There are various ways to 27 quantify forces at the cellular or tissue level [8,9], however mechanical forces experienced by Integrins are a major and highly conserved force bearing protein family. They connect 38 the actomyosin cytoskeleton to the extracellular matrix and are essential for numerous 39 mechanically regulated processes in vivo or in vitro [21,22]. However, in vivo it is particularly 40 unclear how integrin-based structures are mechanically loaded since forces have so far been 41 analysed in focal adhesions, which typically are not found in soft tissues [11][12][13]16]. 42 Therefore, we chose to investigate Drosophila muscle attachment sites in vivo, which 43 experience high mechanical forces during development [23] and depend on integrin-based 44 attachment of muscle fibers to tendons cells [21,24]. For the molecular force measurements 45 we selected the integrin activator and mechanotransducer Talin, which is essential for all 46 integrin mediated functions and binds with its globular head-domain to the tail of β-integrin 47 and with its rod-domain to actin filaments [25,26]. Thus, Talin is in the perfect position to 48 sense mechanical forces across integrin-dependent adhesive structures. Surprisingly, we find 49 that less than 15% of the Talin molecules experience significant forces at muscle attachments 50 in vivo suggesting that high tissue forces are rather sustained by recruiting a large excess of 51 Talin molecules to muscle attachments. This may have important impact for the robustness of 52 muscle attachment under peak mechanical load in muscles. 53 54

91
To assess the functionality of the Talin tension sensor protein ( Talin-TS) more rigorously, we  92   first analysed Talin-TS localization in adult hemi-thoraxes and found that Talin-TS localizes  93 to myofibril tips as expected ( Fig. 1c-f). Second, we performed western blot analysis to 94 ensure that the tension sensor module is incorporated into Talin protein isoforms as designed 95 (Fig. 1g). Third, we quantified sarcomere length in flight muscles and found the expected 96 length of 3.2 µm in wild type (WT) [28] and talin-TS flies (Fig. 1h-j). Forth, we tested flight 97 ability [29] and found that the insertion of neither the sensor module or the individual 98 fluorescent proteins into the internal position nor the sensor module at the C-terminus caused 99 flight defects (Fig. 1k). Finally, we confirmed that Talin-TS (or Talin-I-YPet) is expressed 100 correctly at all developmental stages (embryo, larva and pupa) and is detected most 101 prominently at muscle attachment sites as previously reported for endogenous Talin (Fig. 1l (Fig. 2a, b). In these cells, Talin-I-YPet localises to adhesions at the fiber tips and at 113 myofibril ends as well as to costameres, which connect myofibrils at the sarcomeric Z-discs to

133
For establishing force measurements using these primary fiber cultures, we performed 134 fluorescence lifetime imaging microscopy (FLIM) to determine the FRET efficiency of the 135 Talin tension sensor containing the HP sensor module (TS) compared to the zero-force 136 control (C-TS). We created distinct masks for Talin FRET signals either in the entire fiber, or 137 only in cell-substrate adhesions at the fiber tips, or in costameres along myofibrils (Fig. 2g-j). 138 Consistent with earlier Talin force measurements, we observed a reduction in FRET 139 efficiency of TS compared to the control C-TS within the entire fiber, indicating that Talin 140 indeed experiences mechanical forces in these adherent, primary muscle fibers. As expected, 141 we find higher average forces across Talin at muscle-substrate adhesions compared to the rest 142 of the cell. In costameres, which are not fixed to the plastic substrate, the FRET efficiency of 143 TS is indistinguishable from the control, indicating that forces across Talin at costameres are 144 lower and do not exceed 6-8 pN. Together, these data demonstrate that the Drosophila Talin 145 tension sensor reports similar Talin forces at adhesions of cultured muscle fibers as were 146 previously described for Talin in focal adhesions of mammalian fibroblasts [11,12,16].

194
To substantiate these findings, we compared flies carrying the HP-based Talin sensor 195 (6-8 pN) to those with the stable variant HPst (9-11 pN), which only differs in two point 196 mutations. We found similar and highly reproducible differences in FRET efficiency (Fig. 3d,  197 Extended Data Fig. 3d) indicating that at 20-24 h APF, some Talin molecules experience 198 forces of even ≥10 pN at muscle attachment sites. Importantly, comparison of TS to its stable 199 variant (stTS) revealed a significant difference in FRET efficiency at 20 h APF while the 200 respective zero-force controls were indistinguishable (Fig. 3d). This demonstrates that a 201 proportion of the mechanically engaged Talin molecules experience a range of forces between 202 7 and 10 pN at muscle-tendon attachments in vivo, further emphasizing that the observed 203 differences are force-specific. 204 To test whether the remaining Talin molecules experience forces that are too low to be 205 detected by the HP or HPst sensor modules, we generated flies with the F40 sensor module, 206 which is sensitive to forces of 1-6 pN [13]. Again, we quantified a decrease in FRET 207 efficiency relative to the control at 20 h and 24 h APF but FRET efficiency differences 208 remained small and no change was observed at 30 h APF (Fig. 3e). Thus, a large proportion 209 of the Talin molecules at muscle attachment sites are not exposed to significant mechanical 210 forces during development.

Talin concentration at developing muscle attachments 221
Since Talin is thought to play an important mechanical role during tissue formation, we 222 wanted to test whether such a small proportion of mechanically engaged Talin molecules in 223 vivo could still contribute a significant amount of tissue-level tension. We therefore quantified 224 the absolute amount of Talin molecules present at muscle attachment sites by combining in 225 vivo fluorescence correlation spectroscopy (FCS) with quantitative confocal imaging (see 226 workflow in Extended Data Fig. 4a-d). From FCS measurements in the muscle interior we 227 calculated the counts per particle (CPP) value, i.e. the molecular brightness of a single 228 Talin-I-YPet particle in each pupa. Since such a particle may correspond to a Talin monomer 229 or dimer, we compared the Talin-I-YPet brightness to the brightness of free monomeric YPet 230 expressed in flight muscles and found no significant difference (Fig. 4a). We conclude that 231 Talin is mostly monomeric in the muscle interior. 232 Next, we calculated the Talin concentration a muscle attachment sites by calibrating 233 confocal images using the molecular brightness (CPP) information from the FCS 234 measurements. Using a dilution series of Atto488, we ascertained that the fluorescence 235 intensity increases linearly with the concentration over multiple orders of magnitude in our 236 measurements (Extended Data Fig. 4e). The resulting images with pixel-by-pixel Talin 237 concentration values (Fig. 4b) indicate an average concentration at the muscle attachment of 238 5.9 µM (20 h), 10.9 µM (24 h) and 30.9 µM (30 h) (Fig. 4c). Thus, the local concentration of 239 Talin molecules increases approximately two-fold from 20 h to 24 h and five-fold to 30 h, 240 indicating that Talin may contribute to the overall increase in tissue stress by its strong 241 recruitment to maturing muscle attachment sites. 242

Imaging of stainings 367
Samples were imaged on a Zeiss LSM 780 scanning confocal microscope with Plan 368 Apochromat objectives (10x air, NA 0.45 for overview images and 40x oil, NA 1.4 for detail 369 images). For thick samples, a z-stack was acquired and maximum-projected using ImageJ. 370

Sarcomere length quantification 371
Sarcomere length was quantified as previously described using the ImageJ plug-in MyofibrilJ 372 (https://imagej.net/MyofibrilJ) [28]. Briefly, an area with straight, horizontal myofibrils is 373 analysed by Fourier transformation to find the periodicity of the sarcomeres. One area was 374 analysed for each hemithorax stained with phalloidin and imaged at 40x and zoom 4. 375

Western blotting 376
Western blotting was performed according to standard procedures. Specifically, 15 flies each 377 were homogenized in 100 µL 6x SDS loading buffer (250 mM Tris pH 6.8, 30% glycerol, 1% 378 SDS, 500 mM DTT) and heated to 95°C for 5 min. 200 µL of water were added and the 379 equivalent of 0.5 (10 µL) and 1 fly (20 µL) were loaded onto a NuPAGE Novex 3-8% Tris-380 Acetate Gel. The transfer to the membrane was carried out overnight. The membrane was 381 blocked (5% blotting grade blocker, BioRad) and then incubated overnight with a 1:1 mixture 382 of anti-Talin antibodies E16B and A22A (1:1000 in block). For detection, HRP anti-mouse 383 antibody and Immobilon Western Chemiluminescent HRP Substrate (Millipore) were used. 384 collected on apple juice agar plates for 24 hours and dechorionated in 50% bleach (0.024% 392 hypochlorite) for 3 min. Living embryos were mounted in 50% glycerol before imaging. L3 393 larvae from the same cross were immobilised by immersing them in 60°C water for about 1 s 394 [29] and mounted using a plexiglass slide with a groove and one spacer coverslip on each side 395 in 50% glycerol. 5x1 tile scan z-stacks were acquired using a 10x objective to image the 396 entire larva. 397

Isolation and differentiation of primary muscle fibers 398
Primary cells were isolated from Drosophila embryos and differentiated as previously 399 described [31,32] with the following modifications: Embryos (5-7 hours old, aged at 25°C) 400 were collected from smaller cages on only one 9 cm molasses plate per genotype. Embryos 401 were homogenized with a Dounce homogeniser using a loose fit pestle in 4 mL Schneider's 402 Drosophila medium (Gibco 21720-024, lot 1668085) and after several washing steps (using 403 2 mL medium) re-suspended to a concentration of 3x10 6 cells/mL. Finally, cells were plated 404 in 8-well ibidi dishes (1 cm 2 plastic bottom for microscopy with ibiTreat surface) coated with 405 vitronectin (optional) at a density of 3-9x10 5 cells/cm 2 and differentiated for 5-7 days at 25°C 406 in a humid chamber. 407

Fixation, staining and imaging of primary muscle fibers 408
Primary muscle fibers were fixed on day 6 after isolation with 4% PFA in PBS for 10 min at 409 RT on a shaker. Phalloidin-staining (Alexa647-conjugate, Molecular Probes) was performed 410 overnight in the dark at 4°C. Fixed cells were imaged in PBS on a Zeiss LSM 780 with a 40x 411 oil objective (Plan Apochromat, NA 1.4). Live imaging of twitching primary cells was 412 performed on a Leica SP5 confocal with a 63x water objective (HCX PL APO 63x/1.2 W 413 CORR λ BL ), acquiring the transmission light channel and the YPet channel simultaneously. 414

Sample preparation for live imaging of pupae 415
White pre-pupae were collected and aged at 27°C to the desired time point. Before imaging, a 416 window was cut into the pupal case above the thorax and the pupae were mounted on a 417 custom-made slide with a groove as previously described [50]. 418

FLIM-FRET data analysis 427
The FLIM data were analysed using a custom-written MATLAB program [11,12]. First, an 428 intensity image was created to manually draw a region of interest (ROI) around the target 429 structure (adhesions/costameres in primary cells or muscle attachment sites in pupae, also see 430 Extended Data Fig. 2). To create a binary mask of the target structure, Multi-Otsu 431 thresholding with three classes was applied to the signal in the ROI blurred with a median 432 filter (3x3 pixels) and holes in the mask containing the brightest class were filled. Photon 433 arrival times of all photons inside the mask were plotted in a histogram and the tail of the 434 curve was fitted with a monoexponential decay yielding the fluorescence lifetime τ. Fits with 435 more than 5% relative error in lifetime determination were excluded from further analysis. 436 For dimmer samples (primary fiber cultures and intermolecular FRET pupae), we used a 10% 437 relative error cut-off. The FRET efficiency E was calculated according to the following 438 formula with τ DA being the lifetime of the donor in presence of the acceptor and τ D the 439 lifetime of the donor alone: 440 For all measurements, τ D was determined using Talin-I-YPet. Experiments were repeated 2-5 442 times on different experiment days with 10-15 pupae/cells imaged per genotype and day. 443

Calculation of the proportion of mechanically engaged Talin 444
We determined the number of mechanically engaged (=open) tension sensor N open relative to 445 the total number of molecules N total at the muscle attachment site using biexponential fitting 446 similarly as previously described [12]. Briefly, we assumed that the fluorescence decay from 447 a tension sensor FLIM measurement can be described by because of the large contour length increase upon opening of the sensor. Thus, we determined 451 the lifetime τ noFRET by using a monoexponential fit on Talin-I-YPet data as described above. 452 The lifetime τ FRET was determined from zero-force control (Talin-C-TS) data. Since the 453 Talin-C-TS sample contains fully fluorescent sensor (τ FRET ) and sensor with non-fluorescent 454 mCherry acceptor (τ noFRET ), we used a biexponential fit with fixed τ noFRET to determine τ FRET . 455 The two lifetimes τ noFRET and τ FRET were then fixed and used to fit Talin-TS and Talin-C-TS  456 data biexponentially, thereby determining the relative contributions of photons from 457 molecules with these two lifetimes. From this, the relative number of molecules with τ noFRET 458 and τ FRET was calculated, taking into account that FRET reduces the number of photons 459 detected in the donor channel. Finally, the ratio N open /N total was determined by normalizing the 460 Talin-TS values to the respective Talin-C-TS For all measurements, the axis ratio of the detection volume S = 5 was consistently fixed [53].

Estimation of Talin density and tissue stress 525
To estimate Talin density on the membrane from pixel-by-pixel concentration values, we 526 divided the average number of molecules in the focal volume at the muscle attachment sites 527 by the membrane area in the focal volume. The focal volume was determined by 528 Rhodamine 6G FCS measurements as described above. For the shape of the focal volume we 529 assumed an ellipsoid with the long axis (z) being 5-times the short axis (x=y). Hence, for a 530 focus volume of 0.32 fL, the membrane area in the z-y-plane is 0.63 µm 2 . Taking into account 531 that there are two membranes (one from the tendon and one from the muscle) and that the 532 membrane is not flat (ruffles approximately increase the area 2-fold as determined from EM-533 images [56]) the total membrane area in the focal volume is about 2.5 µm 2 . 534 To estimate Talin-mediated tissue stress, we calculated force threshold of sensor x Talin 535 density x proportion of mechanically engaged Talin = 7 pN x 400 molecules/µm 2 x 13.2% = 536 0.37 kPa for 20 h APF and 7 pN x 700 molecules/µm 2 x 9.6% = 0.47 kPa for 24 h. Note, that 537 these values are lower estimates since individual molecules might experience forces higher 538 than 7 pN. 539

Statistics 540
Box plots display the median as a red line and the box denotes the interquartile range. 541 Whiskers extend to 1.5 times the interquartile range from the median and are shortened to the 542 adjacent data point (Tukey). In addition, all data points are shown as dots. Tests used for 543 statistical evaluation are indicated in the figure legends. 544 Code availability 545 FLIM-FRET data was analysed using custom-written MATLAB code as published previously 546 [11,12]. The code is available upon request. 547

Supplemental Video 1 -Legend 818
Video of twitching primary muscle fiber shown in Fig. 2f. Talin-I-YPet signal (green) is 819 overlaid with the transmission light channel (grey) acquired simultaneously. The length of the 820 movie is 1 min with a time resolution of 1.29 s played at 10x speed. 821