Single Molecule Mechanics and Kinetics of Cardiac Myosin Interacting with Regulated Thin Filaments

The cardiac cycle is a tightly regulated process wherein the heart generates force to pump blood to the body during systole and then relaxes during diastole. Disruption of this finely tuned cycle can lead to a range of diseases including cardiomyopathies and heart failure. Cardiac contraction is driven by the molecular motor myosin, which pulls regulated thin filaments in a calcium-dependent manner. In some muscle and non-muscle myosins, regulatory proteins on actin tune the kinetics, mechanics, and load dependence of the myosin working stroke; however, it is not well understood whether or how thin filament regulatory proteins tune the mechanics of the cardiac myosin motor. To address this critical gap in knowledge, we used single-molecule techniques to measure the kinetics and mechanics of the substeps of the cardiac myosin working stroke in the presence and absence of thin filament regulatory proteins. We found that regulatory proteins gate the calcium-dependent interactions between myosin and the thin filament. At physiologically relevant ATP concentrations, cardiac myosin’s mechanics and unloaded kinetics are not affected by thin filament regulatory proteins. We also measured the load-dependent kinetics of cardiac myosin at physiologically relevant ATP concentrations using an isometric optical clamp, and we found that thin filament regulatory proteins do not affect either the identity or magnitude of myosin’s primary load-dependent transition. Interestingly, at low ATP concentrations, thin filament regulatory proteins have a small effect on actomyosin dissociation kinetics, suggesting a mechanism beyond simple steric blocking. These results have important implications for both disease modeling and computational models of muscle contraction.

pump blood to the body during systole and then relaxes during diastole. Disruption of 27 this finely tuned cycle can lead to a range of diseases including cardiomyopathies and 28 heart failure. Cardiac contraction is driven by the molecular motor myosin, which pulls 29 regulated thin filaments in a calcium-dependent manner. In some muscle and non-30 muscle myosins, regulatory proteins on actin tune the kinetics, mechanics, and load 31 dependence of the myosin working stroke; however, it is not well understood whether or 32 how thin filament regulatory proteins tune the mechanics of the cardiac myosin motor. 33 To address this critical gap in knowledge, we used single-molecule techniques to 34 measure the kinetics and mechanics of the substeps of the cardiac myosin working 35 stroke in the presence and absence of thin filament regulatory proteins. We found that 36 regulatory proteins gate the calcium-dependent interactions between myosin and the 37 thin filament. At physiologically relevant ATP concentrations, cardiac myosin's 38 mechanics and unloaded kinetics are not affected by thin filament regulatory proteins. 39 We also measured the load-dependent kinetics of cardiac myosin at physiologically 40 relevant ATP concentrations using an isometric optical clamp, and we found that thin 41 filament regulatory proteins do not affect either the identity or magnitude of myosin's 42 primary load-dependent transition. Interestingly, at low ATP concentrations, thin filament 43 regulatory proteins have a small effect on actomyosin dissociation kinetics, suggesting a 44 mechanism beyond simple steric blocking. These results have important implications for 45 both disease modeling and computational models of muscle contraction. 46 47 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi. org/10.1101org/10. /2023 stroke in two substeps, with the first substep associated with phosphate release and the 161 second associated with ADP release (4,17,25,29). To better understand the effects of 162 regulatory proteins on the mechanics of substeps of the β-cardiac myosin working 163 stroke, we used ensemble averaging of individual binding interactions, which enables 164 the detection of subtle substeps that are typically obscured by Brownian motion 165 (4,30,31). Binding interactions were synchronized either upon the initiation of binding 166 (time forward averages) or the termination of binding (time reverse averages) using a 167 changepoint algorithm and then averaged as previously described (29). The difference 168 between the time forward and time reverse averages is indicative of a two-substep 169 working stroke, and we see a two-substep working stroke for both the regulated and 170 unregulated thin filaments (Figs. 2c-d), consistent with previous studies of unregulated 171 thin filaments (4,25,29). The displacement of the time forward and time reverse 172 averages at detachment gives the size of the total working stroke (29). The difference in 173 displacement between the end of the time forward averages and the beginning of the 174 time reverse averages gives the size of the second substep of the working stroke, and 175 the difference between the total working stroke and the second substep gives the size 176 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 1 0 of the first substep. Here, we constructed cumulative distributions of substep 177 displacements from our individual binding interactions, and the data follow the expected 178 shape of a cumulative distribution for a single Gaussian function (Figs. 2f-g). We did 179 not observe a statistically significant difference in the size of either the first (P = 0.36) or 180 second (P = 0.55) substep of the working stroke in the presence or absence of 181 regulatory proteins. information not only on the mechanics of the working stroke but also the kinetics. The 187 distribution of attachment durations can be used to calculate the actomyosin 188 dissociation rate. Cumulative distributions of binding interaction durations were well 189 fitted by single exponential functions to yield the actomyosin detachment rate (Fig. 3a). 190 Interestingly, we observe that the detachment rate measured in the presence of 191 regulatory proteins (3.9 (-0.3/+0.4) s -1 ) was slightly slower than the rate measured 192 without regulatory proteins (5.5 ± 0.4 s -1 ) (P = 0.002). This is consistent with previous 193 studies of skeletal muscle myosin conducted at low ATP concentrations which show 194 slower detachment rates in the presence of regulatory proteins (20). 195 The assertion that dissociation of regulated thin filaments from myosin is slowed 196 at low ATP is further supported by kinetic information extracted from the ensemble 197 averages. The time reverse averages give the rate of transitioning from the second 198 substep to the detached state, and previous studies have shown that this transition is 199 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 1 5

Discussion 281
We used high-resolution optical trapping techniques to examine the mechanics 282 and kinetics of the β-cardiac myosin working stroke in both the presence and absence 283 of regulatory proteins. We found that regulatory proteins do not tune the mechanics of 284 the working stroke or the load-dependent kinetics of β-cardiac myosin contraction at 285 physiologically relevant ATP concentrations; however, we did observe slight differences 286 in the kinetics of actomyosin dissociation in the presence of regulatory proteins at low 287 ATP, suggesting that these regulatory proteins have subtle effects beyond just sterically 288 blocking the interactions between myosin and the thin filament. 289

290
The molecular role of regulatory proteins in cardiac muscle 291 In the healthy heart, the ATP concentration is ~8 mM (40,41), and even in heart 292 failure, the ATP concentration remains in the millimolar range (1-4 mM) (42)(43)(44). Here,293 we found that at physiologically relevant millimolar ATP concentrations, regulatory 294 proteins have no appreciable effect on the mechanics or kinetics of the β-cardiac 295 myosin working stroke or the kinetics of actomyosin dissociation (Fig. 4b). Consistent 296 with previous studies (7,9,10,45), our stopped-flow measurements show that at 297 physiologically relevant ATP concentrations, the rate of ADP release is much slower 298 than the rate of ATP-induced actomyosin dissociation (Figs. 3b-c), and therefore the 299 muscle shortening speed will be limited by the rate of ADP release. We did not observe 300 any differences in the rates of ADP release (Fig. 3b) or the rate of actomyosin 301 dissociation measured in the optical trap at saturating ATP (k 0 , Fig. 4b) in the presence 302 of regulatory proteins. Although we observed subtle changes in the second-order rate of 303 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 1 6 ATP-induced dissociation between regulated and unregulated filaments (Fig. 3d), these 304 changes are irrelevant at [ATP] > 50 μM where the rate of ADP release limits 305 detachment. Therefore, in the functioning heart, regulatory proteins do not modulate the 306 kinetics of the myosin working stroke. Moreover, our data show that the load-dependent 307 rate of ADP release limits actomyosin dissociation, and this is unchanged by regulatory 308 proteins (Figs. 3b-c and 4). Finally, we show that the mechanics of the working stroke 309 are unchanged by regulatory proteins (Fig. 2). Taken together, our data are consistent 310 with tropomyosin primarily serving a role in sterically blocking the calcium-dependent 311 interactions between cardiac myosin and the thin filament under working conditions in 312 the heart (46,47). 313 Although not relevant to working conditions in the heart, our results reveal that 314 regulatory proteins modulate the kinetics of ATP-induced actomyosin dissociation at low 315 ATP concentrations. To measure the substeps of the working stroke, we conducted 316 optical trapping experiments at very low ATP concentrations not experienced in the cell, 317 since this slows mechanical transitions that are too fast to resolve at physiologically 318 relevant ATP (28). While our results show that regulatory proteins do not change the 319 size of the working stroke or the coupling between kinetics and mechanics ( Fig. 2), they 320 also reveal that regulatory proteins slow the dissociation of actomyosin at very low ATP 321 concentrations (Fig. 3a). The time reverse ensemble averages demonstrate that 322 actomyosin dissociation is limited by the transition from the second substep to the 323 detached state (Figs. 2c-d), which is similar to the second-order rate of ATP-induced 324 actomyosin dissociation measured using stopped flow techniques (Fig. 3b) (29-31). All 325 of these rates (actomyosin detachment rate (Fig. 3a), time reverse ensemble average 326 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 1 7 rate ( Fig. 2c-d), and ATP-induced dissociation rate measured in the stopped flow ( Fig.  327 3b)) are similar, consistent with the rate of ATP-induced dissociation limiting 328 detachment at low ATP. Moreover, for each of these measurements, the rates are 329 significantly slower in the presence of regulatory proteins than in the absence (Fig. 3d). 330 Taken together, these data demonstrate that at low ATP concentrations, regulatory 331 proteins slow actomyosin detachment due to slowed ATP-induced dissociation. 332 While not relevant to physiology, the observation that regulatory proteins can 333 tune ATP-induced dissociation has interesting implications. The tuning of detachment 334 kinetics by regulatory proteins cannot be fully explained by a simple steric blocking 335 mechanism, where tropomyosin only blocks the interactions between myosin and the 336 thin filament (46). The exact mechanism of this kinetic tuning is not known; however, it 337 has been proposed that tropomyosin can interact directly with myosin (48) The effects of regulatory proteins on actomyosin appear to be isoform specific 353 Here, we saw that at physiologically relevant ATP concentrations, tropomyosin 354 and troponin do not appreciably affect the β-cardiac myosin working stroke. 355 Interestingly, the effects of regulatory proteins on myosin appear to depend on the 356 specific protein isoforms used. There are many distinct tropomyosin isoforms expressed 357 in eukaryotic cells, with tissue-specific expression patterns, and even within the same 358 cell, different tropomyosin isoforms localize to different subcellular actin pools (54). 359 Moreover, while all myosin isoforms can associate with actin, they preferentially interact 360 with certain actin structures in an isoform-specific manner (55). The biophysical role of 361 tropomyosin appears to vary with both the myosin and tropomyosin isoform, where 362 some tropomyosin isoforms inhibit the interactions of certain myosin isoforms with actin 363 (55,56), while others promote these interactions (19,57). This isoform specificity has 364 been proposed to help localize specialized motors to specific regions of the cell (58). 365 The molecular roles of tropomyosins on myosin motor function can extend 366 beyond steric effects. For example, decoration of actin with Tpm1p slows the rate of 367 ADP release from the myosin-V isoform Myo2p, and this slowing of ADP release 368 kinetics enables Myo2p to processively walk on tropomyosin decorated actin filaments, 369 something it will not do on bare actin (19). Even within the family of myosin-II motors, 370 there is evidence that regulatory proteins could tune the load-dependent kinetics of 371 myosin in an isoform specific manner. For example, non-muscle myosin-IIA's force 372 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 1 9 sensitivity is increased by tropomyosin Tm4.2 (18). In the case of skeletal muscle 373 myosin, some studies suggested that tropomyosin decreases the step size of myosin by 374 inhibiting the ability of both myosin heads to bind to the thin filament (20), while other 375 studies have not seen this effect (21,22). The working stroke size that we measured for 376 β-cardiac myosin in the presence of regulatory proteins is indistinguishable from that 377 measured in the absence of regulatory proteins with both one-headed and two-headed 378 myosin constructs (4,17,24,25), so we do not believe that our inability to observe a 379 difference in mechanics is related to the two-headed nature of the construct used here 380 (59). 381 382

Conclusions 383
Our results clearly demonstrate that under physiologically relevant ATP 384 concentrations, regulatory proteins do not cause appreciable changes in the mechanics 385 or kinetics of the β-cardiac myosin working stroke; however, they can tune myosin's 386 kinetics at low ATP concentrations, suggesting effects beyond a simple steric blocking 387 mechanism. This has important implications for both our understanding of the 388 mechanism of muscle regulation and mathematical modeling of muscle contraction, 389 which relies on accurate, sensitive measurements of these parameters. 390 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 2 0

Protein expression and purification 392
Cardiac actin and myosin were purified from cryoground porcine ventricles as 393 previously described (10). Human troponin and tropomyosin were expressed 394 recombinantly in E. coli, purified, and complexed as described previously (10). Myosin 395 subfragment-1 (S1) for spectroscopic measurement was prepared by limited proteolysis 396 using chymotrypsin as previously described, and N-(1-Pyrene)Iodoacetamide-labeled 397 actin was prepared as previously described (10 ADP release experiments were performed as described previously (7,45,61). 408 Briefly, 1 μM phalloidin-stabilized, pyrene-labeled actin, 1.5 μM tropomyosin (when 409 appropriate), 1.5 μM troponin (when appropriate), 1 μM S1 myosin, and 100 μM 410 Mg*ADP were rapidly mixed with 5 mM Mg*ATP. This caused an increase in 411 fluorescence that was well fit by a single exponential function, where the rate equals the 412 rate of ADP release (35). Each experiment consisted of 5 technically repeated 413 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 2 1 measurements, and the 3 experiments were used to calculate the mean and standard 414 deviation. A two-tailed Student's t-test was used for statistical testing. 415 The rate of ATP-induced actomyosin dissociation was measured as previously 416 described (61). Briefly, 1 μM phalloidin-stabilized, pyrene-labeled actin, 1.5 μM 417 tropomyosin (when appropriate), 1.5 μM troponin (when appropriate), 1 μM S1 myosin, 418 and 0.04 U/mL apyrase were rapidly mixed with varying concentrations of Mg*ATP. The 419 resultant fluorescence transients were best fit by the sum of two exponential functions, 420 as previously described (35). The amplitude of the fast phase was fixed to prevent 421 artifacts due to the dead time of the instrument. As has been shown before, the rates of 422 the fast and slow phases were hyperbolically related to the concentration of ATP. Data The fast phase of ATP-induced dissociation was modelled as the formation of a collision 426 complex between actomyosin and ATP that is in rapid equilibrium (K 1 ') followed by an 427 irreversible isomerization and rapid dissociation (k +2 '). The rate of the fast phase, k fast , is 428 hyperbolically related to the ATP concentration by: 429 At low [ATP], the second order rate of ATP-induced dissociation is given by K 1 ' * k +2 '. 431 The concentration of ATP was measured spectroscopically for all experiments. The rate 432 of ATP-induced dissociation was measured over a full range of concentrations 3 times, 433 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 2 2 and each time, the fitted parameters were extracted. Reported values are the average 434 of these 3 trials and the error is the standard deviation. Statistical testing was performed 435 using a 2-tailed Student's t-test. 436 437

In vitro motility assays 438
In vitro motility assays were performed as described (10) Experiments were performed on a custom-built, microscope free dual-beam 449 optical trap described previously (29). These experiments utilized the three-bead 450 geometry in which a thin filament is held between two optically trapped beads and 451 lowered on to a surface bound bead that is sparsely coated with myosin (27,28). 452 Tropomyosin was dialyzed into KMg25 buffer the night before the experiment. Actin was 453 attached to beads using a biotin-streptavidin linkage, where actin contained 10% 454 biotinylated actin and polystyrene beads were coated with streptavidin, as previously 455 described (28,29). Flow cells were coated sparsely with beads as previously described 456 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 2 3 (28,29). Flow cells were loaded with myosin (1-7 nM in KMg25 with 200 mM KCl to 457 prevent myosin filament formation) for 5 minutes and the surface was blocked with 1 458 mg/mL BSA for 5 minutes. This was followed by activation buffer. For low ATP 459 experiments, the activation buffer contained KMg25 with 1 μM ATP, 192 U/mL glucose 460 oxidase, 48 μg/mL catalase, 1 mg/mL glucose, and ~25 pM biotin-rhodamine-phalloidin 461 actin. For the high ATP experiments, conditions were identical, except 1 mM Mg*ATP 462 was used. When appropriate, troponin and tropomyosin were also included at 200 nM. 463 The concentration of free calcium was calculated using MaxChelator (60). This was 464 followed by 4 μL of streptavidin beads. Flow cells were then sealed with vacuum grease 465 as previously described and data were collected within 90 minutes of sealing (29). 466 Surface-bound beads were probed for binding interactions using small 467 oscillations of the stage position that were stopped once data collection began. Data 468 were collected at 20 kHz and filtered to 10 kHz according to the Nyquist criterion. For 469 each bead-actin-bead assembly, the trap stiffness was calculated from fitting of the 470 power spectrum as previously described (28). 471 472

Implementation of the isometric optical clamp feedback 473
Here, we used an all-digital implementation of an isometric feedback clamp (36). 474 In an isometric optical clamp feedback experiment, the position of one bead (the 475 transducer) is continuously sampled, and deviations from its original setpoint position 476 are compensated for by moving the second (motor) bead using acoustic optical 477 deflectors (AODs, Gooch and Housego). The positions of the beads were recorded 478 using quadrant photodiodes, the feedback calculations were digitally performed on a 479 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 2 4 field programmable gate array (FPGA) board (National Instruments PCIe-7852), and the 480 laser controlling the motor bead was translated using AODs. 481 The error signal used for the feedback, V t , is the time filtered positional error for 482 the current sample period given by: 483 where K p is the user defined proportional gain, K i is the user defined integral gain, E t is 484 the current sample's absolute error from the setpoint, and W is the user defined  The time constant for the feedback response time was set as previously 491 described (36). Briefly, a bead-actin-bead dumbbell was held in the dual beam traps, a 492 square wave was injected into the transducer bead channel, and then movement of the 493 motor bead by the feedback system was monitored. The proportional and integral gains 494 were empirically adjusted to give a response time of ~5 ms without introducing 495 oscillations into the system. 496 497

Analysis of single molecule data 498
All data from optical trapping experiments were analyzed using our custom-built 499 MATLAB program, SPASM, as previously described (29). Briefly, binding interactions 500 between myosin and the thin filament were identified using a peak-to-peak covariance 501 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 2 5 threshold, and the initiation and termination times of binding were identified using a 502 changepoint algorithm. Data traces were excluded if the separation between the bound 503 and unbound populations of the covariance histogram was not well defined. To improve 504 the signal-to-noise ratio, the signal from both beads was summed and divided by 2 as 505 previously described (29). Ensemble averages and histograms of binding interactions 506 were generated as previously described (28-30). Ensemble averages were fit by single 507 exponential functions in MATLAB until the signal plateaued. Cumulative distributions of 508 step sizes and attachment durations were fit with cumulative functions for the Gaussian 509 and exponential functions, respectively. Statistical testing for normality was done using 510 a Shapiro-Wilk test. Statistical testing of the step sizes was done using a 2-tailed 511 Student's t-test of individual binding interactions. 95% confidence intervals for the 512 detachment rate were calculated by bootstrapping of individual binding interaction 513 durations, and statistical significance was calculated according to (62). 514 In the optical trap, actomyosin remains attached until ADP is released and ATP 515 induces actomyosin dissociation (35): 516 517 The attachment duration, t on , is given by 518 At low ATP, the detachment rate, k det , will be dominated by the second order rate of 519 ATP binding, while at saturating ATP, the detachment rate will be limited by the rate of 520 ADP release. 521 For isometric optical clamp experiments, data were collected at saturating ATP 522 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 2 6 concentrations, where the rate of ADP release limits the rate of actomyosin dissociation 523 (63). Binding interactions were identified using a variance threshold, set by the position 524 of the transducer bead, and the force exerted by the motor bead and the attachment 525 duration were measured. The relationship between the force, F, and the load dependent 526 detachment rate, k(F) was modeled using the Bell equation (38): 527 Where k 0 is the rate of the primary force sensitive transition in the absence of force, d is 528 the distance to the transition state, and k B *T is the thermal energy. The distribution of 529 attachment durations is exponentially distributed at each force, and therefore follows the 530 following probability density distribution (37): 531 Maximum likelihood estimation (MLE) was used to determine the most likely values of k 0 532 and d, as previously described (37). 95% confidence intervals for parameter values 533 fitting were determined using 1000 rounds of bootstrapping simulations, and hypothesis 534 testing was performed using the difference in the means and the variances of these 535 distributions as previously described (62) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. packages used include the following: "readr_2.1.3", "readxl_1.4.0", "data.table_1.14.6", 551 "here_1.0.1", "ggpubr_0.5.0", "ggtext_0.1.2", "purrr_0.3.5", "tidyr_1.2.1", "dplyr_1.0.10", 552 "tibble_3.1.8", "cowplot_1.1.1", and "ggplot2_3.4.0". 553 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint 2 9 CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

Figure Legends
The copyright holder for this preprint this version posted January 10, 2023. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023.

696
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023.

744
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023.

840
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted January 10, 2023. ; https://doi.org/10.1101/2023.01.09.522880 doi: bioRxiv preprint