Three variants of Cdh23 as model protein. We chose the first extracellular (EC1) domain of Cdh23 as the model protein. Cdh23 EC1 consists of nine β-strands interconnected by reverse β turns, 310 α-helix, and random coils (Fig. 1C)29,30. We used three variants, Cdh23 EC1 (S47), Cdh23 EC1(V47), and Cdh23 EC1(P47), with different native packing densities, long-range H-bond networks, and different cross-correlated crankshaft types motions among β-strands30 (Fig. 1D). Cdh23 EC1(S47) is a wildtype variant conserved for a majority of the species, including Homo sapiens (Hs), whereas Cdh23 EC1 (V47) is another wildtype variant conserved in lower order vertebrates like Callorhinchus milii (fish), Gekko japonicus (reptiles), anser cygnoides domesticus (Swan goose), Alligator mississippiensis (crocodile reptile), Gallus gallus (ave)30. Evolutionarily these species may be in a lower order than sapiens. However, some of these species possess better hearing sensitivity at a lower frequency range than humans31,32. Further, some of these lower vertebrates require hearing at low-air pressure (at high altitude for swans) or under-water pressure (for alligators) where the noise threshold is high33. Moreover, the evolutionary trend for proteins may not correlate with the ranks of the expressors. In line, Cdh23 EC1(V47) possesses the highest number of native contacts, long-range H-bonds, and thus most robust cross-correlated motions among β-strands 30(Fig. 1D). The last variant is Cdh23 EC1(P47) which is a mutant-variant of Cdh23 EC1(S47) that features a progressive hearing loss (PHL) phenotype in mice34. PHL is an aggressive form of hearing-loss with aging where a patient suffers complete hearing loss at a very early age (less than 20 years in humans)31,35. Regarding native packing, Cdh23 EC1(P47) ranks last among the three variants (Fig. 1D,1E). Reportedly, V47 shows the highest resistance against thermal and chemical denaturants, whereas P47 is the least30. Serine to proline isn’t unique for Cdh23, relatively abundant with phenotypes36. Even for mechanosensitive Titin proteins, Ser22 to Pro mutation is reported with a phenotype of cardiomyopathy37.
Covalent tethering of proteins and MT. To monitor the responses of protein variants to small tensile forces, we force-clamped the chimeric polyprotein constructs using MT. A detailed description of the MT including hardware, Instrument software and analysis software are described in the methods and supplementary Figs. 1, 9–13. Two repeats of Cdh23 EC1 variants were sandwiched between I27 constructs, a trimer of I27 domains at the C-terminus and a monomeric I27 domain at the N-terminus (Fig. 1). The trimer of I27 domains was covalently attached to the glass-coverslip, and the I27 monomer at the N-terminus was attached to the paramagnetic bead, both covalently using sortase-mediated enzymatic stapling as described previously 38 (methods). The I27 domains serve as a spacer, and a standard marker for specific pulling as the mechanical stability of I27 is significantly higher than all three variants of Cdh23 EC1 14,15(Supplementary Fig. 3). The assymmetry in the constructs is primarily to ease out the DNA recombinant process. We clamped the protein variants to varying forces for 5 minutes from a resting force of 4 pN and monitored the change in lengths (ΔL) from the jumps in bead positions in real-time (Supplementary Fig. 4–5). Notably, all protein variants, featuring the behaviour of Worm-like chain (WLC) polymer, showed a gradual increase in the end-to-end extensions, x(F), with force 39 (Fig. 4D, Supplementary Fig. 4). x(F) defines the extent of unfolding and, thus marks the final denaturant states at the respective clamping forces (Supplementary Fig. 4–5).
Number of long-range interactions, more precisely electrostatic H-bond interactions, follow decreasing order from V47 to S47 to P47 variants (Fig. 1, supplementary Fig. 6). Reportedly, the chemical and the thermal stabilities also followed the decreasing order from V47 to P47 variant30. Accordingly, we measured the weakest force-resistance for Cdh23 P47 featuring force-induced extensions between 5 pN – 15 pN, followed by Cdh23 S47 between 13 pN − 25 pN, and strongest for Cdh23 V47 between 19 pN − 38 pN (Fig. 2, Supplementary Fig. 5, 7). Irrespective of the force- resistance, we identified multiple microstates for all the variants during unfolding-refolding transitions at small clamping forces. To characterize the microstates, we measured the length-change (ΔL) for each step and plotted them as distributions (Methods and Fig. 2). The histograms of ΔL for all variants followed comparable trend, narrow distributions at lower forces, wide at intermediate forces reaching saturations, and finally sharp distributions at very high forces. The widths of the distributions indicate the extent of reversible transitions among microstates under tension while the number of peaks in the distributions infers the number of microstates within the spatial limit of the MT.
To further highlight the heterogeneity in the microstates across variants, we plotted the fraction of states or probability of states as they survive at the clamping forces (Fig. 3). For simplicity, we used only 3-states, a native or folded state (with 0 nm ≤ ΔL ≤ 15 nm), a complete denatured or unfolded state (45 nm ≤ ΔL, maximum x(F) is obtained from Fig. 4 unfolding length data), and a cluster of intermediate states (15 nm ≤ ΔL ≤ x(F)). The probability of the states at respective force-clamps was estimated from the dwell time of each state, normalized to the total clamping time. To quantitatively estimate the dwell time of steps, we performed ‘Autostepfinder’, a widely used step-finder protocol that is based on the mean standard deviation model 40. We noticed a sigmoidal transition for the probability of folded state with force for all variants. The transitions from native to unfolded states occurred at different critical forces (Fcrit), lowest for P47 (8.8 ± 0.2 pN) and largest for V47 (25.7 ± 0.2 pN). However, strikingly we noticed different widths (ΔF) or slopes of the transitions across variants. We highlighted these regions with light green boxes (Fig. 3). The green boxes marked the co-existence of all 3 states, indicating the force-range at which the protein-variants undergo reversible transitions between numerous conformational states. While the critical force-range (Fcrit) of transition is a measure of force-resistance, the width of the transitions (ΔF) measures heterogeneity in the microstates during the transition from folded to unfolded states. More the width, more the heterogeneity, and hence more the shock-absorptivity or force-dissemination. Together, the critical force-range of transitions and the width measures the force-adaptation range. Clearly, the V47 variant is not only the most resistant to tensile forces among the three, it also possesses maximum heterogeneous states during phase transitions whereas P47 has least.
Next, we mapped the probability of the open state of all three variants and fit the rise in states to linear equation, excluding the force-value points with no unfolding probability. The slope gives quantitative estimation of protein’s unfolding cooperativity. Conceptually, a completely cooperative unfolding will show a strict two-step transition with an infinite slope, whereas the completely non-cooperative transition will have slope of 0. Here, the slope value for P47 is 0.15 ± 0.01/pN, S47 is 0.09 ± 0.01/pN, and V47 is 0.05 ± 0.01/pN. The order of unfolding cooperativity is, P47 > S47 > V47. Overall, our data from force-clamp experiments indicate that Cdh23 V47, which contains the most densely packed network of intra-domain interactions, is mechanically the most stiff yet most disseminating of tensile forces. We, therefore, infer that higher crankshaft or contact order transforms β-rich proteins to more malleable under tensile forces, however, without causing any permanent damage. The inter β-strand interactions serve as shock-absorber and resist elongation from mechanical inputs.
Effect of intra-domain contacts in folding rates and force tolerance. Contact-orders contribute to the folding dynamics of proteins. Further, mechanoresponsive proteins undergo unfolding-folding periodically under tension in physiology. The folding dynamics of the mechanoresponsive proteins are, thus functionally important. We, therefore, set to decipher the effect of inter domain interactions in the folding and unfolding kinetics.
We performed a dual-step force-clamp spectroscopy with a monomer of Cdh23 EC1 domain parsed between I27 domains as -Nterm-(I27)3-EC1-I27-Cterm (Fig. 5). The use of a monomeric domain enabled us to measure the kinetics for direct transitions from the native state to the complete denatured state and vice versa. The resting force was set to 4 pN. Subsequently, the protein variants were individually clamped at varying constant forces (19–38 pN for S47 variant, 11–19 pN for P47 variant, and 23–38 pN for V47 variant) for 30–60 seconds and monitored the survival time till complete unfolding. We excluded the extensions below 5 nm in the analysis due to the resolution limit in our measurements (Supplementary Fig. 2). After each force-clamp, we subjected the variants to a high tensile force of 38 pN for 15 seconds to ensure complete extension prior to refolding. For refolding, we clamped the proteins at low forces, 11 − 4 pN for S47 variant, 8 − 4 pN for P47 variant, and 19 − 9 pN for V47 variant, respectively. We completed the cycle by quenching the force to 4 pN. For each variant, we monitored a minimum of 30 different beads (Supplementary Table 1).
The survival probability of the folded and unfolded states was estimated from the dwell time (Fig. 5). We deduced the lifetimes of folding and unfolding from the single exponential decay fit to corresponding survival probabilities (Fig. 5B) (Supplementary Table 1) and obtained the force-induced on-rate and off-rate data (Fig. 5D). Subsequently, the variation in rates with force were fit to Bell’s model41 and obtained the intrinsic kinetic parameters, \({k}_{0}^{f/uf}\) (intrinsic-rate at zero force, ‘f’ represents folding and ‘uf’ refers to unfolding) and \({x}_{β}^{f/uf}\) (the width of the potential energy barrier) (Fig. 5C and Supplementary Table 2).
Time-dependent adaptation against fatigue while receiving repetitive force pulse:. The time-dependence of a mechanoresponsive protein is often correlated with the molecular-fatigue which arises from the repetitive stretch and release cycle of the protein42,43. Cdh23 too experiences repetitive force perturbations from the sound. It is thus interesting to monitor the molecular fatigue of Cdh23 and its relation to force-adaptibility. We, therefore, exposed the protein variants to a cycle of high force and low-force (4 pN), repetitively. The proteins were completely unfolded for 15 seconds at the respective critical unfolding forces (For S47 protein 23 pN, P47 protein 15 pN, and V47 protein 38 pN) and then refolded at a refolding force of 4 pN and waited for 15 seconds (Fig. 6A). The critical forces of unfolding and refolding were selected based on the data from the chevron plot of folding-rates and the probability of states plot. (Fig. 5C).
We noticed two distinct unfolding patterns, direct unfolding (unfolds directly along with the initial stretch) and delayed stepwise unfolding (goes through the metastable states before complete unfolding) (Fig. 6A). Counterintuitive to the folding-kinetics pattern, we observed the more frequent direct unfolding of P47 (20.7%) than S47 (11.3%) and V47 (6.25%), respectively. It is thus, logical to relate the direct unfolding events with the delay in folding in the previous force-cycle, a feature of molecular fatigue. Overall, we infer that proteins with low force- adaptibility, suffer faster mechanical fatigue. More importantly, a direct unfolding relates to ‘no force-buffer’ and conveys the input mechanical signal directly, exposing the sensory organs to potential damage.