The distinctive mechanical and structural signatures of residual force enhancement in myofibers

Abstract

In muscle, titin proteins connect myofilaments together and are thought to be critical for contraction, especially during residual force enhancement (RFE) when force is elevated after an active stretch. We investigated titin’s function during contraction using small-angle X-ray diffraction to track structural changes before and after 50% titin cleavage and in the RFE-deficient, mdm titin mutant. We report that the RFE state is structurally distinct from pure isometric contractions, with increased thick filament strain and decreased lattice spacing, most likely caused by elevated titin-based forces. Furthermore, no RFE structural state was detected in mdm muscle. We posit that decreased lattice spacing, increased thick filament stiffness, and increased non-crossbridge forces are the major contributors to RFE. We conclude that titin directly contributes to RFE.

One-Sentence Summary:

Titin contributes to active force production and residual force enhancement in skeletal muscles.

Animal movement through complex environments requires unique mechanical features of not only isometric or shortening (concentric) muscle contractions, but also lengthening (eccentric) contractions (1). Eccentric contraction produces immediate and rapid force enhancement at the level of the sarcomere that is greater than that possible during isometric or concentric contractions and is poorly described by current muscle models (2, 3). Force enhancement improves functional tasks such as counter-movement jumps (4), downhill braking (5), and joint stabilization when negotiating complex terrain (6). Clinically, eccentric-focused training accelerates muscle hypertrophy above conventional training, and is widely used for athletes and also for restoring mobility independence to older adults (7, 8).

During eccentric contraction, both crossbridge and non-crossbridge structures are stretched and produce viscoelastic forces that contribute to force enhancement (9). It has been suggested that crossbridge-based force enhancement plateaus after a short stretch of ~28 nm per sarcomere (9, 10), while further stretch increases force via non-crossbridge, viscoelastic “spring” elements that increase stiffness upon activation (9, 11, 12). After stretch, force enhancement from crossbridges dissipates quickly while the non-crossbridge elastic component remains, leaving a long-lasting enhancement of force above that of pure isometric contractions at the same final length, the so-called residual force enhancement (RFE)(13). The mechanisms underlying RFE are unclear but mounting evidence suggests that the titin protein plays a key role (Fig. 1A). Titin, the largest known protein, extends from the Z-disk to the M-line, where it runs along the thick filament in the A-band, and is extensible in the I-band. Titin extends during sarcomere stretch and becomes 4–6 times stiffer upon Ca²⁺ activation (14–18) via as yet unresolved mechanisms (19). Increasing titin-based forces during passive stretch (20) and submaximal activation (21, 22) also increases the Ca²⁺ sensitivity of force with increasing sarcomere length, implying that increased crossbridge recruitment may also contribute to RFE.

Fig. 1. X-ray diffraction of TC fiber bundles before and 50% titin cleavage during Iso_S, Iso_L, and Iso_R conditions.

(A) Schematic of skeletal half-sarcomere with relevant structural periodicities, I-band titin segments, and TEV_P cleavage site indicated. (B-C) Tension traces of fibers during mechanical experiments before (B) and after (C) 50% titin cleavage. Active tension is total tension during contractions, minus the passive tension at those lengths. (D) Passive stiffness after titin cleavage, normalized to its paired pre-stiffness value at Iso_S. (E) Tension before (blue) and after (red) titin cleavage for passive, active, and total conditions. (F) transmission electron micrographs of TC sarcomeres after full mechanical protocols with sham (−TEV_P) or treatment (+TEV_P) conditions. (G) RFE in fibers before and after titin treatment. (H) Representative X-ray diffraction pattern of skeletal psoas fibers, with labeled reflections indicating relevant periodic structures that are referenced in (A). Connecting letters: different letters are significantly different (post hoc analysis P < 0.05). Data throughout reported as mean ± s.e.m. Full statistical details in Table S1. letters are significantly different (post hoc analysis P < 0.05). Data throughout reported as mean ± s.e.m. Full statistical details in Table S1.

We investigated titin’s function during skeletal muscle contraction using small-angle X-ray fiber diffraction to track structural changes before and after cleavage of 50% of I-band titin in mouse fast-twitch muscle fibers, and in fibers from ‘muscular dystrophy with myositis’ (mdm) mice which lack force enhancement (23). Specific cleavage of I-band titin was achieved by using heterozygotes of a transgenic ‘titin cleavage’ (TC) mouse model (Fig. S1), in which titin is controllably cleaved via an embedded tobacco etch virus protease (TEV_P) cleavage site (17, 24). Mdm carries a complex titin mutation that manifests as a small deletion in the titin gene (25, 26). Our results demonstrate that, during RFE, in addition to increased tension, there are structural-based X-ray diffraction signatures that are distinct from those of pure isometric contraction. Titin cleavage attenuates RFE tension, blunts RFE-associated diffraction signatures, and increases disorder in the myofilament lattice. We further report that when RFE is absent, as in mdm mice, the structural signature of RFE is also absent. Collectively, we posit that RFE is caused, in part, by decreased lattice spacing, stretched inter-myofilament bridge elements (e.g., titin, myosin-binding protein C [MyBPc]), and increased thick filament stiffness, with titin a major regulator of each.

Results and Discussion

Titin cleavage reduces mechanical stability

We conducted RFE experiments on permeabilized heterozygote TC psoas fibers before and after 50% titin cleavage via TEV_P incubation. X-ray diffraction patterns and force traces were obtained under three conditions: isometric contraction at 2.7 μm sarcomere length (SL) (short isometric contraction; Iso_S), isometric contraction at 3.0 μm SL (long isometric contraction; Iso_L), and stretch-hold contraction from 2.7 to 3.0 μm SL (isometric contraction after an active stretch [RFE]; Iso_R). Compared to pre-titin cleavage, post-cleavage fibers produced less tension, but the shape of the tension traces was similar (Fig. 1B–C). For Iso_S and Iso_L, 50% titin cleavage reduced passive stiffness (measured here by sinusoidal oscillations; Fig. 1D), passive tension, active tension, and total tension (P < 0.01, Fig. 1E; Table S1), but the relative difference after treatment was always greater for Iso_S vs. Iso_L (Fig. S2A). Titin-based force and stiffness are known to be critical for thick filament centering in the A-band, with greater stability at longer SLs where titin-based forces are higher (17, 27). Indeed, at a near-slack SL of 2.4 μm, the A-bands of titin-cleaved fibers fell apart during maximal activation (pCa 4) (17), but were maintained here at longer SLs and at submaximal activation (pCa 6; Fig. 1F). Furthermore, RFE (increase in total force from Iso_L to Iso_R) was reduced after titin cleavage (P < 0.01; Fig. 1G, S2B) from 30.69 ± 3.35% to 17.44 ± 2.24% of reference force (Iso_L). We converted RFE to force per thick filament (f_TF), assuming that all forces are transmitted to the thick filament, and that in samples with a d_1,0 = 37.69 nm ~80% of the cross-sectional area is occupied by myofibrils, resulting in a thick filament density of ~488×10⁶ thick filaments/mm². The active tension increased from 48.26 kPa in Iso_L to 68.53 kPa in Iso_R, resulting in f_TF values of 98.95 pN and 140.51 pN for Iso_L and Iso_R, respectively. Therefore, compared to Iso_L, Iso_R increased by 41.56 pN. These results are consistent with numerous previous studies that provided indirect evidence for titin’s involvement in RFE (28–30), and with the absence of RFE in the mdm titin mutant (31, 32). These results demonstrate a direct relationship between I-band titin and RFE.

Lattice structure and order during Iso_R are structurally distinct

We assessed structural changes using small-angle X-ray diffraction (Fig. 1H), which provides information concerning the underlying structural arrangements of sarcomeric proteins (Fig. 1A). We compared diffraction patterns from pure isometric contractions at two different sarcomere lengths (Iso_S vs. Iso_L) and contractions with ramp stretch- hold from the short to the longer length (Iso_L vs Iso_R). Lattice spacing (LS) was calculated as the separation of the 1,0 equatorial reflections from thick filaments (D_1,0 plane; Fig 1H). Computational modeling studies suggest that increasing LS decreases force production and blunts length-dependent activation (LDA) by affecting crossbridge kinetics (33). LS is controlled in part by radial forces imposed on the myofilament lattice by titin (34, 35) and expands in relaxed fibers after titin cleavage (22). Furthermore, myopathic human fibers with abnormally stiff titins have smaller LS than controls (36). We report that LS decreased from Iso_S to Iso_L, as expected, but decreased further in Iso_R (P < 0.0001; Fig. 2A, S3; Table S2). The “excess” lattice shrinkage in Iso_R, noted previously (37), suggests that radial forces acting on the myofilament lattice are enhanced after active stretch, suggesting increased titin-based forces in Iso_R with respect to Iso_L. LS is an important regulator of crossbridge kinetics and therefore cross-bridge based force production (33), especially in permeabilized preparations, where the lattice is generally expanded compared to intact, and shrinks more rapidly upon stretch. Compared to Iso_L, the Iso_R lattice was smaller by 1.08% (0.41 nm). Based on previous assessments (33), we estimated a small but significant RFE contribution of ~5.2% force, or f_TF ~ 5.14 pN.

Fig. 2. Sarcomeric structural parameters of TC fibers before and after 50% titin cleavage.

D_1,0 (A), I_1,1/I_1,0 (B), σ_A (C), σ_D (D) S_M3 (E), I_M3 (F), S_gActin (G), S_T3 (H), and S_M6 (I) were recorded before (blue) and after (red) 50% titin cleavage, at three conditions: Iso_S, Iso_L, and Iso_R. Connecting letters: different letters are significantly different (Tukey HSD P < 0.05). Full statistical details in Table S2 and Table S3. Intra-sample pre-post differences in Fig. S3.

Structural evidence of crossbridges suggest no differences between Iso_R and Iso_L

We next considered whether more crossbridges are present during Iso_R vs. Iso_L. Previous reports using structural, mechanical or ATP consumption assays suggested no differences in crossbridge recruitment between Iso_L and Iso_R (9, 37–42). The equatorial intensity ratio (I_1,1/I_1,0) is widely used as a measure of the transfer of mass (i.e., myosin heads) from the thick to thin filaments, with increasing I_1,1/I_1,0 indicating more myosin heads associated with the thin filaments (43). We found that before titin cleavage, I_1,1/I_1,0 remained consistent across contraction conditions (P = 0.18; Fig. 2B, S3;Table S2). At longer SLs, I_1,1/I_1,0 decreased as filament overlap decreased (44). Iso_L has less thick and thin filament overlap compared to Iso_S, so I_1,1/I_1,0 would be expected to decrease if nothing else changed. Therefore, the finding that I_1,1/I_1,0 is not significantly different between short and long SL contraction conditions suggests a relatively larger degree of crossbridge recruitment at longer SLs (i.e., Iso_L and Iso_R), as expected via myofilament LDA (45). These findings also imply that there is no further crossbridge recruitment in Iso_R vs. Iso_L, a finding that corroborates other studies using different methodologies (9, 38–40).

Titin cleavage by 50% led to an overall increase in I_1,1/I_1,0 across contraction conditions, with a large increase in the spread of the values including some with unrealistically large values greater than the rigor state (2–3; ⁴³) (P < 0.0001; Fig. 2B; Table S2), which was initially puzzling. These results are likely due to increased lattice disorder after titin cleavage as evidenced by increased σ_A, an indicator of thick filament disorder across adjacent myofibrils (46), and σ_D, a measure of LS heterogeneity, after titin cleavage (P < 0.001; Fig. 2C–D; Table S2). Because we cannot easily uncouple the effects of lattice disorder (44, 47) and mass shift due to myosin head movement on I_1,1/I_1,0, it would be unwise to interpret increased I_1,1/I_1,0 after titin cleavage as indicating more transfer of myosin heads towards actin. However, in contrast to the equatorial reflections, meridional patterns are less affected by lattice disorder. Myosin head configuration can be evaluated via the spacing (S_M3) and intensity (I_M3) of the M3 myosin meridional reflection. Increases in S_M3 are typically associated with increasing crossbridge recruitment, or the reorientation of myosin heads in a position that increases the chance of attachment (48). Before titin cleavage, S_M3 varied among conditions as follows: Iso_S < Iso_L = Iso_R (P < 0.0001; Fig. 2E, S3; Table S2), similar to I_1,1/I_1,0 before titin cleavage, both suggesting increased crossbridge recruitment at longer SL, but unaffected by previous active stretch. In comparison, titin cleavage reduced S_M3 across all conditions (P < 0.0001; Fig. 2E, S3; Table S2). Decreased S_M3 across conditions after 50% titin cleavage suggests that titin is important in modulating myosin head action not only in passive muscle, but also during contraction—a point suggested in muscles from mice with titin mutations but never before shown by cutting of titin springs within a sample. Additional support for this notion comes from I_M3, which provides additional details about variation in myosin head orientation. Before titin cleavage, I_M3 decreased from Iso_S to Iso_L (P < 0.001; Fig. 2F; Table S3) as expected due to decreased filament overlap (48, 49). However, I_M3 for Iso_R > Iso_L = Iso_S (P < 0.001; Fig. 2F), even though they are at the same SL, suggesting that the orientation of heads is more ordered in Iso_R. Of note, after titin cleavage, I_M3 was reduced in all conditions to similar values, suggesting that titin-based forces are partially responsible for myosin head order, and the conditional effect with passive or active stretch.

The thin filaments also provide structural information indicative of crossbridge recruitment (50). In passive cardiac and skeletal muscle, thin filaments extend and troponin complexes reorient with sarcomere stretch in a way that has been linked to increasing Ca²⁺ sensitivity (20, 22). The A6 (S_A6) and A7 (S_A7) spacings report on the left- and right-handed actin helical structures within the thin filament (Fig. S4;Table S3) and are used here to calculate the axial spacing of the actin monomers (S_gActin) (51, 52), where S_gActin can be used as a measure of thin filament extension (53, 54). Also relevant for this discussion are structural changes in the troponin complex, indicated by changes in the spacing (S_T3) of the third troponin meridional reflection T3 arising from the troponin complexes spaced every ~37 nm along the thin filaments. We found that before titin cleavage, both S_gActin and S_T3 are Iso_S < Iso_L = Iso_R (P < 0.01; Fig. 2G–H; Table S3). Iso_L = Iso_R is noteworthy because in purely isometric or passive conditions, the thin filament acts as a stiff Hookean spring so that changes in thin filament strain can be directly interpreted as changes in the total (passive + active) force exerted on the filament (22, 50), as observed from Iso_S to Iso_L; Fig. 1E). However, while Iso_R forces are larger than Iso_L, the thin filament strains are equal. This suggests that something distinct is occurring during eccentric contraction that decouples force and thin filament strain and may suggest that crossbridge-based force is decreased while non-crossbridge based force is increased. Another interesting result is that the S_T3 is altered by thin filament length, as expected for a thin filament associated protein. However, ANCOVA showed that S_T3 decreases independently of S_gActin (P < 0.001; Fig. 2H; Table S3), although the physiological relevance of this independent S_T3 action is unknown.

Increased titin-based forces may not necessarily result in increased thin filament strain because titin’s tether points are far away from the tip of the thin filament; titin’s permanent interaction with the thin filament is close to the Z-disk (55, 56), and other potential alternative interaction sites are at or around the N2A region in the I-band (57, 58). Although a definitive mechanism remains elusive, another possibility is disruption of MyBPc thick-thin bridges during the eccentric phase of the Iso_R condition. MyBPc bridges are purported to be important for contraction (22, 59–61) and seem to be forcibly ruptured by a quick passive stretch (22). Therefore, a role for MyBPc to explain the Iso_R seems plausible and could be directly studied using an inducible “cut and paste” MyBPc mouse line (62).

Thick filament strain suggests titin-based forces are enhanced in Iso_R vs. Iso_L

Arguably the most contentious debate in the RFE field is whether titin contributes to RFE by producing more force during Iso_R vs. Iso_L. The hypothesis is as follows: upon activation, titin stiffness increases by 4–6 times (15, 17, 18, 39), which would imply that during and after eccentric contractions, titin-based forces would be greater than if just activated at the longer SL. Mechanically, titin-based force pulls on and stretches the thick filament, with increasing titin force extending the thick filament (21, 22, 34, 63). We found that thick filament strain, quantified via the spacing of the M6 meridional reflection (S_M6), increased progressively as follows: Iso_S < Iso_L < Iso_R (P < 0.0001; Fig. 2I, Table S2), suggesting higher titin-based force in Iso_R vs. Iso_L. Additionally, 50% titin cleavage slightly reduced S_M6 across all conditions (P < 0.0001 for both; Fig. 2I; Table S2) but the general relationship among conditions was preserved. These data indicate titin as a key contributor to thick filament strain.

In addition to titin, MyBPc can also apply a pulling force to strain the thick filament. To date, MyBPc plays an unclear role during contraction, but can also function like a spring and store some level of force (61), so storing force during an eccentric stretch seems plausible. On the other hand, MyBPc is relatively short and so would most likely rupture and reattach to the thin filament during the ~150 nm per half thick filament eccentric stretch conducted here, minimizing its contribution to force exerted on the thick filament during Iso_R, although some contribution cannot be discounted. Apart from MyBPc, crossbridge forces during contraction also strain the thick filament (63). However, our analysis above, along with other mechanical and biochemical studies (9, 38–40), suggest no differences in the degree of crossbridge recruitment between Iso_L and Iso_R so that the contribution of crossbridge-based thick filament strain should be similar between Iso_L and Iso_R. Based on our analysis, we postulate that the thick filament strain in Iso_R relative to Iso_L is primarily due to increased titin-based force after active stretch.

We next used the thick filament strain data to provide estimates of f_TF during Iso_R, which also includes the LS-dependent effects described above. Compared to Iso_L, the thick filament in Iso_R is strained 0.0694% more, with 41.56 pN more f_TF. The excess f_TF is a product of both the forces acting to strain the thick filament, such as parallel elastic elements (e.g., titin and possibly MyBPc), and the changes in force that arise primarily from stiffening of the thick filaments when they are stretched (64, 65). Approximating changes to thick filament stiffness (Stiff_TF) is a non-trivial task, as thick filament stiffness is nonlinear (21). To accomplish this, we derived f_TF as a function of percent thick filament elongation (e_p) using active tension-thin filament strain data from fast twitch mouse muscle (21), and converted to f_TF - thick filament strain. The data were described well by a third-order polynomial: f_TF = 2.98*10⁻⁸e_p + 709e_p³, simplified to f_TF = 709e_p³, with the Stiff_TF-strain relationship equal to the derivative of f_TF, Stiff_TF = 2127e_p². We next calculated the e_p values for Iso_L and Iso_R from S_M6 values as follows: passive S_M6 at 3.0 μm SL is 7.205 nm (21, 22, 34, 63), and the Iso_L and Iso_R S_M6 = 7.247 and 7.252 nm, respectively. Therefore, from the passive state, the thick filaments are strained 0.586% and 0.655% during Iso_L and Iso_R, respectively. Plugging the e_P values into the Stiff_TF-strain equation produces 729.7 and 912.9 pN/e_p for Iso_L and Iso_R, respectively. Therefore, the thick filament stiffness increases by 25.1% from Iso_L to Iso_R, which also reflects an up to ~25.1% increase in f_TF (66). However, the exact relationship between Stiff_TF and f_TF is somewhat less than linear, depending on the properties of thick filament compliance. Because the relationship is not yet well defined (64, 65), we used a conservative lower bound of ~0.5 Stiff_TF to f_TF ratio. Said another way, ~12.4 pN up to a maximum of ~24.8 pN of the measured RFE per thick filament (f_TF ~ 41.6 pN) can be associated with mechanical stiffening of the thick filament.

A loss of RFE also leads to a loss of distinctive diffraction signatures during Iso_R

As described above, we characterized distinctive diffraction signatures during Iso_R, but an open question was whether this structural state is required for RFE. To assess this point, we utilized the muscular dystrophy with myositis (mdm) mouse, which to our knowledge is the only skeletal mouse model in which RFE can be induced in wild-type but not in HOM mdm (23, 32). We assessed whether the structural indicators of the RFE state were also missing in HOM skeletal muscles. Experiments were run similarly to the TC experiments, but at different SLs for the three contraction types: isometric 2.4 μm SL (mdm Iso_S), isometric 3.2 μm SL (mdm Iso_L), and RFE (mdm Iso_R; stretch-hold from 2.4 to 3.2 μm SL). Compared to TC fibers before titin cleavage, WT mdm fibers showed similar diffraction signatures of RFE for LS, S_M6, I_1,1/I_1,0, σ_A, S_M3, and S_A6 (Fig. 3A–F; Table S4). In contrast, mdm fibers showed no statistical difference in any of these parameters between Iso_R and Iso_L except for I_1,1/I_1,0. Here, as in the TC experiments after titin cleavage, HOM mdm produced unusually large I_1,1/I_1,0, with associated increases in σ_A, suggesting reduced thick filament stability in HOM mdm fibers, and so inflated I_1,1/I_1,0 independent of myosin head mass movement (see discussion above). Based on these data, it is reasonable to propose that there is a causative relationship between the distinctive diffraction signature during Iso_R and the presence of RFE.

Fig. 3. Titinopathic mdm fibers produce no distinctive RFE condition.

D_1,0 (A), S_M6 (B), I_1,1/I_1,0 (C), σ_A (D), S_M3 (E), and S_A6 (F) were recorded for WT (blue) and mdm (red) EDL fiber bundles at three conditions: Iso_S, Iso_L, and Iso_R. Connecting letters: different letters are significantly different (Tukey HSD P < 0.05). Full statistical details in Table S4.

A stiffer titin in Iso_R vs. Iso_L contributes to RFE

The underlying mechanism(s) of RFE, here f_TF = 41.6 pN above Iso_L tension, have long been unresolved. With this X-ray diffraction dataset, we thus far estimated two mechanisms that contribute to RFE: shrinking LS and increased thick filament stiffness (f_TF ~ 12.4–24.8 pN). This implies that there is still ~16.8–29.2 pN of f_TF unaccounted for. This most likely comes from the stretch of non-crossbridge myofilament bridge proteins. Titin is thought to become stiffer upon activation, and we can calculate how much stiffer it would become if it did indeed account for the missing component. On this approximation, we assume that titin contributes all forces imposed on the thick filament and that there are 12 titins per thick filament (6 titins per half-thick filament), equating to ~2.8–4.9 pN per titin. Based on previous data (67), passive titin-based forces are ~2.5 and ~5 pN per titin at 2.7 and 3.0 μm SL, respectively, equating to a 2.7 to 3.0 μm SL stretch stiffness (~150 nm stretch per titin) of ~0.016 pN/nm. If we add the excess titin-based forces to the 3.0 μm SL values assuming it is all the unaccounted for f_TF, then the titin-based stiffness after the stretch would be ~0.035–0.049 pN/nm, suggesting that upon activation, titin stiffness increased ~2.2–3.1 times compared to the passive state. This value will be somewhat lower if MyBPc or other parallel elastic components are involved. Presently, no molecular mechanism for this activation-dependent increase in titin stiffness is agreed upon, although studies are ongoing (19, 58). One idea (11, 19, 57, thoroughly discussed in ⁶⁸) relies on evidence of relatively weak titin-thin filament interaction at the N2A and nearby PEVK region in passive muscle and that this binding is stronger in the presence of Ca²⁺. This attachment functionally shortens the titin free-length and allows only the stiffer PEVK region to extend with increasing SL. Fig. 4 presents hypothetical configurations that can explain our data, where titin free-lengths are extended more during active vs. passive stretch, and deserves further empirical study. At present and under an as-yet unclear mechanism, we provide strong evidence that titin-based forces are greater after an active stretch as compared to an isometric contraction at the longer length, leading to structural and mechanical changes that enhance force, which answers the decades-old question as to why RFE exists.

Fig. 4. Mechanism of RFE.

Configuration of sarcomeric proteins during (A) Iso_S, (B) Iso_L, and (C) Iso_R that can account for their distinctive mechanical and structural signatures. In passive muscle, low-level titin-thin filament interactions occur in such a way that passive stretch is enough to detach-reattach and/or drag titin along the thin filament so that titin-based free length and extension are still similar to if they were not attached at all. During contraction, the titin-thin filament interaction becomes stronger, so that during an eccentric contraction, titin extension occurs above that in passive, producing elevated titin-based force and explaining the mechanical and structural signatures in Iso_R. Increased titin-based force contributes to RFE, which also leads to smaller lattice spacing and increased thick filament stiffness, improving force production and force transmission, respectively.

Conclusions

Our data demonstrate that titin is a critical regulator of sarcomeric tension, and as such, an essential determinant of RFE. Furthermore, the presence of RFE seems to align with a distinctive structural state that is affected by titin cleavage and the mdm mutation. Finally, our analysis provides evidence that the generation of RFE is predominately caused by not only increases to titin-based forces, but also via titin’s pull on sarcomeric structures that decrease lattice spacing and increase thick filament stiffness.

Data and materials availability:

All data are available in the main text or the supplementary materials, or available upon reasonable request.