A new experimental model for force enhancement: steady-state and transient observations of the Drosophila jump muscle

Abstract

The increase in steady-state force after active lengthening in skeletal muscle, termed force enhancement (FE), has been observed for nearly one century. Although demonstrated experimentally at various structural levels, the underlying mechanism(s) remain unknown. We recently showed that the Drosophila jump muscle is an ideal model for investigating mechanisms behind muscle physiological properties, because its mechanical characteristics, tested thus far, duplicate those of fast mammalian skeletal muscles, and Drosophila has the advantage that it can be more easily genetically modified. To determine if Drosophila would be appropriate to investigate FE, we performed classic FE experiments on this muscle. Steady-state FE (FE_SS), following active lengthening, increased by 3, 7, and 12% of maximum isometric force, with increasing stretch amplitudes of 5, 10, and 20% of optimal fiber length (FL_OPT), yet was similar for stretches across increasing stretch velocities of 4, 20, and 200% FL_OPT/s. These FE_SS characteristics of the Drosophila jump muscle closely mimic those observed previously. Jump muscles also displayed typical transient FE characteristics. The transient force relaxation following active stretch was fit with a double exponential, yielding two phases of force relaxation: a fast initial relaxation of force, followed by a slower recovery toward steady state. Our analyses identified a negative correlation between the slow relaxation rate and FE_SS, indicating that there is likely an active component contributing to FE, in addition to a passive component. Herein, we have established the Drosophila jump muscle as a new and genetically powerful experimental model to investigate the underlying mechanism(s) of FE.

the force production of muscle relies on the overlap of myosin and actin filaments. This overlap changes with the length of the sarcomeric unit and the overall muscle length. However, when muscle is actively lengthened, the force produced at the new length is greater than predicted from the muscle's force-length relationship, a history-dependent phenomenon known as force enhancement (FE). Steady-state FE (FE_SS) is a well-accepted characteristic of skeletal muscle, demonstrated experimentally across a wide range of structural scales: from whole muscle (34, 50) and multijoint contractions involving several muscles crossing several joints (14, 51) to single-fiber preparations (8, 9, 54) and single myofibrils (26). This phenomenon is also conserved across species, as evidenced by studies demonstrating FE in small mammals (16, 35, 44), amphibians (33, 41), as well as human skeletal muscle (15, 39, 52).

Numerous prior studies have demonstrated that FE_SS increases with rising amplitudes of stretch (1, 5, 8, 9, 54) on both the descending (9, 38) and ascending (20, 21) limb of the force-length relationship. On the other hand, the magnitude of FE_SS is not influenced by speed of stretch (8, 9, 33, 54). Additionally, a strong correlation exists between FE_SS and the magnitude of mechanical work applied during lengthening, such that applied work is a clear predictor of FE (31).

Although FE is well characterized, its fundamental mechanism remains unknown and is still heavily debated (17, 19, 37). Morgan and colleagues (38) proposed the sarcomere nonuniformity theory to explain the increase in force as an increase in the distribution of sarcomere lengths in response to the stretch on the descending limb of the force-length relationship of muscle. Elegantly, this model is capable of justifying the observations of both force depression (FD) and FE. However, no clear example of sarcomere popping, due to instability on the descending limb, has been observed experimentally (43, 45, 48). Moreover, direct measurement of sarcomere lengths during and after active stretch for isolated myofibrils (57) and within a single sarcomere (40) demonstrated that the nonuniformity of sarcomere lengths did not increase with active length changes, thereby suggesting that sarcomere nonuniformity could not solely be responsible for FE_SS.

Recently, the search for FE mechanisms has focused on determining the importance of contributions from passive compared with active sarcomere components. A major passive component of the sarcomere, titin, functions as a parallel elastic spring element (12, 20, 22) and has been proposed to store mechanical energy during lengthening in the presence of calcium (8, 20, 24, 25, 42). A passive element could explain the observations of FE on both the ascending and descending limb of the force-length relationship (41, 43, 49) and within a single sarcomere (26, 35, 43). Whereas these studies suggest that titin can account for a large portion of FE_SS, a passive component alone appears unable to account for the entirety of the higher force seen with FE (3, 26). Evidence for an active, cross-bridge-based FE component comes from studies, such as that of Rassier and Herzog (47) and our previous study (29), which demonstrate a decrease in the relaxation rate following active stretch for increasing amplitudes of stretch. Thus a complex set of active and passive sarcomere mechanisms may be responsible for FE (18, 25, 29).

To help sort out this complex set of mechanisms, a model system enabling easy manipulation of muscle proteins would be very helpful. The Drosophila jump muscle, or the tergal depressor of the trochanter (TDT), is a synchronous muscle that possesses force-pCa, force-length, and force-velocity relationships similar to that of mammalian skeletal muscle (10), as well as similar transient and steady-state FD in response to active shortening (30). If this model also exhibits characteristic FE behavior, then when coupled with the easy genetic mutability of Drosophila, it will make an excellent experimental model to study both passive and active components of the underlying FE mechanism.

To determine if FE is present in jump muscle, we examined force production of Drosophila TDT fibers in response to three stretch speeds and three amplitudes of active muscle lengthening and analyzed the period of force recovery immediately after lengthening to gain insight into the transient aspects of FE.

MATERIALS AND METHODS

Jump muscle preparation and muscle testing apparatus.

Drosophila jump muscle fibers expressing wild-type myosin (56) were dissected and mounted onto a fiber mechanics apparatus, as described previously (10, 30). Briefly, jump muscles were dissected from the thoraces of 2- to 3-day-old female Drosophila and chemically demembranated (skinned) in dissection solution [pCa 8.0, 5 mM MgATP, 1 mM free Mg²⁺, 0.25 mM phosphate, 5 mM EGTA, 20 mM N,N-bis(2-hydroxyethyl)-2-aminoethanesulfonic acid (BES; pH 7.0), 175 mM ionic strength, adjusted with Na methane sulfonate, 1 mM DTT, 50% glycerol, and 0.5% Triton X-100] for 1 h at 4°C. Dissected muscles were pared down to fiber bundles, which were T-clipped and then mounted on a multiwelled single fiber mechanics apparatus capable of measuring millinewton-scale loads and length changes with nanometer resolution at a submillisecond response (55).

Isometric contractions.

Jump muscle fibers were mounted onto the mechanics apparatus in relaxing solution [260 mM ionic strength, 10 mM MgATP, 45 mM creatine phosphate, 1,200 U/ml creatine phosphokinase, 1 mM Mg²⁺ (free), 5 mM EGTA, 20 mM BES (pH 7.0), and 1 mM DTT], maintained at 15°C. While in relaxing solution, muscle fibers were stretched to an optimal fiber length (FL_OPT), corresponding to an average sarcomere length of 3.6 μm, at which interclip dimensions were measured. Jump fibers were transferred to preactivating solution (same as relaxing, except 0.5 mM EGTA) with an automated well exchanger. After 2 min of equilibration, TDT fibers were transferred to activating solution (pCa 5.0), where the isometric curves for the final length (FL_OPT) were recorded. After 60 s of activation, jump muscle fibers were returned to relaxing solution for 5 min. A similar set of procedures was carried out to obtain the isometric tension curve for the shortest initial length (80% FL_OPT: shortened 20% from FL_OPT).

Evaluation of FE.

A similar cyclic process of activation and relaxation was used to investigate FE. However, activated fibers were eccentrically stretched at three different speeds (200, 20, and 4% FL_OPT/s) for the longest amplitude of stretch (20% FL_OPT). Each fiber also underwent an active stretch of 20, 10, and 5% FL_OPT at a constant velocity of 4% FL_OPT/s for a total of five experimental runs (Fig. 1). Active stretching took place within the 1st 30 s of each 60-s run, after full activation was achieved. The start time of each stretch was adjusted so lengthening ended at 30 s from the start of activation, regardless of the amplitude or speed of stretch. Following the stretch (while still in activation solution), muscle length was held constant for an additional 30 s to allow force to recover toward a steady-state value. This is critical to ensure that each fiber was activated for the same amount of time and allowed to recover for the same period following stretch. The muscle was then transferred to relaxing solution (Fig. 2). All runs ended at the same final length (FL_OPT), and the order was systematically randomized for each fiber. FE_SS was found as the difference between the enhanced force at steady state [(F_inf); see Eq. 1] and the isometric force (F_iso) at the corresponding final length (FE_SS = F_inf − F_iso). When normalized to F_iso, FE_SS is converted to a percent increase in force.

Fig. 1.

Activated stretch protocol used to determine force enhancement (FE) in Drosophila jump muscle fibers. For all runs, muscle fibers were stretched to the same final jump muscle sarcomere length, 3.6 μm, that generates maximum isometric tension [optimal fiber length (FL_OPT)]. To determine if steady-state FE (FE_SS) was dependent on amplitude of stretch, the muscle was started at shorter lengths and lengthened by 5, 10, and 20% of FL_OPT to the same final length (FL_OPT) at a velocity of 4% FL_OPT/s. Muscle fibers were also stretched at 3 different velocities, 200, 20, and 4% FL_OPT/s, over the greatest amplitude of shortening, 20%. For all tests, muscles were calcium activated, pCa 5.0, at time 0 s and maintained in the activated state throughout the duration of each test period (60 s).

Fig. 2.

Representative active stretch and reference isometric (FL_OPT: 3.6 μm) force traces from 1 Drosophila jump muscle fiber (A and B). Force levels are normalized (T/T_isometric × 100) to a corresponding reference isomeric tension (average of final 3 s). FE_SS showed no correlation to a change in stretch velocity (A and C) but increased with increasing amplitude of stretch (B and D). Means ± SD; n = 10. *P < 0.05, 2-way ANOVA.

Throughout the protocol, after every two test runs, an isometric reference contraction was recorded at FL_OPT to monitor the integrity of the fiber. These isometric values were compared with the initial isometric reference traces to monitor for damage or fatigue in the fiber. If a >30% decrease in force was observed between a reference isometric contraction and the first isometric measurement, then the fiber was deemed damaged and excluded from the study. All active stretch runs were systematically randomized, from one experiment to the next, to avoid order-dependent biasing. The mechanical work done on the muscle during active lengthening was determined by integrating the force-displacement trace using a trapezoidal approximation method (6, 7, 30). Peak force was found as the maximum value of the force-time trace, which always occurred at the end of active stretch (t = 30 s).

Transient analysis.

We observed a biphasic force recovery consisting of both a rapid drop in force and a slower relaxation to steady state following active lengthening. To quantify this transient force relaxation immediately following stretch, the force-time data were analyzed using a double-exponential function, similar to how we analyzed the transient force redevelopment period of FD in the jump muscle (30). Specifically, the transient relaxation period was analyzed using the following double-exponential function

$$F(t)=F_{\inf }+(A_F{e}^{-k_{F}t}+A_S{e}^{-k_{S}t})$$ (1)

where A_F and A_S are the amount of force recovery associated with both the fast and slow rate of relaxation, respectively (i.e., the magnitude of decreased force during that phase); k_F and k_S are the fast and slow exponential force relaxation rates, respectively; and F_inf represents the asymptotic force at infinite recovery time (29). However, both relaxation rates are influenced by the large initial, fast force recovery response. Values for F_inf, k_F, k_S, A_F, and A_S were obtained using a Levenberg-Marquardt error minimization algorithm in a commercial curve-fitting program (DeltaGraph 5.0; SPSS, Chicago, IL).

Statistical analysis.

A block design, two-way ANOVA, was used to explore differences and determine statistical significance (P < 0.05). Regression analyses were used to determine the correlations among FE_SS, work, and the rates of force relaxation (k_F, k_S).

RESULTS

Steady-state FE.

Jump muscle fibers (length: 133.0 ± 14.6 μm, width: 71.2 ± 11.5 μm, height: 40.7 ± 5.6 μm; average ± SD) exhibited classic FE behavior (Fig. 2, A and B) in response to three amplitudes and three speeds of active lengthening. The jump muscle produced 34.7 ± 5.8 mN/mm² isometric tension at FL_OPT. FE_SS increased (2.6–12.3%) with increasing amplitudes (5–20% FL_OPT) of stretch (Table 1 and Fig. 2D) but showed no correlation with increasing stretch velocities (Table 2 and Fig. 2C). Regression analysis demonstrated a significant link between the mechanical work applied to the muscle during active lengthening and amount of FE_SS (Fig. 3; P < 0.01). Mechanical work increased significantly with increasing amplitudes of stretch (Table 1) but remained constant for increasing velocities of stretch (Table 2).

Table 1.

The influence of stretch amplitude on transient and steady-state characteristics of FE

	FEss, %	Work, nJ	kF, s−1	kS, s−1	AF, mN × 102	AS, mN × 102
5% FLOPT	2.64 ± 1.43*	0.65 ± 0.41*	2.10 ± 1.11	0.15 ± 0.05	0.85 ± 0.60	1.05 ± 0.86
10% FLOPT	7.24 ± 2.22†	1.42 ± 0.66†	1.71 ± 0.66	0.10 ± 0.09	0.94 ± 0.50	0.96 ± 0.25
20% FLOPT	12.25 ± 5.28*†	2.63 ± 0.86*†	1.35 ± 0.52	0.05 ± 0.03†	1.06 ± 0.43	0.97 ± 0.55

The amount of steady-state force enhancement (FE_SS), work, rates of force relaxation [fast (k_F) and slow (k_S)], and the amounts of force relaxation [fast (A_F) and slow (A_S)] for 3 amplitudes of stretch. All values are means ± SD.

^*P < 0.05 from 10% optimal fiber length (FL_OPT; 1-way ANOVA);

^†P < 0.05 from 5% FL_OPT (1-way ANOVA).

Table 2.

The influence of stretch velocity on transient and steady-state characteristics of FE

	FEss, %	Work, nJ	kF, s−1	kS, s−1	AF, mN × 102	AS, mN × 102
4% FLOPT/s	12.25 ± 4.79	2.63 ± 0.86	1.35 ± 0.52*	0.05 ± 0.02	1.06 ± 0.43	0.97 ± 0.55
20% FLOPT/s	14.21 ± 4.26	2.78 ± 1.35	3.05 ± 1.54†	0.22 ± 0.19†	2.03 ± 1.09	0.88 ± 0.45
200% FLOPT/s	14.00 ± 4.95	2.64 ± 1.40	5.33 ± 2.10*†	0.27 ± 0.18†	2.37 ± 0.97†	1.12 ± 0.69

The amount of FE_SS, work, k_F and k_S, and A_F and A_S for 3 velocities of stretch. All values are means ± SD.

^*P < 0.05 from 20% FL_OPT/s (1-way ANOVA);

^†P < 0.05 from 4% FL_OPT/s (1-way ANOVA).

Fig. 3.

The relationship between muscle mechanical work and FE_SS for all active shortenings across all fibers tested. Regression analysis indicates a significant (P < 0.05) correlation between the muscle mechanical work applied during active lengthening and FE_SS.

Transient force relaxation following stretch.

To capture the biphasic transient force relaxation period following active lengthening, force-time data were fit with a double exponential (Eq. 1 and Fig. 4). The double exponential captured both a fast and slow component of the force relaxation, which more accurately (R² = 0.947) represents the complete transient force response (0–30 s) compared with a single exponential (R² = 0.897).

Fig. 4.

Representative force relaxation traces following active stretching. The force response was fit with a double-exponential relaxation function (see Eq. 1), yielding 2 decay rates: fast (k_F) and slow (k_S). A: k_F and k_S increased with increasing speeds of stretch. B: k_F showed no correlation with increasing amplitudes of stretch, but k_S decreased with increasing amplitude.

The amount of force relaxation (i.e., force drop) in the fast phase of force recovery, A_F, increased with increasing stretch velocities (Table 2) but showed no change with increasing amplitudes of active lengthening (Table 1). Furthermore, no differences were observed for the amount of force relaxation in the slow phase, A_S, for either stretch amplitude (Table 1) or velocity (Table 2). Regression analysis showed no correlation between the amount of FE_SS and either k_F (Fig. 5A) or k_S (Fig. 5B).

Fig. 5.

The relationship between FE_SS and both phases of force relaxation, k_F and k_S. Regression analyses across all active stretches found no correlation between FE_SS and k_F (A) or k_S (B).

The examination of the rates of force relaxation for stretch velocity and amplitude showed that k_F increased with increasing stretch velocity (Fig. 6A), but no significant correlation was observed between k_F and stretch amplitude (Fig. 6B). The slow relaxation rate, k_S, increased with faster stretches (Fig. 6C) and decreased significantly from the smallest (5% FL_OPT) to the largest (20% FL_OPT) length changes (Fig. 6D). Therefore, when blocked for the influence of stretching speed, there is a strong negative correlation between k_S and FE_SS (Table 1), indicating a slower relaxation to a more enhanced steady-state force with higher levels of FE_SS. The peak force at the end of the active stretch increased with both increased velocity (Fig. 7A) and increased amplitude (Fig. 7B) of stretch.

Fig. 6.

The pooled results (n = 10) for both k_F and k_S across all fibers and experimental conditions. A: early k_F increases with increasing velocities of stretch and exhibits no difference for increasing amplitudes of stretch (B). C: late k_S increased with velocity of stretch but decreased with increasing amplitudes of stretch (D). Means ± SD; n = 10. *P < 0.05, 2-way ANOVA.

Fig. 7.

A higher peak force at the end of stretch was observed for both increasing velocities (A) and increasing amplitudes (B) of stretch. Means ± SD; n = 10. *P < 0.05, 2-way ANOVA.

DISCUSSION

Steady-state FE.

We have expanded the capabilities of the Drosophila jump muscle for investigating history-dependent phenomena by showing that it possesses substantial FE characteristics and is robust enough to handle the eccentric contractions involved when making these measurements. Eccentric contractions are well known to have a damaging effect on muscle. However, our skinned jump muscle fiber bundles performed very well when subjected to the rigorous eccentric FE protocol (five eccentric and four isometric contractions; Fig. 2), as 56% of the jump muscle fibers tested fulfilled the inclusion parameters for our study. This compares favorably with eccentric measurements made with other skinned fiber preparations (49, 53). For example, in a recent study of skinned human vastus lateralis fiber bundles subjected to a single eccentric stretch, only 73% of the samples were able to be included (4).

The jump muscle exhibited classic FE behavior in response to stretch amplitude (1, 5, 8, 54) and velocity (8, 9, 33). Qualitatively, FE_SS increased with stretch amplitude and showed no correlation with stretch velocity (Fig. 2). Quantitatively, the jump muscle exhibits FE_SS of similar magnitude to previously reported values. Our experiments showed an average FE_SS of 7.2% for a 10% elongation. Herzog and Leonard (20) showed a ratio of 8.5% FE_SS for an amplitude of 9% for in situ cat soleus muscles. Similarly, Lee and Herzog (32, 33) demonstrated, in skinned lumbrical muscles of Rana pipiens, an 11.5% increase in FE_SS for a 10% stretch. This ratio was also conserved for intact frog lumbrical muscles, showing 10.7% FE_SS for 10% of stretch and 18.3% for 20% of stretch (46).

Similarly, our jump muscle results showed no change in FE_SS with stretch velocity, which exactly matches prior findings that robustly demonstrated stretch velocity to have no influence on the amount of FE_SS (16, 33, 35, 44). Additionally, we found that FE_SS strongly correlated with the mechanical work done on the muscle (Fig. 3), which is similar to prior observations in isolated skeletal muscle of mice (31), indicating that mechanical work is a good predictor of FE_SS in the Drosophila jump muscle.

Insights into the FE mechanism.

Whereas the major insights into FE from our new preparation are most likely to come from the in vivo protein engineering possible with Drosophila, our current experiments provide some unique insight into the FE mechanism. Our experiments support the recent conclusion that the amount of FE_SS generated cannot be explained by the amount of force the muscle is subjected to during stretch (29). Whereas our amplitude data show a strong correlation between stretch amplitude and both peak force (force muscle experiences during stretch; Fig. 7B) and FE_SS (Fig. 2D), our velocity data show no correlation between FE_SS (Fig. 2C) and stretch velocity, despite the significant increase in peak force that occurs with faster stretches (Fig. 7A).

Interesting, however, might be the viscous and elastic contributions to this force and what these may suggest regarding FE. The increased force with increased velocity would be primarily viscous in nature, whereas the change with amplitude would likely be elastic. Thus our observations that FE_SS strongly correlates with stretch amplitude but does not correlate with stretch speed suggest that alterations in FE_SS can be induced by performing length changes in a way that perturbs the elastic components of the sarcomere but not by means that primarily perturb the viscous components. In general, the viscous response of the sarcomere is dominated by cross bridges actively cycling [e.g., Bagni et al. (2) and Kawai (28)], whereas passive responses are dominated by elastic structures, such as titin and passive elasticity within cross bridges [e.g., Granzier (13) and Herzog (17)]. One could thus argue that passive elastic components are likely the dominant player in FE (3).

There is currently great interest in resolving this question of the roles of passive and active sarcomere components in the FE mechanism. Besides the general argument mentioned above, there is substantial experimental evidence for a passive component, particularly for titin playing a significant role in the FE mechanism. A recent review by Herzog (17) proposed that titin could account for many experimentally observed FE phenomena, including the passive FE observed following active stretch, and increasing FE_SS with larger stretch amplitudes. This was further supported by the complete elimination of passive FE when titin was removed from sarcomeres (27). However, recent experimental evidence suggests that more than just a passive component, such as titin, is likely involved. Bullimore et al. (3) showed that whereas a passive element could explain a significant portion of FE, it was not able to account for the entire observed FE. Joumaa et al. (27), using myofibrils in which active force generation was eliminated completely via a troponin C depletion, found that titin could only account for 25% of the observed FE_SS. The authors concluded that in addition to titin, FE is influenced by cross-bridge formation and active force generation.

We gained insight into the influence of passive vs. active components involved in the FE mechanism by analyzing the transient period following stretch. A double-exponential curve fit yielded two distinct phases of force relaxation: a fast initial phase of relaxation and a slower component of force relaxation. The interpretation of these transients according to the Huxley and Simmons model (23) would mean that the fast phase is dominated by passive recoil due to the elastic stiffness component of cross bridges and parallel passive sarcomere elements. The second phase, k_S, is primarily set by the rate of cross-bridge cycling (11). Thus we could potentially gain insight into passive vs. active elements by determining if either of these phases correlates with changes in FE_SS.

The comparison of the fast and slow rates with all of the FE_SS measurements we made, regardless of the stretch protocol, did not reveal any significant correlations (Fig. 5). This is similar to the lack of a correlation and a very weak correlation observed by our previous work with cat soleus muscle (29) and thus does not provide insight into the FE mechanism. However, when we analyzed FE_SS when blocked for the different protocols (only velocity or only amplitude variations), there were some insightful correlations. We observed that FE_SS was altered by changing the amplitude of stretch. There was not a significant correlation between this alteration in FE_SS and k_F (Fig. 6B), but k_S did decrease with increasing FE_SS (Fig. 6D). If k_S is interpreted as stated above, then this would support the involvement of an active cross-bridge component and suggests that cross-bridge cycling slows down with increasing FE_SS. This could account for some of the increase in FE_SS if the slowing is due to cross bridges staying attached longer to actin compared with before the stretch. A decrease in cycling rate has been supported previously by findings of a decrease in ATPase rate following stretch (25). Interestingly, when examined in isolated myofilaments by the same laboratory, no increase in cross-bridge dwell time or duty ratio was observed with FE (36). However, as the authors state, caution should be taken when comparing these myofilament results with other studies. They used a very low ATP concentration (0.1 μM ATP), far below physiologic, for testing isolated myofibrils (36) but a physiologic concentration (2.5 mM ATP) for skinned fibers (25), similar to the physiologically relevant concentration (10 mM MgATP) used herein. Furthermore, the laser trap used to evaluate myofibrils was more compliant than the expected in vivo stiffness, which could lead to longer dwell times and lower cross-bridge force (36).

We observed that increasing stretch velocity strongly correlated with increasing k_F (Fig. 6A), as well as increases in A_F (Table 1) and peak force (Fig. 7). However, the velocity protocol resulted in no change in FE_SS. Thus these correlations are unlikely to provide insight into the FE_SS mechanism. Instead, these might be telling us something about how sarcomere components are responding to faster lengthening speeds of eccentric muscle contraction when working on the dynamic portion of the force-velocity curve. However, this is outside the scope of our current investigation.

Whereas our current analysis of the transient component of FE lends support to an active cross-bridge component being involved in the FE mechanism, the genetic mutability of the Drosophila system will enable more targeted and unique experiments to help test the FE mechanism. Now that we have established the jump muscle as an excellent model system, in the future, we will alter proteins, such as titin and myosin, that contribute to passive and active properties of muscle to explore the influence of sarcomere stiffness and contractile kinetics on FE.

GRANTS

Support for this work was provided by the National Science Foundation (NSF Career Award CBET-0954990; to D. T. Corr) and the National Institute of Arthritis and Musculoskeletal and Skin Diseases (R01 AR064274; to D. M. Swank).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

Author contributions: R.A.K., D.M.S., and D.T.C. conception and design of research; R.A.K. performed experiments; R.A.K. analyzed data; R.A.K., D.M.S., and D.T.C. interpreted results of experiments; R.A.K. prepared figures; R.A.K. drafted manuscript; R.A.K., D.M.S., and D.T.C. edited and revised manuscript; R.A.K., D.M.S., and D.T.C. approved final version of manuscript.