An embryonic myosin converter domain influences Drosophila indirect flight muscle stretch activation, power generation and flight

Small insects require high wing beat frequencies to generate sufficient aerodynamic power for flight. Many flight muscle adaptations have evolved to enable power generation at high muscle speeds. Some insect flight muscles are not directly attached to the wings, and these are known as indirect flight muscles (IFMs) as they oscillate the thorax to power flight. A subset of insect IFMs, asynchronous IFMs, contract more frequently than the nerve firing rate (Josephson et al., 2000). Asynchrony and power generation are possible because of stretch activation (SA) and shortening deactivation (SD). SA is a delayed rise in tension following rapid muscle lengthening (Pringle, 1949). The delayed activation increases tension generation by the muscle during active shortening, compared with if it was not lengthened (stretched) prior to shortening. Its opposite and complementary phenomenon SD is a delayed drop in tension following shortening, which decreases resistance during muscle lengthening (Josephson et al., 2000). In muscle types that cyclically lengthen and shorten, SA and SD increase power generation. In the highly stretch activated IFMs, SA and SD take over modulating force levels so that there is no need for calcium cycling with each contraction cycle, thereby enabling the muscle contraction cycle frequency to be faster than the motor nerve stimulation frequency.

IFM proteins have evolved to enable faster muscle speed and cyclical power generation (Bullard et al., 1973; Vigoreaux et al., 1990). There are specialized versions of connecting filaments (titin homologs) (Burkart et al., 2007), thin filament regulatory proteins (Qiu et al., 2003) and thick filament proteins (Peckham et al., 1992). In particular, the most important thick filament protein for setting muscle mechanical properties, myosin, has evolved for high power generation at high speeds. We found that most rate constants of the IFM myosin isoform (IFI) cross-bridge cycle are much faster than slower myosin isoforms (Swank et al., 2006b). In particular, IFI has evolved a very fast detachment rate from actin.

To determine the structural region(s) responsible for the increased cross-bridge rate constants, we previously investigated the role of the four S-1 domains that vary between Drosophila myosin isoforms as a result of alternative splicing from a single gene (Swank et al., 2006a; Swank et al., 2004; Yang et al., 2008). We found that the converter region (Fig. 1) had the largest effect on cross-bridge rate constants (Yang et al., 2010) and power generation (Swank et al., 2002). When we exchanged the converter region from an embryonic myosin isoform with IFI we observed a 30% decrease in power, a 50% lower frequency of maximum power production and 60% slower actin velocity in the motility assay (Littlefield et al., 2003). Surprisingly, the animals could still fly fairly well in spite of these changes in muscle mechanical performance (Swank et al., 2002).

However, our previous study of the influence of the embryonic converter on power generation was likely limited by at least two factors, which probably decreased our IFM power values and perhaps skewed our perspective on its influence on IFI-embryonic converter (EC) flight ability. First, the power measurements were performed using small amplitude sinusoidal conditions: 0.125% muscle length (ML) change. This is much less than the 2–4% ML change estimated to occur in Drosophila virilis from measurements of thorax deformation during flight (Chan and Dickinson, 1996). Unfortunately, no estimate exists for Drosophila melanogaster, but most likely it is substantially greater than 0.125%. Applying longer length changes should increase IFM power values. Second, the ATP concentration of 5 mmol l⁻¹ in the fiber bathing solution may have limited the amount of power generated. Our initial converter study was performed prior to the

investigation in which we found that found IFM kinetics (f_max) continue to speed up until about 10 mmol l⁻¹ ATP (Swank et al., 2006b). This suggests that Drosophila myosins have a low ATP affinity. Third, we did not explore the contribution of the converter to SA or how any changes in SA might impact IFM power generation. We recently discovered that myosin isoforms can significantly alter SA tension, which influences power generation (Zhao and Swank, 2013). Perhaps the converter domain is involved in the structural mechanism that alters SA tension. Finally, we did not measure wing stroke amplitude (WSA), only wing beat frequency (WBF), when interpreting how the changes in muscle mechanics were influencing flight ability. WSA and WBF are the major parameters that flies vary to alter aerodynamic power generation (Lehmann and Dickinson, 1997).

To measure the influence of the converter region on power under conditions that more closely mimic IFM performance during flight, we used the work loop technique (Josephson, 1985). When conditions were optimized for maximum power generation from each fiber type, we found that the embryonic converter produced only 60% of the power generation of control fibers. The conditions that gave the most power from the IFI-EC fibers were a 2-fold increase in amplitude and a 60% decrease in sinusoidal oscillation frequency compared with IFI fibers. This enabled the slower IFI-EC myosin to operate at its optimal cycling rate and tension levels. By observing the tension response to two length step protocols, we found that the converter influenced SA. The lower SA tension was primarily because of a decrease in passive (relaxed) stiffness of the muscle. The IFI-EC flies appeared to compensate for the slower muscle kinetics by decreasing WBF and increasing WSA. This enabled moderate flight ability despite lower SA tension and power generation.

IFI-EC isometric tension (A₀) decreased by 27% and passive isometric tension (P₀) decreased by 39% compared with control IFI fibers (Table 1; Fig. 2). The passive tension decrease accounted for all of the difference in A₀ between the muscle types as net active isometric tension (F₀) was not significantly different.

To determine whether the decrease in passive isometric tension, P₀, of IFI-EC compared with IFI was because of a difference in the length of the two groups of fibers during experiments, we tracked the distance each fiber was stretched during optimal starting length determination. Fiber starting length was determined by lengthening the fiber until small amplitude power generation reached its maximum value. The experimental length of the two groups of fibers was the same. IFI fibers were stretched an average of 17.9±1.3% ML and IFI-EC by 17.0±1.5% ML from the length where tension was first apparent (just taut). This was not significantly different as the P-value from Student's t-test was 0.64 for 16 and 20 fibers tested, respectively.

IFI and IFI-EC fibers both showed the classic stretch-activated response to a 1% ML step increase for the two lengthening time periods tested, 0.5 and 3.5 ms (Fig. 3). Tension rose simultaneously with the length step due to the stretching of fiber elastic components (Huxley and Simmons phase 1) (Huxley and Simmons, 1971). Tension declined after lengthening ceased (phase 2), followed rapidly by a delayed increase in tension (phase 3). The delayed tension increase is also referred to as stretch-activated tension (A_SA).

For both length step durations, 0.5 and 3.5 ms stretch, A_SA and passive stretch-activated tension (P_SA) were both significantly reduced for IFI-EC fibers, but net active stretch tensions (F_SA) were not different (Table 2 and Fig. 3). P_SA and A_SA of IFI-EC were 51% and 40% lower than IFI, respectively, for the 0.5 ms length step. Under the 3.5 ms length step protocol, IFI-EC fibers generated 42%lower P_SA and 50% lower A_SA than IFI. The IFI-EC rate constant for phase 3, k₃, measured from the uncorrected (A_SA) 0.5 ms data, was 34% lower in IFI-EC fibers. This indicates a slower cross-bridge turnover (Table 2).

The different step durations, 0.5 and 3.5 ms, did not produce any significant alterations in phase 3 amplitudes, A_SA, P_SA or F_SA, when compared within the same fiber type group (Table 2). We could not reliably measure k₃ for 3.5 ms steps because of the very brief durations of phase 2 and 3. Thus we could not compare k₃ values between protocols. However, the peak of phase 3 was typically 1–1.5 ms later for the 3.5 ms duration step than for the 0.5 ms step.

To further determine how exchanging the converter influenced muscle stiffness, we measured active (pCa 5.0), passive (pCa 8.0) and net active muscle stiffness by calculating the change in tension during the length step from the amplitude of phase 1. IFI-EC active stiffness was about 40% less than IFI fiber stiffness (Table 3). Passive stiffness was also about 40% less, resulting in no net active stiffness difference.

IFI fibers generated 728±31 W m⁻³ of oscillatory power (Table 4). The optimized conditions that produced this power were an oscillation frequency (f_max) of 136±11 Hz and a ML change amplitude of 0.86±0.05% (Table 4 and Fig. 4). Under these conditions, IFI-EC produced 71±15 W m⁻³ and 0.7 J m⁻³ work, about 10% of IFI power and work. To increase power from IFI-EC, we optimized conditions for IFI-EC and found that the optimal length change and frequency conditions for IFI-EC were 2-fold greater and 60% lower than for IFI (Table 4). These conditions increased work 13-fold and power 6-fold compared with IFI-EC run under optimal IFI conditions. While we could have tried smaller increments of frequency and ML change values to more precisely determine optimal conditions and maximum power output for IFI-EC, it was clear from the increments we used (see Materials and methods) that optimized IFI-EC fibers cannot produce as much power as optimized IFI. Optimized IFI-EC fibers produced about 60% of the power produced by optimized IFI fibers (Table 4 and Fig. 4). Interestingly, work was increased in IFI-EC fibers to be 1.6-fold higher than IFI work (Table 4), but the lower frequency required for the higher IFI-EC work limited power generation. In general, the highest frequency that produced positive power at the optimal length change was about 275 Hz for IFI and 100 Hz for IFI-EC fibers. Muscle length changes greater than about 2% for IFI and 3% for IFI-EC produced negative power.

Increasing ML change from 0.125% to the optimal length for a specific fiber type greatly increased muscle power, but IFI-EC benefited more as its power output increased 5.4-fold compared with 3.2-fold for IFI (Tables 4 and 5, column 1). Not all of the power increase resulted from changing amplitude as optimizing power output also required decreasing ML oscillation frequency (f_max) by 26% for IFI and 60% for IFI-EC (Tables 4 and 5, column 2).

The decrease in IFI-EC muscle performance led to a decrease in flight performance (Table 6). IFI-EC flight index decreased by 34% compared with IFI. We measured a 15% lower WBF for IFI-EC compared with IFI that likely is contributing to the decreased flight ability. In contrast, IFI-EC WSA increased 5% compared with IFI, which should increase aerodynamic power generation (Lehmann and Dickinson, 1997).

The embryonic converter dramatically decreased power generation, especially when run under IFI optimal conditions. Comparing the IFI-EC work loop under IFI conditions with the optimal IFI work loop (Fig. 4) provides insight when interpreted in the context of the decreased IFI-EC SA tension and lower SA k₃ values. The IFI-EC muscle was not able to generate high tension at the start of muscle shortening such that tension was lower at the start of shortening compared with that at the end of lengthening. This caused net negative work during this portion of the cycle. A more substantial SA tension boost at the start of shortening, such as the boost apparent in the IFI work loop, would have increased work and power production. Also contributing to IFI-EC's decreased work was its slower k₃, which caused the smaller IFI-EC SA tension boost to occur too late in the cycle, near the end of shortening. Thus, only near the end of shortening was the muscle producing net positive work.

However, IFI-EC work and power were significantly improved by optimizing the length change conditions to be compatible with its myosin properties. Slowing IFI-EC work loop oscillation frequency matched its slower k₃ so that the timing of the SA response led to a boost in tension at the start of shortening (Fig. 4). A 2-fold increase in change in ML also helped increase power output, but was somewhat unexpected. The greater amplitude during shortening certainly helped increase positive work, but it could have increased negative work as much or more than positive work. The lower IFI-EC stiffness likely enabled the larger amplitude to be beneficial, otherwise tension might have been much higher during lengthening, resulting in more negative work.

While IFI-EC fibers were less stiff than IFI fibers, this stiffness was primarily due to muscle components that contribute to stiffness at pCa 8.0, passive stiffness. Our observations of decreased passive tension P₀ and passive SA tension P_SA (pCa 8.0 measurements) supported this finding. It might seem odd that decreases in passive properties resulted from altering the myosin cross-bridge structure as one might think that cross-bridges only influence muscle stiffness when the muscle is active rather than when it is relaxed. Cross-bridges are described as detached from actin or ‘weakly bound’ at pCa 8.0. However, significant mechanical effects from weakly bound cross-bridges have been observed in relaxed vertebrate muscle (Brenner et al., 1982; Brenner et al., 1984; Hirose et al., 1994) and insect IFM (Schmitz et al., 1994). High-speed stiffness measurements under relaxing conditions, such as we made when measuring SA properties, are influenced by weakly attached cross-bridges. The weakly attached bridges are elastically distorted before they can detach (Brenner et al., 1982). Perhaps weakly bound cross-bridges contribute more significantly to passive properties in IFM than in other muscle types as high passive stiffness is typical of IFM (Granzier and Wang, 1993).

Alternatively, there are some unique protein structures in IFM that could connect cross-bridges to actin when the muscle is relaxed. For example, myosin could interact with actin through the regulatory light chain extension (Moore et al., 2000). Electron microscopy studies of the equivalent essential light chain extension of mouse cardiac myosin suggest it can bind to actin and to the myosin converter domain (Lowey et al., 2007). Electron microscopy experiments by Reedy and colleagues show ‘troponin bridges’ that appear to be myosin binding to troponin (Perz-Edwards et al., 2011; Tregear et al., 1998). For either alternative, Drosophila myosin appears to be interacting with actin and contributing to tension and stiffness at pCa 8.0. If Drosophila myosin is contributing to relaxed stiffness, then one interpretation of our results is that the embryonic converter is decreasing myosin stiffness.

That the converter influences myosin stiffness has been suggested by Kraft and colleagues (Köhler et al., 2002; Seebohm et al., 2009). Their mechanical analyses of point mutations in the converter that cause hypertrophic cardiomyopathy suggest that the converter is a highly compliant element in myosin and that changes to its structure can alter cross-bridge stiffness. They found one mutation, Arg719Trp, that caused a ~47% increase in relaxed fiber stiffness when measured using very rapid length changes (Köhler et al., 2002).

If IFI-EC myosin is more compliant than IFI, this should also cause less net active stiffness and, according to most models of tension generation (Seebohm et al., 2009), less net isometric and net SA tension generation. While we did not observe a significant decrease in these net active muscle properties, they can also be influenced by myosin step size and myosin cross-bridge cycle kinetics (Seebohm et al., 2009). We previously analyzed step size and found no measurable difference between IFI and IFI-EC (Littlefield et al., 2003), but from this study and our previous studies we know that the cross-bridge kinetics of IFI-EC myosin are dramatically slowed compared with those of IFI (Swank et al., 2002; Yang et al., 2010). Specifically, we have previously shown that IFI-EC increases myosin ATPase rate but decreases actin motility velocity. These results suggest that IFI-EC myosin possesses a higher duty ratio (the ratio of the average time a myosin head spends strongly bound to actin versus weakly bound) than IFI (Littlefield et al., 2003). A higher duty ratio means more myosin heads are strongly bound to actin at any given time, thus increasing active tension and stiffness. Thus, one possibility is that an increase in stiffness from the higher IFI-EC duty ratio is offsetting a decrease in myosin cross-bridge stiffness, resulting in no apparent change in net active stiffness, net isometric tension or net SA tension.

Another possibility for the decrease in passive stiffness is an indirect or secondary consequence of the converter exchange. Perhaps the IFI-EC muscles have adapted and become less stiff, which allows for the larger ML change amplitude that helps IFI-EC myosin produce more work and power (Fig. 4). As the IFM is still growing in size during the first 2 days of adult life (Ready and Beall, 1993), there should be ample opportunity for feedback from IFM mechanical performance to alter protein isoform expression. Passive stiffness of the IFM is highly influenced by kettin/sallimus and projectin, members of the titin family of elastic connecting filaments (Bullard et al., 2002; Granzier and Labeit, 2004; Kulke et al., 2001; Vigoreaux et al., 2000). The different isoform lengths of these proteins help set passive stiffness in various Drosophila muscle types (Burkart et al., 2007). A change in one of these proteins' isoforms in IFI-EC is not apparent from protein gels we have run in the past, but a thorough mass spectrometry study would be needed to rule out this possibility.

We previously found that the embryonic converter decreases IFM power generation and slows IFM kinetics, e.g. f_max (Swank et al., 2002). However, these measurements were made using small amplitude, 0.125% ML, sinusoidal analysis. At the time, we were focused on elucidating the influence of the converter on rate constants of the myosin cross-bridge cycle (Swank et al., 2002). Sinusoidal analysis requires that the tension response be linear with the length change, which only holds to about 0.3% change in ML (Swank, 2012). Our 2002 study also differed from our current study in that the skinned fiber bathing solution included only 5 mmol l⁻¹ ATP. A subsequent IFI kinetics study showed this concentration to be sub-saturating for wild-type fiber power generation as f_max increased up to about 10 mmol l⁻¹ ATP (Swank et al., 2006b). Comparing our current study results with our 2002 study reveals new information about IFI-EC and IFI properties. First, to differentiate the influence of ATP concentration from ML amplitude, we re-ran small amplitude, 0.125%, sinusoidal analysis at a higher ATP concentration, 12 mmol l⁻¹, which is saturating for IFM kinetics. Compared with our previous results (Swank et al., 2006b), increased ATP concentration caused f_max to increase for both IFI (1.2-fold) and IFI-EC (1.7-fold), but only resulted in significantly increased power for IFI (Table 5). Why did power generation not increase for IFI-EC? An increase in f_max should benefit power production as it means the fiber can produce work at higher speeds. The reason likely lies in the finding that increased ATP concentration increased f_max but decreased force production. The higher f_max is likely because of an increased detachment rate, as ATP concentration affects detachment of myosin from actin (Lymn and Taylor, 1971). However, an increase in detachment rate also decreases duty ratio, which decreases force generation. Thus, while increasing ATP concentration sped up IFI-EC f_max, this was likely offset by a substantial decrease in IFI-EC force. An alteration in force production is supported by IFI-EC net tension being equal to IFI in this study (12 mmol l⁻¹ ATP) while in our 5 mmol l⁻¹ ATP study, IFI-EC net tension was 2.3-fold higher than IFI (Swank et al., 2002).

While increasing ATP concentration did not improve IFI-EC power generation, larger ML amplitudes definitely boosted power production. Power increased in direct proportion to each fiber's ML increase above the small, 0.125% ML, amplitude used previously (Swank et al., 2002). IFI-EC power increased almost 6-fold because of a 16-fold amplitude increase to 2% ML, while IFI power increased 3-fold because of an 8-fold increase to 1% ML (Tables 4 and 5, compare low and high amplitude powers in the current study). When amplitude was increased, f_max decreased. This keeps the muscle operating at the optimal location on each fiber's force–velocity curve for maximum power generation. However, as optimizing IFI-EC resulted in a muscle length change increase of 2-fold more than IFI, its f_max decreased more than IFI (61% compared with 26% for IFI; Table 4 compared with Table 5).

The changes to SA and power generation by the embryonic converter caused decreased flight performance as flight index was lower for IFI-EC than for IFI flies. One reason for the poor flight is the 15% slower WBF as WBF is a major variable for insect aerodynamic power generation (Laurie-Ahlberg et al., 1985; Lehmann and Dickinson, 1997). However, while slower WBF sacrifices aerodynamic power, slower WBF helps generate more muscle power for flight as it better matches the slowed IFI-EC muscle kinetics. WBF cannot drop to exactly match the slower IFI-EC f_max, as aerodynamic power would be sacrificed to the point of flightlessness.

While WBF decreased, IFI-EC WSA was 5% greater than IFI WSA. Although this might seem like a small increase, aerodynamic power is approximately proportional to WSA³ (Laurie-Ahlberg et al., 1985). This increase helped offset the power lost from the decreased WBF and likely promoted flight ability, although not to the level of IFI flight performance. It is intriguing to speculate that the increase in WSA means that a larger ML change is occurring during flight that would be beneficial for IFI-EC muscle power generation. For this to be true, the increased WSA would have to be a result of increased thorax oscillation amplitude during flight. The two IFM muscle sets move the wings by alternately contracting the thoracic cuticle in the vertical and horizontal directions.

We also do not know the degree to which the changes in WSA and WBF are intentional versus forced to occur by the alterations in muscle mechanical properties caused by the embryonic converter. Flies can voluntarily vary WBF and WSA to change aerodynamic power (Lehmann and Dickinson, 1997). A voluntary mechanism for this variation was suggested by Gordon and Dickenson's findings that IFM calcium concentration correlates with IFM nervous stimulation frequency, WBF and aerodynamic power (Gordon and Dickinson, 2006). Our 2011 study supports this possibility as we found that increasing calcium concentration in IFM increases SA tension, power and muscle speed (Wang et al., 2011). Of particular relevance, we found that a 2-fold increase in IFM calcium concentration can produce a 2- to 3-fold increase in power generation and a 1.2-fold increase in f_max, which matches the aerodynamic power and WBF changes observed by Gordon and Dickinson. These significant and intriguing performance alterations at the whole-animal, fiber and molecular level demonstrate the crucial importance of the converter domain to the functional variation that enables muscle types to perform their specific tasks.

The control line, IFI (pwMhc2 in the original paper), was previously generated by integrating a genomic wild-type myosin heavy chain into the Drosophila genome (Swank et al., 2000). The creation of the experimental line with the native IFM myosin converter replaced with an embryonic version was performed as previously described (Swank et al., 2002). Briefly, the entire exon 11 coding region and flanking introns of pWMhc2, a wild-type Drosophila Mhc gene cloned into a P element vector (pCaSpeR), were replaced by alternative exon version 11d. This genomic construct was microinjected into Drosophila embryos and incorporated into its genome using P element-mediated transformation (Cripps et al., 1994). The transgenic lines were then crossed into an Mhc¹⁰ background, no myosin expressed in IFM (O'Donnell et al., 1989), to produce fly lines expressing only the transgenic myosin in IFM. Flies for all experiments were raised at 25°C.

Fiber preparation for mechanical evaluation was performed as previously described (Swank, 2012). In brief, a bundle of six dorsal longitudinal IFM fibers was removed from a half thorax of a 2–4 day old female fly. Fibers were then separated, split lengthwise, and chemically demembranated in dissection solution [pCa 8.0: 5 mmol l⁻¹ MgATP, 1 mmol l⁻¹ free Mg²⁺, 0.25 mmol l⁻¹ potassium phosphate, 5 mmol l⁻¹ EGTA, 20 mmol l⁻¹ N,N-bis(2-hydroxyethyl)-2-aminoethanesulfonic acid (BES, pH 7.0), 175 mmol l⁻¹ ionic strength, adjusted with sodium methane sulfonate, 1 mmol l⁻¹ DTT, 50% glycerol and 0.5% Triton X-100] for 1 h at 4°C. Aluminium T-clips (MicroConnex, Snoqualmie, WA, USA) were used to secure the ends of the fiber to a tension gauge (Kronex Technologies, Oakland, CA, USA) and a length motor (Physik Instrumente, Waldbronn, Germany). The muscle mechanics chamber was mounted on an inverted compound microscope to enable accurate fiber length and cross-sectional area measurements. A rectangular prism (Edmund Optics, Barrington, NJ, USA) was adhered to the glass bottom of the chamber so that the height of the muscle fiber, in addition to its length and width, can be measured using video analysis software (Ion Optix, Milton, MA, USA). The cross-sectional area was calculated using the equation for an ellipse. Temperature was set at 15°C for all mechanical measurements. The fiber was maintained in relaxing solution [pCa 8.0: 12 mmol l⁻¹ MgATP, 30 mmol l⁻¹ creatine phosphate, 600 U ml⁻¹ creatine phosphokinase, 1 mmol l⁻¹ free Mg²⁺, 5 mmol l⁻¹ EGTA, 20 mmol l⁻¹ BES (pH 7.0), 200 mmol l⁻¹ ionic strength, adjusted with sodium methane sulfonate, 1 mmol l⁻¹ DTT]; the length was adjusted until just taut, and then lengthened by 5% ML above just taut length. The fiber was activated to pCa 5.0 by partial exchange of the relaxing solution with activating solution (same as relaxing solution but with calcium concentration adjusted to pCa 4.0). As the sarcomere length range over which Drosophila oscillates its muscles during flight is not known, we empirically determined the starting length for our experiments by lengthening the fiber, in 2% increments, until the length that produced maximum power, as determined by sinusoidal analysis (see below), was found. Isometric tension measurements, step analysis and work loop analysis were performed at this optimal length. Fibers that did not produce power higher than 150 W m⁻³ for IFI or 60 W m⁻³ for IFI-EC, as measured by small amplitude sinusoidal analysis, were deemed to have been damaged during dissection and were discarded.

Work loop analysis was performed to determine the mechanical power output of fibers at approximate in vivo length changes and oscillation frequencies (Josephson et al., 2000). The muscle fiber was oscillated in a sinusoidal pattern at 25, 75, 100, 125, 150, 175 and 200 Hz and amplitudes of 0.5%, 0.75%, 1%, 1.25%, 1.5%, 1.75%, 2.0% and 2.25% ML to obtain maximum muscle power output. We started the sinusoidal pattern by lengthening the fiber from the shortest point in the length change cycle as this yielded the greatest power output as opposed to starting by shortening the fiber from the longest point or starting at the mid-point (Wang et al., 2011). For each frequency and amplitude, 10 identical sinusoids were applied to the fiber, and fiber length change and the tension response were recorded at a sampling frequency of 8 kHz. The values from the 8th cycle were used to determine the work and power output as power reached a constant value by the 6th or 7th cycle. Negative work absorbed by the fiber during lengthening (calculated as W=∫FdL, where F is tension and dL is fiber length change) was subtracted from positive work produced by the fiber during shortening to obtain net work production per cycle. Power was the product of work and frequency.

Isometric tension (A₀) was measured 1 ms prior to the start of the length step at optimal fiber length (Fig. 2). At pCa 5.0, a 1% ML step stretch over 0.5 ms, the fastest rate that can be applied on our mechanics apparatus, and 3.5 ms, an estimated lengthening rate for Drosophila IFM during flight (Swank et al., 2006a; Wang et al., 2011), were applied to the fibers at 15°C. Our work loop measurements and previous work have shown that 1% ML is optimal for wild-type power generation in the fly (Ramanath et al., 2011; Wang et al., 2011). After holding the fiber at the new length for 300 ms, the fibers were slowly, over 500 ms, returned to their original length. The tension response to the length step was analyzed. The total stretch-activated tension (A_SA) was calculated by subtracting total isometric tension (A₀) immediately before stretch from the delayed increased tension peak after stretch (Fig. 2). For the 0.5 ms length steps, A_SA was measured by averaging the tension trace points (sampled at 8 kHz) over 1 or 2 ms at each fiber's phase 3 peak. The phase 3 peak was typically 3.5–4.5 ms for IFI and 5–7 ms for IFI-EC as the IFI-EC peak tended to be broader than that of IFI. For 3.5 ms length steps, the peak occurred slightly later, at 5–6 ms for IFI and 6–8 ms for IFI-EC. If isometric tension decreased by more than 10% after a single stretch, data from that fiber were discarded.

The fiber was relaxed by replacing the bathing solution with relaxing solution (pCa 8.0), and passive isometric tension (P₀) was measured. Length step increases of the same amplitude and rate were imposed on the relaxed fiber to attain passive stretch characteristics (Fig. 2). Subtracting P_SA from A_SA yielded the increased tension due to active stretch (corrected active stretch tension, F_SA=A_SA−P_SA). P_SA and P₀ were measured at the exact same time as their active stretch counterparts (Linari et al., 2004; Wang et al., 2011).

Muscle stiffness (stress/strain) was determined by measuring the amplitude of phase 1 for 1% length step increases and subtracting the isometric tension value that immediately preceded the length increase. This was performed for both 0.5 and 3.5 ms duration length steps. Similar to SA and tension, stiffness was measured under relaxed (pCa 8.0) and active (pCa 5.0) conditions. Net active stiffness was calculated by subtracting passive stiffness from active stiffness.

Sinusoidal analysis was performed to determine the optimal length of the fiber and the influence of the higher ATP concentration used here compared with our previous study. A sinusoidal length change of 0.125% ML and a frequency set from 0.5 to 650 Hz were applied to the fiber. Power (W m⁻³) was calculated as π f E_v(ΔL/L)², where f is the frequency of the length perturbations (s⁻¹), E_v is the viscous modulus at f, and ΔL/L is the amplitude of the sinusoidal length change divided by the length of the fiber between the two T-clips (Swank, 2012).

To allow direct comparisons of muscle kinetics and flight parameters, flight assays were conducted at 15°C, the temperature at which muscle mechanical measurements were performed. Flies were moved to 15°C at least 1 h prior to the experiments. Flight ability was assayed by observing whether a fly was capable of flying up (U), horizontally (H), down (D) or not at all (N) when released in a Plexiglas flight chamber (Drummond et al., 1990). Flight index equals 6U/T+4H/T+2D/T+0N/T, where T is the total number of flies tested (Tohtong et al., 1995).

WBF of a fly, tethered by gluing monofilament fishing line to the fly's head, was determined using an optical tachometer as previously described (Hyatt and Maughan, 1994).

WSA was determined by measuring the maximum angle that a fly wing makes as it moves through a full wing stroke (Fig. 5). Two to three day old flies were anesthetized and 0.1 mm diameter tungsten wire tethers were attached to the thorax with Loctite Professional Superglue. The flies recovered for at least 3 h after tethering and were kept at 15°C for at least 1 h prior to testing at 15°C. Flies were suspended by adhering the free end of the tether to a custom-modified camera mount using modeling clay. The mount allowed camera distance and orientation to be consistent for each fly. Flies were aligned facing the camera at a slight downward angle such that the camera was at a 90 deg angle to the wing path during flight. This alignment angle was required as the wing path is not perpendicular to the long axis of the body (Fig. 5B). A Canon EOS Digital XTI camera with an EX Sigma 105 mm F2.8 DG Macro lens was used to take 10 pictures of each fly during uninterrupted flight. The camera shutter speed was 1/30 of a second with aperture F14, ISO 1600, and set on auto white balance. This resulted in about five wing strokes per exposure. Wing stroke angle analysis of photographs was performed using ImageJ. The angle made by the upper and lower limits of the composite stroke was measured (Fig. 5A). The two wing stroke angles were added together to determine the total wing stroke amplitude that contributes to aerodynamic power (Lehmann and Dickinson, 1997). The resulting values from each of 10 pictures were averaged to obtain wing stroke angle for each fly.

We thank Bernadette Glasheen for excellent technical assistance.