In Vivo Performance of Trunk Muscles in Tree Frogs During Calling

Sound-producing muscles are among the fastest known skeletal muscles in animals; e.g. the swimbladder muscles in toadfish (Opsanus tau) and in plainfin midshipman (Porichthys notatus), the tymbal muscles in the cicada (Okanagana vanduzeei), the shaker muscles of rattlesnakes (Crotalus atrox) and the trunk and laryngeal muscles in many species of anuran amphibians (Josephson and Young, 1985; Marsh and Taigen, 1987; Lindholm and Bass, 1993; Schaeffer et al. 1996; Rome et al. 1996). These muscles are selected to operate at high frequencies and in many cases must also produce high power outputs. High power output is required because effective signalling requires that the sound produced be loud (Martin and Gans, 1972), and coupling between the mechanical power produced by the muscle and the radiated sound is often inefficient (Prestwich et al. 1989). Because muscle power output is greatly influenced by contractile frequency, these muscles provide suitable models to study the limits of power production at high contractile frequencies. However, to design in vitro experiments that will realistically predict the limits of performance during in vivo movements, a detailed understanding of in vivo performance is necessary. The present study uses in vivo measurements to determine those parameters that may influence the power output of muscles involved in call production in amphibians.

Vocalization is an important aspect of the social behavior in anurans. Most male amphibians produce species-specific mating calls that attract only conspecific mates. These calls are often very loud (100–120 dB sound pressure level, SPL, at a distance of 50 cm) (Gerhardt, 1975; Littlejohn, 1977), high in acoustic energy and entail a very high energy cost. Calling can be sustained for prolonged periods in some species, and the muscles involved in calling appear to be specialized for aerobic metabolism (Marsh and Taigen, 1987). The rate of oxygen consumption during calling is almost twenty times higher than the resting rate (Taigen and Wells, 1985; Prestwich et al. 1989; Prestwich, 1994). In addition to these advertisement calls, anuran amphibians produce many other kind of calls, such as territorial calls and release calls (Rand, 1988). A call can consist of a single pulse of sound, or the sound amplitude can be modulated to produce a series of pulses, known as a trilled call.

Martin (1971, 1972) proposed that amplitude modulation of the sound could be influenced by the strength and pattern of contraction of the trunk muscles. He predicted that in some species the trunk muscles may contract continuously, causing a unidirectional flow of air into the buccal cavity and vocal sacs. In such species, the pulses would be produced as a result of passive vibration of structures within the larynx (classified as Type I calls). He appears to have based this prediction on some similarities between natural pulses and those produced during passive activation of the larynx by passing air through isolated preparations. In other species, the natural pulses of sound are quite different from those produced in vitro, and Martin (1971, 1972) suggested that these pulses are produced actively by repetitive contraction of the trunk muscles (Types II and III calls). Almost no data are available to evaluate this suggested difference in muscle activity between calls of Type I and those of Types II and III. Prior to the present study, no one has demonstrated an active role of the trunk muscles during advertisement calls; however, active involvement of the trunk musculature has been demonstrated during release calls in Bufo valliceps (Martin and Gans, 1972). Laryngeal muscles have been shown to be active during release and advertisement calls in Rana pipiens and H. versicolor, and Schmidt (1965, 1972) suggested that they play an active role in modulating the call structure. The muscles involved in sound production are composed entirely of fast oxidative glycolytic fibers with high aerobic capacities (Marsh and Taigen, 1987; Given and McKay, 1990). This composition is in marked contrast to the majority of anuran skeletal muscles, which consist predominately of fast glycolytic fibers with poor aerobic capacities (Putnam and Bennett, 1983).

Very few studies have been directed towards understanding the mechanical performance of the calling muscles of frogs. Two studies have presented data on the laryngeal muscles. Manz (1975) described the fusion frequency of laryngeal muscles in the European tree frog Hyla arboria, and McLister et al. (1995) presented a comparative study of the contractile properties of a laryngeal muscle (tensor chordarum) in three species of tree frogs (H. versicolor, H. chrysoscelis and H. cinereia). No previously published information is available on the contractile properties of the external and internal oblique muscles, which presumably power the pulsatile movement of air during a call. These trunk muscles must have two characteristic properties. First, they must produce high power, which allows the production of loud calls. Second, they must have a contractile frequency matching the pulse frequency within a trill, which can be very high for some species.

We selected two closely related species of North American gray tree frogs, H. versicolor and H. chrysoscelis. H. versicolor is a tetraploid species presumably derived from H. chrysoscelis (Bogart and Wasserman, 1972; Ralin, 1975). Despite their close morphological similarity, the pulse frequency of their trilled calls differs substantially (20–25 Hz for H. versicolor and 40–50 Hz for H. chrysoscelis at 25 °C) (Ralin, 1977; Gerhardt, 1978). These frogs are particularly interesting because they also provide an opportunity to compare in vivo performance of homologous muscles operating at considerably different contractile frequencies. The present study was designed to provide a more detailed understanding of the call mechanics and to measure the in vivo performance of trunk muscles during natural calling. We determined the amplitude and pattern of muscle length changes and made simultaneous EMG recordings from these muscles.

Hyla versicolor LeConte were collected in the vicinity of Boston, MA, USA, and H. chrysoscelis Cope were collected in the vicinity of Wilson county, TN, USA, during their active breeding season from mid May to early July. Animals were housed in glass aquaria in beds of sphagnum moss and were provided with a water source. Frogs were maintained under conditions similar to those found in their natural habitat during the breeding season (approximately 25 °C with a L:D cycle of 15 h:9 h). They were fed crickets supplemented with calcium carbonate and multi-vitamins every other day. Most of the data on both species were collected within the first week of captivity. Tapes of calls of H. versicolor and H. chrysoscelis recorded in the field were played every night for several hours to establish a chorus in the laboratory. Some of the animals were injected with crude pituitary extract, which is known to have a stimulatory effect on calling behavior (Schmidt, 1972). Rana pipiens pituitary glands were obtained from Carolina Biological Supply (North Carolina, USA). Animals were injected with 0.2 ml of a solution made by dissolving six glands in 1 ml of distilled water.

High-speed video was used to film changes in the diameter of the trunk during active calling. The electrical activity of the trunk muscles and the sound produced were recorded simultaneously with the video films. On the day of the recordings, animals were anesthetized with 3-aminobenzoic acid (MS 222), and EMG electrodes fashioned from stainless-steel wires with a coated diameter of 0.11 mm were surgically implanted. Two leads with approximately 3 mm of bare surface were placed subcutaneously approximately 1 cm apart along the fibers of the external oblique muscle, and a third lead was placed subcutaneously near the sacrum as a ground electrode. The recording electrodes lay on the surface of the external oblique muscle, which is immediately under the skin. Preliminary attempts to implant bipolar hook electrodes among the muscle fibers were not successful because the muscle is thin and easily ripped by such electrodes. Externally, the wires were encapsulated in silastic tubing. Animals were allowed to recover for at least 6 h in small plastic containers.

At the time of recording, the frogs were placed in an acrylic box on a piece of foam-core board marked with a 1 cm×1 cm grid and inclined at 45 °. The trunk movements were recorded with a NAC Visual Systems HV-1000 recorder running at 500 or 1000 fields s⁻¹ for H. versicolor and H. chrysoscelis, respectively. Signals from EMG electrodes were processed by a WPI DAM-50 preamplifier with high- and low-pass filters set to 10 and 3000 Hz, respectively. Preamplified EMG signals were recorded onto the video tape using a NAC Visual Systems wave inserter and simultaneously digitized using a 12-bit MacAdios A/D converter and Superscope software (from GW Instruments) running on a Macintosh IIci computer. Sample frequency for the A/D converter was set at 4000 Hz. Calls on the video tape were analyzed field by field using the still advance mode on a S-VHS Panasonic AG 1860 video recorder. After processing by a Hotronic time-base correction unit, the fields were captured with a Data Translation Quick Capture video board and analysed using NIH Image running on a Macintosh computer.

Changes in muscle length were calculated from measurements of the trunk diameter recorded in a dorsal view (Fig. 1). Preliminary recordings of a lateral view revealed that the shape of the trunk expressed as the ratio of the dorso-ventral to right–left diameters did not change during the call. Following the in vivo recordings, the animals were killed and their skins were removed to expose the external oblique muscle. The lungs were inflated by injecting air to produce trunk diameters similar to those obtained from the films during natural calling. Muscle lengths were measured at three different trunk diameters corresponding to the maximum, intermediate and minimum diameters recorded in vivo (Fig. 1). On the basis of the equation obtained from the relationship between the diameter at these three positions and the length of the muscle, muscle lengths were calculated for the entire call. Muscle lengths were converted to muscle strains (muscle length/starting length, L/L_i) and smoothed using the smoothing spline function in the application Igor (Wave Metrics). This minimal smoothing reduced measurement artifacts but retained all major features of the call. Fourier analysis showed that the smoothing progressively attenuated frequencies above 120 Hz and eliminated frequencies above 300 Hz. The resultant strain waves were further analysed to calculate shortening and lengthening velocities and cycle durations using pulse analysis functions in Superscope.

Sound signals during calling were recorded using a small microphone located in the acrylic chamber. Signals from the microphone were preamplified and then digitized by the MacAdios A/D converter.

All results presented are means ± standard error of the mean for fifteen calls produced by five H. versicolor (three calls per animal) and twelve calls by three H. chrysoscelis (four calls per animal).

The overall structure of the call was similar in both species. For each sound pulse, a corresponding EMG signal was recorded from the external oblique muscle (middle panel of Fig. 2A,B). The cycling frequency of the muscles was the same as the sound pulse frequency (lower panel of Fig. 2A,B). The number of pulses per call varied from 8 to 30 during the experimental sessions. The initial length (L_i) of the oblique muscles at the beginning of the call was longer than the length at which they operated during most of the call (see Fig. 1A).

From L_i, the muscle shortens progressively over the first 3–4 cycles (Fig. 2). For the remainder of the call, the muscle functions with a reduced strain as it cycles between a pulse minimum (L_p,min) and a pulse maximum (L_p,max). We estimated the total strain attained in the first part of the call by subtracting the mean of the first 6–8 values of L_p,min from L_i and dividing by L_i. The pulse strain for any given pulse of the call is defined as (L_p,max-L_p,min)/L_i. At the end of the active pulses, the muscle returns with a variable time course to its starting length.

Pulse frequencies within a typical call recorded in the laboratory for H. versicolor was approximately 25 Hz at approximately 25 °C (Fig. 2A; Table 1). At similar temperatures, H. chrysoscelis called with a pulse frequency of approximately 40–55 Hz (Fig. 2B; Table 1). These frequencies were similar to those reported from field recordings (Gerhardt, 1978; Ralin, 1977).

During calling in H. versicolor, the external oblique muscles experienced a mean total strain of 21.30±2.1 % and a mean pulse strain of 12.2±1.11 % (N=15 calls), while in H.

chrysoscelis, mean total strain and mean pulse strain were 19.13±0.99 % and 7.33±0.58 % (N=12 calls), respectively (Table 1). The 10 % difference in mean total strain between the two species was marginally statistically significant (P=0.043, U=18, Mann–Whitney non-parametric analysis). However, this difference was small compared with the 40 % difference in pulse strain (P=0.0004, U=0).

In both species, the muscle cycled through a length trajectory referred to here as a saw-tooth pattern. Each cycle was divided unequally, with a longer shortening phase compared with the lengthening phase. In H. versicolor, the shortening phase occupied approximately 75 % of the total cycle time, while in H. chrysoscelis the muscle shortened for approximately 65 % of the cycle time (Fig. 3A,B). In vivo shortening and lengthening velocities were calculated by differentiating the muscle length with respect to time and dividing by the maximum muscle length during a pulse (L_p,max), which was also close to the optimal length determined in vitro (M. Girgenrath and R. L. Marsh, unpublished results) (Table 1). The shortening velocities were 3.96±0.75 L_p,max s⁻¹ and 5.28±0.21 L_p,max s⁻¹ in H. versicolor and H. chrysoscelis, respectively. Lengthening velocities were very similar in both species (8.17±0.38 L_p,max s⁻¹ in H. versicolor and 8.37±0.25 L_p,max s⁻¹ in H. chrysoscelis) (Table 1).

Two clear biphasic EMG spikes were recorded corresponding to each strain cycle in H. versicolor (Fig. 3A; Table 1). In H. chrysoscelis, a single biphasic signal was recorded corresponding to each pulse (Fig. 3B; Table 1). The time lag between the beginning of the EMG signal and the beginning of shortening of the muscle is referred to as the electromechanical delay (EMD) (see Fig. 4). Early in the call of H. versicolor the EMD was 8–15 ms (Figs 4, 5). As the call proceeded, this electromechanical delay decreased, with a resultant negative EMD in the later cycles of the call. With the exception of one animal, a significant correlation existed between pulse number and EMD within individuals. Pulses with a negative EMD also experienced an increase in strain (Fig. 6). The EMD also declined in H. chrysoscelis with pulse number, but to a lesser extent than in H. versicolor (Fig. 5).

The time lag between the start of the EMG and the beginning of the sound signal was measured as the electroacoustic delay (EAD) (see Fig. 4). The EAD remained constant when plotted as a function of pulse number for both species (data not shown). The EAD ranged from 10 to 15 ms in H. versicolor and from 4.5 to 9.0 ms in H. chrysoscelis.

Following the last active shortening and corresponding pulse of sound, the muscle lengthened and began to shorten in the absence of electrical activity. Finally, the muscle returned to its starting length as the call terminated (Fig. 4B).

Anurans are known to produce some of the loudest sound signals for their body size compared with other terrestrial animals (Wells and Taigen, 1992). This high acoustic power output must, of course, be derived from the muscles involved in the production of sound, the external and internal oblique and laryngeal muscles. Data from our present investigations confirm that each pulse within the trilled calls of H. versicolor and H. chrysoscelis is produced following an active contraction of the external oblique muscles. Presumably, the internal obliques operate synchronously with the external obliques. If they did not, we would have expected to record a somewhat attenuated EMG signal out of phase with the large signal from the external obliques. Thus, the obliques have a contraction rate identical to the pulse rate within the call. Prior to the beginning of the call, the lungs are inflated with air (Fig. 1A). As the trunk muscles go through cycles of contraction and relaxation, this volume of air is cycled back and forth between the lung and the vocal sac, producing the individual pulses of a trill. Each pulse of sound is produced when the external and internal oblique muscles shorten, pushing the air out of the thoracic cavity through the larynx into the vocal sac (Fig. 1C). This air makes the vocal chords and associated cartilages vibrate and thus produce sound. At the end of each cycle, the muscle relaxes and relengthens passively as the air is returned from the vocal sac. The sound pulse appears to continue into the lengthening part of the cycle as the air passes back through the larynx into the lungs (Fig. 2). This continued sound production is interesting because it is powered by the elastic recoil of the vocal sac. Thus, elastic energy stored in the vocal sac is useful not only in returning air to the lungs but also in sound production.

Since call duration can be limited by lung volume (Martin, 1972), this mechanism of recycling the air by cyclical contraction of the internal and external obliques may allow the animal to produce longer and louder calls. The work done by a moving fluid is proportional to the pressure of the fluid multiplied by its volume. Producing a loud call thus requires moving a substantial volume of air. For example, in H. versicolor, the air moved in just a few pulses would empty the lungs and require reinflation by the relatively slow buccal pump if it were not recycled. Type I calls by small frogs that rely on a unidirectional airflow (Martin, 1971, 1972) must then be very short or of low intensity. An alternative mechanism, in which fresh air from outside would be used to reinflate the lungs between pulses, would increase the energy cost for producing a call of any given length and intensity.

The time between the electrical activity and the sound pulse, referred to as the electroacoustic delay EAD (Fig. 4), remains constant through the entire length of the call, indicating that a direct correlation exists between active contraction and sound production. The constant EAD might reflect that the beginning of sound production is coincident with passive opening of the larynx when a critical level of air pressure is built up in the lungs. Alternatively, the timing of laryngeal opening could be due to activity of the laryngeal muscles occurring with a constant phase relationship to the activation of the obliques.

The first three or four cycles of muscle contraction at the beginning of the call are often associated with little or no sound production (Martin, 1972) (Fig. 2). Over these first few cycles, the vocal sac becomes successively larger in diameter in each cycle (M. Girgenrath and R. L. Marsh, personal observation; Fig. 1). This increase in vocal sac diameter is probably associated with an increase in sound intensity (Gans and De Gueldre, 1992) and thus may be necessary for effective sound radiation. Vocal sac diameter increases to a maximum when the trunk diameter is at its minimum (Fig. 1C). The increased diameter of the vocal sac also increases the pressure in this structure and is probably necessary to provide the energy to refill the lungs quickly during the brief period during which the oblique muscles are relaxed. The EMGs are also smaller during the first few pulses, which could represent reduced recruitment. Alternatively, the increasing EMG amplitude could represent synaptic facilitation. An increase in amplitude of the postsynaptic potentials in response to a train of neural stimulation has been suggested to occur in the laryngeal muscles of male Xenopus laevis during call production (Tobias and Kelly, 1988).

The length pattern followed by the external oblique muscles during the major part of the call could favor enhanced performance of these muscles during each cycle and thus ensure pulses of high intensity. The starting length of the external oblique muscles is approximately 1.13 times the optimal length of the muscle (M. Girgenrath and R. L. Marsh, unpublished observations). These longer lengths result from the inflation of the lungs before the beginning of the call. The maximum strain is achieved progressively during the first 3–4 pulses. Once the maximum strain is achieved, the muscle cycles with a reduced strain for the rest of the call. Our preliminary data suggest that the muscle operates close to its optimal length during the major part of the call (M. Girgenrath and R. L. Marsh, unpublished results). Owing to the length-dependent nature of force generation in the muscle, operation close to the optimal length should allow the muscle to maximize force and thus power production in each cycle. The saw-tooth length cycle employed by the obliques during calling could be of particular advantage to muscles operating cyclically. During cyclical contractions, power is contributed only during the shortening phase of the cycle, while external work is necessary to relengthen the muscle before it can shorten again (Josephson, 1993). The saw-tooth pattern should allow the oblique muscles of the two hylid species to produce power over a longer period in each cycle. Also, in a system where muscle relengthening is passive (as in the present system, where lengthening of the muscle is due to recoil of the stretched vocal sac), it seems advantageous to spend less time in the passive part of the cycle. Adductor muscles in scallops have been reported to follow a somewhat similar length trajectory during swimming. The closure of two valves is accomplished by active shortening of the muscle (occupying 56 % of the total cycle time in Argopecten irradians and 65 % of the total cycle time in Chlamys hastata). Opening of the valves (lengthening of the adductor muscles) is due to the release of elastic energy stored in the hinge ligaments (Marsh and Olson, 1994; Marsh et al. 1992). Recent measurements of muscle length changes during slow flights in pigeons using sonomicrometers have also suggested a longer shortening phase compared with the lengthening phase (A. A. Biewener, personal communication). However, not all locomotor systems can take advantage of this type of cycle because they use symmetrically arranged sets of muscles to drive the movements. For, example, the muscles involved in fish swimming apply sinusoidal length patterns in which the lengthening of one set of muscles is accomplished by the shortening of an antagonistic set (Coughlin et al. 1996).

The quantitative differences between the calls of H. versicolor and H. chrysoscelis are probably important for the function of the muscles that produce the calls. Both total strain and pulse strain varied between the two species (Table 1). In H. chrysoscelis, the muscle experiences reduced total and pulse strain compared with H. versicolor. Because the oblique muscles in H. chrysoscelis operate at a much higher frequency, a reduced strain may allow the muscle to shorten with a velocity more favorable in terms of producing force and power (Josephson, 1993). However, despite the reduced strain, the external oblique of H. chrysoscelis shortens 1.35 times as fast as does this muscle in H. versicolor. These muscles can operate at a similar relative shortening velocity (V/V_max) in the two species only if H. chrysoscelis has faster muscles.

EMG patterns were similar in both species; however, in H. versicolor, there were two stimuli per cycle whereas only one stimulus was observed in H. chrysoscelis (Fig. 3; Table 1). Two stimuli per cycle could possibly generate more power at the operating frequency used by the oblique muscles in H. versicolor, allowing the muscle to be active through a longer part of the shortening phase. As cycle frequencies increase, extra stimuli may not always increase the power output (Stevenson and Josephson, 1990). Rather, the increased duration of activity may result in a lowering of the net power available because a higher residual force will be present during the lengthening phase. This higher force will require more work to relengthen the muscle. In H. chrysoscelis, the operating frequency of the muscle is much higher than in H. versicolor, and just one stimulus may be enough to keep the muscle active through most of the shortening phase.

The phase of the EMG with respect to the beginning of shortening is not constant throughout the call. At the start of the call, the EMG occurs several milliseconds before the beginning of shortening. This delay becomes shorter as the call proceeds, resulting in a negative EMD towards the end of the call. In the pulses with a negative EMD, the muscle starts to shorten prior to any electrical activity. The initial shortening of the muscle late in the call, which occurs before the activation of the muscle, is probably due to elastic recoil of the body wall after the larynx closes. The drift in the EMD may result from a mismatch between the frequency of the EMG and the inherent resonant frequency of the system. However, during these pulses, the muscle also experiences an increased strain compared with the pulses during the earlier part of the call (Fig. 6). The muscle lengthens to a greater extent because of delayed activation (Fig. 4). Thus, the time of active shortening of the muscle with respect to the EMG and the distance shortened actively by the muscle appear to remain constant. Further evidence of elastic recoil is seen after the last active pulse of the call, when a small amount of shortening occurs in the absence of any muscle activity (Fig. 4). Possible elastic structures include the skin and the muscles themselves. However, in the present system, the muscles probably operate at or below the optimal length and are unlikely to store energy elastically.

The functional role of the elastic recoil of the trunk is not clear. Energetic saving could be accomplished if internal kinetic energy could be saved in elastic elements and returned in the subsequent cycle as the body walls oscillate during a call (Alexander, 1988). The kinetic energy required to accelerate the body wall is calculated to be approximately 0.017 J kg⁻¹ muscle (using velocities measured from the video recordings and the total mass of the trunk muscles). However, this accounts for less than 0.5 % of the work done by the muscle during each pulse cycle (M. Girgenrath and R. L. Marsh, unpublished data) and it is therefore unlikely that elastic energy storage provides significant energetic savings during calling. In the absence of significant storage of kinetic energy in the trunk, the energy to stretch the elastic trunk must be provided by the pressure of the air returning from the vocal sac as it recoils elastically. After the glottis closes, this elastic energy could bring about the initial shortening of the trunk muscles. This energy return could increase the initial power output by compensating to some extent for the low power output during the first part of the cycle, which is occupied by the activation of the muscle. Of course, this mechanism would not increase the total power output of the obliques because the energy required to stretch the vocal sac must come from these muscles initially, and this energy is simply returned to the trunk. Further work is required to confirm the mechanism of loading of the elastic elements and how this energy saving contributes, if at all, to enhancing the overall performance of the system.

The saw-tooth pattern of the muscle length trajectory found during natural cycles could play a very important role in enhancing power production in this system and in other similar systems where muscle power is produced during only one phase of the cycle. A longer shortening phase in a length cycle can be of particular advantage for producing high power in muscles operating at high frequencies (where cycle time is a limiting factor), as it may allow a longer period in each cycle for producing positive work. Additionally, it may also allow the muscle to relax sufficiently before the beginning of the lengthening phase. For reproductive success, it is important to maintain sound power as well as the species-specific pulse frequency. Different strains and EMG patterns used by the two species could be the result of adaptations to optimize power while operating at their respective frequencies. However, the significance of this saw-tooth length trajectory, the applied strains and the EMG patterns used by these animals can be only evaluated by future in vitro studies.

We would like to thank Stefan Girgenrath for extending his help with the filming. This work was supported by a grant from the NIH (AR39318) to R.L.M.