Male antiphonal calls and phonotaxis evoked by female courtship calls in the large odorous frog (Odorrana graminea)

Acoustic communication plays a vital role in frog reproduction. In most anuran species, long-distance sound communication is one-way from males to females; during the reproductive season, males produce species-specific advertisement calls to attract gravid females, and females are generally silent but perform phonotactic movements that lead to amplexus. One exception is the concave-eared torrent frog (Odorrana tormota). In this species, females produce courtship calls that elicit antiphonal vocalizations by males, followed by precise phonotactic movements. The large odorous frog O. graminea (previously Odorrana livida) in southern China is subject to the same environmental constraints as O. tormota, with which it is sympatric; it is unclear whether their sound communication is one-way or bidirectional. Here, we provide the first data on female O. graminea vocalizations and their functions. Using playbacks of female calls, we conducted acoustic behavioral experiments in the laboratory in response to which males emitted single- or multi-note antiphonal calls with a varying fundamental frequency. Moreover, they were attracted to female call playbacks, exhibiting precise phonotaxis. The female courtship call–male response interaction thus forms a duet between partners of a receptive pair. These results demonstrate that this unique communication system likely reflects an adaptation to an environment in which short-distance communication is at a premium given the high levels of ambient noise.


Introduction
Acoustic communication plays a vital role in frog reproduction. During the reproductive season, males of most frog species emit species-specific advertisement calls to attract gravid females; females without vocal sacs are generally silent but exhibit positive phonotaxis that leads to amplexus (Endler 1992;Bush 1997;Tobias et al. 1998;Ryan 2001;Narins et al. 2007). One exception is the concave-eared torrent frog (Odorrana tormota). In response to male calls, females of O. tormota produce broadband courtship calls with energy extending into the ultrasonic range; males can detect ultrasound, thus avoiding masking by intense, predominantly low-frequency ambient noise, especially after a heavy rainfall, averaging from 58 to > 70 dB SPL measured between 50 and 4000 Hz (with peak energy around 100 Hz) (Feng et al. 2002;Narins et al. 2004;Arch et al. 2011). Female calls typically elicit one precisely timed antiphonal vocalization from males, followed by male phonotaxis Shen et al. 2008;Shen 2018;Shen and Xu 2019) and the calls from male evoke phonotaxis and vocal responses from gravid females in O. tormota (Shen et al. 2011b). The large odorous frog (Odorrana graminea), a sympatric species, is subject to the same environmental constraints as O. tormota. Previous results show that males produce diverse broadband signals, most of which contain ultrasonic harmonics in O. graminea (Shen et al. 2011) and have the ability to detect ultrasound (up to 24 kHz) Shen et al. 2008;Liu et al. 2014). However, it is unclear whether female O. graminea are silent or produce courtship calls. The present study demonstrates that female O. graminea can produce courtship calls when oviposition is imminent. Female courtship calls are functionally a kind of fertility advertisement that evokes a male's immediate antiphonal responses and phonotactic movements, finally leading to amplexus. Thus, we believe that there is a bidirectional sound communication system between gravid females and receptive males in O. graminea. These results should stimulate further studies to provide new insights into the mechanisms underlying bidirectional acoustic communication in anuran amphibians.

Animal preparation
The large odorous frog (O. graminea Boulenger, previously O. livida) is an arboreal, nocturnal species living near noisy streams and waterfalls in mountainous areas (elevation 450-1200 m) of southern China. These frogs are sexually dimorphic, with adult males having a snout-vent length of ca. 48 mm, and females are much larger, having an S-V length of ca. 91 mm (Fei et al. 2009 Seven gravid females and 16 males of O. graminea were collected from Tau Hua Creek on rainy nights. In the field, an amplexed pair was uncoupled by hand and then kept separately in visually opaque but acoustically transparent plastic cages in a quiet, darkened room approximately 1 km away from their natural habitat.

Frog call recordings
Approximately, 30 min after capture, indoor recordings were carried out overnight prior to oviposition. In a quiet, darkened, anechoic room (3.0 × 2.5 × 2.5 m, with padded walls covered with acoustic foam, thickness 10 cm) under dim infrared illumination, one male was placed in an indoor black gauze-fenced arena (1.5 × 1.3 × 0.86 m), where a captive female in her cage was separated from a male in his cage by approximately 3 m. We conducted four recording sessions each night and each session was approximately 15 min at a temperature of 22 ± 2 °C and humidity of ~ 60%. In this way, most types of calls, including female natural calls, male elicited antiphonal calls and staccato calls, see Shen et al. (2011a), were recorded with a ¼-inch wide-band condenser microphone and a preamplifier (models 40BE and 26CB, respectively, G.R.A.S. Sound & Vibration, Holte, Denmark; frequency response, 4 Hz-100 kHz, ± 3 dB) placed above the cage and a digital audio recorder (Sound Devices model 722, Sound Devices, Reedsburg, WI, USA) with a sampling rate of 96 kHz and 16-bit resolution. The approximate distance between the microphone and the calling subjects was ca. 20 cm. The level of the recorded calls in dB SPL (RMS) was compared to the level of the tone recorded from the calibrator (Bruel and Kjaer 4231) that produces a 1 kHz tone at 94 dB SPL (RMS) with the same microphone.

Laboratory phonotaxis studies
Laboratory phonotaxis experiments were conducted in an anechoic room at a temperature of 22 ± 2 °C and humidity of ~ 60%. The room measured 3 m long, 2.5 m wide, and 2.5 m high, with 10 cm thick acoustic foam on all surfaces for sound absorption. Prior to the start of the experiment, three males were freshly captured and kept individually in small plastic cages each night. A rectangular arena measuring 2.0 m by 1.5 m was located within the room. During the phonotaxis experiments, each male frog was placed on the floor in the center of the short side of the arena (that is, the release site) under a removable glass cover. The frequency response of the loudspeaker (Fostex FE87E) was flat (± 3 dB) within the frequency range of 1-30 kHz; the speaker was placed 1.2 m away on the floor opposite to the release site, broadcasting female courtship calls (abbr. FCC) as the playback stimuli presented at one stimulus every 60 s and at a level of ~ 80 dB SPL, measured at the center of the arena. The duration of each FCC stimulus was 100 ms and that of each two-note stimulus (FCC1 + FCC2) was 350 ms. The general experimental protocol followed that described by . Single-note FCC or a 'two-note' call (FCC1 and FCC2) exemplars were selected for testing each captive male, as shown in Fig. 1a and b, respectively.
The animal's movements were monitored and recorded under infrared illumination by an infrared video camera (Sony DCR-TRV30E or HDR-SR7). All sound recordings from the field and the lab were transferred to the computer as WAV files, analyzed (fast Fourier transform, 1024 points), and displayed using SELENA software, a customdesigned program (S. Andrzheevski, St. Petersburg) (Narins et al. 2004;Feng et al. 2006;Shen et al. 2008), and PRAAT (Boersma and Weenick 2010) programs (see Feng et al. 2009). At the conclusion of each experiment, all animals were released back to their native sites along the creek.

Statistics
The trajectories for each responsive male frog in the phonotactic experiments were illustrated based on the video recordings.
Two parameters were measured: the jump distance (D, in cm) and azimuthal jump angle (α, in degrees). The azimuthal angle of each long-distance jump was calculated using the formula α = arc sin d /D, where d is the shortest distance from the Fig. 1 a- male's present position to the straight line between the male's initial position and the center of the loudspeaker.

Female natural calls in the laboratory
Prior to ovulation at night, captive gravid females (O. graminea) in an indoor arena produce simple calls repeatedly approximately once every 10 min. The female calls immediately stimulate male vocalizations. The audio and video recordings clearly show that the female's natural calls induce receptive male antiphonal responses followed by phonotactic movements, finally even resulting in amplexus. Therefore, such a call has been shown to be a highly attractive stimulus, called a 'female courtship call' (abbr. FCC). The FCC is a single note (Fig. 1a), or sometimes a 'two-note' call ( Fig. 1b), of shallow frequencymodulated stacks. The fundamental harmonic sweeps from 4.47 ± 0.57 kHz downward to 2.68 ± 0.61 kHz with a duration of 26.5 ± 8.3 ms, a mean fundamental frequency of 4.0 kHz and an average level of 85.3 ± 5.5 dB SPL (n = 82), measured at a distance of 50 cm above the female frog's head in the lab setup, much above the ambient noise level at the calling site, which averages from 58 to more than 70 dB SPL in the frequency range between 50 and 4000 Hz. A single-note FCC has several harmonics (Fig. 1a); the frequencies of the fundamental harmonic (F 0 or F 1 ) to fifth harmonic (F 5 ) marked by the red vertical line were 4.9 kHz, 9.8 kHz, 14.7 kHz, 19.6 kHz, and 24.6 kHz, respectively, and the call duration was ca. 40 ms with energy extending into the ultrasonic range (up to 36 kHz). The 'two-note' call ( Fig. 1b) has approximately the same spectrum as a single-note FCC, and the second note (FCC2) has a shorter duration (the difference approximately 8 ms) and is less intense (the difference approximately 5 dB), with an internote interval between FCC1 and FCC2 of 255 ± 47 ms (n = 14).

Male antiphonal calls in the laboratory
When a male frog in captivity in a quiet darkened room was separated by black cotton gauze mesh from a gravid female frog in another small chamber, the male was often induced to vocalize by the female courtship call. Six examples of immediately evoked antiphonal responses (abbr. AR) are shown in Fig. 2. The latencies to the male's first AR were 0.2 s, 0.35 s, 2.7 s, 1.5 s, 1.6 s, and 1.2 s, respectively, corresponding to Fig. 2a-f, measured from the onset of the FCC or the first note of the female call (FCC1).
The AR was generally a short upward-downward tonal call with a duration of approximately 60-85 ms (e.g., Fig. 2a, b), exhibiting a number of harmonics and different F 0 , although recorded from the same male. The mean F 0 of a single AR note was 4.47 kHz, the maximum F 0 was 4.75 kHz, and the minimum F 0 was 4.15 kHz, as shown in Fig. 2a, and the frequencies of the harmonics marked by the red vertical line were 4.9 kHz, 9.8 kHz, 14.7 and 19.6 kHz, respectively. In Fig. 2b, the mean F 0 of a single AR was 5.66 kHz, the maximum F 0 was 6.26 kHz, and the minimum F 0 was 4.89 kHz, and the frequencies of these harmonics marked by the red vertical line were 6.3 kHz, 12.6 kHz, 18.9 and 25.2 kHz, respectively.
Sometimes the ARs were short frequency-modulated (FM) calls, frequently followed by a multi-note long call with a duration of 600-800 ms, including 4-5 separate notes, as shown in Fig. 2c-f. The analysis results show that the mean F 0 of an individual AR note varied. For example, the ARs were often quite complex when evoked by a single FCC (Fig. 2c, f) or a 'two-note' call (FCC1 and FCC2), as shown in Fig. 2d, e.
The mean F 0 of eight elicited AR notes in Fig. 2c was 3.0 kHz, 2.6 kHz, 0.6 kHz, 4.1 kHz, 4.7 kHz, 5.9 kHz, 4.3 and 4.5 kHz. In Fig. 2d, the mean F 0 for seven elicited AR notes was 2.6 kHz, 2.6 kHz, 3.6 kHz, 3.9 kHz, 5.0 kHz, 4.9 and 5.8 kHz. In Fig. 2e, the mean F 0 for five elicited AR notes was 2.6 kHz, 2.7 kHz, 3.1 kHz, 4.0 kHz, and 3.5 kHz. In Fig. 2f, the mean F 0 for eight elicited AR notes was 2.2 kHz, 1.8 kHz, 2.0 kHz, 2.3 kHz, 2.6 kHz, 3.3 kHz, 11.6 and 12.7 kHz. Moreover, the last long call in Fig. 2c-f was similar to a singing tenor characterized by pronounced and varying frequency modulation patterns, the occurrence of nonlinear phenomena (i.e., frequency jumps, subharmonics, and deterministic chaos), and having energy in the ultrasonic range up to ca. 42 kHz. Female O. graminea produced simpler calls, as shown in Fig. 1a-b. We believe that male vocalizations in this species, as shown in Fig. 2c-f, have a highly variable and complex call production mechanism, akin to that of O. tormota (Suthers et al. 2006), which needs further study. To the best of our knowledge, only males of O. graminea have such a complicated and varied antiphonal response. Its functional significance may be an important trait that transmits a male's fitness to the female. These ARs are rare in other anuran species (Ryan 1985;Narins et al. 2000).

Male phonotaxis in the laboratory
We found that in a quiet, dark indoor arena, newly captured males were most responsive during the phonotaxis experiments. In response to the FCC stimuli at a playback level of 85 dB SPL, in addition to vocal responses evoked, males rapidly and accurately localized the loudspeaker. The  Fig. 2 a-  representative phonotactic trajectories elicited from the males by the FCC playback are illustrated in Fig. 3. The FCC playback resulted in the attraction of receptive males to the speaker (for one example, see Supplementary Video). The video shows that upon hearing the FCC playback, a receptive male of O. graminea immediately turned his body towards the speaker and made a long jump ca. 100 cm with a minimum latency of less than 1 s and extraordinary localization acuity, as illustrated by the blue trajectory in the middle part of Fig. 3. Most males would go directly to or near the loudspeaker, sometimes touching the foam on the wall above the loudspeaker and falling. A small number of males jumped a short distance (ca. 30 cm) and then quickly crawled toward the playback speaker. When the male frog heard the second or third acoustic stimulus (FCC playback), it would immediately orient itself, adjust its trajectory, take a large leap, and make contact with the speaker, possibly leading to amplexus if the sound source were a live gravid female frog. Therefore, we believe that FCC is a salient stimulus and has a significant attraction to males.
The precision of the long-distance jumps (more than 30 cm) of receptive male O. graminea was remarkable, with an average azimuthal error of jumps (α) of just 0.7° ± 0.8° (n= 14), even smaller than that in male concave-eared frogs . This accuracy suggests that large odorous frogs have the capacity to localize their mates in noisy, dark habitats.

Discussion
The large odorous frog O. graminea is an arboreal, nocturnal species that inhabits the vegetation along noisy streams, rendering acoustic communication challenging; one solution is to advertise conspicuously. Males of O. graminea have been shown to produce various calls with spectral energy extending into the ultrasonic range to avoid masking from broadband ambient noise and to detect ultrasound up to 24 kHz (Shen et al. 2011a;Liu et al. 2014;Shen 2018). In this study, we found that similar to O. tormota, female gravid frogs of O. graminea also emit courtship advertisement calls when oviposition is imminent. Functionally, female courtship advertisement calls elicit both male vocalization and approach. Therefore, the female call-male answer interaction forms a duet between partners of a receptive pair. In addition, we found that the playback of the male advertisement call can induce accurate phonotactic behavior of female frogs of O. graminea as well (not shown here), similar to female O. tormota (Shen et al. 2011b), and the main characteristics are the short distance jump and rapid crawling due to their larger weight. We believe that bidirectional sound communication indeed plays an important role in both O. graminea and O. tormota reproduction. We speculate that the lack of specialized structures in the external ear and the larger body size of O. graminea relative to O. tormota do not significantly affect the phonotactic results reported, although there are still many differences in auditory frequency sensitivity and spectral characteristics of antiphonal calls in these two sympatric frog species (Shen et al. 2011b;Liu et al. 2014). The present results further show that in comparison with usual advertisement calls in O. graminea (Shen et al. 2011a), receptive males emit distinct antiphonal calls of longer duration, in particular, with drastic changes in pitch to overcome the masking effects of ambient noise, to increase the salience of the communication signal, to attract gravid females, to facilitate amplexus and to ensure successful reproduction in the field.
In the laboratory, the fact that female call stimuli emitted from a point source (speaker) reliably elicit the male's first antiphonal response with unequal latency ranging from 0.2 to 2.7 s (see Fig. 2) suggests that recognition, rather than localization, is likely the primary problem. We observed that when the second or third stimulus was presented to the male frog during his transit toward the loudspeaker, the frog suddenly adjusted its body orientation and then made a precise jump toward the sound source. This suggests that for the male frog, the recognition comes first, and the localization is secondary.
The present studies indicate that males of O. graminea can localize a calling female or a loudspeaker with an extraordinary acuity of less than 1°, despite their small head size (interaural distance less than 2 cm). However, most amphibians are less well endowed, generally showing an acuity of approximately 16-23° (Christensen-Dalsgaard 2005), as they locate a sound source based on low-frequency perception. In contrast, ultrasonic males of O. tormota have the capacity to perceive higher frequency sounds as an adaptation to their noisy habitats, which may underlie their hyperacute sound localization. Additional mechanisms that underlie localization hyperacuity in O. graminea remain to be studied.
It is worth mentioning that we conducted field investigations along Shunyan Creek (Hongyangou) Hejiang, Sichuan Province, China (28.640° N, 106.138° E) on 30 May-10 June 2009 and on 28 May-7 June 2010, respectively, at elevations > 701 m. Figure 4 shows the main members of the research team investigating the habitat of the large odorous frogs (O. graminea) along Shunyan Creek in June 2009. We recorded sporadic calls of male frogs in this very noisy habitat. No gravid female frogs were found. When playing back the female calls recorded from Huangshan, Anhui Province,