The evolution of release and advertisement calls in green toads (Bufo viridis complex)

Abstract

In the present paper we tested the hypothesis that differences in the acoustic communication system of diploid and tetraploid green toads (Bufo viridis complex) might be due to selection for reproductive character displacement. We recorded two acoustic signals of the toad repertoire − the advertisement call (a long range mate-attracting signal) and the release call (a short-range signal mediating male–male interactions) − from six sympatric Central Asian populations (three diploid and three tetraploid populations) as well as from three allopatric diploid populations from Italy, and compared their patterns of variation with the pattern of among-population genetic distances. Although release and advertisement calls share the same morpho-physiological constraints, they show significantly different patterns of variation. Release calls vary congruently with the pattern of genetic distances, suggesting that mutation and genetic drift have been the major forces responsible for their change both in time and space. By contrast, the pattern of advertisement-call variation is not consistent with the phylogeny of the group, because the advertisement calls of Asian diploid and tetraploid populations differ from each other more than their genetic distances would predict. These results strongly support the hypothesis that selection acted on the advertisement calls of either or both Asian taxa, possibly, to favour reproductive isolation.

polyploidy, reproductive character displacement

Introduction

The evolution of mate-recognition systems plays an important role in evolutionary biology because changes in signals and/or in response to signals may produce reproductive isolation among populations and they may therefore result in the origin of new species (Dobzhansky, 1951; Coyne & Orr, 1989, 1997). But factors promoting or constraining evolutionary change in communication systems are often difficult to unravel because of the difficulty in separating cause and effect (Gerhardt, 1994). This is particularly true when investigating the interplay of ecological, genetic and behavioural factors in determining the outcome of interactions between partially differentiated populations (Gerhardt, 1999).

According to the reproductive character-displacement hypothesis, natural selection will operate on signallers, receivers or both to reduce mistakes in the identification of conspecific signals (Dobzhansky, 1951; Butlin, 1987, 1989). The higher the costs of mistakes (reduced hybrid fitness) the stronger the selection for assortative mating. Despite the general consensus this hypothesis received in the past, more recent studies have cast doubt on its evolutionary significance (Butlin, 1987; Gerhardt, 1999). One problem with reproductive character displacement arises from differences in definition: it sometimes being used to refer to the selective process favouring premating isolation in sympatry, and sometimes to the resulting spatial pattern of intra- and interspecific variation, where sympatric populations of the two taxa differ from each other more than they do in allopatry (Gerhardt, 1999). In this paper, we consider reproductive character displacement in the context of the evolutionary (historical) process that caused and maintains diversification of the mate-recognition system in sympatry between two sibling taxa.

From a theoretical point of view, reproductive character displacement is expected to occur only in highly restrictive conditions (Paterson, 1982; Spencer, Mcardle & Lambert, 1986; Butlin, 1989). In fact, when sympatric contact is first established, the two incipient species must have evolved some differences in their signals, and yet they must also make some errors in identification. Moreover, they must show genetic incompatibility or pay sufficiently high costs of misidentification. Finally, they should have limited gene flow with allopatric populations to avoid dilution of the selective effects evolved in sympatry (but see Liou & Price, 1994). The theoretical scepticism accords with that of many empiricists, who argue that examples of reproductive character displacement are not only extremely rare, but are often ambiguous because they cannot reject alternative explanations, in which reproductive isolation, rather than the target of selection, is the side-effect of the different evolutionary histories of the two populations in allopatry (review in Butlin, 1987; Howard, 1993; Gerhardt, 1999).

The lack of good examples of reproductive character displacement may be due to the rarity of the phenomenon or to the difficulty of it being documented. To solve the dilemma of whether sympatry causes character displacement or, alternatively, whether character differentiation in allopatry allows sympatry, we must be able to clearly formulate, rigorously test, and be able to reject several competing hypotheses (Noor, 1999). First, we should be able to reject the phylogenetic hypothesis (Sætre et al., 1997), according to which differences in mate-recognition systems are solely due to the accumulation of random mutations in allopatry. Second, we should be able to reject the pleiotropic hypothesis, according to which selection did not act on mate-recognition systems directly, but on characters to which signals are pleiotropically associated. Third, we should be able to reject the ecological hypothesis, suggesting that differences in mate recognition systems reflect different (allopatric) environmental selection for enhancing signal propagation.

In the present paper, we ask whether differences in the acoustic communication system of diploid and tetraploid green toads (Bufo viridis complex) might be due to character displacement. We test this hypothesis by comparing two functionally different signals of the toad repertoire: advertisement and release calls. Advertisement calls are long-range signals produced by males to attract females (Ryan, 2001). Release calls are short-range signals produced by males in response to erroneous male–male mating attempts (Brown & Littlejohn, 1972; Di Tada, Martino & Sinsch, 2001). Whereas advertisement calls have the potential to act as a premating isolation mechanism and to be under selection to diverge in sympatry, release calls are expected to undergo the opposite process (Gerhardt & Schwartz, 1995; Gerhardt, 2001; and references therein). Independently of receivers, ineffective release calls result in prolonged male–male amplexus, causing wastage of time and energy and, possibly, an increased predation risk (Blair, 1968). Therefore, release calls should be expected to be under selection to converge in sympatry (Leary, 2001).

We compare the observed patterns of call variation with those expected under the character displacement hypothesis (divergent advertisement calls, convergent release calls) and with those expected under the alternative hypotheses of genetic drift, pleiotropy and environmental selection (Table 1). If genetic drift has been the major evolutionary force responsible for call variation, we expect positive association between call and genetic differences among populations and congruency in the patterns of variation of release and advertisement calls. If pleiotropy (i.e. selection on body size, or a direct effect of genome duplication) has been the primary cause of evolutionary changes, we still expect congruency in the patterns of release and advertisement call variation, because pleiotropic effects do not depend on the call function, but we do not necessarily expect a significant association between call variation and genetic differences among populations. Finally, if environmental selection is responsible for the observed patterns of advertisement call variation, we can expect two different scenarios according to whether release calls, which are not directly affected by environmental selection, vary due to the accumulation of random mutations (genetic drift hypothesis) or to the pleiotropic effects of environmental selection on advertisement calls.

Table 1.

Patterns of call variation under different evolutionary processes

		Patterns
Causal hypothesis	Processes	Advertisement calls (Long- range communication)	Release calls (Short-range communication)	Congruency between patterns of advertisement and release call variation
Null hypothesis
Genetic drift	Accumulation of random mutations	Congruency with phylogeny	Congruency with phylogeny	Yes
Alternative hypotheses
Pleiotropy	Selection on genetically correlated phenotypic characters	Congruency with genetically correlated phenotypic traits	Congruency with genetically correlated phenotypic traits	Yes
Environmental selection	Selection on long-range signals	Congruency with environment	None	Yes, if release calls vary pleiotropically with advertisement calls. No, if release calls vary for genetic drift
Reproductive character displacement	Selection on signal and/or on response to signal	Divergence in sympatry	Convergence in sympatry	No

THE SYSTEM

The green toad's range covers much of Europe (except France and the Iberian Peninsula), North Africa, part of the Arabian Peninsula, and Asia as far East as north-western China and Mongolia (Dujsebayeva et al., 1997; Stöck, 1998; Stöck et al., 1999; Borkin et al., 2001). Tetraploid green toad populations have been found in Central Asia over a broad area from Turkmenistan to Kazakstan and Mongolia (Roth & Ràb, 1986; Borkin & Kuzmin, 1988). In these regions, diploid toads are found in lowlands, whereas tetraploids preferentially inhabit the mountains. However, despite the paucity of studies carried out in these regions, a number of lowland tetraploid populations have been also found (Pisanetz, 1978; Dujsebayeva et al., 1997; Borkin et al., 2001).

The advertisement call of green toads is a trill, a 3–6 s train of pulses emitted in a regular rhythm, each pulse showing a tonic structure with a fundamental (and dominant) frequency of 1.1–1.7 kHz (Giacoma, Zugoloro & Beani, 1997). Asian diploid and tetraploid advertisement calls markedly differ with respect to both spectral and temporal properties: diploid calls, on average, have fundamental frequencies 0.3 kHz lower than those of tetraploid calls and are 15% longer, with pulse rates 60% higher than those of tetraploid calls at the same temperature (Castellano et al., 1998). Playback experiments showed that both pulse-rate and fundamental frequency are important for species recognition in green toads (Castellano & Giacoma, 1998), and that Asian diploid females, when given a choice between a typical diploid and typical tetraploid call, significantly preferred the conspecific call (Giacoma & Castellano, 2001).

In a previous study (Castellano et al., 1998), we tested the reproductive character displacement hypothesis by comparing calls from one syntopic locality with calls from what we considered allopatric localities. These analyses did not support the character displacement hypothesis. However, recent findings of tetraploid toads in Kazakstan (Borkin et al., 2001) suggests that most of what we considered to be allopatric populations actually reside in broad areas of sympatry. For this reason, in the present paper we adopt a larger geographical scale and compare, within a clearly defined phylogenetic framework, (sympatric) Asian diploid and tetraploid populations with (allopatric) diploid populations from the westernmost part of the range.

Material and methods

STUDY SITES AND RECORDING METHODS

From 1994 to 1997, we recorded the advertisement and release calls of male green toads at eight breeding localities: five in Central Asia, two in the Italian Peninsula, and one in Sardinia (Fig. 1). All males from the Italian Peninsula (Puglia and Maremma), from Sardinia (Portoscuso), and from two Central Asian localities, Kopa (Kazakstan) and Tulek (Kyrgyzstan) had a diploid karyotype of 2n = 22 chromosomes; whereas all males from Big Lake (Kazakstan) and Isik-kul (Kyrgyzstan) had a tetraploid karyotype of 2n = 44 chromosomes. Finally, at Kok-Jar (Kyrgyzstan), we visited two breeding sites: in the first we found diploid toads only, whereas in the second, a few hundred meters from the first, we found both diploid and tetraploid toads (and several triploids) (see also Dujsebayeva et al., 1997; Castellano et al., 1998). In the present paper, we call ‘Kok-Jar2’ the sample of diploid males from both the first and the second breeding site, and ‘Kok-Jar4’ that of tetraploid males captured at the second site.

Figure 1.

Maps of the localities where green toads were collected. (1) Maremma, (2) Puglia, (3) Portoscuso, (4) Almaty, (5) Isik-kul, (6) Kok-jar2, (7) Kok-jar4, (8) Tulek, and (9) Kopa.

Calls were recorded with a condenser microphone (Sennheiser K3U-ME88) and with either analogue (Marantz CP-230, Marantz CP-430) or digital (SONY TCD-D100) tape recorders. We recorded male calling activity continuously until we got a sufficient number of release and advertisement calls (minimum = 3, maximum = 20, average = 7).

After recording the advertisement call, we caught males and measured their body temperature with either a mercury or a digital thermometer (APPA-model 51) having an accuracy of 0.5°C and 0.1°C, respectively. Most of the animals were taken alive to our laboratories where morphometrical, karyological and biochemical analyses were carried out (for more details see Castellano et al., 1998).

We recorded release calls by gently pressing the sides of a male held between thumb and forefinger directly above a microphone (Leary, 1999). Most animals were recorded once, either in the field, shortly after capture, or under laboratory conditions, within 48 h of being caught. However, to better analyse the effect of temperature on release calls, 11 males from five populations (Puglia, Maremma, Portoscuso, Kok-Jar2 and Kok-Jar4) were recorded repeatedly at different body temperatures. We individually housed these males in 20 × 30 × 30 cm plastic cages in a temperature-controlled chamber. We carried out recording sessions at nine different chamber temperatures (9, 12, 15, 19, 22, 24, 27, 30 and 34°C). Before each session, we let the animals acclimate at the room for 1–3 h. We measured male body temperatures both before and after their call recording, and employed the mean for successive analyses.

CALL ANALYSIS

Recorded calls were digitized (sample rate = 44.1 kHz, 16 bit) and analysed by Canary 1.1 software (Charif, Mitchel & Clark, 1993) on an Apple McIntosh IIic, as well as by Sound Forge 4.0 software on a IBM-compatible PC.

The advertisement calls (Fig. 2a) of the Asian and Sardinian toad populations were described in previous papers (Castellano & Giacoma, 1998; Castellano et al., 1998; Castellano et al., 1999; Castellano, Giacoma & Dujsebayeva, 2000), to which we refer the reader for more details. In the present work, we focus on the toad release calls and compare their pattern of variation with that shown by the advertisement calls.

Figure 2.

Oscillogram of (a) the advertisement call and (b) the release call.

On the waveform of a release call (Fig. 2b), we measured (1) call duration, (2) the number of pulses, (3) the pulse duration and the interpulse duration; whereas from the power spectrum we measured the fundamental frequency (FFT = 1024 points, FFT resolution = 43.07 Hz, and Hamming's sampling window).

ALLOZYME ANALYSIS

We kept tissue frozen at −80°C until used. We separately homogenized an approximately equal volume of liver and skeletal muscle (thigh) in an equal volume of grinding buffer (0.1 M Tris-HCl solution at pH 7.5, with 0.47 mM 2-mercaptoethanol, 1.07 mM Na-EDTA, and 0.52 mM NADP) and we centrifuged it at 17 000 g for 40 min. The supernatant was removed and stored at −80°C until it was employed in standard horizontal electrophoresis.

Allozyme electrophoresis was carried out on Cellogel sheets at 4°C using buffer systems and strains as described by Meera Khan (1971) and Mensi et al. (1992). The enzymes studied were: glucose-6-phospate dehydrogenase (G6PD), glutamate-oxalacetate-transaminase (GOT), hexokinase (HK1), phosphoglucose isomerase (PGI), 6-phosphogluconate dehydrogenase (6PGD), 1, 4, Glucose phosphate dehydrogenase (GPD), phosphoglucomutase (PGM), superoxide dismutase (SOD), isocitrate dehydrogenase (IDh-S), mannose phosphate isomerase (MPI) and malate dehydrogenase (MDh).

STATISTICAL ANALYSIS

Because temperature affects the temporal structure of both advertisement and release calls, we calculated residuals from the linear regression analyses between log-transformed body temperatures and call temporal properties on the whole data set. Release calls residuals were calculated by considering both the males recorded only once in the field and the 11 males recorded repeatedly at different temperatures under laboratory conditions. However, to avoid pseudo-replication, residuals of these 11 specimens were averaged so that all males contributed once to the population mean. Fundamental frequency is mainly influenced by body size rather than body temperature. Asian diploid toads are much larger than both Asian tetraploid and Italian diploid toads, and consequently call at lower frequencies (Castellano et al., 1998). Although these differences in frequency can favour reproductive isolation (Giacoma & Castellano, 2001), they might be the pleiotropic effect of selection for large size in Asian diploid toads, rather than selection for reproductive isolation. To prevent the confounding effects of pleiotropy and reproductive character displacement, we adopted the conservative choice of adjusting call frequencies for body size, by calculating residuals from the linear regression of log-transformed body size and log-transformed fundamental frequency.

To obtain a reduced number of uncorrelated variables to be used in the computation of among-population acoustic distances, we carried out Principal Component Analyses on the call character residuals. We then computed among-population Euclidean distances using these PCA scores. Separate analyses were conducted for release and advertisement calls. Finally, to calculate the correlation between the among-population Nei's genetic distances (Nei, 1978) and either the release call or the advertisement call distances, we employed a non-parametric test, the Mantel test of matrix association (Smouse, Long & Sokal, 1986; Manly, 1991). This analysis calculates the regression coefficients between a dependent distance matrix (either the release or the advertisement call distances) and an independent matrix (the genetic distances); by means of a randomization procedure it estimates the null-hypothesis probability that the regression coefficient does not significantly differ from zero.

Results

EFFECTS OF TEMPERATURE AND BODY SIZE ON CALL STRUCTURE

Advertisement and release calls show a markedly different structure. The advertisement call is a long-duration (3–8 s) train of pulses (minimum number = 30; maximum number = 190; mean = 81) emitted at regular rate (Fig. 2a). Pulses last for 20–30 ms and show a monotonic and regular amplitude modulation with peak of energy at the mid point. With respect to the advertisement call, the release call is a much shorter (0.07–0.41 s) train of pulses (minimum number = 3; maximum number = 14; mean = 7.1) (Fig. 2b). On average, pulses are shorter (minimum duration = 5.7 ms; maximum duration = 30 ms; mean = 11.9), although their duration may vary markedly within a single call, with the final pulse several times longer than the others.

Despite these conspicuous differences in structural design, advertisement and release calls are affected by temperature and body size in a qualitatively similar way. Temperature correlates negatively with duration of both advertisement (N = 158; R = 0.639; b = −0.234; F = 107.779; d.f. = 1, 156; P < 0.001) and release calls (N = 184; R = 0.454; b = −4.534; F = 47.363; d.f. = 1, 182; P < 0.001), it correlates negatively with pulse duration (advertisement call: N = 158; R = 0.782; b = −1.270; F = 246.233; d.f. = 1, 156; P < 0.001; release call: N = 181; R = 0.667; b = −0.475; F = 14.309; d.f. = 1, 179; P < 0.001) and interpulse duration (advertisement call: N = 158; R = 0.830; b = −2.378; F = 344.793; d.f. = 1, 156; P < 0.001; release call: N = 181; R = 0.728; b = −0.866; F = 201.267; d.f. = 1, 179; P < 0.001), whereas it correlates positively with pulse rate (advertisement call: N = 158; R = 0.639; b = 1.494; F = 522.794; d.f. = 1, 156; P < 0.001; release call: N = 182; R = 0.854; b = 2.07; F = 486.912; d.f. = 1, 180; P < 0.001). Body size correlates negatively with the fundamental frequency of both calls (advertisement call: N = 172; R = 0.804; b = −0.012; F = 311.247; d.f. = 1, 170; P < 0.001; release call: N = 95; R = 0.775; b = −0.01; F = 139.766; d.f. = 1, 93; P < 0.001), although the regression lines significantly differ in intercept (ANCOVA: N = 267; d.f. = 1, 264; F = 266.705; P < 0.001). The fundamental frequency of release calls is about 150 Hz lower than that of advertisement calls.

PATTERNS OF CALL VARIATION

Tables 2 and 3 show population means and standard deviations of temperature-adjusted acoustic properties of release and advertisement calls, respectively. Populations significantly differ with respect to all the release and advertisement call properties (ANOVA: P < 0.001). Table 4 shows the canonical loadings of the five principal components extracted from the release call properties. Only the first four differ significantly among populations (ANOVA: P < 0.007) and have been employed in the calculation of the Euclidean distances. Figure 3(a) shows release calls described by the first and second principal components. Asian diploid and tetraploid release calls are more similar to each other than they are to Italian diploid release calls. In fact, release calls of Asian diploid and tetraploid toads differ significantly with respect to the first principal component (absolute mean difference = 0.578; Tukey HSD multiple comparison test: P = 0.05), but not with respect to the second component (absolute mean difference = 0.457; Tukey HSD multiple comparison test: P > 0.05); whereas Asian tetraploid and Italian diploid toads differ with respect to both the first (absolute mean difference = 0.679; Tukey HSD multiple comparison test: P = 0.007) and the second (absolute mean difference = 0.820; Tukey HSD multiple comparison test: P = 0.003) component.

Table 2.

Mean population values of release call acoustic properties adjusted to a body temperature of 16 °C and a body size of 80 mm

		Call duration (s)			Pulse rate (s−1)			Pulse (ms)			Interpulse (ms)			Frequency (kHz)
Populations	N	Mean	SD	N	Mean	SD	N	Mean	SD	N	Mean	SD	N	Mean	SD
Maremma	(2n allopatric)	10	0.13	0.04	10	33.61	3.75	10	13.22	1.39	10	21.02	2.84	13	1.01	0.22
Puglia	(2n allopatric)	7	0.13	0.02	7	38.07	7.57	7	12.57	3.45	7	17.46	4.80	7	1.05	0.10
Portoscuso	(2n allopatric)	24	0.14	0.03	24	39.14	3.38	23	12.68	2.28	23	14.31	2.40	23	1.06	0.09
Kopa	(2n sympatric)	12	0.15	0.03	12	45.66	6.40	12	11.26	2.51	12	12.24	1.99	12	1.09	0.20
Tulek	(2n sympatric)	9	0.16	0.02	9	48.22	3.83	9	10.54	1.57	9	11.78	2.28	10	1.08	0.04
Kokjar 2	(2n sympatric)	4	0.20	0.02	4	40.81	3.43	4	12.39	1.64	4	13.50	1.49	4	0.93	0.05
Issik kul	(4n sympatric)	14	0.16	0.03	14	42.30	6.22	14	11.68	3.38	14	14.27	3.04	13	1.04	0.21
Big Lake	(4n sympatric)	12	0.18	0.04	10	44.07	5.71	10	10.15	2.32	10	13.33	3.64	10	1.03	0.07
Kokjar 4	(4n sympatric)	7	0.21	0.03	7	35.91	4.95	7	11.82	2.30	7	17.80	3.67	7	1.14	0.13

Table 3.

Mean population values of advertizement release call properties adjusted to a body temperature of 16 °C and a body size of 80 mm

		Call duration (s)			Pulse rate (s−1)			Pulse (ms)			Interpulse (ms)			Frequency (kHz)
Populations		N	Mean	SD	N	Mean	SD	N	Mean	SD	N	Mean	SD	N	Mean	SD
Maremma	(2n allopatric)	8	4.15	1.03	8	28.13	0.85	8	28.46	2.80	8	30.56	3.14	18	1.29	0.05
Puglia	(2n allopatric)	10	4.34	1.02	10	28.29	1.81	10	24.45	2.60	10	36.42	3.51	18	1.36	0.09
Portoscuso	(2n allopatric)	70	4.09	1.02	70	33.84	2.31	70	22.37	2.77	70	27.43	3.61	71	1.28	0.08
Kopa	(2n sympatric)	16	4.39	1.29	16	38.63	2.34	16	20.98	2.40	16	22.77	4.44	15	1.24	0.07
Tulek	(2n sympatric)	13	3.54	0.81	13	39.24	4.31	13	19.91	2.68	13	23.76	5.61	12	1.19	0.08
Kokjar 2	(2n sympatric)	12	4.37	1.30	12	36.50	2.70	12	23.53	2.25	12	23.93	2.31	10	1.20	0.06
Issik kul	(4n sympatric)	13	4.08	1.09	13	24.23	2.51	13	31.83	4.23	13	39.40	5.03	13	1.24	0.06
Big Lake	(4n sympatric)	3	3.63	0.50	3	25.70	2.75	3	31.68	4.80	3	33.90	3.43	3	1.14	0.06
Kokjar 4	(4n sympatric)	13	4.29	1.25	13	28.31	3.58	13	26.34	2.98	13	34.39	6.40	12	1.26	0.09

Table 4.

Canonical loadings of the five principal components extracted from the advertisement call properties, and the percentage of variation each component explains

	Principal component canonical loadings
	I	II	III	IV	V
Call duration	−0.255	0.656	0.699	−0.124	0.007
Pulse rate	0.984	0.128	0.009	0.059	0.112
Inter-pulse duration	−0.843	−0.011	−0.251	−0.469	0.078
Pulse duration	−0.747	−0.285	0.223	0.555	0.057
Frequency	−0.237	0.762	−0.524	0.297	−0.002
Percentage	47.147	22.189	17.525	12.698	0.44

Figure 3.

Scatter plot of the first and second principal components of (a) advertisement and (b) release calls (black circles, Asian diploids; white circles, Asian tetraploids; grey triangles, Italian diploids).

Table 5 shows the principal component canonical loadings of the advertisement calls. The first three components differ significantly among populations and have been employed in the computation of the among-population Euclidean distances. In contrast to the release calls, the Asian tetraploid advertisement calls differ more from the Asian diploid calls than from those of the Italian diploid toads (Fig. 3b).

Table 5.

Canonical loadings of the five principal components extracted from the release call properties, and the percentage of variation each component explains

	Principal component canonical loadings
	I	II	III	IV	V
Call duration	0.052	−0.779	0.212	0.585	0.063
Pulse rate	0.937	0.085	−0.071	−0.075	0.324
Inter-pulse duration	−0.813	−0.191	0.353	−0.331	0.259
Pulse duration	−0.566	0.500	−0.425	0.471	0.164
Frequency	0.141	0.626	0.721	0.260	−0.017
Percentage	37.647	25.833	17.520	14.933	4.067

PATTERNS OF GENETIC VARIATION

Table 6 shows the Nei's genetic distances (D) between the nine green toad populations. These distances were employed to generate the cluster shown in Figure 4. Italian, Asian diploid and Asian tetraploid green toad populations form three well distinguished groups. Furthermore, although karyologically different, Asian diploids and tetraploids are more similar to each other than either to Italian diploids.

Table 6.

Between-population Nei's genetic distances (SA, Sardinia; MA, Maremma; PU, Puglia; KO, Kopa; TY, Tulek; KK2, Kok-Jar 2; KK4, Kok-Jar 4; IK, Isik-kul; AA, Almaty; BK, Big Lake)

	SA	MA	PU	KO	TU	KK2	KK4	IK	BK
SA	–
MA	0.014	–
PU	0.121	0.089	–
KO	0.367	0.329	0.440	–
TU	0.343	0.334	0.465	0.008	–
KK2	0.336	0.351	0.499	0.027	0.005	–
KK4	0.468	0.420	0.279	0.102	0.115	0.139	–
IK	0.482	0.468	0.333	0.109	0.103	0.111	0.013	–
BK	0.473	0.454	0.321	0.104	0.101	0.111	0.009	0.002	–

Figure 4.

UPGMA cluster of electrophoretic distances among the nine populations of green toads.

ASSOCIATION BETWEEN PATTERNS OF GENETIC AND CALL VARIATION

There is a significant positive correlation (one-tailed Mantel test: b = 0.718; P = 0.0054, after 10 000 randomizations) between between-population release-call distances and genetic distance. In contrast, between-population advertisement call distances are not significantly associated with both release call distances (one–tailed Mantel test: b = −0.115; P = 0.7082, after 10 000 randomizations) and genetic distances (one-tailed Mantel test: b = 0.306; P = 0.2603, after 10 000 randomizations) (Fig. 5b). However, when we exclude from the analysis the distances between Asian diploid and tetraploid calls (which show higher than expected values), we still find a positive and significant association between call and genetic distances (one-tailed Mantel test: b = 1.01; P = 0.0195, after 10 000 randomizations).

Figure 5.

Scatter plot of Nei's genetic distances against the Euclidean distances on (a) the advertisement calls and (b) the release call. (Grey circles refer to Asian diploid–tetraploid comparisons.)

Discussion

Release and advertisement calls show significantly different structures, but share similar morpho-physiological constraints. In the proximate-cause domain, these design differences might be explained by differences in the muscular control of the anatomical structures involved in signal production (Martin, 1972; Martin & Gans, 1972; McClelland, Wilczynski & Ryan, 1996; McClelland, Wilczynski & Ryan, 1998). However, even though the neural software controlling for the two signals may differ, both the hardware (i.e. the phonatory organs) and the mechanism of signal production (active amplitude modulation) are the same, and this is why advertisement and release calls are similarly affected by body size and body temperature. In anurans, the fundamental frequency of the calls is determined to a large extent by the mass and tension of the vocal cords; the larger the animal the more massive his vocal cords, and the lower the frequencies of its call (Ryan, 1988). By contrast, body temperature influences muscle contractility and has marked effects on the temporal structure of calls (Cocroft & Ryan, 1995). Green toads use active amplitude modulation to produce pulses (Martin, 1972), therefore the higher the temperature the quicker the muscles can contract and the higher the pulse repetition rate of their calls (the active amplitude-modulating frequency).

Release and advertisement calls not only have different design features, but also show different patterns of variation: green toad release calls vary congruently with the phylogeny of the group, whereas advertisement calls do not. Such a different pattern suggests that the two signals might have experienced different evolutionary forces and that these forces operated on the active control of the phonatory organs (i.e. the neural control of laryngeal muscle movements) rather than only on the morphology of passively vibrating structures (i.e. the size and shape of the vocal cords).

One possible reconstruction of the evolutionary history of green toad communication systems, consistent to the present day pattern of variation, is shown in Figure 6(a) and involves genetic drift on release calls and divergent selection on Asian diploid and tetraploid advertisement calls. This hypothesis suggests that, within the green toad group, release calls did not undergo markedly different selective pressures, and they have changed in time and space mostly because of the accumulation of random mutations (genetic drift hypothesis). On the contrary, the Asian green toads experienced selective pressures on their advertisement calls, which were not encountered by the Italian toads. These selective pressures occurred after the polyploid mutation, they operated on either or both the Asian taxa, and caused divergent displacement between their advertisement calls.

Figure 6.

Two hypotheses on the evolution of the advertisement and release calls in green toads: (a) divergent Reproductive Character Displacement (d-RCD) on advertisement calls and genetic drift on release calls; (b) pleiotropic effects of polyploid mutation (Pl) and convergent Reproductive Character Displacement on release calls (c-RCD) (A-2n, Asian diploid toads; A-4n, Asian tetraploid toads; I-2n, Italian diploid toads).

Unfortunately, reconstructing the past is a difficult task because more than one story can often fit present-day observations. Green toads are not an exception to this rule. In fact, the inconsistency between the patterns of release and advertisement call variation could also be explained by a different evolutionary scenario, in which, together with genetic drift and selection, morphological constraints and pleiotropy play a crucial role.

SIGNAL CONSTRAINTS AND THE PLEIOTROPIC HYPOTHESIS

If we use the computer metaphor, viewing acoustic signals as the output from software (the neural module controlling for the muscular activity that regulates air flow through the larynx) installed on hardware (the anatomical organs involved in signal production), we can interpret signal constraints in light of hardware properties, whose variation significantly affects the software performance. Size and temperature might therefore be considered as hardware properties that influence signal production by acting on either the passively (body size) or actively (body temperature) vibrating structures of the larynx.

Such constraints play a crucial role in the evolution of signals (Ryan, 1988) because they establish a correlation between the software output and the hardware properties and therefore they enrich signals with information, which might become meaningful whenever the receivers are able to perceive it and to respond accordingly. In green toads, as in most anurans (review in Ryan, 1988), the fundamental frequency is negatively correlated with body size, and has the potential to encode signaller size information. In some toads and frogs, females prefer lower frequencies (Ryan & Keddy-Hector, 1992) and this results in sexual selection for larger males (Morris, 1991). A second reason why constraints are important in signal evolution is that they may indirectly expose signals to the effects of evolutionary forces. For example, the northern cricket frog, Acris crepitans, shows a longitudinal clinal variation in call frequency, which has been explained as the pleiotropic consequence of a clinal increase of body size due to selection for resistance to desiccation in the more arid western part of the range (Nevo & Capranica, 1985; but see Wilczynski & Ryan, 1999 for an example of size-independent frequency variation in cricket frogs).

Even genome duplication, by affecting muscle and neural responses, has been suggested to have indirect effects on phenotypic characters important for mate choice. Bogart & Wasserman (1972) suggested that tetraploid Odontophrynus americanus males call at lower pulse rates than their diploid counterpart because of their larger cell dimensions or tissue mass, directly due to genome duplication. Keller & Gerhardt (2001) tested the Bogart & Wasserman hypothesis by directly producing autotriploid offspring from diploid Hyla chrysoscelis and by comparing their calls with those of parents, with calls of tetraploid H. versicolor, and with those of triploid hybrids between H. chrysoscelis and H. versicolor. Autotriploid frogs, which showed erythrocyte nuclear areas intermediate between diploids and tetraploids, produced advertisement calls with a pulse rate 13% lower than that typical of diploids, but higher than the pulse rate of triploid hybrids. These results support the hypothesis that polyploidy might directly affect call structure, although the differences in pulse rate between tetraploid and diploid tree frogs are unlikely to have been caused by pleiotropic effects alone.

Because tetraploid advertisement calls show lower pulse rates than calls of Asian diploid toads, the pleiotropic hypothesis apparently holds. In this case, the evolutionary scenario would be as shown in Figure 6(b). Here, the Asian diploid–tetraploid ancestor already evolved different calls from the Italian toad ancestor and genome duplication then caused profound modifications in the tetraploid call structure, producing the present-day inconsistency between the phylogeny and the pattern of advertisement call variation. However, because genome duplication (like body temperature) affects the hardware of signal production (i.e. how quickly muscles can contract, or how quickly neural responses can be transmitted), genome duplication will also modify congruently the structure of the release call. We must therefore assume that selection for convergent character displacement (sensuGrant, 1972) operated either on Asian diploid or on tetraploid release calls (or both) to cancel out the effects of genome mutation.

The hypothesis of divergent character displacement on advertisement calls (Fig. 6a) is more parsimonious than the alternative hypothesis of pleiotropy and convergent character displacement on release calls (Fig. 6b). Furthermore, there is evidence that, in green toads, the ploidy level by itself does not affect the call parameters conspicuously: triploid green toads, in fact, produce advertisement calls that neither differ in pulse rate nor in frequency from those of tetraploid calls (Castellano et al., 1998).

SELECTION FOR ISOLATION OR SELECTION FOR ENHANCING SIGNAL TRANSMISSION?

If, for parsimony's sake, we accept the hypothesis of genetic drift on release calls and divergent selection on advertisement calls and reject the alternative pleiotropic and convergent selection hypothesis, we are still faced with the question of which functional properties of the advertisement call does divergent selection actually work on. Did selection favour signals that allow better discrimination between diploids and tetraploids (reproductive character displacement)? Or did selection simply enhance call transmission, indirectly causing character divergence because diploids and tetraploids experienced acoustically different environments (environmental selection)?

The process of reproductive character displacement is thought to result in a spatial pattern of intra- and interspecific variation where sympatric populations of the two taxa differ from each other more than they do in allopatry (Butlin, 1987, 1989). This work compares allopatric (Italian) and sympatric (central Asian) diploid populations with only sympatric tetraploid populations. Further works that also compare sympatric and allopatric tetraploid populations will permit a more rigorous testing of the reproductive character displacement hypothesis. It is important to notice, however, that the spatial pattern of character displacement is expected only when the two taxa first come into secondary contact (Waage, 1979; Marshall & Cooley, 2000). In fact, if allopatric populations descend from sympatric populations that have recently colonized new habitats, on no account should we expect them to differ from each other less than their ancestors did in sympatry. Therefore, although finding of a consistent spatial pattern of variation may provide strong support for character displacement, not finding it does not necessary mean that the process did not work somewhere and sometime in the past.

The environmental selection hypothesis (Ryan & Kime in press) provides us with less ambiguous predictions than those of character displacement. If differences between diploid and tetraploid advertisement calls are mainly due to selection for enhancing call transmission in highland and lowland habitats, we expect that: (1) highland and lowland habitats were acoustically different environments; (2) these acoustic differences had different effects on diploid and tetraploid calls; resulting in (3) calls that tended to fare better (either in absolute or relative terms) in the habitats where they are typically broadcast. Transmission experiments could easily be carried out to test these predictions (Kime, Turner & Ryan, 2000; Castellano, Giacoma & Ryan, in press). Rejection of the environmental selection hypothesis will further support the reproductive character displacement hypothesis.

Acknowledgement

We thank G. Odierna and G. Aprea for karyological analyses, and T. Dujsebayeva and V. Eremchenko for support in the field.