Molecular patterns of differentiation in canyon treefrogs (Hyla arenicolor): evidence for introgressive hybridization with the Arizona treefrog (H. wrightorum) and correlations with advertisement call differences

Abstract

Detection of genetic and behavioral diversity within morphologically similar species has led to the discovery of cryptic species complexes. We tested the hypothesis that U. S. populations of the canyon treefrog (Hyla arenicolor) may consist of cryptic species by examining mate-attraction signals among three divergent clades defined by mtDNA. Using a multi-locus approach, we re-analyzed phylogenetic relationships among the three clades and a closely related, but morphologically and behaviorally dissimilar species, the Arizona treefrog (H. wrightorum). We found evidence for introgression of H. wrightorum’s mitochondrial genome into H. arenicolor. Additionally, the two-clade topology based on nuclear data is more congruent with patterns of call variation than the three-clade topology from the mitochondrial dataset. The magnitude of the call divergence is probably insufficient to promote isolation of the nuclear-DNA defined clades should they become sympatric, but further divergence in call properties significant in species identification could promote speciation in the future.

Introduction

Theories of speciation lead to the general expectation that genetic differentiation among species or populations of wide-ranging species will be reflected in phenotypic traits. In particular, evolutionary biologists focus on differences in sexually selected traits because of their role in promoting reproductive isolation and speciation (Ryan et al., 1996; Tregenza & Butlin, 1999; Masta & Madison, 2002; Swallow et al, 2005). Indeed, differences in reproductive behaviors in animals breeding at the same time and place have often led to the discovery of cryptic species, defined here as morphologically similar species with high levels of genetic divergence (Stein, 1963; Zink & Johnson, 1984; Shaw, 2000; Bickford, 2007; Lemmon, 2007).

However, this correspondence of genetic and behavioral divergence between species does not always apply, and in fact comparisons of genetic and phenotypic divergence among populations of a single wide-ranging species often show a diversity of patterns (Coyne & Orr, 2004; Tregenza, 2002). Although some studies find the expected correlation (Christianson, 2005; MacDougall-Shackleton & MacDougall-Shackleton, 2001), many others do not (Tregenza & Butlin, 1999). For example, geographic distance was a better predictor of behavioral isolation than genetic distance in the plethodontid salamander, Desmognathus ochrophaeus (Tilley et al., 1990) and in the túngara frog, Physalaemus pustulosus (Ryan et al., 1996; Pröhl et al., 2006). Sexual selection was found to drive trait divergence in both jumping spiders (Masta et al., 2002) and frogs (Boul et al., 2007). Alternatively, sexually selected traits sometimes show less diversification than expected among genetically distinct populations, as in fanged frogs (Emerson & Ward, 1998) and stalk-eyed flies (Swallow et al., 2005). Thus within a species, complex interactions between evolutionary forces acting on local populations, such as counterbalancing natural and sexual selection, can lead to a discordance between genetic and phenotypic divergence.

Our original goal was to discover if behavioral divergence (advertisement calls) among populations of the canyon treefrog, H. arenicolor, corresponds to previously reported genetic divergence. Barber (1999) found three geographically isolated, genetically distinct, mitochondrial lineages within the U.S. portion of this species’ distribution: mean sequence divergences (ranging between 9.2 and 13.3%) were of sufficient magnitude to suggest the presence of three, allopatrically distributed species. Differences in advertisement calls have previously identified cryptic species of gray treefrogs (H. versicolor and H. chrysoscelis: Bogart & Wasserman, 1972; Ptacek et al., 1995; Holloway et al., 2006) and corroborated genetic divergence in chorus frogs (genus Pseudacris: Lemmon et al., 2007). Thus we sought to determine if call properties known to promote reproductive isolation in frogs differed among the three mtDNA-defined clades of H. arenicolor (Barber, 1999). If so, were the differences likely to be sufficient to promote between-clade discrimination or did they merely reflect relatively minor consequences of drift and selection expected in geographically isolated populations?

During the course of sampling calls, several studies were published that reinforced the need to corroborate phylogenetic patterns derived from mtDNA (Shaw, 2002; McGuire et al., 2007; Good et al., 2008). Molecular studies based on a single marker can be confounded by the effects of ancestral polymorphism, lateral gene transfer (introgression), or different rates of evolution among markers. Studies including multiple loci can help to avoid misinterpretation of phylogenetic relationships resulting from these effects (Moore, 1995; Ballard, 2003). We therefore compared mtDNA and nuclear markers.

We will show that the three-clade structure proposed by Barber (1999) is confounded by introgressive hybridization with another treefrog species, H. wrightorum. Moreover, patterns of call differentiation among H. arenicolor better fit a two-clade structure indicated by the nuclear data. Although the magnitude of the behavioral differences between the two nuclear-defined clades would almost certainly be insufficient to promote between-clade reproductive isolation at this time, our results identify the groups of populations with the greatest potential for speciation. Our results serve to emphasize the necessity of using multiple genetic markers when characterizing patterns of genetic differentiation.

Materials and Methods

Recording and tissue sampling

During 2000–2007 we recorded the advertisement calls of 74 males of H. arenicolor in 11 populations from throughout the U.S. range (Table 1; Fig. 1). For the genetic analyses, tissue from 28 individuals of H. arenicolor (see Fig. 1 for localities), H. wrightorum, H. eximia, and three outgroup species, H. femoralis, H. avivoca, and H. chrysoscelis were analyzed (Table 2). We analyzed the calls of three of the sixteen H. arenicolor used in the genetic analysis. Call data were collected from all of the same populations as the remaining individuals sampled for the genetic analysis.

Table 1.

Populations from which male calls were recorded, and number of males recorded (N).

Population	Clade	Latitude	Longitude	N
Davis Mountains, TX	1	30.700	103.500	2
Reynolds Creek, AZ	1	33.874	110.992	6
Lower Workman Creek, AZ	1	33.846	110.972	10
Parker Canyon, AZ	1	33.798	110.967	3
Pine Creek, UT	1	37.217	112.974	7
Madera Canyon, AZ	2	31.735	110.883	11
Box Canyon, AZ	2	31.798	110.777	3
Jacobson’s Creek, AZ	2	32.667	109.803	2
Ruby Road, AZ	2	31.397	111.139	2
Diamond Creek, AZ	3	35.766	113.373	26
Peach Springs, AZ	3	35.586	113.435	2

Figure 1.

Map showing U.S. portion of the H. arenicolor species range. Dashed lines indicate the distribution of Barber’s three mitochondrial clades, whose identities correspond to the number. Locations where tissue and male advertisement calls were sampled are indicated by the boxes. HAR 51, HAR 52, and HAR 3468 correspond to the locality where these tissue samples were collected from individuals other than those in Clade 3 that changed position in the nuclear gene tree relative to the mtDNA topology.

Table 2.

Locality information for specimens used in molecular analysis and Genebank numbers for sequences.

Species	Specimen	Population	Latitude	Longitude	Genebank
	Number				mtDNA	nDNA
					12S, 16S	Rag	Rhod	11T
Hyla arenicolor	HAR1	Peach Springs, AZ	35.586	113.435	GU989087	GU989059	GU944717	GU968612
Hyla arenicolor	HAR8	Diamond Creek, AZ	35.766	113.373	GU989077	GU989049	GU944707	GU968602
Hyla arenicolor	HAR11	Lower Workman Creek, AZ	33.846	110.972	GU989068	GU989040	GU944698	GU968593
Hyla arenicolor	HAR49	Houston Mesa, AZ	34.364	111.282	GU989065	GU989037	GU944695	GU968590
Hyla arenicolor	HAR50	Pine Creek, UT	37.217	112.974	GU989084	GU989056	GU944714	GU968609
Hyla arenicolor	HAR51	Rucker Canyon Road, AZ	31.753	109.418	GU989075	GU989047	GU944705	GU968600
Hyla arenicolor	HAR52	Jacobson's Creek, AZ	32.667	109.803	GU989067	GU989039	GU944697	GU968592
Hyla arenicolor	HAR53	Ruby Road, AZ	31.397	111.139	GU989086	GU989058	GU944716	GU968611
Hyla arenicolor	HAR63	Colorado river mile 246, AZ	35.824	113.648	GU989091	GU989063	GU944721	GU968616
Hyla arenicolor	HAR65	Colorado river mile 273, AZ	36.096	113.921	GU989092	GU989064	GU944722	GU968617
Hyla arenicolor	HAR3468	Davis Mountains, TX	30.685	104.078	GU989080	GU989052	GU944710	GU968605
Hyla arenicolor	HARB43	Santa Rita Field Station, AZ	31.739	110.866	GU989066	GU989038	GU944696	GU968591
Hyla arenicolor	HARA20	Box Canyon, AZ	31.798	110.777	GU989072	GU989044	GU944702	GU968597
Hyla arenicolor	HARB28	Cave Canyon, AZ	31.709	110.771	GU989070	GU989042	GU944700	GU968595
Hyla arenicolor	HARB9a	Peach Springs, AZ	35.586	113.435	GU989076	GU989048	GU944706	GU968601
Hyla arenicolor	HARB33	Diamond Creek, AZ	35.766	113.373	GU989071	GU989043	GU944701	GU968596
Hyla eximia	HEX31	Jalisco, MX	19.978	103.261	GU989088	GU989060	GU944718	GU968613
Hyla eximia	HEX32	Jalisco, MX	19.978	103.261	GU989074	GU989046	GU944704	GU968599
Hyla eximia	HEX370	Jalisco, MX	20.917	103.033	GU989083	GU989055	GU944713	GU968608
Hyla eximia	HEX369	Jalisco, MX	20.917	103.033	GU989078	GU989050	GU944708	GU968603
Hyla wrightorum	HWR58	Cochise Co., AZ	31.494	110.403	GU989085	GU989057	GU944715	GU968610
Hyla wrightorum	HWR59	Coconino Co., AZ	34.286	110.858	GU989069	GU989041	GU944699	GU968594
Hyla wrightorum	HWR60	Gila Co., AZ	34.253	110.844	GU989089	GU989061	GU944719	GU968614
Hyla wrightorum	HWR33	Pinetop, AZ	34.13	109.932	GU989079	GU989051	GU944709	GU968604
Hyla wrightorum	HWR34	Pinetop, AZ	34.13	109.932	GU989081	GU989053	GU944711	GU968606
Hyla avivoca	HAV	Macon Co., AL	32.476	85.607	GU989090	GU989062	GU944720	GU968615
Hyla chrysoscelis	HCH61	Phelps Co., MO	37.616	91.984	GU989082	GU989054	GU944712	GU968607
Hyla femoralis	HFE3858	Chatham Co., GA	31.999	81.12	GU989073	GU989045	GU944703	GU968598

Call Analyses

For each male, at least ten consecutive advertisement calls were recorded between 2100 and 0100 hrs using either a Sony stereo cassette recorder (TC-D5M), a Tascam DAT recorder (DA-P1), or a solid-state Marantz digital recorder (PMD670). A Sennheiser directional microphone (ME-66) with windscreen was positioned at 50–100 cm from the calling frog. The frequency-response of all three recorders was flat within ± 3 dB over the range from 30–17000 Hz, and the speed variation (which limits the accuracy of the measurement of temporal properties) was less than 0.06%. After each recording, we measured body temperature to the nearest 0.1 C using a Weber mercury quick-read cloacal thermometer. Calls recorded on tape cassettes were first digitized at a sampling rate of 44 kHz using Sound Edit 16 v.2 (Macromedia, 1996). We analyzed five sequential advertisement calls from each recording session using the Raven 1.2.1 software package (Cornell Lab of Ornithology, 2003); temporal properties were measured from oscillogram displays and spectral properties, from power spectra (Hamming window, FFT length = 1024 samples).

The advertisement call of H. arenicolor is composed of a series of pulses (Fig. 2a). The acoustic energy is concentrated in two broad spectral peaks; a series of sidebands are spaced at frequency intervals equal to the periodicity of the repeating waveform (Fig. 2c). Each pulse also possesses a sub-pulse structure (Fig. 2b). In addition to the high-frequency peak, which usually contains the greatest amplitude, we analyzed five temporal call variables: number of pulses; pulse duration; call period; pulse rate; and pulse rise-time (Fig. 2). We chose to examine these six call variables because they are known to be important for species recognition in other frogs (Gerhardt & Huber, 2002).

Figure 2.

a. Oscillogram of two H. arenicolor advertisement calls. Call period is defined as the time interval between the beginning of one call and the beginning of the subsequent call. b. Expanded view of two pulses from within a single call showing the distinctive sub-pulse structure. The pulse rate of the call is calculated as the inverse of the pulse period, the time interval between the beginning of one pulse and the beginning of the subsequent pulse. Pulse rise time is the time interval from the beginning of a pulse to the maximum amplitude of the pulse. c. Power spectrum of advertisement call. d. Oscillogram of a train of H. wrightorum advertisement calls. e. Expanded view of three pulses from within a single H. wrightorum call. Time scales are indicated within each oscillogram. A video showing male H. arenicolor and H. wrightorum calling can be seen at: http://www.biosci.missouri.edu/gerhardt/video/aren_wright.m4v.

We analyzed patterns of call variation using STATISTICA 5.5 software (StatSoft 2000). Before applying parametric tests, we tested all trait distributions for normality with a Kolmogorov-Smirnov goodness-of-fit test. Where necessary, data were log-transformed to meet assumptions of homoscedasticity and normality. Several call characteristics were found to be temperature-dependent, and so values for these characteristics were temperature corrected to 18 C using the results of a linear regression of the call variable on temperature.

We used a forward-stepwise discriminant function analysis to determine whether advertisement-call variables were effective in predicting membership in the three mtDNA clades proposed by Barber (1999), because this procedure allows for the a priori consideration of groups (James & McCulloch, 1990). The canonical variates that are generated summarize multivariate trait variation and produce corrected distances among groups. However, because call characteristics are frequently correlated, we also conducted a principal components analysis to generate a smaller number of uncorrelated variables, which were then subsequently used in a multivariate analysis of variance. Examination of the factor loadings allowed us to determine which particular call variables contributed most to behavioral divergence between the mitochondrial clades.

DNA Extraction, Amplification, and Sequencing

Total genomic DNA was extracted from EtOH-preserved toe and liver tissue with the DNeasy Tissue Kit (Quiagen, Inc., Valencia, CA). An approximately 2.5 kb segment of mitochondrial 12S and 16S genes and the intervening valine tRNA were amplified with a set of overlapping primers (Table 3, Goebel et al., 1999). Three nuclear gene fragments (a single exon of Rag 1, a segment from exon 4 of Rhodopsin, and a segment of an anonymous marker G11T) were amplified via polymerase chain reaction (PCR) and selected primer pairs (Table 3).

Table 3.

Primers used for amplification and/or sequencing.

Primer Name	Primer Sequence (5'-3')	Annealing Temperature (° C)	Source
MtDNA
MVZ-59	ATAGCACTGAAAAYGCTDAGATG	47	(Goebel et al., 1999)
tRNAVal	GGTGTAAGCGARAGCTTTKGTTAAG	47	(Goebel et al., 1999)
12SB-H	GAGGGTGACGGGCGGTGTGT	47	*
12L1	AAAAAGCTTCAAACTGGGATTAGATACCCCACTAT	47	(Goebel et al., 1999)
16sh	GCTAGACCATKATGCAAAAGGTA	47	(Goebel et al., 1999)
12sm	GGCAAGTCGTAACATGGTAAG	47	(Pauly et al., 2004)
16sa	ATGTTTTTGGTAAACAGGCG	47	(Goebel et al., 1999)
16sc	GTRGGCCTAAAAGCAGCCAC	47	(Pauly et al., 2004)
16sd	CTCCGGTCTGAACTCAGATCACTGAG	47	(Pauly et al., 2004)
16scr	GTGGCTGCTTTTAGGCCYAC	47	*
1066F	AGTACCGCAAGGGAAATATGA	47	*
nDNA
Rag1-C	GGAGATGTTAGTGAGAARCAYGG	55	(Biju and Bossuyt, 2003)
Rag1-D	GCTGCATTTCCRATRTCACAGTG	55	(Biju and Bossuyt, 2003)
Rhod1A	ACCATGAACGGAACAGAAGGYCC	57	(Bossuyt and Milinkovitch, 2000)
Rhod1C	CCAAGGGTAGCGAAGAARCCTTC	57	(Bossuyt and Milinkovitch, 2000)
11T_84F	TGGAGTACCCCTTTAAATCTGAAT	58	(Holloway et al., 2006)
11T_388R	ATAAAGTGCATAAGTAAGTAAAAGTGAA	58	(Holloway et al., 2006)
11T-A–F	ACCCTAAAAGAGCAAACGTC	55	(this study)
11T-A–R	GGCCCCTGGTCAGAGATAC	55	(this study)

All primers labeled * were designed in the labs of D. M. Hillis and D. Cannatella.

PCR amplifications of the mitochondrial fragments were performed with the following thermocycler protocol: 2 min at 94 C, denaturing for 30 sec at 94 C, annealing for 30 sec at 47 C, extension for 1 min at 72 C, and final extension for 7 min at 72 C; denaturing, annealing and the first extension stage were cycled 35 times. Nuclear fragments were amplified under the following conditions: 10 min at 95 C, denaturing for 1 min at 95 C, annealing at 55 to 58 C (see Table 3) for 1 min, extension for 1 min at 72 C, and a final extension at 72 C for 7 min; denaturing, annealing and the first extension stage were cycled 35–40 times. Products were visualized with a 2% agarose gel, and products were purified using a QIAquick PCR purification kit (Quiagen, Inc., Valencia, CA). Sequencing reactions were completed with an automated DNA sequencer (ABI 3730 and 3700 analyzers).

Alignment and Phylogenetic Analysis

Mitochondrial fragments were combined to give a contig of 2,453 bp. Alignment was performed using Clustal X; within the ingroup, very few regions were difficult to align, and these were adjusted manually to minimize the inferred number of evolutionary changes. Nuclear sequences were aligned and edited using SEQUENCHER 4.5 and rechecked by eye. Heteozygous sites were left as ambiguous and only one sequence per individual was produced per nuclear loci. One H. arenicolor individual, HAR52, was found to be heterozygous between mtDNA Clades 1 and 2. No heterozygous individuals between H. arenicolor and H. wrightorum were observed. The three nuclear fragments were run in separate maximum parsimony and maximum likelihood analyses. Nuclear fragments were then concatenated to give 960 bp.

Phylogenetic analyses were conducted by creating trees using maximum parsimony (MP) and maximum likelihood (ML) methods. Maximum parsimony trees were generated using PAUP 4.0b10 (Swofford, 2002). Both the mtDNA and nuclear data sets were analyzed using heuristic searches with 100 random addition sequences and TBR branch swapping. Nodal support values were obtained through bootstrapping with 2000 pseudoreplicates. For the maximum-likelihood analysis, Modeltest 3.7 (Posada & Crandall, 1998) was used for each dataset to select a best-fit model of nucleotide substitution. For the mtDNA, the selected model according to the Akaike Information Criterion (AIC) was: GTR+G+I. The HKY+I model was selected for the concatenated nuclear dataset. The program Phyml (Guindon & Gascuel, 2003) was used for the maximum likelihood analysis with 2000 bootstrap pseudoreplicates.

Results

Call Analysis

The average values of the call properties produced by males in each of the three mtDNA-defined clades are summarized in Table 4. Several of the call properties are quite similar among clades, with average differences less than 10% (pulse duration, pulse repetition rate, pulse rise time). Three call characteristics are more variable between mitochondrial clades: number of pulses, call period, and high-frequency peak. Nevertheless, there was extensive overlap in call properties between these groups in all pairwise comparisons.

Table 4.

Mean, standard deviation and average differences (given as percentage) between three mitochondrial clades for each call variable.

Character	Clade 1 (28)	Clade 2 (18)	Clade 3 (28)	Avg % Difference
Number of pulses	17.9 ± 2.6	15.0 ± 2.5	15.8 ± 1.8	11.0
Pulse duration (ms)	39.4 ± 4.9	42.3 ± 3.5	37.0 ± 3.4	8.5
Call period (sec)	3.600 ± 0.658	2.816 ± 0.450	3.662 ± 0.736	15.5
Pulse rate (1/sec)	16.30 ± 1.06	15.15 ± 1.12	15.53 ± 0.90	4.7
Pulse rise time (%)	69.1 ± 4.9	69.6 ± 5.2	65.4 ± 3.8	4.0
High-frequency peak (Hz)	2093.2 ± 251.8	2487.5 ± 206.9	2154.8 ± 294.1	10.7

Discriminant function analysis using the six call variables discriminated among the three mitochondrial clades. All six variables were retained in the significant model (Wilks' Λ= 0.308; F_12,132 = 8.81; P < 0.0001). Malahanobis distances confirm that calls produced by males from mtDNA Clade 2 are more distant from calls in Clades 1 and 3 than they are from each other in terms of multivariate call space (Table 5).

Table 5.

Results from discriminant function analysis using six call characteristics. Squared Mahalanobis distances are presented below the diagonal. F- and P-values are presented above the diagonal.

	Clade 1	Clade 2	Clade 3
Clade 1	-	10.3 (<0.00001)	6.3 (0.00003)
Clade 2	6.34	-	11.0 (<0.00001)
Clade 3	3.00	6.78	-

Many of the call characteristics analyzed are significantly intercorrelated (Table S1); therefore, the six call variables were subjected to a principal component analysis. The principal axis method was used to extract the components, and this procedure was followed by a varimax (orthogonal) rotation. The first two components displayed eigenvalues greater than 1, together accounting for 54.7% of the total variance. Corresponding factor loadings are presented in Table 6. In interpreting the rotated factor pattern, an item was considered to load on a given component if the factor loading was 0.60 or greater. Using this criterion, number of pulses, call period and high-frequency peak were found to load on the first component. The second component is comprised of pulse repetition rate and pulse duration.

Table 6.

Factor loadings for two principal components derived from principal components analysis using varimax rotation. Loadings of greater than 0.60 are indicated in bold.

Character	PC1	PC2
Number of pulses	0.62	0.31
Call period	0.76	−0.24
Pulse repetition rate	0.19	0.80
High-frequency peak	−0.69	−0.30
Rise time	−0.39	0.32
Pulse duration	0.05	−0.80
Eigenvalue	1.85	1.43
% Total Variance	30.9	23.8

Examination of the factor scores showed clearly that there was extensive overlap in call structure between mtDNA Clades 1 and 3 and that there was less overlap in the calls of these two clades with mitochondrial Clade 2 (Fig. 3). Using the two principal components as variables in a multivariate analysis of variance, we confirmed that males in the three mtDNA clades produced advertisement calls with significantly different call characteristics (Wilks’ Λ = 0.52, Rao’s R_4,140 = 13.51, P < 0.0001). Subsequent univariate analyses showed that the first principal component differed significantly among mtDNA-defined clades (F_2,71 = 24.6, P < 0.0001, Table 7). Post-hoc analyses indicate that males in Clade 2 produced advertisement calls that were significantly different from calls produced by males in Clades 1 and 3 (Tukey Honest Significant Difference Test for Unequal N, P’s < 0.001). Males in mtDNA Clade 2 produced short calls of high frequency at a rapid rate compared to males in the other clades. The second principal component did not differ significantly among mitochondrial clades (F_2,71 = 2.83, P = 0.066).

Figure 3.

Bivariate plot of principal components derived from six call characteristics.

Table 7.

Mean values for principal components in each of the three clades. Superscript letters indicate significant differences between mitochondrial clades (P < 0.001).

	Clade 1	Clade 2	Clade 3
PC1	0.50 a	−1.10 b	0.21 a
PC2	0.30 a	−0.40 a	−0.04 a

Phylogenetic Analyses

mtDNA analysis

Both analyses from the mtDNA sequence dataset confirm the phylogenetic structure described by Barber (1999), in which individuals fall into one of three distinctive genetic clades. Furthermore, as in Barber’s work, all H. arenicolor Clade 3 individuals group with individuals of H. wrightorum. Clade 3 individuals are confined to the Grand Canyon, with Clade 1 individuals found both north and south of it (Fig. 1). The left side of Figure 4 shows the ML tree for the mitochondrial data set.

Figure 4.

Gene trees based on mitochondrial DNA sequences (left) and nuclear DNA sequences (right). Numbers in parentheses next to each H. arenicolor individual indicates which mitochondrial clade they belong to as defined by Barber (1999). Individuals marked with * are H. arenicolor individuals other than those from Clade 3 that changed position in the nuclear gene tree compared to the mtDNA topology. Values in parentheses at nodes indicate ML bootstrap values and MP bootstrap values, respectively. Nodes that only have one value do not include maximum parsimony support values as these nodes fell into a polytomy in the MP analysis.

Nuclear DNA analysis

Tests for incongruency among the three nuclear fragments were run using the incongruence length difference (ILD; Farris et al. 1994) and Shimodaira-Hasegawa (SH; Shimodaira & Haesgawa, 1999) tests. The ILD test found all three fragments to be congruent (P’s > 0.05). The SH test found the Rhod and Rag fragments to be congruent with one another (P = 0.213); however, both were significantly incongruent with the third fragment, G11T (P’s < 0.05). Despite this inconsistency, all three nuclear fragments strongly support the removal of mtDNA Clade 3 individuals from the mitochondrial grouping with H. wrightorum and place them among the H. arenicolor mtDNA Clade 1 individuals. Some of the discrepancy is probably due to the placement of HAR52, the individual that is heterozygous at one locus (Rag) between H. arenicolor Clades 1 and 2. This individual’s placement has a low bootstrap value of 58. Furthermore the H. arenicolor individual from west Texas, HAR 3468, had a few unique substitutions, making its placement among the other H. arenicolor individuals less certain (bootstrap value 62). Finally, the largest discrepancy was caused by the placement of H. eximia individuals, in which they were either placed with H. wrightroum in a polytomy with the outgroup species (Rag), placed sister to H. wrightorum (Rhod), or placed sister to a group that included all H. wrightroum and H. arenicolor (G11T).

We assert that these incongruencies reflect the paucity of informative sites. Because all H. arenicolor group out from H. wrightorum in all three genes, we decided to concatenate the data, following Weins (1998). He posits that combining datasets, even if incongruent, can increase the accuracy in regions that are already well supported by increasing the number of informative sites. This method may be especially helpful when the number of informative sites is low, as is in our case. Out of 960 base pairs of nuclear sequence data, only 39 base pairs were informative.

Using a concatenated data set of all three nuclear regions, both ML and MP analyses strongly support the monophyly of H. arenicolor. Because MP and ML analyses are similar, we only show the ML tree along with ML and MP support values (Fig. 4). The nuclear analyses reveal that the mtDNA Clade 3 does not group with H. wrightorum, but instead individuals are placed with H. arenicolor Clade 1 individuals. A second group within H. arenicolor is composed of most of the mtDNA Clade 2 individuals. Also differing from the mtDNA topology, the individual HAR51 no longer groups in with Clade 1, but is instead found in the group with mtDNA Clade 2 individuals, and this placement is strongly supported (bootstrap of 84).

Discussion

This study was conducted to test the idea that genetic divergence might be paralleled by biologically significant differences in pre-mating isolating mechanisms in three mtDNA-defined clades, consisting of geographically isolated groups of populations. Our analysis of nuclear DNA, however, led to the discovery of a significant discrepancy between the nuclear and mtDNA gene trees. We hypothesize that the observed incongruity was caused by mitochondrial introgression between H. arenicolor from mtDNA Clade 3 and H. wrightorum.

Considering the extreme phenotypic differences between H. arenicolor and H. wrightorum, the history and ecology of introgression between them is arguably of more evolutionary interest than the potential reproductive isolation of different H.arenicolor groups. H. wrightorum and its sister species, H. eximia, are the closest extant relatives of H. arenicolor (Hedges, 1968; Barber, 1999); they are found throughout Arizona and Mexico, with distributions mostly occurring within the range of H. arenicolor (Fig.1). These two species differ greatly from H. arenicolor morphologically and behaviorally (Duellman, 1970; Gergus, 2004). They are smaller and have a smooth, typically green or dark brown skin compared to the bumpier gray and tan skin of H. arenicolor. Comparison of the waveforms reveals that the advertisement calls produced by these two species are strikingly different in temporal properties, call shape and pulse shape (Fig. 2). Although our study did not address variation in the advertisement calls of H. wrightorum, Gergus et al. (2004) found that this species’ average pulse repetition rate (108 pulses/s versus 15 pulses/s) and call duration (194 ms versus 975 ms) were quite different from those of H. arenicolor. Mixed-species breeding choruses of H. arenicolor and H. wrightorum commonly occur in northern and central Arizona, and field observations indicate that the dominant species at the chorus changes with the time of the breeding season and local patterns of rainfall (Gerhardt, pers. obs.). Mating mistakes involving interception of females by more abundant heterospecific males are well documented in anuran amphibians (Lamb & Avise, 1986; Gerhardt & Huber, 2002). Thus the possibility of mismatings between these two species is likely. Even though H. wrightorum is currently not found within the Grand Canyon, where the introgressed clade is found, dispersal of introgressed individuals from adjacent sympatric areas or the extinction of H. wrightorum from the Grand Canyon are likely scenarios.

Discrepancies between mtDNA and nuclear gene trees are often credited to either ancestral polymorphisms and incomplete lineage sorting, or introgressive hybridization (Moore, 1995; Avise, 2000). The first cause seems unlikely here, because of the long branch lengths separating the various groups. In both the mitochondrial and nuclear analyses, branch lengths subtending each major lineage are longer than branch lengths within groups. Thus, the mtDNA haplotypes of H. wrightorum and Clade 3 H. arenicolor are less divergent from one another than from the other H. arenicolor mtDNA clades. In contrast, nuclear sequences of individuals from the mtDNA Clades 1 and 3 are more similar to one another than to sequences of H. wrightorum. We interpret this pattern as resulting from past hybridization between H. wrightorum and H. arenicolor, which led to mitochondrial introgression. Our sampling suggests that the introgression is unidirectional with the mitochondrial genome of H. wrightorum being introgressed into populations of H. arenicolor and not vice versa, as only one H. arenicolor clade is nested within the mitochondrial group of H. wrightorum and H. eximia.

Evidence of cytoplasmic genome (chloroplast and mitochondrial) introgression in plant and animal taxa is well documented (Dowling & Secor, 1997; Wirtz, 1999; Shaw, 2002; Arnold, 2006; Baack & Riseberg, 2007; McGuire 2007; Good et al., 2008), and it has been found in anurans as well (Splosky & Uzzell, 1984; Lamb & Avise, 1986; Lemmon et al., 2007; Plötner et al., 2008). We did not find evidence of nuclear introgression between H. arenicolor and H. wrightorum, as no individuals shared alleles between these species. Several studies have shown mitochondrial introgression with little or no apparent nuclear introgression (Bernatchez et al., 1995; Glémet et al.,1998; McGuire 2008). Indeed, Bernatchez et al. (1995) found the mtDNA of an allopatric population of brook char to be identical to that of the arctic char, even though these brook char are indistinguishable from other brook char populations in terms of morphology and nuclear DNA, a situation similar to that of the H. arenicolor mitochondrial Clade 3. Furthermore, the calls of H. arenicolor from the introgressed Clade 3 are similar to those of H. arenicolor from other clades, and there is no resemblance to calls of H.wrightorum (Fig. 2), indicating that the introgressed genes have not affected call structure of these populations. Mitochondrial introgression may also have occurred between clades, as one individual from the Chirachuahua mountains of southern Arizona (HAR 51) no longer grouped with mtDNA Clade 1 individuals but was well supported in its placement with mtDNA Clade 2 individuals (Fig. 4). The other H. arenicolor individuals that differed in the nuclear topology (HAR 3468 and HAR 52) had low bootstrap support and so their placement cannot be considered as evidence of introgression. More extensive sampling of H. arenicolor (particularly in areas where clades may potentially overlap), H. wrightorum, and H.eximia, and the use of more highly informative markers is needed to better explore and understand this introgressive event.

Our call analysis found that call differentiation better reflects the nuclear gene topology than that of the mitochondrial gene tree. Calls of males from the southern portion of the U.S. range, mostly mtDNA Clade 2, differ from those of the other two mtDNA-defined clades (1 and 3), which are quite similar. Nevertheless, the magnitude of differences in call properties of known relevance to female anurans is unlikely to be sufficient to promote reproductive isolation should populations of different clades become sympatric in the near future. With regard to call properties useful for species identification in other hylid frogs, for example, consider pulse rate and dominant frequency. The average between-clade difference in mean pulse rate – the property most often used for species recognition in anurans – is of the order of 5%. Experiments using synthetic calls indicate that female hylids require a minimum difference of the order of 20% for effective discrimination (e.g., Gerhardt, 2005; review: Gerhardt & Huber, 2002). Variation in dominant frequency is correlated with body size in most hylids and many other kinds of frogs (review: Gerhardt & Huber, 2002); thus, species differences and geographic variation in this property could be the result of selection on body size as well as an effect of female choice. The mean high-frequency peak in Clade 2 is about 19% higher than that in Clades 1 and 3, but preferences based on frequency in frogs are easily abolished or reversed by differences in intensity that arise because of spatial distributions of males and females in choruses (review: Gerhardt & Huber 2002).

There is also extensive overlap between clades in values of other call properties (Table 4). Thus even if discrimination of differences in any of these properties were far superior in H. arenicolor to that in other frog species, females would usually encounter pairs of individual males of different clades with values that would not be sufficiently different (if at all) for mate identification. Future studies examining female choice between exemplars of pre-recorded calls from the two nuclear-defined groupings should test this conclusion directly. The use of natural vocalizations would also test for the possibility that discrimination could be based on combinations of call properties (e.g., Gerhardt & Brooks, 2009).

Our study adds to the literature on mitochondrial introgression in vertebrates, and supports the view that multiple lines of evidence are needed when inferring phylogenetic histories. We demonstrate inter-specific mitochondrial introgression between the canyon treefrog, H. arenicolor and its sister species H. wrightorum. Further sampling at the population level from these two species are required to explore the possibility of discovering more introgressive events, explore possible areas of current hybridization, and address hypotheses concerning introgression.