Physiological implications of genomic state in parthenogenetic lizards of reciprocal hybrid origin


Michael Kearney, Department of Zoology, The University of Melbourne, Parkville, Vic. 3010, Australia.
Tel.: +613 8344 4864; fax: +613 8344 7909; e-mail:


Parthenogenesis often evolves in association with hybridization, but the associated ecological consequences are poorly understood. The Australian gecko Heteronotia binoei is unusual because triploid parthenogenesis evolved through reciprocal crosses between two sexual lineages, resulting in four possible cytonuclear genotypes. In this species complex, we compared the performance of these parthenogenetic genotypes with their sexual progenitors for a suite of physiological traits (metabolic rate, thermal tolerance, locomotor performance, and in vitro activity and gene sequence divergence of a cytonuclear metabolic pathway, cytochrome C oxidase). Mass-specific metabolic rate scaled differently with body mass for parthenogens and sexuals, while heat tolerance provided the only evidence for cytonuclear incompatibility in hybrid parthenogens. The most prominent phenotypic effects were attributable to nuclear genome dosage. Overall, our results suggest that the hybrid/polyploidy origin of parthenogenetic H. binoei has had surprisingly few negative fitness consequences and may have produced a broader overall niche for the species.


Parthenogenetic organisms should theoretically outcompete and replace sexual lineages through a twofold reproductive advantage over all-female reproduction (Maynard Smith, 1971). It has been difficult to conceive of an advantage to sex so powerful as to cover this twofold cost (West et al., 1999; Birdsell & Wills, 2003), yet naturally parthenogenetic organisms are taxonomically rare, making up one of every thousand species (White, 1978). Resolution to this paradox may in part reflect the constraints on the evolution of parthenogenesis.

Many parthenogenetic organisms evolved in association with hybridization events, including all but one of the known vertebrate cases (Avise, 2008; Kearney et al., 2009; Sinclair et al., 2010). While the causal link between hybridization and parthenogenesis is unresolved, one possible scenario is that hybridization between divergent sexual ancestors disrupted meiotic pathways (Moritz et al., 1989). If this were so, we could expect other disruptions caused by hybrid incompatibility to lower the viability of parthenogenetic individuals produced in such a manner, eroding the twofold advantage to some degree.

The fitness consequences of hybridization can range from complete inviability to enhanced fitness (hybrid vigour) (Arnold, 1997; Crow, 2000; Barton, 2001). Hybrid offspring may have intermediate phenotypes through additive genomic interactions (Schlosser et al. 1998). They may also exhibit novel or extreme phenotypes as a result of nonadditive genomic interactions, which may positively or negatively impact whole-organism physiology (Rieseberg and Carney 1998). Hybrid breakdown is a reduction in fitness caused by incompatibilities between divergent genomes (Coyne and Allen Orr 1989; Coyne and Allen Orr 1997; Cullum, 1997; Grossman et al., 2001; Ballard and Rand 2005; Ellison and Burton 2006; Dowling et al. 2007). Such incompatibilities are likely responsible for the general tendency for hybridizations between species are often found to be unidirectional. This phenomenon has been called ‘Darwin’s Corollary to Haldane’s Rule’ (Turelli & Moyle, 2007) and states that males of one species may successfully hybridize with females of another, yet the reciprocal cross fails (Tiffin et al., 2001). This is frequently the case in hybrid parthenogenesis, with mitochondrial DNA analysis often revealing that all known lineages have the same maternal genome (for animal examples see Moritz et al., 1992a; Kearney et al., 2006; Moritz et al., 1992b; Avise, 2008; Murphy et al., 2000). Recent theoretical and empirical studies suggest that asymmetrical hybridization is due to unequal evolutionary rates between uniparentally inherited genes, such as those of the mitochondria, compared to the biparentally transmitted nuclear genomes (Turelli & Moyle, 2007; Bolnick et al., 2008).

Despite this broad pattern for unidirectional hybrid origins, parthenogenetic lineages are occasionally produced through reciprocal crosses among sexual lineages. One well-described vertebrate example is the Australian gecko Heteronotia binoei (Moritz, 1993; Strasburg & Kearney, 2005; Strasburg et al., 2007). This species complex contains two mitochondrially distinct parthenogenetic lineages (‘3N1’ and ‘3N2’), which are the result of hybridization events between two sexual lineages ‘CA6’ and ‘SM6’ (Fig. 1). The 3N1 lineage resulted from crosses between CA6 females and SM6 males, while the 3N2 lineage resulted from crosses between CA6 males and SM6 females (Moritz, 1993; Figs 1b and 3). In addition, all parthenogenetic lineages of Heteronotia are triploids, containing either two CA6 genomes and one SM6 (‘form A’) or the reverse (‘form B/C’) (Fig. 1). Triploidy is assumed to have arisen when the original hypothetical diploid (and parthenogenetically reproducing) hybrids of CA6 and SM6 backcrossed repeatedly with males of either parental species to generate multiple genetic clones (Moritz et al., 1989). Fortuitously, these reciprocal backcrosses occurred within both 3N1 and 3N2 lineages, resulting in all four possible cytonuclear combinations in parthenogenetic Heteronotia.

Figure 1.

 Diagram showing the diploid sexual lineages, CA6 and SM6, and the parthenogenetic lineages arising from reciprocal hybridization events between them. Names of the genetic types are given in bold type. CA6 and SM6 chromosomes are depicted in grey and white, respectively, as are the CA6 and SM6 mitochondrial backgrounds (background colour of boxes). 3N1 and 3N2 refer to the two mitochondrial lineages of the parthenogens, as determined by the direction of the original cross. Forms ‘A’ and ‘B’ refer to the nuclear genome bias. Form ‘A’ has three nuclear genomes (CA6/CA6/SM6) with a towards CA6, while form ‘B’ has three (CA6/SM6/SM6) with a biased towards SM6, modified from Moritz (1993).

In this study, we exploit the symmetry of cytonuclear backgrounds in parthenogenetic Heteronotia to understand how mitochondrial and nuclear genes interact at the phenotypic level in polyploid hybrids. This provides insight into the constraints acting on the formation of parthenogenetic hybrids and on the influence of genome dosage (ploidy) on hybrid viability. If cytonuclear incompatibilities are a major issue in this complex, we predict that certain combinations will be more metabolically compatible than others. At the molecular level, proteins encoded by nuclear genes and mitochondrial genes interact, which may affect the performance of the electron transport system (Burton et al., 1999, 2006). If divergent mitochondrial and nuclear genomes have been brought together in hybrids, the electron transport chain and other molecular processes may be disrupted, and we would expect to see physiological consequences at the whole-organism level. For example, assuming additive genetic effects, we expect that hybrid parthenogens with two CA6 nuclear genotypes (‘A’ parthenogens) should exhibit higher metabolic compatibility in the presence of CA6-type mtDNA and reduced metabolic compatibility in the presence of SM6-type mtDNA; a reciprocal pattern should hold for ‘B’ forms.

We conducted phenotypic and molecular analyses of Heteronotia parthenogens (the four cytonuclear combinations) and their two most closely related sexual progenitors to investigate their physiological implications. Previous comparisons between sexual and parthenogenetic H. binoei (Kearney & Shine, 2004a,b, 2005; Kearney et al., 2005) found genome dosage– and reproductive mode–related differences in development, physiology and morphology but included only the 3N1 lineage and eastern lineages of CA6 and SM6. Phylogenetic analysis strongly points to a western origin of the parthenogenetic lineages, and here, we extend this previous work to include 3N2 lineages and western CA6 and SM6 lineages.

At the molecular level, we considered the interaction between proteins encoded by one nuclear gene and three mitochondrial genes involved in the electron transport system. The nuclear-encoded cytochrome c (CYC) plays a particularly critical role in mediating electron transfer from complex III (cyc reductase) to complex IV (cyc oxidase) via direct interaction with the mitochondrially encoded cyc oxidase subunit II (COII). The maintenance of the tight interaction in this protein complex likely requires the coevolution of CYC and COII (Burton et al., 1999; Grossman et al., 2001). In addition, mitochondrially encoded COII must also directly interact with ten other nuclear CO subunits in this complex (IV), which must also be coevolving for the system to remain efficient.

We hypothesized that disruption of this COX-cyc oxidase co-adapted gene complex in the hybrid background of parthenogenetic Heteronotia would show decreased activity, resulting in lower relative physiological performance overall. Therefore, at the level of the whole organism, we also expect to find evidence for hybrid breakdown, which may in turn affect the species’ ecology and distribution. In addition, because of increases in genome and cell size, polyploidy may also be expected to decrease the mass-specific metabolic rate (Kamel et al., 1985; Kozlowski et al., 2003). We tested these expectations through a comparative analysis of cold and heat tolerance, metabolic rate and locomotor performance.

Materials and methods

Animal collection and maintenance

Geckos were sourced from the wild and from a captive-bred laboratory colony. Wild-caught specimens were collected from eleven sites throughout Western Australia between 24 June and 10 July 2007 (Table 1). We determined identity (reproductive mode, mitochondrial lineage) by using mitochondrial DNA and microsatellite markers (Strasburg, 2004; Strasburg & Kearney, 2005; Strasburg et al., 2007). Captive-bred individuals were first- and second-generation females derived from animals collected from South Australia and the Northern Territory between 22 August and 4 September 2000 (Kearney & Shine, 2004b).

Table 1.   Sample sizes from each locality (N) and sources of animals used in this study. Genetic type refers to the combination of nuclear genomes and mitochondrial background (refer to Fig. 1).
Reproductive typeGenetic typeLocalityLatitudeLongitudeN
  1. *Captive-bred lineages.

Minnie Creek24°01′48.11″115°41′47.65″3
Mt. Sandiman24°24′07.28″115°24′57.18″9
Neds Creek25°28′46.26″119°38′49.18″3
Useless Loop26°07′54.06″113°25′01.59″1
3N1AVictory Downs (Clone 7)*25°59′17.98″132°58′12.01″7
Wintinna (Clone 11)*27°42′42.01″134°06′54.03″9
Neds Creek25°28′46.26″119°38′49.18″2
3N1BAileron (Clone 4)*22°38′53.97″133°21′05.90″8
Mt. Willoughby (Clone 38)*27°57′29.99″134°08′42.00″8
Neds Creek25°28′46.26″119°38′49.18″2
Parthenogen total    104
Useless Loop26°07′54.06″113°25′01.59″12
Sexual total    42
Grand total    146

The geckos were housed at The University of Melbourne in a constant temperature room at 20 °C with 12-h day/night cycle. Animals were housed in pairs with an individual from the same population (male–female pairs for sexuals, female–female for parthenogens) within plastic tubs partitioned into eight 135 × 170 mm compartments, each with a 20-mm-deep sandy substrate, cardboard hides, a dish of standing water and under-floor heating (Thermofilm; Thermofilm Australia Pty. Ltd, Springvale, Victoria, Australia) to 30 °C substrate temperature. The animals were watered and fed crickets twice each week upon which time the tubs were rotated among shelves to omit any potential position effects. All experiments commenced at least 5 months after capture.

Resting metabolic rate

We measured resting metabolic rates (VO2rest) of fasted adult geckos at 30 °C and 50% relative humidity using a flow-through respirometry system. Per measurement session, up to seven geckos were placed in separate plastic cylindrical chambers (100 × 20 mm) inside a temperature-controlled incubator (Steridium E500, Brendale, Qld, Australia). An eighth identical, empty chamber was used as reference, with a multiplexer (RM8 Intelligent Multiplexer; Sable Systems, Las Vegas NV, USA) automating flow among the eight chambers. All measurements began after at least 1 h to allow each gecko to acclimate to the conditions of the chamber in the incubator. During measurements, room air was drawn through a given chamber by a gas analyser subsampler (SS-3, Sable Systems, < 0.01 mL error) at a constant rate (30 mL min−1) regulated by a mass flow controller (MFC-2 Mass Flow Valve Controller; Sierra, Monterey, CA, USA). The oxygen concentration in the chamber was sampled every second with an Oxilla II Oxygen Analyzer (Sable Systems, < 0.001% error). Air was scrubbed of water vapour prior to entering the oxygen analyser (Drierite desiccant; Krackeler Scientific Inc., Albany, NY, USA). CO2 was not removed because the accuracy of Oxilla II is not affected by CO2 and effects of Drierite on CO2 are negligible (White et al., 2006).

Oxygen consumption measurements were taken for 30 min for each animal chamber while simultaneously measuring oxygen concentrations in the reference chamber. A data acquisition program (Expedata; Sable Systems, San Jose, CA, USA) recorded output from the oxygen analyser and also controlled the multiplexer. We recorded standard metabolic rate in mL O2 g−1 h−1 as the mean of the most level 5-min section after baselining the data (using empty animal chambers pre- and post-animal measurements) and adjusting for barometric pressure, flow rate and body mass.

Heat tolerance

Heat tolerance is assayed in a variety of ways in reptiles, a common index being the critical thermal maximum (body temperature at the ‘onset of spasms’ or the loss of righting response) (Lutterschmidt & Hutchinson, 1997). Prior to this threshold, a number of lizards, including geckos, will also pant (Greer, 1989). We used the panting threshold as an index of heat tolerance of the geckos by placing one individual at a time in a respirometry chamber in the incubator. After a 30-min acclimation period to the conditions of the incubator at 30 °C, we programmed the incubator to slowly increase the temperature to 41 °C over a further 30 min (approximately 0.3 °C min−1). We monitored the gecko in real time with a video camera mounted inside the incubator. When the gecko began to pant, we immediately recorded the air temperature in the chamber and removed the gecko from the incubator.

Cold tolerance

We used the critical thermal minimum CTmin, i.e. the temperature at which the righting response is lost, as a measure of cold tolerance. Following the procedures of Kearney & Shine (2004b), we placed 10–15 individuals in covered individual glass Petri dishes, placed them in an incubator set to an initial temperature of 11 °C and allowed the geckos to acclimate for 30 min. We then dropped the incubator temperature at 0.5 °C intervals every 30 min (approximating natural rates of cooling in nature), each time turning each animal upside down without opening the Petri dish. When an animal was unable to right itself, we opened the dish and gently stroked its abdomen with an artist’s paintbrush. If the animal had not righted itself after 10 strokes, the animal was removed from the experiment. The temperature was monitored inside each Petri dish with a temperature data logger (Thermochron Ibutton, Dallas Identification). The temperature of the logger at the time the animal was removed was recorded as the CTmin.

Sprint speed

We raced geckos along a hand-built rectangular wooden track (2.0 × 0.1 × 0.1) marked with lines 100 mm apart, placed in a temperature-controlled room set to 16 °C. This temperature was chosen because it is near their voluntary lower activity temperature limit (Kearney & Porter, 2004), and aerobic activity can only be sustained at this temperature for around 1 h or less (Kearney et al., 2005). Individual geckos were chased down the track with an artist’s paintbrush, and the trials were recorded with a Sony digital camcorder (24 frames per second) located directly above the track. Each animal was raced three times with at least 24 h between each trial. Speed was determined through a frame-by-frame analysis using AVS Video ReMaker (Online Media Technologies Ltd, London, UK). The fastest running time of the three trials across 10 mm was recorded as the final sprint speed.

Locomotor endurance

We measured locomotor endurance by walking animals on a miniature animal treadmill. At least four individuals (two from each geographic location) from each lineage were randomly selected for the endurance trial, with a total sample size of 32. Before the trial, the animals had been fasted for 2 days and placed in a controlled temperature room at 16 °C overnight. Animals were contained on the treadmill in a 60 × 200 × 60 mm Perspex chamber. The treadmill speed was 0.015 m s−1, which is a slow walking pace similar to the treadmill speed used in the study of Kearney et al. (2005). We used a plastic bristle brush placed at the rear end of the chamber for adverse stimulation and a dark shelter in front for positive stimulation. The endurance was timed with a stopwatch, and each trial terminated when the animal bumped into the brush and struggled or refused to walk for more than 10 s.

Sequencing mitochondrial components of COX and nuclear-encoded CYC

The mitochondrial DNA diversity in sexual Heteronotia is extensive (Strasburg & Kearney, 2005; Fujita et al., 2010). However, mitochondrial diversity of both the 3N1 and 3N2 lineages is very limited and matches western sexual lineages, hence the inference that the origin of parthenogenesis was limited to very local regions in Western Australia (Moritz, 1991). Indeed, phylogeographic analysis indicated single origins of mitochondrial lineages in the parthenogens (Strasburg & Kearney, 2005; Kearney et al., 2006; Strasburg et al., 2007). Thus, despite the extensive mtDNA diversity in the H. binoei complex, our interest is focused at a very limited portion of that diversity.

To quantify divergence between the cytonuclear lineages of Heteronotia, we sequenced the mitochondrial cytochrome c oxidase genes (COX) and the nuclear cytochrome c gene (CYC). We chose to examine these genes because of their involvement in the cytonuclear COX complex, for which we conducted activity assays. We found no amino acid divergence between CA6 and SM6 for CYC and did not attempt to explore this gene further in the parthenogens. We focused our sampling on cox1 and cox3 for the mitochondrial genes as we were able to obtain at least nine sequences each from 3N1 and 3N2. We amplified each marker via PCR, the products of which were purified using ExoSAPIT (USB Corp, Cleveland, OH, USA), sequenced in both directions with the original PCR primers using BigDye version 3.1 chemistry (Applied Biosystems, Inc., Carlsbad, CA, USA) and analysed on an ABI 3730 Genetic Analyzer (Applied Biosystems, Inc.). We generated contigs, edited the sequences in Geneious version 5.0 (Biomatters, Aukland, New Zealand), aligned the consensus sequences with muscle version 3.8 (Edgar, 2004) and conducted subsequent population genetic and divergence analyses in DnaSP version 5.0 (Librado & Rozas, 2009). Our aim was to confirm that 3N1 and 3N2 are divergent at genes that may have an important impact on metabolic rate differences between the parthenogenetic lineages but also to assess whether selection may have played a role in that divergence. We thus calculated nucleotide diversity (within-lineage), pairwise sequence divergence and dN/dS (ratio of nonsynonymous to synonymous substitutions). We also performed a McDonald–Kreitman (MK) test of selection (McDonald & Kreitman, 1991), which examines deviations (either from selection or from demography) from the neutral expectation that the ratio of nonsynonymous to synonymous polymorphisms equals the ratio of nonsynonymous to synonymous substitutions.

COX activities

We used mitochondria extracted from liver in assays that spectrophotometrically quantify the activity of COX by measuring the reduction of cytochrome c. We extracted mitochondria using the Mitochondria Isolation Kit from Sigma (MITOISO1; Sigma-Aldrich, St Louis, MO, USA). Geckos were euthanized with 50 μL of a 1 : 1 v:v ration of MS-222 and water-neutralized with sodium bicarbonate (Conroy et al., 2009). Approximately 50 μL of liver was homogenized using a dounce homogenizer and then subjected to differential centrifugation according to the manufacturer’s protocol. The final mitochondrial extract was resuspended in 40 μL of 1× Storage Buffer (10 mm HEPES pH 7.5, 0.25 m sucrose, 1 mm ATP, 0.08 mm ADP, 5 mm sodium succinate, 2 mm K2HPO4 and 1 mm DTT) and immediately frozen at −80 °C; freezing the samples shears the mitochondrial membrane and helps expose COX to the substrate in subsequent assays (Ellison et al., 2008).

We used the extracts in the COX assays using the Cytochrome c oxidase Assay Kit (CYTOCOX1; Sigma-Aldrich), following the manufacturer’s protocol with some modification detailed here. The assays employ horse heart CYC as substrate for the mitochondrial preparations. This approach has been successful in detecting cytonuclear disruption in organisms as divergent as hybrid Drosophila species (Sackton et al., 2003); thus, we feel confident this kit will provide meaningful results using Heteronotia mitochondrial extracts despite the use of horse CYC. We thawed the frozen samples on ice and used a 1 : 10 dilution of the extract in the assays (1 : 5 dilutions were used for samples 676–678 due to the small yield from the mitochondrial extraction). Forty microlitres of the diluted mitochondrial extracts and 100 μL reduced CYC were mixed by inversion in 2060 μL 1× reaction buffer in a cuvette; this is double the volume specified in the manufacturer’s protocol but necessary due to the machine and cuvette size used to measure absorption. The reactions were immediately measured for absorption using a kinetic program (7 readings at 10 second intervals, 23 °C, wavelength 550 nm, using the WPA Biowave 2100; Cyron, Schindellegi, Switzerland), to capture the decrease in absorption as CYC oxidizes. To standardize the activity measurements across samples, we measured, in duplicate, the protein concentration of the mitochondrial extracts using a TCA-Ponceau-S protocol (Hayner et al., 1982) and a MultiSkan EX microplate reader (Thermo Fischer Scientific, Waltham, MA, USA). Replicate samples were always measured simultaneously on the same plate. Each replicate involved quantifying protein concentration of each sample at 1 : 10 and 1 : 20 dilutions, using the rescaled average to fit into the BSA standard curve. Replicates were of similar magnitude, often with the same reading, demonstrating the repeatability of the assay. The final protein concentration of each sample was an average from the two replicates. The activity was a calculation of changes in absorption over time, corrected for the dilution factor and standardized with the protein concentration of the sample.


Although our study system includes hybrids of reciprocal crosses, it does not lend itself to a completely crossed three-factor design as there are no triploid sexuals or diploid parthenogens. Therefore, to tease apart the effects of hybridization in triploid parthenogens with different mitochondrial backgrounds, compared to their sexual relatives, we performed three separate tests using two-factor anovas. First, within the parthenogenetic lineages, we tested for differences attributable to genome dosage and mtDNA lineage, as well as the interaction between these factors. Genome dosage refers to bias of the nuclear genomes, i.e. whether the individual had more CA6 or SM6 genomes. The second factor, mtDNA lineage, refers to whether the mitochondrial background was CA6 or SM6. Although all individuals measured were mature adult females, the mass of individuals did vary; therefore, mass was included as a covariate as a proxy for age. This also served as a proxy for age, which affects metabolism, because the ages of wild-caught individuals were unknown. If mass was not a significant covariate, it was removed for the comparison of that trait. A significant interaction term for this analysis would suggest the presence of cytonuclear interactions.

Second, we tested for effects of reproductive mode and genome dosage, comparing the western CA6 and SM6 sexual populations with the 3N2 parthenogenetic lineages. Reproductive mode refers to whether the individual was parthenogenetic or sexual. Mass was also included as a covariate but removed if nonsignificant. This analysis parallels the study of Kearney & Shine (2004b), which instead used eastern sexual populations and 3N1 parthenogenetic lineages. We first ran the analysis with only 3N2 individuals because of the potential confounding effect of their different history (most 3N1 were long-term captive or captive-bred specimens) and then ran a further analysis including 3N1 and 3N2 lineages.

Standard metabolic rates and COX activities were log-transformed prior to analysis to equalize variances between groups and improve normality. Transformations were not necessary for other data sets.


Mitochondrial background and genome dosage effects within parthenogens

Mitochondrial background independently influenced cold tolerance (3N1 had lower CTmin than 3N2; anova, Tables 2 and 3, Fig. 2b) and endurance (3N1 had lower endurance than 3N2; anova, Tables 2 and 3, Fig. 2c), while genome dosage had strong, independent effects on cold tolerance (form A had lower CTmin than form B; anova, Tables 2 and 3, Fig. 2b). Interactions between these factors were apparent for heat tolerance (panting temperature) and sprint speed (anova, Tables 2 and 3, Fig. 2a,d). For the 3N1 lineage, form A had a slightly higher panting temperature than form B, while the reverse pattern occurred for the 3N2 lineage (Fig. 2a). For sprint speed, 3N2B was faster than the other three lineages (Fig. 2d). There were no significant effects of mitochondrial background or nuclear genome dosage on metabolic rate or COX activity (anova, Tables 2 and 3).

Table 2.   Results of anovas from three separate comparisons of genetic types (Fig. 1) from the Heteronotia binoei species complex. The first comparison includes only the four parthenogenetic types to test for the presence of cytonuclear interactions using two factors: genome dosage (i.e. the majority or bias of the nuclear genomes; CA6 or SM6) and mtDNA lineage (i.e. mitochondrial background; CA6 or SM6). The second compares 3N2 parthenogens and both sexual types as a parallel of a previous study comparing 3N1 parthenogens and sexuals. The third incorporates both 3N1 and 3N2 parthenogens in comparison with both sexual types. The second and third comparisons include reproductive mode (or whether they reproduce via parthenogenesis or sex) as a factor, which is tightly correlated with ploidy. Statistically significant effects are indicated in bold.
Parthenogens 3N1 vs. 3N2
(a) Resting metabolic rate
  Genome dosage1,951.48E−030.1450.704
  Dosage × mtDNA1,953.27E−030.3200.573
  Mass1,950.31130.448< 0.001
(b) Panting temperature
  Genome dosage1,1000.2200.1970.658
  Dosage × mtDNA1,1006.3235.6540.019
(c) Critical thermal minimum
  Genome dosage1,9438.80722.335< 0.001
  Dosage × mtDNA1,940.0750.0450.833
(d) Endurance
  Genome dosage1,18102.9250.2050.656
  Dosage × mtDNA1,18295.2950.5890.453
(e) Sprint speed
  Genome dosage1,940.1082.4140.124
  mtDNA1,941.16425.895< 0.001
  Dosage × mtDNA1,940.2124.7180.032
(f) COX activity
  Genome dosage1,244.04E−050.1770.677
  Dosage × mtDNA1,243.39E−041.4880.234
Parthenogen (3N2 only) vs. sexuals
(a) Resting metabolic rate
  Genome dosage1,1023.48E−030.2270.635
  Reproductive mode1,1022.82E−030.1840.669
  Dosage × reproductive mode1,1022.76E−030.1800.672
(b) Panting temperature
  Genome dosage1,1022.0291.6140.207
  Reproductive mode1,1024.2333.3670.069
  Dosage × reproductive mode1,1020.6200.4930.484
(c) Critical thermal minimum
  Genome dosage1,9738.55530.639< 0.001
  Reproductive mode1,976.2564.9710.028
  Dosage × reproductive mode1,970.0580.0460.830
(d) Endurance
  Genome dosage1,192460.9416.4510.020
  Reproductive mode1,19763.4272.0010.173
  Dosage × reproductive mode1,19455.91.1950.288
(e) Sprint speed
  Genome dosage1,970.3068.9790.003
  Reproductive mode1,970.69320.311< 0.001
  Dosage × reproductive mode1,970.0621.8190.181
(f) COX activity
  Genome dosage1,242.6E−060.0090.923
  Reproductive mode1,2405.04E−060.998
  Dosage × reproductive mode1,241.09E−040.3970.534
Parthenogens (3N1 and 3N2) vs. sexuals
(a) Resting metabolic rate
  Genome dosage1,1463.25E−040.0240.877
  Reproductive mode1,1463.51E−040.0260.873
  Dosage × reproductive mode1,1461.22E−040.0090.925
(b) Panting temperature
  Genome dosage1,1420.1990.1740.678
  Reproductive mode1,1425.3384.6580.033
  Dosage × reproductive mode1,1420.0600.0520.820
(c) Critical thermal minimum
  Genome dosage1,13347.45730.886< 0.001
  Reproductive mode1,1332.6071.6970.195
  Dosage × reproductive mode1,1330.0050.0030.956
(d) Endurance
  Genome dosage1,282323.6565.6120.025
  Reproductive mode1,28176.090.4250.52
  Dosage × reproductive mode1,28935.5772.260.144
(e) Sprint speed
  Genome dosage1,1320.2685.420.021
  Reproductive mode1,1320.3036.120.015
  Dosage × reproductive mode1,1329.57E−0700.996
(f) COX activity
  Genome dosage1,387.69E−050.2840.597
  Reproductive mode1,382.25E−040.8320.368
  Dosage × reproductive mode1,382.04E−060.0080.931
Table 3.   Trait means ± standard deviation (N in brackets) for physiological measures of sexual (CA6 and SM6) and parthenogenetic (3N1A, 3N1B, 3N2A and 3N2B) lineages of Heteronotia binoei.
LineageResting metabolic rate (mL O2 g−1 h−1)Panting temperature (°C)Critical thermal minimum (°C)Locomotor endurance (min)Sprint speed (m s−1)COX activity
CA60.162 ± 0.1261 (25)39.6 ± 1.14 (26)3.8 ± 1.01 (23)11.5 ± 7.33 (5) 0.7 ± 0.231 (23)0.028 ± 0.0163 (7)
3N1A0.166 ± 0.1113 (20)39.5 ± 0.85 (22)3.7 ± 0.61 (19)23.3 ± 23.23 (4)0.69 ± 0.199 (19)0.041 ± 0.0173 (7)
3N1B0.141 ± 0.0689 (18)38.9 ± 0.84 (18)5.1 ± 2.07 (17)20.2 ± 13.59 (5)0.67 ± 0.321 (17)0.032 ± 0.0122 (7)
3N2A0.163 ± 0.0968 (29)39.1 ± 1.14 (31)4.3 ± 0.96 (29)32.4 ± 18.69 (5)0.82 ± 0.17 (29)0.025 ± 0.0197 (7)
3N2B0.185 ± 0.1484 (33)39.5 ± 1.18 (33)5.6 ± 1.35 (33)44.4 ± 27.42 (8)0.99 ± 0.182 (33)0.029 ± 0.0085 (7)
SM60.164 ± 0.1091 (19)39.8 ± 0.89 (16)5.1 ± 1.02 (16)41.7 ± 9.66 (5)0.76 ± 0.135 (16)0.025 ± 0.0192 (7)
Figure 2.

 Trait means ± standard errors for sexual (CA6 and SM6) and parthenogenetic (3N1A, 3N1B, 3N2A and 3N2B) lineages of Heteronotia binoei. The ‘3N1’ lineages have CA6 mitochondrial DNA, while the ‘3N2’ lineages have SM6 mitochondrial DNA. The ‘A’ lineages have two doses of the CA6 nuclear genome and one dose of the SM6 genome, while the ‘B’ lineages have the opposite. Summary statistics are presented in Table 2.

Reproductive mode and genome dosage effects across parthenogens and sexuals

In comparisons between sexuals and 3N2 parthenogens only, significant effects of reproductive mode were apparent for CTmin and sprint speed (sexuals had lower CTmin and lower sprint speed than parthenogens, Fig. 2b,d). Genome dosage effects were present for CTmin, endurance and sprint speed, with CA6 sexuals and form A (CA6-biased parthenogens) having lower CTmin (Fig. 2b), lower sprint speed (Fig. 2d) and lower endurance (Fig. 2c) than SM6 sexuals and form B (SM6-biased parthenogens). When these comparisons were made more broadly with 3N1 parthenogenetic lineages included, the patterns were consistent except that the reproductive mode effect on CTmin was nonsignificant. There was no significant effect of reproductive mode or genome dosage on metabolic rate or COX activity.

Molecular evolution

We sequenced cyc from cDNA libraries prepared from CA6 and SM6 liver tissue; despite some divergence at synonymous sites (dS = 0.044), amino acid sequences were identical, as expected given the high conservation of cyc seen in animals. We therefore focused our sequencing efforts for the parthenogenetic lineages to the mitochondrial genes cox1, cox2 and cox3. Nucleotide diversities were low, as expected given the restricted mitochondrial origins of each parthenogenetic lineage. For 3N1, nucleotide diversity is 0.0004 for cox1, 0.001 for cox2 and 0.0001 for cox3; for 3N2, nucleotide diversity is 0.0062 for cox1 and 0.001 for cox3 (our sample size for 3N2 cox2 was insufficient to measure nucleotide diversity). Divergence measurements confirm the distinctness of 3N1 and 3N2 mitochondrial genomes. For cox1, there are 127 fixed differences between the two lineages (from an alignment of 1185 base pairs), corresponding to a net nucleotide divergence Da (Nei, 1987) of 0.119. For cox3, there are 58 fixed differences between 3N1 and 3N2 (from an alignment of 472 base pairs) and a Da of 0.123. For cox2, Da = 0.039 (27 fixed differences, although our sample size for 3N2 was = 1). Ratios of dN/dS for both genes were << 1 (0.007 for cox1; 0.062 for cox2, 0.039 for cox3), indicative of strong purifying selection. The MK test examines deviations from neutrality with respect to nonsynonymous and synonymous polymorphisms within a species and nonsynonymous and synonymous substitutions between species. Both cox1 and cox3 did not deviate from neutral expectations based on the MK test (two-tailed P >> 0.05; Fisher’s exact test); cox3 exhibited deviations from neutrality (two-tailed = 0.037), but because of inadequate sample size, this result does not consider polymorphism within 3N2 and should be taken with caution.

Metabolic scaling

While there were no differences between lineages in mean mass-specific metabolic rate, this trait scaled differently with mass for parthenogens and sexuals (heterogeneity of slopes test, F1,140 = 6.164, = 0.014) (Fig. 3). A negative linear relationship between log-transformed mass and mass-specific metabolic rate was observed for the parthenogens, while no significant relationship was found for the sexuals (Fig. 3). To test the robustness of this result, we also analysed the independently collected data set of Kearney & Shine (2004b) and found very similar results (Fig. 3). To further test for consistency across data sets, we performed an anova including data set, reproductive mode, mass and all interactions possible interaction terms. The three-way interaction term was nonsignificant (F1,245 = 0.875, = 0.350). Removing this term from the analysis, there was no significant effect of reproductive mode (F1,246 = 3.703, = 0.055) or data set, but there was a strong reproductive mode-by-mass interaction (F1,246 = 12.527, < 0.001).

Figure 3.

 Linear regression equations for the relationship between log mass-specific metabolic rate and log body mass for Heteronotia binoei. Sexual and parthenogenetic lineages from the present study (‘Expt 2’, solid lines; parthenogens R= 0.265, N = 100, coefficient =  −1.15, CI lower = −1.53, CI upper = −0.76, P < 0.001; sexuals R2 = 0.010, N = 44, coefficient = −0.24, CI lower = −0.98, CI upper = 0.50, P = 0.517) and from Kearney & Shine (2004b) (‘Expt 1’, dashed lines; parthenogens R= 0.069, N = 82, coefficient =  −0.94, CI lower = −1.70, CI upper = −0.17, P = 0.017, sexuals R2 = 0.094, N = 27, coefficient = 0.65, CI lower = −0.18, CI upper = 1.49, P = 0.120) are shown, as well as the standard allometric relationship for squamate reptiles (Andrews & Pough, 1985) (‘predicted’, dotted line).


We found surprisingly little evidence for our prediction that hybrid lineages with greater genomic imbalance (i.e. 3N1B and 3N2A, Fig. 1) would express reduced physiological performance due to cytonuclear incompatibilities. However, we did find strong, parallel effects of genome dosage within each hybrid lineage and an unusual pattern with respect to mass-specific scaling of metabolic rate.

Weak evidence for cytonuclear incompatibilities

The sexual progenitors of the parthenogens, the CA6 and SM6 races, diverged sometime in the Pliocene (Fujita et al., 2010). Because both the 3N1 and 3N2 parthenogens evolved much more recently, in separate events in the late Pleistocene (Strasburg et al., 2007), we might expect significant incompatibilities to arise among alleles between their hybrid genomes because of antagonistic epistasis (known as Dobzhansky-Muller incompatibilties Coyne & Allen Orr, 2004; Turelli & Moyle, 2007). Such incompatibilities may occur between nuclear genomes or between nuclear and mitochondrial genomes because of the rapid evolution of the latter (i.e. cytonuclear interactions) and have been found both between and within sexual species (Coyne & Allen Orr, 1989; Burton et al., 2006; Turelli & Moyle, 2007).

Unisexual taxa of hybrid origin do show evidence for reduced fitness/performance compared to sexual progenitors (Wetherington et al., 1987; Cullum, 1997; Mee et al., 2011). However, in the case of H. binoei, we found surprisingly little evidence for deficient performance in hybrid parthenogens when compared with sexual progenitors despite the probable strong influences of cytonuclear metabolic pathways on many traits (endurance, sprint speed, metabolic rate itself). Indeed, the only trait that showed significant differences in the direction expected, if cytonuclear incompatibilities were an issue, was heat tolerance.

In accordance with our findings for whole-organism traits, we found little evidence at the biochemical or molecular levels for cytonuclear incompatibility. As expected, the dN/dS ratios did not show evidence for positive selection in the mitochondrial genes, as it is one of the most highly conserved gene regions; however, genetic divergence via drift between allopatric populations can lead to the disruption of coevolved gene complexes upon hybridization (e.g. the California copepod, Tigriopus californicus Burton et al., 2006). There were several amino acid differences in cox1, cox2 and cox3 between CA6 and SM6 genomes. While these amino acid differences do not involve positions that directly interact with CYC (Capaldi et al., 1983), they can alter interactions with other nuclear components that ultimately affect COX activity, if not other complexes of the electron transport chain. Identifying the precise functional consequences of these amino acid substitutions – and their effect on physiological traits – will require additional functional tests that measure the activity of other components of the electron transport chain.

Ideally, our measurements of electron transport chain activity would include all four subunits and Heteronotia-specific substrate; however, the lack of genetic and genomic resources limited our assays to COX assays using readily available horse heart CYC. We feel confident in our results because (1) the same treatment and substrate were used across lineages, which should factor out the use of a divergent CYC, and (2) previous studies have successfully used similar assays, including horse heart CYC, with systems as divergent as Drosophila (Sackton et al., 2003). The structure of CYC is highly conserved in animals, with several residues having important protein–protein interaction functions, including positions 9, 13, 16, 70, 72, 73, 81–83, 86 and 87 (Banci et al., 1999). Of these, only positions 9 and 81 differed between Heteronotia and horse, with position 9 composed of two common amino acid variants. Position 81 is isoleucine in horse, a common residue, but methionine in Heteronotia; while methionine is uncommon at this position, it is nonetheless hydrophobic and likely a conservative substitution. Thus, the conservation of sequence and structure between horse heart and Heteronotia CYC is sufficient to impart confidence in our enzymatic assays of COX activity.

Parallel nuclear genome dosage effects in independently derived parthenogens

In this study, we observed consistent phenotypic differences in cold tolerance due to nuclear genome dosage between the sexual populations and both parthenogenetic lineages (3N1 and 3N2). This genetic effect is the simplest physiological consequence of hybridization, with genomes combining additively and phenotypes reflecting the dosage summations of the sexual parental genotypes. We also found dosage effects for sprint speed and endurance. Although we did not observe dosage effects for heat tolerance, 3N1 and 3N1B lineages differed in the same direction as in previous studies (Kearney & Shine, 2004b). In contrast to this study, previous phenotypic comparisons between parthenogenetic and sexual lineages of the H. binoei species complex have been restricted to the 3N1 lineage and eastern lineages of the sexual taxa (Kearney & Shine, 2004a,b, 2005; Kearney et al., 2005). These studies demonstrated differences between the two parental sexual species for body size, scalation, the pattern of reproduction, development time, heat and cold tolerance, metabolic rates and water-loss rates, with corresponding nuclear dosage effects in the parthenogenetic lineages (Kearney & Shine, 2004a, 2005). The differences between the current and previous studies may in part reflect different approaches in measuring some traits (e.g. sprint speed was measured in this study at 15 °C rather than 25 °C) but may also reflect differences between eastern and western sexual lineages of CA6 and SM6.

The phenotypic differences between CA6 and SM6 sexual lineages presumably reflect different selective processes operating in their respective ranges (Kearney & Shine, 2004b). This is particularly clear for cold tolerance, as CA6 sexuals have lower critical thermal minima than do SM6 sexuals and also live in colder climates (Kearney et al., 2003). Genome dosage biases in the triploid parthenogens produce parallel phenotypic patterns to those in the sexual lineages (Kearney & Shine, 2004b). The present study confirms that these phenotypic dosage effects occur in each of the independently derived 3N1 and 3N2 parthenogenetic lineages, with CA6-biased parthenogens also living in colder climates (Kearney et al., 2003). Because each parthenogenetic lineage had highly restricted geographic origins (Moritz, 1991, 1993; Kearney et al., 2006; Strasburg et al., 2007), the present environmental differences between parthenogens of different dosage probably reflect differential success in spreading to and surviving in these environments. This is consistent with the frozen niche variation model, which suggests that hybrid parthenogens have captured the phenotypic diversity from sexual progenitors and occupy a range of narrow ecological niches (Vrijenhoek, 1984). Indeed, because no diploid parthenogenetic H. binoei has ever been observed (despite very extensive sampling, more than 1000 individuals), it is conceivable that the evolution of polyploidy has been favoured in parthenogenetic H. binoei through effects of nuclear genome dosage and partial recovery of the parental niches.

Divergent metabolic scaling in parthenogenetic and sexual lineages

The scaling of mass-specific metabolic rate with body size is a fundamental issue in biology (Kleiber, 1932) but has been unexplored in phenotypic comparisons of parthenogens and their sexual relatives. We found that mass-specific metabolic rate scales almost linearly (negatively) with body mass in parthenogenetic H. binoei but that it was independent of body mass in the sexual H. binoei. This has the effect that mass-specific metabolic rate is relatively higher for parthenogens than for sexuals early in ontogeny but converges as maturity is approached. This result is consistent with previous work showing that juvenile parthenogens exhibit significantly higher mass-specific metabolic rates than juvenile sexuals, while adult parthenogens do not (Kearney & Shine, 2004b).

It has previously been suggested that ploidy should affect metabolic rates via influences on cell size, with elevated ploidy (or genome size) producing larger cells and lower mass-specific metabolic rates (Szarski, 1983; Kamel et al., 1985; Kozlowski et al., 2003). While a quantification of cell size in parthenogenetic and sexual H. binoei is yet to be done, parthenogens do have larger leucocytes (C. Moritz, pers. obs.). However, there is no reason to expect ploidy to differentially affect cell size during ontogeny, and moreover, the pattern we observe is in the opposite direction to the predictions of cell-size-based theories. Our results potentially reflect a greater cost of growth in parthenogenetic than in sexual H. binoei, which is consistent with the observed slower growth rates in parthenogenetic individuals (Kearney & Shine, 2004a). This may have significant ecological consequences for the different genetic types, particularly in environments with scarce or limited resources. Metabolic scaling in parthenogenetic and sexual H. binoei, and in other agamic complexes, appears worthy of further investigation.


This study indicates that the various combinations of divergent sexual genomes of H. binoei during the evolution of parthenogenesis via hybridization and polyploidy have surprisingly low fitness consequences, at least according to the traits that we considered. This indicates that the metabolic pathways in the sexual lineages of H. binoei have remained largely compatible over millions of years of independent evolution, which has facilitated the successful hybrid origin of parthenogenesis in this extremely abundant and ecologically successful organism. Nonetheless, there are clear effects of dosage of the parental genomes on physiological phenotypes of the parthenogens, which correlates with their observed eco-geographic distributions. This strongly indicates that the evolution of polyploidy in H. binoei, through back-crossing events with sexual progenitors, has produced a broader overall niche and correspondingly a broader geographic range, consistent with the frozen niche variation model (Vrijenhoek, 1984).


This work was supported by an Australian Research Council Grant DP0771924 to MK and CM and was undertaken under The University of Melbourne Animal Ethics Permit 0703441. Western Australian specimens were collected under Department of Environment and Conservation research permit no. SF004376. We are also grateful for constructive comments on this manuscript from Natalie Briscoe and two anonymous reviewers.