ASYMMETRIC VIABILITY OF RECIPROCAL-CROSS HYBRIDS BETWEEN CRESTED AND MARBLED NEWTS (TRITURUS CRISTATUS AND T. MARMORATUS)

Authors


Present address: School of Environment and Life Sciences, University of Salford, Manchester M5 4WT, United Kingdom

Present address: Department of Biology, Adnan Menderes University, Aydin 09100, Turkey

Abstract

Hybridization between divergent lineages often results in reduced hybrid viability. Here we report findings from a series of independent molecular analyses over several seasons on four life stages of F1 hybrids between the newts Triturus cristatus and T. marmoratus. These two species form a bimodal hybrid zone of broad overlap in France, with F1 hybrids making up about 4% of the adult population. We demonstrate strong asymmetry in the direction of the cross, with one class (cristatus-mothered) making up about 90% of F1 hybrids. By analyzing embryos and hatchlings, we show that this asymmetry is not due to prezygotic effects, as both classes of hybrid embryos are present at similar frequencies, implicating differential selection on the two hybrid classes after hatching. Adult F1 hybrids show a weak Haldane effect overall, with a 72% excess of females. The rarer marmoratus-mothered class, however, consists entirely of males. The absence of females from this class of adult F1 hybrids is best explained by an incompatibility between the cristatus X chromosome and marmoratus cytoplasm. It is thus important to distinguish the two classes of reciprocal-cross hybrids before making general statements about whether Haldane's rule is observed.

One of the most fundamental questions in evolution concerns the origin and maintenance of the discrete taxa that we see in nature (Dobzhansky 1937). When species are sympatric or parapatric, their future independence is contingent upon the existence of reproductive barriers between them: do they interbreed, and if so, are hybrids viable and fertile (Mayr 1963)? Hybrid zones have long been valued as locations for examining interactions of incipient and closely related species (Kocher and Sage 1986; Hewitt 1988; Barton and Hewitt 1989; Harrison 1993; Jiggins and Mallet 2000). Processes relating to speciation are difficult or impossible to observe within species, but can emerge in analysis of hybrids between them (Wu and Palopoli 1994; Voss and Shaffer 1996).

Molecular genetic markers have allowed detailed analyses of hybridization over the last 30 years. One popular approach has been to combine the study of nuclear markers with maternally inherited cytoplasmic mtDNA. Early analyses of this type showed that cytoplasm from taxon A could cross species boundaries and function on the nuclear background of taxon B (Ferris et al. 1983; Powell 1983; Spolsky and Uzzell 1984, 1986; Gyllensten and Wilson 1987; Tegelström 1987). The widespread dual use of nuclear and cytoplasmic markers has revealed various types of asymmetry in hybridization. First, the progeny of heterospecific matings often show a strong bias in the direction of the cross, inferred from preponderance of one type of mtDNA in hybrids (Patton and Smith 1993; Arnold et al. 1996). This asymmetry may be due to prezygotic (Szymura et al. 1985; Lamb and Avise 1986; Baker et al. 1989; Konkle and Philipp 1992; Wirtz 1999), or postzygotic (Thornton 1955; Wu and Beckenbach 1983; Rakocinski 1984; Hillis 1988; Orr and Coyne 1989; Szymura and Barton 1991; Welch 2004; Bolnick and Near 2005; Johannesen et al. 2006; Turelli and Moyle 2007; Presgraves 2008) effects, or some combination of the two (Scribner and Avise 1994). In the former case, asymmetry may reflect relative abundance of species rather than intrinsic behavioral differences, with the rarer species providing the female parent (Hubbs 1955; Avise and Saunders 1984). In some cases, symmetrical fitness of hybrids has been demonstrated (Gyllensten et al. 1985; Nürnberger et al. 1995; Sites et al. 1996). If both types of hybrids occur, the sexes often differ greatly in fitness (Haldane 1922; Parris et al. 1999), which can lead to further genetic asymmetries (Tegelström and Gelter 1990). Second, the extent of introgression of the cytoplasmic genome can be quite different in the two species (Gyllensten and Wilson 1987; Harrison 1989; Bernatchez et al. 1995; Avise 2004; Johannesen et al. 2006). Asymmetries endogenous to the hybrid class can exist independently of the asymmetries in introgression, although the first is usually invoked to explain the second (Carr et al. 1986; Dowling et al. 1989; Lehman et al. 1991; Thulin et al. 1997; Cathey et al. 1998).

Great advances have been made in our understanding of the genetics of speciation in recent years (Coyne and Orr 2004). General principles that are beginning to emerge include the prevalence of Bateson–Dobzhansky–Muller (BDM) incompatibilities (Bateson 1909; Coyne and Orr 2004), widespread agreement with Haldane's rule (Haldane 1922; Laurie 1997), the large X effect (Coyne and Orr 1989; Orr and Coyne 1989), and isolation asymmetry (Darwin 1859; Turelli and Moyle 2007; Bolnick et al. 2008). This article concerns asymmetry of hybrid parentage and resolution of postzygotic effects in hybrids between two species of newt.

In France, there is a broad region of overlap between the newt species Triturus cristatus and T. marmoratus where hybridization occurs (Arntzen and Wallis 1991). The two forms differ in sexual behavior (Sparreboom 1986; Zuiderwijk 1990), habitat preference (Schoorl and Zuiderwijk 1981), morphology (Vallée 1959), chromosome morphology (Macgregor et al. 1990), albumins (5–17 AID units, [Busack et al. 1988]) and possess numerous species-specific allozymes (D= 0.86 ± 0.27 [Rafiński and Arntzen 1987], D= 0.68 [Arntzen et al. 2007], D= 0.84 [Macgregor et al. 1990]) and mitochondrial restriction sites (d= 0.075–0.083 [Wallis and Arntzen 1989; Arntzen and Wallis 1991]). The fossil record suggests that the most recent common ancestor exceeds 18–15 Ma (millions of years ago) (Estes and Hoffstetter 1976; Estes 1981), redated to > 24.2–23.8 Ma and used for molecular calibration (Steinfartz et al. 2007). F1 hybrids of the two species have high viability but show a range of fitness effects (Table 1). There have been many other such crossing experiments with similar results that used related crested newt species in place of T. cristatus (White 1946), but we exclude these here.

Table 1.  The history of hybridization between Triturus cristatus and T. marmoratus.
de L'Isle du Dréneuf 1862recognition of hybrid as a new taxon, Triton blasii
Peracca 1886first firm statement on the hybrid nature of the new form, cristatus-mothered hybrids described as a taxon different from the reverse cross, Triton trouessarti
Boulenger 1898experimental backcross of hybrid ×cristatus
Wolterstorff 1904experimental demonstration of F1 hybrid origin of intermediate through crosses in both directions
Bataillon and Tcherniakofsky 1932disrupted spermatogenesis in marmoratus-mothered male F1 hybrid
Lantz 1934, 1947marmoratus females ×cristatus males give rise to four female and five male F1 hybrids; F1 hybrids shown to be not completely sterile, mostly through the female hybrid line
Vallée 1959among adults in 104 ponds across Mayenne 4.7% are hybrid (42 females and 20 males); about equal hatchling proportion (7–8%) from interspecific crosses in both directions; seven marmoratus-mothered F1 hybrids (and none of the reverse combination) survive to adulthood
Vallée 1959; Arntzen and Hedlund 1990hatchling proportion of 6–12% from female hybrids against 45–64% from homospecific matings
Schoorl and Zuiderwijk 1981, pers. comm.among adults in 154 ponds across Mayenne 3.1% are hybrid (27 females and 15 males)
Arntzen, unpublished dataamong adults in 51 ponds in area ”C” (Fig. 2) 4.1% are F1 hybrids (60 females and 40 males)
Zuiderwijk and Sparreboom 1986; Zuiderwijk 1988no obvious directionality in interspecific mating display
Arntzen and Hedlund 1990thirty-four F1 hybrids among 484 adult females (7.0%)
Arntzen and Wallis 1991ten F1 female hybrids all cristatus-mothered; low levels of introgression across mosaic hybrid zone among adults in eight ponds 3.8% are hybrid
R. Donovan 1991, pers. comm.a marmoratus male introduced to a cristatus pond sires ≥ 24 F1 hybrids (23 females and one male)

Within the region of overlap in the département Mayenne, detailed ecological, behavioral, and genetic studies have been carried out over the last 30 years. Syntopic (i.e., T. cristatus and T. marmoratus coexisting) ponds contain on average 3.8% adult F1 hybrids and introgression is almost nonexistent (Arntzen and Wallis 1991), so the zone is best characterized as bimodal (Jiggins and Mallet 2000). Significantly, and prompting this current investigation, all 10 F1 hybrids studied for mtDNA were mothered by T. cristatus (Arntzen and Wallis 1991). In this current article, we address several questions that arise essentially from two main initial foci. First, does this asymmetry of parentage hold for a larger sample of F1 hybrids of both sexes from more ponds and is there any geographic pattern to the asymmetry? Second, are its causes prezygotic (e.g., mating preference, gametic incompatibility) or postzygotic (i.e., the result of differential survival of F1 hybrids depending on the direction of mating)? If the latter holds, at what stage(s) does selection occur? We address these questions using a variety of molecular analyses on four different developmental stages (adult, larva, hatchling, embryo) of F1 hybrids: mtDNA of adults identified by morphology; mtDNA of larvae identified by microsatellite loci; allozymes of yolk sac, on hatchlings identified by allozymes; mtDNA of embryos identified by microsatellites.

Methods

ADULT ANALYSIS

Samples of adult F1 hybrid newts (Fig. 1) were collected over many field seasons from 1980 to 1997. Tissues used for DNA extraction were liver or muscle homogenates stored in Tris-EDTA buffer at −70°C, tail tips or single oocytes. Samples were digested with proteinase K in 0.5–1 mL SDS STE lysis buffer (Wallis 1987) at 37°C for at least 2 h (larger tissue samples rotated at 50°C overnight). DNA was extracted by phenol–chloroform, ethanol-precipitated, and resuspended in ultrapure water.

Figure 1.

Ventral coloration of Triturus cristatus (left), T. marmoratus (right) and F1 interspecific hybrids (centre). The top row are females and the bottom row are males.

Partial mtDNA cytochrome b (cyt b) sequences (G. P. Wallis, unpubl. data) from samples of the two species away from the hybrid zone were used to determine species-diagnostic RFLPs. Four restriction enzymes (HaeIII, HinfI, NlaIII, and ThaI) had five completely diagnostic restriction sites in this region. The majority of F1 adults were typed for all five restriction sites but in all cases a minimum of two restriction sites was used (HaeIII). The original 10 F1 hybrid females (Arntzen and Wallis 1991) were tested and confirmed, but are omitted from all tables and analyses in this article.

PCR amplifications of a 315-bp region of cyt b were performed with primers L14841 (5′-AAAAAGCTTCCATCCAACATCTCAGCATGATGAAA-3′) and H15149 (5′-AAACTGCAGCCCCTCAGAATGATATTTGTCCTCA-3′) (Kocher et al. 1989). We used ca. 10 ng of DNA in 10–20 μl volumes using Jeffreys' buffer (Jeffreys et al. 1988) and AmpliTaq (Applied Biosystems, Foster City, CA) in a Perkin-Elmer Cetus (Waltham, MA) heating block. After an initial denaturation at 94°C for 4 min, 40 polymerization cycles were used as follows: 94°C denaturation for 60 sec, 50°C annealing 60 sec, and 73°C extension for 60 sec. Later analyses used primers MVZ15 (5′-GAACTAATGGCCCACACRRTACGNAA-3′) and MVZ16 (5′-AAATAGGAARTATCAYTCTGGTTTRAT-3′) (Moritz et al. 1992) and 35 cycles of: 95°C for 30 sec, 48°C for 30 sec, and 72°C for 30 sec. These primers gave an 830-bp product, which was ethanol-precipitated and resuspended in 20 mL TE. Four microliters aliquots of these products were digested with up to four different restriction enzymes (BRL, Gaithersburg, MD; Boehringer, Mannheim, Germany; according to manufacturers' specifications). Digests were run on 250-mL 2–3% agarose gels with 70 μg of ethidium bromide per 100 mL (gel and electrode buffer), using a 123-bp size ladder (BRL) as a molecular weight standard, and photographed over a 305-nm UV source.

In total, 67 hybrid adult newts (41 females, 26 males) were typed, representing 24 ponds in the Mayenne region (Fig. 2; Table 2). Nineteen of these F1 hybrids were typed for nine or more diagnostic protein loci (Rafiński and Arntzen 1987; Arntzen and Wallis 1991), confirming that they were F1 hybrids. Mitochondrial DNA from three specimens of each parental species was also included for controls.

Figure 2.

Distribution of newts in France. (A) Northern limit of Triturus marmoratus and southern limit of T. cristatus; hatched area is the zone of overlap (Arntzen and Wallis 1991). (B) Département Mayenne. (C) Central area; forest in black. (D) Northern area. Ponds sampled are numbered. Forests indicated by initials are: Bois de Chapelles, Bois des Vaux, Bois d'Hermet, Forêt de Bourgon, La Lande Royale. Refer to Table 2 for sample sizes.

Table 2.  Numbers of F1 hybrids recorded from 58 ponds across four age classes of newt. Cc and Cm indicate T. cristatus and T. marmoratus cytoplasm respectively. Pond numbers are as indicated in Figure 1.
PondAdult hybridsLarvaeHatchlingsLate embryo
FemalesMalesCollectedF1CollectedRearedcristatusmarmoratusHybridsRearedF1
CcCmCcCm Cc    CcCmF1 CcCm
  1. 1Sixty hatchlings could not be identified at both enzyme markers.

1361     40544737      
1371 1   37936122112    
138      340735425     
139  1   405157 157      
157      250241923     
158    14813890134517913    
162      20915213      
163      42242 42      
167      2918314681     
168      309703324 5    
184   2            
219      52860466      
222             89 1
231      2113131       
2328 1 124810001214117   981 
233  1   199251510   27 1
238      19562501011    
246  1             
247      21830 27      
2491     280231310      
2531               
2782 2 934       36  
2922 3             
2A12     460332911  35  
2A7    712          
2B91               
2C83 1   2831118 1121 1
2D57 2112 56029523   78  
2D8      22013 1012    
2E4             35  
2F2      28383 83      
2F8  1   226441951     
2G7      401715       
2G9      34584374 2 30  
2H1      59116311    
2H6             60  
2K5      32945202131    
2N2      20045142515    
2N8             786 
2P7             45  
2R7      10015 14 1    
431   3            
5411               
N01             29  
N03             60  
N06             786 
N07              117  
N08               48 3 
N10               45 1 
N11               66  
N12                8 12
N13               36  
PP12 1               
PP15   1             
PP18 2               
PP39 1               
PP49 7  4             
PP63   1            
Total41019744827963614601463892202411119185

LARVAL ANALYSIS

A total of 448 larvae was collected from five ponds (Fig. 2, Table 2) in the late summers of 1980 (232), 1986 (278, 2A7, 2D5), 1989 (232), and 1997 (158, 232), and stored in 70% ethanol for up to 17 years. Larvae were identified using the diagnostic microsatellite locus Tcri14 as previously described (Jehle et al. 2000; Krupa et al. 2002; Jehle et al. 2005). Maternal origin was determined by PCR using mtDNA primers MVZ15 and MVZ16, as described for adults above.

HATCHLING ANALYSIS

In all, 9636 recently deposited eggs were collected from 29 different syntopic ponds (Fig. 2, Table 2) during March and April 1987. Eggs were collected together with a small section of the plant on which they were deposited. Up to 50 embryos per batch were raised in 5-L buckets filled with pond water. Buckets were inspected at weekly or three-day intervals. Hatchlings at developmental larval stages 30–35 (Epperlein and Junginger 1982) were harvested and stored in liquid nitrogen. Most of these were freshly hatched, but about 25% were manually released from their jelly capsule. The yolk sac was separated from the developing larva under a binocular microscope. Tissue samples were homogenized separately in Tris-EDTA buffer. Using starch gel electrophoresis, hatchlings were scored for Ldh-2 and Mdh-1 to identify F1 hybrids. The yolk sacs of these hybrids were subsequently scored for Gpi. These three loci provide fixed differences for T. cristatus and T. marmoratus (Rafiński and Arntzen 1987; Arntzen and Wallis 1991). Analysis of 29 adult hybrids showed that the loci are codominantly expressed. The buffer systems used were EDTA-Tris-malate pH 7.4 for LDH, Tris-citrate pH 7.0 for MDH, and Tris-citrate pH 6.0 for GPI (Shaw and Prasad 1970).

EMBRYO ANALYSIS

Analysis of 1119 embryos from 21 ponds (corresponding to the 2001 and 2002 samples in Jehle et al. 2005) followed the same collection procedure as above. Embryos were used for analysis once they had reached stage 28 (Epperlein and Junginger 1982). F1 hybrids were distinguished from parentals using the diagnostic microsatellite markers Tcri14, Tcri32, and Tcri36 (Jehle et al. 2000; Krupa et al. 2002; Jehle et al. 2005). Twenty-three embryos were unequivocally identified as F1 hybrids, and their maternal origin was subsequently determined by PCR using mtDNA 12S primers (F: GGGTTGGTAAATCTCGTGC, R: TAGAGCACCGCCAAGTCCTTTG, [Titus and Larson 1995]) and 35 cycles of: 95°C for 30 sec, 55°C for 30 sec, and 72°C for 30 sec, that yielded a 238-bp fragment.

Results

ADULTS

Triturus cristatus and T. marmoratus reference specimens showed species-specific mtDNA profiles when PCR products were digested independently with each of the four restriction enzymes. For 67 adult F1 hybrids, a T. cristatus profile was found in 60 cases and a T. marmoratus profile seven times (Table 2). The sex ratios for these two classes of newts differed substantially (G= 14.6, P < 0.001). F1 mothered by T. cristatus were 32% male (19/60), whereas T. marmoratus-mothered F1 were 100% male (7/7).

LARVAE

In total, 27 of 448 (6%) larvae from four of the five ponds were identified as hybrid larvae (Table 2). All of these had cristatus mtDNA.

HATCHLINGS

A total of 1460 embryos survived through to stage 30; 1400 were genotyped for two enzyme markers. Among these, totals of 463 T. cristatus (33%) and 892 T. marmoratus (64%) were obtained (Table 2). Forty-five hatchlings (3.2%) were heterozygous (hybrid) for T. cristatus and T. marmoratus alleles at both Ldh-2 and Mdh-1 (Table 2).

In these hybrid hatchlings, yolk sac Gpi, indicating maternal genotype, was characteristic of T. cristatus in 20 cases and of T. marmoratus in 24 cases, and one sample was heterozygous (Table 2). The observation of about equal numbers of hybridizations in both directions falls within the 1:2 to 1:1 range of expectations, based upon respectively either a twofold higher fecundity for female T. marmoratus (Arntzen and Hedlund 1990) or equal annual recruitment at hatchling stage for T. cristatus and T. marmoratus.

EMBRYOS

Of 23 (2.1%) F1 hybrid embryos identified from nine of the 21 ponds, 18 had T. cristatus mtDNA and five had T. marmoratus mtDNA (Table 2).

DATA INDEPENDENCE

F1 hybrids of the same year class from the same pond will usually be siblings, given the low frequency of hybridization (Table 1), small effective number of breeders (Jehle et al. 2001) and small effective population number (Jehle et al. 2001, 2005). For embryo, hatchling and larval classes, the most likely number of independent interspecific matings sampled is simply equal to the number of ponds sampled, for each class.

Adult F1 hybrids from different ponds were always considered independent. Within ponds, newts were only considered independent if they were either of demonstrably different ages determined skeletochronologically (Francillon-Viellot et al. 1990), or they were caught ≥ 4 years apart. Such newts are highly unlikely to be siblings and therefore represent different matings.

These considerations lead to a reduced dataset representing conservative estimates of independent successful interspecific matings sampled (Table 3), and it is upon these that we base our main conclusions. With respect to sex ratio in adult F1, each newt is an independent sample, so we use full F1 adult sample sizes for this aspect.

Table 3.  Summary dataset of F1 hybrids across four developmental stages of newt. Cc refers to newts with T. cristatus cytoplasm; Cm refers to newts with T. marmoratus cytoplasm. The final two columns represent estimates of minimum number of interspecific crosses sampled after correcting for potential siblings from the same pond.
ClassNPondsF1%F1Cc obsCm obsCc minCm min
Adultca. 180024673.860 728 4
Larvae448 5276.027 0 4 0
Hatchling140029443.120241211
Embryos111921232.118 5 6 4

Discussion

One obvious feature of our data is the strong evidence for assortative mating. The extensive hatchling data (Table 2) include only 3% F1 hybrids, similar to previous estimates (Arntzen and Wallis 1991).

Although the present study describes naturally occurring marmoratus-mothered hybrids for the first time, it supports the earlier finding (Arntzen and Wallis 1991) that significantly more adult hybrids are cristatus-mothered than we would expect by chance (28/32 = 88%; G test of goodness of fit to 1:1, with correction for continuity = 18.4, P < 0.001) (60/67 = 90% in uncorrected data). The allozyme analysis of hatchlings, however, leads to rejection of the mating asymmetry hypothesis, which we had suggested based on behavioral interactions (Zuiderwijk 1986, 1990; Zuiderwijk and Sparreboom 1986). Hybrid hatchlings mothered by T. cristatus are no more common than marmoratus-mothered F1 hybrid offspring (12/23 = 52%; G test of goodness of fit to 1:1 = 0.04; P > 0.1) (20/44 = 45% in uncorrected data). The adult population of hybrids (28:4) is not the same as the hatchling population (12:11) (G test of independence = 8.5; P < 0.01). So the asymmetry of parentage of F1 hybrid adults cannot be attributed to sperm competition, gamete incompatibility, or mating bias, as we had previously believed (Arntzen and Wallis 1991). The simplest explanation is that the bias results from greater mortality of posthatchling marmoratus-mothered hybrids than cristatus-mothered hybrids.

The inferred mortality of F1 hybrids is in addition to the 50% mortality caused by the extraordinary case of chromosome 1 heteromorphism in both parental species (Macgregor and Horner 1980). This remarkable balanced lethal system expresses itself at the late tail-bud stage of embryonic development (stages 28–30; Epperlein and Junginger 1982). The differential selection that we have identified takes place after this phase, possibly at metamorphosis.

SEX RATIOS OF ADULT F1 HYBRIDS

An additional novel finding in our data involves an apparent dependence between the sex of the F1 hybrid and the maternal parent. That is, marmoratus-mothered hybrids are much more likely to be male than cristatus-mothered hybrids. This is partly why marmoratus-mothered hybrids were not found in the earlier study, because no males were included. Hybrid frequency data summed across several studies in Mayenne (129 females, 75 males; Table 1) support Haldane's Rule (72% excess females), but this is a simplification of a more complex underlying pattern in which sex ratios are skewed in opposite directions and asymmetrically between the two cytoplasmic classes of newt.

DIFFERENTIAL MORTALITY OF HYBRID CLASSES

Given the observations, the best explanation is that differential mortality of F1 hybrids is taking place according to the direction of the interspecific cross. As 11/23 = 48% of hybrid hatchlings (24/44 = 55% uncorrected) have T. marmoratus mothers in contrast to only 4/36 = 11% of larvae + adults (7/94 = 7% uncorrected), this mortality must be taking place after hatching (and probably before metamorphosis) in marmoratus-mothered larvae. If embryos are pooled with hatchlings, the proportion of marmoratus-mothered larval embryos is 15/33 = 45% (29/67 = 43% uncorrected).

Under this scenario, the 3.1–4.7% adult hybrids that we observe in the region (Table 1) underrepresents the amount of interbreeding of the two species. It is difficult to make inferences from percentage of hybrid F1 in the population across developmental stages because sampling methods are necessarily different and there is also spatial and temporal variation among samples. The two best estimates of hybrid proportions are in hatchlings and adults. The observed proportion of F1 hybrid hatchlings is 44/1400 (3.1%) representing hybrids from both crosses. Hybrids from T. cristatus mothers appear to increase in frequency during life (from 1.3% in embryos to about 3.5% in adults) whereas marmoratus-mothered hybrids decrease in frequency during life (from 1.6% in embryos to 0.4% among adults). This suggests that the cristatus-mothered F1 hybrids experience some hybrid vigor, although the effect could be complicated by changes in the amount of hybridization through time (Albert et al. 2006). Skeletochronology of 57 F1 hybrid adults collected across 21 ponds, plus 69 T. cristatus and 88 T. marmoratus (Francillon-Viellot et al. 1990) showed that hybrids tended to be older than both T. cristatus (P < 0.01) and T. marmoratus (P < 0.05). Although the number of years over which hybrids reproduce is expected to be larger than for either parental species, reduced fertility due to failures in spermatogenesis (Bataillon and Tcherniakofsky 1932; White 1946; Lantz 1947; Lantz and Callan 1954) and reduced embryo hatching success (Arntzen and Hedlund 1990) may negate increased longevity in both male and female hybrids.

We have uncovered a case of apparent symmetry in hybrid formation, but large asymmetry in hybrid fitness and inverted asymmetric sex ratio between the two hybrid classes.

EXPLANATORY GENETIC MODELS

BDM incompatibilities involving cytoplasmic factors or sex chromosomes can be invoked to explain asymmetric postzygotic isolation (Orr 1995; Welch 2004). Triturus cristatus and T. marmoratus have XX/XY determination (Sims et al. 1984; Hillis and Green 1990; Schmid and Steinlein 2001). In our data, the biggest fitness discrepancy is between adult females of the two different F1 hybrid classes (41:0). These two classes have the same chromosomal makeup, implicating a nuclear–cytoplasmic interaction (Clark and Lyckegaard 1988; Turelli and Orr 2000; Rand et al. 2001; Dowling et al. 2008) (Fig. 3) or maternal effect. Cytoplasmic incompatibility could involve mitochondrial genes, for which there are precedents involving cytochrome c and cytochrome oxidase I in Tigriopus (Edmands and Burton 1999; Blier et al. 2001; Willett and Burton 2001; Rawson and Burton 2002; Sackton et al. 2003; Rand et al. 2004; Ellison and Burton 2006, 2008). This model is attractive because we know that BDM incompatibilities tend to accumulate on X chromosomes (Coyne and Orr 1989, 2004). Such systems may be a common cause of exceptions to Haldane's rule (Laurie 1997). Such an incompatibility might be expected to act early in development, leading to a greater effect on viability than fertility (Turelli and Moyle 2007). Overlying this incompatibility is a weak Haldane effect with a variety of possible explanations (Coyne and Orr 2004). The greater effect on Cm males suggests at least two additional effects beyond the nuclear–cytoplasmic effect.

Figure 3.

Genetic model for inherited factors for reciprocal-cross interspecific F1 hybrids. Shown are: X, Y, autosomes (A), and cytoplasm (C), each subscripted “c” for cristatus and “m” for marmoratus. Shading indicates entities that are inferred to be involved in incompatibility. In the case of male F1 hybrids, several combinations of uniparentally inherited factors and autosomes may contribute to incompatibility, but at least one of the sex chromosomes is probably involved.

ACCUMULATION OF HYBRID INVIABILITY

Because the species are evolutionarily far apart (Steinfartz et al. 2007), it is to be expected that several postzygotic effects would occur. Some workers have attempted to put a timeframe on the accumulation of hybrid inviability to make an “incompatibility clock” (Bolnick and Near 2005). Data from centrarchid fish yielded a surprisingly low rate of 3.13% loss of viability of F1 hybrids per million years (Bolnick and Near 2005), estimated across taxa up to 34 Ma apart. If we take 24 Ma as the divergence time of our two newt species (Steinfartz et al. 2007), we would expect 75% loss of F1 viability. Taking the 41 cristatus-mothered female F1 hybrids as the expected number for each of the four classes (Fig. 3) of our 67 adult hybrids equates to a ((4 × 41) − 67) × 100/(4 × 41) = 59% loss of hybrids. Despite the assumptions made in making this calculation (life-history stages comparable; linear rate; no positive or negative viability effects in cristatus-mothered female hybrid F1), the observed viability loss is similarly low.

Associate Editor: D. Presgraves

ACKNOWLEDGMENTS

W. van Ginkel (Amsterdam) and S. Delamaire (Sheffield) helped in the laboratory, J. Castanet (Paris) determined the age of some hybrids by skeletochronology and M. Schilthuizen (Leiden) tested newt ovaries for the presence of Wolbachia infection. This work was supported by a UK Natural Environment Research Council grant No. GR9/1360 to TB and GPW. GPW gratefully acknowledges the support of a Temminck Fellowship from the National Museum of Natural History, Leiden, which permitted the completion of this work. The comments of two anonymous reviewers helped to improve the clarity of the article.

Ancillary