Introduction

Sexual reproduction is almost universal as a means of reproduction in a wide range of organisms and has the advantage of allowing gene exchanges to purge detrimental mutations and/or rapidly evolving parasites (Maynard Smith, 1978; Hamilton et al., 1990). Moreover, a two-sex system is generally used by most sexual organisms and has evolutionary advantages in numerous situations (Iwasa and Sasaki, 1987). Despite this, sex determination mechanisms are diverse and can differ even in closely related species (Peichel et al., 2004). In Drosophila melanogaster, sex is determined by the ratio of the chromosome number (X) to the number of the autosomal set (A) (Bridges, 1921). Animals with an X/A ratio of 1 become female, and those with an X/A ratio of 0.5 become male. In the housefly, Musca domestica, however, a male-determining factor (abbreviated as ‘M’) on the Y chromosome is critical for development into a male (Hediger et al., 1998). Without the ‘M’ factor, a female-determining factor (abbreviated as ‘F’) functions maternally, such that the animal develops as a female (Dübendorfer and Hediger, 1998).

Goldschmidt (1934) proposed the theory of sex determination of Lymantria dispar after studying hybrids among Japanese local races. In L. dispar, the heterogametic sex is female (ZW) and the homogametic sex is male (ZZ). Goldschmidt (1955) hypothesized that the female sex-determining factor, ‘F’, is a maternally inherited cytoplasmic factor and that the male sex-determining factor, ‘M’, is located on the Z chromosome (see Appendix; Figure 4). Goldschmidt (1934) suggested that different races of L. dispar possess female- and male-determining factors of different strengths. Within each race these factors are well balanced, but in inter-race crosses changes in the relative levels of these factors can generate intersex individuals or even cause sex reversal. Goldschmidt (1934) inferred that the female- and male-determining factors were weakest in Hokkaido subspecies (L. dispar praeterea) and stronger in Honshu (Tokyo (TK)) subspecies (L. dispar japonica). Goldschmidt (1934) obtained all-male broods from a cross between Hokkaido females and Honshu males. Inversely, Goldschmidt (1930) obtained all-female broods of L. dispar by backcrossing, that is, when F1 females resulting from a cross between Honshu females and Hokkaido males were mated with Hokkaido males (Appendix Table A1). The results of these crosses indicate that presumably weaker Hokkaido sexes disappear in the F1 or the backcrossed F2 generation when crossed with the other sex from the Honshu subspecies. The production of all-female or all-male broods is thus interesting in the context of sex determination in L. dispar.

Higashiura et al. (1999) found all-female broods owing to male death in L. dispar at Bibai (BB) in Hokkaido, Japan, and maternal inheritance of this trait. Higashiura et al. (1999) also observed six all-female broods out of 66 egg batches in the next generation in the same 0.2-ha forest, indicating the persistence of all-female broods. Clarke and Ford (1982) re-examined crosses between Hokkaido females and Honshu males and obtained all-male broods as Goldschmidt (1934) indicated. However, the characteristics of all-female broods obtained by the Goldschmidt backcross have not been re-examined.

If the Goldschmidt backcross indeed generates all-female broods in Hokkaido L. dispar, mitochondrial DNA (mtDNA) analysis of normal and all-female broods could reveal the maternal origin of the all-female broods. Bogdanowicz et al. (2000) described mtDNA variation among worldwide populations of L. dispar and resolved the haplotypes into four groups: (1) Okinawa, Japan, (2) Hokkaido, Japan, (3) Honshu and Kyushu, Japan, and mainland Asia, and (4) Europe, Tunisia, and North America. To test whether normal and all-female broods of Hokkaido L. dispar have different origins, we analysed mtDNA haplotypes in both broods. We also re-examined the Goldschmidt backcross that produces all-female broods at the backcrossed F2 generation. If the hypothesized weaker male sex of Hokkaido L. dispar indeed disappears in the Goldschmidt backcross, then females of the all-female broods mated with the hypothesized strong Honshu male should produce normal broods. In this study, we present the results of such crosses, which show that the origin of all-female broods is the hybridization between Honshu and Hokkaido subspecies.

At present, all-female broods are known to arise from selective male killing at egg or larval stages, particularly in arthropods (Majerus, 2003). Several microbial symbionts, including organisms in the genera Wolbachia, Spiroplasma, Rickettsia and Arsenophonus, are known to produce all-female broods in their host insects, as cytoplasmic elements are inherited only matrilineally (Hurst et al., 1997, 2003; Majerus, 2003). To test whether certain bacteria are present in all-female broods, diagnostic polymerase chain reaction (PCR) was carried out on the larvae of all-female broods of Hokkaido L. dispar using primers that amplify fragments of genes found in the above-mentioned bacteria and in eubacteria in general. As endosymbiont bacteria show up as bright flecks in the cytoplasm by 4′,6-diamidino-2-phenylindole (DAPI) staining, we also stained the eggs of all-female broods of Hokkaido L. dispar with DAPI to test whether certain symbionts were involved in the production of the all-female broods. Given that microsporidians are commonly found in lepidopteran insects (Agnew et al., 2003), we would expect DAPI staining to reveal both prokaryotic and eukaryotic endosymbionts in all-female broods.

The aims of this study were (i) to test whether microbial symbionts are involved in the production of all-female broods by diagnostic PCR and DAPI staining and (ii) to test whether all-female broods of L. dispar originate from hybridization between Honshu and Hokkaido subspecies by mtDNA analyses and the Goldschmidt backcross. This information will be useful in characterizing the molecular mechanisms of sex determination in L. dispar.

Materials and methods

Study species

Inoue (1982) phenotypically classified L. dispar of Japan into five subspecies: L. dispar praeterea Kardakoff for Hokkaido populations, L. dispar japonica (Motschulsky) for the mainland (Honshu, Shikoku and Kyushu) populations (Figure 1) and three other island subspecies. All of these subspecies are univoltine. Schintlmeister (2004) treated the Hokkaido populations as L. dispar dispar, the same as mainland Asian and European populations. However, we adopted Inoue's classifications in this paper, because Hokkaido mtDNA haplotypes are quite different from Asian haplotypes, showing about 2% mtDNA sequence divergence (Bogdanowicz et al., 2000). Pogue and Schaefer (2007) reported that the distribution of L. dispar japonica is in Honshu, Shikoku and Kyushu, and extends north into parts of Hokkaido. They also treated Hokkaido populations as a distinct species, L. umbrosa (Butler). The type locality of the lectotype male for the revised species, L. umbrosa, is the most southwestern region of Hokkaido, or Hakodate (Figure 1). The distributions of the two species in Hokkaido are unclear. Therefore, we did not adopt the classifications described by Pogue and Schaefer (2007).

Figure 1
figure 1

Map showing collection sites for egg masses of L. dispar used in this study. The arrow indicates North. Bibai (BB) (43°17′N, 141°51′E), Shikaoi (SK) (43°06′N, 142°57′E), Abira (AB) (42°45′N, 141°49′E), Niikappu (NP) (42°24′N, 142°14′E), Tomakomai (TM) (42°38′N, 141°37′E) and Asahikawa (AS) (43°49′N, 142°19′E) are in Hokkaido, Japan. Sannai (SN) (40°48′N, 140°40′E), Rifu (RF) (38°19′N, 140°56′E), Kawasaki (KW) (38°10′N, 140°40′E) and Tokyo (TK) (35°38′N, 139°23′E) are in Honshu, Japan. L. dispar praeterae is distributed in Hokkaido, Japan. L. dispar japonica is distributed in Honshu, Shikoku and Kyushu, Japan (Inoue, 1982). The Tsugaru Strait divides these two subspecies into their distributions. Hakodate, Hokkaido, is the type locality of L. umbrosa (Pogue and Schaefer, 2007).

Full size image

In Hokkaido, larvae pupate in July and August. Adults emerge, mate and deposit eggs in August and September. As each female lays its full egg complement in only one mass containing 100–900 eggs, it is possible to recognize the properties of the original female by rearing eggs from individual masses. Eggs diapause during winter and hatch in the following May. Larvae feed on leaves of the Japanese larch Larix leptolepis, the Japanese birch Betula platyphylla and many other broad-leaved trees.

Sample collection and rearing of Hokkaido L. dispar

We collected egg masses at BB (Figure 1) and Shikaoi (SK) in Hokkaido, Japan. The collection sites were two Betula platyphylla forests (designated Betula-1 and Betula-2), Larix leptolepis forest at BB and one Betula platyphylla forest at SK. We collected and analysed seven egg masses from the Larix forest at BB in September 1999, six egg masses from Betula-1 forest at BB in October 1999 and 2001, 41 egg masses from Betula-2 forest at BB in October between 1999 and 2002, 11 egg masses at SK in November 2000 and 3 egg masses at Niikappu (NP) in December 2002. After collection, the egg masses were kept individually in a cold room at 0 °C until May of the following year. Then, about 100 eggs from each of the egg masses were placed in an incubator at 20 °C. After hatching, 20 first instar larvae from each of the egg masses were frozen in liquid nitrogen and stored at –80 °C for mtDNA analysis. The remaining larvae were reared to pupation on an artificial diet (product #F9642B, Bio-Serv, Frenchtown, NJ, USA). We counted unhatched eggs in some of the egg masses to obtain egg-hatch rates. Sex ratios among the pupae were recorded for each of the egg masses. As egg-hatch rates were <60% in all-female (male death) broods in the previous study (Higashiura et al., 1999), we regarded egg masses showing egg-hatch rates of 70% or more as normal sex-ratio broods without rearing.

PCR detection of bacterial symbionts

For PCR detection of bacterial symbionts, we used larvae hatched from three and two egg masses of all-female broods and normal broods, respectively, collected from Betula-2 forest at BB, Hokkaido, in October 1999. All of these broods were subjected to mtDNA sequencing as described in the next section. Five first instar larvae preserved in 99% ethanol from each of the egg masses were individually subjected to DNA extraction using a QIAamp Tissue Mini Kit (Qiagen, Hildeu, Germany). The purified DNA was eluted with 200 μl of TE buffer (10 mM Tris-HCl (pH 8.0), 0.1 mM ethylenediaminetetraacetic acid). For detection of potentially male-killing bacterial symbionts, diagnostic PCR analysis was conducted as described (Tsuchida et al., 2002). The following primers were used: ftsF and ftsR for the ftsZ gene of Wolbachia spp., 16SA1 and TKSSspR for the 16S rRNA gene of Spiroplasma spp., 16SA1 and Rick16SR for the 16S rRNA gene of Rickettsia spp., 16SA1 and Ars16SR for the 16S rRNA gene of Arsenophonus spp., 16SA1 and 16SB1 for the 16S rRNA gene of general eubacteria, and MtrA1 and MtrB1 for the mitochondrial rRNA gene of the host insect (Appendix Table A2). For positive control reactions, DNA samples from Callosobruchus chinensis (Coleoptera: Bruchidae) infected with Wolbachia sp., Drosophila melanogaster (Diptera: Drosophilidae) infected with Spiroplasma sp., Nephotettix cincticeps (Homoptera: Deltocephalidae) infected with Rickettsia sp. and Diaphorina citri (Homoptera: Psyllidae) infected with Arsenophonus sp. were also subjected to diagnostic PCR.

DAPI staining of embryos to detect endosymbionts

We collected the eggs of an all-female brood obtained by crossing a female of an all-female brood (44 females and 0 males) with a male of a normal brood (14 females and 16 males) collected at Abira (AB) in October 2005. As a control, we obtained eggs of normal broods by crossing individuals from two normal broods (a mother from a brood containing 18 females and 16 males, and a father from a brood containing 20 females and 17 males) collected at BB in October 2006. Eggs were collected from both all-female and normal sex-ratio broods at 3 days after oviposition. The collected eggs from each brood were treated with 50% (v/v) aqueous bleach solution for 30 s to 3 min. After washing with phosphate-buffered saline (PBS), eggs were fixed with Carnoy's fixative for 3 h in a siliconized tube at 4 °C. Subsequently, eggs were placed in PBS Tween for 15 min, and were treated with 50% aqueous bleach solution for 1 min before the surgical procedure. The chorion and vitelline membranes were removed with tungsten needles (cat. #500134, World Precision Instruments, Sarasota, FL, USA) under a microscope in PBS Tween solution. The dechorinated eggs were transferred to PBS Tween solution in another siliconized tube. A volume of 1 ml of DAPI solution (1 mg ml−1) was added and the mixture was incubated for 5 min. The stained eggs were observed under a fluorescence microscope using an ultraviolet filter.

As a positive control, Callosobruchus chinensis embryos infected with Wolbachia (strain jC) (Kondo et al., 2002) were stained using DAPI solution. We also evaluated C. chinensis without Wolbachia obtained by tetracycline treatment (Kondo et al., 2002). C. chinensis eggs were collected within 24 h after oviposition. The eggs were submerged in Carnoy's fixative with adzuki beans, and then detached from the beans. After removing the beans from the fixative, the eggs were fixed in a siliconized tube for 2 h at 4 °C. Eggs were rehydrated in PBS Tween solution for 15 min and stained with DAPI solution for 5 min. The observational method for stained C. chinensis eggs was the same as that used for L. dispar eggs.

mtDNA sequencing

In all, 5 or 10 frozen first instar larvae from each of the broods were ground with a pestle in a pre-chilled mortar and subjected to total DNA extraction using a QIAamp DNA Mini kit (Qiagen). In all, 1075 bp of mtDNA segment encoding the parts of three genes, cytochrome c oxidases I and II and nicotinamide adenine dinucleotide dehydrogenase subunit 1, was amplified by PCR from 2 ng of template DNA. The gene-specific primers have been described previously (Bogdanowicz et al., 2000) (Appendix Table A2). PCR (40 cycles of 94 °C for 0 s, 48 °C for 0 s and 72 °C for 15 s) was carried out using a Rapid Cycler (Idaho Technology, Salt Lake City, UT, USA). The amplified products were directly subjected to DNA sequencing using a Big Dye Terminator Cycle Sequencing Ready Reaction Kit and ABI PRISM 3100 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). The DNA sequences determined in this study have been deposited in the DDBJ, EMBL and GenBank nucleotide sequence databases under accession numbers AB244647–244657 and AB244665. Pairwise sequence-based distances among the studied mtDNA haplotypes (a total of 1075 base pairs across the three genes) were estimated using PAUP* 4.0b10 (Swofford, 2003) under the Hasegawa, Kishino and Yano (HKY) model (Hasegawa et al., 1985).

Crossing experiments

We collected two egg masses at Sannai (SN), Aomori in northern Honshu, Japan, in November 2001. The mtDNA haplotypes of these larvae were determined using the protocol described above. Hatched larvae were reared to adulthood and sexed in July 2002. The females were crossed with males from egg masses collected from Betula-2 at BB, Hokkaido, in October 2001. The offspring were reared to adulthood and sexed in July 2003. The F1 females were backcrossed with males with Hokkaido mtDNA haplotypes from egg masses of normal broods collected at NP, AB and Tomakomai (TM), Hokkaido, Japan in December 2002. The backcrossed F2 offspring were reared to adulthood and sexed in July 2004.

Females of all-female broods emerged from egg masses collected from Betula-2 at BB, Hokkaido, in October 2001. In July 2002, they were crossed either with (one) males from Honshu egg masses or (two) males from BB, Hokkaido; the latter cross served as a positive control. The Honshu egg masses were collected in November 2001 at Rifu (RF), Kawasaki (KW) and also at SN. Larvae from these crosses were reared to pupation and sexed in July 2003. An artificial diet (Bio-Serv; product #F9642B) was used for rearing.

A Hokkaido female that emerged from an egg mass collected at Asahikawa (AS) in April 2006 was crossed with a TK male from an egg mass collected at TK in October 2005. The offspring were reared and sexed in July 2007.

The exact P-value for the binomial proportion test (hereafter abbreviated as the binomial test) (SAS Institute, 1999) is used to analyse the egg hatch rate and sex ratio data.

Results

Detection of endosymbionts

Diagnostic PCR analysis of 15 larvae from three all-female broods of L. dispar detected no Wolbachia, Spiroplasma, Rickettsia, Arsenophonus or general eubacteria. These symbionts were also not detected in the 10 larvae from two normal sex-ratio broods. A mitochondrial gene of the host insect was successfully amplified from all of the samples, indicating that the DNA extraction was successful. Each positive control sample gave specific PCR products of the expected sizes for each bacterium and general eubacteria, confirming the specificity of the PCR (data not shown).

No differences were detected between L. dispar embryos of all-female and normal broods by DAPI staining (Figures 2a and b). However, in C. chinensis embryos infected with Wolbachia, we were able to detect a cytoplasm containing large amounts of Wolbachia (Figure 2c). In embryos without Wolbachia, which were obtained through tetracycline treatment, we found a single type of nucleus, lacking all apparent endosymbionts (Figure 2d).

Figure 2
figure 2

Eggs stained with DAPI solution and viewed under a fluorescence microscope using an ultraviolet filter to detect endosymbionts. As DAPI stains nuclei exclusively, we observe a cytoplasm containing many parasitic bacteria. The length of each white bar represents 20 μm. (a) An embryo deposited by a female from an all-female brood of Lymantria dispar at 3 days after oviposition. (b) An embryo from a normal sex-ratio brood of L. dispar at three days after oviposition. (c) An embryo of Callosobruchus chinensis infected with Wolbachia (strain jC) (Kondo et al., 2002) within 24 h after oviposition. The arrow shows cytoplasm with abundant bacteria. (d) An embryo of C. chinensis without Wolbachia obtained by tetracycline treatment.

Full size image

mtDNA sequences of all-female and normal sex-ratio broods of Hokkaido L. dispar

The mtDNA haplotypes of the all-female broods were found to be quite different from those of the normal sex-ratio broods (Figure 3). Pairwise mtDNA sequence divergence values from comparisons of the normal broods ranged from 0 to 0.0028 (substitution per site). The haplotype H1 in Figure 3 had been identified previously and was included in the Hokkaido haplotypes reported by Bogdanowicz et al. (2000). Nevertheless, the mtDNA haplotype (A1) of the all-female broods was the most common among those seen in Honshu (Japan), Kyushu (Japan) and mainland Asia (Bogdanowicz et al., 2000). Pairwise mtDNA sequence divergence values from comparisons of the all-female broods ranged from 0 to 0.0019. However, the sequence divergence between the normal and the all-female broods ranged from 0.016 to 0.019.

Figure 3
figure 3

Variable sites of three genes, cytochrome c oxidases I and II (COI and COII), and nicotinamide adenine dinucleotide dehydrogenase subunit 1 (ND1), in the mtDNA of (the upper) all-female broods and (the lower) normal sex-ratio broods of Lymantria dispar, in Hokkaido, Japan. There are 17 fixed differences between all-female broods and normal sex ratio broods. These fixed different sites are shaded in grey. The 4–5-digit numbers in the upper column indicate the nucleotide position corresponding to the mtDNA sequence of Drosophila yakuba (Clary and Wolstenholme, 1985). ‘Number’ indicates the number of egg masses with the given haplotype among the egg masses collected at the locality. ‘A1’ is the same as the most common haplotype (28 out of 44 individuals) in Russia, China, Japan and Korea, and ‘H1’ is the same as the most common haplotype (3 out of 5 individuals) in Hokkaido, Japan (Bogdanowicz et al., 2000). Non-synonymous substitutions for A1 or H1 are boxed. BB, Bibai; NP, Niikappu; SK, Shikaoi (Figure 1).

Full size image

The 18 all-female broods shown in Figure 3 contained no males at all. In one brood of the minimum rearing number shown in Figure 3, the sex ratio was 6:0 (females:males) (P(1:1)=0.03: the binomial test). In one brood of the maximum rearing number shown in Figure 3, the sex ratio was 29:0 (P(1:1) <0.001: the binomial test). The egg-hatch rates of the all-female broods were between 38 and 58%, indicating that all-female production was probably because of male death. The 50 normal sex-ratio broods in Figure 3 had more than 70% egg-hatch rate, or had a sex ratio of 1:1 (for example, the most skewed ratio of one normal brood in Figure 3, in which females:males=12:4; P(1:1)=0.08: the binomial test).

Crossing experiments

We used five normal sex-ratio broods of Honshu L. dispar in this study (Table 1). The mtDNA haplotype of all of these broods was A1, an Asian type distributed through Asia, including Japan, south of Hokkaido (Bogdanowicz et al., 2000). Of 15 broods of Hokkaido L. dispar used for crossing experiments in this study, 11 broods exhibited normal sex ratios near 1:1, whereas 4 broods consisted of all females. Egg-hatch rates were ⩾67% in the normal broods and ⩽56% in the all-female broods, indicating that all-female production was probably because of male death. In the Hokkaido L. dispar, the mtDNA haplotypes of the normal broods were H1 or H6, which are haplotypes that are found only in Hokkaido.

Table 1 Characteristics of Lymantria dispar broods used for crossing experiments shown in Tables 2–4
Full size table

When the Honshu females and the Hokkaido males shown in Table 2, both of which came from broods with normal sex ratios, were crossed, the F1 offspring exhibited normal sex ratios. However, when F1 females were again crossed with Hokkaido males, the backcrossed F2 offspring were entirely female. Two series of independent experiments gave the same results (Table 2). These results are in perfect agreement with Goldschmidt's finding of all-female brood production in the hybridization crosses between Honshu females (L. dispar japonica) and Hokkaido males (L. dispar praeterea) (Goldschmidt, 1930; Appendix Table A1).

Table 2 Re-evaluation of Goldschmidt’s backcross result
Full size table

When Hokkaido females from all-female broods were mated with Hokkaido males, their offspring were all females in five replicates (Table 3a). These results confirm the result of Higashiura et al. (1999), indicating that the all-female broods produce all-female broods when females of these broods are mated with Hokkaido males.

Table 3 Crossing of the all-female broods and the Hokkaido males or the Honshu males
Full size table

When Hokkaido females of the all-female broods were mated with Honshu males, however, the offspring exhibited normal sex ratios of approximately 1:1 in five replicates (Table 3b).

We confirmed that an AS, Hokkaido, female mated with a TK, Honshu male produced an all-male brood. The egg hatch rate of the brood was 79%, which is significantly higher than 50% (P<0.0001: the binomial test; Table 4).

Table 4 Results of an Asahikawa (AS), Hokkaido, female mated with a Tokyo (TK), Honshu, male (mitochondrial DNA (mtDNA) haplotype is shown in Figure 3)
Full size table

Discussion

Higashiura et al. (1999) found all-female broods in the Hokkaido population of L. dispar and showed that all-female broods mated with Hokkaido males produce all-female broods. Male killing caused by cytoplasmic bacterial symbionts has been reported in a variety of insects (Hurst et al., 1997, 2003; Majerus, 2003). Although we analysed the infection status of all-female and normal broods using PCR tests specific for Wolbachia, Rickettsia, Spiroplasma, Arsenophonus and general eubacteria, we found no bacteria in either all-female or normal broods of L. dispar larvae in Hokkaido. Moreover, we obtained cytological evidence showing that all-female broods of L. dispar do not have obvious prokaryotic or eukaryotic endosymbionts that might cause male killing, as we found no differences in DAPI staining between the embryos of the all-female brood and normal sex ratio. In contrast, the cytoplasm of Callosobruchus chinensis was stained by DAPI because of its high content of bacteria, or Wolbachia. These results suggest that male death in L. dispar may be caused by something other than endosymbionts, but do not exclude the possibility of viral influences. An RNA virus is reported to cause selective male-killing in the oriental tea tortrix, Homona magnanima (Nakanishi et al., 2008).

The mtDNA analysis of L. dispar performed in this study revealed the different origins of the all-female broods and normal sex-ratio broods found in the same populations in Hokkaido, Japan. The normal sex-ratio broods possessed Hokkaido mtDNA haplotypes, whereas the all-female broods had Asian (including Honshu) mtDNA haplotypes. Here we must consider the results of Goldschmidt's backcross (1930), that is, F1 females resulting from a cross between Honshu, Japan, females (L. dispar japonica) and Hokkaido males (L. dispar praeterea) mated with Hokkaido males, as an explanation for the all-female broods observed in Hokkaido.

We successfully reproduced Goldschmidt's result by backcrossing Honshu females with Hokkaido males, both of which were obtained from normal broods, and found that the backcrossed F2 progeny were entirely female (Figure 4; Appendix Equations (A1) and (A2)). On the other hand, when Hokkaido females from all-female broods were crossed with Honshu males, the offspring exhibited normal 1:1 sex ratios (Appendix Equation (A3)). The results of the crosses conform to the prediction made by Goldschmidt's theory. The all-female broods observed in a small fraction of Hokkaido females may have originated from hybridization between Honshu females and Hokkaido males. These all-female broods produce all-female broods as long as the females of these broods are mated with Hokkaido males.

Figure 4
figure 4

Results of hybridization crosses in Lymantria dispar according to Goldschmidt's sex determination theory (Goldschmidt, 1934). Circles indicate the cytoplasm holding the female determinant factor (F). Large ovals indicate the Z chromosome holding the male determinant factor (M). ‘F’ and ‘M’ indicate sex determinant factors, not the chromosome. Small ovals indicate W chromosome. Open and grey symbols, circles and ovals, show weak and strong sex determinants, respectively. Female is ZW and male is ZZ. Honshu individuals have strong sex determinants indicated by ‘st’, and are represented by (Fst) Mst W for females, and (Fst) Mst Mst for males (parentheses indicate a cytoplasmic factor). Hokkaido individuals possess the weakest sex determinants, and are represented by (FHok) MHok W for females, and (FHok) MHok MHok for males. (a) All-female broods obtained by backcrossing, that is, F1 females of a cross between strong Honshu females (L. dispar japonica) and weak Hokkaido (Hok) males (L. dispar praeterea) mated with Hokkaido males. Lethality is indicated by the cross bar. (b) All-male broods obtained by crosses between weak Hokkaido females and strong Honshu males. The dashed cross bar indicates sex-reversed female, or male.

Full size image

The origin of the Honshu females found in Hokkaido is not clear. Bogdanowicz et al. (2000) found that one of six analysed Hokkaido individuals carried an Asian mtDNA haplotype and proposed that this female may be ‘either a migrant from elsewhere in Asia or an individual of mixed ancestry’. We are currently investigating the distribution of mtDNA haplotypes throughout Hokkaido.

The reciprocal hybridization crosses between L. dispar individuals give the opposite results, producing all-male broods (Figure 4; Appendix Equation (A4); Goldschmidt, 1934). Clarke and Ford (1982) reassessed this cross, asking whether Hokkaido females mated with Honshu males would produce all-male progeny, including both sex-reversed males and normal males. Goldschmidt (1934) thought that F1 females of this cross had the weakest female and strong male sex determinants and would be transformed into males. Clarke and Ford (1982), however, found that all of the phenotypically male hybrids lacked the heterochromatic W chromosome found in normal females. They stated that ‘the excess of males in Goldschmidt's classic cross is due to the Haldane (1922) effect of genic imbalance and not to sex reversal’. We obtained an all-male brood by crossing a Hokkaido female with a TK male. The egg-hatching rate of the brood was significantly >50%. This implies that either incompatibilities may manifest before or after oviposition. However, Goldschmidt (1934) likely took the high hatching rate of all-male broods into consideration, as his work mentions the different survival rates of sex-reversed males and females. The survival rate and the presence of the heterochromatic W chromosome during larval stages of such all-male broods should be studied further, as Clarke and Ford (1982) may have evaluated the presence of the heterochromatic W chromosome in late-instar larvae, after the female larvae had died.

Hornett et al. (2006) proposed the existence of a suppressor gene against the male-killing trait caused by Wolbachia in the Southeast Asian and Polynesian nymphalid butterfly Hypolimnas bolina. Polynesian females infected with the bacteria exhibited the male-killing trait; however, Southeast Asian females produced 1:1 sex ratio progenies even when infected with the bacteria. Hornett et al. (2006) concluded that zygotic suppression is controlled at a single, genetically dominant locus because they obtained 1:1, 2:1 and 4:1 sex ratio progenies in the F1, backcrossed F2 and backcrossed F3 generations of Southeast Asian females (backcrossed with Polynesian males). However, in L. dispar, the presence of a suppressor gene only on the Z chromosomes of the Honshu and Asian backgrounds may be able to produce entirely all-female broods in the backcross, but we found no obvious symbionts.

Goldschmidt's theory of sex determination is based on the hypothesis that different races of L. dispar possess female- and male-determining factors of different strength (Goldschmidt, 1934). The balance within each race and the imbalance of sex-determining strength between different races can be considered to be the outcome of a coevolutionary ‘arms race’ between the sexes in a recent theory of sexual conflict (Arnqvist and Rowe, 2005). This theory is based on the presence of antagonistic interactions between females and males, for example, mating frequencies. At the species-specific level, these coevolutionary ‘arms races’ between the sexes are hidden by continuous adaptation and counter-adaptation, but they propel the evolution of the mating system in a group of insects (Arnqvist and Rowe, 2002). In L. dispar, the ‘arms races’ between the sexes are hidden in race-specific crosses, but the outcome of the conflict may occur in crosses between sexual races. The sex-determining genes reflect the outcome of this sexual conflict (Arnqvist and Rowe, 2005) in the manner indicated by Goldschmidt's theory of sex determination.

Goldschmidt (1934) pointed out that ‘an intersex is an individual which starts development with its original, gametic sex, but changes sex during development’. He regarded the timing of this change, or turning point, to be important and formulated the time law of intersexuality. Dietrich (2003) pointed out that ‘Goldschmidt was attempting a synthesis of evolutionary and developmental biology similar to the synthesis that has been emerging during the past 15 years (known as evolutionary developmental biology, or evo-devo)’. As shown in the results of the inter-race crosses in this paper, his theory should be re-evaluated, at least for L. dispar. Goldschmidt (1955) assumed that the ‘F’ factor on the W chromosome was the maternally inherited cytoplasmic factor. In insects such as the blowfly, hessian fly, fungus gnat and house fly, maternal factors are involved in female sex determination (Ullerich, 1984; Inoue and Hiroyoshi, 1986; Stuart and Hatchett, 1991; de Saint Phalle and Sullivan, 1996; Dübendorfer and Hediger, 1998; Sánchez and Perondini, 1999). On the basis of the results presented here, we suggest that the molecular mechanisms of male or female mortality may be revealed by crosses between different races with different sex strengths. Moreover, it is expected that the research investigating these molecular mechanisms will clarify the sexual determinants of L. dispar in the near future.