In two islands located in the southern part of Japan (Tanegashima Island and Okinawa Island), some females of the butterfly Eurema mandarina (Lepidoptera; Pieridae) are known to have a chromosomal constitution of males (ZZ) instead of females (ZW). This incongruence between chromosomal and phenotypic sex can be explained by feminization of genetic males induced by Wolbachia [25,26]. Two distinct strains of Wolbachia, wCI and wFem, have been found in E. mandarina inhabiting these islands. Females having male chromosomes (ZZ) are consistently infected with both wCI and wFem, while females singly infected with wCI are true females (ZW). Despite having complete male chromosomes, ZZ females are morphologically and behaviorally completely female and fully fertile. Moreover, when such individuals are treated with tetracycline hydrochloride, a bacteriostatic antibiotic, during larval development, they develop as butterflies with intersexual morphology. Before carrying out this experiment, the appearance of intersexes in the treated generation was not expected because perturbation of sex determination was believed to occur during early embryogenesis when the sex-determining genes start to be expressed. However, the appearance of intersexes clearly shows that Wolbachia needs to be present during larval development for complete feminization. The fact that butterflies with more male-like characters were generated after longer antibiotic treatments shows that Wolbachia acts continuously during larval development [26] (Figure 2). The presence of Wolbachia during the embryonic stage seems to be necessary for complete feminization since antibiotic treatment during the whole larval stages also generated intersexes instead of complete males. We recently found that the splicing pattern of the sex-determining gene doublesex (dsx) changes according to the Wolbachia infection status (Narita et al., in preparation). Wolbachia-induced feminization has also been found in Eurema hecabe, a sibling species of E. mandarina. The two Wolbachia strains found in feminized E. hecabe are indistinguishable from wCI and wFem derived from E. mandarina based on multi-locus sequence typing [27].

A second case of Wolbachia-induced feminization in diploid insects has been found in the leafhopper Zyginidia pullula (Homoptera; Cicadellidae) [28]. In Z. pullula, females have two X chromosomes (2n = 8AA + XX) while males have only one X chromosome (2n = 8AA + X0). This is contrary to the situation of the Eurema butterflies wherein females have one Z chromosome and males have two Z chromosomes. When Wolbachia-infected Z. pullula females collected in northern Italy were mated with males, they produced exclusively female broods. Close inspection of these female broods revealed that about half of them had an intersexual morphology (i.e., showing upper pygofer appendages, a typical male secondary sexual character), while the rest of the broods had the normal female phenotype. While the karyotypes of the normal females were XX, those with the upper pygofer appendages were X0 and were thus feminized males. Administration of tetracycline hydrochloride to Wolbachia-infected females (phenotypically normal female) resulted in a nearly 1:1 sex ratio in the subsequent generation [28]. The incongruence between the phenotypic sex and genomic sex implies the presence of epigenetic modification. Negri et al. [29] compared the DNA methylation patterns between intersexual and normal Z. pullula by performing a methylation-sensitive random PCR and found that females with the male upper pygofer appendages showed a female methylation pattern. On the other hand, some rare feminized males bore testes instead of ovaries. And these individuals showed a male methylation pattern. These findings suggest that Wolbachia induces feminization by disrupting male imprinting.

Wolbachia can also cause feminization in non-insect arthropods. Wolbachia-induced feminization in the woodlouse Armadillidium vulgare (Isopoda; Armadillidiidae) has a long research history [30]. As known in several insect species, the A. vulgare Wolbachia appear to be quite sensitive to high temperatures. Very young Wolbachia-infected females of A. vulgare reared at 30 °C gradually acquire a male phenotype [31,32]. In addition to Wolbachia, a non-bacterial feminizing factor (f) can also force chromosomal males of A. vulgare to become phenotypic functional females. The f factor is suspected to be a genetic element derived from the Wolbachia genome that becomes inserted into the host nuclear genome [33]. Genes resisting the feminizing effects or the transmission of feminizing elements have been found in natural populations of A. vulgare [34,35], thus illustrating the conflict between the feminizing elements and the rest of the host genome.

Unlike insects, androgenic hormone, a sex hormone secreted from a specific male organ called the androgenic gland, profoundly affects male sexual differentiation in A. vulgare [36,37]. In this respect, sex determination in A. vulgare seems to be more labile compared with that in insects. It is considered that Wolbachia feminize genetic males by disrupting the secretion of androgenic hormone. Before the discovery of feminization in Eurema butterflies, Wolbachia was assumed to be incapable of inducing complete feminization in insects [31]. This assumption appeared reasonable considering the substantial differences in the sex-determining systems between insects and non-insects. The actual occurrence of Wolbachia-induced feminization in both insects and non-insects may imply the presence of some common mechanism of sex determination that is targeted by Wolbachia. Alternatively, the feminization found in insects and non-insects may have distinct underlying mechanisms and only be superficially similar.

The false spider mite, Brevipalpus phoenicis (Acarina; Tenuipalpidae), along with two closely related species, Brevipalpus obovatus and Brevipalpus californicus, is known to reproduce by thelytokous parthenogenesis [38]. Surprisingly, B. phoenicis females have a haploid genome, which was confirmed by fluorescence microscopy and variations at nine microsatellite loci [39]. Treatment of adult females with tetracycline hydrochloride led to a drastic increase in the proportion of male offspring (from 5.6% to 50.6%), suggesting that haploid individuals were feminized by a bacterium [40]. This bacterium is now known to be a member of the genus Cardinium [41]. Haploid individuals of the other species, B. obovatus and B. californicus, were also found to be feminized by Cardinium [42,43]. A feminizing effect of Cardinium was also found in the wasp Encarsia hispida (Hymenoptera; Aphelinidae), in which thelytokous parthenogenesis is known [44] (see Section 7.1).

Sex determination of the amphipod crustacean Gammarus duebeni (Amphipoda; Gammaridae) is more complex compared with that of insects. Basically, males and females are determined by a balance of a polyfactorial system of allelic sex genes located on several pairs of chromosomes, but the photoperiod can profoundly affect the sex determination [30]. Moreover, based on detailed breeding experiments, cytoplasmic factors were considered to act as female determiners in G. duebeni [45]. Subsequently, it was found that several microsporidian parasites such as Octosporea effeminans, Nosema granulosis and Dictyocoela duebenum living in the host cytoplasm alter male G. duebeni to functional females [46,47,48,49,50,51,52]. Rodgers-Gray et al. [50] demonstrated that Nosema manipulate the sex differentiation of G. duebeni by preventing androgenic gland differentiation, androgenic gland hormone production and consequently male differentiation. This is in agreement with observations of Wolbachia-induced feminization in A. vulgare. Although taxonomically unrelated (eukaryotes and prokaryotes), these feminizers may manipulate their crustacean hosts through a common mechanism.

Feminization can also be caused by non-microbes. According to Vance [53], more than 10% of wild-caught adults of the mayfly Baetis bicaudatus (Ephemeroptera; Baetidae) are parasitized by nematode Gasteromermis sp. (Nematoda; Mermithidae). None of the parasitized individuals (n = 126) contained visible eggs, ovaries or testes. Among them, the external morphology of 82 individuals (65%) was indistinguishable from that of normal females. The remaining 44 parasitized individuals showed an array of intersexual morphologies. Moreover, measurement of the DNA contents by flow cytometry suggested that the parasitized individuals having intersexual morphologies (“parasitized intersexes”) were genetically male individuals while parasitized individuals visibly indistinguishable from normal females (“parasitized females”) were composed of both genetically female and genetically male individuals [53]. The behaviors of the mayfly are also changed by the nematode. Unparasitized males form swarms near the river and do not return to the water after they have emerged. Vance [53] found that all 418 swarming individuals were unparasitized males and that the parasitized individuals showed ovipositing behavior, which is never seen in unparasitized males. Laboratory studies demonstrated that parasitized individuals (both parasitized females and parasitized intersexes) became very agitated shortly after emergence as adults. Within 3–6 hours, all the parasitized mayflies had crawled into the water down the side of the rock. The nematodes could then be seen escaping through a puncture wound in the mayfly’s abdomen [53]. Therefore, the Gasteromermis nematode can horizontally transmit to a new host. The mayflies were killed by the emergence of the nematode. The feminization in this case seems to be an adaptive strategy for the Gasteromermis nematodes, which do not have a vertical route of transmission.

Although less conspicuous than the case of the mayfly, perturbation of the development of secondary sexual characters by endosymbionts or parasites has also been found in various insects [54,55], in which the adaptive role of feminization remains unclear.

In woodlice such as Porcellilo dilatatus, Porcellio laevis and A. vulgare, an intersexual trait in genetic females is known to be transmitted from both parents. Viruses are considered to be the causal agent of the masculinization in genetic females because male characters disappeared after heat treatment, male characters appeared after injection of a 0.22-μm-filtered tissue extract and intersexuality was correlated with the presence of cytoplasmic viral particles [56]. The adaptive significance of the masculinizing effect remains unknown.

Drosophila have two X chromosomes in females and only one X chromosome in males. By hypertranscribing the X-linked genes in males, Drosophila dissolve the imbalance in the gene dosage of the X-linked genes, which is often called dosage compensation. This process requires the formation of the dosage compensation complex (DCC), which consists of MSL-1, MSL-2, MSL-3, MLE and MOF [60].

How do male killers discriminate between males and females? Using classic genetic experiments, Veneti et al. [61] tested the ability of Spiroplasma to kill D. melanogaster males carrying mutations in genes encoding the DCC. Interestingly, Spiroplasma failed to kill males lacking any of the five protein components (MSL-1, MSL-2, MSL-3, MLE and MOF) of the DCC. Therefore, although the direct mechanism of male killing is elusive, their study clearly showed that the presence of the functional DCC is necessary for Spiroplasma to cause male killing in D. melanogaster.

A mechanistic implication of male killing was obtained in our study on Wolbachia-induced male killing in the moth Ostrinia scapulalis (Lepidoptera; Crambidae), namely that males are killed by the feminizing effect of Wolbachia. Similar to other lepidopteran species, the sex chromosome constitution of O. scapulalis is ZZ in males and ZW in females. Wolbachia-infected females produce both ZZ and ZW offspring, but only ZZ offspring die during late embryonic and larval development. Interestingly, when adult females were fed with sucrose containing an antibiotic (tetracycline hydrochloride) prior to oviposition, they produced eggs that developed as intersexes as well as normal males and females [62,63,64] (Figure 3). The forewings of these intersexes showed a clear mosaic pattern of male-color and female-color sectors. Some of the intersexes had the bursa copulatrix, the female organ which accepts spermatophores from males during mating. Surprisingly, none of the tissues of the intersexes, including the bursa copulatrix, had a W chromosome, thus representing the pure male genotype (ZZ). Therefore, it is clear that the bursa copulatrix in these intersexes shows a female phenotype under a male genotype. Based on these findings, we concluded that Wolbachia has a feminizing effect on Ostrinia males but its full expression, which occurs in the natural condition, is lethal for genetic males. A reduction in the Wolbachia density during embryogenesis may result in partially feminized individuals (intersexes), probably due to the attenuated expression of the feminizing effect. The sex-determining gene dsx of these intersexes was shown to exhibit both male and female splicing patterns [65]. Reported very recently was conclusive evidence for the feminization as a mechanism of male killing in Ostrinia. In Wolbachia-infected O. scapulalis, female-specific splicing in dsx was observed in all the embryos including those with male genotype (ZZ) which were destined to die [66].

In addition, this case of male killing in Ostrinia showed another aspect when Wolbachia was completely eliminated. While other cases of male killing exhibit a normal 1:1 sex ratio after elimination of the male killers in their mothers, the elimination of Wolbachia from Ostrinia females by treatment with tetracycline hydrochloride during the whole larval stages resulted in the appearance of only males in the subsequent generation. This finding was erroneously interpreted as evidence for the Wolbachia-induced feminization [67,68]. However, later studies clearly showed that half of these offspring having female karyotype (ZW) die during late embryonic and larval development [63,64]. These individuals were recently shown to exhibit male-specific splicing in dsx [66], which implies that a female-determining factor, possibly located on W chromosome is degraded in Ostrinia and the Wolbachia substitutes the sex-determining role.

In some cases, male killing is masked or suppressed. Wolbachia-induced male killing was reported in the butterfly Hypolimnas bolina (Lepidoptera; Nymphalidae) [69,70]. In H. bolina, a strong female-biased sex ratio has been maintained for more than 100 years [71]. Surprisingly, in some populations, host resistance to male killing has recently spread and reached fixation [72,73,74]. In these populations, Wolbachia ceased to kill males but was not excluded from the host population. Hornett et al. [75] revealed that the same Wolbachia was inducing cytoplasmic incompatibility in the butterfly. This finding can explain the persistence of Wolbachia in the H. bolina population without inducing male killing.

In Drosophila recens, the sex ratios are not affected by naturally-occurring Wolbachia. This Wolbachia causes strong cytoplasmic incompatibility (see below) in D. recens and the infection frequency of Wolbachia in this species is 98%. On the other hand, none of the individuals are infected with Wolbachia in Drosophila subquinaria. Interestingly, introgression of the D. recens Wolbachia into D. subquinaria by hybridization and backcrossing resulted in the expression of male killing [76]. Furthermore, crossing experiments have demonstrated that the resistance to male killing is dominant, autosomal, multigenic and dependent on the zygotic, not maternal, genotype [76]. Similar to the case of H. bolina, male killing was masked by the fixation of resistant genes in D. recens. If there were no sister species, we would not have been able to reveal the potential for male killing in this insect.

The pea aphid Acyrthosiphon pisum is known to exhibit cyclical parthenogenesis. Previously, infection of Spiroplasma was only investigated under the asexual reproduction mode [77,78]. The effects of various endosymbionts in A. pisum under the sexual reproduction mode were recently analyzed and Spiroplasma was revealed to cause male killing [79]. The fact that Spiroplasma found in A. pisum was monophyletic with other male-killing Spiroplasma in ladybird beetles, butterflies and moths may indicate a common origin of the male-killing ability.

Why do these microbes kill males? Killing only males during early development (early male killing) is considered advantageous for maternally transmitted microbes for three possible reasons that are not mutually exclusive [80,81]: (i) females can gain extra resources that were to be allocated to their brothers (resource reallocation); (ii) mating with sibling can be avoided (inbreeding avoidance); and (iii) microbes can be transmitted from dead (or dying) males to females by oral intake (horizontal transmission by cannibalism). On the other hand, resource reallocation cannot explain late male killing because males would consume nearly all the resources necessary for their own development before they are killed. Microsporidian parasites that cause late male killing in mosquitoes have intermediate hosts (copepods). Killing the mosquito larvae of large body size (late male killing) is considered to be an optimal strategy for the microsporidians because they can maximize their number and thereby increase their probability of transmission to copepod hosts [58,82].

Thus, the classification of male-killing phenotypes is based not only on the timing of the action, but also on the evolutionary strategies adopted by the symbionts and microbial agents responsible for the phenotypes (i.e., bacteria versus eukaryotes and viruses). Therefore, it has been assumed that early and late male killing are fundamentally different phenomena whose underlying mechanisms are also entirely different [83,84]. In Drosophila, however, a Spiroplasma strain that normally cause early male killing was also found to induce late male killing depending on the maternal host age, in that male-specific mortality of larvae and pupae was more frequently observed in the offspring of young females [85]. Since the lowest Spiroplasma density and occasional male production were also associated with newly emerged females, we proposed a density-dependent hypothesis for the expression of the early and late male-killing phenotypes. Early male killing and late male killing could be regarded as alternative strategies adoptable by microbial reproductive manipulators [85].

The timing of male killing can be artificially changed in the Wolbachia-infected butterfly, H. bolina. Treatment of Wolbachia-infected adult females with tetracycline, a bacteriostatic antibiotic, produced a delay in the timing of male death. On the other hand, treatment of the surviving larvae with rifampicin, a bactericidal antibiotic, rescued the males. Based on these findings, it was hypothesized that Wolbachia possesses the ability to kill males through bacterial activity during larval development [86]. This phenomenon argues against the view that male killing is achieved by specifically targeting an early developmental process within the sex determination pathway. This is in line with the case of Wolbachia-induced feminization in the butterfly E. mandarina, wherein Wolbachia needs to be present during larval development for complete feminization [26].

There are various mechanisms underlying microbe-induced parthenogenesis. Gamete duplication appears to be common in Wolbachia-infected hymenopteran wasps [93,94]. In wasps such as Trichogramma species and Leptopilina clavipes, meiosis is normal, but during the first mitotic division, the chromosomes fail to segregate in metaphase, resulting in diploidization of the nucleus [93,94]. In Muscidifurax uniraptor, on the other hand, meiosis and the first mitotic division are normal, but during the second mitotic division, diploid females are produced by the fusion of two cell nuclei [95]. Each offspring produced by gamete duplication is a homozygote at all loci and is not a genomic copy of its mother. On the other hand, the genotype of all offspring was indistinguishable from their mothers in Rickettsia-induced parthenogenesis occurring in Neochrysocharis formosa (Hymenoptera; Eulophidae) based on the polymorphisms in a microsatellite locus [96]. Moreover, by excluding the possibility of automictic parthenogenesis with central fusion, which has been observed in some ants [97,98], and using cytogenetic observations, Adachi-Hagimori et al. [96] concluded that apomictic parthenogenesis is the underlying mechanism wherein eggs do not undergo meiosis. Similarly, in the Wolbachia-induced parthenogenesis in the mite Bryobia praetiosa, the genotypes of the mothers and offspring are indistinguishable based on the polymorphisms in three microsatellite loci [40]. Therefore, apomictic parthenogenesis is the likely mechanism for this phenomenon as Weeks et al. [40] assumed, but another possibility of automictic parthenogenesis with central fusion cannot be completely excluded owing to the lack of cytogenetic observations.

It has been naturally assumed that diploidization induced by microbes automatically leads to female development. However, a recent finding indicates that, in the wasp E. hispida, Cardinium-induced parthenogenesis induction does not occur by diploidization alone. Feeding antibiotics to infected adult E. hispida females resulted in uninfected male offspring. By karyotype observations and flow cytometry analyses, Giorgini et al. [44] demonstrated that these males were diploid. This finding indicates that at least in E. hispida, diploidy restoration is necessary but not sufficient for female development. Thus, Cardinium is required to feminize diploid male embryos [44]. In this sense, this example should also be added to the list of microbe-induced feminization. As Giorgini et al. [44] argues, it might be necessary to consider the possibility that the mechanism of Wolbachia-induced parthenogenesis is also comprised of two separate steps, i.e., diploidization and feminization.

Cytoplasmic incompatibility is the most common phenotype of Wolbachia and Cardinium infection (Table S1). In diploid organisms, cytoplasmic incompatibility is an embryonic lethality that results in sperm and eggs having different cytoplasmic contents [17,99,100]. The effect arises from changes in the gamete cells caused by cytoplasmic (intracellular) parasites like Wolbachia and Cardinium, which infect a wide range of insect species. Cytoplasmic incompatibility occurs when a Wolbachia-infected male mates with a female that is either uninfected (unidirectional cytoplasmic incompatibility) or infected by another Wolbachia strain (bidirectional cytoplasmic incompatibility). Any other combinations of crosses are compatible. An infected female is compatible with an uninfected male or any infected male of the same Wolbachia strain. On the other hand, an uninfected female is only compatible with an uninfected male.

Cytoplasmic incompatibility in haplodiploid hosts may lead to haploid (male) offspring. Cytoplasmic incompatibility produces distinct phenotypes in three closely related haplodiploid hymenopteran species of the genus Nasonia, namely mortality in Nasonia longicornis and Nasonia giraulti, and conversion to male development in Nasonia vitripennis [101].

In addition to arthropods, some nematodes are also hosts of Wolbachia [102]. In filarial nematodes, Wolbachia does not seem to exhibit selfish behaviors like reproductive manipulations. Instead, Wolbachia is necessary for the host nematodes. Elimination of Wolbachia from filarial nematodes generally results in either death or sterility of the nematodes [103]. Consequently, current strategies for the control of filarial nematode diseases include elimination of Wolbachia via treatment with the antibiotic doxycycline rather than far more toxic anti-nematode medications [102].

In the bedbug Cimex lectularius (Hemiptera; Cimicidae), Wolbachia appears to be an obligate nutritional mutualist. Wolbachia is specifically localized in the bacteriomes, a specialized organ for endosymbiotic bacteria, and vertically transmitted via the somatic stem cell niche of germalia to oocytes. The transmitted Wolbachia infects the incipient symbiotic organ at an early stage of embryogenesis. Elimination of Wolbachia results in retarded growth and sterility of the host insect. However, these deficiencies are rescued by oral supplementation of B vitamins. These findings suggest that Wolbachia provides essential nutrition for this host [104].

In the parasitic wasp Asobara tabida (Hymenoptera; Braconidae), a strain of Wolbachia is necessary for oogenesis [105]. It is also known that Wolbachia rescues the oogenesis defect in a Sex-lethal mutant of D. melanogaster [106].

Increasing attention has been paid to the recent findings that Wolbachia can protect Drosophila against pathogenic RNA viruses such as Drosophila C virus, Cricket Paralysis virus, Flock House virus, Nora virus [107,108,109] and West Nile virus [110], as well as the fungus Beauveria bassiana [111]. Interestingly, in mosquitoes, the presence of transinfected Wolbachia interferes with a wider range of pathogens including nematodes and bacteria [112], Dengue virus and Chikungunya virus [113,114], as well as the avian and rodent malaria parasites Plasmodium gallinaceum [114] and Plasmodium berghei [115]. Biological control of mosquito-borne diseases is at the testing stage [116,117,118].

Spiroplasma also has similar effects on its hosts. Drosophila neotestacea, a fly that feeds on mushrooms, suffers complete loss of fecundity when parasitized by the nematode Howardula aoronymphium (Allantonematidae; Tylenchida). However, flies infected with Spiroplasma have near normal fecundity when parasitized by the nematode [119].