Transgenic Testing Does Not Support a Role for Additional Candidate Genes in Wolbachia Male Killing or Cytoplasmic Incompatibility

DISCUSSION

Previous studies identified genes and gene candidates for reproductive parasitism in Wolbachia and Spiroplasma bacterial endosymbionts (21, 24–26); however, continued investigations are necessary to fully resolve the genetics of reproductive parasitism for both basic and applied purposes, especially since the diversity of affected hosts is considerable (10, 47). In particular, the evolution of male killing by various microorganisms is hypothesized to be the result of convergent evolution of distinct genes due to differences in the timing of lethality (i.e., early versus late male killing) and in sex determination systems across affected hosts (XY and ZW, for example) (47). Further, wmk does not fully recapitulate the phenotype when expressed transgenically. Regarding CI, it also affects a wide variety of hosts, and several diverse phylogenetic types (e.g., types I to IV) of CI genes have been identified (24, 29), leaving open the possibility of the existence of additional types. Here, we evaluated various gene candidates for CI and male killing that were identified in recent studies on reproductive parasitism genetics (Fig. 1). The results and analyses substantiate identification of cifA, cifB, and wmk as the crucial reproductive parasitism genes or gene candidates in phage WO that have been discovered thus far, and the transgenic strains developed here provide a resource to evaluate alternative functions of the tested phage WO genes. Indeed, phage WO genes in the eukaryotic association module are enriched for predicted eukaryotic function or homology, and several of these genes likely interact with the host (22). There are also many genetic aspects of the endosymbiosis that have not been fully explored genetically. These include phage lysis and interactions with the eukaryotic and bacterial host genomes (48), survival of the Wolbachia within host cells, other reproductive parasitism phenotypes such as feminization or parthenogenesis, and altered host fecundity (49). Therefore, despite the lack of evidence for involvement in male killing or CI, there are many other possible phenotypes and functions that these transgenic strains could be used to study.

The WD0633 gene was previously investigated in relation to reproductive parasitism (37, 41, 50). The gene is associated with CI genomes (41), contains predicted ankyrin repeats that may mediate interactions with the eukaryotic host (37), encodes a putative type IV effector (see Fig. S2 in the supplemental material) predicted to be architecturally similar to the SpAID male-killing toxin (Fig. 2A; see also Fig. S1B), occurs in a mobile element like the other parasitism genes (21, 24, 25), and is directly adjacent to cifB in the EAM (22). However, it is crucially not present in the assembled genome of wRec (23) that causes both CI in its native host Drosophila recens and male killing in the sister species Drosophila subquinaria (11). Despite transgenic assays revealing no role of WD0633 in CI or male killing (Fig. 2), it is possible that, like cifA and cifB, WD0633 cannot function on its own and may require coexpression with another gene, which will be the subject of future investigations. A homolog exists in many Wolbachia genomes, including that of the divergent wBif male-killer strain of Drosophila bifasciata, but it is missing from the reduced genome of wRec (23, 25). The maintenance of the gene in divergent strains and loss in a strain with an eroded phage may suggest that the gene has an important function in phage WO biology, particularly given the likelihood of secretion and interaction with the host. However, any putative functions or strict associations of WD0633 with active phage WO remain to be assessed. In addition, the similarity between the predicted protein structures of SpAID and WD0633 (21) remains a mystery (Fig. S1B).

Although WD0633 and SpAID have no full homologs outside Wolbachia and Spiroplasma, respectively, the OTU domain within SpAID shares weak homology with domains in three Wolbachia proteins other than WD0633. Reciprocally, the domain in WD0633 does not have any non-Wolbachia homologs. Thus, it remains possible that there was an old gene transfer event of the OTU domain. The SpAID gene, however, has no known evolutionary connection to sequences within cifA, cifB, and wmk, and there are distinctions between the resulting male killing phenotypes of Spiroplasma and Wolbachia at the molecular level. For example, the presence of Spiroplasma results in neural development defects in D. melanogaster, while Wolbachia shows no such defects in D. bifasciata (51). Given these differences and the apparently unique parasitism genes and candidates, it is unlikely that a putative gene transfer from Spiroplasma to Wolbachia resulted in conferral of a parasitism phenotype.

We also tested the function of wmk sequence homologs residing in the wMel genome. Indeed, wMel contains four additional homologs (Fig. 3A), along with two (untested) partial ones that are significantly shortened by transposons and that contain only one HTH domain each (25). Not all reproductive parasitism strains contain additional homologs (wRec has only one), and some have fewer copies than wMel (wInn and wBor have three and wBif has one full and one partial). In addition, wmk is the only copy in wMel that has direct homologs in all sequenced male-killing strains based on gene synteny and sequence, making it the best candidate of all copies. In the simplest model, a single Wolbachia male-killing gene would be required to kill males. In more-complex models, some or all additional copies may together or individually result in a phenotype. As they are similar homologs and since there are multiple copies in many genomes, we tested the wMel homologs for function. WD0508 (71% sequence homology to wmk) was previously demonstrated to not kill males transgenically (25). Here, we showed that the full-length homologs WD0255 (81% sequence homology to wmk), WD0622 (65% sequence homology to wmk), and WD0623 (81% sequence homology to wmk) also did not result in a transgenic phenotype (Fig. 3B). Their functions, if any, therefore remain undetermined, but there are many possibilities. One is that they do not have a function. Since they vary in number from genome to genome, one or a small number may confer function, and the others may represent copies that have lost or not yet gained a function and may not be maintained over time. Another hypothesis is that they may need to work additively. Phenotypes may emerge or strengthen with expression of multiple copies. For example, the moderate penetrance of transgenic wmk could be enhanced by one or more of these homologs. Notably, the CI phenotype can be transgenically induced only by dual expression of the cifA and cifB genes, and neither can induce the phenotype alone (24). However, previous dual expression of wmk and the adjacent WD0625 gene did not change the phenotype (25). It is possible that, as a consequence of singly expressing most candidates in this work (including both wmk homologs and other candidates), their involvement in a phenotype was missed. However, this is unlikely with the wmk homolog candidates since wmk is the only full homolog in both the wBif and wRec male-killer genomes that has been shown to induce high levels of male killing (11). Alternatively, additional copies (and perhaps wmk as well) may have an unidentified function. Indeed, previous work showed the predicted protein structure of Wmk is similar to that of homologs in other strains despite great amino acid sequence divergence (25). In addition, wmk homologs are ubiquitous in Wolbachia genomes despite male killing being a rare phenotype. The maintenance of the gene in non-male-killing strains and the conserved predicted protein structure suggest that the Wmk protein may have a pleiotropic function beyond male killing. Indeed, Wmk may have another primary function but can serve as a genetic reservoir for development of male killing in some circumstances. These and additional hypotheses remain to be evaluated in future work.

In addition, it remains unclear why wmk has a greater number of divergent copies in some genomes than in others, but this may be important with respect to its ability to adapt to new hosts or to relaxed selection when host resistance suppresses male killing. If so, multiple copies may be correlated with the proclivity of the host for resistance or suppression in cases where male killing may positively affect the fitness of Wolbachia. Indeed, wRec and wBif each have only one full-length homolog and wInn and wBor each have three (there is no documented host resistance in wBif, wInn, and wBor), while wMel has five (and may be resistant) (53). Importantly, wBol1b, the male-killing strain of Hypolimnas bolina butterflies, contains seven different homologs. The butterfly host is known to develop resistance in many populations, and this resistance fluctuates in a population over time (17, 52). Thus, the presence of many homologs in this strain also correlates with a known tendency for the evolution of host resistance. In other cases where male killing exists with few homologs (such as wInn), it may be that Wmk targets a protein or peptide that could be lethal to the host if mutated, so resistance may be futile as has been previously hypothesized (53). Further, these three copies are in the same phage WO region with a synteny that matches many other strains, and it is therefore likely that these three copies existed before insertion into these strains by the phage (25). In contrast, wMel and wBol1b have other wmk copies not shared by most strains, so they may have uniquely arisen in these prophages. Further, some additional copies in the genomes are the simple result of multiple phage WO insertions with their own version of wmk in one Wolbachia genome. Fewer copies may therefore correlate with less host resistance, while more copies may correlate with greater resistance, which is the basis of a copy number-host resistance hypothesis. Notably, we can use the negative transgenic expression results associated with these additional copies to narrow down the range of important residues corresponding to the male death phenotype induced by Wmk (the only functional copy thus far). There are 28 residues unique to Wmk that, alone or in combination, may be important for the onset of the phenotype (Fig. 3C). While they are not clustered in any region of the protein, they may help narrow down the points of interactions with putative host targets to be assessed in the future.

Next, we functionally evaluated male-killing candidates identified in our previous comparative genomic analysis (Fig. 4) (25). These candidates were identified by looking for genes shared across the wBif, wInn, wBor, and wRec male-killer genomes, among other criteria. Seven were identified, among which were wmk (evaluated previously) and cifA (evaluated previously and in the analyses whose results are shown in Fig. 4). Five others remained to be identified (Fig. 4A). Upon transgenic expression, none induced a biased sex ratio (Fig. 4B). They therefore did not recapitulate the phenotype in this system. This does not rule out the possibility of a role in natural parasitism phenotypes conclusively, but it is unlikely they are parasitism genes. It is possible that the transgenic system is not able to fully induce male killing by these genes or that they must be expressed in conjunction with another gene. However, there is no additional evidence (such as homology to toxin-antitoxin systems, etc.) to suggest they function together with wmk or any other gene. Of the five genes, WD0627 and WD0628 may be of the most interest for further parasitism research. Their adjacent positions do not necessarily suggest that they function with wmk since they are on the opposite DNA strand, which is atypical of cotranscribed genes, although the possibility remains. They are, however, located between cifA and wmk in the EAM region and are of interest due to their general proximity with these genes.

Given the coassociations of wmk and cifA in genomics analyses and in physical relation to each other in some divergent genomes (such as wMel and wBif) (24, 25), we hypothesized that the two genes may function together. However, transgenic assays did not demonstrate additive or epistatic effects when they were coexpressed (Fig. 5). Therefore, their relationship with each other, if any, remains unsolved. There are a few possibilities, which are not all mutually exclusive. One is that their colocalization is coincidental due to a putative shared origin. Similarly, they may have close proximity (within ∼5 kb of each other) by chance and not as a consequence of a common origin. Another possibility is that the fitness of Wolbachia in a host is contingent upon expression of multiple parasitism phenotypes. If environments are rapidly fluctuating and these phenotypes are subject to pressure as a consequence of environment interactions, multiple genes may enable Wolbachia to spread in such circumstances without the cost of relaxed selection with respect to the unexpressed genes. For example, male killing may not be penetrant at high temperatures, but CI might still manifest, allowing the bacteria to proliferate under these conditions (32). Moreover, if one phenotype is ablated due to genetic mutation or host resistance, then the other provides a backup. Two or more genes expressing different phenotypes could function in bet-hedging to benefit Wolbachia in complex ecological and environmental scenarios. These or other premises remain to be tested.

As genes beyond wmk were evaluated and did not recapitulate CI or male killing, the issue remains of why transgenic wmk cannot fully induce a male-killing phenotype and why wMel does not naturally kill males in D. melanogaster. It is still possible that other genes may be involved that have not yet been discovered; however, we have now evaluated all of the top candidates identified thus far. In addition, we previously used many different transgenic drivers to test the premise that different expression levels are required for the phenotype (25). However, increasing the level of expression by an order of magnitude did not result in a change, indicating that expression levels likely do not underlie the partial phenotype (25). The problem represented by the lack of a full transgene phenotype remains unsolved, as does that represented by the lack of male killing via natural infection. One possibility that we explored was that alternative transcripts of wmk might underlie a natural lack of phenotype. Inspection of the upstream DNA identified several alternative start codons, in addition to the annotated methionine codon, that are possibly used in embryos (Fig. 6A and B; see also Fig. S3A). Transgenic expression of the most likely alternative transcript yielded no phenotype (Fig. 6C), supporting the hypothesis that alternative transcripts could underlie phenotypic differences.

There is an imperfect correlation with the finding that the male-killer genomes do not have as many alternative start codons as the CI genomes (Fig. 6A). The notable exceptions are the multipotent wRec as well as wBol1b, which infects a butterfly host of ZW sex determination and may not be affected by the same male-killer toxin as drosophilids (XY sex determination) (54). Notably, inspection of the three most commonly used prokaryotic alternative start codons in each of the genomes depicted in Fig. 6A reveals that the wBif and wBor male-killer genomes contain none of these three common codons upstream of the annotated start. Only less-common codons are present. The wInn homolog does contain a GTG codon upstream; in model prokaryotes, however, this is much more commonly used than the annotated ATA codon in this strain (45). Thus, the GTG may be the main start codon, and the ATA codon could represent a misannotation by the Glimmer gene prediction program. All others (the CI strains, the nonparasitic strain, and wBol1b) contain at least one of the top three most commonly used codons upstream of the annotated start. Therefore, of the three natural male-killers in drosophilids, it is possible that none express alternative transcripts or express a very small number. The CI-causing and nonparasitic strains, however, may express higher numbers of longer transcripts that impede function. Indeed, addition of other elements (that do not typically interfere with protein activity) such as an HA tag on either end of the protein was found to ablate the phenotype as well (Fig. S3B and C). This suggests that alternative wmk transcripts, which may appear commonly in genomes with certain start codons, may result in loss of function. This therefore represents a hypothesis underlying the natural inability of wMel to kill males in its host, as it may express nonfunctional forms of the protein. Transgenic expression, however, would not encounter this issue as the transcript has only one start codon optimized for host expression. The potential difference in the transcript lengths is one possible reason that transgenic expression of wmk results in a phenotype whereas natural expression does not.

The basis of the difference in phenotype from the transcripts is unclear but may lie in potential differences in transcriptional or translational speed affecting the amount of protein. The two phenotypes from the two transcripts could potentially result from different RNA secondary structures impeding protein translation or altering protein folding (see Fig. S3D). Indeed, experiments in Escherichia coli demonstrated that the first 5 to 10 codons in an mRNA transcript greatly determine mRNA folding at the translation start and that this region of mRNA structure is the primary determinant of the translation rate (55). Differing resulting translation rates are proposed to be the basis of selection for noncanonical start codons, as codons would be selected based on their effect on translation. Notably, several of the upstream codons are conserved across several strains, in terms of both codon sequence and location. This conservation in a noncoding region supports the notion of the sequences having putative functional importance, potentially as alternative start codons as tested here or as alternative promoters corresponding to the gene. Future work should compare and contrast the transcripts and putative promoters of various strains to further assess these hypotheses in vivo. It remains unclear why wRec would induce male killing in a sister species with the presence of a common alternative codon; however, it is notably unable to kill males in all D. subquinaria strains (11). Therefore, we do not yet understand why the transgenic phenotype is weak, but we present a new hypothesis to account for the difference between the transgenic phenotype and (lack of) native phenotype in which the transgene expresses only the transcript that leads to a sex ratio bias whereas the native strain expresses some number of nonfunctional transcripts.

Here, we evaluated and present evidence on the role of many wMel genes in CI or in male killing or both. The hypotheses according to which those other genes are involved in their tested phenotypes are not supported by the data. Notably, most of the genes were tested singly, and it is possible that they work together to induce a phenotype. Further, Wolbachia strains can induce weak versus strong phenotypes (56, 57) or CI versus male killing (11, 12), depending on different factors, including host background. Thus, testing candidate genes in other host genetic backgrounds will be an important future direction that may yield new or different results. Previously, we tested dual wmk-WD0625 expression, as those genes had the potential to be coexpressed, but this did not result in a change in phenotype (25). The only two gene candidates within this work that were anticipated to be linked were wmk and cifA due to the reasons explained above; however, future work may create new dual-expression lines of different gene combinations to determine if additional genes function in parasitism phenotypes. The results also generated several new hypotheses and analyses relating to the connection between Spiroplasma male killing and Wolbachia, the origin of multiple copies of wmk in a genome, Wmk residues critical for protein function, the correlation between wmk and cifA, and a putative transcriptional basis for some of the complexities of Wolbachia genotype and phenotype that could be tested with future work. This work not only advances our understanding of the role of phage WO genes in eukaryotic host biology but will also spur new research into the unique genetics of this symbiosis.