The Entomopathogenic Bacterial Endosymbionts Xenorhabdus and Photorhabdus: Convergent Lifestyles from Divergent Genomes

1Evolutionary biologists have long sought to distinguish the characteristics that define both convergent and divergent evolutionary history. Understanding divergence in microorganisms, such as Eubacteria, is difficult, because our concept of a bacterial species has undergone radical changes with the advent of whole-genome sequencing. However, our ability to sequence and analyze whole-genomes has begun to provide critical insights into evolutionary patterns. For example, a number of approaches have been used to determine how bacterial genomes reflect evolutionary divergence and convergence, including the exploration of phylogenetic relationships based on average amino acid identity, shared gene orthology, and correlated indel alignments. More recently, clustering analyses of protein domains for sequenced microbes have been used to identify and predict the niches of these organisms. Those organisms with a similar distribution of protein families (Pfams), but different 16S rRNA evolutionary patterns, suggest convergent evolutionary histories, while organisms with similar 16S rRNA sequences, but different niches (both environmental and functional) suggest divergent evolutionary patterns. As genomic, environmental, and functional datasets become more correlated, these distinctions become more apparent,.

2It is now clear that the composition of bacterial genomes is dynamic, and susceptible to many changes through the processes of genome reduction, gene duplication and divergence, vertical inheritance, and horizontal gene transfer, all of which occur at the confluence of multiple pressures, including the environment, mutation, and competition. While it is possible in many bacterial genomes to detect the results of these mechanisms, such as genome reduction in endosymbionts, it remains more difficult to characterize the evolutionary path of those organisms that come from similar niches and have similar phylogenetic relationships. Do they represent a single organism, or have they speciated? One example is the comparison between entomopathogenic bacteria in the genera Xenorhabdus and Photorhabdus. Both types of bacteria are mutualists with nematodes and pathogens of insects. However, genetic and physiological studies reveal that they use functionally different approaches for these roles –, suggesting that Xenorhabdus and Photorhabdus underwent divergent evolution that arrived at convergent lifestyles.

3Xenorhabdus spp. are motile, Gram-negative enterobacteria that form mutualistic associations with entomopathogenic soil nematodes in the genus Steinernema and are pathogenic towards a variety of insects –. In the nematode, Xenorhabdus spp. are carried in a specialized region of the intestine, termed the receptacle, of the third-stage infective juvenile (IJ). The IJs live in the soil until they invade the hemocoel of susceptible insect hosts. The bacteria are released in the insect hemocoel, where they overcome the insect's defense systems and produce numerous virulence factors that participate in suppressing insect immunity and killing the host –. The bacteria proliferate to high levels in the insect cadaver and produce diverse antimicrobial compounds that suppress the growth of antagonistic microorganisms –. Xenorhabdus spp. also secrete an array of exoenzymes that stimulate macromolecular degradation, the products of which, together with the bacteria themselves, are thought to provide a nutrient base for nematode growth and reproduction –. When nematode numbers become high and nutrients become limiting in the insect cadaver, nematode progeny re-associate with bacteria and differentiate into colonized, non-feeding IJs that emerge into the soil to forage for new hosts,,. Thus, the tripartite Xenorhabdus-nematode-insect interaction represents a model system in which both mutualistic and pathogenic processes can be studied in a single bacterial species.

4Photorhabdus species, like Xenorhabdus, are γ-proteobacteria that have evolved a nematode-mutualistic / insect-pathogenic lifestyle. Photorhabdus bacteria colonize the intestines of Heterorhabditis spp. nematodes, which carry them into susceptible insects that are killed and degraded for nutrients (reviewed in, ). Despite their similar lifestyles, Xenorhabdus and Photorhabdus bacteria display differences in the underlying molecular mechanisms that are used to achieve successful host interactions (reviewed in ). For example, both Xenorhabdus and Photorhabdus must be able to survive responses of the insect immune system, such as antimicrobial peptide (AMP) production, but each uses different mechanisms to overcome AMP challenge. For example, Photorhabdus uses lipopolysaccharide (LPS) modification to resist the action of the host-derived AMPs –, but X. nematophila prevents induction of insect AMP expression altogether,. In addition, screens have been conducted in both Xenorhabdus and Photorhabdus to identify mutants defective in colonizing the infective stage juvenile nematode. Thus far, no overlap in genetic determinants required for colonization has been observed between the two genera –. These molecular and genetic differences are underscored by morphological and life-style differences between the two systems. For example, Xenorhabdus and Photorhabdus bacteria are carried by the infective juvenile stage in different locations: in a unique Steinernema spp. intestinal structure called the receptacle, or an extended region of the anterior intestine of Heterorhabditis spp., respectively. Further, the transmission of P. luminescens to H. bacteriophora infective juvenile progeny requires bacterial colonization of maternal rectal glands and hatching of the progeny within the mother (endotokia matricida). No such rectal gland colonization has been observed in S. carpocapsae nematodes, nor is endotokia matricida essential for IJ colonization (Chaston and Goodrich-Blair, unpubl. data).

5To better understand the biology of Xenorhabdus, we sequenced the complete genomes of X. nematophila ATCC 19061 and X. bovienii SS-2004,. Comparison of these Xenorhabdus genomes to the sequenced genomes of Photorhabdus luminescens subsp. laumondii TT01 and P. asymbiotica ATCC 43949 provides evidence for genomic divergence between these two genera even though they share similar lifestyles. Our analysis of these two Xenorhabdus genomes provides insight into the complex lifestyle of these nematode symbionts, is of interest for understanding bacterially mediated insect infections, and is a resource for using these entomopathogens as biocontrol agents of agriculturally-relevant insect pests.

6Both Xenorhabdus and Photorhabdus employ similar mechanisms to complete their lifecycle. Their ability to associate with entomopathogenic nematodes is a key driver in their evolution and likely shaped their respective genomes. Below, we compare their genomes, and illustrate the differences that reflect their genomic divergence despite convergent lifestyles.

7The Xenorhabdus nematophila ATCC 19061 and Xenorhabdus bovienii SS-2004 genomes are circular and composed of 4,432,590 and 4,225,498 bp, respectively (Figure 1). The X. nematophila genome contains 7 ribosomal RNA operons, encodes 79 tRNA genes, has an average GC content of 44.2%, and is predicted to have 4,299 protein-coding open reading frames (Table 1). X. nematophila also contains an extrachromosomal element of 155,327 bp, containing 175 predicted protein-coding open reading frames (Figure 1 and Table 1). The X. bovienii genome contains 7 ribosomal RNA operons, encodes for 83 tRNA genes, has an average GC content of 45% and is predicted to contain 4,260 protein coding regions (Figure 1 and Table 1).

8We performed a number of genomic analyses on these two genomes including their metabolism (Text S1), transposases (Text S2), secretion systems (Text S3), small RNAs (Text S4), Tc toxins and hemolysins (Text S5), and secondary metabolites (Text S6). We also performed a detailed proteomic analysis of secreted proteins in X. nematophila, which we describe in Text S7, and note that a detailed analysis of regions of genome plasticity was performed previously for these two bacteria.

9Xenorhabdus and Photorhabdus are more closely related to each other than to any other known species. Members of these genera are known to associate with specific nematode genera and no cross-associations are known. Specifically, Xenorhabdus bacteria are found associated with Steinernema nematodes whereas Photorhabdus bacteria are found associated with Heterorhabditis nematodes.

10To confirm the phylogenetic divergence of this association with current data, we constructed two phylogenies for the bacteria and nematodes as shown in Figure 2. We first built a 16S rRNA phylogeny that included both Xenorhabdus species in our study and two Photorhabdus species, Photorhabdus luminescens subsp. laumondii TT01 and P. asymbiotica ATCC 43949. This tree shows the close phylogenetic relationship between the Xenorhabdus and Photorhabdus and their placement within the Enterobacteriaceae, relative to other bacteria in the Proteobacteria. This 16S rRNA phylogeny was further confirmed by a multi-locus sequence analysis (Text S8). In contrast, a phylogeny based on the 18S inter-ribosomal sequence of nematodes shows that the nematode hosts of Xenorhabdus and Photorhabdus are not closely related (Figure 2). Specifically, Xenorhabdus species are phylogenetically closer to Photorhabdus than their respective hosts, Steinernema and Heterorhabditis, are to each other even though both nematodes belong to the order Rhabditida.

11Despite the relatively close relationship between these Xenorhabdus/Photorhabdus lineages (their 16S rRNA genes are over 94% identical), each of these genomes has been disrupted by numerous insertions, deletions, inversions and translocations. An orthology analysis comparing the coding sequences of all four genomes reveals a total of 2,313 shared sequences, with each Xenorhabdus genome containing close to 1,000 species-unique genes (Figure 3). Our analysis also reveals that the two Xenorhabdus and Photorhabdus genomes share more genes exclusive with each other (409 and 893, respectively) than between Xenorhabdus-Photorhabdus pairs (62 genes for X. nematophila and P. luminescens; 76 for X. nematophila and P. asymbiotica; 155 for X. bovienii and P. luminescens; 170 for X. bovienii and P. asymbiotica). We also performed a genomic similarity analysis between each pair of genomes using both average nucleotide identity and tetranucleotide frequencies as shown in Figure S1. We found that for all of these similarity metrics, the Xenorhabdus genomes are more similar to each other than to the Photorhabdus genomes or to other closely related bacteria like Yersinia pestis CO92 and Proteus mirabilis HI4320. We found the same trend for the Photorhabdus genomes, which are more similar to each other than to the Xenorhabdus genomes, Y. pestis, or P. mirabilis.

12Further, comparisons of the positions of orthologous genes in these genomes reveals extensive rearrangements in each genome and yields the characteristic X-shaped alignments (data not shown) apparent when inversions encompass and are symmetric to the replication origin,. The synteny between the two Xenorhabdus genomes is also more highly conserved in the first half of the chromosome; however a large inversion spanning nearly 400 kb has occurred within this region in the X. bovienii genome. Although the Xenorhabdus genomes harbor large numbers of IS elements, there is no apparent relationship between the number and location of these translocatable elements and the occurrence of genome rearrangements.

13To begin unraveling the metabolic and physiological differences that may exist among these bacterial entomopathogens, we constructed phylogenomic maps for all four Xenorhabdus and Photorhabdus genomes (Figure S2). Phylogenomics posits that those ORFs sharing a similar evolutionary history will cluster into functional modules corresponding to different aspects of the organism's lifestyle. Construction of a phylogenomic map proceeds by comparing each predicted protein in a genome against a database of predicted proteins from all other completely sequenced genomes. A phylogenetic profile for each protein is thus generated with each cell containing the bit score of the best BLAST hit to a protein in a given microbial genome. These profiles are then clustered to generate a similarity matrix and further visualized as a topographical landscape of mountains where each mountain contains groups of proteins that share phylogenetic history and potentially correspond to putative functional modules (Datasets S1, S2, S3, S4). Overall, we found that all four maps had comparable topography with the X. nematophila and X. bovienii maps more similar to each other than the P. luminescens and P. asymbiotica maps (Figure S2).

14We then annotated these mountains by performing a gene ontology enrichment analysis to determine if individual mountains contained genes associated with a particular function as shown in Tables S1, S2, S3, S4. In general, we found that the mountains across all four maps reflect the general lifestyle of these bacteria, as mountains enriched for genes associated with transcription and translation; metabolism; energy production and conversion; motility and chemotaxis; and transport were detected. We also found that there were a number of functional modules exclusive to either Xenorhabdus or Photorhabdus bacteria. Between the Xenorhabdus we found mountains over-enriched for genes associated with stress response (GO:0006950) and nuclease activity (GO:0004518). Between the Photorhabdus, the most striking over-enriched functional modules are those associated with pathogenesis (GO:0009405), symbiosis, encompassing mutualism through parasitism (GO:0044403), and interspecies interaction between organisms (GO:0044419). An analysis of these two mountains (mountain 35 in P. luminescens, Table S3; and mountain 7 in P. asymbiotica, Table S4) reveals that they contain a large number of type III secretion system proteins, which are known to be important during insect colonization by the Photorhabdus-Heterorhabditis pair. Since neither Xenorhabdus species is known to contain genes encoding for type III secretion (Text S3), it is not surprising that mountains enriched for this known gene ontology designation do not exist.

15To gain predictive insights into genetic components that represent divergent and convergent approaches to insect and nematode host-association, we performed an additional phylogenomic clustering analysis of genes specific to either to the genus Xenorhabdus alone (class X) or to both Xenorhabdus and Photorhabdus (class XP). Genes in class XP were generated by retaining only those homologs found between the Xenorhabdus and Photorhabdus genomes but not in Salmonella typhimurium LT2 or Escherichia coli K12. We reasoned that S. typhimurium LT2 and E. coli K12 are reasonable representations of the genetic content within the Enterobacteriaceae and by filtering the Xenorhabdus and Photorhabdus gene sets against these two genomes, we would potentially identify those genes specific to these two genera. A total of 243 genes were identified in this manner, and subsequent phylogenomic mapping analysis revealed a map with 9 mountains (Table S5 and Dataset S5). Similarly, we constructed a phylogenomic map for the 290 orthologs found between X. bovienii and X. nematophila but not in the Photorhabdus genomes, S. typhimurium LT2, or E. coli K12. This resulted in a phylogenomic map with 15 mountains (Table S6 and Dataset S6). We report our following analysis using X. nematophila gene locus names.

16One of the strengths of phylogenomic mapping is that every gene on the map is clustered according to a phylogenetic profile that determines in what other bacteria homologs of that gene are present. As a result, additional inferences for a gene can be determined by correlating it to known information about those bacteria that define its phylogenetic profile. We used this approach to analyze the genes on both of these maps by tabulating the known environmental and taxonomic associations of each bacterium that comprises each gene's phylogenetic profile. Given that both Xenorhabdus and Photorhabdus are host-associated bacteria, we expect that those mountains enriched for genes found in other host-associated bacteria could infer factors necessary for insect or nematode interactions. As a result, we obtained the organismal information provided for each microbe in the complete microbial genome collection in NCBI and used this to categorize each microbe as either host-associated or unknown- / not- host-associated (Dataset S7). A given bacterial species was scored as host-associated if it is found in association with plants, animals, or protozoans as a pathogen, mutualist, or “commensal”.

17In general, we found several mountains in each of the X and XP classes that were significantly enriched for genes carried by bacteria that are either host-associated or not host-associated (Table 2) relative to all X. nematophila genes. Proteins encoded by the XP class could be necessary for conserved responses to selective pressures encountered in insect hosts or common between Steinernema spp. and Heterorhabditis spp. host environments. On the other hand, X class proteins are expected to be involved in Xenorhabdus-specific responses to Steinernema nematode environments and the insects they infect. These proteins could either represent a convergent response to similar host pressures or divergent responses to unique host habitats. We further determined that for most mountains enriched in genes with homologs in host-associated bacteria, those bacteria are significantly over-represented for γ-proteobacteria. This suggests the possibility that these host-association genes might partition by vertical inheritance.

18An analysis of the XP class phylogenomic map revealed six mountains that were over-represented for genes from host-associated bacteria (Table 2). These mountains contain genes encoding toxins and proteases (mountains XP1, XP4, XP7, and XP8; Table S5), putative membrane transporters including iron and iron-related acquisition transport systems (XP4, XP7, XP8), transcriptional regulators (XP1, XP4, XP7, and XP8), and toxin/antitoxin members or modules (XP1, XP7, and XP10). Many of these genes are well-known in the Xenorhabdus and Photorhabdus lifestyle, including the toxins, which are used to kill their respective insects (e.g. Tc Toxin, (Text S5), XaxAB Toxin,,, and XhlAB hemolysin,. As a result, these shared sets of genes likely represent important factors common between the two genera that may help in stabilizing the nematode-bacteria mutualism in general.

19In addition to those mountains enriched for genes from host-associated bacteria, we found other mountains that may also play potential roles in Xenorhabdus and Photorhabdus interactions with nematodes or insects. For example, mountain XP14 (Table S5) contains members of the Wal lipoolysaccharide (LPS) biosynthesis locus, one of which is induced during X. koppenhoeferi infection of the white grub Rhizotrogus majalis and three other known LPS biosynthesis genes (XNC1_1391, XNC1_2486, and XNC1_2487) that are necessary for both nematode mutualism and pathogenesis in P. luminescens. The presence of LPS biosynthesis genes with XP-class genes fits with the current understanding that bacterial LPS plays a key role in both pathogenic and mutualistic associations –.

20In the X-class phylogenomic map five of the nine mountains contained genes associated with genetic mobility (e.g. transposases) while four did not (Table 2, Table S6). The latter group includes mountain X2, which contains 4 tellurite resistance genes; X5, a relatively large mountain that contains 3 of the known 14 xenocoumacin production genes involved in maintaining cadaver sterility; X6, which contains predominantly genes encoding proteins of unknown function; and X9, which contains 3 groups of 5 phage-encoded genes, each group containing a putative holin protein-encoding gene (identified by manual inspection). Given that this map is specific to only orthologs between the two Xenorhabdus genomes, it is likely that the genes clustered within these mountains, such as those hypothetical proteins in mountain X6, are specific to Xenorhabdus biology.

21The complex association of Xenorhabdus and Photorhabdus with nematodes is a beautiful example of host-microbe symbioses. In this paper, we report the complete sequencing of the X. nematophila ATCC 19061 and X. bovienii SS-2004 genomes. Our analysis reveals that Xenorhabdus bacteria can produce a large arsenal of insecticidal toxins, commensurate with their known entomopathogenic lifestyle. Our comparative analysis of Xenorhabdus and Photorhabdus genome provides insight into how their relationships with different nematodes have shaped their evolutionary history.

22Xenorhabdus and Photorhabdus are more phylogenetically similar to each other than their nematode hosts (Figure 2), suggesting that both Xenorhabdus and Photorhabdus diverged more recently from a common ancestor. This bacterial progenitor may have been capable of colonizing both Steinernema and Heterorhabditis, and long-term association with their host may have independently given rise to Xenorhabdus and Photorhabdus. The delineation of these two genera is marked by the fact that each genus can only colonize specific nematode hosts. Importantly, Xenorhabdus and Photorhabdus are not the only bacteria known to engage in pathogenic symbioses with nematodes. For example, the γ-proteobacterium Moraxella osloensis can associate with the nematode Phasmarhabditis hermaphrodita and parasitize slugs. M. osloensis (family Pseudomonadaceae) is phylogenetically distinct from either Xenorhabdus or Photorhabdus (family Enterobacteriaceae). Since P. hemaphrodita belongs to the same order as both Steinernema and Heterorhabditis (Rhabditida), this suggests that γ-proteobacteria have a long association as nematode symbionts. As a result, it is entirely possible that a progenitor of Xenorhabdus and Photorhabdus differentiated from a more ancient predecessor before associating with their respective nematode hosts. Further divergence would be expected to result in mechanisms that maintain specificity with their respective nematode hosts.

23Under this model, these genomes would partition into genus- or species-specific genes that help maintain their specificity and shared homologs that are general to their similar lifestyles. This is supported by our findings that these bacteria share 3,299 orthologs between any Xenorhabdus-Photorhabdus combination, representing at least 70% of the predicted coding sequences in each genome (Figure 3). As each genus diverged, the number of shared orthologs between genera would be expected to decrease while the number of genera-specific orthologs would increase. This is also supported in our analysis, as we found 2,313 orthologs shared between all four bacteria, representing less than half of the predicted coding sequences in each of their respective genomes. Furthermore, each Xenorhabdus and Photorhabdus pair share more orthologs exclusive to each other than to any Xenorhabdus-Photorhabdus pair (Figure 3). These differences are also underscored by our whole-genome average nucleotide identity and tetranucleotide usage analyses, which show that each Xenorhabdus and Photorhabdus pair is more similar to each other than to any other combination (Figure S1).

24The divergence of Xenorhabdus and Photorhabdus raises two questions: which genes are conserved and which genes have diverged? Our phylogenomic mapping analysis revealed many core physiological pathways are highly conserved across all four bacteria (Table 2 and Tables S1, S2, S3, S4, S5, S6). These include genes encoding toxins and proteases, iron-related transporters, and LPS biosynthesis. Some of these genes, like the toxins, are well known for their interactions with insects, and these toxins have likely been retained across the evolutionary history of these two genera to help maintain the mutualistic relationship they have with their respective nematode hosts (Text S5). Other genes like the LPS biosynthesis cluster are also known for both nematode mutualism and pathogenesis in P. luminescens, and our finding of homologs in P. asymbiotica and the two Xenorhabdus bacteria may indicate their similar role in these three bacteria.

25In addition to these shared homologs, we also identified putative insect-environment responsive elements in our XP class phylogenomic map. Two of these genes, XNC1_2015 and XNC1_2125 (mountain XP1, Table S5), putatively encode galactophilic lectins and a search of other bacterial genomes revealed homologs in only three other bacteria: Enterobacter cloacae, Ralstonia solanacearum, and Pseudomonas aeruginosa. One of these, the PA-IL lectin, mediates P. aeruginosa adherence to a galactose epitope on the surface of epithelial cells, and fibronectin. Similarly, the galactophilic lectin homologs of Xenorhabdus and Photorhabdus may mediate specific adherence to insect or nematode host tissues. One particularly relevant target is insect blood cells (hemocytes), and indeed, Drosophila melanogaster hemocytes express a galactose-containing antigen. Therefore, it is plausible that both Xenorhabdus and Photorhabdus utilize galactophilic-lectin homologs to adhere to insect hemocytes.

26One set of genes revealed in our analysis has likely duplicated and diverged in these two genera. We found that the putative virulence determinants known as invasins have a core set of highly conserved genes found in all four genomes in addition to other invasion genes that are specific to either Xenorhabdus or Photorhabdus (Text S5). The Xenorhabdus invasion proteins are characterized by a domain of unknown function (DUF) domain, whereas the Photorhabdus invasins contain Ig-like domains that are related to those found in E. coli and Yersinia. In Yersinia, these proteins are known to play a role in uptake by their hosts, and it is entirely possible that these genes function in a similar manner in Photorhabdus. Given that both Xenorhabdus and Photorhabdus interact with nematode and insect hosts, these genes may play similar roles and their divergence could be linked to the specificity of their known hosts.

27Convergent pathways are also present in our analysis. For example, our phylogenomic mapping analysis confirmed previous observations that Photorhabdus genomes contain type III secretion system (T3SS) genes that are absent in both Xenorhabdus genomes. The T3SS system is necessary for insect colonization by Photorhabdus, which uses it to secrete its numerous toxins and insect-killing factors,. The presence of this pathway in Photorhabdus is likely preserved within the Enterobacteriaceae, as many closely related bacterial pathogens like Yersinia also use the T3SS to deliver toxins. This would suggest that Xenorhabdus lost these genes as it diverged rather than Photorhabdus acquiring this system horizontally. In Xenorhabdus, delivery of toxins into the insect is not precisely known; however, possible mechanisms include two-partner secretion systems (e.g. XhlAB, Text S5), the flagellar apparatus,, or outer membrane vesicles (Text S7). As a result, Xenorhabdus and Photorhabdus have converged upon parallel strategies for toxin delivery using wholly different mechanisms.

28We also found differences in the way that Xenorhabdus and Photorhabdus overcome oxidative stress. Oxidative stress resistance is important for insect pathogenesis – and has been implicated in both Xenorhabdus, and Photorhabdus, nematode host interactions. Our phylogenomic analysis of Xenorhabdus orthologs revealed a number of genes predicted to confer tellurite resistance (mountain X2, Table S6), a mechanism known to be involved in combating oxidative stress. In contrast, Photorhabdus does not have tellurite resistance genes, and may use other mechanisms to respond to reactive oxygen stress, including catalase and luciferase enzymes, and the autoinducer-2 pathway. As a result, these two genera may have converged upon ways to overcome oxidative stress using entirely divergent pathways.

29A third example of convergence is sterile cadaver maintenance. In addition to selective colonization events that help ensure that the proper symbiont is passed to progeny nematodes (e.g. ), both Xenorhabdus and Photorhabdus produce compounds that prevent other bacteria from thriving within an infected insect cadaver. These products include antibiotics, such as the Xenorhabdus-specific xenocoumacins, and a wide variety of other small molecules produced by Photorhabdus. Importantly, xenocoumacin is the major antibiotic class produced by X. nematophila,, and genes within the cluster encoding this compound were identified in our Xenorhabdus-specific phylogenomic mapping analysis (mountain X5, Table S6). While xenocoumacin production is not known in X. bovienii, several biosynthesis gene clusters have also been identified in X. bovienii and three orthologs of the X. nematophila xenocoumacin-producing genes are present in mountain X5 (xcnADE) (though the postulated natural product resulting from the xcnADE biosynthesis gene cluster is expected to be structurally dissimilar from xenocoumacin (Bode, unpublished data)). This suggests that X. nematophila and X. bovienii may use different variations of the same molecular mechanism for antibiotic production. In contrast, Photorhabdus do not produce Xenocoumacin class antibiotics but produces the antibiotic isoproylstilbene instead and utilizes bacteriocins called lumicins to prevent other bacteria from thriving within the insect cadaver ). Functional assays show that Xenorhabdus genomes encode factors that kill closely related Xenorhabdus species – but these genes have no sequence similarity to the lumicin-producing genes in Photorhabdus with the exception of two X. bovienii genes (XBJ1_1085 and XBJ1_1080) that are similar to the P. luminescens Usp-like/catalytic domain/typO873-like DNAse/RNase components (plu1894, plu0884, and plu4177). Thus, the mechanisms Xenorhabdus and Photorhabdus use in secretion mechanisms, response to oxidative stress, and maintenance of a sterile insect cadaver all represent convergent approaches to help maintain their similar lifestyles.

30Recently, there has been much discussion among microbiologists regarding what constitutes a bacterial species,. Molecular characteristics, such as rRNA sequencing and DNA-DNA hybridization have been used to classify bacteria, but in many cases these are too highly conserved across species to be useful as classification tools. One definition of a species includes the niche of living in another organism. This ecotype model recognizes the special relationship between genes and the environment – and has been proposed to explain bacterial species evolution. One of the basic tenets of this model is that a common bacterial ancestry will be retained among bacterial populations residing within ecological niches. Our analysis of the Xenorhabdus and Photorhabdus genomes here appear to support this ecotype model of speciation. Selective pressures induced by bacterial-nematode interactions would result in periodic selection,, which is expected to give rise to genomic changes that ensure the specificity, stability, and maintenance of this symbiosis. Clearly, the bacterial-nematode lifestyle is successful, given its continued existence and expansion in other pairings that parasitize other organisms like slugs. Our findings support the hypothesis that Xenorhabdus and Photorhabdus diverged from a common ancestor, and, due to the selective pressures of maintaining this symbiosis, evolved different mechanisms to converge upon the same lifestyle.

31X. nematophila ATCC 19061 used in this study was acquired from American Type Culture Collection. The X. bovienii strain used in this study was deposited on Jun. 28, 2000 with the Agriculture Research Culture Collection (NRRL) International Depository Authority at 1815 North University Street, in Peoria, Ill. 61604 U.S.A., according to the Budapest Treaty on the International Recognition of the Deposit of Microorganisms for the Purpose of Patent Procedures and was designated as NRRL-30311.

32Cultures of X. nematophila ATCC 19061 and X. bovienii SS-2004 were grown on LB agar supplemented with ampicillin (150 ug/ml). Cells were scraped into 10 ml of LB broth and 6 mL were subcultured into 500 mL LB for 18.5 hours at 30°C. Four 35-ml aliquots were treated in the following manner. Cells were pelleted and resuspended in 15 mL TE buffer (10 mM tris, pH 8.0; 1 mM EDTA, pH 8.0) prior to adding 4 mg proteinase K and 0.66% final concentration of SDS. The solution was incubated at 48–55°C for 2 h prior to extracting twice with 1∶1 phenol-chloroform solution and twice more in chloroform to remove residual phenol. The samples were precipitated by addition of 0.1 volumes of 3 M sodium acetate and 0.6 volumes of isopropanol, and then frozen overnight at −80°C. Samples were then spun for 10 min at 10,000 RPM, washed in cold 70% ethanol and air dried. All four aliquots were resuspended in 15 mL of TE at 4°C overnight, and gently pipetted to aid in resuspension. RNase was added to a final concentration of 25 µg /mL and the samples were incubated at 37°C for 1 h. Samples were then extracted once in one volume of phenol, once in one volume of 1∶1 phenol-chloroform, twice in equal volumes of chloroform, and precipitated with 0.1 volumes of sodium acetate and 2 volumes of 95% ethanol. Pellets were washed in 40 mL of 70% ethanol, dried completely, and re-suspended in 5 ml TE buffer at 4 C overnight and gentle pipetting to a final concentration of 0.64 mg/ml with an OD260/OD280 ratio of 1.9. A single high molecular weight band was visible by gel electrophoresis.

33A total of 90,000 reads were generated for each genome. These reads came from two DNA libraries (insert sizes 2–4 kbp and 4–8 kbp) prepared using mechanical shearing of DNA and cloning into pUC18, followed by a shotgun sequencing approach. The genome was then assembled and edited using the Phred/Phrap/Consed software package –. Finishing was completed by generating an optical map, as previously described, cut with the restriction enzymes AflIII and EagI and aligning the assembled sequences to the map. Gaps were closed by sequencing specific products. All rRNA operons were amplified with specific flanking primers, sequenced and assembled individually. All positions with Phred scores less than 40 were re-sequenced using an independent PCR fragment as template. The error rate is estimated to be less than 1∶10,000 bp.

34Genome annotation was performed using the MaGe annotation pipeline as previously described, and all subsequent genomic analyses presented in this study were performed using this annotation. The sequence and annotation for the X. nematophila chromosome, plasmid, and X. bovienii chromosome are deposited in GenBank under accession numbers FN667742, FN667743, and FN667741, respectively.

35A representative phylogeny of the phylum Proteobacteria was constructed using the 16S rRNA sequences obtained from completely sequenced genomes as shown in Figure 2. The 16S rRNA sequence from the complete genome sequence of the Actinobacteria Streptomyces coelicolor A3(2) was used as an outgroup. Particular focus on members of the γ-proteobacteria, especially those Enterobacteriaceae known to be closely related to Xenorhabdus and Photorhabdus bacteria, were included. All 16S rRNA sequences were first aligned using MUSCLE and the resulting alignment analyzed using the phylogenetic analysis program phylip. A maximum likelihood tree was generated using this approach and bootstraps were also calculated (100 replicates).

36A representative phylogeny of nematodes was also constructed in a similar manner. All 18S rRNA sequences for nematodes used in a previous study by Blaxter et al. were obtained and aligned using MUSCLE. The resulting alignment was also analyzed using phylip, and a maximum likelihood tree was generated with bootstraps (100 replicates).

37An orthology analysis was performed for X. bovienii, X. nematophila, Photorhabdus asymbiotica, and P. luminescens as shown in Figure 3. This analysis was performed using the Phyloprofile Exploration tool in the MicroScope Microbial Genome Annotation Platform. All four genomes were compared against each other using a minLrap ≥0.6 and identity ≥30% and these data were used to determine all possible ortholog combinations as shown in Figure 3.

38To determine the similarity of all four genomes at the species level, we calculated three different metrics including average nucleotide identity using BLAST (ANIb), average nucleotide identity using MUMmer (ANIm), and tetranucleotide frequencies using the program JSpecies. All analyses were performed using standard parameters as shown in Figure S1.

39Phylogenomic maps were constructed for X. bovienii, X. nematophila, Photorhabdus asymbiotica, and P. luminescens as previously described. Briefly, BLASTP was used to align each predicted protein in each genome against a local database of predicted proteins from 1,173 sequenced bacteria obtained from the National Center for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi, accessed: 09/10/2010). We retained results for each protein match that registered a bit scores >50 and an e-value <1×e-5 in 5 or more of the sequenced genomes. A raw data matrix of bit scores was constructed with each row representing a protein and each column corresponding to a different sequenced genome. Correlations for each pair of proteins were calculated using Spearman's rank correlation. For each protein, the top 50 correlates that had the highest positive correlation scores were retained. Each protein was then assigned an (x, y) coordinate in the plane using a combination of force-directed placement and multi-dimensional scaling. These proteins were then visualized as a three-dimensional topographical map using the computer program VxInsight (Figure S2 and Dataset S1, S2, S3, S4). Each mountain on the map represents those proteins that share similar phylogenetic history, and the height of each mountain is proportional to the density of the proteins within that area. Mountains were determined for each map and gene ontology (GO) assignments were generated for each protein in each mountain using Interpro. GO database files were constructed for each genome, and used to determine the enrichment of GO terms for each mountain using the GO::TermFinder software package. The top three enriched GO terms for each mountain on each map were retained and shown in Tables S1–S4.

40A phylogenomic map was also constructed using only those proteins orthologous between the Xenorhabdus genomes but not in the Photorhabdus genomes, Salmonella typhimurium LT2, and Escherichia coli K12 using the same method described above. Orthologs specific to the Xenorhabdus were determined using the Phyloprofile Exploration tool in the MicroScope Microbial Genome Annotation Platform with an minLrap ≥0.6 and identity ≥30%. A second ortholog-specific phylogenomic map was also constructed using only those proteins orthologous within the Xenorhabdus and Photorhabdus genomes but not in Salmonella typhimurium LT2, and Escherichia coli K12. In all cases, the X. nematophila proteins corresponding to orthologs were used to construct each respective phylogenomic map (Table S5–S6 and Datasets S5–S6).

41Host-association analyses for each mountain on these two phylogenomic maps were performed by identifying the set of microbes for which each protein in the mountain had a significant BLAST hit. The host-association information for each microbe was then determined using the Organism Information data found associated with the complete microbial genome collection in the NCBI (http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi, accessed: 10/25/2010), the Integrated Microbial Genomes Systems (http://img.jgi.doe.gov/cgi-bin/w/main.cgi, accessed: 02/22/2011), and ExPasy (http://au.expasy.org/sprot/hamap/interactions.html, accessed: 02/22/2011). These were further supplemented with our own annotations for those organisms with no host-association entries (Dataset S7). Each species was labeled as “host-associated” or “not host-associated” based on if it was normally found in association with a plant, animal, or protozoan host, as any of a pathogen, mutualist, or “commensal”. Group data were taken directly from NCBI (same as previous). For each mountain, the category for each organism that carried the corresponding protein in the mountain was tabulated. These data were then compared to the same data generated for the entire X. nematophila genome using Fisher's Exact Test to determine if proteins within any given mountain were either over- or under-represented for a particular host-association (Table 2). A similar approach was used to determine the over- and under-representation of proteins within each mountain according to taxonomic identity as shown in Table 2.