Genome Evolution and the Emergence of Fruiting Body Development in Myxococcus xanthus

Background Lateral gene transfer (LGT) is thought to promote speciation in bacteria, though well-defined examples have not been put forward. Methodology/Principle Findings We examined the evolutionary history of the genes essential for a trait that defines a phylogenetic order, namely fruiting body development of the Myxococcales. Seventy-eight genes that are essential for Myxococcus xanthus development were examined for LGT. About 73% of the genes exhibit a phylogeny similar to that of the 16S rDNA gene and a codon bias consistent with other M. xanthus genes suggesting vertical transmission. About 22% have an altered codon bias and/or phylogeny suggestive of LGT. The remaining 5% are unique. Genes encoding signal production and sensory transduction were more likely to be transmitted vertically with clear examples of duplication and divergence into multigene families. Genes encoding metabolic enzymes were frequently acquired by LGT. Myxobacteria exhibit aerobic respiration unlike most of the δ Proteobacteria. M. xanthus contains a unique electron transport pathway shaped by LGT of genes for succinate dehydrogenase and three cytochrome oxidase complexes. Conclusions/Significance Fruiting body development depends on genes acquired by LGT, particularly those involved in polysaccharide production. We suggest that aerobic growth fostered innovation necessary for development by allowing myxobacteria access to a different gene pool from anaerobic members of the δ Proteobacteria. Habitat destruction and loss of species diversity could restrict the evolution of new bacterial groups by limiting the size of the prospective gene pool.


INTRODUCTION
Speciation in higher organisms usually occurs in genetic isolation. Successive rounds of gene duplication and divergence, followed by individual gene loss, is thought to have contributed to morphological diversification [1]. Speciation has not been as extensively studied in bacteria. Small gene duplication events have been noted in many sequenced bacterial genomes [2] but lateral gene transfer (LGT) also contributes to genome diversification. Bacteria exchange genes by conjugation, transformation, and transduction which are widespread in nature and can occur between distantly related organisms [3]. Genome sequencing has revealed substantial rates of foreign gene acquisition [4]. Analysis of all gene phylogenies in the sequenced members of the c Proteobacteria indicated that LGT rather than gene duplication provided most of the diversity in genomic repertoires [5].
LGT also has the potential to drive metabolic innovation. The dissemination of antibiotic resistance genes on a global scale is the paradigm, but examples of disseminated gene clusters for metabolic pathways and pathogenicity determinants also exist [6,7]. These results provide compelling evidence that LGT can influence genotype evolution at the level of sub-speciation and by extension LGT may aid bacterial speciation.
While gene duplication in M. xanthus led to the production of several large multi-gene families [8], the role of LGT in M. xanthus genome evolution has not been evaluated.
LGT is generally supported if a phylogenetic tree for a gene is in disagreement with that of the 16S rDNA gene, although incongruent trees can also arise by gene loss from related lineages. The other method to detect LGT relies on differences in the nucleotide sequence composition of a gene relative to average of that particular host. Given enough time acquired genes converge in codon usage with the bulk of the genome by amelioration [9], which make ancient transfers difficult to detect with compositional algorithms. Furthermore, LGT may be most successful when foreign genes and recipient genomes have similar codon usage [10]. Neverthe-less, genes with codon usage that differs from that of the host are excellent candidates for LGT [11,12].
Prokaryotic evolution is the product of environmental pressures combined with changes in the gene repertoire over time. The myxobacteria provide an opportunity to examine genome evolution within the context of speciation because fruiting body development is a unique trait that defines a phylogenetic order. In this work we examined the evolutionary history of M. xanthus genes required for fruiting body development using compositional and BLAST algorithms. Our results suggest that genes for metabolic enzymes, particularly those involved in polysaccharide biosynthesis, were more likely to be acquired by LGT and genes for sensory systems were more likely to be vertically inherited with numerous examples of duplication and divergence. of d Proteobacteria species are known, but for the sake of brevity, only those genera with a complete genome sequence will be mentioned here. With the exception of the myxobacteria and Bdellovibrio most of the genera, including Anaeromyxobacter, Desulfotalea, Desulfovibrio, Geobacter, Lawsonia, Pelobacter, and Syntrophus are anaerobic (use terminal electron acceptors other than O 2 ). Many are also chemolithotrophs (use inorganic compounds as energy sources) and autotrophs (derive all their carbon from CO 2 ). While the modus operandi for most genera in the d Proteobacteria is metabolic diversity, the myxobacteria are neither chemolithotrophs nor autotrophs and instead devote their resources to social cooperativity directed toward predation and the construction of a unique multicellular structure, the asexual fruiting body.
A phylogenetic approach was initially used to identify possible cases of LGT in M. xanthus. Similarity searches using the BLASTP algorithm, with a 10 210 cutoff, were performed with each putative M. xanthus protein. The top four BLASTP hits were used because the top hit does not always represent the closest homolog and more distant hits can be widely diverged from the M. xanthus query sequence. One or more of the top four hits was a d Proteobacteria for 55.0% of the genes suggesting a vertical descent as the inheritance paradigm for over half of the M. xanthus genes. Genomes also contain unique genes, in this case 22.2%, that have not been found in any other organism (so called ORFans). The remaining gene products (22.8%) have no homologs among d Proteobacteria in the top 4 BLASTP hits, but often striking homology with gene products from other organisms suggestive of an alien origin.
The same phylogenetic analysis was applied to other members of the d Proteobacteria. With the exception of Bdellovibrio, which contains 43.0% d Proteobacteria genes, other genera in the d Proteobacteria contained a higher proportion of d Proteobacteria genes, including Anaeromyxobacter (60.7%), Desulfovibrio (84.1%), Geobacter (90.2%), Lawsonia (83.6%), and Pelobacter (74.7%). Bdellovibrio is thought to have acquired an unusually large number of genes via LGT [13]. These results suggest that M. xanthus has received more alien genes by LGT than most other d Proteobacteria.

Fruiting body development
The myxobacteria represent an evolutionary branch that culminated in a unique and striking form of multicellularity. The evolutionary history of genes known to be required for fruiting body development was examined in order to glean some insight into principles guiding the evolution of this group of organisms. Only genes in which mutations diminish the capacity to form fruiting bodies or cause at least a 10-fold decrease in sporulation are included here. Many genes that affect the timing of development, but not the integrity of the end product, have been reported but were not considered here.
The majority of the genes examined in this study have been published previously and are the subject of ongoing investigations into the mechanism of fruiting body development. This group of genes was supplemented with developmental mutants identified using the mariner-based transposable element magellan-4 [14], which was used previously to identify motility genes in M. xanthus [15][16][17]. 40,000 magellan-4 insertions were screened for loss of fruiting body development. Mutant genes were backcrossed into the wild type DK1622 to confirm that the transposon causes the mutant phenotype. These genes, together with previously identified developmental genes, are given in table 1. The phylogeny of each gene was examined using the BLASTP algorithm. Genes with a codon usage that differed from that of the majority of the host genes were identified with software developed by Karlin and Mrazek [11,12].
Among the genes required for development very few exhibited a codon bias suggestive of LGT (Table 1). The exceptions include a unique gene, MXAN4486, and MXAN0581, which encodes the essential translation initiation factor InfC that is found in other d Proteobacteria. However, the M. xanthus InfC protein is unique in that it contains a 66 amino acid C-terminal extension that is essential for development but not translation [18][19][20]. The precise function of this extension remains unknown, but is thought to aid in generating an essential developmental signal, the D-signal. We were unable to identify another InfC protein with a similar extension in the database.
While the remaining genes had a normal codon bias, the tBLASTn algorithm suggested that some were acquired by LGT because their closest relatives in the database were not members of the d Proteobacteria (table 1). The contributions of LGT and vertical inheritance to the pool of genes for fruiting body development can be approximated from these data. 57 genes (73%) have a d Proteobacteria phylogeny and a normal codon bias suggestive of vertical inheritance. 17 genes (22%) have either abnormal codon bias or produce protein products whose closest relatives are from another bacterial group. The remaining 2 genes (5%) have a normal codon bias but are unique and their phylogenetic source could not be identified. These results suggest that M. xanthus fruiting body development is not possible without alien genes.
When the genes were examined within the context of their function two striking correlations were observed. Genes encoding essential developmental enzymes were often acquired by LGT (table 1, 7/14, 50%). Most notably this includes genes required for exopolysaccharide (eps) and lipopolysaccharide (rfb) biogenesis. On the other hand most of the genes required for signal production and sensory transduction have a phylogeny rooted in the d Proteobacteria and a normal codon bias (table 1, 37/41, 90%). Many of these genes have several paralogs in M. xanthus and seem to be products of duplication and divergence. These include two component systems, serine/threonine protein kinase systems, sigma 54-dependent response regulators, and chemosensory pathways [8].
The phylogenetic sources of the alien developmental genes can suggest commonalities about the pool of donors. In some cases a clear gene source does not exist. For example, the MXAN4621 and MXAN5125 proteins have similar amino acid identities with proteins from widely different phyla (table 1). Nevertheless the phylogenic origin of each alien gene in the genome was explored systematically to determine whether any trends exist. The phyla containing the top four homologs of each alien gene are plotted in figure 1 for Myxococcus, Anaeromyxobacter and Bdellovibrio. Relative to the latter two organisms, Myxococcus has an unusually high abundance of genes whose closest relatives are found in the Actinobacteria and Cyanobacteria. Most soil Actinobacteria (like Streptomyces) are obligate aerobes. Cyanobacteria also live in aerobic habitats since they generate O 2 during photosynthesis. These results suggest that M. xanthus acquired genes from organisms in aerobic habitats.

Aerobic respiration
How alien genes are functionally integrated into existing pathways was examined with a less complex pathway, the electron transport chain. The nearly unique presence of aerobic respiration among the d Proteobacteria, suggested that myxobacterial electron transport genes were derived by LGT. Examination of electron transport is also expected to reveal how recently acquired genes are integrated into functional networks. Was the entire pathway captured as a functional unit or were individual components acquired and used to replace existing components? The protein components of the M. xanthus electron transport system were identified through comparative phylogeny. The upstream elements, NADH dehydrogenase (complex I) and succinate dehydrogenase (complex II), are common to many electron transport pathways (figure 2). The principal Myxococcus quinone is a menaquinone with 8 isoprenoid units [21,22]. The d Proteobacteria along with several other phyla lack the traditional bc 1 -complex III and contain a newly described complex III containing a five-heme cytochrome c, a one-heme cytochrome c, and a three-subunit molybdopterin oxidoreductase [23]. Genes encoding three copies of this complex, potentially one for each cytochrome c oxidase complex (denoted MF1cc in figure 3) are found in M. xanthus [23]. There are two broad classes of cytochrome oxidase, which reduces O 2 to H 2 O (complex IV). Cytochrome c oxidase is coupled to the quinol:cytochrome c oxidoreductase (MF1cc) complex III. M. xanthus contains genes for three cytochrome c oxidases of this type, two coxBAC operons, and a cbb 3 operon (table 2). The quinol oxidases derive electrons directly from the quinone pool and M. xanthus contains genes for both subunits of cytochrome d quinol oxidase (cydAB, table 2).
The origin of each M. xanthus electron transport chain gene was examined using both phylogenetic and compositional approaches. In all cases the codon usage was consistent with other highly    Cytochrome oxidase genes have been subject to intense scrutiny because of their dissemination by LGT over vast phylogenetic distances [7]. The distribution of one subunit from each cytochrome oxidase was examined (bold subunits in table 2). Among the top twenty BLASTP hits to the M. xanthus cytochrome cbb 3 oxidase, Bdellovibrio is the only d Proteobacteria member (not shown) suggesting that this cbb 3 oxidase was acquired by the only two aerobic members of this group.
A neighbor-joining tree of the CydA subunit of the cytochrome d quinol oxidase revealed two versions of this operon within the d Proteobacteria (figure 3). One is found in all d Proteobacteria with the exception of M. xanthus and is presumed to be ancient and vertically transmitted. M. xanthus appears to have lost this operon sometime after its divergence from A. dehalogenans. It is surprising to find this oxidase in anaerobic organisms where O 2 is highly toxic [24]. A distinctly different cydAB operon is found in A. dehalogenans and M. xanthus and was likely incorporated by LGT prior to the divergence of the two organisms (figure 3). The M. xanthus cydAB operon falls in the same clade as diverse phylogenetic groups, most notably Actinobacteria and Planctomycetes.
The coxBAC operon is found in many members of the d Proteobacteria ( figure 4). Again, the presence of this operon is surprising for anaerobic species where the function is not known and suggests an ancient origin in the d Proteobacteria. In M. xanthus and A. dehalogenans the coxBAC operon has a different phylogenetic history from that of the other d Proteobacteria and falls in the same clade as an eclectic mixture of organisms from diverse phyla. The results are consistent with the idea that Myxococcus and Anaeromyxobacter have lost the ancient version of this operon and acquired a different one. In addition, the coxBAC operon has been duplicated in M. xanthus.
In summary, the M. xanthus electron transport chain is a chimeric pathway containing components from diverse phylogenetic sources. M. xanthus has lost ancestral cytochrome oxidase genes common to other d Proteobacteria and acquired other cytochrome oxidase genes. Acquisition of novel cytochrome oxidase and succinate dehydrogenase genes is coincident with the appearance of aerobic growth. Curiously, the only other aerobic organism in the d Proteobacteria, Bdellovibrio, has notable differences from Myxococcus suggesting that it acquired aerobic metabolism using different gene sources.

DISCUSSION
The genome of virtually every free-living bacterial species contains a contingent of genes acquired by LGT [4]. Genes acquired by LGT can provide selective advantages with regard to antibiotic resistance, carbon utilization, and habitat range [4,6,7]. Indeed, most recent changes to the metabolic network of E. coli are due to LGT rather than gene duplication [25]. We estimate that roughly a quarter of the Myxococcus genome has been acquired by LGT based on phylogeny and/or codon usage. Some alien genes came from pathways that are found in diverse phyla, for example succinate dehydrogenase and cytochrome oxidase, and their import would be expected to provide a strong selective advantage by allowing aerobic growth. Other alien genes were fashioned into a unique developmental cycle and presumably played a different role in their previous host.

Developmental innovation is correlated with a habitat shift
What event(s) facilitated the innovation of this unique developmental cycle and the divergent evolution of Myxococcales from other d Proteobacteria? We propose that the ancient myxobacterium gained access to a novel gene pool that was not available to other d Proteobacteria. A phylogenetic analysis of .220,000 proteins from the genomes of 144 prokaryotes identified specific 'highways' of lateral inheritance, some among distantly related organisms that live in similar environments [26]. The succinate dehydrogenase and cytochrome oxidase genes were clearly acquired from aerobic organisms since they mediate aerobic respiration. Over 500 M. xanthus proteins have as their closest relatives proteins from Actinobacteria and Cyanobacteria ( figure 1). This number exceeds by 10-fold the number of Actinobacteria and Cyanobacteria homologs found in Bdellovibrio. Most soil Actinobacteria, like Streptomyces, are aerobes and Cyanobacteria make their habitat aerobic by performing oxygenic photosynthesis. These factors argue that a shift to an aerobic habitat has given the Myxococcales access to a novel gene pool via LGT. Whether this habitat shift predated the emergence of fruiting body development is not clear but this issue could potentially be resolved when the genome sequences of more d Proteobacteria are completed.

Myxococcus electron transport is a chimeric pathway
Alien genes are generally integrated at the periphery of metabolic networks [25] where they could provide an immediate selective advantage. Multiple terminal electron acceptors provide flexibility in the face of ever changing environmental stresses. Of the four new M. xanthus terminal oxidase complexes, one involved the addition of a new oxidase (cco; cbb3-type cytochrome oxidase), one involved replacement of an ancient oxidase (cyd; cytochrome bd quinol oxidase), and one involved replacement of an ancient oxidase followed by duplication of the imported operon (cox; hemecopper cytochrome c oxidase). O 2 has one of the most positive reduction potentials of any terminal electron acceptor (O 2 / H 2 O = 0.82 V) so these cytochrome oxidases are located at the periphery of the electron transport chain. Succinate dehydrogenase (complex II) is located at the other end of the electron transport chain where it, along with NADH dehydrogenase (complex I), provide reducing equivalents. Succinate dehydrogenase catalyzes the interconversion of succinate and fumarate with the reduction of FAD allowing an organism to harvest the full potential of the tricarboxylic acid (TCA) cycle. Succinate dehydrogenase transfers reducing equivalents derived from the TCA cycle to the electron transport chain. In organisms with mixed respiration strategies like E. coli succinate dehydrogenase is produced only during aerobic growth [27]. The phylogenetic sources for the M. xanthus dehydrogenase and oxidases are different suggesting a piecemeal integration into the electron transport pathway.

Predation as a means of acquiring genes
In a microcosm containing 13 C labeled E. coli added to agricultural soil, the heavy carbon isotope was enriched in wild myxobacteria indicating that myxobacteria are predators [28]. Does predation  enhance the acquisition of new genes by the predator? The LGT observed with Bdellovibrio has been proposed to arise, at least in part, through the predatory capabilities of the organism [13]. Indeed, the Bdellovibrio genome is enriched in genes whose closest homolog is in the c Proteobacteria, the group containing the principal prey species [13]. Like Bdellovibrio, the M. xanthus genome is enriched in genes whose closest homolog is in the c Proteobacteria, the group containing the commonly used prey species E. coli [28]. Myxobacteria have a much broader prey range than Bdellovibrio and significantly, a broader distribution of closest gene homologs ( figure 1). Some notable differences from Bdellovibrio include genes from Actinobacteria, Acidobacteria, and Cyanobacteria (i.e. .8-fold increase), which tend to be residents of same type of soil habitats [29]. Acidobacteria are thought, on the basis of 16S gene abundance, to be one of the most prominent members of soil, though few species have been cultivated [30]. It has been recently proposed that Acidobacteria form a sister clade with the d Proteobacteria [31]. While vertical inheritance from a common ancestor could result in the high representation of Acidobacteria genes, this hypothesis does not explain why Bdellovibrio has so few Acidobacteria genes (figure 1). Organisms in the other highly represented bacterial groups are known to be excellent food sources for M. xanthus in the laboratory [32]. M. xanthus is known to acquire genes from E. coli by conjugation [33] and transduction [34] under laboratory conditions but gene transfer in the natural environment has not been experimentally demonstrated.

Conclusions and prospectus
Our results argue that genes acquired from community members influence bacterial evolution. Successful community members may foster the evolution of successful communities by sharing genetic and phenotypic innovations that promote fitness. Cooperative evolution can have strong selective advantages in nature as demonstrated by the widespread emergence of antibiotic resistance or LGT of terminal oxidases in electron transport chains. It follows then that biologically diverse habitats aid the evolution of new bacterial species by providing a larger pool of prospective genes. While studies of the biogeography of microorganisms is in its infancy, bacterial communities can be endemic to certain areas or associated with unique species of higher organisms [35]. Habitat destruction and loss of species diversity could restrict the evolution of new bacterial groups by limiting combinations of LGT.

Isolation of fruiting body deficient mutants
Myxococcus xanthus LS2208 (DfibA) cells were grown to a density of 5610 8 cells/ml in CYE broth [10 g/L Difco Casitone, 5 g/L yeast extract, 10 mM 3-(N-morpholino) propanesulfonic acid (MOPS; pH 7.6), and 4 mM MgSO 4 ]. A 3 ml aliquot was harvested by centrifugation, washed twice with 1 ml of sterile distilled water, and resuspended in 30 ml sterile distilled water. Washed cells were mixed with plasmid pMycoMar [14], and electroporated at 0.65 kV, 400 V and 25 mF [36]. The electroporation cuvette was flushed immediately with 1 ml of CYE broth to recover cells, which were then incubated with shaking at 32uC for 4 h before plating on CF agar [ [37] or generalized transduction with phage Mx4 [38].

Cloning of M. xanthus genomic DNA flanking magellan-4 insertions
To clone magellan-4 insertions, genomic DNA was isolated from vegetative cultures grown in CYE medium containing 50 mg/ml kanamycin. A 1 ml cell culture was harvested by centrifugation and resuspended in 0.2 ml of 1X PBS buffer [8 g NaCl, 0.2 g KCl, 1.44 g Na 2 PO 4 and 0.24 g KH 2 PO 4 in 1000 ml distilled H 2 O, pH 7.4]. Genomic DNA was isolated by using Invitrogen Easy-DNA kit. Genomic DNA (0.5 mg) was digested with BssHII (New England Biolabs) in a total volume of 20 ml and digestions were dialyzed on a 0.025 mm pore size filter (Millipore) against distilled water for 30 min (drop dialysis). 8 ml of this DNA was treated with T4 DNA ligase (Promega) and drop-dialyzed before electroporation into E. coli host CC118 [15]. Electroporants were recovered on LB agar containing kanamycin (50 mg/ml) after incubation at 37uC for 24 h. Plasmid DNA was sequenced with primers Mar1 or Mar2 [15].

Fruiting body formation and sporulation assay
Myxococcus xanthus cells were grown in CYE broth to about 5610 8 cells/ml. Cells were resuspended in TPM buffer [10 mM Tris HCl, pH 7.6, 1 mM KH(H 2 )PO 4 , pH 7.6, 10 mM MgSO 4 ] to a final density of 1610 10 cells/ml. 10 ml of each suspension was spotted onto TPM agar plates and incubated at 32uC. Digital images of fruiting body formation were taken every 24 h for a total of 72 h. Cells were resuspended in 1 ml of TPM buffer, heated at 50uC for 2 hours, and sonicated to kill vegetative cells. Spore production was determined by direct counts using a Petroff-Hauser chamber. Spore viability was determined by plating serial dilutions on CYE agar plates. Plates were incubated at 32uC for 5 days before counting colonies.

Genomic analyses
Phylogenetic and comparative genomic analyses were conducted with data from .300 genomes. Predicted peptides from each of eight completed d Proteobacteria genomes were obtained from NCBI in March and April 2006 including Anaeromyxobacter dehalogenans, Bdellovibrio bacteriovorus, Desulfotalea psychrophila, Desulfovibrio desulfuricans, Geobacter sulfurreducens, Lawsonia intracellularis, Myxococcus xanthus, and Pelobacter carbinolicus. Comparison of peptides was determined using the BLASTP algorithm without filtering against the non-redundant peptide database at NCBI during this time period [39]. The top ten hits for each predicted peptide were identified with an E value,10 210 . The species of each top hit and the corresponding bacterial group were identified from the BLAST algorithm output. The identity of each bacterial group used the taxonomic data provided by NCBI. Alignments of predicted peptides were generated using the Muscle algorithm [40]. Neighbor-Joining trees and tree visualization were performed using MEGA version 3.1 [41]. Tree branches with a bootstrap value,50% were condensed.