Polynucleobacter necessarius, a model for genome reduction in both free-living and symbiotic bacteria

General Features of the Genome.

The circular chromosome (Fig. S1) of the symbiotic P. necessarius strain is 1.56 Mbp long and contains approximately 1,279 protein-coding genes (SI Results and Discussion, Genome Composition Analysis and Table S1). The reduction in genome size that has occurred since the establishment of symbiosis can be estimated by comparing this genome with that of the P. necessarius free-living strain QLW-P1DMWA-1 (see also Fig. S2), an acceptable procedure given that the symbiont’s gene set is largely a subset of that of the free-living strain (SI Results and Discussion, Genome Composition Analysis and Fig. S3). This reduction is more apparent as a decrease in the amount of coding DNA (42.3%) than in genome size (27.7%) because of the massive number of pseudogenes in the symbiont (Fig. 1). There are virtually no genes related to mobile elements and extremely few recently duplicated genes in the symbiotic isolate. Horizontally transmitted genes, mostly originated before the split between the two strains, are present (SI Results and Discussion, Genome Composition Analysis), but there is no sign of DNA exchange between the symbiont and its host Euplotes, nor any other eukaryote. A total of 277 genes in the symbiont genome (105 of which are not shared with the free-living relative) have unknown functions (SI Results and Discussion, Genome Composition Analysis).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Percentage allocation of coding and noncoding DNA in the free-living and symbiotic Polynucleobacter genomes. The graph areas are proportional to genome sizes (2.16 and 1.56 Mbp, respectively). Pseudogenes were identified with a more conservative (*) and a more permissive (**) approach, as detailed in SI Results and Discussion, Genome Composition Analysis.

Metabolism.

Central metabolism and carbon sources.

Both the symbiont and the free-living P. necessarius lack a glycolytic pathway. They do not possess the central regulatory enzyme of the Embden-Meyerhof pathway (6-phosphofructokinase) nor enzymes specific to the Entner-Doudoroff variant, which is used by most Burkholderiaceae bacteria. They also lack any enzyme that could phosphorylate glucose to glucose-6-phosphate or be involved in the assimilation of other monosaccharides. The inability to exploit sugars as carbon or energy sources reflects a general poverty in catabolic pathways.

Genomic analysis (Fig. 2 and SI Results and Discussion, Details on Metabolic Analysis; see also ref. 19) suggests that the principal carbon sources for the free-living strain are pyruvate, acetate, carboxylic acids, and probably compounds convertible to them. These can be metabolized to acetyl-CoA, which is the key intermediate of both energy production and anabolism, thanks to a complete glyoxylate cycle, tricarboxylic acids cycle (TCA), and gluconeogenesis pathways. Acetyl-CoA can additionally be directed to the synthesis of polyhydroxybutyrate (PHB), a storage polymer whose production has been investigated in the bacterium Ralstonia eutropha, also a member of Burkholderiaceae (23). A total of eight amino acids and all TCA intermediates can also be converted to glucose.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Schematic drawing of selected pathways of Polynucleobacter necessarius metabolism as inferred by genomic analysis. Elements in red are exclusive to the free-living strain’s genome. DH, dehydrogenase; cyt, cytochrome.

The symbiont possesses all of the aforementioned enzymatic paths with the exception of the glyoxylate cycle, the metabolic link between acetyl-CoA and biosynthetic pathways. Notably, genes involved in the polymerization, depolymerization, and metabolic regulation of PHB are still present. Thus, non-TCA carboxylic acids are not exploitable as sole carbon sources, and most enzymes acting on related compounds are missing. Only three amino acids can potentially be converted to glucose in the symbiont (SI Results and Discussion, Details on Metabolic Analysis). Thus, it either relies on a very small range of compounds as sole carbon sources, or directly imports various metabolic precursors from its host.

Other Genomic Features.

Physiological Bases of the Polynucleobacter–Euplotes Symbiosis.

The metabolic profile provides a clear explanation for the inability to grow symbiotic P. necessarius strains outside their hosts (15). The symbiont relies on the ciliate at least for carbon sources, organic nitrogen and sulfur, and other essential molecules, including many cofactors. In contrast to ancient or extremely specialized symbionts (e.g., ref. 21), though, it can still perform its own basic anabolic processes and energy production. This condition is similar to that of the other symbiotic member of the family Burkholderiaceae with a severely reduced genome: “Candidatus Glomeribacter gigasporarum” (22), a beneficial endosymbiont of arbuscular mychorrhizal fungi. P. necessarius and “Ca. G. gigasporarum” underwent independent events of genome reduction [see also the phylogenetic tree of Ghignone et al. (22)] that produced the loss of different metabolic pathways. For example, “Ca. G. gigasporarum” has conserved more amino acid degradation pathways than P. necessarius (SI Results and Discussion, Details on Metabolic Analysis) but lacks instead the β-oxidation pathway (22). Both share the loss of glycolytic pathways.

The symbiotic P. necessarius is probably specialized for the intracellular environment in other aspects too, as suggested by its reduced set of genes for sensing and stress resistance. Defensive and regulation mechanisms, as well as membrane and cell wall plasticity, seem to be weakened, providing other plausible reasons for the inability of the bacterium to grow outside its host.

It is more difficult to understand why the symbiont is essential for Euplotes survival (12, 15). Many obligate symbionts of eukaryotes described as mutualists serve as a source of essential molecules (7, 25). The possibility that P. necessarius provides at least some metabolites to its host cannot be completely ruled out, but we consider it unlikely. Euplotes are heterotrophic algal and bacterial feeders and can probably obtain all required amino acids and cofactors from their diet, unlike specialized feeders like sap-feeding insects. More likely, the ciliate host requires Polynucleobacter to fix a catabolic deficiency (e.g., in compound degradation) in a pathway usually conserved in both bacteria and eukaryotes but lost in the clade of Euplotes species harboring Polynucleobacter. Vannini et al. (26) worked on a similar premise and provided evidence for a possible role of these bacteria in glycogen depolymerization; nevertheless, we found nothing in the genome supporting this hypothesis, so the real catabolic pathway involved remains uncertain. The genome of a Polynucleobacter-harboring Euplotes and further experimental investigation will be able to better address the matter, now that the genomic bases of Polynucleobacter biology are established.

It has been pointed out (27) that obligate bacterial symbionts generally do not have “symbiotic genes” coding for exotic functions. Their genomes are fundamentally a subset of those of free-living relatives, and the functional role of symbiosis is better explained by a metabolic cooperation between partners (28, 29). An interesting exception are bacterial secretion systems (BSSs), which are probably involved in the ancestral invasion process and are often found in an active or degraded form in symbiont genomes (25). The symbiotic P. necessarius strain does not encode a complete BSS, contrary to some more parasitic-like bacteria that infect amoebas (8). Nevertheless, we found a region similar to a pathogenicity island (positions 625,503–639,686) including 16 ORFs, 10 of which show similarities with type II and type IV secretion system assembly protein genes (the other 6 ORFs do not share significant similarities with any available sequence). This region is of horizontal origin (SI Results and Discussion, Genome Composition Analysis) and is absent in the free-living strain genome. Most of the genes are considerably shorter (range, 13–78%) than their closest homologs and are possibly nonfunctional. An ancestral free-living strain may have acquired these “invasion genes,” which allowed it to survive ingestion and digestion in a predatory Euplotes, and was then trapped in the cytoplasm. The ubiquity of free-living P. necessarius and the high frequency of BSS genes horizontal transmission, together with no apparent requirement for the evolution of novel functions (at least at the genomic level), may have facilitated the multiple origins and complex pattern of replacement inferred for these essential Euplotes symbionts (14).

Genomic Reduction in the Symbiont.

The symbiotic P. necessarius isolate possesses one of the smallest genomes observed so far in Betaproteobacteria, being surpassed only by two exceptional cases that blur the distinction between organism and organelle: “Candidatus Tremblaya princeps” (0.14 Mbp, 148 genes) (30) and “Candidatus Zinderia insecticola” (0.21 Mbp, 232 genes) (29).

Progressive genome reduction is the rule for obligate symbionts (30, 31). This process has been explained by nonselective mechanisms in insects’ symbionts: either relaxed selection and enhanced genetic drift (32, 33) or an increase in mutation rates (34). The first hypothesis stems from the decreased number of essential functions in symbionts and their small population sizes, reduced gene exchange, and frequent occurrence of bottlenecks. The prediction is for a higher synonymous/nonsynonymous (dN/dS) ratio in symbiotic lineages, which has often been reported (32, 33), although a potential bias resulting from synonymous-site saturation has been suggested (34, 35). In P. necessarius, dN is slightly higher in the symbiotic than in the free-living strain (Fig. 3), in accordance with expectations. Nevertheless, it was impossible to obtain confident estimates of dS in the two lineages despite their unparalleled level of sequence similarity, because synonymous sites were still saturated on the branch leading to the closest outgroup. When more P. necessarius genomes of both free-living and symbiotic strains become available, this problem will almost certainly be circumvented with the help of an intraspecific outgroup.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Nonsynonymous divergence between 904 homolog functional genes in the free-living and symbiotic Polynucleobacter necessarius. The rate of divergence from the outgroup is higher in the symbiont (averages: 0.0556, SE: 0.0049 vs. 0.0479, SE: 0.0035; P = 0.006, two-tailed t test). Axes are in logarithmic scale, and each circle represents the average position of 40 gene-windows projected onto the diagonal.

Moran et al. (31) and McCutcheon and Moran (36) described the steps of genome erosion: the crucial turning point is the loss of DNA repair mechanisms, which brings increased mutation rates, an A+T bias, and massive gene inactivation driven by the spread of mobile elements (MEs). Under these conditions, elimination of DNA in noncoding regions is expected to be due to an intrinsic deletion bias (30, 37). Most of our findings are compatible with this scenario, notwithstanding the differences between the systems studied: P. necessarius does not experience a bottleneck during the host asexual division, and ciliate effective population sizes are arguably larger than those of insects. Where our system stands in striking contrast to the prior model is the virtual absence of MEs, paired with the abundance of pseudogenes. MEs and pseudogenes should be a signature of the first stages of genome erosion and until now have always been found together and in large numbers in recently evolved symbionts (30, 35, 36), but in very low quantity, or entirely missing, in the extremely reduced genomes of more ancient symbionts (36). In this regard, the P. necessarius case is an exception in recent symbionts and may be explained by the preexisting small size of the ancestral genome or by a relatively large effective population size. It demonstrates, however, that the multiplication of MEs is not a required step of genome erosion.

An aspect that has been less investigated is the role of translesion DNA polymerases (TLPs), and we propose here a scenario that may deserve attention in future studies. The symbiotic P. necessarius lacks all TLPs. Although the loss of MMR and homologous recombination may increase the rate of mildly deleterious mutations, the loss of TLPs introduces the risk that a single damaged nucleotide, even in noncoding regions, can entirely block replication. This expectation is supported by the severe reduction in survival of bacteria with artificially impaired translesion synthesis under DNA-damaging conditions (38, 39) and reduced fitness under nonstress conditions (40). A small genome, however, provides less target sites for damage. If the last TLP was lost passively during the early stages of gene inactivation, this would have exerted an additional pressure toward deletions in sequences whose function is not selected for maintenance. The umuC and umuD genes (encoding the two subunits of Pol V, the only TLP present in the free-living strain) are not even recognizable as pseudogenes in the genome of the symbiotic Polynucleobacter, suggesting that the loss of translesion synthesis happened relatively early during the genome erosion process.

Not all proximate causes of genome reduction are known, but among them there is illegitimate recombination, apparently more so when coupled with MMR loss (24). Although the molecular pathways involved are not entirely characterized, TLPs actually participate in deletion accumulation during illegitimate recombination in Salmonella (41). However, other mechanisms have been proposed in different systems (e.g., ref. 42). Illegitimate recombination and absence of MMR likely constitute the main source of deletions in Polynucleobacter through TLP-independent mechanisms.

The probability that an aborted chromosome replication leads to cell death could be alleviated by multiple copies of the genome. It is thus intriguing to also link the loss of TLPs with the presence of multiple nucleoids in Polynucleobacter. Polyploidy has been observed in other intracellular symbionts with reduced genomes, such as Buchnera (27), and this explanation might apply to them too (and is not in contradiction with the possibility that a multiple-copy genome is advantageous because of increased gene expression). An alternative hypothesis is that multiple nucleoids arose in P. necessarius before the loss of the last remaining TLP, perhaps as byproduct of an unbalanced cell cycle, and paved the way for reduced selection on translesion synthesis maintenance. However, the proximal causes of the polyploidy in the symbiont are not immediately apparent from genomic sequences. The symbiont and the free-living P. necessarius share a very similar set of genes involved in chromosome segregation and cell division [including the ParAB-parS, structural maintenance of chromosome (SMC), and filamentous temperature-sensitive proteins (Fts) systems (43)].

Genomic Reduction in the Free-Living Progenitor.

As stressed above, the genome of the symbiotic P. necessarius strain is largely a subset of that of its conspecific relative. However, the free-living P. necessarius strain QLW-P1DMWA-1 already possesses a remarkably small genome (19), comparable in size to that of betaproteobacterial obligate pathogens like Neisseria (2.09–2.28 Mbp) and “Ca. G. gigasporarum” (approximately 1.72 Mbp) and much smaller than those of other free-living or facultative symbiotic Burkholderiaceae (range, 3.75–9.73; Table S2). Thus, a first event of gene loss in the Polynucleobacter lineage occurred before the establishment of the symbiosis, presumably in a free-living ancestor.

Genome streamlining in free-living bacteria is less understood than genome erosion in symbionts but has drawn attention in Prochlorococcus (44) and “Candidatus Pelagibacter ubiquae” (45), two marine taxa with huge global populations. Most authors have argued for adaptive explanations of genome streamlining in free-living bacteria (44, 46), and selection-driven gene loss has been reported for experimental populations (47, 48). In particular, mechanisms requiring relaxed selection and increased drift are generally considered unrealistic because of the huge population sizes of these bacteria.

Nevertheless, the reduced genomes of symbionts and the streamlined genomes of free-living bacteria share many analogies, like a higher AT content and reduced DNA repair mechanisms, in particular MMR loss [reported here for P. necessarius and also observed in strains of Prochlorococcus (44) and “Candidatus Pelagibacter” (49)]. This suggests that nonadaptive mechanisms may have shaped the genomes of these free-living lineages too, perhaps acting in the past when the population sizes were smaller. Although likely detrimental during the first stages, a genome erosion driven by the same mechanisms invoked for symbionts may produce a compact, specialized, and less costly metabolism that could be very successful in the right environment.