Main

Endophytic bacteria reside within the living tissue of plants without substantively harming them. They are of high interest for agrobiotechnological applications, such as the improvement of plant growth and health, phytoremediation1 or even as biofertilizer2. Supply of nitrogen derived from fixation of atmospheric N2 by grass endophytes, such as Gluconacetobacter diazotrophicus and Azoarcus sp. strain BH72, which has been shown to occur in sugarcane3 and Kallar grass2, is a process of potential agronomical and ecological importance.

Although the lifestyle of these endophytes is relatively well documented, the molecular mechanisms by which they interact beneficially with plants have only been poorly elucidated. A combination of features makes Azoarcus sp. strain BH72 an excellent model grass-endophyte4. (i) It supplies nitrogen derived from N2 fixation to its host, Kallar grass (Leptochloa fusca (L.) Kunth); in planta it is usually not culturable, but can be detected by culture-independent methods based on nifH-encoding nitrogenase reductase, the key enzyme for N2 fixation2. (ii) It colonizes nondiseased plants in remarkably high numbers: estimates range from 108 cells (culturable cells per gram root dry weight (RDW) of field-grown Kallar grass5) to 1010 cells (estimated on the basis of abundance of bacterial nifH-mRNA in roots)2. (iii) It is the only cultured grass endophyte shown by molecular methods to be the most actively N2-fixing bacterium of the natural population in roots2. (iv) It also colonizes the roots of rice, a cereal of global importance, in high numbers (109 cells per g RDW) in the laboratory, and spreads systemically into shoots6. Plant stress response is only very limited in a compatible, that is, well-colonized rice cultivar7. Notably, Azoarcus sp. strain BH72 is capable of endophytic N2-fixation inside the roots of rice8.

For a wider application in agriculture, more knowledge is required on mechanisms of interaction and host specificities. Although the genome of a related species, strain EbN1, belonging to a branch of Azoarcus species that typically occurs in soils and sediments but not in association with plants9, is available10, phenotypic differences and phylogenetic distances of 5–6% suggest they might deserve the rank of a separate genus in future9. The plant-associated strain BH72—like many N2-fixing endophytes grass endophytes—has not been detected in root-free soil11. In this study, we present the complete genome sequence of a diazotrophic grass endophyte, Azoarcus sp. strain BH72, and highlight features that may contribute to knowledge of the endophytic lifestyle of these plant-beneficial bacteria, which may be instrumental in developing biotechnological applications.

Results

General features of the genome and mobile elements

The Azoarcus sp. strain BH72 genome sequence was obtained with a whole genome shotgun approach, the assembly being validated by a complete fosmid (Fig. 1b) and a bacterial artificial chromosome (BAC) map12. Characteristics of the single, circular chromosome and the predicted genes are shown in Figure 1 and Table 1.

Figure 1: Circular representation of the Azoarcus sp.
figure 1

strain BH72 genome displaying relevant genome features and validation of the sequence assembly by a fosmid map. In the final consensus sequence each base matched at least phred40 quality. (a) From the outer to the inner concentric circle: circle 1, genomic position in kilobases; the origin of replication was clearly detectable by a bias of G toward the leading strand (GC skew); the start of the dnaA gene located in this region was defined as zero point of the chromosome; circles 2 and 3, predicted protein-coding sequences (CDS) on the forward (outer wheel) and the reverse (inner wheel) strands colored according to the assigned COG (clusters of orthologous groups) classes; leading strand, 2,074 CDS = 52.0%; lagging strand, 1,918 CDS = 48.0%; circle 4, tRNA genes (green) and the four rRNA operons (pink); circle 5, the G+C content showing deviations from the average (67.92%); circle 6, GC skew; a bi-directional replication mechanism suggested by a clear division into two equal replichores. (b) Fosmid map of the Azoarcus sp. strain BH72 chromosome. Each blue arc represents a single fosmid clone mapped onto the assembled sequence; circle 1, CDS with homologs in the chromosome of Azoarcus sp. strain EbN1 (e-value below e−30); circle 2, gene clusters coding for surface-related proteins or other functions not related to proteins of Azoarcus sp. strain EbN1: exopolysaccharide/ lipopolysaccharide-related and pilus-related gene clusters (blue), flagella and chemotaxis related gene clusters (light blue), virulence-related gene clusters (red), proteins related to metabolism (gray), conserved hypothetical proteins or other proteins related to proteins of rhizobia or plant commensals, and various genes not present in Azoarcus sp. strain EbN1 (gold); circle 3, putative genomic islands predicted by the Pai-Ida program 1.1 (score > 3.8); circle 4, transposases and phage-related genes.

Table 1 Genome features of the N2-fixing endophyte Azoarcus sp. strain BH72 in comparison to the denitrifying soil bacterium Azoarcus sp. strain EbN1

The genome contains remarkably few phage- or transposon-related genes, indicating a low degree of lateral transfer and genome rearrangements; just eight loci (Fig. 1b, circle 4) contain genes for integrases, recombinases, transposases or phage-related genes (Supplementary Table 1 online). Only a few loci correspond to predicted anomalous gene clusters or putative pathogenicity islands (Fig. 1b, circle 3). In contrast, the genome of the Azoarcus-related soil isolate strain EbN1 contains 237 transposon-related genes10. Also rhizobial genomes harbor >100 transposases or phage-related genes (http://www.kazusa.or.jp/rhizobase/). Likewise, many plant-pathogenic proteobacteria contain high numbers of mobile elements14. High genomic plasticity might reflect the need for continuous adaptation to changing environments like soil or to host defense mechanisms. For nodule symbionts, soil is an alternative habitat in their life cycle; in contrast typical grass endophytes can not usually be isolated from root-free soil11,13. The comparatively low number of mobile elements in the endophyte BH72 might indicate a low rate of recent gene transfer and genome rearrangements, which is presumably due to adaptation to a stable, low-stress microenvironment inside plants.

Comparative genomics

Genome comparison revealed a surprisingly low degree of synteny between genomes of strain BH72 and the Azoarcus-related strain EbN1 (Fig. 2). At a low cutoff e-value of e−30, the majority of predicted proteins (58%) in strain BH72 have some counterparts in strain EbN1 (Fig. 1b, circle 1). However, only 43% of these proteins were more closely related to those of EbN1 than to proteins of other strains. Other pathogenic or plant symbiotic proteobacteria have even less related genomes (Supplementary Table 2 online). Because strains BH72 and EbN1 have a very different ecology, the differences may give important hints as to which genes are required specifically for the endophytic lifestyle. Several gene clusters of strain BH72 that are lacking in EbN1 or that are more similar to genes of other bacteria (Fig. 1b) harbor genes that encode proteins putatively involved in cell surface components or other features that may be required for the endophytic lifestyle (see below, Fig. 3 and Supplementary Table 1 online).

Figure 2: Synteny between the genomes of Azoarcus sp.
figure 2

strain BH72 and Azoarcus sp. strain EbN1. The genome of Azoarcus sp. strain EbN1 is adjusted to dnaA with its start codon as zero point of the chromosome. The diagram depicts x-y plots of dots forming syntenic regions between the two Azoarcus genomes. Each dot represents an Azoarcus sp. strain BH72 CDS having an ortholog in Azoarcus sp. strain EbN1, with coordinates corresponding to the CDS number in each genome. The orthologs were identified by best BLASTP matches of amino acid sequences (e-value < e−30).

Figure 3: Gene clusters in Azoarcus sp.
figure 3

strain BH72 that are lacking in EbN1 or are more similar to genes of other bacteria. Synteny of selected clusters with gene clusters of other bacteria: gum-cluster, Xanthomonas campestris pv. campestris (locus_tag numbers XC_1671 - XC_1658); lipopolysaccharide-related cluster, Pseudomonas aeruginosa PA01 (locus_tag numbers PA5012 - PA4997); sci-cluster, Pseudomonas aeruginosa PA01 (locus_tag numbers PA0074 - PA0101); pink/white, genes present or not present in the gene cluster of strain BH72, respectively. Numbers refer to genes of Azoarcus sp. strain BH72 listed in Supplementary Table 1 online. Highest similarities to proteins of other bacteria are depicted by the following colors: red, human or animal pathogens; green, root nodule symbionts (rhizobia); yellow-green, root-associated bacteria; turquoise, plant pathogens; other bacteria according to their phylogenetic affiliations: black, Azoarcus sp. strain EbN1; gray, beta-subgroup of Proteobacteria; blue, gamma-subgroup of Proteobacteria; yellow, alpha subgroup of Proteobacteria; brown, others.

Carbon metabolism and signal transduction

Strain BH72 has a strictly respiratory type of metabolism and does not grow on any carbohydrate tested9,15. Aspects of putative carbon metabolism are shown in Figure 4. The inability to utilize common carbohydrates might contribute to a plant-compatible endophytic lifestyle because, in contrast to phytopathogens, the bacteria cannot grow and proliferate on the major cell wall constituents although a cellulase is present16.

Figure 4: Overview of physiological features of Azoarcus sp.
figure 4

strain BH72. Depicted are carbohydrate metabolism, selected transporters and catabolic pathways for organic acids, nitrogen assimilatory and dissimilatory pathways, inorganic ion transport systems, and protein translocation systems. Crosses indicate pathways or reactions that are apparently not present in strain BH72. Details of metabolic features: the genome does not contain genes required for a functional Entner-Doudoroff pathway or the oxidative branch of the pentose phosphate pathway, however, the nonoxidative pentose phosphate pathway is complete. All enzymes required for glycolysis via the Embden-Meyerhoff pathway, gluconeogenesis and the TCA cycle are present, including genes encoding a typical phosphoenolpyruvate: sugar phosphotransferase system (PTS), but no specific outer membrane transporter for carbohydrates. Four complete and two incomplete copies of the TRAP-transport system (DctPQM) for C4-dicarboxylates are present. C4-dicarboxylates are probably metabolized via the glyoxylate shunt pathway, malate being decarboxylated by the malic enzyme (MaeB, two gene copies). Ten putative alcohol dehydrogenases are encoded in the genome (see also Supplementary Table 1 online), five of which are PQQ-dependent enzymes not present in strain EbN1.

The major carbon sources for strain BH72 are dicarboxylic acids and ethanol9. Transport systems for C4-dicarboxylates (Fig. 4) might be of vital importance during the association with host plants, as in symbiotic rhizobia17. Ethanol might be important for association with flooded plants like rice, which accumulate ethanol under anoxic conditions, especially at root tips—one of the typical sites of colonization of strain BH72. Correspondingly, its genome harbors ten genes encoding putative alcohol dehydrogenases.

Strain BH72, despite being adapted to a relatively stable, low-stress microenvironment, shows a remarkable density of signal transduction systems (see details in Supplementary Table 3 online). Thus it may be a good example for sophisticated signal transduction networks.

N2 fixation and nitrogen metabolism

Azoarcus sp. strain BH72 appears to be highly adapted to environments poor in available nitrogen sources, which correlates with its role as an N2-fixing endophyte (Fig. 4). (i) A low-affinity glutamate dehydrogenase (GDH) for ammonium assimilation is lacking, a feature highly unusual in free-living bacteria, whereas it is present in strain EbN1. Only the high-affinity ATP-consuming assimilation system (GS[2x]-GOGAT) is present. (ii) Four genes encoding high-affinity ammonia transporters exist (amtB/Y/D/E), one of them with an additional regulatory domain. (iii) In contrast to the soil strain EbN1, structural genes for the molybdenum-dependent nitrogenase complex and all genes required for cofactor synthesis and maturation of the nitrogenase are present in strain BH72, one of them in two copies (nifY). Several putative low-potential electron donors for N2 fixation were identified including two flavodoxin-encoding genes (nifF1, nifF2), 12 genes for ferredoxin-like proteins; two clusters encoding putative electron transport systems (rnf1, rnf2) might be instrumental for electron supply to ferredoxin during N2 fixation. Several genes likely to be involved in the regulatory cascade are also listed in Supplementary Table 1 online. Although in pure culture, N2-fixing strain BH72 does not excrete substantial amounts of nitrogenous compounds18, it supplies fixed nitrogen to its grass host2. The four ammonia transport proteins are putative candidates for export to the plant. Two transport systems for glutamate or glutamine as well as nine for branched-chain amino acids were also identified; however, the presence of periplasmic substrate–binding proteins suggests that these systems are used for import and not for export.

About 38 genes encoding enzymes and transporters involved in nitrate metabolism were identified (Fig. 4 and Supplementary Table 1 online). As in strain EbN1, genes required for assimilatory nitrate and nitrite transport and reduction are present in strain BH72 in one copy (nasFED, nirC, nasA, nasC, nirBD). In contrast to strain EbN1, strain BH72 cannot conduct denitrification to N2, as genes required for a nitrite reductase are missing in strain BH72. However, genes for subsequent denitrification reactions (norCBQD and nosZRDFYL encoding reductases for nitric oxide and nitrous oxide, respectively) are present. The periplasmic localization of the nitrate reductase components (duplicated nap operon) and detoxification of NO2 via an NO3/NO2 antiporter (NarK) might decrease toxicity of nitrite. In contrast, the denitrification pathway is complete in strain EbN1 for which denitrification appears to be a typical feature10.Footnote 1

Iron-transport related proteins

In Gram-negative bacteria, TonB-dependent, outer-membrane proteins are responsible for the specific uptake of ferric-siderophore complexes, high-affinity iron chelators. They are also important for perception of environmental signals and are associated with pathogenicity of plant pathogens19. Strain BH72 possesses 22 genes encoding proteins related to iron transport (Supplementary Table 1 online), twice as many as described for strain EbN1 and even more than other N2-fixing endosymbionts (Bradyrhizobium japonicum, 13; Sinorhizobium meliloti, 9; and Mesorhizobium loti, 1). Two genes (azo2156, azo3836) are not even present in the P. fluorescens Pf5 genome, a plant-associated bacterium known for its capacity to produce and take up a wide range of siderophores20, which contains 45 such genes. Although putative receptors for hydroxamate- and catechol-type siderophores, ferricitrate, vitamin B12, colicins and unknown substances are present, there was no evidence for biosynthetic pathways for known hydroxamate or catecholate siderophores21. Moreover, production of siderophores was not detected experimentally (Supplementary Fig. 1a online). Apparently, this strain is highly adapted to obtaining chelated iron from other sources, as fungi and monocotyledonous plants also produce siderophores22. The high number of putative receptors suggests a role not only for rhizosphere competence of strain BH72, but also for biocontrol.

Plant-associated lifestyle

Surface characteristics of bacteria are important factors for recognition by and interaction with the host. Several gene clusters putatively related to surface components of strain BH72 are lacking in strain EbN1, or are more highly related to genes of plant-associated or pathogenic bacteria.

Type IV pili are among the few factors known to affect endophytic colonization of grass diazotrophs23. Establishment of microcolonies on roots and fungal mycelium, and systemic spreading in rice are mediated by type IV pili23. The strain BH72 genome harbors 41 genes encoding proteins putatively involved in pilus assembly and regulation, whereas only 30 such genes were found in strain EbN1. Genes highly similar in both species encode proteins with conserved function such as assembly or regulation (for example, PilBCD-PilF-PilMNOPQ-PilTU-PilZ or PilSR-PilGHIL). Other pilus proteins related mostly to phytopathogenic bacteria might be either pseudopilins involved in secretion (PilV/W/X) or putative tip adhesins (PilY1A/B, 31% and 39% sequence identity to Ralstonia solanacearum and Xylella fastidosa, respectively) that are lacking even in strain EbN1 and might be characteristic for interaction with plant surfaces (Supplementary Table 1 online).

Other cell surface components that are often involved in recognition or virulence of pathogens are lipopolysaccharides, exopolysaccharides and capsular material. Intriguingly, many gene products putatively involved in their synthesis in strain BH72 are not highly related to those of the soil isolate EbN1, but to proteins of plant symbionts, pathogens or gamma- or alpha-proteobacteria (Supplementary Table 1 online). Several genes of strain BH72 are similar to the gum operon for exopolysaccharide production in phytopathogenic Xanthomonas campestris24; genes encoding putative glycosyl transferases have considerable similarity to the rhizobial pss gene cluster (Supplementary Table 1 online, 20–23), which is involved in exopolysaccharide polymerization, translocation and thus in plant-microbe interaction; a gene cluster related mainly to lipopolysaccharide synthesis is most similar to genes of gamma-proteobacteria including pathogens; these clusters did not show sufficient synteny to support the assumption of a very recent gene transfer (Fig. 3).

Motility. Flagella are pivotal for motility, adhesion, biofilm formation and colonization of the host. Strain BH72 is highly motile by means of a polar flagellum. At least 48 genes were identified that are generally required for biosynthesis and function of flagella and chemotaxis. They are located in three different noncontiguous clusters and are mostly related to genes of other beta-proteobacteria and a few pathogens of the gamma-subgroup (Supplementary Table 1 online). There are three genes encoding flagellins (fliC1, fliC2, fliC3) and two encoding flagellar motor proteins, suggesting an important role for motility in the plant-associated lifestyle. In contrast, the nonmotile strain EbN1 does not possess a complete flagellar regulon10.

Secretion and communication. Several genes encoding potential protein secretion systems were identified in the genome of strain BH72 (Fig. 4) for a sec-dependent pathway, a signal recognition particle (SRP)-mediated translocation and a twin arginine translocation (Tat) system, all of them targeting proteins through the inner membrane. Secretion of proteins through the entire cell envelope seems to be limited to only three varieties of pathways. Genes were identified that encode one type I secretion system and one autotransporter. Two gene clusters were detected that encode a type II secretion-related system (type IIb secretion system25), consisting only of GspDEFG.

Two other secretion systems are common to plant-associated bacteria, type III and IV secretion systems, which transport a wide variety of effector proteins into the extracellular medium or into the cytoplasm of eukaryotic host cells and affect interaction26,27,28,29. Intriguingly, neither system is present in strain BH72, probably preventing the export of toxic proteins to the host.

'Quorum sensing' is a common way of bacteria to communicate with each other or hosts by means of autoinducers that accumulate in the extracellular environment in a cell density-dependent manner30. Although there is evidence that autoinducer-dependent gene regulation occurs in Azoarcus sp. strain BH72 (Böhm, M. & Reinhold-Hurek, B., unpublished data), this strain appears to escape the usual communication systems. Widespread autoinducers of Gram-negative bacteria are N-acyl homoserine lactones (AHLs)30. There is no evidence for genes encoding an AHL-based quorum-sensing system in strain BH72; genes encoding the autoinducer synthetase (LuxI/LasI–type) or the responsible cytoplasmic autoinducer receptor (LuxR/LasR-type) are lacking. Furthermore, different bacterial sensor strains detecting presence of short- or long-chain AHLs did not yield a positive response toward strain BH72 (Supplementary Fig. 1b online). Also genes encoding the autoinducer-2 synthetase LuxS30 are lacking. Gram-positive bacteria usually use peptides as autoinducers30. Genes expected for this system were not detected in strain BH72 either.

Virulence and interaction factors. The strain BH72 genome stands out by the lack of obvious genes involved in production of toxins. Moreover, common hydrolytic enzymes that macerate plant cell wall polymers and thus contribute to a phytopathogenic lifestyle and plant damage are rare: pectinase-encoding genes are absent; only a few genes encode putative glycosidases (palZ, spr1, ndvC, eglA), some of them with transmembrane helices. Detection of genes for membrane-bound enzymes is in agreement with the observation that strain BH72 does not secrete cellulases into the culture medium, but shows activities of a cell surface-bound endoglucanase (EglA) and exoglycanase that also hydrolyses xylosides16, a major component of primary cell walls in grasses. A low production of macerating enzymes is likely to contribute to compatibility with the plant host, however these hydrolases might assist in endophytic colonization, as shown for the endoglucanase EglA31.

There is no genomic evidence for the central process in the rhizobium-legume symbiosis, the induction of nodulation. Common nodABC genes required for the biosynthesis of the Nod-factor backbone are not present in strain BH72; only a few genes show some sequence similarity to other nod or nol genes (Supplementary Table 1 online). However, like other grass-associated microbes such as Azospirillum a gene similar to nodD is present, in rhizobia encoding a central regulator for flavonoid-inducible gene expression.

Some other gene clusters in strain BH72 are also interesting targets for putative roles in plant-microbe interactions. One cluster shows similarity to genes that are localized in the sci-genomic island, affecting virulence of human pathogens. The genomic organization shows remarkable synteny with genes of P. aeruginosa (Fig. 3), arguing for a more recent gene transfer. Interestingly some homologs are also present in rhizobia. Other noticeable gene clusters code for mainly conserved hypothetical proteins or metabolism-related proteins that have orthologs in rhizobia, or of regulatory proteins (ColRS) important for root colonization of commensals (Supplementary Table 1 online, no. 177–185, 188–196, 198–200).

Discussion

The complete genome sequence of Azoarcus sp. strain BH72 offers insights into genomic strategies for an endophytic life style, and allows identification of various features that may contribute to their interaction with plants. The strain appears to be adapted to a relatively stable, low-stress microenvironment, since its genome contains remarkably few phage- or transposon-related genes in comparison to many soil bacteria or pathogens, indicating a low plasticity of the genome. The lack of the typical communication system of Gram-negative bacteria based on AHLs also argues for a rather exclusive microhabitat.

Strain BH72 appears to be disarmed compared to plant pathogens by its in ability to metabolize carbohydrates coupled with the lack of a massive occurrence of cell-wall degrading enzymes. Moreover, this bacterium lacks known toxins and type III and IV secretion systems that are typically used by pathogens to transport effector molecules to their host. This might be instrumental for avoiding damage to the plant host despite a dense internal colonization, and only the small set of hydrolases identified may be required for penetrating into the plant tissue. Some specific features of nodule symbionts are also lacking, like most nod/nol genes required for nodule induction.

Genome comparison with the Azoarcus-related, nondiazotrophic, nonendophytic soil bacterium strain EbN1 revealed features likely to be important for plant-microbe interaction. Strain BH72 appears to be highly adapted to environments of low nitrogen availability, N2-fixation playing a key role in its ecology. Several gene clusters that were lacking in strain EbN1 or were highly similar to genes of plant-associated or pathogenic bacteria were related to cell surface components that are often involved in recognition—gene products participating in the synthesis of exopolysaccharide, lipopolysaccharide, type IV pilus tips, the flagellar and chemotaxis apparatus. Further targets for studying interaction mechanisms were also identified by comparative genomics, such as virulence-related sci genes or genes encoding conserved hypothetical proteins shared with nodule symbionts. A large and diverse set of TonB-dependent receptors (22) might play a role in iron acquisition and biocontrol. In future functional genomic analyses, the role of these target genes for host compatibility will be elucidated, which is crucial for a wider agro-biotechnological application of N2-fixing endophytes.

Methods

Whole genome shotgun sequencing.

DNA shotgun libraries with insert sizes of 1 kb and 2–3 kb in pGEM-T (Promega), and 8-kb fragments in pTrueBlue-rop (MoBiTec) vectors were constructed by MWG Biotech. Plasmid clones were end-sequenced on ABI 3700 sequencers (ABI) by MWG Biotech AG. Basecalling was carried out using PHRED32,33. High-quality reads were defined by a minimal length of 250 bp with an averaging quality value of ≥20. Finally, 60,715 high-quality reads, a total of 39,266 (5.26 ×), 18,070 (2.32 ×) and 3,379 (0.40 ×) end sequences (×'s indicate genome equivalents) from libraries with 1-kb, 2- to 3-kb and 8-kb inserts, respectively, were obtained.

Sequence assembly and assembly validation.

Basecalling, quality control and elimination of vector DNA sequences of the shotgun-sequences were performed by using the software package BioMake (Bielefeld University) as previously described34. Sequence assembly was performed by using the PHRAP assembly tool (http://www.phrap.org). The CONSED/AUTOFINISH software package35,36, supplemented by the in-house tool BACCardI37, was used for the finishing of the genome sequence.

For gap closure, a BAC library with inserts of 90 kb in pBeloBAC11 was constructed and BAC contigs were assembled12. Remaining gaps of the whole genome shotgun assembly were closed by sequencing on shotgun and BAC clones carried out by IIT GmbH on LI-COR 4200L and ABI 377 sequencing machines. So that we would obtain a high quality genome sequence, all regions of the consensus sequence were polished to at least phred40 quality by primer walking. Collectively, 1,374 sequencing reads were added to the shotgun assembly for finishing and polishing of the genomic sequence. Repetitive elements, that is, rRNA operons, were sequenced completely by primer walking on BAC clones. For assembly validation, a fosmid library with inserts of 35–38 kb was constructed by IIT GmbH using the EpiFOS Fosmid Library Production Kit (Epicentre). End-sequencing of 672 fosmids was carried out on ABI 377 and MegaBACE 1000 (Amersham Biosciences) sequencing machines by IIT GmbH and on ABI 3730XL DNA analyzers by the sequencing group of the Max Planck Institut für Molekulare Genetik. For assembly validation, fosmid end sequences were mapped onto the genome sequence by employing the BACCardI tool.

Genome analysis and annotation.

In a first step automatic gene prediction and annotation were performed using the genome annotation system GenDB 2.0 (ref. 38) as previously described14. In a second step the coding sequences (CDS) prediction was validated: a position weight matrix (PWM) was generated to score all translation starts, and visualization of CDS was done using GeneQuest (DNASTAR Inc.). Reinspection of starts was coupled to recomputing of homology (BlastP) and assessment of function. In this way 4.3% more ORFs were detected, and 15.5% of the start sites were changed in comparison to the prediction obtained by a combined GLIMMER and CRITICA approach39. Intergenic regions were checked again for CDS missed probably by the automatic annotation using the BLAST programs40. During manual annotation, the following criteria were applied: (i) for hypothetical proteins, amino acid identities to other proteins were less than 30% over the entire length of the protein; (ii) for conserved hypothetical proteins, amino acid identities to proteins of unknown function were more than 30%; (iii) for proteins with putative or probable functions, sequence identities to named gene products were >20% or 40%, respectively.

Genomic comparison.

For comparative analyses, the annotated genome sequence of Azoarcus strain EbN1 (acc. nos. CR555306, CR555307, CR555308) was imported into GenDB. Comparisons of chromosomal sequences were carried out with GenDB38.

Detection of regions with atypical G+C content.

For detection of anomalous gene clusters or putative pathogenicity islands in bacterial genomes, the Pai-Ida program 1.1 (http://compbio.sibsnet.org/projects/pai-ida/) based on an iterative discriminant analysis41 was used.

Database submission.

The nucleotide sequence of the Azoarcus sp. strain BH72 chromosome was submitted to EBI under accession number AM406670–Azoarcus sp. BH72.

Note: Supplementary information is available on the Nature Biotechnology website.