Plasmids with a Chromosome-Like Role in Rhizobia

Abstract

Replicon architecture in bacteria is commonly comprised of one indispensable chromosome and several dispensable plasmids. This view has been enriched by the discovery of additional chromosomes, identified mainly by localization of rRNA and/or tRNA genes, and also by experimental demonstration of their requirement for cell growth. The genome of Rhizobium etli CFN42 is constituted by one chromosome and six large plasmids, ranging in size from 184 to 642 kb. Five of the six plasmids are dispensable for cell viability, but plasmid p42e is unusually stable. One possibility to explain this stability would be that genes on p42e carry out essential functions, thus making it a candidate for a secondary chromosome. To ascertain this, we made an in-depth functional analysis of p42e, employing bioinformatic tools, insertional mutagenesis, and programmed deletions. Nearly 11% of the genes in p42e participate in primary metabolism, involving biosynthetic functions (cobalamin, cardiolipin, cytochrome o, NAD, and thiamine), degradation (asparagine and melibiose), and septum formation (minCDE). Synteny analysis and incompatibility studies revealed highly stable replicons equivalent to p42e in content and gene order in other Rhizobium species. A systematic deletion analysis of p42e allowed the identification of two genes (RHE_PE00001 and RHE_PE00024), encoding, respectively, a hypothetical protein with a probable winged helix-turn-helix motif and a probable two-component sensor histidine kinase/response regulator hybrid protein, which are essential for growth in rich medium. These data support the proposal that p42e and its homologous replicons (pA, pRL11, pRLG202, and pR132502) merit the status of secondary chromosomes.

The classical view of replicon architecture in bacteria conceives of one indispensable chromosome and several dispensable plasmids. This conception was shattered 2 decades ago by the discovery of additional chromosomes. Based on information available in GenBank, secondary chromosomes have been described in proteobacteria, including members of the Alphaproteobacteria (Agrobacterium tumefaciens, Agrobacterium radiobacter, Agrobacterium vitis, all the Brucellaceae analyzed, Ochrobactrum anthropi, Paracoccus denitrificans, Rhodobacter sphaeroides, and Sphingobium japonicum), Betaproteobacteria (all the Burkholderiales analyzed, Cupriavidus taiwanensis, Ralstonia eutropha, Ralstonia picketii, and Variovorax paradoxus), and Gammaproteobacteria (Alliivibrio salmonicida, Photobacterium profundum, Pseudoalteromonas haloplanktis, and all the Vibrionales analyzed). Also, secondary chromosomes were found in species as diverse as Butyrivibrio proteoclasticus, a Cyanothece sp., Deinococcus radiodurans, Leptospirales (Leptospira biflexa, Leptospira borgpetersenii, Leptospira interrogans), Prevotella melaninogenica, Sphaerobacter thermophilus, and Thermobaculum terrenum.

Primary chromosomes usually are the largest replicons possessing a dnaA-based replication system and harboring all or nearly all the essential genes that govern cellular processes, such as transcription, translation, DNA replication, and energy metabolism (37). On the other hand, secondary chromosomes harbor some essential genes that are either absent from the main chromosome or duplicated in the secondary chromosome. Most of the cases mentioned before possess duplications on the secondary chromosome of genes encoding ribosomal RNAs or tRNAs, which are likely to contribute to normal cell viability. In just a few examples (A. radiobacter, Burkholderia xenovorans, a Cyanothece sp., the Leptospirales, P. haloplanktis, and T. terrenum), secondary chromosomes lack structural RNAs, possessing instead other genes that participate in essential functions. Secondary chromosomes commonly possess replication systems different from the usual dnaA-based system present in primary chromosomes. Since these novel replication systems may be present in bona fide megaplasmids, a recent analysis suggested the name “chromids” to refer to megaplasmids that have acquired sets of essential genes, thus becoming secondary chromosomes (26).

The most common way to identify secondary chromosomes in bacteria has been by localization of unique rRNA and/or tRNA genes through bioinformatic approaches. For instance, secondary chromosomes of A. tumefaciens and A. vitis possess rRNA operons as well as genes for prototrophic growth (24, 55). Another approach to identify secondary chromosomes has been by experimental demonstration of their requirement for cell growth. For instance, a segment encompassing 12.5% of the pSymB replicon (1.6 Mb) of Sinorhizobium meliloti is required for normal growth on rich medium (9), presumably due to the presence of an arg-tRNA gene in that sector (37). Valuable as bioinformatic approaches are, the identification of essential genes based solely on sequence information remains a daunting task. Among the factors that hinder this effort are significant gaps in knowledge, variability in growth strategies among bacteria, nonorthologous substitutions, and reiteration of genes with important effects on cell metabolism (21, 41).

Rhizobium etli CFN42, an alphaproteobacterium and nitrogen-fixing symbiont of the common bean (Phaseolus vulgaris), harbors a multipartite genome constituted by one chromosome and six large plasmids (p42a to p42f), ranging in size from 184 to 642 kb (23). Plasmid elimination assays demonstrated that five of the six replicons are dispensable for cell viability in rich medium. However, despite repeated attempts, it was not possible to eliminate p42e from the wild type (4, 5). Although this stability might be due to an unusually stable partition system, another interesting possibility is that genes on p42e perform essential functions, thus making it a candidate for a secondary chromosome. To ascertain if this is the case, we decided to perform an in-depth functional analysis of p42e, employing bioinformatic tools, insertional mutagenesis, and programmed deletions.

In this work, we demonstrate the active participation of p42e in the primary metabolism of R. etli. Synteny analysis and incompatibility studies revealed highly stable replicons equivalent to p42e in content and gene order in other Rhizobium species. Also, it was not possible to eliminate p42e or homologous plasmids by incompatibility; recA-independent cointegration of p42e with other replicons of the strain was observed instead. Finally, two genes of p42e were identified as essential for growth in rich medium. These data suggest that p42e and its homologous replicons (pA, pRL11, pRLG202, and pR132502) merit the status of secondary chromosomes.

(This research was conducted by C. Landeta in partial fulfillment of the requirements for a Ph.D. in Ciencias Biomédicas from the Universidad Nacional Autónoma de México, Cuernavaca, Mexico, 2011.)

MATERIALS AND METHODS

Bacterial strains and plasmids.

The strains and plasmids used in this study are listed in Table S1 in the supplemental material.

Growth conditions.

R. etli, R. leguminosarum bv. viciae, and Agrobacterium tumefaciens were grown in PY medium (peptone-yeast extract medium supplemented with CaCl₂ at a final concentration of 4.5 mM) at 30°C (42). Escherichia coli strains were grown in Luria-Bertani (LB) broth at 37°C. When needed, antibiotics were added at the following concentrations (in μg ml⁻¹): kanamycin (Kan), 30 (E. coli) or 15 (R. etli); nalidixic acid (Nal), 20; gentamicin (Gm), 30; tetracycline (Tc), 10 (E. coli) or 3 (R. etli); and spectinomycin (Sp), 100.

All the R. etli strains were also grown at 30°C in Y minimal medium (MMY) with 10 mM succinic acid and 10 mM ammonium chloride (3) as carbon and nitrogen sources, respectively. To prevent growth arrest by metabolic imbalance in subsequent cultures, MMY was also supplemented with biotin at 1 mg liter⁻¹ (19). When needed, amino acids or vitamins were added to MMY at the following concentrations (in μg ml⁻¹): cobalamin, 0.5; methionine, 100; and nicotinic acid, 12. For determinations of growth kinetics, the corresponding strains were grown overnight at 30°C in liquid PY medium. Cells were collected by centrifugation at 13,000 rpm in an Eppendorf microcentrifuge and washed twice with fresh MMY, and appropriate volumes were used to inoculate 50-ml MMY cultures at an initial A₆₂₀ of 0.05. The cultures were incubated for up to 24 h with orbital shaking (200 rpm) at 30°C; culture samples (1 ml) were taken every 3 h, and growth was measured by determining the total cell protein concentration by the Lowry method. For determination of nicotinate auxotrophy, it was necessary to starve the desired strains from nicotinate by previous growth for 24 h in MMY medium lacking nicotinic acid. Cells obtained by this procedure were used to inoculate fresh MMY. The CE3 cyoA mutant was grown in MMY plates and incubated at 30°C in a sealed chamber, where microaerobic conditions were set up by repeated flushing with a gaseous mixture of argon-oxygen (99:1, vol/vol). The CE3 actP mutant was grown in PY medium with copper sulfate at a 1 mM final concentration.

Recombinant DNA procedures.

Total and plasmid DNA isolation, digestion with restriction enzymes, cloning, agarose gel electrophoresis, and E. coli transformation were performed by standard procedures (51).

Specific PCR primers were designed using the Oligo 6.0 software and were purchased from Unidad de Síntesis Química (Instituto Biotecnología, Universidad Nacional Autónoma de México [UNAM]). Oligonucleotides used in this work are listed in Table S2 in the supplementary material. PCR amplifications were done in a TC-312 thermocycler (Techgene, Burlington, NJ). Taq DNA polymerase (Altaenzymes, Alberta, Canada) was used in most PCRs, with a cycling regime that includes a denaturing step at 94°C for 3 min, followed by 30 cycles of 94°C for 1 min and 68°C for 1 min, followed by a final elongation step at 72°C for 5 min. To reduce synthesis errors in amplifications involving the repABC region, Platinum Taq DNA polymerase high fidelity (Invitrogen, Carlsbad, CA) was used under the same cycling regime as before. For some primer pairs, a three-step cycling regime, consisting of a denaturing step at 94°C for 3 min, followed by 30 cycles of 94°C for 1 min, 50 to 56°C (depending on the primer annealing temperature) for 1 min, and 72°C for 1 min, followed by a final elongation step at 72°C for 5 min, was used.

For Southern hybridizations, DNA was digested with appropriate restriction enzymes, electrophoresed in 1% (wt/vol) agarose gels, and blotted onto nylon (Hybond N+). Hybridization was carried out under high-stringency conditions, using Amersham's Rapid-hyb buffer (GE Healthcare, United Kingdom). Specific probes were labeled with [³²P]CTP by random priming using Amersham's Rediprime system.

Vector insertion mutagenesis (VIMS).

All the mutations in this work were introduced into the R. etli CE3 genetic background. The mutagenesis design employed entails the cloning of an intragenic segment for the gene of interest into a small, conjugative plasmid that is unable to replicate in Rhizobium (suicide plasmid). Upon transfer to Rhizobium, single-crossover recombination with the homologous target introduces the whole plasmid as a cointegrate, producing a knockout of the gene of interest. Additionally, if the gene is located at the beginning of an operon, loss of expression of the whole operon will ensue, due to polarity effects. To this end, intragenic PCR products (300 to 800 bp) of the desired genes, derived with the appropriate primers, were digested with BamHI and ligated, in most cases, into BamHI-restricted pK18mob (53). Ligated products were transformed into E. coli DH5α (25). After verification of the clones, each plasmid was transformed in E. coli S17-1 (54), and these were used as donors in plate matings with R. etli CE3. R. etli transconjugants were selected on PY medium with kanamycin and nalidixic acid. PCR amplifications of selected Nal^r Kan^r transconjugants were done to verify that the expected single-crossover event had occurred, using a combination of universal oligonucleotides (M13 forward or reverse primers; Invitrogen) and oligonucleotides corresponding to the insert. Mutants were also verified by analyzing BamHI-restricted genomic DNA by Southern blot hybridization, using the appropriate ³²P-labeled pK18mob recombinant plasmids as probes, as described previously (20). For the construction of cobG and mmuM mutants, appropriate PCR products were digested with BamHI and XbaI and cloned into similarly restricted pVEX1311 and pIC20RDA, respectively. Ligated products were transformed into E. coli EK0610 (2) (for pVEX1311) or DH5α. Transfer of pIC20RDA derivatives was done by conventional biparental matings using E. coli S17-1 as the donor; transfer of pVEX1311 derivatives required a triparental mating, employing the appropriate R. etli strain, E. coli EK0610/pVEX1311 cobG, and HB101/pRK2013 as the conjugation helper.

Incompatibility curing.

To explore the incompatibility relationships of p42e with diverse rhizobial replicons, a plasmid harboring the putative incompatibility determinants of p42e was built. To that end, the repABC region of plasmid p42e was amplified by PCR with appropriate primers (see Table S2 in the supplemental material). The amplified segment included a sector encompassing from 500 bp upstream of the start codon of repA to the middle of repC (codon 238), thus including the two proposed incompatibility regions for plasmids of this family (8). The PCR product was then cloned into pCR2.1-TOPO by T-A ligation and transformed into E. coli DH5α. From the resulting plasmid, the whole repABC_p42e region was excised by digestion with KpnI and XbaI and cloned into similarly restricted pBBR1MCS5 (33). The resulting plasmid (pBBRrep1) was introduced into E. coli S17-1, and this strain was used as a donor in plate matings with different Rhizobium or Agrobacterium strains. The plasmid transfer frequency was obtained by dividing the number of Nal^r Gm^r transconjugants by the total number of output recipient cells.

Selected transconjugants from these matings were verified for plasmid content by a variant of the Eckhardt technique (see below). Transconjugants that presented rearrangements (see Results) were additionally verified by PCR amplification of different p42e sectors (nadB, cobF, minC, cyoA, actP, selA, cls), using specific primers and by Southern blot hybridization of the plasmid profiles with appropriate probes.

Generation of site-specific deletions with the Cre-loxP system.

All the deletions constructed here were generated using a modification of the vector-mediated excision system (2). In this system, the region to be eliminated is flanked by loxP sites, the target of the site-specific Cre recombinase. To introduce these sites, a PCR product (ranging from 600 bp to 1.8 kb), delimiting the left end of the planned deletion, was cloned into the loxP Sp^r suicide plasmid pVEX1311. Upon transfer to Rhizobium, single-crossover recombination with the homologous target introduces the whole plasmid as a cointegrate, introducing the loxP site as well. To delimit the right end of the deletion, a second PCR product was cloned into the loxP Tc^r suicide plasmid pIC20R loxP oriT. This plasmid lacks appreciable sequence identity, other than the oriT sequence, with pVEX1311. Integration of the pIC20R loxP oriT derivative into the strain harboring the previous pVEX1311 derivative generates a double-cointegrate strain in which the region to be deleted is flanked by loxP sites in a direct orientation. The orientation of the insertion of these PCR products has to be planned to ensure a direct orientation of the loxP sites in the final, double-cointegrate strain. To generate an in vivo deletion of the desired region, plasmid pBBRCre, which expresses the site-specific recombinase Cre, was introduced by plate matings into the strain harboring the appropriate double cointegrate, selecting for Nal^r Gm^r transconjugants. Excision of the desired region by Cre-mediated popout recombination occurred in these strains at frequencies approaching 90%. The design of the deletion system permits selection of strains carrying the planned deletion by its Sp^r Tc^s phenotype, since the Tc^r determinant lies within the loxP-flanked region and the Sp^r determinant lies outside it. Elimination of pBBRCre from the deleted derivatives was done by growing them in PY medium lacking any antibiotic, looking for Sp^r Gm^s derivatives. To ensure that the desired deletion had taken place, each strain was checked for the size of p42e, the absence of pBBRCre in Eckhardt-type gels (17) (see below), and the absence of specific segments lying within the deleted region by PCR amplification with appropriate primers (see Table S2 in the supplemental material).

Plasmid and chromosome visualization.

For visualization of the large plasmids in Rhizobiales, the Eckhardt technique (17), as modified by Hynes and McGregor (27), was used. In this technique, a mild, in-gel lysis of bacterial cells, followed by electrophoresis in 0.75% (wt/vol) agarose gels for 20 h, allows the easy separation of circular plasmid molecules from 10 kb up to 2 Mb in length.

Chromosome visualization was achieved by pulsed-field gel electrophoresis (PFGE), as described previously (38), with the following modifications. Plugs were incubated overnight at 37°C in a cell lysis solution (10 mM Tris-HCl [pH 7.2], 50 mM NaCl, 100 mM EDTA, 0.2% deoxycholate, 0.5% N-lauroyl-sarcosine, 1 μg ml⁻¹ lysozyme, and 20 μg of RNase) and washed twice in washing buffer (20 mM Tris-HCl [pH 7.5], 50 mM EDTA) and once with water. After incubation for 1 day at 50°C in proteinase K buffer (100 mM EDTA [pH 8], 0.2% deoxycholate, 0.1% N-lauroyl-sarcosine, and 50 μg ml⁻¹ of proteinase K), the plugs were washed three times for 20 min each time in washing buffer. Gel electrophoresis was done in a Bio-Rad CHEF-DRIII system under the following conditions: one-sixth of the plug; initial switch time (IS), 800 s; final switch time (FS), 800 s; temperature, 13.5°C; field angle,106°; run time, 64 h at 2.2 V cm⁻¹.

Lipid analysis by TLC.

Cultures (1 ml) of R. etli were inoculated from fresh overnight cultures to an A₆₂₀ of 0.1. The cultures were labeled with 1 μCi of [1-¹⁴C]acetate (58 mCi mmol⁻¹; Amersham Biosciences) for 24 h. Lipids were extracted, and the chloroform phase was subjected to lipid analysis in one-dimensional thin-layer chromatography (TLC) plates (high-performance TLC aluminum sheets, Silica Gel 60; Merck) using a chloroform-methanol-glacial acetic acid solvent system as described previously (52).

Transmission electron microscopy (TEM) and cell-length analysis.

Electron microscopy images were obtained from negatively stained preparations on carbon-coated grids. Samples of the desired strains were obtained from cultures grown either in PY medium or in MMY for 9 h, centrifuged, and washed with 10 mM HEPES buffer (pH 8.0). Washed cells were placed over a grid and stained with a 1% uranyl acetate solution for 8 min, dried at room temperature, and observed at 80 kV with a JEM-1200EXII electron microscope (JEOL Ltd., Japan). The length of at least 100 cells for each strain and condition was measured from photographs. Cell length data were ranked every 0.5 μm, and the chi-square test was applied using the CE3 data as the expected values.

Bioinformatic analyses.

To locate genes on p42e that possibly participate in metabolic pathways, the sequence of p42e was subjected to functional automatic annotation of genes and assignation to pathways using the KEGG automatic annotation server (KAAS) (40; http://www.genome.jp/tools/kaas/) running in bidirectional best-hit mode. Results of this analysis were compared to the prediction of metabolic pathways for the complete R. etli CFN42 genome (available in the Metacyc database [7]; http://metacyc.org/) retaining those genes with predictable function, present in p42e and not repeated elsewhere in the genome; the genes chosen were mainly those that participate in central metabolism and cellular functions.

For determination of syntenic relationships, alignments were generated using the predicted protein sequences in each replicon, using the program PROmer, part of the MUMmer program suite (34). Conservation and context analysis of relevant regions was done using the MicrobesOnline database (13) (available at http://www.microbesonline.org/). Searching for distant homologs of predicted proteins was done by PSI-BLAST (1) (available at http://blast.ncbi.nlm.nih.gov/Blast.cgi). Prediction of protein secondary structure and comparison with published protein structures were done at the PSIPRED server (6) (http://bioinf.cs.ucl.ac.uk/psipred/), using pGenTHREADER (35).

RESULTS

A significant fraction of the genes in p42e are involved in primary metabolic functions.

To ascertain which genes in p42e may be involved in primary metabolic functions, genes in this replicon were assigned to metabolic pathways using the KAAS server (40) as described in Materials and Methods. The results of this prediction were compared to the prediction of metabolic pathways for the complete genome of R. etli CFN42, with the aim of identifying those genes in p42e that participate in known pathways and that are not present elsewhere in the genome. Those are shown in Fig. 1.

FIG. 1.

Map of plasmid p42e. Relevant genes are indicated inside the circle. Deletions employed in this work are indicated on the outside arcs.

As befits its plasmid origin, p42e has a replication region (repABC) typical of Rhizobium plasmids (23). Besides the previously recognized minCDE operon, involved in proper septum localization, genes participating in a variety of biosynthetic, degradative, and transport functions were also found. Among these, there is a cobFGHIJKLM operon involved in generation of cobalamin (vitamin B₁₂) through the aerobic pathway, a nadABC operon responsible for the early steps of NAD biosynthesis, a thiMED operon participating in a newly described thiamine salvage pathway (30), a cls gene responsible for cardiolipin synthesis, and the glnT gene, encoding glutamine synthetase III (10). We also found the cyoABCDE operon, encoding the cytochrome O terminal oxidase, one of the three respiratory terminal oxidases present in R. etli.

Regarding degradative functions, p42e harbors an asnRPAB operon, which encodes aspartase and a thermolabile asparaginase (43), part of the pathway for histidine degradation (the hutUGHI operon), genes for melibiose utilization (agpA, agaL1, and agaL2) and two sets of genes that might participate in protocatechuic acid degradation (pcaDCHGB and pcaIJF). Transport functions include a kdpABCD operon, encoding one of the four predicted potassium uptake systems (14), as well as the actP and hmrR genes, involved in copper expulsion. A gene that might be involved in selenocysteine-tRNA formation (selA) was also found.

It is interesting to note that in some of these cases, part of the corresponding pathway is located in p42e, but the rest of the pathway is in another replicon. Examples of this are the cobFGHIJKLM operon, involved in generation of cobalamin, where the rest of the pathway is encoded on the chromosome, the nadABC operon responsible for the early steps of NAD biosynthesis, where the late steps and the salvage pathway for NAD are encoded on the chromosome, as well as the thiMED operon, participating in the thiamine salvage pathway, where the de novo thiamine pathway (thiCOSGE) is encoded on plasmid p42b (39).

Based solely on gene annotation, a significant fraction of the genes in p42e (51 out of 459, 11%) appear to be devoted to primary metabolic functions. Other than the chromosome, this is the second largest genomic element in R. etli containing genes involved in primary metabolism, suggesting that p42e merits the status of secondary chromosome.

Plasmids equivalent to p42e are present in other Rhizobium species.

To merit the status of secondary chromosome, replicons equivalent to p42e should be conserved among related Rhizobium species. In order to demonstrate if this is the case, we made protein sequence alignments of the complete replicons of other sequenced Rhizobium species against p42e, as described in Materials and Methods. As shown in Fig. 2 A, there is an extensive similarity and conservation of gene order between R. etli CFN42 p42e and R. etli CIAT652 pA (22). Synteny between these replicons is broken only in the interval spanning coordinates 120 kb to 200 kb in the p42e sequence. Interestingly, the only spontaneous deletion of p42e (obtained previously during attempts to eliminate p42e) covers this interval (Fig. 1, Δ185). Remarkably, synteny analysis between all the replicons in these two strains revealed extensive similarity only between the respective chromosomes or for the p42e-pA pair and significant, albeit lower similarities, between p42f-pC and p42d-pB (22).

FIG. 2.

Plasmids equivalent to p42e are present in other Rhizobium species. Alignments of R. etli CFN42 plasmid p42e with R. etli CIAT652 plasmid pA (A), R. leguminosarum bv. viciae plasmid pRL11 (B), R. leguminosarum bv. trifolii WSM2304 plasmid pRLG202 (C), and R. leguminosarum bv. trifolii WSM1325 plasmid pR132502 (D) were done using PROmer. The scale is in Mbp.

Extensive similarity of p42e was also found with similar-sized replicons in other Rhizobium species. As shown in Fig. 2B to D, equivalence in order and conservation to p42e was found with pRL11 of R. leguminosarum bv. viciae 3841 (29, 61), pRLG202 of R. leguminosarum bv. trifolii WSM2304 (45), and pRLG132502 of R. leguminosarum bv. trifolii WSM1325 (44). Interestingly, in all these cases, synteny was missing in roughly the same sector (coordinates 120 kb to 200 kb) as in the comparisons between p42e and pA, indicating that this region is dispensable in the corresponding genomes.

Although no extensive synteny was detected between p42e and other replicons in the Rhizobiales, it was previously noted that several gene clusters are shared between p42e and the chromosomes II of A. tumefaciens C58, A. vitis S4, and A. radiobacter K84, as well as with pSymB of Sinorhizobium meliloti (26, 55). Notably, all the genes involved in primary metabolism described in the previous section are located in p42e or equivalent replicons in R. etli and R. leguminosarum biovars viciae and trifolii but in chromosomes or putative secondary chromosomes in other Rhizobiales (see Table S3 in the supplemental material). Taken together, the finding of replicons equivalent in content and gene order to p42e in other Rhizobium species supports our contention that these should be considered secondary chromosomes.

Plasmid p42e evades its elimination through recA-independent cointegration.

Previous attempts to eliminate p42e relied on a positive selection system, based on the use of the sacB gene (4, 5). Although this system has been successful to achieve elimination of the rest of the plasmids in R. etli CFN42 (4, 5), an alternative approach would be to force the loss of p42e through incompatibility. To this end, we cloned, in a conjugative vector replicable in R. etli (pBBR1MCS), the replication sector of p42e (repABC_p42e); the sector chosen includes the two regions known to function as incompatibility factors (8) but lacks a complete repC gene, hence precluding replication from this region. The resulting plasmid should replicate using only the replication system from pBBR1MCS but should exert incompatibility toward p42e. Thus, if elimination of p42e were possible, introduction of this plasmid by conjugation into R. etli would suffice to achieve its dislodgment from the strain.

Introduction by conjugation of both vector pBBR1MCS5 and its derivative harboring repABC_p42e (pBBRrep1) occurred readily with E. coli (Table 1). Interestingly, although introduction of pBBR1MCS5 into either wild-type R. etli CE3 or its recA derivative occurred at frequencies comparable to the ones seen for E. coli, introduction of pBBRrep1 occurred only at vastly lower frequencies (10,000-fold lower than the ones seen with pBBR1MCS5). Since both plasmids share the same conjugative and replicative systems, the restriction in the frequency of transconjugants obtained with pBBRrep1 should be due to difficulties in the maintenance of the incoming plasmid, perhaps caused by a strong selection against the elimination of p42e.

TABLE 1.

Reduced transfer frequency of a plasmid harboring repABC_p42e in Rhizobiales

Recipient strain	Plasmid transfer frequency (10−7) (mean ± SD)a
	pBBR1MCS5b	pBBRrep1c
E. coli DH5α	6,600 ± 3,100	5,800 ± 12,000
R. etli CE3	240 ± 250	0.98 ± 1
R. etli CE3 recA	8,500 ± 1,300	0.1 ± 0.26
R etli CIAT652	51,600 ± 15,000	0.075 ± 0.15
R. leguminosarum bv. viciae 3841	358,000 ± 430,000	1.1 ± 0.81
A. tumefaciens GMI9023	68,000 ± 113,000	0.58 ± 0.57

^aPlasmid transfer frequency is expressed as the number of transconjugant cells with the desired plasmid divided by the total number of output recipient cells. Each datum is the mean of at least three independent determinations.

^bControl plasmid.

^cPlasmid carrying the repABC_p42e operon.

Verification of the transconjugants of both R. etli CE3 and its recA derivative harboring pBBRrep1 by plasmid profiles revealed that, in all cases, a normally sized band corresponding to p42e was absent (Fig. 3 A). The absence of p42e was accompanied by the loss of another plasmid (depending on the transconjugant analyzed), coupled with the finding of new, larger plasmid bands. These changes should be readily explained by invoking cointegration of p42e with other plasmids in the strain, rather than simple losses of p42e. In support of this explanation, Southern blots of the plasmid profiles, hybridized against probes corresponding to each plasmid, revealed the recombinant nature of the new larger plasmid bands (data not shown). Moreover, the presence of p42e in the transconjugants was ascertained by PCR amplification of different sectors scattered over p42e (data not shown).

FIG. 3.

Rearrangements observed in Rhizobium strains harboring plasmid pBBRrepABC_p42e. At the top of each panel, combinations of letters (e.g., b-e) indicate which kind of cointegration has occurred. (A) Plasmid profile of R. etli recA (lane RE) and their recApBBRrep1 derivatives. (B) Chromosomal profile of R. etli CE3 (lane RE) and a recApBBRrep1 derivative (lane e-chr); lane L, molecular size marker (Schizosaccharomyces pombe chromosomes [Bio-Rad]). Chromosomal sizes at the right of this panel are indicated in Mbp. (C) Plasmid profile of R. etli CIAT652 (lane RC) and a CIATpBBRrep1 derivative (lane A-C). (D) Plasmid profile of R. leguminosarum bv. viciae 3841 (lane RL) and a 3841pBBRrep1 derivative (lane 10-11). Letters or numbers on the left of panels A, C, and D mark the locations of plasmids in R. etli CFN42 (panel A), R. etli CIAT652 (panel C), and R. leguminosarum bv. viciae 3841 (panel D). Asterisks indicate the locations of novel cointegrate plasmids.

This analysis allowed us to conclude that, rather than being lost, p42e had been cointegrated with either plasmid p42b, p42c, p42f (Fig. 3A), or p42d (data not shown). Only one isolate appeared to have lost p42e (Fig. 3A, rightmost lane); this isolate, however, still retains most of p42e, as suggested by PCR amplification of different sectors scattered over p42e (data not shown). To evaluate whether cointegration between p42e and the chromosome occurred in this isolate, chromosome size was analyzed by PFGE. As shown in Fig. 3B, chromosome size in this isolate was increased from 4.3 Mb to 4.8 Mb, consistent with the insertion of p42e in the R. etli CE3 chromosome. Thus, p42e evades its elimination by recA-independent cointegration with other replicons.

To ascertain whether plasmids equivalent in order and content to p42e belong to the same incompatibility group, pBBRrep1 was introduced into R. etli CIAT652 and R. leguminosarum bv. viciae 3841. Vector plasmid pBBR1MCS5 was introduced readily in these strains, but transconjugants with the repABC_p42e-harboring derivative (pBBRrep1) were found at only very low frequencies (Table 1). Plasmid profile analyses of these transconjugants revealed that in R. etli CIAT652, pA was cointegrated with pC (Fig. 3C) while in R. leguminosarum bv. viciae 3841, pRL11 was cointegrated with pRL10 (Fig. 3D). These data indicate that plasmids p42e, pA, and pRL11 all belong to the same incompatibility group and evade their elimination by cointegration with other replicons.

Introduction of pBBRrep1 into A. tumefaciens GMI9023 (49) (a plasmid-free strain) also occurred at very low frequencies (Table 1), suggesting that repABC_p42e is incompatible with the only repABC replicon present in this strain, i.e., the A. tumefaciens linear chromosome. However, chromosome sizing of the transconjugants by PFGE failed to reveal any changes in the size of Agrobacterium chromosomes (data not shown).

The inability to eliminate p42e or equivalent replicons by incompatibility, obtaining instead cointegrates with other replicons in the cell, is consistent with the view that loss of these replicons may be lethal even in rich medium, an attribute to be expected in secondary chromosomes.

Participation of p42e in primary metabolic functions.

To ascertain the participation of p42e-localized genes in primary metabolic functions, we inactivated the corresponding genes or operons by vector insertion mutagenesis (VIMS) (see Materials and Methods). We excluded the thiMED (30) and asnRPAB (43) operons and the glnT gene (10) from this analysis, whose functionality was demonstrated previously.

Inactivation of the nadABC operon provokes nicotinate auxotrophy.

The pathway for de novo biosynthesis of NAD in bacteria is the aspartate pathway, encoded by the nadABCDE genes (47). In R. etli, the first three genes are localized in p42e, and the rest are localized on the chromosome. Inactivation of the nadABC operon in R. etli CE3 provoked a nicotinate auxotrophy (see Fig. S1 in the supplemental material), consistent with the participation of the NAD salvage pathway. This phenotype was seen after a previous subculture in minimal medium, due to the need to deplete the endogenous NAD pool.

R. etli CE3 actP is hypersensitive to copper.

Maintenance of copper homeostasis requires a P-type ATPase encoded by actP, which exports copper (46). A mutant on actP in p42e displays hypersensitivity to copper (see Fig. S1 in the supplemental material).

A minCDE mutant is affected in cell length but not in growth.

The products of the minCDE genes participate in the accurate placement of the division site, allowing septum formation in the middle of the cell (50). Hence, we evaluated the growth kinetics and cell length of the CE3 minC mutant by electron microscopy. Despite the fact that growth was not affected by inactivation of this operon (data not shown), significant differences (P ≪ 0.01) were found in distribution of cell length between the mutant and the wild-type strain (see Fig. S2 in the supplemental material). The mutant displayed a shorter cell length than the wild type (median cell length, 1.53 versus 2 μm, respectively); in fact, 13% of the mutant cells were found in the size class of 0.5 to 1.0 μm, whereas this class was absent in the wild type.

A cobalamin auxotrophic phenotype requires mutations in the cobFGHIJKLM operon and in mmuM.

In R. etli, cobalamin biosynthesis proceeds by the aerobic pathway (59). Part of the pathway is encoded on p42e, while the rest is on the chromosome. Cobalamin is a cofactor for essential enzymes, such as methionine synthase and ribonucleotide reductase. Some bacteria have two isozymes for each enzyme, one B₁₂ dependent and another B₁₂ independent (15, 18, 48). R. etli harbors two chromosomally located ribonucleotide reductases, one dependent and another independent of cobalamin. In contrast, there is a potential B₁₂-dependent methionine synthase (metH) in the chromosome, but the classical B₁₂-independent methionine synthase (metE) is absent from this organism. Despite this, a cobG R. etli mutant is still able to grow in minimal medium (see Fig. S3 in the supplemental material). However, R. etli possess a homocysteine S-methyltransferase protein (encoded in RHE_PC00163, localized in p42c) that is an ortholog to the third methionine synthase (mmuM) of E. coli, which uses S-methylmethionine as a methyl donor and zinc as a cofactor (58). Simultaneous inactivation of cobG and mmuM in R. etli leads to an inability to grow in minimal medium, which is corrected upon addition of either cobalamin or methionine to the medium (see Fig. S3 in the supplemental material).

CE3 cyoA mutant is affected in growth in minimal medium under microaerobiosis.

The cytochrome oxidases comprise complex IV in the electron transport chain. The cytochromes of R. etli expressed under free-living conditions are cytochromes aa₃ and o; the former is expressed under aerobic conditions, while the latter is expressed under microaerobic conditions (56). Genes encoding the cytochrome c oxidase (also known as aa₃) are located in the chromosome (ctaBCDE), while the cytochrome o oxidase is encoded in p42e (cyoABCDE). As expected, a CE3 cyoA mutant grows well in both rich and minimal media under aerobic conditions but displays a severe growth reduction in minimal medium under microaerobiosis (see Fig. S3 in the supplemental material).

A CE3 cls mutant fails to produce cardiolipin.

Membrane phospholipids of plant-associated bacteria are mainly phosphatidylglycerol, cardiolipin, phosphatidylethanolamine (PE), and their methylated derivatives (monomethyl-PE and dimethyl-PE) and phosphatidylcholine. Cardiolipin is synthesized from phosphatidylglycerol through cardiolipin synthase (cls) (36). A CE3 cls mutant is unable to produce cardiolipin (see Fig. S3 in the supplemental material). This inability was also seen in strain CFNX185, which also lacks this gene (Fig. 1, Δ185 also, see Fig. S3 in the supplemental material). The inability to produce cardiolipin, however, does not affect growth compared to that of the wild type (data not shown).

No distinctive phenotypes were found in mutants affected in selA, hutI, or pcaD. In the last two cases, this is due to the inability of the wild-type strain to grow on the corresponding metabolites (histidine either as a carbon or nitrogen source; protocatechuic acid as a carbon source). The selA gene codes for an l-seryl-tRNA(Ser) selenium transferase protein, which incorporates selenium into proteins as selenocysteine (11). Although this gene is essential in bacteria that use selenocysteine for recoding, R. etli and related bacteria lack other genes needed for the selenium utilization trait, such as selB, selC, and selD (63, 64).

The results shown in this section clearly demonstrate the participation of several genes located on p42e in different aspects of primary metabolism. Important as these genes are, it is appropriate to stress that none of the mutations analyzed compromise growth on rich medium. Therefore, putative essential genes on p42e must be looked for in other sectors of this replicon.

Deletion analysis of p42e reveals regions essential for growth in R. etli.

In order to find regions on p42e essential for growth on rich medium, we undertook a programmed deletion analysis of this replicon, using a modification of the vector-mediated excision system (2). In this system, the region to be deleted is flanked by loxP sites (introduced by cointegration) in direct orientation; excision of the region of interest is achieved by expression of the Cre recombinase, at frequencies approaching 90% (see Materials and Methods).

Ten deletions affecting predefined sectors of p42e were obtained using this system (Fig. 1, outer arcs). The sizes of the deleted regions range from 27 kb to 447 kb; these deletions, together with the spontaneous deletion harbored by strain CFNX185 (5), provide a nearly complete deletion coverage of p42e. The only regions not covered by deletions include the cobFGHIJKLM operon, the repABC operon, and gene RHE_PE00001 (see below). Nine of these deletions (deletions 1, 2, 3, 4, 5, 9, 10, 11, and 185) do not noticeably affect the growth of the corresponding strains on rich medium. Distinctive phenotypes for strains bearing these deletions were seen only on minimal medium, in a manner consistent with the loss of recognized genes (for instance, the Δ1 strain showed nicotinate auxotrophy and a min phenotype, the Δ185 strain did not produce cardiolipin and was unable to degrade melibiose). The largest deletion strain analyzed was the Δ5 strain (447-kb deletion); thus, it is possible to eliminate nearly 90% of p42e without severely impairing growth in rich medium.

A marked reduction in growth on rich medium was seen for strains harboring either deletion 6 or 7. Upon growth of these strains on rich medium plates, only pinpoint colonies were seen, which took twice as long to acquire a detectable size as colonies formed by the wild-type strain. A precise evaluation of the growth retardation in these strains was hindered both by their slow growth and by the frequent appearance of revertants of unknown origin. Both Δ6 and Δ7 strains have lost a sector encompassing genes RHE_PE00003 to RHE_PE00035 (Fig. 1); since a strain harboring Δ9 (lacking the region comprising RHE_PE00003 to RHE_PE00023) grows normally, we infer that the region responsible for the slow-growth phenotype on rich medium lies in the sector spanning from genes RHE_PE00024 to RHE_PE00035. In fact, deletion of this interval turned out to be impossible to obtain, despite the flanking of this interval with appropriate loxP sites.

To find out which gene on this sector is needed for growth in rich medium, systematic inactivation of the transcriptional units in this interval was done with the VIMS approach (see Materials and Methods). Inactivation of transcriptional units encompassing RHE_PE00025, RHE_PE00026 to RHE_PE00030 (the cyoABCDE operon), RHE_PE00031, RHE_PE00032 to RHE_PE00033, and RHE_PE00034 were readily obtained through this approach. Inactivation of RHE_PE00024 was impossible to obtain by VIMS; interestingly, inactivation of this gene was readily obtained only upon introduction of supernumerary copies of RHE_PE00024. These data indicate that RHE_PE00024 (encoding a probable two-component sensor histidine kinase/response regulator hybrid protein) is an important determinant for growth in rich medium.

Attempts to generate deletions lacking RHE_PE00001 were frustrated by the inability to get insertions on this gene, even by employing the VIMS approach. This inability can be circumvented by introduction of additional copies of RHE_PE00001. As expected, corresponding insertions occurred either on RHE_PE00001 in p42e or in the supernumerary copy. These data are consistent with the proposal that RHE_PE00001 is a gene that is essential for growth of R. etli on rich medium.

Gene RHE_PE00001 encodes a hypothetical protein, containing DUF1612 (a domain of unknown function, conserved in Rhizobiales), as well as a probable winged helix-turn-helix motif (HTH-13). Searching for more-distant homologs using PSI-BLAST revealed similarity with members of the Fic (filamentation induced by cyclic AMP [cAMP]) protein family (data not shown). Interestingly, an analysis of the predicted RHE_PE00001 protein in the PSIPRED server, using pGenTHREADER (data not shown), revealed a structural folding similar to that of the Fic family protein of Shewanella oneidensis (12). The significance of these similarities is explored in Discussion.

DISCUSSION

In this work, we present four lines of evidence in support of our proposal that p42e and related replicons should be considered secondary chromosomes in Rhizobiales: (i) the presence of a significant fraction of genes functionally involved in primary metabolism; (ii) the finding of replicons equivalent to p42e, in content and gene order, in other Rhizobium species; (iii) the inability to eliminate p42e or equivalent replicons by incompatibility; and (iv) the identification of two genes whose presence is needed for normal growth in rich medium.

Regarding the first point, our data show that nearly 10% of the genes in p42e participate in different aspects of primary metabolism, including amino acid biosynthesis and degradation, vitamin and cofactor biosynthesis, membrane lipid formation, sugar utilization, electron transport, and respiratory ability, as well as septum location. Interestingly, several of these aspects reveal a certain degree of functional solidarity among the different replicons, in the sense that part of the pathway is located in p42e and the rest in other replicons, such as the main chromosome. This was clearly evinced for cobalamin, thiamine, and NAD biosynthesis, as well as for septum location.

It is important to note that the number of primary metabolic functions encoded in p42e may be as yet underestimated. In our analysis, we concentrated on genes with a functional annotation which had no additional copies elsewhere in the genome. This leaves out genes for which functional reiteration is commonplace. Among genes on p42e that may fall in this category are those encoding a type 1C penicillin binding protein (pbpC), a peptide methionine sulfoxide reductase (msrAe), four different adenylate cyclases (cyaCe, cyaA, RHE_PE00043, and RHE_PE00045), an extracytoplasmic sigma factor and its corresponding anti-sigma (RHE_PE00003 and RHE_PE00004), a DNA topoisomerase type IB (RHE_PE00209), glutathione S-transferase (RHE_PE00210), and a set of genes that may participate in double-strand break repair via the Ku pathway (RHE_PE00249 to RHE_PE00252). Although the functionality of most of these genes remains to be evaluated, there is evidence that at least the cyaA copy is active in R. etli (57). Analyses of genes that participate in double-strand break repair via the Ku pathway in S. meliloti revealed functional reiteration; genes on the so-called pSymB (orthologs to the ones present in p42e) are required for a full repair activity (32).

The finding of replicons equivalent to p42e in other Rhizobium species is a further argument for its role as a secondary chromosome. Other than the respective chromosome, p42e and their relatives are the largest genomic elements shared between different Rhizobium spp., including R. etli and R. leguminosarum biovars viciae and trifolii. The finding that p42e, pA, and pRL11 belong to the same incompatibility group also argues in favor of a common origin for these replicons.

Our inability to achieve elimination of p42e, pA, or pRL11 by incompatibility curing can be explained by the finding of two loci on p42e (and corresponding replicons) that are essential for growth in rich medium. Based on this view, cells that lose p42e are unable to grow in rich medium. The strong pressure imposed by another replicon of the same incompatibility group in the cell makes it more likely to select low-frequency recA-independent cointegration events of p42e (and pA or pRL11) with other replicons in the cell. These cointegrates should be stable, since p42e and their relatives now employ the replication system from the replicon in which it was cointegrated, while the plasmid used for incompatibility curing harbors a separate replication system (pBBRMCS).

Although the exceptional stability of p42e and its relatives should be explained by their possession of two essential genes, recent data suggest that there are additional factors. Attempts to eliminate p42e providing supernumerary copies of RHE_PE00001 and RHE_PE00024 proved to be unsuccessful (data not shown), suggesting that other factors, such as the presence of an addiction system, may have a role in the high stability of this replicon.

Two regions relevant for growth in rich medium were identified in this work. One of them (RHE_PE00001) encodes a hypothetical protein that contains DUF1612 (a domain of unknown function, conserved in Rhizobiales), as well as a probable winged helix-turn-helix motif (HTH-13). This protein is conserved among all sequenced rhizobial species. As mentioned in Results, this protein shows detectable similarity in both primary structure and folding with members of the Fic protein family. Members of the Fic family are thought to participate together with cAMP in regulatory processes for cell division, probably involving folate metabolism. Some members of this family are death-on-curing (Doc) proteins, involved in toxin-antitoxin systems (12). Intriguing as these similarities are, we believe that RHE_PE00001 does not possess these functions, because it lacks both the well-conserved HPFXXGNG motif (12) and the Fido motif (31), characteristic of this protein family. Thus, it is possible that a Fic folding was recruited on RHE_PE00001 for a different process, perhaps involving DNA interactions.

The second region comprises gene RHE_PE00024, which was annotated as a probable two-component sensor histidine kinase/response regulator hybrid protein, also conserved in the Rhizobiales. Even though its precise function remains unknown, it is plausible to invoke a role as a global regulator, controlling an essential aspect of cell metabolism. In this regard, it is germane to mention that although many two-component systems are dispensable (60, 62), at least the WalK/WalR two-component system in low-G+C Gram-positive bacteria (16) and the two-component sensor histidine kinase/response regulator hybrid protein CckA in Caulobacter crescentus (28) are essential proteins. Although the exact roles of RHE_PE00001 and RHE_PE00024 will be explored in future studies, this work reveals the complexities inherent to infer gene essentiality based solely on sequence data. None of these genes was recognized as essential, based on sequence annotation or through screenings in the Database of Essential Genes (http://tubic.tju.edu.cn/deg/).

Data presented here are consistent with a scenario in which the ancestor of p42e and their relatives was a typical repABC plasmid. Several genes located on the main chromosome underwent migration to this plasmid replicon without gene duplication, thence generating a second essential replicon. The migration process invoked here is probably a consequence of the activity of the recombinational systems of this group of bacteria. The case of p42e and their relatives is notable in the sense that accretion of chromosomal genes was particularly active, entailing the acquisition of some genes essential for growth in rich medium. The exploration of the processes for generation and possible selective advantages of this particular genome architecture constitute an interesting field for further research.