High-resolution transcriptional analysis of the symbiotic plasmid of Rhizobium sp. NGR234

Authors

  • Xavier Perret,

    1. Laboratoire de Biologie Moléculaire des Plantes Supérieures, University of Geneva, 1 chemin de l'Impératrice, 1292 Chambésy, Geneva, Switzerland.,
    Search for more papers by this author
    • †These authors contributed equally to this study

  • Christoph Freiberg,

    1. Institut für Molekulare Biotechnologie, Abteilung Genomanalyse, Beutenbergstrasse 11, 07745 Jena, Germany.
    Search for more papers by this author
    • †These authors contributed equally to this study

  • André Rosenthal,

    1. Institut für Molekulare Biotechnologie, Abteilung Genomanalyse, Beutenbergstrasse 11, 07745 Jena, Germany.
    Search for more papers by this author
  • William J. Broughton,

    1. Laboratoire de Biologie Moléculaire des Plantes Supérieures, University of Geneva, 1 chemin de l'Impératrice, 1292 Chambésy, Geneva, Switzerland.,
    Search for more papers by this author
  • Rémy Fellay

    1. Laboratoire de Biologie Moléculaire des Plantes Supérieures, University of Geneva, 1 chemin de l'Impératrice, 1292 Chambésy, Geneva, Switzerland.,
    Search for more papers by this author

William J. Broughton. E-mail broughtw@sc2a.unige.ch; Tel. (+41) 22 906 1740; Fax (+41) 22 906 1741.

Abstract

Most of the bacterial genes involved in nodulation of legumes (nod, nol and noe ) as well as nitrogen fixation (nif and fix ) are carried on pNGR234a, the 536 kb symbiotic plasmid (pSym) of the broad-host-range Rhizobium sp. NGR234. Putative transcription regulators comprise 24 of the predicted 416 open reading frames (ORFs) contained on this replicon. Computational analyses identified 19 nod boxes and 16 conserved NifA-σ54 regulatory sequences, which are thought to co-ordinate the expression of nodulation and nitrogen fixation genes respectively. To analyse transcription of all putative ORFs, the nucleotide sequence of pNGR234a was divided into 441 segments designed to represent all coding and intergenic regions. Each of these segments was amplified by polymerase chain reactions, transferred to filters and probed with radioactively labelled RNA. RNA was extracted from bacterial cultures grown under various experimental conditions, as well as from bacteroids of determinate and indeterminate nodules. Generally, genes involved in the synthesis of Nod factors (e.g. the three hsn loci) were induced rapidly after the addition of flavonoids, whereas others thought to act within the plant (e.g. those encoding the type III secretion system) responded more slowly. Many insertion (IS) and transposon (Tn)-like sequences were expressed strongly under all conditions tested, while a number of loci other than those known to encode nod, noe, nol, nif and fix genes were also transcribed in nodules. Many more diverse transcripts were found in bacteroids of determinate as opposed to indeterminate nodules.

Introduction

Soil bacteria of the genera Azorhizobium, Bradyrhizobium, Mesorhizobium, Rhizobium and Sinorhizobium (collectively known as rhizobia) are capable of forming symbiotic associations with leguminous plants. Under conditions of limiting soil nitrogen, specialized root structures called nodules may develop. In compatible interactions, rhizobia are released into the nodules, where they form nitrogen-fixing bacteroids which exchange ammonia for nutrients synthesized by the host plant. Symbiotic specificity varies from rhizobia that are devoted to one or a few legumes to Rhizobium sp. NGR234, which nodulates more than 112 genera of legumes (and the non-legume Parasponia andersonii ) (Trinick, 1980; Lewin et al., 1987; Pueppke and Broughton, 1999). Transconjugants of Agrobacterium tumefaciens cured of its Ti plasmid, but harbouring the symbiotic plasmid of NGR234 (pNGR234a ), are able to nodulate Vigna unguiculata (Broughton et al., 1984). This implies that most symbiotic loci are carried on pNGR234a, although homologues of nodEG and nodPQ were later found on the chromosome (Perret et al., 1991). Further analysis revealed that the molecular determinants of promiscuity are clustered in three, dispersed (Perret et al., 1991) hsn (host specificity of nodulation) loci (Broughton et al., 1986; Lewin et al., 1987). In contrast to nod genes, most nif and fix homologues are grouped into a single 55 kb cluster (fixU to y4xE), which has been spared disruption by insertion (IS) or transposon-like sequences (Freiberg et al., 1997). So far, only four of the 24 putative regulators of transcription encoded by pNGR234a have been the subject of detailed investigations: nodD1 (Relićet al., 1993a), nodD2 (Fellay et al., 1998), syrM1 (Hanin et al., 1998) and y4sM (Perret et al., 1994).

Regulation of bacterial and plant symbiotic genes is mediated by the exchange of molecular signals between the two symbionts. Among the many compounds present in root exudates, flavonoids are potent activators of nod gene transcription (Fellay et al., 1995a). Expression of nodulation genes is mediated by NodD proteins, which belong to the LysR family of prokaryotic transcriptional regulators. Although NodD binds to specific promoter sequences called nod boxes even in the absence of an inducer, the presence of flavonoids reinforces this interaction and is required for transcriptional activation (Goethals et al., 1992; Fisher and Long, 1993). In turn, most nod genes are involved in elaboration of another class of bacterial signal molecules, the Nod factors. These lipochito-oligosaccharides permit bacteria to enter the root and induce the formation of nodule meristems (Relićet al., 1993a and b; 1994; Dénariéet al., 1996; D'Haeze et al., 1998; Hanin et al., 1999). Precise modulation of nod gene expression is required in effective symbioses. For instance, NolR, which is present in many rhizobia (Kiss et al., 1998), is needed for optimal nodulation of Medicago sativa by co-ordinating the repression of nodABC in R. meliloti (Kondorosi et al., 1989; Cren et al., 1995). Similarly, down-regulation of the nodABCIJnolOnoeI operon, which occurs after the initial flavonoid induction, is abolished in NGRΩnodD2 and results in modification of the symbiotic properties of the mutant strain (Fellay et al., 1998).

During the later stages of the symbiosis, the expression of many bacterial genes is co-ordinated with nodule morphogenesis via decreasing oxygen pressure in infected nodule cells (Vasse et al., 1990; Fisher, 1994). Although the regulation of nif and fix genes differs between A. caulinodans, B. japonicum and R. meliloti, all require FixL and FixJ (an oxygen-dependent, two-component regulatory system), FixK (a transcriptional regulator whose expression is activated by FixJ), as well as NifA and RpoN. So far, homologues of fixLJ and fixK have not been identified in NGR234, but both the pSym-borne nifA and the chromosomal rpoN genes are required for the formation of fully effective nodules (i.e. Fix+) (van Slooten et al., 1990; R. Fellay, unpublished).

Of the several ways to assay the symbiotic function of the plasmid-borne genes, mutagenesis of 416 open reading frames (ORFs) followed by screening for altered phenotypes on 112 genera of legumes is clearly impractical. Instead, we developed methods of identifying those genes that are differentially expressed during the symbiosis. Previous transcriptional analyses of entire symbiotic replicons were based on the physical maps of pNGR234a (Perret et al., 1991) and the pSym of R. etli strain CFN42 (Girard et al., 1991). DNA of overlapping cosmids was digested with restriction enzymes and probed with radioactively labelled RNA of bacteroids or rhizobia grown under various environmental conditions (Fellay et al., 1995b; Girard et al., 1996). Unfortunately, it was not possible using these methods to determine which of the genes found on a large hybridizing restriction fragment was actively transcribed. To obtain higher resolution, pNGR234a was divided into 441 segments selected to represent all ORFs as well as their intergenic regions. After polymerase chain reaction (PCR) amplification, these segments were separated on agarose gels, transferred to membranes and probed with radioactively labelled RNA.

Results

Transcription of some nodulation loci, such as the nodABC genes, can be detected a few minutes after the addition of flavonoids to the growth medium. Time course experiments showed that the expression of most nod genes reaches a maximum after 1 h of induction with flavonoids and, as in the cases of the nodABCIJnolOnoeI and nodSU transcripts, was no longer detected 24 h later (Fellay et al., 1995b). To compare our results with those of earlier studies based on competitive RNA hybridizations, liquid cultures of NGR234 were grown in the presence of daidzein for 1 h and 24 h. Genes expressed in nodules were identified by labelling RNA extracted from bacteroids purified from plants that form determinate (Vigna unguiculata ) as well as indeterminate nodules (Cajanus cajan ).

Reproducibility and sensitivity of the technique

Background signal artefacts were excluded from the analysis by comparing the positions of PCR products on filters with those of hybridization signals on autoradiographs (Fig. 1). To facilitate interpretation of the data, hybridization intensities were classified into seven categories ranging from none (0) to maximum (M) (see Fig. 1; Fig. 2). In the absence of a reliable internal standard (e.g. a gene whose transcription remains constant in free-living cells as well as in bacteroids), the autoradiographs were exposed until the signal of some loci reached maximum intensity (see Fig. 1, panel sKL to ft1). In this way and in all experiments, several DNA fragments produced signals of similar intensities. Reproducibility of the method was assessed in various ways. Most experiments, including controls in which the cells were grown in liquid cultures without flavonoids, were repeated twice. No significant differences in the hybridization patterns of duplicate experiments were observed, and the results are consistent with previously published hybridization data (Fellay et al., 1995b; Freiberg et al., 1997; Rochepeau et al., 1997; Viprey et al., 1998). Furthermore, observations made using β-galactosidase assays (Berck et al., 1999; Fellay et al., 1998; Hanin et al., 1997) and β-glucuronidase promoter fusion studies (A. Del Greco and X. Perret, unpublished) fully supported the data shown in Fig. 2.

Figure 1.

. Examples of the transcription analyses performed on pNGR234a.The upper panels correspond to ethidium bromide-stained gels of PCR products separated on 1% agarose gels. Letters above each column represent the first and last PCR fragment shown in a panel. The lower panels correspond to hybridization signals in: NI, NGR234 cells grown at 27°C in liquid Rhizobium minimal medium supplemented with succinate (RMM); I-1, cells grown in RMM followed by a 1 h induction with 2 × 10−7 M daidzein; I-24, RMM-grown rhizobia collected 1 day after induction with daidzein; NV, bacteroids purified from Fix+ nodules of V. unguiculata inoculated with NGR234; NC, bacteroids purified from Fix+ nodules of Cajanus cajan inoculated with NGR234; and SH, fragments probed with NGR234 sequences purified by subtractive DNA hybridization against genomic DNA of R. fredii strain USDA257. Intensities of hybridization signals ranged from none (0) to maximum (M). For example, in row NC, panel fq4 to qH, values of 0, 3, 4, 5 and 4 were attributed to signals from left to right respectively. In contrast, the signal in the second lane of the adjacent panel sKL to ft1 (row NC) corresponds to the maximum intensity (M). Non-specific hybridization to PCR by-products are marked with arrows.

Figure 2.

. Transcription map of pNGR234a. Genes (e.g. bF in segment B) and gene fragments (e.g. fb1 in segment B) are represented by arrows showing the direction of transcription and coloured depending on their predicted function. The position and orientation of putative symbiotic promoters is shown with small coloured arrows: nod boxes are marked in red and numbered as in Table 1, NifA-σ54 sequences are shown in yellow. Size and position of PCR-amplified fragments corresponding to gene(s) or intergenic regions are given directly below the genetic map. Overlapping fragments are displaced with respect to adjacent products. Colour-coded intensities of expression range from white (0 = no signal) to purple (M = maximum) in seven steps. RNA probes were prepared from: NI, NGR234 cells grown at 27°C in liquid Rhizobium minimal medium supplemented with succinate (RMM); I-1, cells grown in RMM followed by a 1 h induction with 2 × 10−7 M daidzein; I-24, RMM-grown rhizobia collected 1 day after induction with daidzein; NV, bacteroids purified from Fix+ nodules of V. unguiculata inoculated with NGR234; NC, bacteroids purified from Fix+ nodules of Cajanus cajan inoculated with NGR234; and SH, the 441 fragments were hybridized against pooled sequences made specific by subtractive DNA hybridization against genomic DNA of R. fredii strain USDA257. Hybridization intensities are shown in shades of grey.

Although expression of 169 ORFs was not detected, radioactively labelled RNA hybridized to about 60% (247) of the 416 genes and gene fragments (Fig. 2). Undoubtedly, many factors, including the half-life of the individual mRNAs, affect signal intensity. Also, actively transcribed genes that are duplicated on another replicon add to the hybridization signals. An example of this kind of interference is obvious in the case of dctA1, which is part of an operon that begins with nifQ and is controlled by a NifA-σ54-dependent promoter. This locus is expressed in nodules (Fig. 2, segment V) but, after induction with daidzein, the weak signals detected with the dctA1 fragment are probably caused by transcripts from the chromosomal copy (dctA2) (Fig. 2, I-1 and I-24). Fortunately, non-specific cross-hybridizations are restricted to nearly identical DNA sequences. For example, ORFs y4aM and y4wC encode putative transcriptional regulators of 69% identity and 89% similarity but y4aM (Fig. 2, segment A, I-1 and I-24) is induced early and expressed in V. unguiculata nodules, while y4wC is induced late and is expressed in both determinate and indeterminate nodules (Fig. 2, segment W, NV and NC). It thus seems as though these techniques can discriminate between transcripts that show 82% identity over more than half the length of the genes.

Expression of pNGR234a genes in non-induced conditions

Many of the 22 ORFs expressed in free-living cells (Fig. 2, NI) are also transcribed under all conditions. Among these are repC (y4cK), which is involved in the replication of pNGR234a, y4xL, a protein excreted by the type III secretion system (TTSS)(Viprey et al., 1998), a protein of 272 amino acids of unknown function (y4pM) as well as products of IS-like sequences (y4iO and y4iP; y4pG and y4sC of NGRIS-2a and -2b respectively; fs5; y4sN and fs4; y4uE). Other loci, such as y4lF, y4lG, y4lH and y4lL, are only expressed in non-induced cultures (as well as in cells exposed to a flavonoid for 1 h), suggesting that they might contribute to growth under free-living conditions.

Flavonoid induction of transcription

The addition of flavonoids to the growth medium dramatically changed the patterns of gene expression (Fig. 2, I-1 and I-24). Daidzein enhanced the transcription of approximately 147 previously silent ORFs. Of these, 14 belong to three classes of insertion sequences (NGRIS-3, IS-4 and IS-5). As these ISs are present in multiple and identical copies in pNGR234a as well as in other replicons, it is clearly not possible to determine which of these elements carry the activated gene(s).

Except for several ORFs linked to IS elements, most of the strongly induced genes are found in the three hsn loci and the cluster encoding the type III secretion system (TTSS). With the exception of y4xL, which is thought to be constitutively expressed (Viprey et al., 1998), the expression of all ORFs between y4xI and y4yS occurred 24 h after induction with daidzein. In contrast, expression of most nod genes (marked in red in Fig. 2) reached a maximum after 60 min of induction and declined later. This is particularly clear in the case of the nodABCIJnolOnoeI operon (Fig. 1, I-1 and I-24, first column), transcription of which is repressed in a nodD2-dependent manner 24 h after flavonoid induction (Fellay et al., 1995a; 1998). Interestingly, levels of nodD1 transcripts were increased significantly after flavonoid induction (Fig. 2; I-1, segment A). A similar observation was made in R. fredii USDA191, a strain closely related to R. fredii USDA257. As in NGR234, nodD1 of USDA191 and USDA257 are immediately downstream of a conserved nod box-like sequence (Table 1 and Fig. 2[link]; nod box no. 3). Although present in the reverse orientation, NodD1 and NodD2 bind to this motif and modulate the expression of nodD1 of USDA191 (Machado et al., 1998).

Table 1. . List of predicted nod boxes contained in pNGR234a. Sequences of nod boxes found on pNGR234a are centred around the given positions (Position). Depending on whether they are found on the plus or minus DNA strand, their orientation (Orientation) is marked as + or − respectively. nod boxes are numbered from 1 to 19 by order of appearance, starting at base one of segment A (as in Freiberg et al., 1997).a. Detection of transcriptional activity 1 h (1 h) and 24 h (24 h) after induction with flavonoids (expression levels as in Fig. 2).b. Number of mismatches compared with the consensus sequence.c. Distance to the next downstream gene present in proper orientation (Gene).d. No gene was found in the proper orientation downstream of the nod box.e.nod box consensus sequence as in van Rhijn and Vanderleyden (1995), with R for either A or G; Y replacing U, T or C; N is for any base.Thumbnail image of

Computational analyses of the pNGR234a sequence revealed 19 nod box-like regulatory sequences, at least five of which (nod box nos. 2, 4, 7, 8 and 12; Table 1; Fig. 2[link]) regulate the expression of known nod loci. With the exception of nod boxes nos. 10 and 16, the transcription patterns of downstream ORFs suggest that most of these regulatory elements are functional. The apparent lack of promoter activity of nod boxes 10 and 16 is probably not the result of accumulated mutations, however, as major differences between the DNA sequences of these elements and consensus nod boxes (van Rhijn and Vanderleyden, 1995) were not found (Table 1). In contrast, nod box no. 13 appears to have retained regulatory activity, despite the absence of properly oriented downstream genes. That one of the three identical, pSym-borne copies of NGRIS-5 is adjacent to nod box no. 13 implies that a gene was probably separated from its promoter by rearrangements in the genome.

Transcription of y4gF, y4gG, y4gH and y4gI, which are probably involved in the biosynthesis of rhamnose, was only detected 24 h after induction with daidzein. As with some other flavonoid-inducible loci (e.g. y4fA, y4fB, y4iH, y4jE), their regulation cannot be directly associated with nod boxes. This points to the existence of alternative or indirect regulatory pathways, some of which nonetheless require a functional NodD1. For instance, some genes of the TTSS locus are thought to be under the control of y4xI, a nod box-controlled regulator of transcription (Viprey et al., 1998).

Gene expression in nodules

Important changes in gene expression accompany the differentiation of rhizobia into nitrogen-fixing bacteroids (Fig. 2, NV and NC). Principally, these involve repression of the majority of flavonoid-inducible transcripts and induction of those required for effective nitrogen fixation. Because of the marked sensitivity of nitrogenase to oxygen and the high energy requirements of nitrogen fixation, the expression of nif and fix loci is tightly controlled (for review, see Fischer, 1994). Unlike nod genes, which are dispersed over the complete pSym, nif and fix genes form a 55 kb cluster (fixU to y4xE) that contains 10 NifA-σ54-dependent promoters. Transcriptional analyses suggest that many more loci are specifically expressed in bacteroids, however. Some, such as the cytochrome P450 operon (y4kR to ‘y4lD), are under the control of one of the six other NifA-σ54-like sequences. Two opposing operons, y4nG to y4nJ and y4nL to y4nM, which possibly encode functions linked to sugar metabolism, are also regulated in a similar manner.

As mentioned above, flavonoid-inducible nod boxes do not necessarily precede nod genes. Similarly, a number of NifA-σ54-independent genes were actively expressed in nodules. These include two co-transcribed genes, y4qG and y4qF, which encode a class III pyridoxal-phosphate-dependent aminotransferase and a peptidase of the S9A family respectively. Based on the intensity of hybridization signals, both ORFs are expressed at levels as high as those of the nitrogenase structural genes (see Fig. 1, NV, NC, panels fq4 to qH and vK to vM), suggesting an important role in the functioning of determinate and indeterminate nodules. Although y4sO is clearly homologous to y4qF (70% identity and 92% similarity at the amino acid level), y4sO is apparently not transcribed in nodules.

Differential gene expression in determinate and indeterminate nodules

Significant differences in gene expression were found when comparing the transcription patterns in bacteroids of V. unguiculata with those of C. cajan (Fig. 2, NV and NC). Except for genes directly involved in nitrogen fixation (e.g. nif, fix, and cytochrome P450 loci), which are transcribed in both nodule types, more than 20 ORFs were only expressed in determinate nodules. Among these were nodD1, y4aN (homologous to syrB of R. meliloti ), y4aO, a homologue of A. rhizogenes repB (y4cJ), y4oP to y4oR (coding for components of an ABC transporter system), two pioneer ORFs (y4oV and y4oW), otsB (involved in trehalose synthesis), two co-transcribed genes (y4qC and y4qB), several IS-like sequences (fe7, y4qE, y4rG and y4rJ), y4rN (coding for a transmembrane protein homologous to an antibiotic resistance determinant of Mycobacterium smegmatis ), a transcription regulator (y4xQ), as well as nolU and nolV of the TTSS cluster. In contrast, fewer genes were clearly expressed in C. cajan nodules. Among them are an oxidoreductase (y4hM), peptidases (y4nA and y4tI), a transposase (y4nE), a threonine dehydratase (y4tJ), an ABC transporter binding protein (y4wM) and RhcN, whereas y4mC and y4wO did not have any homologous database entries.

Subtractive DNA hybridization

To identify DNA sequences of NGR234 that are not found in the genome of the closely related strain R. fredii USDA257, a pool of Sau3AI fragments was purified by subtractive DNA hybridization (Perret et al., 1994). Probing of the filter set with these radioactively labelled sequences showed that many of the 441 fragments hybridized (Fig. 2, SH). Signals found with the nifD and nolX fragments (Fig. 2, SH, segments X and Y) raise questions about the ability of the technique to eliminate every common sequence, especially as the nolX fragment of NGR234 is 98% identical to that of the PCR fragment from USDA257.

Nonetheless, many of the 150 hybridizing fragments (i.e. those absent from USDA257) were expected from the known differences between the two genomes. Among these were nolL and noeE (Fig. 1, SH, panel hA to hJ), which are responsible for the acetylation and sulphatation of NodNGR factors, respectively (Hanin et al., 1997; Berck et al., 1999), as well as a series of ORFs that comprise fixF to y4hA (Fig. 2, SH, segment G and H). Others include the NGRRS-1 locus, of which the clustered NGRIS-2, NGRIS-4 and NGRIS-10 elements are absent from USDA257 (Perret et al., 1997). In fact, over 20% of the fragments hybridizing to subtracted sequences correspond to IS- and Tn-like elements, suggesting that a significant proportion of the differences between the NGR234 and USDA257 genomes comprises mobile elements.

Downstream of hsnII, a cluster of genes unique to NGR234 seems to be involved in the transport of sugars and other compounds (y4mI, y4mJ and y4mK), whereas another set of genes (y4tI to y4tN) encodes various classes of enzymes. Other unique genes are possibly under the control of NifA-σ54 promoters. These include y4nG to y4nI, which encode a putative nucleotide sugar epimerase, as well as two transmembrane proteins. Genes encoding other general metabolic functions seem to occur only on pNGR234a, including an aminotransferase (y4uB) and y4uA, an enzyme involved in the biosynthesis of cell wall components.

Discussion

Although sequences unique to NGR234 represent a sizeable fraction of the pSym, most genes found on pNGR234a are also present in USDA257. Obviously, unique Sau3AI fragments are not clustered in a single locus. Rather, they form dispersed islands, highlighting the dynamic structure of this plasmid. Large repeats, IS and Tn sequences are possibly involved in this phenomenon. The transcriptional activation of ORFs belonging to IS elements during flavonoid induction or in bacteroids suggests that transposition may occur at higher frequencies during symbiosis. Apart from the known differences in nod genes (Pueppke and Broughton, 1999) and IS elements (Perret et al., 1997), USDA257 also lacks various enzymes, some of which are expressed in nodules (e.g. y4qB to y4qD). Numerous fragments that hybridize to sequences unique to NGR234 also carry genes of unknown function, many of which are not expressed under the conditions tested.

Similarly, a large number of predicted ORFs apparently remain silent. Perhaps the methods used failed to detect very low-level gene expression, but the conditions tested here represent only a small subset of those encountered by NGR234 during its life cycle. Nonetheless, flavonoid-inducible and bacteroid-specific genes make up 60% of the 441 ORFs and gene fragments encoded by pNGR234a. Several large intergenic regions, such as fa1-nolK, y4cH-y4cI, y4eC-y4eD, y4qC-y4qD and fdxB-y4wA, encode transcripts induced by flavonoids that may or may not be found in bacteroids. Analysis of these regions using a new and more sensitive GeneMark matrix compiled from the genes of pNGR234a failed to detect ORFs however. This is possibly because of genes that: (i) use very distinct codons; (ii) have ORFs that are too short to be detected using current programmes; or (iii) encode symbiotically active but non-translated RNA species (this has been described in B. japonicum ; Ebeling et al., 1991).

In general, the timing of expression of individual genes correlates well with the presence of nod box or NifA-σ54-regulatory sequences. Genes dependent on nod boxes are flavonoid inducible and expressed early in the symbiosis, whereas NifA-σ54-dependent loci are activated in bacteroids. Nevertheless, y4vC is apparently controlled by both (nod box no. 14 and a NifA-σ54-regulatory sequence). Remarkably, Nod factor biosynthetic genes are not organized into a singular regulon. Although most nod gene promoters contain nod boxes, expression of the nodABCIJnolOnoeI and nodSU operons is abolished 24 h after induction with daidzein, whereas transcription of noeE and nolL can still be detected in V. unguiculata nodules. Obviously, several parallel and probably interdependent regulatory pathways control the expression of symbiotically active genes. Daidzein induction of expression of the transcriptional regulators nodD1, y4aM, y4fQ, nodD2, y4xI and syrM2 (the last two of which are under the control of active nod boxes), and the increased transcript levels of y4qH and y4wC found in bacteroids, highlights this complexity.

Unfortunately, the charged membranes used for hybridization could only be reused once, severely limiting the number of analyses that could be performed. Thus, we were unable to analyse transcription patterns in several different determinate and indeterminate nodules. Nevertheless, the available data point to important differences between bacteroids of V. unguiculata and C. cajan nodules. Transcription of a number of nod genes, a gene involved in plasmid replication, ORFs encoding transporters of sugars as well as amino acids/peptides, a regulator of transcription of various biosynthetic and respiratory enzymes suggest that V. unguiculata bacteroids have a higher overall metabolic activity than those found in C. cajan nodules. Although fewer ORFs are expressed in C. cajan nodules, some are nonetheless specific to this plant, suggesting that bacteroid metabolism varies with the host.

Despite its current limitations, our system of transcriptional analyses is a solution to the problem of quickly and accurately delimiting important symbiotic loci and their regulatory elements. Separation of PCR products on gels before Southern blotting followed by careful recording of the positions of correct fragments on membranes helped to eliminate false-positive signals (marked with arrows in Fig. 1). A necessary refinement to the technique would be to produce membranes (or arrays) that could be probed many times. This would permit analyses of regulatory pathways by performing expression studies with RNA extracted from, for example, nodD1, nodD2, syrM1 and nifA mutants. Analyses could also be made using RNA extracted from NGR234 cells grown under low oxygen or high salinity, or induced with root exudates rather than purified flavonoids. Yet the present technique will help to dissect symbiotic molecular genetics. All 19 nod boxes will be cloned, fused to an appropriate reporter gene and their activity in the presence of different flavonoids and/or root extracts assayed. Genes such as y4aN, y4aO, y4qF and y4qG, as well as the regulators of transcription y4aM, mucR and y4qH can be mutated and their effect on symbiosis in determinate versus indeterminate nodules assayed. Loci y4gF to y4gI, which encode enzymes involved in rhamnose synthesis, are transcribed later than most nodulation genes. As mutation of fixF leads to the abolition of rhamnose-rich polysaccharides and ineffective nodules on V. unguiculata (Jabbouri et al., 1996), the symbiotic role of this biosynthetic pathway could be examined using classical methods.

Experimental procedures

Molecular and microbiological techniques

E. coli recombinants were grown at 37°C in Luria–Bertani medium, Terrific broth or in two-fold YT medium (Sambrook et al., 1989). Strains of Rhizobium were raised at 27°C in/on Rhizobium minimal medium supplemented with succinate (RMM; Broughton et al., 1986). Flavonoid induction of nod genes was accomplished by adding daidzein to liquid cultures of NGR234 to a final concentration of 200 nM 1 h and 24 h before harvesting the cells. Cosmid or phage M13 DNA, as well as rhizobial genomic DNA, was prepared using standard procedures (Perret et al., 1991; Freiberg et al., 1997). Sequence analysis of M13 recombinants was performed as described previously (Freiberg et al., 1997).

Construction of the ORF map

The 536 165 bp sequence of pNGR234a was divided into 441 segments representing the 416 predicted genes as well as all intergenic regions. Segments were amplified using specific oligonucleotide pairs (primers of 18–20 bases) and cosmid (Perret et al., 1991) or M13 genomic DNA as templates in standard PCR reactions. Criteria for selecting the target sequences for PCR amplification included the following: (i) fragment size was limited to approximately 1 kb (smallest = 567 bp; to largest = 1953 bp; mean = 1121 bp) to minimize intensity differences in the subsequent hybridizations; (ii) adjacent primers were optimized to avoid overlapping PCR products and to minimize the distance between each amplified fragment (mean gap size = 95 bp); and (iii) for putative operons, 5′ primers were placed upstream of predicted transcription start points. In some cases, it was necessary to use inserts of M13 phages from the pSym sequence library were used in place of the PCR products. The 3′ extremities of M13 clones were verified by sequencing to avoid selection of co-ligation artefacts and to minimize overlap with adjacent, amplified fragments. Before Southern transfer, the size, quantity and overall quality of amplified products were verified. Standardized amounts of DNA were separated on 1% (w/v) agarose gels and vacuum blotted onto GeneScreen Plus nylon membranes (DuPont NEN). Separation of the target fragment on gels permitted accurate determination of the hybridizing band and elimination of false-positive results.

RNA extraction, labelling and hybridizations

Cultures of cells were collected by centrifugation when the absorbance at 600 nm reached 0.4–0.5. Bacteroids were isolated from effective nodules of V. unguiculata and C. cajan grown in large Leonard jars (Pueppke and Broughton, 1999) for 6–8 weeks after inoculation with NGR234. After collection, the nodules were crushed in liquid nitrogen and resuspended in sterile water. Debris was removed by filtration, and bacteroids were recovered by centrifugation (4000 × g for 5 min). Purification, labelling of RNA from bacteroids and rhizobia with [γ-32P]-ATP as well as hybridization were performed as described previously (Fellay et al., 1995b), except that unlabelled competitor RNA was not added to the prehybridization solutions. Pools of DNA sequences purified from NGR234 by subtractive DNA hybridization against R. fredii USDA257 were also obtained and labelled with 32P (Perret et al., 1994). Filters were exposed to X-ray films for 2–21 days depending on the intensity of the signal.

Footnotes

  1. †These authors contributed equally to this study

Acknowledgements

We wish to thank D. Gerber and S. Relić for their help in many aspects of this work. We are grateful to S. Jabbouri for critical reading of the manuscript. Financial assistance was provided by the Fonds National Suisse de la Recherche Scientifique (grant no. 31-45921.95) and the University of Geneva.

Ancillary