Multiple Recombination Events Maintain Sequence Identity Among Members of the Nitrogenase Multigene Family in Rhizobium etli

Corresponding author: David Romero, Depto. de Genética Molecular, Centro de Investigación sobre Fijación de Nitrógeno-UNAM, Apartado Postal 565-A, 62100 Cuernavaca, Morelos, México, dromero{at}cifn.unam.mx (E-mail).

Communicating editor: P. L. FOSTER

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

A distinctive characteristic of the Rhizobium genome is the frequent finding of reiterated sequences, which often constitute multigene families. Interestingly, these families usually maintain a high degree of nucleotide sequence identity. It is commonly assumed that apparent gene conversion between reiterated elements might lead to concerted variation among members of a multigene family. However, the operation of this mechanism has not yet been demonstrated in the Rhizobiaceae. In this work, we employed different genetic constructions to address the role of apparent gene conversion as a homogenizing mechanism between members of the plasmid-located nitrogenase multigene family in Rhizobium etli. Our results show that a 28-bp insertion into one of the nitrogenase reiterations can be corrected by multiple recombination events, including apparent gene conversion. The correction process was dependent on the presence of both a wild-type recA gene and wild-type copies of the nitrogenase reiterations. Frequencies of apparent gene conversion to the wild-type nitrogenase reiterations were the same when the insertion to be corrected was located either in cis or in trans, indicating that this event frequently occurs through intermolecular interactions. Interestingly, a high frequency of multiple crossovers was observed, suggesting that these large plasmid molecules are engaging repeatedly in recombination events, in a situation akin to phage recombination or recombination among small, high-copy number plasmids.

GENE duplication, leading to the formation of multigene families in eukaryotes, has been widely recognized as a mechanism for the generation of new functions (OHNO 1970 ; JOHN and MIKLOS 1988 ). Interestingly, members of multigene families tend to vary in a concerted way, keeping a high similarity between their members (JOHN and MIKLOS 1988 ; DOVER 1993 ). Concerted evolution between members of tandemly-arranged multigene families in eukaryotes has been amply documented. Conservation in a nucleotide sequence is thought to occur predominantly through frequent unequal exchanges between its members (PETES 1980 ; SZOSTAK and WU 1980 ; WILLIAMS and STROBECK 1985 ). A consequence of this mechanism is the frequent expansion and contraction of the tandem array. However, recent determinations show that an alternate mechanism, gene conversion, plays a major role in achieving homogenization in tandem multigene families (GANGLOFF et al. 1996 ). Gene conversion has been defined as the non-reciprocal transfer of sequence information between homologous or homeologous DNA sequences. Frequent events of gene conversion are also responsible for concerted evolution between members of dispersed gene families (JACKSON and FINK 1981 ; KLEIN and PETES 1981 ). Unequal exchanges are less adequate to achieve homogenization in these cases, due to the high likelihood of rearrangement of single-copy DNA flanked by dispersed members.

Studies about the occurrence of similar processes in prokaryotes have been hindered by the paucity of reiterated elements in enterobacterial genomes. A recent determination, based on the sequence of the whole Escherichia coli genome (BLATTNER et al. 1997 ), shows that reiterated elements constitute about 2.5% of the genome. The most conspicuous families of long reiterated elements in E. coli and Salmonella typhimurium are the rrn operons, tuf genes, and different types of insertion sequences (BACHELLIER et al. 1996 ; DEONIER 1996 ; BLATTNER et al. 1997 ), which are commonly arranged as dispersed reiterations. Typically, a high level of nucleotide sequence identity is observed among members of each family. Several reports indicate that homogenization between the reiterated rrn operons in E. coli (HARVEY and HILL 1990 ), duplicated flagellin genes (OKAZAKI et al. 1993 ) or the tuf reiterations in S. typhimurium (ABDULKARIM and HUGHES 1996 ) may be achieved through apparent gene conversion.

Extensive DNA reiteration is found in the genomes of bacteria belonging to the symbiotic nitrogen-fixing genus Rhizobium (reviewed by ROMERO et al. 1997 ). These genomes may carry as much as 700 representatives of long, reiterated elements, belonging to 200 different families (FLORES et al. 1987 ). This high level of reiteration has been observed both in the chromosome and in the large plasmids that are typical of this genus (FLORES et al. 1987 ; GIRARD et al. 1991 ). In fact, an analysis of the sequence of the whole symbiotic plasmid (pSym, 536.1 kb) of Rhizobium sp. NGR234 reveals that reiterated elements constitute about 18% of the plasmid genome (FREIBERG et al. 1997 ).

Besides transposable elements, multigenic families have been observed for several housekeeping genes, such as the fla (BERGMAN et al. 1991 ), ftsZ (MARGOLIN and LONG 1994 ), groEL (FISCHER et al. 1993 ), rpoN (KUNDIG et al. 1993 ) and citrate synthase genes (PARDO et al. 1994 ; HERNANDEZ-LUCAS et al. 1995 ). Several genes involved solely in the symbiotic process are also reiterated, such as nifHDK (BADENOCH-JONES et al. 1989 ; NOREL and ELMERICH 1987 ; QUINTO et al. 1985 ), fixN (DAVID et al. 1987 ; SCHLUTER et al. 1997 ) and the nodD genes (SCHULTZE et al. 1994 ). The reiteration mode is usually dispersed; only the fla gene reiterations are arranged in tandem (BERGMAN et al. 1991 ).

High levels of nucleotide identity are common between these reiterations. A considerable fraction of the reiterated class (about 70%) in the pSym of Rhizobium sp. NGR234 is comprised of identical reiterations (FREIBERG et al. 1997 ). Other multigenic families, such as fla (BERGMAN et al. 1991 ), nifHDK (BADENOCH-JONES et al. 1989 ; NOREL and ELMERICH 1987 ; QUINTO et al. 1985 ) and the citrate synthase genes (PARDO et al. 1994 ; HERNANDEZ-LUCAS et al. 1995 ) also exhibit high identities (over 95%) in nucleotide sequence. A trivial explanation for the occurrence of such high levels of identity would be a relatively recent evolutionary origin. This explanation seems unlikely, because (i) in some cases, such as the nifHDK family of Rhizobium etli, the sequences are ancient enough to be present in every member of the species (SEGOVIA et al. 1993 ), and (ii) although high levels of nucleotide sequence identity are seen between reiterations within a species, a lower level is seen between species themselves. These data have led to the suggestion that some homogenizing mechanism, conceivably gene conversion, is operating between reiterated sequences in Rhizobium.

So far, no studies have been published about the occurrence of gene conversion-like events between reiterations in Rhizobium. The only data concerning this phenomenon were obtained during a study of phage crosses in Rhizobium meliloti, where apparent gene conversion is claimed to occur at a low frequency (OROSZ et al. 1980 ), close to the one for spontaneous mutation. If the low frequency observed also applies to recombinational interactions between reiterations, apparent gene conversion would be an inefficient homogenizing mechanism in Rhizobium.

In this work, we address the role of apparent gene conversion as an homogenizing mechanism between members of the nitrogenase multigene family in R. etli. Our results indicate that multiple recombination events, including apparent gene conversion, play an important role in maintaining sequence identity among members of this family.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Bacterial strains, plasmids and media:
The bacterial strains and plasmids employed are listed in Table 1. E. coli strains were grown at 37° in luria broth medium (MILLER 1972 ), and R. etli strains were grown at 30° in PY medium (NOEL et al. 1984 ). Antibiotics were added at the following concentrations: kanamycin (Km), 30 µg ml^-1 (E. coli) or at variable concentrations depending on the purpose (R. etli, see below); nalidixic acid (Nal), 20 µg ml^-1 (R. etli); spectinomycin (Sp), 100 µg ml^-1 (E. coli or R. etli); and tetracycline (Tc), 5 to 10 µg ml^-1 (R. etli) or 10 µg ml^-1 (E. coli).

Plasmid construction:
All DNA manipulations were carried out under standard protocols (SAMBROOK et al. 1989 ) using restriction enzymes, bacterial alkaline phosphatase and T4 polynucleotide ligase (Amersham Ltd., England). To construct a hybrid Km^r/Sp^r cassette, a 1074-bp fragment, containing the Sp^r gene from pHP45-Sp (Table 1) was obtained through PCR amplification employing oligonucleotides o4 and o5 (Table 2); these oligonucleotides contain BglII sites. This PCR product was digested with BglII and ligated into BamHI-restricted pSUP5011 (a plasmid containing Tn5-mob, SIMON 1984 ), replacing the 1.6-kb mob fragment. The resulting plasmid, which carries a Tn5 Km/Sp, was called pCRS1. From pCRS1, digestion of the BglII sites already existing in the IS50s led to the release of a 3814-bp BglII fragment, containing a promoterless Km^r gene and the Sp^r gene with its own promoter.

To facilitate further subcloning steps of this fragment, the polylinker-containing vector pIC20H (MARSH et al. 1984 ) was digested with SacI and religated, removing a segment carrying XhoI, BglII, XbaI, EcoRV, ClaI and EcoRI sites from the polylinker; this plasmid was called pCRS2. The 3814-bp BglII fragment from pCRS1 was ligated into the unique BamHI site of pCRS2, generating pCRS3. From pCRS3, the promoterless Km^r gene and the Sp^r gene with its own promoter were excised as a 3864-bp HindIII fragment, through digestion of two HindIII sites on the polylinker.

To generate a transcriptional fusion between this fragment and the nifD gene from R. etli, plasmid pEM15 (MORETT et al. 1988 ) was used. This plasmid is a derivative of pSUP205 (SIMON et al. 1983 ), containing one of the nifHDK operons of R. etli as an EcoRI fragment (nifHDK region a). Ligation of the 3864 bp HindIII fragment from pCRS3 into one of the HindIII sites present in pEM15 provokes an interruption of the nifD coding sequence (codon 139) creating, in the proper orientation, a transcriptional fusion between nifD and the promoterless Km^r gene. The resulting plasmid was called pCRS4.

To generate a polar insertion in nifH, a 28-bp double-stranded oligonucleotide (I-SceI, Table 2) with overhanging, compatible BglII ends, was ligated into the unique BglII site of pCRS4, thus interrupting the nifH coding sequence (codon 147); this plasmid was named as pCRS5. Finally, to generate plasmid derivatives able to replicate in R. etli, plasmids pCRS4 and pCRS5 were digested with EcoRI and the resulting fragments were ligated separately onto the unique EcoRI site of the broad-host range plasmid pRK7813 (JONES and GUTTERSON 1987 ); the resulting plasmids were called pCRS6 and pCRS7, respectively.

Construction of Rhizobium etli strains carrying the nifH::I-SceI and nifD::Km/Sp alleles:
Introduction of the nifH::I-SceI and nifD::Km/Sp alleles into R. etli was carried out by an in vivo gene replacement procedure (SIMON et al. 1983 ). To that end, plasmid pCRS4 was introduced by transformation into E. coli S17-1 and the transformants were mated with R. etli CE3 as a recipient. Double recombinants were selected as Nal^r Sp^r Tc^s transconjugants. To verify that the desired gene replacement has occurred, double recombinants were analyzed by Southern blot hybridization against the appropriate nif and Sp probes. This procedure yielded strain CFNX236, which carries the nifD::Km/Sp allele in nif region a. Introduction of the nifH::I-SceI and nifD::Km/Sp allelic combination was carried out in the same way, but employing pCRS5 as the donor. To ensure coinheritance of both markers, double recombinants were analyzed by PCR amplification using oligonucleotides o2 and o3 (see Table 2 and Figure 1); this primer pair only yielded a PCR product upon integration of the nifH::I-SceI allele. A strain containing both the nifH::I-SceI and nifD::Km/Sp allelic combination in nif region a was called CFNX237. A recA::Cm derivative from strain CFNX237 was generated by a homogenotization procedure developed in our laboratory (MARTINEZ-SALAZAR et al. 1991 ; ROMERO et al. 1995 ), using pMS22 as the source of the recA allele. This procedure gave strain CFNX242.

View larger version (13K): In this window In a new window Download PPT slide   Figure 1. —Experimental design. Both parts of the figure represent nif region a. Symbols are as follows: the nifH promoter (p); the nifD::Km/Sp allele (triangle symbol); and the nifH::I-SceI allele (stick and ball symbol). Arrows beneath the figures indicate the expected transcripts in each case. (a) In this case, cells become Kmr due to transcription of the Kmr gene from the nifH promoter. (b) Introduction of the nifH::I-SceI allele leads to a Kms phenotype, due to the introduction of additional termination codons. Small arrows in this figure indicate the location of specific oligonucleotides (o1–o4) used for characterization.

To generate R. etli derivatives carrying these allelic combinations onto small, self-replicating plasmids, E. coli S17-1 derivatives carrying pCRS6 or pCRS7 were mated with either R. etli CE3 (pSym⁺) or R. etli CFNX89 (pSym^-) as recipients. Transconjugants were selected by their resistance to both nalidixic acid and spectinomycin. These crosses produced strains CFNX238 to CFNX241 (see Table 1).

PCR amplification and nucleotide sequencing:
PCR amplifications were carried out using AmpliTaq DNA polymerase in a DNA Thermal Cycler 480 (Perkin-Elmer, Inc., Norwalk, CT). PCR conditions consisted of 30 cycles of 92° for 1 min, 56° for 1 min, and 72° for 1 min, except for amplifications employing primer pairs o1–o4 and o2–o4, that were done by 30 cycles of 95° for 1 min, 55° for 1 min, and 72° for 2 min. PCR products for nucleotide sequencing were purified using Centri-Sep spin columns (Princeton Separations Inc., Adelphia, NJ). Nucleotide sequencing was performed with an Applied Biosystems Inc. model 373A automated DNA sequencer and a Taq DyeDeoxy Terminator cycle sequencing kit as specified by the manufacturer (Applied Biosystems Inc., Foster City, CA).

Filter blot hybridization and determination of plasmid profiles:
Genomic DNA was digested with BamHI, electrophoresed in 1% agarose gels, blotted onto nitrocellulose (Hybond N+), and hybridized under stringent conditions using Amersham's Rapid-hyb buffer as specified by the manufacturer (Amersham Corp.). Plasmid profiles were obtained by an in-gel lysis method (ECKHARDT 1978 ), blotted onto nitrocellulose and hybridized similarly. Hybridizations with oligonucleotide probes were done in a sodium chloride-sodium citrate solution using standard procedures (AUSUBEL et al. 1987 ). Most probes were linearized and labelled with ³²P--CTP by a random priming procedure (FEINBERG and VOGELSTEIN 1983 ) using a Rediprime DNA labelling system (Amersham Corp.). Oligonucleotide probes were labelled with ³²P--ATP using T4 polynucleotide kinase.

Molecular characterization of Km^r derivatives:
To ascertain the molecular events leading to the formation of each Km^r derivative, single-colony isolates were initially screened for the presence and location of the nifH::I-SceI allele. To that end, genomic DNA of each isolate was subjected to PCR employing either the o2–o3 primer pair or the o1–o3 pair (Figure 1). Primer o2 has a sequence that matches the nifH::I-SceI allele, while primers o1 and o3 bind to specific points in the nifH sequence (Table 2). Thus, reactions with the o2–o3 pair gave an amplified product only if the nifH::I-SceI allele was still present, while those with pair o1–o3 served as a positive control for the PCR. To determine the location of the nifH::I-SceI allele in the Km^r derivatives that still carried this allele, further PCR amplifications were made with primer pairs o2–o4 and o1–o4 (Figure 1 and Table 2). The first pair of primers gave a 3665-bp PCR product only if the nifH::I-SceI allele was still coupled to the nifD::Km/Sp allele, while the second pair provided a positive control for these reactions.

Km^r derivatives were also characterized by determining pSym size in Eckhardt-type gels. This analysis allows us to distinguish wild-type plasmids (390 kb) from amplified (510 kb) or deleted (270 kb) derivatives, as described previously (ROMERO et al. 1991 ; ROMERO et al. 1995 ). Additionally, genomic DNA of each derivative was digested with BamHI, blotted, and probed against a nifH-specific probe. Under these conditions, strain CFNX237 shows three hybridizing bands of 13 kb (nif region a harboring the nifH::I-SceI and nifD::Km/Sp allelic combination), 5.6 kb (nif region b) and 4.5 kb (nif region c). Derivatives harboring a tandem amplification preserve the same three bands, but show an additional 8.8-kb band, representing the join point; stoichiometry of these bands is also typical, where the band corresponding to nif region c is more abundant than the rest. Band pattern is also altered in the derivative carrying an inversion, showing three bands of 10.4, 6.5 and 5.6 kb; the first two bands are join points for this rearrangement, while the last corresponds to nif region b. The derivative carrying a deletion show a single, nifH-positive band of 13.6 kb; this band is the join point. Location of the nifH::I-SceI and nifD::Km/Sp alleles was verified through hybridization with allele-specific probes.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Experimental design:
The main objective of this work was to study the relative role of recombination vs. apparent gene conversion in the maintenance of sequence identity in a reiterated multigene family. To that end, we chose the nitrogenase multigene family of R. etli as a model. All members of this family are located in a single 390-kb plasmid, the symbiotic plasmid (pSym). This family is composed of two identical direct reiterations of about 5 kb (nif regions a and b), which are nifHDK operons; these operons are 120 kb apart on the pSym. The third element of this family (nif region c) is located in the middle of this zone, and consists of an identical reiteration, 1.5 kb long, harboring a complete nifH gene and a truncated nifD gene in an inverted orientation vis a vis nif regions a and b. Homologous recombination between nif regions a and b leads to frequent genomic rearrangements, such as deletions and amplifications (ROMERO et al. 1991 ; ROMERO et al. 1995 ). This particular arrangement allows us to study the recombinational dynamics of a plasmidic multigene family.

Since expression of these nif regions under ex planta conditions does not confer any scorable phenotype, we modified nif region a by inserting a promoterless Km^r cassette into the nifD gene, as described in MATERIALS AND METHODS. This cassette also carries a Sp^r gene with its own promoter. As shown in Figure 1A, expression of the Km^r gene in this nifD::Km/Sp allele should be under the control of the nifH promoter. This construct was then modified by the insertion of a 28-bp oligonucleotide into the nifH gene (the nifH::I-SceI allele, see MATERIALS AND METHODS). This nifH::I-SceI allele leads to alterations in the translational reading of the nifH gene, because in-frame reading of this insertion causes misreading of two termination codons (UAG and UAA) present in the oligonucleotide. Additionally, since this insertion provokes a +1 frameshift, two additional stop codons are uncovered (UAA and UGA) at positions matching codons 185–186 and 214–215 of the wild-type nifH sequence, respectively. As shown in Figure 1B, the nifH::I-SceI allele should block, by polarity, the expression of the nifD::Km/Sp allele, thus leading to a Km^s phenotype. Selection for Km^r derivatives give us a positive system to identify events that lead to the loss or relocation of the 28-bp insertion, conceivably via recombination with the other members of this multigene family.

Initial tests of the functionality of this system were done in E. coli. To that end, we introduced plasmids pCRS4 (carrying the nifD::Km/Sp allele; Table 1) and pCRS5 (carrying the nifH::I-SceI and nifD::Km/Sp allelic combination; Table 1) into E. coli HB101. In this system, expression directed by the nifH promoter depends on ⁵⁴ and the NifA activator protein (VALDERRAMA et al. 1996 ). As expected, E. coli strains harboring either pCRS4 or pCRS5 were sensitive to kanamycin (60 µg ml^-1). Upon introduction of a second, compatible plasmid carrying the constitutively-activated Klebsiella pneumoniae nifA gene (pMC71A) into the strain harboring pCRS4, cells became resistant to kanamycin. In contrast, cells carrying both pCRS5 and pMC71A remained sensitive to kanamycin. Thus, these results indicate that expression of the Km^r gene in the nifD::Km/Sp allele is dependent on the nifH promoter and that the introduction of the nifH::I-SceI allele blocks that expression.

Construction of R. etli strains containing on the pSym either the nifD::Km/Sp allele (strain CFNX236, Table 1) or the nifH::I-SceI and nifD::Km/Sp allelic combination (strain CFNX237, Table 1) was done by allelic replacement (see MATERIALS AND METHODS). In this host, maximal expression from the nifH promoter is achieved under microaerobic conditions (VALDERRAMA et al. 1996 ). However, basal transcription from this promoter under aerobic conditions was enough to confer to strain CFNX236 a low-level resistance to kanamycin (3 µg ml^-1). As expected, strain CFNX237, carrying the nifH::I-SceI and nifD::Km/Sp allelic combination was sensitive to kanamycin. Thus, loss or relocation of the nifH::I-SceI allele can be detected in strain CFNX237 by scoring the frequency of Km^r derivatives.

Theoretically, recombinational repair of the nifH::I-SceI allele to yield a Km^r derivative can arise either by sister-strand exchanges or through intramolecular exchanges. As shown in Figure 2, a sister-strand crossover between nif regions a and b leads to the formation of a large tandem duplication, where the join point carries the nifD::Km/Sp allele but lacks the nifH::I-SceI allele. Alternatively, removal of the nifH::I-SceI allele by sister-strand exchanges can come about from either double recombination or gene conversion. Both processes generate a non-rearranged pSym, lacking the nifH::I-SceI allele but maintaining the nifD::Km/Sp allele (Figure 2). This kind of product was called apparent gene conversion, since a formal distinction of which process is participating (double recombination or gene conversion) is not possible when the recombining sequences are in direct orientation (SEGALL and ROTH 1994 ).

View larger version (20K): In this window In a new window Download PPT slide   Figure 2. —Formation of Kmr derivatives by sister-strand exchanges. nif regions a, b and c are shown as rectangles with white, stippled or black shading, respectively, the nifH::I-SceI allele as a stick and ball symbol and the nifD::Km/Sp allele as a triangle symbol. (Left part) Recombination between nif regions a and b leads to the formation of a large tandem duplication, where the join point (indicated as a rectangle with mixed shading) carries the nifD::Km/Sp allele but lacks the nifH::I-SceI allele. (Right part) Both double recombination or gene conversion can generate a non-rearranged pSym, lacking the nifH::I-SceI allele but maintaining the nifD::Km/Sp allele. This kind of product was called apparent gene conversion, since making a formal distinction between which process (double recombination or gene conversion) is participating is not possible when the recombining sequences are in direct orientation. Only the selectable products are shown.

Intramolecular exchanges might also be responsible for the generation of selectable Km^r derivatives. As shown in Figure 3 (left part), an intramolecular crossover generates a true gene conversion recombinant. However, this event cannot be distinguished from the non-rearranged class generated by sister-strand exchanges (see above); therefore, all these are also scored as apparent gene convertants. Intramolecular exchange between nif regions a and c generates an inversion (Figure 3, center). This recombinant has a Km^r phenotype due to the relocation of the nifD::Km/Sp allele, which is now fused to the nifH region c promoter. Finally, intramolecular double recombination may also produce a Km^r recombinant, due to a relocation of the nifH::I-SceI allele (Figure 3, right).

View larger version (22K): In this window In a new window Download PPT slide   Figure 3. —Intramolecular exchanges might also be responsible for the generation of selectable Kmr derivatives. All symbols are as in Figure 2. (Left part) An intramolecular crossover generates a true gene conversion recombinant. However, this event cannot be distinguished from the non-rearranged class generated by sister-strand exchanges (see Figure 2); therefore, all these are also scored as apparent gene convertants. (Center part) An intramolecular exchange between nif regions a and c generates an inversion. This recombinant has a Kmr phenotype due to the relocation of the nifD::Km/Sp allele, which is now fused to the nifH region c promoter. (Right part) Intramolecular double recombination may also produce a Kmr recombinant, due to a relocation of the nifH::I-SceI allele. Only the selectable products are shown.

Correction of a small insertion in the nif multigene family is achieved by multiple recombination events:
To evaluate the frequency of correction in this multigene family, strain CFNX237, carrying the nifH::I-SceI and nifD::Km/Sp allelic combination on the pSym, was grown overnight in rich media and plated in media containing a low concentration of kanamycin (3 µg ml^-1). Km^r derivatives were found at a high frequency (344 x 10^-6). To identify the molecular events responsible for the generation of Km^r derivatives, 51 colonies were randomly chosen from seven independent selection experiments and purified as single-colony isolates. These derivatives were characterized by a combination of PCR, sizing of the resulting pSym and Southern blot hybridization against specific probes and then assigned to specific classes according to the criteria described in MATERIALS AND METHODS.

As shown in Figure 4, derivatives belonging to class I (tandem duplication) were the most abundant, comprising 74% of the observed products. This class may have resulted from a sister-strand exchange (Figure 2, left). Class II derivatives (apparent gene conversions) constitute 14% of the total isolates analyzed. As explained before, these may originate either through interactions involving sister strands (double crossover or gene conversion, Figure 2, right) or from intramolecular exchanges (true gene conversion, Figure 3, left). Inversions (class III, Figure 4) were very scarce in this sample, being represented by a single derivative (ca. 2%). This class is readily explained by assuming a reciprocal intramolecular exchange (Figure 3, center; see DISCUSSION).

View larger version (17K): In this window In a new window Download PPT slide   Figure 4. —Molecular events leading to the correction of a small insertion in the nif multigene family. Roman numerals indicate specific classes. Numbers in bold denote the frequency of clones in each class, while numbers in parentheses are the number of Kmr clones represented in the corresponding class. All symbols are as in Figure 2.

The remaining 10% of the derivatives are divided into three additional classes (IV to VI, Figure 4). These were generated most likely through a combination of apparent gene conversion plus additional recombination events. For instance, class IV is a tandem duplication similar but not identical to class I. Unlike class I, class IV derivatives have additionally lost the nifH::I-SceI allele, thus being also a case of apparent gene conversion. Class V is similar to the simple class II derivatives (apparent gene conversion), but it harbors an additional nifD::Km/Sp allele in nif region b. Class VI is an apparent gene conversion that carries a large deletion of the pSym; deletions similar to this have been observed previously (ROMERO et al. 1991 ). No derivatives attributable to intramolecular double recombination (Figure 3, right) were found, despite the finding of other complex, double-exchange events. Possible reasons for the absence of this class are presented in the DISCUSSION.

Thus, these results indicate that: (i) a variety of recombination events are participating in the correction of a small insertion in a member of this multigene family; (ii) tandem duplication is a major contributor for the observed correction, followed significantly by apparent gene conversion; and (iii) multiple recombination events were frequently found.

Role of the recA gene in recombinational correction:
The participation of the recA gene in the processes leading to correction of the nifH::I-SceI allele was evaluated by introducing the recA::Cm allele into strain CFNX237 as described in MATERIALS AND METHODS. Introduction of the recA::Cm allele provokes a 50-fold reduction in the frequency of isolation of Km^r derivatives (from 344 x 10^-6 in strain CFNX237 to 6 x 10^-6 in strain CFNX242). Characterization of twenty Km^r isolates obtained from strain CFNX242 did not reveal any loss or relocation of the nifH::I-SceI allele. Although not characterized further, these derivatives are more likely due to spontaneous mutations that restore transcription of the Km^r gene. Thus, the recA gene participates in the formation of tandem duplications and apparent gene conversions.

Recombinational correction in the nif multigene family occurs frequently by intermolecular exchanges:
A particular limitation of the system employed so far is that correction of the nifH::I-SceI allele might be generated either by intermolecular exchanges (sister-strand events), by intramolecular recombination or through a combination of both processes. However, a clear distinction of the role of intermolecular exchanges can be achieved through the incorporation of the allele whose correction is to be scored into a separate plasmid. As shown in Figure 5, correction in this system occurs only by intermolecular exchanges. This process has three separate outcomes. Single exchanges provoke the incorporation of the small plasmid into the pSym (Figure 5, left). The other two alternatives (double exchange or gene conversion) maintain the small plasmid and the pSym as separate entities, but the small plasmid now lacks the nifH::I-SceI allele (Figure 5, center and right). Although the expected products for double exchange or gene conversion events are different, it is not feasible to achieve a distinction between these alternatives, due to the possibility of cosegregation between the small plasmid (generated by a double exchange) and a second, uninvolved copy of the pSym.

View larger version (17K): In this window In a new window Download PPT slide   Figure 5. —Formation of Kmr derivatives by intermolecular exchanges. Intermolecular exchanges can form selectable Kmr derivatives either by a double exchange (center part) or by a true gene conversion event (right part). The product shown at the left part (single exchange) is not selectable, because it confers only a low-level resistance to kanamycin.

To evaluate the role of intermolecular exchanges in the observed recombinational correction, small, broad host-range plasmids containing either the nifD::Km/Sp allele (pCRS6, Table 1) or the nifH::I-SceI and nifD::Km/Sp allelic combination (pCRS7, Table 1) were constructed as described in MATERIALS AND METHODS. Conjugative transfer of each plasmid into a wild-type R. etli strain (CE3) generated strains CFNX238 and CFNX239 (Table 1). Strain CFNX238 (harboring plasmid pCRS6) showed a high-level resistance to kanamycin (70 µg ml^-1), while strain CFNX239 (carrying pCRS7) remained Km^s. Km^r derivatives from strain CFNX239 were obtained at a frequency of 28 x 10^-6. A total of 21 separate isolates (obtained from four independent selection experiments) were classified under the criteria described in MATERIALS AND METHODS. All these derivatives showed a structure compatible with apparent gene conversion (Figure 5, right). The absence of the single-exchange class was expected, due to the low resistance to kanamycin observed when the nifH::I-SceI and nifD::Km/Sp allelic combination was integrated on the pSym. Similar to the previous system, the double-exchange class was absent from this sample (see DISCUSSION).

The frequency of the apparent gene conversion class in this system (28 x 10^-6) is similar to the one estimated before (48 x 10^-6). These results suggest that this class is frequently generated through intermolecular exchanges.

This system allows us to determine if the formation of apparent gene convertants depends on the presence of additional copies of the nif region. To that end, plasmid pCRS7 was introduced into strain CFNX89, generating strain CFNX241; since this strain is devoid of the pSym, additional copies of the nif region are absent. The frequency of Km^r derivatives is reduced threefold (to 7.1 x 10^-6); these derivatives did not show the correction of the nifH::I-SceI allele. Thus, these results, coupled to the recA dependency data, indicate that apparent gene conversions are formed through a recombinational process, and not due to a recA-dependent, excision repair process.

Correction in the nif multigene family leads to a precise restoration in nucleotide sequence:
To be useful as a process leading to concerted variation between the nif regions, apparent gene conversion must restore the nifH::I-SceI allele to an otherwise wild-type sequence. To ascertain if this was the case, three independent class II, Km^r derivatives and one class VI derivative (Figure 4) were subjected to PCR amplifications employing oligonucleotides o1 and o4 (Figure 1) as primers. This particular primer pair yielded products of the region encompassing the site where the nifH::I-SceI allele was located. These PCR products were sequenced, using oligonucleotide o1 as a primer. In every case, nucleotide sequence was fully restored to wild type. These data were extended by isolating pCRS7 from four independent Km^r derivatives obtained as explained in the previous section; these plasmids were then transformed into E. coli. Nucleotide sequence from these derivatives also yielded a region identical to the wild-type sequence (data not shown). Thus, apparent gene conversions lead to a precise restoration of the wild-type nucleotide sequence.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

In this study we show that multiple recombination events, including tandem duplication, apparent gene conversion and inversion, lead to the correction of a small insertion in the nif multigene family. Tandem duplications appeared at a frequency of 2.8 x 10^-4, and were formed through recombination between nif region a and nif region b. The abundance of this class (75% of the observed events) was expected, since any crossover in the 1-kb region separating the nifH::I-SceI and nifD::Km/Sp alleles must produce a Km^r derivative. Thus, the frequency at which this class is generated reflects the frequency of recombination per kb of available homology in this organism. This estimate agrees well with previous evaluations in an interval of equivalent size (ROMERO et al. 1995 ).

Apparent gene conversion events in the nif multigene family were readily detected. The recombinational origin of these events is supported by (i) the strict requirement for a functional recA gene, (ii) the dependency on additional copies of the nif region for their formation, and (iii) a precise restoration of the wild-type nucleotide sequence. Apparent gene conversions comprised 24% of the observed events, appearing at a cumulative frequency of 8 x 10^-5. This estimate is remarkably similar to the observed frequencies of apparent gene conversion in both E. coli (HARVEY and HILL 1990 ) and S. typhimurium (SEGALL and ROTH 1994 ). Previous estimates obtained in R. meliloti, based on data from phage crosses, suggested that the efficiency of apparent gene conversion was lower in this organism than in E. coli (OROSZ et al. 1980 ). Our results clearly show that this is not the case, at least for recombinational interactions among members of a multigene family.

It is important to remark that this frequency refers to interactions with a particular outcome, namely, the correction of a specific 28-bp insertion. When all the possible corrections alongside the 5-kb nif regions are considered, the fraction of cells in a population with some change caused by apparent gene conversion in the nif regions should become substantial. Our results show that transfer of genetic information through apparent gene conversion occurs at a frequency that exceeds the frequency of spontaneous mutation. Therefore, this mechanism must contribute significantly to maintain the genetic homogeneity observed among members of the nif multigene family in Rhizobium, resulting in concerted evolution. We suggest that apparent gene conversion should also be relevant to explain other instances of maintenance of sequence identity in Rhizobium, as seen in a variety of multigene families.

Besides demonstrating the occurrence of apparent gene conversion, our work also allows us to draw some useful inferences about the recombination process in Rhizobium. Compared to tandem duplication and apparent gene conversion, inversions are minor contributors to correction of the effects of a small insertion. Only a single inversion, formed by recombination between nif region a and nif region c was isolated, at a frequency of 6 x 10^-6. Although the ocurrence of an inversion allows us to classify this interval as permissive (according to the terminology of SEGALL and ROTH 1994 ), the scarcity of this class is intriguing. Possible deleterious effects of this rearrangement on strain viability can be ruled out, because the strain harboring an inversion grows as well as the parental strain (data not shown). The infrequent occurrence of inversions cannot be explained by restriction in the size of a homologous region available for recombination. The interval used to generate an inversion is similar in location and size (750 bp) to the one used to generate a tandem duplication (1 kb). Despite this similarity, inversions occur 100-fold less frequently than tandem duplications.

Rarity of the inversion class can be explained by assuming, as has been proposed previously, that recombination in bacteria frequently entails the use of the so-called half-crossing over (MAHAN and ROTH 1989 ; KOBAYASHI 1992 ; SEGALL and ROTH 1994 ; ROTH et al. 1996 ; YAMAMOTO et al. 1996 ). Following this proposal, a half-crossover between two recombining sequences would generate a recombinant molecule and two broken DNA ends. Tandem duplications and apparent gene conversions can be readily generated by this mechanism; in contrast, formation of an inversion would require two coincidental half-crossovers to produce a viable recombinant, explaining the low frequency of this class. The occurrence of half-crossovers would also explain two additional aspects of this work, namely the preference of intermolecular exchanges for generation of apparent gene convertants and the absence of derivatives attributable to intramolecular double recombination (Figure 3, right). These features might be a consequence of the fact that a single intramolecular half-crossover generates a linear molecule, thus precluding the isolation of a viable recombinant.

Considering the arguments given above, the occurrence of classes attributable to complex, double-exchange events (classes IV to VI, Figure 4) might seem paradoxical. However, we believe that these classes are formed by successive, rather than coincidental, half-crossover events. Common to these classes is that they appear to be composites of two different recombination events, where the product of each event is selectable. For instance, class IV was apparently generated by two selectable events, one a tandem duplication and the other an apparent gene conversion. It is conceivable that recombinants belonging to this class may have formed as successive recombination events during colony development, in a situation akin to phage recombination or recombination among small, high-copy number plasmids (YAMAMOTO et al. 1988 ).

	ACKNOWLEDGMENTS

We are indebted to SUSANA BROM, JAIME MARTÍNEZ-SALAZAR and RAFAEL PALACIOS for critical reviewing of the manuscript, to JOSÉ ESPÍRITU and CÉSAR HERNÁNDEZ for help in preparing the manuscript, to PATRICIA BUSTOS for help with nucleotide sequencing, and to JAVIER RIVERA and MARÍA DE LA PAZ SALAS for skillful technical assistance. Partial financial support for this research was provided by grants No. 4321-N9406 from the Consejo Nacional de Ciencia y Tecnología (México) and No. 030355 from the Programa de Apoyo a las Divisiones de Estudios de Posgrado (Universidad Nacional Autónoma de México). C. R. was a recipient of a scholarship from the Consejo Nacional de Ciencia y Tecnología (México).

Manuscript received December 3, 1997; Accepted for publication March 2, 1998.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ABDULKARIM, F. and D. HUGHES, 1996  Homologous recombination between the tuf genes of Salmonella typhimurium.. J. Mol. Biol. 260:506-522[Medline].

AUSUBEL, F. M., R. BRENT, R. E. KINGSTON, D. D. MOORE, J. G. SEIDMAN et al., 1987 Current Protocols in Molecular Biology. Massachusetts General Hospital, Harvard Medical School. John Wiley & Sons, Inc., New York.

BACHELLIER, S., E. GILSON, M. HOFNUNG and C. W. HILL, 1996 Repeated sequences, pp. 2012–2040 in Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology, Ed. 2, edited by F. C. NEIDHARDT, R. CURTISS III, J. L. INGRAHAM, E. C. C. LIN, K. B. LOW, JR. et al. American Society for Microbiology Press, Washington, DC.

BADENOCH-JONES, J., T. A. HOLTON, C. MORRISON, K. F. SCOTT, and J. SHINE, 1989  Structural and functional analysis of nitrogenase genes from the broad-host-range Rhizobium strain ANU240. Gene 77:141-153[Medline].

BERGMAN, K., E. NULTY, and L. SU, 1991  Mutations in the two flagellin genes of Rhizobium meliloti.. J. Bacteriol. 173:3716-3723[Abstract/Free Full Text].

BLATTNER, F. R., G. PLUNKETT, III, C. A. BLOCH, N. T. PERNA, and V. BURLAND et al., 1997  The complete genome sequence of Escherichia coli K-12. Science 277:1453-1474[Abstract/Free Full Text].

BOYER, H. W. and D. ROULLAND-DUSSOIX, 1969  A complementation analysis of restriction and modification of DNA in Escherichia coli.. J. Mol. Biol. 41:459-472[Medline].

BROM, S., A. GARCÍA DE LOS SANTOS, T. STEPKOWSKI, M. FLORES, and G. DÁVILA et al., 1992  Different plasmids of Rhizobium leguminosarum bv. phaseoli are required for optimal symbiotic performance. J. Bacteriol. 174:5183-5189[Abstract/Free Full Text].

BUCHANAN-WOLLASTON, V., M. C. CANNON, J. L. BEYNON, and F. C. CANNON, 1981  Role of the nifA gene product in the regulation of nif expression in Klebsiella pneumoniae.. Nature 294:776-778[Medline].

DAVID, M., O. DOMERGUE, P. POGNONEC, and D. KAHN, 1987  Transcription patterns of Rhizobium meliloti symbiotic plasmid pSym: identification of nifA-independent fix genes. J. Bacteriol. 169:2239-2244[Abstract/Free Full Text].

DEONIER, R. C., 1996 Native insertion sequence elements: locations, distributions and sequence relationships, pp. 2000–2011 in Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology, Ed. 2, edited by F. C. NEIDHARDT, R. CURTISS III, J. L. INGRAHAM, E. C. C. LIN, K. B. LOW, JR. et al. American Society for Microbiology Press, Washington, DC.

DOVER, G. A., 1993  Evolution of genetic redundancy for advanced players. Curr. Opin. Genet. Dev. 3:902-910[Medline].

ECKHARDT, T., 1978  A rapid method for the identification of plasmid deoxyribonucleic acid in bacteria. Plasmid. 1:584-588[Medline].

FEINBERG, A. P. and B. VOGELSTEIN, 1983  A technique for radiolabelling DNA restriction endonuclease fragments to high specific activity. Anal. Biochem. 132:6-13[Medline].

FIGURSKI, D. H. and D. R. HELINSKI, 1979  Replication of an origin-containing derivative of plasmid RK2 dependent on a plasmid function provided in trans.. Proc. Natl. Acad. Sci. USA 76:1648-1652[Abstract/Free Full Text].

FISCHER, H.-M., M. BABST, T. KASPAR, G. ACUÑA, and F. ARIGONI et al., 1993  One member of a groESL-like chaperonin multigene family in Bradyrhizobium japonicum is co-regulated with symbiotic nitrogen fixation genes. EMBO J. 12:2901-2912[Medline].

FLORES, M., V. GONZÁLEZ, S. BROM, E. MARTÍNEZ, and D. PIÑERO et al., 1987  Reiterated DNA sequences in Rhizobium and Agrobacterium spp. J. Bacteriol. 169:5782-5788[Abstract/Free Full Text].

FREIBERG, C., R. FELLAY, A. BAIROCH, W. J. BROUGHTON, and A. ROSENTHAL et al., 1997  Molecular basis of symbiosis between Rhizobium and legumes. Nature 387:394-401[Medline].

GANGLOFF, S., H. ZOU, and R. ROTHSTEIN, 1996  Gene conversion plays the major role in controlling the stability of large tandem repeats in yeast. EMBO J. 15:1715-1725[Medline].

GIRARD, M. L., M. FLORES, S. BROM, D. ROMERO, and R. PALACIOS et al., 1991  Structural complexity of the symbiotic plasmid of Rhizobium leguminosarum bv. phaseoli. J. Bacteriol. 173:2411-2419[Abstract/Free Full Text].

HARVEY, S. and C. W. HILL, 1990  Exchange of spacer regions between rRNA operons in Escherichia coli.. Genetics 125:683-690[Abstract].

HERNÁNDEZ-LUCAS, I., M. A. PARDO, L. SEGOVIA, J. MIRANDA, and E. MARTÍNEZ-ROMERO, 1995  Rhizobium tropici chromosomal citrate synthase gene. Appl. Environ. Microbiol. 61:3992-3997[Abstract].

JACKSON, J. A. and G. R. FINK, 1981  Gene conversion between duplicated genetic elements in yeast. Nature 292:306-311[Medline].

JOHN, B., and G. L. G. MIKLOS, 1988 The Eukaryote Genome in Development and Evolution. Allen & Unwin, London.

JONES, J. D. G. and N. GUTTERSON, 1987  An efficient mobilizable cosmid vector, pRK7813, and its use in a rapid method for marker exchange in Pseudomonas fluorescens HV37a. Gene 61:299-306[Medline].

KLEIN, H. L. and T. D. PETES, 1981  Intrachromosomal gene conversion in yeast. Nature 289:144-148[Medline].

KOBAYASHI, I., 1992 Mechanisms for gene conversion and homologous recombination: the double-strand break repair model and the successive half crossing-over model, pp. 81–133 in Advances in Biophysics, edited by M. KOTANI. Japan Scientific Societies Press and Elsevier, Tokyo and Limerick.

KÜNDIG, C., H. HENNECKE, and M. GÖTTFERT, 1993  Correlated physical and genetic map of the Bradyrhizobium japonicum 110 genome. J. Bacteriol. 175:613-622[Abstract/Free Full Text].

MAHAN, M. J. and J. R. ROTH, 1989  Role of recBC function in formation of chromosomal rearrangements: a two-step model for recombination. Genetics 121:433-443[Abstract/Free Full Text].

MARGOLIN, W. and S. LONG, 1994  Rhizobium meliloti contains a novel second homolog of the cell division gene ftsZ.. J. Bacteriol. 176:2033-2043[Abstract/Free Full Text].

MARSH, J. L., M. ERFLE, E. J, and E. JWYKES, 1984  The pIC plasmid and phage vectors with versatile cloning sites for recombinant selection by insertional inactivation. Gene 32:481-485[Medline].

MARTÍNEZ-SALAZAR, J. M., D. ROMERO, M. L. GIRARD, and G. DÁVILA, 1991  Molecular cloning and characterization of the recA gene of Rhizobium phaseoli and construction of recA mutants. J. Bacteriol. 173:3035-3040[Abstract/Free Full Text].

MILLER, J. H., 1972 Experiments in Molecular Genetics. Cold Spring Harbor Laboratory, Cold Spring Harbor, New York.

MORETT, E., S. MORENO, and G. ESPÍN, 1988  Transcription analysis of the three nifH genes of Rhizobium phaseoli with gene fusions. Mol. Gen. Genet. 213:499-504.

NOEL, K. D., F. SÁNCHEZ, L. FERNÁNDEZ, J. LEEMANS, and M. A. CEVALLOS, 1984  Rhizobium phaseoli symbiotic mutants with transposon Tn5 insertions. J. Bacteriol. 158:148-155[Abstract/Free Full Text].

NOREL, F. and C. ELMERICH, 1987  Nucleotide sequence and functional analysis of the two nifH copies of Rhizobium ORS571. J. Gen. Microbiol. 133:1563-1576.

OHNO, S., 1970 Evolution by Gene Duplication. Springer-Verlag, Berlin.

OKAZAKI, N., S. MATSUO, K. SAITO, A. TOMINAGA, and M. ENOMOTO, 1993  Conversion of the Salmonella phase 1 flagellin gene fliC to the phase 2 gene fljB on the Escherichia coli K-12 chromosome. J. Bacteriol. 175:758-766[Abstract/Free Full Text].

OROSZ, L., A. PÁY, and G. DALLMANN, 1980  Heterozygosis of phage 16-3 of Rhizobium meliloti: moderate level of mismatch repair or gene conversion. Mol. Gen. Genet. 179:163-167[Medline].

PARDO, M. A., J. LAGÚNEZ, J. MIRANDA, and E. MARTÍNEZ, 1994  Nodulating ability of Rhizobium tropici is conditioned by a plasmid-encoded citrate synthase. Mol. Microbiol. 11:315-321[Medline].

PETES, T. D., 1980  Unequal meiotic recombination within tandem arrays of yeast ribosomal DNA genes. Cell 19:765-774[Medline].

PRENTKI, P. and H. M. KRISCH, 1984  In vitro insertional mutagenesis with a selectable DNA fragment. Gene 29:303-313[Medline].

QUINTO, C., H. DE LA VEGA, M. FLORES, J. LEEMANS, and M. A. CEVALLOS et al., 1985  Nitrogenase reductase: a functional multigene family in Rhizobium phaseoli.. Proc. Natl. Acad. Sci. USA 82:1170-1174[Abstract/Free Full Text].

ROMERO, D., S. BROM, J. MARTÍNEZ-SALAZAR, M. L. GIRARD, and R. PALACIOS et al., 1991  Amplification and deletion of a nod-nif region in the symbiotic plasmid of Rhizobium phaseoli.. J. Bacteriol. 173:2435-2441[Abstract/Free Full Text].

ROMERO, D., J. MARTÍNEZ-SALAZAR, L. GIRARD, S. BROM, and G. DÁVILA et al., 1995  Discrete amplifiable regions (amplicons) in the symbiotic plasmid of Rhizobium etli CFN42. J. Bacteriol. 177:973-980[Abstract/Free Full Text].

ROMERO, D., G. DÁVILA and R. PALACIOS, 1997 The dynamic genome of Rhizobium, pp. 153–161 in Bacterial Genomes: Physical Structure and Analysis, edited by F. J. DE BRUIJN, J. R. LUPSKI and G. WEINSTOCK. Chapman & Hall, New York.

ROTH, J. R., N. BENSON, T. GALITSKI, K. HAACK, J. G. LAWRENCE et al., 1996 Rearrangements of the bacterial chromosome: formation and applications, pp. 2256–2276 in Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology, Ed. 2, edited by F. C. NEIDHARDT, R. CURTISS III, J. L. INGRAHAM, E. C. C. LIN, K. B. LOW, JR. et al. American Society for Microbiology Press, Washington, DC.

SAMBROOK, J., E. F. FRITSCH and T. MANIATIS, 1989 Molecular Cloning: a Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York.

SCHLÜTER, A., T. PATSCHKOWSKI, J. QUANDT, L. B. SELINGER, and S. WEIDNER et al., 1997  Functional and regulatory analysis of the two copies of the fixNOQP operon of Rhizobium leguminosarum strain VF39. Mol. Plant Microbe Interact. 10:605-616[Medline].

SCHULTZE, M., E. KONDOROSI, P. RATET, M. BUIRÉ, and A. KONDOROSI, 1994  Cell and molecular biology of Rhizobium-plant interactions. Int. Rev. Cytol. 156:1-75.

SEGALL, A. M. and J. R. ROTH, 1994  Approaches to half-tetrad analysis in bacteria: recombination between repeated, inverse-order chromosomal sequences. Genetics 136:27-39[Abstract].

SEGOVIA, L., J. P. W. YOUNG, and E. MARTÍNEZ-ROMERO, 1993  Reclassification of american Rhizobium leguminosarum biovar phaseoli type I strains as Rhizobium etli sp. nov.. Int. J. Sist. Bacteriol. 43:374-377.

SIMON, R., 1984  High frequency mobilization of gram-negative bacterial replicons by the in vitro constructed Tn5-Mob transposon. Mol. Gen. Genet. 196:413-420[Medline].

SIMON, R., U. PRIEFER, and A. PÜHLER, 1983  A broad host range mobilization system for in vivo genetic engineering: transposon mutagenesis in gram-negative bacteria. Bio/Technology 1:784-791.

SZOSTAK, J. W. and R. WU, 1980  Unequal crossing over in the ribosomal DNA of Saccharomyces cerevisiae.. Nature 284:426-430[Medline].

VALDERRAMA, B., A. DÁVALOS, L. GIRARD, E. MORETT, and J. MORA, 1996  Regulatory proteins and cis-acting elements involved in the transcriptional control of Rhizobium etli reiterated nifH genes. J. Bacteriol. 178:3119-3126[Abstract/Free Full Text].

WILLIAMS, S. M. and C. STROBECK, 1985  Sister chromatid exchange and the evolution of rDNA spacer length. J. Theor. Biol. 116:625-636[Medline].

YAMAMOTO, K., H. YOSHIKURA, N. TAKAHASHI, and I. KOBAYASHI, 1988  Apparent gene conversion in an Escherichia coli rec⁺ strain is explained by multiple rounds of reciprocal crossing over. Mol. Gen. Genet. 212:393-404[Medline].

YAMAMOTO, K., N. TAKAHASHI, Y. FUJITANI, H. YOSHIKURA, and I. KOBAYASHI, 1996  Orientation dependence in homologous recombination. Genetics 143:27-36[Abstract].

This article has been cited by other articles:

				G. Santoyo, J. M. Martinez-Salazar, C. Rodriguez, and D. Romero Gene Conversion Tracts Associated with Crossovers in Rhizobium etli J. Bacteriol., June 15, 2005; 187(12): 4116 - 4126. [Abstract] [Full Text] [PDF]

				J. Zuniga-Castillo, D. Romero, and J. M. Martinez-Salazar The Recombination Genes addAB Are Not Restricted to Gram-Positive Bacteria: Genetic Analysis of the Recombination Initiation Enzymes RecF and AddAB in Rhizobium etli J. Bacteriol., December 1, 2004; 186(23): 7905 - 7913. [Abstract] [Full Text] [PDF]

				W. J. Broughton Roses by Other Names: Taxonomy of the Rhizobiaceae J. Bacteriol., May 15, 2003; 185(10): 2975 - 2979. [Full Text] [PDF]

C. Silva, P. Vinuesa, L. E. Eguiarte, E. Martinez-Romero, and V. Souza Rhizobium etli and Rhizobium gallicum Nodulate Common Bean (Phaseolus vulgaris) in a Traditionally Managed Milpa Plot in Mexico: Population Genetics and Biogeographic Implications Appl. Envir. Microbiol., February 1, 2003; 69(2): 884 - 893. [Abstract] [Full Text] [PDF]