Bradyrhizobia from Wild Phaseolus, Desmodium, and Macroptilium Species in Northern Mexico

Abstract

rRNA genetic markers were analyzed in 97 isolates of nodule bacteria from six legume species in Chihuahua, Mexico. The most common genotypes were widely shared across host species and had 16S rRNA sequences identical to those of strains from an eastern North American legume (Amphicarpaea) that are closely related to Bradyrhizobium elkanii.

Phaseolus is a legume genus of ca. 50 species endemic to North and South America, with the greatest concentration of species occurring in Mexico (25). It includes one of the most widely cultivated legumes in the world (common bean, Phaseolus vulgaris), as well as several other economically important species such as lima bean (P. lunatus), tepary bean (P. acutifolius), and scarlet runner bean (P. coccineus). While the symbiotic root nodule bacteria of certain cultivated Phaseolus species have been extensively analyzed (1, 7, 10, 14, 16, 23, 24, 26, 32), virtually nothing is known about nodule symbionts for the majority of undomesticated species in the genus.

This study characterized the diversity and relationships of nodule bacteria from three perennial Phaseolus species within their native geographic range in the Sierra Madre Occidental of Mexico. Two of the species (P. parvulus and P. pauciflorus) are members of a morphologically distinctive clade that primarily occurs in high-altitude habitats (1,000 to 3,000 m) and are diminutive plants with globose taproots, traditionally recognized as Phaseolus section Minkelersia (5). In a large phylogenetic analysis that included all known extant species of Phaseolus (6), species in section Minkelersia formed a sister group to a large clade (>20 species) that included lima bean (P. lunatus). The remaining species sampled (P. pedicellatus) is another high-altitude taxon that apparently diverged early in the evolution of the genus and is not closely related to section Minkelersia or to any of the cultivated species (6).

All of these Phaseolus species occur sympatrically in pine-oak forests of the Sierra Madre of northwestern Mexico. To provide a further perspective on host range and specificity, bacterial isolates were also collected from three other legumes (Macroptilium gibbosifolium, Desmodium grahamii, and Desmodium retinens) that grew intermixed with the Phaseolus species. Macroptilium is a close relative of Phaseolus within the Papilionoideae tribe Phaseoleae (6), while Desmodium is placed in the related tribe Desmodieae (9). Bacterial isolates were characterized by analysis of restriction fragment length polymorphisms (RFLPs) in 16S rRNA genes and by screening for length variants in the 16S-23S rRNA internal transcribed spacer (ITS) (35) and in the 5′ end of 23S rRNA (18, 21). For representative isolates, DNA sequencing was then performed on 16S rRNA and the 5′ 23S rRNA region.

At two sites in the Sierra Madre mountains of western Chihuahua, Mexico, 44 isolates were sampled from Phaseolus, 39 isolates from Desmodium, and 14 isolates from M. gibbosifolium (Table 1). The Areponapuchi site (107°50′W, 27°31′N, 2,200 m elevation) and the Cerocahui site (108°01′W, 27°15′N, 2,100 m) are 33 km apart. P. pauciflorus was the only legume found at both locations, and the remaining five species were sampled at just one of the two sites. P. pedicellatus plants were only observed in one small patch, and very few nodules were obtained, but for all other legume taxa, nodules were collected from many individual plants scattered over a 1-km area within the site. Isolates were named by an abbreviation of the host's name. One isolate was purified from each nodule as described previously (21, 27). All isolates grew slowly on yeast-mannitol agar plates (colonies not visible before 4 days), suggesting that they were members of the genus Bradyrhizobium.

TABLE 1.

Numbers of bacterial isolates with each of eight genotypes associated with six legume hosts

Host	Sitea	Composite genotypeb (no. of isolates)
		A	B	C	D	E	F	G	H
P. parvulus	ARE				2	7
P. pedicellatus	ARE				1	2
P. pauciflorus	ARE	1				8	1
P. pauciflorus	CER		8		1	10	3
M. gibbosifolium	CER		3	1		2	4	2	2
D. grahamii	ARE		8		1	12
D. retinens	ARE				5	13

^aARE, Areponapuchi; CER, Cerocahui, Mexico.

^bBased on 16S rRNA RFLP and length variation in ribosomal ITS and 23S rRNA; also see Table 2.

DNA was purified from bacterial cells by the protocol of Lafay and Burdon (15), and PCR amplification was performed as described previously (18, 19, 20). A nearly full-length portion of 16S rRNA (1,410 bp) was amplified from all isolates with primers fD1d and rP1a (18), and 10-ml aliquots of PCR products were digested with four restriction enzymes (Table 2). Fragments were separated on 1.9% agarose gels stained with ethidium bromide to visualize RFLPs. Six isolates were selected that encompassed the major RFLP types, and their 16S rRNA genes were sequenced on both strands as described previously (18) by using an Applied Biosystems model 310 automated sequencer. Analysis of restriction sites in these sequences resulted in predicted fragment sizes (Table 2) that corresponded exactly to those observed on agarose gels in all cases.

TABLE 2.

RFLP and length variation markers distinguishing Bradyrhizobium genotypes

Geno- type	16S rRNA polymorphic restriction fragments (bp)a				Length variationb
	MspI	HaeIII	HinfI	HhaI	5′ ITS	3′ ITS	23S rRNA
A	156,502	34,66	96,684	796	360	680	260
B	156,502	100	780	796	360	540	260
C	156,502	34,66	780	796	360	540	260
D	43,122,495	100	333,449	344,452	460	540	230
E	43,122,495	100	333,449	344,452	400	540	230
F	156,502	34,66	780	796	460	540	230
G	156,502	34,66	96,684	796	360	520	230
H	156,502	100	780	796	460	520	230

^aFragment lengths calculated from complete 16S rRNA sequences of selected isolates; discrepancies in the sums of fragment lengths for MspI and HinfI are due to two single-base-pair insertion-deletion differences.

^bSize of PCR-amplified fragments (base pairs) estimated relative to marker bands (HaeIII digest of φX174) on agarose gels.

To further characterize diversity among the 97 isolates, PCR was used to screen for length variation in the ITS region between 16S rRNA and 23S rRNA (35), and in a 5′ portion of 23S rRNA (18, 21). The 5′ portion of the ITS region was amplified with primers 16S-1 (5′-TGCGGCTGGATCACCTCCTT) and igsr4 (5′-GATCGAACCGACGACCTCAT), which correspond to the 3′ end of 16S rRNA and a portion of the tRNA^Ala gene within the ITS, respectively. The 3′ portion of the ITS region was amplified with primers tralaf (5′-TTTGCAAGCATGAGGTCGTCGGTT) and 6n23Sdp (5′-CACGTCCTTCATCGCCTCT), which correspond to part of the tRNA^Ala gene and positions 43 to 61 of the 23S rRNA gene, respectively. For both the 5′ and 3′ portions of the ITS region, three distinct length variants were detected that differed by 20 to 160 bp (Table 2). The 5′ portion of 23S rRNA was amplified with primers 23Sup115 and 23SrIII (28). These primers yield a 260-bp DNA fragment in Bradyrhizobium japonicum USDA 110 (GenBank Z35330), but many Bradyrhizobium strains have a 24- to 28-bp deletion in this region (19, 20, 21). All six isolates sequenced for 16S rRNA were also sequenced for a 468- to 496-bp portion of 23S rRNA spanning this region as described previously (18).

Sequences were first aligned by using CLUSTAL W (31), and trees were constructed by maximum parsimony using the PAUP software (version 4.0b1; D. L. Swofford, Smithsonian Institution, Washington, D.C.). To determine the degree of statistical support for branches in the phylogeny (12), 1,000 boostrap replicates of the data were analyzed. 16S rRNA data were compared to the following Bradyrhizobium reference sequences (strains without formal names are identified by host legume genus in brackets): B. japonicum USDA 6^T (GenBank U69638 [3]) and USDA 110 (Z35330); B. elkanii USDA 76^T (U35000); B. liaoningense LMG 18230^T (AJ250813); FN13 [Lupinus] (X87273 [3]); four Bradyrhizobium sp. strains from Australia (15)--5028A [Bossiaea] (Z94811), 5040B [Bossiaea] (Z94812), 5680G [Bossiaea] (Z94804), and 5111P [Daviesia] (Z94805); LMG 9514 [Lonchocarpus] (X70405); jwc91-2 [Amphicarpaea] (AF178437), th-b2 [Amphicarpaea] (AF178438 [18]); Pe1-3 [Platypodium] (AF15936 [21]); and Cj3-3 [Clitoria] (AF321212), Ec3-3 [Erythrina] (AF321213), and Da3-1 [Desmodium] (AF321215 [20]). Azorhizobium caulinodans (X67221) was used as the outgroup (36).

The GenBank accession numbers for the 16S rDNA sequences determined in this work are AF384135 to AF384140, and the 23S rDNA sequences have accession numbers AF384129 to AF384134.

rRNA variation

Analysis of fragments in 16S rRNA digested with four different restriction enzymes with 4-bp recognition sequences revealed only four patterns among the 97 isolates (Table 2). Screens for length variation in ribosomal ITS and 5′ 23S rRNA allowed an additional genotype to be distinguished within each 16S rRNA RFLP type, resulting in a total of eight composite genotypes (designated A to H). The 16S rRNA RFLP pattern of genotypes D and E is identical to that of B. elkanii USDA 76^T (U35000), while the pattern in genotypes B and H is identical to that of B. japonicum USDA 6^T (U69638).

Genotype E was by far the most common (56% of all isolates) and occurred in association with all six legume species sampled (Table 1). Only three other genotypes (B, D, and F) were found in more than two isolates, and these were also shared across at least two host legume genera. Every genotype associated with the two Desmodium taxa also occurred in association with one or more of the Phaseolus species, and three of the six genotypes sampled from M. gibbosifolium also occurred on Phaseolus (Table 1). This suggests limited bacterial host specificity among these legume taxa. Likewise, all genotypes detected in more than two isolates were shared across the two sample locations, suggesting that spatial variation was limited.

Linkage disequilibrium was analyzed for all pairwise comparisons of four polymorphic 16S rRNA restriction sites and the 5′ ITS and 5′ 23S rRNA length variants (Table 2; for one HinfI site and for the 3′ ITS, too few isolates showed variant types to permit statistically reliable analysis of disequilibrium [13]). Of 15 possible pairwise comparisons among the six markers, 14 yielded the maximum possible value of the standardized disequilibrium coefficient (D′ = 1, P < 0.001), and the remaining pair (HaeIII/23S rRNA) yielded a nonsignificant value (D′ = 0.23). Randomization of variants in this last pair may be affected either by recurrent mutation in the HaeIII site or by recombination (20), which is expected to have a greater effect on reducing disequilibrium for more widely separated sites (13). Interestingly, the HaeIII site and the 23S rRNA length polymorphism represent the most distant pair of markers analyzed. Approximately 2,300 to 2,400 bp separate the 16S rRNA HaeIII site from the region of insertion or deletion variation in the 5′ end of 23S rRNA.

Overall, the high disequilibrium across almost all markers is consistent with the interpretation that there has been little or no gene transfer in the 16S-23S ribosomal gene region among strains within this population. This contrasts with some other recent analyses of ribosomal DNA (11, 20, 33). Despite coexisting in the same habitat on a shared set of legume hosts, bacterial groups appear to be maintained as distinct lineages. Evidently, recombinant genotypes either occur too rarely to be detected without a much larger sample of isolates or are rapidly eliminated from populations by selection. However, because this study focused on only a small (∼2,500 bp) ribosomal gene region, the results leave open the possibility that gene transfer events may be observed for less closely spaced chromosomal sites. Thus, it would be informative in future work to analyze additional loci distant from the ribosomal genes.

A nearly full-length portion of 16S rRNA (1,408 to 1,412 bp) was sequenced in one representative isolate from genotypes A, B, D, E, F, and G. A search of Bradyrhizobium sequences in GenBank identified strains that were highly similar or identical to each of the Mexican isolates (Fig. 1). For example, strains Ppar1-21 (genotype E) and Ppar1-31 (genotype D) from P. parvulus had 16S rRNA sequences identical to strains th-b2 and jwc91-2, respectively, which originated from the legume Amphicarpaea bracteata in eastern North America (18). Strain Ppau3-41 (genotype A) from P. pauciflorus was identical to Bradyrhizobium sp. strain 5028A sampled from the Australian legume Bossiaea ensata (15). Strain Ppau3-25 (genotype B) had only one nucleotide substitution relative to B. japonicum USDA 6^T, and strain Mg1-11 (genotype G) from M. gibbosifolium had only one substitution relative to Bradyrhizobium sp. strain 5040B from Bossiaea ensata in Australia (15). A parsimony analysis indicated that all of the strains fell into two well-defined clades (99 to 100% bootstrap support) related to either B. japonicum or B. elkanii (Fig. 1). Strains Ppar1-21 and Ppar1-31 were included in the B. elkanii clade, and the remaining isolates were grouped with B. japonicum and its relatives.

FIG. 1.

Phylogenetic relationships of six Bradyrhizobium isolates from Phaseolus and Macroptilium (shown in boldface) based on parsimony analysis of 16S rRNA sequences. Numbers above the branches are bootstrap percentages (for clarity, only values of >85% are shown). GenBank accession numbers are given after the strain names.

To further analyze relationships, the same six Bradyrhizobium isolates from Phaseolus and Macroptilium were also sequenced for a 5′ portion of 23S rRNA that is highly polymorphic among Bradyrhizobium strains (18, 21). The 23S rRNA phylogeny broadly resembled the 16S rRNA tree (results not shown). Isolates Ppar1-21 and Ppar1-31 again clustered with B. elkanii USDA 76^T and its relatives, while the remaining four isolates grouped with B. japonicum USDA 110. Isolates Mg1-1 and Mg1-11 (genotypes F and G) had a short 23S rRNA length variant (Table 2) resembling that seen in all B. elkanii relatives analyzed to date (18, 19, 20, 28). However, parsimony analysis of 23S rRNA sequence variation did not group these strains with B. elkanii, but rather, with strain Pe1-3 from Panama, which is another B. japonicum relative that also has a short 23S rRNA length variant (21).

Nodulation and nitrogenase activity

No seeds of the six host legume species were available to use for inoculation studies. I thus used plants of P. vulgaris (cultivar Midnight Black Turtle Soup) and M. atropurpureum, which is known to be nodulated by a broad range of bradyrhizobia (21, 29). Twenty isolates picked at random (prior to genetic marker analysis) were used to inoculate four plants of each species according to previously described procedures (34). Seeds were surface disinfected with 50% ethanol and then germinated. Seedlings were planted in pots by using a Bradyrhizobium-free mixture of sand, perlite, and potting soil and then inoculated with ca. 10⁹ cells of a particular isolate grown in yeast-mannitol broth. Plants were grown in a greenhouse for 33 days with precautions to avoid bacterial contamination across inoculation treatments (34). Uninoculated control plants grown simultaneously in the same room were found to be completely free of nodules. Plants were fertilized weekly with a nitrogen-free nutrient solution (21). At harvest, nodule numbers were recorded, and each plant's root system was analyzed for acetylene reduction activity by using a Hewlett-Packard 5890 series II gas chromatograph as described previously (27).

All Bradyrhizobium isolates tested from the six host legume species formed effective, nitrogen-fixing nodules on M. atropurpureum, with most isolates developing 45 to 90 nodules per plant and having acetylene reduction rates of >2 μmol of ethylene plant⁻¹ min⁻¹ (results not shown). In contrast, nodule formation was generally much lower on plants of P. vulgaris, with five isolates completely failing to form nodules. Among the other 15 isolates, 10 showed no detectable acetylene reduction in nodules, and the remaining 5 isolates (Ppar1-31, Ppar1-30, Dgra1-19, Dret2-51, and Dret2-63) all had rates of <0.08 μmol of ethylene plant⁻¹ min⁻¹ on P. vulgaris.

Prevailing views about Phaseolus nodule bacteria are influenced by the large body of research on P. vulgaris, whose bacteria, although highly diverse at the species level, all fall into the genus Rhizobium (1, 7, 10, 14, 16, 23, 24, 26, 32). Nevertheless, P. vulgaris may not be the best model for nodule symbioses in Phaseolus, because it apparently occupies a relatively derived position within the genus (6). In addition to the use of Bradyrhizobium sp. by all of the Phaseolus species of this study (Fig. 1), Bradyrhizobium sp. is also known to be utilized by P. lunatus (8, 29), as well as by closely related genera such as Macroptilium and Vigna spp. (e.g., Fig. 1) (4, 29). These observations support the view that symbiosis with Bradyrhizobium represents the ancestral condition in the genus Phaseolus and that utilization of Rhizobium is a recent innovation that may be restricted to P. vulgaris and some of its close relatives, such as P. coccineus (26). Indeed, P. vulgaris is a highly promiscuous host that can form nodules (of various degree of effectiveness) with Bradyrhizobium and other genera of nodule bacteria (2, 14, 17, 30). To better understand the evolution of nodule symbiosis in the genus Phaseolus, it will be important in future work to characterize bacteria from a broader sample of lineages within Phaseolus and from related genera (6) and also to test the capacity of Bradyrhizobium-utilizing Phaseolus taxa to form nodules with Rhizobium strains.

While relatives of both B. japonicum and B. elkanii were well represented among the isolates from Chihuahua, about two-thirds of the isolates had genotypes related to B. elkanii. Analysis of both 16S rRNA and partial 23S rRNA sequences indicated very high similarity of these isolates to two strains from eastern North America (th-b2, jwc91-2) associated with the legume A. bracteata (Fig. 1). A. bracteata also occurs in Mexico (22), but its root nodule bacteria have not yet been analyzed in that region. Bradyrhizobium strains closely similar to th-b2 and jwc91-2 also associate with species of Apios and Desmodium in eastern North America (18), suggesting that this group has both a wide geographic distribution and a broad potential host range across papilionoid legumes endemic to this region. Future studies of this group should focus on sampling nodule bacteria from additional legume taxa sympatric with these other genera, in order to better delimit patterns of host use.