American origin of Cupriavidus bacteria associated with invasive Mimosa legumes in the Philippines

Authors


Correspondence: Matthew A. Parker, Department of Biological Sciences, State University of New York, Binghamton, NY 13902, USA. Tel.: (607) 777 6283; fax: (607) 777 6521; e-mail: mparker@binghamton.edu

Abstract

To identify the origins of Cupriavidus nodule symbionts associated with two invasive Mimosa species in the Philippines, 22 isolates were sequenced for portions of three chromosomal genes and two symbiotic plasmid loci. Eleven isolates were identical at all gene loci (2713 bp) to a lineage found in Central America. Four other Philippine isolates were identical to a second Cupriavidus lineage distributed both in Central America and in the Caribbean. None of the remaining Philippine strains had more than 0.6% sequence divergence from American Cupriavidus lineages. These results imply that the Philippine population was founded by multiple introductions from the native range of their Mimosa hosts.

Cupriavidus nodule symbionts were first discovered in Taiwan in 2001, from the legumes Mimosa pudica and Mimosa diplotricha, which are indigenous to the Americas (Chen et al., 2001, 2003; Gyaneshwar et al., 2011). Subsequently, Cupriavidus was found associated with invasive Mimosa species in India (Verma et al., 2004), Papua New Guinea (Elliott et al., 2009), and Yunnan Province of China (Liu et al., 2011). Within the native range of M. pudica in the Neotropics, Cupriavidus was identified as a nodule symbiont in 2006 (Barrett & Parker, 2006) and has also been sampled from other Mimosa species native to North and South America (Barrett & Parker, 2006; Andam et al., 2007; Mishra et al., 2012). Based on rRNA gene sequence variation, Cupriavidus strains from India, Papua New Guinea, Taiwan, and Yunnan Province are genetically similar to Costa Rican strains, suggesting that they were introduced into Asian habitats along with their Mimosa hosts (Barrett & Parker, 2006). However, the number of genetic markers studied and the number of locations sampled are currently too limited to be able to establish definitive links between particular Cupriavidus strains in Asia vs. the Americas.

This study analyzed the origins of Cupriavidus symbionts associated with two invasive Mimosa species (M. diplotricha and M. pudica) in the Philippines. We compared Philippine bacteria to Cupriavidus populations from three regions in the native range of their Mimosa hosts (Costa Rica, Puerto Rico, Texas; Barrett & Parker, 2006; Andam et al., 2007) to analyze (1) whether any Philippine strains were identical to American strains, indicating recent trans-Pacific colonization and (2) whether the origin of particular Philippine strains could be traced to specific regions in the Americas.

Nodules were collected from M. pudica and Mdiplotricha in two locations 5 km apart near Mount Makiling in Laguna Province, Luzon. Standard methods were used for bacterial isolation, DNA purification, and PCR (Parker et al., 2002). Portions of two symbiotic plasmid loci, nifH (473 bp) and nodA (488 bp), were sequenced in 54 Cupriavidus strains (see Supporting Information, Table S1 for Cupriavidus strain information). Partial sequences for nifH and nodA sequences were also obtained for two Philippine Burkholderia strains and three Burkholderia strains from Panama (see Table S2 for primers). nifH and nodA loci had identical tree topologies with no homoplasy, so a combined phylogenetic analysis was performed on concatenated sequences. Six reference strains were included in the phylogenetic analysis: Burkholderia tuberum STM678 (AJ302315, AJ302321), Burkholderia phymatum STM815 (NC010627), Burkholderia sp. WSM3937 (EU219869, EU219867), Burkholderia sp. PAS44 (AY883420, EU434822), Cupriavidus taiwanensis LMG19424 (CU633751), and Cupriavidus sp. PAS15 (AY752962, AY752964). Three alpha-rhizobial strains were used as outgroups: Azorhizobium caulinodans ORS571 (NC009937), Bradyrhizobium japonicum USDA110 (NC004463), and Bradyrhizobium elkanii USDA76 (AB094963, AM117554). Trees were inferred using mrbayes 3.1 (Ronquist & Huelsenbeck, 2003) using a HKY substitution model with parameters estimated separately by locus and by codon position. The analysis was run for 500 000 generations with a 250 000 generation burn-in period.

Portions of two chromosomal loci, recA (701 bp) and rpoB (661 bp), were sequenced for all 54 Cupriavidus strains using primers listed in Table S2. GenBank accession numbers are provided in Table S3. dnasp v5 (Librado & Rozas, 2009) was used to compare nucleotide diversity for strains from the four geographic regions. Bayesian analysis followed the same approach used for symbiotic loci with seven data partitions (three codon positions within recA and within rpoB, and 23S rRNA gene sequences assigned to a seventh partition). Both Cupriavidus taiwanesis LMG19424 and the nonsymbionts C. necator N1 and Cupriavidus  necator JMP134 were included as reference strains.

PCR amplification of the 5′ 23S ribosomal RNA region (Barrett & Parker, 2006) detected four length variants among 262 Philippine isolates from Mimosa. A 432-bp-length variant typical of beta-rhizobia (Cupriavidus or Burkholderia) was observed in 67% of the isolates (175/262). The other three length variants (474, 561, 639 bp) were represented by 23, 63, and 1 isolate, respectively. Fifty six isolates, representing all length variants, were sequenced. BLAST searches indicated that Cupriavidus predominated among the 44 strains with the 432-bp-length variant (40 Cupriavidus, 4 Burkholderia). For all 12 sequences from the other length variants, various strains of Rhizobium proved to be the most similar. A nearly full-length portion of the 16S rRNA gene (1452 bp) was sequenced from six strains with the 432-bp 23S rRNA gene length variant to confirm generic identity. All proved to be closely related to symbiotic Burkholderia and Cupriavidus strains from Central America (Barrett & Parker, 2006).

Analysis of two Cupriavidus symbiotic plasmid loci (nifH and nodA) detected a total of only seven haplotypes (Fig. S1). All 22 Philippine Cupriavidus isolates had nifH and nodA sequences identical to two of the American nifH/nodA haplotypes. Most Philippine Cupriavidus strains (21/22) had a nifH and nodA sequence variant (type AA1) common in both Costa Rica and Puerto Rico. This variant is also found in the C. taiwanensis type strain LMG19424 (Amadou et al., 2008). One Philippine strain (pp2.75; Table S1) had a nifH/nodA sequence variant (type BB) present in a few Puerto Rico strains. Chen et al. (2005) also detected this nifH/nodA sequence variant in a Cupriavidus strain from Taiwan (GenBank AY752962, AY752964). The AA1 and BB haplotypes differed by 15/473 bp in nifH and by 23/488 sites within nodA. All of the Cupriavidus nifH/nodA haplotypes clustered into a strongly supported clade that had affinities to nodule symbiotic Burkholderia strains from Central and South America (Fig. S1). nifH and nodA were also sequenced in two Philippine Burkholderia isolates (mpp4 and mpp5). Both isolates proved to have nifH and nodA sequences identical to two Burkholderia strains from Panama (Fig. S1).

For the 54 Cupriavidus strains, combined nucleotide diversity for the three chromosomal loci (23S rRNA, recA, rpoB) was similar across geographic localities (Table S4). For the two symbiotic plasmid loci, strains from Costa Rica had low nucleotide diversity, and the Philippines had modest diversity compared with the two remaining populations. Thus, there was no evidence in the form of reduced nucleotide diversity that a population bottleneck took place during colonization of the Philippines.

Bayesian phylogenetic analysis of concatenated recA, rpoB, and 23S rRNA gene sequences indicated that Philippine Cupriavidus strains had close affinities to five separate lineages of American Cupriavidus symbionts (Fig. 1). The most common Philippine multilocus haplotype (haplotype 1, found in 11/22 Philippine Cupriavidus isolates) was identical to that of two Costa Rican Cupriavidus isolates. Four Philippine isolates shared a second haplotype (haplotype 2), which was also detected in Costa Rica and Puerto Rico. Four multilocus genotypes were seen among the remaining seven Philippine isolates. While none of these was identical to any of the American Cupriavidus strains sampled, each of these genotypes proved to be very closely related (> 99% nucleotide similarity) to Cupriavidus strains from Costa Rica or Puerto Rico (Fig. 1).

Figure 1.

Bayesian tree for concatenated recA,rpoB, and 23S rRNA Cupriavidus gene sequences (1752 bp) from 22 Philippine isolates (marked by bold font) and 35 isolates from other locations. Branches with Bayesian posterior probabilities > 0.95 are marked by dots. nifH/nodA haplotypes are indicated on the right. Composite haplotypes 1 and 2 (defined in Table S1) were shared by 13 and 6 isolates, respectively.

There was substantial symbiont overlap for the two Philippine Mimosa host species. The 11 isolates analyzed from M. diplotricha (Table S1) all had multilocus haplotypes 1 or 2. Both of these common haplotypes also included Philippine isolates sampled from M. pudica as well.

The distribution of symbiotic plasmid variants (nifH/nodA haplotypes) across chromosomal lineages (Fig. 1) suggested that horizontal transfer of plasmid variants has taken place during the evolution of these Cupriavidus strains. The overall tree topology for nifH/nodA variants (Fig. S1) was quite different from the tree for the three chromosomal genes (Fig. 1). In some cases, two or more variant nifH/nodA types were present within a well-defined monophyletic group in the chromosomal gene tree. Also, one particular symbiotic plasmid haplotype (AA1) was distributed across several divergent chromosomal clades. Thus, it appears that one sym plasmid variant has recently spread across Cupriavidus lineages that are not close relatives at other loci.

In conclusion, this is one of the first studies to document precise genotypic matches between rhizobia in a legume's invaded habitat and in its ancestral range. More than half of the Philippine strains analyzed proved to be completely identical for five marker loci (2713 bp) to Cupriavidus strains from the Americas, and the remaining strains were closely similar to American Cupriavidus as well. These data rule out the possibility that Philippine rhizobia associated with Mimosa are descendants from a single migration event, because the Cupriavidus strains comprised five distinct groups that were each linked to different lineages indigenous to Central America or the Caribbean (Fig. 1). Establishment of the Philippine population evidently did not involve new plasmid transfer events, because all of the combinations of chromosomal lineages and plasmid types detected in the Philippines (Fig. 1) were also present in the ancestral range. Further studies involving more widespread sampling will be helpful to increase understanding of bacterial migration and invasion processes.

Acknowledgements

We are grateful to J. Pfeil for assistance with sequencing. We thank A. Salazar for help with sampling in the Philippines. Funding was provided by NSF grant MCB-0640246.

Ancillary