FEMS Microbiology Ecology, 91, 2015, fiv075
doi: 10.1093/femsec/fiv075
Advance Access Publication Date: 30 June 2015
Research Article
RESEARCH ARTICLE
Diversifying selection by Desmodiinae legume species
on Bradyrhizobium symbionts
Matthew A. Parker∗ , Jennifer G. Jankowiak† and Grace K. Landrigan
Department of Biological Sciences, State University of New York, Binghamton, NY 13902, USA
∗ Correspondence author: Department of Biological Sciences, State University of New York, Binghamton, NY 13902, USA. Tel: +607-777-6283;
Fax: +607-777-6521; E-mail: [email protected]
Present address: School of Marine and Atmospheric Sciences, Stony Brook University, Stony Brook, NY 11794, USA.
One sentence summary: Differential use of Bradyrhizobium lineages by ten species of Desmodium and Hylodesmum legumes implies that host interactions
promote genetic diversity in these root-nodule symbionts.
Editor: Wietse de Boer
†
ABSTRACT
Desmodium and Hylodesmum (Papilionoideae Subtribe Desmodiinae) are among the most common herbaceous perennial
legumes native to eastern North America. To analyze the population structure of their Bradyrhizobium sp. root-nodule
bacteria, 159 isolates were sampled from ten host species across a 1000 km region. Phylogenetic analysis of four
housekeeping loci (2164 bp) and two loci in the symbiosis island (SI) chromosomal region (1374 bp) indicated extensive
overlap in symbiont utilization, with each common bacterial clade found on 2–7 species of these legume genera. However,
host species differed considerably in the relative proportion of symbionts belonging to different Bradyrhizobium clades. High
phylogenetic incongruence between trees for housekeeping loci and SI loci suggested that diversification of these
Bradyrhizobium lineages involved substantial horizontal gene transfer. Plant inoculation with strains from six Bradyrhizobium
clades revealed marked disparity in relative bacterial reproductive success across four Desmodium species. Estimated yield
of Bradyrhizobium progeny cells per plant ranged from zero to >109 , and strains with high fitness on one host sometimes
reproduced poorly on other host species. Diversifying selection on bacteria, arising from differential success in habitats
with different Desmodium and Hylodesmum taxa, is therefore likely to affect Bradyrhizobium diversity patterns at the
landscape level.
Keywords: horizontal gene transfer; phylogenetic clustering; reproductive success; symbiotic specificity
INTRODUCTION
It is clear that legume plants must play some role in structuring the genetic diversity among their nodule bacteria (rhizobia),
because different legume hosts commonly utilize distinct subsets of a symbiont community (Bena et al. 2005; Miché et al. 2010;
Stepkowski et al. 2012; Ardley et al. 2013; Ehinger et al. 2014). However, in cases where a legume lineage has speciated into multiple taxa that coexist regionally, little is known about the extent
to which plant diversification is associated with changes in symbiont use (Bena et al. 2005; Servı́n-Garcidueñas et al. 2014).
The goal of this study was to analyze utilization of Bradyrhizobium symbionts by individual species in the legume genera
Desmodium and Hylodesmum (Papilionoideae Tribe Desmodieae).
Desmodium is the largest genus of legumes in eastern North
America (19 species recognized in Gleason and Cronquist 1991),
and some species are among the most commonly encountered
herbaceous perennial legumes native to this region. The genus
Hylodesmum was segregated from Desmodium in 2000 (Ohashi
and Mill 2000), and is represented by three species in North
America. We sought to sample sites across a large portion of the
geographic ranges for multiple taxa, including sites where two
Received: 7 April 2015; Accepted: 24 June 2015
C FEMS 2015. All rights reserved. For permissions, please e-mail: [email protected]
1
2
FEMS Microbiology Ecology, 2015, Vol. 91, No. 7
or more species of Desmodium or Hylodesmum coexisted locally,
in order to evaluate the extent of symbiont overlap.
In Bradyrhizobium, symbiosis-related genes are clustered in
a portion of the chromosome termed the symbiosis island (SI)
region, which has hallmarks of being a mobile genetic element
(Kaneko et al. 2002). Markers in the SI region commonly have
a different genealogical history from other genes (Moulin et al.
2004; Stepkowski et al. 2007; Aserse et al. 2012). Thus, we analyzed four housekeeping loci and two loci in the SI region. The
extent of non-randomness in symbiont utilization by legume
species was quantified by a tree permutation method (Webb,
Ackerly and Kembel 2008). This approach tests whether the set
of bacterial strains used by a particular legume taxon are significantly more similar to one another than strains selected at
random from the entire community (Sachs et al. 2009; Koppell
and Parker 2012; Parker 2012). To complement the field surveys
of symbiont utilization, we also performed an inoculation experiment to analyze whether host legumes differentially affected
Bradyrhizobium reproduction.
MATERIALS AND METHODS
Sampling. In searches of 31 sites in six states, we located 67 populations of Desmodium and Hylodesmum. Individual sites had from
one to seven species, and a total of ten host species was sampled
overall (Table S1, Supporting Information). One to three nodule
isolates were obtained per species per site (144 isolates). These
were augmented with prior data for 15 isolates from seven other
sites in the region (Parker 2012), for a total of 159 bacterial isolates. Four Desmodium species less common in the region (D.
canescens, D. ciliare, D. marilandicum and D. strictum) were only
found at one or two sites. The other six species were each sampled at 6–18 sites spanning at least three states. Except in a few
cases, isolates within a site came from separate host plant individuals.
Nodules were surface-sterilized with ethanol and sodium
hypochlorite, and one bacterial isolate was purified from each
nodule (Parker 2012). Genomic DNA was obtained by heating
bacterial cells at 95 ◦ C for 5 min in a lysis buffer containing 1%
Triton X-100, followed by chloroform extraction (Parker 2012).
Portions of four housekeeping loci (dnaK, rpoB, rplC and 23S rRNA)
and two SI region loci (nifD and nodC) were sequenced as described (Parker 2012), representing 3538 bp of sequence information per isolate. The dnaSP software package v5.10 (Librado and
Rozas 2009) was used to estimate nucleotide diversity (π, the average pairwise number of nucleotide differences per site) for the
housekeeping loci and SI loci data sets.
To measure the magnitude of genetic divergence among host
legumes, we sequenced two chloroplast DNA markers [trnL-F region (1049 bp) and trnK intron (399 bp)] and the nuclear ribosomal ITS region (691 bp), as described (Parker, Doyle and Doyle
2004; Heinze 2007). We analyzed one individual from each of two
separate populations for D. paniculatum and D. perplexum, and a
single individual for each of the other eight Desmodium and Hylodesmum species.
Phylogenetic analyses. Four strains representing well-known
Bradyrhizobium taxa were included as references in the tree analyses: B. elkanii USDA 76T ; B. diazoefficiens USDA 110T ; B. japonicum USDA 6T and B. cytisi CTAW11T . We also included 68 other
Bradyrhizobium strains sampled from diverse legume hosts in
the same regions where the Desmodium and Hylodesmum strains
were collected (Table S1, Supporting Information). These strains
came from 19 legume host species in 13 genera (Amphicarpaea,
Apios, Baptisia, Centrosema, Chamaecrista, Clitoria, Galactia, Lespedeza, Lupinus, Rhynchosia, Stylosanthes, Tephrosia and Zornia),
sampled at 22 sites in NY, CT, RI and NC (Parker 2012).
SI and non-SI region loci typically show considerable phylogenetic incongruence (Stepkowski et al. 2007; Koppell and Parker
2012). Thus, we used concatenated data for the four housekeeping loci in one tree analysis, and performed a separate analysis on concatenated data for the two SI region loci (nifD, nodC).
Trees were inferred using MrBayes (Ronquist and Huelsenbeck
2003) with protein coding loci partitioned by codon position and
separate estimates of rates and nucleotide composition for each
codon position and locus. A HKY substitution model was used.
Some alternative models were tested but all yielded very similar
trees. Analyses were run for two million generations, sampling
every 400 generations, with the last 1000 samples saved for tree
analysis. Replicate runs yielded identical consensus tree topologies. Trees were rooted using Azorhizobium caulinodans ORS571
(NC009937) as an outgroup.
ClonalFrame 1.1 (Didelot and Falush 2007) was used to identify gene transfer events altering the genealogy of SI and nonSI regions of the chromosome. The six loci were concatenated according to their order on the B. diazoefficiens USDA110
chromosome (dnaK, 23S rRNA, nifD, nodC, rplC and rpoB). The
ClonalFrame algorithm infers the relationships of strains while
simultaneously estimating the chromosomal position of recombination events disrupting the clonal inheritance in particular
groups of strains. Default settings were used (50 000 burnin generations followed by tree sampling every 100 generations for another 50 000 generations).
Phylocom 4.0.1 (Webb, Ackerly and Kembel 2008) was used
to analyze whether bacteria were phylogenetically clustered according to legume host species. Analyses were performed separately for the nifD/nodC tree and the housekeeping gene tree.
For each set of bacterial strains from a particular host species,
MPDsample (Webb, Ackerly and Kembel 2008) is defined as the average across all strain pairs of the branch length separating them
on the phylogenetic tree. Significance was assessed by comparing MPDsample values to the null distribution inferred from 1000
random permutations of isolate names across tips of the observed tree. The Net Relatedness Index (NRI), a standardized
measure of the magnitude of clustering, was calculated as the
difference between MPDsample and the MPD value for the same
sample size calculated from the null distribution (MPDnull ), divided by the standard deviation of MPDnull .
Symbiotic compatibility. To analyze whether host legumes differentially affected Bradyrhizobium reproduction, seedlings of
four Desmodium species (D. canadense, D. canescens, D. paniculatum and D. perplexum) were inoculated with six strains. The
strains were selected at random from six divergent lineages
found for the two SI region loci (nifD and nodC). Seeds were scarified and surface-disinfected in concentrated sulfuric acid, and
then planted in a 1:3 mixture of sand and rhizobia-free potting soil (n = 12 for each of the 24 combinations of host plant
and bacterial strain). A separate set of 12 seedlings per host
species was grown in the same greenhouse room as uninoculated controls. Seedlings were inoculated six days after planting
with 5 ml of broth culture (containing approximately 109 bacterial cells/ml). Thirteen seedlings died early in the experiment
for unknown reasons; these were excluded from all analyses.
Standard precautions were taken to prevent bacterial contamination, involving a gravel/perlite barrier on the soil surface, and
watering by subirrigation (Wilkinson, Spoerke and Parker 1996).
None of the uninoculated control plants had nodules when harvested, indicating that these procedures effectively prevented
Parker et al.
contamination across treatments. Plants were rotated to new
room positions every 7–10 days (to randomize exposure across
local light intensity heterogeneity within the greenhouse), and
harvested five weeks after inoculation. At harvest, nodules were
counted, and the diameter of 10 nodules per plant was measured
using an ocular micrometer at 6× magnification (five nodules
nearest to each of two randomly chosen coordinates on the root
system). Total plant dry biomass was then measured.
To characterize the effect of nodule size on Bradyrhizobium
reproduction (Simms et al. 2006), seven nodules of varying diameter were surface-disinfected, crushed in 500 μl of phosphate buffer, serially diluted, and then colony-forming units
were counted after growth on yeast mannitol agar plates. The
linear regression of ln(bacterial population size) on nodule diameter had an R2 value of 0.76, with the estimated slope (2.05 ±
0.52, P = 0.011) indicating that a 1 mm increase in nodule diameter was associated with a 7.75-fold increase in bacterial numbers. This regression was used to estimate mean Bradyrhizobium
numbers per nodule from a plant’s average nodule diameter. Total Bradyrhizobium reproduction per plant was then calculated as
the product of nodule number and inferred bacterial population
size per nodule.
GenBank accession numbers for the 900 new sequence
records obtained in this study, and for the 522 other existing sequence records utilized, are provided in Table S2 (Supporting Information).
3
more recent for the Desmodium species relative to these Hylodesmum taxa.
Higher diversity in Desmodium symbionts
SI loci exhibited higher nucleotide diversity than housekeeping loci (Table 1), consistent with previous studies of Bradyrhizobium in this region and elsewhere (Parker 2012, Parker and
Rousteau 2014). In addition, Bradyrhizobium strains associated
with Desmodium had more than 2-fold higher nucleotide diversity, for both SI and non-SI loci, compared to Hylodesmum symbionts (Table 1). This was not due to the fact that the Desmodium
sample came from more species. Each of the four best-sampled
Desmodium species (D. canadense, D. paniculatum, D. perplexum and
D. rotundifolium) individually had more nucleotide diversity than
was detected in the Hylodesmum symbionts. Remarkably, diversity in the Desmodium symbionts was also substantially higher
than in a pooled set of Bradyrhizobium strains from 13 other
legume genera sampled in the same region (Table 1). Rhizobial
nucleotide diversity is a potential indicator of whether a host is
specialized versus promiscuous in symbiont acquisition. These
results suggest that species of Desmodium utilize more diverse
Bradyrhizobium strains than do other cooccurring legume taxa in
this region.
Symbiont lineages are widely shared across
host species
RESULTS
Host legume relationships
A phylogenetic network using the NeighborNet algorithm (implemented in Splitstree 4.11.3, Huson and Bryant 2006) on concatenated chloroplast and nuclear ribosomal ITS data indicated
that Desmodium and Hylodesmum were strongly differentiated
(Fig. 1). There was limited sequence divergence among the
Desmodium species (mean: 0.5%; range: 0.2–0.9%), and more than
4-fold higher nucleotide divergence between the two Hylodesmum species (2.3%). This suggests that in eastern North America, the time frame for speciation and divergence has been much
Bayesian analysis yielded a well-resolved tree for the concatenated housekeeping gene data (Fig. 2, left). Isolates in eight major clades were grouped to more easily depict the tree topology,
with clade representation on individual host legumes shown
in Table 2. About 80% of the Desmodium and Hylodesmum isolates (128/159) grouped into a set of five clades (a,b,c,d and e) related to B. elkanii (Table 2). All but two of the remaining isolates
(29/159) belonged to clade g, related to B. japonicum USDA 6. Clade
prevalence was highly unequal, with some clades having fewer
than five isolates, and one clade (e) encompassing 77 Desmodium and Hylodesmum isolates. The tree suggested that there was
little symbiotic exclusivity among these organisms. All of the
D. strictum
D. marilandicum
D. ciliare
D. pan.1
D. rotundifolum
D. pan.2
D. canad
D. canes D. perp.1
D. perp.2
H. glutinosum
0.01
H. nudiflorum
Figure 1. Phylogenetic network inferred using NeighborNet for concatenated trnL-F, trnK intron and nrITS (2139 bp) in Desmodium and Hylodesmum legumes. Individuals
from two separate populations were analyzed for D. paniculatum (abbreviated D. pan.1 and D. pan.2) and D. perplexum (D. perp.1 and D. perp.2). Other abbreviations: D.
canad = D. canadense and D. canes = D. canescens. Scale bar depicts 1% sequence divergence.
4
FEMS Microbiology Ecology, 2015, Vol. 91, No. 7
Table 1. Nucleotide polymorphism summary for Bradyrhizobium strains from Desmodium, Hylodesmum and 13 other legume genera in eastern
North America.
Legume host taxon
N
D. canadense
D. paniculatum
D. perplexum
D. rotundifolium
Four rare Desmodium sp.b
All Desmodium pooled
H. glutinosum
H. nudiflorum
All Hylodesmum pooled
strains from thirteen other cooccurring legume generac
20
29
19
17
12
97
35
27
62
68
Four housekeeping loci
π a ± 1 SD
0.057 ±
0.057 ±
0.043 ±
0.050 ±
0.047 ±
0.052
0.016 ±
0.019 ±
0.019
0.040 ±
0.004
0.004
0.009
0.009
0.011
0.007
0.001
0.005
Two symbiotic loci
π ± 1 SD
0.146 ±
0.141 ±
0.092 ±
0.125 ±
0.089 ±
0.134
0.023 ±
0.072 ±
0.058
0.093 ±
0.010
0.007
0.021
0.014
0.028
0.014
0.004
0.010
a
Average pairwise number of nucleotide differences per site.
D. canescens (3), D. ciliare (4), D. marilandicum (2) and D. strictum (3).
Bradyrhizobium strains from 19 legume species in 13 genera (Amphicarpaea, Apios, Baptisia, Centrosema, Chamaecrista, Clitoria, Galactia, Lespedeza, Lupinus, Rhynchosia,
Stylosanthes, Tephrosia and Zornia), sampled at 22 sites in NY, CT, RI and NC (Parker 2012).
b
c
Figure 2. Bayesian trees for four concatenated housekeeping loci (2164 bp; left side) and concatenated nifD/nodC loci (1374 bp; right side) in 231 Bradyrhizobium isolates.
Lines linking the two trees show combinations of housekeeping gene clade and SI gene clade that were found in particular strains. Within-clade disparity in branch
length is indicated by triangle size (see Figs S1and S2, Supporting Information for complete topology, strain identity and clade posterior probability values). All branches
shown had a posterior probability of 0.98–1.00, except for five branches marked with numerals (which had posterior probabilities of 0.58, 0.93, 0.73, 0.79 and 0.81,
respectively).
larger clades included isolates from two or more Desmodium or
Hylodesmum host species, and every well-sampled host species
had symbionts from multiple clades. Moreover, every clade with
Desmodium or Hylodesmum isolates also included strains from
other legume host genera as well (Table 2). Finer-scale analysis
at the level of haplotypes (multilocus sequence variants) reinforced the view that symbionts were widely shared. Ten haplotypes (out of 67 total Desmodium and Hylodesmum haplotypes)
proved to be identical at every locus to particular isolates sampled from other legume genera in the region. The four most common haplotypes in the sample were each found on three or more
Desmodium/Hylodesmum species, as well as on legumes in other
genera.
The Bayesian tree for concatenated SI loci (nifD and nodC)
had a rather different topology (Fig. 2, right), although for
many isolates, their closest relatives were the same as in the
Parker et al.
5
Table 2. Number of Bradyrhizobium isolates from eight Desmodium and two Hylodesmum legumes falling into eight clades in the concatenated
housekeeping gene tree (Fig. 2). The clade distribution of 68 strains from 13 other eastern North American legume genera (Parker 2012) is shown
in the bottom row.
Bradyrhizobium clade
Legume species
a
D. canadense
D. canescens
D. ciliare
D. marilandicum
D. paniculatum
D. perplexum
D. rotundifolium
D. strictum
H. glutinosum
H. nudiflorum
Des./Hylodes. total
13 other legume genera
2
∗
3
5
26
b
c
d
e
f
g
10
4
1
7
2
3
3
12
32
19
3
1
3
1
8
3
1
7
10
9
2
5
3
11
2
h
1
28
12
77∗
10
1
3
9
4
4
1
1
29∗ ∗
5
1
2
Total
20
3
4
2
29
19
17
3
35
27
159
68
for housekeeping gene clade e, 63 isolates had SI genes in clade E, and the others had SI clade D [10 isolates], SI clade F [1 isolate], or SI clade K [3 isolates].
for housekeeping gene clade g, 22 isolates had SI genes in clade G, and the others had SI clade B [2 isolates] or SI clade I [5 isolates].
∗∗
same host species
0.6
fraction of strain pairs
housekeeping gene tree. For example, most strains with housekeeping genes related to B. elkanii (clades a,b,c,d and e) carried SI
genes from four related clades (C, D, E and F). Three clades in the
nifD/nodC tree were very large (D, E and G) and each included isolates from six or more Desmodium and Hylodesmum species. On
the finer scale of individual haplotypes, there were six nifD/nodC
composite haplotypes among the Desmodium and Hylodesmum
isolates (out of 57 total haplotypes) that were each identical to
strains from other legume genera in the region. These six shared
haplotypes represented 43% of all Desmodium and Hylodesmum
isolates (69/159), and included cases of strain identity to symbionts from five other legume genera (Amphicarpaea, Chamaecrista, Rhynchosia, Stylosanthes and Tephrosia). Thus, certain SI
variants common among symbionts of Desmodium and Hylodesmum also associate with legumes in a variety of other genera as
well.
The trees for housekeeping genes and SI loci were highly
incongruent (Fig. 2). Isolates that were close relatives in one
tree were sometimes partitioned into two or more divergent
clades in the other tree. This is consistent with other studies of
Bradyrhizobium (Aserse et al. 2012; Stepkowski et al. 2012; Parker
and Rousteau 2014), and suggests that horizontal gene transfer has acted to redistribute SI variants across chromosomal
backgrounds. This inference was corroborated by a ClonalFrame
analysis (Didelot and Falush 2007), which detected a very high
likelihood of SI region import in many of these strain groups
(Fig. S3, Supporting Information).
The high apparent prevalence of SI lateral transfer (Fig. 2,
Fig. S3, Supporting Information) raises the question of whether
host-associated selection on SI variants may play some role in
facilitating the establishment of strains that have acquired a
novel SI sequence type. To analyze this issue, we selected all isolate pairs that were closely related at housekeeping loci (<0.5%
nucleotide divergence), and examined whether any pairs exhibited unusually high divergence at SI loci. A large majority of the
strain pairs that were close relatives at housekeeping loci also
had very low SI nucleotide divergence as well (median = 0.1%;
interquartile range, 0–0.2%). However, a few strain pairs had
highly divergent SI genes (up to 21.2% different). High SI divergence was found for 3.6% of strain pairs sampled from the same
host species, and for 15.0% of pairs that came from different
legume species (Fig. 3; χ 2 = 54.13, P < 0.0001). Thus, if two closely
di erent host
0.4
0.2
zero
0.1 - 1.0%
1.1 - 13.0% 13.1 - 21.2%
nifD/nodC nucleotide divergence
Figure 3. SI gene divergence among strain pairs that were closely related at
housekeeping loci (nucleotide divergence <0.5%), partitioned by whether the
pair came from the same Desmodium or Hylodesmum host species (n = 548), or
from different hosts (n = 1549).
related strains were associated with different Desmodium or Hylodesmum species, they had a substantially elevated chance of
carrying strongly divergent SI genes. This suggests that transfer
to a novel host may sometimes be associated with acquisition
of a novel SI sequence type.
Unequal phylogenetic clustering for Desmodium
versus Hylodesmum symbionts
Tree permutation analyses indicated that symbionts from the
two Hylodesmum species exhibited substantial phylogenetic
clustering (Table S3, Supporting Information). These legumes
thus harbored bacteria that were significantly more similar, on
average, relative to the null distribution generated by permuting strains across tips on each tree. However, symbiont clustering was not significant in either tree for any of the four
well-sampled Desmodium species (D. canadense, D. paniculatum,
D. perplexum and D rotundifolium). These legumes utilized so
many diverse symbiont lineages that their clustering index
6
FEMS Microbiology Ecology, 2015, Vol. 91, No. 7
DISCUSSION
Figure 4. Reproductive success for six Bradyrhizobium strains on four Desmodium host species. Strains 1–6, with their housekeeping gene clade/SI gene clade
(Fig. 2) in brackets after the strain name, are desb1[e/E], dxfc2[g/B], dpby1[b/C],
drkt4[e/D], dcby2[e/K] and dekt3[g/G]). Data are the mean and 95% CI for the yield
of bacterial cells per plant (the product of nodule number, nodule diameter and
the number of cells per unit nodule diameter).
(NRI) was not distinguishable from the null distribution involving strain pairs selected at random across the entire dataset.
For the four rarer Desmodium species (D. canescens, D. ciliare,
D. marilandicum and D. strictum), no conclusions about phylogenetic clustering are warranted given the limited sample
sizes.
Differential Bradyrhizobum reproductive success
with Desmodium species
Nodule numbers, nodule size and plant biomass all varied significantly when four Desmodium species were inoculated with
six Bradyrhizobium strains representing divergent housekeeping
gene and SI lineages (Table S4, Supporting Information). Nodules developed in all plant/bacterial combinations except for
the interaction of strain desb1 with D. canescens, where nodules were entirely absent. Among the other 23 combinations of
strain and host legume, there was a 5.7-fold range in average
nodule numbers per plant, and a 3.5-fold range in mean nodule diameter. Estimated total bacterial reproduction per plant
varied from 0 to 1.1 billion for bacterial strains interacting with
D. canescens (Fig. 4). On the other three Desmodium species, there
was a smaller range of variation across strains in bacterial reproductive success (100–540 million cells/plant). The relative reproductive success of individual strains clearly differed across the
four Desmodium species (Fig. 4). These results imply that habitat
patches dominated by different host legumes are likely to favor
different sets of bacterial lineages. Across the 24 combinations
of plants and bacterial strains, plant biomass was positively correlated with mean bacterial reproductive success (Fig. S4, Supporting Information; Spearman rs [22 d.f.] = 0.64, P < 0.001).
Thus, plants tended to favor proliferation of those strains that
conferred better growth.
None of the legume species harbored symbionts that were
wholly unique, but hosts clearly differed in the suite of bacterial lineages that were utilized (Table 2). Diversifying selection
on symbiont populations is thus likely to occur across habitats
occupied by different Desmodium and Hylodesmum taxa.
The legume D. canescens, which showed the most striking
disparity in symbiotic outcomes of any host in the inoculation study (Fig. 4), illustrates how hosts may affect Bradyrhizobium population composition. We were only able to obtain nodule samples from a single D. canescens population, and found
only one Bradyrhizobium genotype (housekeeping gene clade
g/SI clade G) on these plants. We found that strain dekt3 from
D. canescens had unusually high reproductive success on this
legume, producing nearly five times more progeny than the average value for other combinations of plants and bacterial strains
(Fig. 4). By contrast, D. canescens plants proved to be unable to
form nodules at all with a strain from the Bradyrhizobium clade
(e/E) that was most common across the sample as a whole. Clade
g/G strains occurred frequently on several Desmodium species
(Table 2), and on three of these species, clade g/G strain dekt3
had either the highest or second-highest reproductive success of
any of the bacteria tested (Fig. 4). Thus, proliferation of Bradyrhizobium clade g/G is likely in habitats occupied by several Desmodium taxa.
By contrast, clade g/G bacteria were rare or absent on both
Hylodesmum taxa (Table 2), which mostly utilized a few specific Bradyrhizobium clades related to B. elkanii. This apparent
symbiotic selectivity in Hylodesmum is illustrated by the finding
of lower symbiont genetic diversity (Table 1), and significantly
higher phylogenetic relatedness of Hylodesmum symbionts relative to Desmodium (Table S3, Supporting Information). It would
be of interest to perform inoculation studies with Hylodesmum
hosts to experimentally characterize their symbiotic behavior.
Unfortunately, the regulation of Hylodesmum seed germination is
currently not well understood, so it was not possible to include
them along with Desmodium hosts in the inoculation study.
There was no evidence that a particular SI sequence type was
required for nodulation of any of these legume species, because
all of the well-sampled hosts associated with multiple SI lineages. Also, all four Desmodium species in the inoculation study
nodulated with strains representing diverse SI clades (Fig. 4).
However, in view of the high phylogenetic incongruence of SI
versus non-SI loci (Fig. 2), it is important to consider whether
host-associated selection on SI variants may contribute to reassortment of SI sequence type across Bradyrhizobium lineages.
A growing body of work indicates that horizontal transfer of
symbiosis-related genes can be an important factor in bacterial adaptation to legume hosts colonizing a new geographic region (Sullivan et al. 1995; Nandasena et al. 2006; Wei et al. 2009;
Horn et al. 2014). Similar processes may also occur across habitats occupied by different hosts within a region (Parker 2012). If
a bacterial strain disperses into a site where its current SI type
makes it unsuccessful in nodulating the legumes in its immediate vicinity, then acquisition of an alternative SI variant from
resident rhizobia may improve its fitness with local hosts. Thus,
host switching may be associated with an increased likelihood
of SI gene transfer. This may explain the observation that related
strains sampled from different hosts had a higher probability of
carrying divergent SI genes, compared to strain pairs from the
same host species (Fig. 3).
As in some previous analyses (Bala and Giller 2001; Bena
et al. 2005; Servı́n-Garcidueñas et al. 2014), we found that host
Parker et al.
legume divergence (Fig. 1) was not correlated with divergence
in symbiont use. For example, H. nudiflorum and D. perplexum
had highly similar symbiont assemblages (Table 2), despite being not at all closely related. Also, the legumes D. canadense and
D. canescens were among the most closely related species pair of
any of the hosts examined (Fig. 1), yet they had strikingly different responses to inoculation with six Bradyrhizobium lineages
(Fig. 4).
If rhizobia that are adapted to one host taxon always have
lowered success interacting with others, then different legumes
will function as distinct niches that each favors a different set
of bacteria. Local variation in legume community composition
will then contribute to diversifying selection on bacteria (Rangin et al. 2008; Parker 2012). These effects will be amplified if the
bacteria that are favored by a particular legume act in turn to
increase the plant’s ecological success (Ehinger et al. 2014).
The current results indicate an apparent lack of bacterial specialization at the level of individual legume species. However,
hosts nevertheless showed wide variation in symbiont use (Table 2, Fig. 4). More complex forms diversifying selection, involving differential success of strains that are shared across multiple legume taxa, may thus occur in these communities. Studies
on how bacterial population composition varies across habitat
patches dominated by different host taxa, combined with inoculation experiments and broader surveys of symbiont utilization,
will provide deeper insights into the processes structuring rhizobial genetic diversity.
SUPPLEMENTARY DATA
Supplementary data is available at FEMSEC online.
ACKNOWLEDGEMENTS
We are grateful to Mark Blumler, Julian Shepherd and Bryan Connolly for assistance in locating Desmodium populations.
FUNDING
This work was supported by the National Science Foundation
[grant MCB-0640246].
Conflict of interest. None declared.
REFERENCES
Ardley JK, Reeve WG, O’Hara GW, et al. Nodule morphology,
symbiotic specificity and association with unusual rhizobia are distinguishing features of the genus Listia within
the Southern African crotalarioid clade Lotononis s.l. Ann Bot
2013;112:1–15.
Aserse AA, Räsänen LA, Aseffa F, et al. Phylogenetically diverse
groups of Bradyrhizobium isolated from nodules of Crotalaria
spp., Indigofera spp., Erythrina brucei and Glycine max growing
in Ethiopia. Mol Phylogenet Evol 2012;65:595–609.
Bala A, Giller KE. Symbiotic specificity of tropical tree rhizobia
for host legumes. New Phytol 2001;65:495–507.
Bena G, Lyet A, Huguet T, et al. Medicago—Sinorhizobium symbiotic
specificity evolution and the geographic expansion of Medicago. J Evolution Biol 2005;18:1547–58.
Didelot X, Falush D. Inference of bacterial microevolution using multilocus sequence data. Genetics 2007;175:
1251–66.
7
Ehinger M, Mohr TJ, Starcevich JB, et al. Specializationgeneralization trade-off in a Bradyrhizobium symbiosis with
wild legume hosts. BMC Ecol 2014;14:8.
Gleason HA, Cronquist A. Manual of Vascular Plants of Northeastern
United States and Adjacent Canada. New York, USA: Botanical
Garden Press, 1991.
Heinze B. A database of PCR primers for the chloroplast genomes
of higher plants. Plant Methods 2007;3:4.
Horn K, Parker IM, Malek W, et al. Disparate origins of Bradyrhizobium symbionts for invasive populations of Cytisus scoparius (Leguminosae) in North America. FEMS Microbiol Ecol
2014;89:89–98.
Huson DH, Bryant D. Application of phylogenetic networks in
evolutionary studies. Mol Biol Evol 2006;23:254–67.
Kaneko T, Nakamura Y, Sato S, et al. Complete genomic sequence of nitrogen-fixing symbiotic bacterium Bradyrhizobium japonicum USDA110. DNA Res 2002;9:189–97.
Koppell JH, Parker MA. Phylogenetic clustering of Bradyrhizobium
symbionts on legumes indigenous to North America. Microbiology 2012;158:2050–9.
Librado P, Rozas J. DnaSP v5: A software package for comprehensive analysis of DNA polymorphism data. Bioinformatics
2009;25:1451–2.
Miché L, Moulin L, Chaintreuil C, et al. Diversity analyses
of Aeschynomene symbionts in Tropical Africa and Central
America reveal that nod-independent stem nodulation is not
restricted to photosynthetic bradyrhizobia. Environ Microbiol
2010;12:2152–64.
Moulin LG, Bena G, Boivin-Masson C, et al. Phylogenetic analyses of symbiotic nodulation genes support vertical and lateral gene co-transfer within the Bradyrhizobium genus. Mol
Phylogenet Evol 2004;30:720–32.
Nandasena KG, O’Hara GW, Tiwari RP, et al. Rapid in situ evolution of nodulating strains for Biserrula pelecinus L. through lateral transfer of a symbiosis island from the original mesorhizobial inoculant. Appl Environ Microb 2006;72:7365–7.
Ohashi H, Mill RR. Hylodesmum, a new name for Podocarpium
(Leguminosae). Edinburgh J Bot 2000;57:171–88.
Parker MA. Legumes select symbiosis island sequence variants
in Bradyrhizobium. Mol Ecol 2012;21:1769–78.
Parker MA, Doyle JL, Doyle JJ. Comparative phylogeography of
Amphicarpaea legumes and their root-nodule symbionts in
Japan and North America. J Biogeogr 2004;31:425–34.
Parker MA, Rousteau A. Mosaic origins of Bradyrhizobium legume
symbionts on the Caribbean island of Guadeloupe. Mol Phylogenet Evol 2014;77:110–5.
Rangin C, Brunel B, Cleyet-Marel JC, et al. Effects of Medicago truncatula genetic diversity, rhizobial competition, and
strain effectiveness on the diversity of a natural sinorhizobium species community. Appl Environ Microb 2008;74:
5653–61.
Ronquist F, Huelsenbeck JP. MRBAYES 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 2003;19:
1572–4.
Sachs JL, Kembel SW, Lau AH, et al. In situ phylogenetic structure and diversity of wild Bradyrhizobium communities. Appl
Environ Microb 2009;75:4727–35.
Servı́n-Garcidueñas LE, Zayas-Del Moral A, Ormeño-Orrillo E,
et al. Symbiont shift towards Rhizobium nodulation in a group
of phylogenetically related Phaseolus species. Mol Phylogenet
Evol 2014;79:1–11.
Simms EL, Taylor DL, Povich J, et al. An empirical test of partner
choice mechanisms in a wild legume–rhizobium interaction.
P R Soc Biol Sci 2006;273:77–81.
8
FEMS Microbiology Ecology, 2015, Vol. 91, No. 7
Stepkowski T, Hughes CE, Law IJ, et al. Diversification of lupine
Bradyrhizobium strains: evidence from nodulation gene trees.
Appl Environ Microb 2007;73:3254–64.
Stepkowski T, Watkin E, McInnes A, et al. Distinct Bradyrhizobium communities nodulate legumes native to temperate
and tropical monsoon Australia. Mol Phylogenet Evol 2012;63:
265–77.
Sullivan JT, Patrick HN, Lowther WL, et al. Nodulating strains
of Rhizobium loti arise through chromosomal symbiotic
gene transfer in the environment. P Natl Acad Sci USA
1995;92:8985–9.
Webb CO, Ackerly DD, Kembel SW. Phylocom: software for the
analysis of phylogenetic community structure and trait evolution. Bioinformatics 2008;24:2098–100.
Wei G, Chen W, Zhu W, et al. Invasive Robinia pseudoacacia in China is nodulated by Mesorhizobium and Sinorhizobium species that share similar nodulation genes with
native American symbionts. FEMS Microbiol Ecol 2009;
68:320–8.
Wilkinson HH, Spoerke JM, Parker MA. Divergence in symbiotic
compatibility in legume-Bradyrhizobium mutualism. Evolution
1996;50:1470–7.