Invasive legumes can associate with many mutualists of native legumes, but usually do not

Abstract Mutualistic interactions can strongly influence species invasions, as the inability to form successful mutualisms in an exotic range could hamper a host's invasion success. This barrier to invasion may be overcome if an invader either forms novel mutualistic associations or finds and associates with familiar mutualists in the exotic range. Here, we ask (1) does the community of rhizobial mutualists associated with invasive legumes in their exotic range overlap with that of local native legumes and (2) can any differences be explained by fundamental incompatibilities with particular rhizobial genotypes? To address these questions, we first characterized the rhizobial communities naturally associating with three invasive and six native legumes growing in the San Francisco Bay Area. We then conducted a greenhouse experiment to test whether the invasive legume could nodulate with any of a broad array of rhizobia found in their exotic range. There was little overlap between the Bradyrhizobium communities associated with wild‐grown invasive and native legumes, yet the invasive legumes could nodulate with a broad range of rhizobial strains under greenhouse conditions. These observations suggest that under field conditions in their exotic range, these invasive legumes are not currently associating with the mutualists of local native legumes, despite their potential to form such associations. However, the promiscuity with which these invading legumes can form mutualistic associations could be an important factor early in the invasion process if mutualist scarcity limits range expansion. Overall, the observation that invasive legumes have a community of rhizobia distinct from that of native legumes, despite their ability to associate with many rhizobial strains, challenges existing assumptions about how invading species obtain their mutualists. These results can therefore inform current and future efforts to prevent and remove invasive species.

The set of organisms with which a host could form mutualistic associations-its potential mutualistic associates (PMA)-could critically determine whether an exotic species becomes invasive (McGinn et al., 2016;Nunez et al., 2009;Pringle et al., 2009;Traveset & Richardson, 2014). While a promiscuous invader might adopt the existing community of mutualists available within its novel range (Dickie, Bolstridge, Cooper, & Peltzer, 2010;Parker, 2001b;Rodriguez-Echeverria, 2009;Rodriguez-Echeverria, Le Roux, Crisostomo, & Ndlovu, 2011), an invader with a narrow set of PMA might require familiar, closely coevolved mutualists (i.e., those the host has previously encountered in its native range). If an invading host has a narrow set of PMA and does not encounter familiar mutualists in its exotic range, it might fail to form mutualistic partnerships, which could dramatically decrease its performance (S. Porter and E. Simms, in prep for resubmission) and may limit its invasion success . Thus, successful invaders are expected to be generalists in terms of the number and phylogenetic diversity of mutualists with which they can associate, yet few studies have tested this hypothesis (but see McGinn et al., 2016).
The PMA of an invading host can be contrasted with the composition of mutualistic symbionts with which it actually associates in a localized area-its realized mutualistic associates (RMA) (Ehinger et al., 2014). The RMA of an invader in a novel exotic range depends on a combination of its own PMA and the community of available mutualists. Mutualist community composition, in turn, depends on mutualist biogeography and the PMA of local hosts.
If an invading host has a large set of PMA and can adopt many mutualists that are locally abundant in its exotic range, then the invader may exhibit a set of RMA that closely resembles that of native hosts in the same region (Pringle et al., 2009;van der Putten et al., 2007;. Alternatively, the RMA of an invasive host could differ from that of native hosts in the same region; this could occur in two ways. First, the exotic range may have been coinvaded by an invading host's familiar mutualists from its home range (Dickie et al., 2010;Pringle et al., 2009;van der Putten et al., 2007;. Second, mutualists familiar to the invasive host might have cosmopolitan distributions and therefore be ready F I G U R E 1 Leguminous plants are pernicious invaders globally, threatening native diversity and disrupting ecosystem function and services. In California, (a) French broom (Genista monspessulana), (b) Spanish broom (Spartium junceum), and (c) gorse (Ulex europaeus) are invasive legumes that utilize a community of mutualists distinct from native legumes in the same range and waiting for the invader when it arrives in the new range (van der Putten et al., 2007). Previous studies provide some evidence for all of the aforementioned possible structures of invasive hosts' RMA (e.g., Weir, Turner, Silvester, Park, & Young, 2004;Leary, Hue, Singleton, & Borthakur, 2005;Lafay & Burdon, 2006;, Seifert et al. 2009, Nunez et al., 2009Dickie et al., 2010;Rodriguez-Echeverria, 2010, Porter, Stanton, & Rice, 2011Ndlovu, Richardson, Wilson, & Le Roux, 2013). The complexity of the observed patterns demands research that relates the PMA and RMA of invasive species.
Plant species in the family Fabaceae (legumes) comprise an excellent system with which to study how the PMA and RMA of a host can influence the trajectory of its invasion. Many legumes form mutualistic associations with rhizobial bacteria, which infect their roots and endo-symbiotically fix atmospheric di-nitrogen (Sprent, 2007).
Rhizobial symbionts are horizontally (infectiously) transmitted to their leguminous hosts. Legume seeds disperse independently of rhizobia, resulting in aposymbiotic (uninfected) legume seedlings; rhizobia are released into soil from senescing nodules and live independently in the soil until they encounter and infect a legume root (Sprent, 2007).
This horizontal mode of symbiont transmission in legumes leaves opens many possible pathways by which invading legumes could obtain rhizobia outside their home ranges. Although legumes are globally distributed (Yahara, Javadi, Onoda, & de Queiroz, 2013) and comprise some of the world's most noxious invasive species (Daehler, 1998;Yahara et al., 2013), the influence of rhizobial mutualists on legume invasion success is still debated (Richardson & Pyšek, 2000).
Here, we examine the RMA and PMA of three invasive legumes to address the role of mutualism in the invasion process. Specifically, we ask (1) Sachs, Kembel, Lau, & Simms, 2009;Ehinger et al., 2014). The combined collections comprise 715 isolates (see Table S1 for a list of all isolates, collection information, and Genbank Accession Numbers for representative isolates of each genotype identified). Because invasive hosts generally produce dense monocultures, collection sites for the nine legumes examined here were often nonoverlapping, however all collections occurred within a 350 km 2 region (Table 1; Fig. S1).
To obtain rhizobial isolates, legume individuals were carefully un- 1.5% agar) (Somasegaran & Hoben, 1994), incubated in the dark at room temperature, and twice restreaked onto new YMA plates from single-cell initiated colonies. A single-cell initiated colony was picked from each final restreak plate, inoculated into sterile YM broth, and incubated at 25°C and 120 rpm. Late-log-phase cultures were divided into two aliquots, one archived in 50:50 v:v culture:60% sterile glycerol at −80°C; the other pelletized and stored at −20°C for DNA extraction.

| Rhizobium identification and characterization
We characterized rhizobia isolated from wild-collected plants obtained for this study ( Isolates that had been field collected from the invasive hosts in their native range were included in this molecular analysis for comparison. We could find only two such isolates that had been sequenced at either ITS or nifD. One was associated with U. europaeus in its Trimmed ITS and nifD sequences, as well as concatenated ITS and nifD sequences, were aligned using the MAFFT v7 online alignment tool (Katoh & Standley, 2013). Distance matrices were generated using the Jukes-Cantor distance metric in the dnadist package of phylip v. 3.694 (Felsenstein, 2005). Genotypes were identified using the cluster function in Mothur v. 1.36.0 (Schloss et al., 2009).
Consensus sequences were generated using Mothur at 97% similarity for ITS sequences, 99% similarity for nifD sequences, and 98% similarity for concatenated sequences.
Separate phylogenetic trees for each locus and for the concatenated loci were generated using MrBayes v. 3.2.2 (Ronquist & Huelsenbeck, 2003), each with two parallel runs of 2,000,000 generations starting from random trees, three heated and one "cold" chain (heating temperature = 0.1), and a burnin fraction of 25%. Majority rule consensus trees were reconstructed from a sample of the post- T A B L E 1 Number of rhizobial isolates and genotypes identified from field collections of six native and three invasive legumes in the San Francisco Bay Area nature of the network structure observed using the neighbor-nets (see below) and are therefore presented only to illustrate relationships to known reference strains (Fig. S2).
A separate molecular network was generated for each individual locus and for the concatenated loci using the neighbor-net algorithm in SplitsTree v. 4.14.2 (Huson & Bryant, 2006

| Nodulation assay
We assessed the promiscuity of the three invasive plants in their invasive range by determining their ability to associate with a broad range of 117 rhizobial isolates originally collected from 12 different leguminous hosts (both native and invasive, including hosts not studied here, but all growing in the SF Bay Area;  (Somasegaran & Hoben, 1994) containing 7-ppm nitrogen was added to each tube.
The 117 rhizobial isolates used in the nodulation assay were obtained from two sources: (1) many isolates were obtained from the collection described above prior to genotyping (99 isolates); (2) several isolates were obtained from the investigators' additional research collections to represent strains associated with other native and invasive legumes common in the San Francisco Bay Area (18 isolates; see Table S2 for a list of the isolates used in the nodulation assay and their sources). Isolates were chosen to span a broad range of host species and collection sites. Inoculum from each isolate was prepared from 50 μl of −80°C glycerol stock prepared from field-collected nodules (see above), grown in YM broth at 25°C shaken at 120 rpm to a density of 1 × 10 6 per ml, as measured by optical density at 600 nm. Each rhizobial isolate was inoculated onto one seedling of each legume species. Seedlings were randomly assigned rhizobial isolates and inoculated 17 days after planting by adding 1 ml of the appropriate inoculum to the base of the plant stem in each tube. An additional ten plants per legume species were inoculated with sterile YM broth as negative controls; none of the control plants were nodulated at harvest. Plants were harvested 47 days after planting (30 days after inoculation), the roots thoroughly washed, and the presence of nodules recorded.
Successful association was defined as the formation of at least one robust nodule that appeared to be effectual (i.e., not <1 mm and/or white or clear). Reanalysis of our results increasing the cutoff for defining successful nodulation to two nodules did not qualitatively alter our findings.

| Realized mutualistic associates-field collections
All analyses were performed in R v. 3.2.2 (R Core Team, 2014). For each individual locus and the concatenated loci, rank abundance curves were generated for the relative abundances of genotypes associated with native versus invasive legumes under field conditions using the vegan package (Oksanen et al., 2013). Chao estimates for genotype richness (Gotelli & Colwell, 2001) associated with each legume host under field conditions (i.e., sample richness for each legume species) were determined using the vegan package (Oksanen et al., 2013). Phylogenetic diversity of genotypes associated with each legume host under field conditions was calculated as the mean pairwise molecular distance using the Jukes-Cantor metric between all pairs of genotypes associated with that legume species (note, mean pairwise molecular distance for the concatenated ITS and nifD loci of A. glaber and A. micranthus and the nifD locus of A. glaber were set to 0 for this analysis, as all rhizobia isolated from these species were identified as the same genotype; qualitatively similar results were obtained in separate analyses that excluded these species). Students' t tests were used to test for differences in genotype richness and phylogenetic diversity between native and invasive legume species, using legume species as replicates.

| Potential mutualistic associates-nodulation assay
For each test host legume species grown in the greenhouse nodulation assay, we categorized the test rhizobial isolates into "isolate origin" groups based on the relationship between the host species on which they were tested and the wild-grown host species from which they were originally isolated. The categories were as follows: (1) those originally isolated from the same species as the test host species (conspecific isolate) and (2) those originally isolated from a legume species other than the test host species (allospecific isolate). The allospecific isolates were further split into two subgroups based on the invasion status of the host species from which they were isolated: (1) those originally isolated from a native legume (native allospecific isolate) and (2) those originally isolated from an invasive legume (invasive allospe-  The other, conc 001, comprised nearly 70% of the isolates from native hosts but only ~8% of isolates from invasive hosts (Figure 2).

| Realized mutualistic associates-field collections
When considering the concatenated genotypes, the richness and phylogenetic diversity of the Bradyrhizobium communities associated with wild-collected legumes did not significantly differ between the native and invasive hosts (t 7 = 2.151, p = .068 and t 7 = −0.155, p = .881, respectively), but there was a trend for the invasive hosts to associate with a greater number of genotypes than the native hosts ( Figure 4). This trend was driven by two factors: (1) The high number of genotypes found on U. europaeus and (2) dominance by the common conc 001 genotype of the Bradyrhizobium community associated with the native legumes (Figure 3) S4). Only three of the 21 (14%) ITS genotypes (Fig. S3) and three of the 25 (12%) nifD genotypes (Fig. S4) were found on both invasive and native hosts. Nevertheless, one ITS genotype (ITS 001; a subset of which corresponds to conc 001) was found on all nine legume species and was the most common ITS genotype on both native and invasive legumes (Fig. S5). In contrast, the nifD genotype that dominated the Bradyrhizobium communities associated with native legumes (nifD 002) was not found associated with any of the invasive legumes in our study (Fig. S6).
Patterns identified when examining the ITS and nifD loci separately generally supported the observation that, in nature, the Bradyrhizobium communities associated with native and invasive legume hosts did not significantly differ in richness or phylogenetic diversity. The one exception was that categorizing rhizobial communities by nifD genotype revealed significantly greater phylogenetic diversity in invasive than native legumes ( Fig. S7; ITS: t 7 = 1.735, p = .126 and t 7 = −0.634, p = .546, for richness and phylogenetic diversity, respectively; nifD: t 7 = 1.971, p = .089 and t 7 = 2.615, p = .035, for richness and phylogenetic diversity, respectively).
Genotyping by each locus separately did produce different conclusions about community evenness, based on rank abundance curves of genotypes associated with either native or invasive legume hosts. For both native and invasive legumes, Bradyrhizobium communities were dominated by a few ITS genotypes (Fig. S5). In contrast, categorizing rhizobia by nifD genotype revealed different degrees of evenness between communities associated with native versus invasive legumes.
Bradyrhizobium communities of native legumes were dominated by a few common nifD genotypes, whereas nifD genotypes were relatively evenly represented within the communities associated with invasive legumes (Fig. S6).
Finally, ITS genotypes that had been isolated from Europeangrown U. europaeus (UU22sfb) and S. junceum (Sj4) were more than 97% similar to genotypes ITS 019 and ITS 003, respectively. In our field collection, ITS 019 was only found associated with U. europaeus F I G U R E 3 Among rhizobial communities of wild-grown hosts, genotypes of rhizobia associated with native legumes are less evenly distributed than those associated with invasive legumes. Rank abundance curves depicting relative abundances of genotypes associated with (a) native and (b) invasive legumes collected from the field. Genotypes were identified from concatenated ITS and nifD sequences. Asterisks indicate genotypes found associated with both native and invasive legumes whereas ITS 003 was found associated with both native and invasive legumes (Fig. S3). The nifD genotype isolated from European-grown U. europaeus (UU22sfb) was more than 99% similar to genotype nifD 006, which in our field collection was found associated with two of the invasive legumes (G. monspessulana and U. europaeus) but none of the native legumes (Fig. S4).

| DISCUSSION
Contrary to our expectations, the communities of rhizobia associated with wild-grown native and invasive legumes overlapped very little. Only a small percentage of Bradyrhizobium genotypes associated with both native and invasive legumes under field conditions, which suggests that the surveyed invaders are not currently forming novel associations with local mutualists in their exotic range. This result is surprising because, when tested under greenhouse conditions, the invasive legumes in our study could associate with many of the mutualists isolated from native legumes in nature. Our results parallel growing evidence at sites worldwide that invasive legumes utilize rhizobial communities that differ from those of native legumes (Chen et al., 2005;Lafay & Burdon, 2006;Rodriguez-Echeverria, 2010;Weir et al., 2004), although the opposite trend was observed in an Australian Mimosa invasion (Parker, Wurtz, & Paynter, 2006).
Rhizobial communities associated with wild-grown native versus invasive legumes in our study tended to differ in genotype dominance and evenness. Native legumes were dominated by one rhizobial genotype that is found throughout the state of California (Hollowell et al., 2016). In contrast, the invasive legumes in our study were less dependent on a few dominant rhizobial genotypes (i.e., had more even communities of rhizobial partners). The latter trend was primarily driven by one invader, U. europaeus, which associated with a particularly high number of rhizobial genotypes in the field. Interestingly, one of the F I G U R E 4 The diversity of rhizobia associating with wild-grown legumes does not differ between native and invasive hosts. (a) Chao richness and (b) phylogenetic diversity estimates for genotypes sharing 98% sequence identity across concatenated ITS and nifD sequences associated with native and invasive legume species growing in the field. Plant species codes are as defined in Figure 2 native legumes in our study, L. arboreous, has invaded other regions of the world. Although L. arboreous' RMA has not yet been evaluated in its invasive range, in our study, its community of RMA overlapped with that of the other native legumes. The broader and more even communities of RMA of the invasive species in our study could be a factor promoting their invasion success. Alternatively, a stronger or longer history of positive plant-soil feedbacks by the native legumes than the invasive legumes in this region may have favored a community of rhizobial mutualists associated with native legumes that is dominated by a few potentially highly beneficial rhizobial strains. Testing these hypotheses will require research into the mutualistic benefits provided by the different rhizobial genotypes when associating with native and invasive legumes.
The native and invasive legumes in our study primarily occurred at different field sites, although all were within a 350 km 2 region of the San Francisco Bay Area. It is, therefore, possible that the differences in rhizobial communities associated with native and invasive legumes observed in our study were due to geographic distance rather than host origin. We believe this to be unlikely for two reasons. First, at the one site where we collected sympatric individuals of one native (A. glaber) and two invasive (G. monspessulana, U. europaeus) legumes, the community of rhizobia associated with G. monspessulana overlapped very little with that of the native (five of 31 [16%] isolates shared based on concatenated genotypes), and the community of rhizobia associated with U. europaeus was completely distinct from that of the native (0 of 32 isolates shared based on concatenated genotypes). Second, the rhizobial communities among host species within the same collection site were generally as dissimilar as the rhizobial communities across collection sites; this was particularly true of the more even rhizobial communities associated with the invasive legumes. Further investigation into the host and geographic causes of these patterns, particularly in situations in which invasive hosts occur sympatrically with natives, is necessary to elucidate how mutualist acquisition influences biological invasion success.
Are the Bradyrhizobium strains associating with legumes invading the San Francisco Bay Area related to those that associate with conspecifics growing in their native European ranges? We could find remarkably little data with which to address this question, but the two isolates for which we were able to obtain ITS and/or nifD region sequence information suggest that the Bradyrhizobium genotypes associating with these invasive legumes in their home ranges are closely related to those they associate with in their exotic range. Two hypothesis could explain this result: (1) These genotypes had a pre-existing cosmopolitan distribution or (2) they have recently invaded the SF Bay Area from Europe, either coincident with or subsequent to the introduction of their legume hosts.
The cosmopolitan hypothesis derives from the notion that "everything is everywhere, but, the environment selects" (Baas-Becking 1934, as translated by deWit and Bouvler 2006). Certain rhizobial strains are indeed widely distributed (Stepkowski et al., 2007;Hollowell et al. 2016). For example, rhizobia associated with invasive Acacia and native legumes in the Mediterranean belong to cosmopolitan clades (Rodriguez-Echeverria, 2010). Similarly, in our study, the fifth most common Bradyrhizobium genotype associated with the invasive legumes (conc 001, which also dominated the community of Bradyrhizobia associated with native legumes) is widely distributed throughout California (Hollowell et al. 2015). Several studies have attributed successful legume invasions, particularly by woody shrubs, to such widely distributed rhizobia (Parker, 2001b;van der Putten et al., 2007;. However, recent studies dispute the idea that all microbes occur everywhere, acknowledging that many microbes are dispersal-limited, which could drive observed geographic patterns of microbial distributions (Litchman, 2010;Martiny et al., 2006). Indeed, there are many examples of symbiont limitation during agricultural legume introductions that necessitated the use of deliberate rhizobium inoculation (Coburn, 1907;Nunez et al., 2009;Pringle et al., 2009;Schwartz et al., 2006).
There are several mechanisms by which microbial mutualists could be introduced into an exotic range, but the primary modes by which rhizobia arrive are unclear. Rhizobia may arrive with their hosts.
For example, invasive plants are occasionally introduced with intact root systems, which would certainly harbor symbionts (Pringle et al., 2009). Additionally, seed companies frequently distribute rhizobium inoculum  and deliberate soil transport has often accompanied or closely followed agricultural legume introduction, which could disperse rhizobia into the surrounding environment.
Finally, although rhizobia are not transmitted maternally (Sprent, 2007), methods of seed harvesting in which soil contacts the seeds may deposit rhizobia on seed surfaces (M. Zafar, personal observation; Perez-Ramirez et al. 1998, Stepkowski et al., 2005. Future observational and experimental research is sorely needed to better understand rhizobium dispersal.
Unfortunately, the native microbiota associated with noncrop species is often poorly characterized (but see, e.g., Thrall et al. 2007, Hollowell et al. 2016, which hampers efforts to discover routes of rhizobium invasion. Indeed, we cannot definitively determine whether the rhizobia associated with G. monspessulana, S. junceum, and U. europaeus in their exotic range have a cosmopolitan distribution or co-invaded the San Francisco Bay Area, because we lack detailed information regarding the region's rhizobial community prior to invasion.
To distinguish co-invasion of previously endemic microbial mutualists from those with cosmopolitan distributions, areas that have not previously been invaded must be thoroughly sampled, including greenhouse experiments involving repeated planting of non-native hosts into soil from uninvaded areas to amplify potentially cosmopolitan but rare rhizobial genotypes.
Although the invasive legumes in our study are generally not currently associating with novel rhizobial mutualists in their exotic range, their potential to associate with a wide variety of rhizobia could have promoted successful establishment early in the invasion process.  (Lafay & Burdon, 2006;Parker, 2001a;Rodriguez-Echeverria et al., 2012), but these novel associations may provide less benefit than associations with familiar rhizobial strains (Rodriguez-Echeverria et al., 2012;Thrall, Burdon, & Woods, 2000). Selection pressure on the soil rhizobium community imposed by a successful legume invader might amplify the soil density and/or relative abundance of more beneficial rhizobia through a positive feedback process (Wolfe & Klironomos, 2005; but see Birnbaum and Leishman 2013). Future work is needed to determine the relative magnitudes of fitness benefits exchanged by different combinations of rhizobial genotypes and legume hosts species.
Through time, as highly beneficial symbionts are either introduced or naturally selected from diverse extant soil populations, invasive legumes may obtain greater mutualistic benefits by switching from novel mutualists to co-evolved symbionts. Positive feedbacks between invaders and these preferred mutualists may then propel invasions (Wolfe & Klironomos, 2005), akin to an invasional meltdown (Rodriguez-Echeverria, 2010;Simberloff & Von Holle, 1999). We therefore hypothesize that, in our system, the relatively high diversity and abundance of native legumes and the ability of the invaders to form associations with the rhizobial symbionts of these native legumes could have provided early generations of invading legumes with enough and sufficiently compatible native symbionts to survive prior to the population expansion of familiar, more beneficial, rhizobial symbionts. As these familiar rhizobial associates were encountered, either as rare individuals in the existing soil rhizobium population or through subsequent introduction, their numbers were amplified by positive plant-soil feedbacks. The end result of such a temporally staged invasion process would be the distinct rhizobial communities associated with native and invasive legumes observed in this study.
The use of distinct symbiont communities by native and invasive hosts has important conservation implications. For example, the mutualisms on which native hosts depend may be degraded if soil-borne mutualists compete with each other and invasive hosts promote population growth of their preferred mutualists. Whether such interactions occur, and their ecological importance, remains to be determined in many systems, including our own (Leary et al., 2005;Nuñez & Dickie, 2014;van der Putten et al., 2007;Rodriguez-Echeverria et al., 2011).
However, our work suggests that management informed by the existing distribution patterns of mutualist symbionts could aim to reduce the benefits invasive hosts derive from their preferred mutualists (Litchman, 2010). Future research on the mechanisms by which mutualists promote and/or inhibit species invasions could help prevent future biological invasions and inform efforts to restore invaded communities.

ACKNOWLEDGMENTS
We are grateful to Marin County Parks, the City of San Rafael, the S.S.P., and NSF-DEB 1457508 and NSF-DEB 0645791 to E.L.S.

CONFLICT OF INTEREST
None declared.

AUTHOR CONTRIBUTIONS
KJL, ELS, and SSP conceived the ideas; KJL, MT, MZ, and SSP carried out the research; KJL analyzed the data; KJL wrote the manuscript with editorial input from all co-authors.