Growth of an invasive legume is symbiont limited in newly occupied habitats
ABSTRACT
Mutualisms may play an important role in the establishment and invasion success of introduced species, but their influence is little studied. To test whether a lack of root nodule symbionts may limit the performance of invasive legumes, seedlings of Cytisus scoparius were introduced to an old-field habitat and then either inoculated with Bradyrhizobium strains from existing C. scoparius populations, or left uninoculated. In two separate years, inoculation more than doubled average plant biomass. For uninoculated transplants, nodule formation was positively correlated with proximity to plants of the native legume Desmodium canadense, but not related to distance from a second legume species, Apios americana. Polymerase chain reaction assays and DNA sequencing confirmed that bacteria isolated from uninoculated C. scoparius plants were indistinguishable from Bradyrhizobium strains in root nodules of D. canadense. By contrast, bacterial strains associated with A. americana were never found in C. scoparius nodules. Transplants in seven other habitats across a 160 km region also showed a highly significant, fivefold biomass increase in response to inoculation. Thus, colonizing legumes can suffer from a scarcity of nodule symbionts. However, certain indigenous legumes may create favourable microhabitats for invasion, by increasing symbiont availability in their vicinity.
INTRODUCTION
Plants in the family Leguminosae rank among the most widely disseminated taxa spread to new regions by human activities (Hughes & Styles, 1989; Lonsdale, 1994; Cronk & Fuller, 1995). Although symbiosis with nitrogen-fixing nodule bacteria (rhizobia) is a key feature of legume ecology, biologists currently lack a clear understanding of symbiotic interactions during range expansion. Legumes must obtain rhizobia from their immediate environment at the start of each generation, since the bacteria are not transmitted within seeds (Perez-Ramirez et al., 1998). Thus, if soil from ancestral habitats has not been transported to provide a source of co-adapted rhizobia, legumes entering a new habitat may suffer reduced performance owing to symbiont scarcity, unless rhizobia can be acquired from other legume taxa indigenous to the site (Parker, 2001).
Numerous cases have been reported in which legumes fail to form nodules when first introduced to a new geographical region (Wilson, 1934; Hely, 1957; Bottomley & Jenkins, 1983; Halliday & Somasegaran, 1983; Kueneman et al., 1984; Woomer et al., 1988; Lindstrom et al., 1990; Odee et al., 1995; Sullivan et al., 1995). Thus, it is clear that legume colonists can sometimes experience a scarcity of compatible rhizobia when they first arrive in a new habitat. If a site has too few bacteria for an optimal degree of nodule formation, then introduction of rhizobia should cause improved plant demographic performance. Addition of rhizobia has been shown to increase legume growth or survival in agricultural fields and pastures (Young & Mytton, 1983; Thies et al., 1991) and revegetation projects (Herrera et al., 1993; Jha et al., 1995). However, inoculation experiments in natural habitats are extremely few (Parker, 1995).
Variation in colonizing ability may be related to opportunities for symbiont acquisition from legume taxa that are indigenous to the invaded habitat (Richardson et al., 2000; Parker, 2001). However, due to the variously restricted host ranges of rhizobia (Young & Johnston, 1989; Perret et al., 2000; Parker et al., 2004), only a subset of native legume taxa are likely to be potential sources of symbionts for any particular invasive legume. Most invaders will therefore encounter high spatial variability in symbiont availability owing to the heterogeneous distribution of native legumes across the landscape. However, no inoculation studies have been carried out to experimentally measure the extent to which invasive legumes may be symbiont limited in different sites within their naturalized range. In order to understand which plants have been the primary source of symbiont strains utilized by invaders, molecular analyses are also needed to characterize the genetic relationships of the rhizobial populations on invasive legumes and on co-occurring native legume taxa.
In this study, we analysed nodule symbiosis in experimental populations of Cytisus scoparius (L.) Link (Scotch broom). Cytisus scoparius is native to Europe, but is an aggressive invader on four other continents as well as on several oceanic islands (Cronk & Fuller, 1995; Peterson & Prasad, 1998). Within the USA, naturalized populations of C. scoparius have been present for around 140 years (Parker, 1997; Peterson & Prasad, 1998), and have spread to at least 25 states (USDA, 2001). We experimentally introduced C. scoparius seedlings into sites in central New York and Pennsylvania that are > 100 km from existing naturalized populations in New York's Hudson River valley, in order to analyse the following issues. First, do seedlings with access only to rhizobia naturally present in the soil develop fewer root nodules than experimentally inoculated plants, and does inoculation improve C. scoparius transplant growth? Second, can the origin of nodule symbionts acquired by C. scoparius plants be identified by using molecular traits to compare them with bacterial strains from co-occurring native legumes? Finally, for uninoculated plants, is nodule formation correlated with proximity to particular taxa of native legumes?
METHODS
Study organisms
Cytisus scoparius (Papilionoideae, tribe Genisteae; sometimes referred to as Sarothamnus scoparius) is a shrub that can grow to a height of 3 m and flourishes best in sunny, open habitats with poor rocky or sandy soils (Bossard & Remjanek, 1994; Parker, 1997, 2002; Paynter et al., 1998). Nodule bacteria associated with C. scoparius fall into the genus Bradyrhizobium Jordan (Sajnaga & Malek, 2001; Sajnaga et al., 2001; Rodriguez-Echeverria et al., 2003). At present, there are only five described species of Bradyrhizobium that form symbioses with legumes (Euzeby, 2005). However, Bradyrhizobium is a large and diverse taxon (with many currently undescribed species) that includes the predominant symbionts of a very large array of legumes in habitats throughout the world (e.g. Lafay & Burdon, 1998; Parker, 1999, 2002; Doignon-Bourcier et al., 2000). Inoculation tests show that Bradyrhizobium isolates from native European populations of C. scoparius can form effective nitrogen-fixing symbioses with other legumes also placed in the tribe Genisteae (e.g. Lupinus L., Genista L.), as well as with certain legume genera from other papilionoid tribes (Sajnaga & Malek, 2001; Rodriguez-Echeverria et al., 2003). Certain North American isolates from Lupinus have also been shown to develop nodules on C. scoparius (Wilson, 1939). However, the number of distinct bacterial species represented among C. scoparius rhizobia, their host range, and their affinities to other Bradyrhizobium taxa all remain poorly characterized (Rodriguez-Echeverria et al., 2003).
Common garden experiments
Year 1
In late May 2003, 180 C. scoparius seedlings (4 weeks old and with shoots of 3–4 cm) were transplanted into a field at Binghamton University's Ecological Research Facility. The site was dominated by a mixture of grasses and native perennial herbs (Solidago L., Aster L., Asclepias L., Galium L.) and included a few scattered individuals of the native legumes Desmodium canadense (L.) DC and Apios americana Medicus. The field was mown before planting to reduce vegetation height to about 5 cm. C. scoparius seedlings were planted in a grid with 0.9 m spacing between adjacent plants. Cytisus scoparius seeds were initially induced to germinate by scarification in concentrated H2SO4 for 120 min (which killed any contaminant rhizobial cells adhering to seeds), and then planted in a Bradyrhizobium-free soil mixture (containing peat moss, vermiculite, and quartz sand) in bedding plant containers (3 × 3 × 5 cm soil volume per seedling). At planting, soil plugs containing the seedling root mass were inserted into 3 cm diameter holes at the garden site. The established vegetation of the field was not otherwise disturbed. Plants were watered several times during the first 2 weeks after transplanting, but subsequently received natural rainfall only. The C. scoparius seedlings were derived from three source populations: CK (Carkeek Park, Seattle, WA, USA), JP (Johnson Prairie, Thurston County, WA, USA), and TC (Kanawauke Traffic Circle, Orange County, NY, USA; n = 60 seedlings per population). Plants from the three populations were assigned at random to planting locations in the garden. For each seedling, its distance from the closest individual of each of the two native legumes in the habitat (D. canadense, A. americana) was recorded.
Half of the C. scoparius plants were inoculated with a diverse mixture of bacterial symbionts from the eastern USA on 5 June. An inoculation solution was prepared by mixing an equal volume of four liquid cultures, each containing a single Bradyrhizobium isolate sampled from C. scoparius populations in Cape Cod (MA), Wyoming (RI), Kanawauke Traffic Circle (NY), or Lake Tiorati (NY). Cultures were grown for 7 days in YM broth (Spoerke et al., 1996) and contained approximately 109 cells per millilitre. The four Bradyrhizobium isolates were chosen arbitrarily without any prior knowledge of their symbiotic properties, but sequencing of a 5′ portion of the 23S rRNA gene indicated that the four strains could be discriminated from one another (and from Bradyrhizobium strains indigenous to the garden habitat) by sequence differences in this region (see Results). Half of the seedlings from each source population were chosen at random and 5 mL of the inoculation solution was slowly dripped around the base of each plant's stem with a pipet. The remaining seedlings were exposed only to whatever rhizobial strains were naturally present in the soil of the garden habitat. However, to control for nutrient addition in the inoculation solution, a batch of the inoculation mixture was autoclaved for 2 h to kill bacterial cells, and the uninoculated seedlings each received 5 mL of this solution.
In late June, C. scoparius transplants began to suffer high mortality caused by vertebrate herbivory (presumably from meadow voles, Microtus pennsylvanicus Ord, whose runways were common throughout the garden). Remaining transplants were thus protected by enclosing them in cylindrical cages. The cages were made out of plastic mesh (1 × 1 cm mesh size) and effectively excluded the vertebrate herbivores, since no further damage was seen on caged plants. By the time cages could be made and installed on transplants, 103 of the original C. scoparius plants had been killed. These plants were replaced with new seedlings during the first 2 weeks of July and were inoculated on 14 July. Because the new plants experienced a growth period that was more than a month shorter than the original transplants, all statistical analyses were performed separately for the early vs. late transplant groups. During the first week of October, plants were harvested by digging up a cylindrical volume of soil (20 cm diameter × 15 cm deep). The soil was washed away from the root systems, and root nodules were counted. For 11 uninoculated plants that developed nodules, a subsample of one to four nodules per plant was saved to isolate bacterial occupants for later determination of symbiont identity. All plants were then dried at 100 °C for 48 h and the total dry mass was recorded. Four plants in the later transplant group (2 inoculated, 2 uninoculated) died for unknown reasons prior to harvest. These plants were excluded from the analysis of nodule numbers because the plants had disappeared and nodules could not be counted. The dead plants were included in the analysis of plant biomass (with a mass of zero), to provide an overall assessment of performance for all subjects in the two inoculation treatments. However, results of statistical analyses on plant biomass were the same if these four dead plants were excluded.
Year 2
The garden experiment was repeated in the same habitat in 2004 using a paired design. Pairs of C. scoparius seedlings were planted 0.6 m apart at 78 sites during the first week of June. Plants were protected from vertebrate herbivory using the same cylindrical mesh cages employed in 2003. The plant pairs were scattered over a 35 × 40 m area that encompassed the smaller 2003 garden plot, with the minimum distance separating adjacent pairs ranging from 1.5 m to 8.4 m. The distance from each pair to the nearest D. canadense individual was measured. One member of each pair selected at random was inoculated 1 week after planting. Inoculation procedures were identical to the 2003 experiment and used the same four Bradyrhizobium strains. Plants were harvested in early October, and processed as described above. For three of the 78 pairs, one or both plants died for unknown reasons and disappeared prior to harvest; these three plots were excluded from the analysis.
Statistical analyses
Results of nonparametric tests (Mann–Whitney, Wilcoxon paired sample, or Spearman rank correlation) are reported because these tests require fewer assumptions for validity than parametric statistical procedures. Qualitatively similar results were obtained in all cases when parametric procedures were used (t-tests or Pearson correlation).
Bacterial strain identification
A single-colony isolate was purified from each of 25 nodules on 11 uninoculated 2003 C. scoparius transplants by plating surface-sterilized nodules on YM agar (Spoerke et al., 1996). For comparison, bacterial isolates were also purified from the two native legumes on the plot (29 isolates from D. canadense and 16 from A. americana). These 70 bacterial isolates were initially characterized by two polymerase chain reaction (PCR) assays for sequence variation in the ribosomal RNA region. All currently known Bradyrhizobium strains have 16S rRNA sequences that cluster into two basic lineages, involving groups related to the species B. japonicum (Kirchner) Jordan or to B. elkanii Kuykendall (Lafay & Burdon, 1998; van Berkum & Fuhrmann, 2000; Parker et al., 2002; Qian et al., 2003; Parker, 2004). A multiplex PCR assay targeting the 3′ 16S rRNA region differentiates these lineages based on a 43-bp (base pair) difference in the size of the alternative amplification products (Parker, 2003). A second assay using the primers 23Sup115 and 23SrIII targeted the 5′ intervening sequence region of the 23S rRNA gene, where Bradyrhizobium strains from eastern North America that are related to B. japonicum show a 260-bp amplification product, compared to a 232-bp product for relatives of B. elkanii (Sterner & Parker, 1999; Parker, 1999). Six strains representing both variants detected in these assays (three from D. canadense, one from A. americana and two from C. scoparius transplants) were then sequenced for a larger portion of the 5′ 23S rRNA region (468–495 bp) using primers 23Sup6n and 23Sr#2 (Sterner & Parker, 1999). To characterize further the lineages detected in these analyses, the highly polymorphic internal transcribed spacer (ITS) region between 16S rRNA and 23S rRNA was also sequenced in four isolates (828–831 bp). Two distinct sequences were found in these four isolates which differed by 15 nucleotide substitutions and by two short insertion/deletion polymorphisms (GenBank AY826684–AY826687). Two pairs of PCR primers were then designed for differential amplification of DNA templates corresponding to these two ITS sequence types. Primers cs64f (5′-GTGCCGCAAGGTAATTCTGC) and cs64r (5′-ATCGAACCCCACACCAATGTCT) yield a 250-bp amplification product with D. canadense strain EDC2.8 and C. scoparius strain ERFr1-7CK2. Primers csits.f3 (5′-ATGTAGCTCACAAGGCTGCGT) and csits.r2 (5′-CAGAATGTTGTCTGTAAGAACTG) yield a 185-bp product with template DNA from D. canadense strain EDC3.6 and C. scoparius strain ERFr4-13TC1. Each of these pairs of strains failed to yield an amplification product with the alternate primer pair, indicating that the primers had adequate specificity to discriminate between these bacterial lineages.
Regional transplants
Cytisus scoparius seedlings were also transplanted in early June 2003 into unaltered vegetation at 13 sites spanning 160 km in New York and Pennsylvania. No known invasive population of C. scoparius was established within 130 km of any of the sites, which included locations in Susquehanna County, PA, USA and five New York counties (Broome [5 sites], Tioga [2], Tompkins [3], Delaware [1], and Oneida [1]). All transplant sites were in partially open habitats such as roadsides, power line corridors, river bluffs, or old fields, which are typical of locations occupied by invasive C. scoparius in the north-eastern USA. The transplants were 5 weeks old at the time of planting and were rooted in a Bradyrhizobium-free soil mix and planted as described above. Twenty-four seedlings were transplanted at each site, spaced 0.8 m apart. Transplants were watered at planting, and then 12 seedlings per site chosen at random were inoculated with 5 mL of the same inoculum mix (four Bradyrhizobium strains) used for the common garden plants. Surviving plants were dug up between 29 September and 6 October, and processed as described previously to determine nodule numbers and total dry biomass.
RESULTS
Common garden experiments
2003 experiment
The distribution of nodule numbers was substantially different for inoculated and uninoculated plants (Fig. 1). All inoculated plants had nodules, whereas 33–45% of the uninoculated plants completely lacked nodules and most of the remaining plants had relatively few. Mann–Whitney tests indicated that inoculated plants had significantly higher nodule numbers (P < 0.0001) both for the original set of C. scoparius plants introduced to the garden in May, and also for among plants established in July to replace those killed by herbivores.
Distribution of nodules developing on uninoculated C. scoparius plants (hatched bars) and plants inoculated with Bradyrhizobium (black bars) in the 2003 garden experiment. Top two panels represent the initial group of transplants, and individuals planted in July to replace those killed by herbivores are shown in the bottom two panels.
There was a strong pattern of increased growth in plants inoculated with Bradyrhizobium (Fig. 2). Although final dry mass was about five times greater for the original set of C. scoparius planted in May compared to those planted in early July, both sets of plants showed inoculation effects. For plants established in July, inoculation caused a highly significant increase in growth for all three source populations. For plants established at the earlier date, two of the three populations displayed significantly higher growth of inoculated plants. Among the uninoculated plants, there was a positive relationship between biomass and the number of nodules they developed (rs = 0.81, P < 0.0001, and rs = 0.48, P = 0.0002, for the early and later groups of transplants, respectively).
Mass of Cytisus scoparius transplants (mean and 1 SE) in the 2003 garden experiment when inoculated with Bradyrhizobium (black bars) or uninoculated (hatched bars). The initial group of transplants (a) and the individuals planted in July to replace those killed by herbivores (b) are plotted separately, and sample size is given below each bar. Results from Mann–Whitney tests to compare inoculation treatments are symbolized: ns, P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001.
For uninoculated plants, the availability of nodule symbionts appeared to have a patchy spatial distribution that was related to their proximity to one of the two native legumes in the habitat. Eight plants of the legume D. canadense were present in and around the plot, and for the initial set of transplants, there was a significant negative correlation between the number of nodules on uninoculated plants and their distance from the nearest D. canadense plant (Fig. 3; Spearman rs = −0.45, P < 0.01, n = 33). Among uninoculated plants introduced to the transplant site in July, the correlation between nodule numbers and distance to D. canadense plants was not significant although it also had a negative sign (rs = −0.19, P = 0.17, n = 55). Both the early and late groups had a similar distribution of distances to D. canadense plants (early group: mean = 4.1 m, range = 0.6–9.3 m; late group: mean = 4.5 m, range = 0.5–8.5 m; Mann–Whitney test, P > 0.70). Thus, the difference in magnitude of the correlation for the two groups probably reflects the smaller range of values for nodule numbers in the later group (Fig. 1, bottom left panel). For both the early and the late groups pooled, the correlation between nodule numbers on uninoculated transplants and distance to D. canadense was highly significant (rs = −0.30, P < 0.005).
Number of root nodules developing on uninoculated Cytisus scoparius transplants in relation to their distance from plants of the legume Desmodium canadense in the 2003 garden experiment.
A second native legume also occurred on the transplant site (A. americana). However, there was no significant relationship between nodule numbers on uninoculated transplants and their proximity to A. americana plants (early group, rs = −0.08, P = 0.67; later transplants, rs = −0.06, P = 0.68; pooled rs = −0.05, P = 0.64). These results are consistent with PCR assays and DNA sequencing results indicating that D. canadense was the primary source for the bacteria forming nodules on uninoculated transplants (see below).
2004 experiment
Outcomes in the second year confirmed results of the 2003 garden experiment. Plants experimentally inoculated with Bradyrhizobium developed about six times more nodules on average than C. scoparius transplants exposed only to the rhizobia indigenous to the habitat (mean = 68.1 vs. 11.6 nodules per plant, respectively, Mann–Whitney test, P < 0.0001). The biomass of each inoculated plant plotted against its paired uninoculated neighbour showed considerable scatter across the 75 plant pairs (Fig. 4), indicating significant environmental heterogeneity in the habitat. Nevertheless, the plant receiving Bradyrhizobium inoculation had higher biomass in 59 of the 75 pairs, with the ratio between inoculated biomass/uninoculated biomass averaging 2.1. A Wilcoxon paired sampled test indicated that inoculation significantly increased plant growth (P < 0.0001). Among uninoculated plants, biomass was again positively related to the number of nodules they developed (rs = 0.49, P < 0.0001).
Mass of each inoculated Cytisus scoparius transplant plotted against the uninoculated plant from the same plot in the 2004 garden experiment. The diagonal line marks equality of the two treatments.
Because plants were distributed across a broader area in the 2004 experiment, the range of distances between C. scoparius transplants and plants of the native legume D. canadense (0.7–35.4 m) varied over a larger scale than in 2003 (proximity to A. americana was not analysed in 2004 since the 2003 results indicated it was not a source of rhizobia for transplants). As in 2003, a significant negative correlation was observed between the number of nodules developing on uninoculated transplants and distance to the nearest D. canadense plant (Fig. 5; rs = −0.43, P < 0.0001).
Number of root nodules developing on uninoculated Cytisus scoparius transplants in relation to their distance from plants of the legume Desmodium canadense in the 2004 garden experiment.
Identification of nodule bacteria on transplants
All 16 isolates of nodule bacteria from the legume A. americana and four of 29 isolates from D. canadense plants at the garden site yielded 16S rRNA and 23S rRNA PCR length variants corresponding to strains related to Bradyrhizobium elkanii (Parker, 2003). A 468-bp portion of 23S rRNA was sequenced in two isolates, and blast searches confirmed that these isolates (GenBank AY826692, AY826693) were closely related to strains in the B. elkanii lineage (data not shown). This set of strains was not studied further, because none of nodule bacteria from C. scoparius transplants had these markers. The remaining 25 isolates from D. canadense and all 25 isolates from C. scoparius transplants had 16S rRNA and 23S rRNA markers indicative of strains related to B. japonicum. Sequencing of a 495-bp 5′ portion of the 23S rRNA gene in four of these isolates confirmed that they were relatives of B. japonicum. Two 23S rRNA sequence variants were found among these four strains that differed at two of the 495 nucleotides. One D. canadense strain and one C. scoparius strain shared an identical sequence (strains EDC2.8 and ERFr1-7CK2, AY826688, AY826689), as did a second pair of D. canadense and C. scoparius strains (EDC3.6 and ERFr4-13TC1; AY826690, AY826691).
To study this set of strains in further detail, the 16S rRNA-23S rRNA ITS region was also sequenced. Strains EDC2.8 and ERFr1-7CK2 were identical for the entire 828-bp ITS, even though this region is highly polymorphic among most strains of Bradyrhizobium (van Berkum & Fuhrmann, 2000; Parker, 2003; Parker et al., 2004). Strains EDC3.6 and ERFr4-13TC1 were identical for a different ITS sequence of 831 bp that had 15 nucleotide substitutions and two short insertions relative to the ITS sequence of strains EDC2.8 and ERFr1-7CK2. Two sets of PCR primers were designed to selectively amplify template DNA of each of these two alternative ITS sequence variants (primers ‘cs64f/cs64r’ for the EDC2.8/ERFr1-7CK2 variant, and primers ‘csits.f3/csits.r2’ for the EDC3.6/ERFr4-13TC1 variant). PCR assays with both primer pairs were carried out with all 50 isolates in the B. japonicum group from D. canadense and C. scoparius. Twenty-two of 25 D. canadense isolates and 24 of 25 isolates from C. scoparius transplants yielded a PCR amplification product only with the cs64f/cs64r primers. The remaining four isolates (three from D. canadense and one strain from C. scoparius [ERFr4-13TC1]) yielded a PCR product only with the csits.f3/csits.r2 primer pair. These results imply that most of the symbionts acquired by uninoculated C. scoparius transplants belong to a Bradyrhizobium lineage that predominates among nodules on plants of D. canadense in this habitat (exemplified by strain EDC2.8). Furthermore, a second less common Bradyrhizobium lineage associated with plants of D. canadense in this habitat is also responsible for a small fraction of the nodules developing on uninoculated C. scoparius transplants. All of the bacterial isolates in this less common group came from plants growing within 0.8 m of one another in one local area of the garden.
The four Bradyrhizobium strains used for inoculation of half of the C. scoparius plants in the common garden experiment could be differentiated from all 25 of the isolates recovered from the uninoculated plants by various combinations of PCR assay results (16S rRNA multiplex, 5′ 23S rRNA, cs64f/cs64r, csits.f3/csits.r2 and one other primer set; data not shown). Thus, the four Bradyrhizobium strains introduced at high density on inoculated plants throughout the plot were not a source for the nodules developing on plants in the uninoculated treatment.
Regional transplants
Overall transplant survivorship to the end of the growing season was roughly 20% and did not vary significantly for inoculated (37/156) and uninoculated plants (27/156, G = 1.972, P > 0.10). At six of the 13 sites there was complete mortality of C. scoparius transplants. Herbivory by both vertebrates and invertebrates appeared to be a major cause of transplant mortality, since substantial grazing damage was evident on many of the plants that survived. All 37 of the surviving inoculated transplants had root nodules (mean, 16.4 nodules per plant; range, 3–55), while 25 of the 27 C. scoparius plants in the uninoculated treatment completely lacked nodules on their root system (mean, 0.4 nodules per plant; range, 0–8). At six of the seven sites that had surviving transplants, the mean mass of inoculated plants exceeded that of uninoculated plants, and for all sites pooled, the average mass of inoculated plants was more than fivefold greater than among uninoculated plants (mean ± 1 SE = 0.132 g ± 0.038 vs. 0.024 g ± 0.007, respectively; Mann–Whitney test, P < 0.0001). Most locations had too few surviving plants in one or both treatments to provide sufficient statistical power for meaningful analysis at the level of individual sites. However, the two sites that had at least five surviving plants in each treatment (in Tioga County, NY, USA and Susquehanna County, PA, USA) both showed significantly higher mass for plants that were inoculated with Bradyrhizobium symbionts (Mann–Whitney tests, P < 0.05).
DISCUSSION
Our studies at two scales demonstrate that C. scoparius seedlings transplanted into new habitats grew substantially better if provided with Bradyrhizobium than when they had access only to rhizobia indigenous to the sites. Within one site, C. scoparius plants responded positively to inoculation in two separate years by producing more nodules and growing faster. The same responses were seen in seedlings planted across sites spanning 160 km. Although mutualisms were largely ignored as factors in invasion success until recently (Richardson et al., 2000; Parker, 2001), these results add to a growing body of evidence that positive interactions operating on a local scale may play a role in invasions.
To understand the importance of rhizobia to invasion dynamics, we need to know not only how seedling growth is influenced by access to rhizobia, but also how increased seedling growth translates into increases in C. scoparius population persistence or spread. Longer-term experiments are needed to address this question, but there are reasons to believe that factors influencing seedling establishment are likely to be important to invasion success. Population models showed seedling growth to be the most important driver of differences in population growth rate between habitats in Washington State (Parker, 2000). Rhizobia, along with other factors such as herbivores, which took a heavy toll on the transplants at most of our sites, are likely to influence colonization success because they have a strong effect on seedling growth.
This is the first study where native legumes that were the source of nodule symbionts for an invasive legume have been definitively identified by molecular characters. Uninoculated transplants developed far fewer nodules than inoculated plants overall. Yet indigenous rhizobia compatible with C. scoparius were not uniformly scarce. Within the site of the intensive transplant experiment, nodule formation on uninoculated plants was positively correlated with their proximity to plants of the legume D. canadense. Subsequent PCR assays and DNA sequencing indicated that bacterial isolates from these uninoculated C. scoparius plants were indistinguishable from Bradyrhizobium strains associated with nearby D. canadense plants. A second common legume indigenous to the transplant site (A. americana) had nodule symbionts that were never detected in the root nodules of C. scoparius transplants. These results emphasize that invasive legumes are likely to encounter a locally patchy distribution of compatible rhizobia, as a result of variation in the identity and abundance of native legumes that have inhabited different microsites in the recent past. This study is one of the first to illustrate how the influence of symbionts on community invasibility depends on the details of host specificity and the particular identity of resident legume taxa.
Populations of compatible rhizobia are expected to be as dynamic as their hosts. Uninoculated C. scoparius seedlings clearly suffered from a lack of nodule bacteria over the course of a single growing season. However, symbiont scarcity may be alleviated locally in a year or two by bacterial proliferation induced by the colonists themselves (Parker, 2001). Therefore, limitation by rhizobia may contribute to positive density-dependence (i.e. an Allee effect) early in the process of invasion. Allee effects are significant because they affect both spatial and temporal dynamics of invasion. For example, small nascent populations isolated from the main geographical range may experience a higher extinction risk. If such populations do survive, there may be an extended lag phase with little population growth followed by an accelerating rate of spread (Lewis & Kareiva, 1993; Cappuccino, 2004; Parker, 2004). To demonstrate such Allee effects, studies are needed that follow populations of both invasive legumes and their rhizobia from the earliest stages of invasion through time.
Although outside of the scope of this study, the quality of symbionts available to colonizing legumes may be just as important as their quantity. Invaders may be able to acquire symbiotic partners from other legume species present at a site, but there is no guarantee that such bacteria will be optimally adapted for efficient nitrogen fixation with the invading species. Evolutionary changes in efficiency of the symbiosis could also contribute to an Allee effect in legume invasions. It would be of interest to compare how legume invaders perform with bacterial strains acquired in newly occupied habitats vs. bacterial strains from long-established populations. Such studies are in progress and should help to better our understanding of symbiotic interactions affecting legume colonization success.
ACKNOWLEDGEMENTS
We thank Ami Lam and Kevin Horn for field assistance, and Joanne Pfeil for expert help with DNA sequencing. Financial support was provided by NSF grant DEB-0212369.
