The ability to form effective mutualisms with nitrogen-fixing bacteria (rhizobia) is implicated in the success of introduced leguminous plant species, such as Acacia. While Acacia appear to associate with rhizobia where introduced, there is evidence that the extent of this may limit success during early stages of colonization.
We examine three Australian Acacia species that have been introduced to New Zealand and ask whether variation in their ability to form rhizobial associations can explain differences in the degree to which they have established and spread since introduction.
In both Australia and New Zealand, we used glasshouse experiments to measure growth and nodulation of Acacia seedlings grown under two soil treatments: soils taken from underneath conspecifics (Host+ soils) and soils taken from the same sites but away from Acacia trees (Host−). We predicted that suitable rhizobia would be widespread in Australia leading to similar growth and nodulation in Host+ and Host− soils. However, we predicted lower growth and nodulation in New Zealand Host− soils, relative to New Zealand Host+ soils, due to limited availability of suitable rhizobia away from established conspecifics. We also predicted that differences between Host+ and Host− soils would be less marked in Acacia that were more widespread in New Zealand. Finally, we examined whether the establishment of one Acacia species might facilitate the establishment of other species by planting seedlings into soils associated with each of the two congeners.
As predicted, seedling growth and nodulation were lower in Host− than Host+ soils in New Zealand but there was no significant difference in Australia. In both countries, the difference between Host+ and Host− soils was similar for all three species and in conspecific and congeneric soils.
Synthesis. In New Zealand, Acacia seedlings that colonize sites away from established conspecifics or congeners are likely to suffer reduced growth and nodulation, which may limit their ability to establish and spread away from introduction sites. However, this limitation was the same for all three species, implying that interactions with soil biota cannot explain differences in the degree to which these Acacia have spread in New Zealand.
Some plant species introduced to new regions appear to leave behind soil pathogens that can regulate populations in the native range, which may give them an advantage over other species in establishment and spread following introduction (Klironomos 2002; Reinhart et al. 2003; Reinhart & Callaway 2006). For species that form mutualistic associations with soil biota, however, introduction to new regions may be a disadvantage unless suitable mutualists are also present in the introduced range soils (Burdon et al. 1999; Parker, Malek & Parker 2006). One example is leguminous species that form symbioses with nitrogen-fixing bacteria (rhizobia), which can facilitate establishment into poor soils. Although studies have found that many invasive legumes have successfully formed mutualistic associations with rhizobia in introduced ranges (Rodríguez-Echeverría et al. 2003, 2011; Parker, Wurtz & Paynter 2007; Callaway et al. 2011; Porter, Stanton & Rice 2011), there is evidence that rhizobia may be more limiting when species first colonize new sites (Parker, Malek & Parker 2006; Stanton-Geddes & Anderson 2011), which could limit their ability to spread in novel environments. If so, variation in the extent to which species experience such limitation may help explain variation in the extent to which introduced species spread away from introduction sites.
Although rhizobia are present in many soils, the ability of plant species to form viable symbioses (effective nodules) is dependent on both the identity and density of bacteria available. Highly promiscuous plant hosts are able to nodulate with a wide range of rhizobia strains and at low bacterial densities, while less promiscuous hosts show greater strain specificity and require higher bacterial densities before they nodulate (Roughley 1987; Bhuvaneswari, Lesniak & Bauer 1988; Thrall, Burdon & Woods 2000; Thrall et al. 2005, 2007). In addition, plant hosts themselves influence the availability of rhizobia in the soil (Thrall, Burdon & Woods 2000), with rhizobia population numbers increasing rapidly in response to compatible plant hosts (Purchase & Nutman 1957; Parker 2001) and declining when they are absent (Thrall et al. 2001).
The Australian Acacia are a diverse group of leguminous trees and shrubs that are widely cultivated outside their native range, primarily for forestry and horticulture. A relatively high proportion (around 6%) of species introduced to new regions has succeeded in establishing outside of cultivation and spreading into native ecosystems (Richardson et al. 2011). This may partly be due to their ability to establish into nutrient-poor soils, which is facilitated by interactions with rhizobia. Several studies have focused on this genus as a model system with which to study the role of rhizobia in determining invasion outcomes (Rodríguez-Echeverría et al. 2009, 2011; Birnbaum et al. 2012).
Rhizobia with which Acacia can nodulate appear to be widespread in their native Australian range (Barnet & Catt 1991), which may be because congeneric species often share rhizobia (Thrall, Burdon & Woods 2000; Thrall et al. 2007; Birnbaum et al. 2012) and Acacia are a dominant component of many habitats. Compatible rhizobia also appear to be present in many locations where Acacia have been introduced, with species recorded nodulating in Europe (Rodríguez-Echeverría et al. 2009), Asia (Midgley & Vivekanandan 1987; Le Roux et al. 2009; Ma et al. 2012), Africa (Mohamed et al. 2000; Joubert 2002; Rodríguez-Echeverría 2010; Boukhatem et al. 2012) and the Americas (Aronson, Ovalle & Avendano 1992), as well as outside their native range in Australia (Birnbaum et al. 2012). Although it is not clear what facilitates nodulation outside the native range, the widespread occurrence of cosmopolitan rhizobia (Weir et al. 2004; Birnbaum et al. 2012) and the co-introduction of compatible rhizobia from the native range have both been implicated (Rodríguez-Echeverría 2010; Birnbaum et al. 2012; Ndlovu et al. 2013).
Despite the apparent ubiquity of suitable rhizobia, there are many Acacia species that have failed to establish beyond their initial introduction sites. A potential reason for this is that although suitable rhizobia are geographically widespread, their availability in the soil is limiting. Low densities and a patchy distribution of compatible rhizobia have been suggested to limit the establishment of other introduced legumes (Parker, Malek & Parker 2006), and there are examples of Acacia failing to nodulate or performing poorly in some introduced soils (Turk, Keyser & Singleton 1993; Weir 2006). Previous studies that examined interactions between introduced Acacia and rhizobia have focused on Acacia species already known to be invasive and largely assess performance using soils taken from beneath established individuals, where we would expect the plants to have successfully encountered and cultivated suitable rhizobia (e.g. Mohamed et al. 2000; Rodríguez-Echeverría 2010; Birnbaum et al. 2012). This approach may have underestimated the potential for rhizobial availability to limit Acacia establishment in new locations.
Here, we examine the extent to which rhizobial availability in the introduced range of Acacia could influence species' ability to establish beyond introduction sites and thus determine the differential patterns of spread observed. We do this by comparing the growth and nodulation of Acacia seedlings grown in soil collected both beneath and away from adult conspecifics in both the native and introduced ranges of three Acacia species that differ in the degree to which they have established and spread since their introduction to New Zealand. Moreover, since many Acacia species share compatible rhizobia, we also test whether the presence of an established congener can facilitate the growth and nodulation of other Acacia species. By comparing plant performance in soils from both beneath and away from established Acacia in both native and introduced ranges, we undertake the first direct test of the hypothesis that a lack of the soil biota associated with established individuals could limit seedling performance at new sites. By examining species that vary in the extent to which they have established and spread in New Zealand, this is also the first study to directly test whether plant–rhizobia interactions might explain those differences and thus influence invasion outcomes.
New Zealand has no native Acacia species but at least 150 species of Australian Acacia have been introduced (Diez et al. 2009). Although the majority have only naturalized close to introduction sites, eight species are more widespread and are considered environmental weeds (Howell 2008). Of these environmental weeds, all except one are considered invasive in other parts of their global introduced range (Richardson & Rejmánek 2011). There is little information on interactions between Acacia and rhizobia in New Zealand. Although nodules were collected from an established population of Acacia longifolia, a species that is invasive in some parts of New Zealand, this species failed to nodulate in soils collected away from established plants (Weir 2006). Therefore, while cosmopolitan rhizobia capable of nodulating Acacia species appear to be present in New Zealand (Weir et al. 2004; Weir 2006), they may occur at low densities and their availability could be limiting for species in the early stages of colonization.
We tested the following hypotheses:
The growth and nodulation of Acacia seedlings should be lower in soil collected away from adult conspecifics relative to beneath adult conspecifics in the introduced (New Zealand) range due to low rhizobial availability at new sites. The widespread presence of compatible rhizobia in the native (Australian) range means that seedling growth and nodulation should be similar regardless of where soils are collected.
Acacia species that are more widespread in New Zealand should be less limited by the potentially low rhizobial availability and show less of a reduction in seedling performance when grown in soils collected away from adult conspecifics relative to beneath adult conspecifics.
As Acacia can often share rhizobia, rhizobial populations associated with naturalized congeners should improve plant performance, relative to soils taken away from any species of Acacia, and could therefore facilitate the establishment of Acacia arriving in new locations.
Materials and methods
The three Acacia species we selected have all established and exhibit seedling recruitment at introduction sites in New Zealand, indicating that compatible rhizobia are present to some extent. However, the species differ in the degree to which they have spread to establish beyond introduction sites (Table 1). The species are all native to south-eastern Australia, a region with a close climate match with New Zealand (Kriticos 2012). Acacia dealbata Link is widespread and common throughout south-eastern Australia (Maslin 2001) and is highly invasive in New Zealand, where it forms extensive monocultures along agricultural margins and in riverbeds. It is also considered invasive in other parts of the world, particularly southern Africa, the Americas and Mediterranean Europe (Richardson & Rejmánek 2011). This species has been recorded nodulating throughout its introduced range, including South Africa (Joubert 2002), Chile (Aronson, Ovalle & Avendano 1992), Sri Lanka (Midgley & Vivekanandan 1987) and China (Ma et al. 2012). The native range of A. dealbata overlaps that of the two other species we selected. Acacia baileyana F. Muell. is native to a small area around Cootamundra in New South Wales where it occurs in the forest understorey or in forest gaps (Maslin 2001), although it has been widely planted and is now naturalized beyond this range in Australia. In New Zealand, A. baileyana is a popular cultivated tree and, although widely naturalized, rarely establishes far from source populations. Acacia baileyana is considered invasive in Africa (Richardson & Rejmánek 2011). Although it may nodulate with a variety of strains within Australia (Roughley 1987), its association with rhizobia does not appear to have been studied outside its native range. Acacia pravissima F. Muell. ex Benth is native to higher elevation zones of the southern Australian Great Dividing Range where it is found in open eucalypt forest and moist areas (Maslin 2001). In New Zealand, it can reproduce successfully close to parent trees but has not established beyond garden plantings and is currently classed as a casual (Howell & Sawyer 2006). There are no records of it being invasive elsewhere in the world (Richardson & Rejmánek 2011) and little information on the degree to which it nodulates in either its native or introduced range.
Table 1. Invasion status and introduction date for the three species of Acacia included in this study: Acacia baileyana, Acacia dealbata and Acacia pravissima
To quantify the extent to which Acacia seedling performance is limited by rhizobial availability in New Zealand, relative to Australia, we conducted glasshouse experiments in both countries. We grew seedlings in soil collected from the field, measured seedling growth and nodulation and explored the relationship between these variables.
For each Acacia species in each country, we had two soil treatments. First, we used soils collected from beneath established conspecifics (Host+). Because populations of compatible rhizobia increase in the presence of their host plants, seedlings grown in Host+ soils should not be limited by the availability of rhizobia. Secondly, we collected soils from the same sites but 20 m away from conspecifics and any other Acacia trees (Host−). We considered this distance sufficient to escape any effects of Acacia trees and their root systems on soil communities (see also Callaway et al. 2011), but close enough to ensure that other soil properties were similar. The Host− soils therefore represent the soil conditions encountered by seedlings that spread away from parent populations. We used the difference in seedling growth between the Host+ and Host− soils as a relative measure of the extent of any rhizobia limitation seedlings would experience when colonizing new sites. To test whether the presence of established congeners could facilitate seedling growth, we also carried out a cross-inoculation experiment and planted seedlings of each species into the Host+ and Host− soils from populations of the other two study species in both Australia and New Zealand.
Study sites and soil collection
Because nodulation with rhizobia can be influenced by soil properties and environmental conditions (Vincent 1965; Habish & Khairi 1970), we collected soils from multiple populations (sites) in Australia and New Zealand to ensure we sampled a range of soil conditions. We identified four sites per species in Australia and five sites per species in New Zealand (Appendix S1 in Supporting Information). In Australia, populations were located within the species' known geographic range (Maslin 2001) but for logistical reasons, we limited ourselves to searching for these within a three-hour drive of Canberra (35°16′ S 149°7′ E). We located four populations each of A. baileyana and A. pravissima that contained five or more adult individuals. Acacia dealbata was more widespread around Canberra, and for logistical ease, we chose four populations located close to the A. baileyana and A. pravissima populations (see Appendix S1). None of the three species co-occurred at our Australian study sites. In New Zealand, study populations were located within a two-hour drive of Christchurch (43°31′ S 172°38′ E). We included all A. baileyana and A. pravissima populations with more than one individual we could find, five of which contained fewer than five individuals, and included the five largest A. dealbata populations (Appendix S1). In New Zealand, A. baileyana and A. pravissima co-occurred at two sites where they had been planted as ornamentals (Appendix S1). Here, we ensured that soils for 1 species were sampled at least 20 m away from any other species of Acacia.
For the Host+ treatment, we collected soil from beside the base of up to five haphazardly selected trees of each species at each site. Soils were collected to a depth of about 10 cm, excluding the litter layer and bulked to form one Host+ soil sample per site. For the Host− treatment, we took between 3 and 5 samples per site to a depth of about 10 cm and bulked these to form one Host− soil sample per site. We did not avoid other legume species when collecting Host− soils because a key aim of this study was to examine the growth of Acacia seedlings in soils they would typically encounter when establishing away from parent plants. In Australia, we collected a total of 24 soil samples: 2 host treatment soils (Host+ and Host−) × 3 host species (A. baileyana, A. dealbata, A. pravissima) × 4 sites per species. In New Zealand, we collected a total of 30 soil samples (5 sites per species).
After collection, soils were stored in large paper bags for transfer to the glasshouse where they were air-dried for up to 48 hours. Soils were then sieved to remove any stones and other dry matter and stored in paper bags at room temperature until use (< 2 weeks).
We conducted glasshouse experiments separately in Australia and New Zealand. Conducting the experiments separately means that the results cannot be directly compared between countries due to potential differences in glasshouse conditions. For this reason, when testing for between-country differences, we use the difference between Host+ and Host− soils as a relative measure of plant performance, rather than absolute variation.
We obtained seed from the Australian Seed Company (http://www.ausseed.com.au) and used seeds from the same seed lot in both the native and the introduced range to control for any differences that may arise from fitness variation in seeds of different provenances. Seeds were germinated by placing them in boiling water for 1 min and then removing them from the heat and leaving them in the water to imbibe overnight, following the Australian Seed Centre Manual (Gunn 2001). They were then transferred to germination trays containing a 1 : 1 mixture of sterile vermiculite/sand and watered as required until germination.
When seedlings reached the first leaf stage, we transplanted them into each of the treatment soils. To do this, we filled pots (150 mm height, 80 mm diameter) to ¾ with 1 : 1 sterilized vermiculite/sand mixture. We then covered each pot with 100 mL of one of the field soil treatments to serve as an inoculant for the seedlings and covered this with a further 1-cm layer of sterile soil. This protocol follows Thrall et al. (2007), and the relatively small amount of field soil used is intended to minimize the effects of any differences in soil chemistry and nutrient status on plant growth. The soils of each pot were covered with polyurethane beads (c. 2 mm in diameter) to prevent cross-contamination during watering. We also included a sterile control treatment for each species in each country, where seedlings were planted into pots filled only with the sterile vermiculite/sand mixture and covered with polyurethane balls.
Pots were arrayed in the glasshouse in a randomized block design. Each block consisted of one seedling of each species planted into each soil treatment, as well as two sterile controls per species. In Australia, this resulted in 78 seedlings per block: 24 field soil samples (2 host treatment soils × 3 species × 4 sites as above) × 3 seedling species (A. baileyana, A. dealbata, A. pravissima), plus 6 sterile controls. This was replicated 6 times giving a total of 468 seedlings in Australia. In New Zealand, there were 96 seedlings per block: 30 field soil samples × 3 seedling species, plus 6 sterile controls. Each of these was replicated seven times giving a total of 672 seedlings in New Zealand.
Seedlings were grown under an 18–24°C temperature regime with ambient light conditions. In Australia, there were several days when temperatures exceeded this due to a combination of hot weather and problems with the air conditioning system. Seedlings were grown for 14–16 weeks in Australia and 16–18 weeks in New Zealand. The difference in growth period was due to time constraints, and seedlings were large enough by 14 weeks for treatment differences to be observed. Seedling mortality was low. However, if seedlings died within the first few days, they were replaced. Any seedlings that died after the first few days were not replaced and were omitted from the analysis. Seedlings were watered with N-free 1 : 20 diluted McKnight's solution (McKnight 1949) three times a week and tap water if needed otherwise. Pots were weeded regularly to ensure that seedling growth was not affected by competition with other plants.
We used growth rate as our measure of plant performance across treatments. To calculate this, we harvested the above-ground parts of the plants and oven-dried them at 70 °C for 48 h before weighing. The growth rate of each plant was then calculated as above-ground plant dry weight/the number of days since planting into treatment soils (g day−1). To assess the importance of interactions with rhizobia for plant growth and identify whether patterns of nodulation were linked to plant performance, we used the total number of effective nodules as a measure of symbiotic success (Thrall et al. 2007). We considered any nodule to be effective if it was pink to red in colour (Corbin, Brockwell & Gault 1977; Thrall et al. 2007), indicating the occurrence of nitrogen fixation. To count nodules, we separated roots from above-ground plant parts and scored roots for the numbers of effective nodules using the following categories: < 5, 5–10, 10–25, 25–50 and 50–100. For analysis, the numbers of nodules recorded for each seedling were assigned the midpoint of each category (e.g. 2.5, 7.5).
We analysed our data using mixed models fitted in a Bayesian framework. This allowed us to include ‘site’ as a random effect, which we did as a further precaution to control for any differences due to site-specific variation in soil type that might affect seedling growth and nodulation. Site was included as a random effect by specifying a different mean for each site with those means modelled as drawn from a normal distribution with mean zero and variance estimated from the data. We assigned the overall intercept and regression coefficients normal prior distributions with mean 0 and variance 1000, and for the ‘site’ term, we specified a non-informative uniform prior (0–100) on the standard deviation following Gelman (2006).
We analysed our data in two stages. First, we included either growth rate or the number of effective nodules as the response variable and had four categorical treatment variables as main effects: host soil treatment (Host+/−), Acacia host species for the soil, Acacia seedling species and country. Each treatment was included as a two- or three-level factor variable, and we included all main effects and interaction terms. Each term was included in the model by coding them as dummy variables and choosing a reference class with coefficient set to zero. We did not include the sterile controls in our analysis as our central questions relate specifically to relative differences between the Host+ and Host− soils in each country and between species, rather than absolute growth and nodulation values. However, for visual comparison of performance in the experimental soils relative to performance in sterile soils, we present the overall mean values for growth and nodulation of each species in the results section.
Secondly, in order to directly examine the influence of nodulation on plant growth in each country and for each species, we again set growth as the response variable but this time included nodulation as a continuous variable in a model that included all other treatment variables and their interactions. To visualize the relationship between growth and nodulation for each combination of the treatments, we used this model to calculate the incremental increase in growth that resulted from the addition of one nodule under each of the treatment combinations, that is, the slope of the line describing the relationship between seedling growth rate and the number of nodules produced.
Models were fitted using Markov chain Monte Carlo methods as implemented in OpenBugs (Thomas et al. 2006) called from the BRugs library in R v. 2.13.1 (R Development Core Team 2011). We ran three chains each with a burn-in of 10 000 iterations. The posterior distributions were then sampled from a further 10 000 iterations of each chain, which were checked for convergence.
For each analysis, we tested the overall effect of each treatment on growth or nodulation by calculating the difference between the two classes in each treatment having the most extreme coefficient values. We calculated this difference for each of the 30 000 iterations (10 000 from each chain) and then calculated the median and 95% credible intervals of these differences. We considered that imposition of a treatment had a significant effect on growth or nodulation if the 95% credible intervals of the differences did not overlap zero, implying a significant difference in growth or nodulation between at least two classes in that treatment.
Seedling growth rate was consistently higher in field-inoculated soils than in the sterile controls in both Australia and New Zealand (Fig. 1). There was a significant interaction between Host+/− soil treatment and country (Fig. 2a), with seedlings growing more than twice as fast in Host+ soils relative to Host− soils in New Zealand (0.013 g day−1 compared with 0.005 g day−1, respectively), while in Australia, growth rate was not strongly influenced by Host+/− soil treatment. There were no other significant interactions indicating that irrespective of country, the relative difference in growth between Host+ and Host− soils was similar across species and was not influenced by whether they grew in their own or congeneric Host+ soils.
Seedlings formed effective nodules in all field-inoculated soils (Fig. 3), with the exception of Host− soils from two A. dealbata sites in New Zealand, where some seedlings failed to form nodules or formed only non-effective nodules. Around 40% of sterile controls formed nodules, usually forming only 1 or 2 with the exception of one A. dealbata and one A. pravissima seedling in Australia that each formed 10–25 functional nodules, highlighted by the slightly higher mean values for the sterile controls of these species in Australia (Fig. 3).
As with plant growth rate, there was a significant influence of Host+ soils on levels of nodulation and this influence varied between Australia and New Zealand (Host+/− × country interaction in Fig. 2b). In New Zealand, seedlings grown in Host+ soils formed over three times as many nodules as those grown in Host− soils (on average, 29 compared with 9, respectively). In Australia, there was a tendency towards lower nodulation in the Host− soils compared with Host+, but this effect was non-significant (an average of 18 nodules were formed in the Host+ soils compared with 11 in the Host−).
In contrast to plant growth response, differences in nodulation between Host+ and Host− soils varied depending on the seedling species and the host species, that is, there was a significant Host+/− × seedling species × host species interaction (Fig. 2b). This was due to A. dealbata seedlings varying in their response to Host+ versus Host− soils, depending on the host species (Fig. 3). Seedlings showed the greatest increase in the number of nodules produced in the Host+ soils taken from underneath A. pravissima when compared to the Host− soils taken from the same A. pravissima sites (on average, 65 nodules compared with 25, respectively). Seedlings showed the least increase in the Host+ soils of conspecifics when compared to the Host− soils from the same conspecific sites (on average, 39 vs. 20, respectively). There were no further influences of seedling species, host species or country on nodulation.
The importance of nodulation for plant growth was confirmed by the significant effect of the number of effective nodules on growth rate when this was included as a variable in the model (Fig. 4). However, this model highlighted that factors other than nodulation may also contribute to variation in growth, because the interaction between Host+/− and country remained significant (Fig. 4a). In other words, lower growth in Host− soils relative to Host+ soils in New Zealand could not be entirely explained by differences in the number of effective nodules. In addition, there was some indication that nodule effectiveness for growth varied depending on the seedling species (Host+/− × seedling species × nodulation interaction in Fig. 4b) and the identity of the host species (Host+/− × host species × nodulation interaction). The interaction with seedling species seemed due to A. baileyana showing a more marked growth response to increasing nodulation in all soils compared with the other two species, as well as a more marked increase in the influence of a nodule on growth between the Host+ and Host− soils (Fig. 5). Specifically, each nodule formed by A. baileyana seedlings in the Host− soils resulted in a 0.0002 g day−1 greater increase in growth rate than those formed in the Host+ soils compared with an increase of 0.0001 g day−1 shown by A. dealbata and 0.00008 g day−1 shown by A. pravissima. The interaction with host species seemed due to all seedlings showing a limited growth response to increasing nodulation in the Host+ soils taken from A. pravissima populations. There were no further interactions between nodulation and the other variables examined.
In this study, we tested three hypotheses to determine the extent to which interactions with soil biota, specifically rhizobia, could influence species growth rates when introduced to new locations. Our first hypothesis was supported: seedlings grew more slowly in soil collected away from established conspecifics in their introduced New Zealand range, but not in their native Australia. Our second hypothesis was not supported: all species showed a similar reduction in growth when grown in soils collected away from established conspecifics in New Zealand. Finally, our third hypothesis was supported: for all species, growth in soils associated with a congener was greater than in soils collected away from any Acacia.
Rhizobial availability limits plant performance in the introduced relative to the native range
In Australia, the growth of Acacia seedlings does not appear to be constrained by lack of rhizobia at sites away from established individuals, most likely due to the widespread availability of rhizobia in the region (Barnet & Catt 1991). Although seedling growth was reduced away from conspecifics, which we might expect if soil rhizobia populations decline in the absence of a host plant (Parker 2001; Thrall et al. 2005), the reduction in growth was not strong or statistically significant. Other studies that have examined the influence of soil biota on species performance have found that interactions with natural enemies in the soil are generally more pronounced in species' native ranges (e.g. Klironomos 2002; Reinhart et al. 2003; Reinhart & Callaway 2006), particularly in association with conspecifics (MacKay & Kotanen 2008). However, we found that positive interactions dominated seedling performance, since seedling growth in the treatment soils was considerably higher than in the sterile control soils. In addition, the positive relationship between seedling growth rate and degree of root nodulation implied a direct relationship between rates of rhizobial infection from the soil and seedling growth rate. Therefore, although we were unable to separate the positive and negative effects of soil biota on seedling growth, due to using whole soils that contained the suite of organisms present at any location, it is unlikely that negative interactions with soil pathogens in the native range masked a stronger positive influence of rhizobia present in the Host+ soils in Australia than was suggested by our data.
In New Zealand, most seedlings showed an increase in performance in experimental soils, relative to the sterile controls, and nodulated to some degree in most soils. Nevertheless, in contrast to Australia, seedling growth rate in New Zealand did appear to be limited by low availability of rhizobia in soils away from established individuals. Both seedling growth and nodulation were reduced in the absence of an Acacia host, suggesting that compatible rhizobia were present at much lower densities. Therefore, although our findings support previous work that suggests rhizobia with which Acacia can nodulate are widespread (Rodríguez-Echeverría et al. 2009; Birnbaum et al. 2012; Boukhatem et al. 2012), they also highlight that growth rate will be limited by rhizobial availability away from established populations. Rhizobial limitation of plant performance has been demonstrated in other introduced legumes, including Cytisus scoparius and Chamaecrista fasciculata in the United States (Parker, Malek & Parker 2006; Stanton-Geddes & Anderson 2011), but not all. Specifically, Robinia pseudoacacia performed equally well in its invaded European as its native American range (Callaway et al. 2011). These conflicting findings could reflect differences in the rhizobial densities at which species can nodulate or in species' ability to share the same rhizobia as native or other introduced legumes (Parker, Malek & Parker 2006). For example, the limitation observed in our study could be because in New Zealand, introduced plant species nodulate with different rhizobial strains to native plants (Weir et al. 2004; Weir 2006), but where introduced plant species can share rhizobia with native plants, limitation may be less pronounced.
The role of rhizobia in plant invasion
There was no species-specific variation in seedling growth rate, whether in Host+ or Host− soils, in either the native or introduced range. While there was some interspecific variation in the influence of nodulation on growth, this was limited. The lack of significant species × country interactions indicates that any differences in glasshouse conditions (e.g. temperature and photoperiod) are likely to have had little influence on the outcome of the study. Therefore, since the absence of a strong species effect in growth and nodulation contrasts with the different degrees to which the three species have established and spread in New Zealand, our findings suggest that interactions with rhizobia are not the prime drivers of differences in Acacia invasion success.
Residence time and introduction effort are both widely recognized to be important determinants of the distribution range of alien species (Williamson et al. 2009; Aikio, Duncan & Hulme 2010). It is therefore conceivable that the varying invasiveness of the three Acacia species simply reflects the different dates of introduction, so that all three species are equally invasive and rhizobia are similarly important for the invasion of these taxa. However, the species also differ in their local abundance. Despite a residence time of around 100 years and frequent planting as a landscape tree, A. baileyana does not develop extensive populations in New Zealand. Thus, for this species at least, the observed differences in its current distribution, relative to the more widespread A. dealbata, appear to be neither a function of introduction date or rhizobial limitation.
Although interactions with rhizobia cannot explain the different patterns of establishment and spread in these three Acacia species, the increase in seedling growth associated with an increase in nodulation highlights the importance of rhizobia for plant performance in the introduced as well as the native range. Moreover, our finding that all three species of Acacia performed equally well in soils collected from beneath congeners indicates that the presence of native or naturalized congeners may facilitate the establishment of other rhizobia-dependent species, which has been proposed in other systems (Richardson et al. 2000; Dickie, Koide & Steiner 2002; Hill & Kotanen 2012). Nevertheless, the limitation we observed at a distance of only 20 m away from focal plants suggests that this facilitation could promote only localized spread.
There was considerable unexplained variation in seedling growth rate in our data even when nodulation was included as a covariate in the analysis. This will be due partly to natural variation in seedling fitness and variation in soil properties beneath parent trees (Klironomos 2002; Rodríguez-Echeverría 2010), including potential interactions with soil pathogens and mycorrhizal fungi. However, the fact that all three species showed a comparable reduction in growth away from established individuals in New Zealand indicates that regardless of the mechanism underlying the patterns we found, interactions with soil biota more generally do not underpin differences in the degree the three species have established and spread since introduction to New Zealand. Furthermore, that this pattern was comparable for each species and across the soils of each congener highlights the generality of this finding, since together these soils represent a range of experimental sites, populations of varying sizes and soil both pre- and post-colonization by Acacia.
By comparing species' abilities to form associations when first introduced to an area in both their native and introduced range, we have shown that the availability of rhizobia has the potential to limit seedling growth of introduced Acacia. In addition, by examining species that differ in the degree to which they have established and spread since introduction to New Zealand, we have demonstrated that interactions with rhizobia and soil biota in general cannot explain these differences and are therefore unlikely to underpin the variable invasion success shown by Acacia in New Zealand. Finally, although the presence of established congeners may facilitate seedling establishment and growth, this influence was highly localized. Thus, the role of rhizobia and soil biota in general in determining invasion outcomes may not be as important as previously thought.
This research was funded by a New Zealand Tertiary Education Commission PhD scholarship, the Bio-Protection Research Centre and the Commonwealth Scientific and Industrial Research Organisation (CSIRO). We thank Amar Singh, Annette Girardin, Carolin Weser, Inkeri Lehtinen, Nicola Day and Ruth Aveyard for field and glasshouse assistance and Pete Thrall and his team at CSIRO for valuable advice in setting up our experiments. For site access and permissions, we thank Christchurch City Council, Environment Canterbury, Glen Colwyn Estates, New South Wales National Parks and Wildlife Services, Orton Bradley Park, Jock Bulman, Peter Chamberlain, Lynda and Will Collins, Bruce Deans, Matt Latham, Kevin and Gill Morton, Geoff and Jill Miners and Mack Wilson.