Legume-nodulating rhizobia are widespread in soils and plants across the island of O‘ahu, Hawai‘i

Legumes and their interaction with rhizobia represent one of the most well-characterized symbioses that are widespread across both natural and agricultural environments. However, larger distribution patterns and host associations on isolated Pacific islands with many native and introduced hosts have not been well-documented. Here, we used molecular and culturing techniques to characterize rhizobia from soils and 24 native and introduced legume species on the island of O’ahu, Hawai‘i. We chose two of these isolates to inoculate an endemic legume tree, Erythina sandwicensis to measure nodulation potentials and host benefits. We found that all rhizobia genera can be found in the soil, where only Cupriavidus was found at all sites, although at lower abundance relative to other more common genera such as Rhizobium (and close relatives), Bradyzhizobium, and Devosia. Bradyrhizobium was the most common nodulator of legumes, where the strain Bradyrhizobium sp. strain JA1 is a generalist capable of forming nodules on nine different host species, including two native species. In greenhouse nursery inoculations, the two different Bradyrhizobium strains successfully nodulate the endemic E. sandwicensis; both strains equally and significantly increased seedling biomass in nursery inoculations. Overall, this work provides a molecular-based framework in which to study potential native and introduced rhizobia on one of the most isolated archipelagos on the planet.


Introduction
The Hawaiian Islands due to their isolation have developed a biogeographic landscape that is prone to invasion of introduced organisms, including soil microbes that range from pathogenic to mutualistic lifestyles.Although most microbial introductions were not recorded, the Nitrogen Fixation for Tropical Agricultural Legumes (NifTAL) Project, which started in Hawai'i in 1975, introduced a large number of non-native rhizobia to the islands [1].The project expanded research on legumes, rhizobia, and inoculation methods that would increase the biological nitrogen fixation efficiency of legumes.At least some of the 1774 non-native rhizobia strains

Sampling and isolation
Soil and legume nodules were sampled across the island of O'ahu, Hawai'i with permission from landowners (Fig 1).The sampling location and site characteristics are shown listed in S1 Table .At each site, at least 3 soil samples were haphazardly sampled from the top 15 cm and combined to make one sample for a total of 377 samples.The soil samples come from diverse soil environments, some of which may not necessarily have leguminous plants actively growing.For sites that have leguminous plants, we targeted a diversity of species, and when possible, multiple plants of the same species were sampled.For each plant, the flowers, leaves, and pods were photographed for identification.Roots of the collected plants were washed thoroughly, and nodules were separated from the larger rooting system and washed until all apparent soil had been removed.Both soil and nodules were transported back to the laboratory for further processing.
For rhizobia isolation, one root nodule from each plant was washed in 95% ethanol for seven seconds, then soaked in a 2.5% sodium hypochlorite (30% bleach solution) with several drops of Tween 20 for 3 minutes and washed in five successive baths of sterile water.The surface-sterilized root nodules were crushed in 1 mL of sterile water and the nodule solution was then diluted serially to 10 −6 in water.100 μL of 10 −6 diluted root nodule solution was plated on yeast mannitol agar (YMA, 1 g yeast extract, 10 g mannitol, 0.5 g potassium phosphate dibasic, 0.2 g magnesium sulfate, 0.1 g sodium chloride, and 30 g agar, 1 L water) and incubated for at least 8 days at 27˚C.All morphologically distinct colonies from each nodule were further purified and maintained on YMA.

Molecular methods
Amplicon library preparation and sequencing followed standard protocols [17].Briefly, soil samples were homogenized and 0.25 g was used to extract DNA using the PowerSoil Max kit (Qiagen, USA).The amplicon libraries were prepared using a two-step, dual barcoding method [17], using the Earth Microbiome Project primer pairs 515F [18] and 806R [19].The libraries were sequenced with Illumina MiSeq.The DNA sequences were analyzed using QIIME2 [20] for data quality control following our established pipeline [21].The ASVs were identified using the SILVA v138 database [22], and the known genera of rhizobia [23] were separated from the main dataset for further analyses.
To identify the cultures, DNA was extracted from actively growing cells using a standard Cetyltrimethylammonium bromide (CTAB) buffer/chloroform extraction protocol and amplified using the primer pair 515FB (5' GTGYCAGCMGCCGCGGTAA 3') [18] and 926RB (5' CCGYCAATTYMTTTRAGTTT 3') [19] that target the V4-V5 variable region of the 16S rRNA gene using the following PCR parameters: 95˚C for 30s, 95˚C for 30s, 49˚C for 30 s, 68˚C for 60s, repeat steps 2-4 34x, 68˚C for 300 s.PCR products were cleaned using solid phase reversible immobilization (SPRI) magnetic beads prior to sequencing using Sanger technology.The resulting chromatograms were manually edited for correct base calls and sequences were clustered at the 99% sequence similarity using the online CD-HIT Suite [24] and representative sequences of each cluster were identified using BLAST that matches to GenBank's curated type strain database.

Inoculation assay
To determine the efficiency of rhizobia inoculations on the endemic and threatened tree E. sandwicensis, we selected two different Bradyrhizobium strains that were found with this host: Bradyrhizobium sp.strain JA1 (isolate UHC282) is a generalist found across the island of O'ahu across multiple hosts, and Bradyrhizobium strain JA9 (isolate UHC283) was found only in nodules associated with E. sandwicensis.A nutrient-poor Oxisol, Wahiawa Series soil collected from the Poamoho Research Station (21.5750803˚, -158.10241˚), was sieved using a 1 mm mesh, mixed 1:1 (v/v) with perlite to improve drainage, and autoclave sterilized at 121˚C for 30 minutes twice, with thorough mixing in between.Fresh cultures of the two isolates above were suspended in double deionized water and immediately inoculated onto the soil to achieve 10 6 rhizobia cells/mL of soil.Sterile plastic pots were filled with 480 mL of inoculated soil, or sterile uninoculated soil watered with 5 mL of 0.5% KNO 3 weekly as a positive control and watered with tap water as a negative control.Seeds were surface sterilized and scarified, prior to planting into each pot, and grown for 10 weeks under greenhouse conditions with daily watering.There were five replicates per treatment (N = 20).
To measure inoculum efficiency, we used a combination of several non-destructive metrics that allowed us to preserve these threatened plants.Stem height was measured from the soil line of the plant to the tip apical meristem; approximate stem volume was calculated using the formula for cone volume V = (1/3)πr 2 h using basal stem radius (r) and height (h).The total number of nodules was counted, and some were cut open to confirm active nodulation through color.Wet weight of the whole seedling was taken by first washing the root system clean of soil and dabbed dry prior to weighing, taking care to perform these steps as consistently as possible.Together, these data provided a general measure of inoculum efficiency of the inoculated rhizobia strains [25].

Statistical analyses
All statistical analyses were conducted by using R [26].The sampling map was drawn using the package ggmap [27].For each sample in the soil dataset, the sequence data were condensed at the genus level, and the data was collapsed to presence and absence.The data was then normalized across each sampling site by percentage.In other words, the prevalence of taxa in this study was based on its presence relative to total number of samples.This prevalence was plotted as a "bubble chart" using the ggplot2 package [28].To measure whether there is differential occurrence between site disturbance, we used the "adonis" function in the vegan package [29].The mean differences between E. sandwicensis growth were conducted via the Least Significant Distance test with a 0.95 confidence interval with the agricolae package [30], and graphs were plotted by using the ggplot2 package [28].When calculating the averages for each experimental group, the seedlings that did not have nodules were not incorporated.Only two of the seedlings in the positive (fertilized) control survived, and thus this treatment was removed from further analyses.

Rhizobia from soil sequences
In this study, we identified 18 genera of rhizobia on O'ahu from soils and nodules (Fig 2).Some genera could not be separated from each other (e.g. the Rhizobium complex = Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium; the Burkholderia complex = Burkholderia-Paraburkholderia; the Methylobacterium complex = Methylobacterium-Methylorubrum) based on the short V4 segment of the 16S rRNA gene.We note that it was not possible to clearly detect which members of Burkholderia and the Methylobacterium complex nodulate legumes based on the sequence data, but they have been included for completeness.Therefore, we report 13 operational genera.Of these, the Rhizobium complex, Bradyrhizobium, Devosia, Mesorhizobium, and Microvirga appear to be the most widespread, but only Cupriavidus was found across all sites sampled, although often at lower abundance than other genera.
Although the rhizobia genera do not seem to cluster according to sites, soil or land-use, agricultural sites tend to have most, if not all genera, whereas non-agricultural sites tend to have fewer (Fig 2).Of note is the Waimea Ridge site that is only accessible through helicopter.This site has the lowest genus richness with only Bradyrhizobium, the Burkholderia complex, Cupriavidus, and the Methylobacterium complex.There was no significance difference (p = 0.545) between the genera that occurred in agricultural soils compared to less disturbed or natural soils.
The overwhelming majority of individual plants associated with multiple rhizobia strains, whether they were collected from the same location (e.g.Arachis pintoi) or from different locations (e.g.Crotalaria juncea, Leucaena leucocephala) (Table 1).It is worth noting that the two endemic plant species E. sandwicensis and A. koa were both associated with Bradyrhizobium sp.strain JA1, but in addition, E. sandwicensis was associated with Bradyrhizobium sp.strain JA9.Generally, an individual nodule contains only one isolate of rhizobia (although they can contain other bacteria).We found that in two cases, a single nodule contained two different rhizobia strains.An Arachis pintoi nodule contained Bradyrhizobium JA1 and Rhizobium sp.strain JA4, and a Desmodium tortuosum nodule, contained both Rhizobium sp.strain JA4 and Agrobacterium sp.strain JA13.

Inoculation assays
Inoculation of seedlings using the cell slurry method successfully produced nodules and had positive effects on the host in certain measurements (Fig 3).80% of inoculated plants formed nodules, although the number of nodules was relatively few and tended to cluster close the main taproot.There was no significant number of nodules formed between the two isolates.Nodulated plants produced noticeably larger root systems compared to the control.Although we did not quantify the density of these roots, the contribution of the larger inoculated root system likely contributed to the overall significant increase in plant wet weight biomass   compared to the control (p < 0.001).Colonization by isolate UHC283 (strain JA9) significantly increased total wet plant biomass compared to isolate UHC282 (strain JA1) and the control.There were no significant differences between stem height (p = 0.662) and stem volume (p = 0.299) of either isolate compared to each other and the negative control.

Discussion
In this study, we characterized plant-nodulating rhizobia from 12 sites across the island of O'ahu, Hawai'i from both soil and active nodules and found that all plant-nodulating rhizobia genera are present on O'ahu, although only a subset of these genera were found forming nodules.We did not detect any discernable patterns in the distribution of these genera, suggesting that they are ubiquitous in disturbed habitats.Of these habitats, agricultural sites tend to have the largest number of genera compared to the undisturbed Waimea Ridge site that has the fewest genera.This observation is consistent with our understanding of the history of agriculture and introduction of microbes to agricultural sites, but the lower rhizobia genera richness reflects the lower richness of native Hawaiian plant species in intact habitats.Of the five most common genera, Cupriavidus can be found at all sites but only Bradyrhizobium can nodulate a variety of host plants.Cupriavidus (Order Burkholdariales) were isolated exclusively from Mimosa pudica nodules and is consistent with their preference for only Mimosa species [31].Here we found the genus to be present in the broad array of soils from nutrient poor agricultural soils to undisturbed native soils, and soils with thick fungal mycelium mats.Many of these soils do not have M. pudica plants, but there is evidence showing that Cupriavidus necator can nodulate various leguminous species [32,33], including Leucaena leucocephala that is widely distributed across O'ahu, with the exception of the more undisturbed sites.In contrast, the most common strain that associates broadly with many hosts was Bradyrhizobium JA1.These findings reflect other works showing that Bradyrhizobium species are ubiquitous in tropical soils due to their ability to nodulate a broad range of tropical legumes [9,34,35].We also isolated Ensifer, Mesorhizobium, and Rhizobium, but these occurred in lower frequency than Bradyrhizobium, although they are also widespread.The exceptions are Devosia and Microvirga (Order Hyphomicrobiales), which we were not able to isolate, perhaps due to their low level of occurrence.The occurrence of Cupriavidus and other rhizobia such as Mesorhizobium in the rhizosphere of many non-legume plant hosts and fungal-rich habitats suggests that they have an external "free living" phase in the rhizosphere where they consume root and fungal exudates [36][37][38], and a more well-known symbiotic phase inside of nodules.A much more detailed sampling and species-level identification these rhizobia strains will be required to confirm these associations.
We observed an occurrence of multiple rhizobia strains within a single nodule ("mixed nodulation") of the ground peanut Arachis pintoi.The nodule contained Bradyrhizobium sp.strain JA1 and Rhizobium sp.strain JA4.This phenomenon of mixed nodulation is unusual because through host sanctioning, legumes preferentially select strains that have higher rates of nitrogen fixation.Certain rhizobia species are inefficient at fixing nitrogen or cannot fix nitrogen; thus, these rhizobia would be sanctioned from nodulating the respective legume [39].However, mixed nodulation of different strains with varying nitrogen fixation efficiencies does not necessarily impact the plant negatively [40].Therefore, less efficient rhizobia strains have evolved mechanisms that allow them to escape sanction and cohabit the same nodule with efficient strains.The fact that we have discovered multiple strains associating with a single nodule with limited sampling suggests that this occurrence is more common in nature than previously thought and deserves further attention in future works.
We showed that soil inoculations with fresh cultures, followed by planting E. sandwicensis seeds, is a relatively effective way to introduce rhizobia to seedlings in a nursery setting.At the soil inoculation rate of 1×10 6 cells/ml of soil, only 80% of seedlings were successfully nodulated.Various factors could have contributed to this, including the deactivation active cells prior to seedlings forming colonizable roots.The time of inoculation to germination of E. sandwicensis usually ranges about 5-10 days.Perhaps this time delay was enough to render enough rhizobia cells dead or inactive so that they are no longer able to nodulate seedlings.Seedlings that were successfully inoculated had significantly larger biomass.Since there were no significant differences in stem height and volume, the difference was likely driven by a larger and denser root system.Measurements using dry weight would have been the best practice, but we chose to preserve these threatened plants and found that a combination of various measurements provided acceptable data to distinguish the treatments apart [25].We found that inoculation using Mesorhizobium strain JA1 wasn't as effective as the less common Bradyrhizobium strain JA9.This indicates that perhaps strain JA9 is more specific to E. sandwicensis, although much more rigorous sampling will be required to confirm this relationship.
Our understanding of the patterns of rhizobia and host associations are limited in this study because of several factors.Although our soil dataset allowed some inferences into the occurrence of rhizobia across the island, our modest sampling of host-connected nodules and the low resolution of 16S rRNA gene did not allow us to make strong conclusions about the specific patterns of host and symbiont associations, nor did they provide enough insights into whether some strains were native or introduced.However, as Hawai'i is the "invasive capital" of the world, it is likely that many of the strains we found with non-native plants also came by ways of human introductions.What is clear, however, is that the NifTAL project brought in many non-native strains to Hawai'i [5] and these strains likely still exist, especially in agricultural soils such as the ones we sampled in this study.The strains isolated from the two species of endemic plants came from somewhat disturbed habitats, and thus it is difficult to determine whether they are truly native strains.On the island of O'ahu where we sampled, the few intact habitats left are found in upper elevation ridges with relatively minor disturbance from anthropogenic activities.Future sampling of native legumes from these undisturbed habitats will be more meaningful to isolate native strains, as there is a general pattern of native microbes occurring with native plants, exemplified by some species of mushroom-forming fungi in native forest habitats [41].

Conclusions
In this study, we showed that the soils across the island of O'ahu contain diverse rhizobia, some of which can be isolated from legume nodules.We gained insights into some general patterns such as the richness of rhizobia genera, and the generalist strains that occur widely across our sampling sites.This suggests that there is a constant source of rhizobia inoculants in these soils, and cultivation of leguminous plants do not need inoculants, except for those that associate with specific strains of rhizobia.Using strains that were isolated from E. sandwicensis, we showed that E. sandwicensis seedlings can be inoculated with fresh cultures, resulting in seedlings with significantly larger biomass.Although both strains resulted in positive seedling growth, there appears to be differences in plant biomass, so therefore we recommend that seedlings be inoculated with Bradyrhizobium strain JA9 in the nursery before transplanting into the ground.Future research to test delivery systems, such as applying lyophilized inoculum directly into the soil, mixing directly into water and applying to the medium surface with growing plants, or coating of seeds for field inoculations will be helpful to the restoration efforts of the threatened and endangered legumes in Hawai'i.

Fig 1 .
Fig 1. Map of sampling locations across the island of O'ahu, Hawai'i.The size of the dots indicates the number of samples collected and sequenced at that site.https://doi.org/10.1371/journal.pone.0291250.g001

Fig 2 .
Fig 2. Genera of plant-nodulating rhizobia that occur on the island of O'ahu, Hawai'i.The presence of each genus (or genus complex) is indicated by a dot where the size corresponds to the normalized abundance.The sites are clustered based on the presence of genera and their normalized abundance.https://doi.org/10.1371/journal.pone.0291250.g002