Genotypic and phenotypic analyses reveal distinct population structures and ecotypes for sugar beet‐associated Pseudomonas in Oxford and Auckland

Abstract Fluorescent pseudomonads represent one of the largest groups of bacteria inhabiting the surfaces of plants, but their genetic composition in planta is poorly understood. Here, we examined the population structure and diversity of fluorescent pseudomonads isolated from sugar beet grown at two geographic locations (Oxford, United Kingdom and Auckland, New Zealand). To seek evidence for niche adaptation, bacteria were sampled from three types of leaves (immature, mature, and senescent) and then characterized using a combination of genotypic and phenotypic analysis. We first performed multilocus sequence analysis (MLSA) of three housekeeping genes (gapA, gltA, and acnB) in a total of 152 isolates (96 from Oxford, 56 from Auckland). The concatenated sequences were grouped into 81 sequence types and 22 distinct operational taxonomic units (OTUs). Significant levels of recombination were detected, particularly for the Oxford isolates (rate of recombination to mutation (r/m) = 5.23 for the whole population). Subsequent ancestral analysis performed in STRUCTURE found evidence of six ancestral populations, and their distributions significantly differed between Oxford and Auckland. Next, their ability to grow on 95 carbon sources was assessed using the Biolog™ GN2 microtiter plates. A distance matrix was generated from the raw growth data (A 660) and subjected to multidimensional scaling (MDS) analysis. There was a significant correlation between substrate utilization profiles and MLSA genotypes. Both phenotypic and genotypic analyses indicated presence of a geographic structure for strains from Oxford and Auckland. Significant differences were also detected for MLSA genotypes between strains isolated from immature versus mature/senescent leaves. The fluorescent pseudomonads thus showed an ecotypic population structure, suggestive of adaptation to both geographic conditions and local plant niches.


| INTRODUC TI ON
The health and well-being of plants is to a large extent determined by the microorganisms with which they coexist (Compant, Samad, Faist, & Sessitsch, 2019). While some bacteria can cause disease, others have the capacity to confer disease resistance and promote plant growth. A typical example concerns bacteria from the genus of Pseudomonas, which are ubiquitously found in soil and waters, but also form intimate associations with plants (Silby, et al., 2011).
Sugar beet is commercially grown in Europe for sugar production.
Early studies performed in the late 1980s with field-grown sugar beets at the University of Oxford farm (Wytham, Oxford) showed that fluorescent pseudomonads are the largest group of bacteria inhabiting the phyllosphere, and their species composition changes during the growing season (Rainey, Bailey, & Thompson, 1994). As a representative of sugar beet-associated pseudomonads, P. fluorescens SBW25 was used as a model for further genetic and biological analysis of plant-bacterial interactions (Bailey, Lilley, Thompson, Rainey, & Ellis, 1995;Rainey, 1999;Silby et al., 2009). First of all, it is interesting to note that P. fluorescens SBW25 is able to aggressively colonize other crops such as wheat, maize, and peas, suggesting that the interactions are not species-specific (Humphris et al., 2005;Jaderlund, Hellman, Sundh, Bailey, & Jansson, 2008). This bacterium has, thus, likely evolved functional traits for successful plant colonization in general (Rainey, 1999). Both in vivo and in vitro studies indicated that SBW25 can protect sugar beet seedlings against damping-off disease caused by the soilborne fungal pathogen Pythium ultimum (Ellis, Timms-Wilson, & Bailey, 2000). A nonproteinogenic amino acid (l-furannomycin) was identified as one of the antimicrobial compounds produced by SBW25 (Trippe, McPhail, Armstrong, Azevedo, & Banowetz, 2013). Furthermore, promoter trapping techniques were developed for the SBW25/sugar beet model, and their subsequent application led to identification of 139 loci whose expression is elevated during bacterial colonization in planta (Rainey, 1999;Silby et al., 2009). Some plant-inducible genes, particularly those involved in biofilm formation and histidine utilization (hut), have been investigated in great detail (Gal, Preston, Massey, Spiers, & Rainey, 2003;Liu, Rainey, & Zhang, 2015). However, our understanding of the genetic diversity and population structure of fluorescent Pseudomonas is limited.
In a previous study, 30 fluorescent pseudomonads were isolated from the phyllosphere of field-grown sugar beet in Oxford from where P. fluorescens SBW25 originated (Rainey et al., 1994). These isolates represented Pseudomonas present during a single growing season. They were subjected to restriction fragment length polymorphism (RFLP) analysis and phenotypic characterization of fatty acid methyl ester (FAME), biochemical properties and carbon source assimilation. The phenotypic data consistently showed that these isolates were grouped according to their time of sampling and leaf type (immature, mature, and senescent). While the RFLP data were complicated by the presence of megaplasmids, the derived genotypic groups were closely correlated with clusters generated on the basis of the phenotypic data (Tett et al., 2007). The data thus implicated niche adaptation of pseudomonads to the local plant conditions. This initial finding prompted further analysis of the Pseudomonas population structure whereby a total of 108 isolates were collected in a single sampling occasion in the same field in Oxford (Haubold & Rainey, 1996). These isolates were genetically and phenotypically characterized using 10 allozyme and 23 biotype markers. The allozyme data indicated that the Pseudomonas population was in overall linkage disequilibrium and showed an ecotypic structure. There was a significant correlation between isolate distribution and habitat, that is, leaf type and plot. Moreover, the data also suggested a probability of frequent large-scale recombination among certain isolates.
Multilocus sequence analysis (MLSA) has become a universal technique for studying the population genetics of bacteria, including Pseudomonas (Bennasar, Mulet, Lalucat, & Garcia-Valdes, 2010;Castaneda-Montes et al., 2018b;Ogura, Shimada, & Miyoshi-Akiyama, 2019). It involves a comparative sequence analysis of three or more housekeeping genes, which together provide higher resolution of the phylogenetic relationships, when compared with analysis of 16S rRNA genes. Nucleotide sequences can be obtained from DNAs amplified by PCR or directly extracted from genome sequences if available. While whole-genome sequencing (WGS) can provide information about the entire gene content, and thus an idea of the pan-genome (Karasov, Barrett, Hershberg, & Bergelson, 2017;McCann et al., 2017), inferences on parameters governing molecular evolution and geographic structure can readily be obtained from a detailed analysis of a small set of conserved genes (Ogura et al., 2019;Straub et al., 2018). For Pseudomonas populations, MLSA has most frequently been used for analysis of the plant pathogenic bacterium P. syringae (Akira & Hemmi, 2003;Straub et al., 2018), and the opportunistic human and animal pathogen P. aeruginosa (Castaneda-Montes et al., 2018b;Kidd et al., 2012). MLSA schemes have also been developed for P. putida and P. fluorescens (Andreani et al., 2014;Garrido-Sanz et al., 2016;Ogura et al., 2019). However, MLSA has rarely been applied to plant-associated fluorescent Pseudomonas (Alvarez-Perez, de Vega, & Herrera, 2013), which comprise several phylogenetically distinct species with a common feature of pyoverdine production. Pyoverdines are siderophores secreted by fluorescent pseudomonads for iron acquisition .
They are normally used as a marker for strain identification because of the distinguishable fluorescent yellow-green color.
Here, we describe the population structure and diversity of fluorescent Pseudomonas inhabiting the phyllosphere of sugar beet (Beta vulgaris var. Amethyst). The same plant cultivar was grown in two geographic locations (Oxford, United Kingdom and Auckland, New Zealand), and bacterial samples were taken from three leaf types (immature, mature and senescent). We first performed MLSA analysis and obtained complete sequences of three genes (gapA, gltA, and acnB) for a total of 152 isolates. The MLSA data indicated that the Pseudomonas population was primarily associated with geographic location and leaf type from where they were isolated. We found evidence of significant recombination and identified six ancestral genotypes. Next, we performed Biolog assays to determine the ability of Pseudomonas to grow on 95 unique carbon sources, including histidine and its derivate urocanate. The data allowed assessment of the potential correlations between the observed genotypes and phenotypes, and a discussion of the underlying mechanisms of bacterial diversification using the dissimilation of histidine and urocanate as an example.

| Isolation and culture of pseudomonads
A total of 170 leaf-colonizing Pseudomonas strains were isolated from nine sugar beet plants (Beta vulgaris var. Amethyst) in three 5-m 2 experimental plots at the University of Oxford farm, Wytham, Oxford, UK in 1993. Some of these strains (108 isolates) were subjected to a previous multilocus enzyme electrophoresis (MLEE) analysis wherein details of the sampling methods are provided (Haubold & Rainey, 1996). Briefly, three plants were randomly sampled per plot, and each plant was divided into three leaf types (immature, mature and senescent). For each leaf type, two replicate leaves were placed in a sterile plastic bag containing 5 ml sterile water. After a 1 min "massage," the bacterial suspensions were dilution-plated onto King's Medium B (KB) supplemented with CFC (10 μg/ml cetrimide, 10 μg/ml fucidin, and 50 μg/ml cephalosporin) from Oxoid (Hampshire, UK). After incubation at 28°C for 48 hr, two colonies were randomly picked for each leaf sample and purity of the isolates was further checked by streaking onto KB agar plates. The obtained isolates were subcultured in lysogeny broth (LB) and subsequently stored frozen at −80°C by mixing 1.0 ml culture with 0.8 ml glycerol saline solution (70% glycerol, 0.85% NaCl).
In 2004, five plants of the same B. vulgaris variety were grown in native soil at one site located in West Auckland, New Zealand, and Pseudomonas strains were isolated similarly using the KB + CFC selective plates. Isolates were coded similarly as strains from Oxford, indicating their origin of isolation: plot (1), plant (1 to 5), leaf type (1, senescent; 3, mature; 5, immature) and replicate leaf (a or b). A prefix X was assigned to distinguish the Auckland isolates from those from Oxford. For example, isolate X131b1 identifies isolate 1 from Auckland plot 1, from plant 3, from senescent (1) leaf "b." Isolates from Oxford were additionally assigned a more simplified code from U100 to U269.

| PCR amplification and DNA sequencing
Total DNA was extracted from bacterial cells using the CTAB-based method as previously described (Zhang, Kosier, & Priefer, 2001).
Briefly, bacteria were resuspended in 567 μl TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0) to which 30 μl of 10% SDS and 3 μl of 20 mg/ml proteinase K were added. After lysis at 37°C for 1 hr, 100 μl of 5 M NaCl and 80 μl of CTAB/NaCl solution (10% hexadecyltrimethyl ammonium bromide, 0.7 m NaCl) were added and incubated at 65°C for 10 min. DNAs were subsequently extracted with equal volume of phenol/chloroform and precipitated with 0.6 volume of isopropanol. PCR was performed in a 25 μl reaction containing 1x PCR buffer, 0.2 mM dNTPs, 1.6 mM MgCl2, 2% DMSO, 10 pmol for each primer, 30-100 ng template DNA and 0.2 units Taq polymerase (Invitrogen). After an initial denaturation at 94°C for 3 min, DNAs were amplified in 30 cycles of denaturation at 94°C for 30 s, annealing at 58°C for 30 s, elongation at 72°C for I min, followed by a 3-min final extension at 72°C. The PCR products were purified using the Exo-CIP Rapid PCR Cleanup Kit (New England Biolabs), before they were sent to Macrogen Inc (South Korea) for DNA sequencing.

| Assays for bacterial growth
The GN2 microtiter plates (Biolog Inc., Hayward, CA) were used following the manufacturer's instructions. Inoculants were prepared by first growing stored bacteria in LB broth and subcultured once in R2A broth (Reasoner & Geldreich, 1985). Cells were spun down and resuspended in the same volume of sterile deionized water; 100 μl of this bacterial suspension was then inoculated into 15 ml of the IF-0 inoculating fluid from Biolog and subsequently starved at 28°C for 2 hr. Next, 150 μl of the starved cells was pipetted into each well of the GN2 MicroPlate and incubated at 28°C for 48 hr. The initial and final cell density was estimated by measuring absorbance at the wavelength of 660 nM (A 600 ) in a Synergy 2 plate reader equipped with the Gen5 software (BioTek Instruments). Growth on histidine and urocanates was assessed with the minimal M9 salt medium as previously described (Zhang et al., 2012).

| Data analysis
Geneious (Biomatters Ltd; Auckland, New Zealand) was used to manipulate the DNA sequences, including multiple sequence alignment, trimming, and concatenation. Unique sequence types (STs) were grouped into operational taxonomic units (OTUs) that included all STs that varied by the mean pairwise distance (0.06) of the total sample. Rarefaction analysis was performed using MOTHUR v.1.34.4 (Schloss et al., 2009). The numbers of STs and OTUs were determined by selecting the isolates at random from the population, and the procedure was repeated 1,000 times. The Simpson's index (1-D) of diversity was calculated to compare variability in isolates from Oxford and Auckland, and it takes into account both the number of taxa (i.e., STs or OTUs) and the relative abundance of each taxa (Simpson, 1949 (Anderson, 2001). Pairwise distances were used as input, and the tests were run with 9,999 permutations. Visual comparison of isolates from different ecological sources (location, plot, plant and leaf type) was performed by multidimensional scaling (MDS) using PRIMER v6. Ancestral subpopulations of Pseudomonas were determined using STRUCTURE software (Pritchard, Stephens, & Donnelly, 2000).
Recombination was detected using the likelihood permutation test in LDhat v2.1 (McVean, Awadalla, & Fearnhead, 2002) as previously described (Kidd et al., 2012). Evidence of recombination breakpoints was further obtained from the aligned multilocus sequences using the SBP/GARD server (Kosakovsky Pond, Posada, Gravenor, Woelk, & Frost, 2006). GARD examined 2,717 models in 00:16:25 wallclock time, at a rate of 2.76 models/second. The alignment contained 340 potential breakpoints, translating into the search space of 6,550,950 models with up to 3 breakpoints, of which 0.04% was explored by the genetic algorithm. Significance of the topological incongruence was inferred by KH test (Kishino & Hasegawa, 1989).
The Biolog growth data (A 660 ) were analyzed by first eliminating the 28 carbon sources which did not support any bacterial growth for all tested Pseudomonas isolates. To cluster the growth phenotypes, a distance matrix was generated from absorbance data of the remaining 67 carbon sources using PRIMER v6. The matrix was then subjected to PERMANOVA implemented in PRIMER V6 as described above for genotypic data. Correlation between phenotype and genotype was performed using principal component analysis (PCA) and canonical variate analysis (CVA) (Renaud, Dufour, Hardouin, Ledevin, & Auffray, 2015).

| Multilocus sequencing of pseudomonads isolated from the phyllosphere of sugar beet
A total of 230 isolates (164 from Oxford, 66 from Auckland) were initially subjected to sequencing of seven housekeeping genes (gapA, gltA, acnB, gyrB, coxC, pgi, and ropD) using the universal primers (Hwang et al., 2005). However, the rates of PCR or sequencing failure were very high, suggesting that the leaf-colonizing Pseudomonas populations are highly diverse. After modification of the primer sequences (Table 1), high-quality DNA sequences were obtained for three genes (gapA, gltA, and acnB) in a total of 152 isolates (96 from  (Figure 1a), indicating that the sample was a good representation of the genetic diversity. As summarized in  Recombination, as mentioned above, was detected in a single gene (gapA) but not in the concatenated sequence by the likelihood permutation test. However, when the analysis was performed with data from Oxford and Auckland separately, significant levels of recombination were noted in the concatenated sequence, particularly for the Oxford population; these were further supported by single breakpoint recombination analysis using the 10,340 model (Table 3).

| Genetic diversity of the Pseudomonas populations
Further GARD analysis revealed two breakpoints with significant topological incongruence, indicating recombination breakpoints between the MLSA genes ( Figure 4a).
Recombination events can be utilized to separate the isolates into populations on the basis of ancestry; that is, the presence of linked alleles in multiple isolates mandates that they have shared ancestry to some degree. Obviously, multiple sources of ancestry can be present in any one isolate. Ancestral analysis performed in STRUCTURE found evidence of six ancestral populations, and their distributions at the level of the population and isolate are presented in Figure 4b,c, respectively. Few isolates were "highly recombinant" without any predominant source of their ancestry (8% in total, mostly from Oxford).

| Phenotypic analysis of the Pseudomonas populations
Growth data were obtained for a total of 95 carbon substrates in- Finally, it should be noted that minor but significant differences were detected between genotypes for strains isolated from immature versus mature leaves (ANOSIM, r = .103, p = .01), and also for immature versus senescent leaves (r = .075, p = .01), but not for mature versus senescent leaves (r = .015, p = .74). Similar results were found in terms of growth phenotypes on histidine and urocanate: immature versus mature leaves (r = .185, p = .01), immature versus

TA B L E 3 Analysis of recombination in
Pseudomonas populations by location senescent leaves (r = .136, p = .01), and mature versus senescent leaves (r = .05, p = 1.7). No significant differences were detected for the potential effects of plants and plots. The two subpopulations did not share any unique sequence types, and only have four common OTUs clustered according to the mean pairwise distance of the population. Given the large geographic distance between Oxford and Auckland, this finding is not surprising, as it is generally consistent with our current knowledge of microbial biogeography (Nemergut et al., 2013). The observed difference can be explained by the combined effects of historical contingencies and contemporary environmental disturbances (Martiny et al., 2006;Sun et al., 2014).  -Perez et al., 2013;Andreani et al., 2014;Frapolli et al., 2007).  Wang et al., 2001). Therefore, further larger-scale MLSA analyses with multiple cultivars and multiple geographic locations are necessary to verify the roles of geographic factors and local plant environmental conditions in shaping Pseudomonas population structure.

| D ISCUSS I ON
Recombination is a major driving force in shaping bacterial genetic diversity, but its relative importance to mutation varies greatly among different species (Gonzalez-Torres, Rodriguez-Mateos, Anton, & Gabaldon, 2019). A previous survey showed that the highest and lowest r/m values differed by three orders of magnitude in bacteria and archaea (Vos & Didelot, 2009). The ability of recombination to cause changes in the genome exceed that of mutation (r/m > 1) in more than half of the analyzed bacterial and archaeal species (56%, 27 out of 48 Furthermore, we found that utilization of eight carbon substrates was primarily responsible for separating the Oxford and Auckland subpopulations. These included urocanic acid (or urocanate).
Urocanate is the first intermediate of the histidine degradation pathway (Zhang & Rainey, 2007). Both histidine and urocanate are included in the GN2 MicroPlate, and their utilization patterns were tested in this study. If a strain can grow on histidine (His + ), it must have all the catabolic enzymes required for the utilization of urocanate.
However, certain pseudomonads can grow on histidine, but not on urocanate (His + , Uro -), suggesting that these strains lack a functional transport system for urocanate uptake. This led to a hypothesis that variation in histidine and urocanate utilization is attributable to genetic differences in transport systems (Zhang et al., 2012). This hypothesis was confirmed in a previous study by heterogeneous complementation after the urocanate-specific transporter (HutTu) was identified in the model strain of P. fluorescens SBW25 (Zhang et al., 2012). Here, we observed a significant association of genotypes with the utilization of urocanate but not histidine. While almost all Oxford strains (95%) were capable of growing on urocanate, only one-third could grow on urocanate in the Auckland subpopulation ( Figure 7). Given the absence of historical sugar beet cultivation in the Auckland soil, our finding sits in accord with the previously proposed niche-specific accumulation of urocanate in planta (Zhang, Ritchie, & Rainey, 2013). Urocanate may act not only as a nutrient but also an important signal for successful bacterial colonization.
Based on this, it is logical that urocanate utilization is widespread in the Oxford subpopulation as a result of adaptation to the local conditions of sugar beet phyllosphere. Fitness improvements associated with urocanate utilization are likely to occur through changes in the uptake systems (Dean, 1995;Zhang et al., 2012). Together, our data provide an example of the genetic basis of phenotypic variation for plant-associated fluorescent pseudomonads.

ACK N OWLED G M ENTS
We thank Annabel Gunn and Yunhao Liu for technical assistance.
XXZ appreciates financial support by MBIE Catalyst Fund, New Zealand (project no. 92846082). RWJ was supported by a British Society for Plant Pathology fellowship award.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.