Invasive Lactuca serriola seeds contain endophytic bacteria that contribute to drought tolerance

The mutualistic relationship between alien plant species and microorganisms is proposed to facilitate or hinder invasive success, depending on whether plants can form novel associations with microorganisms in the introduced habitats. However, this hypothesis has not considered seed endophytes that would move together with plant propagules. Little information is available on the seed endophytic bacteria of invasive species and their effects on plant performance. We isolated the seed endophytic bacteria of a xerophytic invasive plant, Lactuca serriola, and examined their plant growth-promoting traits. In addition, we assessed whether these seed endophytes contributed to plant drought tolerance. Forty-two bacterial species were isolated from seeds, and all of them exhibited at least one plant growth-promoting trait. Kosakonia cowanii occurred in all four tested plant populations and produced a high concentration of exopolysaccharides in media with a highly negative water potential. Notably, applying K. cowanii GG1 to Arabidopsis thaliana stimulated plant growth under drought conditions. It also reduced soil water loss under drought conditions, suggesting bacterial production of exopolysaccharides might contribute to the maintenance of soil water content. These results imply that invasive plants can disperse along with beneficial bacterial symbionts, which potentially improve plant fitness and help to establish alien plant species.


Results
Identification of seed endophytic bacteria and their PGP traits. We collected wild L. serriola seeds from four natural populations in the vicinity of the Gwangju Institute of Sciences and Technology, Gwangju, South Korea (Fig. 1, Supplementary Table S1). Seeds from 4-12 plants from each population were composited, and 80 seeds from each composite sample were used to isolate endophytic bacteria.
A total of 129 bacterial strains were isolated, and their 16S rDNA sequences were compared with previously reported sequences of bacterial type strains in the EZBioCloud database (Chunlab, Seoul, South Korea). The 16S rDNA sequences of all isolates exhibited more than 99% similarity with the type strains. A total of 18 genera were identified (Fig. 2, Table 1): Acidovorax, Bacillus, Cellulosimicrobium, Chryseobacterium, Cronobacter, Curtobacterium, Enterobacter, Enterococcus, Erwinia, Exiguobacterium, Kosakonia, Paenibacillus, Pantoea, Pseudomonas, Rhizobium, Saccharibacillus, Stenotrophomonas, and Xanthomonas. K. cowanii was detected in all four populations, and Cr. dublinensis was detected in three populations (Fig. 3). Across all plant populations, most isolates belonged to Gammaproteobacteria (Fig. 2). Firmicutes constituted 13-25% of the isolates, whereas no Firmicutes specimens could be detected in the YJ population. Alpha-and Betaproteobacteria, Actinobacteria, and Bacteroidetes were detected in either MS or GG populations, but not in YJ and SD populations.
PGP traits were examined for 42 representative isolates (Table 1; Supplementary Fig. S1). All tested isolates produced ACC deaminase, and 95.2% of the isolates could grow without a nitrogen source. In addition, 85.7% and 69% of the isolates could solubilize insoluble phosphate and produce siderophores, respectively. Less than half of the isolates produced IAA. IAA production was not observed for K. cowanii isolated from the GG site, whereas it was detected for those from other sites. Thus, even though isolates were assigned to the same species based on their 16S rDNA sequence similarity, their PGP activities likely depended on their source population. These 42 isolates have been deposited in the Korean Collection for Type Cultures (KCTC) ( Table 1).
Bacterial growth under drought conditions and EPS production. All 42 representative isolates produced capsular materials outside their cells ( Supplementary Fig. S1). When the isolates were grown in media with a water potential of -0.73 MPa, Pa. ananatis SD16 presented the highest OD value (1.65), whereas Pa. nicotianae MS14 presented the lowest OD value (0.27) (Supplementary Fig. S1). For EPS quantification, K. cowanii and the two isolates with the highest OD values were selected from each source population.
The bacterial isolates grown in media with a water potential of − 0.73 MPa tended to produce more EPS than those grown under non-stress condition (F = 8.09, P < 0.01). Notably, isolates responded differently to the stress condition, as indicated by the significant isolate × condition interaction (F = 47.62, P < 0.001) (Fig. 4). EPS production increased under stress condition in K. cowanii YJ4 (t = 6.04, adjusted P < 0.001), K. cowanii SD1 (t = 5.33, adjusted P < 0.001), K. cowanii MS1 (t = 17.12, adjusted P < 0.001), K. cowanii GG1 (t = 16.95, adjusted P < 0.001), and Pa. ananatis GG19 (t = 13.91, adjusted P < 0.001). In contrast, Erwinia tasmaniensis YJ6 exhibited lower EPS production under stress condition (t = 9.00, adjusted P < 0.001). K. cowanii MS1 exhibited the highest EPS production (719.00 µg/mL), whereas Pa. hunanensis GG17 presented the lowest EPS production (76.67 µg/ mL) under stress condition.  www.nature.com/scientificreports/ Effects of isolates on the drought tolerance of A. thaliana. After the discontinuation of water supply for 10 days, the soil water content in the drought treatment decreased to 3-5%. Notably, the effects of bacterial treatments on the soil water content differed between the water treatments, as indicated by the significant bacteria × water treatment interaction (F = 2.03, P < 0.05). Under moist conditions, the soil water content of the bacteria-inoculated treatments was similar to that of the control without inoculation. In contrast, the water content of soil inoculated with K. cowanii GG1 was higher than that of the control without bacteria under drought conditions (t = 3.50, adjusted P < 0.05) (Fig. 5a). In plants grown under drought conditions, the bacterial treatment affected shoot fresh weight (F = 2.52, P < 0.05) and MDA content (F = 2.62, P < 0.05) (Fig. 5b, and e). In particular, A. thaliana inoculated with K. cowanii GG1 produced heavier shoots (t = 2.97, adjusted P < 0.05) than the untreated control (Fig. 5b). Root

Figure 2.
A phylogenetic tree constructed based on the 16S rDNA sequences of seed endophytic bacteria isolated from Lactuca serriola seeds and their closely related type strains. A total of 167 nucleotide sequences were analyzed using MEGA software (Version 7.0, https:// www. megas oftwa re. net) 30 . The evolutionary distances were computed using the Kimura 2-parameter method and trees were constructed using the neighbor-joining method. The optimal tree with the sum of branch length = 1.11 is shown. The positions containing gaps and missing data were eliminated. A total of 1,176 positions were included in the final dataset. www.nature.com/scientificreports/ dry weight (F = 1.56, P = 0.14) and relative water content (F = 1.10, P = 0.38) did not differ between the bacterial treatments under drought condition (Fig. 5c, and d), but MDA contents did (F = 2.62, P = 0.01) (Fig. 5e). The soil water content under drought conditions was positively correlated with the shoot fresh weight and negatively with the MDA concentration ( Fig. 6a, b, and c). EPS production in media with water potential of − 0.73 MPa was not correlated with soil water content (Fig. 6d), plant shoot fresh weight, or MDA concentration.

Discussion
Invasive L. serriola plants possess diverse seed endophytic bacteria that produce PGP molecules. Some of them survived and increased EPS production in the growth medium under water stress conditions. When grown under drought conditions, A. thaliana plants inoculated with K. cowanii GG1 presented a higher shoot biomass than those without bacterial inoculation. A total of 129 bacterial strains were isolated from L. serriola seeds and were assigned to 42 species. All isolated species, except Ce. aquatile and Cr. dublinensis, have previously been reported as plant-associated bacteria in the rhizosphere, roots, leaves, stems, flowers, or seeds of crop plant species (see Supplementary Table S2 for references). Notably, some of these species have been shown to produce PGP molecules and promote plant growth or stress tolerance. For instance, Pa. ananatis increases the root weight of pepper seedlings 31 , and Pa. dispersa ameliorates salt tolerance in wheat 32,33 . Ba. altitudinis and En. hormaechei subsp. steigerwaltii increase root length in wheat and maize [34][35][36] . All 42 representative strains in this study exhibited at least one of the tested PGP traits (Table 1), indicating that the invasive plant L. serriola harbors diverse endophytic bacteria with the potential to promote plant growth.
All bacterial isolates produced ACC deaminase. Under environmental stress, the ethylene level increases in plants, resulting in the inhibition of plant growth 37 . ACC is the immediate precursor of ethylene. ACC deaminase cleaves ACC into α-ketobutyrate and ammonia. Thus, bacteria producing ACC deaminase can block ethylene production, which facilitates plant growth under diverse environmental stresses, including drought, salt stress,     39,40 . One species of the Gammaproteobacteria, K. cowanii, was detected in all L. serriola populations. K. cowanii (Enterobacter cowanii) was first isolated from the leaf tissues of Eucalyptus plants exhibiting signs of blight disease. Although it is known to cause bacterial spots in Mabea fistulifera 41 , endophytic K. cowanii isolated from Tylosema esculentum exhibited diverse PGP traits 42 . In addition, EPS production by K. cowanii was examined for its application as a plant growth promoter 43,44 .
K. cowanii isolated in the present study produced more EPS when grown in media with a negative water potential than in control conditions. This is a characteristic of bacterial strains that facilitate plant tolerance to environmental stress 28,29 . K. cowanii are known to produce α-type heteropolysaccharides, and the polysaccharide composition depends on growth conditions 43,44 . As observed for Pseudomonas aeruginosa, environmental stresses, including drought, can induce the accumulation of diguanylate cyclase, which increases EPS production, although these mechanisms in K. cowanii are unknown 45 .
We hypothesized that L. serriola seeds might contain bacterial strains that potentially contribute to the drought tolerance of the species. Although bacterial strains isolated from the rhizosphere, roots, and leaves have been shown to promote plant performance under drought stress [46][47][48][49] , it is not known if seed endophytic bacteria also have similar activities. In this study, a bacterium isolated from the seeds, K. cowanii GG1, improved the aboveground biomass in A. thaliana under drought stress conditions, when compared with untreated A. thaliana.
The mutualistic relationship between alien plants and microorganisms has been suggested to facilitate or hinder invasive success [13][14][15] . The development of a novel beneficial association between the pre-existing microorganisms in the introduced habitats and an alien plant promotes the establishment of the alien species. In contrast, invasive success would diminish if no mutualistic microorganisms are available in the introduced habitats. In www.nature.com/scientificreports/ this context, differential soil microbiota between the native and introduced communities would act as a barrier to the invading alien plant species. Soil microbiota in the introduced habitats have been postulated as a major source of mutualistic microorganisms, but our results suggest that seed endophytic bacteria might be another source. Microorganisms within seeds of alien plants can disperse with the plant germplasm to novel habitats, and at least some of them, like K. cowanii in this study, might be able to promote the plant performance [18][19][20] . Seed endophytic bacteria, similar to the soil microbiota, might influence the invasive dynamics of alien plants. For instance, if seeds and their beneficial endophytes dispersed simultaneously from their native to the introduced range, seed endophytic bacteria might facilitate the initial establishment of alien plants, a critical step of plant invasion 12 .
Inoculation with K. cowanii GG1 increased soil water content under drought conditions as well as the aboveground biomass of A. thaliana. Moreover, the soil water content in the bacteria-inoculated treatments was positively correlated with the aboveground biomass and negatively with the MDA content of plants under drought conditions (Fig. 6). These results indicate that isolated bacteria influence soil water retention under drought stress, consequently affecting the production of shoot biomass and controlling physiological damage to the host plants, as indicated by the MDA concentration 46,50 .
As bacterial EPS production likely improves soil water retention 26 , we expected that bacterial EPS production in the media would be positively correlated with the water content of soil with bacterial inoculation. K. cowanii GG1 produced a relatively high concentration of EPS when grown in media with a negative water potential, and increased the soil water content under drought conditions when it was used to inoculate the seeds and soil. However, no correlation was detected between the EPS production in the media and soil water content of the tested bacterial strains (Fig. 6). In particular, EPS production by K. cowanii MS1 was higher than that of K. cowanii GG1; however, the water content of soil inoculated with K. cowanii MS1 was lower than that of the control. The EPS production of P. hunanensis GG17 was lower than that of K. cowanii GG1, whereas the water content of the soil inoculated with these two strains was similar.
Considering the diverse factors that affect bacterial EPS production, the absence of a correlation between bacterial EPS production in the liquid medium and the water content of bacteria-inoculated soil under drought conditions is not surprising. For instance, the amount and composition of EPS are highly dependent on the growth conditions 43,44 . The different conditions of the liquid media used for EPS quantification and the soil used for plant growth likely caused this discrepancy. In addition, bacterial growth in the rhizosphere is known to depend on the host species, likely influenced by the differential chemical compositions of root exudates 51,52 . As we applied the bacteria isolated from L. serriola to A. thaliana, the effects of these plants on the bacterial growth in the rhizosphere might be different. Lastly, K. cowanii strains from plant populations might be distinctive genotypes that might affect plant performance differentially.
A caveat of this study is that we used A. thaliana instead of host plant L. serriola although model plant species have often been used in similar studies 49,50,53 and K. cowanii GG1 was able to colonize the root of A. thaliana ( Supplementary Fig. S3). Assessing physiological changes of L. serriola in response to the endophyte treatment would be the next step for more complete understanding of invasive host plant-seed endophyte interactions. Transcriptome analysis of L. serriola plants and endophytic bacteria will be a useful approach for the physiological study. Metagenome study would provide information on the genetic mechanisms of the effects of endophytic bacteria on plant performances.
In conclusion, L. serriola, a xerophytic invasive plant species, possesses diverse seed endophytic bacteria. Notably, one isolate, K. cowanii GG1, increased EPS production in media with a highly negative water potential, increased soil water content, and promoted the growth of A. thaliana under drought conditions. These results imply that invasive plants can disperse along with their symbiotic bacteria, which may promote successful establishment in the introduced area.

Methods
Seed sources and endophyte isolation. We collected wild L. serriola seeds from four natural populations in the vicinity of the Gwangju Institute of Sciences and Technology, Gwangju, South Korea (Fig. 1, supplementary table S1). Permission to collect L. serriola was given by the Ministry of Environment, Korea. Experimental research on plants including the collection of plant material complied with relevant institutional and national guidelines and legislation. Seeds from 4-12 plants from each population were composited, and 80 seeds from each composite sample were used to isolate endophytic bacteria. To sterilize the seed surface, we removed the seed pappi and soaked the seeds in 70% ethanol for 1 min and 3% sodium hydrochloride for 3 min. The seeds were washed three times with sterile distilled water. To confirm surface sterilization, 100 μL of the rinsing water was spread on R2A (#218,263, Difco) and Luria-Bertani (LB) agar (#7279, Acumedia) and incubated at 25 °C for 7 days.
To extract bacterial genomic DNA, a single colony was inoculated into the LB broth and incubated in a shaking incubator at 25 °C and 180 rpm. DNA was extracted using the Exgene Cell SV kit (GeneAll, Seoul, South Korea) following the manufacturer's instructions. The 16S rDNA gene was amplified by polymerase chain reaction (PCR) using the universal primers 27F (GAG TTT GATCMTGG CTC AG) and 1492R (TAC GGY TAC CTT GTT ACG ACTT), as previously described 54  www.nature.com/scientificreports/ Laboratories, CA, USA) using the following program: initial denaturation, 95 ℃ for 3 min; denaturation, 95 ℃ for 30 s; annealing, 55 ℃ for 30 s; elongation, 72 ℃ for 1 min; and additional extension for 5 min at 72 ℃. The PCR products were purified using the QIAquick PCR Purification Kit (Qiagen, Venlo, the Netherlands). Sequencing was performed using the Sanger method with universal 16S rDNA 27F and 1492R primer (Macrogen Inc; Seoul, Korea). If required, additional 16S rDNA primers 518F (CCA GCA GCC GCG GTA ATA CG) and 800R (TAC CAG GGT ATC TAA TCC ) were used for sequencing. The nucleotide sequences were aligned using MEGA 7.0 software 30 and compared with previously reported sequences of bacterial type strains using EZBioCloud (Chunlab, Seoul, South Korea). A phylogenetic tree was constructed using the sequences isolated in the present study and those of the type strains with the highest sequence similarity, using MEGA 7.0. The distance matrix was calculated using Kimura's two-parameter method, and the phylogenetic tree was constructed using the neighbor-joining method 55,56 . The 16S rDNA sequences of the isolates have been deposited in the NCBI GenBank, and their accession numbers are provided in Table 1.
Bacterial phenotyping. Of the 129 identified bacterial isolates, 42 representative isolates were selected and their in vitro PGP traits were examined (Supplementary Fig. S1). A representative isolate was randomly selected from a pool of isolated strains in the same phylogenetic clade (Fig. 2). When isolated strains from different plant populations grouped into the same clade, one representative isolate for each population was selected. To examine their ability to solubilize inorganic phosphate, all isolates were inoculated on the National Botanical Research Screening for drought-tolerant bacteria. Before examining the drought tolerance of the isolated bacteria, we tested if the isolates produced capsular material outside the cells as a quick screening procedure of drought-tolerant bacteria 60 . All isolates cultured on the KB medium were stained using a negative capsule staining method with 10% nigrosine and 1% crystal violet 61 . We observed clear halo zones around all isolated bacterial cells, indicating capsular material production ( Supplementary Fig. S1).
To examine the drought tolerance of the 42 selected isolates, we measured their growth in trypticase soya broth (TSB; #211,825, Difco) with a water potential of -0.73 MPa following Sandhya et al. 28,62 . The water potentials were adjusted with 25% polyethylene glycol 6000 (#817,007, Merck), as described by Michel and Kaufmann 63 . After incubating the bacterial isolates in TSB overnight, 1% volume aliquots of the incubated cultures were transferred to TSB with -0.73 MPa and incubated in a shaking incubator at 28 °C and 180 rpm for 24 h. The growth of each bacterial isolate was estimated by measuring the optical density of the sample at 600 nm (OD 600 ) using BioSpectrometer basic (Eppendorf, Hamburg, Germany). The Escherichia coli DH5α strain was used as a control to confirm the experimental protocol of the drought tolerance test and the EPS quantification because this strain is known to produce colonic acid, a kind of EPS 64 . The OD 600 of triplicate samples was measured for each isolate.

Quantification of EPS in drought-tolerant bacteria.
Of the seed endophytes isolated from each plant population, we selected the two isolates with the highest OD 600 values at − 0.73 MPa and one isolate (Kosakonia cowanii) detected in all four plant populations for EPS quantification. A total of 12 bacterial isolates were incubated under non-stress (TSB) and stress (TSB at -0.73 MPa) conditions and EPS production was measured following Liu et al. 65 . After culturing the isolates at 28 °C with shaking at 180 rpm for 3 days, the bacterial cells were harvested by centrifugation at 9,000 rpm for 15 min. The bacterial pellet was dissolved in sterilized deionized water for the measurement of dry bacterial biomass. Three volumes of cold ethanol were added to the separated supernatant and the samples were incubated at 4 °C overnight to facilitate EPS precipitation. After centrifugation at 15,000 rpm for 30 min at 4 °C, the precipitated EPS were washed twice with cold absolute ethanol and dissolved in heated deionized water. The concentration of the completely dissolved EPS was measured by the phenol-sulfuric acid method, using D-glucose as a standard 66 .  46 . The same amount of Hoagland's solution without bacteria was used as a negative control. Ten plants were randomly assigned to each bacterial treatment. The pots were randomly positioned in a growth chamber (Hanbaek, Co. LTD., Kyunggi-do, South Korea) and maintained at 22 °C under a 16/8 h light/ dark photoperiod throughout the experiment. Sterilized deionized water was applied to each pot once or twice per week, and 20 mL of Hoagland's solution was applied to each pot once per week during the growth 46 . Half of the plants were subjected to drought conditions 5 weeks after sowing by discontinuing their watering.

Effects of isolates on the growth of
After 10 days of drought treatment, the plants were harvested and shoot fresh weight was measured immediately. The roots were dried in a drying oven (Hanbaek Co. LTD., Kyunggi-do, South Korea) at 70 °C for 72 h and their dry weight was measured. The leaves were frozen in liquid nitrogen and stored at -80 °C to measure their malondialdehyde (MDA) content, following Zhang and Huang 68 . In response to drought stress, MDA accumulates in plant leaves as a byproduct of oxidative damage to membrane lipids. The leaves were homogenized in 0.1% (w/v) trichloroacetic acid (TCA; T6399, Sigma-Aldrich), and the mixture was reacted with 20% TCA containing 0.5% thiobarbituric acid (TBA; T5500, Sigma-Aldrich) by boiling at 95 °C for 15 min. The absorbance of the resulting mixture was measured at 532 nm using BioSpectrometer basic (Eppendorf). The relative water content of the plants was determined following the methods of Türkan et al. 69 . Leaf fresh weight was measured immediately after harvesting. Leaf turgid weight was measured after submersion in sterilized deionized water for 24 h at 4 °C in a refrigerator, and the dry weight was measured after drying at 70 °C for 72 h. The gravimetric soil water content at the time of harvest was measured after drying the soil at 105 ℃ for 72 h.
Statistical analysis. All statistical analyses were performed using R software 4.0.1 (R Foundation for Statistical Computing, Austria). To compare the EPS production of the isolated bacteria, a two-way analysis of variance (ANOVA) was performed, with bacterial isolates, stress conditions, and their interaction as independent variables and EPS production as the dependent variable. Pair-wise differences between the stress conditions of each isolate were evaluated based on the Bonferroni method. To assess the effect of bacterial isolates on plant performance and soil water content, a two-way ANOVA was used, with the drought treatment, isolates, and their interaction as independent variables, and plant traits and soil water content as the dependent variables. Soil water and MDA content was log-transformed to normalize the data, and outliers for all measured traits were excluded from the dataset. As bacterial effects on soil moisture depend on soil water conditions (see results), the control and drought treatments were further assessed separately using ANOVA. Differences between the effects of the negative control and those of the isolated bacteria were evaluated based on Dunnett's adjustment, under control and drought conditions. To determine the relationship between the EPS production of isolates, soil water content, and plant traits, Pearson's correlation coefficients were calculated.