The ecological role of bacterial seed endophytes associated with wild cabbage in the United Kingdom

Abstract Endophytic bacteria are known for their ability in promoting plant growth and defense against biotic and abiotic stress. However, very little is known about the microbial endophytes living in the spermosphere. Here, we isolated bacteria from the seeds of five different populations of wild cabbage (Brassica oleracea L) that grow within 15 km of each other along the Dorset coast in the UK. The seeds of each plant population contained a unique microbiome. Sequencing of the 16S rRNA genes revealed that these bacteria belong to three different phyla (Actinobacteria, Firmicutes, and Proteobacteria). Isolated endophytic bacteria were grown in monocultures or mixtures and the effects of bacterial volatile organic compounds (VOCs) on the growth and development on B. oleracea and on resistance against a insect herbivore was evaluated. Our results reveal that the VOCs emitted by the endophytic bacteria had a profound effect on plant development but only a minor effect on resistance against an herbivore of B. oleracea. Plants exposed to bacterial VOCs showed faster seed germination and seedling development. Furthermore, seed endophytic bacteria exhibited activity via volatiles against the plant pathogen F. culmorum. Hence, our results illustrate the ecological importance of the bacterial seed microbiome for host plant health and development.

Moreover, interspecific interactions of phylogenetically different bacteria can also alter the volatile blend composition, affecting the activity of volatiles Tyc et al., 2015). The effects of the emitted microbial VOCs on the host plants and their antagonists can vary from negative, positive to neutral (van Dam, Weinhold, & Garbeva, 2016). For instance, plantgrowth promoting effects were reported for volatiles emitted by bacteria (Park, Dutta, Ann, Raaijmakers, & Park, 2015;Ryu et al., 2003) and fungi (Cordovez et al., 2017). In addition, volatiles from an endophyte of maize (Zea mays), Enterobacter aerogenes have been shown to alter the host plant's resistance to a fungal pathogen and an insect pest (D'Alessandro et al., 2014), suggesting that volatiles also exhibit plant protection against a broad range of attackers. Interestingly, volatiles emitted by the nectar-inhabiting yeast Metschnikowia reukaufii influenced the nectar preference of a generalist bee (Rering, Beck, Hall, McCartney, & Vannette, 2018). However, it is unknown so far whether volatiles emitted by seed endophytes in particular benefit the associated host plant and whether interspecific interactions between endophytes change volatile emission with consequences for the host in terms of growth, development, and resistance.
Here, we aimed to investigate the potential role of volatiles produced by seed endophytic bacteria associated with wild cabbage (Brassica oleracea L.) on plant growth, development and resistance against a leaf chewing insect herbivore and two pathogenic fungi. These wild cabbage populations are considered to be the ancestors of current cultivated cabbage. Seeds originated from five populations growing along the rugged coastline of Dorset, United Kingdom (Gols, Dam, Raaijmakers, Dicke, & Harvey, 2009;Van Geem, Harvey, Cortesero, Raaijmakers, & Gols, 2015;Wichmann, Alexander, Hails, & Bullock, 2008). Previous work has shown that there is considerable population-related variation in the expression of primary and secondary metabolites (glucosinolates) in British populations of wild cabbage. These differences have an effect on the behavior and development of several species of insect herbivores and their natural enemies associated with these plants both in the laboratory and in the field (Gols, Bullock, Dicke, Bukovinszky, & Harvey, 2011;Gols et al., 2008;Harvey, Dam, Raaijmakers, Bullock, & Gols, 2011;Moyes, Collin, Britton, & Raybould, 2000;Newton, Bullock, & Hodgson, 2009;Van Geem et al., 2015). However, this previous research ignored the possibly important role played by the plant microbiome on plant traits that affect growth, fitness, and defense. We hypothesize that seeds of wild cabbage contain cultivable endophytic bacteria whose volatiles are beneficial for the host plant. Here, we aim to isolate endophytic bacteria from five different populations of wild cabbage plant populations. We hypothesize that the five different plant populations harbor different endophytic bacterial strains, each producing its specific volatile blend, which in turn differentially affect their interaction with the host plant. Seeds were surface-sterilized by a modified protocol by Araujo et al. (2002). To this end, seeds (1 g) of each plant population were subsequently incubated for 3 min in 2% NaOCl, 3 min in 80% ethanol, and rinsed five times with sterile distilled water. The sterilized seeds were transferred to a sterile mortar with 1 ml of 10 mM phosphate buffer (pH 6.5) and crushed using a sterile pestle. A volume of 100 µl was taken and transferred to 900 µl of 10 mM phosphate buffer. A serial dilution was made from this solution, and each dilution was plated in triplicates on 1/10th TSBA plates (5.0 g/L NaCl, 1.0 g/L KH 2 PO 4 ; 3 g/L Oxoid Tryptic Soy Broth; and 20 g/L BACTO agar, pH 6.5) (Tyc et al., 2015). Plates were incubated for one week at 24°C and examined regularly for visible bacterial growth.

| Enumeration of bacterial colony-forming units (CFU) and preparation of glycerol stocks
For the enumeration of colony-forming units (CFU) of the isolated endophytic bacteria an aCOlyte Colony Counter (Don Whitley Scientific, Meintrup DWS Laborgeräte GmbH, Germany) was used.
After one week of incubation, the CFUs of each petri dish containing the bacteria were enumerated. The CFU numbers were based on three replicates per dilution series per plant population. Single bacterial colonies were picked from plates and transferred to 10 ml liquid 1/10th Tryptic Soy Broth (TSB) (5.0 g/L NaCl, 1.0 g/L KH 2 PO 4 ; and 3 g/L Oxoid Tryptic Soy Broth) and incubated overnight at 24°C, 190 rpm. The next day a volume of 750 µl culture was mixed with 750 µl 50% (v/v) glycerol. Prepared glycerol stocks were transferred to a −80°C freezer for long time storage. for 45 s, 72°C for 1 min, and a final round of amplification at 72°C for 5 min. After amplification, a volume of 5 µl of each PCR reaction was loaded on a 1% (w/v) agarose gel and checked after electrophoresis for presence of correct-sized PCR fragments. Positive PCR products were cleaned using the Qiagen PCR purification kit (Cat# 28,104; Qiagen Benelux BV, Venlo, the Netherlands) and sent to Macrogen (Macrogen Europe, Amsterdam, the Netherlands) for sequencing. The obtained sequences of the 16S rRNA gene were examined for quality and trimmed to approximately the same size (~700 bp) using BioEdit 7.2.5 (Hall, 1999 33-8. The Alignments and the tree were generated with ClustalW and bootstrap analysis was performed with 10,000 resamplings.

| Taxonomic identification of endophytic bacteria by 16S rRNA PCR
Phylogenetic tree images were created by using the phylogeny.fr platform (www.phylo geny.fr) (Dereeper et al., 2008) using standard settings. The sequences obtained during this study are submitted to NCBI GenBank under submission number SUB5675460 and the accession numbers MN079062 -MN079072 (Table 1).

F I G U R E 1
The five different wild cabbage plant (B. oleracea) populations grown at their natural location, their seeds, and their isolated cultivable microbiome. (A) Overview of the five different used plant populations, seeds, and the isolated microbiome from 1 gram of seeds. (B) Number of bacterial colony-forming units (CFU) obtained from 1 gram of surface-sterilized seeds of each plant population. Bars represent standard deviation (SD). No significant differences in CFU/g seed were observed among the seeds of the five plant populations (ONE-WAY ANOVA post hoc Tukey tests). The same letter above the bars indicates no significant difference between the samples with p > .05  Figure 1). The seeds were surface-sterilized as described above, dried on filter papers in a flow cabinet for 15 min, and stratified for 3 days at 4°C. An overnight inoculum of each bacterial isolate (Table 1) was prepared. For this, a single colony of each bacterial isolate was picked from plate and grown in 20 ml 1/10th Tryptic Soy Broth at 190 rpm and 20°C. Each bacterial inoculum was diluted to an OD 600 of 0.005 (monoculture or mixtures) in 20 ml 10 mM phosphate buffer. Fifty µl was plated on 1/10th Tryptic Soy Agar (pH = 6.5) in a two-compartment Petri dish (9 cm diameter; Greiner bio-one B.V., Alphen a/d Rijn, the Netherlands, Cat# 635,102) and incubated at 20°C for 48 hr. After three days of stratification, 8 seeds of each plant population were placed on 0.8% plant agar medium (P1001 Duchefa Biochemie, pH = 5.8) opposite the inoculated bacterial isolates in the two-compartment Petri dish. Plates containing the bacteria and seeds were incubated for a week. For the control, seeds were placed on one side of the two-compartment Petri dish without bacterial inoculum being added to the growth medium. All Petri dishes were sealed with Parafilm and stored in climatic chamber (20°C; 180 µ mol light/m 2 /s at plant level; 16:8 hr (light: dark); 60%-70% R.H.). Images were captured starting from the 3rd day to the 7th day to record radicle emergence, primary root length. For the estimation of the seedling fresh weight on days three, five and seven the seedlings were weighed on a microbalance (Mettler-Toledo MT5 Electrobalance). Primary root length of seedlings (cm) was analyzed using SmartRoot plugin in Fiji, image analysis software (Schindelin et al., 2012). Three technical replicates were prepared.  prepared and seeds were treated as described above. In total, 16

TA B L E 1 Organisms used in this study
Petri dishes (4 per treatment (3) and the control) were prepared and incubated for one week. For the control, seeds were placed on one side of the two-compartment Petri dish without added bacterial inoculum (n = 4). All Petri dishes were sealed with Parafilm and stored in climatic chamber (20°C; 300 µ mol light m -2 s -1 at plant level; 16:8 hr (light:dark)) for seed germination and pregrowth of the plants. A total of 64 one-week-old seedlings that were either exposed or not exposed to bacterial volatiles were transferred to

| Effects of bacterial volatiles on fungal growth (mycelial expansion)
To test the effect of the emitted bacterial volatiles on fungal hyphal extension, the two plant pathogenic model fungi, Rhizoctonia solani (AGII) 2.2IIIB (Garbeva, Silby, Raaijmakers, Levy, & Boer, 2011) and Fusarium culmorum were used (de Rooij-van der Goes, 1995). The fungi were precultured on 1/5th Potato Dextrose Agar (PDA) (29 g/L Oxoid CM 139) (Fiddaman & Rossall, 1993) and incubated at 24°C for 7 days prior to the experiment. The assays were performed in Petri dishes (9 cm diameter, Greiner bio-one B.V., Alphen a/d Rijn, the In the lid of the Petri dish, 12.5 ml of water-agar medium (WA) (20 gL -1 BACTO agar) was added and inoculated in the middle with a 6-mm-diameter PDA agar plug containing R.solani or F.culmorum hyphae. The plates were sealed with Parafilm and incubated at 24°C for five days.
This allowed us to test fungal exposure to the volatiles produced by the bacteria grown in the bottom compartment without the fungi being in direct physical contact with the bacteria. On the fifth day, the extension of the hyphae was measured and compared to the hyphae extension in the control plates (fungi exposed to 1/10th TSBA growth medium without bacteria). For the analysis, digital photographs were taken. The digital images were analyzed using the AXIO VISION v4.8 imaging Software (Carl Zeiss Imaging Solutions GmbH).

| Effects of bacterial volatile exposure on plant herbivory resistance
We also tested the effect of volatiles produced by P. marginalis, P.

| S TATIS TIC AL ANALYS IS
The effect of bacterial volatiles on plant growth and development

| Effects of bacterial volatiles on seed germination, primary root length and plant biomass
Volatile exposure treatments significantly affected seed germination (Chi-Square = 38.94; df = 3; p < .001; Figure 3). Exposure to volatiles of all bacterial monocultures promoted seed germination of all plant populations but this was only significant for the seeds exposed to volatiles of the monocultures of P. agglomerans and P. azotoformans, as well as to the mixture of P. marginalis and P. azotoformans (Figure 3).
Seed germination was faster and more seeds were germinated when F I G U R E 3 Germination (proportional) of wild cabbage (B. oleracea) seeds on the 5th day following continuous exposure to bacterial volatiles emitted by (A) P. marginalis, P. azotoformans, and the combination of both compared with the control (B. oleracea without exposure to bacterial volatiles). (B) when exposed to bacterial volatiles emitted by monocultures of S. rhizophila, P. orientalis and P. agglomerans or control (no bacterial volatile exposure) for five days. Significant differences between the treatments and the control are indicated by different letters above bars based on ONE-WAY ANOVA, post hoc Tukey multiple comparison tests (n = 8) exposed to volatiles of P. agglomerans monocultures in comparison to seed germination in the controls (Figure 3b). When comparing exposure to volatiles produced by a single bacterium species, only volatiles emitted by P. agglomerans also strongly promoted primary root length (Figure 4a, b) and seedling fresh biomass (Figures 3a, b, 4a, b) compared with the root length and seedling biomass of the controls and of seeds exposed to volatiles emitted by the other monocultures

| Effects of bacterial volatiles on the growth of two plant pathogenic model fungi
Volatiles produced by Pseudomonas azotoformans D1 were strongly inhibiting (p = .015) the growth of the plant pathogenic fungus F I G U R E 4 Primary root length (mean ± SE) of all wild cabbage (B. oleracea) population seedlings when exposed for five days to bacterial volatiles emitted by (A) P. marginalis, P. azotoformans, and the combination of both compared with the control (B. oleracea without exposure to bacterial volatiles).
(B) when exposed to bacterial volatiles emitted by monocultures of S. rhizophila, P. orientalis, and P. agglomerans or control (no bacterial volatile exposure) for five days. Different letters above bars are based on Tukey HSD multiple comparison tests in general linear model (n = 15) and indicate significant differences between the treatments and the control and Pantoea agglomerans E44 (p < .001) were able to strongly inhibit the growth of the plant pathogenic fungus Fusarium culmorum ( Figure 6b).

| Effects of volatiles emitted by monocultures and mixtures of P. marginalis and P.azotoformans on plant herbivory resistance and larval performance and survival
Plants from the Winspit population exposed to bacterial volatiles did

| Detected headspace volatile compounds and effect of interspecific interactions on bacterial volatile blend composition
GC/MS-Q-TOF analysis revealed a total number of 9 volatile organic compounds that were not detected in the noninoculated controls (Table 2). The 9 detected compounds belonged to different chemical classes including acids, alcohols, alkenes, terpenes, and sulfides.
Each bacterium emitted its specific blend of compounds and the emitted individual volatiles compounds differed between each bacterial inoculum (Table 2, Figure 7a). The PLSDA analysis could clearly separate the blends. Clear separations between controls, monocultures, and the combination of P. marginalis with P. azotoformans were obtained in PLSDA score plots (Figure 7a, b). The volatile composition of the blend emitted by the bacterial mixture resembled that of the blends emitted by the monocultures of these bacteria. Three compounds, cyclohexane, dimethyl disulfide, dimethyl trisulfide were emitted by all bacterial inocula. We could tentatively identify 7 compounds emitted by monocultures of P. agglomerans E44, 6 for P. marginalis B1, 7 for P. azotoformans D1, 6 for S. rhizophila D5, and 4 for P. orientalis E8. For the combinations of P. marginalis with P.
azotoformans, we obtained a total number of 7 volatile organic compounds. The most prominent detected headspace volatile organic compounds were the two sulfur-containing compounds dimethyl disulfide (C 2 H 6 S 2 ) and dimethyl trisulfide (C 2 H 6 S 3 ) that were produced by all tested bacteria (Table 2). Interestingly, 1-undecene and the unknown compound produced by the monoculture of P. marginalis were not detected in the blend produced by the bacterial mixture (Table 2).

| D ISCUSS I ON
Seeds and plant seedlings are clearly a crucial stage of a plant's development: failure to germinate is lethal. However, thus far, little is known about seed-associated microorganisms and their impacts on plant growth and development (Nelson, 2018). Furthermore, there is not much knowledge about the metabolites produced by the microorganisms that reside inside seeds and their effect on plant TA B L E 2 Tentatively identified volatile organic compounds (VOCs) produced by endophytic bacteria isolated from seeds of B. oleracea RT*, retention time, the RT value stated is the average retention time of three replicates.
ELRI**, experimental linear retention index value, the RI value stated is the calculated average of three replicates.
p-value***, statistical significance (peak area and peak intensity). development, growth, and health. Despite increasing awareness of the importance of the plant holobiont to plant evolution and ecology, the importance of the seed microbiome has generally been neglected (Berg & Raaijmakers, 2018;Hacquard, 2016;Rosenberg & Zilber-Rosenberg, 2016). This is one of the few studies investigating the beneficial effects of seed-associated bacteria and the metabo- investigating the abundance of endophytic bacteria in plant tissues (Compant, Mitter, Colli-Mull, Gangl, & Sessitsch, 2011;Ferreira et al., 2008;Graner, Persson, Meijer, & Alstrom, 2003;Rosenblueth et al., 2012;Truyens et al., 2015). Many of the bacteria isolated from the seeds of B. oleracea belonged to the genera Chyrseobacterium, Stenotrophomonas, Sphingomonas, Pseudomonas and Pantoea, which are known bacterial endophytes of many plant species (Graner et al., 2003;Nelson, 2004;Truyens et al., 2015). However, our study focused on culturable bacteria and, therefore, only a subset of the total seed-associated microbiome was assessed. Further metagenome-based studies need to be performed to detect the other nonculturable microorganisms associated with plant seeds. of these species are well-known for their plant-growth promoting effects (Park et al., 2015;Raza, Yousaf, & Rajer, 2016;Santoyo et al., 2012). The mechanism underling growth-promoting effects of bacterial volatiles are largely unknown. It has been proposed that bacterial volatiles may modulate phytohormonal networks in the host plants, such as those involving ethylene (Ryu et al., 2003), cytokinin (Ortiz-Castro, Valencia-Cantero, & Lopez-Bucio, 2008), ABA (Zhang et al., 2008) or auxin (Bailly et al., 2014). However, the target tissues of bacterial volatiles and how these are recognized and activate plant signaling are still being investigated (Bailly & Weisskopf, 2012;Sharifi & Ryu, 2018).
In general, whereas all five of the cabbage populations performed better when exposed to bacterial volatiles, three of the populations Old Harry (plant population C) population showed better seed germination and seedlings produced longer primary roots (data not shown).

Furthermore, The St. Aldhelms Head (plant population D) and the Old
Harry population yielded higher seedling fresh biomass compared with the three other populations. Overall, the Old Harry population showed the best plant performance regardless of the bacterial volatile blend it had been exposed to. These results suggest that seeds of the various cabbage populations differ in their responsiveness to growth promotion by bacterial volatiles. Interestingly, we could only isolate one bacterial species that could be cultured from the Old Harry plant population. However, this might be due to the applied culture-dependent approach and, most probably, only subsets of the total seed microbiome of each plant population has been assessed.
Furthermore, this study investigated how the exposure to volatiles emitted by the endophytic bacteria influenced the resistance of wild B. oleracea plants to M. brassicae larvae. The highest mortality was found when M. brassicae larvae were exposed to volatiles emitted by the bacterial mixture of P. marginalis and P. azotoformans after three days and by the monocultures of P. marginalis after seven days. These results suggest that plants exposed to bacterial volatiles has only marginal and transient effects on the larval performance. It is also possible that the larvae are less affected by increased plant resistance as their development advances (Jeschke et al., 2017).
Previous work demonstrated that exposure of Arabidopsis to volatiles emitted by Bacillus amyloliquefaciens GB03 transcriptionally induced sulfate assimilation, and this resulted in increased total shoot glucosinolates and reduced larval performance of Spodoptera exigua (Aziz et al., 2016). In the study (Aziz et al., 2016), larval performance was also determined when caterpillars had been feeding on the plants for 7 and 9 days, respectively, and did not cover complete immature development (larvae may compensate for initial reduced feeding later in their development).
The five British wild cabbage populations studied here grow along a linear transect along the often rugged chalky coastline of Dorset and geographic formations known as the "Purbeck Hills".
These populations are discrete and apparently have been stable for many decades and perhaps centuries (Wichmann et al., 2008).
Previous studies have shown that concentrations and types of secondary metabolites in them (glucosinolates) differ markedly among the different populations, even those growing within a few km of each other (Gols et al., 2009(Gols et al., , 2008Moyes et al., 2000). This suggests that there may be little gene flow between them (Wichmann et al., 2008). The five populations also exhibit varying degrees of exposure to prevailing winds from the south to west, which are often persistent and reach gale force in the more exposed locations (e.g. St. Aldhelms Head and Kimmeridge). Moreover, some of the plant populations are not that large: Old Harry, for instance, contains ~ 50-100 plants, many of them at least several years old (Mitchell & Richards, 1979). The vegetation has been classified as maritime grassland and the floral diversity largely depends on the degree of exposure to harsh conditions (Mitchell & Richards, 1979;Wichmann et al., 2008).
The species is considered a poor competitor and seedlings are easily shaded out by grasses in spring such as Festuca rubra and Lolium perenne (Mitchell & Richards, 1979). Therefore, the presence of endophytic bacteria on seeds may play a crucial role in enabling wild cabbage to persist in the face of intense competition with grasses for germination sites.
This is the first report showing how wild cabbage populations respond toward bacterial volatiles coming from their own seed microbiome. Our study clearly shows that seeds endophytes may play an important role in early development of the plant (seed germination and seedling growth). This study indicates the importance to further explore the seed-associated microbiome and the interactions within the seed microbiome and between the seed microbiome and the host plant. Further studies should combine both metagenomics and culturable approaches in order to comprehensively understand the underlying mechanism of positive impacts of the seed microbiome on plant growth, development, and resistance in wild cabbage plants.

ACK N OWLED G M ENTS
This work was financially supported by The Netherlands December 2019 after first online publication: the Acknowledgment section has been updated on this version].

CO N FLI C T O F I NTE R E S T S
None declared.