Drought Tolerant Wheat Varieties Enrich Seed Microbiomes For Benecial Microbes Under Drought Conditions

Climate change is predicted to increase the incidence and severity of drought conditions, posing a signicant challenge for agriculture globally. Plant microbiomes have been demonstrated to aid crop species mitigate drought stress. In this study the wheat seed microbiomes from lines with contrasting drought tolerances were interrogated for microbes that alleviate drought stress. Metagenomic and culture-based methods were used to prole and characterise the seed microbiome composition of four drought tolerant and three drought susceptible wheat lines under rainfed and drought conditions. Curtobacterium accumfaciens (Cf D3-25) and Arthrobacter sp. (Ar sp. D4-14) were isolates enriched in drought tolerant lines under drought conditions and demonstrated the ability to promote wheat growth under drought conditions. Members of Triticeae inoculated with Cf D3-25 and Ar sp. D4-14 showed a biostimulation affect. This study indicates seed microbiomes from genetically distinct wheat lines enrich for benecial bacteria in ways that are both line-specic and responsive to environmental stress. As such, seed from stress-phenotyped lines represents an invaluable resource for the identication of benecial microbes with plant growth promoting activity that could improve commercial crop production, particularly under drought stress. representative isolate of two cultured genera that demonstrated signicant differences between the contrasting genotypes of wheat seed microbiomes were selected and assessed for a biostimulation effect in Triticeae. Bacterial strains Cf D3-25 and Ar. sp D4-14 were cultured in Lysogeny Broth (Oxoid, Australia) overnight at 26°C. After 24 hours (h), seeds of Triticum aestivum (cv Bob White Red Haplotype) were sterilised by soaking in 80% ethanol for 3 minutes (min), then washed ve times in sterile distilled water. To determine the CFU ml -1 , plate dilutions combined with OD (optical density) readings were performed. The cultures were centrifuged and washed in PBS twice before being resuspended in their original volume of overnight culture. These cultures were then diluted step-wise with PBS to concentrations of 10 -1 , 10 -2 , 10 -3 , 10 -4 and 10 -5 respectively. Seeds were soaked in undiluted, or diluted solutions for four h at 26°C in a shaking incubator. As a control, seeds were soaked in PBS without bacteria under the same conditions. Fifteen inoculated seeds were then placed on moist sterile lter paper in sterile petri plates and allowed to grow for seven days at room temperature. There were four replicates per treatment. The same assay was performed using ryecorn, barley (cv Hindmarsh), oat, and spelt using


Introduction
Drought poses signi cant challenges to agricultural production in arid and semi-arid regions of the globe.
Increases in mean global temperatures and decreased frequency and intensity of precipitation, as a result of climate change, are extending periods of drought and decreasing crop yields, highlighting the increasing need for abiotic stress tolerance in crop species [2]. Triticum aestivum (wheat) is grown on approximately 220 million hectares of land globally, producing 20% of the world's caloric requirements [3]. In Australia, drought conditions have had severe impacts on wheat production, reducing the winter harvest to 35% below the ten year average in 2019 [4,5]. Research into the mechanisms that increase drought tolerance in crop plants focused primarily on improving crop genetics through techniques such as QTL mapping, marker assisted breeding and introgression from wild species [6][7][8][9]. The identi cation of drought tolerance with high heritability is confounded by complex genotype by environment interactions [7][8][9][10]. Envirotyping is a comprehensive method of measuring environmental factors with the capacity to affect phenotypic variation in plant growth designed to aid genotype by environment (GEI) and phenotype prediction [11]. Analysis of arid and drought envirotypes has identi ed microbiome composition as an environmental factor that affects drought tolerance in crop plants [11,12]. As a result, there is increasing interest in interrogating the plant microbiome with the goal of identifying microbes that augment the phenotype conferred by the genome of the host plant and hence confer drought tolerance [13][14][15][16][17].
Plants form intimate relationships with microbes that colonise their tissues and organs, forming a single ecological unit known as a holobiont [18]. These microbiomes have been demonstrated to in uence a myriad of plant traits such as germination, biomass accumulation, pathogen resistance and abiotic stress tolerance [19]. Seeds harbor the initial inoculum of microbes, playing a vital role in the transmission of microbial resources between plant generations [20]. As the initial colonisers of plant tissue, the seed microbiome is believed to shape the overall composition of the plant microbiome and therefore have a competitive advantage over microbes recruited from soil and root [21,22]. A comparatively limited number of bacterial genera have been detected in seed microbiomes compared to other plant microbiomes, such as root microbiome [23]. It is postulated that the seed microbiome evolves as a result of host selection for traits that complement the plant genome [18,23]. While root microbiomes have been the focus of signi cant study, the seed microbiome has largely been overlooked [24].
Studies into the in uence of drought on plant microbiomes have focused almost exclusively on root microbiomes [12,16,25,26]. Several gram-positive rhizosphere bacteria have been demonstrated to be effective in improving drought stress tolerance in crop plants [27][28][29]. Timmusk et al, demonstrated that wheat treated with Paenibacillus and Bacillus species isolated from arid environments increased plant biomass by up to 78% and increased the plant survival rate ve-fold under severe drought conditions [30].
Notably, Paenibacillus and Bacillus species isolated from moderate drought environments did not increase drought tolerance in wheat [30,31]. The consensus amongst many plant microbiome studies is that drought conditions have a signi cant impact on the composition and diversity of the microbiome, often leading to a marked increase in Actinobacteria in root microbiomes [12,16,26]. It is clear from these studies that recruitment and enrichment of certain bacterial elements in root microbiomes were driven by host-speci c metabolic and phenotypic factors [12,26]. Further to this, microbiomes have also shown cultivar-speci c enrichment in root microbiomes of rapeseed, cannabis and wheat [14,17,32]. We hypothesise that seed microbiomes from plants exposed to abiotic stress enrich for bacteria that are more tolerant of abiotic stress, and that the composition of seed microbiomes differ between drought susceptible and drought tolerant wheat lines.
In this study, we present the rst analysis of seed microbiomes from contrasting drought tolerant and drought susceptible wheat lines under drought and rainfed conditions, with the aim of identifying microbes that are capable of increasing drought tolerance in a host plant. By interrogating the seed microbiome in silico, endophytes can be identi ed and isolated that have the potential to increase the drought tolerance of crop species at a commercial level.

Seed source
Seeds from eleven lines of Triticum aestivum were sourced from the Grains Innovation Park (DJPR) in Horsham, Victoria. These seeds were harvested in 2017 from a eld trial where 43 wheat lines were examined for drought response. Each line was subjected to rainfed and drought conditions, simulated using rainout shelters. The drought susceptibility index (DSI) was calculated for each line, using yield differences under drought and rainfed conditions. The lines were subsequently classed as either drought tolerant, or drought susceptible.
Culturing the seed microbiome Microbial isolation For microbial isolation, a subset of six lines was chosen from the eleven lines that had been pro led. Bacteria were isolated from the seeds of four drought tolerant wheat lines (lines 1, 2, 3 and 4) and two drought susceptible wheat lines (lines 10 and 11). Ten seeds from each treatment were plated onto stacked pieces of sterile lter paper soaked in Nystatin (50mg L -1 ). These seeds were germinated in the dark for two days, then grown for a further four days under light conditions. The seedlings were harvested and seed husks discarded. Plant tissues from the pooled seedlings were then immersed in phosphate buffered saline (PBS) and ground using a QIAGEN Mixermill for up to one minute at 30 Hertz. A 10µL aliquot of the resulting macerate was added to 90µl of PBS. Further 1:10 dilutions were performed to generate 10 -3 and 10 -4 solutions. Reasoners 2 Agar (R2A, Oxoid, Australia) agar plates were then inoculated with 10 -3 and 10 -4 solutions from each treatment and allowed to grow for 24-48 hours. Individual bacterial colonies were then streaked onto single R2A plates to isolate and purify single bacterial strains. A total of 438 bacterial strains were isolated from the four wheat lines and stored in 20% glycerol at -80 °C.

Microbe identi cation using MALDI-TOF
The microbial isolates were putatively identi ed using the Bruker MALDI Biotyper system. The bacterial strains were taken from the -80 °C glycerol stock, plated onto R2A and grown for 48 h. Single colonies were taken from the isolate plates and prepared for analysis using the manufacturer's Extended Direct Transfer method. The plate was analysed using the Bruker MALDI-TOF ultra eXtreme in accordance with the manufacturer's instructions. Escherichia coli strain ATCC 25922 was used as a quality control and as an internal standard. The resulting protein spectra were processed using the MALDI BioTyper automation 2.0 software at default settings. The microbes were then assigned a preliminary identi cation by comparing raw protein spectra against known spectra in the MALDI BioTyper library, which contained 2,750 species from 471 genera as of January 2019. The Biotyper library was supplemented with an inhouse database generated from previous endophyte studies of this research group (D. Auer, pers comm.) The protein spectra were processed using an in-house Re ner pipeline (GeneData 13.5). A hierarchical clustering algorithm in Analyst (GeneData 13.5) was used to create a phenogram that grouped bacterial strains based on the similarity of protein pro les.
Microbiome pro ling 16S rRNA library preparation and sequencing The microbiomes of the four drought tolerant lines (line 1, 2, 3, 4) and three drought susceptible lines (line 9, 10, 11) were pro led. Ten seeds from each line were plated onto lter paper soaked with sterile water and germinated in the dark at room temperature for two days. The seedlings were grown for a further four days under light conditions. Seedlings of approximately equal size were harvested and the seed husks discarded. Ten replicates, consisting of the plant tissues from ve pooled seedlings, were used for each line. Each replicate was snap frozen in liquid nitrogen. DNA was then extracted from each sample. Minor modi cations were made to the QIAGEN MagAttract 96 DNA Plant Core Kit during DNA extraction to allow for use of the Biomek FX liquid handling station. The V4 hyper variable region of the 16S rRNA gene was targeted for microbiome pro ling using the Illumina 16S Metagenomic Sequencing Library Preparation protocol (Methods S2) in conjunction with PNA PCR blockers to reduce ampli cation from plant organelles [33]. Paired end sequencing was performed on an Illumina HiSeq3000 Ò using the 2 x 150 bp v3 chemistry cartridge.

Bioinformatics
Microbiome bioinformatics were performed with a combination of Pandaseq and QIIME2 2019.10. The paired-end Illumina reads were trimmed and assembled using PANDAseq [34]. Using QIIME2 (v 2019.10), the 16S sequences were ltered and denoised using Deblur (via q2-deblur) [35,36]. There was a total of 18,220,959 quality-ltered reads. There were 182 total samples with a minimum of 489 and a maximum of 695,151 reads per sample. Taxonomy was assigned to ASVs using SILVA132 99% v4 (via q2-featureclassi er) [37,38]. All amplicon sequence variants (ASVs) were aligned using MAFFT (via q2-alignment) and a phylogeny was created using fasttree (via q2-phylogeny) [39,40]. Alpha-diversity metrics such as observed features and Faith's phylogenetic diversity and beta-diversity metrics such as RPCA [41,42] The rarefaction curve reached an asymptote, indicating that sampling had captured the diversity of the seed microbiomes. In order to capture the diversity of the microbiomes, a sampling depth of 15,000 reads was applied across samples. Analysis and visualisation were performed using QIIME2 pipelines. The QIIME2 feature table was exported using in biom formatted and ANOVA analysis performed on individual ASVs [43].
Plant growth promotion effects of the cultured microbiome Evaluation of microbiome bacteria for biostimulation of Triticeae A representative isolate of two cultured genera that demonstrated signi cant differences between the contrasting genotypes of wheat seed microbiomes were selected and assessed for a biostimulation effect in Triticeae. Bacterial strains Cf D3-25 and Ar. sp D4-14 were cultured in Lysogeny Broth (Oxoid, Australia) overnight at 26°C. After 24 hours (h), seeds of Triticum aestivum (cv Bob White Red Haplotype) were sterilised by soaking in 80% ethanol for 3 minutes (min), then washed ve times in sterile distilled water. To determine the CFU ml -1 , plate dilutions combined with OD (optical density) readings were performed. The cultures were centrifuged and washed in PBS twice before being resuspended in their original volume of overnight culture. These cultures were then diluted step-wise with PBS to concentrations of 10 -1 , 10 -2 , 10 -3 , 10 -4 and 10 -5 respectively. Seeds were soaked in undiluted, or diluted solutions for four h at 26°C in a shaking incubator. As a control, seeds were soaked in PBS without bacteria under the same conditions. Fifteen inoculated seeds were then placed on moist sterile lter paper in sterile petri plates and allowed to grow for seven days at room temperature. There were four replicates per treatment. The same assay was performed using ryecorn, barley (cv Hindmarsh), oat, and spelt using the methods described above. Data was statistically analysed using a one-way ANOVA and Tukey test to determine any signi cant difference (p ≤ 0.05) between treatments using OriginPro 2018 (v b9.5.1.195).

Greenhouse drought pot trial
The wheat seeds (cv Bob White Red Haplotype) were sterilised according to the in vitro method described in section 'Evaluation of microbiome bacteria for biostimulation of Triticeae'. The seeds were then soaked in overnight cultures of either Cf D3-25, or Ar. Sp for four hours at 26°C in a shaking incubator. For control seedlings, seeds were soaked in sterile LB for four hours at 26°C in a shaking incubator. These seeds were planted in a glasshouse in a potting medium containing a mixture of 25% Biogrow potting mix, 37.5% Fertool vermiculite and 37.5% Fertool perlite. For each treatment, eight seeds were planted at a depth of one centimetre (cm) around the edge of each pot for a total of 12 pots. The endophyte-treated seeds were subjected to one of three watering conditions. Pots under the well-watered, mild drought, or severe drought condition received 300mL, 150mL, or 50mL of water respectively, every 48 h. After the rst week of growth, seeds that had not germinated were removed, reducing the total number of plants per pot to four. After six weeks of growth, the plants were photographed and measured (shoot and root length in cm). Aerial and root tissue were separated then weighed. This data was analysed using OriginPro 2018 (Version b9.5.1.195) as described in "Evaluation of microbiome bacteria for biostimulation of Triticeae.
Pantoea and Pseudomonas were identi ed in both drought susceptible lines (DSL) and drought tolerant lines (DTL). Pantoea and Pseudomonas were the dominant genera in DSL isolates comprising 32.56% and 22.79% respectively, as opposed to 27.56% and 2.67% of DTL isolates. Pantoea and Pseudomonas were more prevalent under rainfed conditions (RC) comprising 43.8% and 26.8%, respectively, rather than 12.59% and 9.54% of all isolates from drought conditions (DC).
Arthrobacter and Curtobacterium isolates were exclusive to DTL and were the dominant genera accounting for 24.00% and 25.78% of isolates, respectively ( Figure 1). Arthrobacter isolates were also exclusive to DC, while Curtobacterium isolates dominated under DC (22%) as opposed to RC (1.68%). Furthermore, Curtobacterium was the only bacteria isolated from seeds of DT line 3 under DC, while Arthrobacter was the only bacteria isolated from DT line 4 under DC (Appendix 3).
Clavibacter and Rathayibacter were exclusively isolated from DTL under DC but only accounted for 2.22% and 0.89%, respectively (Figure 1). Erwinia and Paenibacillus were exclusive to DSL under RC, but only accounted for 0.47% and 0.47% of the population, respectively. Bacteria that were unable to be identi ed using MALDI-TOF were isolated from both DTL and DSL, being 16.5% and 43.7% of all isolates, respectively. Unknown isolates were also identi ed from DC and RC, at 33.2% and 24.7%, respectively.

Microbiome pro ling
Variation in diversity Samples were collected from the seeds of wheat cultivars of differing levels of drought tolerance that had been subject to drought and rainfed conditions. The microbiomes of these seeds were characterised using 16S rRNA amplicon sequencing. To characterise the diversity of the wheat seed microbiomes, alpha and beta diversity analyses were performed. Shannon index comparison of DC microbiomes (8.813) was signi cantly more diverse than RC microbiomes (7.035) (q = 2.49e-18, H=76.35, Kruskal-Wallis test). However, the number of observed OTUs and Shannon's index of DTL microbiomes and DSL microbiomes, when compared there was no signi cant difference (p>0.05, Kruskal-Wallis test). Robust Aitchison PCA (RPCA) was performed on the wheat seed microbiomes, using the QIIME2 DEICODE module [42]. The microbiome pro les of DTL and DSL showed signi cant differences (Figure 2a, PERMANOVA pseudo-F = 20.95, p = 0.001). There was also a signi cant difference between the microbiomes of wheat seeds subjected to either DC and RC (Figure 2b, PERMANOVA pseudo-F = 8.76, p = 0.001).
The microbiomes of DTL showed enrichment of certain microbes when compared with DSL microbiomes, under both DC and RC (Figure 3). At a family level, Enterbacteriaceae and Microbacteriaceae were highlighted. DTL had a higher abundance of Microbacteriaceae (1.045%) compared to DSL (0.359%), under DC. DTL also had higher abundance of Microbacteriaceae (0.514%) compared to DSL (0.274%), under RC. When ranking the relative abundance of OTUs at a family level, Microbacteriaceae were ranked at 25 for DTL under DC and 27 for DTL under RC (Figure 4). Comparatively, Microbacteriaceae were ranked at 50 for DSL under DC and 48 for DSL under RC. Genera from the order Microbacteriaceae were signi cant in an ANOVA of the microbiome of DTL and DSL under DC, namely Curtobacterium (p=3.29E-3), Agromyces (p=0.032) and an unknown genera (p=0.048) (Supplementary Table 3).

Plant growth promotion effects of the cultured microbiome Evaluation of microbiome bacteria for biostimulation of Triticeae
To assess the growth promotion effect of Curtobacterium accumfaciens novel strain Cf D3-25 and Arthrobacter sp. novel strains Ar. sp D4-14 on plants, a seedling assay was established with members of the tribe Triticeae (wheat -Triticum aestivum; spelt -Triticum spelta, durum -Triticum durum; ryecorn -Secale cereale; oats -Avena sativa; barley -Hordeum vulgare). Wheat seeds inoculated with different concentrations of Cf D3-25 or Ar. sp D4-14 were germinated and allowed to grow for seven days to evaluate potential biostimulation activity. There was a root lengthening effect observed in microbetreated wheat seedlings (see Figure 5). In early wheat seedlings inoculated in Cf D3-25 solutions diluted to 10 0 , 10 -2 and 10 -3 (containing 7 x 10 8 , 7 x 10 6 and 7 x 10 5 CFU mL -`1 , respectively), the root lengthening was signi cant compared to the uninoculated control, increasing root length by 7.94%, 9.05% and 7.95%.

Greenhouse drought pot trial
To assess the ability of Curtobacterium accumfaciens D3-25 and Arthrobacter sp. D4-14 to aid drought tolerance, an in planta assay was established in seed inoculated wheat exposed to varying levels of drought. At the end of six-weeks of growth, the wheat plants were harvested and the wet shoot and root weight were measured. Wheat plants inoculated with Cf D3-25 and Ar. sp D4-14 had signi cant increases in shoot and root weight, compared to control, across a range of conditions. Shoot weight of Cf D3-25inoculated wheat plants, under well-watered conditions, increased by 46.82% compared to the control (Tukey test, p=0.003, Figure 6a). Under drought conditions, there was a 7.71% increase in shoot length of wheat plants inoculated with Cf D3-25 under mild drought conditions, but this increase was not signi cant. The shoot weight of Ar. sp D4-14-inoculated wheat plants did not signi cantly increase, compared to control. Root weight was signi cantly increased in six-week old wheat plants inoculated with Cf D3-25 under mild and severe drought conditions, compared to Ar. sp D4-14-inoculated and control plants (Figure 6b) After six weeks of growth, the shoot length and number of leaves on each plant were measured. In the well-watered condition, the endophyte-treated plants had signi cantly more fully-expanded leaves than the control (Figure 6c). Plants that had been inoculated with either Cf D3-25 or Ar. sp D4-14 had 35% and 27% more fully-expanded leaves, respectively, than the control plants (Tukey test, p=0.001 and p=0.02). Ar. sp D4-14-treated plants exhibited signi cantly more full-expanded leaves under severe drought. Wheat plants inoculated with Cf D3-25 had an 8.47% increase in shoot length under well-watered conditions, compared to the control (Tukey test, p=0.001, Figure 6d). Similarly, wheat plants inoculated with Ar. sp D4-14-inoculated had signi cantly longer shoot lengths, under well-watered condition, with a 6.68% increase compared to the control (Tukey test, p=0.016). However, this effect was not evident under moderate or severe drought conditions. Discussion A combination of metagenomic and culture methods were used to characterize the seed microbiome composition of four drought tolerant and two drought susceptible wheat lines under rainfed and drought conditions. Water availability and line-speci c selection were responsible for signi cant variation in both diversity and abundance of microbes. A collection of microbes was cultured from the seeds of six wheat lines. From the collection, genera that had been enriched in microbiomes under drought conditions in drought tolerant lines were selected and screened for biostimulation effects and screened for effects on drought tolerance.

Variation of seed microbial diversity is in uenced by drought and wheat cultivar
The collection of microbes cultured from the seeds of six wheat lines showed considerable differences in genera diversity and abundance under drought stress. Genera belonging to Gammaproteobacteria, namely, Pantoea and Pseudomonas dominated the seed microbiomes of drought susceptible lines (DSL), but were depleted in drought tolerant lines (DTL). Pseudomonas, in particular, was depleted signi cantly in DTL compared to DSL. A similar trend was also observed between seed microbiomes under rainfed conditions and drought conditions (DC), where Pantoea and Pseudomonas were depleted under drought stress. In contrast, it was clear that genera belonging to Actinobacteria, had undergone line-dependent selection by drought tolerant wheat lines.
The microbiomes of these seeds were pro led using 16S rRNA amplicon sequencing and analysed using QIIME2. As observed in the cultured portion of the six wheat seed microbiomes, drought stress and wheat lines had signi cant effects the abundance of Gammaproteobacteria in microbiome pro ling. Under drought conditions, all lines showed a marked depletion of Gammaproteobacteria families. In DTL, Enterobacteriaceae and Pseudomonadaceae the two most dominant families in RC were signi cantly depleted under DC. This depletion was not observed to the same extent in DSL under DC. Interestingly, previous studies of the wheat microbiome under drought stress have not observed the marked depletion in Gammaproteobacteria observed across wheat lines in this study. In a study by Naylor et al, there was no detectable change to the abundance of Gammaproteobacteria under drought stress [12]. It is possible that the depletion of Gammaproteobacteria under drought conditions is a phenomenon that is speci c to the seed microbiome. Actinobacteria families were selected for and enriched by DTL lines under DC. Under rainfed conditions, the Microbacteriaceae family made up 1.045% of DTL and only 0.274% of DSL. Under drought stress, Microbacteriaceae was enriched in both DTL and in DSL, though the enrichment was more pronounced in DTL. The enrichment of Actinobacteria in the microbiomes of wheat and other crop species has been consistently observed in multiple drought studies [12,16,25,44,45]. A study of 30 genetically divergent plant species, subjected to drought stress, observed an increase in relative abundance of Actinobacteria six-fold, although results varied by plant species [44]. Similarly, studies of microbiomes under drought conditions have highlighted the enrichment of Actinobacteria by 3.1 fold in wheat as well as a signi cant increase in agave [12,45]. The enrichment of Actinobacteria is particularly prevalent in the root microbiome of wheat plants undergoing the early stages of drought stress [12]. It has been proposed that the enrichment of Actinobacteria genera under drought could be driven by one or more conserved properties of the Actinobacterial lineage [46].
Seed microbiomes of DTL demonstrated higher levels of Actinobacteria enrichment under DC, compared to the seed microbiomes of DSL. Cultivar-speci c enrichment of speci c genera (and occasionally speci c strains) has been demonstrated in potato, common bean, cannabis, sorghum and wheat [47][48][49][50][51]. Wheat cultivars Lewjain, Penawawa and Symphony have been demonstrated to select species uorescent pseudomonads from soil microbiomes [52][53][54]. In root microbiomes, striga-resistant sorghum cultivars enriched for Acidobacteria GP1, Burkholderia, Cupriavidus (Burkholderiaceae), Acidovorax and Albidiferax (Comamonadaceae) OTUs, when grown in unfertilised soil [51]. In fact, the relationship between strain and cultivar can be so speci c that a bacterial strain that is bene cial in one cultivar can be detrimental in another. In Brassica napus, Paenibacillus polymyxa Sb3-1 increased growth in the cv. Avata but caused leaf yellowing in the cv. Traviata [32]. In DT Line 3 and DT Line 4, Arthrobacter was enriched 1.5 and 1.9-fold, respectively. This was not observed in drought susceptible lines, suggesting that the enrichment of Arthrobacter is dependent on cultivar genetics. Previous studies have demonstrated cultivar-speci c selection of an Arthrobacter OTU in the rhizosphere of wheat cultivar PI561725 and enrichment of the genera in the presence of 2,4-diacetylphloroglucinol producing species [50]. This study suggests that drought tolerant lines select and enrich for Actinobacteria genera when they recruit the seed microbiome of the next generation.
The seed microbiomes under DC had a higher Shannon's index than their rainfed counterparts, indicating higher levels of diversity. There is con icting data on the effects of drought on microbial diversity. In a study by Jochum et al, multiple generations of drought stress on wheat microbiomes lead to an increase in alpha and beta diversity [25]. However, in other studies the microbiomes of sorghum and wheat exhibited a marked decrease in diversity [26,55]. It has been suggested that increased diversity could be explained by selective enrichment for functionality, rather than taxonomy [25]. It is worth noting that an increase in diversity under drought could be due to the depletion of dominant taxa, such as Gammaproteobacteria, allowing a greater number of OTUs with low abundance to be captured by sequencing.
Seed microbiomes enrich for microbes that promote root growth and drought tolerance To determine if bacteria enriched in DTL lines under drought stress had a biostimulation effect, members of the tribe Triticeae were inoculated with representative isolates from the two cultured Actinobacteria genera. Wheat and oat seeds inoculated with different concentrations of Cf D3-25 and Ar. sp D4-14 had signi cantly longer roots, compared to control plants. In wheat, Cf D3-25-treated seedlings had an increase in root length between 7.94-9.05% and Ar. sp D4-14-treated seedlings had a 21% increase in root length. In oat, Cf D3-25-treated seedlings solutions had an increase in root length between 63.85-104.68 and Ar. sp D4-14-treated plants had between 63.85-80.14% increase in root length. The inoculation concentration of Cf D3-25 did not appear to signi cantly impact the root lengthening effect seen in oat and wheat. High concentrations of Ar. sp D4-14 inhibited the growth of barley and oats. Curtobacterium has been found to associate with roots and promote plant growth in Arabidopsis, lettuce, basil, red clover and cucumber [56][57][58].
Representative isolates from Curtobacterium and Arthrobacter were inoculated into wheat seedlings which were then exposed to varying levels of drought to determine if they could increase drought tolerance. Cf D3-25-treated wheat plants showed signi cant increases in root weight under both mild (26%) and severe drought conditions (27%), compared to the control. Wheat plants inoculated with Ar. sp D4-14 had no signi cant increase in wet root weight. Cf D3-25 treated wheat plants also showed a signi cant increase in wet shoot weight (47%) under well-watered conditions, although this effect was not evident under either drought conditions. Both Cf D3-25 and Ar. sp D4-14 had signi cantly more fully expanded leaves than the control under well-watered conditions, showing an increase of 35% and 27%, respectively. There was no signi cant difference in the number of fully expanded leaves in isolate-treated wheat plants under drought conditions. Ar. sp D4-14-treated wheat plants did not show a clear drought response in the cv. Bob White. However, as the action of these isolates may be cultivar-speci c, it may be prudent to investigate the effects of Ar. sp D4-14-inoculation on other wheat cultivars under drought conditions.
There are a number of microbial traits that are associated with increasing drought tolerance in host plants, speci cally the ability to produce 1-aminocyclopropane-1-carboxylate deaminase (ACCd), indole-3acetic acid (IAA) and siderophores [12]. ACCd regulates ethylene, preventing it from reaching inhibitory levels and thus allowing normal root growth under drought conditions [59]. Auxin analogue IAA enhances shoot and root growth under drought conditions by regulating stomatal aperture [60]. Under drought conditions, siderophore synthesis allows for nutrient cycling [61]. In lettuce under drought conditions treatment, Curtobacterium herbarum strain CAH5 increased shoot length, root length and wet biomass by 1.54, 1.23 and 3.84 fold, respectively, compared to control plants [62]. C. herbarum CAH5 demonstrated the ability to produce IAA, ACC deaminase, siderophores and solubilise phosphate [62]. C. accumfaciens colonizes xylem system of the host plants, where it can be either pathogenic (causing vascular wilt in bean) or endophytic [63]. It has been suggested that the prominence of Actinobacteria in plant microbiomes could also be due to their method of infection [63]. Colonisation by xylem-limited bacteria increases resistance to water ow, causing stomatal closure and lowering transpiration [64]. Inoculation of A. nitroguajacolicus, the species of Arthrobacter whose 16S rRNA gene pro le is the most closely related to Ar. sp D4-14, into wheat increased the total dry weight by more than 200% in conditions of high salinity, a condition at often occurs in tandem with drought [65]. A. nitroguajacolicus demonstrated the ability to produce auxin, ACC deaminase, siderophores and solubilize phosphate [65]. Another A. pokkalii strain P3B162 T has been observed to grow in 10% and 20% polyethylene glycol (PEG) solutions, indicating the strain is drought tolerant [66]. In previous studies, Curtobacterium and Arthrobacter isolates have shown a number of traits that increase abiotic stress tolerance i.e. drought tolerance under a range of conditions.

Final remarks
This study indicates seed microbiomes from genetically distinct wheat lines enrich for bene cial bacteria in ways that are both line-speci c and responsive to environmental stress. As such, they represent an invaluable resource for the further identi cation of bene cial microbes with plant growth promoting activity and hence improving commercial crop production. The composition of the seed microbiomes differed signi cantly between lines and environmental conditions. It was clear that drought tolerant lines selected for and enriched certain microbial families when exposed to either rainfed or drought conditions. Both Cf D3-25 and Ar. sp D4-14, cultured microbes belonging to taxa enriched by drought tolerant lines under drought conditions, demonstrated the ability to promote plant growth. Cf D3-25 increased drought tolerance in wheat. Microbe pro ling suggested the enrichment of microbes was line-speci c, therefore, it would be worthwhile to assay key microbes in other wheat cultivars. Combinations of genetically tolerant and susceptible lines under different stresses could be exploited in a similar manner to nd other useful bacteria e.g. in highlighting N e cient microbes in microbiomes subjected to environment low in nitrogen.