Biofilms Positively Contribute to Bacillus amyloliquefaciens 54-induced Drought Tolerance in Tomato Plants

Drought stress is a major obstacle to agriculture. Although many studies have reported on plant drought tolerance achieved via genetic modification, application of plant growth-promoting rhizobacteria (PGPR) to achieve tolerance has rarely been studied. In this study, the ability of three isolates, including Bacillus amyloliquefaciens 54, from 30 potential PGPR to induce drought tolerance in tomato plants was examined via greenhouse screening. The results indicated that B. amyloliquefaciens 54 significantly enhanced drought tolerance by increasing survival rate, relative water content and root vigor. Coordinated changes were also observed in cellular defense responses, including decreased concentration of malondialdehyde and elevated concentration of antioxidant enzyme activities. Moreover, expression levels of stress-responsive genes, such as lea, tdi65, and ltpg2, increased in B. amyloliquefaciens 54-treated plants. In addition, B. amyloliquefaciens 54 induced stomatal closure through an abscisic acid-regulated pathway. Furthermore, we constructed biofilm formation mutants and determined the role of biofilm formation in B. amyloliquefaciens 54-induced drought tolerance. The results showed that biofilm-forming ability was positively correlated with plant root colonization. Moreover, plants inoculated with hyper-robust biofilm (ΔabrB and ΔywcC) mutants were better able to resist drought stress, while defective biofilm (ΔepsA-O and ΔtasA) mutants were more vulnerable to drought stress. Taken altogether, these results suggest that biofilm formation is crucial to B. amyloliquefaciens 54 root colonization and drought tolerance in tomato plants.


Introduction
Tomato plants (Solanum lycopersicum), as an important part of the agricultural industry, are widely cultivated across the world. World production of tomatoes was 177 million tons in 2016. Abiotic stresses are various adverse environmental factors that cause severe declines in yields of tomato and other plant crops. In terms of plant physiology, drought is a common stress environmental factor that limits growth and leads to mortality of crops. Moreover, due to global warming, drought stress type B. amyloliquefaciens 54. The pellicle formations and colony morphologies of the wild-type strain and its derivatives were examined in defined media (LBGM). As shown in Figure 1, compared to wild 54, null mutations in abrB or ywcC caused formation of hyper-robust biofilm pellicles ( Figure  1A, C) in LBGM liquid medium and more complex colony patterns ( Figure 1B) on LBGM agar plates, whereas epsA-O or tasA mutants formed pellicles with unwrinkled or featureless surfaces and colony patterns with apparent defects ( Figure 1A-C). These results suggest that these genes are related to B. amyloliquefaciens 54 biofilm formation.  In this study, we compared the impact of root inoculation with wild-type B. amyloliquefaciens 54 and its hyper-robust biofilm (∆abrB and ∆ywcC) and defective biofilm (∆epsA-O, and ∆tasA) mutants on drought tolerance in tomato plants.

Beneficial Rhizobacteria Significantly Improve Drought Tolerance
Of the bacterial isolates, three showed strong drought tolerance, with survival rates exceeding 40% after the water deprivation period, while control bacteria without PGPR root inoculation had a survival rate of only 6.25% (Table S2). Of these three isolates, B. amyloliquefaciens 54 conferred the strongest drought tolerance, and the survival rate of treated seedlings reached 61.54%.

B. Amyloliquefaciens 54 Forms Robust Biofilms Both in LBGM Medium and on The Surface of Tomato Roots
Previous studies have mentioned that the bacteria can form multicellular biofilm communities that are crucial to efficient plant root colonization and can protect plant roots against environmental hazards [45,46]. The strong drought tolerance induced by B. amyloliquefaciens 54 could be related to formation of biofilms. Therefore, we evaluated its biofilm formation by monitoring pellicle characteristics and colony morphologies in bofilm medium LBGM (lysogenic broth [LB] supplemented with 1% glycerol and 100 µM MnSO4). As expected, B. amyloliquefaciens 54 formed robust and wrinkled pellicles in LBGM liquid medium ( Figure 1A,C) and complex colony patterns on agar plates ( Figure 1B). According to the regulatory genetic circuitry that regulates biofilm formation ( Figure 1D) [31][32][33], null mutations of these genes which result in hyper-robust biofilms (∆abrB and ∆ywcC) and strong defective biofilms (∆epsA-O and ∆tasA) were introduced into wild-type B. amyloliquefaciens 54. The pellicle formations and colony morphologies of the wild-type strain and its derivatives were examined in defined media (LBGM). As shown in Figure 1, compared to wild 54, null mutations in abrB or ywcC caused formation of hyper-robust biofilm pellicles ( Figure 1A,C) in LBGM liquid medium and more complex colony patterns ( Figure 1B) on LBGM agar plates, whereas epsA-O or tasA mutants formed pellicles with unwrinkled or featureless surfaces and colony patterns with apparent defects ( Figure 1A-C). These results suggest that these genes are related to B. amyloliquefaciens 54 biofilm formation.
Considering that B. amyloliquefaciens 54 forms strong biofilms in defined medium, we wondered whether it might also form biofilms on the surface of plant roots. Therefore, we examined biofilm formation of B. amyloliquefaciens 54 and its biofilm mutants on tomato plant roots. We introduced a constitutively expressed GFP (Green Fluorescent Protein) reporter into the wild-type strain and its biofilm mutants to better view the biofilms on roots. As expected, wild-type B. amyloliquefaciens 54 colonized the roots and formed strong biofilms, and ∆abrB and ∆ywcC mutants formed larger cell clusters on the roots and more obvious biofilm-like structures ( Figure 2). Conversely, knockout of epsA-O or tasA induced a sharp decrease in colonization, and almost no biofilm formation was detected on the roots ( Figure 2). formation of B. amyloliquefaciens 54 and its biofilm mutants on tomato plant roots. We introduced a constitutively expressed GFP (Green Fluorescent Protein) reporter into the wild-type strain and its biofilm mutants to better view the biofilms on roots. As expected, wild-type B. amyloliquefaciens 54 colonized the roots and formed strong biofilms, and ΔabrB and ΔywcC mutants formed larger cell clusters on the roots and more obvious biofilm-like structures ( Figure 2). Conversely, knockout of epsA-O or tasA induced a sharp decrease in colonization, and almost no biofilm formation was detected on the roots ( Figure 2).

Biofilm Formation Contributed to B. Amyloliquefaciens 54-Induced Drought Tolerance
To reveal whether biofilm formation regulates drought stress tolerance, we root-inoculated wild-type B. amyloliquefaciens 54 and its biofilm mutants. After being deprived of water for 15 days before being watered again, plants treated with hyper-robust biofilm mutants (∆abrB and ∆ywcC) showed stronger drought tolerance than that displayed by plants treated with wild-type B. amyloliquefaciens 54, while plants treated with defective biofilm mutants (∆epsA-O and ∆tasA) showed weaker drought tolerance ( Figure S1).
The relative water content (RWC) of leaves is a crucial indicator of the water status in plants and plays a vital role in protecting plants from over-dehydration stress. Although root inoculation with wild-type B. amyloliquefaciens 54 and all tested biofilm mutants improved RWC over that of control plants, hyper-robust biofilm mutant (∆abrB and ∆ywcC) treatment increased the RWC of leaves more than did wild-type treatment, whereas defective biofilm mutant (∆epsA-O and ∆tasA) treatment resulted in lower leaf RWC ( Figure 3A). resulted in lower leaf RWC ( Figure 3A).
A previous study showed that root vigor is a sharp and reliable biomarker of plant drought stress tolerance. Similar to the RWC results, all B. amyloliquefaciens 54 treatments improved root vigor over that of control plants. Hyper-robust biofilm mutant (ΔabrB and ΔywcC) treatment, especially with the ΔabrB mutant, resulted in higher root vigor that did wild-type treatment, while defective biofilm mutant (ΔepsA-O and ΔtasA) treatment resulted in lower root vigor ( Figure 3B).
Abiotic stress is known to promote peroxidation of membrane lipid and electrolyte leakage in plant tissues, and malondialdehyde (MDA) is frequently used as an indicator of lipid peroxidation. Therefore, leaf MDA content was determined. The data indicated that wild-type B. amyloliquefaciens 54 and all mutants reduced leaf MDA content. Furthermore, hyper-robust biofilm mutant (ΔabrB and ΔywcC) treatment reduced MDA content more than did wild-type treatment, while defective biofilm mutant (ΔepsA-O and ΔtasA) treatment reduced MDA content ( Figure 3C). The above results suggest that B. amyloliquefaciens 54 biofilm formation is positively related to drought stress tolerance in tomato plants.  The relative transcript levels of stress-responsive genes lea (D), tdi65 (E), ltpg2 (F). SlActin was used as an internal control to normalize individual gene expression values. The data were measured after microbe inoculation and subsequently withholding water for 11 days. Data represent the mean ± SD (n = 3). Different letters on top of the bars indicated significant differences among different treatments. Tukey's studentized range (HSD) test was applied in the statistical analysis.
A previous study showed that root vigor is a sharp and reliable biomarker of plant drought stress tolerance. Similar to the RWC results, all B. amyloliquefaciens 54 treatments improved root vigor over that of control plants. Hyper-robust biofilm mutant (∆abrB and ∆ywcC) treatment, especially with the ∆abrB mutant, resulted in higher root vigor that did wild-type treatment, while defective biofilm mutant (∆epsA-O and ∆tasA) treatment resulted in lower root vigor ( Figure 3B).
Abiotic stress is known to promote peroxidation of membrane lipid and electrolyte leakage in plant tissues, and malondialdehyde (MDA) is frequently used as an indicator of lipid peroxidation. Therefore, leaf MDA content was determined. The data indicated that wild-type B. amyloliquefaciens 54 and all mutants reduced leaf MDA content. Furthermore, hyper-robust biofilm mutant (∆abrB and ∆ywcC) treatment reduced MDA content more than did wild-type treatment, while defective biofilm mutant (∆epsA-O and ∆tasA) treatment reduced MDA content ( Figure 3C). The above results suggest that B. amyloliquefaciens 54 biofilm formation is positively related to drought stress tolerance in tomato plants.

Biofilm Formation Was Conductive to B. Amyloliquefaciens 54-Regulated Stress-Related Genes
To further elucidate the role of B. amyloliquefaciens 54 in drought stress, we analyzed the expression of several stress-related genes, including lea, ltpg2, and tdi-65. As shown in Figure 3, pre-treatment with wild-type B. amyloliquefaciens 54 increased expression of these genes relative to expression after mock treatment ( Figure 3D-F). To clarify whether biofilm formation is involved in the regulatory mechanism, we inoculated plants with biofilm mutants of B. amyloliquefaciens 54. As shown in Figure 3, compared with wild-type B. amyloliquefaciens 54 inoculation, hyper-robust biofilm mutant (∆abrB and ∆ywcC) treated plants produced higher gene expression, while defective biofilm mutant (∆epsA-O and ∆tasA) treatment produced lower gene expression. This suggests that B. amyloliquefaciens 54 biofilm formation is positively involved in the observed induction of stress-related gene expression.

Biofilm Formation Was Involved in B. Amyloliquefaciens 54-Mediated Antioxidant Enzyme Activities
Reactive oxygen intermediates (ROIs) are widely used signaling molecules that control various processes, such as pathogen defense, abiotic stress responses and programmed cell death [47]. SOD, CAT, POD, and APX are referred to as the major ROI-scavenging enzymes. Therefore, we investigated the activities of these antioxidant enzymes in the leaves of tomato plants pre-inoculated with B. amyloliquefaciens 54 and its biofilm mutants. Under drought stress, B. amyloliquefaciens 54 treatment elevated the activities of SOD, CAT, POD, and APX of tomato plants at 7 and 11 days compare to mock-inoculated controls ( Figure 4A-C and Figure 3D). Furthermore, hyper-robust biofilm mutant (∆abrB and ∆ywcC) treatment elevated SOD, POD, CAT, and APX activities compared to wild-type treatment at 11 days and 7 days (Figure 4A-C and Figure 3D). In addition, defective biofilm mutant (∆epsA-O and ∆tasA) treatment decreased SOD and POD activities compare to wild-type treatment at 11 days ( Figure 4A,B). However, CAT and APX activities displayed similar trends between the defective biofilm mutant (∆epsA-O and ∆tasA) treatment and wild-type treatment ( Figures 4C and 3D).
Tukey's studentized range (HSD) test was applied in the statistical analysis.

Biofilm Formation Was Conductive to B. Amyloliquefaciens 54-Regulated Stress-Related Genes
To further elucidate the role of B. amyloliquefaciens 54 in drought stress, we analyzed the expression of several stress-related genes, including lea, ltpg2, and tdi-65. As shown in Figure 3, pretreatment with wild-type B. amyloliquefaciens 54 increased expression of these genes relative to expression after mock treatment ( Figure 3D-F). To clarify whether biofilm formation is involved in the regulatory mechanism, we inoculated plants with biofilm mutants of B. amyloliquefaciens 54. As shown in Figure 3, compared with wild-type B. amyloliquefaciens 54 inoculation, hyper-robust biofilm mutant (ΔabrB and ΔywcC) treated plants produced higher gene expression, while defective biofilm mutant (ΔepsA-O and ΔtasA) treatment produced lower gene expression. This suggests that B. amyloliquefaciens 54 biofilm formation is positively involved in the observed induction of stressrelated gene expression.

Biofilm Formation Was Involved in B. Amyloliquefaciens 54-Mediated Antioxidant Enzyme Activities
Reactive oxygen intermediates (ROIs) are widely used signaling molecules that control various processes, such as pathogen defense, abiotic stress responses and programmed cell death [47]. SOD, CAT, POD, and APX are referred to as the major ROI-scavenging enzymes. Therefore, we investigated the activities of these antioxidant enzymes in the leaves of tomato plants pre-inoculated with B. amyloliquefaciens 54 and its biofilm mutants. Under drought stress, B. amyloliquefaciens 54 treatment elevated the activities of SOD, CAT, POD, and APX of tomato plants at 7 and 11 days compare to mock-inoculated controls ( Figure 4A-C, and 3D). Furthermore, hyper-robust biofilm mutant (ΔabrB and ΔywcC) treatment elevated SOD, POD, CAT, and APX activities compared to wild-type treatment at 11 days and 7 days ( Figure 4A-C, and 3D). In addition, defective biofilm mutant (ΔepsA-O and ΔtasA) treatment decreased SOD and POD activities compare to wild-type treatment at 11 days ( Figure 4A,B). However, CAT and APX activities displayed similar trends between the defective biofilm mutant (ΔepsA-O and ΔtasA) treatment and wild-type treatment ( Figure 4C and 3D).

Biofilm Formation Was Involved in B. Amyloliquefaciens 54-Induced Stomatal Closure in Tomato Plants
Root-associated B. subtilis restricts stomatal openings in response to drought and pathogen challenges [48]. To clarify whether B. amyloliquefaciens 54 induces stomatal closure under drought stress conditions, we pre-treated tomato plant roots with B. amyloliquefaciens 54 and measured the stomatal apertures of the epidermis via microscopy. As shown in Figure 5, root inoculation with B. amyloliquefaciens 54 reduced stomatal apertures at 8 and 24 h compared to that in mock-inoculated controls ( Figure 5A,B). Furthermore, both hyper-robust biofilm mutant (∆abrB and ∆ywcC) and defective biofilm mutant (∆epsA-O, and ∆tasA) treatments induced stomatal closure ( Figure 5A,B). Interestingly, both 8 h and 24 h post-inoculation, plants treated with hyper-robust biofilm mutants (∆abrB and ∆ywcC), especially with the ∆abrB mutant, showed smaller stomatal sizes than did those inoculated with the wild-type; on the other hand, those treated with defective biofilm mutants (∆epsA-O and ∆tasA), especially with the ∆tasA mutant, showed larger apertures ( Figure 5A,B). These results indicate that B. amyloliquefaciens 54 induced stomatal closure, with robust biofilm mutants showing enhanced ability and defective biofilm mutants showing decreased ability to mediate such closure.

Biofilm Formation Was Involved in B. Amyloliquefaciens 54-Induced Stomatal Closure in Tomato Plants
Root-associated B. subtilis restricts stomatal openings in response to drought and pathogen challenges [48]. To clarify whether B. amyloliquefaciens 54 induces stomatal closure under drought stress conditions, we pre-treated tomato plant roots with B. amyloliquefaciens 54 and measured the stomatal apertures of the epidermis via microscopy. As shown in Figure 5, root inoculation with B. amyloliquefaciens 54 reduced stomatal apertures at 8 and 24 h compared to that in mock-inoculated controls ( Figure 5A,B). Furthermore, both hyper-robust biofilm mutant (ΔabrB and ΔywcC) and defective biofilm mutant (ΔepsA-O, and ΔtasA) treatments induced stomatal closure ( Figure 5A and 5B). Interestingly, both 8 h and 24 h post-inoculation, plants treated with hyper-robust biofilm mutants (ΔabrB and ΔywcC), especially with the ΔabrB mutant, showed smaller stomatal sizes than did those inoculated with the wild-type; on the other hand, those treated with defective biofilm mutants (ΔepsA-O and ΔtasA), especially with the ΔtasA mutant, showed larger apertures ( Figure  5A,B). These results indicate that B. amyloliquefaciens 54 induced stomatal closure, with robust biofilm mutants showing enhanced ability and defective biofilm mutants showing decreased ability to mediate such closure. amyloliquefaciens 54, and its derived mutants. SlActin was used as an internal control to normalize nced1 expression value. Data represent the mean ± SD (n = 3). Different letters on top of the bars indicate significant differences among different treatments. Tukey's studentized range (HSD) test was applied in the statistical analysis.

ABA Is Required for B. Amyloliquefaciens 54-Mediated Stomatal Closure
Under drought stress, plants accumulate the hormone ABA, which plays a major role in stomatal closure to prevent water loss by transpiration [49][50][51]. Therefore, to investigate whether ABA contributes to B. amyloliquefaciens 54-triggered stomatal closure, we examined the transcript levels of the nced1 gene that encodes 9-cis-epoxycarotenoid dioxygenase, which is a key enzyme in ABA biosynthesis, and the ABA contents in tomato leaves 8 and 24 h post-inoculation with wild-type B. amyloliquefaciens 54 and its biofilm mutants. The results showed a significant increase in nced1 gene transcription ( Figure 5C) and ABA content ( Figure 5D) compared to those in non-inoculated controls.

ABA Is Required for B. Amyloliquefaciens 54-Mediated Stomatal Closure
Under drought stress, plants accumulate the hormone ABA, which plays a major role in stomatal closure to prevent water loss by transpiration [49][50][51]. Therefore, to investigate whether ABA contributes to B. amyloliquefaciens 54-triggered stomatal closure, we examined the transcript levels of the nced1 gene that encodes 9-cis-epoxycarotenoid dioxygenase, which is a key enzyme in ABA biosynthesis, and the ABA contents in tomato leaves 8 and 24 h post-inoculation with wild-type B. amyloliquefaciens 54 and its biofilm mutants. The results showed a significant increase in nced1 gene transcription ( Figure 5C) and ABA content ( Figure 5D) compared to those in non-inoculated controls. Furthermore, hyper-robust biofilm mutant (∆abrB and ∆ywcC) treated plants displayed higher nced1 gene expression and ABA content than did wild-type-treated plants, whereas defective biofilm mutant (∆epsA-O and ∆tasA) treated plants showed lower nced1 gene expression and ABA content. These findings demonstrated that ABA is involved in B. amyloliquefaciens 54-mediated stomatal closure and that biofilm formation contributes to ABA biosynthesis regulation by B. amyloliquefaciens 54.

Discussion
In recent years, the functions of PGPR in improving plant health by promoting resistance to pathogens, insect pests, and abiotic stressors, such as drought and salinity, have been well illustrated [10,11,13]. B. amyloliquefaciens and other members of the genus Bacillus have been implicated in a range of plant functions from plant growth and development to biotic and abiotic stress tolerance. In this study, we isolated wild strains of B. amyloliquefaciens 54, which has been demonstrated to enhance drought tolerance. However, how this strain increases drought tolerance in the rhizosphere is still not well understood.
Biofilms encase populations of different organisms and are pivotal to colonization and maintenance of long-term associations between microbes and plant hosts [52,53]. Biofilms protect against environmental stressors [54,55]. Timmusk et al. showed that biofilm formation ability of Paenibacillus polymyxa contributes to enhancing wheat drought tolerance [56]. Lu et al. showed that the EPS-encoding gene epsc plays a vital role in Bacillus amyloliquefaciens FZB42 to boost the drought tolerance of Arabidopsis thaliana [57]. In our study, we paid attention to the function of biofilm formation in plant drought tolerance and present several pieces of evidence suggesting its importance. First, we demonstrated that the isolated wild-type B. amyloliquefaciens 54, which has an excellent ability to help tomato plants against drought stress, forms robust biofilms ( Figure 1A-C). Furthermore, we applied its biofilm mutants to demonstrate that its biofilm formation ability is positively related to colonization ability and drought tolerance efficacy (Figure 2). That is, hyper-robust biofilm mutants significantly improved colonization ability and drought tolerance efficacy, while defective biofilm mutants showed substantially decreased colonization ability and drought tolerance efficacy compared to the same parameters in plants treated with the wild-type strain. These results demonstrate that the capacity of B. amyloliquefaciens 54 to form biofilms in LBGM medium is positively related to its capacity to do so on tomato plant roots.
Leaf RWC was introduced as a criterion of a plant's water status that accurately presents the balance between water absorption through root hairs and water consumed through transpiration [58]. The present study showed that tomato plants pre-inoculated with B. amyloliquefaciens 54 had relatively higher RWC under drought stress conditions than did untreated controls ( Figure 3A). These results confirm previous reports, in which PGPR was shown to increase RWC of plants [59,60]. Furthermore, the biofilm formation ability was positively related to leaf RWC; hyper-robust biofilms increased RWC, while defective biofilms decreased RWC compared to that in plants treated with the wild-type strain. The results demonstrate that biofilm formation is involved in B. amyloliquefaciens 54-induced RWC of tomato plants.
Root vigor has been widely used as a sharp and reliable biomarker of drought tolerance and reflects the capacity of water and nutrition absorption under drought stress conditions [61]. A previous study showed that a consortium of three PGPR strains could increase root vigor and promote drought tolerance in cucumber plants [4]. In our study, B. amyloliquefaciens 54-treated plants displayed sharply increased root vigor compared with mock-inoculated controls during the drought stress period. In addition, hyper-robust biofilms increased root vigor, while defective biofilms decreased root vigor compared to that of plants treated with the wild-type strain. This suggests that biofilm formation contributes to B. amyloliquefaciens 54-induced root vigor enhancements.
A general component of the plant response to drought stress is the accumulation of ROS compounds which can harm the production of biomolecules, such as lipids, nucleic acids, and proteins, and which therefore need to be neutralized [62,63]. Peroxidation of membrane lipids is a major damaging effect of ROS. MDA is a widely used indicator of oxidative lipid injury in plants [64]. In our investigation, leaf MDA content was significantly lower in plants pre-inoculated with B. amyloliquefaciens 54 ( Figure 3C). The result is consistent with previous studies, in which PGPR was shown to decrease the MDA content of plants under abiotic stress conditions [65,66]. Moreover, hyper-robust biofilms increased leaf MDA content, while defective biofilms decreased leaf MDA content compared to that in plants treated with the wild-type strain ( Figure 3C), suggesting that B. amyloliquefaciens 54 biofilm formation protects cell membranes from ROS and decreases lipid peroxidation, thereby enhancing drought tolerance. In addition, to prevent oxidative stress and cellular damage, plants produce ROS-scavenging enzymes, including SOD, POD, CAT, and APX [12,67]. Under drought stress conditions, plants pre-inoculated with B. amyloliquefaciens 54 displayed relatively higher SOD, POD, CAT, and APX activities than did mock-inoculated controls ( Figure 4A-D). Furthermore, hyper-robust biofilms increased these activities, while defective biofilms decreased the activities relative to those in plants treated with the wild-type strain ( Figure 4A-D). These results indicate that B. amyloliquefaciens 54 induced antioxidant enzyme activities, with robust biofilm mutants showing enhanced ability and defective biofilm mutants showing decreased ability to mediate such activities.
Stomatal closure is the primary response to drought stress in most plants that curtail water loss from transpiration. Therefore, the stomatal aperture is a major factor that contributes to drought tolerance [68,69]. The present study demonstrated that plants pre-inoculated with B. amyloliquefaciens 54 had smaller stomatal aperture than did control plants ( Figure 5A,B). The results confirm previous reports that PGPR could induce stomatal closure [70]. In addition, biofilm formation was positively related to stomatal aperture. That is, hyper-robust biofilms decreased stomatal aperture, whereas defective biofilms increased stomatal aperture relative to that in plants treated with the wild-type strain.
ABA was reported as a central regulator of stomatal development and aperture [71,72]. Previous studies reported that water shortage in plant roots would increase more pronounced ABA biosynthesis levels in roots than in leaves. However, ABA produced in leaf tissues plays a more important role than ABA produced in root tissues, although root ABA was also transported to leaf tissues to maintain stomatal closure [64]. Thus, nced1, an ABA biosynthesis gene, and leaf ABA content were measured. B. amyloliquefaciens 54 treatment significantly increased nced1 gene expression and ABA content compared with levels in uninoculated controls. These results were in agreement with previous studies suggesting that PGPR could induce ABA biosynthesis [70,73]. Additionally, hyper-robust biofilms increased both nced1 gene expression and ABA content, while defective biofilms decreased both compared to levels in plants treated with the wild-type strain.
In summary, our results demonstrated that B. amyloliquefaciens 54 is able to form biofilms on the surface of plant roots and that biofilm is positively associated with B. amyloliquefaciens 54-induced drought stress tolerance. It also provides a new theoretical basis for plant drought resistance induced by PGPR and new theoretical support for the application of PGPR in arid environments. Besides, we hypothesize that chemical communication is a crucial link between the plant roots and bacterium. Chen et al. showed that B. subtilis can sense the chemical signal released from plant roots and induce biofilm formation [74]. Therefore, it will be interesting to identify those chemical signal molecules that trigger Bacillus sp. to form biofilm and to understand the mechanism how the bacterium senses and responds to the signal in regulating biofilm formation.

Bacterial Strains and Culture Conditions
All 30 examined bacterial strains in this study came from our biocontrol bacteria library, which consisted of beneficial bacteria isolated by our lab from the field. They were cultured at 28 • C for 24 h on Luria-Bertani (LB) agar medium. A single colony from a freshly streaked plate was then selected and inoculated into LB broth and incubated at 28 • C for 16 h in a shaker at 200 rpm. The broth culture was spun at 6000× g in a centrifuge for 15 min, and the resulting pellet was resuspended in sterile water and adjusted to a final concentration of 10 6 colony forming units (CFU)/mL (OD 600 = 0.5) for further experiments.
Genomic DNA from B. subtilis 3610 mutants was used to construct corresponding B. amyloliquefaciens 54 mutants according to the established protocol for transformation of Bacillus spp. [75]. To view the biofilms surrounding the roots of tomato plants, the genomic DNA of B. subtilis strain YC21 was used to construct a GFP fusion containing a constitutive expression GFP reporter according to a method described previously [30].

Plant Growth and Treatments
Tomato seeds (S. lycopersicum cv. Shanghai 903) were sterilized for 30 s with 95% (v/v) ethanol followed by a further 5 min in 5% (w/v) NaClO, and then washed with sterile distilled water three times. Afterwards, the seeds were spread onto solid Murashige and Skoog (MS) medium to germinate. The seedlings were transplanted into natural clean soil and incubated in a greenhouse with a relative humidity of 60% and a 16/8 h day/night photoperiod. The tomato plantlets were transplanted into pots after two weeks. One week later, 20 mL of each adjusted suspension was applied to the plants by root irrigation. Sterile distilled water was applied as a control in this experiment. For each treatment, 24 seedlings were applied. The plants were then deprived of water for 15 days as drought stress treatment. After 15 days, normal watering was resumed. To characterize drought tolerance, survival rate of each treatment was measured after 5 days rewatering.

Biofilm Formation Assay
To evaluate biofilm formation of wild-type B. amyloliquefaciens 54 and its derivatives, we monitored colony morphology on LBGM agar plate and pellicle formation on LBGM liquid media. Cells were first grown in 3 mL LB broth to late exponential growth phase. For colony morphology, 2 µL of LB pre-culture (OD 600 = 0.5) was spotted onto LBGM agar plates, which were incubated at 30 • C for 36 h prior to analysis. To analyze pellicle biofilm formation, 10 µL of LB pre-culture (OD 600 = 0.5) was added to 10 mL of LBGM liquid medium and cultured for 48 h at 30 • C. Images were taken by a Canon digital camera (EOS Digital 600D, Canon Co, Tokoyo, Japan).

Biofilm Formation of B. Amyloliquefaciens 54 and Its Biofilm Mutants on Tomato Roots
Wild-type B. amyloliquefaciens 54 and its derivatives that contained gfp reporters were used to view biofilm formation on the root surfaces of tomato plants. Sterilized tomato seeds were grown on solid MS medium for 4 days at 25 • C. Tomato seedlings were then transplanted into MS liquid medium for 48 h at 25 • C with shaking at 60 rpm. Microbe suspensions were then added into the medium with initial OD 600 = 0.2. After a further 48 h culture, the tomato roots were examined for biofilm formation with confocal laser scanning microscopy (CLSM; LSM710 microscope; Carl Zeiss AG, Oberkochen, Germany).

Stomata Assays
To measure the stomatal apertures of tomato plants pre-treated with B. amyloliquefaciens 54 or its derivatives, fully expanded tomato leaves were detached, and then incubated in MES buffer (pH 6.2) for detection [76]. Stomatal aperture images were recorded under microscopy (Zeiss Axio Observe 3m and Zen Standard Software, Carl Zeiss AG, Oberkochen, Germany). For each treatment, at least 60 randomly selected stomata were measured in triplicate.

Leaf ABA Content Determination
To determine ABA content, 1 g fresh leaf tissue was soaked in liquid nitrogen, and then extracted in 80% (v/v) methanol. The extract was centrifuged at 4000× g for 20 min twice, and the supernatant was recovered and evaporated under a vacuum to remove the methanol. Afterwards, ethyl acetate was added to combine with the organic fraction at pH 3.0. Under a vacuum, the dry combined fractions were extracted and dissolved in Tris-buffered saline (150 mM NaCl, 1 mM MgCl2, 50 mM Tris-HCl, pH 7.8), and ABA content was measured by an enzyme-linked immunosorbent assay (ELISA) kit (Agdia, Elkhart, IN, United States) [70].

RNA Isolation, RT-PCR, and qRT-PCR
The total RNA of tomato plant leaves was isolated using a Plant RNA Kit (Tiangen Bio-Tech, Beijing, China). The first cDNA strand from 1 µg total RNA was synthesized using a PrimeScript RT reagent kit (TaKaRa Bio Inc., Kyoto, Japan) and oligo (dT) primers. qRT-PCR was conducted on an ABI 7500 system (Applied Biosystems, Foster City, CA, USA) with the SYBR premix ExTaq mixture (TaKaRa) according to a standard protocol: 95 • C for 10 min, followed by 40 cycles of 95 • C /15 s, 60 • C/20 s, and 72 • C /30 s, and ending with 72 • C for 5 min. In this study, the tomato actin2 gene was chosen as a quantitative control. The relative gene expression was calculated using the 2 −∆∆Ct formula. A detailed description of all PCR primers used in this study is shown in Supplementary Table S1. To detect antioxidant enzyme activities and MDA levels under drought stress conditions, leaves at comparable developmental stages were collected. SOD, POD, CAT, and APX activities were analyzed according to established methods [77]. MDA content was detected as previously described [78]. A total of 0.3 g fresh leaf tissue was homogenized in 50 mM phosphate buffer (pH 7.8). The extract was centrifuged at 8000× g for 15 min, and the supernatant was added with 2.5 times volume thiobarbituric acid (TBA). Afterwards, the mixture was boiled for 20 min and then placed in the ice-bath. The sample was centrifuged for 5 min at 10000× g and the absorbance spectra of supernatant at wavelength 532 nm and 600 nm was monitored. Relative water content (RWC) was measured as previously described [79]. Fresh weight (FW) of tomato leaf was excised from the plants and immediately recorded. The leaf was soaked in distilled water at room temperature with constant light for 4 h, and the turgid weight (TW) was recorded. Dry weight (DW) was recorded after drying at 80 • C for 24 h. RWC was determined using the following formula: RWC (%) = [(FW − DW)/(TW −DW)] × 100. Root vigor was determined according to the triphenyltetrazolium chloride (TTC) method [80]. A total of 0.5 g fresh root tissue was soaked with 5 mL 0.4% v/v TTC and 5 mL 0.06 mol·L −1 phosphate buffer (pH 7.0) and incubated in darkness at 37 • C for 3 h. The reaction was terminated by adding 2 mL of 1 mol L −1 sulfuric acid. The mixture was subsequently centrifuged and absorbance of supernatant was measured at wavelength 485 nm. Root vigor was calculated as the quantity of TTC reduced per gram of root sample per hour (µg g −1 h −1 ).

Statistical Analysis
All experiments were repeated three times under the same conditions, and significant differences were analyzed by Tukey's studentized range (HSD) test (p < 0.05).