Bacillus subtilis environmental isolates and B. subtilis NCIB3610 (hereafter referred as 3610) were routinely grown in Luria-Bertani (LB) broth. For colony biofilm development, LBGM (Shemesh and Chai, 2013) and MSgg (Branda et al., 2001) were used. All strains used in this study are listed in Table 1. When necessary, antibiotics were used at the following concentrations: 5 μg ml^-1 chloramphenicol, 0.5 μg ml^-1 erythromycin, 10 μg ml^-1 kanamycin, 12.5 μg ml^-1 lincomycin, 50 μg ml^-1 spectinomycin, and 5 μg ml^-1 tetracycline for B. subtilis strains. Chemicals were purchased from Sigma. Restriction enzymes were purchased from New England Biolabs (NEB). Oligonucleotides were purchased from Integrated DNA Technologies (IDT). DNA sequencing was performed at Genewiz (NJ, USA).

All insertion deletion mutations were generated by long-flanking PCR mutagenesis (Wach, 1996). The mutation was first constructed in the B. subtilis laboratory strain PY79 (Schroeder and Simmons, 2013) and then introduced into various environmental isolates of B. subtilis by genetic transformation following a protocol described previously (Gryczan et al., 1978). Primers used to generate deletion mutations in the pdgS gene and the pgs operon are described in Supplementary Table S1. To construct strains with the P_pgsB-lacZ fusion integrated into the chromosome at the amyE locus, genomic DNA containing P_pgsB-lacZ was prepared from YC1275 that was constructed in a previous study (Gao et al., 2015) and was introduced into different strain backgrounds by transformation. A similar promoter fusion (P_SM21pgsB-lacZ) was created by using the primers P_pgsB-F1 and P_pgsB-R1 (Supplementary Table S1) and the genomic DNA prepared from the B. subtilis strain SM21 due to extreme similarity in the DNA sequences of the pgs promoter region from SM21 and 3610 (Supplementary Figure S1). PCR amplified promoter sequences were then cloned into the vector pDG1730 to make a promoter-lacZ fusion. The integration of the reporter fusion to the amyE locus of the strain SM21 and verification of such integration followed a protocol described previously (Gao et al., 2015)

For comparison of colony mucoidy, B. subtilis strains were streaked out on LB agar plates (+1.5% agar). The LB agar plates were freshly poured, air-dried in the Lamina Flow hood for 40 min, and sealed and kept at 4°C prior to use. After streaking inoculation, the LB plates were incubated at 37°C for 12 h. Images of the colonies were taken using a Leica MSV269 stereoscope. For comparison of colony biofilms, B. subtilis cells were first grown to exponential growth phase in LB broth and 2 μl of the culture was spotted to the biofilm inducing media LBGM or MSgg. The plates were incubated at 30°C for 2–3 days. Images of the colony biofilms were recorded similarly.

The method for γ-PGA isolation was modified from a previously published paper (Stanley and Lazazzera, 2005). Briefly, cells were grown in LB at 37°C for 24 h. The supernatant was collected, brought to pH 2.0 with concentrated sulfuric acid, and then incubated at 4°C overnight. Ethanol was added to the supernatant to a final concentration of 80% and the sample was incubated at -20°C for 10 min for γ-PGA precipitation. Samples were centrifuged at 4°C and 5000 rpm for 30 min. The resulting pellet containing γ-PGA was suspended in dH₂O prior to analysis by gel electrophoresis. The γ-PGA samples were size-fractionated on 8% acrylamide gels. The gels were stained with 0.05% methylene blue in 3% acetic acid for 30 min, and de-stained sequentially in dH₂O.

Assays were conducted as previously described (Chai et al., 2009). Cells were cultured in LB or LBGM medium at 37°C in a water bath with shaking. One milliliter of culture was collected at each indicated time point and cells were centrifuged down at 5000 rpm for 10 min. Cell pellets were suspended in 1 ml Z buffer (40 mM NaH₂PO₄, 60 mM Na₂HPO₄, 1 mM MgSO₄, 10 mM KCl, and 38 mM β-mercaptoethanol) supplemented with 200 μg ml^-1 lysozyme. Resuspensions were incubated at 37°C for 15 min. Reactions were started by adding 200 μl of 4 mg ml^-1 ONPG (2-nitrophenyl-β-D-galactopyranoside) and stopped by adding 500 μl of 1 M Na₂CO₃. Samples were briefly centrifuged down at 5000 rpm for 1 min. The soluble fractions were transferred to cuvettes (VWR), and absorbance of the samples at 420 nm was recorded using a Bio-Rad Spectrophotometer. The β-galactosidase specific activity was calculated according to the equation (Abs₄₂₀/time × OD₆₀₀) × dilution factor × 1000. Assays were conducted at least in triplicate.

Cell colonization in the tomato rhizosphere by a 3610 derivative (CY49), a CYBS54 derivative (CY50) and the corresponding Δpgs mutants (CY150 and CY151) was monitored. The tomato seeds were surface-sterilized with a 30 s treatment in 70% ethanol, followed by 15 min in sodium hypochlorite (10% active chlorine), and by three subsequent washing steps with sterile water for at least 15 min each. The seeds were pre-germinated on sterile moist filter paper and were then incubated in a growth chamber (16:8 h light/dark photoperiod). Tomato plantlets were transplanted 2 weeks later into pots filled with sterilized soil (by autoclave). The pots were arranged in a randomized block design in a greenhouse. The temperature was maintained at 25°C and the photoperiod was 16:8 h light/dark. B. subtilis strains were grown in LB broth at 28°C for 24 h in a shaker at 200 rpm. Cell suspensions were adjusted to a cell density of 5 × 10⁷ cells per ml. Twenty milliliters of the B. subtilis suspension was applied as irrigation to each pot. Rhizosphere samples were collected at 1, 3, 5, 7, 13, 21, and 30 days after inoculation. Four rhizosphere samples were analyzed per treatment at each sampling time. Each rhizosphere sample consisted of the total root system with tightly adhering soil of four individual plants, which were thoroughly mixed and immediately processed for C.F.U. counts. To determine the C.F.U. counts of the inoculants by the plating method, serial dilutions of the cell suspension were spread-plated on LB agar plates with appropriate antibiotics and 100 μg ml^-1 cycloheximide. Plates were incubated at 30°C overnight. The number of individual colonies was counted and the collected data were analyzed.

In our previous studies, we showed that environmental isolates of B. subtilis were capable of robust biofilm formation and that there was a correlation between biofilm robustness and the biocontrol efficacy of those isolates (Chen et al., 2012, 2013). As the starting point of this work, we wanted to test whether those environmental isolates produce γ-PGA, and if so, how γ-PGA production influences biofilm formation. A total of 13 different B. subtilis environmental isolates (Table 1), together with the model strain 3610 (Branda et al., 2001), were chosen due to their higher than average biocontrol efficacy shown in previous studies (Chen et al., 2012, 2013). The B. subtilis 3610 is a commonly used, undomesticated model strain for biofilm studies (Branda et al., 2001). γ-PGA-producing B. subtilis strains are known to form mucoid colonies on LB agar plates (Stanley and Lazazzera, 2005). In multiple previous studies, colony mucoidy was used as a qualitative measurement for γ-PGA production (Stanley and Lazazzera, 2005; Cairns et al., 2013; Chan et al., 2014). We therefore streaked out cells on LB plates and visually compared colony mucoidy after 12 h of incubation at 37°C. As shown in Figure 2A, large variations in colony mucoidy were observed among the tested strains; colonies of some environmental isolates (e.g., SM21, CYBS11, CYBS12, CYBS21, and CYBS54) appeared to be much more mucoid and bigger in size than others (e.g., CYBS14, CYBS22, and CYBS23). Interestingly, the strain 3610 was much less mucoid compared to other environmental isolates (Figure 2A). This result indicated that γ-PGA production varies significantly in different environmental isolates of B. subtilis. To confirm that colony mucoidy was indeed due to γ-PGA production, we introduced a deletion mutation in the pgs operon, which is responsible for the biosynthesis of γ-PGA in B. subtilis, into those isolates. Our results showed that virtually all the deletion mutants had a dry and near flat colony phenotype on LB plates (Figure 2A), confirming that the colony mucoidy was solely due to γ-PGA production. Colonies of strong γ-PGA producers (e.g., CYBS54) often expanded further on the agar plates over time, explaining why those colonies were larger in size (Figure 2A). In extreme cases, colonies became semi-fluidic and eventually covered the entire plate (data not shown), which implies that under certain conditions, secreted γ-PGA may help the colony expand on a solid surface. This is somewhat similar to the observation in a previous report, in which another secreted polymeric substance, an EPS, was shown to facilitate colony expansion on the solid surface due to the buildup osmotic pressure in the extracellular space of a colony (Seminara et al., 2012).

To further compare γ-PGA production, we followed a published protocol to purify secreted γ-PGA from cells grown in LB broth (Stanley and Lazazzera, 2005). We then ran a SDS-PAGE and visualized those γ-PGA molecules by staining the gel with methylene blue. Our results show that preparations from different environmental isolates contained varied amounts of γ-PGA as judged by the methylene blue staining SDS-PAGE (Figure 2B). This observation also largely matched the colony mucoidy phenotype of those isolates (Figure 2A). B. subtilis is known to produce other types of secreted polymers, such as EPSs and TasA amyloid fibers (Kearns et al., 2005; Branda et al., 2006). To test the specificity of this dye-staining-based technique, we also compared the γ-PGA preparations from the two environmental isolates (CYBS26 and CYBS54) and their corresponding double mutants of ΔepsH ΔtasA. Our results showed only a mild decrease in the methylene dye staining in the double mutants of ΔepsH ΔtasA when compared to the preparations from the wild type cells (Supplementary Figure S1). This suggests that the methylene dye staining-based qualitative measurement of γ-PGA production is largely valid.

Although, varied γ-PGA production in different environmental isolates could be due to many different factors, one simple possibility we can think of is the variation in the expression of the pgs operon (e.g., due to altered regulation) in different isolates. To test this hypothesis, we applied a promoter lacZ fusion for the pgs operon [P_pgsB-lacZ; note that the promoter sequence of pgs was amplified by using 3610 genomic DNA as the template (Gao et al., 2015)], and introduced the reporter fusion into those environmental isolates by genetic transformation. We then compared the activities of the reporter fusion in those isolates grown to late exponential phase (OD₆₀₀ = 1) under shaking conditions. Activities of the P_pgsB-lacZ reporter fusion varied in the tested strains. Interestingly, they largely matched their colony mucoidy phenotype with only a few exceptions (Figure 3A). For example, the strain CYBS21 showed much more colony mucoidy than 3610 (Figure 2A), yet the activity of the reporter fusion was only marginally higher in CYBS21 than in 3610 (Figure 3A).

To further explore the cause for the differential production of γ-PGA in different isolates, we sequenced the pgs promoters from five selected environmental isolates of strong (CYBS54 and SM21), intermediate (CYBS9 and CYBS26), or weak (CYBS14) γ-PGA producers based on colony mucoidy (Figure 2A). The sequencing results revealed that in the core regions of the pgs promoters (∼150-bp upstream of the transcription start covering the -10 and -35 motifs, and the putative DegU binding site in the promoters) (Ohsawa et al., 2009), no variations in DNA sequence was seen among those different strains (Supplementary Figure S2). We did observe some sequence variations further upstream and next to an experimentally characterized CcpA binding site (Ishii et al., 2013). Nevertheless, it will not be straightforward to evaluate the potential significance of these variations in differential γ-PGA production since there are cases that same variations appear in both weak and strong γ-PGA producers (Supplementary Figure S2). As one more interesting observation, the weak γ-PGA producing strain CYBS14 contains a few nucleotide variations in the 5′-untranslated leader sequence of the pgs mRNA, which could impact ribosome binding or folding of the mRNA secondary structure. However, and again, this cannot explain why 3610 is also a weak γ-PGA producer but shows no sequence variation in the mRNA leader sequence (Supplementary Figure S2).

To summarize, it is still unclear why γ-PGA production varies significantly in different environmental isolates of B. subtilis. It may be in part due to differential expression of the γ-PGA biosynthesis genes as shown in Figure 3A (even though all the different strains bear the same promoter fusion), since this result largely matches the colony mucoidy phenotype. This may in turn be due to different activities of the regulatory proteins involved in the pgs operon regulation, or due to sequence variations in the promoter regions, which somehow impact transcriptional activation of the pgs operon by those proteins. Alternatively, γ-PGA production also depends on the activity of a depolymerase PdgS, whose gene is regulated separately from that of the pgs operon (Figure 1B; Suzuki and Tahara, 2003).

The regulation of γ-PGA production in B. subtilis is still not completely understood. Previous studies showed that the pgs biosynthesis genes were activated by DegU, which was in turn activated by the ComQXPA quorum-sensing system (Figure 1B). The colony of the ΔdegSU double mutant was completely flat with little mucoidy on the LB plate (Figure 4B). Interestingly, when working on the B. subtilis model strain 3610, we frequently observed that the spo0A mutant had a colony mucoidy phenotype on the LB plate (Figure 4A), suggesting that Spo0A may negatively regulate γ-PGA production. This colony mucoidy phenotype was completely reversed in the Δspo0AΔpgs double mutant, confirming that Spo0A regulates the pgs genes on colony mucoidy (Figure 4A). Spo0A is collectively activated by multiple sensory histidine kinases directly or indirectly (from KinA to KinD, Figure 1A), whose activities are linked to environmental or plant signals (López et al., 2009; Chen et al., 2012; Beauregard et al., 2013; Shemesh and Chai, 2013). We thus tested whether any of the histidine kinase mutants had a mucoid colony phenotype. Indeed, both the ΔkinC and ΔkinD mutants in 3610 formed more mucoid colonies on the plates when compared to 3610, especially the double mutant of ΔkinCΔkinD (Figure 4A). On the contrary, the ΔkinA or ΔkinB deletion mutation had little influence on the colony phenotype (Figure 4A). Spo0A∼P represses the abrB gene (Greene and Spiegelman, 1996). A double mutant of Δspo0AΔabrB also lacked the colony mucoidy phenotype, indicating that AbrB is downstream of Spo0A in this regulation. We further ruled out the possibility of a masked effect from overexpression of AbrB-repressed matrix genes since the ΔabrBΔepsH double mutant still showed a flat colony on the LB plate.

To summarize, our results indicate that upon sensing environmental signals, KinC and KinD indirectly and negatively regulate γ-PGA production by first activating Spo0A, which subsequently down-regulates AbrB activities. AbrB is a transition stage regulator and a known transcription repressor (Chumsakul et al., 2011). It will be interesting to learn how AbrB positively regulates γ-PGA production. The KinC/KinD-mediated signaling mechanism was previously shown to activate the biofilm pathway in B. subtilis; a double mutant of ΔkinCΔkinD is severely defective in biofilm formation (Shemesh and Chai, 2013). Thus both biofilm formation and γ-PGA production are co-regulated by the KinC/KinD-mediated signal transduction mechanism but in an opposite fashion.

YwcC is a TetR-type transcription repressor (Kobayashi, 2008; Chai et al., 2009). Previous studies showed that in B. subtilis, YwcC negatively regulates matrix genes and thus biofilm formation via SlrA and SlrR, two antagonist proteins of the biofilm master repressor SinR (Figure 1A; Kobayashi, 2008; Chai et al., 2009, 2010). Interestingly, we found that the ΔywcC mutant in 3610 also had a colony mucoidy phenotype (Figure 4B), indicating that like KinC/KinD and Spo0A, YwcC is also involved in co-regulation of both matrix genes and genes for γ-PGA biosynthesis. Again, the colony mucoidy of the ΔywcC mutant was not likely due to overexpression of YwcC-controlled matrix genes, since the colony of the ΔywcCΔepsH double mutant was still very much mucoid when compared to the ΔepsH mutant (Figure 4B). On the other hand, colonies of the ΔywcCΔepsHΔpgs triple mutant completely lost colony mucoidy, suggesting that the colony mucoidy phenotype seen in the ΔywcCΔepsH double mutant was due to over-production of γ-PGA (Figure 4B).

YwcC negatively regulates the slrR gene, which encodes a key regulatory protein controlling matrix production (Kobayashi, 2008; Chai et al., 2009). To test whether SlrR is involved in γ-PGA production controlled by YwcC, we introduced a ΔslrR deletion mutation into the double mutant of ΔywcCΔepsH. The triple mutant formed colonies that were less mucoid than that of ΔywcCΔepsH (Figure 4C), indicating that SlrR may lie downstream of YwcC in regulating γ-PGA production and colony mucoidy. To further confirm that SlrR (positively) regulates γ-PGA production, we applied a ΔslrR derivative with an inducible slrR gene (under the control of the hyperspank promoter) at an ectopic locus (Chai et al., 2010). When the inducer IPTG was added to the LB agar media, the resulting colonies were more mucoid than the ones in the absence of IPTG (Figure 4C). This confirms that SlrR is a positive regulator for γ-PGA production. To conclude, we believe that YwcC and SlrR mediate another pathway co-regulating both biofilm matrix genes and the γ-PGA biosynthetic genes. The activity of YwcC is predicted to be regulated by binding to a small ligand (a chemical signal) from the environment or plant host (Kobayashi, 2008; Chai et al., 2009). While the signal is unknown, we speculate that it may be important for both biofilm formation and γ-PGA production.

Finally, to test whether the colony mucoidy phenotype seen in various mutants of B. subtilis 3610 is caused by altered expression of the γ-PGA biosynthetic genes in those mutants, we introduced a promoter-lacZ transcriptional fusion for the pgs operon (P_pgsB-lacZ) into various mutants and performed assays of the β-galactosidase activities for the resulting strains (Gao et al., 2015). As shown in Figure 3B, the ΔkinCΔkinD double mutant, the Δspo0A mutant, and the ΔywcC mutant all had higher activities of the transcriptional fusion whereas the ΔdegUS mutant and the ΔabrBΔepsH mutant showed a reduced activity of the reporter (Figure 3B; Ohsawa et al., 2009; Chan et al., 2014; Marlow et al., 2014). The ΔslrR mutant did not show a clearly reduced activity in the reporter fusion (Figure 3B). However, it is known from previous studies that in LB, the slrR gene is expressed weakly and only in a small proportion of cells (Chai et al., 2010). Indeed, our unpublished RNA-seq results suggest in SlrR-overproducing (P_hyspank-slrR) cells, the pgs operon is among the highly induced genes (data not shown). Overall, the colony mucoidy phenotype in different mutants of 3610 matched well the levels of the expression of the pgs operon in those strains.

Given that γ-PGA is produced in abundance by many environmental isolates of B. subtilis (Figure 2A) and that the same isolates are capable of forming robust biofilms (Chen et al., 2013), it is reasonable for us to ask whether γ-PGA production contributes to biofilm formation in those strains. Although, several previous studies already investigated the possible role of γ-PGA in biofilm formation in B. subtilis (Stanley and Lazazzera, 2005; Branda et al., 2006; Morikawa et al., 2006), due to different strains, medium conditions, and experimental settings used in those studies, the conclusions were inconsistent. Therefore, it remains unclear whether γ-PGA is part of the biofilm matrix and whether its production is important for biofilm formation. Here, we tested the colony biofilm phenotype for the 13 wild strains and the corresponding Δpgs mutants in the biofilm-inducing medium (LBGM) (Shemesh and Chai, 2013). Most environmental isolates as well as B. subtilis 3610 formed colony biofilms with distinct surface architectures (Figure 5A). Many of the corresponding Δpgs mutants, however, formed colony biofilms with less structural complexity (e.g., CYBS11, CYBS13, CYBS22, and CYBS25). In the well-studied B. subtilis model strain 3610, lack of complex surface features is often used as an indication of weaker or defective biofilms. For example, the colonies biofilms formed by the eps or tasA mutants that are deficient in biofilm matrix production are also featureless (Figure 5B). Our results thus suggest that in general, γ-PGA production plays a notable role in increasing complex morphology and robustness of colony biofilms in many of the tested environmental isolates, although less important than the EPSs and TasA fibers. As one more piece of evidence, the colony biofilm by the pdgS mutant of B. subtilis 3610 (which is deficient in production of the γ-PGA depolymerase and shown to have a higher yield of γ-PGA) demonstrated an increased structural complexity (Figure 5B), similar to what was seen in the ΔsinR mutant (Subramaniam et al., 2013). In terms of the putative role of γ-PGA in biofilm formation, since there is no evidence that γ-PGA can influence expression of the matrix genes such as the epsA-O and tapA operons (data not shown), we speculate that secreted γ-PGA may become part of the biofilm matrix or facilitate assembly of other matrix components. In a previous study (Branda et al., 2006), it was shown that the TasA fibers and the EPS can complement extracellularly in that a mixture of the ΔtasA and ΔepsH mutant cells are capable of forming wild type-like biofilms. We did similar complementation experiments by mixing the pgs mutant cells with an equal number of either ΔtasA or ΔepsH cells (Supplementary Figure S3). Our results seem to indicate that like TasA and EPS, γ-PGA can also be shared by cells within the biofilm.

A puzzle thus arose since in a previous study (Branda et al., 2006), it was reported that the Δpgs mutant of B. subtilis 3610 showed no difference in the biofilm phenotype from that of the wild type strain. However, it is important to note that biofilm assays in that study were conducted in MSgg, another commonly used biofilm-inducing medium for B. subtilis (Branda et al., 2001). We repeated the assay for B. subtilis 3610 and the corresponding Δpgs mutant in MSgg. Not surprisingly, little difference in colony biofilm phenotype was observed between the two strains in MSgg (Figure 6A). We also repeated this test by using SM21, a stronger γ-PGA producer (Figure 2A) and its corresponding Δpgs mutant. The result was similar although this time a mild difference was seen between the two strains (Figure 6A). We already showed that in LBGM, the difference in colony morphology and robustness was quite clear between these two wild type strains and the corresponding Δpgs mutants (Figure 5A). It seems to us that the role of γ-PGA in biofilm formation in tested B. subtilis strains somehow depends on what media we use for the biofilm assays.

Although, the above results may address much of the discrepancy in previous results about the role of γ-PGA in biofilm formation, the cause of the observed medium-dependence is still unknown. One possibility could be that in MSgg, the expression of the pgs operon is significantly lower than in LBGM and thus the importance of γ-PGA in B. subtilis biofilm formation diminishes in MSgg. To test that, we performed two assays, comparing both the pgs gene expression and γ-PGA production in LBGM and MSgg using two selected strains (SM21 and 3610). In the first experiment, we prepared secreted γ-PGA from cells grown in either LBGM or MSgg and examined the samples on SDS-PAGE similarly as described above. Our results confirmed that cells of the strong γ-PGA producer SM21 produced much higher amounts of γ-PGA in LBGM than in MSgg, while the weaker γ-PGA producer 3610 did not seem to have a strong production of γ-PGA in either medium (Figure 6B). In the second experiment, we tested whether the pgs genes were differentially expressed in LBGM and MSgg. A promoter-lacZ reporter fusion (P_SM21pgsB-lacZ) was similarly constructed and introduced into SM21 as described in a previous study (Gao et al., 2015) except that this time the promoter sequence was PCR amplified using SM21 genomic DNA as the template. As shown in Figure 6C, in SM21, the pgs reporter fusion were expressed significantly higher in LBGM (unfilled squares) than in MSgg (filled diamonds). A similar difference in B. subtilis 3610 between the two different media was previously reported by us (Gao et al., 2015). In conclusion, we believe that γ-PGA production plays a notable role in biofilm formation in B. subtilis and that more importantly conditions strongly influencing the yield of γ-PGA production may dictate the importance of γ-PGA in B. subtilis biofilm formation.

The physiological function of γ-PGA in B. subtilis has been investigated in previous studies (Ogunleye et al., 2015) but remains unclear. Since B. subtilis is a soil bacterium and can secrete γ-PGA in large quantities, we suspect that one of the important functions of γ-PGA may have to do with environmental fitness of the bacterium or interactions with the plant, the natural host of B. subtilis in the rhizosphere. This is analogous to the role of γ-PGA in B. anthracis in bacteria–host interactions and virulence (Jang et al., 2011). We further postulate that γ-PGA may directly contribute to colonization of plant roots by the bacteria or do so indirectly through increasing biofilm robustness [In a previous study, we demonstrated that biofilm formation plays an essential role in root colonization by B. subtilis cells (Chen et al., 2013)]. To test our hypothesis, we performed colonization assays on the tomato plant root using the wild type strains and the Δpgs mutants from both a 3610 derivative (a weak γ-PGA producer, Figure 2A) and a CYBS54 derivative (a strong γ-PGA producer, Figure 2A). The assays were done over a period of 30 days under green house conditions by following a protocol that we established in the previous study (Chen et al., 2013). On various days following bacterial inoculation, root-associated bacterial cells were isolated and total C.F.U was counted by plating. Our results showed that for the CYBS54 derivative (a stronger γ-PGA producer), the wild type cells had a 12-fold higher attachment efficiency than the Δpgs mutant at the end of the experiment (Figure 7A). For the 3610 derivative, the number of the wild type cells attached to the roots was only marginally higher than that of the Δpgs mutant (Figure 7B). This indicates that for strong γ-PGA-producing strains, γ-PGA may play an important role in bacterial attachment to the plant root surface.