Herbaspirillum seropedicae rfbB and rfbC genes are required for maize colonization


E-mail rose.adele@terra.com.br; Tel. (+55) 41 3361 1581; Fax (+55) 41 3361 1578.


In this study we disrupted two Herbaspirillum seropedicae genes, rfbB and rfbC, responsible for rhamnose biosynthesis and its incoporation into LPS. GC-MS analysis of the H. seropedicae wild-type strain LPS oligosaccharide chain showed that rhamnose, glucose and N-acetyl glucosamine are the predominant monosaccharides, whereas rhamnose and N-acetyl glucosamine were not found in the rfbB and rfbC strains. The electrophoretic pattern of the mutants LPS was drastically altered when compared with the wild type. Knockout of rfbB or rfbC increased the sensitivity towards SDS, polymyxin B sulfate and salicylic acid. The mutants attachment capacity to maize root surface plantlets was 100-fold lower than the wild type. Interestingly, the wild-type capacity to attach to maize roots was reduced to a level similar to that of the mutants when the assay was performed in the presence of isolated wild-type LPS, glucosamine or N-acetyl glucosamine. The mutant strains were also significantly less efficient in endophytic colonization of maize. Expression analysis indicated that the rfbB gene is upregulated by naringenin, apigenin and CaCl2. Together, the results suggest that intact LPS is required for H. seropedicae attachment to maize root and internal colonization of plant tissues.


Herbaspirillum seropedicae, which belongs to the β-Proteobacteria (Krieg and Holt, 1984), is a diazotrophic endophyte known to associate with many agriculturally important plants, such as maize, rice, sorghum, wheat and sugar-cane (Baldani et al., 1986; Olivares et al., 1996). This microorganism may stimulate plant growth by producing and secreting phytohormones, protecting the host against pathogenic microorganisms, or by supplying fixed nitrogen to the plant (Baldani et al., 1986; Elbeltagy et al., 2001; Gyaneshwar et al., 2002). The association of H. seropedicae with poaceous crops apparently initiates with attachment of the bacteria to root surfaces, followed by colonization of the emergence points of secondary roots and penetration through discontinuities of the epidermis. Rapid occupation of root intercellular spaces then occurs, along with colonization of aerenchyma, root xylem and aerial portions (James et al., 2002; Roncato-Maccari et al., 2003; Monteiro et al., 2008). Although large numbers of H. seropedicae cells can be found in internal plant tissues (James and Olivares, 1998; James et al., 2002; Roncato-Maccari et al., 2003), the molecular mechanisms of colonization are not yet understood.

Bacterial surface components are involved in early stages of the plant colonization process and the attachment step probably depends on components of the bacterial cell envelope (Gao et al., 2001; Campbell et al., 2002; Jofréet al., 2004; Broughton et al., 2006). Lipopolysaccharides (LPS) are indispensable components of the cell surface of Gram-negative bacteria (Eidels and Osborn, 1971). In Rhizobium, the expression of cell surface antigens is modulated in symbiosis, and several polysaccharides are at least partially modified during the transition from free-living cell to bacteroid form (Broughton et al., 2006). These changes are induced by plant derived compounds, mainly by flavonoids present in root exudates (Duelli and Noel, 1997). LPS mutants of several rhizobia were shown to remain in the infection threads without developing into nitrogen-fixing bacteroids (Noel et al., 1986), and not completing nodule development on many host plants (Noel et al., 1986; Campbell et al., 2002; Broughton et al., 2006). Also, in the non-symbiotic γ-Proteobacteria Pseudomonas putida, the O-antigen from LPS is necessary for efficient colonization of potato roots (de Weger et al., 1989).

Rhamnose is a monosaccharide frequently found as an integral part of LPS, mainly in the O-antigen. The bacterial biosynthetic pathway of rhamnose requires the product of four genes: rfbA, which codes for glucose-1-phosphate thymidylyltransferase, rfbB (dTDP-d-glucose 4,6-dehydratase), rfbC (dTDP-4keto-6deoxy-d-glucose 3,5-epimerase) and rfbD (dTDP-4keto-l-rhamnose reductase). In many bacteria these genes are found in acluster containing other genes necessary for O-antigen biosynthesis (Samuel and Reeves, 2003), suggesting that they are specifically involved in LPS biosynthesis. In support of this hypothesis, the disruption of rhamnose biosynthesis affected LPS composition in many Gram-negative bacteria such as Pseudomonas aeruginosa, Azospirillum brasilense and Azorhizobium caulinodans (Rahim et al., 2000; Gao et al., 2001; Jofréet al., 2004). Furthermore, the host plant colonization patterns of A. brasilense and A. caulinodans were severely affected by disruption of rfb (Gao et al., 2001; Jofréet al., 2004).

The rfb genes were found in the H. seropedicae genome sequence and knockout mutant strains were constructed to determine their role in LPS composition and host interaction. The results clearly show the involvement of rfbB and rfbC in LPS biosynthesis and also suggest that efficient attachment to maize root surface and colonization of internal tissues by H. seropedicae requires intact LPS.


Identification, genomic organization and mutagenesis of H. seropedicae rfb genes

Analysis of the H. seropedicae genome data base (http://www.genopar.org) revealed the presence of four genes coding for proteins similar to RfbB, RfbC and RfbD (Fig. S1), responsible for dTDP-rhamnose biosynthesis (Table S1). In H. seropedicae, two rfbD-like genes were found; the most conserved of them is located 36.3 kb upstream from rfbBC. Upstream from rfbBC was found galE which codes for UDP-glucose 4-epimerase and rfbG (codes for CDP-glucose 4,6-dehydratase). The intergenic regions of this cluster are very short, with rfbB and rfbC overlapping by one base, suggesting that they comprise a single operon. Although a rfbA-like gene was not found in the rfb cluster, there is a gene located 3.8 kb downstream from rfbC with high similarity to a phospho-sugar nucleotidyltransferase (Table S1) whose product could catalyse the first step of the rhamnose biosynthesis pathway.

Analysis with the STRING software (http://string.embl.de) showed that the organization rfbGgalErfbBC is unique among Proteobacteria. The most common organization of microbial rhamnose biosynthesis genes is rfbBDAC. Furthermore, in many microorganisms the genes for the O-antigen biosynthesis pathway are in a cluster containing genes for monosaccharide biosynthesis, glycosyltranferases, ABC-type sugar transport systems, O-antigen polymerase, O-antigen flippase and other related genes (Reeves and Wang, 2006). In H. seropedicae, these genes are dispersed in the genome, comprising small operons such as rfbGgalErfbBC.

To determine the role of H. seropedicae rfb genes in LPS biosynthesis and host interaction, knockout mutant strains were constructed. Plasmids pHSRAMEBB and pHSRAMEBC containing, respectively, the rfbB and rfbC were mutagenized with the EZ:Tn5<TET-1> (Epicentre Laboratories) transposon cassette. The double recombinant strains were named H. seropedicae RAMEBB (rfbB) and RAMEBC (rfbC). To construct of a strain containing reporter–gene fusion, a lacZ::nptI cassette was inserted into the rfbB of the wild type by homologous recombination. A double recombinant strain was isolated and named H. seropedicae LPEB10.

LPS characterization of H. seropedicae wild-type and mutant strains

SDS-PAGE was used to determine the LPS profile of the H. seropedicae strains (Fig. 1). The rfbB and rfbC mutant strains contained quantitatively and qualitatively different LPS profiles when compared with the wild type grown in NFbHPN medium. The wild-type strain profile had strong low-molecular-weight (LMW) bands and a series of high-molecular-weight (HMW) bands in a ladder-like pattern. Comparisons with known LPS profiles from other bacteria suggested that the LMW bands corresponded to the lipid-A plus core oligosaccharide, and the HMW bands to LPS molecules with a different number of O-antigen units linked (Carlson, 1984; Braun et al., 2005). The LPS of the mutant strains apparently lacked the O-antigen, and their lipid-A plus core portion had an electrophoretic shift when compared with wild-type profile.

Figure 1.

Electrophoresis patterns of LPS isolated from H. seropedicae strains. SDS-PAGE was performed with total LPS extracted from wild-type (lane 1), rfbB- and rfbC- mutant strains (lanes 2 and 3 respectively) grown in NFbHPN medium.

GC-MS monosaccharide analysis of the LPS oligosaccharide chain from the wild-type strain of H. seropedicae revealed the presence of rhamnose as the major sugar (35.9%), followed by glucose and N-acetyl glucosamine (Table S2). On the other hand, in the LPS oligosaccharide chain of the rfbB and rfBC strains, heptose and galactose were the quantitatively predominant sugars (Table S2). As expected, rhamnose was not present in the LPS oligosaccharide chain of the mutant strains, nor, interestingly, was N-acetyl glucosamine.

When the mutant strains were grown with rhamnose as the single carbon source, the wild-type LPS phenotype was not restored (data not shown), indicating that this monosaccharide is incorporated into the LPS only when it is biosynthesized by the rfb pathway.

Bacterial attachment and colonization of maize roots

Plant–bacteria interaction assays were performed to compare the endophytic association capacity of the rfbB and rfBC mutant strains with that of the wild type. The number of wild-type H. seropedicae cells attached to the maize root surface was approximately 100-fold higher than that of the mutant strains (Fig. 2). To investigate if these differences influenced the colonization of internal tissues, the number of endophytic bacteria was determined 1, 4, 7 and 10 days after inoculation (d.a.i.) (Fig. 2). Twenty-four hours after inoculation the number of endophytic wild-type bacteria was also 100-fold higher than that of the mutant strains. This 100-fold difference remained 4, 7 and 10 d.a.i.

Figure 2.

Maize root colonization by H. seropedicae wild-type (black bars), rfbB (grey bars) and rfbC (dark grey bars) mutant strains. Results are shown as means of Log10 (number of bacteria g−1 of fresh root) ± standard deviation. Asterisks indicate significant differences at P < 0.01 (Duncan multiple range test) between the wild-type and the mutant strains attachment to maize roots and colonization of inner root tissues. d.a.i., days after inoculation.

H. seropedicae wild-type strain out-competes the rfbB strain for maize colonization

The mutant strains were also tested in competition experiments with the wild type. Herbaspirillum seropedicae RAM4, a strain expressing the Ds-RED-fluorescent protein, was used in these experiments. The colonization pattern of this strain is identical to that of the wild-type SmR1 (Monteiro et al., 2008). Herbaspirillum seropedicae RAM4 and RAMEBB (rfbB:EZ:Tn5<TET1>) strains were incubated with maize roots separately or simultaneously in different proportions [total of 105 colony-forming units (cfu) ml−1] and then the number of cells of each strain attached or in the internal tissues was determined.

In an inoculum proportion of 1:1 with RAM4, the number of RAMEBB (rfbB) cells attached was 50-fold lower than when the strains were inoculated separately (compare Fig. 3B with Fig. 3A). On the other hand, the number of H. seropedicae RAM4 cells attached when co-inoculated was similar to that of the wild type (Fig. 2) and over 1000-fold higher than that of the co-inoculated rfbB mutant strain. Strain RAM4 also out-competed the H. seropedicae RAMEBB for colonization of internal tissues: the number of rfbB strain cells in internal plant tissues was again approximately 1000-fold lower than that of strain RAM4 at 1, 3, 7 or 10 d.a.i. when the strains were co-inoculated (Fig. 3B). Compared with the internal colonization when only the mutant strain was inoculated, competition from the wild type reduced RAMEBB cell count by 50-fold. In contrast, attachment and internal plant colonization by H. seropedicae RAM4 was identical either in the presence or in the absence of competing mutant strain.

Figure 3.

Competition between H. seropedicae RAM4 (black bars) and RAMEB-B (grey bars) strains for maize colonization.
A. The strains were inoculated separately in maize and colonization was followed over time.
B. The strains were co-inoculated at 1:1 proportion in maize and colonization was followed over time.
C. Same as (B), but the RAMEB-B:RAM4 proportion 100:1 was used to inoculate the maize roots.
Results are shown as means of Log10 (number of bacteria g−1 of fresh root) ± standard deviation. One asterisk indicates significant differences at P < 0.01 (Duncan multiple range test) between the competing strains. Two asterisks indicate significant differences at P < 0.01 (Duncan multiple range test) between the mutant strain colonization when it is inoculated separately and when it is co-inoculated.

When a cell proportion of RAM4:RAMEB-B of 100:1 was used to inoculate maize roots, no cells of the mutant strain were recovered from either the root surface or inner tissues. Again, attachment and internal plant colonization by H. seropedicae RAM4 was similar with or without the competing strain (data not shown). When RAM4:RAMEB-B was inoculated at 1:100 proportion (Fig. 3C), the number of attached cells was similar for both strains. One day after inoculation, the bacterial cell number was also similar for both strains. However, 4 d.a.i. the number of RAM4 cells reached 106 g−1 of root which is over 100-fold higher than that of the co-inoculated rfbB mutant strain. In the subsequent days, the number of RAM4 cells remained at the same level. In contrast, the number of RAMEB-B cells decreased, getting over 1000-fold lower than that of the RAM4. It is interesting to note that the number of rfbB strain cells was lower 7 and 10 d.a.i. when co-inoculated with RAM4 than when inoculated separately.

Bacterial attachment to an inert matrix

To test if the attachment phenotype of rfbB and rfbC mutant strains was specific towards the root surface, attachment assays to a glass fibre matrix were also performed. Under the tested conditions, no difference of glass fibre attachment was observed between the wild-type and mutant strains (Table S3), suggesting that the reduced attachment to maize roots depends on recognition of this host and that an important component for recognition is modified in the mutant strains.

Attachment of H. seropedicae to maize root surface is inhibited by glucosamine and isolated LPS

To test the hypothesis that altered LPS was responsible for the decrease in attachment to maize roots we performed attachment assays using purified wild-type LPS (1 mg of glucose equivalent ml−1) as a competitor during incubation of the bacteria with maize roots. Under this condition, the attachment of the wild-type H. seropedicae decreased to levels similar to those of the mutant strains, whereas the attachment of the mutant strains was not affected (Fig. 4). When LPS extracted from the rfbB mutant strain (RAMEB-B) was used as competitor no difference in attachment of any strain was observed, confirming that the wild-type LPS is required to block the attachment sites on the maize root surface.

Figure 4.

Isolated H. seropedicae LPS inhibited attachment to maize root surface. The compounds were added at the moment of inoculation of the maize roots by H. seropedicae wild-type (black bars), rfbB (grey bars) and rfbC (dark grey bars) mutant strains; monosaccharides were added at 1 mg ml−1 and purified LPS at 1 mg of equivalent glucose ml−1. Results are shown as means of Log10 (number of bacteria g−1 of fresh root) ± standard deviation. Asterisks indicate significant differences at P < 0.01 (Duncan multiple range test) between wild-type attachment to maize roots when in the presence or absence of competing compounds.

To determine which LPS monosaccharides are involved in the binding to the plant root surface, we performed attachment assays using an excess of several monosaccharides as competitors. Only glucosamine and N-acetyl-glucosamine (1 mg ml−1) were capable of reducing the attachment of the wild type to similar levels as the mutant strains, while the attachment of the mutant strains was not affected by either sugar (Fig. 4). The attachment assay was also performed using a lower concentration (10 µg ml−1) of N-acetyl-glucosamine. In this biologically relevant concentration, the attachment of the wild type was also reduced to levels similar to that of the mutant strains, while the attachment of mutant strains was not affected. The results strongly support the involvement of N-acetyl-glucosamine in the attachment of H. seropedicae to maize roots.

rfbB or rfbC knockout decreases H. seropedicae resistance to SDS, polymyxin B and to plant metabolites

Since changes in LPS structure often result in increased sensitivity to cationic peptide antibiotics, detergents and other chemical stress (Lerouge and Vanderleyden, 2001; Campbell et al., 2002; Jofréet al., 2004), we tested the sensitivity of the mutant strains to SDS, polymyxin B and the plant-derived compounds naringenin, quercetin and salicylic acid. The rfbB and rfbC mutant strains were very sensitive to SDS. Concentrations as low as 0.5% abolished growth of the mutant strains on LA plates completely, whereas growth of the wild type was not affected by concentrations up to 1% (Fig. 5A). The mutant strains were also significantly (P < 0.05, Student's t-test) more sensitive to lower concentrations of polymyxin B sulfate (Fig. 5B) and salicylic acid (Fig. 5C) compared with the wild type.

Figure 5.

Sensitivity of H. seropedicae wild-type (black squares), rfbB (lozenges) and rfbC (triangles) strains to chemical stress. Equal numbers of cells (approximately 100–1000) were plated on agar plates in the presence of variable concentrations of SDS (A), polymyxin B sulfate (B) and salicylic acid (C). LA medium was used for SDS sensitivity test; and solid NFbHPN was used to test the other compounds. After 24–48 h of incubation at 30°C, the colony-forming units were determined. The results are reported as percentage of colony-forming units relative to the number of colonies grown in the absence of the tested compounds and represent the average of at least three independent experiments.

The flavonoids naringenin and quercetin (up to 500 µg ml−1) had no negative effect on the growth of H. seropedicae mutant strains, nor did sugar-cane extract at concentrations up to 20% (data not shown).

Pleiotropic effects of rfbB or rfbC mutations in H. seropedicae

The knockout of capsular monosaccharides biosynthesis genes is often correlated to alterations in motility, growth rate and modifications in exopolysaccharide production (Ormeño-orrilo et al., 2008). These traits, which may play important roles in plant colonization, were also analysed in the mutant strains.

Herbaspirillum seropedicae rfbB and rfbC mutant strains showed the same motility pattern observed for wild-type strain in semi-solid medium (Fig. S2). The growth curve of H. seropedicae strains in NFbHPN medium revealed that the maximum growth rates (µmax) of the mutants (µmax = 0.170 ± 0.005 h−1) were lower than the wild type (µmax = 0.305 ± 0.003 h−1). Herbaspirillum seropedicae rfbC and rfbB mutant strains also had smaller colonies on LA and NFbHPN agar medium compared with the wild-type strain (data not shown). The mutant strains colonies in NFbHPN agar plates were as gummy as those of the parental strain (Fig. S3A). Moreover, the mutants and wild-type strains had similar fluorescence intensities under UV (365 nm) when grown on NFbHPN agar plates containing 0.02% calcofluor (Fig. S3B), indicating that the production of exopolysaccharides was not affected.

rfbB gene expression is modulated by plant-derived signals

To study rfb expression, we constructed the strain LPEB10 bearing a rfbB::lacZ chromosomal fusion.

The rfbB gene expression in this strain grown under different conditions is shown in Table 1. None of the tested carbon sources (2% malate, rhamnose, mannose or glucose), quercetin (50 µg ml−1), NaCl (100 mM) or MgCl2 (5 mM) affected the expression of rfbB. However, when the medium was supplemented with sugar-cane extract (5%), polymyxin B sulfate (2.5 µg ml−1), salicylic acid (25 µg ml−1) or in the presence of one maize seedling (and its root exudate) per ml of culture, the expression increased twofold compared with the control condition (NFbHPG medium). Addition of calcium chloride (5 mM), naringenin (50 µg ml−1) or apigenin (100 µg ml−1) increased the expression fourfold.

Table 1.  Regulation of H. seropedicae rfbB gene expression.
Conditionβ-Galactosidase activity
SmR1 (wild type)LPEB 10 (rfbB::lacZ::nptI)
  • a. 

    There was a significant difference (P < 0.01; Duncan multiple-range test) between the control and tested conditions of rfbB expression.

  • Values are expressed as nmol ONP min−1 (mg protein)−1 ± standard deviation.

  • Different letters after the values indicate statistic difference between the values (P < 0.01; Duncan multiple-range test).

Control0 ± 0.0 a55.0 ± 1.0 b
Malate (2%)0 ± 0.0 a60.5 ± 5.6 b
Glucose (2%)0 ± 0.0 a58.2 ± 6.9 b
Mannose (2%)0 ± 0.0 a54.3 ± 3.7 b
Rhamnose (2%)1.2 ± 2.5 a51.8 ± 3.5 b
Maize seedling (one seedling ml−1)a0 ± 0.0 a110.2 ± 15.1 c
Sugar-cane extract (5%)a0 ± 0.0 a128.6 ± 9.7 c
Quercetin (50 µg ml−1)5.0 ± 4.3 a54.0 ± 2.7 b
Naringenin (50 µg ml−1)a0 ± 0.0 a220.7 ± 17.2 d
Apigenin (100 µg ml−1)a0 ± 0.0 a227.9 ± 9.3 d
CaCl2 (5 mM)a0 ± 0.0 a216.0 ± 0.2 d
MgCl2 (5 mM)0 ± 0.0 a70.0 ± 8.1 b
NaCl (100 mM)0 ± 0.0 a66.7 ± 8.0 b
Polymyxin B sulfate (5 µg ml−1)a0 ± 0.0 a115.2 ± 0.7 c
Salicylic acid (25 µg ml−1)a0 ± 0.0 a107.2 ± 12.9 c

To test whether the electrophoretic pattern of the LPS was altered under these conditions, the H. seropedicae wild-type strain was grown either in the presence or in the absence of CaCl2 (5 mM), and the total LPS was extracted and separated by SDS-PAGE. The LPS profile of cells grown in the presence of calcium ions showed an increase of moities containing 12, 13, 14 and 15 units of the O-antigen (Fig. S4) as compared with the profile obtained in the absence of CaCl2. This result suggests that the observed increase in rfbB transcription by these signals is related to enhanced production of the O-antigen and incorporation into the LPS and is consistent with an increase in the expression of genes related to LPS oligomerization. Similar results were observed when the H. seropedicae wild type was grown in the presence of naringenin (50 µg ml−1) (data not shown).


In this study we disrupted two H. seropedicae genes, rfbB and rfbC, responsible for rhamnose biosynthesis and showed that they are necessary for LPS biosynthesis and endophytic association.

SDS-PAGE analyses of total LPS showed that the mutant strains apparently lack the O-antigen, and the lipid-A plus core portion of the mutated LPS showed an electrophoretic shift, suggesting that it is was either smaller or had a more negative charge than the wild-type lipid-A plus core portion. Monosaccharide composition analysis of the wild-type LPS showed that rhamnose is the predominant sugar, followed by glucose and N-acetyl glucosamine. The LPS of the mutant strains had a completely different sugar composition from the wild type: heptose and galactose being the quantitatively predominant sugars and rhamnose was not detected. These results show that rfb mutations drastically altered the LPS chemical composition. In P. aeruginosa, rhamnose links the core and O-antigen, and the rfbC mutant strain of this organism synthesized a truncated core which was unable to act as an attachment point for the O-antigen (Rahim et al., 2000). A similar function for rhamnose in H. seropedicae could account for the observed LPS structural reorganization in the mutant strains. In order to investigate if these modifications had biological consequences, the sensitivity of the mutant strains to SDS, polymyxin B and plant metabolites was determined and plant assays were performed to compare endophytic association proficiency.

Since the LPS of the mutants seems to lack the O-antigen it is likely that the changes in the saccharide moiety of LPS may have increased cell surface hydrophobicity and, consequently, decreased resistance to detergents. Unlike the H. seropedicae wild-type strain, the rfbB and rfBC mutant strains were unable to grow in medium supplemented with 0.5% SDS. This effect has also been observed in rfb mutant strains of A. brasilense (Jofréet al., 2004).

Salicylic acid, a plant phenolic metabolite, is a key signal molecule in regulating plant defences in response to a wide variety of pathogens (Martínez-Abarca et al., 1998; Prithiviraj et al., 2005; Stacey et al., 2006). Upon infection, salicylic acid triggers either a localized or systemic-acquired response in which the plant gains long-lived resistance to pathogens (Durrant and Dong, 2004). Salicylic acid can also act as a bactericidal agent (Prithiviraj et al., 2005). Our results revealed that rfbB and rfbC are necessary for the resistance of H. seropedicae to low concentrations of salicylic acid (up to 25 µg ml−1), similar to the concentrations accumulated in bacterial infection sites of plants (16–23 µg ml−1) (Huang et al., 2006). If the LPS acted as a barrier to salicylic acid, preventing it entering into the bacteria, the truncated LPS of the mutant strains would presumably render them more sensitive to this compound. Our results also showed that the mutation in rfbBC increased the H. seropedicae sensitivity to the cationic peptide polymyxin B, suggesting that the intact LPS has a protective effect against this antibiotic. Many plant-derived flavonoids also have bactericidal activity (Xu and Lee, 2001): we tested H. seropedicae sensitivity to quercetin and naringenin at several concentrations with no apparent negative effects on the mutant strains viability under tested conditions.

In vivo assays showed that efficient attachment of H. seropedicae depends on cell surface molecules, since the number of wild-type bacteria attached to maize root surfaces was approximately 100-fold higher than that of the mutant strains. The fact that the number of wild-type cells bound to the roots was similar to the number of mutant cells bound when LPS, N-acetyl glucosamine or glucosamine were used as competitors confirms our hypothesis that the LPS alterations are responsible for the observed phenotypic differences in rfb mutant strains. Our results suggest that H. seropedicae LPS participates in the attachment of the bacteria possibly by anchoring the bacterium to plant receptors by its N-acetyl glucosamine residues. Moreover, LPS seems to have a specific involvement in attachment of H. seropedicae to plant roots, since attachment to glass fibre was not altered by the mutations.

The number of wild-type bacteria colonizing the internal plant tissues was significantly higher (100-fold) than that of the mutant strains. Co-inoculation assays of the wild-type and mutant strains revealed that the defect in rhamnose biosynthesis decreased competitiveness considerably. Competitiveness is an indispensable property for rhizobacteria in their niche, where competition to colonize plants is very high. The decrease in competitiveness of the rfbB mutant strain could be explained by the observed attachment deficiency and by its increased sensitivity to basal plant chemical defences. An alternative explanation is that the presence of bacterial cells induced systemic defence response in the plant, to which the wild-type LPS confers resistance. Thus, the rfbB mutant strain would suffer significantly more growth inhibition in the presence of wild-type strain than in its absence.

The observed decrease in growth rate of the mutant strains in liquid medium could explain in part the colonization and competitiveness phenotype. If this was the case, the number of the mutant strains colonizing the internal tissues should be lower than the wild type at the initial stages, but as the number of cells per gram of tissue stabilizes, the number of cells of the rfb mutants would progressively increases and reach to the same level as the wild type. The comparison of the time-course colonization pattern of the wild-type and mutant strains do not support this assumption: the number of cells of wild type were 106 per gram at 3 d.a.i. whereas the mutant strains stabilizes in 104 per gram even 10 d.a.i. In the competition assays, this trend becomes even clearer: when a ratio of RAM4:RAMEB-B of 1:100 was used to inoculate maize roots, the wild type-like strain reached 106 cells per gram at 3 d.a.i. but the numbers of the mutant strain was still 104 per gram, and even decreased to 103 per gram at 7 d.a.i. These results indicate that the ineffective colonization of the rfbB mutant strain is not a simple effect of lower growth rate.

The involvement of LPS in Rhizobium NGR234–legume symbiosis has been shown previously (Noel et al., 1986; Campbell et al., 2002; Broughton et al., 2006). Also, Jofré and co-workers (2004) showed decreased root maize colonization by A. brasilense rfbD and, recently, Ormeño-Orrilo and colleagues (2008) observed a marked decrease in internal colonization of maize by Rhizobium tropici CIAT899 strains impaired in LPS biosynthesis. Therefore, LPS seems to be a key molecule on the bacterial surface necessary for attachment, penetration and colonization of internal plant tissues by a broad range of endophytes in addition to their role in endosymbiotic associations. It remains to be determined whether their role in these two types of plant–bacteria interactions are the same.

Since intact LPS is required to maximize the H. seropedicae–plant interaction, its biosynthesis might be regulated by plant-derived signals. In Rhizobium NGR234, a very well-studied rhizobacteria, flavonoids exuded by legumes induce the nodulation genes, which promote the production and secretion of lipochitooligosaccharides (Nod factors), leading to the development of infection threads, symbiosomes and root nodules (Broughton et al., 2006; Le Quéréet al., 2006). This communication between Rhizobium NGR234 and their hosts includes various symbiotic factors, such as exopolysaccharides, capsular polysaccharides, the type three secretion system and LPS (Marie et al., 2004; Broughton et al., 2006; Le Quéréet al., 2006). Therefore, we analysed the expression of the rfbB gene in the presence of effectors potentially present in the rhizosphere, using a rfbB::lacZ::nptI strain. Different carbon sources did not induce differential rfbB expression, nor did the flavonoid quercetin affect the expression of this gene. On the other hand, the flavonoids naringenin and apigenin increased rfbB expression fourfold and sugar-cane extract increased expression twofold. Duarte-Almeida and colleagues (2006) showed that sugar-cane extract contains many flavonoids, which could act as inducers of the rfbB gene. Expression was also increased twofold by the presence of maize seedlings; presumably its exudate also contains active factors. Since naringenin and apigenin did not affect the survival of the rfb mutant strains of H. seropedicae, these compounds might signal to the bacteria the presence of the plant host and induce expression of genes necessary for colonization, including rfbB. Interestingly, transcriptional activation of rfbB gene in Rhizobium sp. NGR234 is greatly induced by the flavonoid apigenin, and this induction results in the production of rhamnose-rich LPS, which is required for the release of bacteria from infection threads and for protection of the bacteria against plant defence metabolites (Marie et al., 2004). These results suggest that plant metabolites can modulate the synthesis of bacterial LPS to allow the establishment of the association.

Polymyxin B sulfate and salicylic acid also increased rfbB expression twofold, suggesting that the biosynthesis of LPS may be triggered by chemically aggressive compounds. Some evidence suggests that salicylic acid may directly affect bacterial gene expression: in P. aeruginosa PA14 it was shown to downregulate fitness and virulence factor production, and affect bacterial attachment and biofilm formation at concentrations that did not inhibit growth (Prithiviraj et al., 2005). In Agrobacterium tumefaciens, salicylic acid inhibited the induction of virulence genes (Yuan et al., 2007). The response of bacteria to salicylic acid may be important to guarantee their survival upon entering the plant.

Surprisingly, addition of calcium ions (5 mM) increased H. seropedicae rfbB transcription fourfold. This induction capacity was not caused by osmotic influence or bivalent cations, since NaCl or magnesium ions had no effect on rfbB expression. This result suggests that calcium ions possibly act as an inducer of expression of genes necessary for H. seropedicae colonization. Calcium ions and Ca2+ transporters play an important role as factors of host recognition and specificity in many bacteria (Economou et al., 1990; Ehrhardt et al., 1996; Gehring et al., 1997; Felle et al., 1998; and Broughton et al., 2006). In addition, Ca2+ is involved in a variety of bacterial cellular processes, including cell cycle and division, motility, pathogenesis and chemotaxis (Michiels et al., 2002; Dominguez, 2004). Changes in intracellular Ca2+ concentration also alter stability and activity of several enzymes, thus Ca2+ can act as a metabolic regulator (Rampersaud et al., 1991; Holland et al., 1999; Michiels et al., 2002; Dominguez, 2004). Under natural conditions, H. seropedicae could access calcium ions from the plant apoplast, where its concentration varies from 1 to 10 mM (Bush, 1995; Lecourieux et al., 2006), suggesting that rfbB transcription is upregulated in the bacterium when it enters living plant tissues. An alternative hypothesis is that plant signals could act to increase intracellular Ca2+ concentration triggering transcription of genes involved in plant colonization even under lower extracellular Ca2+ levels. The mechanism of Ca2+-dependent transcription regulation in H. seropedicae remains to be determined, since no putative promoter or regulatory sequence could be identified upstream of the rfbGgalErfbBrfbC gene cluster.

The finding that H. seropedicae rfbB and rfbC knockout decreases endophytic association proficiency suggests that attachment and recognition of the bacteria by the plant involves bacterial surface polysaccharide molecules, most probably LPS. The defect in colonization may be due to a combination of modification of LPS constitution, decreased attachment capacity, and increased susceptibility to plant-derived defence metabolites.

Experimental procedures

Growth of bacterial strains

The bacterial strains and their relevant characteristics are listed in Table 2. Herbaspirillum seropedicae strains were grown at 30°C and 120 rpm in NFbHPN medium (Klassen et al., 1997). Escherichia coli strains were grown at 37°C in LB medium (Sambrook et al., 1989). Antibiotics were added at the following concentrations when required: ampicillin (Ap) 100 µg ml−1; kanamycin (Km) 100 µg ml−1; tetracycline (Tc) 10 µg ml−1; streptomycin (Sm) 80 µg ml−1.

Table 2.  Bacterial strains and plasmids used in this study.
Strain/plasmidsRelevant characteristicsReference
  1. Ap, ampicillin; Cm, chloramphenicol; Km, kanamycin; Sm, streptomycin; Tc, tetracycline; superscript r, resistant.

 E. coli Top 10F-mcrAΔ(mcrr-hsdRMS-mcrBC) ϕ80lacZΔM15 ΔlacX74 deoR recA1 endA1 araΔ139 Δ(ara,leu) 7697 galU galK λ- rpsL nupG λ-Invitrogen, San Diego, CA, USA
 H. seropedicae SmR1Spontaneous Smr, wild typePedrosa et al. (1997)
 H. seropedicae RAM4Ds-RED protein producing strain, Smr, KmrMonteiro et al. (2008)
 H. seropedicae RAMEBBrfbB mutant, Smr, TcrThis work
 H. seropedicae RAMEBCrfbC mutant, Smr, TcrThis work
Plasmids and vectors  
 pHSRAMEBBpUC19 containing H. seropedicae SmR1 rfbB gene, AprThis work
 pHSRAMEBCpUC19 containing H. seropedicae SmR1 rfbC gene, AprThis work
 pKOK 6.1Apr, Cmr, lacZ::Kmr cassetteKokotek and Lotz (1989)
 pUC18/19AprInvitrogen, San Diego, CA, USA

DNA manipulations and mutagenesis

The plasmids used in this study are listed in Table 2. Plasmid and total DNA preparations, agarose gel electrophoresis, restriction endonuclease digestion and cloning were performed according to standard protocols (Sambrook et al., 1989). Bacterial strain transformation was carried out by electroporation with a Gene Pulser apparatus (Bio-Rad Laboratories, Hercules, CA, USA). Southern blots were performed using a radioactive probe labelled with [α-32P]-dCTP, or ‘Gene Images Random Prime DNA Labelling Kit’ in conjunction with ‘Gene Images CDP-Star Detection Kit’, according to the manufacturer's instructions (GE HealthCare, Pollards Wood, UK).

For rfb mutagenesis, plasmids pHSRAMEBB and pHSRAMEBC containing, respectively, the cloned rfbB and rfbC were disrupted by the EZ:Tn5<TET-1> (Epicentre Biotechnologies, Madison, WI, USA) transposon cassette that confers resistance to tetracycline (Tc). These constructions were electro-transformed in H. seropedicae SmR1. The mutant strains were selected and named H. seropedicae RAMEBB (rfbB) and RAMEBC (rfbC). For reporter–gene fusion strains, the lacZ::nptI cassette isolated from pKOK 6.1 (Kokotek and Lotz, 1989) was inserted in the rfbB gene of the plasmid pHSRAMEBB. This construction was then electro-transformed into the H. seropedicae wild type. A rfbB::lacZ::nptI chromosomal strain was selected and named H. seropedicae LPEB 10. Insertion of the cassettes in the genomes of all the mutants and the double-cross-over events were confirmed by Southern blot analyses.

LPS preparation for SDS-PAGE analysis

The total LPS extraction for electrophoretic analysis was performed according to Broughton and colleagues (2006), with some modifications. Briefly, the bacterial cells obtained by centrifuging 2 ml of liquid cultures in NFbHPN medium (OD600 = 1.5) were lysed in 80 µl of lysis buffer (1 M Tris pH 6.8, 4% β-mercaptoethanol, 10% glycerol, 0.005% bromophenol blue) at 100°C for 10 min. After cooling to room temperature, 2% SDS and 0.1 mg ml−1 proteinase K were added, and the mixture incubated at 60°C for 16 h. Finally, two volumes of sample buffer (120 mM Tris pH 6.8, 3% SDS, 9% β-mercaptoethanol, 30% glycerol, 0.03% bromophenol blue) were added. Five microlitres of final mixture was separated by SDS-PAGE (16% acrylamide) and visualized by silver periodate oxidation staining (Tsai and Frisch, 1982).

LPS extraction, purification and monosaccharide composition

Following growth for 18 h in NFbHPN medium bacterial cells were removed by centrifugation (5 min, 8000 g), washed three times with buffered saline and once with distilled water, and then lyophilized. Four and a half grams of lyophilized H. seropedicae strains cells were utilized for LPS extraction by the hot-aqueous phenol method of Westphal and Jann (1965). Both aqueous and phenol phases were dialysed exhaustively in 3500 Da membranes. The fractions were centrifuged to remove insoluble materials and the soluble material containing LPS was lyophilized (Ridley, 2000). Ten milligrams of these fractions were soft-hydrolysed according to Lüderritz and colleagues (1971), to allow the separation of the lipid-A from the oligosaccharide chain. The soluble fraction containing the oligosaccharide chain (500 µg) was then lyophilized, converted to trimethylsilyl methyl glycoside derivatives according to York and colleagues (1985), and analysed by a gas chromatograph HP 5890 GC equipped with a DB-1 capillary column, coupled to an ion trap 5970 MSD mass spectrometer [50°C (2 min) to 160°C at 20°C min−1; to 200°C at 2°C min−1, and then to 250°C at 10°C min−1]. The monosaccharide compositions are expressed as a percentage of the total detected carbohydrate represented by each of the sugar moieties.

Plant assays

Seeds of Zea mays cv. SHS-3031 were surface-sterilized with 6% sodium hypochlorite and 0.01% Tween-20 (USB, Cleveland, OH, USA) solution for 20 min, and shaken in 70% ethanol for 5 min. The seeds were then washed four times with sterile distilled water by shaking and transferred to 96-well deep-well blocks (Greiner Bio-One, Kremsmünster, Austria) containing 3 M filter paper holders and 2 ml of plant medium (Egener et al., 1999) and grown at 25°C with a 12 h light period and 1250 lux illumination provided by white-type fluorescent tubes (Aqua Glo, Tokyo, Japan). After 3 days of growth, each seedling was inoculated with 105 cfu of H. seropedicae strains for 30 min at 30°C and 50 rpm. The bacterial counts were made immediately (attachment) and 1, 4, 7 or 10 d.a.i. (internal colonization). For attachment assays, approximately 0.05 g of fresh root was cut, weighed and washed twice by immersion in sterile saline (0.9% NaCl), and then vortexed vigorously for 20 s in 1 ml of sterile saline. The supernatant was used to determine the number of bacteria attached per gram of fresh maize roots. For internal colonization assays, the roots were cut 1, 4, 7 or 10 d.a.i., weighed and surface sterilized by a 2 min wash with 1% sodium hypochlorite containing 0.01% Tween-20, followed by 2 min in 70% ethanol, and then washed three times with sterile distilled water for 2 min. The samples were then homogenized using a sterile pestle and mortar, and the extracts diluted in 1 ml of sterile saline. The diluted extracts were used to determine the number of bacteria colonizing internal plant tissues. The results reported represent the average of at least three independent experiments.

Competition assays were as described above, but using H. seropedicae RAM4 (wild-type colonization phenotype) and H. seropedicae RAMEBB (rfbB) strains, at 1:1 and 1:100 proportions. The total number of cells in the inoculum was kept at 105 cells ml−1. Determination of cfu was as before and antibiotic resistance and Ds-RED production were used to identify and distinguish the strains. Values represent the average of three independent experiments.

Attachment assays were also performed as described above, but in the presence of purified wild-type or mutant strains LPS (1 mg of glucose equivalents ml−1), or 1 mg ml−1 of monosaccharide solution (mannose, glucose, rhamnose, galactose, glucosamine or N-acetyl glucosamine) as competitors during inoculation of the plant with bacteria. After selection of strong competitors, the assay was repeated using a biologically relevant concentration of these competitors (10 µg ml−1). The results reported represent the average of three independent experiments.

Biofilm formation on glass fibre

Herbaspirillum seropedicae strains were grown in 10 ml of NFbHPN medium with addition of 50 mg of glass fibre, at 30°C and 120 rpm. Glass fibre samples were taken at 6, 12 and 16 h after inoculation. The glass fibre was stained with 200 µl of crystal violet 1%, and washed three times with 0.9% saline solution. Then, 1 ml of absolute ethanol was added to remove the dye, and this solution was used to determine the OD550. The negative control of this procedure was without bacterial inoculation. The values are expressed as OD550 of the samples subtracted from the OD550 of the negative control. The results reported represent the average of three independent experiments. Each assay was performed in duplicate.

Resistance assays

Herbaspirillum seropedicae strains were grown in liquid NFbHPN medium to OD600 ≈ 1.0, and 102–103 cells were plated on NFbHPN agar plates in the presence of variable concentrations of polymyxin B sulfate, salicylic acid, naringenin, quercetin or sugar-cane extract. After 24 h incubation at 30°C, the number of cfu was determined. For SDS resistance assay cells were plated on LA medium and incubated at 30°C for 24–72 h. The results reported represent the average of at least three independent experiments.

Phenotypic characterization of pleiotropic effects

To evaluate growth, H. seropedicae strains were grown at 30°C and 120 rpm in NFbHPN medium, for 16 h, until OD600 ≈ 1. The cultures were then adjusted to OD600 = 0.2 in NFbHPN medium and incubated at 30°C and 120 rpm. A sample was taken every hour from the culture to determine the OD600. The results reported represent the average of two independent experiments. To evaluate motility, 105 cells of H. seropedicae strains were inoculated into the centre of a glass tube, which was partially immersed in NFbHPN medium containing 0.5% agar. The culture was incubated at 30°C and the movement of the cells down the glass tube and formation of surface biofilm in the surrounding medium were monitored over time. Production of EPS was visually evaluated by observing gum production on solid media, and by fluorescence intensity on solid media containing calcofluor (0.02%) under UV light.

rfbB gene expression

The H. seropedicae LPEB10 mutant strain (rfbB::lacZ::nptI) was grown in NFbHP medium, containing glutamate (5 mM) as the nitrogen source (NFbHPG) for 16 h. To evaluate the effect of carbon sources, the cells were collected and the cell densities adjusted to an OD600 = 1.0 in NFbHPG medium containing either malate (2%), glucose (2%), mannose (2%) or rhamnose (2%) as the carbon source, grown for 6 h and then the β-galactosidase activity was determined (Miller, 1972). The effect of added effectors on rfbB expression was tested in a similar fashion: after adjusting the cultures to OD600 = 1.0 in NFbHPG medium the following were added to separate cultures: the flavonoids quercetin (50 µg ml−1), apigenin (100 µg ml−1) or naringenin (50 µg ml−1), CaCl2 (5 mM), MgCl2 (5 mM), NaCl (100 mM), sugar-cane extract (5%), a maize seedling (one seedling ml−1, approximately 50 mg of fresh roots), polymyxin B sulfate (2.5 µg ml−1) or salicylic acid (25 µg ml−1). The cultures were grown for 6 h and the β-galactosidase activity was then measured (Miller, 1972). Protein determination was carried out according to Bradford (1976). The values are expressed as β-galactosidase activity standardized by protein concentration. The results reported represent the average of at least five independent experiments. The control with a maize seedling and without bacteria had no detectable β-galactosidase activity.


This work was supported by the Brazilian agencies CAPES, Instituto do Milênio/CNPq, Fundo Paraná and Fundação Araucária. The authors thank Roseli Prado and Julieta Pie for technical assistance, and William Deakin for critical reading of the manuscript and suggestions.