Growth of Azospirillum irakense KBC1 on the Aryl β-Glucoside Salicin Requires either salA or salB

Abstract

The rhizosphere nitrogen-fixing bacterium Azospirillum irakense KBC1 is able to grow on pectin and β-glucosides such as cellobiose, arbutin, and salicin. Two adjacent genes, salA and salB, conferring β-glucosidase activity to Escherichia coli, have been identified in a cosmid library of A. irakense DNA. The SalA and SalB enzymes preferentially hydrolyzed aryl β-glucosides. A Δ(salA-salB) A. irakense mutant was not able to grow on salicin but could still utilize arbutin, cellobiose, and glucose for growth. This mutant could be complemented by either salA or salB, suggesting functional redundancy of these genes in salicin utilization. In contrast to this functional homology, the SalA and SalB proteins, members of family 3 of the glycosyl hydrolases, show a low degree of amino acid similarity. Unlike SalA, the SalB protein exhibits an atypical truncated C-terminal region. We propose that SalA and SalB are representatives of the AB and AB′ subfamilies, respectively, in glycosyl hydrolase family 3. This is the first genetic implication of this β-glucosidase family in the utilization of β-glucosides for microbial growth.

β-Glucosidases are a heterogeneous group of enzymes, present in eukaryotic and prokaryotic organisms, which catalyze the hydrolysis of cellobiose and chemically related β-glucosides of which the aglycone can be an aromatic compound, such as in arbutin and salicin. Because of the abundance of cellulose in plant cell walls, most microbial β-glucosidases have been investigated in soil and digestive microflora and have been described as the ultimate enzymatic step in the biological conversion of cellulose into glucose. These β-glucosidases were formerly called cellobiases. Nevertheless, some β-glucosidases preferentially hydrolyze aryl β-glucosides, such as arbutin and salicin.

β-Glucosidase activities have been also found in plant growth-promoting rhizobacteria such as Rhizobium (24), Azoarcus (31), and Azospirillum (3) spp., but their corresponding genes have not been characterized. We investigated the genetic determinants of these enzymes in bacteria of the genus Azospirillum which preferentially colonize the rhizosphere of graminae. In the genus Azospirillum, both A. lipoferum and A. brasilense are known for their morphogenic effects on plant roots and for beneficial effects on plant growth (28). The root-colonizing properties and ability to produce phytohormones such as indole-3-acetic acid have been investigated particularly in A. brasilense (5, 6, 7, 29, 30, 38–40). On the other hand, A. irakense, in contrast, is a low indole-3-acetic acid producer (46) and is unique among the Azospirillum species for its ability to grow on pectin (21).

We report on the characterization of two β-glucosidases in the type strain of A. irakense and the genetic analysis of the two corresponding genes. The two purified enzymes, SalA and SalB, show a higher specific affinity for aromatic β-glucosides, such as salicin and arbutin, than for cellobiose. Analysis of the deduced amino acid sequences of salA and salB allowed their unambiguous classification in family 3 of the glycosyl hydrolases. This is the first genetic implication of this β-glucosidase family in the utilization of β-glucosides as carbon sources.

MATERIALS AND METHODS

Strains, vectors, and growth conditions.

The Escherichia coli and A. irakense strains and plasmids used are listed in Table 1. E. coli strains were grown in Luria-Bertani medium (LB) (33) at 37°C. Azospirillum was grown in LB supplemented with 2.5 mM CaCl₂ and 2.5 mM MgSO₄ (LB*) at 30°C. For solid media, 15 g of agar liter⁻¹ was added. Conjugations were performed on LB* plates, and the A. irakense transconjugants were selected on MMAB minimal medium (41) with 0.5% malate as the C source and 18 mM NH₄Cl as the N source. Antibiotics were used at the following concentrations: ampicillin, 50 μg ml⁻¹; kanamycin, 25 μg ml⁻¹; and tetracycline, 10 μg ml⁻¹. Growth rates in liquid minimal medium (MMAB) supplemented with 18 mM NH₄Cl and with the appropriate C source (cellobiose, salicin, or arbutin) at 8 mM was measured by monitoring the optical density 595 nm. Malate (0.5%) was used as the C source in all Azospirillum precultures.

TABLE 1.

Bacterial strains and plasmids

Strain or plasmid	Relevant characteristics	Reference or source
A. irakense
KBC1	Wild type, LMG 10653, isolated from rice rhizophere	20
FAJ0691	KBC1 (salA salB)::kan Kmr	This work
E. coli DH5α	F80 dlacZΔM15 Δ(lacZYA-argF) recA endA hsdR supE	Gibco BRL
Plasmids
pUC19	Cloning vector, Apr	45
pBluescript SK(−)	Cloning vector, Apr	Stratagene
pSUP202	Mobilizable plasmid, suicide vector for A. irakense, Cmr Tcr Apr	35
pRK2013	Tra+ helper plasmid	11
pLAFR1	IncP1 broad-host-range plasmid, Tcr	12
pLAFR3	pLAFR1 derivative containing the pUC8 multiple cloning site, Tcr	36
pHP45Ω-Km	Kmr cassette	10
pFAJ0650	pLAFR1 with 24-kb EcoRI genomic fragment from A. irakense KBC1	This work
pFAJ0651	pLAFR1 with 15-kb EcoRI insert from pFAJ0650	This work
pFAJ0652	pUC19 with 6.3-kb KpnI-EcoRI insert from pFAJ0651, expressing salA	This work
pFAJ0654	pUC19 with 4.5-kb SphI-EcoRI insert from pFAJ0652, expressing salA	This work
pFAJ0655	pUC19 with 2.7-kb HindIII-EcoRI insert from pFAJ0651	This work
pFAJ0660	pLAFR3 with 5.5-kb BamHI-HindIII insert from pFAJ0651	This work
pFAJ0662	pUC19 with 6.5-kb HindIII-SalI insert from pFAJ0650, expressing salB	This work
pFAJ0663	pUC19 with 1.9-kb PstI insert from pFAJ0650	This work
pFAJ0665	SK− with 2.7-kb HindIII-EcoRI from pFAJ0655 and 0.4-kb EcoRI-PstI insert from pFAJ0663, expressing salB	This work
pFAJ0666	pFAJ0660 with 1.2-kb HindIII fragment containing the 3′ region of salA, expressing salA	This work
pFAJ0667	pFAJ0660 with blunted 2.7-kb HindIII-BamHI insert from pFAJ0665, expressing salB	This work

DNA techniques and sequence analysis.

Standard methods were used for plasmid isolation, chromosomal DNA preparation, transformation, Southern blotting, and hybridization (33). DNA fragments were recovered from agarose gels by using a Nucleotrap kit (Macherey-Nagel). For Southern hybridization, DNA was transferred to a Hybond-N membrane (Amersham), and DNA probe was labelled with digoxigenin-dUTP by using a random labelling kit (Boehringer Mannheim). The hybridization signal was detected with a chemiluminescence detection kit (Boehringer Mannheim). For sequence analysis, recombinant plasmids were purified with a Qiagen kit. DNA sequencing of pUC18/19 subclones, using the chain-terminating dideoxynucleoside triphosphate method (34), was carried out with an AutoRead sequencing kit (Pharmacia-LKB) on an automated sequencer (ALF; Pharmacia-LKB). DNA sequence analysis was done on both strands. Sequence data were processed and analyzed by using the PCGene software package (Intelligenetics). The prediction of the amino-terminal signal sequence was done with the SignalpWWW server (27). The classification of glycosyl hydrolases is available on the ExPASy server (17).

Enzyme purification and N-terminal amino acid sequence analysis.

Overnight cultures of E. coli clones harboring pFAJ0654 (salA gene) or pFAJ0662 (salB gene) were centrifuged and resuspended in 0.5× phosphate-buffered saline (50 mM sodium phosphate, 75 mM sodium chloride [pH 7.2]) supplemented with DNase I (Sigma) at 200 U/ml. The lysis was performed in FastRNA Tubes-Blues (Bio 101) in a FastPrep machine (Bio 101-Savant) for 30 s at speed 6. The lysate was cleared by two centrifugations (5 min at 13,000 rpm) and filtration through Millex 0.22-μm-pore-size filters. For preparation of SalB enzyme, the lysate was cleared by centrifugation for 2.5 h at 120,000 × g at 4°C and concentrated on Microcon10 concentrators (Amicon). The two enzymatic samples (500 μl of extract per run) were applied to a gel filtration column (Superdex 200 Prep Grade HR16/50; Pharmacia), which was eluted with 0.5× phosphate-buffered saline at a flow rate of 1 ml/min. Positive fractions were pooled and adjusted to pH 5.0 (SalA preparation) or pH 4.6 (SalB preparation) with glacial acetic acid and then filtered through Millex 0.22-μm-pore-size filters. Then cation-exchange chromatography was performed with gradients of buffers A (50 mM sodium acetate) and B (50 mM sodium acetate, 1 M sodium chloride) which were adjusted at pH 5 (SalA purification) or pH 4.6 (SalB purification). The SalA preparation was applied to cation-exchange column (S Sepharose High Load XK16/50; Pharmacia). The column was rinsed with 90% buffer A–10% buffer B and eluted with 75% buffer A–25% buffer B. In the case of SalB purification, the sample was loaded onto MonoS HR5/5 column (Pharmacia) and eluted with a gradient of 0 to 25% buffer B at 0.5 ml/min for 25 min. In both cases, the positive fractions were pooled and purity was estimated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). N-terminal amino acid sequence analysis of the purified SalA and SalB was accomplished by Edman degradation using an Applied Biosystems 477A protein sequencer (Applied Biosystems/Perkin Elmer, Foster City, Calif.).

Enzyme assays and zymograms.

To detect β-glucosidase or cellobiohydrolase activity in E. coli clones containing A. irakense DNA, cells were grown overnight on LB plates and then incubated with 4-methylumbelliferyl-β-d-glucoside (MUG; Sigma) or 4-methylumbelliferyl-β-d-cellobioside (MUC; Sigma) in an overlay (0.2 mg ml⁻¹ in 0.7% agarose buffer–0.1 M K₂HPO₄-KH₂PO₄ [pH 7.0]) for 10 min to 1 h at 37°C. Colony fluorescence was detected with a transilluminator at 302 nm.

p-Nitrophenyl-β-d-glucoside (PNPG; Sigma), p-nitrophenyl-β-d-cellobioside (PNPC; Sigma), and p-nitrophenyl-β-d-xyloside (PNPX; Sigma) were also used as synthetic substrates. Ranges of pH from 5.0 to 8.0 and temperature from 30 to 50°C were first tested with PNPG. Then a sample of SalA was added to 1 ml of 50 mM morpholineethanesulfonic acid buffer (pH 6.5) and allowed to equilibrate to 45°C for 5 min. In the case of SalB, N-tris(hydroxymethyl)methyl-2-aminoethanesulfonic acid (TES) buffer was used at 50 mM, pH 7.0. The reaction was started by the addition of a preheated substrate solution to give a final concentration of 10 mM of PNPG, PNPX, or PNPC. Enzymatic reaction was terminated by the addition of 1 ml of 1 M Na₂CO₃, and the absorbance at 420 nm was immediately measured and compared against a p-nitrophenol standard curve. The hydrolysis of cellobiose, gentiobiose, or salicin was measured by glucose production in a total volume of 400 μl. The reaction was terminated by the addition of 400 μl of 0.5 M Na₂CO₃ and neutralized by 200 μl of 1 M HCl. Glucose assays were performed by monitoring the reduction of NADP⁺ to NADPH at 340 nm in the presence of hexokinase and glucose-6-phosphate dehydrogenase (G6PDH). The assays were carried out as follows. A 500-μl reaction mixture containing 100 mM TES buffer (pH 7.4), 8 mM MgCl₂ · 6H₂O, 1.5 mM ATP, 1 mM NADP⁺, and 10 U of a hexokinase-G6PDH coupled enzymatic preparation (Sigma) was added to 400 μl of each stopped and equilibrated sample. The absorbance values were compared to those obtained for a standard curve of glucose. The hydrolysis of arbutin was measured by the release of hydroquinone. The absorbance at 520 nm was measured in an alkaline solution after the addition of 400 μl of 0.5 Na₂CO₃ and compared against a hydroquinone standard curve.

For separation of the SalA and SalB enzymes by isoelectric focusing (IEF), PhastGel IEF in the range of pH 5 to 8 (Pharmacia) was used with a PhastSystem electrophoresis chamber (Pharmacia). Electrophoresis was carried out according to the manufacturer’s instructions. Agarose overlays contained 0.7% agarose in 0.1 M K₂HPO₄-KH₂PO₄ buffer (pH 7.0) and 0.4 mg of MUG per ml. After IEF migration, the IEF gel was covered by a cooled overlay and incubated at 37°C. MUGase activity was detected under UV light at 302 nm.

Nucleotide sequence accession number.

Sequence data have been submitted to the GenBank database under accession no. AF090429.

RESULTS

Cloning of two adjacent loci encoding β-glucosidase activity.

Approximately 3,000 E. coli clones from a genomic cosmid library of A. irakense KBC1 have been tested for β-glucosidase and cellobiohydrolase activities with MUG or MUC, using an overlay technique. Seven clones showed both MUG- and MUC-hydrolyzing activities, and similar EcoRI, BamHI, and HindIII restriction patterns of the corresponding recombinant cosmid DNAs were obtained (data not shown). One of these cosmids, pFAJ0650, was retained for further characterization. The restriction map of the pFAJ0650 cosmid containing a 24-kb insert is shown in Fig. 1. From this cosmid, two groups of MUG/PNPG-positive subclones have been obtained. The first group shared the 4.5-kb SphI-EcoRI fragment (pFAJ0651, pFAJ0652, and pFAJ0654); the second group shared the 3.1-kb HindIII-PstI (pFAJ0662 and pFAJ0665). The 2.7-kb HindIII-EcoRI overlapping fragment (pFAJ0655) from these two subcloning groups was MUG/PNPG negative, suggesting that two loci encoded β-glucosidase activity. Clones of the first group also exhibited PNPX activity. In the E. coli pUC19 subclones, transcription of the salA and salB genes is driven by the lacZ promoter of the pUC19 vector since β-glucosidase activity was observed for only one orientation of the insert DNA.

FIG. 1.

Restriction map of pFAJ0650 cosmid and its derivative subclones harboring the salA or salB gene. The expression of β-glucosidase and β-xylosidase activities in each of the E. coli clones harboring pFAJ0650 or its subclones was tested with the substrates PNPG and PNPX, respectively. B, BamHI; E, EcoRI; H, HindIII; K, KpnI; P, PstI; S, SalI; Sp, SphI.

Sequence analysis of the DNA region encoding β-glucosidase activity.

The region which contained the two putative β-glucosidase encoding loci was sequenced. The deduced amino acid sequence revealed two open reading frames, one from nucleotides 121 to 2319 and the second from nucleotides 2342 to 4291. These open reading frames, named salA and salB, encoded putative proteins of 732 and 649 amino acids, respectively. Sequence analysis did not reveal a Rho-independent transcriptional terminator between salA and salB or downstream of salB. Each putative translation start codon is preceded by a sequence which is similar to the ribosome-binding sequence of gram-negative bacteria.

The N-terminal regions of SalA and SalB displayed characteristics of signal peptides, suggesting transport through the plasma membrane. Cleavage sites predicted by the SignalpWWW server, and confirmed by N-terminal amino acid sequencing of the purified SalA and SalB proteins, occurred between amino acid residues 26 and 27 in SalA and between amino acid residues 21 and 22 in SalB protein. The N-terminal sequences of SalA and SalB are MKVHQLFKAALATSLCLTAFAGGAMA/QAKGAWQNTSL and MRRLPHLSLLALMLYSGTALA/APQQPALPEGQ, respectively (sequenced amino acids of the mature protein are in boldface; slash signs indicate the cleavage site).

Comparison of the deduced amino acid sequences of SalA and SalB with other known β-glucosidases in data banks revealed that SalA and SalB are not similar to the β-glucosidases of glycosyl hydrolase families 1 and 4. However, a good match was found with β-glucosidases of family 3; 36% identity was observed between SalA and Cbg1 of Agrobacterium tumefaciens, and 52% identity was observed between SalB and BgxA of Erwinia chrysanthemi. Identity between SalA and SalB of A. irakense was only 12%. Consistent with these results, the structures of mature SalA and SalB exhibited all of the main traits of the family 3 β-glucosidases. Nevertheless, the SalB protein is, like BgxA of E. chrysanthemi, an atypical member of this family. As for most members of glycosyl hydrolase family 3, the mature SalA and SalB proteins contain an N-terminal catalytic domain (A domain) and a C-terminal domain (B domain) (Fig. 2). In the C-terminal region of the A domain, a putative catalytic aspartate residue is conserved in all proteins of this family (2), as well as in SalA (position 254) and SalB (position 334). However, the C-terminal region of the B domain of SalB is truncated and is very similar to that of BgxA of E. chrysanthemi. In this shorter B domain, the C-terminal region conserved among family 3 members is also present in SalB but appears more compressed than in the other enzymes (Fig. 2). Moreover, the 45 amino acids of the SalB C-terminal region are required for the expression of β-glucosidase activity in E. coli (E. coli with pFAJ0655 was negative for MUG and PNPG activity [Fig. 1]), suggesting a functional role of this conserved region in SalB.

FIG. 2.

Schematic representation of the AB and AB′ organization of SalA and SalB. The A domain of the glycosyl hydrolase family 3 (in grey) is the largest conserved region and contains the catalytic site (triangle). The B domain contains different highly conserved regions (black bars) which are present in SalA as well as in the atypical truncated B domain of SalB.

Purification and characteristics of the two β-glucosidases.

The two β-glucosidases SalA and SalB were purified from E. coli clones harboring either salA in pFAJ654 or salB in pFAJ662. The two β-glucosidases could be distinguished by molecular weight, pI, and substrate specificity. The apparent molecular weight of the purified SalA protein determined by SDS-PAGE was 78,500, consistent with the deduced value of 75,000. In contrast, the SalB protein exhibited a lower molecular weight 64,600, on SDS-PAGE and a deduced value of 66,600. As previously described, the N-terminal amino acid sequences of SalA and SalB were experimentally determined and compared with obtained DNA sequences. The pI values of SalA (6.0) and SalB (5.7) also exhibited an unambiguous difference which allowed their efficient separation by IEF. The substrate specificity of the two β-glucosidases was evaluated in the presence of synthetic (PNPG, PNPX, and PNPC) or natural (cellobiose, gentiobiose, arbutin, and salicin) substrates. The optimal conditions for PNPG hydrolysis by SalA and SalB extract were 45°C and pH 6.5 or 7.0, respectively. With both SalA and SalB, the K_m measurements revealed their greater affinity for natural or synthetic aryl β-glucosides (PNPG, arbutin, and salicin) than for cellobiose or gentiobiose (Table 2). These enzymes are therefore not cellobiase or gentiobiase types of β-glucosidases. SalA also exhibited the capacity to cleave PNPX and PNPC.

TABLE 2.

Biochemical characteristics of SalA and SalB

Substrate	SalA			SalB
	Km	Sp acta	Relative activityb	Km	Sp act	Relative activity
PNPG	0.11	10.80	100	0.05	19.68	100
PNPX	0.40	6.16	57	NDc	—d	ND
PNPC	NDc	0.97	9	ND	—	ND
Salicin	0.40	10.82	100	0.06	17.10	86
Arbutin	0.25	11.28	104	0.06	26.13	132
Cellobiose	16.40	0.08	0.7	5.56	1.61	9
Gentiobiose	11.11	0.34	3	1.12	0.65	3

^aMicromoles of glucose, p-nitrophenol, or hydroquinone (see Materials and Methods for details of experiments) produced per minute per milligram of protein. 

^bPercentage of specific activity in the presence of PNPG. 

^cND, not determined. 

^d—, no activity detected. 

Phenotype and complementation of a salA-salB deletion mutant.

A kanamycin resistance (Km^r) cassette from pHP45Ω-Km was inserted into the blunted ends of pFAJ0654 digested with PstI (Fig. 3). The construction was subcloned into the suicide vector pSUP202 and then transferred into A. irakense KBC1 by a triparental mating. The Km^r transconjugants were purified, and insertion of the Km^r gene in the A. irakense chromosome as well as the absence of pSUP202 were verified by hybridization. The Δ(salA-salB) mutant (strain FAJ0691) was unable to grow on salicin after 24 h of culture at 30°C and showed a slightly delayed growth on arbutin. However, growth on cellobiose was similar to that of the wild-type strain. No effect of this deletion was observed for growth on malic acid, gentiobiose, and glucose (data not shown).

FIG. 3.

Phenotype and complementation of a Δ(salAsalB) A. irakense mutant. (A) Restriction map of the salA salB region of A. irakense and schematic representation of the salA salB deletion constructed by Km^r cassette exchange. (B) Restriction map of A. irakense DNA cloned in the three plasmids: pFAJ0660 (without salA or salB), pFAJ0666 (expressing salA), and pFAJ0667 (expressing salB). The inverted triangle in pFAJ0667 indicates the mutation in salA created by blunted ligation. The genotype and phenotype of each constructed A. irakense strain are also indicated. The dashed line represents the out-of-scale DNA. B, BamHI; Bs, BssHII; Cl, ClaI; E, EcoRI; H, HindIII; P, PstI.

We investigated the ability of each β-glucosidase to restore the Sal⁻ phenotype by complementing FAJ0691 with either salA or salB. First, plasmid pFAJ660, which contains the presumed promoter region of the salA and salB genes (data not shown) and part of SalA, was introduced into FAJ0691. As expected, this plasmid could not restore growth on salicin (Fig. 3). Then plasmid pFAJ0666 was obtained by the addition of the complete 3′ end of salA in pFAJ0660 digested with HindIII. The E. coli clones harboring pFAJ0666 showed both PNPG and PNPX activities, suggesting that SalA was functional. The Δ(salA-salB) mutant which contained pFAJ0666 was able to grow on salicin as the sole carbon source. To test the role of salB, plasmid pFAJ0667 was obtained by the addition of the complete sequence of the salB gene in HindIII-digested and blunted pFAJ0660. By this blunted cloning, a nonsense mutation in salA was created, as indicated in Fig. 3. The presence of salB in pFAJ0667 was first verified by restriction mapping. Second, in situ detection of MUG activity after IEF revealed that the unique β-glucosidase activity of pFAJ0667 comigrated with that of pFAJ0665 (harboring salB) and not with that of pFAJ0654 and pFAJ0666 (harboring salA). Finally, because only SalA exhibited both PNPX and PNPG activities, the lack of PNPX activity in E. coli clones harboring pFAJ0667 was additional evidence that the SalA enzyme is not functional in this construct. Therefore, plasmid pFAJ0667 conferred only SalB β-glucosidase to the host cells. pFAJ0667 was introduced into FAJ0691, and growth on salicin was restored (Fig. 3), suggesting that either salA or salB is required and sufficient for salicin utilization by A. irakense.

DISCUSSION

The rhizosphere nitrogen-fixing bacterium A. irakense KBC1 exhibits pectinolytic activity (21) and is able to grow on several β-glucosides such as cellobiose (20) and, as described in this work, arbutin and salicin. We investigated in this rhizosphere bacterium the genes and encoded enzymes which allow the utilization of such plant-derived β-glucosides. From a cosmid library of A. irakense expressed in E. coli, we identified and characterized at genetic and biochemical levels two β-glucosidases, SalA and SalB, which are required for the growth of A. irakense on salicin.

Based on their deduced amino acid sequences, these β-glucosidases were classified as members of family 3 of the glycosyl hydrolases (15–17). All bacterial β-glucosidases of family 3 are organized into two main domains, a catalytic A domain and a noncatalytic B domain (18). In contrast to the AB organization of SalA and most β-glucosidases of this family, the SalB structure is unusual (Fig. 2). SalB shows an atypical truncated B domain and a high degree of similarity with the β-glucosidase BgxA of E. chrysanthemi in which the truncated C-terminal B domain was renamed B′ (42). Because of the surprisingly high degree of similarity between SalB and BgxA, the hypothesis of gene transfer between these two gram-negative bacteria which exhibit the similar host range of monocotyledonous plants can be proposed. However, the GC contents of salB and bgxA, 65 and 57%, respectively, are quite similar to those of their corresponding host genomes, 66% in A. irakense and 57% in E. chrysanthemi. Because this is the first report of two different copies of family 3 β-glucosidases in the same bacterium, the phylogenetic relationship between some of the best-known bacterial β-glucosidases of this family was evaluated. By using the whole amino acid sequence of these enzymes, their most conserved A domain, or their A domain and the most conserved residues of the B domains, similar phylogenetic trees were obtained (Fig. 4). The SalB of A. irakense and the BgxA of E. chrysanthemi constitute an unambiguous AB′ cluster which is separated from SalA of A. irakense and all the other representative AB bacterial β-glucosidases of family 3. It is of interest that the percentage of similarity between SalB and BgxA of these taxonomically unrelated α and γ subclass proteobacteria is higher (52%) than the similarity between SalA and Cbg1 of the two α subclass proteobacteria A. irakense and A. tumefaciens (36%). This feature could suggest that the amino acid sequence of the AB′ group is highly conserved. In conclusion, analysis of their deduced amino acid sequences indicate that the adjacent salA and salB genes are two paralogous genes which encode AB and AB′ β-glucosidases of glycosyl hydrolase family 3.

FIG. 4.

Phylogenetic tree of several bacterial members of the glycosyl hydrolase family 3. The tree was constructed by the neighbor-joining method. SalB and BgxA constitute an unambigous AB′ cluster, while the AB subfamily seems to be a more heterogeneous group within the glycosyl hydrolase family 3. BglB Clos., BglB of Clostridium thermocellum (13); Cbg1 Agro., Cbg1 of A. tumefaciens (4); SalA Azos., SalA of A. irakense; BglX Esch., BglX of E. coli (44); CelD Pseu., CelD of Pseudomonas fluorescens subsp. cellulosa (32); BgxA Erwi., BgxA of E. chrysanthemi (42); SalB Azos., SalB of A. irakense.

In gram-positive and gram-negative bacteria, the β-glucosidases required for the utilization of β-glucosides belong to family 1 or 4 of the glycosyl hydrolases (17, 37). However, no role in β-glucoside assimilation has been assigned to the β-glucosidase BglX of E. coli K-12 (44), the β-glucosidase BgxA of E. chrysanthemi (42), and the other known bacterial β-glucosidases of family 3. The constructed Δ(salA-salB) A. irakense mutant was not able to grown on salicin but could still utilize arbutin, cellobiose, gentiobiose, and other carbon sources such as glucose or malate. This is the first genetic evidence that β-glucosidases of family 3 are involved in the assimilation of aryl β-glucosides. In A. irakense, the β-glucosidases of family 3 could be evolutionary alternatives to the β-glucosidases of families 1 and 4 in the assimilatory pathways of β-glucosides. The structure of the β-glucosidases implicated in the utilization of arbutin or cellobiose in A. irakense is under investigation.

The presence of signal sequences predicts a periplasmic location of SalA and SalB, as has been suggested for BglX of E. coli (44) and BgxA of E. chysanthemi (42). Indeed, supernatants of A. irakense or E. coli, containing the salA and salB genes of A. irakense, did not have PNPG- or MUG-hydrolyzing activity. This raises the question on how β-glucosides are transported through the outer membrane of gram-negative bacteria. Recently Andersen et al. (1) identified the bglH gene of the E. coli bgl operon, encoding a carbohydrate-specific outer membrane porin with high specificity for arbutin, salicin, gentiobiose, and cellobiose. Whether a similar situation exists for A. irakense remains to be determined.

The Sal⁻ phenotype of the Δ(salA-salB) A. irakense mutant is in agreement with the greater affinities of the SalA and SalB enzymes for aromatic β-glucosides than cellobiose (Table 2). SalB has strict specificity for glucose in the aryl β-glucoside, and its active site appears to exclude substrates of a particular size, explaining the finding of no activity with PNPC but weak activity with cellobiose. SalA has a broader substrate spectrum, including xylose in the β linkage. For the best-known bacterial β-glucosidases of family 3, cellobiose is also a less efficient as a substrate than aromatic natural or synthetic β-glucosides or cellodextrins (13, 14, 23, 32, 43). However, because SalA and SalB exhibited similar K_ms with arbutin or salicin, and because we observed a slightly delayed growth of the Δ(salA-salB) mutant on arbutin (data not shown), these two proteins could be also involved in arbutin utilization. Nevertheless, it is clear that A. irakense contains a second assimilatory pathway for arbutin. This arbutin pathway could to some extent be involved in the utilization of salicin, because of the slight growth of the Δ(salA-salB) mutant on salicin after 36 h (data not shown).

Since either salA or salB can complement the Δ(salA-salB) A. irakense mutant for growth on salicin (Fig. 3), these two AB and AB′ β-glucosidase encoding genes may be functionally redundant for the utilization of salicin. The salA and salB genes most likely are part of an operon because their coding regions are separated by only 22 nucleotides and because the presumed promoter which controls their expression in A. irakense is located at least 2 kb upstream of salA (data not shown). The latter can be deduced from the minimum length of DNA required 5′ upstream of the salA coding region for complementation of the Δ(salA-salB) A. irakense mutant (Fig. 3). Nevertheless, we cannot at this stage exclude the possibility that SalA and SalB have a specific function for the hydrolysis of still unknown plant-derived aromatic β-glucosides. The ability of SalA to hydrolyze a higher range of substrates than SalB may support this hypothesis. Plant-derived aryl β-glucosides or the aglycones have been found to have a role in signalling or saprophytic relationships between plants and bacteria. Arbutin and salicin, but not their aglycones, induce the production of a cyclic lipodepsinonapeptide toxin, called syringomycin, by phytopathogenic strains of Pseudomonas syringae (25). P. syringae cannot grow on arbutin and salicin as sole carbon sources (19). In the case of some virulent Agrobacterium strains, the aryl β-glucoside coniferin, purified from Pseudotsuga menziesii shoot extract, has been identified as the major inducer of a virE-lacZ fusion introduced in these Agrobacterium strains (26). However, most of these strains exhibit high β-glucosidase activity, and one of them, A. tumefaciens B3/73, produces a β-glucosidase which in vitro hydrolyzes coniferin to the common vir inducer coniferyl alcohol (4). In this case, the aglycone could be the true active compound. Finally, the glucose released after biological cleavage of aryl β-glucosides is a suitable carbon source for many bacteria, and this nutritional capacity could contribute to the competitiveness and survival of bacteria in the rhizosphere. This catabolic property has been well studied in the phytopathogen E. chrysanthemi 3665 (8, 9), in the enteric E. coli K-12 (see reference 22 for a review), and in other gram-negative and gram-positive bacteria (37).