APPLIED AND ENVIRONMENTAL MICROBIOLOGY,
`May 2001, p. 2276–2283
`0099-2240/01/$04.0010 DOI: 10.1128/AEM.67.5.2276–2283.2001
`Copyright © 2001, American Society for Microbiology. All Rights Reserved.
`
`Vol. 67, No. 5
`
`Intra- and Extracellular b-Galactosidases from Bifidobacterium bifidum
`and B. infantis: Molecular Cloning, Heterologous Expression,
`and Comparative Characterization
`PETER L. MØLLER,† FLEMMING JØRGENSEN, OLE C. HANSEN,
`SØREN M. MADSEN, AND PETER STOUGAARD*
`Biotechnological Institute, DK-2970 Hørsholm, Denmark
`
`Received 30 October 2000/Accepted 27 February 2001
`
`Three b-galactosidase genes from Bifidobacterium bifidum DSM20215 and one b-galactosidase gene from
`Bifidobacterium infantis DSM20088 were isolated and characterized. The three B. bifidum b-galactosidases
`exhibited a low degree of amino acid sequence similarity to each other and to previously published b-galac-
`tosidases classified as family 2 glycosyl hydrolases. Likewise, the B. infantis b-galactosidase was distantly
`related to enzymes classified as family 42 glycosyl hydrolases. One of the enzymes from B. bifidum, termed
`BIF3, is most probably an extracellular enzyme, since it contained a signal sequence which was cleaved off
`during heterologous expression of the enzyme in Escherichia coli. Other exceptional features of the BIF3
`b-galactosidase were (i) the monomeric structure of the active enzyme, comprising 1,752 amino acid residues
`(188 kDa) and (ii) the molecular organization into an N-terminal b-galactosidase domain and a C-terminal
`galactose binding domain. The other two B. bifidum b-galactosidases and the enzyme from B. infantis were
`multimeric, intracellular enzymes with molecular masses similar to typical family 2 and family 42 glycosyl
`hydrolases, respectively. Despite the differences in size, molecular composition, and amino acid sequence, all
`four b-galactosidases were highly specific for hydrolysis of b-D-galactosidic linkages, and all four enzymes were
`able to transgalactosylate with lactose as a substrate.
`
`Since they were first discovered by Tissier (33), the bi-
`fidobacteria have been investigated extensively by several sci-
`entists (e.g., references 23 and 27). In recent years, bifidobac-
`teria have attracted particular attention due to their promising
`health-promoting properties, for example, reduction of harm-
`ful bacteria and toxic compounds in the intestine, prevention
`of dental caries, reduction of total cholesterol and lipid in
`serum, and relief of constipation (2, 5, 10, 17, 36, 41). There-
`fore, live probiotic bifidobacteria, which may improve the mi-
`crobial balance of the human gastrointestinal tract, have been
`used to supplement dairy products for many years. Another
`approach to increase the number of beneficial bacteria in the
`human intestine is to selectively stimulate their growth by sup-
`plementing food with ingredients which can only be metabo-
`lized by such bacteria. Certain oligosaccharides, the so-called
`prebiotics, have been shown to exert this growth-stimulating
`effect on probiotic bacteria, including bifidobacteria.
`So far, most of the probiotic bacteria and the prebiotic
`oligosaccharides have been used in combination with dairy
`products, and since these products often contain large amounts
`of lactose, much attention has been focused on the enzyme
`b-galactosidase (EC 3.2.1.23), which is involved in the bacterial
`metabolism of lactose. In addition to normal hydrolysis of the
`b-D-galactoside linkage in lactose, some b-D-galactosidase en-
`zymes may catalyze the formation of galactooligosaccharides
`
`through transfer of one or more D-galactosyl units onto the
`D-galactose moiety of lactose. This transgalactosylation reac-
`tion (12) has been shown to be a characteristic of b-galactosi-
`dase enzymes from a great variety of bacterial and fungal
`species (7, 19, 21, 40).
`Galactooligosaccharides produced from lactose by trans-
`galactosylation specifically stimulate growth of bifidobacteria
`(39), and recently Van Laere et al. (37) have described a novel
`b-galactosidase from Bifidobacterium adolescentis that pref-
`erentially hydrolyzes galactooligosaccharides. Therefore, it
`is generally accepted that a structural and catalytic character-
`ization of the b-galactosidase enzymes of probiotic bacteria is
`of central importance for an understanding of their health-
`promoting effects.
`Lactose hydrolysis and transgalactosylation properties of the
`enzyme have been studied in several probiotic bacteria includ-
`ing bifidobacteria (6, 7, 25, 26, 30, 35, 37), but so far only one
`DNA sequence of a bifidobacterial b-galactosidase gene has
`been published (Bifidobacterium longum; EMBL accession no.
`AJ242596) (24), and another sequence has been deposited in a
`database (Bifidobacterium breve; EMBL accession no. E05040).
`In this paper, we describe the molecular cloning, sequencing
`and characterization of three different b-galactosidase en-
`zymes from Bifidobacterium bifidum (DSM20215) and one en-
`zyme from Bifidobacterium infantis (DSM20088).
`
`MATERIALS AND METHODS
`
`* Corresponding author. Mailing address: Biotechnological Insti-
`tute, Kogle Alle´ 2, DK-2970 Hørsholm, Denmark. Phone: 45 45160444.
`Fax: 45 45160455. E-mail: pst@bioteknologisk.dk.
`† Present address: Department of Dairy and Food Science, The
`Royal Veterinary and Agricultural University, DK-1958 Frederiksberg
`C, Denmark.
`
`Bacterial strains, plasmids, and culture conditions. Fifteen different bifido-
`bacterial strains were purchased from Deutsche Sammlung von Mikroorganis-
`men und Zellkulturen GmbH, Braunschweig, Germany, and analyzed for their
`ability to synthesize galactooligosaccharides. Two strains, B. bifidum DSM20215
`and B. infantis DSM20088, which were able to synthesize oligosaccharides, were
`selected for this study. The strains were grown anaerobically using TPY medium
`
`2276
`
`

`

`VOL. 67, 2001
`
`b-GALACTOSIDASES FROM BIFIDOBACTERIA
`
`2277
`
`(28) at 37°C with BBL GasPak Anaerobic systems (Becton Dickinson and Co.,
`Cockeysville, Md.). DNA cloning was performed using the Escherichia coli
`strains (i) MT102, a derivative of MC1000 (hsdR-K12) (4), (ii) XL-1–5, an F2
`derivative of XL1-Blue (3), and (iii) ER1458 (22). The E. coli strains were grown
`in Luria-Bertani (LB) medium (18) supplemented with 100 mg of ampicillin/ml
`and 40 mg of 5-bromo-4-chloro-3-indolyl-b-D-galactopyranoside (X-Gal)/ml
`when appropriate. The plasmid pBluescript KS(2) (Stratagene Cloning Systems,
`La Jolla, Calif.) was the vector used for DNA fragment cloning.
`Chemicals and enzymes. Chemicals were purchased from Sigma Chemical Co.
`(St. Louis, Mo.). Restriction enzymes and other enzymes used for DNA manip-
`ulation were from New England Biolabs, Inc. (Beverly, Mass.) and were used
`according to the instructions of the manufacturer.
`Isolation of b-galactosidase genes. Chromosomal DNA was prepared from a
`cell pellet harvested from 500 ml of TPY culture. The cells were resuspended in
`4.4 ml of lysis solution (20 mM Tris-HCl, 20 mM MgCl2, 20% glucose, 5 mg of
`lysozyme/ml, and 350 U of mutanolysin/ml [pH 6.5]) and incubated at 37°C for
`60 min. Cells were lysed by the addition of 12 ml of TEN buffer (100 mM
`Tris-HCl, 1 mM EDTA, and 100 mM NaCl [pH 7.5]), 1.6 ml of 10% sodium
`dodecyl sulfate (SDS), 1.6 ml of 0.5 M EDTA (pH 7), and 0.3 ml of proteinase
`K (20 mg/ml), followed by incubation at 37°C for 60 min. Five milliliters of
`phenol and 5 ml of chloroform were added, and the extraction was repeated until
`the water phase could easily be separated from the interphase. The genomic
`DNA was precipitated with isopropanol, resuspended in 10 mM Tris-HCl–1 mM
`EDTA (pH 8.0), and treated with RNase. The genomic DNA was then digested
`with restriction enzymes, ligated into pBluescript KS(2), digested with the same
`enzymes, and treated with alkaline phosphatase. Digestion of B. bifidum genomic
`DNA was performed using BamHI, EcoRI, HindIII, PstI, SacI, KpnI, ApaI, and
`SalI, whereas B. infantis DNA was digested with KpnI. Ligation mixtures were
`used to transform E. coli MT102, and b-galactosidase-producing clones were
`identified as blue colonies on X-Gal-containing plates.
`Preparation of cell lysates. E. coli cells harboring the recombinant b-galacto-
`sidase genes were lysed with a French pressure cell. Harvested cells from a
`750-ml culture of E. coli ER1458 (optical density at 450 nm 5 1) were washed
`with 50 ml of 50 mM sodium phosphate–10 mM MgCl2 (pH 6.8) and then
`resuspended in 7 ml of the same buffer. The French pressure cell was operated
`at 196 MPa. Alternatively, the cells were resuspended in Z buffer (60 mM
`Na2HPO4, 40 mM NaH2PO4, 10 mM KCl, 1 mM MgSO4, and 50 mM b-mer-
`captoethanol [pH 7.0]), mixed with glass beads (33% [vol/vol], 212 to 300 mm;
`Sigma), and lysed in an ultrasonic bath by incubation twice for 5 min. Cell debris
`was removed by centrifugation, and the supernatant was used directly for enzyme
`activity measurements and enzyme characterization.
`Assays for b-galactosidase activity. Hydrolysis of o-nitrophenyl (ONP)-b-D-
`galactopyranoside at 37°C and pH 7.0 followed by measurement of absorbance at
`420 nm was used for determination of b-galactosidase activity (18). Assays were
`performed with Z buffer (for b-galactosidases BIF1, BIF2, and INF1) or Z buffer
`containing 0.5% Triton X-100 (for BIF3), and the reactions were stopped by the
`addition of 1 M Na2CO3. Transgalactosylation assays were performed with 0.4 M
`lactose, 50 mM Na citrate, and 100 mM Na2HPO4 (pH 6.0), and the 50-ml
`reaction volumes were incubated for approximately 20 h at 40°C. A 5-min
`incubation at 95°C was used to stop the enzyme reaction. The reaction mixtures
`were analyzed by thin-layer chromatography on Silica Gel 60 plates (Merck) in
`a solvent containing butanol, 2-propanol, and water (3:12:4). Samples of 1 ml of
`diluted sample (1:1 dilution in water) were subjected to three runs. After being
`dried, the sugars were visualized by spraying with an orcinol reagent, followed by
`incubation at 100°C for 5 to 10 min.
`Molecular mass determination. The native molecular mass of b-galactosidases
`expressed in E. coli was determined by analytical gel filtration on a Superdex 200
`HR 10/30 column (Pharmacia) followed by b-galactosidase assay of collected
`fractions. The molecular mass markers used for calibration of the column were
`thyroglobulin (669 kDa), ferritin (440 kDa), human immunoglobulin G (160
`kDa), transferrin (81 kDa), and ovalbumin (43 kDa). The molecular mass of the
`b-galactosidase subunits was determined by SDS-polyacrylamide gel electro-
`phoresis (SDS-PAGE) (15). Samples were reduced with dithiothreitol and
`loaded onto 7.5 or 10% minigels which were stained with Coomassie brilliant
`blue or silver stained.
`Enzyme purification. The BIF2 b-galactosidase was purified from E. coli
`extract by anion-exchange chromatography on a 5-ml HiTrap Q column (Phar-
`macia) at pH 7.5, followed by gel filtration on a Sephacryl S-200 HR column
`(Pharmacia, 1.6 by 60 cm). The INF1 b-galactosidase was purified from E. coli
`extract by gel filtration on a Superdex 200 HR 10/30 column (Pharmacia) fol-
`lowed by anion-exchange chromatography at pH 7.5 on a Mini-Q column (Phar-
`macia). The BIF3 b-galactosidase was purified from E. coli extract by gel filtra-
`tion on a Superdex 200 HR 10/30 column, run on a SDS–7.5% polyacrylamide
`
`gel as described above, transferred to polyvinylidene difluoride (PVDF) mem-
`brane (Immobilon P; Millipore), and stained with Coomassie brilliant blue.
`Digestion of BIF3 polypeptide bound to PVDF with endoproteinase Lys-C and
`extraction of the resultant proteolytic peptides were performed as described (8).
`The peptide fragments were purified by reversed-phase chromatography on a
`SMART system (Pharmacia) equipped with a mRPC C2/C18 SC2.1/10 column
`(Pharmacia) using a gradient of 0 to 80% acetonitrile in 0.1% trifluoroacetic
`acid. Peptide sequencing was performed as described below. The expression level
`of BIF1 was too low to permit purification. Therefore, crude cell extract con-
`taining the BIF1 enzyme was used to determine the molecular weight by gel
`filtration.
`N-terminal amino acid sequence analysis. Enzyme samples were run on SDS-
`polyacrylamide gels as above, transferred to PVDF membrane (Problott; PE
`Biosystems), stained with Coomassie brilliant blue, and analyzed by Edman
`degradation with a protein microsequencer (Procise; PE Biosystems).
`DNA sequence analysis. DNA sequencing was carried out using Cy-5-labeled
`primers. Vector specific primers, T3 and T7, were used in the first sequencing
`reaction mixture, followed by reactions with sequence-specific primers. The
`reaction mixtures were run with an ALF Express sequencer (Pharmacia). Data-
`bases were searched for homologous proteins with the BLAST facility (1). Com-
`parison of amino acid sequences was performed as a BestFit analysis with the
`Wisconsin software package, version 10.0 (Genetics Computer Group, Madison,
`Wis.). The gap creation and gap extension penalty parameters for the BestFit
`analysis were 8 and 2, respectively. Alignment of b-galactosidase protein se-
`quences was performed with the ClustalX program (34). The aligned sequences
`were subsequently imported into the PAUP 4.0b4 program (32), where phylo-
`genetic trees were generated using a neighbor-joining algorithm.
`Nucleotide sequence accession numbers. The four b-galactosidase sequences
`were deposited in the EMBL nucleotide sequence database with the accession
`numbers AJ272131 (BIF1), AJ224434 (BIF2), AJ224435 (BIF3), and AJ224436
`(INF1).
`
`RESULTS
`
`Isolation of b-galactosidase genes from B. bifidum DSM20215.
`Genes encoding b-galactosidase from B. bifidum DSM20215
`were cloned by shotgun cloning. Chromosomal DNA was iso-
`lated, cut with restriction enzymes, and inserted into cloning
`vectors as described in Materials and Methods. Ligation mix-
`tures were transformed into b-galactosidase-deficient E. coli
`cells, and b-galactosidase-producing transformants were iden-
`tified on X-Gal indicator plates. Ligation mixtures with PstI-
`restricted B. bifidum DSM20215 chromosomal DNA gave rise
`to five positive blue clones out of approximately 1,500 screened
`transformants, and mixtures with KpnI-restricted DNA result-
`ed in one positive clone out of approximately 600 transfor-
`mants. Restriction enzyme analysis indicated that four of the
`five PstI clones were identical, whereas the fifth PstI clone was
`different from the four identical clones. The PstI clones were
`denoted pBIF1 and pBIF3, respectively, and the single KpnI
`clone was denoted pBIF2. The positions of the b-galactosidase
`genes on the cloned fragments were examined by subcloning
`the inserts of the plasmids pBIF1, pBIF2, and pBIF3, respec-
`tively, as described below.
`Plasmid pBIF1 contained a 7.6-kb insert. Deletion of 3 kb
`from one end of the fragment to a BamHI site (Fig. 1) and 1.8
`kb from the other end to an EcoRI site totally eliminated
`b-galactosidase activity measured on X-Gal indicator plates.
`Another plasmid construct, in which a 1-kb PstI-to-KpnI frag-
`ment was deleted, showed increased b-galactosidase activity,
`indicating that the KpnI site was close to the structural b-ga-
`lactosidase gene. Therefore, this deletion mutant was chosen
`as an anchor during DNA sequencing by so-called primer
`walking (Fig. 1). The resulting 3.5-kb DNA sequence con-
`tained an open reading frame with a coding capacity of 1,020
`
`

`

`2278
`
`MØLLER ET AL.
`
`APPL. ENVIRON. MICROBIOL.
`
`ing frame of 690 amino acid codons corresponding to a mo-
`lecular mass of 77 kDa (EMBL accession number AJ224436).
`Characterization of b-galactosidase genes from B. bifidum
`and B. infantis. A comparison of the open reading frames in
`plasmids pBIF1, pBIF2, pBIF3, and pINF1 showed only a
`minor degree of protein sequence similarity between the en-
`coded b-galactosidases (Fig. 2). Especially, the pINF1 se-
`quence seemed to be distantly related to the other three genes
`and to the E. coli lacZ gene, as indicated by the short stretches
`of homology that the BestFit analysis returned (data not
`shown). Despite the fact that all four genes are bifidobacterial
`b-galactosidase genes and that three of them were derived
`from the same strain, they showed surprisingly little resem-
`blance to each other.
`Database searches with full-length amino acid sequences as
`queries showed homology to other known b-galactosidase se-
`quences. The highest scores of amino acid identity found by
`comparison to other b-galactosidase sequences deposited in
`databases were 36, 53, 35, and 50% for the pBIF1, pBIF2,
`pBIF3, and pINF1 reading frames, respectively. Subsequences
`around the catalytic domains—corresponding to the sequence
`around the glutamic acid residue at position 461 in the E. coli
`lacZ gene—were selected from 30 b-galactosidases previously
`deposited in databases and aligned with the four sequences
`from this work. The alignment shown in Fig. 2A was subse-
`quently used to generate a phylogenetic tree by neighbor-
`joining analysis. The resulting tree (Fig. 2B) showed that the
`INF1 b-galactosidase from B. infantis was located in a group of
`enzymes previously designated as family 42 glycosyl hydrolases
`(shown as class I in Fig. 2) (http://afmb.cnrs-mrs.fr/;pedro/
`CAZY/db.html and reference 11) and that the three b-galac-
`tosidases, BIF1, BIF2, and BIF3, from B. bifidum belonged to
`the group of enzymes classified as family 2 glycosyl hydrolases
`(shown as class II in Fig. 2). However, a closer examination of
`the alignments showed that BIF3 was deeply rooted within the
`group of family 2 glycosyl hydrolases and that the enzymes
`BIF1 and BIF2 were only distantly related to other family 2
`glycosyl hydrolases. A phylogenetic analysis using full-length
`sequences confirmed the results obtained with the catalytic
`domain subsequences (data not shown). As expected, the
`alignment analysis placed the four known 6-phospho-b-galac-
`tosidases (family 1) in the same subgroup (class III in Fig. 2).
`The b-galactosidase encoded by the pBIF3 sequence was
`found to be quite large compared to what is normally observed
`for lacZ group enzymes. Sequence analysis showed that the
`homology to known b-galactosidases was located in the N-
`terminal part of the reading frame, whereas no homology be-
`tween the C-terminal half of the BIF3 enzyme and other b-ga-
`lactosidases could be detected. A separate BLAST search with
`the C-terminal part revealed homology to enzymes known to
`contain a galactose binding domain, e.g., sialidase from Mi-
`cromonospora viridifaciens (9), galactose oxidase from Dacty-
`lium dendroides (14) and sialidase from Clostridium septicum
`(EMBL accession no. X63266). As shown in Fig. 3, the amino
`acid residues known to bind galactose in sialidase and galac-
`tose oxidase are conserved in the BIF3 sequence, implying that
`the BIF3 b-galactosidase contains a galactose binding site.
`Signal peptide prediction using the SignalP program de-
`scribed by Nielsen et al. (20) (http://www.cbs.dtu.dk/services/
`SignalP/) showed that the first 32 amino acid residues of the
`
`FIG. 1. Map of plasmids pBIF1, pBIF2, pBIF3, and pINF1. Black
`boxes indicate the cloning vector pBluescript KS(2), grey boxes show
`the position of b-galactosidase genes, and white boxes symbolize
`cloned sequences outside the b-galactosidase genes. Restriction en-
`zyme sites used for mapping the b-galactosidase genes are shown
`above the maps, and the sites used as anchors in DNA sequencing by
`primer walking are indicated by arrows below the maps.
`
`amino acid codons corresponding to a molecular mass of ap-
`proximately 112 kDa (EMBL accession number AJ272131).
`Plasmid pBIF2 was similarly subcloned in order to map the
`position of the b-galactosidase gene. b-Galactosidase activity
`was mapped to a 4.3-kb NruI-to-BamHI fragment in the mid-
`dle of the original 13.6-kb KpnI fragment. Further subcloning
`of a NruI-to-EcoRI fragment resulted in transformants without
`b-galactosidase activity, indicating that the EcoRI site was lo-
`cated in the b-galactosidase gene (Fig. 1). Therefore, DNA
`sequencing of the b-galactosidase gene on plasmid pBIF2 was
`initiated at the EcoRI site and completed by primer walking.
`The resulting 3,700-bp DNA sequence contained an open
`reading frame of 1,044 amino acid codons corresponding to a
`polypeptide with a molecular mass of 117 kDa (EMBL acces-
`sion number AJ224434).
`Plasmid pBIF3, containing an insert of approximately 20 kb,
`was further subcloned and the b-galactosidase activity of trans-
`formants harboring the deletion plasmids was determined.
`Since cleavage at one of the internal KpnI sites in the 20-kb
`fragment abolished b-galactosidase activity, this site was cho-
`sen as an anchor for DNA sequencing (Fig. 1). The resulting
`5.5-kb DNA sequence contained an open reading frame of
`1,752 amino acid codons corresponding to a molecular mass of
`188 kDa (EMBL accession number AJ224435).
`Isolation of b-galactosidase genes from B. infantis DSM20088.
`Genes from B. infantis DSM20088 encoding b-galactosidase
`were isolated as described above for B. bifidum. Chromosomal
`DNA was restricted with KpnI, inserted in a cloning vector, and
`transformed into b-galactosidase-deficient E. coli cells as de-
`scribed in Materials and Methods. Nine b-galactosidase pro-
`ducing clones out of approximately 5,000 transformants were
`isolated. DNA sequencing showed that all the clones were
`identical. One of the clones, pINF1, was selected for further
`analysis. DNA sequencing of a 4.3-kb KpnI fragment by primer
`walking from the ends of the fragment revealed an open read-
`
`

`

`VOL. 67, 2001
`
`b-GALACTOSIDASES FROM BIFIDOBACTERIA
`
`2279
`
`FIG. 2. Amino acid sequence comparison of active-site regions in
`selected b-galactosidases. (A) Sequences corresponding to the region
`around the catalytic Glu461 in the lacZ enzyme of E. coli were aligned
`using the ClustalX program. The BIF1, BIF2, BIF3, and INF1 se-
`quences were obtained in this study. The other sequences are identi-
`fied by database accession numbers. The conserved glutamic acid res-
`idue (E) is shown in frames, and conserved residues within classes I, II,
`and III are shaded. (B) Neighbor-joining analysis of the alignment
`shown in Fig. 2A. The Sulfolobus sequences were used as an outgroup.
`Results from a bootstrap analysis (n 5 100) are shown for the junctions
`with a value above 80.
`
`BIF3 reading frame constituted a potential signal peptide. N-
`terminal protein sequencing of BIF3 b-galactosidase expressed
`in E. coli confirmed that the predicted signal peptide was
`indeed cleaved off when BIF3 was expressed in E. coli (see
`
`below). The three other b-galactosidase sequences showed no
`signs of a signal peptide, when analyzed similarly (data not
`shown).
`The untranslated sequences (UTS) (20) upstream of the
`
`

`

`2280
`
`MØLLER ET AL.
`
`APPL. ENVIRON. MICROBIOL.
`
`FIG. 3. Identification of a galactose binding domain in BIF3. The amino acid sequence of the BIF3 b-galactosidase (this study) was aligned to
`(i) galactose binding domains in sialidase from M. viridifaciens (accession no. D01045) and galactose oxidase from D. dendroides (accession no.
`M86819), (ii) sialidase from C. septicum (accession no. X63266), and (iii) a protein of unknown function from Streptomyces coelicolor (accession
`no. AL031155). Amino acid residues, which have been found by X-ray crystallography to interact with the bound galactose moiety in sialidase from
`M. viridifaciens, are marked with an asterisk below the sequences.
`
`open reading frames of the b-galactosidase genes were exam-
`ined for putative transcription and translation signals. UTS
`from other bifidobacterial genes were compared to the UTS
`from the four b-galactosidase genes described here, but no
`obvious transcription initiation signals were identified. How-
`ever, when sequences immediately upstream of the translation
`initiation ATG codon were compared, potential base pairing to
`the 39-end of Bifidobacterium 16S rRNA was evident (Fig. 4).
`Heterologous expression of B. bifidum and B. infantis b-ga-
`lactosidases in E. coli. The Bifidobacterium b-galactosidase
`genes on plasmids pBIF1, pBIF2, pBIF3, and pINF1 were
`expressed in E. coli under growth conditions which would nor-
`mally repress expression from the inducible E. coli lacZ pro-
`moter located in the flanking region in the cloning vector. This
`observation indicated that endogenous, internal bifidobacterial
`sequences upstream of the b-galactosidase genes may serve as
`transcription initiation signals in E. coli. Similarly, initiation of
`translation may be facilitated by the putative E. coli ribosome
`binding sites (AGGA) (Fig. 4) which were observed immedi-
`ately upstream of the open reading frame in all the b-galacto-
`sidase genes. The b-galactosidase activity of E. coli cells ex-
`pressing BIF1, BIF2, BIF3, or INF1 was exclusively found in
`cell extracts (BIF1, BIF2, and INF1) or in cell extract and
`membrane fraction (BIF3), whereas no activity was found in
`the growth medium.
`The native molecular masses of the recombinant b-galacto-
`sidases were determined by gel filtration, and the subunit sizes
`were determined by SDS-PAGE of the purified enzymes
`and/or calculated from the DNA sequence (Table 1). The
`native molecular mass of the BIF1 b-galactosidase was 620
`kDa, and the size of the open reading frame corresponded to
`a subunit molecular mass of 112 kDa (Table 1). Taken to-
`gether, these data suggest that BIF1 is a hexameric b-galacto-
`sidase.
`Recombinant BIF2 b-galactosidase produced in E. coli ex-
`hibited a native molecular mass of approximately 236 kDa and
`a subunit size of approximately 130 kDa. Since the length of
`the open reading frame in the DNA sequence corresponded to
`117 kDa, the BIF2 b-galactosidase is probably a dimeric en-
`zyme. The BIF2 b-galactosidase was purified from E. coli and
`subjected to N-terminal sequence analysis. The N-terminal
`amino acid sequence (MNTTDDQRKN) confirmed that the
`predicted open reading frame of the DNA sequence was ac-
`tually translated.
`The BIF3 b-galactosidase produced in E. coli showed a
`native molecular mass of approximately 180 kDa when soni-
`cation with glass beads was used for cell lysis. Homogenization
`with a French press, however, led to a molecular mass of 360
`kDa for the active enzyme. In both cases, the subunit size
`
`determined by SDS-PAGE was approximately 182 kDa. Since
`both extraction procedures resulted in an enzymatically active
`BIF3 enzyme, the BIF3 b-galactosidase is apparently active as
`a dimeric molecule and also—contrary to almost any other
`b-galactosidase—as a monomeric molecule. Further experi-
`ments are required, however, to determine whether the en-
`zyme exists as a monomer or a dimer in vivo. The N-terminal
`amino acid sequence of BIF3 b-galactosidase produced in
`E. coli was VEDATRSDSTTQMS. This sequence corre-
`sponded to the sequence V33-E34-D35-A36 found in the N-
`terminal part of the BIF3 open reading frame, implying that
`the first 32 amino acids of the BIF3 b-galactosidase constitute
`a signal peptide which is cleaved off posttranslationally. The
`processing site observed between amino acid residues Ala32
`and Val33 is identical to the one predicted by the SignalP
`computer program (20). The exceptionally large open reading
`frame of BIF3 was further verified by amino acid sequence
`analysis of internal peptide fragments derived from the re-
`combinant enzyme. BIF3 b-galactosidase was purified to
`homogeneity and digested with endoproteinase Lys-C. The
`peptide fragments were separated by high-pressure liquid
`chromatography, and selected peptide peaks were then ana-
`lyzed with a protein sequencer. Six peptide sequences, span-
`ning a wide range of the BIF3 amino acid sequence, were
`identified (F73-Q82, M384-H393, W433-N442, I906-S915, T1317-
`Q1326, and V1418-T1425). All the peptide sequences completely
`matched the amino acid sequence deduced from the DNA
`sequence, thus confirming that the large BIF3 reading frame
`was indeed translated.
`Recombinant INF1 b-galactosidase expressed in E. coli
`showed a native molecular mass of approximately 140 kDa,
`and SDS-PAGE of the purified enzyme indicated a subunit
`molecular mass of 73 kDa, which is in agreement with the
`subunit size of 77 kDa predicted from the DNA sequence.
`Therefore, we conclude that the INF1 b-galactosidase is prob-
`
`TABLE 1. Structural properties of recombinant B. bifidum and
`B. infantis b-galactosidases produced in E. coli
`
`Enzyme
`
`Molecular massa determined by:
`
`Gel
`filtration
`
`SDS-
`PAGE
`
`DNA
`sequence
`analysis
`
`Proposed
`molecular
`structure
`
`Signal
`peptide
`
`BIF1
`BIF2
`BIF3
`
`INF1
`
`620
`236
`180 and 360
`
`140
`
`ND
`130
`182
`
`73
`
`112
`117
`188
`
`77
`
`Hexamer
`Dimer
`Monomer and
`dimer
`Dimer
`
`2
`2
`1
`
`2
`
`a Molecular mass values are in kilodaltons. ND, not determined.
`
`

`

`VOL. 67, 2001
`
`b-GALACTOSIDASES FROM BIFIDOBACTERIA
`
`2281
`
`FIG. 4. Comparison of UTS immediately upstream of the ATG start codon in Bifidobacterium genes and the 39-terminal consensus sequence
`in Bifidobacterium genes encoding 16S rRNA. Potential base pairing to the 39-terminal rRNA sequence is indicated by underlined nucleotides.
`Possible sequences facilitating ribosome binding in E. coli are shaded in the BIF1, BIF2, BIF3, and INF1 sequences. The boldface type indicates
`the ATG start codon. The sequences are identified by database accession numbers.
`
`ably a dimeric enzyme. N-terminal amino acid sequence anal-
`ysis of the INF1 enzyme showed the sequence AQRRAHR-
`WPK, which perfectly matched residues 2 to 10 in the amino
`acid sequence predicted from the DNA sequence. The protein
`sequence analysis confirmed that the open reading frame of
`the DNA sequence was indeed translated and indicated that
`Met1 was cleaved off during expression of the enzyme in E. coli.
`Substrate specificity and transgalactosylation properties.
`The substrate specificity of the four b-galactosidases was ex-
`amined in enzyme assays with the following chromogenic sub-
`strates: ONP-b-D-galactose, ONP-b-D-glucose, ONP-b-D-xy-
`lose, ONP-b-D-fucose, and ONP-b-D-galactose-6-phosphate.
`All the b-galactosidases predominantly hydrolyzed ONP-b-D-
`galactose, and less than 10% of the activity observed with
`ONP-b-D-galactose was measured with the other substrates.
`Transgalactosylation assays with total cell
`lysates from
`E. coli strains harboring the plasmids pBIF1, pBIF2, pBIF3,
`and pINF1 were performed as described in Materials and
`Methods. As shown in Fig. 5, all four b-galactosidases were
`able to synthesize galactooligosaccharides. The observed ratio
`between oligosaccharides and monosaccharides suggested that
`the BIF1 and BIF2 enzymes were superior to BIF3 and INF1
`with respect to transgalactosylation.
`
`DISCUSSION
`
`In this paper, we present three new b-galactosidase genes
`from B. bifidum DSM20215 and one new b-galactosidase gene
`from B. infantis DSM20088. In a previous report, Dumortier et
`al. (7) described the purification of one of the three b-galac-
`tosidases that they found in B. bifidum DSM20215, and they
`characterized the transgalactosylating activity of the enzyme.
`We have now extended these studies by isolation, cloning,
`and sequencing of the three different b-galactosidase genes of
`B. bifidum, followed by heterologous expression in E. coli and
`characterization of the individual enzymes. Although the mo-
`lecular structure and the amino acid sequence of the three
`
`enzymes differed widely, they were all highly specific for hy-
`drolysis of the b-D-galactosidic bond in lactose as judged from
`assays with ONP substrates. Furthermore, we found that all
`three b-galactosidases were able to catalyze the formation of
`galactooligosaccharides by transgalactosylation. This result,
`however, does not correspond to the data presented by Du-
`mortier et al. (7), who found that B. bifidum DSM20215 con-
`tains one transgalactosylating and two hydrolysing b-galacto-
`sidases. The discrepancy might be explained by the use of
`different assay conditions, e.g., pH and lactose concentration.
`The BIF3 enzyme presented here probably corresponds to the
`transgalactosylase characterized by Dumortier et al. (7), since
`almost identical molecular weights were observed.
`The enzymes BIF1 and BIF2, described in this study, both
`have a multimeric subunit structure which is very similar to
`other family 2 b-galactosidases, e.g., a subunit molecular mass
`of approximately 115 kDa, no signal sequence, and no carbo-
`hydrate binding domain. In contrast, BIF3 is very different
`from other enzymes classified as family 2 glycosyl hydrolases.
`This b-galactosidase is active as a monomeric 180-kDa mole-
`cule and contains a putative signal sequence which is cleaved
`off when the enzyme is expressed heterologously in E. coli,
`resulting in transport of the enzyme to the periplasma, either
`in a free form or bound to the membrane. Generally, attempts
`to translocate the E. coli lacZ protein to the extracellular space
`or to the periplasma of E. coli have failed so far, probably
`because of membrane jamming or periplasmic toxicity (16, 29,
`31). However, since th