INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Bifidobacteria are important components of the human intestinal microflora, in which they occur at concentrations of 10⁹ to 10¹⁰ cells/g of feces (45), and of fermented milk products, to which they are added mainly because of their supposed health-promoting activities (2, 20). However, molecular genetic research on bifidobacteria is still in its infancy. In addition to many 16S rRNA sequences, only six genes (see Table 6) and two plasmids of members of the genus Bifidobacterium have been cloned and sequenced. Only recently has development of cloning vectors and transformation techniques started (19).

One important metabolic activity which is exhibited by almost all bifidobacteria is deconjugation of bile salts, which occurs naturally in the intestines of humans (44). The responsible enzyme, bile salt hydrolase (BSH) (cholylglycine hydrolase; EC 3.5.1.24), catalyzes the hydrolysis of glycine- and/or taurine-conjugated bile salts into amino acid residues and deconjugated bile salts (bile acids). The function of this enzyme in the producing bacterium is not known, although various observations have been made and various theories about its role in the producer have been proposed; these theories include a role in utilization of the liberated amino acid or a role in increased resistance to the toxic levels of bile salts in the gastrointestinal environment (10). A recent screening analysis of more than 300 lactic acid bacterial strains performed in our laboratory showed that BSH activity is found primarily in organisms isolated from the gastrointestinal tracts of mammals (Bifidobacterium spp., Lactobacillus acidophilus, Lactobacillus gasseri, Lactobacillus johnsonii, some strains of Lactobacillus plantarum), while organisms isolated from fermented milk preparations and vegetables (Lactococcus lactis, Lactobacillus delbrueckii, Lactobacillus helveticus, Streptococcus thermophilus) do not exhibit BSH activity (44).

In a mammalian host deconjugation of bile salts takes place in the small and large intestines. The exact location of this process depends on the distribution of the bacterial flora in the host species; for example, in mice a Lactobacillus flora is present in the small intestine, while in humans a significant flora starts only at the end of the ileum and is fully developed in the large intestine. This means that bile salt deconjugation in mice starts in the small intestine (47), while significant deconjugation in humans begins at the end of the ileum and is completed in the large bowel (28, 34). The function of bile salt deconjugation in mammalian hosts is also not understood. However, in experiments performed with germfree and conventional mice, it has been observed that the presence of a microflora increases bile acid excretion and decreases serum cholesterol levels (5, 45). More is known about the adverse effects of bile salt metabolism, such as extensive deconjugation of bile salts in the small bowel in humans, which leads to steatorrhoea. Furthermore, enhanced concentrations of secondary bile acids in the colon can have tumor-promoting effects (29).

So far, BSH has been isolated from Bacteroides fragilis subsp. fragilis (41), Clostridium sp. (13), Lactobacillus sp. (25), and Bifidobacterium longum (15) and characterized. The bsh genes of Clostridium perfringens (7), L. plantarum (6) and L. johnsonii (11) have been cloned and sequenced, and BSH-negative mutants of L. plantarum (24) and Lactobacillus sp. (46) have been constructed.

In recent years the possibility of using bile salt deconjugation by lactic acid bacteria to lower serum cholesterol levels in hypercholesterolemic humans or to prevent hypercholesterolemia in individuals with normal cholesterol levels has received increased attention (8, 9). This possibility is surrounded by controversy concerning its validity and the mechanisms involved (28). To resolve the questions, a mechanistic approach which works with defined bacterial strains and can be used to investigate the effect of BSH in experimental animals and in humans is necessary. To create a basis for this investigation, we isolated and purified the BSH of B. longum SBT2928, a strain which has been isolated from human feces and which exhibits very high BSH activity. We cloned the BSH gene of this strain and selected and characterized a number of mutants, including a spontaneous BSH-negative strain. The latter strain could be used as a negative control in both animal experiments and human studies without the regulatory problems usually encountered with strains constructed by genetic engineering techniques.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Strains, plasmids, and media. The strains and plasmids used in this study are listed in Table 1. The strains were stored in 20% glycerol at 50°C. Escherichia coli was grown in TY medium (37) at 37°C with vigorous agitation or on TY medium solidified with 1.5% agar. For kanamycin resistance selection 50 µg of kanamycin per ml was added to the media. B. longum was grown in Briggs Liver Broth (3), which was prepared as follows. Tomato juice was mixed with an equal volume of distilled water and steamed for 1.5 h. Then the pH was adjusted to 7.0 with 2.5 M NaOH, and the preparation was filtered through a paper filter. Liver extract was prepared by using Bacto Liver (Difco Laboratories, Detroit, Mich.). Ten grams of liver powder was extracted with 170 ml of distilled water at 50 to 60°C for 1 h and then heated for 2 to 3 min at 100°C. After the preparation cooled, the pH was adjusted to 7.0, and the resulting extract was filtered through a paper filter. To prepare a 100-ml ascorbate-cysteine solution, 0.2 g of L-cysteine and 34 g of ascorbate were dissolved in water, and the pH was adjusted to 7.0 by adding ca. 11 g of sodium carbonate. The base medium was prepared as described by Briggs (3), and 75 ml of liver extract was added per 1,000 ml of medium. Shortly before the culture was started, 15 ml of the ascorbate-cysteine solution was added per 1,000 ml of medium. The strains were grown in liquid cultures at 37°C under anaerobic conditions (airtight containers with AnaeroGen AN35 bags, [Oxoid, Hampshire, United Kingdom]) for 16 to 20 h and on agar plates for 48 h. Alternatively, especially in the genetic experiments, M17 medium (48) supplemented with 1% glucose and 2.5 mM L-cysteine was used, and cells were grown as described above. To select mutants, M17 agar (1.5% agar) supplemented with 0.5% lactose and 15 ml of the ascorbate-cysteine solution per 1,000 ml of agar was used as the base medium and as the overlay agar for BSH activity detection.

Preparation of cell extracts. Cells in an overnight culture were sedimented by centrifugation for 10 min at 10,000 × g and 4°C, washed twice in 0.1 M sodium phosphate buffer (pH 7.0), and resuspended in the same buffer to give a density of approximately 10 to 15 optical density units at 600 nm (path length, 1 cm). To reduce oxidation of the enzyme, 10 mM dithiothreitol (DTT) was routinely added. Seven milliliters of the cell suspension was sonicated (Vibracell; Sonics & Materials, Danbury, Conn.) for 3 min at level 5 by using a 50% duty cycle and constant cooling. To remove residual DNA, 1% streptomycin sulfate and 0.1% polyethylenimine were added, and the suspension was stirred for 15 min. Then the mixture was centrifuged for 10 min at 20,000 × g and 4°C. The supernatant was stored as a cell extract at 20°C.

BSH assay and protein assay. A fast BSH assay was carried out as described by Coleman and Hudson (7), with several modifications. Portions (10 to 20 µl) of cell extract or partially purified BSH were added to wells of a 96-well microtiter plate with a flat bottom, and 200 µl of reaction mixture (50 mM sodium phosphate, 10 mM DTT, 1 mM EDTA, 10 mM sodium glycodeoxycholate; pH 5.5) was added. To ensure that the reaction conditions were (semi)anaerobic and to prevent evaporation during longer incubations, the wells were overlaid with 50 µl of light paraffin oil (0.85 g/ml; Merck, Darmstadt, Germany). Precipitation of deoxycholic acid was monitored with a microtiter plate reader (model 3550-UV microplate reader; Bio-Rad, Hercules, Calif.) at 37°C. The time until precipitation was an indication of the enzyme activity in the sample.

Chromatography methods. MonoQ anion-exchange chromatography (MonoQ HR5/5; Pharmacia Biotech, Uppsala, Sweden) was carried out as described by Grill et al. (15), with the following modifications. The solvent used was 0.1 M sodium phosphate buffer (pH 7.0), and the elution buffer was 0.1 M sodium phosphate (pH 7.0) to which 1 M NaCl was added, which resulted in a pH of 6.5. After proteins were loaded onto the column, they were eluted with a 0 to 0.5 M NaCl gradient in 20 min by using a flow rate of 1 ml/min. One-milliliter fractions were collected.

Immunostaining of BSH with anti-C. perfringens BSH antibodies. Anti-C. perfringens BSH antibodies were a kind gift from J. P. Coleman of East Carolina University (7). After a preparation was blotted onto a polyvinylidene difluoride (PVDF) membrane by using the Mini-blot system (Bio-Rad) (49), the membrane was immunostained with a Proto Blot II kit (Promega, Madison, Wis.) by following the manufacturer's instructions. The first antibody was diluted 1:2,000.

Determination of native and subunit molecular weights. The native molecular weight was determined by gel chromatography with HPLC; triplicate preparations were chromatographed at 21 or 4°C. The subunit molecular weight was determined by performing four independent sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) analyses (21) in which low-molecular-weight standard proteins (Bio-Rad) were used.

Determination of the pH and temperature optima for BSH activity. The BSH activity that was purified with consecutive MonoQ and MonoS chromatography steps was determined between pH 3 and pH 9 under standard conditions (see above) without DTT. The following buffers were used: 0.1 M sodium citrate-sodium phosphate buffer (pH 3 to 5); 0.1 M sodium phosphate buffer (pH 5.5 to 8); and 0.1 M sodium citrate-sodium phosphate buffer (pH 9).

Determination of the pH and temperature stability of BSH. We examined the stability of BSH at different pH values by using the buffers mentioned above at pH 3 to 10. The pH 10 buffer was prepared with 0.1 M sodium borate. Samples were incubated at the different pH values for 30 min at 37°C, and then the residual activity was determined at pH 6 by using the ninhydrin method without DTT.

Inhibition and activation of the enzyme. Inhibition or activation was studied by using the following compounds: iodoacetate, N-ethylmaleimide, periodic acid, EDTA, phenylmethylsulfonyl fluoride (dissolved in 70% ethanol), para-chloromercuribenzoic acid (dissolved in dimethyl sulfoxide), MgCl₂, MgSO₄, HgCl₂, CaCl₂, CuCl₂, and NaCl. Because some of these substances interact with phosphate, 20 mM MES (morpholineethanesulfonic acid) buffer (pH 6) was used for this part of the study. Before the substrate was added, the enzyme was incubated for 30 min at 37°C with the compounds mentioned above. Then the standard enzyme assay was carried out to determine the residual enzyme activity.

Determination of the N-terminal amino acid sequence. BSH was purified by consecutive MonoQ and MonoS chromatography steps. The resulting fractions that exhibited BSH activity were combined, concentrated, and applied to a mini electrophoresis unit (Bio-Rad). After separation, the proteins were electroblotted onto a PVDF membrane (Bio-Rad) and stained with Coomassie brilliant blue R-250 by using the method of Matsudaira (30), and the BSH protein was cut out and used for sequencing. N-terminal amino acid sequencing was performed with a Perkin-Elmer/Applied Biosystems model 476A protein sequencer (Perkin-Elmer/Applied Biosystems, Foster City, Calif.).

General genetic techniques, DNA sequencing, and sequence analysis. B. longum DNA was isolated by the method of Leenhouts et al. (23), with several modifications. The cells in 1.5 ml of an overnight culture were sedimented, washed once with distilled water, and resuspended in 20% sucrose-10 mM Tris-HCl (pH 8.1)-10 mM EDTA-50 mM NaCl containing 5 mg of lysozyme per ml. After incubation at 55°C for 10 min, first 20 µl of proteinase K (20 mg/ml) and then 25 µl of 10% SDS were added, and the suspension was mixed gently by inversion. The mixture was incubated at 60°C for at least 1 h to obtain complete lysis. The lysate was cleared by several phenol-chloroform extraction steps, and the DNA was precipitated by using standard methods. After the ethanol was removed without drying, the pellet was dissolved in 200 µl of 10 mM Tris-HCl (pH 7.4)-1 mM EDTA, and 5 µl of RNase (10 mg/ml) was added. The method could easily be scaled up for a culture volume of 15 ml.

Plasmid construction. To construct pBH1322 and pBH1351, pBH13 was cut with EcoRV and Asp718, the overhanging ends were filled in with the Klenow polymerase, and the fragments were separated in and isolated from an agarose gel and then cloned into the blunt end cloning vector pCR-Blunt (Invitrogen, Carlsbad, Calif.). The BSH regions of mutant strains were amplified by PCR performed with PWO polymerase (Roche Diagnostics Nederland b.v., Almere, The Netherlands) and appropriate primers and cloned into pCR-Blunt (Invitrogen).

PCR and RT-PCR. Table 2 shows the primers which were used in PCR and/or reverse transcription PCR (RT-PCR). Each PCR was carried out by using PWO polymerase as recommended by the manufacturer (Roche Diagnostics Nederland). RNA was isolated by using the macaloid method described by van Asseldonk et al. (50). First-strand cDNA synthesis was carried out with a first-strand cDNA synthesis kit (Roche Diagnostics Nederland) as recommended by the manufacturer by using either primer C or primer E (Table 2 and Fig. 1). Subsequently, PCRs were carried out by using Taq polymerase (HT Biotechnology, Ltd., Cambridge, United Kingdom), appropriate primer pairs, and both cDNAs.

View larger version (38K): [in this window] [in a new window]   FIG. 1.   Genetic and transcriptional organization of the bsh site (A) and RT-PCR analysis of the transcript of the bsh region (B and C). Pbsh, putative bsh promoter; A through E, primers A through E used for RT-PCR (Table 2); MWM, molecular weight marker; DNA, control reaction; +AMV, first-strand reaction with AMV; AMV, first-strand reaction without AMV. (B) First-strand cDNA synthesis with primer E. (C) First-strand cDNA synthesis with primer C. In the lanes in which primers A and C were used and in the lanes in which primers A and F were used three times as much RT-PCR product was applied in order to obtain bands whose intensity was similar to that of the DNA control.

Selection of spontaneous and UV-induced BSH-negative mutants. A spontaneous BSH-negative mutant was isolated during development of a selection procedure for detection of UV-induced bsh mutants. An M17 agar plate containing B. longum SBT2928 colonies was overlaid with M17 agar containing 0.5% lactose, 15 ml of a cysteine-ascorbate solution per liter, and 5 mM glycodeoxycholic acid. One BSH-negative colony which did not produce a precipitation halo was detected, and the organism was purified.

6

Nucleotide sequence accession number. The bsh nucleotide sequence has been deposited in the GenBank database under accession number AF148138.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Isolation and purification of B. longum BSH. B. longum SBT2928 was selected during a broad screening analysis of lactic acid bacteria (44) because of its high level of BSH activity. Since no major differences in enzyme activity during different growth phases or in the presence and in the absence of a 0.1 mM bile salt mixture (see Materials and Methods) were observed (results not shown), BSH was routinely isolated from overnight cultures grown in Briggs Liver broth.

View larger version (42K): [in this window] [in a new window]   FIG. 2.   Purification and immunostaining of B. longum BSH. (A) SDS-PAGE (10% polyacrylamide) after silver staining. Lane MW, molecular weight marker; lane CFE, cell extract; lanes Q, S, and GF, BSH active fraction after MonoQ chromatography, MonoS chromatography, and gel filtration, respectively. (B) Immunostaining with anti-C. perfringens BSH antibodies. Lane CFE, cell extract; lanes Q, S, and GF, BSH active fraction after MonoQ chromatography, MonoS chromatography, and gel filtration, respectively; lanes P and N, cell extract of E. coli with and without C. perfringens BSH, respectively.

View larger version (63K): [in this window] [in a new window]   FIG. 3.   Alignment of BSH of various bacteria, of a hypothetical protein of Bacillus subtilis, and PVA of Bacillus sphaericus. Asterisks, identical amino acids; dots, similar amino acids. The boxes indicate amino acids involved in the active site of PVA, including C-1, D-20, Y-82, N-175, and R-228 (positions based on the B. sphaericus DNA sequence). The amino acid sequence which was determined with purified BSH is indicated by boldface type. The enzymes examined were the BSH of B. longum (BSHBLO), L. johnsonii BSH (BSHLBJ), L. plantarum (BSHLBP), and C. perfringens (BSHCLP); the hypothetical protein of Bacillus subtilis (HYPBSU); and Bacillus sphaericus PVA (PENABSP).

Biochemical characterization of B. longum BSH. SBT2928 BSH is an enzyme that has a broad substrate range and can hydrolyze not only human bile salts but also tauroursodeoxycholic acid (3, 7-OH) and taurohyodeoxycholic acid (3, 6-OH) (Table 4). The highest levels of activity were observed with glycochenodeoxycholic acid (defined as 100% activity) and with glycine-conjugated bile acids. Furthermore, the enzyme exhibited a slight preference for conjugated chenodeoxycholic acid over conjugated cholic and deoxycholic acid. The substrate specificity of the purified enzyme and the BSH activity in the cell extract were the same (results not shown). Of the six major human bile salts tested (Table 4), the enzyme exhibited the highest affinity for glycochenodeoxycholic acid and, in general, higher affinities for glycine-conjugated bile acids than for taurine-conjugated bile acids. This is in good agreement with the results of the substrate specificity analysis.

Construction of a gene bank for B. longum SBT2928 and isolation of the BSH gene. A gene bank of the chromosome of B. longum SBT2928 was constructed in plasmid pUK21 and was transferred into E. coli NM522 (see Materials and Methods). A total of 30,000 colonies of transformants were plated onto BSH selection agar. Two colonies which produced a white precipitation halo after 2 to 3 days were identified. The plasmids in the cells in these colonies were designated pBH13 and pBH16 and used for further analysis. The inserts of both plasmids were analyzed with various restriction enzymes. pBH13 had a 2.7-kb insert, and pBH16 had a 8-kb insert (Fig. 4). Using the EcoRV site of the insert in pBH13, we constructed two subclones, pBH1351 (1.5-kb insert) and pBH1322 (1.2-kb insert) (Fig. 4). As E. coli(pBH1351) was positive in BSH activity assays, either most or all of the BSH genetic information was present on the 1.5-kb insert of this clone.

View larger version (11K): [in this window] [in a new window]   FIG. 4.   Plasmid maps of plasmids pBH13 and pBH16, which confer BSH activity, and two subclones. Heavy lines, plasmid pUK21; boldface restriction sites, sites present in the insert; Plac, E. coli lacZ promoter.

Nucleotide sequencing and analysis. The nucleotide sequence of the insert in pBH13 comprises 2,604 bases. In this sequence one complete open reading frame (ORF) and one partial ORF were detected, and these ORFs could encode a 317-amino-acid protein and a 447-amino-acid partial protein, respectively (Fig. 1A and 5). The complete ORF was present in the pBH1351 insert. The N-terminal amino acid sequence of the deduced 317-amino-acid protein starting from the third amino acid was identical to the N-terminal amino acid sequence determined by using purified BSH (Fig. 5). These data show that the complete ORF encodes BSH, and, accordingly, the gene was designated bsh. Taking into account the fact that f-Met was processed, the deduced protein had a molecular weight of 35,024 and a pI of 4.51. The derived molecular weight is in good agreement with the molecular weight obtained from the biochemical characterization of the enzyme. The N terminus of the deduced protein did not contain a typical leader peptide (56) or a membrane-spanning domain.

View larger version (55K): [in this window] [in a new window]   FIG. 5.   Nucleotide sequence of the bsh gene region of B. longum. IR, inverted repeat; DR, direct repeat; 10, putative 10 promoter region; SD, putative SD sequence. The N-terminal amino acid sequence determined for purified BSH is underlined. Mutations are indicated above the nucleotide sequence by boldface letters which indicate the exchanged or inserted nucleotides. The amino acids of the supposed active site (C-1, D-20, N-81, N-172, and R-225) are double underlined.

Homology studies with B. longum BSH. A BlastX search (1) revealed high levels of homology between B. longum BSH and BSH of several Lactobacillus species (37 to 48% identity) (GenBank accession number AF091248 [6, 11]), as well as C. perfringens BSH (36% identity) (7), penicillin V acylase (PVA) of Bacillus sphaericus (30% identical amino acids) (36), and a hypothetical protein of Bacillus subtilis (29% identity) (GenBank accession number P54948). Recently, the crystal structure of the PVA was elucidated (42). The amino acids of the active site were identified as Cys-1, Asp-20, Tyr-82, Asn-175, and Arg-228. A comparison of the deduced amino acid sequences of B. longum BSH and other BSH with the deduced amino acid sequence of the PVA showed that except for Tyr-82 all of these amino acids are conserved (Fig. 3).

bsh and glnE are transcriptionally coupled. In order to examine whether bsh and glnE are transcriptionally coupled, total RNA was isolated from B. longum SBT2928, and RT-PCR was carried out with several primer pairs that probed three different regions (Fig. 1A). Primers B and C were used for an internal control, primers D and E were used to detect transcripts running from bsh to glnE, and primers A and F and primers A and C were used to detect whether mRNA extended upstream beyond the possible transcription start site of bsh. First-strand cDNA synthesis was carried out both with primer E and with primer C. The results were the same for both cDNAs (Fig. 1B and C). In addition to the positive DNA control, a clear signal was obtained with primers D and E, which showed that bsh and glnE were present on the same transcript. Products were also obtained with primers A and F and with primers A and C, but the yields were lower. This indicated that a transcript extended upstream from bsh, but it may have been less abundant than the bsh-glnE transcript. Apparently, the stem-loop structure upstream of bsh (see above) could not eliminate transcription completely.

Selection, construction, and analysis of mutants with mutations in the BSH gene. BSH-negative mutants were selected after spontaneous mutation or after UV treatment. One spontaneous mutant (A3) and two UV-induced mutants (2001 and 2503) were obtained. The bsh genes of the mutants were amplified with a high-fidelity polymerase by using primers A and E (Table 2 and Fig. 1) and were sequenced. Table 7 shows the mutations in bsh (Fig. 5) and some of the properties of the mutants. The mutation in A3 resulted in a stop codon that resulted in deletion of 117 amino acids. The frameshift mutation in 2001 replaced 50 amino acids at the C terminus with 25 amino acids of reading frame +1. Both truncated forms of BSH were inactive, demonstrating that the C terminus is important for enzyme activity. Surprisingly, the PCR-amplified mutated bsh gene of A3 was active in E. coli, and the complete protein was detected in this organism (Table 7) (results not shown). In mutant 2503 Leu-80 was replaced by Ser. This amino acid was immediately upstream of the putative active site residue Asn-81. In cell extracts of this B. longum mutant and of E. coli harboring this recloned mutated bsh gene, no activity was detected, but the protein was still present.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Characterization of B. longum SBT2928 BSH showed that this enzyme is similar to the B. longum BB536 enzyme described by Grill et al. (15), but the two enzymes may differ in minor characteristics. However, Grill et al. did not provide any genetic data about B. longum BSH; such data are provided in this paper.

BSH is known to be very oxygen sensitive (41). The optimized isolation conditions described here, in which 10 mM DTT is included during sonication, allow workers to isolate highly active BSH under normal atmospheric conditions. Optimization of the sonication conditions is crucial since even in the presence of DTT, BSH activity is lost after extended sonication (Table 3).

The reported native molecular weights of BSH of different species vary greatly. Using HPLC, which is the most accurate method, we obtained a value for B. longum SBT2928 BSH of between 125,000 and 130,000, while Grill et al. (15) reported a value of 250,000 for strain B. longum BB536 BSH. Native molecular weights of 250,000 have also been reported for C. perfringens and Bacteroides fragilis enzymes (13, 41). Another C. perfringens strain has a native molecular weight of 147,000 (7). Interestingly, the molecular weight of the enzyme of Lactobacillus sp. strain 100/100 (now L. johnsonii [11]) is only 105,000 or 115,000. The subunit molecular weights of BSH have been reported to be between about 32,000 and 56,000. Consequently, native enzymes are octamers (Bacteroides fragilis [41]), hexamers (B. longum BB536 [15]), tetramers (C. perfringens [7, 13]), or trimers (L. johnsonii [25]). Supporting evidence that the BSHs of B. longum and perhaps also all other bacteria are tetramers comes from the high levels of homology of all BSHs whose primary amino acid sequences are known with PVA (42), which is a tetramer. Additionally, four of the five putative active-site amino acids of PVA are conserved in all BSHs, suggesting that these enzymes are also related in their tertiary and quaternary structures (Fig. 3). How this conjecture can be reconciled with the previously described heterotrimeric structure of the well-characterized BSH of L. johnsonii 100/100 (25) is not clear, but the problem could be approached by molecular modelling in which the three-dimensional structure of PVA is used as a template (see below).

In summary, biochemical characterization of B. longum SBT2928 BSH has shown that this enzyme has a tetrameric structure, a molecular weight of about 130,000, and broad substrate specificity, is active under pH and temperature conditions that occur in the human body, and requires a free ---SH group for catalytic activity.

The following facts prove that the BSH encoded by the gene cloned in this work and the enzyme which we purified and biochemically characterized are identical: (i) the 1.5-kb insert conferring BSH activity to E. coli contains an ORF for a protein whose deduced N-terminal amino acid sequence from the third amino acid onward is identical to the deduced N-terminal amino acid sequence of purified BSH, (ii) during MonoQ chromatography BSH activity produced in E. coli elutes at the same salt concentration as the BSH isolated from B. longum (results not shown), (iii) BSH activity isolated from E. coli, like B. longum BSH activity, cross-reacts with polyclonal antibodies raised against C. perfringens BSH, and (iv) various BSH-negative strains of B. longum SBT2928, obtained by UV-induced or spontaneous mutagenesis, have mutations in the bsh gene. The mutant bsh genes lead to a loss of BSH activity in E. coli (except for mutant A3 [see below]).

From the genetic data it is clear that BSH is an intracellular enzyme. This is consistent with the observation that no enzyme activity was present in the supernatants of overnight cultures, while activity was released either by sonication or other cell disruption methods or by lysis in assays performed with whole cells due to the lytic properties of the bile salts (data not shown).

In RT-PCR experiments we obtained products which contained almost the entire known sequence upstream of bsh, although the yield was less. This could indicate that bsh is transcribed both by its own promoter and to a lesser degree from an upstream region. Another piece of supporting evidence that bsh could have its own promoter comes from the fact that BSH activity is found in pBH13 and related constructs in which bsh is cloned in the direction opposite that of the lac promoter in these plasmids. Primer extension experiments have failed so far and need to be carried out to definitely establish whether bsh has a promoter directly upstream of its coding sequence.

Alignment of the 3' end of the 16S rRNA of bifidobacteria with the regions upstream of the ATG start codon of all known bifidobacterial genes revealed that only the AGG of the canonical GGAGG sequence (54) is conserved (Table 6). Also, bsh is preceded by two possible SD sequences with central AGG motives. This could imply that the translation start in bifidobacteria has slightly different requirements than the translation start in, for instance, Bacillus subtilis (51, 54). The gene following bsh, glnE, does not have this AGG motive and has only a very short putative SD sequence (GGA) that has a free energy of 7.2 kcal, which is very low for the gram-positive translation machinery. However, the stem-loop structure in front of glnE might facilitate a translational restart by bringing the stop codon of bsh close to the start codon of glnE (54).

B. longum BSH exhibits high levels of homology with BSH of various lactobacilli and C. perfringens, with a hypothetical protein of Bacillus subtilis, and with PVA of Bacillus sphaericus (42). Recently, the crystal structure of PVA has been determined. A homology analysis of BSH and PVA revealed that four of the five amino acids at the active site of PVA are conserved in BSH, while Tyr-82 is replaced by Asn-81 in BSH. The catalytically important part of Tyr-82 in PVA is the NH group of the peptide bond. Therefore, probably, Tyr-82 could be replaced easily by another amino acid if the NH group is kept in the right position (for BSH, Asn-81). The fact that Asn-81 is conserved in the different BSHs could have to do with different steric requirements for binding of bile salts compared to binding of penicillin V. The striking similarities in the primary structure and in the active-site amino acids indicate that B. longum BSH and all other known BSHs belong to the same group of N-terminal nucleophile hydrolases. Therefore, the secondary and tertiary structures are probably also very similar. It is not clear, however, how much the tertiary structure is conserved in L. johnsonii 100/100 BSH, for which a heterotrimeric structure has been proposed (26).

Downstream of bsh a gene with a very high level of homology to glutamine synthetase adenylyltransferase (glnE) was found. This enzyme transfers adenylyl residues from ATP to glutamine synthetase and is part of the nitrogen regulation cascade (27, 53) in E. coli and, by analogy, probably also in Haemophilus influenzae, various mycobacteria, and pseudomonads. The intergenic region between bsh and glnE is rather short and lacks a strong termination signal. RT-PCR experiments have shown that the two genes are transcriptionally coupled, indicating that in B. longum bsh is part of an operon. The extent of this operon is not known, since only the 5' part of glnE has been cloned and the transcript extends upstream of bsh (see above). Homology searches for the region 250 bp upstream of bsh produced no results. This genetic organization is different from that found in two Lactobacillus strains, while no adjacent sequences have been found for C. perfringens bsh (7). In L. plantarum bsh is monocistronic (6), and in L. johnsonii bsh is preceded by two genes that encode bile salt transport systems, but the exact extent of this operon is not known (11).

The combination of bsh and glnE in one operon is rather surprising, since there is no obvious functional relationship between the two enzymes. However, the coupling of bsh to a gene of the nitrogen regulation cascade reminds us of the hypothesis that hydrolysis of bile salts makes the amino acid nitrogen atoms of the released amino acids available for cells (18, 52).

Interestingly, the A3 bsh mutation, which did not produce a protein in B. longum, did produce activity and a protein in E. coli (Table 7). Apparently, in E. coli the stop codon created was overread in a translational readthrough process (12), which resulted in an intact molecule. One UV-induced mutation, the mutant 2503 bsh mutation, resulted in a leucine-to-serine change at the position next to active site residue Asn-81 which probably resulted in a substantial steric change which led to inactivation of the protein. With the other UV-induced mutation, the mutant 2001 bsh mutation, 50 C-terminal amino acids were removed, activity was lost, and there was rapid turnover of BSH, as the protein could not be detected in B. longum or in E. coli.

The N-terminal amino acid sequence determined for purified BSH did not include a Met residue. This means that formylmethionyl is processed and that Cys is the first amino acid of the mature protein. These findings are identical to findings obtained with PVA, in which Cys-1 plays a central role at the active site (42). The important role of Cys-1 was confirmed by the finding that a change of Cys-1 to Ala led to complete inactivation of BSH.

The biochemical and genetic data reported in this paper provide a sound foundation for investigating the role of bile salt hydrolysis in both the bacterial producers and their mammalian hosts. Furthermore, we created BSH-negative mutants which can be used in animal experiments and human studies to investigate the effects of enhanced bile salt deconjugation in the small intestine and to evaluate in a mechanistic way the hypothesis that BSH lowers serum cholesterol.