J. Rouvinen, Department of Chemistry, University of Eastern Finland, PO Box 111, 80100 Joensuu, Finland Fax: +358 13 2513390 Tel: +358 13 2513318 E-mail: firstname.lastname@example.org
Crystal structures of native and α-d-galactose-bound Bacillus circulans sp. alkalophilusβ-galactosidase (Bca-β-gal) were determined at 2.40 and 2.25 Å resolutions, respectively. Bca-β-gal is a member of family 42 of glycoside hydrolases, and forms a 460 kDa hexameric structure in crystal. The protein consists of three domains, of which the catalytic domain has an (α/β)8 barrel structure with a cluster of sulfur-rich residues inside the β-barrel. The shape of the active site is clearly more open compared to the only homologous structure available in the Protein Data Bank. This is due to the number of large differences in the loops that connect the C-terminal ends of the β-strands to the N-terminal ends of the α-helices within the (α/β)8 barrel. The complex structure shows that galactose binds to the active site as an α-anomer and induces clear conformational changes in the active site. The implications of α-d-galactose binding with respect to the catalytic mechanism are discussed. In addition, we suggest that β-galactosidases mainly utilize a reverse hydrolysis mechanism for synthesis of galacto-oligosaccharides.
Database The coordinates for free and α-d-galactose-bound Bca-β-gal structures have been deposited in the Research Collaboratory for Structural Bioinformatics (RCSB) Protein Data Bank under accession codes 3TTS and 3TTY, respectively.
β-galactosidase from Bacillus circulans sp. alkalophilus
X-ray fluorescence NPG
The enzyme β-galactosidase (EC126.96.36.199) catalyzes the hydrolysis of β(1-3) and β(1-4) galactosyl bonds in oligo- and disaccharides, but also catalyzes the reverse reaction of the hydrolysis, often called transglycosylation. Because of these enzymatic characteristics, β-galactosidase is able to hydrolyze milk sugar, lactose, that consists of one glucose and one galactose molecule, and therefore, the enzyme is usually called a lactase. Reduced or absent activity of lactase induces inefficient lactose digestion, and ∼ 65% of the population worldwide are estimated to suffer from various degrees of lactose intolerance . Thus, these enzymes are very important for the dairy industry in order to manufacture low-lactose and lactose-free milk products [2,3]. Furthermore, β-galactosidases are also widely used in experiments to enzymatically synthesize of galacto-oligosaccharides . Galacto-oligosaccharides are prebiotic and impart health benefits to humans, for example, by selectively stimulating the growth of beneficial bacteria in the colon [5,6]. β-galactosidases from Bacillus circulans, in particular, have been widely used in such experiments [7–11].
β-galactosidases belong to sub-families 1, 2, 35 and 42 of the GH-A superfamily of glycoside hydrolases . All members of this superfamily have an (α/β)8 barrel as the catalytic domain, with two glutamic acid residues acting as an acid/base catalyst and a nucleophile . At least one native structure of β-galactosidase is available for each of the four sub-families in the Protein Data Bank: from Sulfolobus solfataricus  (family GH-1), from Escherichia coli  and Arthrobacter sp. C2-2  (family GH-2), from Penicillium sp.  and Trichoderma reesei  (family GH-35), and from Thermus thermophilus A4  (family GH-42). Moreover, there are also unpublished structures of β-galactosidases from Bacteroides fragilis (family GH-2) and Bacteroides thetaiotaomicron (families GH-2 and GH-35) in the PDB. In addition to the native structures of β-galactosidase, several complex and mutant structures have also been reported whose enzymatic function is well known, mostly from Escherichia coli [20–22]. Of all the determined structures, only that for Thermus thermophilus A4 β-galactosidase (A4-β-gal) shows some homology (29% identity) with Bacillus circulans sp. alkalophilusβ-galactosidase (Bca-β-gal).
Enzymes from alkaliphilic microorganisms are stable and active in alkaline conditions. Because of this, alkaliphilic enzymes, especially from Bacillus species, are often used in industrial applications . However, deeper knowledge of the structural biology of aIkaliphilic enzymes is required in order to better understand their enzymatic characterictics. Here, we report the enzymatic characteristics and first crystal structures of an alkaliphilic β-galactosidase from Bacillus circulans sp. alkalophilus. The crystal structures of a β-glucosidase  and phosphoserine aminotransferase  from the same organism have previously been reported. The reported structures assist in understanding of the enzymatic function and mechanisms of this alkaline enzyme, and also provide a good starting point for rational protein engineering of the enzyme.
Results and Discussion
Biochemical analysis of Bca-β-gal: hydrolysis and transglycosylation
The optimal pH for the reaction of Bca-β-gal with o-nitrophenyl-β-d-galactoside (oNPG) as a substrate was maximal from pH 6 to 7.5, ∼ 50% at pH 9 and 5% at pH 10, while ∼ 20% activity remained at pH 4.5. Thus Bca-β-gal had a wide pH activity and was also active in alkaline buffers. This is in accordance with the internal pH of alkaliphilic bacteria, which is between 7.5 and 9.5 . The activity for homologous A4-β-gal was also highest at pH 6.5, but decreased much more at higher pH than for Bca-β-gal . The optimal temperature for maximal activity with oNPG as a substrate was 55 °C. The activity was 80% and 13% of maximal activity at 40 and 10 °C, respectively, and 20% and 2% at 70 and 80 °C, respectively. No activity was observed with p-nitrophenyl-α-d-galactopyranoside as a substrate. The Km was 2.2 ± 0.2 mm, and the Vm was 10 ± 0.8 μmol·s−1 per mg at optimal pH and temperature for oNPG as a substrate. The Km was 15 ± 1 mm and Vm was 41 ± 1.6 μmol·s−1 per mg for lactose as a substrate under the same conditions. Inhibition of lactose hydrolysis by isopropyl β-d-1-thiogalactopyranoside was competitive, with an inhibition constant (Ki) of 50 ± 3 mm.
To determine the transglycolysation activity of Bca-β-gal, an enzyme assay was performed using 5, 18 and 38 mg·mL−1 lactose. The enzyme had low transglycolysation activity, which was seen only at concentrations of 18 and 38 mg·mL−1. The amount of galactose and glucose released from lactose was much higher than the amount of transglycolysation product formed. Depending on the amount of lactose, only 0.5–4% of all products formed during lactose hydrolysis were transglycolysation products.
Calculation of the Matthews coefficient  suggested the presence of six monomers per asymmetric unit (VM = 2.62 Å3·Da−1), with a solvent content of 53.1% for both structures (native and α-d-galactose-bound). The statistics of the data collection and the structure refinements are presented in Table 1. The final models consist of the whole mature form of Bca-β-gal, which comprises 675 amino acid residues per monomer. In a Ramachandran plot, 89.4% of the residues were found in the most favored regions, 9.5% in additional allowed regions, 0.8% in generously allowed regions, and 0.3% in disallowed regions. Ile545 and Asn625 were the only residues found in disallowed regions. Ile545 is located in a short turn between two β-strands and its structure is stabilized by main-chain hydrogen bonds. Asn625 is located in the β-hairpin loop at the protein surface, and the electron density for this residue is weak. The molecules in an asymmetric unit are similar: the root mean square difference (RMSD) is 0.16 Å between monomers in the native structure and 0.14 Å in the complex structure. α-d-galactose binding caused changes in the overall structure. The same monomers of the native and complex structures have an 0.18 Å difference, but when the whole hexamers are compared, the difference is 0.330 Å, indicating changes in the relative positions of the monomers in the quaternary structure.
Table 1. Data collection and structure refinement statistics. Numbers in parentheses are for the outer shell.
In the crystal structure, the Bca-β-gal polypeptide chain forms three domains (Fig. 1A). The most homologous structure in the Protein Data Bank is β-galactosidase from Thermus thermophilus A4 (A4-β-gal) . The amino acid identity between Bca-β-gal and A4-β-gal is low (29%), and the RMSD is high (1.69 Å for 601 Cα atoms), which is reflected in clear structural differences in both the general structure (Fig. 1B) and local regions, even the active site.
The first domain (Met1–Ser395) is the catalytic domain containing an (α/β)8 barrel. Among the three domains of Bca-β-gal and A4-β-gal the RMSD for the first domain is the smallest, 1.03 Å. An interesting feature in the (α/β)8 barrel of Bca-β-gal is the cluster of sulfur-containing residues: four methionines and one cysteine in the hydrophobic interior of the β-barrel (Fig. 2H). Overall, the loops connecting the C-terminal ends of the α-helices and the N-terminal ends of the β-strands are short, making the N-terminal end of the β-barrel relatively smooth. However, Bca-β-gal has nine additional residues that pack against the hydrophobic N-terminal end of the β-barrel (Fig. 2H). The loops connecting the C-terminal end of the β-strands to the N-terminal end of the α-helices (Fig. 2G) are longer, containing secondary structure elements in many places. In addition, there are clear differences between Bca-β-gal and A4-β-gal in these loops. The first loop after the first β-strand (Tyr15–Trp20) and the second loop (Asn43–Asp58) contain a short α-helix. The third loop (Thr80–Asn117) is 38 residues long, containing two short α-helices and an irregular loop structure that participate in Zn binding. The overall structure of the loop is similar in Bca-β-gal and A4-β-gal. The fourth loop (Val147–Phe214) is 68 residues long, containing two α-helices. This loop also participates in Zn binding, and is important for quaternary structure formation because it forms major interactions with the other monomer. There are clear differences in the position of this loop between Bca-β-gal and A4-β-gal: the greatest distinction is at the C-terminal end of the loop, in which Bca-β-gal has an additional small two-stranded β-sheet. Furthermore, there is a four-residue deletion at the N-terminus of this loop in Bca-β-gal compared to A4-β-gal. This opens up the active site considerably, which may affect substrate binding. The fifth loop (Met254–Asp261) is short. The sixth loop (Asn276–Pro283) is also short in Bca-β-gal, but contains two α-helices and is much longer (23 residues) in A4-β-gal. This longer loop interacts with the second monomer, thus stabilizing the quaternary structure. The seventh loop (Gln308–Pro324) and the eighth loop (Phe345–Thr370) do not have secondary structural features. These are quite similar in Bca-β-gal and A4-β-gal, except at the C-terminal end of the eighth loop.
The second domain (Glu396–Pro609) has a complicated α/β structure, with a central seven-stranded mixed β-sheet and five α-helices. The C-terminal part of the domain forms a small β-barrel. The domains of Bca-β-gal and A4-β-gal are similar in terms of topology, which is reflected by the fact that the RMSD is 1.21 Å. Bca-β-gal is 19 residues longer: these extra residues are located in three β-hairpin loops. The function of the second domain is structural. It forms a major interface with a second monomer, thus stabilizing the tertiary structure assembly.
The third domain (Leu610–Ala675) is the smallest, and has an anti-parallel β-sandwich structure comprising two β-sheets, of which one is clearly smaller. Its structure resembles a jelly roll fold. The overall topologies of Bca-β-gal and A4-β-gal are similar, but this domain has 13 more residues in Bca-β-gal, which are located in loops. The RMSD is 1.53 Å. The function of this domain is probably also structural, because there are no clear clefts on the surface or cavities that could have functionality, for example in ligand binding.
Metal binding site
Four cysteine residues (Cys115, Cys155, Cys157 and Cys160) form a metal binding site in the catalytic domain of Bca-β-gal (Fig. 1A and Fig. S1). The metal atom was identified as zinc by X-ray fluorescence (XRF) analysis. The XRF spectra (Fig. S2) show the presence of zinc, and the energy scan (Fig. S3) across the absorption edge of zinc shows a signal (9.6685 keV) that corresponds to the previously published value (9.66 keV) . The metal binding site is similar to that in A4-β-gal, in which zinc was only observed in the native structure. Thus, it was assumed that the zinc is loosely bound in A4-β-gal . Neverthless, based on the electron density, zinc clearly exists in both Bca-β-gal structures, with low B-factors, suggesting that the zinc is tightly bound. We assume that the zinc has no role in enzymatic reactions, but the metal binding site has a stabilizing effect upon the structure of the catalytic domain.
Quaternary structure of Bca-β-gal
The asymmetric unit of a Bca-β-gal crystal contains six monomers (Fig. 1E,F) arranged in a symmetrical fashion, resembling a barrel whose upper and lower parts are made up of trimers (Fig. 1C,D). According to PISA analysis , the trimer structures are compact because the interface areas between the monomers are extensive (2278 Å2), consisting of a number of hydrogen bonds and hydrophobic interactions. In the trimer, the interface area is therefore 6834 Å2. In A4-β-gal, due to differences in the structure as discussed above, the interface area between monomers is larger (2385 Å2). However, the interface areas between the two trimers are smaller, and far fewer interactions take place (Fig. 1E). Nevertheless, there is a twofold axis of symmetry between monomers that is perpendicular to the threefold symmetry axis. The interface area between the two monomers is 614 Å2, resulting in a 1842 Å2 interface between two trimers. Interestingly, the crystallographic packing of A4-β-gal reveals formation of a similar hexameric arrangement as in Bca-β-gal, but the interface between monomers related by a twofold axis is only 304 Å2. Altogether, the crystal structure of Bca-β-gal suggests that Bca-β-gal is able to form a stable trimer in solution. This is supported by the results of dynamic light scattering, which showed a unimodal distribution with a mean hydrodynamic radius (RH) value of 6.2 nm, corresponding to a protein of ∼ 242 kDa (Fig. S4). Because the size of the Bca-β-gal monomer is 76.8 kDa, this result suggests that Bca-β-gal forms a stable trimeric structure in solution. However, at higher protein concentrations, Bca-β-gal may form a weak hexameric structure, the so-called ‘transient oligomer’ .
In addition to the four cysteine residues of the metal binding site, there are seven non-conserved cysteine residues per monomer. These cysteines are distant from each other, and are thus unable to create any disulfide bonds in the monomer or in the quaternary structure. In addition, no covalent bonds between the monomers were found.
Active site of Bca-β-gal
The omit electron density map (Fig. 2A) calculated for the galactose-soaked crystals showed a clear density for one galactose molecule in each catalytic domain. Unexpectedly, every galactose molecule had the α-anomeric configuration. Seven residues (Arg111, Asn149, Glu150, Glu307, Gln313, Glu355 and His358) form hydrogen bonds with α-d-galactose (Fig. 2B). In addition, Phe345 in particular packs against C4 carbon (Fig. 2B and Fig. S5), which has a hydroxyl group in the axial position in α-d-galactose, thus explaining the specificity for galactose rather than glucose at this binding site. The hydrogen bonding network includes interactions with other surrounding residues and water molecules, for example Trp187 forms a hydrogen bond with α-d-galactose via a water molecule (Fig. 2B). Based on comparison with the structures of A4-β-gal, Glu150 and Glu307 were identified as an acid/base catalyst and a nucleophile, respectively. Asp275 is located near both residues, forming a short hydrogen bond (2.5 Å) with Glu150. Evidently, Asp275 is important in determining the catalytic properties of both glutamates.
When monomers of the native and α-d-galactose-bound structures of Bca-β-gal were superimposed, conformational changes were observed in two loop regions (Fig. 2C and Fig. S6). The first is located in the sixth loop that connects the C-terminal end of the β-strand and the N-terminal end of the α-helix in the region Tyr277–Thr282, which position has changed by 1–3 Å. This loop interacts with the seventh loop, in which there are also conformational changes in the region Gln313–Gln316, especially in the positions of the side chains of Trp315 and Gln313. Because of reorganization of the interactions, Trp315 has moved towards α-d-galactose. Movement of the Trp residue (Trp811) in the active site has also been reported in Trichoderma reeseiβ-galactosidase , so Trp315 may play an important role in substrate binding . Interestingly, the observed conformational changes of Bca-β-gal occur in the same locations, i.e. loops 6 and 7 of the (α/β)8 barrel, as previously found for β-galactosidase from Trichoderma reesei .
Comparing the active sites of Bca-β-gal and A4-β-gal, it is apparent that, despite clear differences in the loop structures (Fig. 2G), most of the residues forming a hydrogen bond with α-d-galactose are the same. However, Gln313, which forms a hydrogen bond with α-galactose, is replaced by Val318 in A4-β-gal, and the hydrogen bond is replaced by a hydrogen bond to His363 in A4-β-gal. The different residue (Gln/Val) in this position may also affect the conformation of the conserved residue Trp315.
Although the α-galactose-binding residues were mainly similar, the overall shapes of the active sites are different. The greatest differences are in loops 4, 5 and 6, making the active site entrance smaller for A4-β-gal compared to Bca-β-gal, thus restricting ligand access. The complex structure reveals binding of only one monosaccharide to the active site, and there is no direct information on binding of the second monosaccharide. However, it is known that carbohydrates often bind to the flat surfaces of aromatic rings. Three such residues exist in the structure of Bca-β-gal. Tyr277 is the most probable binding site for the glucose ring of lactose, and it also exists in A4-β-gal. In addition, there are two tryptophans, Trp315 and Trp207, in the surface of the active site area. Interestingly, the residue corresponding to Trp315 has a different conformation in A4-β-gal and the residue corresponding to Trp207 is totally absent in A4-β-gal. The role for these tryptophans may be in transient binding of carbohydrate substrates, guiding these towards the active site.
An interesting feature of the Bca-β-gal complex structure was the unusual binding of galactose in the α-anomeric form. A4-β-gal was also found to bind galactose in the α-anomeric form, but no explanation was given . Enzymatic hydrolysis of the β-glycosidic bond  is probably initiated by protonation of glycosidic oxygen by an acid/base catalyst, Glu150 (Fig. 3A). This is followed by nucleophilic attack by Glu307, resulting in formation of a covalent bond between galactose and Glu307, and release of glucose. In this covalent intermediate structure, galactose is in the α-configuration. When hydrolysis is completed and β-d-galactose is released, α-d-galactose is formed in solution by mutarotation. Then, α-galactose is able to bind to the active site. The presence of α-galactose in the active site of Bca-β-gal and A4-β-gal could thus resembles binding of a covalent reaction intermediate. Indeed, the distance between O1 of α-d-galactose and OE1 of Glu307 is unusually short (2.4 Å). This complex structure thus indirectly supports the existence of a covalent intermediate in the reaction mechanism.
In several studies [7–11], β-galactosidases, especially from Bacillus circulans, have been found to catalyze the synthesis of galacto-oligosaccharides, which are of considerable interest as prebiotics. However, Bca-β-gal did not generally show formation of galacto-oligosaccharides, but by using a high substrate concentration, weak formation of galacto-oligosaccharides was detected. It has been supposed that some β-galactosidases may be more suitable for galacto-oligosaccharide synthesis than others , and it would be useful to estimate the factors that affect this. van Rantwijk et al. suggested that galacto-olicosaccharides can be synthesized either by thermodynamically controlled reverse hydrolysis or kinetically controlled transglycosylation .
Reverse hydrolysis is the opposite reaction to hydrolysis of lactose. The reaction steps in lactose hydrolysis include an intermediate in which a covalent bond is formed between C1 of galactose and Glu307 (Fig. 3A). This covalent intermediate can be released by a water molecule, leading to complete hydrolysis, or by an acceptor molecule, which may be glucose (leading back to formation of the substrate lactose), another (oligo)saccharide or an alcohol. The reaction is retaining, and the product has a similar β-glycosidic bond as the substrate. Because a water molecule is formed, this is also a condensation reaction. However, lactose hydrolysis is an exothermic reaction, and the chemical equilibrium strongly favors formation of monosaccharides. Because of entropy, the final equilibrium state also contains small amounts of galacto-oligosaccharides.
In transglycosylation, the reacting group (alcohol or (oligo)saccharide) directly replaces the glucosyl of lactose, and a water molecule is not formed. Therefore, it is an SN2 type of substitution reaction leading to formation of a product with an α-glycosidic bond. By examination of the complex crystal structure of Bca-β-gal, it is apparent that a transglycosylation reaction cannot take place in the active site because there is no space for the reacting monosaccharide, which is expected to bind in approximately the same position as the nucleophile Glu307. Because other β-galactosidases also utilize a similar reaction mechanism with glutamic acid residues, it is likely that the transglycosylation mechanism is not possible in these enzymes. In fact, common procedures to synthesize galacto-oligosaccharides include the use of extremely high lactose concentrations (up to 700 g·L−1) and very high temperatures, favoring the formation of endothermic products (oligosaccharides) and implying that the reaction is strongly thermodynamically controlled.
Therefore, we believe that β-galactosidases utilize mainly reverse hydrolysis rather than transglycosylation for synthesis of galacto-oligosaccharides, and the reaction is mainly thermodynamically controlled. Hovewer, the shape of the active site of the β-galactosidases may be very different, as observed for the homologous enzymes Bca-β-gal and A4-β-gal. This affects the kinetics of hydrolysis and reverse hydrolysis. Before final equilibrium is achieved, there are differences in the relative concentrations of substrate, monosaccharides and oligosaccharides that may be utilized in the enzymatic synthesis of galacto-oligosaccharides. Thus some β-galactosidases may be more suitable for synthesis of oligosaccharides, but this requires careful tuning of reaction conditions. However, removal of oligosaccharides from the reaction medium  could be the most effective method for driving a thermodynamically controlled enzymatic reaction towards synthesis of oligosaccharides.
Enzyme production, purification and biochemical analysis
The β-galactosidase gene of Bacillus circulans sp. alkalophilus was cloned into NdeI and XhoI sites in a pET32b vector (Novagen, Madison, WI) by using specific primers in a PCR reaction. Six histidine amino acids were added to the 3′ end of the Bca-β-gal gene for protein purification. The PCR fragment of the Bca-β-gal gene was ligated into the pET32b vector, and the vector was transformed into Escherichia coli. Transformants were randomly chosen and transferred from Luria broth (LB) plates to LB medium, and allowed to grow overnight to obtain plasmid preparations. After growth, plasmids were isolated and verified by restriction enzyme digestions and DNA sequencing to obtain the correct reading frame. The correct clone was transfered to Escherichia coli OrigamiB(DE3) cells (Novagen, Madison, WI) for Bca-β-gal expression. Four liters of LB medium were cultivated in four 2 L Erlenmeyer flasks at 37 °C, 250 rpm, in a shaker. The culture medium was inoculated with a pre-culture, grown overnight at 37 °C. Ampicillin (100 μg·mL−1), kanamycin (1.5 μg·mL−1) and tetracycline (7.5 μg·mL−1) were added at the start of growth, and growth was followed until the cell density was 0.5–0.6, when more ampicillin (100 μL·mL−1) and isopropyl-β-d-1-thiogalactopyranoside (IPTG; final concentration 1 mm) were added. Simultaneously, the temperature was decreased to 20 °C, and the culture was allowed to grow overnight. The next day, cells were removed by centrifugation (6000 g for 15 min at 4 °C) and suspended in 40 mL of cell lysis buffer containing 20 mm Hepes, pH 7.2, 0.1 m KCl, 5 mm MgCl2. This suspension was sonicated on ice for 30 s, and allowed to cool for another 30 s. This cycle was repeated four times. After sonication, the Bca-β-gal sample was centrifuged at 19 000 g for 60 min at 4 °C. To reduce viscosity caused by DNA, 20 U·mL−1 DNase was added to the sample, which was incubated for 1 h at 4 °C and filtrated before purification. An Ni-agarose column was stabilized using 30 mm imidazole buffer, pH 7.3, containing 20 mm Hepes, 0.5 m KCl and 2.5 mm MgCl2. Imidazole and KCl were added to the sample at final concentrations of 30 mm and 0.5 m, respectively. After adding the sample to the column, unbound protein was rinsed out using a stabilization buffer (30 mm imidazole, 20 mm Hepes, 0.5 M KCl and 2.5 mm MgCl2, pH 7.3), and Bca-β-gal was eluted using 200 mm imidazole, 20 mm Hepes, 0.5 m KCl, 2.5 mm MgCl2, pH 7.3. The affinity tag was not removed before crystallization.
After the Ni-agarose column, Bca-β-gal was purified using a column containing 4-aminobenzyl 1-thio-β-d-galactopyranoside agarose (A0414; Sigma, St Louis, MO). This column was equilibrated with 50 mm Hepes buffer, pH 7.3, containing 0.5 m NaCl and 1 mm MgCl2. The Bca-β-gal sample was applied to the column, and unbound protein was rinsed off using equilibration buffer (50 mm Hepes, 0.5 M NaCl and 1.0 mm MgCl2, pH 7.3) and water. Bca-β-gal was eluted from the column using 0.1 m borate buffer, pH 3, containing 500 mm NaCl and 1 mm MgCl2. The Bca-β-gal fraction was neutralized using 500 mm Hepes buffer, 1 mm MgCl2, pH 7.3. The sample was dialyzed in 20 mm Hepes, 1 mm MgCl2, pH 7.3, and concentrated after dialysis to 10 mg·mL−1 using an Amicon Ultra Ultracel-10k centrifugal filter (Millipore, Billerica, MA). The purity of the protein was examined by 10% SDS/PAGE, and was determined to be > 98%.
Characterization of β-galactosidase activity
Unless otherwise stated, Bca-β-gal was assayed by measuring the hydrolysis of o-nitrophenyl-β-d-galactoside (oNPG) in 100 mm citric acid/sodium phosphate buffer, pH 6.0 (McIlvaine buffer). The reaction was started by addition of a preincubated enzyme solution to the sample. The assay mixture was incubated at 30 °C for 5 min, and the reaction was stopped by addition of 1 m Na2CO3. The total volume of the sample was 1 mL, and the final concentration of oNPG was 4 mm. Absorbance was measured at 420 nm against a blank sample. To determine the activity of Bca-β-gal on lactose, the following procedure was used: Bca-β-gal was incubated with appropriate amounts of lactose in 100 mm Hepes buffer, pH 7, instead of McIlvaine buffer, because lactose and phosphate displayed similar retention times in HPLC analysis. The total volume of the sample was 1 mL, and the incubation time was 1–3 h at 50 °C. The reaction was stopped by boiling the incubation mixture at 95 °C for 10 min, and, after cooling on ice, samples were centrifuged in an Eppendorf centrifuge (15 000 g for 10 min at RT). The amount of galactose and glucose liberated were determined by HPLC by using an Aminex HPX-87H column (Bio-Rad Labs., Richmond, CA) with 4 mm H2SO4 as the mobile phase at a flow rate of 0.6 mL·min−1. The concentrations of galactose and glucose were calculated from the total concentration of galactose, glucose and lactose.
In order to determine Km and Vmax, the activity of Bca-β-gal with oNPG or lactose was assayed at six substrate concentrations in an appropriate range. Each measurement was performed in triplicate.
In order to determine the optimal pH and temperature, the following buffers were used: 0.2 m citric acid/Na2HPO4 (McIlvaine buffer) for pH 3.0, 4.0, 5.0, 6.0 and 7.0; 0.2 m Hepes/NaOH for pH 7.0, 7.5, and 8.0; 0.2 m glycine/NaOH for pH 9 and 10; 0.2 m Na2HPO4/NaOH buffer for pH 11 and 12. The final concentration of oNPG was 4 mm. Samples were incubated at 30 °C for 5 min before addition of ice-cold Na2CO3. To determine the optimal temperature, temperatures from 10 to 80 °C in 0.1 m McIlvaine buffer, pH 6.0, were used. The incubation time necessary to achieve the optimum temperature was 2.5 min.
The inhibition of lactose hydrolysis by IPTG was tested in 0.1 m Hepes buffer, pH 7.0, at 50 °C. The lactose concentration was 5 or 10 mm, and the IPTG concentration ranged from 0 to 60 mm, and samples were incubated for 1 or 2 h. After incubation, samples were boiled for 10 min, cooled and analyzed by HPLC, as explained previously. The occurrence of transglycolysation was tested in various lactose concentrations and analyzed by HPLC. Standard methods were used for SDS/PAGE and for protein concentration measurements.
Dynamic light scattering
Dynamic light scattering analysis of Bca-β-gal was performed at 22 °C using a DynaPro99 dynamic light scattering system (Wyatt Technology Corp., Santa Barbara, CA) with a temperature-controlled micro-sampler. The Bca-β-gal sample (4 mg·mL−1) was scanned 20 times using this system.
Crystallization and data collection
The Bca-β-gal sample (10 mg·mL−1) was filtered by using a MILLEX-GV 0.22 μm filter unit (Millipore) before crystallization. After filtration, the sample was saturated to 0.1% with NaN3 in order to prevent growth of the bacteria Bacillus circulans sp. alkalophilus. Crystallization was performed using the hanging-drop vapour diffusion technique at 22 °C. A 2 μL drop of Bca-β-gal solution (10 mg·mL−1) was mixed with an equal volume of crystallization solution, and equilibrated against 500 μL crystallization solution. Two crystallization conditions were found for Bca-β-gal crystals (Fig. S7). Twinned hexagonal plate crystals were obtained using 0.1 m MgSO4, 0.1 m bicine, pH 8.5, 10–15% w/v poly(ethylene)glycol 3350 (Hampton Reseach, Aliso Viejo, CA). When the concentration of poly(ethylene) glycol 3350 was decreased to 6–9% w/v, stick crystals were obtained. The cryosolution used contained 0.1 m bicine (pH 8.5) and 25% v/v 2-methyl-2,4-pentanediol, as well as 10% w/v poly(ethylene) glycol 3350 for native crystals or 20% w/w β-d-galactose for complex crystals. The stick-like crystals were quickly (5 s) soaked and cryocooled to 100 K in a cold nitrogen stream, and used in data collection for structure determination. The native and α-d-galactose complex data were collected from beamlines ID23-1 and ID14-4, respectively, of the European Synchrotron Radiation Facility (Grenoble, France). Both datasets were collected using a 0.5° oscillation range with the ADSC Quantum Q315r detector. Finally, all data were processed using the XDS program .
X-ray fluorescence analysis
The X-ray fluorescence (XRF) analysis was performed on beamline ID29 at the European Synchrotron Radiation Facility using a Rontec Silicon drift diode XRF detector.
Crystal structure determination and model refinement
The three-dimensional structure of Bca-β-gal was determined by molecular replacement using PHASER . A4-β-gal (Protein DataBank code 1KWG)  was used as a search model (29% amino acid identity). The R- and R-free factors of the initial Bca-β-gal structure were very high: 52.3% and 52.8%. Refinement of the native and α-d-galactose-bound structures of Bca-β-gal was performed using PHENIX . Non-crystallographic symmetry was not used in the refinement. The protein structure was refined first, then the ligand α-d-galactose was added, and finally water molecules were introduced. Manual building was performed using Coot . Validation of the structures was performed using PROCHECK .
The authors acknowledge the European Synchrotron Radiation Facility for provision of synchrotron facilities Ritva Romppanen is thanked for skilled technical assistance. This work was supported by a grant from the Finnish Funding Agency for Technology and Innovation (SymBio Technology Program).