High Order Quaternary Arrangement Confers Increased Structural Stability to Brucella sp. Lumazine Synthase*

The penultimate step in the pathway of riboflavin biosynthesis is catalyzed by the enzyme lumazine synthase (LS). One of the most distinctive characteristics of this enzyme is the structural quaternary divergence found in different species. The protein exists as pentameric and icosahedral forms, built from practically the same structural monomeric unit. The pentameric structure is formed by five 18-kDa monomers, each extensively contacting neighboring monomers. The icosahedrical structure consists of 60 LS monomers arranged as 12 pentamers giving rise to a capsid exhibiting icosahedral 532 symmetry. In all lumazine synthases studied, the topologically equivalent active sites are located at the interfaces between adjacent subunits in the pentameric modules. The Brucella sp. lumazine synthase (BLS) sequence clearly diverges from pentameric and icosahedric enzymes. This unusual divergence prompted us to further investigate its quaternary arrangement. In the present work, we demonstrate by means of solution light scattering and x-ray structural analyses that BLS assembles as a very stable dimer of pentamers, representing a third category of quaternary assembly for lumazine synthases. We also describe by spectroscopic studies the thermodynamic stability of this oligomeric protein and postulate a mechanism for dissociation/unfolding of this macromolecular assembly. The higher molecular order of BLS increases its stability 20 °C compared with pentameric lumazine synthases. The decameric arrangement described in this work highlights the importance of quaternary interactions in the stabilization of proteins.

Riboflavin, an essential cofactor for all organisms, is biosynthesized in plants, fungi, and microorganisms. The penultimate step in the pathway is catalyzed by the enzyme lumazine synthase (LS). 1 One of the most distinctive characteristics of this enzyme is the structural quaternary divergence found in different species. The protein exists as pentameric and icosahedral forms, built from practically the same structural monomeric unit. The structure of the monomer consists of four repeated ␤-strand/␣-helix motifs producing a sandwich of four parallel ␤-strands surrounded by four ␣-helices, two on each face of the ␤ sheet (1). In all LS studied, the topologically equivalent active sites are located at the interfaces between adjacent subunits in the pentameric modules (2).
Bacilliaceae express a bifunctional enzyme complex with lumazine synthase and riboflavin synthase activity. Three ␣-subunits (riboflavin synthase) enclosed by 60 ␤-subunits (lumazine synthase) form a protein particle of ϳ1 MDa (1,3). The three-dimensional structure of the lumazine synthase/riboflavin synthase from Bacillus subtilis complexed with a substrate analogue has been determined (4). This structure consists of 60 ␤-subunits (lumazine synthase monomers) arranged as 12 pentamers giving rise to a capsid exhibiting icosahedral 532 symmetry. Two other icosahedric LS have been described by x-ray crystallography. Spinach and thermophilic bacterial Aquifex aelicus LS also exhibit icosahedral 532 symmetry. The three proteins form 1-MDa spherical capsids of 60 LS subunits with icosahedral symmetry. The icosahedral LS structures superimpose very well, highlighting the conservation of the overall folding and quaternary arrangement despite the low sequence homology between them (5).
Four structurally characterized LS assemble as pentamers in their native form and do not further associate to form an icosahedral capsid. These include fungal Magnaporthe grisea, yeast Saccharomyces cerevisiae, and Schizosaccharomyces pombe and bacterial Brucella abortus LS (5). Although the four pentameric enzymes fold in a similar arrangement, the postulated reasons for the lack of icosahedral order differ. Superposition of the pentameric LS shows that the loop connecting the helices ␣4 and ␣5 is critical for preventing the formation of capsids. In all icosahedral LS a pentapeptide kink located in this loop is essential for pentamer-pentamer contacts (5,6). The yeast and fungal pentameric enzymes have insertions of different length in this loop that change its overall orientation and disrupt potential contacts to neighboring subunits (7). We have previously analyzed the divergence in macromolecular assembly between pentameric and icosahedral LS enzymes. In this regard, a high degree of divergence was observed between the sequence from B. abortus LS and the other pentameric structurally characterized members of the family (8). BLS has a 3-residue insertion between helices ␣4 and ␣5 that contribute to form a continuous undistorted helix, unable to form the capsid-stabilizing kink (6). Thus, the BLS structure clearly diverges from pentameric and icosahedric enzymes because of the lack of this critical loop. The different orientation of this straight helix is compensated by a longer loop bridging the helix structure with the contiguous sheet. This unusual divergence prompted us to further investigate the quaternary arrangement of BLS. In the present work, we demonstrate by means of solution light scattering and x-ray structural analysis that BLS assembles as a very stable dimer of pentamers, representing a third category of quaternary assembly for LS. We also describe, by spectroscopic studies, the thermodynamic stability of this oligomeric protein and postulate a mechanism for dissociation/unfolding of its macromolecular assembly.

Expression of the LS Protein
The Brucella sp. LS gene was cloned in pET11a vector (Novagen) as reported previously (9). The plasmid was used to transform BL21(DE3) strain Escherichia coli-competent cells (Stratagene, La Jolla, CA). Ampicillin-resistant colonies were grown until A 600 ϭ 1.0 in LB medium containing 100 g/ml ampicillin, at 37°C with agitation (300 rpm). Five milliliters of this culture was diluted to 500 ml and grown to reach an A 600 of 1.0. At this point the culture was induced adding 1 mM isopropyl-1-thio-␤-D-galactopyranoside and incubated for 4 h at 37°C with agitation (300 rpm). The bacteria were centrifuged at 15,000 ϫ g during 20 min at 4°C.

Protein Purification and Refolding
BLS protein was successfully expressed as inclusion bodies by transformation of strain BL21(DE3) E. coli-competent cells (9). The inclusion bodies were solubilized in 50 mM Tris, 5 mM EDTA, 8 M urea pH 8.0 at room temperature overnight with agitation. The solubilized material was refolded by dialysis during 72 h against phosphate-buffered saline containing 1 mM dithiothreitol. This preparation was purified in a Mono-Q column in a fast-protein liquid chromatography apparatus (Amersham Biosciences, Uppsala, Sweden) using a linear gradient of buffer B (50 mM Tris, 1 M NaCl, pH 8.5). The peak enriched with BLS was further purified on a Superdex-200 column with phosphate-buffered saline buffer, 1 mM DTT. The purity of the BLS preparation was determined on SDS-15% (w/v) polyacrylamide gels. Purified BLS was concentrated (10 mg/ml), frozen in liquid N 2 , and stored at Ϫ20°C.

Circular Dichroism
BLS samples were diluted in 50 mM sodium phosphate, pH 7.0, 1 mM DTT, with increasing concentration of denaturants (urea or GdnHCl). All experiments were performed at 25°C, and samples were incubated at least 2 h before taking CD measurements. Spectra were measured on a spectropolarimeter (JASCO J-810) using either 0.1-or 0.5-cm path length quartz cells. Unfolding was monitored by far-UV CD (260 -200 nm) and expressed as the percentage of molar ellipticity at 222 nm as a function of denaturant concentration. The molar ellipticity of the protein incubated without GdnHCl was taken as 100%. For renaturation tests, BLS was first denatured in high concentration of GdnHCl (6 M), and renaturation was induced by overnight dialysis against 50 mM sodium phosphate, pH 7.0, 1 mM DTT. After this treatment, the CD and SLS signals typical of the native protein were recovered.

Intrinsic Fluorescence Measurements
Emission spectra were carried out by excitation of the samples at 295 nm, and data collection from 300 to 450 nm, using 3-nm band passes for both excitation and emission. Experiments were carried out in 50 mM sodium phosphate, pH 7.0, 1 mM DTT in the presence of increasing concentrations of urea or GdnHCl. Samples were incubated for at least 2 h prior to taking fluorescence measurements. All fluorescence emission spectra were measured at 25°C on a Jasco FP-770 spectrofluorometer.

Thermal Denaturation Monitored by CD
BLS samples were incubated in 50 mM sodium phosphate, 1 mM DTT, pH 7.0 (in the presence or absence of 2 M GdnHCl) or in 50 mM buffer citrate, 1 mM DTT, pH 4.5. Thermal denaturation was conducted by slowly increasing the temperature with a Peltier system (Jasco). The range of temperature scanning was 25-95°C at a speed of 4°C/min. Molar ellipticity at 220 nm was measured every 0.5°C. Fast or slow cooling back to 25°C (from 95 to 25°C at a speed of 1°C/min) did not show a recovery of ellipticity demonstrating the irreversibility of the thermal unfolding. Thus the temperature midpoint of the thermal transition was considered as an apparent T m .
pH Dependence of LS Unfolding pH-induced BLS dissociation and unfolding was evaluated by diluting protein samples (5 M monomers) in 50 mM citrate, 1 mM DTT, pH 6.0 -2.5 or in 50 mM sodium phosphate, 1 mM DTT, pH 7.5-7.0. After 1-2 h of incubation at room temperature, CD and SLS signals were determined. ANS fluorescence emission was measured in the same buffer conditions as CD and SLS experiments, aggregating to the samples 50 M ANS. Values were normalized to ANS fluorescence emission in the same buffer. Excitation and emission were set at 365 and 470 nm, respectively.

Determination of Molecular Weight of BLS by Static Light Scattering
The weight average molecular weight (M w ) of BLS under different conditions was determined on a Precision Detectors PD2010 lightscattering instrument tandemly connected to an high-performance liquid chromatography system and an LKB 2142 differential refractometer. In general, 20 -100 l of LS (0.3-1 mg/ml) was loaded on a Superdex 200 HR-10/30 (24 ml) or a Sephadex G-25 (1 ml) column and eluted with 50 mM phosphate buffer, 1 mM DTT under different pH, urea, GdnHCl, and NaCl conditions. The 90°light scattering and refractive index signals of the eluting material were recorded on a PC computer and analyzed with the Discovery32 software supplied by Precision Detectors. The 90°light scattering detector was calibrated using bovine serum albumin (M w : 66.5 kDa) as a standard. Prior to the injection in a size exclusion chromatography column (SEC), each sample was preincubated for 1-2 h at room temperature in the elution buffer.

Thermodynamic Parameters
Thermodynamics evaluation of GdnHCl-induced unfolding of BLS Ku1 Ku2 was fitted to a two-step model: . The first step (N 10 7 2N 5 ) represents the dissociation of decameric BLS (N 10 ) in two folded pentamers (N 5 ), whereas the second step (N 5 7 5U) represents the concomitant dissociation and unfolding of the pentameric structure (N 5 ) in monomeric subunits (U). Because both steps are well resolved from each other, their thermodynamic parameters were independently analyzed using the lineal extrapolation method (Equation 1) assuming a two-state transition model for each step (10), where ⌬G U is the free energy of unfolding of a protein at a given denaturant concentration, ⌬G H 2 O is the free energy of unfolding in the absence of denaturant, and m the dependence of the free energy on denaturant concentration ([D]).
Step 1-The first step was monitored by SLS in the range of 1.5-2.2 M GdnHCl. The concentration at equilibrium of the N 10 and N 5 species of BLS were calculated using Equations 4 and 5 derived from rearrangement of Equations 2 and 3, where M w represents the weight average molecular weight of BLS. ccN 10 , ccN 5 , and cc T represent the decamer, pentamer, and total protein concentration in millgrams/ml, respectively. [N 10 ] and [N 5 ] represent the molar concentration and M w10 and M w5 the molecular weight of the decameric (174.4 kDa) and pentameric (87.2 kDa) species of BLS, respectively. The equilibrium constant K U(1) and the free energy change ⌬G U(1) for Step 1 are defined in Equations 6 and 7.
The Step 3-The second step was monitored by CD and FT in the range of 2.4 and 3.5 M GdnHCl. The equilibrium constant K U(2) and the free energy change ⌬G U(2) for this transition are defined in Equations 8 and 9.
The total protein concentration in monomer units (P T ) and the fractional population in the native (F N ) and unfolded (F U ) states were calculated using Equations 10 -12, (Eq. 10) where Y is the experimental spectroscopic value, [D] is the GdnHCl concentration, F and U are the intercepts, and m f and m u are the slopes of the pre-and post-unfolding baselines, respectively. Combining Equations 1 and 8 -12, we obtain the general equation (Equation 13) as follows.
⌬G H 2 O (2) and m (2) values of Step 2 were obtained fitting the experimental data measured by CD and FT to Equation 13 by a non-linear square fit method. The midpoint of this transition [D] 50% was calculated as in Equation 14.

Crystallization, Data Collection, and Structure Determination
In addition to the previously obtained BLS crystals, which led to the resolution of its three-dimensional structure (6), two further crystal forms of BLS were obtained in this work by means of the hanging drop, vapor diffusion method. The first form was diamond-like crystals obtained using 30% (w/v) polyethylene glycol 400, 0.1 M sodium acetate in 0.1 M MES buffer, pH 6.5, which diffracted to 2.9 Å and belongs to the trigonal space group P3 1 21. Furthermore, we obtained plate-like crystals using 12% (w/v) polyethylene glycol 4000, 0.1 M sodium acetate in 0.1 M HEPES buffer, pH 7.5, which diffracted to 3.0 Å and belongs to the monoclinic space group P2 1 . X-ray diffraction data were collected both at our in-house x-ray source, a Bruker M18XH6 MAC Science rotating anode interfaced to a Siemens X-1000 multiwire area detector and at the D03B protein crystallography beamline at the Laboratorio Nacional de Luz Síncrotron, Campinas, Brazil (11). Data reduction and processing were carried out with the programs MOSFLM, Scala, and Truncate from the CCP4 suite (12). Crystal packings were determined using the molecular replacement procedure as implemented in the AMoRe package (12), with the previously solved BLS structure as a search model. Crystallographic symmetry construction was carried out using the program O (13).

Structural Analysis
Total accessible surface area and buried surface areas of interaction (⌬ASA) between monomers and pentamers in pentamer and decamer structures, respectively, were calculated with Surface Racer 1.2 program (14) using an implementation of the Lee and Richards (15) algorithm and a probe radius of 1.7 Å. Intermolecular polar and non-polar interactions were calculated with the Molmol 2k.2 (16) and Contacts of Structural Units (17) programs. Any pair of atoms is considered to be in contact if the distance between them is less than 4 Å.

BLS Is a Stable Dimer of Pentamers in Solution-Previous
studies (6,9) have shown that BLS does not assemble as a 1-MDa capsid oligomer, having a retention time in SEC compatible with an apparent molecular mass of 90 kDa. This behavior suggested that the oligomeric structure of this protein in solution was a pentamer. However, the estimation of the molecular weight of a protein by its retention time on a SEC column is a method prone to artifacts. Additionally, the sequence divergence of BLS, compared with other pentameric LS, prompted us to re-analyze the quaternary structure of the protein by spectroscopic techniques such as static light scattering (SLS).
SLS experiments show that the protein has a molecular mass of 180 kDa in solution, corresponding to an assembly of two pentamers (decameric arrangement of the 18-kDa polypeptide chain). To characterize this new quaternary arrangement, we studied the thermodynamic stability of BLS, evaluating the unfolding of the protein induced by the common chemical denaturants (urea and GdnHCl) and pH.
Preincubation of BLS with increasing concentrations of urea shows that the enzyme remains as a stable dimer of pentamers, with no detectable change in its quaternary structure (180 kDa, determined by SLS, data not shown) as well as in its tertiary and secondary structures as followed by tryptophan fluorescence and CD (Fig. 1). The absence of structural changes, even in 8 M urea, indicates that the quaternary arrangement of BLS is very stable.
Conversely, GdnHCl produces a cooperative and reversible change in the tertiary structure reflected by a decrease in tryptophan fluorescence emission (Fig. 1A). This measure senses the environment of Trp-22, the unique tryptophan in BLS monomers that is located on the active site at the monomer-monomer interface (6). In addition, GdnHCl incubation (6 M) produces a complete loss of secondary structure of BLS as monitored by CD spectra (Fig. 1B). The differential effect of GdnHCl and urea cannot be explained by their differences in ionic strength. The behavior of BLS in 8.0 M urea and in the presence of 1 M NaCl is superimposable with its described stability in absence of salt (data not shown), implying that unfolding with GdnHCl seems to be due to more specific interactions of the guanidinium cation with the protein (18,19). Thus, we used GdnHCl-induced denaturation to study the mechanism of BLS unfolding. ( Fig. 2A). In the first step, observed between 1.5 and 2.2 M GdnHCl, the SLS signal intensity of the protein is reduced to half the value measured in the absence of denaturant. However, no changes in the far-UV CD spectra and tryptophan fluorescence of the protein are observed in this range (Figs. 1A and 2A). Thus, intrinsic tryptophan fluorescence is completely insensitive to the change in the quaternary structure of the protein, indicating that the dissociation does not modify significantly the environment of Trp-22. These results point to the existence of a dissociation phenomenon of the decameric structure of BLS into two pentameric subunits, with no further changes in tertiary and secondary structure. On the other hand, the second step observed between 2.4 and 3.5 M GdnHCl, shows a 5-fold decrease in the SLS signal of BLS concomitantly to the disappearance of its far-UV CD signal at 222 nm, and a significant decrease in the fluorescence intensity of its single tryptophan residue (Figs. 1A and 2A). These results are interpreted as an unfolding and loss of the tertiary and secondary structures of the protein coupled to the dissociation of its pentameric arrangement in five monomeric subunits (18 kDa). The overlapping of the changes observed in the second step by SLS, CD ( Fig. 2A) and fluorescence (Fig. 1A) clearly supports our model. These results were further verified by SLS coupled to gel filtration chromatography (Fig. 2B), where it was possible to isolate a 90-kDa intermediate at 2 M GdnHCl. The fact that an intermediate of 90 kDa is stable enough to be detected by this methodology suggests that the decameric assembly is composed of two previously associated pentamers.
The dissociation of the decamer to folded pentamers is also observed when BLS is incubated at acidic pH (Fig. 3). This phenomenon is evidenced by a 2-fold decrease in the SLS signal intensity in the range of pH 4.0 -5.0 as compared with the value measured at pH 7.0. ANS binds to exposed hydrophobic surfaces in partially folded intermediates with higher affinity than to native or completely unfolded proteins (20 -22). This binding result in a marked increase in fluorescence emission compared with the free ANS. We found that the dissociation of BLS at pH in the range 6.0 -4.0 is a reversible process that occurs without significant exposure of hydrophobic patches and changes in secondary structure as monitored by ANS fluorescence and CD spectra (Fig. 3).
At pHs below 4.0 the protein shows an irreversible unfolding, evidenced by a loss of ellipticity at 222 nm and a dramatic increase in binding to ANS. Clearly, both changes are coupled, implying that at acidic pHs the protein exposes hydrophobic surfaces when secondary structure is lost. The light scattering response cannot be accurately monitored below pH 4.0, presumably because of aggregation.

Thermodynamic Stability of BLS Measured by Chemical
Denaturation-To describe the thermodynamic stability of BLS, we studied the dissociation and unfolding steps taking advantage of the differential effect of GdnHCl at distinct concentration ranges. Both transitions (Steps 1 and 2) were shown to be highly cooperative and reversible. The dissociation step was characterized by SLS using GdnHCl up to 2.2 M, in a condition that does not disturb the tertiary and secondary structures of the pentamer. This assumption is supported by the facts that the circular dichroism spectrum is not modified (see Fig. 2A) and the intrinsic tryptophan fluorescence does not vary upon addition of GdnHCl (see Fig. 1A). The SLS determination of the equilibrium between the dimer of pentamers and the pentamer gives a ⌬G of 90 Ϯ 20 kJ/mol decamer as estimated by the LEM method assuming a two-state transition (see "Experimental Procedures"). This value indicates that the protein remains as a decamer under physiological conditions (estimated K D ϭ 2.48 ϫ 10 Ϫ16 M).
On the other hand, the transition from folded pentamer to unfolded monomers is a highly cooperative process that can be measured by GdnHCl denaturation at higher concentrations followed by tryptophan fluorescence (Fig. 1A), SLS, and circular dichroism (Fig. 2A). These signals show a sharp and overlapping change around 2.4 -3.5 M GdnHCl. Thus, all three spectroscopic analyses rule out the existence of a populated intermediate during this transition. The dependence of [D] 50% with protein concentration (Table I) clearly supports the model of a two-state transition from a folded pentamer to unfolded monomers. Thermodynamic analysis of this equilibrium shows that the ⌬G is 330 Ϯ 30 kJ/mol pentamer (Table I; see "Experimental Procedures" for details). In agreement with this analysis, GdnHCl denaturation of BLS previously incubated at pH 5.0 (as a dissociated pentamer, see Fig. 3), gives a ⌬G of 300 Ϯ 30 kJ/mol pentamer, clearly indicating the accuracy of the determined thermodynamic parameter for the stability of the pentamer.
Thermal Denaturation of BLS-The stability of BLS to thermal denaturation was followed by measuring the molar ellipticity at 222 nm as a function of increasing temperature, as shown in Fig. 4. The enzyme shows a sharp decrease of ellipticity between 85 and 95°C with an apparent T m of 88 Ϯ 2°C. The loss of secondary structure is not recovered after slow cooling of the samples, indicating that an irreversible unfolding  phenomenon of BLS takes place under these conditions. Thermal denaturation of the protein, previously incubated in two different conditions that produce the dissociation of the decamer into pentamers (2.0 M GdnHCl and pH 4.5), produces a very similar decrease of about 20°C in the thermal stability of BLS (see Fig. 4), further demonstrating that dissociation and denaturation can be uncoupled under these conditions. Thus, this shift of 20°C can be attributed to the contribution of the pentamer-pentamer interface to the overall stability of BLS.
Quaternary Arrangement of BLS Is Confirmed by X-ray Crystallography-X-ray structure analysis confirmed the quaternary arrangement of BLS. Brucella sp. LS has been crystallized in three different forms (Refs. 6 and 23 and this work). One of these forms (space group P3 1 21) has a high content of water (around 70%) and diffracts x-rays to medium resolution (2.9 Å) with a single pentamer in the asymmetric unit. Modifying the reagent conditions we obtained a second crystalline form (space group P2 1 ) that has two pentamers in the asymmetric unit. The previously obtained crystal form (space group R32) is more densely packed and diffracted to 2.7-Å resolution with a single pentamer in the asymmetric unit, allowing for the resolution of the three-dimensional structure of the enzyme (RCSB Protein Data Bank code 1DI0). Despite the different types of crystal arrangement (Fig. 5A), all crystal forms show the same quaternary arrangement. The previously described pentamer forms a tightly packed dimer of pentamers (Fig. 5B), with the N terminus of the straight ␣ helix (composed of helices ␣4 and ␣5, as named in other LS) and the loop connecting the contiguous sheet making a protuberant surface that tightly fits in the neighboring pentamer. This region of the polypeptide chain has a very high content of histidines and phenylalanines (7 His and 4 Phe on a stretch of 18 residues) (Fig. 5C). Tables II and III show the structural analysis about the nature of the monomer-monomer and pentamer-pentamer interfaces. Each of the monomers that gives rise to the decameric structure buries ϳ45% of its ASA, established mainly by hydrophobic contacts with two neighboring monomers (pentamer assembly) and between pentamers (decamer assembly). Each monomer buries 35.5% of its ASA to form the pentamer assembly. The resulting pentamer is an intertwined structure with the interface predominantly making non-polar contacts (64.8% of non-polar ⌬ASA, 245 van der Waals contacts per monomer). Each monomer also makes 6 hydrogen bonds and 1 salt bridge that stabilize the pentamer assembly. Two to three bridging waters per monomer make an additional 6 hydrogen bonds. Each monomer buries 9.2% of its ASA to form the decamer assembly. This interface is also mainly hydrophobic (61.5% of non-polar ⌬ASA, 92 van der Waals contacts per pentamer). Each pentamer also makes 2 hydrogen bonds that stabilize the decamer assembly. Noteworthy, 10 phosphates make 20 bridging hydrogen bonds stabilizing this interface (the details to be published elsewhere).
Integrated Mechanism of Decameric BLS Unfolding-The crystallographic structural analysis, together with the solution FIG. 5. Quaternary assembly of Brucella sp. LS. A, three different crystal forms of BLS show the same quaternary arrangement. BLS folds as a dimer of pentamers (each decamer is depicted in a different color) with spool-like shape. B, three-dimensional structure of the decameric BLS. Three monomers are shown in different colors to highlight the nature of this oligomeric intertwined structure. The ribbons and surface diagrams clearly show how the continuous helix 4 protrudes into the adjacent pentamer. C, the surface of BLS forming the pentamer-pentamer interface is shown. Histidine residues are colored in red, phenylalanines in yellow. The 5 C-terminal tips of helix 4 are highlighted in the periphery, whereas the internal circle of histidines is shown in the center, close to the pore that goes through the center of the pentamer. studies described above, allowed us to postulate a comprehensive model for BLS unfolding (Fig. 6). As shown, the decameric assembly dissociates in two different conditions to stable folded pentamers, with an estimated ⌬G of 90 Ϯ 20 kJ/mol decamer. Thus, BLS would be capable to shift from a decameric to a pentameric quaternary state at acidic pH, conserving the stability of the protein. The effect of the pH on dissociation would be explained by the high density of histidines at the pentamerpentamer interface (see Fig. 5C). Histidine residues become protonated at pH levels below 6.0, thus producing charge repulsion and loss of contacts resulting in the dissociation of the dimer of pentamers. In contrast, the acidic pH does not disturb the stability of the monomer-monomer interface, because ϳ35% of the accessible surface area of the monomeric polypeptide is buried in monomer-monomer contacts of hydrophobic nature (Tables II and III). In agreement, GdnHCl-induced pentamer unfolding at neutral and acidic pHs gives approximately the same value of ⌬G (300 -330 kJ/mol). This high value of free energy would be the product of the tight and intertwined nature of the monomer-monomer interface, stabilized for a very high number of van der Waals contacts (Table III) resistant to the effect of the low pH. DISCUSSION Common cellular proteins exhibit only marginal stabilities, with free energies of stabilization of the order of 50 kJ/mol (24). In contrast, quaternary interactions have a dominating role in the stabilization of small oligomeric proteins (25). Unfolding pathways for oligomeric proteins have been shown to vary significantly. Subunit dissociation could occur before or after polypeptide unfolding, or the two reactions could occur simultaneously without significant population of equilibrium intermediates (26). Examples of dissociation to folded monomers followed by monomer unfolding have been described (27). Several examples of dissociation coupled to unfolding have also been described (28 -30).

TABLE II
Analysis of the surface areas at the monomer-monomer and the pentamer-pentamer interfaces Polar and non-polar buried surface areas (⌬ASA) of a monomer in monomer-monomer and pentamer-pentamer interactions are shown. All the calculations were made using a probe radius of 1.7Å. The percentage of the buried surface area of each monomer in a hypothetical situation being "free" in solution with respect to the total accessible surface area is represented in parentheses. TABLE III Polar and non-polar contacts, bridging phosphate ions, and water molecules in monomer-monomer and pentamer-pentamer interfaces An acceptor and a donor atom form a hydrogen bond if the distance between them is up to 3.4 Å with an angle of more than 90°. Oppositely charged atoms in close proximity are defined as forming a salt bridge if they are less than or equal to 3.9 Å apart. All the non-polar contacts have been calculated using 4.0-Å cut-off. of structural adaptation during quaternary divergence from pentameric to decameric lumazine synthases. In this sense, the continuous undistorted helix 4 of BLS (see introduction) produces a complementary surface between pentamers composed of hydrophobic interactions, in which there is a marked presence of histidine and phenylalanine residues (Fig. 5C). Additionally, phosphate ions help to cement this interface by means of bridging hydrogen bonds. The fact that we have been able to isolate the intermediate folded pentamer in different conditions would allow for detailed mutagenesis and kinetic folding studies of this unusual protein interface. It would allow also for protein engineering procedures in the use of BLS as protein carrier for the development of immunogens (see below). In contrast, the pentamer presents an intertwined structure with each monomer interacting extensively with its adjacent monomers via multiple hydrophobic, hydrogen bond, and salt bridge interactions. As a consequence, dissociation of the pentamer and unfolding of the resulting monomers occurs as a concerted mechanism. Even though the macroscopic methods employed in this study support a two-state model, we cannot rule out the existence of intermediates. Because residues of neighboring monomers form the active site of the enzyme, there are no structural reasons for the existence of folded monomers.
The apparent transition temperature (T m ) of BLS is very high, typical of a protein from a thermophilic organism. Unexpectedly high T m from other LS were previously described (5). Thermal unfolding of BLS shows an intermediate behavior between pentameric and icosahedric LS, as expected for its decameric order. The pentameric S. cerevisiae LS melting curve has a maximum at 74.1°C, whereas icosahedric B. subtilis showed a sharp melting transition at 93.3°C as determined by differential scanning calorimetry (DSC) (5). Circular dichroism experiments described in this work show that the melting point of BLS is 88 Ϯ 2°C, indicating a thermal stability closer to that of higher order LS. Preliminary differential scanning calorimetry experiments show a similar melting transition temperature (results not shown). Zhang et al. (5) postulated that the predominantly hydrophobic nature of BLS monomer-monomer interface as compared with other LS, would produce a higher thermal stability. In addition, previous dissociation of BLS to pentamers (see Figs. 2 and 3) yields a T m of 70 Ϯ 2°C, close to the value of the pentameric S. cerevisiae LS. Thus, additional thermal stability of BLS can be attributed to the formation of the pentamer-pentamer interface. Our results and those of Zhang et al. show a clear relationship between quaternary order of pentameric assemblies and thermal stability. Moreover, the results shown here make BLS an excellent model to study the mechanisms of folding and thermodynamic stability of an oligomeric stable protein.
The new quaternary arrangement presented here raises the question about the biological implications of decameric LS. Moreover, a close inspection on the sequences of LS (8) suggests that the homologues from Rhodococcus erythropolis, Sinorhizobium meliloti, Mesorhizobium loti, Pseudomonas fluorescens, and possibly others are also decameric LS. A striking feature of the BLS structure is that the association of pentamers creates an internal space, delimitated by the external pores (8.5 Å of internal diameter) and two internal pores of around 23 Å of diameter (see Fig. 5C). These pores delimitate an empty space with the shape of a truncated cone and with a height of 60 Å. Although substrate channeling has to be ruled out because of the lack of an internal riboflavin synthase as in the case of icosahedric enzymes (31), the channels along the 5-fold axes (pores of 8.5 Å of diameter) could allow the passage of substrates but appear too narrow for the exit of lumazine (32).
Thus, this structure could be acting as a reservoir of recently synthesized 6,7-dimethyl-8-ribityllumazine. The fact that BLS dissociates at acidic pHs, like those encountered at phagolysosomes, suggest a putative role of BLS as a transporter and reservoir of lumazine for riboflavin synthesis in the demanding conditions for surviving found inside eukaryotic cells (33). Experimental work is needed to demonstrate the validity of this hypothesis, but the low activity of this enzyme (9) and the existence of fluorescent products associated to recombinantly expressed purified BLS (resistant to urea dissociation) support this idea.
BLS folds as a highly stable dimer of pentamers and is a highly immunogenic protein (23,34,35). These characteristics resemble that of the B subunits of E. coli heat-labile enterotoxin (EtxB) and cholera toxin (CtxB). Both toxins assemble in vivo into exceptionally stable homopentameric complexes, which maintain their quaternary structure in a range of conditions that would normally be expected to cause protein denaturation (36). These remarkable stability properties, as well as the inherent immunogenicity of EtxB and CtxB pentamers, have prompted considerable interest in their use as vaccine delivery vehicles (36). Noteworthy, BLS is more stable to thermal denaturation than EtxB and CtxB toxins (T m of 84 and 75°C, respectively) (37,38). Thus, BLS is a promising potential carrier for the polymeric delivery of antigens or epitopes. The presence of ten sites of linkage and the natural disordered conformation of its N termini (6) linked to its high stability indicate that BLS is a potential candidate for the development of subunit vaccines.