E. Chiancone, CNR Center of Molecular Biology, Department of Biochemical Sciences A. Rossi Fanelli, University La Sapienza, Piazzale Aldo Moro, 5, 00185 Roma, Italy. Fax: +39 064440062, Tel.: +39 0649910761, E-mail: email@example.com
The stability of the dodecameric Listeria innocua ferritin at low pH values has been investigated by spectroscopic methods and size-exclusion chromatography. The dodecamer is extremely stable in comparison to the classic ferritin tetracosamer and preserves its quaternary assembly at pH 2.0, despite an altered tertiary structure. Below pH 2.0, dissociation into dimers occurs and is paralleled by the complete loss of tertiary structure and a significant decrease in secondary structure elements. Dissociation of dimers into monomers occurs only at pH 1.0. Addition of NaCl to the protein at pH 2.0 induces structural changes similar to those observed upon increasing the proton concentration, although dissociation proceeds only to the dimer stage. Addition of sulfate at pH values ≥ 1.5 prevents the dissociation of the dodecamer. The role played by hydrophilic and hydrophobic interactions in determining the resistance to dissociation of L. innocua ferritin at low pH is discussed in the light of its three-dimensional structure.
The effect of low pH on protein structure is related to the electrostatic forces that originate from the changes in net charge and charge distribution attendant with the addition of protons. In monomeric proteins the progressive protonation of amino acid residues on decreasing the pH gives rise to repulsive interactions that lead to loss of secondary and tertiary structure and hence can be exploited to study protein stability . Typically, a relatively fully unfolded conformation is attained in the vicinity of pH 2 because stabilizing salt-bridges are removed and buried side chains become protonated. Several proteins do not unfold completely, however, and adopt relatively compact, soluble conformational states that differ from the native structure as indicated by their spectroscopic properties . In oligomeric proteins the acid denaturation is not limited to the destruction of secondary and tertiary structure of each subunit, but entails the perturbation of intersubunit interactions. In oligomeric proteins, therefore, the structural changes induced by low pH provide additional information on the forces involved in subunit recognition and association, and on the intermediates that may be formed during the assembly/disassembly processes.
The mechanism of acid-induced denaturation has been studied extensively for several monomeric [2,3] and a few oligomeric proteins [4–9]. Most oligomeric proteins were found to be fully disassembled at pH 2.6 . However, two systems of extreme pH stability have been described that undergo dissociation at pH values lower than 2.0, namely the B subunit pentamer of Escherichia coli heat-labile enterotoxin  and the glutamate dehydrogenase hexamer from the hyperthermophile Pyrococcus furiosus. In both systems the molecular basis for the extreme stability of the quaternary assemblage has been ascribed to the presence of specific intersubunit interactions revealed by their crystal structure.
Ferritin, the ubiquitous iron storage protein, represents a well known polymeric assemblage that is highly resistant to chemical and physical denaturants. The quaternary structure is highly conserved and consists of a protein shell that can harbour a micellar iron core and is composed of 24 subunits assembled with 432 symmetry . Natural vertebrate ferritins are copolymers of two different subunits, the L and H chains, that dissociate under extreme pH conditions, namely below pH 2.8 and above pH 10.6 [13–15]. However, the recombinant L homopolymer begins to dissociate at pH values below 1.5 due to the presence of a specific salt-bridge within the subunit tertiary structure . Recently, an unusual dodecameric ferritin has been extracted from the Gram-positive bacterium Listeria innocua. The monomer sequence shows no similarity with ferritin sequences but approximately 30% identity to the proteins of the Dps family (DNA-binding proteins from starved cells) .
The L. innocua ferritin monomer folds into a four-helix (A–D) bundle that resembles closely that of E. coli Dps and of all other ferritins, but lacks the short C-terminal E helix, which in typical ferritins provides the stabilizing interactions at the fourfold symmetry axes. Helices B and C are connected by a long loop containing a short helix (BC) as in the E. coli Dps monomer .
The structural similarity between L. innocua ferritin and the Dps proteins extends to the quaternary assembly of the 12 subunits that form a protein shell with 23 symmetry [18,19]. The L. innocua ferritin dodecamer can, therefore, be envisaged as four trimers placed at the vertices of a tetrahedron, with six two-fold axes and four threefold axes as symmetry elements. Along the three-fold axes there are two nonequivalent environments that have been designated ‘ferritin-like’ and ‘Dps-like’ as the former corresponds to the three-fold interactions in the ferritin 24-mer and the latter is present in Dps, but not in typical ferritins. Despite the different assemblage, the apoferritin shell of L. innocua is endowed with the high temperature stability that characterizes all other ferritins and is at the basis of their isolation procedure . This observation prompted the present study on the acid stability of L. innocua apoferritin. Because the protein shell is assembled from a single subunit, the analysis of the denaturation and renaturation processes is facilitated, without the possible difficulties involved in studying intermediates formed by different subunit types.
The L. innocua ferritin dodecamer exhibits a remarkable acid stability, greater than that of typical ferritins. It is one of the very few oligomers known to date that undergo disassembly at pH values below 2.0. Dimers are the prevalent dissociation products. Neutralization of the acid-dissociated protein leads to the reacquirement of the quaternary structure and of the functional properties that characterize the native protein.
Materials and methods
L. innocua ferritin purification and characterization
L. innocua ferritin was purified from packed bacterial cells according to the method described by Bozzi et al. 1997 . Iron was removed from native ferritin, which contains about 50–100 iron atoms per dodecamer, by incubation for 24 h in 50 m m Mes/NaOH, at pH 6.0, containing 0.3% (w/v) sodium dithionite and subsequent chelation with 2,2′-bipyridyl. Apoferritin concentration was determined spectrophotometrically by using the molar absorptivity 2.59 × 105m−1·cm−1 at 280 nm , based on the dodecamer molecular mass of 216 kDa. The iron content of the native and/or reconstituted ferritin was determined as the 2,2′-bipyridyl complex at 520 nm  and/or by atomic absorption.
Iron incorporation experiments were performed using solutions of ferrous ammonium sulfate prepared in Thunberg tubes, kept anaerobically under a nitrogen atmosphere and used within few hours from their preparation. The iron incorporation kinetics were followed spectrophotometrically at 310 nm and 25 °C upon addition of iron to apoferritin solutions equilibrated in air. The solutions were maintained under continuous stirring during the course of the experiments. The extinction coefficient of micellar iron at 310 nm was taken as 450 (ε 1%, 1 cm). As a control, the rate of Fe(II) autoxidation was measured in parallel. After iron incorporation, the samples were subjected to gel electrophoresis under nondenaturing conditions in 6% (w/v) polyacrylamide gels. The gels were stained for iron with potassium ferrocyanide and for protein with Coomassie Blue.
Analysis of the state of association
Analysis of the state of association was carried out by size-exclusion chromatography experiments at 20 °C on a Superose 12 column (Pharmacia) eluted with 20 m m sodium phosphate, pH 7.0 containing 0.15 m NaCl at a flow rate of 0.5 mL·min−1 controlled by a Dionex gradient pump. After 24 h incubation at pH 1.0–2.0, the samples were diluted 20-fold into the column injection loop. The Superose column was calibrated with horse spleen apoferritin (440 kDa, elution volume Ve = 7.8 mL), rabbit muscle aldolase (161 kDa, elution volume Ve = 9.2 mL), horse liver alcohol dehydrogenase (80 kDa, elution volume Ve = 9.8 mL), bovine serum albumin (66 kDa, elution volume Ve = 10.0 mL), ovalbumin (45 kDa, elution volume Ve = 10.3 mL), and cytochrome c (12 kDa, elution volume Ve = 12.4 mL).
Sedimentation velocity experiments were conducted at 49 000 g and 20 °C on a Beckman Optima XL-A analytical ultracentrifuge equipped with absorbance optics. The protein concentration was 0.3 mg·mL−1. Data were collected at 280 and 310 nm at a spacing of 0.005 cm with three averages in a continuous scan mode and were analyzed with the program DCDT (Walter Stafford, Boston Biomedical Research Institute). The sedimentation coefficients were corrected to s20,w using standard procedures. In all experiments the buffer used was 20 m m sodium phosphate, pH 7.0.
pH dependence experiments
L. innocua ferritin (0.03–1.0 mg·mL−1) was incubated for 24 h at 20 °C at pH 1.0 (100.0 m m HCl), pH 1.5 (31.6 m m HCl), pH 2.0 (10.0 m m HCl) and at pH 2.0 in the presence of 31.6 m m NaCl (pCl 1.5) or 100.0 m m NaCl (pCl 1.0). The pH of the solutions was measured with an InLab 422 electrode (Mettler-Toledo AG) connected to a Corning P 507 ion meter before and after the addition of the protein. After 24 h incubation at 20 °C, the samples were analyzed by CD and fluorescence spectroscopy and by size-exclusion chromatography. In reassociation studies, the protein (30 µg·mL−1) was incubated for 24 h at pH 1.0 or at pH 2.0 pCl 1.0 and then diluted 20-fold in 20 m m phosphate pH 7.0 containing 0.15 m NaCl. After 24 h at 20 °C the protein was concentrated by ultrafiltration on a YM 30 membrane (Amicon) and analyzed by fluorescence and CD spectroscopy, by size-exclusion chromatography and by analytical ultracentrifugation. The iron incorporation capacity was also determined.
Intrinsic fluorescence emission and light scattering measurements were carried out with a LS50B PerkinElmer spectrofluorimeter using a 1 cm pathlength quartz cuvette. Intrinsic fluorescence emission spectra were recorded at 300–400 nm (1 nm sampling interval) with the excitation wavelength set at 295 nm. Light scattering was measured with both excitation and emission wavelength set at 480 nm. Circular dichroism spectra were recorded on a Jasco J-720 spectropolarimeter. Far UV (200–250 nm) and near UV CD (250–310 nm) measurements were performed in a 0.1 cm and 1.0 cm pathlength quartz cuvette, respectively. The results are expressed as mean residue ellipticity ([Θ]) assuming a mean residue weight of 110 per amino acid residue. Experiments with the fluorescent dye anilinonaphthalene-8-sulfonic acid (ANS) were performed by incubating the protein and ANS at a 1 : 5 molar ratio; after 10 min, fluorescence emission spectra were recorded at 400–600 nm (390 nm excitation wavelength). The maximum fluorescence emission wavelength and intensity of the hydrophobic probe ANS depend on the environmental polarity, that is, on the hydrophobicity of the accessible surface of the protein .
Fluorescence quenching was carried out by adding increasing amounts of acrylamide (0–200 m m) to a 35 µg·mL−1 protein solution. Emission spectra (300–400 nm) were recorded 5 min after each acrylamide addition with the excitation wavelength set at 290 nm. The effective quenching constants were obtained from modified Stern–Vollmer plots by analyzing F0/ΔF versus 1/[acrylamide] (20 data points) . All the spectroscopic measurements were performed at 20 °C.
Effect of pH on the dodecameric structure of L. innocua ferritin
The stability of the dodecameric structure of L. innocua ferritin at acid pH values has been studied after incubation of the apoprotein in the pH range 2.0–1.0 at 20 °C for 24 h, a time that was established to be sufficient to reach equilibrium.
L. innocua apoferritin maintains its quaternary structure when incubated at pH values above 2.0 as indicated by the elution volume from a Superose 12 column that is coincident with that of the native protein at pH 7.0 (Ve = 8.7 mL). Upon incubation at pH 2.0 (10 m m HCl), less than 10% of the protein gives rise to a slow moving component with the elution volume of dimers (Ve = 10.7 mL) ( Fig. 1). The secondary structure of native apoferritin is preserved as shown by the far UV CD spectrum, which resembles that measured at pH 7.0 ( Fig. 2A). In contrast, the near UV CD spectrum points to alterations of the native-like tertiary structure ( Fig. 2C). Thus, the characteristic positive dichroic band centered at 255 nm due to phenylalanine residues becomes negative, the ellipticity of the negative peak at about 292 nm attributed to tryptophan residues  decreases by about 30%, and the remaining part of the spectrum displays a small generalized decrease in ellipticity. As a matter of fact, this last feature may reflect either tertiary structure perturbation and/or the small amount of dimers present as their CD spectrum in the aromatic region is essentially featureless (data not shown). Consistent with the near UV CD data, the fluorescence emission of apoferritin incubated at pH 2.0 is red shifted relative to the native protein, λmax= 344 nm and 335 nm, respectively, upon excitation at 295 nm ( Fig. 2B). The observed fluorescence shift suggests that tryptophan residues become more exposed to solvent.
Incubation of L. innocua apoferritin at pH 1.5 (31.6 m m HCl) and 1.0 (100 m m HCl) results in a marked weakening of the quaternary structure as shown by the size-exclusion chromatography data ( Fig. 1). Thus, after incubation at pH 1.5, the amount of protein eluting at a retention time corresponding to dimers is increased significantly with respect to pH 2.0. After incubation at pH 1.0, only subunits are observed and a peak with the elution volume of monomers (Ve = 11.7 mL) is apparent. The disassembly of L. innocua apoferritin incubated at pH 1.5 and 1.0 is paralleled by a significant loss in secondary structure as indicated by the far UV CD spectra included in Fig. 2A. The spectra at these two pH values are indistinguishable and are characterized by a significant decrease in ellipticity relative to pH 2.0, a blue shift of the zero intercept and a change in the ratio between the 208 and the 222 nm bands. In particular, the molar ellipticity ratio ([Θ222]/[Θ208]) shifts to 0.97 from 1.21, the value measured at pH 7.0 and 2.0. The near UV CD spectra at pH 1.5 and 1.0 are comparable and characterized by the absence of dichroic activity over the entire spectral range, pointing to the disappearance of all the contributions of the aromatic residues ( Fig. 2C). The loss in protein structure induced by the addition of protons between pH 2.0 and 1.0 is reflected also in the intrinsic fluorescence spectrum, which displays a further red shift of the emission maximum (λmax= 348–346) and a moderate increase in intensity ( Fig. 2B).
A set of experiments was performed on native ferritin containing about 100 iron atoms per dodecamer. The presence of an iron core does not influence the acid-induced dissociation of the protein. Thus, the elution profile from a Superose 12 column and the spectral properties of native ferritin after incubation for 24 h at pH 2.0 and pH 1.0 correspond to those of the iron-free protein (data not shown).
Effect of anions on the protein at pH 2.0
The effect of proton addition to apoferritin at pH 2.0 has been compared to that exerted by NaCl. The changes in the far UV CD spectrum induced by incubation for 24 h at pH 2.0 in the presence of 31.6 m m NaCl (pCl 1.5) or 100 m m (pCl 1.0) are the same as those induced by equimolar concentrations of protons ( Fig. 3A). On the other hand the fluorescence spectra at pH 2.0 pCl 1.0 and at pH 2.0 pCl 1.5 are characterized by an increased intensity relative to the spectra measured at pH 1.0 and 1.5, although the emission maximum is the same (λmax= 349 nm at pCl 1.5, λmax= 346 nm at pCl 1.0) ( Fig. 3B). At pH 2.0 pCl 1.0 and at pH 2.0 pCl 1.5 the near UV CD spectra show no dichroic activity due to aromatic side chains ( Fig. 3C).
Differences in the effects of proton and NaCl addition to the protein at pH 2.0 are apparent also from Figs 1,4, which depict the elution patterns from a Superose 12 size-exclusion chromatography column. NaCl is less effective than protons in promoting dissociation, the difference being especially marked if one compares pH 1.0 to pH 2.0 pCl 1.0. Thus, at pH 1.0 the protein is fully dissociated with the formation of dimers (Ve = 10.7 mL, 40%) and monomers (Ve = 11.7 mL, 60%), whereas at pH 2.0 pCl 1.0 about 10–15% of the protein is still dodecameric (Ve = 8.7 mL) and dissociation proceeds only to the dimer stage.
In order to establish whether the effect of NaCl is due to selective anion binding to positively charged groups on the protein, so as to effectively shield the electrostatic interactions, or to the effect of the anion on the water structure, NaBr and Na2SO4 were tested. At pH 2.0, the effect of NaBr (pBr 1.0 and pBr 1.5) on the protein is comparable to that exerted by NaCl as apparent from the Superose 12 elution profile and from the spectral properties (data not shown). In contrast, the addition of Na2SO4 in the concentration range 10–100 m m does not induce protein dissociation, as indicated by the elution volume from a Superose 12 column ( Fig. 4), which is identical to that of the protein at pH 7.0, and by the far UV CD spectra ( Fig. 3A). The maximum fluorescence emission wavelength of the protein at pH 2.0 containing 10–100 m m Na2SO4 is blue shifted relative to the protein without Na2SO4 (λmax 341 nm and 344 nm, respectively) ( Fig. 3B). Similar results were obtained when Na2SO4 was added to the protein at pH 1.5. In accordance with this finding the near UV CD spectrum of the protein at pH 2.0 pSO4 1.0 is intermediate between that at pH 7.0 and at pH 2.0 ( Fig. 3C).
Accessibility of hydrophobic residues
The accessibility of hydrophobic residues upon incubation of L. innocua ferritin at pH 2.0 in the presence and absence of 100 m m NaCl was analyzed by the fluorescent probe ANS. The fluorescence emission spectrum of ANS shows an increase in intensity and a blue shift from 510 to 472 nm in the presence of the protein at pH 2.0 relative to the spectrum in the presence of apoferritin at pH 7.0 ( Fig. 5). This observation suggests that incubation at pH 2.0 leads to an increased exposure of hydrophobic surface area. Addition of 100 m m NaCl to the protein at pH 2.0 further increases the fluorescence intensity of the dye without affecting the wavelength of maximum emission, pointing to a further enhancement of the exposed hydrophobic surface area.
The accessibility of the hydrophobic core and the dynamic properties of apoferritin incubated at pH 2.0, pH 2.0 pCl 1.0 and pH 1.0 were studied also by means of the uncharged quencher acrylamide and compared to those of the protein at pH 7.0. The effective quenching constants obtained from Stern–Vollmer plots  were 3.1, 4.5 and 6.1 m−1 for the protein at pH 7.0, pH 2.0 and at pH 2.0 pCl 1.0, respectively. Consistent with the ANS data, these results indicate that exposure to pH 2.0 and the addition of NaCl at this pH value result in exposure of tryptophanyl residues to collisional quenching. At pH 1.0 a further increase of the effective quenching constants to 16.8 m−1 is observed. However, the heterogeneity of the emitting system does not allow a quantitative treatment of the data in structural terms.
Reversibility of the dissociation process
The reversibility of L. innocua apoferritin disassembly was investigated by monitoring the changes in the elution profile of the acid-dissociated protein brought back to neutral pH by a 20-fold dilution into 20 m m phosphate buffer containing 150 m m NaCl. The reassociation process starts upon dilution, although the appearance of the dodecameric structure occurs after about 8 h. Over 90% of the protein elutes as dodecamer from a Superose 12 column after 24 h with less than 10% of aggregates. The far and near UV CD spectra of the refolded apoferritin resemble those of the native protein measured at pH 7.0.
The recovery of the iron incorporation capacity of the reassociated protein was tested by measuring the kinetics of iron uptake at pH 7.0 in 50 m m Mops/NaOH buffer after the addition of 300 iron atoms per dodecamer. All the added iron is incorporated inside the apoferritin shell as indicated by sedimentation velocity experiments. The time course of the reaction is similar to that of the native protein measured in parallel (data not shown).
The present study of L. innocua apoferritin at low pH reveals the extreme stability of the dodecameric assemblage and brings out distinct features of the molecule. Thus, at pH 2.0 significant alterations in tertiary structure take place without changes in quaternary and secondary structure. The first dissociation step, which is apparent at pH 1.5, leads to the formation of extremely stable dimers.
Before discussing the behaviour of the Listeria ferritin dodecamer in detail, a description of relevant features of the subunit interfaces is in order, such as the position of hydrogen bonds and of aromatic residues, in particular of the two tryptophan residues that have been used as spectroscopic probes. The two nonequivalent environments at the three-fold axes are stabilized mainly by hydrophilic interactions. The ‘ferritin-like’ interfaces ( Fig. 6A) contain a salt-bridge between Asp 140 and Arg 63 (at a distance of 2.9 Å). This salt-bridge is close to Trp 144, which in turn faces the internal cavity of the dodecamer. The ‘Dps-like’ interfaces ( Fig. 6B) contain two hydrogen bonds involving His 37 (peptide O) and Asn 38 (ND2), at a distance of 3.0 Å and Gly 36 (peptide N) and Ala 148 (peptide O), at a distance of 3.0 Å. Analysis of the dimer interface ( Fig. 6C) reveals the presence of a large number of hydrophobic residues, Leu 75 and Leu 79 belonging to the BC helix, and Trp 32 and Tyr 33 belonging to the A helix. Of interest, in view of the extreme acid stability of the dimer, is the presence of a hydrogen bond involving Glu 86 (OE2) and Leu 75 (peptide N).
At pH 2.0 the molecular mass and far UV CD spectrum are essentially unaltered relative to pH 7.0, whereas the near UV CD spectrum points to changes in the environment of aromatic amino acids, in particular tryptophan and phenylalanine residues. Consistent with the near UV CD data, the red shift in the intrinsinsic fluorescence maximum, the binding of ANS and the increase in effective acrylamide quenching constant all indicate that the dodecameric structure is loosened and that the aromatic residues become more exposed to solvent. It may be envisaged that the weakening of the salt-bridge interactions at the ‘ferritin-like’ interface leads to an exposure of Trp 144 to the solvent.
The similarity of the far UV CD spectral changes upon addition of NaCl, or of an equimolar amount of HCl to the protein at pH 2.0, suggests that the anion is responsible for the secondary structure alterations and that it does not affect the interactions that maintain the dimeric assembly at pH 2.0, pCl 1.0. The partial dissociation into monomers observed at pH 1.0 but not at pCl 1.0 may support the hypothesis of unusual pKa values for carboxylic residues involved in ionic interactions. A likely candidate is the carboxylate of Glu 86 ( Fig. 6C).
The analysis of the effect exerted by different anions on the protein at pH 2.0 may help in finding a possible explanation for the stability of the L. innocua ferritin assembly at low pH values. The dodecamer dissociation into dimers, occurring at pH 2.0 in the presence of increasing NaCl and NaBr concentrations, indicates that the interactions stabilizing the ‘ferritin-like’ and ‘Dps-like’ three fold interfaces are mainly hydrophilic, in agreement with the X-ray data ( Fig. 6A,B). The similarity of the effect exerted by the two anions suggests that depolymerization may be due to shielding of the electrostatic interactions that are still effective at pH 2.0. The persistence of the dodecameric assembly upon incubation at pH values ≥ 1.5 in the presence of increasing Na2SO4 rules out any possible selective anion binding to the protein and points to a prevalent effect of sulfate on the water structure leading to the stabilization of hydrophobic interactions.
The dodecamer dissociation is always accompanied by marked changes in the far UV CD spectrum, as indicated by the blue shift of the zero intercept for dichroic activity and by the increase of the molar ellipticity ratio ([Θ222]/[Θ208]). This finding suggests that formation of α-helices is not prevented, although the increased charge repulsion due to protonation of carboxyl residues or to charge shielding by chloride ions destabilizes interhelical interactions. Notably, the secondary structure changes are always similar in the pH range 2.0–1.0 and at pCl 1.5–1.0 and independent of the association state of the protein. The pH decrease from 2.0 to 1.5 promotes the dissociation of the dodecamer into dimers and at the same time determines the exposure of the aromatic residues to solvent and a partial loosening of the secondary structure. The change in the fluorescence spectrum consists of a small red-shift of the maximum emission relative to pH 2.0, which has to be contrasted with the marked changes resulting from incubation at pH 2.0. The featureless near UV CD spectrum indicates the loss of tertiary contacts, suggesting that the aromatic residues at the trimeric interfaces become completely exposed to solvent and that Trp32 and Tyr33 at the dimeric interfaces are fluctuating.
The stability of Listeria ferritin toward acid-induced dissociation is shared by horse spleen apoferritin. The latter protein loses its tetracosameric assemblage after incubation at pH values below 2.8 , about 1.5 pH units higher than L. innocua ferritin. In horse spleen apoferritin dissociation gives rise to dimers that are stabilized mainly by hydrophobic interactions . Like the ferritin 24-mer, the L. innocua apoprotein dissociates at acid pH into extremely stable dimeric species. The antiparallel BC helices ( Fig. 6C), which are absent in classic ferritins , are likely to contribute to the unusual stability of the L. innocua ferritin dimer interface due to the presence of hydrophobic interactions between Leu75 and Leu79.
In conclusion, the basic structural unit of the dodecameric L. innocua ferritin is dimeric and is stabilized by hydrophobic interactions. At pH 2.0 the dodecameric ferritin is in a compact state with an altered tertiary structure and a more accessible hydrophobic core. The loss of the dodecameric assembly is always accompanied by the loosening of the tertiary contacts and by a significant loss of the secondary structure elements. These marked tertiary changes, however, do not prevent the formation of dimers. The presence of a quaternary assembly in the absence of a rigid tertiary structure suggests that the formation of intersubunit contacts may be preliminary to the formation of a compact hydrophobic core.
This work was supported in part by a grant from the Ministero per l’Università e la Ricerca Scientifica e Tecnologica, Progetto ‘Biologia strutturale e dinamica di proteine redox’ to E.C. The authors are grateful to Dr Laura Giangiacomo for performing the analytical ultracentrifuge experiments.