PEDS Advance Access originally published online on October 24, 2007
Protein Engineering Design and Selection 2007 20(10):481-490; doi:10.1093/protein/gzm034
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Folding of an antibody variable domain in two functional conformations in vitro: calorimetric and spectroscopic study of the anti-ferritin antibody VL domain
1 Institute of Bio-Organic Chemistry, National Academy of Sciences of Belarus, Minsk 220141, Belarus 2 Research and Production Center for Hematology and Transfusiology, Ministry of Health of the Republic of Belarus, Minsk 220053, Belarus
4 To whom correspondence should be addressed. E-mail: tsybovs{at}musc.edu
 |
Abstract |
|---|
 Top Abstract IntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Understanding refolding pathways of recombinant antibody fragments is essential for efficient production of these proteins of high biomedical significance. The recombinant VL domain of mouse anti-human ferritin antibody F11 formed two distinct functional conformations obtained by refolding from bacterial inclusion bodies using two different procedures. Involvement of a dialysis step at pH 2–3 resulted in the VL-1 conformation with fluorescence of the highly conserved Trp-35 residue quenched by the spatially proximal disulfide bond. This conformation was identical to the ‘native’ VL domain folded in host cells and purified from the cytoplasm. In the absence of the acidic dialysis step, the VL domain adopted a previously unreported conformation, VL-2, that demonstrated prominent fluorescence due to a local structural disorder around Trp-35. Furthermore, VL-2 showed changes in secondary structure and significantly lower stability as determined by differential scanning calorimetry and denaturant-induced unfolding. While more flexible VL-2 binds human ferritin both in solution and after surface adsorption of the antibody domain, the VL-1 conformer needs an adsorption-induced conformational change to allow the access of ferritin to the antigen-binding site. Noteworthy, the two macroscopic conformations constitute kinetically trapped dimers and do not interconvert at elevated temperatures (3 weeks at 37°C or 15 min at 60°C), which indicates a high energetic barrier between them. As a major finding, this paper provides the first description for two stable and functional conformations of an antibody domain.
Keywords: conformational stability/conformers of folding/protein refolding/recombinant antibody/VL domain
|
Introduction |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
More than a decade ago, a limited number of proteins were found to exist in more than one stable conformation under physiological conditions, which challenged the thermodynamic hypothesis of protein folding proposed by Christian Anfinsen (Anfinsen, 1973
). The best known examples are serine proteases (
-lytic protease and subtilisin) (Zhu et al., 1989; Baker et al., 1992), the serpin family of protease inhibitors (Franke et al., 1990; Stein and Chothia, 1991; Devlin and Bottomley, 2005) and luciferase (Sinclair et al., 1994). Alternative conformations of a protein may significantly differ in local and non-local interactions, thermodynamic stability and activity.
A few multi-domain proteins from the immunoglobulin family, including lymphocyte cell adhesion protein CD2 (Murray et al., 1995) and 
1-type human light chain Loc (Huang et al., 1996), adopted multiple compact structures as demonstrated by X-ray crystallography. In the latter case, three distinct spatial structures were obtained at different ionic strengths and pHs of the crystallization solution. CD2 is capable of folding into two equilibrated but unevenly populated conformations, with the less populated state representing a metastable dimer formed through domain swapping (Murray et al., 1995). Conformational diversity was also demonstrated for three anti-phenyloxazolone antibodies by bi- and triphasic hapten-binding kinetics that indicated antibody isomerization (Foote and Milstein, 1994). Recently, James and co-authors reported different conformations of the antibody SPE7; these conformations were capable of binding unrelated antigens, as judged from crystal structures and pre-steady-state kinetics (James et al., 2003). In the two studies discussed earlier (Foote and Milstein, 1994; James et al., 2003), the antibody conformers existed in equilibrium and differed due to relatively small rearrangements of the molecule in the antigen binding site.
Recombinant antibody fragments have found manifold applications as universal tools for selective recognition of a large variety of target molecules [see (Hudson, 1998) for a review]. The scFv fragments, which comprise variable domains of light (VL) and heavy (VH) immunoglobulin chains linked through a flexible peptide, represent the smallest entities that retain the antigen-binding site of the full-length antibody (Bird et al., 1988; Huston et al., 1988). Furthermore, several single-domain antibodies comprising either isolated VH or VL domains were shown to preserve biological activity and stability, which makes single-domain antibodies a promising protein design to construct structurally minimal recognition modules of immunotoxins, abzymes and other antibody-derived proteins for therapeutic, diagnostic, and industrial applications (Gavish et al., 1977; Ward et al., 1989; Brinkmann et al., 1993; Davies and Riechmann, 1996; Martsev et al., 2000). Hence, knowledge of the inter-relationships among folding, stability and binding activity of isolated variable domains of immunoglobulins is important for engineering new proteins with clinical and industrial applications. Recombinant antibody fragments are frequently expressed as bacterial inclusion bodies, raising a demand for simple and effective refolding procedures that would result in a homogeneous, thermodynamically stable and functional macroscopic state (Tsumoto et al., 1998, 2003; Umetsu et al., 2003). Studies of refolding pathways are therefore essential for successful production of ‘robust’ recombinant antibodies.
In this paper, we provide the first description of the two stable and functional conformations of an antibody domain. We employed the recombinant VL domain derived from the anti-human ferritin monoclonal antibody F11 (Martsev et al., 2000; Nymalm et al., 2002). Refolding of the denaturant-solubilized VL domain from bacterial inclusion bodies using two distinct protocols resulted in two functional conformations, VL-1 and VL-2, which possess long-term stability and do not equilibrate under physiological conditions. Using differential scanning calorimetry, CD and fluorescence spectroscopy, and immunoassays, we demonstrated that VL-1 and VL-2 differ in the amount of structure, thermodynamic and pH-stability, and arrangement of the antigen-binding site. A high degree of structural similarity among variable domains of light chain of immunoglobulins indicates that refolding of other recombinant VL domains may result in similar conformational diversity.
|
Material and methods |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Purification of the VL domain
Expression of the VL domain was performed as we described earlier (Martsev et al., 2000). Inclusion bodies, which contained the major (
95%) fraction of the recombinant protein, were separated from the soluble fraction of the cell lysate by centrifugation for 30 min at 35 000 g.
The pellet containing inclusion bodies was solubilized with 6 M guanidine hydrochloride (Gdn–HCl) for 2 h at room temperature. After centrifugation for 20 min at 35000 g, the supernatant was dialyzed against 7 M urea in 25 mM Tris–HCl, pH 7.0 and applied to a Ni2+-NTA Sepharose column (Novagen, USA) equilibrated with the same buffer. After washing the column with 10 volumes of the same buffer, 0.1 M of imidazole was used to elute the VL domain. The purity of the VL domain was >95% as determined by SDS–PAGE. Typically, the overall yield was 50 mg per 1 l of the cell culture.
To purify the minor cytoplasmic fraction of the VL domain, the soluble fraction of the cell lysate was dialyzed against buffer A (0.1 M Tris–HCl, pH 8.1, 200 mM NaCl, 10 mM imidazole). The sample was applied on the Ni2+-NTA-Sepharose column equilibrated with the same buffer. Bound protein was eluted by increasing concentrations of imidazole. The fraction containing the VL domain was dialyzed against 50 mM sodium phosphate, pH 7.4.
Refolding of the two conformers
The VL domain purified from inclusion bodies was refolded by stepwise dialysis procedure according to the two refolding protocols (Fig. 1). To obtain the VL-1 conformer, the refolding protocol-1 was applied in which the purified VL domain was sequentially dialyzed against 5, 3 and 1 M urea in 0.1 M sodium phosphate, pH 2–3, and then the same buffer without the denaturant. At this stage, some variations in the refolding buffer composition, pH and ionic strength (Fig. 1, upper line) did not change the final conformation. The protein was then dialyzed against 0.1 M Tris, pH 8.1, and, finally, against 0.1 M sodium phosphate, pH 7.4. To obtain the VL-2 conformer, we applied the refolding protocol-2 (Fig. 1, bottom line). For this, the VL domain after the Ni2+-NTA chromatography was dialyzed against 5, 3 and 1 M urea in 0.1 M Tris, pH 8.1, or 0.1 M sodium phosphate, pH 7 and, finally, against 0.1 M sodium phosphate, pH 7.4, without the denaturant.
|
Molecular mass determination
Gel filtration chromatography was carried out on a Sephacryl S200 column (1.0 x 106 cm) equilibrated with 50 mM sodium phosphate, pH 7.4. The following molecular mass markers were used: IgG (150 kDa); BSA (66 kDa); ovalbumin (45 kDa); lysozyme (14 kDa). Samples of 0.5 ml containing 0.7–2.0 mg of the VL domain were loaded on the column. All experiments were performed at room temperature with a flow rate of 10 ml h–1. Mean values of three independent experiments are reported.
Interconversions of VL-1 and VL-2
The ability of VL-1 and VL-2 to convert into each other was studied in two lines of experiments. First, both conformers were incubated in 0.1 M sodium phosphate, pH 7.4, at (i) 37°C for 3 weeks or (ii) 60°C for 15 min in a water bath. Second, urea or Gdn–HCl in 0.1 M sodium phosphate, pH 7.4, was added to the samples containing VL-1 and VL-2 to final concentrations of 7 M and 6 M, respectively. After incubation for 12 h, the sample of VL-1 was refolded according to the refolding protocol of VL-2, whereas the sample of VL-2 was refolded according to the refolding protocol of VL-1. In both experiments, intrinsic tryptophan fluorescence was used to monitor the conversions.
CD spectra were recorded on a J-20 spectropolarimeter (‘Jasco’, Japan) at a protein concentration 0.5–0.7 mg/ml in 50 mM sodium phosphate, pH 7.4, using a thermostated cuvette with a 1 mm (far-UV CD) or 10 mm (near-UV CD) quartz cell. Spectra were scanned from 250 to 200 nm (far-UV CD) or 300 to 250 nm (near-UV CD) at a scan speed of 5 nm/min. The averaged spectra of three scans were corrected for the buffer blank.
All spectra were recorded in triplicates. Intrinsic fluorescence spectra of the proteins were recorded in a 1 cm pathlength quartz cuvette in an SFL-1211 fluorometer (‘Solar’, Minsk, Belarus) at a protein concentration of 0.05 mg/ml in 0.2 M KCl–HCl (pH 1.0–1.8), 0.1 M citrate-phosphate (pH 2.0–6.0), 0.1 M sodium phosphate (pH 6.2–7.8) and 0.1 M Tris–HCl (pH 8.0–10.0). Fluorescence was excited at 280 or 295 nm. Spectra were corrected for the buffer blank.
Fluorescence of protein-bound 8-anilino-1-naphthalene sulfonic acid (ANS) was recorded from 400 to 600 nm after excitation at 360 nm. The protein concentration was 0.05 mg/ml in the above buffers, and the molar ANS-to-protein ratio was 10:1. Spectra were corrected for the ANS fluorescence without a protein.
In equilibrium unfolding experiments, samples of the VL domain (0.025–0.1 mg/ml) in 0.1 M sodium phosphate, pH 7.4, containing varying concentrations of urea, were incubated at room temperature for 24 h. Intrinsic fluorescence excited at 280 or 295 nm was measured at 340 nm using a 1 cm pathlength quartz cuvette. Spectra were corrected for the solutions containing no protein. The baselines in the pre- and post-transition regions were assumed to depend linearly on urea concentration. The data were fitted to a two-state N
U model assuming a linear relationship of the Gibbs free energy change of unfolding, 
G, and urea concentration, [D]:
|
G on urea concentration, R is the universal gas constant and T is the absolute temperature. Differential scanning calorimetry
Measurements were performed with a DASM-4 scanning calorimeter (‘Biopribor’, Puschino, Russia), equipped with a computer interface, in a temperature range of 10–80°C at a scan rate of 60 K·h–1. The reference cell was filled with the buffer used to dialyze the protein sample. 0.05–0.1 M sodium phosphate, pH 7.4, was used for the measurements. The heat capacity curves were corrected for the instrumental baseline that was determined with both cells filled with buffer. The protein concentrations varied between 0.5 and 1.8 mg/ml. The averaged heat capacity profiles of five temperature scans were analyzed and deconvoluted using the algorithm of Privalov and Potekhin (1986) and TERMCALC software supplied by the DASM-4 manufacturer. Variations of individual measurements for the midpoint transition temperature, Tm, were within 0.4°C.
Determination of antigen-binding activity
Antigen binding affinity was determined in an enzyme-linked immunosorbent assay (ELISA) using the VL domain directly bound to the surface of polystyrene plates. The assays were performed at room temperature in triplicates, and the mean values are reported.
Prior to performing binding assays, we compared the adsorption propensity of VL-1 and VL-2 onto the wells of a polystyrene plate. Increasing amounts of the VL domain (0–1 µg/ml) in 50 mM sodium borate, pH 8.5, were immobilized onto the wells of a polystyrene plate. After 12–16 h of incubation, the wells were washed twice with distilled water. Remaining binding sites on the polystyrene surface were blocked by incubation for 40 min with PBS–BSA, followed by washing. Coated wells were incubated with 1.6 µg of rabbit polyclonal antibodies, raised against the VL domain of the F11 antibody, in 0.2 ml of PBS–BSA for 1.5 h. After two washings, 250 ng of anti-rabbit antibody conjugated with horseradish peroxidase were added in 0.2 ml of PBS–BSA and incubated for 1.5 h. 0.2 ml of 0.02 M o-phenylenediamine and 0.02 M H2O2 in 0.1 M sodium citrate, pH 5, were then added and the solution was incubated with shaking for 5 min. The reaction was stopped by adding 0.05 ml of 10% H2SO4, and the absorbance was determined at 492 nm.
In the direct ELISA, conformers of the VL domain were immobilized by adsorption onto the wells of a polystyrene plate and allowed to react with increasing concentrations of human spleen ferritin. Stock solutions of the VL domain (0.7–0.8 mg/ml) in 50 mM sodium phosphate, pH 7.4, were diluted to 10 µg/ml with sodium borate, pH 8.5, and adsorbed onto the wells of polystyrene plate overnight. The wells were washed with the above buffer, and the remaining binding sites on the polystyrene surface were blocked by incubation for 40 min with PBS–BSA, followed by washing. Coated wells were incubated with increasing amounts of ferritin in 0.2 ml of PBS–BSA. After 2 h and two washings, 200 ng of mouse anti-ferritin monoclonal antibody G10 conjugated with horseradish peroxidase were added in 0.2 ml of PBS–BSA and incubated for 1.5 h. Bound peroxidase activity was measured as above.
Comparison of binding activities of soluble and polystyrene-adsorbed VL-1 and VL-2
To demonstrate the binding capacity of soluble and bound VL domain conformers, an inhibitory assay was constructed. In this assay, the conformers of the VL domain were immobilized by adsorption onto the wells of a polystyrene plate and allowed to compete with increasing concentrations of the soluble VL domain for human spleen ferritin. The VL domain was diluted to 10 µg/ml with sodium borate, pH 8.5, and adsorbed onto the wells of a polystyrene plate overnight. The wells were washed with the above buffer, and remaining binding sites were blocked by incubation for 40 min with PBS–BSA followed by washing. Coated wells were incubated with increasing amounts of one of the VL domain conformers and 5 µg of human spleen ferritin in 0.2 ml of PBS–BSA. After 2 h and two washings, 200 ng of mouse anti-ferritin monoclonal antibody G10 conjugated with horseradish peroxidase were placed into each well. After another 1.5 h and additional washings, bound peroxidase activity was measured as above. The assays were performed at room temperature in triplicates, and the mean values are reported.
Methods for preparation of human spleen ferritin and anti-ferritin monoclonals were reported previously (Martsev et al., 1995; Kravchuk et al., 1998). Native electrophoresis was carried out in a 10% polyacrylamide running gel without a stacking gel. The following buffers were used: 0.35% acetic acid-NaOH, pH 5, for the running gel; 0.035 M ß-alanine-acetic acid, pH 4.5, as the upper electrode buffer; and 0.1 M sodium acetate, pH 4.5, as the lower electrode buffer. The concentration of the VL domain was determined from the UV absorbance at 278 nm using extinction coefficient of 1.20 mg–1 ml cm–1. Disulfide groups and free thiols were determined in triplicates by the methods of Thannhauser et al. (1984) and Ellman (1959), respectively, in a protein solution containing 3 M guanidine thiocyanate.
|
Results |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Refolding from inclusion bodies
Refolding of the VL domain from 7 M urea by stepwise dialysis according to the two protocols (Fig. 1) resulted in two conformations denoted as VL-1 and VL-2. Each of the two protocols tolerates a small variation in pH, ionic strength and the buffer composition at different stages of refolding. The key difference between the protocols is exposure to low pH (pH 2–3) in refolding protocol-1, which was critical to obtain the conformation denoted as VL-1. When exposure to the low pH was excluded from the procedure to give the refolding protocol-2, the stepwise dialysis at near-neutral pH yielded the conformer VL-2. Refolding at slightly acidic pHs (pH 4–5) led to a mixture of VL-1 and VL-2, as judged by intrinsic fluorescence spectra (Supplementary data are available at PEDS online, Figure 1). From these data, we cannot rule out the possibility that our VL domain may refold into more than two conformations. However, we failed to isolate any of them, except for VL-1 and VL-2, using various buffer systems, ionic strengths and pH values during refolding.
VL-1 and VL-2 do not comprise wrong aggregates, non-native disulfides or sulfur derivatives
Refolding from inclusion bodies is a complex process that may be accompanied by multiple off-pathway folding intermediates that constitute misfolded conformations (Clark, 1998; Lilie et al., 1998; Tsumoto et al., 2003). To rule out the contribution of miss-folded species, e.g., the aggregated ones and those with non-native disulfide bridges, to our VL-1 and VL-2 samples, we applied gel permeation chromatography, native electrophoresis, and quantitative analysis of disulfides and free thiols.
The gel filtration chromatography (Fig. 2A) showed that VL-1 and VL-2 are homodimers, which was generally observed for isolated VL domains (Maeda et al., 1976; Gavish et al., 1977; Azuma et al., 1978; Ionescu-Zanetti et al., 1999). Each protein was eluted as a single peak with an apparent molecular mass of 24.8 ± 0.35 kDa (VL-1) or 26.5 ± 0.42 kDa (VL-2). We failed to isolate any oligomers or highly aggregated species from the samples. Likewise, VL-1 and VL-2 migrated as a single band in the native electrophoresis (Fig. 2B).
|
Determination of the number of disulfide groups under denaturing conditions resulted in 0.97 ± 0.03 and 0.96 ± 0.03 disulfides per protein monomer for VL-1 and VL-2, respectively. This is fully consistent with the two Cys residues that form the single disulfide bridge in the VL domain. The maximal content of free thiols determined by Ellman's method (Ellman, 1959
) did not exceed the value of 0.05 per protein monomer for both VL-1 and VL-2. These measurements strongly suggest that our VL-1 and VL-2 did not comprise appreciable amounts of free thiols and their stable oxygen derivatives. The purification and refolding procedure seems therefore efficient enough for correctly forming the native disulfide bond present in the two alternative conformations of the VL domain. Together, the thiol/disulfide determination and gel permeation chromatography strongly suggest that the samples of VL-1 and VL-2 did not contain considerable amounts of either aggregated species or those with non-native disulfides, nor did they comprise oxydized derivatives of Cys residues. VL-1 and VL-2 do not equilibrate
From intrinsic fluorescence spectra (see below), we observed no interconversion of the two alternative conformations at room temperature or at 4°C (data not shown). Likewise, incubation in 0.1 M sodium phosphate, pH 7.4, at 37°C for 3 weeks or, alternatively, at 60°C for 15 min, revealed no transition of VL-1 into VL-2 or VL-2 into VL-1 (not shown). Furthermore, we established that neither VL-1 
VL-2 nor VL-2 VL-1 conversion can be induced by adding 7 M urea to VL-1 or VL-2 with subsequent refolding according to the alternative protocol. To successfully convert VL-1 to VL-2 or VL-2 to VL-1 required strongly denaturing conditions (6 M Gdn–HCl) to provide complete unfolding (see Materials and Methods).
Secondary structure and environment of aromatic residues
The far-UV CD spectra of VL-1 and VL-2 at pH 7.4 suggest that the secondary structure is predominantly ß-sheet (Fig. 3A), as expected for proteins of the immunoglobulin family. However, the spectrum of VL-1 significantly deviates from that of a typical ß-sheet protein due to a prominent contribution from aromatic side chains. Experimental and theoretical studies have demonstrated that aromatic residues clamped in rigid clusters are characterized by a positive CD band in the region 180–220 nm (Bolotina and Lugauskas, 1985; Bolotina, 1987; Manning and Woody, 1989; Chakrabartty et al., 1993) and can therefore strongly influence far-UV CD spectra of proteins (Bychkova et al., 1992; Uversky and Ptitsyn, 1996; Abramov et al., 2001). The presence of such a band in the difference (VL-1–VL-2) spectrum (Fig. 3A) indicates that the two conformers of the VL domain are different with respect to the environment of aromatic residues. The cluster of aromatic residues that is responsible for the distorted far-UV CD spectrum of VL-1 may be less rigid or absent in VL-2.
|
The near-UV CD spectra of VL-1 and VL-2 are distinct from that of the denatured VL domain, indicating asymmetric environment of aromatic residues, i.e. the presence of a tertiary structure (Fig. 3B). However, the spectrum of VL-1 is clearly different from the spectrum of VL-2, which is another indication of alterations in the arrangement of aromatic residues.
The intrinsic fluorescence spectra of both VL-1 and VL-2 suggest compact conformations with aromatic fluorophores protected from quenching by solvent, as judged from the location of the emission maxima (Fig. 4). Unfolding of the VL domain with 6 M Gdn–HCl leads to exposure of the single and highly conserved tryptophan residue (Trp-35) to the polar environment of the solvent, thus inducing a red shift in the maximum of the fluorescence spectrum (Fig. 4A and B). A compact structure of the conformers was confirmed by the lack of fluorescence emission of the bound hydrophobic dye ANS (Fig. 5), the generally recognized probe for partially folded protein conformations. Furthermore, the intrinsic fluorescence spectra of VL-1 are distinct from those observed for VL-2. On excitation at 280 nm, the spectrum of VL-1 exhibited a single emission maximum at about 305 nm, which corresponds to the fluorescence of Tyr residues in proteins devoid of Trp fluorophores (Fig. 4A). Excitation at 295 nm resulted in only background level of emission of Trp-35 (Fig. 4B). This indicates that the fluorescence of Trp-35 is strongly quenched by the disulfide bond which is the commonly observed internal quencher of tryptophan fluorescence in isolated VL domains (Ohage and Steipe, 1999; Martsev et al., 2002; Ewert et al., 2003). In contrast to the intrinsic fluorescence spectra of VL-1, which are similar to those obtained for many other VL domains, VL-2 demonstrated prominent tryptophan emission with the maximum at around 338 nm (Fig. 4B). This, in combination with the correctly linked disulfide bond, is a strong indicator of an altered environment of Trp-35 in the VL-2 conformation. Furthermore, intrinsic fluorescence studies suggest that VL-1 did not contain considerable amounts of VL-2, as the emission of the single tryptophan of VL-1 is almost completely quenched.
|
|
pH-stability
The two VL conformers demonstrated distinct pH-stability. On lowering pH from 8 to 2, the environment of Trp-35 within VL-1 remained unaltered, as judged by the Trp fluorescence that remained completely quenched over this pH range (Fig. 5A). Likewise, binding of the hydrophobic dye ANS was on the background level between pH 8 and 2 (Fig. 5B). A sharp increase in the Trp-35 fluorescence of VL-1 was observed below pH 2 and accompanied by enhanced binding of ANS (Fig. 5B), thus indicating partial unfolding of the tertiary structure with concomitantly removing the Trp residue from the vicinity of the internal quencher. In contrast to VL-1, the intrinsic fluorescence of VL-2 did not significantly change between pH 8 and pH 4, and demonstrated a biphasic change between pH 4 and pH 2 with only about 1.5-fold increase of fluorescence intensity at pH 1.5 compared to pH 8 (Fig. 5A). In ANS binding studies (Fig. 5B), the VL-2 conformer demonstrated a broad transition with low cooperativity over the pH range 2–8, thus indicating gradually increased binding of ANS due to the exposure of hydrophobic clusters to the solvent. The complicated pH-induced unfolding curves obtained for VL-2 indicate that unfolding might occur through formation of intermediate partially structured states. This was confirmed by the analysis of the intrinsic fluorescence spectra using the ‘phase diagram’ approach (Kuznetsova et al., 2004), which revealed two unfolding intermediates of VL-2 formed at pH 2.3–2.5 and pH 4–4.5 versus no intermediates in the case of VL-1 (Supplementary data are available at PEDS online, Figure 3). Together, the results of pH-induced unfolding studies indicated lower pH-stability of tertiary structure for the VL-2 conformer.
Our calorimetric measurements revealed that thermodynamic stability of both VL-1 and VL-2 is consistent with stability of fully compact proteins. Thermal unfolding of VL-1 and VL-2 resulted in a sharp heat absorption peak (Fig. 6A), as previously observed for many globular proteins. Repetitive cycles of heating and cooling revealed a variable extent of reversibility depending on the end temperature of the scan and the buffer pH and composition (data not shown). The maximal reversibility at pH 7.4 was 80% for VL-1 and 60% for VL-2. This makes the calorimetric profiles of the two conformers amenable to equilibrium thermodynamic analysis (Privalov and Potekhin, 1986). The VL-1 conformer demonstrated higher stability in terms of the midpoint transition temperature (62 versus 59°C observed for VL-2) and transition enthalpy, h (22.2 ± 1.1 J·g–1 versus 14.7 ± 0.7 J·g–1). The heat absorption peak obtained for each of the two conformations was well approximated by a single two-state transition (Fig. 6A, inset). The approximation data, together with clearly seen differences in Tm and h, strongly suggest that VL-1 and VL-2 each constitutes a major macroscopic state of the VL domain, with contamination by the alternative conformer being less than 10%.
|
To compare thermodynamic stability of VL-1 and VL-2 at room temperature in terms of the Gibbs free energy change on unfolding,
G, we analyzed urea-induced unfolding curves monitored by intrinsic fluorescence (Fig. 6B). Both VL-1 and VL-2 showed reversible unfolding, and the transition curves did not change on changing the protein concentrations (data not shown). The experimental data were well fit by the equation based on the two-state approximation (see Materials and Methods). Consistent with the calorimetric data, the VL-1 conformer demonstrated higher thermodynamic stability in terms of G (30.7 ± 1.6 versus 24.5 ± 1.9 kJ·mole–1 for VL-2). The m values calculated for the two conformers (7.6 ± 0.4 kJ·mole–1·M–1 for VL-1 and 5.1 ± 0.5 kJ·mole–1·M–1 for VL-2) were typical of two-state transitions induced by urea (Myers et al., 1995). Phase diagram method (Kuznetsova et al., 2004) showed no intermediates in urea-induced unfolding of VL-1 and VL-2 (data not shown).
The ferritin-binding capacities of the two conformations were first compared by the direct binding assay with the VL domain immobilized by adsorption onto the wells of a polystyrene plate (Fig. 7). To provide reliable estimates, we showed that VL-1 and VL-2 demonstrate the same level of adsorption on the polystyrene surface, as determined with rabbit polyclonal antibodies raised against the VL domain (Fig. 7A). Despite differences in folding and stability of VL-1 and VL-2, the antigen binding curves of the adsorbed VL domain (Fig. 7B) yielded similar affinity constants for the VL-1 and VL-2 conformers (2.8 ± 0.4·107 M–1 and 1.2 ± 0.2·107 M–1, respectively). Given that presumed differences in the spatial arrangement of the antigen-binding site in VL-1 and VL-2 might be lost or partially hidden due to adsorption-induced conformational changes shown for many proteins (Friguet et al., 1984; Jemmerson, 1987), we additionally considered functionality of the two VL conformers in the inhibitory immunoassay (Fig. 8) that allows antigen-binding measurements in solution. In the soluble immunoassay system, less stable VL-2 was capable of efficiently binding to the antigen irrespective of whether VL-1 (Fig. 8A) or VL-2 (Fig. 8B) was used to immobilize ferritin. In contrast, the more stable VL-1 conformer demonstrated significantly lower ferritin-binding capacity in the two formats of the soluble immunoassay (Fig. 8A and B). These results strongly suggest that differences in spatial arrangement of the binding sites of VL-1 and VL-2 do exist.
|
|
Comparison with the cytoplasmic VL domain
A minor fraction of the VL domain containing less than 5% of the total recombinant protein was expressed into the cytoplasm of Escherichia coli cells (data not shown). Using Ni2+-NTA chromatography without denaturants, we purified this cytoplasmic VL domain that was folded into soluble and stable conformation in vivo, and compared it with the VL-1 and VL-2 forms that were refolded in vitro from inclusion bodies. In the gel permeation chromatography, intrinsic fluorescence and calorimetric studies, the properties of the cytoplasmic VL domain were virtually identical to those obtained for the VL-1 conformer (intrinsic fluorescence spectra are shown in Figure 2, Supplementary data are available at PEDS online). These data suggest that the conformation and stability of our VL-1 obtained by in vitro refolding protocol-1 was identical to that produced in the minimal amount in vivo by expression in E. coli.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
The VL domain of antibody F11 refolds in two distinct macroscopic states
When refolded from insoluble inclusion bodies, the VL domain of the mouse antibody F11 adopted two alternative conformations, VL-1 or VL-2, depending on the refolding protocol. Importantly, each of the two conformations constituted the major macroscopic state of the VL domain attainable under the respective refolding protocol. Furthermore, the prominent structural and functional differences between the two alternative forms did not result from contributions from unfolded and aggregated species as well as those with non-native disulphide bridges, free thiols or their oxygen derivatives.
The most evident discriminatory feature of the VL-1 conformer is the environment of Trp-35, which is distinct from that of VL-2, as demonstrated by fluorescence spectroscopy. As commonly observed for VL domains (Ewert et al., 2003), in VL-1 the emission of Trp-35 is quenched due to close proximity to an intramolecular quencher which is generally thought to be the disulfide bond (Ohage and Steipe, 1999; Ewert et al., 2003). In contrast, the prominent tryptophan fluorescence of VL-2 with the maximum at 338 nm seems to be a unique feature that indicates an altered local conformation in the environment of the single Trp residue.
The key difference between the two refolding protocols is the exposure to the low pH (pH 2–3) in refolding protocol-1. Additional studies are needed to reveal the exact mechanisms that lead to formation of either of the two conformations; this work is underway. One possible explanation is compensation of the positive net charge of the VL domain (pI 9) by chloride anions on lowering pH with HCl down to pH 2. The effect of charge compensation by various anions was shown to significantly change the conformation of many proteins (Goto et al., 1990). This compensation of electrostatic repulsion forces in protocol-1 might allow formation of the rigid cluster of aromatic residues detected by CD in VL-1 (Fig. 3A), whereas uncompensated positive charges in refolding protocol-2 might interfere with hydrophobic interactions of the residues constituting the cluster and lead to formation of the alternative environment of Trp-35 in VL-2.
Environment of Trp-35 strongly contributes to folding and stability of the VL domain
Our fluorescence and calorimetric measurements are consistent with a hypothesis that the rigid cluster formed by a number of aromatic residues of the VL domain, including Trp-35, strongly stabilizes the structure of VL-1. This cluster, which we call a ‘conformational lock’, is absent or distorted in VL-2. Indeed, interactions forming the ‘conformational lock’ in VL-1 are remarkably stable between pH 8 and 2 and might therefore be attributed to predominantly hydrophobic forces. This suggestion is consistent with the crystal structure of the VL-1 conformer [Protein Data Bank (PDB) ID 1F6L, (Nymalm et al., 2002)]. In this structure, Trp-35 is surrounded by hydrophobic and aromatic amino acids and is within 5 Å from the disulfide bridge (Supplementary data are available at PEDS online, Figure 4). The cluster around Trp-35 becomes compact on in vitro refolding using protocol-1 as early as at pH 2 in the absence of urea (unpublished data) and survives at neutral pH to give the ‘mature’ VL-1 conformer. This type of structure around Trp-35 is not formed on removing urea by dialysis at near-neutral pHs under refolding protocol-2, which yields the VL-2 conformer.
Involvement of Trp-35 and its environment in stabilizing interactions is supported by at least two lines of evidence. First, the VL-2 conformer lacking the ‘conformational lock’ around Trp-35 demonstrated significantly lower pH-stability. Both the local environment of Trp-35 (Fig. 5A) and the global conformation (Fig. 5B) of VL-1 remained unaltered between pH 8 and 2. In contrast to VL-1, the VL-2 conformer demonstrated a broad and apparently non-cooperative transition over this pH range, as demonstrated by ANS binding studies. This transition indicates a gradual loss of structure. Notably, intrinsic fluorescence of the Trp residue in VL-2 does not change significantly in the pH 2–8 range, indicating that the local environment of Trp-35 in VL-2 is relatively autonomous from the rest of the molecule and may not be tightly associated with global changes of the tertiary structure. Thus, the highly ordered environment of Trp-35 in VL-1 provides higher rigidity and cooperativity, while the lack of this structure in VL-2 provides higher flexibility and lower cooperativity. Second, although the Gibbs free energies of unfolding of the two conformers fall within the range observed for many globular proteins (Neet and Timm, 1994), VL-1 demonstrates significantly higher thermodynamic stability. The calorimetric enthalpy of thermal unfolding of VL-1 is nearly 50% higher (22.2 J·g–1 versus 14.7 J·g–1) while the Gibbs free energy change, G, is 25% higher (30.7 ± 1.6 kJ·mole–1 versus 24.5 ± 1.9 kJ·mole–1) compared to the values obtained for VL-2. It should be noted that, in the case of VL-2, one cannot rule out the presence of a minor fraction of the more thermodynamically stable VL-1 species; therefore our calorimetric and urea unfolding measurements gave the highest possible value for the thermal unfolding enthalpy and G of VL-2.
The VL-1 and VL-2 conformers are kinetically trapped
A distinctive feature of the major part of protein conformers described thus far is their ability to equilibrate under physiological conditions or at least convert into the alternative state, in particular at elevated temperatures or at moderate concentrations of a denaturant. This is the case for several antibodies (Foote and Milstein, 1994; Murray et al., 1995; James et al., 2003) and other proteins that demonstrate conformational diversity, e.g. rabbit muscle adenylate kinase (Zhang et al., 1998), RNAse A and bovine seminal RNAse (Giancola et al., 2000; Sorrentino et al., 2000). In contrast to these previous studies, we could not detect interconversion for the two VL conformations using transient exposure to elevated temperature (15 min at 60°C or 2 weeks at 37°C) or denaturant (up to 7 M urea). No appreciable conversion of VL-1 VL-2 or VL-2 VL-1 type occurred unless the conformers were fully unfolded in 6 M Gdn–HCl with subsequent refolding into either VL-1 or VL-2 conformation according to the respective protocols. Therefore, we conclude that the two VL conformations are separated by a high energetic barrier that prevents interconversion. The scheme given in Fig. 9 illustrates the free energy landscape for the two conformations. Several other examples of kinetically trapped conformers of folding have been reported, e.g. -lytic protease (Baker et al., 1992), subtilisin (Zhu et al., 1989) and ß subunit of luciferase (Sinclair et al., 1994). In these cases, however, formation of the alternative non-functional conformation resulted from either proteolytically cleaving the pro-region (-lytic protease, subtilisin) prior to (re)folding, or from lacking the subunit (luciferase) during the folding process.
|
Functional activity of VL-1 and VL-2
The two alternative conformations of the VL domain are capable of binding human spleen ferritin with closely similar affinities, as determined by ELISA with the VL conformers immobilized by adsorption on the polystyrene surface (Fig. 7). However, the binding affinities of the two conformers were distinct when the assay comprised soluble VL conformers (Fig. 8). In contrast to the more ‘flexible’ VL-2 form, VL-1 in solution showed a dramatic decrease in ferritin-binding capacity. Thus, the VL-1 conformer required surface adsorption to exert its binding activity. One could presume that adsorption-induced conformational changes of VL-1 allowed the binding site to fit the structure of the antigenic epitope on ferritin. The difference in antigen-binding activity between the two conformers is not unexpected given that the environment of Trp-35 is involved in formation of the antigen-binding site of the VL domain (Nymalm et al., 2002). In terms of the highly ordered structure around Trp-35, the surface adsorption of VL-1 induces conformational changes that overcome sterical constraints posed by the Trp-35 environment. The requirement of adsorption-induced conformational changes might provide an explanation as to why binding activity has not been reported thus far for other mouse VLs that presumably adopted the VL-1 conformation. In previous studies of VL domains, functional activity was either not addressed or performed with soluble VLs (e.g.Worn and Pluckthun, 1998; Nishi et al., 2003).
Biological significance of the VL-2 conformer
Whether VL-2 or other alternative VL conformations can be formed in vivo remains to be established. One essential note in this regard is that VL domains are probably not evolutionally optimized for folding in the absence of the partner domains, CL and VH. Furthermore, in vitro refolding protocols involve conditions that are not likely to occur in vivo. In particular, in vitro refolding of proteins extracted from inclusion bodies excludes interactions with intracellular chaperones.
In the recent years, however, the growing use of truncated recombinant antibodies for medical and industrial applications has brought into attention possible pathways of refolding of these non-natural proteins. Antibody VL domains are known to share a high degree of structural similarity. Importantly, both Trp-35 and its environment are highly conserved in other VL domains (e.g. see 1IVL, 1KIR, 1MAJ structures from PDB and references (Constantine et al., 1994; Essen and Skerra, 1994; Fields et al., 1996)). Therefore, one cannot exclude the possibility that other isolated VL domains could exist in macroscopic conformations similar to VL-1 and VL-2 when refolded from a denatured state, e.g. in large scale industrial purification of proteins solubilized from inclusion bodies. The high antigen-binding activity of VL-2 in combination with simplicity of the purification procedure makes this conformer a good candidate for biomedical and biotechnological use. Recently, we constructed and characterized a potential immunotoxin VL-barnase that comprises the VL domain of antibody F11 and barnase, a bacterial RNAse (Martsev et al., 2004; Tsybovsky et al., 2004). Comparison of thermal unfolding profiles and calorimetric enthalpies of VL-barnase and conformers of the VL domain indicates that the VL domain within the fusion protein folds into the VL-2 rather than VL-1 conformation, which supports the hypothesis that alternative functional conformations of recombinant proteins may be of significant biomedical and industrial importance.
|
Footnotes |
|---|
3 Present address: Department of Biochemistry and Molecular Biology, Medical University of South Carolina, Charleston, SC 29425, USA.
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
We are grateful to Dr Christopher Davies (Medical University of South Carolina, USA) for carefully reading the manuscript and for helpful remarks. We would also like to thank anonymous reviewer 1 for the comment on the contribution of aromatic residues to far-UV CD spectra of proteins as well as for the suggestion to apply the phase diagram approach for additional conformational discrimination between VL-1 and VL-2. This work was supported, in part, by the INTAS grant 00-0554 and Belorussian State Research and Technology Programs ‘Genetic Engineering’ (grant 2002-19.12) and ‘Proteomics’ (grant 2003-31).
|
References |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Abramov V.M., et al. Biochemistry (2001) 40:6076–6084.[CrossRef][Medline]
Anfinsen C.B. Science (1973) 181:223–230.
Azuma T., Kobayashi O., Goto Y., Hamaguchi K. J. Biochem. (Tokyo) (1978) 83:1485–1492.
Baker D., Sohl J.L., Agard D.A. Nature (1992) 356:263–265.[CrossRef][Medline]
Bird R.E., Hardman K.D., Jacobson J.W., Johnson S., Kaufman B.M., Lee S.M., Lee T., Pope S.H., Riordan G.S., Whitlow M. Science (1988) 242:423–426.
Bolotina I.A. Mol. Biol. (Mosk.) (1987) 21:1625–1635.[Medline]
Bolotina I.A., Lugauskas V. Mol. Biol. (Mosk.) (1985) 19:1409–1421.[Medline]
Brinkmann U., Lee B.K., Pastan I. J. Immunol. (1993) 150:2774–2782.[Abstract]
Bychkova V.E., Berni R., Rossi G.L., Kutyshenko V.P., Ptitsyn O.B. Biochemistry (1992) 31:7566–7571.[CrossRef][Medline]
Chakrabartty A., Kortemme T., Padmanabhan S., Baldwin R.L. Biochemistry (1993) 32:5560–5565.[CrossRef][Medline]
Clark E.D.B. Curr. Opin. Biotechnol. (1998) 9:157–163.[CrossRef][Web of Science][Medline]
Constantine K.L., Friedrichs M.S., Metzler W.J., Wittekind M., Hensley P., Mueller L. J. Mol. Biol. (1994) 236:310–327.[CrossRef][Web of Science][Medline]
Davies J., Riechmann L. Protein Eng. (1996) 9:531–537.
Devlin G.L., Bottomley S.P. Front. Biosci. (2005) 10:288–299.[CrossRef][Web of Science][Medline]
Ellman G.L. Arch. Biochem. Biophys. (1959) 82:70–77.[CrossRef][Web of Science][Medline]
Essen L.O., Skerra A. J. Mol. Biol. (1994) 238:226–244.[CrossRef][Web of Science][Medline]
Ewert S., Huber T., Honegger A., Pluckthun A. J. Mol. Biol. (2003) 325:531–553.[CrossRef][Web of Science][Medline]
Fields B.A., Goldbaum F.A., Dall'Acqua W., Malchiodi E.L., Cauerhff A., Schwarz F.P., Ysern X., Poljak R.J., Mariuzza R.A. Biochemistry (1996) 35:15494–15503.[CrossRef][Medline]
Foote J., Milstein C. Proc. Natl Acad. Sci. USA (1994) 91:10370–10374.
Franke A.E., Danley D.E., Kaczmarek F.S., Hawrylik S.J., Gerard R.D., Lee S.E., Geoghegan K.F. Biochim. Biophys. Acta (1990) 1037:16–23.[CrossRef][Medline]
Friguet B., Djavadi-Ohaniance L., Goldberg M.E. Mol. Immunol. (1984) 21:673–677.[CrossRef][Web of Science][Medline]
Gavish M., Dwek R.A., Givol D. Biochemistry (1977) 16:3154–3159.[CrossRef][Medline]
Giancola C., Del Vecchio P., De Lorenzo C., Barone R., Piccoli R., D'Alessio G., Barone G. Biochemistry (2000) 39:7964–7972.[CrossRef][Medline]
Goto Y., Takahashi N., Fink A.L. Biochemistry (1990) 29:3480–3488.[CrossRef][Medline]
Huang D.B., Ainsworth C.F., Stevens F.J., Schiffer M. Proc. Natl Acad. Sci. USA (1996) 93:7017–7021.
Hudson P.J. Curr. Opin. Biotechnol. (1998) 9:395–402.[CrossRef][Web of Science][Medline]
Huston J.S., et al. Proc. Natl Acad. Sci. USA (1988) 85:5879–5883.
Ionescu-Zanetti C., Khurana R., Gillespie J.R., Petrick J.S., Trabachino L.C., Minert L.J., Carter S.A., Fink A.L. Proc. Natl Acad. Sci. USA (1999) 96:13175–13179.
James L.C., Roversi P., Tawfik D.S. Science (2003) 299:1362–1367.
Jemmerson R. Proc. Natl Acad. Sci. USA (1987) 84:9180–9184.
Kravchuk Z.I., Chumanevich A.A., Vlasov A.P., Martsev S.P. J. Immunol. Methods (1998) 217:131–141.[CrossRef][Web of Science][Medline]
Kuznetsova I.M., Turoverov K.K., Uversky V.N. J. Proteome Res. (2004) 3:485–494.[CrossRef][Web of Science][Medline]
Lilie H., Schwarz E., Rudolph R. Curr. Opin. Biotechnol. (1998) 9:497–501.[CrossRef][Web of Science][Medline]
Maeda H., Engel J., Schramm H.J. Eur. J. Biochem. (1976) 69:133–139.[CrossRef][Web of Science][Medline]
Manning M.C., Woody R.W. Biochemistry (1989) 28:8609–8613.[CrossRef][Medline]
Martsev S.P., Chumanevich A.A., Vlasov A.P., Dubnovitsky A.P., Tsybovsky Y.I., Deyev S.M., Cozzi A., Arosio P., Kravchuk Z.I. Biochemistry (2000) 39:8047–8057.[CrossRef][Medline]
Martsev S.P., Dubnovitsky A.P., Vlasov A.P., Hoshino M., Hasegawa K., Naiki H., Goto Y. Biochemistry (2002) 41:3389–3395.[CrossRef][Medline]
Martsev S.P., Preygerzon V.A., Mel'nikova Y.I., Kravchuk Z.I., Ponomarev G.V., Lunev V.E., Savitsky A.P. J. Immunol. Methods (1995) 186:293–304.[CrossRef][Web of Science][Medline]
Martsev S.P., Tsybovsky Y.I., Stremovskiy O.A., Odintsov S.G., Balandin T.G., Arosio P., Kravchuk Z.I., Deyev S.M. Protein Eng. Des. Sel. (2004) 17:85–93.
Murray A.J., Lewis S.J., Barclay A.N., Brady R.L. Proc. Natl Acad. Sci. USA (1995) 92:7337–7341.
Myers J.K., Pace C.N., Scholtz J.M. Protein Sci. (1995) 4:2138–2148.[Web of Science][Medline]
Neet K.E., Timm D.E. Protein Sci. (1994) 3:2167–2174.[Web of Science][Medline]
Nishi K., Takai M., Morimune K., Ohkawa H. Biosci. Biotechnol. Biochem. (2003) 67:1358–1367.[CrossRef][Medline]
Nymalm Y., et al. J. Struct. Biol. (2002) 138:171–186.[CrossRef][Web of Science][Medline]
Ohage E., Steipe B. J. Mol. Biol. (1999) 291:1119–1128.[CrossRef][Web of Science][Medline]
Privalov P.L., Potekhin S.A. Methods Enzymol. (1986) 131:4–51.[Medline]
Sinclair J.F., Ziegler M.M., Baldwin T.O. Nat. Struct. Biol. (1994) 1:320–326.[CrossRef][Web of Science][Medline]
Sorrentino S., Barone R., Bucci E., Gotte G., Russo N., Libonati M., D'Alessio G. FEBS Lett. (2000) 466:35–39.[CrossRef][Web of Science][Medline]
Stein P., Chothia C. J. Mol. Biol. (1991) 221:615–621.[CrossRef][Web of Science][Medline]
Thannhauser T.W., Konishi Y., Scheraga H.A. Anal. Biochem. (1984) 138:181–188.[CrossRef][Web of Science][Medline]
Tsumoto K., Ejima D., Kumagai I., Arakawa T. Protein Expr. Purif. (2003) 28:1–8.[CrossRef][Web of Science][Medline]
Tsumoto K., Shinoki K., Kondo H., Uchikawa M., Juji T., Kumagai I. J. Immunol. Methods (1998) 219:119–129.[CrossRef][Web of Science][Medline]
Tsybovsky Y.I., Kedrov A.A., Martsev S.P. FEBS Lett. (2004) 557:248–252.[CrossRef][Web of Science][Medline]
Umetsu M., Tsumoto K., Hara M., Ashish K., Goda S., Adschiri T., Kumagai I. J. Biol. Chem. (2003) 278:8979–8987.
Uversky V.N., Ptitsyn O.B. J. Mol. Biol. (1996) 255:215–228.[CrossRef][Web of Science][Medline]
Ward E.S., Gussow D., Griffiths A.D., Jones P.T., Winter G. Nature (1989) 341:544–546.[CrossRef][Medline]
Worn A., Pluckthun A. Biochemistry (1998) 37:13120–13127.[CrossRef][Medline]
Zhang H.J., Sheng X.R., Niu W.D., Pan X.M., Zhou J.M. J. Biol. Chem. (1998) 273:7448–7456.
Zhu X.L., Ohta Y., Jordan F., Inouye M. Nature (1989) 339:483–484.[CrossRef][Medline]
Received February 15, 2007; revised June 4, 2007; accepted June 5, 2007.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||