- Email: [email protected]
Structure-based Stability Engineering of the Mouse IgG1 Fab Fragment by Modifying Constant Domains
doi:10.1016/j.jmb.2006.06.073
J. Mol. Biol. (2006) 361, 687–697
Structure-based Stability Engineering of the Mouse IgG1 Fab Fragment by Modifying Constant Domains Tuija Teerinen 1 , Jarkko Valjakka 2 , Juha Rouvinen 2 and Kristiina Takkinen 1 ⁎ 1
VTT Biotechnology, P.O. Box 1000, 02044 VTT Espoo, Finland 2
Department of Chemistry, University of Joensuu, P.O. Box 111, 80101 Joensuu, Finland
A semi-rational approach based on structural data was exploited in a search for CH1 and CL domains with improved intrinsic thermodynamic stabilities. Structural and amino acid level comparisons were carried out against known biophysically well-behaving and thermodynamically beneficial scFv and Fab fragments. A number of mutant Fab fragments were constructed by site-directed mutagenesis of regions in the CH1 and CL domains expected to be most sensitive under physical stress conditions. These mutations were located on three sites in the Fab constant domains; a mobile loop in the CH1 domain, residues surrounding the two largest solvated hydrophobic cavities located in the interface of the CH1 and CL domains and the hydrophobic core regions of both CH1 and CL. Expression levels of functional Fab fragments, denaturant-induced unfolding equilibria and circular dichroism spectroscopy were used to evaluate the relative stabilities of the wild-type and the mutant Fab fragments. The highest thermodynamic stability was reached through the mutation strategy, where the hydrophobicity and the packing density of the solvated hydrophobic cavity in the CH1/CL interface was increased by the replacement of the hydrophilic Thr178 in the CL domain by a more hydrophobic residue, valine or isoleucine. The midpoint of the transition curve from native to unfolded states of the protein, measured by fluorescence emission, occurred at concentrations of guanidine hydrochloride of 2.4 M and 2.6 M for the wildtype Fab and the most stable mutants, respectively. Our results illustrate that point mutations targeted to the CH1/CL interface were advantageous for the overall thermodynamic stability of the Fab fragment. © 2006 Elsevier Ltd. All rights reserved.
*Corresponding author
Keywords: Fab fragment; CH1/CL domain; stability engineering; sequence alignment; structure comparison
Introduction Fab fragments, derivatives of immunoglobulins, are four-domain proteins consisting of variable VH
Present address: J. Valjakka, Institute of Medical Technology, University of Tampere, and Tampere University Hospital, Biokatu 6-8, 33520 Tampere, Finland. Abbreviations used: CMO, carboxy-methyloxime; Fab, antibody fragment consisting of the light and the VH and the CH1 domains of the heavy chain; GdnHCl, guanidine hydrochloride; Ig, immunoglobulin; RMSD, RMS atom-positional fluctuation; tm, temperature midpoint of thermal denaturation. E-mail address of the corresponding author: [email protected]
and VL domains and constant CH1 and CL domains. Recombinant antibody fragments, both scFvs containing only VH and VL domains connected by a flexible linker, and Fab fragments, possess optimal characteristics that can be used in several fields of biotechnology, and their potency as research and diagnostic tools and pharmaceutical agents has been proven.1–4 They are expressed easily in bacteria, also in larger scale bioprocesses, and can be modified to improve or control binding specificity, affinity, production and purification yields, detection and immobilization methods and thermodynamic stabilities.5,6 In particular, the constant domains of a Fab fragment offer the possibility for a universal stable platform, which can be combined with variable domains with different antigen-binding specificities. Mutations of single amino acids in the constant domains of a Fab fragment should not alter
0022-2836/$ - see front matter © 2006 Elsevier Ltd. All rights reserved.
688 the antigen-binding properties, since constant domains locate spatially at a distance sufficiently far from the antigen-binding site. However, it has been indicated that Fab constant domains do affect the binding properties and, more specifically, antigen binding may be adjusted by the CH1 and CL domains by suppression of the local conformational changes of the variable domains, which leads to a higher binding affinity as compared to the respective antibody fragments that consist of only Fv domains.7 Instead of the constant domains, structurally very different protein domains have been linked to Fv fragments, such as heterodimeric coiled-coil domains,8 and a heterotetrameric cytoplasmic enzyme, 9 in order to stabilize the Fv fragment part of the protein fusion, increase production yields and enable the construction of bispecific antibody fragments. There is little published data on how the constant domains contribute to the conformational stability of Fab fragments, since much of the work concerning this field has focused on the design of more stable scFv fragments, probably due to the expected broader applicability and the higher expression yields.10–13 Even smaller antibody-like structures than scFvs, i.e. single-domain camelid heavy-chain antibodies, have been developed, which exhibited a conformational stability comparable to that of the most stable scFvs.14,15 Generally, Fab fragments are considered to be more stable, mostly because the hydrophobic interface, which is exposed in scFvs, is buried by the constant domains in Fab fragments. Recently, more detailed knowledge about the effect of constant domains on the stability of variable domains was obtained in studies where Fab domains VL, VH, CH1 and CL with and without disulfide bonds were expressed in different combinations, and the conformational stabilities of the expression units were determined. It was shown that both CH1 and CL constant domains are needed for the stabilization of the variable domains and, further, that only the disulfide-bridged CH1/CL constant domains reached the levels of stability of the most stable variable domains.16 Stability engineering of scFv fragments has been carried out either by means of rational design of the most favourable mutations or random mutagenesis of selected stretches of amino acid residues followed by selection after denaturing stress conditions.12,17 The biophysical characterization of the human scFv frameworks in vitro showed that the most stable variable domains of heavy chains belong to the VH3 subfamily and the most stable variable domains of light chains to the Vκ 3 subfamily. 10, 11, 18 The structural features explaining the stability of the human VH3 and Vκ3 domains were described to include the following: tight packing of the hydrophobic core, hydrophobicity of the core, suitable βsheet propensity, beneficial hydrogen bonding interaction, replacement of less favourable residues by glycine, formation of the charge cluster both within VH and VL domains and intradomain or interdomain disulfide bonds. Following these prin-
Stability Engineering of the Mouse Fab CH1 and CL
ciples, an increase of 20 kJ/mol in the free energy value and a fourfold increase in the expression yield was reported, when six critical residues in a less stable VH6 framework were replaced by those from the most stable VH3.18 For certain applications, e.g. for intrabody expression, where due to the reducing environment of the Escherichia coli cytoplasm, correct disulfide bond formation may be restricted. scFv fragments without intradomain disulfide bonds but with stability levels equal to the respective disulfide-bonded scFvs were developed by in vitro evolution.19,20 In our study, the constant domains of the Fab77 fragment (CL and CH1) were chosen as targets for rational mutagenesis. The anti-testosterone Fab77 fragment with high affinity and specificity has been developed from a monoclonal antibody by random mutagenesis of the CDR regions combined with phage display selections.21,22,24 The three-dimensional structure of Fab77 has been resolved by X-ray crystallography with (Protein Data Bank23 structure 1VPO, formerly 1L7S) and without testosterone (PDB structure 1L7T) at a resolution of 2.15 Å and 2.10 Å, respectively.24 Rational design was preferred over random mutagenesis and directed evolution methods, since the three-dimensional structure of the Fab77 could be used as a starting point for stability engineering. In the site-directed mutagenesis of Fab77 constant domains, knowledge obtained from the analysis of the subfamilies of antibody human variable domains with high stability and good folding properties were utilized and structural alignments with the most stable human germline consensus sequences were performed. In order to improve the stability of the anti-testosterone Fab77 fragment, the regions most critical for thermodynamic stability were evaluated and potentially destabilizing residues were replaced, simultaneously maintaining the immunoglobulin-like structure of the CH1/CL domains in order to use generally applied purification and detection methods. Thus, β-strands known to interact with the protein G immunoglobulin-binding domain i.e. the first and last β-strands of the CH1 domain,25 were not targeted to mutagenesis. Additional cysteine residues were not introduced into the Fab77 constant domains in order to retain the native disulfide bonding.
Results Mutations of the mobile loop in the CH1 domain (residues 128–133) The stretch of amino acid residues SAAQTN located between the first two β strands of the Fab77 CH1 domain, comprising residues 128–133 (the numbering scheme is that of the Kabat database26) (Figure 1), was subjected to rational mutagenesis in order to restrict the structural mobility of this loop in mutations numbered from
Stability Engineering of the Mouse Fab CH1 and CL
689
Figure 1. (a) A ribbon diagram of the Fab77 crystal structure based on B values, (the higher the B value, the stronger the line) and suggested mutations in the CH1/CL. CH1 is shown in green, CL is shown in blue, the testosterone antigen and disulphide bonds are shown in yellow. Sites for mutagenesis and the three largest solvated hydrophobic cavities (C1–C3) are displayed; mutations located in close proximity to the interfacial cavities 1 and 2 are shown in orange, mutations in the hydrophobic core of CH1 and CL domains are shown in red, and the loop between the two first β-strands of CH1 is shown in magenta. The ribbon diagram was generated by PyMOL [http://pymol.sourceforge.net/].36 (b) The amino acid sequences of Fab77 CH1 and CL domains. The numbering is according to Kabat.26 The mutated residues are in bold and shown in colours. The colour coding is the same as in (a).
690
Stability Engineering of the Mouse Fab CH1 and CL
M1 to M8 (Table 1). The larger variation in RMSD values of atomic positions was illustrated by the higher B values that occur in this region in the analysis of the three-dimensional structure of Fab77. Among published murine CH1 domain sequences there is considerable diversity in this solvent-exposed loop region, implying that a modification of the amino acid sequence would be possible also in Fab77, which belongs to the murine IgG1 subclass (Figure 2). In M1, a deletion of all the residues SAAQTN was carried out. In addition to this loop deletion, Pro126 was changed to Ser in M2 and to Thr in M3. In mutations M4 to M8, combinations of deletions of Ala129, Ala 130 and Glu131 were performed, which, according to the structural data do not form stabilizing hydrogen bonds with the neighbouring side chains or with the protein backbone. Pro126 was mutated to the smaller hydrophilic Ser or Thr in order to allow the elongation of the preceding β strand in mutants M7 and M8 (Table 1). Expression levels of these mutated Fab fragments were determined by measuring the amount of active Fab fragment from the periplasmic and culture supernatant samples by ELISA (Figure 5). The expression system used in this study produces Fab fragments with the interchain disulphide bond. Deletion of the whole loop region (residues 128– 133), mutants M1–M3, decreased the expression level around sixfold when compared to that of Fab77. Expression levels of mutants M4, M6 and M7 were comparable to that of Fab 77, whereas mutants M5 and M8 were produced in slightly smaller
amounts. From these loop mutations, M6 was selected over M4 and M7 for further characterization due to its better performance in an ELISA, where the binding activity of culture supernatant samples incubated at different temperatures were analysed (data not shown). The thermodynamic stability of the Fab77 stability mutants was examined by guanidine hydrochloride (GdnHCl) equilibrium denaturation experiments. Unfolding of the Fab77 and selected Fab77 mutants was monitored by the shift of the fluorescence emission maximum as the function of denaturant concentration (Figure 6). The denaturant-induced unfolding curves of the CH1 loop mutation M6 shows that this mutation is denatured at lower concentrations of GdnHCl than the original Fab77 indicating its lower thermodynamic stability. Mutations targeted to the interface between the CH1 and CL domains In the hydrophobic cavity located in the interface between CH1 and CL of Fab77, Phe166 in CH1, Phe118 and Phe135 in CL point into the hydrophobic core and are believed to be essential for the tight packing of the core (Cavity1 in Figure 3). Three water molecules are accommodated in this cavity and one of these participates in hydrogen bonding with the residues Leu160, Ser176 and Thr178 from the VL domain, the second with Ser178 and the third with Thr176 and Pro167 from the CH1 domain. In the mutations denoted M11–M14, M18 and M19 the main objective was to
Table 1. Summary of the Fab77 stability engineering mutations Construct M1 M2 M3 M4 M5 M6 M7 M8 M11 M12 M13 M14 M18 M19 M20 M21 M22 M23 M24 M28 M29 M30
Mutation
Mutation site
Putative structural consequence
Shortening of the CH1 loop Shortening of the CH1 loop, elongation of β-sheet Shortening of the CH1 loop, elongation of β-sheet Shortening of the CH1 loop, not disturbing hydrogen bonding Shortening of the CH1 loop, not disturbing Deletion of A129-Q131 CH1 loop 128-133 hydrogen bonding Shortening of the CH1 loop, increase in Deletion of A129-A130 and Q131N CH1 loop 128-133 hydrophilicity as M6 and elongation of β-sheet Deletion of A129-A130 and Q131N, P126S CH1 loop 128-133 as M5 and elongation of β-sheet Deletion of A129-Q131 and P126S CH1 loop 128-133 Increase in hydrophobicity, remove water L160F CL cavity 1 Increase in hydrophobicity, remove water T178V CL cavity 1 Increase in hydrophobicity, remove water T178 L CL cavity 1 Increase in hydrophobicity, remove water T178 I CL cavity 1 Increase in hydrophobicity, remove water T178V and L160F CL cavity 1 Increase in hydrophobicity, remove water T178V and L160M CL cavity 1 Increase in hydrophobicity, remove water, T178V and deletion of P119 CL cavity 1 and cavity 2 release rigidity Increase in hydrophobicity, remove water T178V and S121L CL cavity 1 and cavity 2 Increase in hydrophobicity, remove water T178V and Q124L CL cavity 1 and cavity 2 Increase in hydrophobicity, remove water T178V and Y122F CL cavity 1 and CH1 cavity 2 Increase in hydrophobicity, remove water T178V and Y122M CL cavity 1 and CH1 cavity 2 Increase in hydrophobicity L177F and S179V CH1 upper core Increase in hydrophobicity N161G and S177F CL upper core Increase in hydrophobicity V115L and N161G and S177F CL upper core Deletion of SAAQTN Deletion of SAAQTN and P126S Deletion of SAAQTN and P126T Deletion of A130-Q131
CH1 loop 128-133 CH1 loop 128-133 CH1 loop 128-133 CH1 loop 128-133
CH1 loop 128-133 denotes a loop region between the first two β-strands of Fab fragment CH1 domain. Cavity 1 and cavity 2 are located in the interfacial region of CH1/CL. The upper cores of CH1 and CL consist of residues in the hydrophobic cores of both CH1 and CL and are superimposable on the antibody variable region upper core (Figure 4).
Stability Engineering of the Mouse Fab CH1 and CL
691
Figure 2. The alignment of the mouse Ig CH1 domains. The sequence and structure of Fab77 (PDB entry 1VPO, H chain) aligned with eight sequences of different Ig subclasses of mouse searched from the IMGT database (IMGT accession number shown).37 The alignment was calculated by 3Dcoffee (regular mode),38 and secondary structure elements of 1VPO were displayed by ESPript.39 The numbering is according to Kabat.26 Conserved positions are shown in red blocks. In 3Dcoffee calculations, data from the following Fab fragment structures available in the PDB database were used in the sequence-structure alignment: 1ZEA for L35252, 2MCP for AF175973, 1C12 for K00689 and 1UZ6 for D78343, 1HI6 for J00479, 4FAB for V00798, and 1IBG for V00799.
increase the hydrophobicity of the core domain and to exclude water molecules from the cavity. Consequently, Leu160 was mutated to the more hydrophobic Phe, and Thr178 was mutated to Val, Leu or Ile in the CL (mutations M11–M14; Table 1). To increase hydrophobicity further, double mutations containing Thr178Val combined with Leu160Phe (mutation M18) or Met (mutation M19) were constructed. The same principles were followed in the design of mutations targeted to the other cavity (Cavity2 Figure 3) in the CH1/CL interface. In the CL, Phe118, Val133 and in CH1, Leu141 line the cavity, contributing to the spacefilling of the hydrophobic core. This cavity contains two entrapped water molecules; one interacts with CL residues Ser121 and Gln124 and the other with Pro119 and Ser131. The following residues were mutated in the interface cavity 2: Pro119 was deleted to decrease the structural constraint (M20), Ser121 (M21) and Gln124 (M22) to Leu to increase hydrophobicity and to hinder the hydrogen bonding between water molecules, and Tyr122 in the CH1 to Phe (M23) or Met (M24). The expression levels of cavity 1 CL mutations, the mutants M12 (Thr178 Val), M14 (Thr178Ile) and the double mutation M19 (Leu160Met and Thr178Val) were higher when compared to that of Fab77 (Figure 5). Interestingly, the Thr178Leu mutation of mutant M13 caused a clear decrease in the expression of functional Fab. The double mutant M23 (Thr178Val in CL and Tyr122Phe in CH1) showed an increased production level. The thermodynamic stability analysis showed that the midpoint for GdnHCl-induced unfolding was 2.6 M for the well-expressed single mutations M12 and M14 and for the double mutation M19, compared to 2.4 M of Fab77 (Figure 5). For the double mutations M18 and M23, with slightly lower expression yields, the midpoint for transition was 2.5 M.
The thermal unfolding transition of Fab77 and one of the thermodynamically most stable mutant, M12, was monitored by CD spectroscopy at 206.5 nm in the far-UV region. The Fab77 and M12 fragments were heated at a constant rate of 1 °C/min. Figure 7 shows the change in the ellipticity of the two Fab fragments as a function of temperature. The strongest decrease in ellipticity occurred at 74.6 °C for Fab77 and at 75.7 °C for M12, indicating loss of secondary structure. Mutations designed by the structural comparison of the most stable human variable domains and Fab77 constant domains The third mutagenesis strategy was based on the structural comparison between the most stable human variable domains and Fab77 constant domain structures. This comparison showed that these domains share substantial structural homology, regardless of the fact that the two β-strands, denoted in a structural context as c′ and c″, which occur in variable domains, are absent from the constant domains. An upper core, as in the hydrophobic interface between VH and VL with a crucial role in the conformational stability, can be identified in the Fab77 interface between CH1 and CL The CH1 and CL domains of Fab77 were separately aligned with human VH3 and Vκ3 (PDB entries 1DHU and 1DH5, respectively), structures with the most favourable biophysical properties. The human variable and Fab77 constant domain structures were superimposed by a least-squares fit of the conserved intradomain disulfide bridges formed between cysteine residues 140 and 195 in CH1 and 134 and 194 in CL, and the Cα atoms of tryptophan residues 154 in CH1 and 148 in CL (Figure 4). Accordingly, Leu177in CH1 was mutated to Phe and Ser179 to Val (M28), based on the occurrence of these residues in respective positions in the most stable human VH3
692
Stability Engineering of the Mouse Fab CH1 and CL
expressed in much smaller amounts than Fab77 (Figure 5) and, for this reason, they were not characterized further.
Discussion Schematically, three different approaches were exploited in the rational stability engineering mutagenesis of the Fab77 fragment. First, the loop located
Figure 3. A detailed overview of the polar contacts between accommodating water molecules in the CH1/CL interface cavities and surrounding residues of Fab77. The cavity calculations were done with pCAST.40 The calculated size for cavity 1 and cavity 2 was 100 Å3. According to the crystallographic structure, the presence of five ordered water molecules was determined, three in cavity 1 and two in cavity 2, respectively. Colour codes: backbone traces of CH1 are shown in green and backbone traces of CL are shown in blue; all the polar contacts between water molecules and the surrounding residues (labelled, backbone and side-chains shown) are shown as red dotted lines; positions of the structured water molecules are shown as dark red spheres. The area occupied by the two cavities is represented by pale grey surfaces. The following residue positions were selected as possible sites of optimization by rational mutagenesis: in CL, Leu160 to Phe or Met, Thr178 to Val, Leu or Ile, deletion of Pro119, Ser121 to Leu and Gln124 to Leu, and in CH1, Tyr122 to Phe or Met. This figure was generated by PyMOL.36
subfamily. In the CL domain, two combination mutations were carried out; Asn161Gly and Ser177Phe (M29) and Val115Leu, Asn161Gly and Ser177Phe (M30). Replacement of Asn161 by Gly was performed in the belief that more space would be available to the bulky phenylalanine in position 177 and that this could increase the hydrophobic nature of the residue. All these mutants were
Figure 4. (a) Superposition of the CH1 domain of Fab77 on a member of the published11 most stable human variable VH3 framework (PDB entry 1DHU, modelled structure). (b) Superposition of Fab77 CL on a member of the most stable human variable Vκ3 framework (PDB entry 1DH5, modelled structure). Cα atoms of the conserved cysteine residues (Cys140 and Cys195 in CH1; Cys134 and Cys194 in CL) and tryptophan residues (Trp154 in CH1 and Trp148 in CL) were used in the least-squares fit. Only the residues of Fab77 are labelled. The side-chain arrangements of the residues located in the hydrophobic upper core of the human VH/VL domain, proposed to be most critical for the conformational stability of human frameworks,11 are shown in grey, superimposed Fab77 CH1 residues are in green and CL residues are in blue, residues selected for mutagenesis are shown in red. This figure was generated by PyMOL.36
Stability Engineering of the Mouse Fab CH1 and CL
693
Figure 5. Expression levels of Fab77 and stability engineering mutants in the RV308 strain. The expression levels of the functional, testosterone-binding, Fab fragments were determined by ELISA from periplasmic and culture supernatant samples. The expression levels in culture supernatant samples (white columns) and in periplasmic samples (hatched columns) were determined from a standard curve of purified Fab77 (concentration based on absorbance at 280 nm). The mean values of four replicate expression level measurements are shown.
between the first two β-strands of the CH1 domain was targeted to mutations, since the B values derived from the three-dimensional structure analysis by X-ray crystallography indicated a high degree of mobility in this loop region of the Fab77 structure. The second strategy for site-directed mutagenesis was based on the fact that the packing of the core regions and sizes of the cavities are critical for the correct conformation and overall stability of folded proteins. The third approach was to compare the structures of CH1 and CL with the most stable human VH and VL frameworks to identify residues that adversely affect the conformational stability of Fab77. The mobile loop in CH1 (residues 128–133) was chosen for mutagenesis due to its structural fluctuation in the B value analysis of the Fab77 structure, indicating that this loop would be the first region to lose its structure during denaturation. Obviously, the fluctuation of the loop structures to a certain degree is allowed or is necessary for the correct folding of the Fab fragment and a too rigid structure hinders the correct arrangement of the β-sheets during the folding process in vivo. The loop located between the two first β-strands in the CH1 of Fab77 is surface-exposed and, in principle, residues hydrophilic by nature would enhance the stability of the loop structure. The stabilizing effect of the strengthened conformational rigidity of the proteins has been realized with proteins other than antibodies.27 However, a rigid protein backbone, especially in the antibody domains, leads to decreased conformational flexibility, a property that may be deleterious with respect to their biological functions. Sequence comparison of CH1 domains among mouse immunoglobulin subclasses, as is shown in Figure 2, indicates that CH1 domains of IgA, IgG1, IgG3, IgG2a IgG2b, IgG2c, IgD and IgE have a quite high sequence variation in the loop comprising residues from 126 to 132, whereas the preceding residues Pro 123 and Leu 124 are highly conserved, as are the residues in the flanking β-strands. Furthermore, the
murine and human IgG1 and Ig2A are the only Ig subclasses not containing a cysteine residue, which participates in the interdomain disulphide bonding between CH1 and CL. Instead, the cysteine residue is located in the C terminus of CH1 in these subclasses. As reported earlier, the overall structure of murine CH1 domains is quite similar but, in addition to segment 127–135 in subtype IgG2b, there is another segment, 163–168, that deviates in structure, and in isotype Ig2A there are several differences compared to other Ig CH1 structures.28 A considerable increase in the thermal stability of the antibody CH3 domain, which has an immunoglobulin β-sheet structure as well, was achieved by residue frequency analysis of CH3 sequences from different species.29 Here, the thermal stability of dimeric CH3 was increased to 86 °C by three mutations, mainly by increasing sidechain packing between opposite dimer sub-units and by a replacement of glycine in the tight turn of a β-strand. The mutations in the mobile loop region of CH1 of Fab77 did not have beneficial properties: Mutations M1–M3, in which the whole loop region was deleted, were expressed in very small amounts (Figure 5). M6 Fab was expressed at the same level as Fab77 but showed decreased thermal stability when compared to that of Fab77 (Figure 6). Fab77 contains three cavities with volumes greater or equal to those sizes of other proteins of the same size,30 one being located in the variable Vκ domain and two of them in the interface of CH1 and CL domains. Some of the mutations targeted to the two largest hydrophobic cavities of the constant domain resulted in an increase in both the thermal stability and the expression level of Fab77. The improved stability could be explained by the fact that more hydrophobic residues in the mutated sites would protrude into the cavity and thus minimise the cavity volume, leading to a more tightly packed hydrophobic core that consequently restricts the movement of the residues in the protein core. These mutations in CL involved the replacement of Thr178 with hydrophobic valine (M12) or isoleucine (M14)
694
Figure 6. GdnHCl denaturation curves of the CH1 loop mutation M6, and the cavity-filling mutations M12, M14, M18, M19, M23 and Fab77. The unfolding transitions of Fab fragments were measured by following the shift in fluorescence emission maximum as a function of the concentration of GdnHCl at the excitation wavelength of 280 nm. The emission was measured from 310 nm to 440 nm, and background fluorescence was subtracted. All final spectra were the average of six scans.
and the combination of Leu160Met and Thr178Val mutations in M19 and in the CH1 replacement of Tyr122 with the more hydrophobic phenylalanine in combination with the CL Thr178Val mutation (M23). Tyr122 in CH1 is assumed to point into the solvent phase and its replacement by phenylalanine at this position may change the orientation of this residue to protrude more towards the cavity. Both Leu160 and Thr178 in CL participate in hydrogen bonding to water molecules in the CL/CH1 interface cavity 1 and, likewise, in the CL domain Ser121 and Gln124 in cavity 2 (Figure 3). It was presumed that by replacing these residues the formation of hydrogen bonds between water molecules would be hindered and so, consequently, would the accommodation of water molecules in these two cavities. The thermal stability of M12 (Thr178Val) and M14 (Thr178Ile) and the double mutant M19 (Leu160Met and Thr178Val) was increased when compared to that of Fab77. It is noteworthy that the double mutant in the CL/CH1 interface Leu160 to Phe and Thr178 to Val in CL (M18) showed a lower level of stability than the M19 mutant (Thr178Val and Leu160Met) and the respective stability-increasing single mutations (M11 and M12), which is presumably caused by the over-packing of the cavity (data not shown for the Leu160Phe mutant). The important stabilizing role of hydrophobic residues in close proximity to cavities has been shown recently for proteins other than antibodies, the change of the hydrophobic residues to more hydrophilic residues resulted in considerable destabilization of the proteins.31,32 In an another study, where small empty cavities of a small globular protein were filled with hydrophobic residues, it was suggested that the packing effects contribute negatively to the protein stability and weaken the benefits of increased hydrophobicity. These cavities were somewhat smaller, 20 Å3, than the two largest cavities in the constant domains of the Fab77 fragment, which have a size of 100 Å3. It
Stability Engineering of the Mouse Fab CH1 and CL
was concluded that even smaller structural rearrangements needed to accommodate larger residues can destabilize proteins.33 On the basis of the fact that both antibody variable and constant domains are structurally quite homologous, even though variable domains contain two additional β-sheets generally denoted as c′ and c″, alignments with the most stable human framework sequences were accomplished in order to reveal if the same guidelines hold for the constant domains and the variable domains. Particular attention was paid to the hydrophobic interface between the CH1 and CL domains, and the hydrophobic cores of constant domains (Figure 4(a) and (b)). Five mutations were carried out, which all destabilized the Fab77 structure significantly and led to decreased expression levels. A possible explanation for this is that, despite their overall structural similarity, they may lack common features in detailed substructures, such as the orientation of certain bond angles, local hydrophobicity and the effect of van der Waals' forces. Fab77 unfolded at a temperature of 75–76 °C, when the thermal unfolding was monitored by circular dichroism spectroscopy in the far-UV region (Figure 7). The thermal denaturation behaviour of Fab77 was compared to that of the published mouse immunoglobulin isotype 2b and its Fab counterpart that had been studied in similar denaturing conditions and protein concentrations. The Fab of isotype 2b derived by papain digestion denatured at ∼61 °C and was proposed to unfold as independent domains, according to circular dichroism and differential scanning calorimetry results.34,35 This example indicates that variation in the thermal stability of Fab fragments may be quite high. Another reason may be that Fab fragments produced by enzymatic digestion are more vulnerable to thermal denaturation than bacterially expressed Fab fragments. More recently, it was shown that a variable domain, which, as such, was observed to be quite unstable, could become more stable and more applicable in the Fab format. This was reasoned by the kinetic stabilization and mutual interaction of the constant
Figure 7. CD temperature scans of Fab77 (open squares) and the mutant M12 (filled diamonds) (150 μg/ml) in 10 mM sodium phosphate (pH 8.1) at 206.5 nm.
695
Stability Engineering of the Mouse Fab CH1 and CL
CH1/CL domain and by the cooperation between the constant CH1/CL and variable VH/VL units.16 In summary, increased thermodynamic stability of the anti-testosterone Fab77 fragment with retained binding affinity was achieved by mutating selected residues that line the two hydrophobic cavities in the interface of the CH1 and CL domains to more hydrophobic residues. Stability engineering focusing only on the constant domains probably increases the overall Fab fragment stability to only a certain degree, because it depends also on the intrinsic stability of the variable domains. We are currently evaluating the performance of the stability engineered constant domains of Fab77 in combination with variable domains with different antigen-binding specificities. A universal Fab fragment constant domain structure with high conformational stability to be used as a stable fusion partner with any variable domain would be greatle advantageous, especially in applications where antibody fragments are exposed to harsh conditions, such as elevated temperatures, non-aqueous solutions or temporary dryness.
Materials and Methods Plasmid constructions The expression vector for the Fab77 wild-type and the stability mutants was a derivative of pKKtac,21 which contains NheI and AscI sites for cloning the VLCL fragment, an intergenic region, and SfiI and NotI sites for cloning the VHCH1 fragment. Mutated codons introduced into primer sequences and PCR reactions were carried out with Vent (NEB) or Dynazyme Ext (Finnzymes, Finland) polymerases using the expression vector of Fab77 as the template. The full-length VLCL or VHCH1 fragments were obtained by PCR assembly reactions. The PCR products were isolated from 1% (w/v) agarose in 40 mM Tris, 20 mM NaAc, 1 mM EDTA (pH 8.0) and purified by the QIAquick gel-extraction kit (Qiagen). The PCR fragments were ligated into the pKKtac expression vector and transformed into E. coli XL-Blue strain (Stratagene). The sequences of the mutated Fab expression units were verified by an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems) according to the manufacturer's protocol (ABI PRISM BigDye Terminator Cycle Sequencing Kit v.1.1, Applied Biosystems). Expression and purification of the Fab fragments The Fab fragments were expressed in the E. coli RV308 strain (ATCC 31608). The pKKtac expression constructs of the mutant Fab fragments were transformed into the RV308 strain. LB medium containing 100 μg/ml of ampicillin and 1% (w/v) glucose was inoculated with a single bacterial colony and incubated overnight at 37 °C. Either 100 ml (small-scale) or 1 l (for Fab fragment purification) of LB medium with 100 μg/ml of ampicillin was inoculated (1:50 dilution) with the overnight culture and incubated at 37 °C. When the cell density reached an absorbance at 600 nm (A600) of 1.0, the expression of Fab fragments was induced by the addition of isopropyl-β-Dthiogalactopyranoside (IPTG from Sigma) to a final
concentration of 1 mM. Cells were grown for 18 h at 30 °C and were harvested by centrifugation (3000g, 15 min at 4 °C). The culture supernatant was used for ELISA analysis and for Fab fragment purification. Periplasmic fractions were isolated from overnight bacterial cultures by sequential incubations (repeated three times) of the harvested cells (from 1 ml) in 0.5 ml of PBS in a solid CO2/ ethanol bath for 5 min followed by thawing at 37 °C. The cell debris was pelleted by centrifugation (20,000g, 5 min at 4 °C) and the clear supernatant containing the periplasmic proteins was used for ELISA. Fab fragments were purified batch-wise by passage through a protein G Sepharose Fast Flow (Pharmacia Biotech) affinity chromatography column. One litre of the supernatant was incubated at 4 °C with gentle shaking overnight with a 5 ml bed-volume of protein G gel and loaded onto the chromatography column. After washing with 20 mM sodium phosphate (pH 7.0), elution was performed with 0.1 M glycine–HCl (pH 2.7) and the eluted fractions were neutralized with 1 M Tris–HCl (pH 9.0). The concentration of Fab was calculated from the A280 of purified samples using an extinction coefficient of 1.558 (1.0 g/l solution). The purity of the protein was confirmed by SDS-PAGE stained with Coomassie brilliant blue. The presence of the interchain disulphide bond between the light chain and the heavy chain of the expressed Fab fragments was confirmed by SDS-PAGE under nonreducing conditions (data not shown). Enzyme-linked immunosorbent assay (ELISA) A testosterone-3-CMO-bovine serum albumin conjugate (Sigma) was coated at a concentration of 2 μg/ml in 0.1 M NaHCO3 (pH 9.6) on 96-well plates overnight at 4 °C. Plates were blocked with 1% (w/v) bovine serum albumin (Sigma) in PBS for 2 h at room temperature. After incubation of the supernatant samples (1:100 (v/v) or 1:200 (v/v) dilutions) or periplasmic fractions (1:50 (v/v) and 1:100 (v/v) dilutions) for 1 h, the Fab fragments were detected by a goat anti-mouse IgG Fab specific alkaline phosphatase conjugate (Sigma). Equilibrium denaturation experiment Fluorescence spectra were collected on a Cary Eclipse spectrofluorimeter (Varian). The excitation wavelength was 280 nm and the slit-width was 5 nm for excitation and for emission. The fluorescence of samples, 0.5 μM protein and concentrations of GdnHCl from 0 to 4.0 M in a 20 mM Hepes buffer (pH 7.0) incubated overnight at 4 °C, was recorded between 310 nm and 440 nm. Blank subtractions were made in all spectra, which were averaged from six scans. The final concentration of the GdnHCl stock solution (8 M) was determined from its refractive index. The fluorescence emission maximum as a function of the concentration of denaturant was calculated using a Gaussian fit from the fluorescence emission data. Circular dichroism measurements Temperature-induced unfolding was monitored by a JASCO J-715 CD spectropolarimeter, equipped with a JASCO PTC-348WI thermocouple (JASCO International). A quartz cuvette of 0.1 cm path-length sealed with a cap to avoid evaporation was used. The temperature was increased from 35 °C to 90 °C at a rate of 1 °C/min (resolution of 0.2 °C and a time constant of 16 s). Fab77
696 and M12 Fab fragments were measured in 10 mM sodium phosphate (pH 8.1) at a final concentration of protein of 150 μg/ml. The wavelength during the temperature scan was 206.5 nm.
Acknowledgements Financial support from the National Technology Agency (Finland) is gratefully acknowledged. Berg is thanked for excellent technical assistance and Dr Timo Pulli for helpful discussions.
References 1. Borrebaeck, C. A. & Carlsson, R. (2001). Human therapeutic antibodies. Curr. Opin. Pharmacol. 1, 404–408. 2. Kontermann, R. E. (2004). Intrabodies as therapeutic agents. Methods, 34, 163–170. 3. Pavlickova, P., Schneider, E. M. & Hug, H. (2004). Advances in recombinant antibody microarrays. Clin. Chim. Acta, 343, 17–35. 4. Yau, K. Y. F., Lee, H. & Hall, J. C. (2003). Emerging trends in the synthesis and improvement of haptenspecific recombinant antibodies. Biotechnol. Advan. 21, 599–637. 5. Roque, A. C., Lowe, C. R. & Taipa, M. A. (2004). Antibodies and genetically engineered related molecules: production and purification. Biotechnol. Prog. 20, 639–654. 6. Holliger, P. & Hudson, P. J. (2005). Engineered antibody fragments and the rise of single domains. Nature Biotechnol. 23, 1126–1136. 7. Adachi, M., Kurihara, Y., Nojima, H., Takeda-Shitaka, M., Kamiya, K. & Umeyama, H. (2003). Interaction between the antigen and antibody is controlled by the constant domains: normal mode dynamics of the HEL-HyHEL-10 complex. Protein Sci. 12, 2125–2131. 8. Arndt, K. M., Muller, K. M. & Pluckthun, A. (2001). Helix-stabilized Fv (hsFv) antibody fragments: substituting the constant domains of a Fab fragment for a heterodimeric coiled-coil domain. J. Mol. Biol. 312, 221–228. 9. Petrov, K., Dion, M., Hoffmann, L., Dintinger, T., Defontaine, A. & Tellier, C. (2004). Bivalent Fv antibody fragments obtained by substituting the constant domains of a Fab fragment with heterotetrameric molybdopterin synthase. J. Mol. Biol. 341, 1039–1048. 10. Ewert, S., Honegger, A. & Pluckthun, A. (2004). Stability improvement of antibodies for extracellular and intracellular applications: CDR grafting to stable frameworks and structure-based framework engineering. Methods, 34, 184–199. 11. Ewert, S., Huber, T., Honegger, A. & Pluckthun, A. (2003). Biophysical properties of human antibody variable domains. J. Mol. Biol. 325, 531–553. 12. Worn, A. & Pluckthun, A. (2001). Stability engineering of antibody single-chain Fv fragments. J. Mol. Biol. 305, 989–1010. 13. Tan, P. H., Sandmaier, B. M. & Stayton, P. S. (1998). Contributions of a highly conserved VH/VL hydrogen bonding interaction to scFv folding stability and refolding efficiency. Biophys. J. 75, 1473–1482.
Stability Engineering of the Mouse Fab CH1 and CL
14. Ewert, S., Cambillau, C., Conrath, K. & Pluckthun, A. (2002). Biophysical properties of camelid V(HH) domains compared to those of human V(H)3 domains. Biochemistry, 41, 3628–3636. 15. Dumoulin, M., Conrath, K., Van Meirhaeghe, A., Meersman, F., Heremans, K., Frenken, L. G. et al. (2002). Single-domain antibody fragments with high conformational stability. Protein Sci. 11, 500–515. 16. Rothlisberger, D., Honegger, A. & Pluckthun, A. (2005). Domain interactions in the Fab fragment: a comparative evaluation of the single-chain Fv and Fab format engineered with variable domains of different stability. J. Mol. Biol. 347, 773–789. 17. Jung, S., Honegger, A. & Pluckthun, A. (1999). Selection for improved protein stability by phage display. J. Mol. Biol. 294, 163–180. 18. Ewert, S., Honegger, A. & Pluckthun, A. (2003). Structure-based improvement of the biophysical properties of immunoglobulin VH domains with a generalizable approach. Biochemistry, 42, 1517–1528. 19. Proba, K., Worn, A., Honegger, A. & Pluckthun, A. (1998). Antibody scFv fragments without disulfide bonds made by molecular evolution. J. Mol. Biol. 275, 245–253. 20. Worn, A. & Pluckthun, A. (1998). An intrinsically stable antibody scFv fragment can tolerate the loss of both disulfide bonds and fold correctly. FEBS Letters, 427, 357–361. 21. Hemminki, A., Niemi, S., Hautoniemi, L., Soderlund, H. & Takkinen, K. (1998). Fine tuning of an antitestosterone antibody binding site by stepwise optimisation of the CDRs. Immunotechnology, 4, 59–69. 22. Hemminki, A., Niemi, S., Hoffren, A. M., Hakalahti, L., Soderlund, H. & Takkinen, K. (1998). Specificity improvement of a recombinant anti-testosterone Fab fragment by CDRIII mutagenesis and phage display selection. Protein Eng. 11, 311–319. 23. Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T. N., Weissig, H. et al. (2000). The Protein Data Bank. Nucl. Acids Res. 28, 235–242. 24. Valjakka, J., Hemminki, A., Niemi, S., Soderlund, H., Takkinen, K. & Rouvinen, J. (2002). Crystal structure of an in vitro affinity- and specificity-matured antitestosterone Fab in complex with testosterone. Improved affinity results from small structural changes within the variable domains. J. Biol. Chem. 277, 44021–44027. 25. Derrick, J. P. & Wigley, D. B. (1994). The third IgGbinding domain from streptococcal protein G: an analysis by X-ray crystallography of the structure alone and in a complex with Fab. J. Mol. Biol. 243, 906–918. 26. Kabat, E. A., Wu, T. T., Reid-Miller, M., Perry, H. M. & Gottesman, K. S. (1991). Sequence of Proteins of Immunological Interest, 5th edit, NIH Publication no. 91-3242, National Institute of Health, Bethesda, MD. 27. Schimmele, B. & Plückthun, A. (2005). Engineering proteins for stability and efficient folding. In Protein Folding Handbook (Buchner, J. & Kiefhaber, T., eds), vol. 5, pp. 1281–1333, Wiley Verlag GmbH & Co. KGaA., Weinheim. 28. Sheriff, S., Jeffrey, P. D. & Bajorath, J. (1996). Comparison of CH1 domains in different classes of murine antibodies. J. Mol. Biol. 263, 385–389. 29. Demarest, S. J., Rogers, J. & Hansen, G. (2004). Optimization of the antibody CH3 domain by residue frequency analysis of IgG sequences. J. Mol. Biol. 335, 41–48. 30. Cuff, A. L. & Martin, A. C. (2004). Analysis of void volumes in proteins and application to stability of the
Stability Engineering of the Mouse Fab CH1 and CL
31.
32.
33.
34.
p53 tumour suppressor protein. J. Mol. Biol. 344, 1199–1209. Kadonosono, T., Chatani, E., Hayashi, R., Moriyama, H. & Ueki, T. (2003). Minimization of cavity size ensures protein stability and folding: structures of Phe46-replaced bovine pancreatic RNase A. Biochemistry, 42, 10651–10658. Adamek, D. H., Guerrero, L., Blaber, M. & Caspar, D. L. D. (2005). Structural and energetic consequences of mutations in a solvated hydrophobic cavity. J. Mol. Biol. 346, 307–318. Bueno, M., Cremades, N., Neira, J. L. & Sancho, J. (2006). Filling small, empty protein cavities: structural and energetic consequences. J. Mol. Biol. 358, 701–712. Vermeer, A. W. P., Norde, W. & van Amerongen, A. (2000). The unfolding/denaturation of immunogammaglobulin of isotype 2b and its Fab and Fc fragments. Biophys. J. 79, 2150–2154.
697 35. Vermeer, A. W. & Norde, W. (2000). The thermal stability of immunoglobulin: unfolding and aggregation of a multi-domain protein. Biophys. J. 78, 394–404. 36. DeLano, W. (2002). The PyMOL User's Manual. DeLano Scientific, San Carlos, CA. 37. Lefranc, M.-P. (2004). IMGT-ontology and IMGT databases, tools and web resources for immunogenetics and immunoinformatics. Mol. Immunol. 40, 647–659. 38. Poirot, O., Suhre, K., Abergel, C., O'Toole, E. & Notredame, C. (2004). 3DCoffee@igs: a web server for combining sequences and structures into a multiple sequence alignment. Nucl. Acids Res. 32, W37–W40. 39. Gouet, P., Robert, X. & Courcelle, E. (2003). ESPript/ ENDscript: extracting and rendering sequence and 3D information from atomic structures of proteins. Nucl. Acids Res. 31, 3320–3323. 40. Binkowski, T. A., Naghibzadeh, S. & Liang, J. (2003). CASTp: computed atlas of surface topography of proteins. Nucl. Acids Res. 31, 3352–3355.
Edited by R. Huber (Received 26 January 2006; received in revised form 19 June 2006; accepted 21 June 2006) Available online 28 July 2006
J. Mol. Biol. (2006) 361, 687–697
Structure-based Stability Engineering of the Mouse IgG1 Fab Fragment by Modifying Constant Domains Tuija Teerinen 1 , Jarkko Valjakka 2 , Juha Rouvinen 2 and Kristiina Takkinen 1 ⁎ 1
VTT Biotechnology, P.O. Box 1000, 02044 VTT Espoo, Finland 2
Department of Chemistry, University of Joensuu, P.O. Box 111, 80101 Joensuu, Finland
A semi-rational approach based on structural data was exploited in a search for CH1 and CL domains with improved intrinsic thermodynamic stabilities. Structural and amino acid level comparisons were carried out against known biophysically well-behaving and thermodynamically beneficial scFv and Fab fragments. A number of mutant Fab fragments were constructed by site-directed mutagenesis of regions in the CH1 and CL domains expected to be most sensitive under physical stress conditions. These mutations were located on three sites in the Fab constant domains; a mobile loop in the CH1 domain, residues surrounding the two largest solvated hydrophobic cavities located in the interface of the CH1 and CL domains and the hydrophobic core regions of both CH1 and CL. Expression levels of functional Fab fragments, denaturant-induced unfolding equilibria and circular dichroism spectroscopy were used to evaluate the relative stabilities of the wild-type and the mutant Fab fragments. The highest thermodynamic stability was reached through the mutation strategy, where the hydrophobicity and the packing density of the solvated hydrophobic cavity in the CH1/CL interface was increased by the replacement of the hydrophilic Thr178 in the CL domain by a more hydrophobic residue, valine or isoleucine. The midpoint of the transition curve from native to unfolded states of the protein, measured by fluorescence emission, occurred at concentrations of guanidine hydrochloride of 2.4 M and 2.6 M for the wildtype Fab and the most stable mutants, respectively. Our results illustrate that point mutations targeted to the CH1/CL interface were advantageous for the overall thermodynamic stability of the Fab fragment. © 2006 Elsevier Ltd. All rights reserved.
*Corresponding author
Keywords: Fab fragment; CH1/CL domain; stability engineering; sequence alignment; structure comparison
Introduction Fab fragments, derivatives of immunoglobulins, are four-domain proteins consisting of variable VH
Present address: J. Valjakka, Institute of Medical Technology, University of Tampere, and Tampere University Hospital, Biokatu 6-8, 33520 Tampere, Finland. Abbreviations used: CMO, carboxy-methyloxime; Fab, antibody fragment consisting of the light and the VH and the CH1 domains of the heavy chain; GdnHCl, guanidine hydrochloride; Ig, immunoglobulin; RMSD, RMS atom-positional fluctuation; tm, temperature midpoint of thermal denaturation. E-mail address of the corresponding author: [email protected]
and VL domains and constant CH1 and CL domains. Recombinant antibody fragments, both scFvs containing only VH and VL domains connected by a flexible linker, and Fab fragments, possess optimal characteristics that can be used in several fields of biotechnology, and their potency as research and diagnostic tools and pharmaceutical agents has been proven.1–4 They are expressed easily in bacteria, also in larger scale bioprocesses, and can be modified to improve or control binding specificity, affinity, production and purification yields, detection and immobilization methods and thermodynamic stabilities.5,6 In particular, the constant domains of a Fab fragment offer the possibility for a universal stable platform, which can be combined with variable domains with different antigen-binding specificities. Mutations of single amino acids in the constant domains of a Fab fragment should not alter
0022-2836/$ - see front matter © 2006 Elsevier Ltd. All rights reserved.
688 the antigen-binding properties, since constant domains locate spatially at a distance sufficiently far from the antigen-binding site. However, it has been indicated that Fab constant domains do affect the binding properties and, more specifically, antigen binding may be adjusted by the CH1 and CL domains by suppression of the local conformational changes of the variable domains, which leads to a higher binding affinity as compared to the respective antibody fragments that consist of only Fv domains.7 Instead of the constant domains, structurally very different protein domains have been linked to Fv fragments, such as heterodimeric coiled-coil domains,8 and a heterotetrameric cytoplasmic enzyme, 9 in order to stabilize the Fv fragment part of the protein fusion, increase production yields and enable the construction of bispecific antibody fragments. There is little published data on how the constant domains contribute to the conformational stability of Fab fragments, since much of the work concerning this field has focused on the design of more stable scFv fragments, probably due to the expected broader applicability and the higher expression yields.10–13 Even smaller antibody-like structures than scFvs, i.e. single-domain camelid heavy-chain antibodies, have been developed, which exhibited a conformational stability comparable to that of the most stable scFvs.14,15 Generally, Fab fragments are considered to be more stable, mostly because the hydrophobic interface, which is exposed in scFvs, is buried by the constant domains in Fab fragments. Recently, more detailed knowledge about the effect of constant domains on the stability of variable domains was obtained in studies where Fab domains VL, VH, CH1 and CL with and without disulfide bonds were expressed in different combinations, and the conformational stabilities of the expression units were determined. It was shown that both CH1 and CL constant domains are needed for the stabilization of the variable domains and, further, that only the disulfide-bridged CH1/CL constant domains reached the levels of stability of the most stable variable domains.16 Stability engineering of scFv fragments has been carried out either by means of rational design of the most favourable mutations or random mutagenesis of selected stretches of amino acid residues followed by selection after denaturing stress conditions.12,17 The biophysical characterization of the human scFv frameworks in vitro showed that the most stable variable domains of heavy chains belong to the VH3 subfamily and the most stable variable domains of light chains to the Vκ 3 subfamily. 10, 11, 18 The structural features explaining the stability of the human VH3 and Vκ3 domains were described to include the following: tight packing of the hydrophobic core, hydrophobicity of the core, suitable βsheet propensity, beneficial hydrogen bonding interaction, replacement of less favourable residues by glycine, formation of the charge cluster both within VH and VL domains and intradomain or interdomain disulfide bonds. Following these prin-
Stability Engineering of the Mouse Fab CH1 and CL
ciples, an increase of 20 kJ/mol in the free energy value and a fourfold increase in the expression yield was reported, when six critical residues in a less stable VH6 framework were replaced by those from the most stable VH3.18 For certain applications, e.g. for intrabody expression, where due to the reducing environment of the Escherichia coli cytoplasm, correct disulfide bond formation may be restricted. scFv fragments without intradomain disulfide bonds but with stability levels equal to the respective disulfide-bonded scFvs were developed by in vitro evolution.19,20 In our study, the constant domains of the Fab77 fragment (CL and CH1) were chosen as targets for rational mutagenesis. The anti-testosterone Fab77 fragment with high affinity and specificity has been developed from a monoclonal antibody by random mutagenesis of the CDR regions combined with phage display selections.21,22,24 The three-dimensional structure of Fab77 has been resolved by X-ray crystallography with (Protein Data Bank23 structure 1VPO, formerly 1L7S) and without testosterone (PDB structure 1L7T) at a resolution of 2.15 Å and 2.10 Å, respectively.24 Rational design was preferred over random mutagenesis and directed evolution methods, since the three-dimensional structure of the Fab77 could be used as a starting point for stability engineering. In the site-directed mutagenesis of Fab77 constant domains, knowledge obtained from the analysis of the subfamilies of antibody human variable domains with high stability and good folding properties were utilized and structural alignments with the most stable human germline consensus sequences were performed. In order to improve the stability of the anti-testosterone Fab77 fragment, the regions most critical for thermodynamic stability were evaluated and potentially destabilizing residues were replaced, simultaneously maintaining the immunoglobulin-like structure of the CH1/CL domains in order to use generally applied purification and detection methods. Thus, β-strands known to interact with the protein G immunoglobulin-binding domain i.e. the first and last β-strands of the CH1 domain,25 were not targeted to mutagenesis. Additional cysteine residues were not introduced into the Fab77 constant domains in order to retain the native disulfide bonding.
Results Mutations of the mobile loop in the CH1 domain (residues 128–133) The stretch of amino acid residues SAAQTN located between the first two β strands of the Fab77 CH1 domain, comprising residues 128–133 (the numbering scheme is that of the Kabat database26) (Figure 1), was subjected to rational mutagenesis in order to restrict the structural mobility of this loop in mutations numbered from
Stability Engineering of the Mouse Fab CH1 and CL
689
Figure 1. (a) A ribbon diagram of the Fab77 crystal structure based on B values, (the higher the B value, the stronger the line) and suggested mutations in the CH1/CL. CH1 is shown in green, CL is shown in blue, the testosterone antigen and disulphide bonds are shown in yellow. Sites for mutagenesis and the three largest solvated hydrophobic cavities (C1–C3) are displayed; mutations located in close proximity to the interfacial cavities 1 and 2 are shown in orange, mutations in the hydrophobic core of CH1 and CL domains are shown in red, and the loop between the two first β-strands of CH1 is shown in magenta. The ribbon diagram was generated by PyMOL [http://pymol.sourceforge.net/].36 (b) The amino acid sequences of Fab77 CH1 and CL domains. The numbering is according to Kabat.26 The mutated residues are in bold and shown in colours. The colour coding is the same as in (a).
690
Stability Engineering of the Mouse Fab CH1 and CL
M1 to M8 (Table 1). The larger variation in RMSD values of atomic positions was illustrated by the higher B values that occur in this region in the analysis of the three-dimensional structure of Fab77. Among published murine CH1 domain sequences there is considerable diversity in this solvent-exposed loop region, implying that a modification of the amino acid sequence would be possible also in Fab77, which belongs to the murine IgG1 subclass (Figure 2). In M1, a deletion of all the residues SAAQTN was carried out. In addition to this loop deletion, Pro126 was changed to Ser in M2 and to Thr in M3. In mutations M4 to M8, combinations of deletions of Ala129, Ala 130 and Glu131 were performed, which, according to the structural data do not form stabilizing hydrogen bonds with the neighbouring side chains or with the protein backbone. Pro126 was mutated to the smaller hydrophilic Ser or Thr in order to allow the elongation of the preceding β strand in mutants M7 and M8 (Table 1). Expression levels of these mutated Fab fragments were determined by measuring the amount of active Fab fragment from the periplasmic and culture supernatant samples by ELISA (Figure 5). The expression system used in this study produces Fab fragments with the interchain disulphide bond. Deletion of the whole loop region (residues 128– 133), mutants M1–M3, decreased the expression level around sixfold when compared to that of Fab77. Expression levels of mutants M4, M6 and M7 were comparable to that of Fab 77, whereas mutants M5 and M8 were produced in slightly smaller
amounts. From these loop mutations, M6 was selected over M4 and M7 for further characterization due to its better performance in an ELISA, where the binding activity of culture supernatant samples incubated at different temperatures were analysed (data not shown). The thermodynamic stability of the Fab77 stability mutants was examined by guanidine hydrochloride (GdnHCl) equilibrium denaturation experiments. Unfolding of the Fab77 and selected Fab77 mutants was monitored by the shift of the fluorescence emission maximum as the function of denaturant concentration (Figure 6). The denaturant-induced unfolding curves of the CH1 loop mutation M6 shows that this mutation is denatured at lower concentrations of GdnHCl than the original Fab77 indicating its lower thermodynamic stability. Mutations targeted to the interface between the CH1 and CL domains In the hydrophobic cavity located in the interface between CH1 and CL of Fab77, Phe166 in CH1, Phe118 and Phe135 in CL point into the hydrophobic core and are believed to be essential for the tight packing of the core (Cavity1 in Figure 3). Three water molecules are accommodated in this cavity and one of these participates in hydrogen bonding with the residues Leu160, Ser176 and Thr178 from the VL domain, the second with Ser178 and the third with Thr176 and Pro167 from the CH1 domain. In the mutations denoted M11–M14, M18 and M19 the main objective was to
Table 1. Summary of the Fab77 stability engineering mutations Construct M1 M2 M3 M4 M5 M6 M7 M8 M11 M12 M13 M14 M18 M19 M20 M21 M22 M23 M24 M28 M29 M30
Mutation
Mutation site
Putative structural consequence
Shortening of the CH1 loop Shortening of the CH1 loop, elongation of β-sheet Shortening of the CH1 loop, elongation of β-sheet Shortening of the CH1 loop, not disturbing hydrogen bonding Shortening of the CH1 loop, not disturbing Deletion of A129-Q131 CH1 loop 128-133 hydrogen bonding Shortening of the CH1 loop, increase in Deletion of A129-A130 and Q131N CH1 loop 128-133 hydrophilicity as M6 and elongation of β-sheet Deletion of A129-A130 and Q131N, P126S CH1 loop 128-133 as M5 and elongation of β-sheet Deletion of A129-Q131 and P126S CH1 loop 128-133 Increase in hydrophobicity, remove water L160F CL cavity 1 Increase in hydrophobicity, remove water T178V CL cavity 1 Increase in hydrophobicity, remove water T178 L CL cavity 1 Increase in hydrophobicity, remove water T178 I CL cavity 1 Increase in hydrophobicity, remove water T178V and L160F CL cavity 1 Increase in hydrophobicity, remove water T178V and L160M CL cavity 1 Increase in hydrophobicity, remove water, T178V and deletion of P119 CL cavity 1 and cavity 2 release rigidity Increase in hydrophobicity, remove water T178V and S121L CL cavity 1 and cavity 2 Increase in hydrophobicity, remove water T178V and Q124L CL cavity 1 and cavity 2 Increase in hydrophobicity, remove water T178V and Y122F CL cavity 1 and CH1 cavity 2 Increase in hydrophobicity, remove water T178V and Y122M CL cavity 1 and CH1 cavity 2 Increase in hydrophobicity L177F and S179V CH1 upper core Increase in hydrophobicity N161G and S177F CL upper core Increase in hydrophobicity V115L and N161G and S177F CL upper core Deletion of SAAQTN Deletion of SAAQTN and P126S Deletion of SAAQTN and P126T Deletion of A130-Q131
CH1 loop 128-133 CH1 loop 128-133 CH1 loop 128-133 CH1 loop 128-133
CH1 loop 128-133 denotes a loop region between the first two β-strands of Fab fragment CH1 domain. Cavity 1 and cavity 2 are located in the interfacial region of CH1/CL. The upper cores of CH1 and CL consist of residues in the hydrophobic cores of both CH1 and CL and are superimposable on the antibody variable region upper core (Figure 4).
Stability Engineering of the Mouse Fab CH1 and CL
691
Figure 2. The alignment of the mouse Ig CH1 domains. The sequence and structure of Fab77 (PDB entry 1VPO, H chain) aligned with eight sequences of different Ig subclasses of mouse searched from the IMGT database (IMGT accession number shown).37 The alignment was calculated by 3Dcoffee (regular mode),38 and secondary structure elements of 1VPO were displayed by ESPript.39 The numbering is according to Kabat.26 Conserved positions are shown in red blocks. In 3Dcoffee calculations, data from the following Fab fragment structures available in the PDB database were used in the sequence-structure alignment: 1ZEA for L35252, 2MCP for AF175973, 1C12 for K00689 and 1UZ6 for D78343, 1HI6 for J00479, 4FAB for V00798, and 1IBG for V00799.
increase the hydrophobicity of the core domain and to exclude water molecules from the cavity. Consequently, Leu160 was mutated to the more hydrophobic Phe, and Thr178 was mutated to Val, Leu or Ile in the CL (mutations M11–M14; Table 1). To increase hydrophobicity further, double mutations containing Thr178Val combined with Leu160Phe (mutation M18) or Met (mutation M19) were constructed. The same principles were followed in the design of mutations targeted to the other cavity (Cavity2 Figure 3) in the CH1/CL interface. In the CL, Phe118, Val133 and in CH1, Leu141 line the cavity, contributing to the spacefilling of the hydrophobic core. This cavity contains two entrapped water molecules; one interacts with CL residues Ser121 and Gln124 and the other with Pro119 and Ser131. The following residues were mutated in the interface cavity 2: Pro119 was deleted to decrease the structural constraint (M20), Ser121 (M21) and Gln124 (M22) to Leu to increase hydrophobicity and to hinder the hydrogen bonding between water molecules, and Tyr122 in the CH1 to Phe (M23) or Met (M24). The expression levels of cavity 1 CL mutations, the mutants M12 (Thr178 Val), M14 (Thr178Ile) and the double mutation M19 (Leu160Met and Thr178Val) were higher when compared to that of Fab77 (Figure 5). Interestingly, the Thr178Leu mutation of mutant M13 caused a clear decrease in the expression of functional Fab. The double mutant M23 (Thr178Val in CL and Tyr122Phe in CH1) showed an increased production level. The thermodynamic stability analysis showed that the midpoint for GdnHCl-induced unfolding was 2.6 M for the well-expressed single mutations M12 and M14 and for the double mutation M19, compared to 2.4 M of Fab77 (Figure 5). For the double mutations M18 and M23, with slightly lower expression yields, the midpoint for transition was 2.5 M.
The thermal unfolding transition of Fab77 and one of the thermodynamically most stable mutant, M12, was monitored by CD spectroscopy at 206.5 nm in the far-UV region. The Fab77 and M12 fragments were heated at a constant rate of 1 °C/min. Figure 7 shows the change in the ellipticity of the two Fab fragments as a function of temperature. The strongest decrease in ellipticity occurred at 74.6 °C for Fab77 and at 75.7 °C for M12, indicating loss of secondary structure. Mutations designed by the structural comparison of the most stable human variable domains and Fab77 constant domains The third mutagenesis strategy was based on the structural comparison between the most stable human variable domains and Fab77 constant domain structures. This comparison showed that these domains share substantial structural homology, regardless of the fact that the two β-strands, denoted in a structural context as c′ and c″, which occur in variable domains, are absent from the constant domains. An upper core, as in the hydrophobic interface between VH and VL with a crucial role in the conformational stability, can be identified in the Fab77 interface between CH1 and CL The CH1 and CL domains of Fab77 were separately aligned with human VH3 and Vκ3 (PDB entries 1DHU and 1DH5, respectively), structures with the most favourable biophysical properties. The human variable and Fab77 constant domain structures were superimposed by a least-squares fit of the conserved intradomain disulfide bridges formed between cysteine residues 140 and 195 in CH1 and 134 and 194 in CL, and the Cα atoms of tryptophan residues 154 in CH1 and 148 in CL (Figure 4). Accordingly, Leu177in CH1 was mutated to Phe and Ser179 to Val (M28), based on the occurrence of these residues in respective positions in the most stable human VH3
692
Stability Engineering of the Mouse Fab CH1 and CL
expressed in much smaller amounts than Fab77 (Figure 5) and, for this reason, they were not characterized further.
Discussion Schematically, three different approaches were exploited in the rational stability engineering mutagenesis of the Fab77 fragment. First, the loop located
Figure 3. A detailed overview of the polar contacts between accommodating water molecules in the CH1/CL interface cavities and surrounding residues of Fab77. The cavity calculations were done with pCAST.40 The calculated size for cavity 1 and cavity 2 was 100 Å3. According to the crystallographic structure, the presence of five ordered water molecules was determined, three in cavity 1 and two in cavity 2, respectively. Colour codes: backbone traces of CH1 are shown in green and backbone traces of CL are shown in blue; all the polar contacts between water molecules and the surrounding residues (labelled, backbone and side-chains shown) are shown as red dotted lines; positions of the structured water molecules are shown as dark red spheres. The area occupied by the two cavities is represented by pale grey surfaces. The following residue positions were selected as possible sites of optimization by rational mutagenesis: in CL, Leu160 to Phe or Met, Thr178 to Val, Leu or Ile, deletion of Pro119, Ser121 to Leu and Gln124 to Leu, and in CH1, Tyr122 to Phe or Met. This figure was generated by PyMOL.36
subfamily. In the CL domain, two combination mutations were carried out; Asn161Gly and Ser177Phe (M29) and Val115Leu, Asn161Gly and Ser177Phe (M30). Replacement of Asn161 by Gly was performed in the belief that more space would be available to the bulky phenylalanine in position 177 and that this could increase the hydrophobic nature of the residue. All these mutants were
Figure 4. (a) Superposition of the CH1 domain of Fab77 on a member of the published11 most stable human variable VH3 framework (PDB entry 1DHU, modelled structure). (b) Superposition of Fab77 CL on a member of the most stable human variable Vκ3 framework (PDB entry 1DH5, modelled structure). Cα atoms of the conserved cysteine residues (Cys140 and Cys195 in CH1; Cys134 and Cys194 in CL) and tryptophan residues (Trp154 in CH1 and Trp148 in CL) were used in the least-squares fit. Only the residues of Fab77 are labelled. The side-chain arrangements of the residues located in the hydrophobic upper core of the human VH/VL domain, proposed to be most critical for the conformational stability of human frameworks,11 are shown in grey, superimposed Fab77 CH1 residues are in green and CL residues are in blue, residues selected for mutagenesis are shown in red. This figure was generated by PyMOL.36
Stability Engineering of the Mouse Fab CH1 and CL
693
Figure 5. Expression levels of Fab77 and stability engineering mutants in the RV308 strain. The expression levels of the functional, testosterone-binding, Fab fragments were determined by ELISA from periplasmic and culture supernatant samples. The expression levels in culture supernatant samples (white columns) and in periplasmic samples (hatched columns) were determined from a standard curve of purified Fab77 (concentration based on absorbance at 280 nm). The mean values of four replicate expression level measurements are shown.
between the first two β-strands of the CH1 domain was targeted to mutations, since the B values derived from the three-dimensional structure analysis by X-ray crystallography indicated a high degree of mobility in this loop region of the Fab77 structure. The second strategy for site-directed mutagenesis was based on the fact that the packing of the core regions and sizes of the cavities are critical for the correct conformation and overall stability of folded proteins. The third approach was to compare the structures of CH1 and CL with the most stable human VH and VL frameworks to identify residues that adversely affect the conformational stability of Fab77. The mobile loop in CH1 (residues 128–133) was chosen for mutagenesis due to its structural fluctuation in the B value analysis of the Fab77 structure, indicating that this loop would be the first region to lose its structure during denaturation. Obviously, the fluctuation of the loop structures to a certain degree is allowed or is necessary for the correct folding of the Fab fragment and a too rigid structure hinders the correct arrangement of the β-sheets during the folding process in vivo. The loop located between the two first β-strands in the CH1 of Fab77 is surface-exposed and, in principle, residues hydrophilic by nature would enhance the stability of the loop structure. The stabilizing effect of the strengthened conformational rigidity of the proteins has been realized with proteins other than antibodies.27 However, a rigid protein backbone, especially in the antibody domains, leads to decreased conformational flexibility, a property that may be deleterious with respect to their biological functions. Sequence comparison of CH1 domains among mouse immunoglobulin subclasses, as is shown in Figure 2, indicates that CH1 domains of IgA, IgG1, IgG3, IgG2a IgG2b, IgG2c, IgD and IgE have a quite high sequence variation in the loop comprising residues from 126 to 132, whereas the preceding residues Pro 123 and Leu 124 are highly conserved, as are the residues in the flanking β-strands. Furthermore, the
murine and human IgG1 and Ig2A are the only Ig subclasses not containing a cysteine residue, which participates in the interdomain disulphide bonding between CH1 and CL. Instead, the cysteine residue is located in the C terminus of CH1 in these subclasses. As reported earlier, the overall structure of murine CH1 domains is quite similar but, in addition to segment 127–135 in subtype IgG2b, there is another segment, 163–168, that deviates in structure, and in isotype Ig2A there are several differences compared to other Ig CH1 structures.28 A considerable increase in the thermal stability of the antibody CH3 domain, which has an immunoglobulin β-sheet structure as well, was achieved by residue frequency analysis of CH3 sequences from different species.29 Here, the thermal stability of dimeric CH3 was increased to 86 °C by three mutations, mainly by increasing sidechain packing between opposite dimer sub-units and by a replacement of glycine in the tight turn of a β-strand. The mutations in the mobile loop region of CH1 of Fab77 did not have beneficial properties: Mutations M1–M3, in which the whole loop region was deleted, were expressed in very small amounts (Figure 5). M6 Fab was expressed at the same level as Fab77 but showed decreased thermal stability when compared to that of Fab77 (Figure 6). Fab77 contains three cavities with volumes greater or equal to those sizes of other proteins of the same size,30 one being located in the variable Vκ domain and two of them in the interface of CH1 and CL domains. Some of the mutations targeted to the two largest hydrophobic cavities of the constant domain resulted in an increase in both the thermal stability and the expression level of Fab77. The improved stability could be explained by the fact that more hydrophobic residues in the mutated sites would protrude into the cavity and thus minimise the cavity volume, leading to a more tightly packed hydrophobic core that consequently restricts the movement of the residues in the protein core. These mutations in CL involved the replacement of Thr178 with hydrophobic valine (M12) or isoleucine (M14)
694
Figure 6. GdnHCl denaturation curves of the CH1 loop mutation M6, and the cavity-filling mutations M12, M14, M18, M19, M23 and Fab77. The unfolding transitions of Fab fragments were measured by following the shift in fluorescence emission maximum as a function of the concentration of GdnHCl at the excitation wavelength of 280 nm. The emission was measured from 310 nm to 440 nm, and background fluorescence was subtracted. All final spectra were the average of six scans.
and the combination of Leu160Met and Thr178Val mutations in M19 and in the CH1 replacement of Tyr122 with the more hydrophobic phenylalanine in combination with the CL Thr178Val mutation (M23). Tyr122 in CH1 is assumed to point into the solvent phase and its replacement by phenylalanine at this position may change the orientation of this residue to protrude more towards the cavity. Both Leu160 and Thr178 in CL participate in hydrogen bonding to water molecules in the CL/CH1 interface cavity 1 and, likewise, in the CL domain Ser121 and Gln124 in cavity 2 (Figure 3). It was presumed that by replacing these residues the formation of hydrogen bonds between water molecules would be hindered and so, consequently, would the accommodation of water molecules in these two cavities. The thermal stability of M12 (Thr178Val) and M14 (Thr178Ile) and the double mutant M19 (Leu160Met and Thr178Val) was increased when compared to that of Fab77. It is noteworthy that the double mutant in the CL/CH1 interface Leu160 to Phe and Thr178 to Val in CL (M18) showed a lower level of stability than the M19 mutant (Thr178Val and Leu160Met) and the respective stability-increasing single mutations (M11 and M12), which is presumably caused by the over-packing of the cavity (data not shown for the Leu160Phe mutant). The important stabilizing role of hydrophobic residues in close proximity to cavities has been shown recently for proteins other than antibodies, the change of the hydrophobic residues to more hydrophilic residues resulted in considerable destabilization of the proteins.31,32 In an another study, where small empty cavities of a small globular protein were filled with hydrophobic residues, it was suggested that the packing effects contribute negatively to the protein stability and weaken the benefits of increased hydrophobicity. These cavities were somewhat smaller, 20 Å3, than the two largest cavities in the constant domains of the Fab77 fragment, which have a size of 100 Å3. It
Stability Engineering of the Mouse Fab CH1 and CL
was concluded that even smaller structural rearrangements needed to accommodate larger residues can destabilize proteins.33 On the basis of the fact that both antibody variable and constant domains are structurally quite homologous, even though variable domains contain two additional β-sheets generally denoted as c′ and c″, alignments with the most stable human framework sequences were accomplished in order to reveal if the same guidelines hold for the constant domains and the variable domains. Particular attention was paid to the hydrophobic interface between the CH1 and CL domains, and the hydrophobic cores of constant domains (Figure 4(a) and (b)). Five mutations were carried out, which all destabilized the Fab77 structure significantly and led to decreased expression levels. A possible explanation for this is that, despite their overall structural similarity, they may lack common features in detailed substructures, such as the orientation of certain bond angles, local hydrophobicity and the effect of van der Waals' forces. Fab77 unfolded at a temperature of 75–76 °C, when the thermal unfolding was monitored by circular dichroism spectroscopy in the far-UV region (Figure 7). The thermal denaturation behaviour of Fab77 was compared to that of the published mouse immunoglobulin isotype 2b and its Fab counterpart that had been studied in similar denaturing conditions and protein concentrations. The Fab of isotype 2b derived by papain digestion denatured at ∼61 °C and was proposed to unfold as independent domains, according to circular dichroism and differential scanning calorimetry results.34,35 This example indicates that variation in the thermal stability of Fab fragments may be quite high. Another reason may be that Fab fragments produced by enzymatic digestion are more vulnerable to thermal denaturation than bacterially expressed Fab fragments. More recently, it was shown that a variable domain, which, as such, was observed to be quite unstable, could become more stable and more applicable in the Fab format. This was reasoned by the kinetic stabilization and mutual interaction of the constant
Figure 7. CD temperature scans of Fab77 (open squares) and the mutant M12 (filled diamonds) (150 μg/ml) in 10 mM sodium phosphate (pH 8.1) at 206.5 nm.
695
Stability Engineering of the Mouse Fab CH1 and CL
CH1/CL domain and by the cooperation between the constant CH1/CL and variable VH/VL units.16 In summary, increased thermodynamic stability of the anti-testosterone Fab77 fragment with retained binding affinity was achieved by mutating selected residues that line the two hydrophobic cavities in the interface of the CH1 and CL domains to more hydrophobic residues. Stability engineering focusing only on the constant domains probably increases the overall Fab fragment stability to only a certain degree, because it depends also on the intrinsic stability of the variable domains. We are currently evaluating the performance of the stability engineered constant domains of Fab77 in combination with variable domains with different antigen-binding specificities. A universal Fab fragment constant domain structure with high conformational stability to be used as a stable fusion partner with any variable domain would be greatle advantageous, especially in applications where antibody fragments are exposed to harsh conditions, such as elevated temperatures, non-aqueous solutions or temporary dryness.
Materials and Methods Plasmid constructions The expression vector for the Fab77 wild-type and the stability mutants was a derivative of pKKtac,21 which contains NheI and AscI sites for cloning the VLCL fragment, an intergenic region, and SfiI and NotI sites for cloning the VHCH1 fragment. Mutated codons introduced into primer sequences and PCR reactions were carried out with Vent (NEB) or Dynazyme Ext (Finnzymes, Finland) polymerases using the expression vector of Fab77 as the template. The full-length VLCL or VHCH1 fragments were obtained by PCR assembly reactions. The PCR products were isolated from 1% (w/v) agarose in 40 mM Tris, 20 mM NaAc, 1 mM EDTA (pH 8.0) and purified by the QIAquick gel-extraction kit (Qiagen). The PCR fragments were ligated into the pKKtac expression vector and transformed into E. coli XL-Blue strain (Stratagene). The sequences of the mutated Fab expression units were verified by an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems) according to the manufacturer's protocol (ABI PRISM BigDye Terminator Cycle Sequencing Kit v.1.1, Applied Biosystems). Expression and purification of the Fab fragments The Fab fragments were expressed in the E. coli RV308 strain (ATCC 31608). The pKKtac expression constructs of the mutant Fab fragments were transformed into the RV308 strain. LB medium containing 100 μg/ml of ampicillin and 1% (w/v) glucose was inoculated with a single bacterial colony and incubated overnight at 37 °C. Either 100 ml (small-scale) or 1 l (for Fab fragment purification) of LB medium with 100 μg/ml of ampicillin was inoculated (1:50 dilution) with the overnight culture and incubated at 37 °C. When the cell density reached an absorbance at 600 nm (A600) of 1.0, the expression of Fab fragments was induced by the addition of isopropyl-β-Dthiogalactopyranoside (IPTG from Sigma) to a final
concentration of 1 mM. Cells were grown for 18 h at 30 °C and were harvested by centrifugation (3000g, 15 min at 4 °C). The culture supernatant was used for ELISA analysis and for Fab fragment purification. Periplasmic fractions were isolated from overnight bacterial cultures by sequential incubations (repeated three times) of the harvested cells (from 1 ml) in 0.5 ml of PBS in a solid CO2/ ethanol bath for 5 min followed by thawing at 37 °C. The cell debris was pelleted by centrifugation (20,000g, 5 min at 4 °C) and the clear supernatant containing the periplasmic proteins was used for ELISA. Fab fragments were purified batch-wise by passage through a protein G Sepharose Fast Flow (Pharmacia Biotech) affinity chromatography column. One litre of the supernatant was incubated at 4 °C with gentle shaking overnight with a 5 ml bed-volume of protein G gel and loaded onto the chromatography column. After washing with 20 mM sodium phosphate (pH 7.0), elution was performed with 0.1 M glycine–HCl (pH 2.7) and the eluted fractions were neutralized with 1 M Tris–HCl (pH 9.0). The concentration of Fab was calculated from the A280 of purified samples using an extinction coefficient of 1.558 (1.0 g/l solution). The purity of the protein was confirmed by SDS-PAGE stained with Coomassie brilliant blue. The presence of the interchain disulphide bond between the light chain and the heavy chain of the expressed Fab fragments was confirmed by SDS-PAGE under nonreducing conditions (data not shown). Enzyme-linked immunosorbent assay (ELISA) A testosterone-3-CMO-bovine serum albumin conjugate (Sigma) was coated at a concentration of 2 μg/ml in 0.1 M NaHCO3 (pH 9.6) on 96-well plates overnight at 4 °C. Plates were blocked with 1% (w/v) bovine serum albumin (Sigma) in PBS for 2 h at room temperature. After incubation of the supernatant samples (1:100 (v/v) or 1:200 (v/v) dilutions) or periplasmic fractions (1:50 (v/v) and 1:100 (v/v) dilutions) for 1 h, the Fab fragments were detected by a goat anti-mouse IgG Fab specific alkaline phosphatase conjugate (Sigma). Equilibrium denaturation experiment Fluorescence spectra were collected on a Cary Eclipse spectrofluorimeter (Varian). The excitation wavelength was 280 nm and the slit-width was 5 nm for excitation and for emission. The fluorescence of samples, 0.5 μM protein and concentrations of GdnHCl from 0 to 4.0 M in a 20 mM Hepes buffer (pH 7.0) incubated overnight at 4 °C, was recorded between 310 nm and 440 nm. Blank subtractions were made in all spectra, which were averaged from six scans. The final concentration of the GdnHCl stock solution (8 M) was determined from its refractive index. The fluorescence emission maximum as a function of the concentration of denaturant was calculated using a Gaussian fit from the fluorescence emission data. Circular dichroism measurements Temperature-induced unfolding was monitored by a JASCO J-715 CD spectropolarimeter, equipped with a JASCO PTC-348WI thermocouple (JASCO International). A quartz cuvette of 0.1 cm path-length sealed with a cap to avoid evaporation was used. The temperature was increased from 35 °C to 90 °C at a rate of 1 °C/min (resolution of 0.2 °C and a time constant of 16 s). Fab77
696 and M12 Fab fragments were measured in 10 mM sodium phosphate (pH 8.1) at a final concentration of protein of 150 μg/ml. The wavelength during the temperature scan was 206.5 nm.
Acknowledgements Financial support from the National Technology Agency (Finland) is gratefully acknowledged. Berg is thanked for excellent technical assistance and Dr Timo Pulli for helpful discussions.
References 1. Borrebaeck, C. A. & Carlsson, R. (2001). Human therapeutic antibodies. Curr. Opin. Pharmacol. 1, 404–408. 2. Kontermann, R. E. (2004). Intrabodies as therapeutic agents. Methods, 34, 163–170. 3. Pavlickova, P., Schneider, E. M. & Hug, H. (2004). Advances in recombinant antibody microarrays. Clin. Chim. Acta, 343, 17–35. 4. Yau, K. Y. F., Lee, H. & Hall, J. C. (2003). Emerging trends in the synthesis and improvement of haptenspecific recombinant antibodies. Biotechnol. Advan. 21, 599–637. 5. Roque, A. C., Lowe, C. R. & Taipa, M. A. (2004). Antibodies and genetically engineered related molecules: production and purification. Biotechnol. Prog. 20, 639–654. 6. Holliger, P. & Hudson, P. J. (2005). Engineered antibody fragments and the rise of single domains. Nature Biotechnol. 23, 1126–1136. 7. Adachi, M., Kurihara, Y., Nojima, H., Takeda-Shitaka, M., Kamiya, K. & Umeyama, H. (2003). Interaction between the antigen and antibody is controlled by the constant domains: normal mode dynamics of the HEL-HyHEL-10 complex. Protein Sci. 12, 2125–2131. 8. Arndt, K. M., Muller, K. M. & Pluckthun, A. (2001). Helix-stabilized Fv (hsFv) antibody fragments: substituting the constant domains of a Fab fragment for a heterodimeric coiled-coil domain. J. Mol. Biol. 312, 221–228. 9. Petrov, K., Dion, M., Hoffmann, L., Dintinger, T., Defontaine, A. & Tellier, C. (2004). Bivalent Fv antibody fragments obtained by substituting the constant domains of a Fab fragment with heterotetrameric molybdopterin synthase. J. Mol. Biol. 341, 1039–1048. 10. Ewert, S., Honegger, A. & Pluckthun, A. (2004). Stability improvement of antibodies for extracellular and intracellular applications: CDR grafting to stable frameworks and structure-based framework engineering. Methods, 34, 184–199. 11. Ewert, S., Huber, T., Honegger, A. & Pluckthun, A. (2003). Biophysical properties of human antibody variable domains. J. Mol. Biol. 325, 531–553. 12. Worn, A. & Pluckthun, A. (2001). Stability engineering of antibody single-chain Fv fragments. J. Mol. Biol. 305, 989–1010. 13. Tan, P. H., Sandmaier, B. M. & Stayton, P. S. (1998). Contributions of a highly conserved VH/VL hydrogen bonding interaction to scFv folding stability and refolding efficiency. Biophys. J. 75, 1473–1482.
Stability Engineering of the Mouse Fab CH1 and CL
14. Ewert, S., Cambillau, C., Conrath, K. & Pluckthun, A. (2002). Biophysical properties of camelid V(HH) domains compared to those of human V(H)3 domains. Biochemistry, 41, 3628–3636. 15. Dumoulin, M., Conrath, K., Van Meirhaeghe, A., Meersman, F., Heremans, K., Frenken, L. G. et al. (2002). Single-domain antibody fragments with high conformational stability. Protein Sci. 11, 500–515. 16. Rothlisberger, D., Honegger, A. & Pluckthun, A. (2005). Domain interactions in the Fab fragment: a comparative evaluation of the single-chain Fv and Fab format engineered with variable domains of different stability. J. Mol. Biol. 347, 773–789. 17. Jung, S., Honegger, A. & Pluckthun, A. (1999). Selection for improved protein stability by phage display. J. Mol. Biol. 294, 163–180. 18. Ewert, S., Honegger, A. & Pluckthun, A. (2003). Structure-based improvement of the biophysical properties of immunoglobulin VH domains with a generalizable approach. Biochemistry, 42, 1517–1528. 19. Proba, K., Worn, A., Honegger, A. & Pluckthun, A. (1998). Antibody scFv fragments without disulfide bonds made by molecular evolution. J. Mol. Biol. 275, 245–253. 20. Worn, A. & Pluckthun, A. (1998). An intrinsically stable antibody scFv fragment can tolerate the loss of both disulfide bonds and fold correctly. FEBS Letters, 427, 357–361. 21. Hemminki, A., Niemi, S., Hautoniemi, L., Soderlund, H. & Takkinen, K. (1998). Fine tuning of an antitestosterone antibody binding site by stepwise optimisation of the CDRs. Immunotechnology, 4, 59–69. 22. Hemminki, A., Niemi, S., Hoffren, A. M., Hakalahti, L., Soderlund, H. & Takkinen, K. (1998). Specificity improvement of a recombinant anti-testosterone Fab fragment by CDRIII mutagenesis and phage display selection. Protein Eng. 11, 311–319. 23. Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T. N., Weissig, H. et al. (2000). The Protein Data Bank. Nucl. Acids Res. 28, 235–242. 24. Valjakka, J., Hemminki, A., Niemi, S., Soderlund, H., Takkinen, K. & Rouvinen, J. (2002). Crystal structure of an in vitro affinity- and specificity-matured antitestosterone Fab in complex with testosterone. Improved affinity results from small structural changes within the variable domains. J. Biol. Chem. 277, 44021–44027. 25. Derrick, J. P. & Wigley, D. B. (1994). The third IgGbinding domain from streptococcal protein G: an analysis by X-ray crystallography of the structure alone and in a complex with Fab. J. Mol. Biol. 243, 906–918. 26. Kabat, E. A., Wu, T. T., Reid-Miller, M., Perry, H. M. & Gottesman, K. S. (1991). Sequence of Proteins of Immunological Interest, 5th edit, NIH Publication no. 91-3242, National Institute of Health, Bethesda, MD. 27. Schimmele, B. & Plückthun, A. (2005). Engineering proteins for stability and efficient folding. In Protein Folding Handbook (Buchner, J. & Kiefhaber, T., eds), vol. 5, pp. 1281–1333, Wiley Verlag GmbH & Co. KGaA., Weinheim. 28. Sheriff, S., Jeffrey, P. D. & Bajorath, J. (1996). Comparison of CH1 domains in different classes of murine antibodies. J. Mol. Biol. 263, 385–389. 29. Demarest, S. J., Rogers, J. & Hansen, G. (2004). Optimization of the antibody CH3 domain by residue frequency analysis of IgG sequences. J. Mol. Biol. 335, 41–48. 30. Cuff, A. L. & Martin, A. C. (2004). Analysis of void volumes in proteins and application to stability of the
Stability Engineering of the Mouse Fab CH1 and CL
31.
32.
33.
34.
p53 tumour suppressor protein. J. Mol. Biol. 344, 1199–1209. Kadonosono, T., Chatani, E., Hayashi, R., Moriyama, H. & Ueki, T. (2003). Minimization of cavity size ensures protein stability and folding: structures of Phe46-replaced bovine pancreatic RNase A. Biochemistry, 42, 10651–10658. Adamek, D. H., Guerrero, L., Blaber, M. & Caspar, D. L. D. (2005). Structural and energetic consequences of mutations in a solvated hydrophobic cavity. J. Mol. Biol. 346, 307–318. Bueno, M., Cremades, N., Neira, J. L. & Sancho, J. (2006). Filling small, empty protein cavities: structural and energetic consequences. J. Mol. Biol. 358, 701–712. Vermeer, A. W. P., Norde, W. & van Amerongen, A. (2000). The unfolding/denaturation of immunogammaglobulin of isotype 2b and its Fab and Fc fragments. Biophys. J. 79, 2150–2154.
697 35. Vermeer, A. W. & Norde, W. (2000). The thermal stability of immunoglobulin: unfolding and aggregation of a multi-domain protein. Biophys. J. 78, 394–404. 36. DeLano, W. (2002). The PyMOL User's Manual. DeLano Scientific, San Carlos, CA. 37. Lefranc, M.-P. (2004). IMGT-ontology and IMGT databases, tools and web resources for immunogenetics and immunoinformatics. Mol. Immunol. 40, 647–659. 38. Poirot, O., Suhre, K., Abergel, C., O'Toole, E. & Notredame, C. (2004). 3DCoffee@igs: a web server for combining sequences and structures into a multiple sequence alignment. Nucl. Acids Res. 32, W37–W40. 39. Gouet, P., Robert, X. & Courcelle, E. (2003). ESPript/ ENDscript: extracting and rendering sequence and 3D information from atomic structures of proteins. Nucl. Acids Res. 31, 3320–3323. 40. Binkowski, T. A., Naghibzadeh, S. & Liang, J. (2003). CASTp: computed atlas of surface topography of proteins. Nucl. Acids Res. 31, 3352–3355.
Edited by R. Huber (Received 26 January 2006; received in revised form 19 June 2006; accepted 21 June 2006) Available online 28 July 2006