- Email: [email protected]
Crystal Structure of HEL4, a Soluble, Refoldable Human VH Single Domain with a Germ-line Scaffold
doi:10.1016/j.jmb.2004.02.013
J. Mol. Biol. (2004) 337, 893–903
Crystal Structure of HEL4, a Soluble, Refoldable Human VH Single Domain with a Germ-line Scaffold Laurent Jespers1, Oliver Schon1, Leo C. James1, Dmitri Veprintsev2 and Greg Winter1* 1 Laboratory of Molecular Biology, Medical Research Council Centre, Hills Road Cambridge CB2 2QH, UK 2
Centre for Protein Engineering Medical Research Council Centre, Hills Road, Cambridge CB2 2QH, UK
The antigen binding site of antibodies usually comprises associated heavy (VH) and light (VL) chain variable domains, but in camels and llamas, the binding site frequently comprises the heavy chain variable domain only (referred to as VHH). In contrast to reported human VH domains, VHH domains are well expressed from bacteria and yeast, are readily purified in soluble form and refold reversibly after heat-denaturation. These desirable properties have been attributed to highly conserved substitutions of the hydrophobic residues of VH domains, which normally interact with complementary VL domains. Here, we describe the discovery and characterisation of an isolated human VH domain (HEL4) with properties similar to those of VHH domains. HEL4 is highly soluble at concentrations of $ 3 mM, essentially monomeric and resistant to aggregation upon thermodenaturation at concentrations as high as 56 mM. However, in contrast to VHH domains, the hydrophobic framework residues of the VH:VL interface are maintained and the only sequence changes from the corresponding human germ-line segment (V3-23/DP-47) are located in the loops comprising the complementarity determining regions (CDRs). The crystallographic structure of HEL4 reveals an unusual feature; the sidechain of a framework residue (Trp47) is flipped into a cavity formed by Gly35 of CDR1, thereby increasing the hydrophilicity of the VH:VL interface. To evaluate the specific contribution of Gly35 to domain properties, Gly35 was introduced into a VH domain with poor solution properties. This greatly enhanced the recovery of the mutant from a gel filtration matrix, but had little effect on its ability to refold reversibly after heat denaturation. Our results confirm the importance of a hydrophilic VH:VL interface for purification of isolated VH domains, and constitute a step towards the design of isolated human VH domains with practical properties for immunotherapy. q 2004 Elsevier Ltd. All rights reserved.
*Corresponding author
Keywords: antibody; single domain; human VH; folding; crystal structure
Introduction Present address: L. Jespers & O. Schon, Domantis Limited, Granta Park, Abington, Cambridge CB1 6GS, UK. Abbreviations used: CDR, complementarity determining regions; dAb, domain antibody; HEL, hen egg lysozyme; Kd , dissociation constant; Mr , molecular mass; PBS, phosphate buffer saline (50 mM phosphate, 150 mM NaCl, pH 7.4); rmsd, root mean square deviation; TMB, 3,30 ,5,50 -tetramethylbenzidine; VH, variable domain of classical heavy chain; VL, variable domain of classical light chain; VHH, variable domain of “heavy chain” antibody from camels and llamas. E-mail address of the corresponding author: [email protected]
The antigen binding sites of human and mouse antibodies comprise six hypervariable loops (complementarity determining regions or CDRs) mounted on the b-sheet scaffolds of paired heavy (VH) and light (VL) chain variable domains. Both domains usually contribute to antigen binding, but in some cases single VH domains isolated from natural antibodies may have comparable affinities and good specificities.1 However such isolated VH single domains (termed VH domain antibodies or VH dAbs) can be difficult to express in Escherichia coli and prone to aggregate upon concentration or
0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
894 prolonged standing at 4 8C.1 – 5 In addition several human VH and VL dAbs show a tendency to bind to the agarose – dextran matrix during sizeexclusion chromatography, thereby resulting in poor recovery and/or aberrant elution profiles under native buffer conditions.5,6 These properties, often referred to as “stickiness”, are thought to derive from exposure of the side-chains forming the hydrophobic interface onto which the VL domain would normally pack.1 The most plausible model for retardation and/or poor recovery from columns involves transient or permanent binding of folded dAbs on the agarose– dextran matrix, through the exposed surface. However in some cases irreversible deposition on the gel filtration matrix could be mediated via the (partially) unfolded state.6 By contrast, camels and llamas produce a class of “heavy chain” antibodies, thought to have lost their light chains and CH1 domains during evolution,7 from which the single variable domains (termed VHHs) are well expressed and form highly soluble monomeric species in solution.8,9 These highly desirable properties were attributed to a series of germ-line substitutions (Leu11 ! Ser, Val37 ! Tyr, Gly44 ! Glu, Leu45 ! Arg, and Trp47 ! Gly/Ile, Trp103 ! Arg/Gly) in the b-sheet scaffold of VHH dAbs, which increase the hydrophilicity of the VHH surfaces that would normally contact the VL and the CH1 domains in conventional antibodies.10 Some features of these VHH sequences of camels have been used to help improve the solution properties of human VH dAbs (a process referred to as “camelisation”).2 By inserting VHH amino acids in the framework 2 (at positions 44, 45 and 47) of a human VH dAb (aOX-13), two dAbs with improved behaviour in solution were generated,2 thereby enabling NMR studies.11 However, camelising VHH substitutions were shown to reduce the expression levels of dAbs,2,3 proved to be thermodynamically destabilising3,12 due to framework deformations,11 and did not completely alleviate aggregation in solution.4,13,14 In another approach, which exploits the natural diversity of the immunoglobulin repertoire, computer databases were screened for candidate VH dAbs bearing hydrophilic mutations at the former VL interface.15 A murine VH dAb bearing an unusual substitution in framework 2 (Gly44 ! Lys) was found that is monomeric, highly soluble ($ 2 mM) and does not stick to the gel filtration matrix. Further work has indicated that other sets of solvent-exposed residues may improve the solution properties of dAbs. For example, several dAbs derived from the heavy chains of conventional llama antibodies expressed well in bacteria, and did not aggregate in solution or stick to the matrix during size-exclusion chromatography.16 Their properties were attributed to a set of five framework substitutions (at positions 6, 74, 83, 84, and 108) unique to llama VH dAbs. Moreover, the CDRs of dAbs from
Structure of a Human Single Domain
camels and llamas can also contribute to improved solution properties. Indeed, the CDR3 loop of VHH domains is on average much longer (17 residues) than the human (12 residues) or mouse (nine residues) CDR3 loops,10,17 and often folds back on the region corresponding to the hydrophobic interface of conventional VH dAbs (as shown by several crystal structures).18 – 21 Subsequently, the expression and the solution properties of camelised and human VH dAbs were improved by extending the length of the CDR3 loops.5,6,22 A further property of the VHH domains is that they often undergo reversible unfolding, thus regaining native state and antigen-binding specificity,5,23 even at concentrations as high as 100 mM24 and after prolonged incubation at temperatures in the range 80 8C to 92 8C.23,25 Several of the llama VH dAbs had similar properties.16,26 By contrast, antibodies of other species and their fragments, including VH dAbs, usually aggregate irreversibly upon heating at low concentrations.25,27 – 30 The small size of camel and llama dAbs, their high solubility, resistance to aggregation in solution, lack of stickiness and tendency to unfold reversibly upon thermal denaturation are highly desirable properties for biotechnological application. However, for therapy it is also desirable that the dAbs are of human origin. Although various attempts have been made to endow human VH dAbs with all the biophysical properties of VHH dAbs (see above), success has been limited. In the course of characterising the properties of human VH dAbs we discovered a domain without camelising mutations but with several properties similar to camel and llama VHH dAbs.
Results Biophysical characterisation of the HEL4 dAb A phage library of human VH dAbs (1.1 £ 1010 clones) (see Materials and Methods) was selected for binding to HEL. After three rounds of selection, individual dAbs ðn ¼ 80Þ were expressed as soluble fragments in bacteria and screened by ELISA. A total of 21 HEL-specific clones were sequenced, revealing six unique sequences of which one (HEL4) was chosen for further study. The HEL4 dAb comprised an 11 amino acid residue CDR3 loop, and 11 amino acid mutations in the VH3 V3-23/DP-47 germ-line segment, which were located within the loops comprising the CDRs consistent with the library design (Figure 1). The HEL4 dAb was purified by affinity chromatography on protein A-Sepharose and then by size-exclusion chromatography on Superdex G75, eluting essentially in a single peak at the expected size for a 13 kDa monomer (Figure 2(a)). This behaviour contrasted with that of other human dAbs studied,5,6 which eluted in small peaks with most of the material aggregating to the chromatographic support. Furthermore, the HEL4 dAb
Structure of a Human Single Domain
895
Figure 1. Sequences of VH domain antibodies presented in this study. Residue numbering according to Kabat et al.38 Residues that differ between HEL4 and DP47d dAb, are underlined on the HEL4 line; the sites of the DP47d single mutants Ser35 ! Gly and Trp47 ! Arg are marked in bold on the DP47d line.
Figure 2. Biophysical properties of the HEL4 dAb. (a) Analytical gel-filtration chromatogram of the HEL4 dAb (10 mM in PBS) applied on a Superdex G75 column. (b) Thermally induced denaturation curves of dAb HEL4 (5 mM in PBS) recorded by CD at 235 nm: (X) mean residual ellipticity upon first heating, (S) mean residual ellipticity upon second heating. Inset: CD spectra of dAb HEL4 (5 mM in PBS) in the far-UV region at different temperature: (W) 25 8C before unfolding; (B) 85 8C (unfolded protein); (O) 25 8C after sample cooling.
896
could be concentrated to 38 mg/ml (3.0 mM) without signs of aggregation. Analytical ultracentrifugation was performed over a range of HEL4 dAb concentrations. At 2 mM and 20 mM, the equilibrium data were consistent with a single species model, whereas at 200 mM, the data could equally fit to a monomeric model or to a monomer –dimer equilibrium with a best fit dissociation constant ðKd Þ of 5 mM (data not shown). The Kd of HEL4 dAb for HEL was determined to be 0.13 mM using the method described by Friguet et al.31 Thermal unfolding of the HEL4 dAb was monitored by circular dichroism (CD) analysis in the far-UV, using a concentration range of 5 –56 mM protein concentration. Secondary structure was lost in a cooperative unfolding transition (Tm ¼ 62:1 8C) and at 85 8C the protein was fully denatured as shown by the 200 – 250 nm spectrum (Figure 2(b), inset). After cooling to 25 8C, the spectrum could be superimposed on that prior to heating, indicating quantitative refolding. The refolded HEL4 dAb was functional as shown by binding to
Structure of a Human Single Domain
protein A- and HEL-coated wells by ELISA (data not shown). Further indications of a reversible two-state mechanism were obtained by re-heating the HEL4 dAb. An almost identical denaturation curve was observed (Figure 2(b)), with marginal loss of ellipticity (# 5%, which may have resulted from aggregation to the quartz surface and/or chemical degradation due to prolonged incubation above 80 8C). The thermodynamic stability (DGN-U) of the HEL4 dAb at 25 8C was calculated as 27.0 kJ/mol using the thermal unfolding data, and is similar to that for urea-denaturation (28.4 kJ/mol) (data not shown). Crystal structure of the HEL4 domain antibody The HEL4 dAb was crystallised in space group I4122 with two molecules in the asymmetric unit. ˚ resolution and The structure was refined to 2.0 A shown to adopt the standard fold of an immunoglobulin variable domain with nine antiparallel b-strands (Figure 3). The root mean
Figure 3. Stereo views of the crystal structure of HEL4. (Top) Superimposed main-chain conformations of the CDR3comprising loop (residues 94 –103) of HEL4 dAb crystallised in dimeric form (dAbs HEL4-A and HEL-B in pink and purple, respectively). Side-chains at positions 96 (Leu96) and 98 (Pro98) in each CDR3-comprising loop are shown to highlight their positions. Residue numbering according to Kabat et al.38 (Bottom) Main-chain conformation of HEL4 dAb with b-strand symbolised as arrows, and coloured CDR-comprising loops: H1 (yellow), H2 (red), and H3 (blue).
897
Structure of a Human Single Domain
square deviation (rmsd) calculated with framework Ca atoms of the 1FGV VH domain (used for ˚ with most of molecular replacement) was 0.53 A the important divergences around residues 40 –43. In the asymmetric unit, two HEL4 dAbs form a homo-dimer (HEL4-A:HE4-B) but are not exactly symmetry-related, with the main differences located in CDR3 (Figure 3). Thus, the CDR3 of the HEL4-A dAb interacts with the HEL4-B dAb (through Leu96 and Leu100c) but not vice versa. Furthermore, Pro98 of CDR3 is either buried (in the HEL4-A dAb) or exposed (in the HEL4-B dAb); this may play a role in “freezing” the CDR3 in two different conformations. The main-chain conformation of the H1 loop (residue Gly26 to Gly35) is characterised by a sharp turn at position 26 and a hydrophobic cluster at positions 24, 27, 29, 34, and fits the type-1 ˚ rmsd with 1FGV-H1 canonical structure32 (0.86 A loop) (Figure 4(a)). The main-chain conformation of the H2 loop (residues 52– 56) is characterised by a six-residue hairpin and Gly54 and Arg71 as key structural elements, and fits the type-3 canonical ˚ rmsd with 1FGV-H2 loop) structure loop32 (0.76 A (Figure 4(b)). By comparing each of the HEL4 framework residues with those of VH domains from conventional VH:VL antibody pairs, we observed that only one side-chain had undergone a major reorientation: Trp47 had rotated by 1808 about its Ca –Cb bond (Figure 5(a)). This unusual rotation brings the polar N1 atom of the side-chain towards solvent, and tucks the phenyl ring into a hydrophobic cavity formed by residues Gly35, Val37 and Ala93. Consequently, the face of the HEL4 dAb that
would normally contact a VL domain in conventional antibodies, becomes significantly more hydrophilic than that of a conventional VH3 domain (Figure 5(b) and (c)), thereby mimicking that of VHH dAbs (Figure 5(d)).21 Of the three amino acid residues forming the hydrophobic cavity into which Trp47 locks, only Gly35 (which is in CDR1) differs in the germline 3-23 segment (where it is serine) and it is the lack of a sidechain at position 35 in the HEL4 dAb, that helps to create the hydrophobic cavity. Biophysical analysis of mutants To investigate the role of the Trp47, two mutants, Ser35 ! Gly and Trp47 ! Arg, were made on the human DP47d dAb from which the HEL4 dAb had been derived (see Materials and Methods) (Table 1). Unlike HEL4 dAb, the DP47d dAb exhibited poor solubility, aggregation on gel filtration with poor recovery (, 5%), and irreversible denaturation upon heating (data not shown). When the Trp47 ! Arg dAb was loaded onto a Sephadex G75 column, 80% of the material was recovered from the column, of which 90% was monomeric. Similarly, with the Ser35 ! Gly dAb, 60% was recovered of which 85% was monomeric (Table 2). However, neither of these mutants survived heat denaturation at 5 mM concentration (data not shown). The mutation of Ser35 to Gly appeared to be less destabilising than that of Trp47 to Arg, as shown by measurement of thermodynamic stabilities of both mutants at 25 8C by urea denaturation (Table 2).
Figure 4. Canonical CDR-comprising loop conformations in the VH segment of the HEL4 dAb. Superimposition of the (a) type-1 H1 loop (residues 26 – 35) and (b) type-3 H2 loop (residues 50 – 58) of the HEL4 dAb (in pink) and the VH3 domain of H52, a humanised anti-CD18 single-chain Fv fragment (PDB entry 1FGV)40 (in blue). Side-chains of residues 27, 29 and 34 in loop H1, and 51, 54 and 58 in loop H2 are shown to highlight their respective positions in both molecules. Residue numbering according to Kabat et al.38
898
Structure of a Human Single Domain
Figure 5. Hydrophobic character of the (former) VL-contacting interfaces in VH and VHH domains. (a) Close-up view of the orientation of the side-chains at positions 35, 37 and 47 in HEL4 dAb (in pink) and in the humanised VH domain of H52 (PDB entry 1FGV) (in blue). Computed molecular surfaces at (b) the face of the HEL4 dAb that would be expected to contact a VL-domain in conventional antibodies, (c) the VL-contacting face of the humanised VH domain of H52, and (d) the former VL-interface of the AMD9 camel VHH domain specific for porcine pancreatic a-amylase (PDB entry 1KXQ).21 The hydrophobic potentials of each molecular surface were calculated by GRID and AESOP.47,48 Areas of marked hydrophobicity are marked in green. The side-chains of residues pivotal to VHH domains (37, 44, 45, 47 and 103) are shown to highlight their respective orientation. Residue numbering according to Kabat et al.38
Discussion Table 1. Data collection and structure refinement statistics Molecule name Space group Buffer ˚) Resolution (A Unique reflections Rmerge Redundancy Completeness (%) Average I=sI Final Rcryst Final Rfree Bond rmsd Angle rmsd (deg.) PDB code
HEL4 I4122 0.1 M Hepes (pH 7.2), 1.6 M MgSO4, 6% glycerol 2.0 19,067 0.143 (0.667) 7.0 (6.0) 99.7 (96.0) 4.3 (2.2) 0.177 (0.183) 0.256 (0.259) 0.007 1.4 1OHQ
Numbers in parentheses are for highest resolution shell ˚ ). (2.11–2.0 A
Our study of the HEL4 dAb, a single domain antibody based on a human germ-line VH3 framework, has shown that excellent solution properties and reversible unfolding on heating are not exclusive to dAbs from camels and llamas. The use of human rather than animal dAbs is clearly preferred for therapeutic use, and these properties may find particular application where harsh conditions such as transient heating, sterilisation, surface deposition and lyophilisation are part of the manufacturing process. Although the HEL4 dAb crystallises as a dimer (see below), we demonstrated by two independent techniques, gel filtration and analytical ultracentrifugation, that it is essentially monomeric at most of the concentrations used in the assays (up to 200 mM). No signs of aggregation were detected after concentrating the samples up to 3.0 mM and incubation in PBS at 37 8C for several days. Above
899
Structure of a Human Single Domain
Table 2. Biophysical properties of human VH3 dAbs dAb DP47d HEL4 Ser35 ! Gly Trp47 ! Arg
Tm (8C)e
DGN-U (kJ/mol)
% Recoverya on Superdex G75
Monomeric (M) or dimeric (D)
Yieldb (mg/ml)
61.4 62.1 61.6 54.8
34.7c 27.0d –28.4c 29.1c 24.0c
,5 .90 60 80
n.d. M M þ 15% D M þ 10% D
2.9 9.5 10.3 6.1
n.d., not determined. a Obtained by integrating the areas of peak(s) eluted from Superdex G75. b Yield of purified protein obtained from one litre of bacterial culture supernatant and normalised to 5 A600 nm : c Thermodynamic stability value at 25 8C, obtained by analysis of urea-induced denaturation. d Thermodynamic stability value at 25 8C, obtained by analysis of thermo-denaturation. e Temperature of mid-point transition upon reversible unfolding (HEL4) or upon aggregation (DP47d, Ser35 ! Gly, Trp47 ! Arg).
55 8C, the HEL4 dAb unfolds cooperatively without aggregating (at concentrations up to 56 mM and temperature up to 85 8C) and upon cooling renatures almost quantitatively. The Tm value of 62.1 8C for the HEL4 dAb compares well with those reported for camel VHH and human VH domains5,23 but its conformational stability (27.0 – 28.4 kJ/mol) is in the lower range of values reported for VHH domains and other human VH3 domains (21.1 – 52.7 kJ/mol),5,23 but still higher than those for human VH domains belonging to other families (13.7 –26.0 kJ/mol).6 The fact that other human dAbs with identical framework sequences to the HEL4 dAb have poor biophysical properties (DP47d dAb and Ewert et al.5) locates the determinants for the desirable properties of HEL4 exclusively in the loops comprising the CDRs. When the structure of HEL4 dAb was solved by X-ray crystallography we found two HEL4 molecules packing onto each other in a manner similar to that of a VL:VH pair in a conventional antibody and to the packing VL:VL domains in Bence-Jones proteins.33 The buried surface area of each HEL4 monomer upon packing ˚ 2 for HEL4-A dAb and HEL4-B ˚ 2 and 895 A (988 A dAb, respectively) compares favourably to that of ˚ 2) and VH (793 A ˚ 2) domains of the the VL (988 A 1FGV scFv. However, as the HEL4 dAb is a monomer in solution, the packing observed in the crystal cannot reflect its state other than at very high concentrations. A similar observation was reported for human thioredoxin, which crystallised as a dimer but displayed all the characteristics of a monomer in solution.34 The structures of the H1 and H2 loops in the HEL4 dAb proved to be typical of this family of human VH domains. However this was not true for the HEL4 CDR3. It has been shown that the backbone conformations of residues at the base of antibody CDR3 loops can fall into one of two classes, kinked or extended. The kinked class is more common.35 In HEL4, we see two conformations of the CDR3 loop, and they are both extended. We do not know if any of these CDR3 conformations are present in the monomeric HEL4 dAb; nor do we know therefore whether the conformation of the CDR3 contributes to its solution
properties as with camel antibodies.18 – 21 It is not clear whether the multiple CDR3 conformations reflect the lack of a constraining light chain, or are driven by the association of the two VH domains together in the crystal. A further unusual feature of the HEL4 structure was the rotation of the Trp47 side-chain. In conventional VH:VL antibody pairs, this highly conserved framework residue lies flat on the VH surface, such that its hydrophobic phenyl ring engages with the VL domain. In the HEL4 structure, this side-chain re-orients itself within the cavity formed by residues Gly35, Val37 and Ala93. The effect of this is to increase the hydrophilicity of the face of the domain usually involved in packing with the VL domain, and we suggest that this may be responsible in part for the properties of the HEL4 dAb in solution and upon gel filtration. This is supported by our observations that two mutations in the DP47d dAb, Trp47 ! Arg (where a hydrophobic residue is replaced by a hydrophilic) and Ser35 ! Gly (where a cavity is created into which Trp47 may flip), lead to a considerable increase in the recovery of the dAb from gel filtration. However the improvements came at a cost: the Ser35 ! Gly mutation reduced the thermodynamic stability of the DP47d dAb by 5.6 kJ/mol, whereas 10.7 kJ/mol was lost upon substitution of Trp47 by Arg. The Trp47 flip is unique among known VH structures and suggests that surface plasticity may contribute to modulating hydrophility. However, we believe that the Trp47 flip is only part of the story; other residues such as those of the loop comprising the CDR3, may also make contributions to its solution properties. It was proposed earlier that rotation of another framework residue, Trp103, contributes to increasing the hydrophilicity of VHH domains.18,36 A similar orientation of Trp103 is observed in the crystal structure of the HEL4 dAb but our structural analysis of several human and murine VH domains (not shown) reveals that the “VHH-orientation” of Trp103 side-chain is common, if not the rule among VH:VL pairs. Like several camel and llama dAbs, the human HEL4 dAb combines the properties of high solubility, monomeric state, lack of stickiness and reversible thermal unfolding. It seems likely that
900
avoidance of aggregation is a common aspect of each of these properties. Protein aggregation is mediated by intermolecular interactions involving either the folded state, or the unfolded state of the molecule (or indeed intermediate unfolded states). We suggest that the Trp47 flip is responsible for the ability of the native state of the HEL4 dAb to avoid aggregation. However, we do not believe that the flip contributes to the ability of the HEL4 dAb to survive thermal unfolding, as evidenced by the study of single-point mutants of the DP47d dAb (Ser35 ! Gly and Trp47 ! Arg) which were soluble but aggregated upon thermal denaturation. Those features of the HEL4 dAb conferring reversible unfolding on thermal denaturation have yet to be identified, and we are not therefore in a position to transpose all the desirable biophysical properties of HEL4 to other more biomedically relevant dAbs. However, we believe that our observations will help in understanding some of the biophysical properties of dAbs and thereby to the engineering of therapeutics based on human dAbs.
Structure of a Human Single Domain
Crystallisation, data collection and structure determination Crystal needles of HEL4 dAb were obtained with the hanging-drop method by mixing 1 ml of protein solution (at 22.3 mg/ml) and 1 ml of reservoir solution (0.1 M Hepes (pH 7.2), 1.6 M MgSO4, 6% (v/v) glycerol) (Hampton Research). Crystals were frozen at 100 K and data were collected at the European Synchroton Radiation Facility (Grenoble, France) on a CCD detector. Data were indexed and integrated using MOSFLM and scaled in SCALA in space group I422 (Table 1). Data ˚ . The strucwith at least an I=sI . 2:0 extended to 2.0 A ture was determined by molecular replacement using the VH3 domain of a humanised anti-CD18 single-chain Fv fragment H52 (PDB entry 1FGV)40 as a model in the program AMORE.41 Initial maps were calculated in REFMAC using the rigid-body model.41 All CDR loops were deleted and the model rebuilt over four rounds of refinement using REFMAC without TLS. Water molecules were added in ARP/wARP over the last two rounds.41 The final model has an Rfree of 0.256 and an Rfactor of 0.177. All residues are located in allowed regions in a Ramachandran plot. Affinity measurement
Materials and Methods Selection, production and purification of HEL4 dAb The human HEL4 dAb (Figure 1) was obtained from a phage-displayed library of human VH dAbs (1.1 £ 1010 clones) after biopanning on hen egg lysozyme (HEL). This repertoire was based on a template gene, DP47d, comprising the germ-line 3-23 segment of the human VH3 family (DP-47)37 and the human JH4b segment (Figure 1); it was built by oligonucleotide-mediated diversification of several codons in DP47d as follows (Kabat numbering38 for the amino acid positions, IUPAC-IUB code39 for the nucleotides): 27, KWT; 28, ANS; 29, NTT; 30, ANC; 31, NMT; 32, NAS; 33, DHT; 35, RSC; 50, RSC; 52, NNK; 52a, RNS; 53, VVW; 54, CGT; 94, RSW; 101, NVS; 102, THT; and NNK codons for all positions from 95 to 100a,b,…,k. For expression in E. coli HB2151 cells, the two tags, (His)6 and VSV, at the C-terminal end of the HEL4 dAb were first removed from the phagemid vector. Five liters of 2 £ TY medium containing ampicillin (100 mg/ml) and 0.1% (w/v) glucose were incubated at 37 8C until A600 nm , 0:8: Expression was induced with 1 mM IPTG at 30 8C. After four hours, the cells were harvested, and the periplasmic fraction was extracted with 150 ml of ice-cold TES buffer. After centrifugation (6000g for 15 minutes followed by 17,000g for 30 minutes), the cleared supernatant was loaded onto a 15 ml protein A-Sepharose column (Amersham Biosciences). After washing the column with 150 ml of PBS (50 mM phosphate, 150 mM NaCl, pH 7.4), 0.5 M NaCl, bound HEL4 dAb was eluted in 0.1 M glycine– HCl (pH 3.0), and neutralised with 1 M Tris – HCl (pH 7.4). The protein sample was concentrated and further purified on Superdex G75 (Amersham Biosciences) in 20 mM Tris (pH 7.4). Protein purity was estimated by visual analysis after SDS-PAGE on 12% (w/v) acrylamide Tris – glycine gel (Invitrogen). Protein concentration was measured at 280 nm, assuming an extinction coefficient of 26,150 M21 cm21.
The apparent dissociation constant ðKd Þ of HEL4 dAb from HEL was determined as described by Friguet et al.31 by mixing a solution of 78 nM His-tagged HEL4 dAb in PBST with increasing amounts of HEL (from 0 mM to 2 mM). After overnight incubation, the solutions were added to HEL-coated wells in PBS buffer containing 2% Tween-20. Bound HEL4 dAb was detected with Ni2þ-peroxidase conjugate (Perbio), and developed with TMB as substrate. The absorbances at 450 nm were corrected for background signal and processed into a Scatchard plot to calculate Kd : Fluorescence measurements The samples were prepared by mixing a 5.0 mM solution of purified dAb in PBS to increasing concentrations of urea (ranging from 0 M to 8 M) and by overnight incubation at 25 8C. The samples were transferred to a quartz cuvette (1 cm path-length), and placed into a F4500 fluorimeter (Hitachi). Fluorescence emission intensities at 320 nm (for HEL4 dAb) or at 350 nm (for DP47d dAb, and mutants Ser35 ! Gly and Trp47 ! Arg) were then recorded at 25 8C using an excitation wavelength of 280 nm, and a slit width of 2 nm for excitation and emission. Unfolding curves and thermodynamic stabilities at 25 8C were obtained by fitting the data of fluorescence intensity versus urea concentration, assuming a two-state cooperative equilibrium as described.42,43 Circular dichroism measurements CD cuvettes (1 cm path-length) were filled with a 5 mM solution of dAb in PBS and transferred to a J-720 polarimeter (Jasco). CD spectra at 25 8C and 85 8C were recorded in the far-UV (200 – 250 nm) with a 2 nm bandwidth, a one second integration time and a heating rate of 50 8C per hour. Unfolding curves from 25 8C to 85 8C were monitored at 235 nm. After unfolding, the sample was cooled down to 25 8C, a spectrum was recorded, and a new thermal-unfolding curve was recorded. For the HEL4 dAb, the unfolding curves were assumed to
901
Structure of a Human Single Domain
be two-state and fitted as described44 using a DCp contribution of 12 cal per amino acid residue.45 The values obtained for Tm (midpoint transition temperature in Kelvin), DHm (enthalpy change for unfolding at Tm ) were then used to calculate the thermodynamic stability ðDGN-U Þ of the protein at 25 8C as described.44 Alternatively, dAb unfolding upon addition of urea at 25 8C was monitored by CD measurements on a polarimeter 215 (AVIV Instruments). Protein samples (Ser35 ! Gly and Trp47 ! Arg) were prepared as described (see Fluorescence measurements). The changes of ellipticity value (D1) were plotted versus the urea concentration and the unfolding curves were fitted as described42,43 to obtain the thermodynamic stabilities at 25 8C. Analytical ultracentrifugation Sedimentation equilibrium experiments were done in a Beckman Optima XLI analytical ultracentrifuge with Ti-60 rotor using interference and absorbance at 280 nm and 230 nm, at 25 8C. The HEL4 dAb was loaded into six-sector 12 mm path length cells at three different concentrations in PBS: 2 mM, 20 mM, and 200 mM. The samples were centrifuged until they reached equilibrium as judged by the changes in the subsequent scans at speeds of 30,000 rpm and 40,000 rpm. Data were analysed using UltraSpin software†. Mutant construction and protein purification The gene encoding the synthetic VH DP47d dAb was obtained from Ian Tomlinson, and used to assemble two single-point mutants (termed Ser35 ! Gly and Trp47 ! Arg, Figure 1) by SOE-PCR.46 Next, the three genes were transferred into the expression vector and for each dAb, an overnight E. coli culture was used to inoculate five litres of 2 £ TY medium containing ampicillin (100 mg/ml). When A600 nm reached ,0.6, induction was started with 1 mM IPTG and the culture was further incubated at 30 8C for 16 hours. After centrifugation, the supernatant was filtered and incubated overnight with 30 ml of Streamline-protein A beads (Amersham Biosciences) at 4 8C. The beads were then packed into a column, washed with 300 ml of PBS, and bound dAbs were eluted in 0.1 M glycine– HCl (pH 3.0). After neutralisation with 1 M Tris – HCl (pH 7.4), the protein samples were dialysed in PBS and concentrated before storage at 4 8C. Protein purity and concentration were estimated as described above. Atomic coordinates The model and structure factors were deposited in the RCSB Protein Data Bank with accession codes 1OHQ and 1OHQSF, respectively.
Acknowledgements We are grateful to Roger Williams, Olga Perisic, Amrik Basran, Chris Johnson and Neil Ferguson for technical help, and to Ian Tomlinson and Kevin Moulder for advice. L.J. and O.S. were † http://www.mrc-cpe.cam.ac.uk
funded by Domantis Limited (Cambridge, UK) under a collaborative research programme with the Medical Research Council.
References 1. Ward, E. S., Gu¨ssow, D., Griffiths, A. D., Jones, P. T. & Winter, G. (1989). Binding activities of a repertoire of single immunoglobulin variable domains secreted from Escherichia coli. Nature, 341, 544– 546. 2. Davies, J. & Riechmann, L. (1994). “Camelising” human antibody fragments: NMR studies on VH domains. FEBS Letters, 339, 285– 290. 3. Davies, J. & Riechmann, L. (1995). Antibody VH domains as small recognition units. Biotechnology (NY), 13, 475– 479. 4. Kortt, A. A., Guthrie, R. E., Hinds, M. G., Power, B. E., Ivancic, N., Caldwell, J. B. et al. (1995). Solution properties of Escherichia coli-expressed VH domain of anti-neuraminidase antibody NC41. J. Protein Chem. 14, 167–178. 5. Ewert, S., Cambillau, C., Conrath, K. & Plu¨ckthun, A. (2002). Biophysical properties of camelid VHH domains compared to those of human VH3 domains. Biochemistry, 41, 3628–3636. 6. Ewert, S., Huber, T., Honegger, A. & Plu¨ckthun, A. (2003). Biophysical properties of human antibody variable domains. J. Mol. Biol. 325, 531– 553. 7. Hamers-Casterman, C., Atartouch, T., Muyldermans, S., Robinson, G., Hamers, C., Bajyana Songa, E. et al. (1993). Naturally occurring antibodies devoid of light chains. Nature, 363, 446–448. 8. Arbabi Ghahroudi, M., Desmyter, A., Wyns, L., Hamers, R. & Muyldermans, S. (1997). Selection and identification of single domain antibody fragments from camel heavy-chain antibodies. FEBS Letters, 414, 521– 526. 9. Lauwereys, M., Arbabi Ghahroudi, M., Desmyter, A., Kinne, A., Holzer, W., De Genst, E. et al. (1998). Potent enzyme inhibitors derived from dromedary heavy-chain antibodies. EMBO J. 17, 3512 –3520. 10. Muyldermans, S., Atarhouch, T., Saldanha, J., Barbosa, J. A. R. G. & Hamers, R. (1994). Sequence and structure of VH domain from naturally occurring camel heavy chain immunoglobulins lacking light chains. Protein Eng. 7, 1129– 1135. 11. Riechmann, L. (1996). Rearrangement of the former VL interface in the solution structure of a camelised, single antibody VH domain. J. Mol. Biol. 259, 957– 969. 12. Davies, J. & Riechmann, L. (1996). Single antibody domains as small recognition units: design and in vitro antigen selection of camelized, human VH domains with improved stability. Protein Eng. 9, 531 –537. 13. Martin, F., Volpari, C., Steinkuhler, C., Dimasi, N., Brunetti, M., Biasiol, G. et al. (1997). Affinity selection of a camelized V(H) domain antibody inhibitor of hepatitis C virus NS3 protease. Protein Eng. 10, 607 –614. 14. Voordijk, S., Hansson, T., Hilvert, D. & van Gunsteren, W. F. (2000). Molecular dynamics simulations highlight mobile regions in proteins: a novel suggestion for converting a murine VH domain into a more tractable species. J. Mol. Biol. 300, 963– 973. 15. Reiter, Y., Schuck, P., Boyd, L. F. & Plaksin, D. (1999). An antibody single-domain phage display library of a native heavy chain variable region: isolation of
902
16.
17. 18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28. 29.
30.
31.
functional single-domain VH molecules with a unique interface. J. Mol. Biol. 290, 685– 698. Tanha, J., Dubuc, G., Hirama, T., Narang, S. A. & MacKenzie, C. R. (2002). Selection by phage display of llama conventional V(H) fragments with heavy chain antibody V(H)H properties. J. Immunol. Methods, 263, 97 – 109. Wu, T. T., Johnson, G. & Kabat, E. A. (1993). Length distribution of CDRH3 in antibodies. Proteins: Struct. Funct. Genet. 16, 1 – 7. Desmyter, A., Transue, T. R., Ghahroudi, M. A., Thi, M. H., Poortmans, F., Hamers, R. et al. (1996). Crystal structure of a camel single-domain VH antibody fragment in complex with lysozyme. Nature Struct. Biol. 3, 803– 811. Spinelli, S., Frenken, L., Bourgeois, D., de Ron, L., Bos, W., Verrips, T. et al. (1996). The crystal structure of a llama heavy chain variable domain. Nature Struct. Biol. 3, 752– 757. Spinelli, S., Frenken, L. G., Hermans, P., Verrips, T., Brown, K., Tegoni, M. & Cambillau, C. (2000). Camelid heavy-chain variable domains provide efficient combining sites to haptens. Biochemistry, 39, 1217– 1222. Desmyter, A., Spinelli, S., Payan, F., Lauwereys, M., Wyns, L., Muyldermans, S. & Cambillau, C. (2002). Three Camelid V(HH) domains in complex with porcine pancreatic a-amylase. J. Biol. Chem. 277, 23645– 23650. Tanha, J., Xu, P., Chen, Z., Ni, F., Kaplan, H., Narang, S. A. & MacKenzie, C. R. (2001). Optimal design features of camelized human single-domain antibody libraries. J. Biol. Chem. 276, 24774– 24780. Dumoulin, M., Conrath, K., Van Meirhaeghe, A., Meersman, F., Heremans, K., Frenken, L. G. J. et al. (2002). Single-domain antibody fragments with high conformational stability. Protein Sci. 11, 500– 515. Perez, J. M., Renisio, J. G., Prompers, J. J., van Platerink, C. J., Cambillau, C., Darbon, H. & Frenken, L. G. J. (2001). Thermal unfolding of a llama antibody fragment: a two-state reversible process. Biochemistry, 40, 74 – 83. van der Linden, R. H., Frenken, L. G., de Geus, B., Harmsen, M. M., Ruuls, R. C., Stok, W. et al. (1999). Comparison of physical chemical properties of llama V(HH) antibody fragments and mouse monoclonal antibodies. Biochim. Biophys. Acta, 1431, 37 – 46. Vranken, W., Tolkatchev, D., Xu, P., Tanha, J., Chen, Z., Narang, S. & Ni, F. (2002). Solution structure of a llama single-domain antibody with hydrophobic residues typical of the VH/VL interface. Biochemistry, 41, 8570– 8579. Young, N. M., MacKenzie, C. R., Narang, S. A., Oomen, R. P. & Baenziger, J. E. (1995). Thermal stabilization of a single-chain Fv antibody fragment by introduction of a disulphide bond. FEBS Letters, 377, 135– 139. Wirtz, P. & Steipe, B. (1999). Intrabody construction and expression III: engineering hyperstable V(H) domains. Protein Sci. 8, 2245– 2250. Wo¨rn, A. & Plu¨ckthun, A. (1999). Different equilibrium stability behaviour of ScFv fragments: identification, classification, and improvement by protein engineering. Biochemistry, 38, 8739– 8750. Vermeer, A. W. P. & Norde, W. (2000). The thermal stability of immunoglobulin: unfolding and aggregation of a multi-domain protein. Biophys. J. 78, 394– 404. Friguet, B., Chaffotte, A. F., Djavadi-Ohaniance, L. &
Structure of a Human Single Domain
32.
33.
34. 35. 36.
37.
38.
39.
40.
41. 42.
43. 44.
45.
46.
47.
Goldberg, M. E. (1985). Measurements of the true affinity constant in solution of antigen-antibody complexes by enzyme-linked immunosorbent assay. J. Immunol. Methods, 77, 305–319. Chothia, C., Lesk, A. M., Gherardi, E., Tomlinson, I. M., Walter, G., Marks, J. D. et al. (1992). Structural repertoire of the human V(H) segments. J. Mol. Biol. 227, 799– 817. Epp, O., Colman, P., Fehlhammer, H., Bode, W., Schiffer, M., Huber, R. & Palm, W. (1974). Crystal and molecular structure of a dimer composed of the variable portions of the Bence-Jones protein REI. Eur. J. Biochem. 45, 513– 524. Gronenborn, A. M., Clore, G. M., Louis, J. M. & Wingfield, P. T. (1999). Is human thioredoxin monomeric or dimeric? Protein Sci. 8, 426– 429. Al-Lazikani, B., Lesk, A. M. & Chothia, C. (1997). Standard conformations for the canonical structures of immunoglobulins. J. Mol. Biol. 273, 927– 948. Desmyter, A., Decanniere, K., Muyldermans, S. & Wyns, L. (2001). Antigen specificity and high affinity binding provided by one single loop of a camel single-domain antibody. J. Biol. Chem. 276, 26285– 26290. Tomlinson, I. M., Walter, G., Marks, J. D., Llewelyn, M. & Winter, G. (1992). The repertoire of human germ-line VH sequences reveals about fifty groups of V(H) segments with different hypervariable loops. J. Mol. Biol. 227, 776– 798. Kabat, E. A., Wu, T. T., Perry, H. M., Gottesman, K. S. & Foeller, C. (1991). Sequences of Proteins of Immunological Interest, 5th edit., U.S. Department of Health and Human Services, Public Health Service National Institute of Health, Bethesda. Cornish-Bowden, A. (1985). Nomenclature for incompletely specified bases in nucleic acid sequences: recommendations 1984. Nucl. Acids Res. 13, 3021– 3030. Eigenbrot, C., Gonzalez, T., Mayeda, J., Carter, P., Werther, W., Hotaling, T. et al. (1994). X-ray structures of fragments from binding and nonbinding versions of a humanized anti-CD18 antibody: structural indications of the key role of VH residues 59 to 65. Proteins: Struct. Funct. Genet. 18, 49 – 62. Collaborative Computational Project Number 4 (1994). Collaborative Computational Project 4. Acta. Crystallog. sect. D, 760– 763. Santoro, M. M. & Bolen, D. W. (1988). Unfolding free energy changes determined by linear extrapolation method. 1. Unfolding of phenylmethane-sulfonyl alpha-chymotrypsin using different denaturants. Biochemistry, 27, 8063– 8068. Pace, C. N. (1990). Measuring and increasing protein stability. Trends Biotechnol. 8, 93 – 98. Pace, C. N. & Scholtz, J. M. (1997). Measuring the conformational stability of a protein. In Protein Structure, A practical Approach (Creighton, T. E., ed.), pp. 299– 321, Oxford University Press, New York. Myers, J. K., Pace, C. N. & Scholtz, J. M. (1995). Denaturant m values and heat capacity changes: relation to changes in accessible surface areas of protein unfolding. Protein Sci. 4, 2138– 2148. Horton, R. M., Hunt, H. D., Ho, S. N., Pullen, J. K. & Pease, L. R. (1989). Engineering hybrid genes without the use of restriction enzymes: gene splicing by overlap extension. Gene, 77, 61 – 68. Goodford, P. J. (1996). Multivariate characterization of molecules for QSAR analysis. J. Chemom. 10, 107– 117.
Structure of a Human Single Domain
48. Owen, D. J., Vallis, Y., Noble, M. E., Hunter, J. B., Dafforn, T. R., Evans, P. R. & McMahon, H. T. (1999). A structural explanation for the binding of
903
multiple ligands by the alpha-adaptin appendage domain. Cell, 97, 805– 815.
Edited by I. Wilson (Received 23 September 2003; received in revised form 9 January 2004; accepted 2 February 2004)
J. Mol. Biol. (2004) 337, 893–903
Crystal Structure of HEL4, a Soluble, Refoldable Human VH Single Domain with a Germ-line Scaffold Laurent Jespers1, Oliver Schon1, Leo C. James1, Dmitri Veprintsev2 and Greg Winter1* 1 Laboratory of Molecular Biology, Medical Research Council Centre, Hills Road Cambridge CB2 2QH, UK 2
Centre for Protein Engineering Medical Research Council Centre, Hills Road, Cambridge CB2 2QH, UK
The antigen binding site of antibodies usually comprises associated heavy (VH) and light (VL) chain variable domains, but in camels and llamas, the binding site frequently comprises the heavy chain variable domain only (referred to as VHH). In contrast to reported human VH domains, VHH domains are well expressed from bacteria and yeast, are readily purified in soluble form and refold reversibly after heat-denaturation. These desirable properties have been attributed to highly conserved substitutions of the hydrophobic residues of VH domains, which normally interact with complementary VL domains. Here, we describe the discovery and characterisation of an isolated human VH domain (HEL4) with properties similar to those of VHH domains. HEL4 is highly soluble at concentrations of $ 3 mM, essentially monomeric and resistant to aggregation upon thermodenaturation at concentrations as high as 56 mM. However, in contrast to VHH domains, the hydrophobic framework residues of the VH:VL interface are maintained and the only sequence changes from the corresponding human germ-line segment (V3-23/DP-47) are located in the loops comprising the complementarity determining regions (CDRs). The crystallographic structure of HEL4 reveals an unusual feature; the sidechain of a framework residue (Trp47) is flipped into a cavity formed by Gly35 of CDR1, thereby increasing the hydrophilicity of the VH:VL interface. To evaluate the specific contribution of Gly35 to domain properties, Gly35 was introduced into a VH domain with poor solution properties. This greatly enhanced the recovery of the mutant from a gel filtration matrix, but had little effect on its ability to refold reversibly after heat denaturation. Our results confirm the importance of a hydrophilic VH:VL interface for purification of isolated VH domains, and constitute a step towards the design of isolated human VH domains with practical properties for immunotherapy. q 2004 Elsevier Ltd. All rights reserved.
*Corresponding author
Keywords: antibody; single domain; human VH; folding; crystal structure
Introduction Present address: L. Jespers & O. Schon, Domantis Limited, Granta Park, Abington, Cambridge CB1 6GS, UK. Abbreviations used: CDR, complementarity determining regions; dAb, domain antibody; HEL, hen egg lysozyme; Kd , dissociation constant; Mr , molecular mass; PBS, phosphate buffer saline (50 mM phosphate, 150 mM NaCl, pH 7.4); rmsd, root mean square deviation; TMB, 3,30 ,5,50 -tetramethylbenzidine; VH, variable domain of classical heavy chain; VL, variable domain of classical light chain; VHH, variable domain of “heavy chain” antibody from camels and llamas. E-mail address of the corresponding author: [email protected]
The antigen binding sites of human and mouse antibodies comprise six hypervariable loops (complementarity determining regions or CDRs) mounted on the b-sheet scaffolds of paired heavy (VH) and light (VL) chain variable domains. Both domains usually contribute to antigen binding, but in some cases single VH domains isolated from natural antibodies may have comparable affinities and good specificities.1 However such isolated VH single domains (termed VH domain antibodies or VH dAbs) can be difficult to express in Escherichia coli and prone to aggregate upon concentration or
0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
894 prolonged standing at 4 8C.1 – 5 In addition several human VH and VL dAbs show a tendency to bind to the agarose – dextran matrix during sizeexclusion chromatography, thereby resulting in poor recovery and/or aberrant elution profiles under native buffer conditions.5,6 These properties, often referred to as “stickiness”, are thought to derive from exposure of the side-chains forming the hydrophobic interface onto which the VL domain would normally pack.1 The most plausible model for retardation and/or poor recovery from columns involves transient or permanent binding of folded dAbs on the agarose– dextran matrix, through the exposed surface. However in some cases irreversible deposition on the gel filtration matrix could be mediated via the (partially) unfolded state.6 By contrast, camels and llamas produce a class of “heavy chain” antibodies, thought to have lost their light chains and CH1 domains during evolution,7 from which the single variable domains (termed VHHs) are well expressed and form highly soluble monomeric species in solution.8,9 These highly desirable properties were attributed to a series of germ-line substitutions (Leu11 ! Ser, Val37 ! Tyr, Gly44 ! Glu, Leu45 ! Arg, and Trp47 ! Gly/Ile, Trp103 ! Arg/Gly) in the b-sheet scaffold of VHH dAbs, which increase the hydrophilicity of the VHH surfaces that would normally contact the VL and the CH1 domains in conventional antibodies.10 Some features of these VHH sequences of camels have been used to help improve the solution properties of human VH dAbs (a process referred to as “camelisation”).2 By inserting VHH amino acids in the framework 2 (at positions 44, 45 and 47) of a human VH dAb (aOX-13), two dAbs with improved behaviour in solution were generated,2 thereby enabling NMR studies.11 However, camelising VHH substitutions were shown to reduce the expression levels of dAbs,2,3 proved to be thermodynamically destabilising3,12 due to framework deformations,11 and did not completely alleviate aggregation in solution.4,13,14 In another approach, which exploits the natural diversity of the immunoglobulin repertoire, computer databases were screened for candidate VH dAbs bearing hydrophilic mutations at the former VL interface.15 A murine VH dAb bearing an unusual substitution in framework 2 (Gly44 ! Lys) was found that is monomeric, highly soluble ($ 2 mM) and does not stick to the gel filtration matrix. Further work has indicated that other sets of solvent-exposed residues may improve the solution properties of dAbs. For example, several dAbs derived from the heavy chains of conventional llama antibodies expressed well in bacteria, and did not aggregate in solution or stick to the matrix during size-exclusion chromatography.16 Their properties were attributed to a set of five framework substitutions (at positions 6, 74, 83, 84, and 108) unique to llama VH dAbs. Moreover, the CDRs of dAbs from
Structure of a Human Single Domain
camels and llamas can also contribute to improved solution properties. Indeed, the CDR3 loop of VHH domains is on average much longer (17 residues) than the human (12 residues) or mouse (nine residues) CDR3 loops,10,17 and often folds back on the region corresponding to the hydrophobic interface of conventional VH dAbs (as shown by several crystal structures).18 – 21 Subsequently, the expression and the solution properties of camelised and human VH dAbs were improved by extending the length of the CDR3 loops.5,6,22 A further property of the VHH domains is that they often undergo reversible unfolding, thus regaining native state and antigen-binding specificity,5,23 even at concentrations as high as 100 mM24 and after prolonged incubation at temperatures in the range 80 8C to 92 8C.23,25 Several of the llama VH dAbs had similar properties.16,26 By contrast, antibodies of other species and their fragments, including VH dAbs, usually aggregate irreversibly upon heating at low concentrations.25,27 – 30 The small size of camel and llama dAbs, their high solubility, resistance to aggregation in solution, lack of stickiness and tendency to unfold reversibly upon thermal denaturation are highly desirable properties for biotechnological application. However, for therapy it is also desirable that the dAbs are of human origin. Although various attempts have been made to endow human VH dAbs with all the biophysical properties of VHH dAbs (see above), success has been limited. In the course of characterising the properties of human VH dAbs we discovered a domain without camelising mutations but with several properties similar to camel and llama VHH dAbs.
Results Biophysical characterisation of the HEL4 dAb A phage library of human VH dAbs (1.1 £ 1010 clones) (see Materials and Methods) was selected for binding to HEL. After three rounds of selection, individual dAbs ðn ¼ 80Þ were expressed as soluble fragments in bacteria and screened by ELISA. A total of 21 HEL-specific clones were sequenced, revealing six unique sequences of which one (HEL4) was chosen for further study. The HEL4 dAb comprised an 11 amino acid residue CDR3 loop, and 11 amino acid mutations in the VH3 V3-23/DP-47 germ-line segment, which were located within the loops comprising the CDRs consistent with the library design (Figure 1). The HEL4 dAb was purified by affinity chromatography on protein A-Sepharose and then by size-exclusion chromatography on Superdex G75, eluting essentially in a single peak at the expected size for a 13 kDa monomer (Figure 2(a)). This behaviour contrasted with that of other human dAbs studied,5,6 which eluted in small peaks with most of the material aggregating to the chromatographic support. Furthermore, the HEL4 dAb
Structure of a Human Single Domain
895
Figure 1. Sequences of VH domain antibodies presented in this study. Residue numbering according to Kabat et al.38 Residues that differ between HEL4 and DP47d dAb, are underlined on the HEL4 line; the sites of the DP47d single mutants Ser35 ! Gly and Trp47 ! Arg are marked in bold on the DP47d line.
Figure 2. Biophysical properties of the HEL4 dAb. (a) Analytical gel-filtration chromatogram of the HEL4 dAb (10 mM in PBS) applied on a Superdex G75 column. (b) Thermally induced denaturation curves of dAb HEL4 (5 mM in PBS) recorded by CD at 235 nm: (X) mean residual ellipticity upon first heating, (S) mean residual ellipticity upon second heating. Inset: CD spectra of dAb HEL4 (5 mM in PBS) in the far-UV region at different temperature: (W) 25 8C before unfolding; (B) 85 8C (unfolded protein); (O) 25 8C after sample cooling.
896
could be concentrated to 38 mg/ml (3.0 mM) without signs of aggregation. Analytical ultracentrifugation was performed over a range of HEL4 dAb concentrations. At 2 mM and 20 mM, the equilibrium data were consistent with a single species model, whereas at 200 mM, the data could equally fit to a monomeric model or to a monomer –dimer equilibrium with a best fit dissociation constant ðKd Þ of 5 mM (data not shown). The Kd of HEL4 dAb for HEL was determined to be 0.13 mM using the method described by Friguet et al.31 Thermal unfolding of the HEL4 dAb was monitored by circular dichroism (CD) analysis in the far-UV, using a concentration range of 5 –56 mM protein concentration. Secondary structure was lost in a cooperative unfolding transition (Tm ¼ 62:1 8C) and at 85 8C the protein was fully denatured as shown by the 200 – 250 nm spectrum (Figure 2(b), inset). After cooling to 25 8C, the spectrum could be superimposed on that prior to heating, indicating quantitative refolding. The refolded HEL4 dAb was functional as shown by binding to
Structure of a Human Single Domain
protein A- and HEL-coated wells by ELISA (data not shown). Further indications of a reversible two-state mechanism were obtained by re-heating the HEL4 dAb. An almost identical denaturation curve was observed (Figure 2(b)), with marginal loss of ellipticity (# 5%, which may have resulted from aggregation to the quartz surface and/or chemical degradation due to prolonged incubation above 80 8C). The thermodynamic stability (DGN-U) of the HEL4 dAb at 25 8C was calculated as 27.0 kJ/mol using the thermal unfolding data, and is similar to that for urea-denaturation (28.4 kJ/mol) (data not shown). Crystal structure of the HEL4 domain antibody The HEL4 dAb was crystallised in space group I4122 with two molecules in the asymmetric unit. ˚ resolution and The structure was refined to 2.0 A shown to adopt the standard fold of an immunoglobulin variable domain with nine antiparallel b-strands (Figure 3). The root mean
Figure 3. Stereo views of the crystal structure of HEL4. (Top) Superimposed main-chain conformations of the CDR3comprising loop (residues 94 –103) of HEL4 dAb crystallised in dimeric form (dAbs HEL4-A and HEL-B in pink and purple, respectively). Side-chains at positions 96 (Leu96) and 98 (Pro98) in each CDR3-comprising loop are shown to highlight their positions. Residue numbering according to Kabat et al.38 (Bottom) Main-chain conformation of HEL4 dAb with b-strand symbolised as arrows, and coloured CDR-comprising loops: H1 (yellow), H2 (red), and H3 (blue).
897
Structure of a Human Single Domain
square deviation (rmsd) calculated with framework Ca atoms of the 1FGV VH domain (used for ˚ with most of molecular replacement) was 0.53 A the important divergences around residues 40 –43. In the asymmetric unit, two HEL4 dAbs form a homo-dimer (HEL4-A:HE4-B) but are not exactly symmetry-related, with the main differences located in CDR3 (Figure 3). Thus, the CDR3 of the HEL4-A dAb interacts with the HEL4-B dAb (through Leu96 and Leu100c) but not vice versa. Furthermore, Pro98 of CDR3 is either buried (in the HEL4-A dAb) or exposed (in the HEL4-B dAb); this may play a role in “freezing” the CDR3 in two different conformations. The main-chain conformation of the H1 loop (residue Gly26 to Gly35) is characterised by a sharp turn at position 26 and a hydrophobic cluster at positions 24, 27, 29, 34, and fits the type-1 ˚ rmsd with 1FGV-H1 canonical structure32 (0.86 A loop) (Figure 4(a)). The main-chain conformation of the H2 loop (residues 52– 56) is characterised by a six-residue hairpin and Gly54 and Arg71 as key structural elements, and fits the type-3 canonical ˚ rmsd with 1FGV-H2 loop) structure loop32 (0.76 A (Figure 4(b)). By comparing each of the HEL4 framework residues with those of VH domains from conventional VH:VL antibody pairs, we observed that only one side-chain had undergone a major reorientation: Trp47 had rotated by 1808 about its Ca –Cb bond (Figure 5(a)). This unusual rotation brings the polar N1 atom of the side-chain towards solvent, and tucks the phenyl ring into a hydrophobic cavity formed by residues Gly35, Val37 and Ala93. Consequently, the face of the HEL4 dAb that
would normally contact a VL domain in conventional antibodies, becomes significantly more hydrophilic than that of a conventional VH3 domain (Figure 5(b) and (c)), thereby mimicking that of VHH dAbs (Figure 5(d)).21 Of the three amino acid residues forming the hydrophobic cavity into which Trp47 locks, only Gly35 (which is in CDR1) differs in the germline 3-23 segment (where it is serine) and it is the lack of a sidechain at position 35 in the HEL4 dAb, that helps to create the hydrophobic cavity. Biophysical analysis of mutants To investigate the role of the Trp47, two mutants, Ser35 ! Gly and Trp47 ! Arg, were made on the human DP47d dAb from which the HEL4 dAb had been derived (see Materials and Methods) (Table 1). Unlike HEL4 dAb, the DP47d dAb exhibited poor solubility, aggregation on gel filtration with poor recovery (, 5%), and irreversible denaturation upon heating (data not shown). When the Trp47 ! Arg dAb was loaded onto a Sephadex G75 column, 80% of the material was recovered from the column, of which 90% was monomeric. Similarly, with the Ser35 ! Gly dAb, 60% was recovered of which 85% was monomeric (Table 2). However, neither of these mutants survived heat denaturation at 5 mM concentration (data not shown). The mutation of Ser35 to Gly appeared to be less destabilising than that of Trp47 to Arg, as shown by measurement of thermodynamic stabilities of both mutants at 25 8C by urea denaturation (Table 2).
Figure 4. Canonical CDR-comprising loop conformations in the VH segment of the HEL4 dAb. Superimposition of the (a) type-1 H1 loop (residues 26 – 35) and (b) type-3 H2 loop (residues 50 – 58) of the HEL4 dAb (in pink) and the VH3 domain of H52, a humanised anti-CD18 single-chain Fv fragment (PDB entry 1FGV)40 (in blue). Side-chains of residues 27, 29 and 34 in loop H1, and 51, 54 and 58 in loop H2 are shown to highlight their respective positions in both molecules. Residue numbering according to Kabat et al.38
898
Structure of a Human Single Domain
Figure 5. Hydrophobic character of the (former) VL-contacting interfaces in VH and VHH domains. (a) Close-up view of the orientation of the side-chains at positions 35, 37 and 47 in HEL4 dAb (in pink) and in the humanised VH domain of H52 (PDB entry 1FGV) (in blue). Computed molecular surfaces at (b) the face of the HEL4 dAb that would be expected to contact a VL-domain in conventional antibodies, (c) the VL-contacting face of the humanised VH domain of H52, and (d) the former VL-interface of the AMD9 camel VHH domain specific for porcine pancreatic a-amylase (PDB entry 1KXQ).21 The hydrophobic potentials of each molecular surface were calculated by GRID and AESOP.47,48 Areas of marked hydrophobicity are marked in green. The side-chains of residues pivotal to VHH domains (37, 44, 45, 47 and 103) are shown to highlight their respective orientation. Residue numbering according to Kabat et al.38
Discussion Table 1. Data collection and structure refinement statistics Molecule name Space group Buffer ˚) Resolution (A Unique reflections Rmerge Redundancy Completeness (%) Average I=sI Final Rcryst Final Rfree Bond rmsd Angle rmsd (deg.) PDB code
HEL4 I4122 0.1 M Hepes (pH 7.2), 1.6 M MgSO4, 6% glycerol 2.0 19,067 0.143 (0.667) 7.0 (6.0) 99.7 (96.0) 4.3 (2.2) 0.177 (0.183) 0.256 (0.259) 0.007 1.4 1OHQ
Numbers in parentheses are for highest resolution shell ˚ ). (2.11–2.0 A
Our study of the HEL4 dAb, a single domain antibody based on a human germ-line VH3 framework, has shown that excellent solution properties and reversible unfolding on heating are not exclusive to dAbs from camels and llamas. The use of human rather than animal dAbs is clearly preferred for therapeutic use, and these properties may find particular application where harsh conditions such as transient heating, sterilisation, surface deposition and lyophilisation are part of the manufacturing process. Although the HEL4 dAb crystallises as a dimer (see below), we demonstrated by two independent techniques, gel filtration and analytical ultracentrifugation, that it is essentially monomeric at most of the concentrations used in the assays (up to 200 mM). No signs of aggregation were detected after concentrating the samples up to 3.0 mM and incubation in PBS at 37 8C for several days. Above
899
Structure of a Human Single Domain
Table 2. Biophysical properties of human VH3 dAbs dAb DP47d HEL4 Ser35 ! Gly Trp47 ! Arg
Tm (8C)e
DGN-U (kJ/mol)
% Recoverya on Superdex G75
Monomeric (M) or dimeric (D)
Yieldb (mg/ml)
61.4 62.1 61.6 54.8
34.7c 27.0d –28.4c 29.1c 24.0c
,5 .90 60 80
n.d. M M þ 15% D M þ 10% D
2.9 9.5 10.3 6.1
n.d., not determined. a Obtained by integrating the areas of peak(s) eluted from Superdex G75. b Yield of purified protein obtained from one litre of bacterial culture supernatant and normalised to 5 A600 nm : c Thermodynamic stability value at 25 8C, obtained by analysis of urea-induced denaturation. d Thermodynamic stability value at 25 8C, obtained by analysis of thermo-denaturation. e Temperature of mid-point transition upon reversible unfolding (HEL4) or upon aggregation (DP47d, Ser35 ! Gly, Trp47 ! Arg).
55 8C, the HEL4 dAb unfolds cooperatively without aggregating (at concentrations up to 56 mM and temperature up to 85 8C) and upon cooling renatures almost quantitatively. The Tm value of 62.1 8C for the HEL4 dAb compares well with those reported for camel VHH and human VH domains5,23 but its conformational stability (27.0 – 28.4 kJ/mol) is in the lower range of values reported for VHH domains and other human VH3 domains (21.1 – 52.7 kJ/mol),5,23 but still higher than those for human VH domains belonging to other families (13.7 –26.0 kJ/mol).6 The fact that other human dAbs with identical framework sequences to the HEL4 dAb have poor biophysical properties (DP47d dAb and Ewert et al.5) locates the determinants for the desirable properties of HEL4 exclusively in the loops comprising the CDRs. When the structure of HEL4 dAb was solved by X-ray crystallography we found two HEL4 molecules packing onto each other in a manner similar to that of a VL:VH pair in a conventional antibody and to the packing VL:VL domains in Bence-Jones proteins.33 The buried surface area of each HEL4 monomer upon packing ˚ 2 for HEL4-A dAb and HEL4-B ˚ 2 and 895 A (988 A dAb, respectively) compares favourably to that of ˚ 2) and VH (793 A ˚ 2) domains of the the VL (988 A 1FGV scFv. However, as the HEL4 dAb is a monomer in solution, the packing observed in the crystal cannot reflect its state other than at very high concentrations. A similar observation was reported for human thioredoxin, which crystallised as a dimer but displayed all the characteristics of a monomer in solution.34 The structures of the H1 and H2 loops in the HEL4 dAb proved to be typical of this family of human VH domains. However this was not true for the HEL4 CDR3. It has been shown that the backbone conformations of residues at the base of antibody CDR3 loops can fall into one of two classes, kinked or extended. The kinked class is more common.35 In HEL4, we see two conformations of the CDR3 loop, and they are both extended. We do not know if any of these CDR3 conformations are present in the monomeric HEL4 dAb; nor do we know therefore whether the conformation of the CDR3 contributes to its solution
properties as with camel antibodies.18 – 21 It is not clear whether the multiple CDR3 conformations reflect the lack of a constraining light chain, or are driven by the association of the two VH domains together in the crystal. A further unusual feature of the HEL4 structure was the rotation of the Trp47 side-chain. In conventional VH:VL antibody pairs, this highly conserved framework residue lies flat on the VH surface, such that its hydrophobic phenyl ring engages with the VL domain. In the HEL4 structure, this side-chain re-orients itself within the cavity formed by residues Gly35, Val37 and Ala93. The effect of this is to increase the hydrophilicity of the face of the domain usually involved in packing with the VL domain, and we suggest that this may be responsible in part for the properties of the HEL4 dAb in solution and upon gel filtration. This is supported by our observations that two mutations in the DP47d dAb, Trp47 ! Arg (where a hydrophobic residue is replaced by a hydrophilic) and Ser35 ! Gly (where a cavity is created into which Trp47 may flip), lead to a considerable increase in the recovery of the dAb from gel filtration. However the improvements came at a cost: the Ser35 ! Gly mutation reduced the thermodynamic stability of the DP47d dAb by 5.6 kJ/mol, whereas 10.7 kJ/mol was lost upon substitution of Trp47 by Arg. The Trp47 flip is unique among known VH structures and suggests that surface plasticity may contribute to modulating hydrophility. However, we believe that the Trp47 flip is only part of the story; other residues such as those of the loop comprising the CDR3, may also make contributions to its solution properties. It was proposed earlier that rotation of another framework residue, Trp103, contributes to increasing the hydrophilicity of VHH domains.18,36 A similar orientation of Trp103 is observed in the crystal structure of the HEL4 dAb but our structural analysis of several human and murine VH domains (not shown) reveals that the “VHH-orientation” of Trp103 side-chain is common, if not the rule among VH:VL pairs. Like several camel and llama dAbs, the human HEL4 dAb combines the properties of high solubility, monomeric state, lack of stickiness and reversible thermal unfolding. It seems likely that
900
avoidance of aggregation is a common aspect of each of these properties. Protein aggregation is mediated by intermolecular interactions involving either the folded state, or the unfolded state of the molecule (or indeed intermediate unfolded states). We suggest that the Trp47 flip is responsible for the ability of the native state of the HEL4 dAb to avoid aggregation. However, we do not believe that the flip contributes to the ability of the HEL4 dAb to survive thermal unfolding, as evidenced by the study of single-point mutants of the DP47d dAb (Ser35 ! Gly and Trp47 ! Arg) which were soluble but aggregated upon thermal denaturation. Those features of the HEL4 dAb conferring reversible unfolding on thermal denaturation have yet to be identified, and we are not therefore in a position to transpose all the desirable biophysical properties of HEL4 to other more biomedically relevant dAbs. However, we believe that our observations will help in understanding some of the biophysical properties of dAbs and thereby to the engineering of therapeutics based on human dAbs.
Structure of a Human Single Domain
Crystallisation, data collection and structure determination Crystal needles of HEL4 dAb were obtained with the hanging-drop method by mixing 1 ml of protein solution (at 22.3 mg/ml) and 1 ml of reservoir solution (0.1 M Hepes (pH 7.2), 1.6 M MgSO4, 6% (v/v) glycerol) (Hampton Research). Crystals were frozen at 100 K and data were collected at the European Synchroton Radiation Facility (Grenoble, France) on a CCD detector. Data were indexed and integrated using MOSFLM and scaled in SCALA in space group I422 (Table 1). Data ˚ . The strucwith at least an I=sI . 2:0 extended to 2.0 A ture was determined by molecular replacement using the VH3 domain of a humanised anti-CD18 single-chain Fv fragment H52 (PDB entry 1FGV)40 as a model in the program AMORE.41 Initial maps were calculated in REFMAC using the rigid-body model.41 All CDR loops were deleted and the model rebuilt over four rounds of refinement using REFMAC without TLS. Water molecules were added in ARP/wARP over the last two rounds.41 The final model has an Rfree of 0.256 and an Rfactor of 0.177. All residues are located in allowed regions in a Ramachandran plot. Affinity measurement
Materials and Methods Selection, production and purification of HEL4 dAb The human HEL4 dAb (Figure 1) was obtained from a phage-displayed library of human VH dAbs (1.1 £ 1010 clones) after biopanning on hen egg lysozyme (HEL). This repertoire was based on a template gene, DP47d, comprising the germ-line 3-23 segment of the human VH3 family (DP-47)37 and the human JH4b segment (Figure 1); it was built by oligonucleotide-mediated diversification of several codons in DP47d as follows (Kabat numbering38 for the amino acid positions, IUPAC-IUB code39 for the nucleotides): 27, KWT; 28, ANS; 29, NTT; 30, ANC; 31, NMT; 32, NAS; 33, DHT; 35, RSC; 50, RSC; 52, NNK; 52a, RNS; 53, VVW; 54, CGT; 94, RSW; 101, NVS; 102, THT; and NNK codons for all positions from 95 to 100a,b,…,k. For expression in E. coli HB2151 cells, the two tags, (His)6 and VSV, at the C-terminal end of the HEL4 dAb were first removed from the phagemid vector. Five liters of 2 £ TY medium containing ampicillin (100 mg/ml) and 0.1% (w/v) glucose were incubated at 37 8C until A600 nm , 0:8: Expression was induced with 1 mM IPTG at 30 8C. After four hours, the cells were harvested, and the periplasmic fraction was extracted with 150 ml of ice-cold TES buffer. After centrifugation (6000g for 15 minutes followed by 17,000g for 30 minutes), the cleared supernatant was loaded onto a 15 ml protein A-Sepharose column (Amersham Biosciences). After washing the column with 150 ml of PBS (50 mM phosphate, 150 mM NaCl, pH 7.4), 0.5 M NaCl, bound HEL4 dAb was eluted in 0.1 M glycine– HCl (pH 3.0), and neutralised with 1 M Tris – HCl (pH 7.4). The protein sample was concentrated and further purified on Superdex G75 (Amersham Biosciences) in 20 mM Tris (pH 7.4). Protein purity was estimated by visual analysis after SDS-PAGE on 12% (w/v) acrylamide Tris – glycine gel (Invitrogen). Protein concentration was measured at 280 nm, assuming an extinction coefficient of 26,150 M21 cm21.
The apparent dissociation constant ðKd Þ of HEL4 dAb from HEL was determined as described by Friguet et al.31 by mixing a solution of 78 nM His-tagged HEL4 dAb in PBST with increasing amounts of HEL (from 0 mM to 2 mM). After overnight incubation, the solutions were added to HEL-coated wells in PBS buffer containing 2% Tween-20. Bound HEL4 dAb was detected with Ni2þ-peroxidase conjugate (Perbio), and developed with TMB as substrate. The absorbances at 450 nm were corrected for background signal and processed into a Scatchard plot to calculate Kd : Fluorescence measurements The samples were prepared by mixing a 5.0 mM solution of purified dAb in PBS to increasing concentrations of urea (ranging from 0 M to 8 M) and by overnight incubation at 25 8C. The samples were transferred to a quartz cuvette (1 cm path-length), and placed into a F4500 fluorimeter (Hitachi). Fluorescence emission intensities at 320 nm (for HEL4 dAb) or at 350 nm (for DP47d dAb, and mutants Ser35 ! Gly and Trp47 ! Arg) were then recorded at 25 8C using an excitation wavelength of 280 nm, and a slit width of 2 nm for excitation and emission. Unfolding curves and thermodynamic stabilities at 25 8C were obtained by fitting the data of fluorescence intensity versus urea concentration, assuming a two-state cooperative equilibrium as described.42,43 Circular dichroism measurements CD cuvettes (1 cm path-length) were filled with a 5 mM solution of dAb in PBS and transferred to a J-720 polarimeter (Jasco). CD spectra at 25 8C and 85 8C were recorded in the far-UV (200 – 250 nm) with a 2 nm bandwidth, a one second integration time and a heating rate of 50 8C per hour. Unfolding curves from 25 8C to 85 8C were monitored at 235 nm. After unfolding, the sample was cooled down to 25 8C, a spectrum was recorded, and a new thermal-unfolding curve was recorded. For the HEL4 dAb, the unfolding curves were assumed to
901
Structure of a Human Single Domain
be two-state and fitted as described44 using a DCp contribution of 12 cal per amino acid residue.45 The values obtained for Tm (midpoint transition temperature in Kelvin), DHm (enthalpy change for unfolding at Tm ) were then used to calculate the thermodynamic stability ðDGN-U Þ of the protein at 25 8C as described.44 Alternatively, dAb unfolding upon addition of urea at 25 8C was monitored by CD measurements on a polarimeter 215 (AVIV Instruments). Protein samples (Ser35 ! Gly and Trp47 ! Arg) were prepared as described (see Fluorescence measurements). The changes of ellipticity value (D1) were plotted versus the urea concentration and the unfolding curves were fitted as described42,43 to obtain the thermodynamic stabilities at 25 8C. Analytical ultracentrifugation Sedimentation equilibrium experiments were done in a Beckman Optima XLI analytical ultracentrifuge with Ti-60 rotor using interference and absorbance at 280 nm and 230 nm, at 25 8C. The HEL4 dAb was loaded into six-sector 12 mm path length cells at three different concentrations in PBS: 2 mM, 20 mM, and 200 mM. The samples were centrifuged until they reached equilibrium as judged by the changes in the subsequent scans at speeds of 30,000 rpm and 40,000 rpm. Data were analysed using UltraSpin software†. Mutant construction and protein purification The gene encoding the synthetic VH DP47d dAb was obtained from Ian Tomlinson, and used to assemble two single-point mutants (termed Ser35 ! Gly and Trp47 ! Arg, Figure 1) by SOE-PCR.46 Next, the three genes were transferred into the expression vector and for each dAb, an overnight E. coli culture was used to inoculate five litres of 2 £ TY medium containing ampicillin (100 mg/ml). When A600 nm reached ,0.6, induction was started with 1 mM IPTG and the culture was further incubated at 30 8C for 16 hours. After centrifugation, the supernatant was filtered and incubated overnight with 30 ml of Streamline-protein A beads (Amersham Biosciences) at 4 8C. The beads were then packed into a column, washed with 300 ml of PBS, and bound dAbs were eluted in 0.1 M glycine– HCl (pH 3.0). After neutralisation with 1 M Tris – HCl (pH 7.4), the protein samples were dialysed in PBS and concentrated before storage at 4 8C. Protein purity and concentration were estimated as described above. Atomic coordinates The model and structure factors were deposited in the RCSB Protein Data Bank with accession codes 1OHQ and 1OHQSF, respectively.
Acknowledgements We are grateful to Roger Williams, Olga Perisic, Amrik Basran, Chris Johnson and Neil Ferguson for technical help, and to Ian Tomlinson and Kevin Moulder for advice. L.J. and O.S. were † http://www.mrc-cpe.cam.ac.uk
funded by Domantis Limited (Cambridge, UK) under a collaborative research programme with the Medical Research Council.
References 1. Ward, E. S., Gu¨ssow, D., Griffiths, A. D., Jones, P. T. & Winter, G. (1989). Binding activities of a repertoire of single immunoglobulin variable domains secreted from Escherichia coli. Nature, 341, 544– 546. 2. Davies, J. & Riechmann, L. (1994). “Camelising” human antibody fragments: NMR studies on VH domains. FEBS Letters, 339, 285– 290. 3. Davies, J. & Riechmann, L. (1995). Antibody VH domains as small recognition units. Biotechnology (NY), 13, 475– 479. 4. Kortt, A. A., Guthrie, R. E., Hinds, M. G., Power, B. E., Ivancic, N., Caldwell, J. B. et al. (1995). Solution properties of Escherichia coli-expressed VH domain of anti-neuraminidase antibody NC41. J. Protein Chem. 14, 167–178. 5. Ewert, S., Cambillau, C., Conrath, K. & Plu¨ckthun, A. (2002). Biophysical properties of camelid VHH domains compared to those of human VH3 domains. Biochemistry, 41, 3628–3636. 6. Ewert, S., Huber, T., Honegger, A. & Plu¨ckthun, A. (2003). Biophysical properties of human antibody variable domains. J. Mol. Biol. 325, 531– 553. 7. Hamers-Casterman, C., Atartouch, T., Muyldermans, S., Robinson, G., Hamers, C., Bajyana Songa, E. et al. (1993). Naturally occurring antibodies devoid of light chains. Nature, 363, 446–448. 8. Arbabi Ghahroudi, M., Desmyter, A., Wyns, L., Hamers, R. & Muyldermans, S. (1997). Selection and identification of single domain antibody fragments from camel heavy-chain antibodies. FEBS Letters, 414, 521– 526. 9. Lauwereys, M., Arbabi Ghahroudi, M., Desmyter, A., Kinne, A., Holzer, W., De Genst, E. et al. (1998). Potent enzyme inhibitors derived from dromedary heavy-chain antibodies. EMBO J. 17, 3512 –3520. 10. Muyldermans, S., Atarhouch, T., Saldanha, J., Barbosa, J. A. R. G. & Hamers, R. (1994). Sequence and structure of VH domain from naturally occurring camel heavy chain immunoglobulins lacking light chains. Protein Eng. 7, 1129– 1135. 11. Riechmann, L. (1996). Rearrangement of the former VL interface in the solution structure of a camelised, single antibody VH domain. J. Mol. Biol. 259, 957– 969. 12. Davies, J. & Riechmann, L. (1996). Single antibody domains as small recognition units: design and in vitro antigen selection of camelized, human VH domains with improved stability. Protein Eng. 9, 531 –537. 13. Martin, F., Volpari, C., Steinkuhler, C., Dimasi, N., Brunetti, M., Biasiol, G. et al. (1997). Affinity selection of a camelized V(H) domain antibody inhibitor of hepatitis C virus NS3 protease. Protein Eng. 10, 607 –614. 14. Voordijk, S., Hansson, T., Hilvert, D. & van Gunsteren, W. F. (2000). Molecular dynamics simulations highlight mobile regions in proteins: a novel suggestion for converting a murine VH domain into a more tractable species. J. Mol. Biol. 300, 963– 973. 15. Reiter, Y., Schuck, P., Boyd, L. F. & Plaksin, D. (1999). An antibody single-domain phage display library of a native heavy chain variable region: isolation of
902
16.
17. 18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28. 29.
30.
31.
functional single-domain VH molecules with a unique interface. J. Mol. Biol. 290, 685– 698. Tanha, J., Dubuc, G., Hirama, T., Narang, S. A. & MacKenzie, C. R. (2002). Selection by phage display of llama conventional V(H) fragments with heavy chain antibody V(H)H properties. J. Immunol. Methods, 263, 97 – 109. Wu, T. T., Johnson, G. & Kabat, E. A. (1993). Length distribution of CDRH3 in antibodies. Proteins: Struct. Funct. Genet. 16, 1 – 7. Desmyter, A., Transue, T. R., Ghahroudi, M. A., Thi, M. H., Poortmans, F., Hamers, R. et al. (1996). Crystal structure of a camel single-domain VH antibody fragment in complex with lysozyme. Nature Struct. Biol. 3, 803– 811. Spinelli, S., Frenken, L., Bourgeois, D., de Ron, L., Bos, W., Verrips, T. et al. (1996). The crystal structure of a llama heavy chain variable domain. Nature Struct. Biol. 3, 752– 757. Spinelli, S., Frenken, L. G., Hermans, P., Verrips, T., Brown, K., Tegoni, M. & Cambillau, C. (2000). Camelid heavy-chain variable domains provide efficient combining sites to haptens. Biochemistry, 39, 1217– 1222. Desmyter, A., Spinelli, S., Payan, F., Lauwereys, M., Wyns, L., Muyldermans, S. & Cambillau, C. (2002). Three Camelid V(HH) domains in complex with porcine pancreatic a-amylase. J. Biol. Chem. 277, 23645– 23650. Tanha, J., Xu, P., Chen, Z., Ni, F., Kaplan, H., Narang, S. A. & MacKenzie, C. R. (2001). Optimal design features of camelized human single-domain antibody libraries. J. Biol. Chem. 276, 24774– 24780. Dumoulin, M., Conrath, K., Van Meirhaeghe, A., Meersman, F., Heremans, K., Frenken, L. G. J. et al. (2002). Single-domain antibody fragments with high conformational stability. Protein Sci. 11, 500– 515. Perez, J. M., Renisio, J. G., Prompers, J. J., van Platerink, C. J., Cambillau, C., Darbon, H. & Frenken, L. G. J. (2001). Thermal unfolding of a llama antibody fragment: a two-state reversible process. Biochemistry, 40, 74 – 83. van der Linden, R. H., Frenken, L. G., de Geus, B., Harmsen, M. M., Ruuls, R. C., Stok, W. et al. (1999). Comparison of physical chemical properties of llama V(HH) antibody fragments and mouse monoclonal antibodies. Biochim. Biophys. Acta, 1431, 37 – 46. Vranken, W., Tolkatchev, D., Xu, P., Tanha, J., Chen, Z., Narang, S. & Ni, F. (2002). Solution structure of a llama single-domain antibody with hydrophobic residues typical of the VH/VL interface. Biochemistry, 41, 8570– 8579. Young, N. M., MacKenzie, C. R., Narang, S. A., Oomen, R. P. & Baenziger, J. E. (1995). Thermal stabilization of a single-chain Fv antibody fragment by introduction of a disulphide bond. FEBS Letters, 377, 135– 139. Wirtz, P. & Steipe, B. (1999). Intrabody construction and expression III: engineering hyperstable V(H) domains. Protein Sci. 8, 2245– 2250. Wo¨rn, A. & Plu¨ckthun, A. (1999). Different equilibrium stability behaviour of ScFv fragments: identification, classification, and improvement by protein engineering. Biochemistry, 38, 8739– 8750. Vermeer, A. W. P. & Norde, W. (2000). The thermal stability of immunoglobulin: unfolding and aggregation of a multi-domain protein. Biophys. J. 78, 394– 404. Friguet, B., Chaffotte, A. F., Djavadi-Ohaniance, L. &
Structure of a Human Single Domain
32.
33.
34. 35. 36.
37.
38.
39.
40.
41. 42.
43. 44.
45.
46.
47.
Goldberg, M. E. (1985). Measurements of the true affinity constant in solution of antigen-antibody complexes by enzyme-linked immunosorbent assay. J. Immunol. Methods, 77, 305–319. Chothia, C., Lesk, A. M., Gherardi, E., Tomlinson, I. M., Walter, G., Marks, J. D. et al. (1992). Structural repertoire of the human V(H) segments. J. Mol. Biol. 227, 799– 817. Epp, O., Colman, P., Fehlhammer, H., Bode, W., Schiffer, M., Huber, R. & Palm, W. (1974). Crystal and molecular structure of a dimer composed of the variable portions of the Bence-Jones protein REI. Eur. J. Biochem. 45, 513– 524. Gronenborn, A. M., Clore, G. M., Louis, J. M. & Wingfield, P. T. (1999). Is human thioredoxin monomeric or dimeric? Protein Sci. 8, 426– 429. Al-Lazikani, B., Lesk, A. M. & Chothia, C. (1997). Standard conformations for the canonical structures of immunoglobulins. J. Mol. Biol. 273, 927– 948. Desmyter, A., Decanniere, K., Muyldermans, S. & Wyns, L. (2001). Antigen specificity and high affinity binding provided by one single loop of a camel single-domain antibody. J. Biol. Chem. 276, 26285– 26290. Tomlinson, I. M., Walter, G., Marks, J. D., Llewelyn, M. & Winter, G. (1992). The repertoire of human germ-line VH sequences reveals about fifty groups of V(H) segments with different hypervariable loops. J. Mol. Biol. 227, 776– 798. Kabat, E. A., Wu, T. T., Perry, H. M., Gottesman, K. S. & Foeller, C. (1991). Sequences of Proteins of Immunological Interest, 5th edit., U.S. Department of Health and Human Services, Public Health Service National Institute of Health, Bethesda. Cornish-Bowden, A. (1985). Nomenclature for incompletely specified bases in nucleic acid sequences: recommendations 1984. Nucl. Acids Res. 13, 3021– 3030. Eigenbrot, C., Gonzalez, T., Mayeda, J., Carter, P., Werther, W., Hotaling, T. et al. (1994). X-ray structures of fragments from binding and nonbinding versions of a humanized anti-CD18 antibody: structural indications of the key role of VH residues 59 to 65. Proteins: Struct. Funct. Genet. 18, 49 – 62. Collaborative Computational Project Number 4 (1994). Collaborative Computational Project 4. Acta. Crystallog. sect. D, 760– 763. Santoro, M. M. & Bolen, D. W. (1988). Unfolding free energy changes determined by linear extrapolation method. 1. Unfolding of phenylmethane-sulfonyl alpha-chymotrypsin using different denaturants. Biochemistry, 27, 8063– 8068. Pace, C. N. (1990). Measuring and increasing protein stability. Trends Biotechnol. 8, 93 – 98. Pace, C. N. & Scholtz, J. M. (1997). Measuring the conformational stability of a protein. In Protein Structure, A practical Approach (Creighton, T. E., ed.), pp. 299– 321, Oxford University Press, New York. Myers, J. K., Pace, C. N. & Scholtz, J. M. (1995). Denaturant m values and heat capacity changes: relation to changes in accessible surface areas of protein unfolding. Protein Sci. 4, 2138– 2148. Horton, R. M., Hunt, H. D., Ho, S. N., Pullen, J. K. & Pease, L. R. (1989). Engineering hybrid genes without the use of restriction enzymes: gene splicing by overlap extension. Gene, 77, 61 – 68. Goodford, P. J. (1996). Multivariate characterization of molecules for QSAR analysis. J. Chemom. 10, 107– 117.
Structure of a Human Single Domain
48. Owen, D. J., Vallis, Y., Noble, M. E., Hunter, J. B., Dafforn, T. R., Evans, P. R. & McMahon, H. T. (1999). A structural explanation for the binding of
903
multiple ligands by the alpha-adaptin appendage domain. Cell, 97, 805– 815.
Edited by I. Wilson (Received 23 September 2003; received in revised form 9 January 2004; accepted 2 February 2004)