Chemical Basis for the Affinity Maturation of a Camel Single Domain Antibody*

Affinity maturation of classic antibodies supposedly proceeds through the pre-organization of the reactive germ line conformational isomer. It is less evident to foresee how this can be accomplished by camelid heavy-chain antibodies lacking light chains. Although these antibodies are subjected to somatic hypermutation, their antigen-binding fragment consists of a single domain with restricted flexibility in favor of binding energy. An antigen-binding domain derived from a dromedary heavy-chain antibody, cAb-Lys3, accumulated five amino acid substitutions in CDR1 and CDR2 upon maturation against lysozyme. Three of these residues have hydrophobic side chains, replacing serines, and participate in the hydrophobic core of the CDR1 in the mature antibody, suggesting that conformational rearrangements might occur in this loop during maturation. However, transition state analysis of the binding kinetics of mature cAb-Lys3 and germ line variants show that the maturation of this antibody relies on events late in the reaction pathway. This is reflected by a limited perturbation of ka and a significantly decreased kd upon maturation. In addition, binding reactions and the maturation event are predominantly enthalpically driven. Therefore, maturation proceeds through the increase of favorable binding interactions, or by the reduction of the enthalpic penalty for desolvation, as opposed to large entropic penalties associated with conformational changes and structural plasticity. Furthermore, the crystal structure of the mutant with a restored germ line CDR2 sequence illustrates that the matured hydrophobic core of CDR1 in cAb-Lys3 might be compensated in the germ line precursor by burying solvent molecules engaged in a stable hydrogen-bonding network with CDR1 and CDR2.

Affinity maturation of classic antibodies supposedly proceeds through the pre-organization of the reactive germ line conformational isomer. It is less evident to foresee how this can be accomplished by camelid heavychain antibodies lacking light chains. Although these antibodies are subjected to somatic hypermutation, their antigen-binding fragment consists of a single domain with restricted flexibility in favor of binding energy. An antigen-binding domain derived from a dromedary heavy-chain antibody, cAb-Lys3, accumulated five amino acid substitutions in CDR1 and CDR2 upon maturation against lysozyme. Three of these residues have hydrophobic side chains, replacing serines, and participate in the hydrophobic core of the CDR1 in the mature antibody, suggesting that conformational rearrangements might occur in this loop during maturation. However, transition state analysis of the binding kinetics of mature cAb-Lys3 and germ line variants show that the maturation of this antibody relies on events late in the reaction pathway. This is reflected by a limited perturbation of k a and a significantly decreased k d upon maturation. In addition, binding reactions and the maturation event are predominantly enthalpically driven. Therefore, maturation proceeds through the increase of favorable binding interactions, or by the reduction of the enthalpic penalty for desolvation, as opposed to large entropic penalties associated with conformational changes and structural plasticity. Furthermore, the crystal structure of the mutant with a restored germ line CDR2 sequence illustrates that the matured hydrophobic core of CDR1 in cAb-Lys3 might be compensated in the germ line precursor by burying solvent molecules engaged in a stable hydrogen-bonding network with CDR1 and CDR2.
The success of the humoral immunity relies heavily on the fast generation of highly specific and tight binding antibodies against virtually any foreign target molecule. These antibodies occur concomitantly as membrane-bound receptors on B-cells or as soluble proteins. The antigen-combining sites of classic antibodies are composed of two variable domains (VH and VL) 1 located at the N-terminal end of the heavy and light chains (1). Both domains display high sequence diversity, mainly concentrated in three complementarity-determining regions (CDRs) in each V domain. Structurally, these regions form loops, clustered at the N-terminal side of the folded domain, and provide the antigen contacts. Upon antigenic challenge, somatic hypermutation or gene conversion mechanisms further diversify the V sequences as well as the affinities of reactive B-cell receptors. A clonal selection system then enriches for those B-lymphocytes that display receptors with a higher affinity for the antigen. This process of somatic hypermutation and clonal selection is known as antibody (or B-cell receptor) affinity maturation (2)(3)(4).
It has been reported on several occasions that antibody paratopes for haptenic molecules or small peptides lose conformational plasticity upon maturation. Wedemayer et al. (5) showed by x-ray crystallography that the anti-hapten antibody 48G7 acquired mutations that result in the pre-organization of the reactive conformation of a highly flexible germ line paratope, in the mature antibody. The observed conformational changes include large rigid body movements between VH and VL as well as backbone and side-chain rearrangements within the hypervariable CDR loops. Manivel et al. (6) observed that large entropic penalties were associated with antibodies from primary responses binding to a synthetic peptide. These penalties were significantly reduced upon maturation. In sharp contrast, a detailed analysis of the effects of the adaptations introduced during the affinity maturation of paratopes against larger antigens remains elusive. However, by comparing HyHEL10 variants at different maturation stages, Li et al. (7) indicated that the antigen maturation for this antibody results in an increased complementarity involving a fraction of the apolar buried surface area of the paratope but not by its pre-organization. * This work was supported by the Vlaams Interuniversitair Instituut voor Biotechnologie, the Onderzoeksraad of the Vrije Universiteit Brussel, and the Fonds voor Wetenschappelijk Onderzoek Vlaanderen. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The Camelidae possess, in addition to classic IgG molecules, a large fraction of naturally occurring heavy-chain antibodies that are devoid of light chains (8,9). The variable domain of camel heavy-chain antibodies (VHH) is assembled by V-D-J gene rearrangement but is distinct from the VH of classic antibodies. Different sets of germ line VH and VHH genes encode the V-germ line repertoire in the camelid genome, whereas the D and J genes are shared between both antibody classes. The affinity maturation process in camelid antibodies remains unexplored apart from the comparisons between cDNA and germ line V-gene segments that revealed the presence of a somatic hypermutation mechanism in camel heavychain antibodies (10).
Due to the absence of the VL domain in camelid heavy-chain antibodies, it is evident that inter-domain movements cannot contribute to the flexibility of the paratope. In addition, the frequent occurrence of interloop disulfide bridges and the high percentage of backbone atoms involved in antigen contacts (8) may considerably reduce the paratope flexibility. Therefore, the occurrence of a rigidification process of the paratope in heavy-chain antibodies during affinity maturation, as observed for classic antibodies, remains an open question.
In this report we constructed germ line revertants of the hen egg white lysozyme (HEWL)-specific VHH, cAb-Lys3 (11,12). The matured VHH and the germ line revertants were subsequently subjected to a transition state analysis using surface plasmon resonance technology. The enthalpic and entropic contributions in the cAb-Lys3-HEWL association were determined to evaluate the plasticity and interaction strengths of the interface residues during the course of affinity maturation. In addition, the crystal structure of the CDR2 loop germ line revertant was solved to provide a structural understanding of the biophysical parameters of the affinity maturation process.

MATERIALS AND METHODS
Reagents-All reagents were analytical grade. HEWL was purchased from Roche Applied Science. PDEA, EDC, NHS, ethanolamine, and 1,4-dithioerythritol were obtained from Biacore AB, and cystamine was obtained from Sigma Aldrich. Restriction enzymes were purchased from Invitrogen.
Design and Construction of the Germ Line Revertants-We chose to evaluate the maturation event by restoring the complete or separate CDR1 and CDR2 germ line sequences of cAb-Lys3. This "cassette" reversion was chosen over the evaluation of individual mutations, because the sequential order of the mutations in the antigen maturation event cannot be traced, and some combinations of amino acid mutations might impose unacceptable loop foldings. The germ line revertants are referred to as cAb-Lys3 gl2 , cAb-Lys3 gl1 , and cAb-Lys3 gl1&2 , for the reversion of the CDR2, CDR1, and the complete germ line sequences, respectively. The antibody carrying the affinity matured sequence, is referred to as cAb-Lys3 m .
The cAb-Lys3 gl2 , cAb-Lys3 gl1 , and cAb-Lys3 gl1&2 were all generated by PCR using mutagenic primers. The cAb-Lys3 gene cloned into a pHEN4 vector (13) was used as template for the PCR reactions. The cAb-Lys3 gl2 mutant was generated by PCR using a mutagenic primer P1 (5Ј-GCA-GCA-ATT-AAT-AGT-GGT-GGT-GGT-AGC-ACA-TAC-3Ј, an AseI restriction site is underlined) annealing at the end of the framework 2 up to codon 58 of CDR2 in the cAb-Lys3 sequence and the primer AM006 (5Ј-CGT-TAG-TAA-ATG-AAT-TTT-CTG-TAT-GAG-G-3Ј) annealing in the vector sequence downstream of the cAb-Lys3 gene. An AseI generated fragment of cAb-Lys3 m -pHEN4 that contained the 5Ј-end of cAb-lys3 and part of the pHEN4 vector sequence was annealed with the AseI-digested PCR product and further amplified with universal reverse primer and AM006. The amplification product was then digested with HindIII and BstEII, ligated into a HindIII/BstEII double digested cAb-Lys3-pHEN4 vector and transformed into Escherichia coli.
The cAb-Lys3 gl1 reversion mutant was constructed using splicing by overlap extension from two PCR fragments with a 15-nucleotide overlap in their CDR1 coding areas. One fragment resulted from a PCR with reverse primer annealing in the pHEN4 vector sequence and a mutagenic forward primer SV2 (5Ј-A-GCT-ACT-GTA-GGT-GTA-TCC-TGA-G-GC-TG-3Ј) annealing in the CDR1 area of cAb-Lys3 m (the Bsu36I site, introduced for screening purposes is underlined). The second fragment resulted from amplification between a universal forward primer annealing in the pHEN4 vector and a mutagenic primer SV1 (5Ј-AC-ACC-TAC-AGT-AGC-TAC-TGT-ATG-GG-3Ј) annealing in the CDR1 codons of cAb-lys3 m . The spliced fragments were amplified with reverse primer and forward primer, digested with HindIII and BstEII, and ligated in a HindIII/BstEII double digested cAb-Lys3 m -pHEN4 vector (13) and transformed into E. coli. The strategy to construct cAb-Lys3 gl1&2 was similar to that of cAb-Lys3 gl1 except that the cAb-Lys3 gl2 -pHEN4 was used as template.
Finally, all mutants were recloned into a cAb-Lys3-pHEN6 vector (containing a C-terminal His 6 -encoded tag) (14) by digestion with HindIII and BstEII, gel purification and ligation in a HindIII/BstEII double digested cAb-Lys3-pHEN6 vector and transformation into E. coli. All mutant genes were confirmed by nucleotide sequencing.
Expression and Purification-Periplasmic expression, extraction, and purification by immobilized metal affinity chromatography and gel filtration of mutants and cAb-Lys3 m were performed according to Lauwereys et al. (14). The concentration of the mutants and wild type cAb-Lys3 was determined spectrophotometrically at 280 nm using the calculated sequence based extinction coefficient (15), which is equal to 27,220 M Ϫ1 cm Ϫ1 for the cAb-Lys3 variants: cAb-Lys3 m and cAb-Lys3 gl2 , or 25,730 M Ϫ1 cm Ϫ1 for the cAb-Lys3 variants: cAb-Lys3 gl1 and cAb-Lys3 gl1&2 .
Immobilization of PDEA-modified Lysozyme-Lysozyme was modified with PDEA (2-(2-pyrinidyldithio)ethaneamine) to introduce reactive disulfide groups before immobilization via surface thiol coupling, as described in the BIAapplications handbook. This immobilization method was chosen based on the closest approximation to a 1:1 binding model for the binding of mutants and cAb-Lys3 m to the immobilized ligand (16). PDEA-modified lysozyme dissolved in 10 mM NaAc buffer, pH 5.5, was coupled to CM5 chips at a rate of 5 l/min. The first flow cell was incubated with 35 l of EDC/NHS and subsequently with 35 l of 1 M ethanolamine, pH 8.5, and served as a reference surface for the kinetic measurements. The second flow cell (fc2) was used to immobilize PDEA-modified lysozyme at 200 resonance units. This was performed by treating fc2 with 10 l of EDC/NHS and subsequently with 15 l of cystamine dihydrochloride, followed by 15 l of 1,4-dithioerythritol. Manual injection of the PDEA-modified lysozyme was then performed until the aimed immobilization level was obtained. Excess of reactive thiols on the chip were deactivated by 20 l of 20 mMPDEA-1 M NaCl.
Measurements/Fittings/Calculations-A Biacore® 3000 instrument was used to obtain binding of cAb-Lys3 m and the germ line mutants at 20, 25, 30, and 35°C, at 30 l/min, with PBS supplemented with 3 mM EDTA and 0.005% surfactant P20 as running buffer. Replicate runs were performed on a different chip, which made the measurements independent. Each VHH was measured at four protein concentrations ranging from 12 M to 31 nM, depending on the affinity of the mutant for HEWL. The mutants were injected for 2 min, and dissociation data were collected during 10 min. Regeneration after each cycle was performed using 10 mM glycine-HCl, pH 1.0, for 2 min. Control injections using 1500 nM cAb-Lys3 m in PBS was performed before and after five kinetic cycles to confirm that the activity of the immobilized HEWL did not change over time. Zero concentration data (injection of buffer alone) was always subtracted from the sensorgrams before fitting. Kinetic parameters were obtained by global fitting of the sensorgrams to a 1:1 model using BIAevaluation 3.1 software. The R max parameter, representing the binding capacity of the chip, was fixed to the average equilibrium response of the control injections.
Calculation of the Transition State Enthalpy and Entropy-The enthalpy and entropy of equilibrium and activation for the interaction of each cAb-Lys3 variant was estimated using the equations: ⌬G°ϭ RT where R is the Rydberg gas constant; the Boltzmann constant; T the absolute temperature; k a (M Ϫ1 s Ϫ1 ) and k d (s Ϫ1 ), respectively, the association and dissociation rate constant of binding; K D (M) the dissociation constant; ⌬G°, ⌬ ‡ G ass°, and ⌬ ‡ G diss°a re the standard Gibbs free energies of binding, association, and dissociation, respectively. By plotting ln(k d h/T) versus 1/T and ln(k a h/T) versus 1/T, better known as Eyring plots, we obtain from the slopes -⌬ ‡ H diss°/ R and -⌬ ‡ H ass°/ R and from the intercepts ⌬ ‡ S diss°/ R and ⌬ ‡ S ass°/ R, respectively (17).
Crystallization and Structure Solution of the cAb-Lys3 gl2 Revertant-Crystals of cAb-Lys3 gl2 in complex with HEWL were grown at 4 mg/ml by hanging drop vapor diffusion with 2 M sodium formate and 100 mM sodium HEPES, pH 7.5, as precipitant solution. The crystal used for data collection was frozen after transfer into the precipitant solution supplemented with 20% glycerol. Synchrotron data for the crystals were obtained at the BW7B beamline of EMBL synchrotron facility at DESY (Hamburg, Germany) (project number: PX-02-16), using a Marresearch image plate detector. Data were processed to 1.5-Å using DENZO and Scalepack (18). The AMORE program (19) generated a molecular replacement solution using the wild type cAb-Lys3-HEWL complex (pdb code: 1JTT) (11) as a model. CNS (20) was used to refine the structure. Data collection, space group, and unit cell dimensions as well as the refinement statistics are shown in Table I. The coordinates and structure factors have been deposited in the Protein Data bank at Research Collaboratory for Structural Bioinformatics (www.rcsb.org/pdb/) as entry 1XFP.
Structural Analysis and Preparation of Figures-The program Contact from the CCP4 software package identified inter-molecular contacts, and the protein interaction server (www.biochem.ucl.ac.uk/bsm/ PP/server) (33) determined changes in solvent-accessible surface areas. Figures were prepared using the program PyMOL. 2

RESULTS
Ontogeny and Somatic Hypermutation of cAb-Lys3-A search in the VHH germ line data base (10) revealed that cAb-Lys3 originated most probably from the cvhhp11 germ line gene (AJ245159, 93.7% identity). From the known camelid J sequences 3 JH3 was identified as the most likely recombined J segment to arrive at the cAb-Lys3 gene. The original D gene used to generate the CDR3 of cAb-Lys3 remains elusive at this stage, because the D gene data base of camelids is not yet available.
The generation of the cAb-Lys3 sequence required 18 nucleotide substitutions in the cvhhp11 germ line gene. Approximately two-thirds of these substitutions are located in the CDR areas (Fig. 1). In the coding strand, the purines are more frequently mutated than pyrimidines (Table II). Such purine/ pyrimidine bias has also been reported for the somatic hypermutation in mouse and human VHs and reflects the strand preference for mutation (2). In sharp contrast, the transition versus transversion mutation bias, which equals 0.5 in mouse VHs is significantly different from the 0.33 value found for the cvhhp11 to cAb-Lys3 substitutions (2).
The "mutation hotspot" (Pu-G-PY) sequences are very susceptible to substitution during the somatic maturation process (2). The cvhhp11 germ line gene contains nine of these mutation hotspots in the CDR, and fourteen occur in the framework regions. Four mutation hotspots of the CDR area are actually modified in the cAb-Lys3 gene. Remarkably, only two hotspots residing within the area encoding the framework regions are altered.
The alignment of the cvhhp11 and cAb-Lys3 amino acid sequences revealed seven amino acid replacements. Five of these replacements are located in the CDRs, of which three (Tyr-29 3 Ile, Ser-30 3 Gly, and Ser-31 3 Pro) occur in the CDR1 and two (Ser-52a 3 Met and Ser-56 3 Ile) are located in the CDR2 (Figs. 1 and 2A).
Localization of Substituted Residues in the VHH Paratope Structure-All five CDR residues that were replaced during affinity maturation of cAb-Lys3, cluster in the folded domain and are located at the periphery of the VHH-HEWL interface (Fig. 2B). The change in solvent-accessible surface area (⌬ASA) of the "matured amino " acids Ile-29, Gly-30, Pro-31, Met-52a, and Ile-56 upon antigen binding equals 122 Å 2 . This represents nearly 50% of the total ⌬ASA of the V-germ line encoded part of the paratope. Remarkably, out of these five amino acids only Ile-29 and Ile-56 have atoms within a distance of 4.0 Å of the antigen (Fig. 2C). Ile-29 contacts the HEWL side chains of Trp-62 L and Leu-75 L , and Ile-56 has atoms within 4 Å of Asp-48 L and Thr-47 L .
In contrast to the Ile-29 and Ile-56, the remaining three matured amino acids do not participate in direct antigen recognition and seem to have rather a structural role. The and angles adopted by Gly-30 allow Pro-31 (in a cis-configuration) to pack against Tyr-27 and Met-34, thereby forming a stable hydrophobic core (Fig. 2C). Residue Met-52a of the CDR2 loop packs against Met-34 and Pro-31 (Fig. 2C), thereby extending the hydrophobic core of the CDR1 loop.
Temperature Dependence of the VHH-HEWL Kinetics and Affinity-The kinetic rate constants k a and k d of the interactions between HEWL and the cAb-Lys3 variants were determined at four different temperatures (20,25,30, and 35°C) by surface plasmon resonance (Fig. 3). All binding curves fitted well to a simple 1:1 "lock and key" binding model, consistent with recognition of two rigid molecules deprived from large conformational changes upon binding (22). These data are graphically represented in a RaPID plot (Fig. 3), which is a convenient two-dimensional display of the k a and k d values of each interacting pair. The diagonal lines of the plot reflect the iso-affinity interactions. A glance at this plot immediately reveals that the largest gain during affinity maturation originates from the improvement of k d , whereas the k a constant was only moderately affected (at most by a factor 4) during maturation from germ line to cAb-Lys3 m . Therefore, the observed ⌬⌬G°values for the reversion to the CDR2 (cAb-Lys3 gl2 ), CDR1 (cAb-Lys3 gl1 ), and to the complete V-germ line sequence (cAb-Lys3 gl1&2 ), which is, respectively, equal to 1.1, 2.6, and 3.5 kcal/mol, are primarily determined by differences in off-rates.
The ⌬H and T⌬S Contributions to Activation, Dissociation, and Equilibrium Free Energies-Eyring plots derived from the kinetic association and dissociation data are linearly decreasing (R 2 of more than 0.95) for all cAb-Lys3 variants (Fig. 4, A  and B). From the slopes and intercepts of these Eyring plots we retrieve the activation enthalpy and entropy contributions to the association free energy (⌬ ‡ G ass°) , dissociation free energy (⌬ ‡ G diss°) and the free energy of binding (⌬G°) for each cAb-Lys3 variant (Fig. 4, C-E).
The formation of the VHH-HEWL transition state for each cAb-Lys3 variant, has a large entropy and enthalpy penalty (Fig. 4F). The enthalpy barriers, however, are significantly higher than those of the entropy, except for the cAb-Lys3 gl1 -HEWL association, where both energies are equally unfavorable. The entropy barriers in the cAb-Lys3 gl1&2 -HEWL and cAb-Lys3 m -HEWL associations are equal. The slight difference in transition state energy (⌬ ‡ G ass°ϭ ⌬ ‡ H ass°Ϫ T⌬ ‡ S ass°) between both variants originates from a 0.6-kcal/mole difference in enthalpy penalty in favor of the cAb-Lys3 m -HEWL transition state. Conversely, the HEWL interactions with the intermediate matured variants, cAb-Lys3 gl1 and cAb-Lys3 gl2 , show significant differences in both ⌬ ‡ H ass°a nd T⌬ ‡ S ass°. However, the overall transition state energy (⌬ ‡ G ass°ϭ ⌬ ‡ H ass°Ϫ T⌬ ‡ S ass°) for these variants is almost equal to those of cAb-Lys3 m and cAb-Lys3 gl1&2 . High enthalpy and entropy barriers (⌬ ‡ H diss°a nd T⌬ ‡ S diss°) (Fig. 4D) are also observed upon dissociation of the VHH-HEWL complexes for all variants. Here again, the enthalpy barriers are significantly higher than those of the entropy. These enthalpy barriers of dissociation increase slightly upon maturation (⌬⌬ ‡ H diss°ϭ 1.4 kcal/mol for cAb-Lys3 gl1&2 3 cAb-Lys3 m ). The entropy penalties (T⌬ ‡ S diss°) in the dissociation event differ between cAb-Lys3 gl1&2 and cAb-Lys3 m by 1.0 kcal/mole.
Enthalpy and entropy contributions to the ⌬G°of the equilibrium binding reaction of each mutant to HEWL are shown in Fig. 4E. For all variants, except for the cAb-Lys3 gl1 -HEWL interaction, which has a slightly negative T⌬S°, favorable enthalpy, and entropy contributions to ⌬G°are observed. As can be deduced from the association and dissociation phases, the enthalpy contributions are higher than those of the entropy in the VHH-HEWL equilibrium reactions for all cAb-Lys3 variants. The difference of 3.5 kcal/mole in ⌬G°between cAb-Lys3 gl1&2 and cAb-Lys3 m is the result of a 2.1 and 1.1 kcal/mole difference in ⌬H°and T⌬S°, respectively. The cAb-Lys3 gl1 has the most favorable enthalpy component but is countered by an unfavorable entropy penalty. The most favorable binding entropy is observed in the cAb-Lys3 gl2 -HEWL equilibrium reaction. The T⌬S°term differs from cAb-Lys3 m by 0.7 kcal/mole for this mutant. The binding enthalpy for cAb-Lys3 gl2 is, however, 1.8 kcal/mole less favorable than in the cAb-Lys3 m -HEWL interaction.
In summary, the difference in binding energy (⌬G°) of 3.5 kcal/mole between cAb-Lys3 gl1&2 and cAb-Lys3 m is primarily due to an increase in favorable enthalpy (2.1 kcal/mole) (Fig. 4, E and F). This increased favorable enthalpy is essentially obtained in the stabilization phase of the transition state complex into the final VHH-HEWL hetero-dimer (1.4 kcal/mole). The difference in T⌬S°between both mutants is equal to 1.1 kcal/ mole originating exclusively from the stabilization phase. In the association event, the on-rate constants as well as the activation energies (⌬ ‡ G ass°) do not differ much among the variants, in contrast to the relative enthalpy and entropy activation barriers for the cAb-Lys3 gl2 -HEWL and cAb-Lys3 gl1 -HEWL association reactions.
Crystal Structure of the cAb-Lys3 gl2 Variant in Complex with HEWL-Crystallizations of the maturation variants in complex with HEWL were undertaken to clarify the thermodynamic and kinetic data. The cAb-Lys3 gl1 and cAb-Lys3 gl1&2 variants in complex with HEWL resist crystallization, possibly because matured CDR1 sequences are critical for the crystal lattice contacts as shown for the matured cAb-Lys3 in complex with HEWL (11). The cAb-Lys3 gl2 -HEWL complex was successfully crystallized, and the superposition of the cAb-Lys3 m and cAb-Lys3 gl2 structures revealed that all backbone atoms occupy identical positions in both complexes. Furthermore, the C␣, C␤, and O␥ atoms of Ser-52a and Ser-56 in the germ line variant possess a structurally equivalent position to the C␣, C␤, and C␥ atoms of the matured-type residues Met-52a and Ile-56.  The absence of atoms due to the Met-52a 3 Ser substitution is compensated by the insertion of two buried water molecules between CDR1 and CDR2. The hydrogen bond network created by these two buried water molecules and the O␥ of Ser-52a and the backbone carbonyl oxygen of residue Pro-31 in cAb-Lys3 gl2 (Fig. 5A), is replaced in cAb-Lys3 m by the Met-52a side-chain that contacts the CDR1 (Fig. 5B).
In contrast to the Met-52a that does not contact HEWL, the other maturated amino acid of the CDR2, Ile-56, is involved in extensive interactions with HEWL (Fig. 2C). The germ line encoded Ser-56 side chain of the cAb-Lys3 gl2 variant is also engaged in HEWL interactions, e.g. the O␥ of Ser-56 is hydrogen bonded to the O␥1 of Thr-47 L (not shown).

DISCUSSION
In this report we analyzed the functional and structural effects of the affinity maturation event of cAb-Lys3, the antigen-binding domain of a camelid heavy chain antibody against HEWL (11,12). Sequence alignment with the dromedary VHH germ line genes revealed that cAb-Lys3 emerged from a recombination with the cvhhp11 germ line gene. Both gene fragments share a critical Cys codon at position 33 (Fig. 1) and a high sequence identity of 93.7%.
Somatic hypermutations in mouse antibodies are biased toward nucleotide transitions (2). Such a bias was not observed between the cvhhp11 germ line and cAb-Lys3 sequence, possibly because the mutations were under a high antigen selection pressure. However, an observed 1/1 transition/transversion bias for the mutations in other VHHs 4 suggests that a conventional genetic mechanism is employed for in vivo somatic hypermutation of camelid heavy chain antibodies.
The replacement mutations in the CDR loops occur at the paratope periphery of the VHH-HEWL interface (Fig. 2B). The diversification at the periphery of the paratope during the affinity maturation is a characteristic feature for the somatic hypermutation and selection process (23). Due to the solventexposed environment of this region, mutagenesis of these residues is believed to have only a moderate and non-cooperative (additive) effect on the overall binding energy (24). Consequently, a typical 10-to 100-fold increase in binding affinity is accomplished by affinity maturation. However, larger affinity differences between germ line and mature antibodies require alternative mechanisms. For instance in the case of the antibody 48G7 and 48G7 g (the germ line precursor), where a 30,000-fold difference in affinity is observed, the effects of the amino acid replacements are highly co-operative with an important influence on the structural isomerism of the antibody (5).
The construction of cAb-Lys3 variants, where the CDR1 and CDR2 were reverted simultaneously or separately to the original V germ line counterparts, allowed us to assess the effects of maturation of the CDR loop structure and on HEWL binding. All association and dissociation reactions fitted well to a simple bimolecular "lock-and-key" reaction model. The VHH-HEWL interactions for the cAb-Lys3 variants differed significantly in off-rate, whereas their on-rates were almost invariant. The 300-fold difference in affinity between the germ line precursor FIG. 2. Structural localization of the amino acid replacements within the paratope of cAb-Lys3. A, residue replacements from the cvhhp11 germ line gene to cAb-Lys3 m . B, antigen eye-view of the molecular surface of the cAb-Lys3 paratope (colored region). The residues of CDR1 and CDR2, which were substituted during affinity maturation, are shown in green and cyan, respectively. The remaining solventinaccessible area upon antigen complexing is colored in pink. C, detailed view of the CDR1 (green) and CDR2 (cyan) region of cAb-Lys3 (pdb code: 1JTT), showing all atoms of the mutated residues in sticks (atoms are color-coded: carbon, green (CDR1) or cyan (CDR2); nitrogen, blue; oxygen, red; sulfur, yellow). The residues of HEWL (blue) interacting with Ile-29A and Ile-56A of cAb-Lys3 are also shown in sticks (atoms are color-coded: carbon, blue; nitrogen, blue; oxygen, red). Only backbone atoms (except for carbonyl oxygens) of the surrounding HEWL and VHH residues are shown.  4. Panels A and B are the Eyring plots for, respectively, the association and dissociation reaction. The following symbols were used to discriminate between variants: f, for cAb-Lys3 m ; OE, for the cAb-Lys3 gl2 mutant; , for cAb-Lys3 gl1 ; and •, for cAb-Lys3 gl1&2 . C-E, the T⌬ ‡ S°(gray bars) and ⌬ ‡ H°(white bars) contributions to the ⌬ ‡ G°of association (C), dissociation (D), and T⌬S°(gray bars) and ⌬H°(white bars) contributions to the ⌬G°of equilibrium binding (E) for each cAb-Lys3 variant (indicated below the bars) with HEWL at 35°C and in the 1 M standard state. F, reaction pathway of the cAb-Lys3 gl1&2 -HEWL and cAb-Lys3 m reaction. The ⌬G°, ϪT⌬S°, and ⌬H°along the reaction coordinate are given for the cAb-Lys3 gl1&2 -HEWL reaction (left) and for the cAb-Lys3 m reaction (right). and the matured form of cAb-Lys3 therefore arises primarily from a decreased k d value. In addition, no detectable co-operativity between the CDR loops upon maturation was observed. Indeed, for the antigen complexation, the ⌬⌬G°values of cAb-Lys3 m versus cAb-Lys3 gl1&2 equaled the sum of ⌬⌬G°m -gl1 and ⌬⌬G°m -gl2 . These results are consistent with association reactions that lack large conformational changes upon binding. This is further corroborated by the fact that the enthalpy dominates the association, dissociation, and equilibrium energies of the VHH-HEWL interaction of all variants, whereas large entropy penalties are typical for induced fit mechanisms or isomeric equilibrium reactions (25). The difference in binding energy (⌬G°) of 3.5 kcal/mole between cAb-Lys3 gl1&2 or cAb-Lys3 m and HEWL is primarily due to a 2.1 kcal/mole favorable enthalpy increase essentially obtained from the stabilization phase (1.4 kcal/mole). The T⌬S°difference of 1.1 kcal/mole between both variants for HEWL originates exclusively from the stabilization phase.
The absence of pronounced conformational changes in the binding of the cAb-Lys3 germ line variants to HEWL is surprising. It was anticipated from the cAb-Lys3-HEWL crystal structure that three targeted residues (Ser-30 3 Gly, Ser-31 3 Pro, and Ser-52a 3 Met) would fix an otherwise flexible CDR1 loop. In the matured antibody, Met-52a belonging to the CDR2 loop packs against Met-34 of the CDR1 loop, thereby extending the hydrophobic core of CDR1. However, the replacement of Met-52a by a more hydrophilic Ser as encoded by the V-germ line sequence is neither detrimental for antigen recognition nor for the stabilization of the CDR1 and CDR2 loop structures (Figs. 4, C-E, and 5). The crystal structures proved that the interaction between Met-52a and Met-34 in the mature antibody is effectively compensated in the cAb-Lys3 gl2 -HEWL structure by the insertion of two water molecules mediating a hydrogen bond network. This intricate H-bond cluster further involves the O␥ of Ser-52a and the main-chain carbonyl oxygen atoms of Pro-31, Asn-72, and Thr-78.
In the mature cAb-Lys3, the Gly-30 peptide bond allows Pro-31 to pack tightly against Tyr-27 and Met-34, thereby forming a stable hydrophobic core that stabilizes the CDR1 loop conformation. This loop configuration, referred to as canonical loop structure type 4, has never been observed in human or mouse VHs (26). In the germ line antibody, serine residues occupy positions 30 and 31 that were expected to enforce a classic type I canonical structure (27) for the H1 loop of cAb-Lys3 gl1 or cAb-Lys3 gl1&2 mutants whereby Tyr-29 would be at the interior of the loop (instead of Pro-31 in cAb-Lys3 m ) and Ser-31 side chain toward the solvent (Fig. 6B). However, adopting a type I canonical loop structure implies that the Tyr-32 leaves the aromatic core with Tyr-99 and Tyr-100b of the CDR3 loop, a critical element to stabilize this loop in cAb-Lys3 m (Fig. 6). In addition, major structural rearrangements within the antibody and the antigen should be introduced to avoid steric clashes and to arrive at a productive HEWL recognition. Such conformational changes would likely be reflected in a slower on-rate for the cAb-Lys3 gl1&2 mutant compared with the mature antibody, and larger entropy penalties upon binding, which are not observed. It is therefore conceivable that the CDR1 adopts a canonical type 4 structure even for the germ line variants. We expect that the atoms of Ser-31 in the germ line precursor are located at equivalent positions in the hydrophobic core of the CDR1 loop as the Pro-31 residue in the matured cAb-Lys3. The formation of a hydrogen bond with Ser-52a (Fig. 5A) and the two buried water molecules would be sufficient to replace the packing contacts of Pro-31 in the hydrophobic core in the CDR1 loop of cAb-Lys3 m . However, a difference of 1 kcal/mole for the dissociation entropy between cAb-Lys3 m and cAb-Lys3 gl1&2 suggests that the CDR loops of the germ line variant exhibit a minor difference in flexibility.
This structural model for cAb-Lys3 gl1&2 is compatible with the enthalpy and entropy contributions in the association phase of both single loop reversion mutants (Fig. 4C). In the cAb-Lys3 gl1 mutant Ser-31 would be sandwiched between residues Met-52a and Tyr-27. This might destabilize the CDR1 loop by introduction of a hydrophilic residue in the hydrophobic core region, which is not compensated by a favorable hydrogen bond network as observed in the cAb-Lys3 gl2 -HEWL complex. Such a destabilization might therefore cause flexible transitions of the loop conformation, which become fixed upon HEWL binding. A higher entropy barrier during transition state formation compared with the other cAb-Lys3 variants should reflect this event. The cAb-Lys3 gl2 -HEWL variant displays the lowest transition state entropy barrier of all the cAb-Lys3 variants. This suggests that the introduced mutations have a stabilizing effect on the reactive paratope conformation. The observed hydrogen bonding network between the CDR1 and CDR2 loop might account for this effect. In both mutants, the differences in enthalpy barriers may originate from differences in solvation potential (28).
Our results show that the affinity maturation in the CDR1 and CDR2 loop of a VHH focuses on the annealing step of the initial encounter complex. These results are opposite to those made by Manivel et al. (6) for panels of antibodies from primary and secondary responses to a peptide antigen. These authors showed that the antibody maturation influences primordially the antigen association phase. In their analysis, the germ line antibodies displayed large entropic penalties in the association reaction, which were eliminated upon maturation. Such results are consistent with a rigidification process of the germ line paratope upon maturation. In the case of the anti-hapten antibody 48G7, maturation was achieved through the reduction of the off-rate (5). Furthermore, the germ line precursor of this antibody had to overcome a 30,000-fold difference in affinity, which was obtained by energetically highly co-operative mutations (29). These mutations resulted in a pre-organization of the reactive conformation of a flexible germ line paratope in the mature antibody. Although the affinity maturation of this antibody is essentially accomplished by reducing the off-rate, it has been suggested that the dissociation event should reveal a large entropic penalty difference between germ line precursor and mature antibody (30).
Two explanations can be proposed for the lack of congruence between the maturation mechanism of cAb-Lys3 and that of conventional antibodies. In a first possibility it can be argued that the maturation mechanism of antibodies binding to large, proteinaceous antigens may deviate fundamentally from those that are specific for small haptens or peptides. To our knowledge, this is the first time that a transition state analysis of antibody maturation was performed for an antibody specific for a protein antigen. However, Li et al. (7) showed for HyHEL10 variants, representing different maturation stages, that affinity maturation of these antibodies binding to HEWL proceeded through the burial of a higher proportion of apolar paratope surface area and by increasing epitope-paratope complementarity. Comparison of the antibody complexes to the free structures of HyHEL63 crystallized in different space groups, showed minor conformational differences. Furthermore, no correlation was found between the amount of buried surface area or number of contacts and the observed affinity difference, albeit that such a correlation has been observed in the comparison of the structures of the antibodies D.44.1 and the more matured F10.6.6 in complex with HEWL (31). However, one should note that all these antibodies were already considerably matured and therefore might not represent early events in the maturation pathway.
A second explanation for the different maturation mechanism between camelid heavy chain and conventional antibodies is based on the absence of a VL partner in the paratope of the former immunoglobulins. An antigen binding entity consisting of a single domain only will obviously lack the VH-VL flexibil-FIG. 6. The proposed structural model of the cAb-Lys3 gl1&2 mutant. Except when indicated all side-chain atoms and carbonyl oxygen of the backbone were removed for clarity. A, stereo representation of a cAb-Lys3 gl1&2 structural model in which the CDR1 loop and CDR2 loop have identical backbone conformations as seen in the cAb-Lys3 gl2 and cAb-Lys3 m crystal structures. All backbone and side-chain atoms of the CDR1 loop are shown (carbon, green; oxygen, red; nitrogen, blue; sulfur, yellow). In addition, backbone and side chain atoms of Ser-52a, Ser-56 (carbon, cyan; nitrogen, blue; oxygen, red; sulfur, yellow), and the mainchain and side-chain atoms of Tyr-99A and Tyr-100bA (magenta) are represented as sticks, and the corresponding residues are labeled. Residues Tyr-27A, Tyr-29A, Ser-31A, and Met-34A are additionally labeled. B, stereo representation of a cAb-Lys3 gl1&2 structural model in which the CDR1 loop adopts a type I canonical structure. Color coding, displayed atoms, and labels are identical to those in A. Atoms of the VHH molecule within 2 Å of atoms of the CDR1 loop are represented as purple spheres.
ity, which will inevitably reduce the plasticity of the germ line paratope. A germ line-encoded cysteine forming a disulfide bond with a cysteine in the CDR3 loop might add considerable rigidity to the CDR loops. The CDR3 loop of VHHs, which is on average longer than those found in VHs, often folds back onto the framework 2 region of the ␤-barrel where it is stabilized by the formation of a hydrophobic core (32). Furthermore, a high proportion of VHH main-chain atoms are involved in the VHHantigen recognition, which reduces the contribution of side chain immobilization to the binding entropy. Therefore, it is likely that the VHH-antigen affinity has to be matured by adding favorable contacts, and by increasing complementarity and percentage of apolar buried surface area. As can be calculated from the proposed structure of cAb-Lys3 gl1&2 in complex with HEWL, the somatic hypermutations do not change the total amount of ⌬ASA (219 Å 2 compared with 221 Å 2 in the matured antibody). However, an increase in the fraction of apolar surface area is evident, as the polar residues in the germ line antibody are replaced by hydrophobic residues in the mature antibody. This observation is consistent with the results of Li et al. (7) for the affinity maturation variants of the HEWLspecific HyHEL8 antibody.