doi:10.1006/jmbi.2001.5075 available online at http://www.idealibrary.com on
J. Mol. Biol. (2001) 313, 473±478
COMMUNICATION
Degenerate Interfaces in Antigen-Antibody Complexes K. Decanniere*, T. R. Transue, A. Desmyter, D. Maes, S. Muyldermans and L. Wyns Vrije Universiteit Brussel Dienst Ultrastructuur, Vlaams Instituut voor Biotechnologie Paardenstraat 65, B-1640 St.-Genesius Rode, Belgium
In most of the work dealing with the analysis of protein-protein interfaces, a single X-ray structure is available or selected, and implicitly it is assumed that this structure corresponds to the optimal complex for this pair of proteins. However, we have found a degenerate interface in a high-af®nity antibody-antigen complex: the two independent complexes of the camel variable domain antibody fragment cAb-Lys3 and its antigen hen egg white lysozyme present in the asymmetric unit of our crystals show a difference in relative orientation between antibody and antigen, leading to important differences at the protein-protein interface. A third cAb-Lys3-hen lysozyme complex in a different crystal form adopts yet another relative orientation. Our results show that proteinprotein interface characteristics can vary signi®cantly between different specimens of the same high-af®nity antibody-protein antigen complex. Consideration should be given to this type of observation when trying to establish general protein-protein interface characteristics. # 2001 Academic Press
*Corresponding author
Keywords: immunoglobulin; heavy chain antibody; VHH; interface; antigen binding
Interfaces Understanding the properties of protein-protein interaction surfaces is an important issue in protein structure and function research. Jones & Thornton compared the interface involved in protein-protein interactions with other surface patches of the same protein.1,2 They showed that the shape and chemical nature of protein interfaces are barely different from other surface patches of the same protein. Nevertheless, subtle tendencies can be observed if the nature of the complex is taken into account, e.g. constitutive multimeric proteins tend to be slightly more hydrophobic at their interface. This correlation between type of complex and interface properties was con®rmed by Lo Conte et al.3 An important concept for understanding interfaces de®ned as early as 1945 is the requirement for shape complementarity.4 Assessment of shape complementarity can be conducted in various Present address: T. R. Transue, Laboratory of Structural Biology, MD F3-05, Niehs, NIH. PO Box 12233, Research Triangle Park, NC 27709, USA. Abbreviations used: CDR, complementary determining region; PDB, Protein Data Bank. E-mail address of the corresponding author:
[email protected] 0022-2836/01/030473±6 $35.00/0
ways. One way is the Sc parameter introduced by Lawrence & Colman.5 With Sc-values between 0.64 and 0.68, antigen-antibody interfaces were shown to be less complementary than other types of protein-protein complexes. This conclusion was con®rmed using the ``gap volume'' to measure shape complementarity.2 Shape complementarity can also be assessed by calculating Voronoi volumes around interface atoms: the volume occupied by an atom is limited by its neighbours and can be calculated using a triangulation algorithm.3,6,7 In the case of perfect shape complementarity, the volume occupied by interface atoms is as small as the volume occupied by similar atoms situated in the core of the protein (the atoms have a high packing density). Using this method, differences in shape complementarity between antibody-antigen complexes and other types of complexes are not apparent.3 Besides shape complementarity, speci®c binding requires complementary patterns of surface polar groups.8 Moreover, a substantial fraction of the free energy gain upon complexation comes from the burial of hydrophobic residues.9,10 The relative importance of these types of interactions depends on the nature of the complex. The biological potential of imperfect interfaces should also be considered. The interface of a T-cell receptor and a self peptide has been described to # 2001 Academic Press
474
Degenerate Interfaces in Antigen-Antibody Complexes
be imperfect, with an Sc-value of 0.45 and large conformational changes upon binding. These characteristics are linked to the biological function of the system: it was proposed as a mechanism to eliminate auto-reactive T-cells during thymic development and to activate a large number of TCRs with one peptide-MHC complex.11 Here we investigate the interface of an antibodyantigen complex, and show that different specimens of one and the same complex can differ signi®cantly at the interaction surface, even within the same crystal. The cAb-Lys3-lysozyme complex The protein pair under investigation here is the complex of a variable domain of a dromedary heavy-chain antibody (VHH), named cAb-Lys3, with hen egg white lysozyme. Heavy chain homodimers are produced by Camelidae as a major fraction of their naturally circulating antibody molecules.12 Isolated variable domains obtained from these heavy chain antibodies are highly soluble and bind their antigen speci®cally and with high af®nity using up to three complementary determining regions (CDRs). The CDR3-loop of cAb-Lys3 is extremely long (24 amino acid residues). The N-terminal part of this loop (ten residues) inserts deeply into the active site cleft of lysozyme and provides approximately 70 % of the antigen contacting surface. The C-terminal part of the CDR3-loop folds over a surface patch interacting with the VL-chain in normal antibodies, and shields part of this hydrophobic surface patch from the solvent. More details of the cAb-Lys3 and its peculiar binding to lysozyme can be found elsewhere.13,14 The cAb-Lys3-lysozyme complex crystallizes in space group P212121, with two complexes in the asymmetric unit (referred to as hel1 and hel2), and Ê resolution (PDB entry code was solved at 2.5 A 1JTO). rms deviations for the antigen and antibody Ê and 0.59 A Ê , respectively. The separately are 0.57 A rms deviation for the superposition of all mainÊ . Both chain atoms of these two complexes is 1.04 A the rms deviation between the complexes and visual inspection of the superimposed structures (Figure 1(a)) indicate that the hel1 and hel2 complexes differ. These differences are not distributed randomly throughout the complex, as can be seen in Figure 1(b). For this Figure, only the main-chain atoms of the cAb-Lys3 molecules are used for the superposition, with the resulting operator applied also to the lysozyme molecules. This procedure reveals a translational and rotational difference Ê , respectively) in relative orientation (9.5 and 1.3 A of antibody versus antigen between the hel1 and hel2 complexes, with the antibody and antigen molecules behaving as rigid bodies (see the legend to Table 2 below, for the exact method of calculation). We extensively studied the effect of this difference in orientation on the cAb-Lys3-lysozyme protein-protein interface.
Figure 1. Superposition of the hel1 and hel2 complexes: (a) using all main-chain atoms; (b) using mainchain atoms of cAb-Lys3 only; (c) superposition as in (b), comparing the hel1 complex with the hel-C2 complex; (d) superposition as in (b), comparing the tel1 and tel2 complexes. cAb-Lys3 molecules are in grey and dark green, lysozyme is in yellow and blue.
Various parameters used in the literature to characterize protein-protein interfaces were calculated (Table 1). These parameters depend, to a varying degree, on the set of atom radii and the cut-off distances used for the calculations, as exempli®ed by the buried surface area obtained with two different programs (last 2 rows in Table 1), and this should be taken into account when interpreting the results listed. Nevertheless, taken together, these parameters provide a quantitative characterisation of the protein-protein interaction surface. Comparing the hel1 and hel2 complexes, differences of approximately 11 % in buried surface area
475
Degenerate Interfaces in Antigen-Antibody Complexes Table 1. Selected ``global''parameters describing the antibody-antigen interface Planarity of interface Length/breath ratio Interface by polar atoms (%) Interface by non-polar atoms (%) Number of hydrogen bonds Ê 3) Gap volume (A Gap volume index Sc value for complex Ê2 Buried surface area (web) A Ê2 Buried area (NACCESS) A
hel1
hel2
hel-C2
tel1
tel2
3.58 0.84 28.1 71.9 8 2438 1.46 0.705 890 837
3.31 0.97 28.3 71.6 8 3289 2.22 0.639 800 742
3.36 0.98 36.1 63.8 6 3044 2.03 0.654 798 741
3.41 0.72 47.8 52.1 5 2997 1.91 0.666 737 783
3.35 0.71 42.5 57.5 6 2974 1.93 0.640 718 768
Most parameters where obtained from the ``protein-protein interaction server'' (www.biochem.ucl.ac.uk/bsm/PP/server).1,2 The Sc-value was obtained using the Sc program.5 The buried surface area from NACCESS was calculated with version 2.1.1 of the program.23
and 25 % in gap volume were observed, resulting in a variation of the gap volume index between 1.46 (very good shape complementarity as found, e.g. in proteins always occurring as heterodimers) and 2.2 (good shape complementarity as found, e.g. in enzyme-inhibitor complexes), which is better than the average (3.02 0.8) for antibody-antigen complexes.2 This observation is more or less con®rmed with the Sc value varying between 0.70 and 0.64 (very good complementarity as in a proteaseinhibitor complex versus good complementarity typical for antibody-antigen complexes). These numbers invalidate the imperfect interface hypothesis for our cAb-Lys3-lysozyme complex, which is not surprising given that the antibody originates from an immunized camel and has therefore been subjected to af®nity maturation (Kd 11 nM; De Genst et al., unpublished results). Table 1 also lists the interface characteristics for new crystals of the cAb-Lys3-lysozyme complexes we have obtained. The cAb-Lys3-hen egg white lysozyme complex can be crystallized in space group C2, with one complex in the asymmetric unit (referred to as the hel-C2 complex), and data Ê resolution (PDB entry were collected up to 2.1 A code 1JTP). The complex of cAb-Lys3 with turkey egg white lysozyme was crystallized in space group P212121, and is isomorphous to the original hen lysozyme complex (two complexes per asymmetric unit, referred to as tel1 and tel2). Data were Ê (PDB entry code 1JTT). The crysobtained to 1.9 A tallisation conditions for the hel-C2 crystal form and the turkey lysozyme complex crystals are very similar to each other, and differ from the original conditions used to obtain the P212121 hen lyzosyme complex crystals (2 M sodium formate and 0.1 M sodium citrate (pH 5.6) as reservoir solution for the C2 and turkey crystals versus 15-20 % polyethylene glycol 8000 and 0.1 M potassium phosphate solution (pH 6.4) for the original crystals). Some of the parameters describing the interface characteristics (buried surface area, gap volume index, Sc) suggest that the hel-C2 complex interface closely resembles that of the hel2 complex, whereas others (polar/non-polar surface, hydrogen bonds) indicate that the hel-C2 complex interface differs
from both the hel1 and hel2 complex interfaces (Table 1, columns 1, 2 and 3). The orientation of antigen versus antibody in the hel-C2 complex corresponds to neither of the orientations found in the P212121 crystal form, with differences in orientation of 9.5 (hel1-hel-C2, Figure 1(c)) and 6.0 (hel2-helC2). The axis of rotation is different than that observed for the hel1-hel2 superposition (compare Figure 1(c) with Figure 1(b)). For the cAb-Lys3-turkey lysozyme complexes (tel1 and tel2), the differences for the selected interface characteristics parameters are less extreme (Table 1, columns 4 and 5). Also, the rms deviation between tel1 and tel2 complexes is slightly smaller than for the hel1 and hel2 complexes (0.78 versus Ê ). However, the difference in orientation 1.04 A between the tel1 and tel2 complexes is of the same magnitude as between the hel1 and hel2 complexes (8.3 versus 9.5 ; Figure 1(d)). So for the tel1 and tel2 complexes, a comparable extent of difference in orientation is linked to smaller differences of the selected interface characteristics. This could be related to the low temperature data collection used for the turkey lysozyme complex leading to a slight unit cell contraction, or to the differences in primary structure between hen and turkey lysozyme. There are seven sequence differences between hen and turkey egg white lysozyme. Only two of these amino acid substitutions, R73 K and D101G, are part of the interaction surface, with marginal in¯uence on the interface. Local (per residue or per atom) differences in interface characteristics can also be analysed, e.g. by looking at differences in accessible surface area and Voronoi volumes. Figure 2 shows the variation of local interface characteristics for cAb-Lys3 in the hel1 and hel2 complexes, both for the buried accessible surface area and for Voronoi volumes. There is little or no difference in accessible surface area or Voronoi volumes for atoms in the central part of the interface, including the atoms involved in hydrogen bonds with the antigen. Towards the periphery, larger differences become apparent. This re¯ects the classi®cation of ``recognition area'' atoms into interface atoms, contact atoms and buried atoms.3 We also note that, although the trends
Table 2. Comparisons of independent copies of antibody-antigen complexes in the RCSB Protein Data Bank rms Ab (number atoms) Ê) (A
rmsd Ag (number atoms) Ê) (A
rmsd all (number atoms) Ê) (A
Interface Polar Non-polar Hydrogen planarity groups (%) (%) bonds
Gap volume Ê 3) (A
Gap volume index
Sc
Surface buried Ê 2) (A
Type
Resolution Ê) (A
1nfd
Fab/protein
2.8
0.32 (476) 0.95 (1640) 2.63 (2116)
7.1, 19.6
3.0 3.14
50.7 44.8
49.2 55.1
8 6
4688 4613
2.89 3.18
0.60 0.56
750 681
1jto
VHH protein
2.5
0.57 (524)
0.59 (515) 1.04 (1039)
1.3, 9.5
3.58 3.31
28.1 28.3
71.9 71.6
8 8
2438 3289
1.46 2.22
0.70 0.64
890 800
1jtp
VHH/protein
1.9
0.36 (508)
0.38 (516) 0.78 (1002)
1.3, 8.3
1bvk
Fv/protein
2.7
0.46 (460)
0.55 (516)
0.70 (976)
1.1, 5.1
2ap2
Fv/peptide
2.4
0.43 (480)
0.50 (44)
0.46 (524)
2.0, 5.1
1dzb
scFv/protein
2.0
0.42 (460)
0.60 (516)
0.64 (976)
1.2, 3.4
1bj1
Fab/protein
2.4
0.15 (468)
0.27 (376)
0.39 (844)
1.1, 3.1
3.41 3.35 1.72 1.62 2.46 2.18 2.49 2.54 2.8 2.84
47.8 42.5 53.8 53.8 49.9 47.9 48.7 50.5 39.8 39.4
52.1 57.5 46.1 46.2 56.0 52.1 51.3 49.5 60.1 60.5
5 6 12 11 3 6 8 9 10 10
2997 2974 3470 4112 1347 1493 4968 4899 2088 1989
1.91 1.93 2.47 3.11 1.49 1.37 2.59 2.99 1.25 1.19
0.66 0.64 0.64 0.68 0.49 0.60 0.61 0.61 0.74 0.74
737 718 716 679 389 457 795 775 795 796
1afv
Fab/protein
3.7
0.09 (476)
0.10 (604) 0.27 (1080)
1.0, 2.1
2.59 2.61
36.9 38.8
63.0 61.1
11 10
5255 5366
3.74 3.82
0.58 0.60
720 720
1fbi
Fab/protein
3.0
0.70 (484)
0.88 (516)
0.81 (100)
0.2, 2.1
2.09 2.09
48.6 47.6
51.3 52.4
12 11
3441 3517
2.03 2.03
0.65 0.66
858 872
1fj1
Fab/protein
2.7
0.14 (448) 0.03 (1004) 0.20 (1454)
0.2, 1.98
2.55 2.53
45.4 44.2
54.5 55.7
11 8
4036 4053
2.48 2.45
0.70 0.73
755 769
2hrp
Fab/peptide
2.2
0.31 (496)
0.35 (40)
0.32 (536)
0.3, 0.8
2.91 3.18
45.0 43.8
54.9 56.2
8 6
1201 1073
1.02 0.91
0.76 0.76
530 542
1mlc
Fab/protein
2.1
0.34 (460)
0.60 (516)
0.50 (976)
0.3, 0.5
1.71 1.79
43.4 51.8
56.5 48.2
8 5
2519 2735
1.81 1.95
0.58 0.59
668 669
1ahw
Fab/protein
3.0
0.27 (464)
0.33 (800) 0.32 (1264)
0.2, 0.5
1nmc
scFv/protein
2.5
0.0 (484)
0.37 (1552) 0.32 (2036)
0.0, 0.0
2.78 3.08 1.48 1.48
60.3 58.3 54.7 54.7
39.7 41.7 45.2 45.2
15 13 9 9
4555 4555 6678 6765
2.40 2.32 5.08 5.15
0.56 0.57 0.60 0.63
953 986 652 652
1bzqa
VHH/protein
2.8
0.05 (120)
0.04 (124)
0.1-0.4, 0.73-2.0
1.45 1.44 1.44 1.42
41.9 41.0 41.6 43.7
58.1 59.0 58.4 56.3
7 7 8 7
2467 2315 2152 2333
2.24 2.04 1.92 2.08
0.65 0.62 0.54 0.66
541 556 551 557
0.13 (244) 0.17
trans., rot. Ê , deg.)b (A
The type of antibody and antigen, the resolution available, the rms deviations (rmsd) for antibody, antigen and complex within the same asymmetric unit, and the differences in relative orientation between antigen and antibody are given. The number of atoms used for each superposition is indicated in parentheses. Also, for each complex, the basic interface characteristics are given. To calculate the difference in relative orientation between antibody and antigen, a least-squares superposition of the antigens was done ®rst, and the resulting operator was also applied to the antibody molecules. Starting from that position, the translation and rotation needed for a perfect least-squares superposition of the core of the heavy-chain of the antibody molecules (the Cys22Cys92 disulphide bond plus three amino acid residues before and after each cysteine) were calculated: the rotation angle is obtained from the trace of this second operator matrix, and the distance is calculated from the translational component. a For 1bzq, all four complexes were superimposed concurrently using the program POLYPOSE (Ca atoms). b Where appropriate, the range of translations and rotation angles for the possible combinations is given.
477
Degenerate Interfaces in Antigen-Antibody Complexes
Figure 2. Differences in (a) buried surface area and (b) Voronoi volume for the cAb-Lys3 interface atoms between the hel1 and hel2 complex. Dark colours represent little or no change, light colours represent large differences. CDR1 atoms are coloured blue, CDR2 atoms are yellow, and the CDR3 atoms are red. Green halos around CDR3 atoms indicate atoms making hydrogen bonds with the antigen. Framework atoms are grey.
cases. Besides for the ab-TCR receptor, ¯exibility similar to our case has been observed, but not emphasized further, for a scFv-turkey egg white lysozyme (1dzb) antibody-antigen complex.16 From a mechanical point of view, one can easily conceive how the cAb-Lys3-lysozyme complex can adopt these different relative orientations: the Nterminal part of the CDR3-loop protrudes into the active site of hen lysozyme in a fashion somewhat similar to a ball and socket joint. In the 1nfd and 1dzb structures, a protruding antigen loop inserts between the antigen binding loops L3 and H3 of a ``classical'' antibody, creating a ``reversed'' situation as compared to the cAb-Lys3-lysozyme case. For 1bvk and 2ap2, with larger changes in relative orientation and surface characteristics than 1dzb, the ball and socket model is not valid (Table 2).
Conclusion are similar for buried surface area and Voronoi volume, the exact details on local variation obtained from the two methods of calculation are different (Figure 2). Other antibody-antigen complexes To assess the relevance of the different orientations for complexes within the same asymmetric unit as found for hel1-hel2 and tel1-tel2 complexes, we turned to the RCSB Protein Data Bank. Of the numerous antibody-antigen complexes available, 14 entries have multiple complexes in the asymmetric unit. We analysed these complexes with respect to the relative position (translation and rotation) between antibody and antigen as described for the cAb-Lys3-lysozyme complex (Table 2). For each of these structures, we calculated the rms deviations for antibodies, antigens and whole complexes for the protein molecules within the same asymmetric unit, as well as the translation and rotation expressing differences in relative orientation. Table 2 also lists the interface characteristics introduced previously (Table 1) for these complexes. In most cases, there is little or no difference in orientation for complexes within the same asymmetric unit, e.g. for PDB entry 1fj1, small rms deviations and small differences in relative orientation can be linked to small variations in the Sc-value and the gap index. For this entry, the interface is expected to be identical for the two complexes within one asymmetric unit. For other entries (top rows of Table 2), large differences in orientation are linked to large variations of both the Sc-value and the gap volume index. The rotation listed for the cAb-Lys3 VHH molecule is the second largest, and is preceded only by the ab T-cell receptor in complex with its antibody (PDB entry code 1nfd).15 Therefore, the cAb-Lys3lysozyme complex is one of the more extreme
Many systematic studies of protein-protein interfaces start with the assumption that there is only one optimal complex for a pair of proteins. Our results show that this assumption is not always valid even within the con®nement of a single crystal structure. We have shown that the parameters characterising the protein-protein interfaces can vary signi®cantly for one and the same complex. As the hel1 and hel2 complexes are both present in the same crystal, the differences between the complexes can not be attributed to different solvent compositions as described by Huang et al.17 Also, in order to allow crystal growth, each of the conformations present in the crystal must be present as a major fraction in the protein solution from which the crystals grow,18,19 with crystal packing interactions locking the structure in a particular sub-state20 upon incorporating the protein in the crystal. Therefore, our observations of differences in relative orientation and the consequences for the interface is a re¯ection of the properties of our complex, and can not simply be dismissed as an artifact introduced by crystallisation. At ®rst sight, degeneracy at the antigen binding interface complicates the analysis of this interaction area and of protein-protein interfaces in general. However, analysis of the cAb-Lys3-lysozyme complexes and similar cases leads to estimates of the intrinsic variance of the interface characteristics parameters, and therefore to a better understanding of the interfaces themselves.
Acknowledgements We thank the Prodex programme and ESA for the use of the protein crystallisation facilities on space ¯ight STS 95. We are also grateful to Yves Geunes for keeping our partition tables healthy.
478
Degenerate Interfaces in Antigen-Antibody Complexes
References 1. Jones, S. & Thornton, J. M. (1997). Analysis of protein-protein interaction sites using surface patches. J. Mol. Biol. 272, 121-132. 2. Jones, S. & Thornton, J. M. (1996). Principles of protein-protein interactions. Proc. Natl Acad. Sci. USA, 93, 13-20. 3. Lo Conte, L., Chothia, C. & Janin, J. (1999). The atomic structure of protein-protein recognition sites. J. Mol. Biol. 285, 2177-2198. 4. Pauling, L. & Pressman, D. (1945). The serological properties of simples substances. IX. Hapten inhibition of precipitation of antisera homologous to the o-, m-, and p-azophenylarsonic acid groups. J. Am. Chem. Soc. 67, 1003-1012. 5. Lawrence, M. C. & Colman, P. M. (1993). Shape complementarity at protein-protein interfaces. J. Mol. Biol. 234, 946-950. 6. Richards, F. M. (1974). The interpretation of protein structures: total volume, group volume distributions and packing density. J. Mol. Biol. 82, 1-14. 7. Richards, F. M. (1985). Calculation of molecular volumes and areas for structures of known geometry. Methods Enzymol. 115, 440-464. 8. Gabb, H. A., Jackson, R. M. & Sternberg, J. E. (1997). Modelling protein docking using shape complementarity, electrostatics and biochemical information. J. Mol. Biol. 272, 106-120. 9. Chothia, C. & Janin, J. (1975). Principles of proteinprotein recognition. Nature, 256, 705-708. 10. Janin, J. (1995). Principles of protein-protein recognition from structure to thermodynamics. Biochimie, 77, 497-505. 11. Ysern, X., Li, H. & Mariuzza, R. A. (1998). Imperfect interfaces. Nature Struct. Biol. 5, 412-414. 12. Hamers-Casterman, C., Atarhouch, T., Muyldermans, S., Robinson, G., Hamers, C., Songa, E. B. et al. (1993). Naturally occuring antibodies devoid of light chains. Nature, 363, 446-448. 13. Desmyter, A., Transue, T. R., Ghahroudi, M. A., Dao, Thi M.-H., Poortmans, F., Hamers, R. et al. (1996). Crystal structure of a camel single-domain
14.
15.
16.
17.
18.
19. 20. 21.
22. 23.
VH antibody fragment in complex with lysozyme. Nature Struct. Biol. 3, 803-811. Transue, T. R., De Genst, E., Ghahroudi, M. A., Wyns, L. & Muyldermans, S. (1998). Camel singledomain antibody inhibits enzyme by mimicking carbohydrate substrate. Proteins: Struct. Funct. Genet. 32, 515-522. Wang, J.-H., Lim, K., Smolyar, A., Tseng, M.-K., Liu, J.-H., Tse, A. G. D. et al. (1998). Atomic structure of an ab T cell receptor (TCR) heterodimer in complex with an anti-TCR Fab fragment derived from a mitogenic antibody. EMBO J. 17, 10-26. Ay, J., Keitel, T., Wessner, H., Scholz, C., Hahn, M. & HoÈhne, W. (2000). Crystal structure of a phage library-derived single-chain Fv fragment complexed Ê resolution. with turkey egg-white lysozyme at 2.0 A J. Mol. Biol. 301, 238-246. Huang, D.-B., Ainsworth, C. F., Stevens, F. J. & Schiffer, M. (1996). Three quaternary structures for a single protein. Proc. Natl Acad. Sci. USA, 93, 70177021. Muesser, T. C., Rogers, P. H. & Arnone, A. (2000). Interface sliding as illustrated by the multiple quaternary structures of liganded hemoglobin. Biochemistry, 39, 15353-15364. Tame, J. R. (1999). What is the true structure of liganded haemoglobin. Trends Biochem. Sci. 24, 372376. Brunger, T. A. (1997). X-ray crystallography and NMR reveal complementary views of structure and dynamics. Nature Struct. Biol. 4, 862-865. Decanniere, K., Desmyter, A., Lauwereys, M., Ghahroudi, M. A., Muyldermans, S. & Wyns, L. (1999). A single-domain antibody fragment in complex with RNase A: non-canonical loop structures and nanomolar af®nity using two CDR loops. Structure, 7, 361-370. Wilson, I. A. & Garcia, K. C. (1997). T-cell receptor structure and TCR complexes. Curr. Opin. Struct. Biol. 7, 839-848. Hubbard, S. J. & Thornton, J. M. (1993). NACCESS, Computer Program Department of Biochemistry and Molecular BiologyUniversity College London.
Edited by J. Thornton (Received 15 May 2001; received in revised form 29 August 2001; accepted 5 September 2001)
