Part 5 Antibody to protein recognition- structural principles in antibody design.

Transcription

1 Monday, January 14, 2013 Biophysics 204 Lecture Notes Robert Fletterick January 14, 2013 Part 5 Antibody to protein recognition- structural principles in antibody design. This section is about the structure, evolution and functions of antibodies, and molecules of the immune system that are related to antibodies. Antibodies are nature s molecules designed to bind all possible proteins, macromolecules and small molecules. What is so special about them? What principles provide this adaptability? We want to understand why antibodies recognize proteins so efficiently and we will learn that their binding scheme is recapitulated in related proteins that are used by cells in the immune system to bind and report features of cellular activation or cellular inactivity to other cells. The so-called Fab domains of antibodies are built to bind antigens- the name given to the molecules that they capture. The antigens may be small organic molecules or sugars, or large proteins or nucleic acids. Fab molecules have antigen-combining sites with special properties due to their structure and amino acid sequences.

2 Two Fab s, and one Fc domain comprise a binding antibody like an IgG molecule, the intact antibody. The antibody is made from two copies each of large and small proteins, covalently linked by SS bonds between the chains. We will look at the engineering of Fab s at the topological and atomic levels. Structural data are essential for discovering the principles of recognition and affinity. Structures are known for over 100 Fab -protein complexes at near atomic resolution. Structures for only a few intact antibodies, are known presumably because most of these are dynamically flexing structures about the domains linkages. Features of a Fab structure- To understand the nature of these interactions, we shall consider the structure of the Fab combining site. The Fab molecule is comprised of two chains, heavy (H) and light (L) according to their lengths. The heavy chain is cleaved in the lab, by the protease papain, to prepare the Fab s from the IgG molecule; the L chain of Fab is made of two domains called C and V for constant and variable, referring to degree of variance in sequences found there for the family of antibodies. The heavy chain fragment also contains two domains C and V. The antigen is held between the VL and VH domains. The contacts are primarily made through three loops in each of the V domains. These are the hyper variable loops and the binding regions of the loop are referred to as complementarity determining regions, CDR s 1-3. Example- 2

3 Fab complex with neuraminidase Tulip WR; Varghese JN; Laver WG; Webster RG; Colman PM. Refined crystal structure of the influenza virus N9 neuraminidase-nc41 Fab complex. Journal of Molecular Biology, 1992 Sep 5, 227(1): The crystal structure of the complex between neuraminidase from influenza and the monoclonal Fab NC41 and a few bound water molecules shows fine but not perfect shape complementarity of the protein antigen and the antibody. [Viral neuraminidase is an enzyme responsible for cleaving sialic acid residues on newly formed virions as they bud off from the host cell. Roche makes a successful drug, Tamiflu, which blocks this enzyme and lessens the impact of influenza. This inhibition results in aggregation of virions on the surface of the host cell, which limits the extent of infection and speeds recovery from illness] Antibodies derive biosynthetically from a single immune cell. If many cells are activated to secrete antibodies, they are polyclonal. If a single cell is selected and clonally expanded the 3

4 antibodies have a unique sequence and are monoclonal and can be used as reagents. This cell was selected to grow, rather than its cousins, because it has on its surface an Fab domain, that by chance binds the antigen. The continued selection of cells for strong binding to antigen leads to somatic mutations in the Fab domain that further enhances the affinity for antigen. Ultimately, the remarkable process is one of evolutionary selection on a very short time scale, days not eons. By carefully cataloging the types and density of contacts between atoms of the Fab and antigen neuraminidase, Peter Coleman showed that the packing density of atoms at the interface can be quantified. These calculations show that the density is somewhat looser than in the interior of a protein. The packing density is imperfect, but good. We might expect charge interactions to be rare since taking a charge out of solvent and into a buried interaction surface is unfavorable without charge complementarity. Complementarity can be of two types, formal charge balance or formal to partial charge balance such as three carbonyl oxygen atoms from three main chains or side chains converging in space across from a Lys NZ+. Both events should be rare in short-term evolution. In this case, the antigen and antibody has one buried salt-link or charged pair, two partly solvated salt-links and 12 hydrogen bonds. The antibody-binding site for neuraminidase is discontinuous and comprises five chain segments and 19 contact residues. The other partner in the complex has thirty-three neuraminidase residues in eight segments that contact the Fab and together have a buried surface area of 900 Å 2. 4

5 Seventeen residues in NC41 are in the five CDR s and contact the neuraminidase. The interface is more extensive than those of the three lysozyme-fab complexes whose crystal structures have been determined. There are only small differences (less than 1.5 Å) between the complexed and uncomplexed neuraminidase structures. We will soon see that five of the six CDR s, not H3, show similar predictable structures, regardless of the particular amino acid sequences in the variable segments. The shape complementarity of protein/protein interfaces for antibody/antigen interfaces is poorer than is observed in other protein/protein interactions. There is a different evolutionary history of antibody/antigen associations compared to other systems considered. Also, the chemical natures of the interfaces are variable, but typically polar like the normal surface of a soluble protein. (Lawrence MC; Colman PM. Shape complementarity at protein/protein interfaces. Journal of Molecular Biology, 1993 Dec 20, 234(4): ) Other structures determined for monoclonal Fab show similar features- Lysozymes, protein G complexes with their respective monoclonal Fab s are broadly similar. General Conclusions: 1. Physical chemistry determining the interactions is not unusual. 2. Comparing the structures three dimensions of three antibodies, different but randomly discovered, for lysozyme showed that no part of the lysozyme surface is special. An Fab can cover virtually any part of the entire surface of lysozyme no part of a protein antigen s surface is special. 5

6 3. Solvent accessible surface lost in complex formation is about 1500Å 2 (750 Å 2 per protein). This area turns out to surprisingly economical- many critical protein protein interfaces are much larger- two to five times this size. If protein binding can be super strong with Å 2, why are other interfaces much larger? 4. The surfaces are complementary and exclude water. Two water molecules are observed in HyHEL5, one water molecule is found in D1.3 and none in HyHEL-10. Structural Changes Do loops of Fab s change on binding? Does the main chain conformation of Fab s change on binding? Structures of main chain for lysozyme alone and in complex agree to 0.5Å, side chains to 1-2Å. Some entropy change is certainly lost to side chain ordering when the complex forms. There is a large structural change however! 6

7 6. The structures of Fv D1.3 alone and in complex are identical to 0.5Å but the VL and VH domain relationship changes by 1Å. In other cases more significant changes are observed on the relative positioning of the V domains. The key to Fab s binding proteins is the disposition of the hyper variable regions that are six surface loops, CDR1-3 for the heavy and light chains. These are called L1, L2, L3 for the light chain and H1, H2, H3 for the heavy chain. CDR1 is in the 30 s loop, CDR2 in the 50 s and CDR3 is made from the 90 s loop, for both L and H. The general description of the IgG fold for V, C, CD4 and CD8 domains, homologs in the immune system is as follows: Do antibodies make hydrogen peroxide? Until 2001, the only known function of antibodies was to bind antigens and direct a conformational change to other molecules of the immune system. An unusual finding was that antibodies are also found to have enzymatic activity! [Antibody catalysis of the oxidation of water. Wentworth P Jr, Jones LH, Wentworth AD, Zhu X, Larsen NA, Wilson IA, Xu X, Goddard WA 3rd, Janda KD, Eschenmoser A, Lerner RA. Science, Recently we reported that antibodies can generate hydrogen peroxide (H 2 O 2 ) from singlet molecular oxygen (1O 2 *). We now show that this process is catalytic, and we identify the electron source for a quasi-unlimited generation of H 2 O 2. Antibodies produce up to 7

8 500 mole equivalents of H2O 2 from 1O 2 *, without a reduction in rate, and we have excluded metals or Cl - as the electron source. On the basis of isotope incorporation experiments and kinetic data, we propose that antibodies use H 2 O as an electron source, facilitating its addition to 1O 2 * to form H 2 O 3 as the first intermediate in a reaction cascade that eventually leads to H 2 O 2. X-ray crystallographic studies with xenon point to putative conserved oxygen binding sites within the antibody fold where this chemistry could be initiated. Our findings suggest a protective function of immunoglobulins against 1O 2 * and raise the question of whether the need to detoxify 1O 2 * has played a decisive role in the evolution of the immunoglobulin fold.] Structure Basics CD4, CD8 and VL and VH all have two β sheets (all strands antiparallel) one with 4 strands and one with 5. All immunoglobulin folds are about 100 residues. Each is a two-layer stack of β sheets. There is a hinge linking the Fc to the Fab domains that is susceptible to papain protease. [Protein engineering can be used to make Fab into a variant an artificial Fv single chain.] After binding antigen, the antibody transmits a signal to the Fc portion to signal the cell that the complex is formed. The nature of the Fab to Fc signal is not known. The immunoglobulin fold comes in two types within the antibody. In other molecules further variations in topology are found. There are two major classes of domain folds defined by the topological linkage and number of secondary structure elements. The constant domain is made from two sheets of four and three strands. The second is the variable domain, with nine strands instead of seven, in two sheets of 4 and 5 strands. The topology figure below is for the variable domain. If strands C and C are 8

9 incorporated into the loop connecting C to D, it is a constant domain of seven strands. In this case the analog of CDR2 would be absent. The strand nomenclature for the VL and VH domains is as shown below in a topology diagram. The beta strands are shown end-on as rectangles with the connecting loops and N and C- termini providing information about the travel direction of the polar peptide chain. Note that CDR1 is perpendicular, bridging the sheets, with respect to CDR2 and CDR3. The letters for the beta strands are in alphabetical order according to the positions of the secondary structural elements as found in the sequence: N-term A B E D CDR3 CDR1 CDR2 G F C C' C'' C-term Names of beta strands and topological positions of CDR's in V domains. In the following figure, the plan view of the positions of the three interacting loops in the two juxtaposed variable domains shows L3 across from H2 and vice versa, with the CDR1 s distal to the domain interface: 9

10 10

11 In three dimensions, the VL and VH chains come together to form the combining site so that the CDR s are adjacent and the most variable structure is found in the center of the combining site. On binding do side chains flex or loops bend? In considering the shapes of the CDR s, the most likely guess would be that the variable sequences leads to variable conformations in each CDR and each Fab. A second model would be that there are few main chain conformations and that the variable side chains adjust to the antigen. The latter is the case. Lesk, Chothia and Tramontano [Chothia C; Lesk AM; Tramontano A; Levitt M; Smith-Gill SJ; Air G; Sheriff S; Padlan EA; Davies D; Tulip WR; et al. Conformations of immunoglobulin hypervariable regions. Nature, 1989 Dec 21-28, 342(6252): )] have examined the structures of the CDR s and showed, surprisingly, that these loops have predictable structures! Comparative studies of known antibody structures and sequences show there is a small repertoire of main-chain conformations for five of the six CDR s of antibodies, and that the particular conformation adopted is determined by a few key conserved residues. By observing a family of sequences, the authors noticed that the hypervariable regions had similar residues at special positions. They reasoned that the special positions were the determinants of the conformation of the loop. More than one class of conserved residues is found within some of the loops, each forming a distinct structural family. These observations 11

12 imply that changes in the identity of residues at non-special sites so not affect the conformations of the canonical structures. Deciding on which special residues are responsible for the structures is the key observation. Lesk and Chothia looked for important sites in packing, for critical hydrogen bonding patterns and for unusual backbone torsional angles. They found critical amino acids in specific sites in the hypervariable regions and in the framework itself. The sequence variation includes changes in identity of amino acids and insertions of amino acids as shown in the table below. The H1 and H2 segments take canonical structures that are close in amino acid position to the variable sequence in the CDR s. The CDRs are not just loops connecting the strands. The CDRs involve parts of the strands and frequently sequence variations do not translate to structure variations. The table is a summary of loops and their subtypes and key residues taken from the Chothia paper. 12

13 Loop SEQUENCE Key Framework Insertions- L1-4 types ,25,33,71 insertions at 31 L2-1type , 64 none L3 3 types , 90 variable at 96 H1-1 type ,94 None H2-4 types ,54, 55,71 Insertions at 52 What about H3? The case is more complex and structures are more variable. [ Morea V; Tramontano A; Rustici M; Chothia C; Lesk AM. Conformations of the third hypervariable region in the VH domain of immunoglobulins. Journal of Molecular Biology, 1998 Jan 16, 275(2): ] 13

14 14

15 Domain Associations The contacts between CL and CH2 are extensive and drive dimerization of the heavy chain. Characteristically, these associations are mainly hydrophobic with the strands being at nearly right angles. Outside the IgG superfamily, this type of association is not common. Interface residues are conserved. Unlike the C domain contacts, those between the V domains are subject to change in registration with each other on binding antigen. Even changing the buffer can alter the association. In fact, for one Fab (loc), M. Schiffer has shown the VLVH domain association changes by 3.5Å depending on whether the crystals are grown with ammonium sulfate or PEG. The antigens interact to keep the 6 CDRs adjacent. Figures and show the relationships of the VV and CC domains in an immunoglobulin. 15

16 16

17 The six loops contact a protein antigen in slightly different ways depending on size and shape of the antigen. The table compares numbers of contacts for several lysozyme Fab pairs and a peptide Fab pair. Table: contact amino acids in the three CDRs of the L and H chains of FAB with lysozymes and peptide. Fab L 30 L 50 L 90 H 30 H 50 H 90 No. of lysozyme or peptide side chains in contact HyHel D HyHEL gp V3-14AA peptide The immunoglobulin fold and its topological variants (check the PDB web site and look for SCOP- and scan the all-beta 17

18 folds) are found in an enormous number of molecules that are involved in cell recognition events. It is a very versatile structure and the tertiary structures are slightly altered. Provacative experiment with interfacial residues in What are the restrictions of amino acid type distribution to get specificity and affinity? Sihdu and colleagues tested the interaction surface by limiting the types of amino acids that could appear in the interface. Directed evolution was used to change antibody CDRs by limiting the residues to either just four or two types of amino acids. Phage display was used to make Fabs for protein antigens. The amino acids and their colors in the figure panel A below are: alanine (magenta), aspartate (green), serine (blue), and tyrosine (cyan). The crystal structure of an Fab-VEGF pair (Protein Data Bank entry 1tzh) has VEGF in gray and the mutated residues of Fab that form interaction are shown as sticks. Surprisingly, for the case of these 4, the new residues form 98% of the buried surface area. In panel B only two amino acids [serine (blue) and tyrosine (cyan) are needed to bind the death receptor, DR5. Protein Data Bank entry 1za3 shows the residues of serine and tyrosine to form 80% of the buried surface area. It was known that tyrosine side chains are very effective in building specific and tight interfaces. These complexes are nanomolar in dissociation constrant. See Fellouse, F. A., Wiesmann, C., and Sidhu, S. S. (2004) Synthetic antibodies from a four-amino-acid code: A dominant role for tyrosine in antigen recognition. Proc. Natl. Acad. Sci. U.S.A. 101, Dissecting Protein Protein Interactions Using Directed Evolution. Daniel A. Bonsor and Eric J. Sundberg Biochemistry, 2011, 50 (13), pp

19 Topic) 19

20 Homologs of IgG domains CD8 MHC binds CD8, a disulfide linked dimer. D.J. Leaky et al. determined the structure of a deglycosylated soluble from (Cell : ) The structure was determined by molecular replacement using a V domain composite of 9 structures. The variable domain is recapitulated in CD4 and CD8. The sequence identity between CD8 and the homolog REI antibody is 20%. The structures of CD4 and CD8 are compared with REI in the paper by Leahy, et al., Cell 1992, 68: The CDR analog is different as shown in Figure 8: 20

21 The conservation of backbone structure is good including the bulges in strands C' and G, even though the sequences are not conserved in CD8α. The protomers of the dimer are related 21

22 within 10 to those on REI. The contacts of the dimer are similar using the G and C' strands (and bulges). The G and C' loops are 2 amino acids longer and move the CDR3 loops closer together. Both CD4 and CD8 show long CDR2 loops. CDR2 is away from the body of the protein. CD8 in the T cell recognition complex In the immune response, the T -cell receptor is responsible for recognizing an antigen peptide presented by the antigen-presenting cell to the surface of the plasma membrane. The complex of proteins meeting at the surfaces of the two cells is the T -cell receptor, HLA-peptide complex, β 2 microglobulin and CD8. The role of CD8 is to mediate the formation of the complex; CD8 is needed in the clonal expansion of cytotoxic T cells in the thymus. The schematic for the inteaction between the T cell and the antigen, AG is shown is the schematic from Merck. 22

23 The structure of the entire complex is not known, but structures are known for a T cell receptor binding to the distal side of HLA and separately for CD8 binding to the proximal side of HLA. Gao, GF; et al. Crystal structure of the complex between human CD8alpha and HLA-A2. Nature, 1997 Jun 5, 387(6633): The homodimer stabilizes the interaction of the T-cell antigen receptor (TCR) with major histocompatibility complex (MHC) class I/peptide by binding to the class I molecule. The crystal structure at 2.7 A resolution of a complex between CD8 and the human MHC molecule HLA-A2 associated with peptide shows that CD8 dimer binds one HLA-A2/peptide molecule, forming an interface with the α2 and α 3 domains of HLA-A2 and also contacting beta2-microglobulin. A flexible loop of the α3 domain (residues ) is clamped between the CDR-like loops of the two CD8 subunits in the classic mode of an antibody-antigen interaction. The position of the α 3 domain is different from that in uncomplexed HLA-A2. However it is similar to the position taken in the complex of the T-cell antigen receptor with HLA-A2, TCR/Tax/HLA-A2, suggesting that the HLA complex formation with T-cell antigen receptor require no further change from CD8 induced changes. The HLA-A2/peptide surface presented for TCR recognition shows no conformational change. The molecular complex of CD8 in red and blue with HLA and peptide in green gray and microglobulin in yellow is: 23

24 The interaction surface showing between the HLA-A2 and the CDR s of both CD8 s shows: the α 3 domain is rotated 7 from average. the contributions from CD8 s is asymmetric all CDRs interact. Overall, 1000Å 2 of surface are buried by each CD8 dimer and HLA. α 3 buries 700 Å 2 CD8 blue buries about 300 Å 2 CD8 red buries about 700 Å 2, shared with the α 2 domain and that its major interaction is with the loop HLA-A2 [ ], predicted by mutagenesis, which inserts between the CD8 s and is held by main and side chain interactions. Most interactions (80 %) are polar and involve side chains. CD8 s do not change structure. 24

25 These contacts are distributed in space and each is of marginal strength. A human mutant HLA- Aw68 Ala 245 to Val, that is just two methyl groups different- binds CD8 poorly and its structure was determined in the Wiley lab. The structure shows Val 245 does not interact with CD8 but twists the loop. 25

26 A clearer view of the complex of HLA and CD8 shows the CDR loops and associations of the two CD8 s to be like the two V domains of the Fab. CD8-1 α3 domain and loop at right From GEORGE F. GAO Nature 387, (1997) Crystal structure of the complex between human CD8 and HLA-A2 CDR like loops of CD8 26

27 A good review of the interactions in recognition of the complex of HLA and peptide is recently published. Nature Reviews Immunology 6, (December 2006) Structural determinants of T-cell receptor bias in immunity. Stephen J. Turner, Peter C. Doherty, James McCluskey & Jamie Rossjohn 27