New comprehensive biochemistry vol 17 molecular genetics of immunoglobulin

complementarity determining region diversity segment antigen binding fragment from papain digestion of immuno- globulin antigen binding fragment from pepsin digestion of immuno- globulin

MOLECULAR GENETICS OF IMMUNOGLOBULIN

New Comprehensive Biochemistry

Molecular Genetics of Immunoglobulin

Edifo rs

Medical Research Council Ltihoratory of Molecular Biology,

Hills Road, Combridge C B 2 2 Q H , U K

1987 ELSEVIER

All rights reserved No part of this publication may hc rcprocluced stored i n a retrieval system, or transmitted i n any form or by any means, electronic mechanical photocopying recording or other- wise, without the prior written permission of the Publisher Elsevicr Science Publishers B.V (Biomed- ical Division) P O Box 1527, 1000 BM Amsterdam T h e Nethcrlands

No responsihility is assumed hy the Publisher for any injury and/or damage t o persons or property as

a matter of products liahility negligence o r otherwise or frcm any use or operation of any methods products, instructions or ideas contained in the material herein Because of the rapid advances in the

medical sciences, the Publisher recommends that independent verification o f diagnoses and drug dos-

ages should be made

S p e d rc7guluriotz.s f i i r reciders i i i rlic U S A This publication has been registered with the Copyright

Clearance Center Inc (CCC) Salem, Massachusetts Information can he obtained from the CCC about

conditions under which the photocopying o f parts of this publication may he madc in the USA All

other copyright questions including photocopving outside of the USA should be referred to the Pub- lisher

ISBN 0-444-XO915-5 (volume)

Sole distributors lor thc USA and C;ln;da:

Elsevier Science Publishing Company Inc

51- Vandcrbilt Avenuc

New York, NY 10017

USA

Library of Congress Cataloging in publication Data

Molecular genetics of inimunoglohulin

(New comprehensive hiochcmistry ; v 17)

Includes bibliographies and index

1, Immunoglobulins Genetics 2 Gene exprcssion

V

Preface Immunoglobulin genes are not just of interest to immunologists An understand-

ing of the way in which DNA rearrangement and somatic mutation contribute to

antibody diversity is of importance to a wide range of biologists The cell-type specificity of immunoglobulin gene expression is of concern to many who are in- terested in gene expression in mammals Furthermore, the immunoglobulin su- perfamily itself presents important questions to those interested in evolution The analysis of immunoglobulins and of their genetics has advanced rapidly since the mid-l970s, mainly as a result of the application of recombinant DNA and

monoclonal antibody technologies The essential features of the molecular anat- omy of both antibodies and their genes have been largely identified; this has re- sulted in significant insights into the way antibody diversity is generated Clearly, much still remains to be elucidated in these areas, whilst studies both of regulation and of phylogeny are still in their infancy We felt nevertheless that it was a good time to draw together what we do know about the molecular genetics of immu- noglobulin

We wish to thank the authors for contributing to this volume and the publisher for prompt publication

Michael S Neuberger

Contents

Preface v

List of abbreviations XII Chapter I Structure and function of antibodies D.R Burton (Sheffield, U K ) 1

I Introduction 1

2.1 General considerations 3

2.2 Domain structure 6

2.3 Structure of Fab 6

2.4 Antigen recognition 8

2.5 Structure of Fc 11

2.7 Isotypes, allotypes and idiotypes

3.1 Introduction 21

3.2 Interaction with protein A 21

3.3 Complement activation 22

3.4 Interaction with cellular 24

3.5 Other functions of IgG 26

3.6 Rheumatoid factors 27

3.7 Membrane or surface IgG 27

3.8 Structure-function relationships in IgG: domain hypothesis , , , , , , , , , 28

Structure o f other immunoglobulins in relation t o IgG

5.1 Structure of I g M

5.2 Functions of IgM 37

39 6.1 Structure of serum 39 6.2 Structure of secret 41 2 Structure of IgG 3

2.6 T h e hinge: IgG subclasses

3 Functions of IgG

4 5 Structure and function of I g M 36

5.3 Membrane IgM 39

Structure and function of IgA ,

6 6.3 Functions of IgA 41

7 Structure and function of IgD 41

X Structure and function of IgE 43

X l Structure o f IgE 43 8 2 Functions of IgE

9 Summary

Acknowledgements 46

References 47

Chapter 2 Genes encoding the immunoglobulin constant regions M Briiggemann (Cambridge UK) 51

1 Introduction

2 Chromosomal localization

3 Organization of constant region genes 52

3.1 Mouse heavy chain genes 52

3.2 Mouse light chain genes 53

3 3 Human heavy chain gencs 54

3.4 Human light chain genes 55

3 5 Other species 55

3 6 Switch regions 58

3.7 Membrane exons 61

J chain 62

62

1 constant region genes 64

4.1 Heavy chain genes 64

4.2 Light chain genes 67

4.3 Polymorphism 68

4.4 Pseudogenes 69

4.5 Evolution 70

4.6 Aberrations and malignancies 72

Acknowledgements 75

References 75

Chapter 3 Genes encoding the immunoglobulin variable regions P.H Brodeur (Boston MA USA) 81

1 Introduction 2 3 G e n e families

V gene structure

3.1 Mouse VI, families

3.2 Human V,, families 86

3.3 Mouse V, families 87

3.5 Mouse V, families 89

3.6 Human V, families 89

3.4 Human V, families 88

IX

4 G e n e n u m b e r

4.1 Number of mouse V genes

4.2 Number of human V genes , , , , , , , , ,

5 Chromosome assignment

6 Gene organization

6.1 Igh locus organization , , , , , , , , ,

6.2 6.3 Igh locus organization

7 Polymorphism

8 Conclusions

I g K locus organization

90 90 02 94 95 95 99 101 101 105 Acknowledgements , , , , , , , _ _ 105

References 106

Chapter 4 Assembly of immunoglobulin vuriuble region gene segmeitls M Reth and L Leclercq (Cologne, FRG and Paris, France) 111

1 Introduction 111 112 112 I 1 4 I17 mechanism

2.1 Joining signals , , , , , , , , ,

2.2 Joining models ,

2.3 Control of joining , , , , , , ,

Order of rearrangement events during B cell tlcvelopment

3.1, Rearrangements at the IgH locus

3.2 Rearrangements at the light chain loci ,

Allelic exclusion of immunoglobulin gcne expression

3 4 129

Acknowledgements , , , 13 1 Kefcrences 131

Chapter 5 Immunoglobulin heavy chain cluss switching U Krawinkel and A Radbruch (Cologne, FRG) 135 1 Introduction 135

,

139

3.2 Isotype commitment

Molecular analysis ,

4

4.2 Long transcripts ,

4.3 Class switch recombination

4.4 Switch sequences

4.5 Switch recombination sites , , , , , , , ,

140

142

145

146

5 Conclusion

Reviews

References

Chapter 6 Immunoglobulin gene expression G.P Cook J.O Mason and M.S Neuberger (Cambridge UK)

I Introduction

2 Tumours as models

3

Patterns of immunoglobulin gene expression during B cell ontogeny

3.1 Changes in chromatin structure

3.2 Expression of productively rearranged lo

3.3 Expression of aherrantly rearranged loci

Expression of unrearranged loci

Processes regulating immunoglobulin gene expression

4.1 Promoter upstream elements

4.2 Enhancer elements

4.3 Other promoter elements

4.4 Transcription termination

4.5 RNA cleavageipolyadenylation

4.6 RNA splicing

3.4 4 4.7 Messenger R N A turnover

4.8 Translational and posttransl ation

5 Major aspects of cell-type specificity

5.1 Restricted cell-type specificity of immunoglobulin gene transcription

5.2 Control of the difference in m R N A abundance between B and plasma cells

5.3 The relative abundance of membrane and secreted immunoglobulin

5.4 Co-expression of two immunoglobulin classes

References

Chapter 7 The generation and utilization of antibody variable region diversity T Manser (Princeton NJ USA)

1 Introduction

2 Antigen independent diversity

2.2 Junctional diversity

2.3 T h e multiplicative potential of combinatori 3.1 T h e evidence for somatic mutation: a histor

3.2 Somatic mutation and the immune respons 3.3 Mechanistic considerations regarding somatic mutation

3.5 T cells and somatic mutation

3.6 2.1 Combinatorial diversity

3 Antigen dependent diversity

3.4 Somatic mutation and clonal selection theory

Antibody diversity and B cell subsets

147

149

149

153

153

153

154

154

155

156

156

157

157

159

164

164

165

166

167

167

168

168

169

170

173

173

177

177

178

178

182

185

185

186

188

190

194

197

197

XI

4 Summary 19X

Acknowledgements 198

Referenccs 198

Chapter 8 The immunoglobulin superfamily F Calabi (Cambridge UK) 203

1 Introduction 203

2 The immunoglobulin 3 The T cell receptor 206

3.1 T h e a/p T cell receptor 207

3.2 T h e $8 T cell receptor 215

3.3 Relationship between the expression of the aip and of the $8 T cell receptors 218

4 T h e major histocompatibility complex 219

4.1 Overall structure 221

4.2 Structure of the variable region 221

4.3 Structure of the constant region 224

4.4 Genetic basis of diversity 224

5 Accessory molecules of the a/p T cell receptor (CD4 and CDX) 226

6 Other members 228

6.2 Poly-immunoglobulin receptor 230

6.3 Neuronal cell adhesion molecule (NCAM) 230

6.4 Myelin associated glycoprotein/ncuronal cytoplasmic protein 3 (MAG/Ncp 3) 231

6.5 Viral members 231

7 Evolutionary considerations 231

6.1 Thy-1 229

Acknowledgements 233

References 233

Subject index 241

complementarity determining region diversity segment

antigen binding fragment from papain digestion of immuno- globulin

antigen binding fragment from pepsin digestion of immuno- globulin

crystallizable fragment from papain digestion of immunoglo- bulin

immunoglobulin Fc receptor framework region

immunoglobulin heavy chain membrane form of immunoglobulin heavy chain membrane form of immunoglobulin heavy chain hypoxanthine, aminopterin, thymidine medium human immunodeficiency virus

immunoglobulin interleukin 4 joining segment joining chain immunoglobulin light chain bacterial lipopolysaccharide long terminal repeat myelin associated glycoprotein major histocompatibility complex nucleotide region

neuronal cell adhesion molecule natural killer cell

N P 4-hydrox y-3-nitrophenylacetyl

Poly-IgR poly-immunoglobulin receptor

RS rearranging sequence ( K locus)

S region switch region

T c R T cell receptor

V regionisegment variable regionisegment

(ii) to trigger the elimination of foreign material In molecular terms this in- volves the binding of certain molecules (effector molecules) t o antibody-coated foreign material to trigger complex elimination mechanisms, e.g the complement system of proteins, phagocytosis by cells such as neutrophils and macrophages The effector systems are generally triggered only by antibody molecules clustered to- gether as on a foreign cell surface and not by free unliganded antibody This is crucial considering the high serum concentration of some antibodies

The requirements imposed on the antibody molecule by the functions (i) and (ii) are in a sense quite opposite Function (i) requires great antibody diversity Func- tion (ii) requires commonality, i.e i t is not practical for Nature to devise a differ- ent molecular solution for the problem of elimination for each different antibody molecule In fact the conflicting requirements are elegantly met by the antibody structure represented in Fig 1 The structure consists of three units Two of the units are identical and involved in binding to antigen - the Fab (fragment antigen binding) arms of the molecule These units contain regions of sequence which vary greatly from one antibody t o another and confer on a given antibody its unique binding specificity The existence of two Fab arms greatly enhances the affinity of antibody for antigen in the normal situation where multiple copies of antigenic de- terminants are presented to the host The third unit - Fc (fragment crystalline) -

is involved in binding to effector molecules As shown in Fig 1 the antibody mol- ecule has a four-chain structure consisting o f two identical heavy chains spanning Fab and Fc and two identical light chains associated only with Fab

Antigen binding Antigen binding

Complement triggering Cell receptor binding (Rheumatoid factor binding)

Fig I A schcinatic rcprcsentation o f antibody structure emphasising the relationship between struc-

ture ;ind function The antibody molecule can be thought of in terms ol three structural units Two Fab

arms bind antigen and a r c therefore crucial for antigen recognition The third unit (Fc) binds effector inolecules triggering antigen elimination The antihody niolccule thus links antigen recognition and an- tigen elimination The structure is composed of four chains Two identical heavy (H) chains span Fah and Fc regions and two identical light ( L ) chains arc a w x i a t c d with Fah alonc

The five classes o f antibodies o r immunoglobulins termed immunoglobulin G (IgG), IgM, IgA IgD and IgE differ in their heavy chains termed y, p, a , 6 and

tibody classes and this leads to the triggcring of different effector functions on binding to antigen, e g IgM recognition o f antigen might lead to complement ac- tivation whereas IgE recognition (possibly of the same antigen) might lead to mast cell degranulation and anaphylaxis (increased vascular permeability and smooth muscle contraction) Structural differences also lead to differences in the poly- merisation state of the monomer unit shown in Fig 1 Thus, IgG and IgE are gen- erally monomeric whereas IgM occurs as a pentarner IgA occurs predominantly

as a monomer in serum and as a dimer in seromucous secretions

T h e major antibody in the serum is IgG and as this is the best-understood an- tibody in terms of structure and function we shall consider it shortly T h e other antibody classes will then be considered in relation to IgG First, however a very brief overview of the structure and function o f the different immunoglobulins will

be presented [ 11

is the major antibody class in normal human serum forming about 70%

of the total immunoglobulin I t is evenly distributed between intra- and extravas- cular pools IgG is a monomeric protein and can be divided i n t o four subclasses

in humans It is the major antibody of secondary immune responses

represents about 105k of total serum immunoglobulin and is largely con- fined t o the intravascular pool I t forms a pentameric structure and is the predom- inant antibody produced early in an immune response, serving as the first line of

defence against bacteraemia As a membrane-bound molecule on the surface of B lymphocytes it is important as an antigen receptor in mediating the response of these cells to antigenic stimulation

IgG

IgM

3

1gA forms about 15-20% of total serum immunoglobulin where it occurs largely

as a monomer In a dimeric complex known as secretory IgA (sIgA) it is the major antibody in seromucous secretions such as saliva, tracheobronchial secretions, co- lostrum, milk and genitourinary secretions

represents less than 1 % o f serum immunoglobulin but is widely found on the cell surfaces of B lymphocytes where it probably acts as an antigen receptor analogously to IgM

though a trace immunoglobulin in serum, is found bound through specific receptors on the cell surface of mast cells and basophils in all individuals It is in- volved in protection against helminthic parasites but is most commonly associated with atopic allergies

IgD

IgE

2 Structure of I g G

2.1 Gerierul considerutions

I n IgG the Fab arms are linked t o thc Fc via a region of polypeptide chain known

as the hinge This region tends t o be sensitive to proteolytic attack generating the basic units of the molecule as distinct fragments The discovery of this action by Porter in 1959 [2] provided the first great insight into antibody structure In 1962

Porter proposed a four-chain structure for the IgG molecule [ 3 ] Since then, chem- ical and sequence analyses, notably by Edelman [4] have confirmed a four-chain structure consisting of two identical heavy (H) chains of inolecular weight approx- imately 50000 and two identical light (L) chains of molecular weight approxi- mately 25 000 The molecular weight of IgG is thus typically approximately 150000 The light chains are solely associated with the Fab arms of the molecule whereas the heavy chains span Fab and Fc parts as shown in Fig 2 A single disulphide bond connects light and heavy chains and a variable number, depending on IgG subclass (see below), connects the two heavy chains The latter connection is made

in the hinge region of the molecule Papain cleaves heavy chains to the amino-ter- minal side of these hinge disulphides producing two Fab and one Fc fragment Pepsin cleaves to the carboxy-terminal side producing a single F(ab’), fragment and smaller fragments of Fc including a carboxy-terminal pFc’ fragment

The light chains exist in two forms known as kappa ( K ) and lambda (A); the forms arc distinguished by their reaction with specific antisera In humans, K chains are somewhat more prevalent than A in mice A chains are rare [ S ] The heavy chains

can also be grouped into different subclasses the number depending upon the spe- cies under consideration I n humans there are four subclasses having heavy chains labelled y l , y2, y3 and 74 which give rise to the I g G l , IgG2, IgG3 and IgG4 sub- classes In mouse there are again four subclasses denoted IgG1 IgG2a, IgG2b and lgG3 The subclasses - particularly in humans - have very similar primary se- quences, the greatest differences being observed in the hinge region The cxist- ence of subclasses is an important feature as they show marked differences in their ability to trigger effector functions I n a single molecule, the two heavy chains are identical as are the two light chains; hybrid molecules are not found

5

Sequence comparison [6] of monoclonal IgG proteins, either myeloma proteins

or more recently antibodies generated by hybridoma technology, indicates that t h e carboxy-terminal half of the light chain and roughly three quarters of the heavy chain, again carboxy-terminal, show little sequence variation between different IgG molecules In contrast, the amino-terminal regions of about 100 amino acid resi- dues show considerable sequence variability in both chains Within these variable regions there are relatively short sequences which show extreme variation and are designated hypervariable regions There are three of these regions or ‘hot spots’

on the light chain and three on the heavy chain Since the different IgGs in the comparison recognise different antigens, these hypervariable regions are expected

to be associated with antigen recognition and indeed are often referred to as com- plementarity determining regions (CDRs) The structural setting for the involve- ment of the hypervariable regions in antigen recognition is discussed below Sequence comparison also reveals the organisation of IgG into 12 homology re- gions or domains [4] each possessing an internal disulphide bond The basic do- main structure is central to an understanding of the relation between structure and function in the antibody molecule and will shortly be taken up in some detail However, the structure in outline form is shown in Fig 2(b,c) It is seen that the light chain consists of two domains, one corresponding to the variable sequence region discussed above and designated the V , (variable-light) domain and the other corresponding to a constant region and designated the C, (constant-light) domain The IgG heavy chain consists of four domains, the VH and CHI domains of the Fab arms being joined to the CH2 and CH3 domains of Fc via the hinge Antigen binding is a combined property o f the V L ~ and V, domains at the extremities of the Fab arms and effector molecule binding a property of the C,2 and/or cH3 do- mains of Fc

From Fig 2(b,c) it is also clear that all of the domains except for cH2 are in close lateral association with another domain: a phenomenon described as domain pairing o r truns-interaction The CI,2 domains have two N-linked branched car-

(b) Domain representation Each heavy chain (shaded) is folded into two domains in the Fab arms, forms a region of extended polypeptide chain in the hinge and is thcn folded into two domains in the

Fc region T h e light chain forms two domains associated only with an Fab arm Domain pairing leads

to close interaction of heavy and light chains in the Fab arms supplemented by a disulphide bridge

T h e two heavy chains are disulphide bridgcd in the hinge (the number of bridges depending on IgG subclass) and are in close domain-paired interaction at their carboxy-tcrmini

(c) Domain nomenclature T h e heavy chain is composed of V , , , C,,1, C,,2 and Cl13 domains T h e light chain is composed of V, and C , domains All the domains are paired except for the CF,2 domains which have two branched N-linked carbohydrate chains interposed between them Each domain has a molecular weight of approximately 13000 Icading to a molecular weight of -50000 for Fc and Fab and

1.50000 for the whole IgG molecule Antigen recognition involves residues from the V I , and V, do-

mains, complement triggering the C,2 and Fc receptor binding the C,,2 and possibly the C,,3 domain

(see text)

bohydrate chains interposed between them The domains also exhibit weaker cis- interactions with neighbouring domains on the same polypeptide chain Fig 2 shows human IgGl in a Y-shaped conformation with the Fab arms roughly coplanar with the Fc This choice is for illustration only - the relative orientations of Fc and Fab and the involvement of the hinge in such orientations is a complex problem with possible significance for the function of IgG subclasses as discussed below

As described in detail in Chapter 2 the three constant domains and hinge of the

heavy chain and the constant domain of the light chain are encoded by separate exons The variable domains arise from genetic recombination events

2.2 Domain structure

Crystallographic studies of whole IgG and fragments of IgG [7-91 have revealed that each domain has a common pattern of polypeptide chain folding depicted in Fig 3 This pattern, the 'immunoglobulin fold', consists of two twisted stacked p- sheets enclosing an internal volume of tightly packed hydrophobic residues The arrangement is stabilised by an internal disulphide bond linking the two sheets in

a central position O n e sheet has four and the other three antiparallel p-strands These strands are joined by bends o r loops which generally show little secondary structure Residues involved in the P-sheets tend to be conserved while there is a greater diversity of residues in the joining segments Fig 3 shows the chain folding for a constant domain The @-sheets of the variable domain are more distorted than those of the C domain and the V domain possesses an extra loop

2.3 Structure of Fab

The four individual domains are paired in two types o f close truns-interaction (Fig

4) [103,17] T h e V, and V L , domains are paired by extensive contact between the

N

el

b6

Face Y /

Face Y ( f y )

Fig 3 Peptide chain folding o f a constant domain The segments fxl-4 (unshaded) and fyl-3 (shaded) form two roughly parallel faces of antiparallel /3-pleated sheet linked by an intra-chain disulphide bridge

(filled rectangle Cys-L31-Cys-2(K) (C,I, human IgG I ) Cys-261-Cys-321 (CJ), Cys-367-Cys-425 (CJ)) Between the p-pleated segments arc other scgments (bl-6) forming helices, bends and other structtires Segments fx3 1x4, f y l and b4 arc foreshortencd in this three-dimensional representation after B e a k and Feinstein [76]

7

-I1

Fig 4 Structure of a complex of the Fab fragment of an antibody molecule with antigen T h e complex

was formed between hen egg-white lysozyme and the Fab fragment of a mouse monoclonal anti-lyso- zyme antibody T h e diagram shows the crystal structure at 2.8 resolution Alpha carbon atoms only are shown thick lines being used for lysozymc and the heavy chain o f Fab and a thin line for the light

chain T h e tight pairing of V,, and V, and of C,,1 and C, domains of Fdb is clearly seen in this dia- gram T h e area of interaction between antihody and antigen is large approximately 20 X 30 A T h e region of Fab in contact with the antigen includes hypervariable loops from both heavy and light chains

with more interactions involving the former However the combining site is not a simple cleft enclosed

by the hypervariable loops but extends beyond them The antigenic site recognised o n lysozyme is not

a linear amino acid sequence - rather it is an arrangement of amino acids in three dimensions provided

by different parts of the linear sequence There is no cvidence for significant conformational changes occurring in either antigen or F a b on complex formation (This diagram was very kindly provided by

of these residues from the aqueous environment The Fab arrangement is further

stabilised by a disulphide bond between CHI and C, domains This bond cova- lently links the carboxy-terminal region of the C, domain with the e2 segment (hu- man IgG1) o r the b l segment (human IgG2,3,4) of the CH1 domain These latter positions, although widely separated in terms of CHI sequence, are close in space conserving the essential structure of the molecule

V,-CH1 and V,-C, cis-interactions are very limited allowing flexibility about the V-C switch region or 'elbow bending' [11,12] In the crystallographic analyses of Fab structures this is reflected in an elbow angle, i.e angle between the V,-V, and CH1-C, pseudo two-fold axes, varying between about 137" and 180" [8]

2.4 Antigen recognition

Further contact between V,, and V, domains in Fab is made by loops from each domain - the hypervariable loops or complementarity determining regions (CDRs)

- which come together in space to constitute the antigen binding site [14,15,17]

It is essentially the extreme variability of these loops on the common framework

of the immunoglobulin fold which provides for the enormous diversity of antigen recognition by antibodies while retaining the same basic structure Fig 5 illus-

trates variability in human light and heavy chain V regions to highlight the rela- tionship between framework regions and CDRs

This traditional view of antigen surrounded by CDRs may be somewhat simpli- fied Thus, it is predicted for example that one of the light chain CDRs is nat gen- erally in the antigen binding site [18], that framework residues can be involved in binding (see below) and that buried residues distant from the binding site may contribute to specificity [ 181 Nevertheless a 'CDR replacement' experiment using genetic engineering techniques indicates, for hapten binding at least, thk over- whelming importance of CDRs [112] Thus, the CDRs from the heavy chain var- iable region of a mouse monoclonal antibody binding NP-cap (4-hydroxy-3-nitro- phenacetyl caproic acid) were substituted for the corresponding CDRs of a human myeloma protein and the hapten affinity of the 'humanised' mouse antibody found

to be very similar to that of the original mouse antibody This elegant experiment also opens up the possibility of constructing human monoclonal antibodies from the corresponding mouse antibodies

Antigen binding sites have also traditionally been viewed as clefts in the anti- body structure There is n o a priori reason why this should be so - the antigen binding site could, for example, protrude and the cleft be found in the antigen molecule itself The traditional view has probably arisen from studies on myeloma proteins where the natural antigen was not known but, by extensive screening, small molecules were found which bound to the antigen binding site This view is chal- lenged by the recently solved crystal structure of the complex of the Fab fragment

of a mouse monoclonal anti-lysozyme antibody and lysozyme [16,17] shown in Fig

4 The area of interaction between antibody and antigen is large, approximately

20 x 30 A, and is formed by relatively flat complementary surfaces on the two proteins The region of Fab in contact with antigen includes CDRs from both heavy and light chains, with more interactions involving the former, but also extends to

in both plots and the four framework regions (FRs) as separating regions of relatively low variability (This diagram from [6] is reproduced with the kind permission of D r E A Kabat.)

some framework residues The third CDR of the heavy chain makes a particularly large contribution to antigen contact The antigenic site recognised on lysozyme is not a linear amino acid sequence - rather it is an arrangement of amino acids in three dimensions provided by different parts of the linear sequence (‘topographi-

cal’ determinant) Whilst in this case there is no evidence for significant confor- mational changes occurring in either antigen or Fab upon complex formation, lim- ited changes have been reported in a different Fab-protein antigen complex [130]

At the present time, the structure of only two Fab-antigen complexes have been

solved and one has to be wary in drawing general conclusions However, a number

Fig 6 Structure of the Fc fragment of human IgG (*), Alpha carbon positions: (0) approximate centres

of carbohydrate hexose units Coordinates were obtained from the Brookhaven Data Bank (after Dei- senhofcr [ 2 3 ] ) T h e pairing of C,,3 domains and the position o f carbohydrate between C,,2 domains is

clcarly seen in this view T h e contact between carbohydratc chains is much more extensive in rabbit

Fc Note that the heavy chains are described only from residue 238: residues 725-238 d o not show well-

defined electron density

11

of other complexes are currently being studied, and it is likely that general prin- ciples for the details of antibody-antigen recognition will emerge in the next 2-3 years One general principle that has been suggested is that antibodies recognise mobile regions of protein antigens [ 19,201 Others have suggested that it is surface accessibility or protrusion which is the primary requirement for antigenicity [21,22] The correlation of accessibility or protrusion and mobility then leads to the re- ported correlation of mobility and antigenicity

2 5 Structure of Fc

In the Fc of IgG [9,23] the two cH3 domains are paired in a pattern similar to that found for the CHl-C,- interaction (Fig 6) The two CH2 domains show no close interaction but have interposed between them two branched N-linked carbohy- drate chains which make little contact between one another in human Fc [23] but more extensive contact in rabbit Fc [30] I n the pairing of the C,3 domains, ap- proximately 1000 A' of surface per domain is involved in the interaction In the Ck,2 case the carbohydrate provides a substitute for the domain-domain contact and helps to stabilise the C,2 domain However, the CH2-carbohydrate contact area

is only about half that of, for example, the cH3-cH3 contact so that one might ex- pect a lower inherent stability for the cH2 domain Indeed the cH2 domain is more sensitive t o proteolytic degradation than the other domains of IgG [24] Domain stability has also been related to the apparent 'softness' of parts of the cH2 domain most remote from the CH2-cH3 interface as indicated by large temperature factors

or missing electron density in the crystal analysis of human Fc [ 2 3 ] A correspond- ing 'softness' has not, however, been found for rabbit Fc (Sutton B.J personal communication) Tables 1 and 2 show a comparison of known cH3 and c H 2 se- quences

The carbohydrate chains of the IgG C,2 domains are not a single oligosacchar- ide moiety but consist of a set of about 20 structures based on a mannosyl-chito- biose core which can be represented [13] as:

T A B L E 1

Comparison of IgG C,,3 domain sequences

Human IgGl [113], IgG2 [114], IgG3 11201, IgG4 (1151, mouse l g G l (1161, lgG2a ‘a’ allotype (1171, IgG2b ‘a’ allotype [1181, IgG3 (1261 and rabbit IgG [119] are translated from nucleotide sequences Guinea pig (G.pig) IgGl and IgG2 [ 5 9 ] were protein sequenced

lgGl E u numbering [6] is used throughout this chapter A dot indicates no residue at the position corresponding to the numbered human IgGl residue T h e right-hand column ( e l , f x l , etc.) indicates the approximate domain location of residues in human Fc and should be compared with Fig 3 Alter- nate (mostly hydrophobic) residues i i i the P-strands tend to be buried and show greater degree of con- servation than residues o n bends Cys-367 forms an intrachain bridge with Cys-425 close t o Trp-381 The asterisked arginine (R’) at position 435 in IgG3 highlights the involvement o f this position as an allotypic marker important in protein A (Section 3.2.) and rheumatoid factor (Section 3.6.) binding Other positions at which allotypic substitutions occur in human IgGs are 356, 358, 379, 384, 392, 431 and 436 [123]

> fx3

< b4

> fx4

<

As shown, four types of mannosyl-chitobiose cores are found ( 5 'bisecting' N -

acetylglucosaminei? fucose) and outer-chain variants include the presence o r ab- sence of galactose and sialic acid The heterogenous mixture of oligosaccharides released from an Fc preparation contains molecular species present in identical molar proportions to one another: a finding which led Rademacher and Dwek [13]

to propose a non-random pairing of certain structures Any possible correlation between carbohydrate heterogeneity and IgG subclass has to date neither been es- tablished nor definitively ruled out

15

I A B L E 2

Comparison of IgG C,,2 domain sequence\

N@ indicate$ that an N-linked carhohydrate chain is attached to Asn-2Y7 (note signal sequence Asn- x-Thr) u indicates unsequcnccd A dot indicates no rcsidue at the position corrcsponding to the num-

hcred IgG 1 residue Cys-261 bridges to Cys-321 close to Trp-277 Allotypic substitutions occur at po-

H u m a n Mouw G pig Rabbit

Sequence sources as in Table 1

The Fc carbohydrate chains have been suggested to adopt both structural and functional roles [ 13,251 Structurally they are important in conferring resistance to proteolysis, and possibly in maintenance of CH2-CH2 domain orientation and as- sembly Loss of carbohydrate has minimal effect on protein A binding [26] but a profound effect on monocyte binding [27] C l q binding [27] is only slightly af- fected although whole complement activation is abolished [28] (see also below) Recently, changes in the glycosylation pattern of total serum IgG have been as- sociated with rheumatoid arthritis and primary osteoarthritis [29],

In terms of exon sequence, the cH2 domain begins at Ala 231 (human IgG1, Table 2) However, in the crystal structures of human and rabbit Fc the heavy chain

is first clearly identified at Pro-238 This suggests that the residues between Ala-

23 1 and Gly-237 are disordered, presumably reflecting flexibility In this respect [25] the first few residues genetically defined as belonging to the cH2 domain be- long more, in structural terms, to the hinge which will now be discussed in some detail

2.6 The hinge: IgC subclusses

Mutant human IgGl myeloma proteins lacking the hinge (as encoded by the hinge exon, i.e the genetic hinge) have been crystallised and the structures solved to an intermediate resolution [31-331 These molecules display Fab and Fc structures similar or identical to the structures of the isolated fragments and are approxi- mately T-shaped They have been described as the 'structure of the antibody mol- ecule' However, some caution is necessary here as, for example, the hinge-de- leted IgGl Dob protein, unlike the intact IgGl molecule, neither activates complement [34] nor binds to monocyte Fc receptor [35] Unfortunately, crystal diffraction patterns of whole 1gG have been characterised by a lack of electron density associated with part of the hinge and the whole of the Fc, a phenomenon which has been related to hinge flexibility [36,37] In considering IgG structure, therefore, one needs to consider the hinge Further, since the principal difference between subclasses tends to be in the nature of the hinge, this also serves t o in- troduce IgG subclass structure

The term hinge arose from electron micrographs of rabbit IgG complexed to small bivalent haptens, which showed Fab arms assuming angles relative t o one another from nearly 0" (acute Y-shaped) to 180" (T-shaped) [38,39] The function

of this hinge flexibility has generally been seen as allowing divalent recognition of variably spaced antigenic determinants Other physical techniques have demon- strated hinge flexibility in solution, although the degree of flexibility differs be- tween IgG subclasses [40,42] This is to be expected, given the differences in the primary structures of the hinges from different IgG subclasses illustrated in Table

3 The hinge shown, the 'structural' hinge [25], consists of the genetic hinge and the connecting region to the C,,2 domain described above (hinge-link or lower

hinge) This hinge is readily divided into three regions as illustrated [25,41] The upper hinge can be seen as allowing flexibility of the Fab arms relative to one an- other (Fab-Fab flexibility) and allowing rotation of the Fab arms The middle hinge contains the interheavy cysteine disulphide bridges and a high content of proline, and probably adopts a relatively rigid double-stranded structure In the case of hu- man IgG3, this is an extremely elongated structure of about 50 residues containing

11 cysteines and 19 prolines The lower hinge is probably responsible for the flex- ibility of Fc relative t o Fab (Fc-Fab flexibility)

Segmental flexibility of IgG is correlated with hinge length Thus, for a series of mouse monoclonal antibodies showing the same (anti-dansyl) combining site, na- nosecond fluorescence polarisation spectroscopy indicates the order of flexibility IgG2b > IgG2a > IgGl [40] Similarly guinea pig lgG2 shows greater segmental flexibility than IgGl [42] The flexibility has been associated particularly with the upper hinge [25,41] There is an association between hinge flexibility and effector function seen in its extreme form by the lack of flexibility and loss of effector func- tion associated with hinge-deleted IgG However flexibility is closely related to the proximity of Fab and Fc and it is difficult to establish the causal factor in mod- ulating effector function This will be discussed in greater detail below

The equilibrium conformation of the monomeric human IgG subclasses have

TABLE 3

Comparison of hinge sequences of IgG

Sequence sources as in Table 1 Residues are aligned so that the positions of the first and last interheavy disulphide bridge cysteines coincide In human

IgGl, mouse IgGl and rabbit IgG, Cys-220 (Eu numbering) forms a disulphide bridge to the light chain and thus the residues 216220 belong structurally

to Fab rather than the hinge Crystallography indicates that Pro-238 is the first residue forming part of the folded C,2 domain so that the residues from

Ala-231 to Gly-237 are assigned structurally to the hinge Papain cleaves human IgGl at position 224, pepsin at position 234 The sequence of human

IgG3 corresponding to the above is given in Table 6

Fig 7 Models proposed for the solution conlormations o f human IgG subclasses The models are pro- posed on the basis of sedimentation data with supporting small-angle X-ray scattering data [48] T h e hinge-delcted I g G l Doh protcin is used as a reference IgG2 and IgG4 arc shown to resemble Dob in the close approach of Fat7 and Fc: for IgG2 thc Fab arms are suggested to fold back In IgGl the hinge

is apparent and the Fab arms are non-colinear as illustrated I n IgG3 the central or middle hinge (Ta- ble 6) is about 90 A long i n agreement with electron microscope studies 1491

been investigated by physical techniques including hydrodynamic and small-angle X-ray and neutron scattering studies [44,48] There is more or less general agree- ment that IgG3 is a very extended molecule as predicted from modelling studies [9,49] with a long middle hinge (Table 6), although an alternative model has been suggested [46] We have recently proposed the structures presented in Fig 7 [48] These structures, although average conformations for monomeric IgGs, do show interesting correlates with the functional activity of IgG subclasses in an associated state For example, complement activation is most efficient for IgG3 especially and also for IgGl where the Fab arms are less likely to interfere with binding sites on

Fc than in IgG2 or IgG4

21

2.7 Isotypes, allotypes und idiotypes

T h e variability of antibodies is often conveniently divided into three types Iso- types a r e variants present in all healthy members of a species: immunoglobulin

classes a n d subclasses are examples o f isotypic variation involving the constant re- gion of t h e heavy chain Allotypes are variants that are inherited as alternatives

(alleles) with not all healthy members of a species therefore inheriting a particular allotype Allotypes occur mostly as variants of heavy-chain constant-region genes,

in man in all four IgG subclasses, IgA2 a n d IgM T h e nomenclature of human im- munoglobulin allotypes is based o n the isotype o n which it is found (e.g G l m de- fines allotypes on a n IgGl heavy chain, Km defines allotypes o n K light chains) followed by a n accepted WHO numhering system The positions of some allotypes

a r e noted in t h e legends t o Tables 1-7

T h e variable region of an antibody can act as a n antigen, and the unique deter- minants of this region that distinguish it from most other antibodies of that species

a r e termed its idiotypic determinants T h e idiotype o f an antibody, therefore, con- sists of a set of idiotypic determinants which individually are called idiotopes Po- lyclonal anti-idiotypic antibodies generally recognise a set of idiotopes whilst a monoclonal anti-idiotype recognises a single idiotope ldiotypes are usually spe- cific for a n individual antibody clonc (private idiotypes) but are sometimes shared between different antibody clones (public, recurrent o r cross-reacting idiotypes)

A n anti-idiotype may react with determinants distant from the antigen binding site

it may fit t h e binding site a n d express the image of the antigen o r it may react with determinants close to the binding site and interfere with antigen binding Sequenc- ing of an anti-idiotypic antibody generated against an antibody specific for the polypeptide GAT antigen in mice revealed a CDR3 identical to that of the anti- gen, i.e the anti-idiotype contains a true image of the antigen [ l l l ]

Fc interaction which is understood in detailed molecular terms

3.2 Interaction with protein A

Protein A is a major cell wall component of most strains of Staphylococcus aiivem

which binds to the Fc regions of immunoglobulins of a variety of subclass and spe-

cies with varying affinity (251 This reaction has been widely exploited in immuno- and histochemistry The protein consists of five homology regions binding to Fc and another region that does not bind to Fc but binds to cell walls [ 1241 Trypsin digestion can be used to isolate active fragments each corresponding to an ho- mology region and of molecular weight about 7000 (SO] The fragments bind to Fc with a stoichiometry of 2: 1 The multivalency of both IgG and protein A with re- spect to one another means that if the two are added to one another in the correct proportions they will form extended complexes and precipitate The structure of

a complex of one of the fragments (fragment B) with human IgG Fc has been solved crystallographically to 2.8 A [23] and indicates that protein A binds at a site be- tween C,2 and C,3 domains This binding is observed for the human IgG sub- classes IgG1, IgG2 and IgG4 and also for IgG3 proteins bearing allotypic markers characteristic of Mongoloid populations ( 5 I ] I n IgG3 proteins from Caucasian populations the protein A contact residue His-435 of the above IgGs is replaced

by Arg This lengthy side chain prevents the formation of favourable IgG-protein

A contact so that such IgG3 proteins do not bind protein A

Protein A binding at the C1.,2-C,,3 interface is not perturbed by deletion o r re- duction and alkylation of the hinge or by aglycosylation of IgG [25,26,24]

3.3 Coniplement activation

The classical pathway of complement is a cascade system generating a variety of potent biological molecules including anaphylatoxins and chemoattractants and leading ultimately to lysis of antibody-coated cells [ S 2 , S 3 ] In health foreign cells

will be the primary target I n disease, host tissue may be attacked on a large scale e.g in autoimmune disorders The pathway is triggered by the interaction of the first complement component, C1, with IgG in an associated state, i.e coating a target cell or aggregated by antigen in an immune complex The pathway is clearly not triggered by monomeric IgG which is at high concentration in the serum C1 is a complex of the complement components C l q , C l r and Cls It is the sub- component C l q which interacts with the C,2 domain of IgG to initiate the enzy- matic process of the pathway Clq is a molecule having the appearance of a ‘bunch

of tulips’ [52] and it is multivalent in its binding to IgG (Fig 8) This multivalency

is probably the key to why complement is only triggered by IgG in an associated form Binding of C l q to monomeric IgG is only weak ( K , - lo4 M-’), whereas

binding to associated IgG and the consequent use of two or more of the tulip heads makes binding much tighter ( K ; , - 10’ M-’) and allows t h e activation process to proceed [25] Theories of activation which involve binding of antigen to the Fab arms of IgG and the induction of conformational changes which are passed down the molecule to Fc, thus affecting the interaction with C l q , are now rejected by most workers [54,25] The reasons for this rejection are many In the first in- stance, C l q multivalency in itself appears an adequate explanation of activation by associated IgG and this is highlighted by the observation that isolated C l q heads bind to IgG aggregates with a similar affinity as intact C l q to monomeric IgG (5.51 Further, C l q binding and CI activation occur whether the association of IgG is

23

Fig s(a.1~) ( a ) A schematic view of the binding of complement Clq to two IgG molecules on a cell

surface C l q is a hexavalent molecule o l molecular weight approximately 460000 It adopts a structure likened to a hunch o f tulips in which six collagenous stalk regions are connected to six globular head regions which contain thc IgG binding site The dimensions for C l q used here correspond to longer

'arm' regions than those originally proposcd from electron microscope studies There is evidence from physical measurements that the arms of C I q posscss some tlexibility although this appears less in C1 than isolated C l q Any flexibility of Clq may complement that of Fc in reducing steric requirements

in the Clq-IgG interaction

(h) A model of C l q binding to dislocated IgG molecules In this speculative model [lOS] it is sug-

gested that hinge flexibility allows dislocation ol the Fah arms of the IgG molecule out of the plane of

the Fc allowing Fc-Fc interaction The model has similarities to that envisaged for Clq binding to IgM based on electron microscope studies

achieved by antigen, heat or chemical cross-linking and, in the case of chemically

cross-linked IgG, they are unaffected by combining site occupancy [25] Attempts

to identify a conformational change in IgG upon antigen binding which correlates with Clq binding, have not been successful [54] Finally, in the case of human IgG3,

it is particularly hard to visualise II common conformational change being passed through the extended hinge region as the result of the binding of a wide variety of different antigens at the extremities of the Fab arms

The flexibility of Fc is complemented by some flexibility in the arms of the C l q molecule [56,58] which may be important in complement triggering Thus, flexi-

bility reduces the stringency of steric requirements when Clq binds to an array of IgG molecules

There is general agreement about the importance of charged groups in the in- teraction of IgG and Clq but controversy over the precise location of the Clq

binding site on IgG [ 2 5 ] The domain responsible for C l q binding is concluded to

be c H 2 [25] as the Facb fragment of rabbit IgG (lacking the C,3 domains) binds C1 with an affinity comparable to IgG and isolated CH2 domains bind C1 with an affinity comparable to Fc Early interest in the cH2 domain centred on Trp-277 and nearby residues, but this was quenched by the demonstration that Trp-277 is

buried in the crystal structure of Fc Three Clq binding sites have more recently

been proposed The first [S9] involves residues in an extended chain region, be- tween Lys-290 and (3111-295 in human IgG1 the second [60] involves residues in the overlapping extended chain region between His-285 and Arg-292, and the third [82] involves residues on the last two anti-parallel /?-strands of the C,2 domain (Gln-318, Lys-320, Lys-322, Pro-331, (3111-333, Thr-335, Ser-337) IgGs of differ- ent subclass and species show differing affinity for C l q and the above proposals attempt to account t o some extent for this in terms of sequence differences be- tween isotypes However, there is an added complication in that it appears that the proximity of Fab arms can modulate the expression of the C l q binding site on

Fc [25] Thus hinge-deleted IgGl does not bind C l q [34] Further and more strik- ingly, IgG4 does not bind C1 whereas its Fc fragment (Fc4) does [61] This is con- sistent with the close approach of Fc and Fab in IgG4 visualised in Fig 7

In comparing isotype behaviour with respect to complement it is important to distinguish C l q binding C1 binding, C1 activation and whole complement acti- vation The most widely used assay is the measurement of the end-product of the whole complement cascade, i.e cell lysis The inability of an IgG to promote ef- ficient lysis does not necessarily indicate an inability to bind C l q A later stage may be implicated For example, it appears that C l q binding is not always directly related to C1 activation [62,63] and furthermore later components of comple- ment, e.g C4b, C3b also interact with IgG A further complication is that a small change in C l q binding affinity may, through the amplification nature of the com- plement cascade process produce a large change i n whole complement activation measured as cell lysis

Comparison of the human IgG subclasses provides the following view [64] All the subclasses in a monomeric state bind C l q with measurable affinity with the or- der of binding constants IgG3 > lgG1 > lgG2 > lgG4 IgG3 and IgGl activate C1 and whole complement efficiently IgG2 is less efficient in complement acti- vation IgG4 does n o t appear to bind C1 and does not activate complement In other species, mouse IgG3, IgG2a and IgG2b, guinea pig IgG2, rabbit IgG, rat IgG2b, lgGl and IgG2a bind C l q and activate complement Mouse IgG1, guinea pig IgGl and rat IgG2c bind C l q weakly if at all, and do not significantly activate complement [25,65,66]

Whilst it is clear that C1 interacts with associated IgG primarily through C l q , there have been suggestions that C l r and/or C l s may also interact weakly with sites

on IgG (e.g [67]) but this has not been clearly demonstrated [25] Activated forms

of C3 and C4 also bind to IgG [53] via covalent interaction with residues in the

heavy chain of Fab The former interaction is important not only in the classical pathway but also for IgGs such as rabbit IgG which activate the alternate pathway

3.4 lnteractiori with cellular Fc receptors

Receptors for the Fc region of IgG are found on a number of cell types and are associated with a variety of functions including phagocytosis (monocytes, macro- phages, neutrophils), antibody-dependent cellular cytotoxicity (monocytes, mac-

25

rophages, lymphocytes), maternofoetal transport (trophoblast) and possibly im- munomodulation (lymphocytes) ‘Fc receptor’ is an operational term and does not imply that the same molecular species is found o n the different cell types [68] In- deed, there is good evidence that a number of molecular species are involved Human leucocyte IgG Fc receptors (FcR) fall into three categories [69] as de- fined by a number of criteria but most especially by reactivity with specific mon- oclonal antibodies FcRI is a 72000 molecular weight receptor found on mono- cytes which binds monomer IgG with high affinity ( 5 x 10’ M - ’ ) FcRII is a 40000 molecular weight receptor found o n monocytes, granulocytes, platelets and B cells which binds well to associated IgG but only very weakly to monomer IgG ( < l o 6 M-’) FcR,, is a 50-70000 molecular weight receptor found on granulocytes, mac- rophages, K and NK cells, again binding associated but not monomer IgG FcRI binds the human IgG subclasses with the rank order IgGl = IgG3 > IgG4 The subclasses are shown to bind to the same receptor by competition experiments IgG2 does not bind The subclass specificity of FcRII and FcR,, has not been de- finitively demonstrated at this stage

Murine leucocyte Fc receptors show both similarities to and differences from human Fc receptors [69] FcRl found on mononuclear phagocytes, shows rela- tively high affinity (10’ M-I) for monomeric mouse IgGZa FcRII, found o n mononuclear phagocytes, granulocytes and B cells, preferentially binds aggre- gated IgG2 and I g G l This receptor has recently been cloned and found to belong

to the immunoglobulin supergene family [129] A third receptor binds mouse IgG3

specifically

Until recently, the conventional wisdom was that complement interacts with the cH2 domain and cellular Fc receptors with the cH3 domain of IgG This has been challenged by a number of workers [25] The binding of IgG to the human mono- cyte Fc receptor (FcRI) will be briefly discussed

The cH3 domain was originally favoured, owing to the (weak) ability of the pFc’ fragment of IgG (cH3 domain dimer) to inhibit the interaction of IgG and mono- cyte Fc receptor Using domain-specific anti-human IgG monoclonal antibodies this has been shown to arise from contamination of pFc’ preparations by small amounts

of parent IgG [3S] In contrast, cH2 domain involvement in receptor binding is favoured by the loss of binding associated with hinge deletion [35] and aglycosy- lation of IgG [27] Further, anti-human IgG monoclonal antibodies specific for the cH3 domain and the CH2icH3 domain interface do not inhibit the Fc receptor-IgG interaction and are still able to bind to receptor-bound IgG [70] Antibodies spe- cific for a cH2 epitope are inhibitory and do not bind to receptor-bound IgG It has, therefore, been suggested that cH2 is the critical domain for monocyte re- ceptor binding [70,71] A similar conclusion is reached for mouse IgG2b binding

to mouse macrophage Fc receptor based on the use of mutant deleted proteins [72] There is, however, still controversy in this area, which is complicated by the existence of the different types of Fc receptor [2S,69] Based on a sequence com- parison of IgGs showing differing affinity for the monocyte Fc receptor, the model shown in Fig 9 has been proposed [71] The IgG binding site is suggested to coni- prise residues of the lower hinge i.e those between the hinge disulphides and the