THE MOLECULAR STRUCTURE OF CELL ADHESION...

P1: JER/MKV P2: RPK/PLB QC: RPK

May 14, 1997 10:36 Annual Reviews AR032-27 AR032-27

Annu. Rev. Biochem. 1997. 66:823–62Copyright c© 1997 by Annual Reviews Inc. All rights reserved

THE MOLECULAR STRUCTURE OFCELL ADHESION MOLECULES

Cyrus ChothiaMRC Laboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, England

E. Yvonne JonesLaboratory of Molecular Biophysics, Oxford Centre for Molecular Sciences, SouthParks Road, Oxford OX1 3QU, England

KEY WORDS: immunoglobulin classification, binding sites, adhesion complexes

ABSTRACT

Considerable advances have been made in our knowledge of the molecular struc-ture of cell adhesion molecules, their binding sites, and adhesion complexes. Forthe cadherins, protein zero, and CD2, additional experimental data support theinsights obtained from structural analysis of their domains and molecular modelsof their adhesion complexes. For neural cell adhesion molecules, L1, fibronectin,tenascin-C, integrins, and vascular cell adhesion molecules, the molecular struc-ture of domains, and in most cases their binding sites, have been elucidated. Thesubstrate recognition sites in some of these molecules possess rate constants forassociation and dissociation that permit both rapid cell migration and, throughavidity, high-affinity cell-cell interactions.

CONTENTS

PERSPECTIVES AND OVERVIEW. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 824

IMMUNOGLOBULIN SUPERFAMILY DOMAINS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 825The V, C1, C2, and I Sets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 825Prediction of the Structure of IgSF Domains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 828Evolution of the Immunoglobulin Superfamily. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 828

CADHERINS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 829Structure of Cadherin Domains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 829Calcium-Binding Sites. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 831Dimer Formation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 832The Adhesion Complex. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 832Calcium Regulation and the Specificity of Adhesion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 833

8230066-4154/97/0701-0823$08.00

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



824 CHOTHIA & JONES

NEURAL CELL ADHESION MOLECULES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 833Structure of the NCAM Domains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834General Structure of NCAM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834The NCAM Adhesion Complex. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 836

L1 AND NEUROGLIAN CELL ADHESION MOLECULES. . . . . . . . . . . . . . . . . . . . . . . . . 836Structure of Immunoglobulin Superfamily Domains in L1. . . . . . . . . . . . . . . . . . . . . . . . . 838Structure of the Fibronectin Type III Domains in L1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 838Mutations in L1 that Cause Neurological Diseases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 839

PROTEIN ZERO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 839Structure of the Extracellular Domain of Protein Zero. . . . . . . . . . . . . . . . . . . . . . . . . . . . 840Oligomerization and Adhesive Interactions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 840Mutations in Protein Zero that Cause Neurological Diseases. . . . . . . . . . . . . . . . . . . . . . 842

FIBRONECTIN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 843Structure of the Fibronectin Type I and Type II Domains. . . . . . . . . . . . . . . . . . . . . . . . . . 843Structure of Fibronectin Type III Domains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 845Connections Between FnIII Domains in Fibronectin and Other Proteins. . . . . . . . . . . . . 846The Integrin- and Heparin-Binding Sites. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 846

TENASCIN-C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 847

INTEGRINS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 847Structure of the N-Terminal Domain of the Integrinα-Subunits. . . . . . . . . . . . . . . . . . . . . 848The Metal and Ligand Binding Site inα-Subunits. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 848Structure of the I Domains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 850The Metal-Binding Site in the I Domains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 850The I-Domain Adhesion Site. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 851

VASCULAR CELL ADHESION MOLECULE 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 851Structure of the Two N-Terminal Domains of VCAM-1. . . . . . . . . . . . . . . . . . . . . . . . . . . . 851Integrin-Binding Sites in VCAM-1 and ICAMs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 853

CD2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 853Structure of the CD2 Extracellular Domains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 854Connections Between IgSF Domains in CD2 and Other Proteins. . . . . . . . . . . . . . . . . . . 854CD2 Adhesion Site. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 856

CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857

PERSPECTIVES AND OVERVIEW

In metazoa, differentiation of cell structures during development, tissue forma-tion, and coordinated dynamic interactions of cells in various processes (such asthe immune response) requires direct contact between cell surfaces and betweencells and the extracellular matrix. The proteins through which these contactsare made are the cell adhesion molecules.

The first biochemical studies of cell adhesion molecules occurred more thantwo decades ago (1, 2). Since then, these molecules have been the subject ofthousands of papers, so a critical review must be restricted to only one aspectof the subject. An article on this topic appeared in this series in 1991 andexamined the implications of cell adhesion molecules for molecular histology(3). Here, the focus is on what is currently known about the molecular structure,binding sites, and adhesion complexes of cell adhesion molecules.

Most of the molecules discussed are built, at least in part, from immunoglo-bulin-like domains; hence, we begin with a general introduction to the structureof this superfamily. We then review the proteins involved in the formation and

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



CELL ADHESION MOLECULES 825

architecture of various solid tissues, including cadherins, neural cell adhesionmolecules (NCAMS), the L1 cell adhesion molecule, and protein zero (P0).Once complexes between these molecules are formed during differentiation,these contacts are generally stable for the life of the cells. We then discussthe extracellular proteins fibronectin and tenascin. Finally, we review proteinsinvolved in the movement and interactions of leukocytes: integrins, vascularcell adhesion molecules (VCAMs), and CD2. All the structural information wediscuss has become available during the past five years, most within the pastyear. Two other recent reviews also present some aspects of this work (4, 5).

Not discussed here is the complex between the T-cellαβ receptor and MHCmolecules. Recent work on the Vα-Cα (6) and Vβ (7) domains of T-cellαβreceptors has provided the first direct information on their molecular struc-ture. Although these results are of great interest, we expect further work to bepublished in the near future that makes inclusion of this topic somewhat prema-ture. Also not discussed here are proteins involved in carbohydrate recognition,which have been described in three excellent and recent reviews (8–10).

IMMUNOGLOBULIN SUPERFAMILY DOMAINS

The V, C1, C2, and I SetsOf the nine cell adhesion molecules discussed here, five (NCAMs, L1, P0,CD2, and VCAMs) are built wholly or largely of domains that belong to theimmunoglobulin superfamily (IgSF). Therefore, we discuss the classificationand structural relationships of the different IgSF domains in these proteins.

Sequences of IgSF members are diverse: Residue identity between domainswithin individual proteins, and among related domains in different proteins, isoften in the 10–30% range. Although these domains all have a similar corestructure formed by twoβ-sheets packed face-to-face, other features of thestructures can display considerable variation. The peripheral regions—the edgestrands of theβ-sheets and the loops that link strands—can have quite differentconformations. For the most distantly related domains, more than half of eachstructure is comprised of peripheral regions that have different folds (11).

Although IgSF members are diverse, inspection of the first structures to bedetermined showed that they could be grouped into related sets (12). Membersof a given set have peripheral regions whose conformations are very similar,even though sequence identity is low. In the original classification of IgSFdomains, three sets were identified: V, C1, and C2. The first two were classifiedon the basis of known structures, the third on the basis of certain sequencecharacteristics that did not fit the known structures (12). Subsequent structuredeterminations showed that the sequence characteristics originally defined for“C2” domains are shared by proteins that actually form two sets of distinctstructures: a redefined C2 set and a new I set (13).

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



826 CHOTHIA & JONES

(a)

Figure 1 (Continued)

Figure 1 shows schematic diagrams of the structure usually common to themembers of each set. There are distinct differences among the common struc-tures of the different sets, as can be seen if the diagrams are superimposed(Figure 1). In terms of strands only, and taking the I-set structure as a referencepoint, the V set has an additional C′′ strand, the C1 set lacks the A′ strand, andthe C2 set lacks the A′ and the D strands. There are also differences in thelength of strands common to different sets. For example, as shown in Figure 1,the strands extend further toward the upper part of the structure in the C1 setthan in the other sets, and the strands extend further toward the lower part ofthe structure in the I and V sets than they do in the other sets.

Each of the IgSF domains in the cell adhesion molecules discussed here canbe readily assigned to one of these sets. The NCAM and L1 domains, anddomain 1 of VCAM, belong to the I set; P0 and domain 1 of CD2 belong tothe V set; and domain 2 of CD2 and of VCAM belongs to the C2 set. Whena domain in a cell adhesion molecule is assigned to the I, V, or C2 set, this

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




(b)

Figure 1 The structural sets of the immunoglobulin superfamily. The structure that is usuallyconserved in the V, C1, C2, and I sets is shown (a, top of b). Solid circles representβ-sheet residues,open circles represent residues in conserved loops, and horizontal broken lines represent hydrogenbonds.β-sheet strands are labeled A, A′, B, C, C′, C′′, D, E, F, and G. The C1-set structures donot have strands A′ and C′′; the C2-set structures do not have strands A′, C′′, and D; and the I-setstructures do not have the C′′ strand. To help compare the structures of the different sets, the sitesusually occupied by the Cys residues that form the intersheet disulphide bridge are indicated bydiamonds. These Cys residues are strongly conserved in the V and C1 sets but less so in the I andC2 sets. To illustrate the differences in the structures, the I, V, and C1 sets are superimposed (b,bottom). In this part, filled circles represent sites common to the I, V, and C1 sets; V, C1, and Irepresent sites conserved in those sets.

classification means that its structure is similar to one of those illustrated inFigure 1, except that, in some cases, the extent of some strands is somewhatgreater. In those instances, any important differences are described.

A broader classification than that described above has been proposed (14)that encompasses IgSF members and all of the other proteins that have the samefold (i.e. proteins that have the same major secondary structure in the samegeometrical arrangement and with the same connections as those found in an

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



828 CHOTHIA & JONES

IgSF member). This new classification includes fibronectin type III domains,the chaperone PapD, and domains from certain bacterial enzymes. The struc-tural similarity of these proteins may reflect evolutionary relationships or mayarise fortuitously from the fact that the physics and chemistry favor certaintypes of folds in proteins. Four classes are defined: v, which corresponds tothe V set; c, which corresponds to the C1 set; s, which corresponds to the C2set but includes the fibronectin type III domains and PapD; and h, which coverscertain bacterial domains (14).

Prediction of the Structure of IgSF DomainsConformation of a protein is determined mainly by the set of key residues thatare involved in intramolecular packing and hydrogen bonding and/or those thatare able to take up an unusual configuration. If two proteins have homologoussequences, they will have similar structures in their core regions (13, 14a). Thedegree of structural similarity beyond the core depends on the extent to which thedifferent proteins conserve sequence length and key residues in other regions.If these other regions are the same size, and if the key residues that determinethe known conformation in one protein are also present in a homologous pro-tein (or if suitable alternative residues are present), the region in question willhave the same or a very similar three-dimensional structure in both proteins.If the key residues (or suitable alternatives) are not present, or if insertionsor deletions occur, the regions in question will have different conformations(11–13, 15).

Knowledge of the key residues within a particular structure allows a reason-able deduction to be made about whether that same structure is conserved inhomologs that have only low sequence identity (i.e. 10–20%) (12, 13). De-termination of the key residues in an I-set protein, “telokin,” led to the predic-tion that many of the corresponding IgSF domains in cell adhesion molecules,surface receptors, and muscle proteins would also have I-set structures (13).Subsequent structure determinations showed this conjecture to be true for thedomain 1 segments of VCAMs and NCAMs, and for the IgSF domains in themuscle proteins, titin and twitchin (see below; 15, 16).

Evolution of the Immunoglobulin SuperfamilyComparison of the IgSF sets illustrated in Figure 1 shows that the I set has astructure that combines features of the structures of the V, C1, and C2 sets: TheI set resembles the V set on the A′-side of the structure and the C1 set on theC′-side of the structure. This intermediate nature of the I set suggests that itcould be the ancestor of the V, C1, and C2 sets.

This view is supported by evidence from phylogenetic comparison of IgSFmembers. For example, bacterial members belong to the I set (17), as does adomain found in marine sponges (18; A Bateman & C Chothia, unpublished

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




data), and IgSF domains in the proteins fromCaenorhabditis elegans(15).The V-, C1-, and C2-set structures, on the other hand, seem to be confined todescendants of the common ancestor of cartilaginous fishes (sharks and rays)and humans. Conclusions drawn from such phylogenetic comparisons aretentative because of the incomplete nature of currently available sequence data.Nevertheless, the I-set structure appears to be the precursor to the modern IgSF(see also 3, 12).

CADHERINS

Cadherins are adhesion molecules involved in embryonic morphogenesis andthe formation of stable interactions among the cells in solid tissues and in ep-ithelia (19, 19b). Nearly all forms have five homologous extracellular domains(Figure 2a), a single membrane–spanning segment, and a cytoplasmic regionthat is anchored through catenins to cytoskeletal actin microfilaments (20).

Cadherins are found in animals ranging from nematodes to humans. Higherorganisms as a group contain several types of cadherins. Each cadherin typeconfers on the cell in which it is expressed specific cell adhesion properties(21). In adhesion, cadherin dimers on one cell make homophilic contacts,through domain 1, with cadherin dimers on another cell (22). Thus, domain1 dictates binding specificity. Both dimerization and adhesion are calciumdependent. In the absence of calcium, a cadherin molecule is monomeric (withthe extracellular domains displaying extensive motion relative to each other)and is unable to form adhesion complexes (23).

Structure of Cadherin DomainsStructures have been determined for domains 1 (24, 25) and 2 (25) of murineepithelial (E) cadherin and for domain 1 of murine neural (N) cadherin (26).All three domains have the same overall structure: seven strands ofβ-sheet(A′, B, C, D, E, F, and G). In the case of domain 1 of N-cadherin and domain2 of E-cadherin, there is also a small A strand. These strands form two twistedβ-sheets—one with strands A′, G, F, and C, and the other with strands (A),B, E, and D. The twoβ-sheets pack face-to-face, like two slices of bread in asandwich (Figure 2b). The arrangement of these strands is similar to that foundin the I-set members of the IgSF (24, 27).

Domain 1 of N-cadherin has 57% residue identity to domain 1 of E-cadherin;correspondingly, these domains have essentially the same conformation (exceptfor four residues at the N terminus and small shifts in the BC and FG loops).Domains 1 and 2 of E-cadherin share only 25% identity; as expected, thereare more structural differences between them. However, the core regions indomains 1 and 2 consisting of all, or almost all, of the A′, B, C, D, E, F, andG strands and a few residues adjacent to the strands (some 66 residues in all)

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



830 CHOTHIA & JONES

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




have the same conformation. It is only the residues forming the loops that linkthe strands in domains 1 and 2 that differ in conformation to varying degrees.

The sequences of domains 3, 4, and 5 have decreasing numbers of identitiesin comparison to the sequences of domains 1 and 2: as little as 10–16% in thecase of domain 5. Nevertheless, alignments of the sequences of domains 3,4, and 5 onto the known structures of domains 1 and 2 indicate that the coreβ-sheet structure is most likely conserved in all three domains (24, 27). Theregions involved in binding calcium (see below) are also conserved. On theother hand, the loop regions at equivalent positions in different domains showvery little conservation in sequence, length, or predicted conformation.

Calcium-Binding SitesThe most complete picture of the sites for binding of Ca2+ and the effect ofCa2+-binding on the organization of cadherins is provided by the structure ofdomains 1 and 2 of E-cadherin (25). Domain 1 (residues 1–99) is linked bya seven-residue segment (residues 100–106) to domain 2. Three Ca2+ bind inthis linker region (Figure 2c). One ion is bound by residues Glu-11, Asp-67,and Glu-69 in domain 1 and by Asp-103 in the linker. The second is boundby residues Asp-134 and Asp-136, the main chain oxygen of residue 143, andAsp-195 in domain 2 and by Asn-102 and the main chain oxygen of residue 104in the linker. The third Ca2+ is bound by all three regions: Glu-11 and Glu-69in domain 1; Asp-100, the main chain oxygen of residue 101, and Asp-103 inthe linker; and Asp-136 in domain 2. Of the twelve residues that bind to Ca2+,four bind to two ions. These extensive interlocking interactions fix the relativeorientation of the two domains (25). Examination of the sequences of domains2, 3, 4, and 5 shows that equivalent calcium-binding residues are conserved (25).Conservation is particularly striking in the regions linking the domains. It islikely, therefore, that calcium occupies similar sites at the interfaces betweendomains 2 and 3, 3 and 4, and 4 and 5. A proposal that these regions bindcalcium was made in 1987 (28). The interactions provided by Ca2+ confer arigid conformation on the five domains. Removal of calcium eliminates theseinteractions, leaving the domains connected by flexible linkers.

←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−Figure 2 Structure and interactions of cadherins. (a) A model of the cadherin dimer (25). Domainsare represented by small cylinders; calcium ions are represented by small filled circles. Two otherdomain-1 pairs are shown making adhesion contacts. The contacts that these two would make toother cadherin dimers creates a zipper-like structure (26). (b) Theβ-sheet structure of a domain 1(24–26). Strands are shown as ribbons and labeled A to G. (c) Two orthogonal views of the structureof domains 1 and 2 in a dimer of E-cadherin (25). Three calcium ions bound at the interface betweendomains 1 and 2 are shown as filled circles. (d) Dimerization interface formed by domains 1 ofN-cadherin in an adhesion complex (26). The regions making interdomain contacts are labeled(see text). Partsb–dof this figure, and parts of Figures 3–9, were drawn using MOLSCRIPT (164).

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



832 CHOTHIA & JONES

Dimer FormationIn crystals of domains 1 and 2 of E-cadherin, dimer formation involves parallelassociation of two molecules (25; Figure 2c). The contacts at the interface aremade by the A-AB-B region of domain 1, the linker region, and the BC andFG turns of domain 2 (some 19 residues in all) (25). Each molecule buries 940A2 of surface at the interface (i.e. the total buried surface is 1880A2). Of thistotal, 45% comes from residues in domain 1, 20% from residues in the linkerregion, and 35% from residues in domain 2.

Given the structure of this interface, and the assumption that the relativeorientations of the other three domains within a native molecule is the sameas that for domains 1 and 2, a dimer of the whole molecule would yield aV-shaped structure for the extracellular portion of the protein, with associationonly occurring at the domain 1–domain 2 interface (Figure 2a). In this model,each arm would be 240A in length, close to the 220A determined by electronmicroscopy (25).

The Adhesion ComplexThe molecular contacts in the crystal of domain 1 of N-cadherin provide amodel for the interactions in the homotypic adhesion complex (26). The domain1–domain 1 association is antiparallel, in which 15 residues in each domain arein contact with, or very close to, the other domain. These residues are in the Cstrand, the CD loop, the D strand, the F strand, and the FG turn (Figure 2d). TheD strand of one domain packs against the FG turn of the other domain, and viceversa. Likewise, the C strand and CD loop residues are in contact with eachother. There are also a number of water molecules in or around the interface thatmake contacts or hydrogen bonds that bridge both domains (26). The residuecontacts at the interface bury 1300A2 of surface; if the contributions of watermolecules bound in and around the interface are considered, the buried area is3300A2 (26).

Given its structure, each cadherin dimer on one cell can make adhesioncontacts with two cadherins on a second cell (26; Figure 2a). This lateralassociation between the domain-1 regions of different molecules can be repeatedto form a zipper-like structure in which multiple cadherins on a pair of cells areinterlocked via domain 1–domain 1 interactions.

In addition to its function in homophilic cell-cell adhesion, E-cadherin is aligand for the integrinαEβ7 (28a). This integrin is expressed on the surface ofT cells found in the vicinity of mucosal epithelia. The binding site region inE-cadherin has been identified by domain deletion experiments and engineeredmutations (28b). These show that E-cadherin requires domains 1 and 2, andGlu-31 (a residue in the BC loop of domain 1), to bind to the integrin. Glu orAsp residues in exposed or extended loops are common features of sites thatbind integrins (see below; Figures 6 and 8).

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




Calcium Regulation and the Specificity of AdhesionThe recognition sites of close-packed protein-protein complexes usually bury anarea of 1650± 350A2 at the interface, in those cases that do not involve a largeconformational change (29). When complex formation does involve a largeconformational change, or a transition from disordered to ordered structures,much larger contact areas are required (29, 30). The buried areas at the interfacesof the cadherin dimer (1880A2) and adhesion complexes (1300A2) are typicalof simple protein-protein complexes.

As discussed above, the dimer interface is formed by the linker regionsand adjacent parts of domains 1 and 2, and calcium ions are essential to thestructure of this region. The Ca2+ brings together carboxylate side chains thatwould otherwise repel each other, and thus give the region the ordered structurerequired for formation of the dimer interface. Formation of the dimer is, inturn, a prerequisite for formation of intermolecular adhesion contacts (31).

As mentioned above, each cadherin subtype has a unique homophilic bindingspecificity that confers on the cell in which it is expressed a specific adhesionpropensity. Alignment of the domain-1 sequences of different cadherin sub-types shows that adhesion specificity is almost certainly produced by differencesin the residues that form the proposed adhesion interface (26; Figure 2d). Of the15 residues involved in adhesion contacts in N-cadherin, some are conserved,but most are very different, at the equivalent sites in other cadherins. Indeed,for these 15 residues, in N- and R (retinal)-cadherins, which are unusual be-cause they display heterophilic interactions with each other, there are only threeconservative substitutions; in N- and E-cadherins, which do not interact, thereare two conservative and four nonconservative differences (26).

NEURAL CELL ADHESION MOLECULES

NCAMs are cell-surface proteins that, through homophilic interactions withmolecules on other cells and heterophilic interactions with molecules on thesame cell, play a central role in regulation, organization, and maintenance ofthe associations made by embryonic and adult neurons and their target tissues(32). NCAM sequences from vertebrates, ranging from humans to frogs, shareresidue identities of 70% to 98%. More distantly related are the fasciclins II ofgrasshopper and fruit fly (Drosophila melanogaster) and theAplysia californicacell adhesion molecule (ApCAM), which have only 25% identity to vertebrateNCAMs (33–35).

Examination of the gene structure, cDNA sequence, and amino acid sequenceof these proteins shows that different isoforms arise from differential exonusage (36, 37). The simplest structure, the 120-kDa form, consists of five IgSFdomains (Ig 1–Ig 5) followed by two fibronectin type III (FnIII) domains anda 30-residue C-terminal segment. This isoform is anchored to the cell surface

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



834 CHOTHIA & JONES

through linkage to glycosyl phosphatidylinositol. The other major isoformshave a transmembrane helix (the 140-kDa form) or a transmembrane helix anda cytoplasmic domain (the 180-kDa form).

There are also two isoforms that involve changes within the Ig and FnIIIdomains. One contains a 10-residue segment (VASE orπ peptide) (38) insertedin the region that links the C′ and D strands of Ig 4. This insertion changesthe function of NCAMs from molecules that promote morphological plasticityto molecules that maintain stable cell-cell contacts (39). The other isoform ismuscle specific, in which a 31- or 36-residue segment is inserted in the regionlinking the two FnIII domains (37, 40); this isoform may be involved in earlymyofibrillogenesis (41, 42).

Structure of the NCAM DomainsThe structure of the Ig-1 of human NCAM has been determined (43). It has anIgSF I-set structure with oneβ-sheet containing four strands (A, B, E, and D)and the otherβ-sheet having five strands (A′, G, F, C, and C′) (Figure 3). Theoverall conformation of this domain is that found in other known I-set structures(43). Alignment of the sequences of the Ig-2–Ig-5 domains of NCAM with thatof the Ig-1 domain shows that all have the same (or similar) residues at mostof the sites that are key for determining the observed conformation of the Ig-1domain. These close matches at key positions imply that the structures of theIg-2–Ig-5 domains are likely to be very similar to that of the Ig-1 domain overat least 80% of their tertairy fold (NK Thomsen, Y Harpaz, FM Poulsen &C Chothia, unpublished observations).

No direct structural information is available for the two FnIII domains inNCAMs. Structures have been determined, however, for two FnIII domainsfound in neuroglian, aD. melanogasterhomolog of the L1 cell adhesionmolecule (45; see below). Alignment of the NCAM FnIII sequences withthose of the neuroglian domains shows sufficient identity and similarity at thepositions critical for FnIII structure to imply that the NCAM FnIII domainswill most likely have largely the same conformation as those in neuroglian (NKThomsen, Y Harpaz, FM Poulsen & C Chothia, unpublished observations).

General Structure of NCAMThe main features of the arrangement of the domains in NCAM have beenobtained from electron microscopy of rotary-shadowed preparations (46). Sev-eral NCAM isoforms were examined. The isoform that consists of the five Igdomains, the two FnIII domains, and the 30-residue C-terminal peptide givesimages with the appearance of a bent rod approximately 45A thick, with onearm of approximately 180A, and the other of approximately 100A. The longerarm was identified as the N-terminal part of the molecule. The conclusiondrawn from these data is that the Ig domains form a rod structure (180A long)

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




Figure 3 The neural cell adhesion molecule. (a) The domain structure (32) and a model for theadhesion complex (47). The extracellular portion of the molecule is formed by five immunoglobulinI-set domains (I1–I5) and two fibronectin type III domains. (b) The structure of domain 1 of NCAM(43). Theβ-sheet strands, A–G, are represented by thick ribbons. (c) A schematic drawing of thestructure of domain 1. Black lines show the nearβ-sheet (strands A′, G, F, C, and C′); grey linesshow the farβ-sheet (strands A, B, E, and D). Horizontal thin lines between residues represent themain chain hydrogen bonds that form theβ-sheets. Residues at key sites are identified with theone-letter code. The other NCAM domains conserve many of these key residues and are predictedto have conformations very similar to domain 1 (see text).

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



836 CHOTHIA & JONES

linked to another rod structure (100A long) formed from the FnIII domainsand the C-terminal peptide. The angle between the two arms varied from 50◦

to 140◦, suggesting that the two arms are linked by a flexible hinge (46).

The NCAM Adhesion ComplexThe general structure of the homophilic complex of NCAMs has been deducedfrom measurements of the ability of single Ig domains to bind to each otherand to the whole molecule (47). In solution, the highest affinity found is thatbetween two Ig-3 domains, followed by the affinity of Ig-2 for Ig-4 and thenthat of Ig-1 for Ig-5. Other domain pairings show little or no affinity. Theseresults suggest that homophilic recognition by NCAMs involves antiparallelassociation of the five Ig domains of one molecule with the five Ig domains inanother NCAM (Figure 3a).

L1 AND NEUROGLIAN CELL ADHESION MOLECULES

The L1 cell adhesion molecule (L1) is a cell-surface glycoprotein that has animportant role in neurogenesis, nerve growth, and fasciculation. The activitiesof this molecule are mediated by its homophilic interactions with other L1molecules on other cells, as well as by its heterophilic interactions with axonin1, a protein that has a sequence similar to that of L1 (48, 49). In humans,mutations in the L1 gene cause X-linked hydrocephalus and other clinicallyrelated genetic diseases (50).

The extracellular portion of L1 is composed of six IgSF domains followed byfive FnIII domains (Figure 4a). These domains are linked by a transmembranehelix to a small intracellular domain (51). L1 sequences from a number of verte-brates share sequence identities of 50–85% with that of human L1 (52, 53). The

−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−→Figure 4 The L1 and neuroglian cell adhesion molecules. (a) The domain structure of the ex-tracellular portion of both of these homologous proteins is formed by six IgSF I-set domains andfive FnIII domains (51). (b) The outline structure of the fourth IgSF domain of human L1. Thisdrawing is an adaptation of one in (55) that contains similar descriptions of the structure of otherL1 domains. Black lines show the nearβ-sheet; grey lines show the farβ-sheet. Horizontal thinlines between residues represent the main chain hydrogen bonds that form theβ-sheets. Residuesat key sites are identified with the one-letter code and can be compared with those of the firstNCAM domain shown in Figure 3. (c) The structure of the first two FnIII domains of neuroglian(45). Theβ-sheet strands are represented by thick ribbons. The metal ion in the interface betweenthe domains is shown as a filled circle. The central part of the G “strand” shown at the top of thefigure forms three turns of polyproline helix (45). (d) The outline structure of the first FnIII domainof neuroglian using the same conventions as in (b). Residues at key sites are identified with theone-letter code (55).

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



838 CHOTHIA & JONES

D. melanogasterprotein, neuroglian, whose function is related to that of L1,has the same domain structure and possesses 25% identity to human L1 (54).

Structure of Immunoglobulin Superfamily Domains in L1A model for the tertiary structure of the extracellular domains of human L1was derived on the basis of their homology to proteins of known structure (55).The sequences of the L1 Ig domains were aligned with that of telokin, an IgSFI-set domain of known structure. At many of the positions that determine theconformation of telokin, the equivalent sites in the L1 domains have the sameresidues (or suitable alternatives). The alignments suggest that≥75% of thestructure of each L1 Ig domain will have the same conformation as telokin: twoβ-sheets, one with strands A, B, E, and D, and one with strands A′, G, F, C, andC′ (Figure 4b). Conservation of other key residues implies that certain loops(A′B, BC, EF, and FG) in all or some of the L1 domains will have conformationssimilar to those found in telokin (55).

Structure of the Fibronectin Type III Domains in L1The crystal structure of a neuroglian fragment containing the first two FnIIIdomains has been determined (45). Each domain has a basic structure commonto other FnIII domains: seven strands that form twoβ-sheets—one with threestrands (ABE), and one with four (GFCC′) (Figure 4d). On a more detailedlevel, the neuroglian FnIII domains have some unusual features. In the firstdomain, the G strand is interrupted by nine residues that form three turns ofpolyproline II helix (Figure 4c); the second domain has one such turn in thisregion. The C′-strand in the first domain also has three adjacentβ-bulges(Figure 4d).

The interface between the two domains is formed by a four-residue segmentthat links the two domains, residues in the EF and CC′ turns of domain 1,and the BC turn of domain 2 (45). The interface between the domains buries770A2 of protein surface. A metal ion also is present in the interface betweenthe domains (Figure 4c). The ion has five ligands: Oγ of Ser-679 and themain-chain O of Pro-680 from the EF loop of domain 1, Oδ of Asn-743 andthe main-chain O of Ile-740 from the BC loop in domain 2, and the O of awater molecule. Bond lengths and the protein crystallographic data indicatethat the metal is sodium. The residues whose side chains ligate to Na+ in thisstructure are not conserved in the other FnIII domains of neuroglian, suggestingthat the metal binding site is unique to the interface between domains 1 and 2. Inthe section on Fibronectin, below, we compare the geometry of the connectionbetween the two neuroglian FnIII domains to that found in other FnIII proteins.

Alignment of the amino acid sequences of the FnIII domains in human L1with the two neuroglian FnIII sequences indicates that all or part of strands A,

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




B, C, E, F, and G (and several of the EF and FG loops) in all of the L1 FnIIIdomains will most likely have conformations similar to those in neuroglian(55). In domains 1 and 2 of L1, the residues that constitute the metal bindingsite in neuroglian are conserved.

Mutations in L1 that Cause Neurological DiseasesMutations in the human L1 gene (located on the X chromosome) give riseto X-linked hydrocephalus and related disorders (50). Among the mutationscharacterized to date are frameshifts, stop codons, base changes that preventproper mRNA processing, and missense mutations that cause single amino acidsubstitutions. Of the 26 missense mutations identified (56), 22 occur in the Igand FnIII domains of L1. Seventeen of the mutations are spread throughoutthe six Ig domains, and five mutations are found in various FnIII domains. Thelikely effect of these 22 mutations can be assessed using the outline structure ofL1 (55). Thirteen substitution mutations occur at positions that are predictedto disrupt domain stability because these sites are highly conserved in L1 ho-mologs ranging fromD. melanogasterto humans, and because the mutationsproduce drastic changes in the size or chemical character of the correspondingresidue.

The other nine mutations alter surface-exposed residues and produce othereffects. For example, two mutations introduce a surface Cys residue that mayaffect L1 function through the inappropriate formation of intra- or intermolec-ular disulphides. One mutation introduces a Pro, which presumably disruptsβ-sheet structure. The other six mutations alter the surface properties (charge,hydrophobicity) of the domains.

Nine mutations are clustered in parts of the structure closest to the N terminusof a domain, and seven of these may affect interactions between domains. Fourmutations, three in the Ig-2 domain and one in the Ig-4 domain, are in the C′Dregion.

The effects on homophilic binding activity of two mutations (R184Q andH210Q) in the Ig-2 domain of L1 have been measured (57). The R184Q mu-tation, which occurs at a residue expected to be important for domain stability,eliminates homotypic binding, whereas the H210Q mutation, which is predictedto affect a surface-exposed residue, has a modest effect on binding.

PROTEIN ZERO

Protein zero (P0) is the major protein component (up to 80%) of the cell surfaceof the myelin sheaths of peripheral nerve cells in vertebrates (58). In sharks,P0 is found in the myelin sheaths of both the peripheral and central nervoussystems (59). The P0 molecule contains an extracellular region consisting of

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



840 CHOTHIA & JONES

a single IgSF domain, a single membrane–spanning helix, and an intracellulardomain (60, 61).

P0 is a homophilic cell adhesion molecule. Interactions among the extra-cellular domains of P0 molecules in adjacent membranes are believed to beresponsible, in part, for holding together individual wraps of the myelin sheath(62); likewise, interactions among the intracellular domains of P0 moleculesare thought to mediate the apposition of the cytoplasmic faces of the myelinmembrane (63).

Structure of the Extracellular Domain of Protein ZeroThe crystal structure of the extracellular region of rat P0 has been determined(64). It consists of a single IgSF V-set domain with the standard twoβ-sheetsformed from strands A, B, E, and D, and strands A′, G, F, C, C′, and C′′,respectively (Figure 5b). The structure is very similar to the V domain ofa typical antibody. Of the 119 residues in the P0 extracellular domain, 90have the same conformation as the equivalent residues in an immunoglobulinV domain. All the differences are in loop regions, and the only major oneis an extended BC loop in P0 that protrudes beyond the body of the domaininto solution and has a highly exposed Trp-28 at its tip (64). The extracellulardomain is tethered to the transmembrane helix by a five-residue connector.

Oligomerization and Adhesive InteractionsThe extracellular domain of rat P0 crystallizes under near physiological condi-tions. This fact, along with data from analytical ultracentrifugation and electronmicroscopy, strongly supports the proposal that the two types of molecular con-tacts observed in crystals of P0 correspond to two modes of interaction involvedin the cell adhesion function of P0 (64).

In the first type of contact, four molecules are clustered in parallel about acrystallographic fourfold axis to form a hollow ring (Figure 5c). Within these

−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−→Figure 5 Protein zero (P0), a myelin adhesion molecule (64). (a) A model for the P0 adhesioncomplex. On cell surfaces, the extracellular domains of P0, represented here as globular units,associate to form tetramers. Tetramers on different cell surfaces then associate in an antiparallelmanner to form the adhesion complex. Subunits in the tetramers are linked to their cell surfaceand transmembrane helix by a five-residue connector and insert a Trp side chain (open ellipse) intothe membrane of the adjacent cell surface. Shown here is part of one row of adhesion molecules;rows with the same structure are also formed normal to the plane of the drawing. (b) The IgSFV-set structure of the extracellular domain of P0. β-sheet strands are represented by thick ribbons.(c) The P0 tetramer as viewed from an opposing cell. The bars indicate the surfaces used by eachmonomer to make adhesion contacts. (d) The adhesion dimer. The two domains are from tetramerson different cell surfaces viewed in the same direction as that shown in (a) and orthogonal to that in(c). Trp-28, which is believed to penetrate the membrane on the opposing cell (see text), is depictedwith a ball-and-stick representation in (c) and (d).

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



842 CHOTHIA & JONES

tetramers, the BC loop of one molecule interacts with the C′′D and EF loops ofthe next. These contacts bury some 600A2 of solvent-accessible area at eachinterface. If envisioned as a disk, the tetramer has, on one side, the C terminiof each monomer, which are all at the same level and pointing in the samedirection, and on the other side, the protruding BC loops with their exposedTrp-28. Given that the C terminus of each extracellular domain is attachedto its cognate transmembrane anchor, the observed spatial arrangement wouldallow the four molecules of a tetramer to be tethered to the same cell surface(Figure 5a). Given their abundance, similar tetramers of P0 are likely to coatthe surface of the myelin membranes (64).

The second form of interaction seen in P0 crystals provides a model forhomophilic cell adhesion. Individual monomers tethered to one cell membraneassociate in an antiparallel manner with monomers tethered to an opposing cellmembrane (Figure 5a andd). This means that a tetramer on one membranemakes contacts with four tetramers on the opposing cell membrane and viceversa (Figure 5c). As in CD2 (see below), the interaction is symmetric, mediatedby the surface of the GFCC′C′′ sheet in both molecules. However, the geometryof the interaction in the P0 contacts differs from that of CD2 in that the relativedirections of the opposed strands are now antiparallel rather than orthogonal.accessible surface area buried at this interface is some 980A2 with a majorcontribution being made by an antiparallel pairing of the C′-strands. In addition,the Trp-28 residue in the extended BC loop in each monomer is positioned topenetrate the membrane of the opposing cell (Figure 5a andd).

The areas buried in the P0 interaction (600 and 980A2) are smaller thanthose found in high-affinity complexes between soluble proteins (29). How-ever, attachment of P0 to the membrane, and the very high concentration of P0

on the intermembrane interface, most likely compensates for weaker interac-tions between individual monomers (see also the discussion in the Conclusionsection).

The dimensions of the stack of two antiparallel P0 tetramers in the proposedadhesive interaction are similar to the intermembrane spacing (46A) observedin myelin sheaths (65, 66). The 46-A gap is equivalent to the distance betweenthe prominently exposed Trp-28 side chains on opposed BC loops (Figure 5a);it is argued, therefore, that insertion of the Trp-28 side chain that protrudes froma P0 monomer on one membrane into the surface of the opposite membrane(64) determines the intermembrane distance in the myelin sheath.

Mutations in Protein Zero that Cause Neurological DiseasesTwo human genetic diseases, Dejerine Sottas syndrome and Charcot-Marie-Tooth disease type 1B, have been linked to mutations in the P0 gene (67, 68).Currently, 16 mutations in the extracellular domain are known. Mapping these

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




mutations on the structure suggests that they disrupt folding of the extracellulardomain (e.g. deletions in theβ-strands or large changes in buried residues) ordisrupt contacts at the oligomer and the adhesive interfaces (64). A particularlysevere disease phenotype results from a Ser-34-Cys substitution (69) that wouldresult in a surface-exposed Cys. As also observed in genetic diseases linkedto L1 (see above) and the fibroblast growth factor (FGF) receptor (70), it isproposed (64) that the Cys residue causes dysfunction through formation ofinappropriate disulphide bonds.

FIBRONECTIN

Fibronectin is deposited in a variety of extracellular matrices (ECMs). The pri-mary role of fibronectin is to attach cells to the ECM by binding to members ofthe integrin family of cell-surface proteins; however, fibronectin is a multifunc-tional protein that also binds to other constituents of the ECM, including fibrin,collagen, and heparin. Fibronectin is involved in embryonic differentiation,cell morphology, cell migration, and thrombosis (71).

Fibronectin is a long flexible molecule (72, 73). It is usually a dimer with itssubunits joined by two disulphide bridges near their C termini. The structureof the dimer contact region around the disulphides has been determined and isformed by 10 residues that coil and associate in an antiparallel manner (74).Each fibronectin chain is comprised of four types of structural elements: fi-bronectin type I (FnI), type II (FnII), and type III (FnIII) domains, and the Vdomain. In the major fibronectin species, these domains are arrayed so thatsix type I domains are followed by two type II domains, three type I domains,fourteen type III domains, the V domain, one type III domain, and three type Idomains (71; Figure 6). Alternative gene splicing can give rise to fibronectinmolecules that contain additional type III domains (one after the seventh andone after the eleventh Fn III) or in which the size of the V domain is modifiedor removed altogether (75).

Structure of the Fibronectin Type I and Type II DomainsThe 12 FnI domains (each with∼40 residues) share sequence identities ofabout 35%. Structures have been determined for the first and seventh FnI inisolation (76, 77) and for a fragment harboring both the fourth and fifth (4FnI-5FnI) (78). The structure of an FnI domain found in tissue-type plasminogenactivator has been determined in isolation and in a fragment also containing itsneighbouring epidermal-growth-factor (EGF)–like domain (79, 80). All fiveof these FnI structures have two smallβ-sheets packed face-to-face. The firstβ-sheet is formed by the first two strands (A and B), and the secondβ-sheet isformed by the next three strands (C, D, and E) (Figure 6). Conserved disulphide

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



844 CHOTHIA & JONES

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




bridges join strands A and D and also strands D and E. The residues that packto form the interior in the four fibronectin structures are highly conserved in thesequences of the other eight domains; it is expected, therefore, that the other FnIdomains have structures very similar to those that are known. In the differentFnI domains, however, the loops joining the strands show small variations insize and conformation.

In the structure of4FnI-5FnI, the secondβ-sheet of the4FnI packs againstthe firstβ-sheet of the5FnI. Thus, the fourβ-sheets in the two domains stackon top of each other to produce a rod-like structure (78; Figure 6). In otherFnI sequences, the residues equivalent to those that form the interface between4FnI and5FnI are not strongly conserved; therefore, it is uncertain whether otherpairs of FnI domains pack in the same manner (78). Although FnI domains havevery similar folds, their target recognition properties differ. Ligand binding alsousually involves more than one domain.4FnI-5FnI is the main fibrin-binding site(82).8FnI-9FnI and2FnII-7FnI are the major collagen-binding sites (83, 84).

No structure has yet been determined for either of the two FnII domains.However, the 40-residue collagen-binding b domain of the bovine seminal fluidprotein, PDC-109, is clearly homologous to FnII. The structure of the collagen-binding domain in PDC-109 has five short strands ofβ-sheet (81); strands B, A,and D form oneβ-sheet that packs at nearly right angles to the second, whichis formed by strands C and E (Figure 6). The cavity formed between the twoβ-sheets is thought to be involved in binding collagen (81).

Structure of Fibronectin Type III DomainsStructures have been determined for the10FnIII in isolation (85, 86) and fora segment containing7FnIII-8FnIII-9FnIII-10FnIII (87). All of these domainshave a similar structure: twoβ-sheets packed face-to-face—one sheet formedby strands A, B, and E, and the other by G, F, C, and C′ (Figure 4d). In thefour domains, theβ-sheets are very similar; the largest difference is whetherthe A strand contains a secondβ-bulge and where that bulge is positioned. Inan alignment of all the FnIII sequences in fibronectin (86), it was suggestedthat (a) each would have aβ-sheet structure and, in most cases, conformationsfor the C′E and EF turns very similar to those seen in10FnIII; and (b) the other

←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−Figure 6 Structure of fibronectin. The middle of the figure shows schematically the domainstructure of fibronectin. The FnI domains are represented by rectangles with bars; FnII domainsare represented by shaded circles; and FnIII domains are represented by open squares. The top leftof the figure shows the structure of the fourth and fifth type I domains (78); the bottom left showsa type II domain from another protein (see text). The right side of the figure shows the structure ofthe segment of fibronectin formed by the seventh to tenth FnIII domains (87). The integrin bindingsite involves the RGD tripeptide in10FnIII, and the synergy region in9FnIII (see text).

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



846 CHOTHIA & JONES

Table 1 Fibronectin type III domains: interdomain geometriesa

Tilt Rotation Buried surfaceProtein Domains (◦) (◦) area (A2)

Fibronectin 7FnIII-8FnIII 52 112 590Fibronectin 8FnIII-9FnIII 48 160 530Fibronectin 9FnIII-10FnIII 12 −43 330Neuroglian 1FnIII-2FnIII 62 175 770

aData are from (87).

turns would vary in size and conformation. These predictions were confirmedfor 7FnIII–9FnIII by their subsequent structure determination (87).

Connections Between FnIII Domains in Fibronectinand Other ProteinsIn Table 1, the interdomain geometries of pairs of FnIII domains are describedin terms of (a) a tilt angle (the angle between the long axes of adjacent domains)and (b) a rotation angle (the rotation required to superimpose adjacent domains)(87). Theβ-sheet frameworks of the7FnIII, 8FnIII, 9FnIII, and10FnIII domainsare connected by five-residue linkers. The connection between the two FnIIIdomains in neuroglians is four residues. Although the lengths of the connectionsbetween FnIII domains in fibronectin are conserved, the interdomain geometriesvary (Figure 6; Table 1). The geometries of three of the four connections areroughly similar in terms of tilt angle (48◦–62◦) and rotation angle (112◦–175◦)(Table 1). For9FnIII-10FnIII, however, the tilt angle and the rotation angle aremuch smaller (Table 1).

The interface between pairs of FnIII domains involves the linker peptide,residues in the AB and EF loops on one FnIII, and residues in the BC or FGloop on the other FnIII. Variations in the volume or chemical character of theresidues in these interfaces explain the observed differences in the interdomaingeometries (87).

The Integrin- and Heparin-Binding SitesThe function of fibronectin as a cell adhesion molecule involves its recognitionby integrins. The sites for fibronectin interaction with integrins reside in9FnIIIand10FnIII. The site in10FnIII involves an Arg-Gly-Asp (RGD) sequence in theFG turn (88). In10FnIII, the FG turn is three to five residues longer than in anyother FnIII domains, and the RGD tripeptide is located at the end of the loop,some 10A away from the body of the domain (85–87; Figure 6). A second siteof interaction with integrins, the “synergy” site, has been identified on9FnIIIin the C′E region (89, 90; Figure 6). The different elements in this region are

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




important for discrimination by different integrins. Residues Asp-1373 andArg-1374 are essential for interaction with integrinαIIbβ3 but not with integrinα5β1 (90), whereas Arg-1379 is important for interactions with integrinα5β1

but not with integrinαIIbβ3 (89).The site for heparin-binding on13FnIII has been examined by site-directed

mutagenesis and involves six Arg and one Lys situated mainly on the A and Bstrands of oneβ-sheet (91, 92).

TENASCIN-C

Tenascin-C is one of a family of three tenascins, which are homologous hexam-eric proteins found in the ECM (93). Each tenascin-C polypeptide contains anN-terminal domain, 13.5 EGF-like domains, 8–15 FnIII domains (dependingon the splice variant), and a C-terminal domain (94). The six subunits are joinedtogether at their N-terminal ends to give a hexabrachion structure resembling thespokes of a wheel projecting from a central hub (95). The biological functionof tenascin is uncertain, but it does have integrin-binding sites, one of which issituated on the third FnIII domain that contains an RGD sequence. The struc-ture of this FnIII domain is very similar to the FnIII domains in fibronectin (87,96). The RGD tripeptide in the FG turn of the third FnIII domain in tenascin-Cis like that in10FnIII of fibronectin (Figure 5). In tenascin-C, however, the FGloop is four residues shorter, which brings the RGD tripeptide 7A closer tothe body of the domain. This RGD site in tenascin-C can bind to integrinα5β3

but not to integrinα5β1 (97). A second site for cell binding is found in theC-terminal FnIII domain of tenascin-C and is probably recognized by integrinα2β1 (97, 98).

INTEGRINS

Integrins constitute a large, widespread family of cell adhesion molecules im-plicated in biological functions such as embryogenesis, the immune response,haemostasis, inflammation, and maintenance of tissue integrity. Integrinsbind not only to components of the ECM, including fibronectin (see above),collagens, and laminin, but also to other cell adhesion molecules, includingVCAM-1 and the intercellular adhesion molecules (ICAMs) (see below) (99).Some integrins also are exploited as cell surface receptors for certain viruses(100).

Members of the integrin superfamily are heterodimeric cell surface glyco-proteins comprised of subunitsα andβ that each span the membrane once. Todate, 16α-subunits and 8β-subunits have been identified. A givenβ-subunitcan combine with one or several differentα-subunits; 21 differentαβ pairings

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



848 CHOTHIA & JONES

are known. The extracellular part of theα-subunits has up to 1114 residues andthat of theβ-subunits has up to 678 residues. In each subunit the N-terminalportions form a globular ligand-binding domain that is connected to the mem-brane by a stalk. A subset ofα chains contain an additional domain of about190 amino acids, variously termed either the inserted (I) domain or, by analogywith homologous domains in von Willebrand Factor, the A domain (101, 102).Mg2+ and Ca2+ ions are required for ligand binding.

Integrins are competent to signal in response to both adhesive interactionsoutside the cell (outside-in) and events inside the cell (inside-out) (99, 103).Outside-in signaling triggers processes within the cell mediated by the cytosolicdomains of theα- andβ-subunits. Inside-out signaling is manifest through whatis believed to be a conformational switch in the extracellular region betweenactive (adhesive) and inactive states.

Structure of the N-Terminal Domain of the Integrinα-Subunitsα-subunits have a ligand-binding N-terminal domain that consists of residues1 to∼440, except in those subunits that have I domains, in which it consists ofresidues 1 to∼130 and∼320 to∼600. In allα-subunits, the sequences in thisregion have a sevenfold repeat, with each repeat consisting of∼60 residues. Astructure has been predicted for this domain on the basis of (a) the conservationof the sevenfold repeat in evolution, (b) sequence patterns within the repeats, (c)symmetry arguments, (d) secondary structure predictions, (e) model building,and (f ) topology of disulphide bridges formed by cysteines (103a). Each repeatis predicted to form a four-strandedβ-sheet, and the sevenβ-sheets, from theseven repeats, are predicted to have a pseudosymmetric radial arrangement thatgives aβ-propeller fold (103a; Figure 7).

The Metal and Ligand Binding Site inα-SubunitsAll integrins have a Ca2+ binding motif in repeats 5, 6, and 7, and some alsohave the motif in repeat 4 (103b). In the predicted structure, these motifsare in the link between the first and secondβ-sheet strands in the respectiverepeats. These positions put the Ca2+ ions close to one another at the bottomof the domain, as shown in Figure 7. The most likely site for Mg2+ ions is at

−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−→Figure 7 α-integrin. Top: the observed structure of the I domain (104–107).Bottom:a schematicdrawing of theβ-propeller structure predicted for the ligand-binding N-terminal domain (103a).Thick ribbons represent strands ofβ-sheet, and coiled ribbons representα-helices. In the predictedstructure of the N-terminal domain, there are long interstrand loops at the top of the structure(the region closest to the viewer); for reasons of clarity, they are shown here as being short. Theregions known to be involved in ligand recognition (103c, 103d; see text) are indicated by∗. Theconnections between the I domain and the N-terminal domain are shorter than shown here, and itis expected that the I domain actually sits on the top of the N-terminal domain (103a).

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




N

76

5

4

3

2

1

C

❋

❋

Ligand binding�site

I DOMAIN

N-TERMINAL�DOMAIN

Mg2+

Ca2+

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



850 CHOTHIA & JONES

the narrowest aperture of the channel that goes through the center of the structure(103a). This places it at the top of the domain (Figure 7) where it could bind theGlu or Asp residues that are a prominent feature in the binding sites of integrinligands (see sections above and below).

Mutations in the regions 153–165 and 187–190 abolish or reduce ligandbinding (103c, 103d). In the predicted structure, the two regions are in in-terstrand loops; close to each other at the top of the domain, and at one sideof the Mg2+ site (103a; Figure 7). In the integrins that contain an I domain,it is inserted in the link between repeats 2 and 3. The position of this linkputs the I domain close to the predicted binding site of the N-terminal domain(Figure 7).

Structure of the I DomainsHigh-resolution crystal structures have been determined for the I domains oftwo integrins [from theαM subunit ofαMβ2 (104, 105) and theαL subunit ofαLβ2 (106, 107)]. The I domains ofαM andαL share 33% sequence identity,and their structures are very similar. Each has anα/β fold consisting of sevenα-helices packed against a central six-strandedβ-sheet (Figure 7). The strandsin theβ-sheet have the order CBADEF. This topology is the same as that ofthe widespread dinucleotide-binding (or Rossmann) fold first observed in theintracellular enzyme, lactate dehydrogenase (108).

The Metal-Binding Site in the I DomainsA divalent cation–binding site is situated on the surface of the I domain at oneend of theβ-sheet, in the position classically occupied by a dinucleotide cofac-tor in other proteins that contain a domain with this fold. In I domain crystalstructures (105–107), the divalent metal ion is directly coordinated at the cationbinding site by residues Ser-139, Ser-141, and Asp-239; Thr-206 participates in-directly through a hydrogen bond to a coordinating water molecule (the residuenumbering here is forαL; identical residues, but with different numbering, areused inαM). A total of six ligands coordinate the metal ion with an octahedralgeometry; the remaining ligands are water molecules or a chloride ion (106,107). The residues involved are highly conserved in the sequences of all known Idomains, and this arrangement has been termed the metal ion–dependent adhe-sion site (MIDAS) motif (105). The key features of this motif are the sequenceDXSXS (residues 137–141) and residues Thr-206 and Asp-239.

Crystal structures have been determined for the Mg2+- (104) and Mn2+-bound (105) forms of theαM I domain. Crystal structures also have beendetermined for the metal-free, Mg2+-, and two Mn2+-bound (106, 107) formsof theαL I domain. For both I domains, comparisons of the various structuresreveal significant shifts in the position ofα7, the C-terminal helix of the domain(105, 107). The data forαL indicate that this conformational variation may be

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




triggered by differences in molecular contacts within the crystal lattices, ratherthan by internal changes in the protein due to metal binding state (107).

The I-Domain Adhesion SiteThe I domains ofαL andαM contribute to the interaction of these integrins withIgSF-type cell adhesion molecules (109, 110). In addition to residues in theMIDAS motif, several solvent-exposed residues in theαL I domain (Met-140,Glu-146, Thr-243, and Ser-245) have been implicated by mutagenesis studieson integrin association with ICAM-1 (111–113). These residues map to thesurface surrounding the divalent cation–binding site. Recent structural resultsindicate that presence of a divalent cation is not requisite for a stable I-domainfold (107); thus, the functional importance of the metal ion may be to interactdirectly with residues on the ICAM-1 molecule. Indeed, in many integrin-binding molecules, an acidic residue (Asp or, less frequently, Glu) has beenidentified as a key functional element (99, 114), and in the Mg2+-bound crystalstructure of theαM I domain, coordination of the metal ion is completed by aGlu side chain donated by a neighboring molecule in the crystal lattice (104).

VASCULAR CELL ADHESION MOLECULE 1

VCAM-1 (115) is one of a subgroup of IgSF-type cell adhesion moleculeswhose members include the intercellular adhesion molecules, ICAM-1 (116,117), ICAM-2 (118), and ICAM-3 (119, 120), and the mucosal vascular ad-dressin cellular adhesion molecules (MAdCAMs) (121). VCAMs, ICAMs, andMAdCAMs bind members of the integrin superfamily and play a central role inleukocyte trafficking and in extravasation (122). In addition, the extracellularregions of several of these molecules act as cellular receptors for various virusesand parasites (123–126).

Each of these cell adhesion molecules consists of an N-terminal extracellularregion built from IgSF domains, a transmembrane helix, and a cytoplasmicdomain. The number and exact nature of the domains making up the extra-cellular regions of these molecules vary, but they all contain a two-domainintegrin-binding fragment that is homologous to the two N-terminal domainsof VCAM-1.

Structure of the Two N-Terminal Domains of VCAM-1The extracellular region of VCAM-1 consists of seven (or less commonly for asplice variant, six) IgSF domains. Crystal structures for a functional fragmentcomprising the two N-terminal domains have been determined independentlyby two groups (127, 128; Figure 8). Domain 1 has an IgSF I-set structure withtwoβ-sheets—strands A, B, E, and D, and strands A′, G, F, and C, respectively.It also has two intersheet disulphides: One is the canonical disulphide between

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



852 CHOTHIA & JONES

1

2

Integrin�binding site

Asp 40

Figure 8 Structure of the two N-terminal domains of VCAM-1 (127, 128). The region involvedin binding integrin is indicated. Theβ-sheet strands are shown as ribbons.

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




strands B and F; the other is a distinctive feature of this subgroup and links theBC and FG loops. The structure is similar to the typical immunoglobulin I-setdomain except for the CD loop, which extends from the body of the domainand displays a solvent-accessible hydrophobic residue (Ile-39) and a highlyexposed acidic residue (Asp-40) (see below).

To date, domain 2 of VCAM-1 is the largest IgSF C2-set domain for which astructure has been determined. The core is structurally the same as the C2-setdomains in CD2 and CD4, with twoβ-sheets [strands A, B, and E, and strandsG, F, C, and C′, respectively (Figure 1)], but the interstrand loops, most notablythe C′E and FG loops, are much longer. These two loops appear flexible; the FGloop shifts to accomodate small differences in the relative positions of domains1 and 2.

The interdomain connection between VCAM-1 domains 1 and 2 is struc-turally homologous to that of CD2 (see below). As in CD2, this portion of thepolypeptide chain is closely associated with domain 2 through hydrogen bondsto the side chains of residues in the FG loop of this domain. Again, as observedin CD2, this form of linker permits some flexibility in the relative orientationof the domains.

Integrin-Binding Sites in VCAM-1 and ICAMsVCAM-1 binds to the integrinα4β1. The residues implicated in this adhesivefunction by mutagenesis (129–133) map to the GFC-strand face and the CDloop of the N-terminal domain. Residues that play a role in binding of integrininclude Ile-39 and Asp-40, which are prominently displayed in the CD loop(Figure 8), and nearby Leu-70 on strand F. Acidic residues, in particular Asp,have long been implicated in adhesive interactions to integrins (88, 114).

The ICAMs bind to integrinsαLβ2 andαMβ2. In ICAM-3, residues impli-cated in adhesion include Glu-37, Leu-66, Ser-68, and Gln-75 (134). Alignmentof ICAM-3 with VCAM-1 (C Chothia, unpublished observations) shows thatLeu-66 is in a position equivalent to that of Leu-70 in VCAM-1. Glu-37, how-ever, is not in the CD loop (unlike Asp-40 in VCAM-1), but in the C strand.Examination of a recent structure for ICAM-2 leads to the same conclusion(J Wang, personal communication). This difference suggests that theαLβ2 andαMβ2 integrins bind via an acidic residue in a flat region (rather than one in aprotruding loop, as is common with other integrins).

CD2

CD2, a protein found on the surface of cytotoxic and helper T-cells (135, 136),enhances the cell-cell interaction mediated by binding of theαβ receptor ofthe T-cell to the major histocompatibility complex class I and II molecules on

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



854 CHOTHIA & JONES

antigen-presenting cells. Human CD2 binds to CD58 (136), and rat CD2 bindsto CD48 (137, 138). CD2 is homologous to both CD58 and CD48; thus, theinteractions are pseudohomophilic.

CD2 is a representative of a large group of homologous cell surface receptors,which includes CD58, CD48, SLAM, 2B4, and LY-9 (139). Except for LY-9,all have an N-terminal extracellular region formed by two IgSF-type domainsthat are tethered to the cell surface by either a transmembrane helix and acytoplasmic domain or by a glycosylphosphatidylinositol anchor (CD48 andan isoform of CD58). LY-9 contains four Ig domains (perhaps as the result ofgene duplication). Sequence analysis also indicates that CD2 is homologousto the N-terminal two domains of members of the carcinoembryonic-antigen(CEA) subset of IgSF molecules (140).

Structure of the CD2 Extracellular DomainsThe crystal structures of rat and human CD2 were the first determined for theentire extracellular portions of a cell adhesion molecule (141, 142; Figure 9).Both structures are very similar and agree well with solution structures deter-mined by nuclear magnetic resonance (NMR) (143, 144).

The N-terminal domain of CD2 has an IgSF V-set structure with twoβ-sheets—one formed by strands A′, G, F, C, C′, and C′′, and the other by strandsB, E, and D. This domain in CD2 differs from canonical V-set structures in tworespects: absence of both an initial A strand and a disulphide between strandsB and F (141–144). The disulphide is absent in all members of the CD2 subset.Normally, in V-set domains, strand A hydrogen-bonds to the B strand beforeswitching to hydrogen-bond, as strand A′, to the G strand. Absence of an Astrand may not be a general feature of the CD2 group because sequence analysisindicates appropriately positioned residues before a Pro at the predicted startof strand A′ in, for example, CD58 and SLAM.

The second domain in CD2 has a typical IgSF C2-set structure (12) with twoβ-sheets—one formed by strands A, B, and E, and the other by strands G, F,C, and C′ (141–142). In this domain of CD2, an intersheet disulphide linksstrands B and F, as found in standard Ig domains; but a distinctive feature ofthe second domain in CD2 is a second disulphide linking the ends of strands Aand G. The arrangement of a V-set domain followed by a C2-set domain seenin CD2 is also found in CD4 (146–148).

Connections Between IgSF Domains in CD2and Other ProteinsIn CD2, the V-set domain is connected to the C2-set domain by a discrete linkerof six residues. Sequence analysis indicates that homologous connections,characterized by the residues ExVskP, are present in other members of thisIgSF subgroup. Comparison of the two available rat CD2 structures indicates

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




Figure 9 Structure of CD2 domains and a model for the adhesion complex (141, 142). The figureshows orthogonal views of two pairs of CD2 molecules. The mode of association shown is seenin crystals of the human and rat molecules. There is considerable experimental evidence that CD2and its homologs, CD58 or CD48, associate in the same manner in actual adhesion complexes. Theβ-sheet strands are shown as ribbons.

that the domains can flex about 7◦ relative to one another. A second regionof flexibility in the extracellular portion of the CD2 molecule is provided bya short stalk of some six residues that tethers the second Ig domain to thetransmembrane helix.

In Table 2, we describe the interdomain geometries of other linked IgSFdomains. The first and second domains of VCAM are linked in a mannerclosely related to that in CD2 and have a similar geometry. Links of a similartype are also predicted, on the basis of sequence homology, to be present incertain other IgSF-type cell surface receptors (70).

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



856 CHOTHIA & JONES

Table 2 Immunoglobin superfamily: interdomain geometries

Tilt a Rotationa Buried surfaceProtein Domains (◦) (◦) areab (A2)

CD4 1V-2C2 45 160 1050CD4 3V-4C2 45 150 1200CD2 1V-2C2 45 160 500VCAM-1 1I-2C2 30 130 975

aSource: (128a).bSource: (C Chothia & EY Jones, unpublished data).

CD4 has four domains: The first and third have a V-set structure; the secondand fourth have a C2-set structure. Unlike in CD2, the V domains in CD4 arelinked directly to the C2 domains. In other words, at the boundaries betweendomains 1 and 2 and between domains 3 and 4 in CD4, the last residue instrand G of the V-set domain is followed by the first residue in strand A of theC2-set domain (rather than connected by the linker found in CD2). This form ofinterdomain linkage would appear to be rigid (145–149). Although the structureof the interdomain links in CD4 is quite different from that in CD2, the V do-mains and C2 domains have similar relative interdomain geometries (Table 2).In antibodies, the connection between the V and C domains includes a molec-ular “ball and socket joint” that allows the VL-VH dimer to rotate relative to theCL-CH1 dimer and give tilt angles that range between∼135 and∼180◦ (150).

CD2 Adhesion SiteCD2 does not exhibit true homophilic adhesion at physiological conditions(151), but given the homology between CD2 and its ligands, the functionalinteractions between members of this IgSF subgroup are presumably pseu-dohomophilic. Indeed, the only intermolecular lattice contacts common tocrystals of both rat and human CD2 are “head-to-head” interactions betweenthe GFCC′C′′ β-sheets of domain 1 (142; Figure 9). This association places thetwo molecules in an orientation appropriate to their function as intercellular ad-hesion molecules. Thus, it was proposed that this binding mode provides a goodmodel for the adhesive interaction (141). This hypothesis is now supported bya broad range of data.

Mutagenesis data implicate residues on the GFCC′C′′ face of the N-terminaldomain of both rat and human CD2 in adhesion in vivo (152–155). Likewise,mutagenesis of homologous residues in CD48 (156) confirms the location of afunctional adhesive site at this same position on this molecule. Analysis of theinteractions of rat CD2 and CD48 by solution NMR showed good agreementwith the crystallographic model for the adhesive area on CD2 (157).

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




Taken together, these data indicate that the adhesive area on domain 1of CD2 extends over the length of strands C and C′, the C-terminal halfof strand F, and the N-terminal half of strand G. This surface has an areaof approximately 750A2, is flat, and is made up predominantly of chargedor polar residues. In particular, both crystallographic (141, 142) and solutionNMR (157) data implicate rat CD2 residues Glu-29, Arg-31, Glu-33, Thr-37,Leu-38, Glu-41, Lys-43, Phe-49, Thr-79, Tyr-81, and Gly-85 as central to theadhesive interaction.

The crystallographic model for adhesion has been further tested using com-plementary mutagenesis of rat CD2 and CD48 (156). The ability of chargereversal mutations in one molecule to compensate for charge reversal muta-tions in the other molecule indicates that residues Glu-41 and Lys-43 of ratCD2 interact with residues Arg-31 and Glu-44 of CD48, respectively. Theseresults are in good agreement with the model, in which the CC′ and FG loopsof opposing molecules are paired, and the angle between the strand directionsof the twoβ-sheets is approximately 90◦ (Figure 9).

CONCLUSIONS

Among proteins of known structure, cell adhesion molecules have quite novelfunctions. These cell-surface proteins use specific homophilic or heterophilicrecognition to either bind cells together or facilitate their movement on a sub-stratum. To what extent are these novel functions reflected in novel struc-tural properties? For an initial answer to this question, the cell adhesionmolecules described here can be compared with other molecules whose struc-tures are known, particularly those whose function involves protein-proteinrecognition.

The folds of the individual domains in the cell adhesion molecules describedhere are in no way novel. They either haveβ-sheets packed face-to-face orα-helices packed onto the faces of aβ-sheet. Folds of this kind are found inmany proteins with a great variety of functions [see the entries in the StructuralClassification of Proteins (SCOP) database (158)].

The arrangement of domains as linear arrays, as observed in cell adhesionmolecules and other cell surface proteins, is rare in proteins of the cytoplasm,but does occur. In certain structural proteins, the arrays can be much longerand can be built of domains related to those found in cell adhesion molecules.For example, the muscle protein titin contains 112 IgSF I-set domains and 132FnIII domains (16, 159). In the gene sequences of cell adhesion molecules,the coding regions for domains are usually separated at the domain boundariesby introns. This finding suggests that cell adhesion molecules have evolvedthrough gene duplication and interintronic recombination; it has been argued

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



858 CHOTHIA & JONES

that this process was crucial to form new large proteins in the development ofthe metazoa (160).

The recognition sites seen in the cadherins, CD2, and to a lesser extent P0

have structures very similar to those seen in stable complexes between solubleproteins. For example, the areas buried in recognition sites of protease-inhibitorand antibody-protein antigen complexes have the same range of values (29). Thevery extended binding surface that seems to be formed in the NCAM adhesiondimer is very different from typical soluble protein-protein complexes, but itdoes occur in muscle proteins.

The kinetics of adhesive interactions of leukocyte adhesion molecules aredifferent from those found for the high-affinity complexes formed by solubleproteins. Measurements of the rates of association and dissociation have beenmade for both the soluble extracellular portions of these adhesion molecules,and the intact molecules attached to cell membranes (161–163). In both cases,the dissociation rate constant (koff ) is much faster than those characteristicallyfound for high-affinity interactions between soluble proteins:∼5 s−1 ratherthan∼10−4 s−1. Adhesion molecules on the surface of leukocytes facilitateboth rapid migration over endothelial surfaces and stable adhesion to targetcells. It has been argued that the dynamic role of a cell adhesion moleculerequires rapid association and disassociation of its adhesive bonds (161).

Measurements in solution of the association rate constant (kon) for leukocyteadhesion molecules give values similar to, or a little less than, those for high-affinity interactions between soluble proteins,∼10−5 M−1s−1 (161). Therefore,for a leukocyte cell adhesion molecule, the equilibrium constant in solution (Kd)is 10−5–10−4 M. This range of affinity is much smaller than that found for thetight-binding complexes between soluble proteins, which usually have aKd inthe range 10−8–10−13 M. However, adhesion molecules on cells are tethered tothe surface and therefore restricted to movement in two dimensions (unlike sol-uble molecules in solution). Also, lateral mobility in the membrane facilitatesconcentration of cell adhesion molecules at sites of cell-cell contact (163). Inexperiments using CD2 on T-cells and purified CD58 in artificial phospholipidbilayers, the affinity constant was shown to be so high that complex forma-tion was strongly favored under physiological conditions (162, 163). Thus,cell adhesion molecules have evolved novel binding sites that possess intrinsicassociation and dissociation constants that permit both high-affinity cell-cellinteractions and rapid cell migration.

ACKNOWLEDGMENTS

We thank Drs. WA Hendrickson, FM Poulson, J Rini, L Shapiro, and NKThomas for providing information prior to publication.

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




Visit the Annual Reviews home pageathttp://www.annurev.org.

Literature Cited

1. Thiery J-P, Brackenbury R, RutishauserU, Edelman GM. 1977.J. Biol. Chem.252:6841–45

2. Takeichi M. 1977.J. Cell Biol.75:464–74

3. Edelman GM, Crossin KL. 1991.Annu.Rev. Biochem.60:155–90

4. Vaughn DE, Bjorkman PJ. 1996.Neuron16:261–73

5. Jones EY. 1996.Curr. Opin. Cell Biol.8:602–8

6. Bentley GA, Boulot G, Karjalainen K,Mariuzza RA. 1995.Science267:1984–87

7. Fields BA, Ober B, Malchiodi EL, Lebe-deva MI, Braden BC, et al. 1995.Science270:1821–24

8. Rini JM. 1995. Annu. Rev. Biophys.Biomol. Struct.24:551–77

9. Weis WI, Drickamer K. 1996.Annu. Rev.Biochem.65:441–73

10. Rini JM. 1995.Curr. Opin. Struct. Biol.5:617–21

11. Lesk AM, Chothia C. 1982.J. Mol. Biol.160:325–42

12. Williams AF, Barclay AN. 1988.Annu.Rev. Immunol.6:381–405

13. Harpaz Y, Chothia C. 1994.J. Mol. Biol.238:528–39

14. Bork P, Holm L, Sander C. 1994.J. Mol.Biol. 242:309–20

14a. Chothia C, Finkelstein AV. 1990.Annu.Rev. Biochem.59:1007–39

15. Sun F, Hammill SJ, Proctor M, FreundSMV, Benian GM, et al. 1996.J. Mol.Biol. 264:624–39

16. Pfuhl M, Pastore A. 1995.Structure3:391–401

17. Bateman A, Eddy SR, Chothia C. 1996.Protein Sci.5:1939–42

18. Gamulin V, Rinkevich B, Schacke H,Kruse M, Muller IM, et al. 1994.Biol. Chem. Hoppe-Seyler375:583–88

19. Takeichi M. 1990.Annu. Rev. Biochem.59:237–52

19a. Drubin DG, Nelson WJ. 1996.Cell84:335–44

19b. Gumbiner BM. 1996.Cell 84:345–5720. Grunwald GB. 1993.Curr. Opin. Cell

Biol. 5:797–80521. Nose A, Nagafuchi A, Takeichi M. 1988.

Cell 54:993–1001

22. Nose A, Tsuji K, Takeichi M. 1990.Cell61:147–55

23. Pokutta S, Herrenknecht K, KemlerR, Engel J. 1994.Eur. J. Biochem.223:1019–26

24. Overduin M, Harvey TS, Bagby S, TongKI, Yau P, et al. 1995.Science267:386–89

25. Nagar B, Overduin M, Ikura M, Rini JM.1996.Nature380:360–64

26. Shapiro L, Fannon AM, Kwong PD,Thompson A, Lehmann, MS, et al. 1995.Nature374:327–37

27. Shapiro L, Kwong PD, Fannon AM, Col-man DR, Hendrickson WA. 1995.Proc.Natl. Acad. Sci. USA92:6793–97

28. Ringwald M, Schuh R, Vestweber D,Eistetter H, Lottspeich F, et al. 1987.EMBO J.6:3647–53

28a. Cepek KL, Shaw SK, Parker CM, Rus-sell GJ, Morrow JS, et al. 1994.Nature372:190–93

28b. Karecla PI, Green SJ, Bowden SJ, Coad-well J, Kilshaw PJ. 1996.J. Biol. Chem.272:30909–15

29. Janin J, Chothia C. 1990.J. Biol. Chem.265:16027–30

30. Janin J, Miller S, Chothia C. 1988.J.Mol. Biol. 204:155–64

31. Tomschy A, Fauser C, Landwehr R,Engel J. 1996.EMBO J. 15:3507–14

32. Cunningham BA, Hemperly JJ, MurrayBA, Prediger EA, Brackenbury R, et al.1987.Science236:799–806

33. Harrelson AL, Goodman CS. 1988.Sci-ence242:700–8

34. Grenningloh G, Rehm EJ, Goodman CS.1991.Cell 67:45–57

35. Mayford M, Barzilai A, Keller F,Schacher S, Kandel ER. 1992.Science256:638–44

36. Owens GS, Edelman GM, CunninghamBA. 1987. Proc. Natl. Acad. Sci. USA84:294–98

37. Prediger EA, Hoffman S, Edelman GM,Cunningham BA. 1988.Proc. Natl.Acad. Sci. USA85:9616–20

38. Small SJ, Haines SL, Akeson RA. 1988.Neuron1:1007–17

39. Doherty P, Moolenaar CECK, AshtonSV, Michalides RJAM, Walsh FS. 1992.Nature356:791–93

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



860 CHOTHIA & JONES

40. Dickson G, Gower HJ, Barton CH, Pren-tice HM, Elsom VL, et al. 1987.Cell50:1119–30

41. Peck D, Walsh FS. 1993.J. Cell. Biol.123:1587–95

42. Byeon MK, Sugi Y, Markwald RR,Hoffman S. 1995.J. Cell Biol. 128:209–21

43. Thomsen NK, Soroka V. Jensen PH,Berezin V, Kiselyov VV, et al. 1996.Nat.Struct. Biol.3:581–85

44. Deleted in proof45. Huber AH, Wang YE, Bieber AJ, Bjork-

man PJ. 1994.Neuron12:717–3146. Becker JW, Erickson HP, Hoffman

S, Cunningham BA, Edelman GM.1989. Proc. Natl. Acad. Sci. USA86:1088–92

47. Ranheim TS, Edelman GM, Cunning-ham BA. 1996.Proc. Natl. Acad. Sci.USA93:4071–75

48. Rathjen FG, Schachner M. 1984.EMBOJ. 3:1–10

49. Sonderegger P, Rathjen FG. 1992.J. CellBiol. 119:1387–94

50. Kenwrick S, Jouet M, Donnai D. 1996.J. Med. Genet.33:59–65

51. Moos M, Tacke R, Scherer H, TeplowD, Fruh K, et al. 1988.Nature 334:701–3

52. Hlavin ML, Lemmon V. 1991.Genomics11:416–23

53. Reid RA, Hemperly JJ. 1992.J. Mol.Neurosci.3:127–35

54. Bieber AJ, Snow PM, Hortsch M, PatelNH, Jacobs JR, et al. 1989.Cell59:447–60

55. Bateman A, Jouet M, MacFarlane J, DuJ-S, Kenwrick S, et al. 1996.EMBO J.15:6048–57

56. Van Camp G, Fransen E, Raes G,Willems P. 1996. http://alt-www.uia.ac.be/u/dnalab/l1.html

57. Zhao X, Siu C-H. 1996.J. Biol. Chem.271:6563–66

58. Doyle JP, Colman DR. 1993.Curr. Opin.Cell. Biol.5:779–85

59. Saavedra RA, Fors L, Abersold RH,Arden R, Horvath S, et al. 1989.J. Mol.Evol.29:149–56

60. Lemke G, Axel R. 1985.Cell 40:501–8

61. Lai C, Brow MA, Nave K, NoronhaAB, Quarles RH, et al. 1987.Proc. Natl.Acad. Sci. USA84:4337–41

62. D’Urso D, Brophy PJ, Staugaitis SM,Gillespie CS, Frey AB, et al. 1990.Neu-ron 4:449–60

63. Ding Y, Brunden KR. 1994.J. Biol.Chem.269:10764–70

64. Shapiro L, Doyle JP, Hensley P, Colman

DR, Hendrickson WA. 1996.Neuron17:435–49

65. Kirschner DA, Inouye H, Ganser AL,Mann V. 1989.J. Neurochem.53:1599–609

66. Kirschner DA, Blaurock AE. 1992. InMyelin: Biology and Chemistry, ed.RE Martinson. Boca Raton, FL: CRCPress

67. Patel PI, Lupski JR. 1994.Trends Genet.10:128–33

68. Warner LE, Hilz MJ, Appel SH, KillianJM, Kolodny EH, et al. 1996.Neuron17:451–60

69. Hayasaka K, Himoro M, Sawishi Y,Nanao K, Takahashi T, et al. 1993.Nat.Genet.5:266–68

70. Wilkie AOM, Morriss-Kay GM, JonesEY, Heath JK. 1995.Curr. Biol.5:500–7

71. Hynes RO. 1990.Fibronectins. NewYork: Springer-Verlag

72. Erickson HP, Carrell N, McDonagh J.1981.J. Cell. Biol.91:673–78

73. Odermatt E, Engel J, Richter H,Hormann H. 1982.J. Mol. Biol. 159:109–23

74. Kar L, Lai C-S, Wolff CE, NettesheimD, Sherman S, et al. 1993.J. Biol. Chem.268:8580–89

75. Schwarzbauer JE. 1991.BioEssays13:527–33

76. Potts JR, Phan I, Williams MJ, CampbellID. 1995.Nat. Struct. Biol.2:946–50

77. Baron M, Norman D, Willis A, Camp-bell ID. 1990.Nature345:642–46

78. Williams MJ, Phan I, Harvey TS, Ros-tagno A, Gold LI, et al. 1994.J. Mol.Biol. 235:1302–11

79. Downing AK, Driscoll PC, Harvey TS,Dudgeon TJ, Smith BO, et al. 1992.J.Mol. Biol. 225:821–33

80. Smith BO, Downing AK, Driscoll PC,Dudgeon TJ, Campbell ID. 1995.Struc-ture3:823–33

81. Constantine KL, Madrid M, Banyai L,Trexler M, Patthy L, et al. 1992.J. Mol.Biol. 223:281–98

82. Matsuka YV, Medved LV, Brew SA, Ing-ham KC. 1994.J. Biol. Chem.269:9539–46

83. Ingham KC, Brew SA, MiglioriniMM. 1989. J. Biol. Chem.264:16977–80

84. Litvinovich SV, Strickland DK, MedvedLV, Ingham KC. 1991.J. Mol. Biol.217:563–75

85. Main AL, Harvey TS, Baron M, Boyd J,Campbell ID. 1992.Cell 71:671–78

86. Dickinson CD, Veerapandian B, Dai X-P, Hamlin RC, Xuong N-H, et al. 1994.J. Mol. Biol.236:1079–92

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.




87. Leahy DJ, Aukhil I, Erickson HP. 1996.Cell 84:155–64

88. Pierschbacher MD, Ruoslahti E. 1984.Nature309:30–33

89. Aota S-I, Nomizu M, Yamada KM.1994.J. Biol. Chem.269:24756–61

90. Bowditch RD, Hariharan M, TominnaEF, Smith JW, Yamada KM, et al. 1994.J. Biol. Chem.269:10856–63

91. Barkalow FJB, Schwarzbauer JE. 1991.J. Biol. Chem.266:7812–18

92. Busby TF, Argraves WS, Brew SA,Pechik I, Gilliland GL, et al. 1995.J.Biol. Chem.270:18558–62

93. Erickson HP. 1993.Curr. Opin. CellBiol. 5:869–76

94. Jones FS, Hoffman S, Cunningham B,Edelman GM. 1989.Proc. Natl. Acad.Sci. USA86:1905–9

95. Erickson HP, Inglesias JL. 1984.Nature311:267–69

96. Leahy DJ, Hendrickson WA, AukhilI, Erickson HP. 1992. Science258:987–91

97. Joshi P, Chung C-Y, Aukhil I,Erickson HP. 1993. J. Cell Sci.106:389–400

98. Siriramarao P, Mendler M, BourdonMA. 1993.J. Cell Sci.105:1001–12

99. Hynes RO. 1992.Cell 69:11–25100. Wickham TJ, Mathias P, Cheresh DA,

Nemerow GR. 1993.Cell 73:309–19

101. Larson RS, Corbi AL, Berman L,Springer T. 1989.J. Cell Biol.120:703–12

102. Colombatti A, Bonaldo P. 1991.Blood77:2305–15

103. Diamond MS, Springer TA. 1994.Curr.Biol. 4:506–17

103a. Springer TA. 1997.Proc. Natl. Acad.Sci. USA94:65–72

103b. Tuckwell DS, Brass A, Humphries MJ.1992.Biochem. J.285:325–31

103c. Zhang X-P, Puzon-McLaughlin W, IrieA, Kovach N, Prokopishyn NL, et al.1997.J. Biol. Chem.In press

103d. Irie A, Kamata T, Puzon-McLaughlinW, Takada Y. 1995. EMBO J.14:5550–56

104. Lee JO, Rieu P, Arnaout MA, Lidding-ton R. 1995.Cell 80:631–38

105. Lee JO, Bankston LA, Arnaout MA,Liddington RC. 1995.Structure3:1333–40

106. Qu A, Leahy DJ. 1995.Proc. Natl. Acad.Sci. USA92:10277–81

107. Qu A, Leahy DJ. 1996.Structure4:931–42

108. Rossman MG, Moras D, Olsen KW.1974.Nature250:194–99

109. Landis RC, McDowall A, Holness CLL,Littler AJ, Simmons DL, Hogg N. 1994.J. Cell Biol.126:529–37

110. Diamond MS, Garcia-Aguilar J, Bick-ford JK, Corbi AL, Springer TA. 1993.J. Cell. Biol.120:1031–43

111. Huang C, Springer TA. 1995.J. Biol.Chem.270:19008–16

112. Edwards CP, Champe M, Gonzalez T,Wessinger ME, Spencer SA, et al. 1995.J. Biol. Chem.270:12635–40

113. Kamata T, Wright R, Takada Y. 1995.J.Biol. Chem.270:12531–35

114. Ruoslahti E, Pierschbacher MD. 1987.Science238:491–97

115. Osborn L, Hession C, Tizard R, Vas-sallo C, Luhowskyj S, et al. 1989.Cell59:1203–11

116. Simmons DL, Makgoba MW, Seed B.1988.Nature331:624–27

117. Staunton DE, Marlin SD, Stratowa C,Dustin ML, Springer TA. 1988.Cell52:925–33

118. Staunton DE, Dustin ML, Springer TA.1989.Nature339:361–64

119. Fawcett J, Holness CLL, Needham LA,Turley H, Gatter KC, et al. 1992.Nature360:481–44

120. Vazeux R, Hoffman PA, Tomita JK,Dickinson ES, Jasman RL, et al. 1992.Nature360:485–88

121. Briskin MJ, McEvoy LM, Butcher EC.1993.Nature363:461–64

122. Springer TA. 1990.Nature 346:425–34

123. Staunton DE, Merluzzi VJ, Rothlein R,Barton R, Marlin SD, et al. 1989.Cell56:849–53

124. Greve JM, Davis G, Meyer AM,Forte CP, Yost SC, et al. 1989.Cell56:839–47

125. Berendt AR, Simmons DL, Tansey J,Newbold CI, Marsh K. 1989.Nature341:57–59

126. Huber SA. 1994.J. Virol. 68:3453–58

127. Jones EY, Harlos K, Bottomley MJ,Robinson RC, Driscoll PC, et al. 1995.Nature373:539–44

128. Wang JH, Pepinsky RB, Stehle T, LiuJH, Karpusas M, et al. 1995.Proc. Natl.Acad. Sci. USA92:5714–18

128a. Bork P, Downing AK, Kieffer B, Camp-bell ID. 1996. Quart. Rev. Biophys.29:119–67

129. Renz ME, Chin HH, Jones S, Fox J, KimKJ, et al. 1994.J. Cell Biol.125:1395–406

130. Osborn L, Vassallo C, Browning BG, Ti-zard R, Haskard DO, et al. 1994.J. CellBiol. 124:601–8

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.



862 CHOTHIA & JONES

131. Vonderheide RH, Tedder TF, SpringerTA, Staunton DE. 1994.J. Cell Biol.125:215–22

132. Clements JM, Newham P, Shepherd M,Gilbert R, Dudgeon TJ, et al. 1994.J.Cell Sci.107:2127–35

133. Chiu HH, Crowe DT, Renz ME, PrestaLG, Jones S, et al. 1995.J. Immunol.155:5257–67

134. Holness CL, Bates PA, Littler AJ, Buck-ley CD, McDowall A, et al. 1995.J. Biol.Chem.270:877–84

135. Hunig T. 1985.J. Exp. Med.162:890–901

136. Selvaraj P, Plunkett ML, Dustin ML,Sanders ME, Shaw S, Springer TA.1987.Nature326:400–3

137. Kato K, Koyanagi M, Okada H,Takanashi T, Wong YW, et al. 1992.J.Exp. Med.176:1241–49

138. van der Merwe PA, McPherson DC,Brown MH, Barclay AN, Cyster JG,et al. 1993.Eur. J. Immunol.23:1373–77

139. Davis SJ, van der Merwe PA. 1996.Im-munol. Today17:177–87

140. Killeen N, Moessner R, Arvieux J, WillisA, Williams AF. 1988.EMBO J.7:3087–91

141. Jones EY, Davis SJ, Williams AF, Har-los K, Stuart DI. 1992.Nature360:232–39

142. Bodian DL, Jones EY, Harlos K, Stu-art DI, Davis SJ. 1994.Structure2:755–66

143. Driscoll PC, Cyster JG, Campbell ID,Williams AF. 1991. Nature 353:762–65

144. Withka JM, Wyss DF, Wagner G, Aru-lanandam ARN, Reinherz EL, RecnyMA. 1993.Structure1:69–81

145. Ryu S-E, Kwong PD, Truneh A, PorterTG, Arthos J, et al. 1990.Nature348:419–26

146. Wang J, Yan Y, Garrett PJ, Liu J,Rodgers DW, et al. 1990.Nature348:411–18

147. Garrett TPJ, Wang J, Yan Y, Liu J, Harri-son SC. 1993.J. Mol. Biol.234:763–78

148. . Ryu S-E, Truneh A, Sweet RW, Hen-drickson WA. 1994.Structure2:59–74

149. Brady RL, Dodson EJ, Dodson GG,Lange G, Davis SJ, et al. 1993.Science260:979–83

150. Lesk AM, Chothia C. 1988.Nature335:188–90

151. van der Merwe PA, Brown MH, DavisSJ, Barclay AN. 1993.Biochem. Soc.Trans.21:340S

152. Peterson A, Seed B. 1987.Nature329:842–46

153. Wolff HL, Burakoff SJ, Bierer BE. 1990.J. Immunol.144:1215–20

154. Arulanandam ARN, Withka JM, WyssDF, Wagner G, Kister A, et al. 1993.Proc. Natl. Acad. Sci. USA90:11613–17

155. Somoza C, Driscoll PC, Cyster JG,Williams AF. 1993. J. Exp. Med.178:549–58

156. van der Merwe PA, McNamee PN,Davies EA, Barclay AN, Davis SJ. 1995.Curr. Biol. 5:74–84

157. McAlister MSB, Mott HR, van derMerwe PA, Campbell ID, Davis SJ,Driscoll PC. 1996.Biochemistry 35:5982–91

158. Murzin AG, Brenner S, Hubbard T,Chothia C. 1995.J. Mol. Biol.247:536–40

159. Labeit S, Kolmerer B. 1995.Science270:293–96

160. Patthy L. 1994.Curr. Opin. Struct. Biol.4:383–92

161. van der Merwe PA, Barclay AN. 1994.Trends Biochem. Sci.19:354–58

162. Chan P-Y, Lawrence MB, Dustin ML,Ferguson DE, Golan DE, et al. 1991.J.Cell. Biol.115:245–55

163. Dustin ML, Ferguson LM, Chan P-Y,Springer TA, Golan DE. 1996.J. Cell.Biol. 132:465–74

164. Kraulis PJ. 1991.J. App. Cryst.24:946–50

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.

Annual Review of Biochemistry Volume 66, 1997

CONTENTSFrom Chemistry to Biochemistry to Catalysis to Movement, William P. Jencks 1

Mechanistic Aspects of Enzymatic Catalysis: Lessons from Comparison ofRNA and Protein Enzymes, Geeta J. Narlikar and Daniel Herschlag 19

Replication Protein A: A Heterotrimeric, Single-Stranded DNA-Binding Protein Required for Eukaryotic DNA Metabolism, Marc S. Wold 61

Bacterial Cell Division and the Z Ring, Joe Lutkenhaus and S. G. Addinall 93

Basic Mechanisms of Transcript Elongation and Its Regulation, S. M. Uptain, C. M. Kane, and M. J. Chamberlin 117

Polyadenylation of mRNA in Prokaryotes, Nilima Sarkar 173Molecular Basis for Membrane Phospholipid Diversity: Why Are There So Many Lipids, W. Dowhan 199

Molybdenum-Cofactor Containing Enzymes: Structure and Mechanism, Caroline Kisker, Hermann Schindelin, and Douglas C. Rees 233

Structure-Based Perspectives on B12-Dependent Enzymes, Martha L. Ludwig and Rowena G. Matthews 269

Dynamic O-Linked Glycosylation of Nuclear and Cytoskeletal Proteins, Gerald W. Hart 315

D-Amino Acids In Animal Peptides, G. Kreil 337Herpes Simplex Virus DNA Replication, Paul E. Boehmer and I. R. Lehman 347

MODELS OF AMYLOID SEEDING IN ALZHEIMER'S DISEASE AND SCRAPIE:Mechanistic Truths and Physiological Consequences of the Time-Dependent Solubility of Amyloid Proteins, James D. Harper and Peter T. Lansbury Jr.

385

Mitochondrial DNA Maintenance in Vertebrates, Gerald S. Shadel and David A. Clayton 409

Target Site Selection in Transposition, Nancy L. Craig 437Regulation of Eukaryotic Phosphatidylinositol-Specific Phospholipase C and Phospholipase, William D. Singer, H. Alex Brown, and, Paul C. Sternweis

475

Clathrin-Coated Vesicle Formation and Protein Sorting: An Integrated Process, Sandra L. Schmid 511

Protein Folding: The Endgame, Michael Levitt, Mark Gerstein, Enoch Huang, S. Subbiah, and Jerry Tsai 549

Regulation of Phosphoenolpyruvate Carboxykinase (GTP) Gene Expression, Richard W. Hanson and Lea Reshef 581

Acyl-Coenzyme A:Cholesterol Acyltransferase, T. Y. Chang, Catherine C. Y. Chang, and Dong Cheng 613

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.

G PROTEIN MECHANISMS: Insights from Structural Analysis, Stephen R. Sprang 639

Ribosomes and Translation, Rachel Green and Harry F. Noller 679

The ATP Synthase-A Splendid Molecular Machine, Paul D. Boyer 717

Subtractive Cloning: Past, Present, and Future, C. G. Sagerström, B. I. Sun, and H. L. Sive 751

Force Effects on Biochemical Kinetics, Shahid Khan and Michael P. Sheetz 785

Transcriptional Regulation by Cyclic AMP, Marc Montminy 807The Molecular Structure of Cell Adhesion Molecules, Cyrus Chothia and E. Yvonne Jones 823

Protein Import into Mitochondria, Walter Neupert 863

Ann

u. R

ev. B

ioch

em. 1

997.

66:8

23-8

62. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by E

urop

ean

Mol

ecul

ar B

iolo

gy L

ab -

Hei

delb

erg

on 0

9/26

/05.

For

per

sona

l use

onl

y.

THE MOLECULAR STRUCTURE OF CELL ADHESION...

Documents

Transcript of THE MOLECULAR STRUCTURE OF CELL ADHESION...