P1: LDI

July 29, 2000 14:55 AR102 CHAP13

Annu. Rev. Biochem. 2000. 69:373–98Copyright c© 2000 by Annual Reviews. All rights reserved

PROTEIN TYROSINE KINASE STRUCTURE

AND FUNCTION

Stevan R. Hubbard and Jeffrey H. TillSkirball Institute of Biomolecular Medicine and Department of Pharmacology, New YorkUniversity School of Medicine, New York, New York 10016;e-mail: hubbard@tallis.med.nyu.edu; till@saturn.med.nyu.edu

Key Words tyrosine phosphorylation, signal transduction, enzyme, growth factorreceptor, X-ray crystallography

■ Abstract Tyrosine phosphorylation is one of the key covalent modifications thatoccurs in multicellular organisms as a result of intercellular communication duringembryogenesis and maintenance of adult tissues. The enzymes that carry out thismodification are the protein tyrosine kinases (PTKs), which catalyze the transfer ofthe γ phosphate of ATP to tyrosine residues on protein substrates. Phosphorylationof tyrosine residues modulates enzymatic activity and creates binding sites for therecruitment of downstream signaling proteins. Two classes of PTKs are present incells: the transmembrane receptor PTKs and the nonreceptor PTKs. Because PTKs arecritical components of cellular signaling pathways, their catalytic activity is strictlyregulated. Over the past several years, high-resolution structural studies of PTKs haveprovided a molecular basis for understanding the mechanisms by which receptor andnonreceptor PTKs are regulated. This review will highlight the important results thathave emerged from these structural studies.

CONTENTS

INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374Protein Tyrosine Kinases in Cellular Signaling. . . . . . . . . . . . . . . . . . . . . . . . . . 374Overall Protein Architecture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376

REGULATION OF RECEPTOR TYROSINE KINASES. . . . . . . . . . . . . . . . . . . . 379Tyrosine Autophosphorylation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379Dimerization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380Additional Mechanisms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381

REGULATION OF NONRECEPTOR TYROSINE KINASES. . . . . . . . . . . . . . . . 381Src and Abl. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382Zap70/Syk and Jaks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382

STRUCTURAL STUDIES OF RECEPTOR TYROSINE KINASES. . . . . . . . . . . . 383Ligand-Binding Domains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383Cytoplasmic Domains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387

0066-4154/00/0707-0373/$14.00 373

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

374 HUBBARD ¥ TILL

STRUCTURAL STUDIES OF NONRECEPTOR TYROSINE KINASES. . . . . . . . 392Src Family Tyrosine Kinases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392

CONCLUSIONS AND PROSPECTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395

INTRODUCTION

One of the fundamental mechanisms by which cells in multicellular organismscommunicate is the binding of polypeptide ligands to cell surface receptors thatpossess tyrosine kinase catalytic activity. Receptor tyrosine kinases (RTKs) aretransmembrane glycoproteins that are activated by the binding of their cognateligands, and they transduce the extracellular signal to the cytoplasm by phospho-rylating tyrosine residues on the receptors themselves (autophosphorylation) andon downstream signaling proteins. RTKs activate numerous signaling pathwayswithin cells, leading to cell proliferation, differentiation, migration, or metabolicchanges (1). The RTK family includes the receptors for insulin and for manygrowth factors, such as epidermal growth factor (EGF), fibroblast growth factor(FGF), platelet-derived growth factor (PDGF), vascular endothelial growth factor(VEGF), and nerve growth factor (NGF). In addition to the RTKs, there exists alarge family of nonreceptor tyrosine kinases (NRTKs), which includes Src, theJanus kinases (Jaks), and Abl, among others. The NRTKs are integral componentsof the signaling cascades triggered by RTKs and by other cell surface receptorssuch as G protein-coupled receptors and receptors of the immune system. The spe-cific reaction catalyzed by PTKs is the transfer of theγ phosphate of ATP to thehydroxyl group of a tyrosine in a protein substrate. The need for tight regulationof PTK catalytic activity is underscored by the numerous PTKs that have beenidentified as oncogenes (2, 3).

The focus of this review is on the insights into molecular mechanisms of PTKregulation gained through high-resolution structural studies. Before reviewing thestructural studies, a brief section on the roles of various PTKs in cellular signaling ispresented, followed by a general discussion of some of the regulatory mechanismsto which PTKs are subject.

Protein Tyrosine Kinases in Cellular Signaling

Several examples are cited to illustrate the importance of PTKs in embryonic de-velopment, metabolism, and immune system function. The development of thevascular system relies on the concerted action of several subfamilies of RTKs andtheir cognate ligands (4). The vascular system is formed in a two-step process.In the first step, referred to as vasculogenesis, endothelial cells differentiate toform a crude network of interconnected vessels. In the second step, termed an-giogenesis, the vessels are remodeled and extended, and nonendothelial supportcells are recruited to the maturing vasculature. Vasculogenesis requires the growthfactor VEGF and one of the RTKs through which it acts, KDR. Angiogenesis

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

PTK STRUCTURE AND FUNCTION 375

requires another VEGF receptor, Flt1, as well as the angiogenic factor angiopoietin1, which is a ligand for the RTK Tie2. More recently implicated in angiogenesis(in the demarcation of arteries and veins) is the ligand ephrinB2 and the RTKEphB4 (5). Ephrins and the Eph receptors are better known for their roles in axonguidance (6).

The effects of the hormone insulin are mediated by the insulin receptor, anRTK family member. Insulin binding to its receptor results in receptor activationand the recruitment of a family of downstream signaling molecules, the IRS pro-teins, to the activated receptor (7). The IRS proteins are adaptor proteins, i.e. theyhave no identifiable catalytic function, which are phosphorylated on multiple ty-rosine residues by the insulin receptor. Activation of phosphoinositide 3-kinase(PI-3K) through binding to phosphorylated IRS is a critical step in the translo-cation of glucose transporters to the cell membrane to facilitate glucose uptake(8, 9).

The largest subfamily of NRTKs, with nine members, is the Src family. Srcfamily members participate in a variety of signaling processes, including mito-genesis, T- and B-cell activation, and cytoskeleton restructuring. Multiple in vivosubstrates have been described for Src and include, among others, the PDGF andEGF receptors; the NRTK focal adhesion kinase (Fak); p130Cas, an adapter pro-tein involved in integrin- and growth factor-mediated signaling; and cortactin, anactin-binding protein important for the proper formation of cell matrix contactsites (10). Src has also been implicated in several human carcinomas, includingbreast, lung, and colon cancer (10).

NRTKs are critical components in the regulation of the immune system. TheJak family of NRTKs are noncovalently associated with the cytoplasmic domainof cytokine receptors, such as the interferon-γ receptor, and they are activatedby ligand-induced receptor oligomerization. Activated Jaks then phosphorylatethe cytokine receptors with which they are associated, providing binding sites forthe Stat family of transcription factors. Phosphorylation of Stats by Jaks leadsto Stat dimerization, translocation to the nucleus, and transcription of specificgenes (11).

Signaling by activated T and B cells of the immune system is dependent onmultiple NRTKs (12). The Src family member Lck is constitutively associatedwith the CD4 or CD8 receptor on T lymphocytes. Upon engagement of the T-cellreceptor with antigen, Lck becomes autophosphorylated and phosphorylates theζ chain of the T-cell receptor, providing binding sites for another NRTK, Zap70.Zap70 is recruited to the T-cell receptor, activated via phosphorylation by Lck, andthen participates in downstream signaling events that mediate transcriptional acti-vation of cytokine genes. Likewise, upon stimulation of the B-cell receptor, the Srcfamily member Lyn is activated, which leads to the recruitment and phosphoryla-tion of Syk, a Zap70-related NRTK. Another NRTK, Btk, also plays an importantrole in B-cell signaling. Mutations in the Btk gene are responsible for X-linkedagammaglobulinemia (13, 14), a disease characterized by the lack of mature Bcells.

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

376 HUBBARD ¥ TILL

Overall Protein Architecture

RTKs consist of an extracellular portion that binds polypeptide ligands, a trans-membrane helix, and a cytoplasmic portion that possesses tyrosine kinase catalyticactivity (Figure 1). The vast majority of RTKs exist as a single polypeptide chainand are monomeric in the absence of ligand. Exceptions include Met and its familymembers, which comprise a shortα chain disulfide-linked to a membrane-spanningβ chain, and the insulin receptor and its family members, which consist of twoextracellularα chains disulfide-linked to two membrane-spanningβ chains. Theα chains are also disulfide-linked to one another, forming anα2β2 heterotetramer.Most polypeptide ligands for RTKs are soluble. Exceptions include the ephrins, theligands for the Eph receptor family, which either span the cell membrane or are teth-ered to the membrane via a GPI (glycosyl-phosphatidylinositol) linkage (15, 16).

The extracellular portion of RTKs typically contains a diverse array of discreteglobular domains such as immunoglobulin (Ig)-like domains, fibronectin type III–like domains, cysteine-rich domains, and EGF-like domains. In contrast, the do-main organization in the cytoplasmic portion of RTKs is simpler, consisting of ajuxtamembrane region (just after the transmembrane helix), followed by the tyro-sine kinase catalytic domain and a carboxy-terminal region. Some receptors, mostnotably members of the PDGF receptor family, contain a large insertion of∼100residues in the tyrosine kinase domain. The juxtamembrane and carboxy-terminalregions vary in length among RTKs. Along with the tyrosine kinase insert, theseregions contain tyrosine residues that are autophosphorylated upon ligand binding.

NRTKs lack receptor-like features such as an extracellular ligand-binding do-main and a transmembrane-spanning region, and most NRTKs are localized in thecytoplasm (17). Some NRTKs are anchored to the cell membrane through amino-terminal modification, such as myristoylation or palmitoylation. In addition toa tyrosine kinase domain, NRTKs possess domains that mediate protein-protein,protein-lipid, and protein-DNA interactions (Figure 2). The most commonly foundprotein-protein interaction domains in NRTKs are the Src homology 2 (SH2) and 3(SH3) domains (18). The SH2 domain is a compact domain of∼100 residues thatbinds phosphotyrosine residues in a sequence-specific manner. The smaller SH3domain (∼60 residues) binds proline-containing sequences capable of forming apolyproline type II helix.

Some NRTKs lack SH2 and SH3 domains but possess subfamily-specific do-mains used for protein-protein interactions. For example, members of the Jak fam-ily contain specific domains that target them to the cytoplasmic portion of cytokinereceptors. The NRTK Fak possesses two domains that mediate protein-protein in-teractions: an integrin-binding domain and a focal adhesion-binding domain. TheNRTK Abl contains a nuclear localization signal and is found in both the nucleusand the cytoplasm. In addition to SH2 and SH3 domains, Abl possesses an Factin–binding domain and a DNA-binding domain.

Another modular domain, present in the Btk/Tec subfamily of NRTKs andin many other signaling proteins, is the pleckstrin homology (PH) domain. PH

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

PTK STRUCTURE AND FUNCTION 377

Met

Ron

Sea

Axl

Eyk

Tyro

3N

yk

Ros

Ret

Ror

1R

or2

Flt1

KDR

Flt4

InsR

IGF1

RIR

RC

SF1R

Kit

Flk2

PDG

FRα

PDG

FRβ

FGFR

1FG

FR2

FGFR

3FG

FR4

EGFR

ErbB

2Er

bB3

ErbB

4

TrkA

TrkB

TrkC

DD

R1

DD

R2

Tie

Tek

EphA

1

EphB

1

cyst

eine

-ric

h

fibro

nect

inty

pe II

I

Ig EG

F

cadh

erin

krin

gle

leuc

ine-

rich

L disc

oidi

n

tyro

sine

ki

nase

Ryk

SA

M

. . . . . .

MuS

K

Fig

ure

1D

omai

nor

gani

zatio

nfo

ra

vari

ety

ofR

TK

s.T

heex

trac

ellu

lar

port

ion

ofth

ere

cept

ors

ison

top

and

the

cyto

plas

mic

port

ion

ison

botto

m.S

ome

RT

Ks

(e.g

.PD

GF

rece

ptor

)co

ntai

na

larg

ein

sert

inth

ety

rosi

neki

nase

dom

ain,

whi

chis

repr

esen

ted

asa

brea

kin

the

rect

angu

lar

sym

bol.

The

leng

ths

ofth

ere

cept

ors

assh

own

are

only

appr

oxim

atel

yto

scal

e.

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

378 HUBBARD ¥ TILL

Abl

Z

ap70

Ja

k B

tk

Fak

Src

C

sk

Fes

SH

3

SH

2

kina

se

pseu

do-k

inas

e

plec

kstr

in h

omol

ogy

inte

grin

bin

ding

foca

l adh

esio

n bi

ndin

g

DN

A b

indi

ng

F-a

ctin

bin

ding

Jak

hom

olog

y

Fig

ure

2D

omai

nor

gani

zatio

nfo

rth

em

ajor

subf

amili

esof

NR

TK

s.T

heam

ino

term

inus

ison

the

left

and

the

carb

oxy

term

inus

ison

the

righ

t.T

hele

ngth

sof

the

NR

TK

sas

show

nar

eon

lyap

prox

imat

ely

tosc

ale.

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

PTK STRUCTURE AND FUNCTION 379

domains bind to phosphatidylinositol (PtdIns) lipids that have been phosphorylatedat particular positions on the head group (19). Concentrations of specific PtdInslipids in the cell membrane, such as PtdIns-3,4,5-P3, increase as a consequence ofPI-3K activation. Proteins can be recruited to activated signaling complexes at themembrane through PH domain interactions with phosphorylated PtdIns lipids.

REGULATION OF RECEPTOR TYROSINE KINASES

Tyrosine Autophosphorylation

Activation of RTKs typically requires two processes: enhancement of intrinsiccatalytic activity and creation of binding sites to recruit downstream signalingproteins. For the majority of RTKs, both of these processes are accomplishedby autophosphorylation on tyrosine residues, a consequence of ligand-mediatedoligomerization. In general, autophosphorylation of tyrosines in the activation loopwithin the kinase domain results in stimulation of kinase activity, and autophos-phorylation of tyrosines in the juxtamembrane, kinase insert, and carboxy-terminalregions generates docking sites for modular domains that recognize phosphoty-rosine in specific sequence contexts. The two well-established phosphotyrosine-binding modules present within signaling proteins are the SH2 domain and thephosphotyrosine-binding (PTB) domain (18).

Two examples serve to illustrate the recruitment roles of phosphotyrosineson activated RTKs. The juxtamembrane region of TrkA, the receptor for NGF,contains a tyrosine in an NPXY motif that, upon autophosphorylation, becomesan interaction site for the PTB domain of Shc (20, 21). Engagement of Shc viathis autophosphorylation site leads to Shc phosphorylation by TrkA, recruitmentof Grb2 and Sos to phosphorylated Shc, and Ras activation. Similarly, within thecarboxy-terminal tail of FGF receptor 1 resides a tyrosine autophosphorylation sitethat serves as a high-affinity binding site for the SH2 domain of phospholipase Cγ

(PLCγ ). The recruitment of PLCγ to FGF receptor 1 via this autophosphorylationsite leads to phosphorylation and activation of PLCγ by the receptor (22, 23).

All RTKs thus far identified contain between one and three tyrosines in thekinase activation loop, which comprises subdomains VII and VIII of the proteinkinase catalytic core (24). Phosphorylation of these tyrosines has been shownto be critical for stimulation of catalytic activity and biological function for anumber of RTKs, including the insulin receptor (25), FGF receptor (26), VEGFreceptor (27), PDGF receptor (28), Met (hepatocyte growth factor receptor) (29),and TrkA (30). The major exception to catalytic enhancement via activation loopautophosphorylation is the EGF receptor. Although a tyrosine in the activationloop is conserved in this RTK subfamily, substitution with phenylalanine has nodemonstrable effect on the signaling properties of the receptor (31).

In principle, receptor autophosphorylation could occur incis (within a recep-tor) or in trans (between receptors). In the first case, ligand-induced dimeriza-tion would cause a conformational change in the receptor that would facilitate

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

380 HUBBARD ¥ TILL

cis-autophosphorylation. In the second case, no conformational change need occurupon dimerization; a simple proximity effect would provide sufficient opportunityfor trans-autophosphorylation to occur. Based on structural studies of the insulinreceptor kinase domain (32, 33), steric considerations indicate that activation looptyrosines in PTKs can only be phosphorylated intrans. Other autophosphorylationsites (e.g. in the juxtamembrane region or carboxy-terminal tail) could potentiallybe autophosphorylated incis.

Dimerization

As discussed above, tyrosine autophosphorylation is the essential modification thatoccurs during RTK activation. Ligand-induced oligomerization of RTKsis the mechanism by which tyrosine autophosphorylation is triggered (34, 35).Ligand binding to the extracellular portion of RTKs mediates the noncovalentoligomerization of monomeric receptors or induces a structural rearrangementin heterotetrameric receptors (e.g. the insulin receptor), facilitating tyrosine au-tophosphorylation in the cytoplasmic domains.

Whether receptor dimerization is sufficient for signal transmission or whetherhigher-order oligomerization is required has not been fully resolved; it likely de-pends on the particular ligand-RTK system. For receptors that bind dimeric ligands,such as the PDGF receptor, a receptor dimer is likely to be a competent signalingunit. However, not all dimeric configurations of a receptor are capable of signaling.The introduction of cross-linking cysteine residues into the transmembrane helixof ErbB2, an EGF receptor family member, indicates that ErbB2 activation is de-pendent on the relative orientation of the two receptors in the dimer (36). For Ephreceptors, biochemical studies show that although a dimeric ephrin is sufficientfor receptor autophosphorylation, a tetrameric ephrin is necessary to elicit the fullrange of biological responses in cells (37). In most cases, RTK dimerization isprobably sufficient for transducing the biological signal.

Ligand binding stabilizes a dimeric configuration of the extracellular domains ofRTKs, but the spatial relationship between the tyrosine kinase-containing cytoplas-mic domains within the dimer is not well understood. The cytoplasmic domainsmay associate only transiently, acting as enzyme and substrate for the other, orthey may interact stably to form a symmetric (or asymmetric) dimer before and/orafter autophosphorylation. For those RTKs whose kinase activity is stimulatedvia activation loop phosphorylation, the transient association model appears tobe consistent with the available biochemical data. In this model, all sites couldconceivably be autophosphorylated within the dimer (i.e. higher-order receptorinteractions would not be required).

If the two cytoplasmic domains in the ligand-mediated dimer form a stablecomplex before autophosphorylation, steric constraints would precludetrans-autophosphorylation of a subset of sites (those nearest the kinase domains), inwhich case higher-order receptor association would be necessary to complete au-tophosphorylation. For the EGF receptor, which does not undergo activation loopautophosphorylation, biochemical evidence suggests that a cytoplasmic domain

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

PTK STRUCTURE AND FUNCTION 381

dimer is required for catalytic enhancement (38, 39); autophosphorylated but mo-nomeric EGF receptors are not activated. Interestingly, all of the identified au-tophosphorylation sites in the EGF receptor are in the long carboxy-terminal tailof the receptor. It is conceivable (from steric considerations) that all of these sitescould be autophosphorylated by the cytoplasmic domain dimer, although evidenceexists for autophosphorylation occurring between pairs of EGF receptor dimers(40). A further level of positive regulation may come from Src phosphorylation ofTyr-845 in the activation loop (41, 42).

Additional Mechanisms

In the simple RTK activation model, ligand-mediated dimerization of the ex-tracellular domains promotes transient association of the cytoplasmic domainsto facilitate trans-autophosphorylation. In this model, tyrosine to phenylalaninesubstitutions of nonactivation loop autophosphorylation sites should not adverselyaffect autophosphorylation of the remaining sites. For some RTKs (e.g. the PDGFreceptor), the activation process appears to be more complicated. Substitution oftwo tyrosines with phenylalanine in the juxtamembrane region of the PDGFβ re-ceptor drastically reduces autophosphorylation of the numerous other sites in thereceptor, which are readily autophosphorylated in the wild-type receptor (43, 44).One possible explanation for this phenomenon is that autophosphorylation of thejuxtamembrane sites relieves an inhibitory restraint, similar to the effect of activa-tion loop autophosphorylation. Interestingly, autophosphorylation of the activationloop tyrosine in the PDGFβ receptor has been shown to be critical for phosphory-lation of exogenous substrates but not for autophosphorylation (44).

Downregulation of RTKs occurs via several processes, including receptor-mediated endocytosis (45), ubiquitin-directed proteolysis (46), and the action ofprotein tyrosine phosphatases (PTPs) (47). Although it has proven somewhat dif-ficult to determine whether specific PTPs are responsible for dephosphorylatingspecific RTKs in vivo, examples are beginning to emerge. InCaenorhabditis ele-gans, loss-of-function mutations in the CLR-1 PTP give a phenotype that mimicsconstitutive activation of the FGF receptor ortholog, EGL-15, implicating this PTPin the regulation of FGF receptor-mediated signaling in nematodes (48). Disrup-tion of the gene encoding PTP1B in mice results in a phenotype that indicates thatat least one of the functions of PTP1B is to dephosphorylate the insulin receptor(49).

REGULATION OF NONRECEPTOR TYROSINE KINASES

The most common theme in NRTK regulation, as in RTK regulation, is tyrosinephosphorylation. With few exceptions, phosphorylation of tyrosines in the acti-vation loop of NRTKs leads to an increase in enzymatic activity. Activation loopphosphorylation occurs viatrans-autophosphorylation or phosphorylation by adifferent NRTK. Phosphorylation of tyrosines outside of the activation loop can

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

382 HUBBARD ¥ TILL

negatively regulate kinase activity. PTPs restore NRTKs to their basal state ofactivity or, in some cases, positively regulate NRTK activity (47).

Src and Abl

Regulation of Src catalytic activity has been studied extensively (50). Src and itsfamily members contain a myristoylated amino terminus, a stretch of positively-charged residues that interact with phospholipid head groups, a short region withlow sequence homology, an SH3 domain, an SH2 domain, a tyrosine kinasedomain, and a short carboxy-terminal tail (Figure 2). Src possesses two impor-tant regulatory tyrosine phosphorylation sites. Phosphorylation of Tyr-527 in thecarboxy-terminal tail of Src by the NRTK Csk represses kinase activity (51). Theimportance of this phosphorylation site is underscored by v-Src, an oncogenic vari-ant of Src that is a product of the Rous sarcoma virus. Owing to a carboxy-terminaltruncation, v-Src lacks the negative regulatory site Tyr-527 and is constitutivelyactive, leading to uncontrolled growth of infected cells (52). Moreover, substitu-tion of this tyrosine with phenylalanine in c-Src results in activation (53). A secondregulatory phosphorylation site in Src is Tyr-416, an autophosphorylation site inthe activation loop. Maximal stimulation of kinase activity occurs when Tyr-416is phosphorylated, and a Tyr-416→Phe mutation can suppress the transformingability of the activating Tyr-527→Phe mutation (53).

Both the SH2 and SH3 domains have been implicated in the negative regulationof Src activity (50); mutations in the SH2 and SH3 domains that disrupt bindingof phosphotyrosine and proline-rich sequences, respectively, activate Src. Themechanisms by which the SH2 and SH3 domains repress Src kinase activity havebeen elucidated through X-ray crystallographic studies and are discussed below.

Although the NRTK Abl contains SH3, SH2, and kinase domains in the samelinear order as in Src (Figure 2), regulation of Abl differs from that of Src. Abllacks the negative regulatory phosphorylation site that is present in the carboxyterminus of Src. In contrast to Src, the carboxy terminus of Abl does not have afunctional role in the control of kinase activity, and mutations in the SH2 domainof Abl that abrogate phosphotyrosine binding do not activate Abl in vivo (54).The SH3 domain of Abl, however, does play a role in the repression of kinaseactivity; mutations in the SH3 domain result in activation of Abl and cellulartransformation (55). Significantly, the mutations in the SH3 domain have no effecton tyrosine kinase activity in vitro, which suggests that the SH3 domain binds to acellular inhibitor in vivo. A leading candidate for such an inhibitor is Pag/MSP23,a member of the peroxiredoxin family of antioxidant enzymes. Pag/MSP23 iscapable of complexing with the SH3 domain of Abl and inhibiting Abl kinaseactivity when overexpressed (56).

Zap70/Syk and Jaks

The SH2 domains of Syk have been implicated in the regulation of Syk kinase activ-ity. Engagement of the two SH2 domains of Syk with the tyrosine-phosphorylatedITAM (immunoreceptor tyrosine-based activation motif) sequences in theζ chain

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

PTK STRUCTURE AND FUNCTION 383

of the T-cell receptor is thought to relieve an inhibitory restraint on the kinasedomain, leading to stimulation of catalytic activity (57). Phosphorylation of Tyr-493 in the Zap70 activation loop by Src family member Lck has been shown toincrease Zap70 catalytic activity (58). Curiously, mutation of the adjacent Tyr-492to phenylalanine results in Zap70 hyperactivity, which suggests that phosphory-lation of Tyr-492 is inhibitory (59). It remains to be demonstrated, however, thatthis hyperactivity is caused by a negative effect of Tyr-492 phosphorylation ratherthan a positive (aberrant) effect of phenylalanine vs tyrosine.

In addition to a fully functional tyrosine kinase domain, Jak family memberspossess a pseudo-kinase domain in which substitution of several key catalyticresidues renders the domain inactive (60). Although enzymatically nonfunctional,the pseudo-kinase domain may play a role in the regulation of Jak activity. Amutant of the Jak family member Tyk2, in which the pseudo-kinase domain isdeleted, lacks catalytic activity in vitro and is not capable of interferon-mediatedsignal transduction (61). In contrast, a Jak2 mutant lacking the pseudo-kinasedomain was able to mediate growth hormone signaling (62). Thus, the role of thepseudo-kinase domain in Jak regulation is not fully understood.

Like Zap70/Syk, Jaks possess twin tyrosine phosphorylation sites within theactivation loop. Autophosphorylation of the first of these tyrosines is important forstimulation of tyrosine kinase activity and biological function (63, 64). The role ofthe second tyrosine is less clear. In Jak2, substitution of the second tyrosine withphenylalanine had no obvious effect (63), whereas the same substitution in Jak3resulted in an increase in kinase activity (64), akin to the effect of the Tyr-492→Phemutation in Zap70.

Jaks are also regulated by SOCS (suppressor of cytokine signaling) proteins.These proteins contain a pseudo-substrate sequence thought to interfere with Jaksubstrate binding and phosphoryl transfer (65). In addition to a pseudo-substratesequence, SOCS proteins possess an SH2 domain that binds to a phosphotyrosine inthe Jak activation loop (66), which may facilitate interaction between the pseudo-substrate sequence and the kinase domain. Binding of the SH2 domain to theactivation loop could also block substrate access directly or alter the conformationof the activation loop to repress catalytic activity.

STRUCTURAL STUDIES OF RECEPTORTYROSINE KINASES

Ligand-Binding Domains

Several crystal structures of ligand-binding domains of RTKs have been reportedin the last few years, providing a structural basis for understanding dimerizationmechanisms and ligand-receptor specificity. In general, only a subset of domainsin the extracellular portion of an RTK are involved in ligand binding. The VEGFreceptor Flt1 contains seven Ig-like domains in its extracellular region (Figure 1),with domains 2 and 3 responsible for the interaction with VEGF. The crystal

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

384 HUBBARD ¥ TILL

N N

NN

CCC C

N

N

N

N

C

C

C

C

Figure 3 Mechanism of dimerization of VEGF and Flt1 (67). A ribbon diagram is shownwith the two protomers of disulfide-linked VEGF shown inorangeandpurple, and Ig-likedomain 2 of Flt1 shown ingreen. The view in the bottom panel is orthogonal to that in thetop panel, as indicated. Figures 3, 4, 6, 8, and 9 were prepared with MOLSCRIPT (114)and RASTER3D (115).

structure of VEGF in complex with Ig-like domain 2 (D2) of Flt1 (67) providesa picture of the simplest dimerization scenario: a dimeric ligand engaging tworeceptors. The structure shows a nearly twofold symmetric arrangement of VEGFand two molecules of D2 (Figure 3). The interface between VEGF and D2 consistsmainly of hydrophobic residues, and the two equivalent receptor-binding sites onthe VEGF dimer comprise residues from both of the VEGF protomers, i.e. thereceptors bind at the junctions of the VEGF protomers. Domain deletion studiesindicate that VEGF binds to D2-D3 with 20-fold higher affinity than to D2 alone(67). This result, along with the observed positioning of VEGF near the carboxy-terminal end of D2, suggests that VEGF also makes contact with D3 over a limitedarea.

Another example of how a dimeric ligand engages two RTK molecules is illus-trated by the crystal structure of NGF bound to the ligand-binding domain of itsreceptor TrkA (68). The extracellular portion of TrkA comprises two cysteine-richdomains separated by a leucine-rich domain, followed by two Ig-like domains

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

PTK STRUCTURE AND FUNCTION 385

(Figure 1). The Ig-like domain proximal to the cell surface, D5, is sufficient forbinding NGF. Analysis of the ligand-receptor interactions observed in the struc-ture, together with sequence comparisons with related neurotrophic factors (BDNFand NT3) and RTKs (TrkB and TrkC), indicates that there is one NGF-D5 interac-tion site that is common in all neurotrophin-Trk binding and a second interactionsite that determines specificity (68). As in the VEGF-D2 interaction, the two D5molecules bind at the interface between the two NGF protomers.

The crystal structure of basic FGF (FGF2) bound to the ligand-binding domainof FGF receptor 1 (69) provides another distinct example of how ligand bindinginduces RTK dimerization. The extracellular portion of FGF receptors containsthree Ig-like domains, the last two of which (D2-D3) are sufficient for FGF binding.FGFs bind to their receptors to form a 1:1 ligand:receptor complex and requireheparin sulfate proteoglycans for receptor dimerization (70). Although no heparinanalog was included in the crystallization solution, the asymmetric unit comprisestwo 1:1 FGF2:D2-D3 complexes related by a pseudo-dyad axis (Figure 4). FGF2engages residues in D2, D3, and the linker that connects the two domains.

Compared with the extensive interactions between FGF2 and D2-D3 in the 1:1complex, the dimer is stabilized by relatively few contacts, which is consistentwith the inability of FGF under normal physiological conditions to dimerize its

Figure 4 Structure-based model of FGF receptor dimerization (69). Shown in full-sphererepresentation are two FGF2 molecules (orange) and two ligand-binding domains of FGFreceptor 1 (greenandcyan). Ig-like domains 2 (D2) and 3 (D3) in the ligand-binding domainare labeled. A heparin dodecasaccharide was manually docked into the putative heparin-binding site and is shown instickrepresentation. Carbon atoms of the dodecasaccharide areshown inblack, oxygen atoms inred, nitrogen atoms inblue, and sulfur atoms inyellow.The view in the right panel is orthogonal to that in the left panel, as indicated.

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

386 HUBBARD ¥ TILL

receptor in the absence of heparin. Stabilization of the dimer comes from contactsbetween FGF2 and the other D2-D3 molecule in the dimer and from receptor-receptor contacts at the base of D2. In contrast to disulfide-linked VEGF and NGF,the two FGF2 molecules are located distally in the dimer and do not contact eachother. Calculation of the electrostatic potential at the surface of the dimer revealsan extensive area of positive potential along the inward faces of D2, extendingonto the top surface of the adjoining FGF2 molecules. Many of the lysine andarginine residues that contribute to the positive potential in both the receptor andligand have been implicated in heparin binding. Manual docking of a heparindodecasaccharide into this region (Figure 4) suggests that a heparin molecule ofsufficient length (octasaccharide or longer) could interact with the heparin-bindingresidues of the receptors and also the adjoining ligands.

Three other structures of RTK ligand-binding domains have been reported,although they do not include bound ligand. A crystal structure of the first threedomains of the insulin-like growth factor 1 (IGF1) receptor (71) provides cluesto the mode of binding of IGF1 and insulin to their respective, related receptors.Domains 1 and 3 of the IGF1 and insulin receptors are so-called L domains (L1and L2), and domain 2 is rich in cysteines (Figure 1). The extracellular portionof members of the EGF receptor family also contains these same three domainsfollowed by another cysteine-rich domain. The L domains feature aβ-helix-typefold, withα helices flanking the amino- and carboxy-terminal ends (Figure 5). Thecysteine-rich domain comprises an array of disulfide-linked modules resemblingthose in tumor necrosis factor receptor (72) and laminin (73).

An important structural feature is the interaction between L1 and the adja-cent cysteine-rich domain. A tryptophan residue from the cysteine-rich domainis embedded in the hydrophobic core of L1. This interaction presumably fixesthe orientation of L1 with respect to the cysteine-rich domain. Interestingly, thistryptophan is conserved in both of the cysteine-rich domains of the EGF receptor.In contrast, no contacts are observed between the cysteine-rich domain and thefollowing L2, which suggests an inherent flexibility between these two domainsthat may be important for ligand binding.

A large cleft is formed by the L domains and the cysteine-rich domain(Figure 5), which is the putative ligand-binding site. Aβ sheet on the bottom of L1and a loop from the cysteine-rich domain line the cleft. Residues important for IGF1and insulin binding have been mapped to these regions (71). The cleft as observedin the crystal structure is too large to bind IGF1 productively, but a rotation of∼25◦

of L2 toward L1 would be sufficient to form a cleft of the appropriate size for IGF1.A low-resolution, three-dimensional reconstruction of the insulin receptor with

bound insulin has been determined by electron microscopy, which provides amodel for the spatial organization of the various domains in the intact receptor(74). The electron density envelope, at a nominal resolution of 20A, was fit withthe structures of individual domains determined at higher resolution by X-raycrystallography. The model indicates that one insulin molecule engages L1 andthe cysteine-rich domain from one receptorα chain and L2 from the otherα chain,

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

PTK STRUCTURE AND FUNCTION 387

Figure 5 Structure of the first three do-mains of the IGF1 receptor (71). Shownis a ribbon diagram of the L1, cysteine-rich, and L2 domains. The putative ligand-binding cleft lies between the threedomains. Residues important for ligandbinding are on the underside of the L1domain (indicated by thearrow) and theloop in the cysteine-rich domain contain-ing residues 255–265. [Reprinted with per-mission from Garrett et al (71).]

consistent with the observed negative cooperativity of insulin binding and the lowaffinity of insulin for the monomeric form of the receptor (αβ). The model alsoindicates that in the insulin-bound state the tyrosine kinase domains are poised fortrans-autophosphorylation.

The extracellular portion of Eph receptors contains two fibronectin type IIIrepeats, a cysteine-rich region, and an amino-terminal domain that is sufficient forbinding ephrins. The crystal structure of the unliganded amino-terminal domain ofEphB2 shows that this domain adopts a jellyrollβ-sandwich fold consisting of twofive-strandedβ sheets (75). The two classes of Eph receptors, A and B, containeither 13 or 17 residues, respectively, in the loop connectingβ-strands H and I,which suggests that this loop may be important in conferring ligand specificity.Mutagenesis studies support this hypothesis (75).

Cytoplasmic Domains

Crystal structures of the tyrosine kinase domains from several RTKs have beenreported. These followed the structure determinations of several related protein ser-ine/threonine kinases (76), the first of which was cyclic AMP-dependent protein

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

388 HUBBARD ¥ TILL

Figure 6 Ribbon diagram of the tyrosine kinase domain of the insulin receptor (33).The α helices are shown inred, the β strands inblue, the nucleotide-binding loop inyellow, the catalytic loop inorange, the activation loop ingreen, and the tyrosine-containingpeptide substrate inpink. The ATP analog (AMP-PNP) is shown inblack ball-and-stickrepresentation. The amino and carboxy termini are denoted by N and C.

kinase (PKA) (77). The overall architecture of the tyrosine kinase domain is sim-ilar to that of the serine/threonine kinases: an amino-terminal lobe comprising afive-strandedβ sheet and oneα helix, and a larger carboxy-terminal lobe that ismainlyα-helical (Figure 6). ATP binds in the cleft between the two lobes, and thetyrosine-containing peptide substrate binds to the carboxy-terminal lobe. Severalresidues are highly conserved in all protein kinases, including several glycinesin the nucleotide-binding loop, a lysine inβ-strand 3, a glutamic acid inα-helixC, an aspartic acid and asparagine in the catalytic loop, and a DFG motif in thebeginning of the activation loop (76; Figure 6).

As discussed previously, autophosphorylation of tyrosines in the activationloop typically leads to stimulation of catalytic activity. Crystal structures of theunphosphorylated forms of the insulin receptor kinase domain (IRK) (32) andthe FGF receptor kinase domain (FGFRK) (78) provide a molecular basis forunderstanding how catalytic activity is repressed before receptor activation. Theactivation loop of the insulin receptor contains three tyrosine autophosphorylationsites. In the crystal structure of unphosphorylated IRK, one of the tyrosines (Tyr-1162) is bound in the active site, hydrogen-bonded to a conserved aspartic acidand arginine in the catalytic loop (Figure 7).

Although Tyr-1162 is seemingly in position to be autophosphorylated incis,the conserved aspartic acid of the DFG motif (Asp-1150) in the beginning of the

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

PTK STRUCTURE AND FUNCTION 389

Figure 7 Comparison of the activation loop conformations in unphosphorylated IRK (32)and tris-phosphorylated IRK (IRK3P) (33). The activation loop is shown ingreen, thecatalytic loop inorange, and the peptide substrate inpink. The rest of the protein in eachcase is represented by a semitransparent molecular surface. Also shown is the ATP analogAMP-PNP, which is partially masked by the amino-terminal lobe of IRK3P. Carbon atomsare shown inwhite, nitrogen atoms inblue, oxygen atoms inred, phosphorus atoms inyellow, and magnesium ions inpurple. Hydrogen bonds between Tyr-1162 and Asp-1132(IRK) and between the substrate tyrosine Y(P) and Asp-1132 (IRK3P) are indicated byblack lines. Figure prepared with GRASP (116).

activation loop, which is involved in Mg-ATP binding (33, 76), is not properlypositioned for catalysis. It appears that the activation loop is too short to positionsimultaneously Asp-1150 for Mg-ATP binding and Tyr-1162 in the active site.Biochemical studies support atrans-autophosphorylation mechanism for Tyr-1162(and Tyr-1158/1163) (79). Substitution of Tyr-1162 with phenylalanine results inan increase in basal-level (no insulin) catalytic activity in the full-length receptor(25, 80), consistent with an autoinhibitory role for Tyr-1162. Thus, the structuraland biochemical data indicate that before autophosphorylation, Tyr-1162 competeswith protein substrates (neighboringβ chain and exogenous substrates) for theactive site.

The FGFRK activation loop contains two tyrosine autophosphorylation sites,Tyr-653/654, corresponding to Tyr-1162/1163 in IRK. Despite>50% sequenceidentity in the FGFRK and IRK activation loops, the conformation of the unphos-phorylated FGFRK activation loop as seen in the crystal structure is significantlydifferent from that in IRK (78). In the FGFRK structure, neither of the activationloop tyrosines are bound in the active site. Rather, the tyrosine kinase-invariantproline at the end of the activation loop and nearby residues are positioned tointerfere with the binding of a substrate tyrosine. Furthermore, in contrast to IRK,the ATP-binding site in the FGFRK structure is not obstructed by the beginningof the activation loop.

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

390 HUBBARD ¥ TILL

The temperature factors (B-factors) derived during crystallographic refinementindicate that portions of the unphosphorylated IRK and FGFRK activation loopsare quite mobile, suggesting that an equilibrium between multiple conformationsexists in solution. A subset of these (e.g. those observed in the IRK and FGFRKcrystal structures) will hinder substrate (protein and/or ATP) binding. For a givenRTK, the conformational equilibrium for the unphosphorylated activation loopmay have been “tuned” through amino acid variation to provide adequate autoin-hibition to deter substrate phosphorylation prior to receptor activation, yet allowtrans-autophosphorylation between receptors that have been dimerized via lig-and binding. Point mutations in the activation loop of several RTKs that result inconstitutive activation may alter this equilibrium (81).

Crystal structures of the phosphorylated kinase domains of the insulin recep-tor (33) and Lck (82), as well as structures of several phosphorylated proteinserine/threonine kinases (77, 83, 84), reveal the role of activation loop phospho-rylation in the stimulation of catalytic activity. Autophosphorylation in the IRKactivation loop brings about a dramatic repositioning of the loop (Figure 7). Theconformation of the tris-phosphorylated IRK activation loop is stabilized in part byinteractions involving the phosphotyrosines, particularly phosphorylated Tyr-1163(pTyr-1163), which is hydrogen-bonded to a conserved arginine in the beginningof the activation loop (Arg-1155) and to a backbone amide nitrogen in the lat-ter half of the loop. The activation loop of mono-phosphorylated Lck adopts asimilar conformation, with pTyr-394 in the same spatial position as pTyr-1163in IRK. In addition to phosphotyrosine interactions, the activation loops of IRKand Lck are stabilized by shortβ-strand interactions with other segments of thecarboxy-terminal lobe.

The crystal structure of the tyrosine kinase domain of the VEGF receptor KDRhas also been reported (85). The VEGF receptors (KDR and Flt1), like the PDGFreceptors, possess a large kinase insert between helices D and E in the carboxy-terminal lobe. After failing to obtain crystals of KDR with an intact kinase insert,McTigue et al (85) were successful in crystallizing a protein in which 50 residues ofthe insert were deleted. KDR possesses two tyrosine autophosphorylation sites inits activation loop, Tyr-1054/1059, which correspond in sequence to Tyr-1158/1163in the insulin receptor. The crystallized protein was phosphorylated on Tyr-1059.As with the insulin and FGF receptors, phosphorylation of the KDR activationloop has a stimulatory effect on kinase activity. As expected from the high degreeof sequence similarity between the kinase domains of KDR and FGF receptor 1(∼55% identity), the two structures are quite similar overall.

An unexpected feature of the KDR kinase structure is the configuration ofthe activation loop. Although phosphorylated, the activation loop is disorderedin the region that includes pTyr-1059. This is in contrast to the structures ofmono-phosphorylated Lck and tris-phosphorylated IRK, in which pTyr-394 andpTyr-1163, respectively, are well ordered and participate in specific interactions.The end of the KDR activation loop in the vicinity of conserved Pro-1168, whichis ordered, adopts the same inhibitory conformation as that seen in the struc-ture of unphosphorylated FGFRK. This suggests that crystallization conditions

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

PTK STRUCTURE AND FUNCTION 391

(e.g. high concentrations of salt or polyethylene glycol) or crystal packing forcesmay, in some instances, influence the conformational equilibrium of the activationloop.

The crystal structure of tris-phosphorylated IRK in complex with an ATP analogand a peptide substrate reveals the mode of substrate binding to tyrosine kinases(33). The cocrystallized 18-residue peptide contains a YMXM motif, which isa favorable substrate sequence for the insulin receptor (86). Only 6 of the 18residues are ordered in the structure, and the residues of the YMXM motif bind asan antiparallelβ strand to the end of the activation loop. The backbone-backboneinteractions between the peptide and the activation loop ensure that only a tyrosineside chain can reach the active site; a serine or threonine side chain is too short. Themethionine side chains fit into two shallow hydrophobic pockets on the surface ofthe carboxy-terminal lobe of the kinase. PTKs exhibit a modest degree of substratesequence selectivity in the vicinity of the tyrosine (87). The variability across thePTK family in the residues comprising the methionine-binding pockets presumablyaccounts for this selectivity, at least in part.

Stimulation of tyrosine kinase activity after activation loop autophosphorylationappears to result from two related effects: (a) removal of steric hindrances tosubstrate binding (Mg-ATP and peptide substrate) and (b) proper positioning ofresidues involved in substrate binding and catalysis. The beginning and end of theactivation loop are the two important regions of the loop. The invariant aspartic acidof the DGF motif (involved in Mg-ATP binding) is found at the beginning of theactivation loop, and the peptide substrate binds to the end of the loop. Additional,nonsteric effects on catalytic rate enhancement from autophosphorylation havebeen reported (88), the nature of which remain to be elucidated.

The precise phosphoryl transfer mechanism used by PTKs, whether associativeor dissociative (89), is not fully understood. Related to this question is the role of theconserved aspartic acid in the catalytic loop (Asp-1132 in IRK), i.e. whether the as-partic acid acts as a catalytic base to abstract the proton from the tyrosine hydroxylgroup or simply positions the hydroxyl group. Biochemical studies on the NRTKCsk indicate that the substrate tyrosine must be neutral for phosphoryl transfer tooccur, an observation that is consistent with a dissociative mechanism (90, 91).

Besides the tyrosine kinase domain, perhaps the only other bona fide globu-lar domain in the cytoplasmic portion of RTKs is the sterile alpha motif (SAM)domain, which is present in the carboxy-terminal region of Eph receptors as wellas in a wide variety of signaling and scaffolding proteins. SAM domains func-tion as homo- and hetero-oligomerization modules. The crystal structures of theSAM domains from EphA4 (92) and EphB2 (93) both reveal a globular domainconsisting of fiveα helices, but the interactions between SAM domains differin the two structures. In the structure of the EphA4 SAM domain, there existsa crystallographic dimer whose interface comprises residues from the extendedamino-terminal segment and the latter part of the carboxy-terminalα helix (92). Incontrast, in the EphB2 SAM domain structure, the asymmetric unit contains eightSAM domains, each of which interacts with its neighbors via two contact regions(distinct from the contact region in the EphA4 dimer), resulting in a continuous

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

392 HUBBARD ¥ TILL

oligomeric structure (93). Thus, clarification of the precise mode of oligomericinteraction between SAM domains will require additional biochemical studies.

STRUCTURAL STUDIES OF NONRECEPTORTYROSINE KINASES

Src Family Tyrosine Kinases

The crystal structures of Src and Hck (another Src family member) provide amolecular framework for understanding the mechanism of regulation of this classof NRTKs (94–96). As outlined above, phosphorylation of Tyr-416 in the activationloop positively regulates catalytic activity, whereas phosphorylation of Tyr-527 inthe carboxy-terminal region negatively regulates activity. The crystallized proteinswere phosphorylated at Tyr-527 and thus represent the downregulated, repressedforms.

In the crystal structures, pTyr-527 is engaged in an intramolecular interac-tion with the SH2 domain (Figure 8), as had been anticipated. Unexpected wasthe intramolecular interaction between the SH3 domain and the segment link-ing the SH2 domain with the kinase domain (SH2-kinase linker), which forms a

Figure 8 Ribbon diagram of downregulated Src (104). The amino- and carboxy-terminalkinase lobes are shown inlight anddark green, respectively. The activation loop in the kinasedomain, containing Tyr-416 (ball and stick), is shown ingray. The SH2 and SH3 domainsare shown indark blueandcyan, respectively. The SH2-kinase linker, which contains ashort stretch of polyproline type II helix (PPII helix), is shown inred. The carboxy-terminaltail, which contains pTyr-527 (ball and stick), is shown inorange. The amino and carboxytermini are denoted by N and C.

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

PTK STRUCTURE AND FUNCTION 393

short polyproline type II helix. The SH2-kinase linker in Hck contains a PXXPmotif, the minimum consensus sequence for SH3 binding, but the linker in Srccontains only one proline. Furthermore, high-affinity binding of SH3 domains toproline-rich sequences generally requires an arginine or lysine on either side ofthe PXXP motif (97), which is absent in the linker. However, the proximity of theSH3 domain to the SH2-kinase linker (on the same polypeptide chain), togetherwith stabilizing effects from the engagement of the SH2 domain with pTyr-527,favors the SH3 interaction with the SH2-kinase linker.

The crystal structures show that the mechanism by which Src kinase activity isrepressed is not by occlusion of the active site by the SH2 or SH3 domains, as mighthave been expected. In fact, these two domains are situated on the back side ofthe kinase domain, opposite the catalytic cleft (Figure 8). Instead, intramolecularinteractions stabilize the misalignment of residues important for catalytic activityand also restrict the relative motion of the two kinase lobes to hinder productiveATP binding.

Structures of protein kinases exhibit considerable diversity in the positioning ofthe amino-terminal kinase lobe with respect to the carboxy-terminal lobe (95). Inthe crystal structure of PKA in complex with a peptide inhibitor and ATP (ternarycomplex), the two lobes of the kinase domain are in close apposition, with ATPbound in between (98, 99). This configuration represents an active kinase state. Incontrast, the two lobes in the unphosphorylated, low-activity form of the insulinreceptor kinase are comparatively far apart (32).

In the crystal structures of downregulated Src/Hck, the two kinase lobes are alsoin a closed configuration. However, a major structural difference between Src/Hckand PKA is the positioning ofα-helix C in the amino-terminal lobe. Within helixC is a glutamic acid (Glu-91 in PKA, Glu-310 in Src) that is conserved in allprotein kinases. In the active form of PKA, the side chain of Glu-91 is salt-bridgedto another protein kinase-invariant residue, Lys-72 (Lys-295 in Src), located inβ-strand 3 of the amino-terminalβ sheet. This lysine coordinates theα andβphosphate groups of ATP in the ternary PKA complex, and its position is affectedby its interaction with Glu-91. The salt bridge between the lysine and glutamic acidis also present in the crystal structure of the phosphorylated (active) kinase domainof Lck (100). In the structures of downregulated Src/Hck, however, helix C is notpacked against theβ sheet in the same manner as in Lck, and Glu-310 is displaced>12 A from Lys-295, pointing away from the ATP-binding site (Figure 9).

The position of helix C (and hence Glu-310) in Src/Hck is influenced by residuesin the SH2-kinase linker. In particular, Trp-260 of the linker inserts into a hydropho-bic pocket at the carboxy-terminal end of the helix, stabilizing the nonproductiveposition of the helix. In the Lck structure, the side chain of Trp-260 adopts aconformation that favors the catalytically productive position of helix C. Muta-tion of Trp-260 to alanine in Src/Hck results in constitutive activation (101, 102).Another residue in the SH2-kinase linker, Leu-255, has been shown to be impor-tant in repression of Src activity (103). In the crystal structure of Src, Leu-255,which is located two residues after the PXXX motif in the SH2-linker, is insertedinto a hydrophobic pocket formed by Trp-286 and Tyr-326 in the amino-terminal

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

394 HUBBARD ¥ TILL

Figure 9 Misalignment of helix C in downregulated Src. Theβ sheets in the amino-terminal lobes of Src (104) and Lck (82) have been superimposed. The protein kinase-invariant lysine inβ-strand 3 (β3) and glutamic acid inα-helix C (αC) are shown inball-and-stickrepresentation. In the phosphorylated (active) Lck structure, the salt bridgebetween Lys-273 and Glu-288 is shown as adotted line. In the structure of downregulatedSrc, the salt bridge is not formed because of the misalignment of helix C.

kinase lobe. Hydrophobicity at residue 255 is conserved in other Src family mem-bers, as is the hydrophobic pocket, which suggests an important role for thisinteraction. Indeed, even conservative substitutions of Leu-255 (Leu-255→Alaor Leu-255→Val) completely deregulate the enzyme, rendering it constitutivelyactive (103).

Autophosphorylation of Tyr-416 in the Src activation loop is required for max-imal kinase activity. In the original crystal structures of Src and Hck, the activationloop is largely disordered (94–96). Therefore, it appeared that the activation loopplayed a passive role in repression, and that autophosphorylation merely stabilizesthe active conformation of the loop. More recent structures of Src and Hck in-dicate that the activation loop is indeed involved in stabilization of the repressedform of Src/Hck (104, 105). In these structures, the segment of the activation loopcontaining Tyr-416 forms a short 310 helix, with Tyr-416 situated in a hydrophobicpocket within the catalytic cleft of the kinase domain. The interactions involv-ing Tyr-416 serve to occlude the substrate-binding site and protect Tyr-416 fromautophosphorylation.

Src can be activated by disruption of the intramolecular restraints on the kinasedomain that are mediated by the SH2 and SH3 domains. Displacement of the SH2and SH3 domains from the carboxy-terminal tail and SH2-kinase linker, respec-tively, can occur through binding of high-affinity ligands. For example, Src can beactivated through SH2 domain binding to the autophosphorylated PDGF recep-tor (43, 106) or through SH3 domain binding to the HIV protein Nef (107). DualSH2-SH3 domain binding to sequences within the NRTK Fak potently activatesSrc (108). Disruption of the intramolecular restraints in Src is thought to destabi-lize the nonproductive position of helix C and the inhibitory conformation of the

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

PTK STRUCTURE AND FUNCTION 395

activation loop, which exposes Tyr-416 for autophosphorylation. The culminationof these structural changes is the optimal positioning of kinase residues that arecritical for catalysis.

CONCLUSIONS AND PROSPECTS

High-resolution structural studies of RTKs and NRTKs, together with a wealth ofbiochemical and genetic data, are continuing to reveal the precise molecular mech-anisms underlying (a) ligand recognition by RTKs, (b) dimerization and activationof RTKs, (c) regulation of catalytic activity in NRTKs, and (d) specificity in PTKsignaling. Future challenges will be determining the structures of additional ligand-RTK complexes and NRTKs, as well as gaining a more detailed understanding ofthe regulatory mechanisms for PTKs whose structures are known.

In addition to providing a fundamental understanding of this important class ofenzymes, structural studies also provide an opportunity for the rational design ofinhibitors capable of interfering with PTK catalytic activity. Given the importanceof PTKs in signaling pathways that lead to cell proliferation, it is not surprisingthat numerous RTKs and NRTKs have been implicated in the onset/progression ofdiseases such as cancer, diabetic retinopathy, atherosclerosis, and psoriasis (109).Potential targets of PTK inhibitors include the ligand-binding sites on RTKs, theATP- and peptide substrate-binding sites in the tyrosine kinase domain, and therecruitment sites on PTKs to which downstream signaling proteins bind. Thusfar, the ATP-binding site in the tyrosine kinase domain has proven to be the mosttractable target, and ATP-competitive inhibitors exhibit a surprising degree ofspecificity given the general sequence conservation in the ATP-binding pocket ofprotein kinases. Numerous structures of PTKs in complex with ATP-competitiveinhibitors are now available (105, 110–113), which reveal the structural basis forthis specificity.

ACKNOWLEDGMENTS

Support is acknowledged from the National Institutes of Health (DK52916) andthe Sidney Kimmel Foundation for Cancer Research (Kimmel Scholar Award toSRH).

Visit the Annual Reviews home page at www.AnnualReviews.org

LITERATURE CITED

1. Schlessinger J, Ullrich A. 1992.Neuron9:383–91

2. Rodrigues GA, Park M. 1994.Curr. Opin.Genet. Dev.4:15–24

3. Hunter T. 1997.Cell 88:333–464. Yancopoulos GD, Klagsbrun M, Folkman J.

1998.Cell 93:661–64

5. Wang HU, Chen ZF, Anderson DJ. 1998.Cell 93:741–53

6. Henkemeyer M, Orioli D, Henderson JT,Saxton TM, Roder J, et al. 1996.Cell86:35–46

7. White MF. 1998.Mol. Cell. Biochem.182:3–11

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

396 HUBBARD ¥ TILL

8. Cheatham B, Vlahos CJ, Cheatham L,Wang L, Blenis J, et al. 1994.Mol. Cell.Biol. 14:4902–11

9. Okada T, Kawano Y, Sakakibara T, HazekiO, Ui M. 1994.J. Biol. Chem.269:3568–73

10. Biscardi JS, Tice DA, Parsons SJ. 1999.Adv. Cancer Res.76:61–119

11. Darnell JE Jr. 1997.Science277:1630–3512. Weiss A, Littman DR. 1994.Cell 76:263–

7413. Vetrie D, Vorechovsky I, Sideras P, Holland

J, Davies A, et al. 1993.Nature361:226–3314. Tsukada S, Saffran DC, Rawlings DJ,

Parolini O, Allen RC, et al. 1993.Cell 72:279–90

15. Flanagan JG, Vanderhaeghen P. 1998.Annu. Rev. Neurosci.21:309–45

16. Holland SJ, Peles E, Pawson T, Schles-singer J. 1998.Curr. Opin. Neurobiol.8:117–27

17. Neet K, Hunter T. 1996.Genes Cells1:147–69

18. Kuriyan J, Cowburn D. 1997.Annu. Rev.Biophys. Biomol. Struct.26:259–88

19. Lemmon MA, Ferguson KM. 1998.Curr.Top. Microbiol. Immunol.228:39–74

20. Obermeier A, Lammers R, WiesmullerKH, Jung G, Schlessinger J, et al. 1993.J. Biol. Chem.268:22963–66

21. Stephens RM, Loeb DM, Copeland TD,Pawson T, Greene LA, et al. 1994.Neuron12:691–705

22. Mohammadi M, Dionne CA, Li W, Li N,Spivak T, et al. 1992.Nature358:681–84

23. Peters KG, Marie J, Wilson E, Ives HE,Escobedo J, et al. 1992.Nature358:678–81

24. Hanks SK, Quinn AM, Hunter T. 1991.Sci-ence241:42–52

25. Ellis L, Clauser E, Morgan DO, Edery M,Roth RA, et al. 1986.Cell 45:721–32

26. Mohammadi M, Dikic I, Sorokin A,Burgess WH, Jaye M, et al. 1996.Mol. Cell.Biol. 16:977–89

27. Kendall RL, Rutledge RZ, Mao X, TebbenAJ, Hungate RW, et al. 1999.J. Biol. Chem.274:6453–60

28. Fantl WJ, Escobedo JA, Williams LT. 1989.Mol. Cell. Biol.9:4473–78

29. Longati P, Bardelli A, Ponzetto C, NaldiniL, Comoglio PM. 1994.Oncogene9:49–57

30. Mitra G. 1991.Oncogene6:2237–4131. Gotoh N, Tojo A, Hino M, Yazaki Y,

Shibuya M. 1992.Biochem. Biophys. Res.Commun.186:768–74

32. Hubbard SR, Wei L, Ellis L, HendricksonWA. 1994.Nature372:746–54

33. Hubbard SR. 1997.EMBO J.16:5572–8134. Ullrich A, Schlessinger J. 1990.Cell 61:

203–1235. Heldin CH. 1995.Cell 80:213–2336. Burke CL, Stern DF. 1998.Mol. Cell. Biol.

18:5371–7937. Stein E, Lane AA, Cerretti DP, Schoeckl-

mann HO, Schroff AD, et al. 1998.GenesDev.12:667–78

38. Mohammadi M, Honegger A, Sorokin A,Ullrich A, Schlessinger J, et al. 1993.Bio-chemistry32:8742–48

39. Sherrill JM. 1997.Biochemistry36:5677–84

40. Sherrill JM. 1997.Biochemistry36:12890–96

41. Sato K, Sato A, Aoto M, Fukami Y.1995. Biochem. Biophys. Res. Commun.215:1078–87

42. Biscardi JS, Maa MC, Tice DA, CoxME, Leu TH, et al. 1999.J. Biol. Chem.274:8335–43

43. Mori S, Ronnstrand L, Yokote K, EngstromA, Courtneidge SA, et al. 1993.EMBO J.12:2257–64

44. Baxter RM, Secrist JP, Vaillancourt RR,Kazlauskas A. 1998.J. Biol. Chem.273:17050–55

45. Sorkin A, Waters CM. 1993.BioEssays15:375–82

46. Mori S, Claesson-Welsh L, Okuyama Y,Saito Y. 1995.Biochem. Biophys. Res.Commun.213:32–39

47. Tonks NK, Neel BG. 1996.Cell87:365–6848. Kokel M, Borland CZ, DeLong L, Horvitz

HR, Stern MJ. 1998.Genes Dev.12:1425–37

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

PTK STRUCTURE AND FUNCTION 397

49. Elchebly M, Payette P, Michaliszyn E,Cromlish W, Collins S, et al. 1999.Science283:1544–48

50. Superti-Furga G, Courtneidge SA. 1995.BioEssays17:321–30

51. Nada S, Okada M, MacAuley A, CooperJA, Nakagawa H. 1991.Nature351:69–72

52. Cooper JA, Gould KL, Cartwright CA,Hunter T. 1986.Science231:1431–34

53. Kmiecik TE, Shalloway D. 1987.Cell49:65–73

54. Mayer BJ, Baltimore D. 1994.Mol. Cell.Biol. 14:2883–94

55. Van Etten RA, Debnath J, Zhou H, Casas-novas JM. 1995.Oncogene10:1977–88

56. Wen ST, Van Etten RA. 1997.Genes Dev.11:2456–67

57. Shiue L, Zoller MJ, Brugge JS. 1995.J.Biol. Chem.270:10498–502

58. Chan AC, Dalton M, Johnson R, Kong GH,Wang T, et al. 1995.EMBO J.14:2499–508

59. Kong G, Dalton M, Wardenburg JB, StrausD, Kurosaki T, et al. 1996.Mol. Cell. Biol.16:5026–35

60. Wilks AF, Harpur AG, Kurban RR, RalphSJ, Zurcher G, et al. 1991.Mol. Cell. Biol.11:2057–65

61. Velazquez L, Mogensen KE, Barbieri G,Fellous M, Uze G, et al. 1995.J. Biol.Chem.270:3327–34

62. Frank SJ, Yi W, Zhao Y, Goldsmith JF,Gilliland G, et al. 1995.J. Biol. Chem.270:14776–85

63. Feng J, Witthuhn BA, Matsuda T, Kohlhu-ber F, Kerr IM, et al. 1997.Mol. Cell. Biol.17:2497–501

64. Zhou YJ, Hanson EP, Chen YQ, MagnusonK, Chen M, et al. 1997.Proc. Natl. Acad.Sci. USA94:13850–55

65. Starr R, Willson TA, Viney EM, Murray LJ,Rayner JR, et al. 1997.Nature387:917–21

66. Sasaki A, Yasukawa H, Suzuki A, Kami-zono S, Syoda T, et al. 1999.Genes Cells4:339–51

67. Wiesmann C, Fuh G, Christinger HW,Eigenbrot C, Wells JA, et al. 1997.Cell91:695–704

68. Wiesmann C, Ultsch MH, Bass SH, de VosAM. 1999.Nature401:184–88

69. Plotnikov AN, Schlessinger J, HubbardSR, Mohammadi M. 1999.Cell98:641–50

70. Spivak-Kroizman T, Lemmon MA, DikicI, Ladbury JE, Pinchasi D, et al. 1994.Cell79:1015–24

71. Garrett TPJ, McKern NM, Meizhen L,Frenkel MJ, Bentley JD, et al. 1998.Na-ture394:395–99

72. Banner DW, D’Arcy A, Janes W, Gentz R,Schoenfeld HJ, et al. 1993.Cell73:431–45

73. Stetefeld J, Mayer U, Timpl R, Huber R.1996.J. Mol. Biol.257:644–57

74. Luo RZ, Beniac DR, Fernandes A, Yip CC,Ottensmeyer FP. 1999.Science285:1077–80

75. Himanen JP, Henkemeyer M, Nikolov DB.1998.Nature396:486–91

76. Taylor SS, Radzio-Andzelm E. 1994.Structure2:345–55

77. Knighton DR, Zheng J, Ten Eyck LF,Ashford VA, Xuong N, et al. 1991.Science253:407–14

78. Mohammadi M, Schlessinger J, HubbardSR. 1996.Cell 86:577–87

79. Wei L, Hubbard SR, Hendrickson WA,Ellis L. 1995.J. Biol. Chem.270:8122–30

80. Wilden PA, Kahn CR, Siddle K, White MF.1992.J. Biol. Chem.267:16660–68

81. Hubbard SR, Mohammadi M, SchlessingerJ. 1998.J. Biol. Chem.273:11987–90

82. Yamaguchi H, Hendrickson WA. 1996.Na-ture384:484–89

83. Russo AA, Jeffrey PD, Pavletich NP. 1996.Nat. Struct. Biol.3:696–700

84. Canagarajah BJ, Khokhlatchev A, CobbMH, Goldsmith EJ. 1997.Cell 90:859–69

85. McTigue MA, Wickersham JA, Pinko C,Showalter RE, Parast CV, et al. 1999.Struct. Fold. Des.7:319–30

86. Shoelson SE, Chatterjee S, Chaudhuri M,White MF. 1992.Proc. Natl. Acad. Sci.USA89:2027–31

87. Zhou S, Carraway KL 3rd, Eck MJ,Harrison SC, Feldman RA, et al. 1995.Na-ture373:536–39

P1: LDI

July 11, 2000 17:24 AR102 CHAP13

398 HUBBARD ¥ TILL

88. Saylor P, Hanna E, Adams JA. 1998.Bio-chemistry37:17875–81

89. Knowles JR. 1980.Annu. Rev. Biochem.49:877–919

90. Cole PA, Grace MR, Phillips RS, BurnP, Walsh CT. 1995.J. Biol. Chem.270:22105–8

91. Kim K, Cole PA. 1998.J. Am. Chem. Soc.120:6851–58

92. Stapleton D, Balan I, Pawson T, SicheriF. 1999.Nat. Struct. Biol.6:44–49

93. Thanos CD, Goodwill KE, Bowie JU.1999.Science283:833–36

94. Xu W, Harrison SC, Eck MJ. 1997.Na-ture385:595–602

95. Williams JC, Weijland A, Gonfloni S,Thompson A, Courtneidge SA, et al.1997.J. Mol. Biol.274:757–75

96. Sicheri F, Moarefi I, Kuriyan J. 1997.Na-ture385:602–9

97. Feng S, Chen JK, Yu H, Simon JA,Schreiber SL. 1994.Science266:1241–47

98. Zheng J, Knighton DR, Ten Eyck LF,Karlsson R, Xuong N, et al. 1993.Bio-chemistry32:2154–61

99. Bossemeyer D, Engh RA, Kinzel V,Ponstingl H, Huber R. 1993.EMBO J.12:849–59

100. Yamaguchi H, Hendrickson WA. 1996.Nature384:484–89

101. Gonfloni S, Williams JC, Hattula K,Weijland A, Wierenga RK, et al. 1997.EMBO J.16:7261–71

102. LaFevre-Bernt M, Sicheri F, Pico A,Porter M, Kuriyan J, et al. 1998.J. Biol.Chem.273:32129–34

103. Gonfloni S, Frischknecht F, Way M,Superti-Furga G. 1999.Nat. Struct. Biol.6:760–64

104. Xu W, Doshi A, Lei M, Eck MJ, HarrisonSC. 1999.Mol. Cell 3:629–38

105. Schindler T, Sicheri F, Pico A, Gazit A,Levitzki A, et al. 1999.Mol. Cell 3:639–48

106. Alonso G, Koegl M, Mazurenko N,Courtneidge SA. 1995.J. Biol. Chem.270:9840–48

107. Moarefi I, LaFevre-Bernt M, Sicheri F,Huse M, Lee CH, et al. 1997.Nature385:650–53

108. Thomas JW, Ellis B, Boerner RJ, KnightWB, White GC 2nd, et al. 1998.J. Biol.Chem.273:577–83

109. Levitzki A, Gazit A. 1995.Science267:1782–88

110. Mohammadi M, McMahon G, Sun L,Tang C, Hirth P, et al. 1997.Science276:955–60

111. Mohammadi M, Froum S, Hamby JM,Schroeder MC, Panek RL, et al. 1998.EMBO J.17:5896–904

112. Lamers MB, Antson AA, Hubbard RE,Scott RK, Williams DH. 1999.J. Mol.Biol. 285:713–25

113. Zhu X, Kim JL, Newcomb JR, Rose PE,Stover DR, et al. 1999.Struct. Fold. Des.7:651–61

114. Kraulis PJ. 1991.J. Appl. Crystallogr.24:946–50

115. Merritt EA, Bacon DJ. 1997.Methods En-zymol.277:505–24

116. Nicholls A, Sharp KA, Honig B. 1991.Proteins11:281–96