Feb 1, 1994 - cation binding domains V and VI contain an ICAM-1 binding site. ..... fragment IH (even at 5 mg/ml). The fact that the dom 5. 120- e.-100-. U 8080 a ..... et al., 1992), using serum-free medium (ADC 1, Biological Industries Ltd).
The EMBO Journal vol. 13 no.8 pp. 1790 - 1798, 1994
Integrin LFA-1 Ol subunit contains site in domains V and VI
Paula Stanley, Paul A.Bates', Joanna Harvey, Robert l.Bennett and Nancy Hogg2 Leukocyte Adhesion Laboratory and IBiomolecular Modelling Laboratory, Imperial Cancer Research Fund, 44 Lincoln's Inn Fields, London WC2A 3PX, UK 2Corresponding author Communicated by M.Fried
In order to identify a binding site for ligand intercellular adhesion molecule-1 (ICAM-1) on the (2 integrin
lymphocyte function-associated antigen-1 (LFA-1), protein fragments of LFA-1 were made by in vitro translation of a series of constructs which featured domain-sized deletions starting from the N-terminus of the a subunit of LFA-1. Monoclonal antibodies and ICAM-1 were tested for their ability to bind to these protein fragments. Results show that the putative divalent cation binding domains V and VI contain an ICAM-1 binding site. A series of consecutive peptides covering these domains indicated two discontinuous areas as specific contact sites: residues 458-467 in domain V and residues 497-516 in domain VI. A three-dimensional model of these domains of LFA-1 was constructed based on the sequence similarity to known EF hands. The two regions critical for the interaction of LFA-1 with ICAM-1 lie adjacent to each other, the first next to the nonfunctional EF hand in domain V and the second coinciding with the potential divalent cation binding loop in domain VI. The binding of ICAM-1 with the domain V and VI region in solution was not sensitive to divalent cation chelation. In short, a critical motif for ICAM-1 binding to the a subunit of LFA-1 is shared between two regions of domains V and VI. Key words: adhesion/cation binding site/ICAM-l/integrin/ LFA-1
Introduction Lymphocyte function-associated antigen-I (LFA- 1) is a leukocyte adhesion receptor belonging to the (2 family of integrins (Springer, 1990). It is expressed on all leukocytes and, by binding to ligand intercellular adhesion molecule-I (ICAM-1), provides the major adhesive force between leukocytes and their target cells. Integrins are oa( heterodimers which require bound divalent cations in order to function (Hynes, 1992). At the N-terminus of the ae subunit are seven homologous tandemly repeated domains (I - VII) of which the last three or four are suggested to contain EFhand-type divalent cation binding sequences (Tufty and Kretsinger, 1975). The integrin divalent cation binding motif differs from classical EF hand sequences in that it lacks the
1790
an
ICAM-1 binding
crucial -z coordinating residue at position 12. Whether these sequences form structurally complete metal binding sites has been questioned, leading to the suggestion that ligands containing peptide motifs such as RGD may supply a crucial aspartate residue to complete the coordination geometry of the divalent cation (Corbi et al., 1987; Humphries, 1990). Alternatively, such a residue could be contributed from elsewhere in the integrin as in galactose binding protein (Vyas et al., 1987). Several studies have now provided evidence for the divalent cation binding capabilities of the integrins. The covalent coupling of 58Co(III) to vitronectin receptor is a direct demonstration of the binding of metal ions (Smith and Cheresh, 1991). This is further supported by the binding of Ca2+ to a protein fragment spanning the EF hand-type domains (IV -VII) of the platelet integrin gpIIbIlla (Gulino et al., 1992). Finally, a modelling study of hybrid integrin -calmodulin EF hands predicts the integrin loop to be a divalent cation chelation site (Tuckwell et al.,
1992). In general, there has been little information about the ligand binding sites on integrins with the exception of the
33 integrin, gpIIbllIa. Specifically, the binding site of an 1 lImer peptide from the fibrinogen 'y chain has been mapped to domain V of the gplbIJEla ca subunit (D'Souza et al., 1990, 1991). Similarly, cross-linking an RGD-containing peptide to domains II-VI of the vitronectin receptor oa subunit has also indicated the importance of the repeated domains for ligand recognition by this second (3 integrin (Smith and Cheresh, 1990). In addition to these repeated domains, the (2 integrins such as LFA-1 (and (31 integrins VLA-1 and VLA-2) contain a 200 residue sequence called the 'inserted' or 'I' domain (positioned between domains II and III) (Larson et al., 1989). Monoclonal antibodies (mAbs) which affect 32 integrin function have now been localized to the 'I' domain which suggests that this domain participates in the process of ligand binding and in fact might contain a ligand binding site (Diamond and Springer, 1993; Landis et al., 1993). LFA- 1 would be predicted to interact with ICAM- 1 differently from the way in which the (3 integrins interact with their ligands as human ICAM-I does not contain an RGD sequence (Simmons et al., 1988; Staunton et al., 1988) and RGD analogues apparently do not interfere with LFA- I binding to ICAM-I (Marlin and Springer, 1987). In addition, it is known that the LFA-I 'binding footprint' on ICAM- 1 is broad, covering the first and possibly part of the second domain of ICAM-1 (Staunton et al., 1990; Berendt et al., 1992). Analysis of hybrid receptors indicates that recognition of ICAM-1 resides with the LFA-1 a subunit (Johnston et al., 1990). In this study, we show that one binding site for human ICAM-1 is located within domains V and VI on the LFA-1 at subunit. -
© Oxford University Press
L.A.\- 1i . N
N
I
I
_
11±
it
,
ICAM-1 binding site in LFA-1 domains V and VI
C
I
III
Porn1
1
IVVI
V II
t;>:
I
I S.E
1'-:: -- -;;
.- 4
Ej o:
'' F1)''
,.
..
.-
..
,.
.l'
.-
.-'
.-
."
,r
."
,r
.-
,-
."
,;s,-
Al !'
1f1
t
IV-I \-\I \-. :II-
AS.E('!77
:
-
Fig. 1. LFA-l a subunit deletion series: cDNAs consist of a series of five fragments deleted sequentially from the N-terminus. The fragments are named according to the first expressed domain, namely I, II, 'I' Dom, HI and IV, and terminate at the beginning of domain VII. Fragments containing domains V - VI, V - VII and a fragment covering the rest of the a subunit to the C-terminus (C) were also made. The position of the 'I' domain is indicated as 2 and the putative divalent cation binding sequences as E.
Results In vitro expression of a nested series of domaindeleted protein fragments from the LFA- 1 a subunit and monoclonal antibody binding The aim of this study was to define the binding site for ICAM-1 on the LFA-1 a subunit. The rationale was to test the ICAM- 1 binding activity of a series of protein fragments from the LFA-1 a subunit which featured domain-sized deletions, starting from the N-terminus and ending at domain VII. In addition, two further cDNAs were made which coded for domains V -VII and the region spanning the region from the end of domain VII to the transmembrane region
(C-terminal fragment, C). This series of fragments is illustrated in Figure 1. The cloned PCR constructs were transcribed and translated in vitro; the translation bands in each case corresponded to proteins of the expected sizes as shown in Figure 2B. In order to establish whether the protein fragments resembled their domain counterparts in native LFA-1, we mapped two specific mAbs prior to testing the fragments for their ability to interact with ICAM-1. The mAbs 38 (Dransfield and Hogg, 1989) and MHM24 (Hildreth et al., 1983), specific for the LFA-1 a subunit and able to inhibit the LFA-1 -ICAM-1 interaction, were tested for their ability to bind to the protein fragments. mAbs 38 (Figure 2A) and MHM24 (data not shown) both immunoprecipitated the three protein fragments which contained the 'I' domain, namely I, II and 'I' Dom. As they did not immunoprecipitate fragments from which the 'I' domain was absent, namely III, IV and C, it was concluded that these mAbs recognized epitopes within the 'I' domain. Non-specific binding was minimal as indicated by immunoprecipitation with a control CD8 mAb 14 (and other control mAbs, data not shown). In all experiments, the translation efficiency of each fragment was checked by SDS -PAGE in order to ensure that equivalent amounts of each fragment were being translated (Figure 2B). As the fragments were recognized in a specific manner the translated protein fragments were considered to retain conformational features of native LFA- 1 oa subunit.
ICAM-1 binds to a fragment which includes domains V and VI of the LFA-1 a subunit The translated LFA-1 a subunit protein fragments were then tested for their ability to be precipitated by ICAM-1 using an ICAM-lFc construct consisting of the first three domains of ICAM-1 fused to human Fc as previously described (Berendt et al., 1992). As the Fc component was of human IgGl isotype, the IgGI myeloma protein CRI was used for control precipitations. ICAM-lFc bound reproducibly to seven of the LFA-1 a subunit fragments, namely I, II, 'I' Dom, III, IV, V-VI and V-VII but not to C fragment (Figure 3A-C). Control precipitations with protein CRI and Sepharose -protein A alone (results not shown) gave background binding. Thus the minimum sequence to which ICAM-1 specifically bound consisted of domains V and VI which indicated that a binding site was located within these
132 residues. Further localization of the ICAM- 1 binding site within domains V and VI Further characterization of the ICAM-1 binding site was continued through the use of a series of consecutive peptides corresponding to the sequence of the LFA- 1 ae subunit from domain IV to domain VII. Figure 4 illustrates the peptide boundaries which were designated in accordance with the predicted secondary structure of domains IV - VII in an attempt to preserve peptide conformation in solution (Eliopoulos et al., 1982). As well as demonstrating binding of ICAM- 1 to LFA- 1 fragments in solution, it was considered important to confirm and extend the preceding observations in a more physiological context by testing the effect of the peptides on the binding of T cell LFA- 1 to purified ICAM-1. The two peptides, dom 5 (428-467: GTQIGSYFGGELCGVDVDQDGETELLLIGAPLFYGEQRGG) and dom 6.1 (497-516: GEAITALTDINGDGLVDVAV) showed significant inhibition at 1 mg/ml. Dom 5.3 (458-477: PLFYGEQRGGRVFIYQRRQL) also significantly inhibited T cell binding to ICAM-1 at 0.5 mg/ml which was the maximum concentration at which this peptide was fully soluble (Table IA). The remaining nine peptides
1791
P.Stanley et al.
A
A 97--
C9 7
69 -
69-
46 -
w
*6
46-
3030.I
I_
l1 I
lI
T
III
''i'
li
1\-
C
B
\
B
97-
466997-
69-
46
*
-
30-
3046 -
30
21
5-
i;!
-
I
11 1'V
f
III
i
I1
1'
111
I\
t
Fig. 2. Anti-LFA-1 ca mAb 38 precipitates 'I' domain-containing fragments. (A) A series of LFA-1 protein fragments obtained by in vitro transcription and translation in the presence of [35S]methionine were precipitated with mAb 38 (+). Control precipitations were carried out with CD8 mAb 14 (-). (B) The translation efficiency of a sample of each of the fragments was checked by SDS-PAGE. Protein fragments corresponding to the expected molecular size for full-length translation product are indicated by arrowheads. Molecular weight
Fig. 3. ICAM-l precipitates LFA-l a subunit fragments containing domains V and VI. Using protein fragments as described in the legend to Figure 2, ICAM-lFc (+) precipitated fragments I, II, 'I' Dom, mI, (A) III, IV but not C-terminal fragment (C) (B) and also precipitated fragments V-VI and V-VII (C). The second of each pair of lanes shows control precipitations (-) using IgGI protein CRI, all of which are negative.
a
markers in kDa
are
indicated
on
titration curves. Titration of peptide dom 5.3 also showed a steep dose -response curve with a loss of activity by 125
(50 jtM) (data not shown). Reversed sequences of the blocking peptides, namely dom Srev, dom 6. lrev and dom 5.3rev, were also tested and 1792
IV
367
483
421
VIl
VI
V
542
604
the ordinate.
and an unrelated 39mer control had no effect on ICAM- 1 binding of T cells at 1 mg/ml (Table IA), and even when tested at 5 mg/mi (data not shown), indicating that positive blocking activity was not simply a non-specific effect of peptides in solution. Figure 5 illustrates dose-response curves for two of the inhibitory peptides, dom 5 and dom 6.1 with both peptides showing characteristically steep
jig/ml
DOMAIN LFA-la
dom dom dom dom dom dom dom dom dom dom dom 7 4.1 4.2 4.3 5.1 5.2 5.3 5/6 6.1 6.2 6/7
dom 5
Poptlde Cation Binding Motif
Blocking PopUde
Fig. 4. Designation of LFA-l subunit peptides and their location within domains IV-VII. The precise amino acid boundaries of individual peptides are given in Table I. Peptides which blocked the LFA-1-ICAM-1 interaction are indicated. a
ICAM-1 binding site in LFA-1 domains V and VI Table L. Inhibition of the binding of T cell LFA-1 to ICAM-I by LFA-I peptides Peptide
Sequence
% Control (z SD)
Molarity at 1 mg/ml (mM)
Asn367-Ser388 Thr384-Met4O5 Pro400-Ser421 Gly428-Gly467
104.3 98.5 94.0 48.8 106.3 57.5 116.8 61.2 107.6 92.0 101.8 3.4 10.2
(6.8) (16.6) (14.0) (11.1) (6.9)
0.406 0.396 0.195 0.236 0.486 0.201 0.464 0.515 0.480 0.44 0.499
38.5 128.3 57.8 114.5 55.7 107.3 4.4
(3.9) (12.5) (6.6) (23.3) (6.7) (9.6) (1.9)
(A) dom 4.1 dom 4.2 dom 4.3 dom 5 dom 5.2 dom 5.3 dom 5/6 dom 6.1 dom 6.2 dom 6/7 dom 7 mAb 38 (anti LFA-1 a) mAb 15.2 (anti ICAM-1) (B) dom S dom Srev dom 5.3 dom 5.3rev dom 6.1 dom 6.1rev mAb 38 (anti LFA-1 a)
Glu438-Ala457 Pro458-Leu477
Gly478-Gly497 Gly497-Val516 GlyS17-LeuS36 Ser537-PheSS6 Val563 -Met582
Gly428-Gly467 Gly467-Gly428 Pro458-Leu477 Leu477-Pro458
Gly497-Val516 Val516-Gly497
(15.3) (5.8) (10.8) (6.4) (14.9) (7.8) (1.78) (11.1)
The effect of each peptide on the LFA-1 -ICAM-1 interaction is presented as a percentage of the control level of T cell binding obtained in the absence of peptide (% control + SD) (A). Consecutive peptides from domains IV-VII were tested in triplicate at 1 mg/ml, with the exception of dom 4.3 and dom 5.3 which were fully soluble only at 0.5 mg/ml (n = 4). (B) blocking peptides dom 5, dom 5.3 and dom 6.1 were tested with their 'reversed' sequences at 2 mg/ml in a separate set of experiments (n = 3). Inhibition by peptides dom 5, dom 5.3 and dom 6.1 is significant at the concentrations shown: dom 5, P < 0.005; dom 5.3, P = 0.006; dom 6.1, P = 0.018 using a t-test for independent samples. Monoclonal antibodies against LFA-la (mAb 38, CD1la) and ICAM-1 (mAb 15.2, CDS4) are used as positive controls for the inhibition of T cell binding. All non-blocking peptides also failed to inhibit at 5 mg/ml.
consistently had no blocking activity at 2 mg/ml (0.5 mg/ml for dom 5.3 and dom 5.3rev) (Table IB) showing that inhibition was due to specific sequences, and not simply to a combination of residues. Therefore in a second assay system measuring LFA-1/ICAM-1 binding, domains V and VI of LFA-1 are shown to be involved in ICAM-1 recognition and the use of blocking peptides has highlighted two particular areas within these domains as contact sites. As dom 5 and dom 5.3 peptides are both active in blocking, the shared residues 458-467 must contain the 'blocking site'. Additionally, residues 497-516 in domain VI (covered by dom 6.1 peptide) also form part of the ICAM-1 binding site on the LFA-1 ox subunit.
120-
e.100-
U
8080
* control- 39mer dom 5 -X9*- dom 7 dom 6.1
a 'Z
60
.0
0-
-4
40
i_. 2
Blocking of ICAM-1 interaction with protein fragment 111 by LFA-1 peptides In order to show that both the T cell and solution based assays were measuring the same event, peptides corresponding to the two areas which blocked binding of T cell LFA-1 to ICAM-1, were tested for their ability to inhibit ICAM-1 binding of in vitro translated LFA-1 protein fragments. Precipitation of fragment III (domains III-VI) by ICAM-lFc was performed in the presence of dom 5, dom 6.1 and control peptides. Dom 5 peptide reduced the precipitation of fragment III compared with ICAM-lFc alone and the control peptide at 1 mg/ml (Figure 6). Neither dom 6.1 nor
the dom 5/6 (and dom 7 data not shown) control peptides (all at 1 mg/ml) blocked the precipitation of ICAM-lFc by fragment IH (even at 5 mg/ml). The fact that the dom 5
Concentration (mg/ml)
Fig. 5. Dose-response curves for inhibition of LFA-l-mediated T cell binding via LFA-1 to ICAM-1 by peptides dombars5, dom 6.1,thedom 7 and a 39mer control peptide (see text). Error represent standard deviation for three separate experiments where each concentration was tested in triplicate. Results are given as a percentage of binding in the absence of peptide.
peptide was able to interfere with the ICAM-1 precipitation of LFA-1 a subunit fragments in solution confirms its 6.1 blocking activity by a second method. Why peptide dom one did not block the solution based assay is uncertain; possibility is that it is less potent than dom S (as seen in the T cell assay). 1793
P.Stanley et al.
A
Precipitation
A
Precipitation
B
Control
AM-.
B
Control
_____b I
t f
.,
_ _4 I
)
.
i
L
IC0
r: o
Ir
21
r,
Ti,
r-
1
0
f.
fx
n
4
+
4±
H
H
H
,:
H
;N
ciL 0 11
OS
CrI
Fig. 6. Peptide dom 5 blocks binding of ICAM-1 to in vitro translated protein fragment III (domains HI-VI). (A) ICAM-lFc was coincubated with peptides dom 5, dom 5/6, dom 6.1 and 39mer control. Only peptide dom 5 (lane 3) interfered with ICAM-lFcmediated precipitation of the LFA-1 a fragment compared with ICAM-lFc alone (lane 1). CRI protein served as a control (lane 2). (B) Control represents an aliquot of each sample, showing equivalent translation in each reaction.
Binding of ICAM- 1 to LFA- 1 domain fragment is not divalent cation dependent The divalent cation Mg2+ is required for LFA-1 to function and as LFA-1 domains V -VII are EF hand-like in structure, they have been implicated in metal binding. We therefore investigated whether the observed ICAM-1 binding to this region was dependent upon divalent cation. The ability of ICAM-1 to bind to the protein fragment III (domains mII-VI) was assessed in the presence and absence of 2.5, 5 and 10 mM EDTA with concentrations compensating for the divalent cation present in the transcription/translation mixture. It can be seen that even the highest concentration of EDTA had no effect on ICAM-1 binding (Figure 7). Therefore the divalent cation does not appear to be required for ICAM- 1 binding to domains V and VI of the LFA-let subunit in solution. Molecular modelling of LFA- 1 subunit domains V and VI In order to have a structural representation of the regions of LFA-1 to which ICAM-l binds, domains V and VI were modelled. The homologous sequences of the 60-63 residues in LFA- 1 domains V, VI and VII have been suggested to contain EF hand motifs (Larson et al., 1989). Multiple sequence alignment indicates that each domain contains two EF hand-like motifs (Figure 8). The first half of each domain bears strong sequence similarity to the classical EF hand (helices A and B). However, the second half of each LFA-l domain shows less conservation of this motif in that the metal binding residues are only weakly conserved although the two helices (C and D) are predicted to remain. This suggests that the overall fold of each of domains V and VI resembles the EF hand-containing domains of several solved crystal a
1794
Fig. 7. Divalent cation chelation fails to interfere with ICAM-l binding to protein fragment IH (domains HII-VI). (A) EDTA (2.5, 5, 10 mM) (lane 3-5) fails to inhibit ICAM-1 binding to fragment HI compared with ICAM-lFc alone (lane 1). Control peptide CRI plus fragment Ill (lane 2). (B) Control samples of the precipitations showing equivalent translation.
structures in containing a pair of EF hands (with one potentially functional and the other non-functional in
integrins). A ribbon diagram illustrating the packing of the two domains V and VI is shown in Figure 9. The two domains were packed together in the same orientation as the domains in calmodulin with helix D running smoothly into helix A'. The two peptides which interfered with the LFA-l -ICAM-1 interaction occupy adjacent positions on the same face of the model, specifically interacting with the loop between the 'functional' and 'non-functional' EF hand in domain V and with the loop of the putative metal binding EF hand in domain VI. Although speculative, this modelling exercise provides a structural perspective to the in vitro experiments by suggesting that ICAM-l forms contacts with LFA-l along one face of domains V and VI.
Discussion Little is known about the physical interaction of the 32 integrin LFA-1 with its ligand ICAM-1. In this study we have mapped a binding site for ICAM-1 to domains V and VI of the subunit. Initially the site was localized to the domain V-VI subregion using in vitro translated protein fragments, which were shown to resemble features of native LFA-1 in that they reacted with specific monoclonal antibodies (also see Landis et al., 1993; R.C.Landis, personal communication). As a 'domain subtraction' procedure was used to localize this ICAM-1 binding site, further ICAM-1 binding sites may exist which are N-terminal to the domain V -VI site. Fine mapping of the ICAM-l binding site using short 'subdomain' peptides showed that two discontinuous regions of domains V and VI contributed to the binding site. The domain V region (by extrapolation 458-467: PLFYGa
ICAM-1
I I I
I
LFA-1 (V) (435) LFA-1 (VI) (496) LFA-1 (VIl) (556) VLA-2 (V) (462) VLA-2 (VI) (525) VLA-2 (VII) (589)
Vtt
FGGE LCGV FGEAI TALT FGRSAI-HGVK
FGSAIAALS FGRSLDGYG
GD]S
4CPV(1)(43) VKKAFA I 3CLN (1) (12) FKEAFSLF3CLN (2) (85) I R E ATF R VF-
KS GF GDGT
5TNC (1) (22) F K A AF D MF 5TNC (2) (98) L E D CF R IF SS F- H (A) M2+
GGGD
X
LLLIGAPLFYGE-QRGGRVFIYQRROQL.GFE EV S E -- L GDPGGYPLGR (495) -VAVGAPLEE---QGAVYIFNGRHGGLSPOPS QjR--IEETQVLSGI (555) E
D GE T GDG L V G D G LA KDT T MDG F N
FGSVLCSV,-
binding site in LFA-1 domains V and VI
-
OW
-VAVG,AESQMI--VLSSRPVVDMVTL,MSFSPA IPVHEVECSYSTnSNI( L GPEG.IE.E N TR VLL.VGAPMYMS--DLKKEEGRVY.LFT.ILKKGIL G QHOFL
(618) (524) -VIVG-SPLENQN-SGAVYIYNGHQGT l.RTKYS OKILGSDGAFRSHLQY (588) -VSIGAFG:QVVQLWSQSIADVA I EASF.TPEKI T L -VNK N AO IL.KLC.F (652) EDEL:KL-FLONFKADARALTDGETKTFLKAGDC S D --GIDIGKIGV,D[EFT (103) TKIELGTVMRS --- LGaNPTEAELODM I NEV DjA D ---- G N GT C F'P E F:L`(68) AAIELRHVMTN -- -LGEKLTDEE.VDEMWIREA N D -- -G (142) D E D ---- GSGTIDFE E FIL (79) TKIELG.TVMRM---LGQNPTKEElL=:DAII.EEV IE E LGEILRA-- - TGEHVTEEDI-EDLUMKDSDJ K N ND G RI F:D ElF L (155)
T
GUNGY
A[gG F
-
H (B)
*
*
*
*
*
1
3
5
9
12
DGQ
D
H (C)
EIEY,IE FV. H
1
(D)-
*
*
*
*
3
5
9
12
Fig. 8. Multiple sequence alignments of the putative divalent cation binding domains V-VII from integrins LFA-1 and VLA-2 plus the EF hand cation binding domains from the X-ray crystal structures for carp parvalbumin (4CPV 1) (Kumar et al., 1990) [names given are the Brookhaven database codes for the X-ray crystal structures used (Bernstein et al., 1977)] and the two domains of calmodulin (3CLN 1 and 3CLN 2) (Babu et al., 1988), and of turkey troponin-C (5TNC 1 and 5TNC 2) (Herzberg and James, 1988). Each domain consists of a pair of EF hands. The average secondary structure (SS) for the crystal structures is shown. Residues involved in direct binding of divalent cation are denoted by M2+ with the positions of Ca2+ coordinating residues indicated by numbered asterisks. Sequence similarities between the integrins and the crystal structures are boxed, hydrophobic residues (F, W, L, V, I, M, A, G, P, R and K) involved in tertiary interactions (shaded boxes) (residues R and K are considered to belong to the hydrophobic set if their long hydrophobic CH2 chains are involved in hydrophobic interactions with other members of the set and their hydrophilic heads point out towards the solvent) and residues that have the potential to bind M2+ (D, E, N, Q, S, T) (open boxes). Lines connecting columns of residues (generally conserved hydrophobics) indicate the stronger tertiary interactions within the model for the LFA-1 domain repeats as well as within the crystal structures. The overall topology is expected to be the same for all domains, that is a pair of EF hands per domain. The divalent cation binding potential of the second EF hand (helices C and D) is obviously diminished considerably as there is only limited sequence similarity with the cation binding residues (i.e. essentially only the second and third M2+ residues, in positions 3 and 5, show sequence similarities). The second repeat is therefore only distantly related to the first more classical EF hand motif.
EQRGG) follows immediately on from the putative EF hand in domain V, whereas the domain VI sequence (497-516: GEAITALTDINGDGLVDVAV) encompasses the divalent cation binding motif of domain VI (see bold residues). Neither the reversed sequences nor peptides from domains IV and VII, including those within the putative metal binding sequence, were inhibitory. Thus the peptide blocking data both supported and extended the initial localization of the binding site carried out with the in vitro translated a subunit protein fragments. Although relatively high concentrations of peptides were required to achieve blocking ( - 200 ItM), this is in line with recent measurements of the weak binding affinity of one other pair of adhesion partners, CD2 and CD48 (van der Merwe et al., 1993). The most immediate comparison can be made with the 03 integrin gpIIblla binding site for fibrinogen oy chain peptide which maps to a single site in domain V (D'Souza et al., 1990, 1991). The gpIlbEla ligand binding site differs from that described for LFA- 1 in that it covers the EF hand site (residues 296-306) next to the LFA-l binding site, i.e. analogous to the dom 5.2 region. However, it is of interest that a ,32 and (3 integrin utilize similar areas of the a subunit to interact with their respective ligands in spite of the very dissimilar nature of these ligands. ICAM-1 which is a member of the immunoglobulin superfamily and a cell membrane protein, shares no obvious sequence similarity with fibrinogen which is a soluble plasma protein. The sequences corresponding to the ICAM- 1 binding region in domain V are only weakly conserved in other integrins. This area has been highlighted in rat VLA- 1 as a potential ligand binding site because of its variant and extended sequence (Ignatius et al., 1990). As another method of characterizing the means by which ICAM-l interfaces with LFA-1, we have modelled domains V and VI. The model suggests for the first time that each domain contains not one but a pair of
DOM V
+
dom 5/5.3
do
DOM VlI
Fig. 9. Ribbon diagram for the model of domains V and VI of LFA-1. Plus signs indicate the positioning of the peptides that interfere with LFA-l -ICAM-1. The two regions are located on the same side of the two domain model but are interrupted by the cavity between the two domains. The diagram was produced with the aid of the display programme MOLSCRIPT (Kraulis, 1991). Helices are labelled A to D for the first repeat and A' to D' for the second. The dashed line indicates the boundary between the domain repeats. Hatched circles suggest positioning of the putative divalent cations (their exact position was not modelled).
1795
P.Stanley et al.
EF hand motifs, only one of which has retained the potential to bind divalent cation. This pairing of EF hands in which one partner is redundant is seen in other divalent cation binding molecules and is thought to maximize the binding affinity of the 'active' partner (van Eerd and Takahashi, 1976; Cook et al., 1991). In the model the two stretches of sequence representing the blocking peptides are juxtaposed along one face of domains V and VI. Although such a modelling exercise must be considered to be speculative, the derived structure does illustrate how both peptides might compose parts of a single binding site on LFA-1. In solution, the binding of ICAM-1 to this region was found not to be dependent on the presence of divalent cation. Thus bound cation may make little difference to the tertiary structure of the immediate site as shown for annexin V (P.Freemont and H.Driessen, personal communication) but have its effect on longer range interactions as has also been seen for annexin V (Lewit-Bentley et al., 1992) and other Ca2+ binding proteins (Gariepy and Hodges, 1983). It may be relevant that a theoretical modelling study with a chimeric integrin-calmodulin EF hand suggests the Ca2+ binding loops of annexin V to resemble those of integrin (Tuckwell et al., 1992). In seeming conflict with this finding is the fact that the intact LFA-1 heterodimer requires Mg2+ to bind ICAM-1 in a stable complex (Marlin and Springer, 1987; Dransfield et al., 1992). A possible resolution of these observations is that divalent cation may be required for ligand binding not to the isolated a subunit, but in order to alter the tertiary relationship between at and ,3 subunits in the intact heterodimer to facilitate ligand binding. In support of this hypothesis, the 24 epitope which is a marker of activated LFA- 1, requires Mg2 + for expression on intact LFA- 1 molecules but is exposed on isolated a subunits in the absence of Mg2+ (Dransfield and Hogg, 1989; Dransfield et al., 1990). Another possibility is that initial binding of ligand might facilitate divalent cation binding by contributing a coordinating residue such as the aspartate residue of RGDcontaining ligands (Humphries, 1990). However, as discussed in the Introduction, recent studies with the (3 integrins suggest that the sites are structurally complete in terms of divalent cation binding (Smith and Cheresh, 1991). For LFA-1, the number and identity (e.g. Mg2+ binding) of occupied sites is not known and it is possible that the domain VI (and domain V) site does not normally bind divalent cation. However, studies carried out on a mixture of ,B2 integrins suggest the potential for full occupancy of the sites at least by Ca2+ (Gahmberg et al., 1988). Recently the role of divalent cation in integrin function has become more complex with the suggestion that the 'I' domain contains a metal binding site and is involved in an undefined manner in ligand binding (Michishita et al., 1993). In summary we have defined an ICAM-l binding site on the LFA-1 a subunit with contact points within domains V and VI. More information will be required about the structure of intact LFA-l in order to understand further how this site participates in ligand binding during the process of LFA-l activation.
Materials and methods PCR amplification of fragments To construct the LFA-1 (x subunit deletion series, fragments were amplified using PCR strategy from a cDNA clone, 3R1 with domain boundaries
1796
assigned according to the designation of Larson and colleagues (Larson et al., 1989). The 5' PCR primers were designed such that they contained a BamHI restriction enzyme site, followed by a Kozak sequence to maximize translational efficiency (Kozak, 1987), an in-frame initiating methionine and a 24 bp hybridizing sequence. The 3' primer consisted of the hybridizing sequence plus a HindUI restriction site. The primers were as follows (with restriction enzyme sites given in bold type): 11-5', 5'-CGGGATCCCATGCCAGTCACCCTGAGAGGTTCCAAC-3'; 'I' Dom-5', 5'-CGGGATCCCATGGACCTGGTATTTCTGTTTGATGGT-3'; III-5', 5'-CGGGATCCCATGAAACAGGACCTGACTTCCTrCAAC-3'; IV-5', 5'-CGGGATCCCATGAATGAACCATTGACACCAGAAGTG-3'; V-5', 5 '-CGGGATCCCATGCAGGTCCAGACAATCCATGGGACC-3'; Common-3', 5'-CCCAAGCTTAATTCCTGAGAGCACTTGGGTCCC-3'. The V-VII and C fragments were amplified from a cDNA library, HPB.ALL. The primers for the C fragment PCR differed from all of the others by containing a 5' EcoRI site and a 3' Clal site. Primers for the V-VII and C fragment were as follows: V-VII-5', 5'-CGGGATCCCATGCAGGTCCAGACAATCCATGGGACC-3'; V-VII-3', 5'-CCCAAGCTTCAACCCCCGTCTTCTGGTCCGGTG-3'; C-5', 5'-CGGAATTCCCATGTTCCCAGGAGGGAGACATGAACTC-3'; C-3', 5'-CCAT-
CGATCTGCTTCTCATACACCACGTCAAC-3'. PCR amplifications were carried out using a GeneAmp DNA kit with native Taq DNA polymerase (Perkin Elmer Cetus). All reactions were performed according to the manufacturer's instructions. Thirty amplification cycles consisting of 2 min at 94°C, 2 min at 590C and 9.9 min at 70°C were used, followed by a 9.9 min chase at 70°C. All PCR fragments were cloned into the phagemid vector pBluescript II KS+ (Stratagene). Sequencing was carried out by dideoxy chain termination analysis using the Sequenase version 2.0 sequencing kit (USB). In vitro transcription and translation Transcription and translation was carried out using the TNT T7- coupled rabbit reticulocyte lysate system (Promega) according to the manufacturer's instructions. 1 Ag of cloned DNA in Bluescript KS+ phagemid was added to each reaction in a total volume of 50 1d containing 50% reticulocyte lysate. Reactions were carried out at 30°C for 90 min in the presence of [35S]methionine (Amersham). Initial experiments were performed using separate transcription (Stratagene) and translation (Promega) reactions yielding comparable results.
Immunoprecipitation and SDS - PAGE Following transcription and translation, 50 Al of 2 x EIA buffer (500 mM NaCl, 100 mM HEPES, pH 7.0) containing 0.2% NP40 was added to each 50 41 translation reaction and the mixture was divided equally. 3 ,ug of ICAM- IFc were added to one half of the reaction mixture and 3 ,Ig of CRI myeloma protein to the other half. mAbs were used either in the form of tissue culture supernatant at 150 t1 per sample or purified protein at 3 yg per sample. The total volume of the reaction was made up to 200 1.l with 1 x EIA (250 mM NaCl, 50 mM HEPES pH 7.0). Reactions were incubated at 4°C for 2 h and then added to 20 ,ul packed volume of protein A-Sepharose CL-4B (Pharmacia) which had been pre-washed with 1 x EIA. The reactions were incubated at 4°C overnight with constant agitation. The protein A-Sepharose was pelleted by brief centrifugation and 10 u1 of supernatant were removed for SDS-PAGE analysis. The remaining supernatant was discarded and the protein A -Sepharose washed three times with 1 x EIA + 0.1% NP40 and twice with 1 x EIA. 20 /Il of SDS-PAGE reducing buffer were added to the protein A beads and boiled, and samples were electrophoresed on 9% polyacrylamide gels. Gels were treated with EN3HANCE (NEN) prior to autoradiography. In the peptide blocking experiments, peptides were added to a final concentration of mg/mil. Peptide dom 6.1 was tested up to 5 mg/nil. In other experiments the divalent cation chelator EDTA was added to final concentrations of 2.5, 5 and 10 mM. Preparation and purification of ICAM- lFc and mAbs The ICAM-lFc fragment contains the first three domains of ICAM-1 fused to a human IgGI Fc tail and was prepared as previously described (Berendt et al., 1992), using serum-free medium (ADC 1, Biological Industries Ltd) to exclude the possibility of contamination with bovine immunoglobulin. The purified human IgGI myeloma protein CRI was a gift from Dr Roy Jefferis, University of Birmingham. The mAbs used in this study were LFA-l a specific mAbs 38 (Dransfield and Hogg, 1989), MHM24 (Hildreth et al., 1983), an ICAM-1 (CD54) specific mAb 15.2 and a control mAbl4 specific to CD8 (Dransfield et al., 1992). mAbs were purified by protein A chromatography (Ey et al., 1978).
ICAM-1 binding site in LFA-1 domains V and VI
Synthetic peptides Preparation and characterization of synthetic peptides. The peptides were synthesized on a model 430A Applied Biosystems Solid Phase Synthesizer on 4-hydroxymethylphenoxymethyl resin using 9-fluorenylmethyloxycarbonyl for temporary a-amino group protection (Carpino and Han, 1992). Each amino acid was coupled as a hydroxybenzatriazole active ester, automatically formed immediately prior to use. Cleavage from the resin and deprotection of the peptide were achieved with trifluoroacetic acid containing phenol, ethanediol, thioanisole and water at 20'C for 2 h. The purity and molecular weight of individual peptides were analysed by plasma desorption spectrometry and the sequences by gas phase amino acid sequencing. A peak corresponding to the expected molecular weight was found for each peptide and in all cases peptides were found to contain the correct sequence. Synthesized peptides were desalted using Sephadex G25 (Pharmacia) to remove potentially cytotoxic compounds and were stored for short periods prior to use as lyophilized powders under desiccating conditions at room temperature. mass
Peptide solubility. Amino acid analysis (Applied Biosystems Model 420H) was used to establish the solubility of all peptides which were tested at concentrations where they were known to be fully soluble. For most peptides this was at least 5 mg/ml, but for dom 6.1 it was 2 mg/ml, and for dom 4.3 and dom 5.3 it was 0.5 mg/ml. Peptide dom 5.1 was essentially insoluble. As a further precaution, all peptides were either centrifuged at 60 g or filtered using low protein binding filters (0.22 /m, Millipore) prior to use in the binding assays. Thus any non-specific inhibition by insoluble peptide was avoided. No difference in experimental results was observed between filtered and centrifuged peptides.
T cell binding to ICAM- 1 T lymphoblastoid cells were expanded from unstimulated PBMC by 1-2 weeks' culture in rIL-2 (20 ng/ml; Cletus) with details as previously described (Dransfield et al., 1992). Purified ICAM-lFc protein (40 l1 of a concentration of 20 ,tg/ml in PBS-A) was added to each well of flat-bottomed 96-well plates (Immulon 1, Dynatech) before overnight incubation at 4°C. Prior to the assay, the plates were saturated with 2.5% BSA in PBS-A (lacking in Ca2+ and Mg2+) (100 1l/well) by incubation for 2 h at room temperature and finally washed four times with PBS-A and once in HEPES buffer (20 mM HEPES, 140 mM NaCl, 2 mg/ml glucose, pH 7.4). Detection of T cell binding to ICAM-lFc was carried out using the technique described previously (Cabanias and Hogg, 1991). Briefly, 5 x 107 cultured T cells were labelled with 200 ACi 5ICrO42-, for 90 min and resuspended in HEPES buffer containing 4 mM MgCl2, 4 mM EGTA and 100 nM PdBu (2 x final concentration). 50 ul of T cells plus 50 1l of peptide or mAb at designated concentrations were added to each well of an ICAM-lFc-coated 96-well plate. Plates were incubated on ice for 20 min, centrifuged at 30 g for 1 min and then incubated for a further 35 min at 37°C. T cells which remained bound after five washes in warmed RPMI were then lysed in 1% Triton X-100 in water and the incorporated radioactivity measured using a Betaplate counter (LKB Instruments). Monoclonal antibodies against LFA-1 ax subunit (mAb 38) and ICAM-1 (mAb 15.2) were included in assays as controls for assessing
LFA-1/ICAM-1-mediated adhesion. Cell viability assays
peptides used in these experiments were tested for cytotoxic activity using an MTT assay (Plumb et al., 1989) with modification (Ross et al., 1992). Briefly T cells at 50 Id per well, were plated out in 96-well flatbottomed plates at a concentration of 5 x 105 cells/ml in RPMI plus 10% FCS. 50 1l of peptide in RPMI were added to the T cells to give a final peptide concentration of 5 mg/ml (the highest peptide concentration used in the assays). 20 11 of MTT (at 5 mg/mi in RPMI) were then added to each well and the plate incubated at 37°C for 2 h (note: T cell assays were for 55 min only). The plate was then centrifuged at 30 g for 1 min and All
the MTT and medium were carefully removed. 200 yl of dimethyl sulfoxide
plus 25 1d of Sorensen's glycine buffer were then added to each well and the optical density (OD) was read at 570 nm (Titertek Multiskan, Flow Laboratories). A titration of cell concentration was included to establish the expected OD reading under non-toxic conditions. None of the peptides used gave OD readings lower than the expected value for the equivalent concentration of cells in the absence of peptide. Procedures for molecular modelling Conserved features of the metal binding domains V-VII of LFA-lIa were identified by comparison with the equivalent domains of a second integrin, VLA-2 (Takada and Hemler, 1989) (Figure 8). The alignment shows that
the first half of each domain bears strong sequence similarity to an EF hand (helices A and B) in that four of the five divalent cation and all hydrophobic residues of the first amphipathic helix are conserved. However, the second half of each LFA-1 domain shows less conservation of this motif in that the metal binding residues are only weakly conserved although the two helices (C and D) are predicted to remain. These sequences were then compared with the EF hand-containing domains of three calcium binding proteins for which X-ray crystal structures are known. Each X-ray structure has one pair of Ca2+ binding EF hands per domain, a feature which appears to maximize Ca2+ binding affinity over individual EF hands (Sekharudu and Sundaralingam, 1988). Such EF hand pairing appeared also to be a feature of the integrins as key tertiary interactions are predicted to be conserved between the hydrophobic residues of helices A and B of the first half of the domain with predicted helices C and D of the second half of the domain. Each domain was modelled from the X-ray crystal coordinates of carp parvalbumin (Kumar et al., 1990) using the technique of replacing nonregular regions of the template, usually the loop regions, by database fragment searches (Bates and Stemnberg, 1992). Helices B and B' are bent due to the proline residue midway along each helix. Although proline residues are rarely found in this position (MacArthur and Thornton, 1991) they do occur (Dempsey et al., 1991; Dekker et al., 1993). A similar distortion not involving proline is seen in the B helix of parvalbumin which has been attributed to the requirements of the EF hand to readjust following the dynamic binding of calcium (Kumar et al., 1990). It is possible that the proline residue in the B helix of some of the integrin repeats serves the same purpose. The domains are linked together in the same way as the domains in calmodulin (Babu et al., 1988) except that the long helix between each globular domain has been substantially truncated such that the last helix of each repeating domain is made to run directly into the first helix of the next. This packing arrangement is the most logical in terms of maintaining a self-avoiding protein chain and an absence of 'linker residues' between domains, i.e. the residues of helix D/A' must contribute to the tertiary contacts within each of the two domains. The complete model was energyminimized using the CHARMM program (Brooks et al., 1983) with default values as incorporated into QUANTA in the POLYGEN software. Several points validate the model presented here. There is substantial sequence similarity within the first half of each domain to a classical EF hand and EF hands usually occur in pairs. There are no steric clashes in the model and conserved hydrophobic residues pack well within and between each domain. There are no unbalanced charges that lack the potential to be solvated. Finally, the model, which was independently derived, allows a simple interpretation of the peptide analysis.
Acknowledgements We are extremely grateful to Nicola O'Reilly for the synthesis of the peptides used in this study. We thank Dr David Simmons (Oxford) for the HPB.ALL cDNA library and Dr T.S.Springer (Boston) for LFA-1 ax subunit construct 3R1; Drs Alister Craig and Tony Berendt (Oxford) for the ICAM-lFc construct; Alison McDowall and Dr Paul Hessian for help with the T cell assays and the preparation and purification of the ICAM-lFc; Drs Alex Law and A.McMichael (Oxford) for mAb MHM24, Dr Roy Jefferis (Birmiingham) for myeloma protein CRI and Dr Fiona Watt for control 39mer peptide; Dr Mike Fried for discussion about in vitro transcription/translation systems; Dr M.J.E.Stemberg for discussion about the modelling procedures. We thank our colleagues Paul Freemont, Paul Hessian, Clive Landis, Mike Stemnberg and Mairi Stewart for their many helpful comments on the manuscript and Louise Dewhurst for her assistance with its preparation.
References Babu,Y.S., Bugg,C.E. and Cook,W.J. (1988) J. Mol. Biol., 204, 91-204. Bates,P.A. and Sternberg,M.J.E. (1992) In Rees,A.R., Sternberg,M.J.E. and Wetzel,R. (eds), Protein Engineering. A Practical Approach. IRL Press, Oxford, pp. 17-141. Berendt,A.R., McDowall,A., Craig,A.G., Bates,P.A., Sternberg,M.J.E., Marsh,K., Newbold,C.I. and Hogg,N. (1992) Cell, 68, 71-81. Bernstein,F.C., Koetzle,T.F., Williams,G.J.B., Meyer,E.F., Brice,M.D., Rodgers,J.R., Kennard,O., Shimanouchi,T. and Tasumi,M. (1977) J. Mol. Biol., 112, 535-542. Brooks,B.R., Bruccoleri,R.E., Olafson,B.D., States,D.J., Swaninathan,S. and Karplus,M. (1983) J. Comp. Chem., 4, 187-217. Cabafnas,C. and Hogg,N. (1991) FEBS Lett., 292, 284-288. Carpino,L.A. and Han,G.Y. (1992) J. Org. Chem., 22, 3404-3409.
1797
P.Stanley et al. Cook,W.J., Ealick,S.E., Babu,Y.S., Cox,J.A. and Vijay-Kumar,S. (1991) J. Biol. Chem., 266, 652-656. Corbi,A.L., Miller,L.J., O'Connor,K., Larson,R.S. and Springer,T.A. (1987) EMBO J., 6, 4023 -4028. Dekker,N., Cox,M., Boelens,R., Verrijzer,C.P., van der Vliet,P.C. and Kaptein,R. (1993) Nature, 362, 852-853. Dempsey,C.E., Bazzo,R., Harvey,T.S., Syperek,I., Boheim,G. and Campbell,I.D. (1991) FEBSLett., 281, 240-244. Diamond,M.S. and Springer,T.A. (1993) J. Cell Biol., 120, 545-556. Dransfield,I. and Hogg,N. (1989) EMBO J., 8, 3759-3765. Dransfield,I., Buckle,A.-M. and Hogg,N. (1990) Immunol. Rev., 114, 29-44. Dransfield,I., Cabauias,C., Craig,A. and Hogg,N. (1992) J. Cell Biol., 116, 219-226. D'Souza,S.E., Ginsberg,M.H., Burke,T.A. and Plow,E.F. (1990) J. Bio. Chem., 265, 3440-3446. D'Souza,S.E., Ginsberg,M.H., Matsueda,G.R. and Plow,E.F. (1991) Nature, 350, 66-68. Eliopoulos,E., Geddes,A.J., Brett,M., Pappin,D.J.C. and Findlay,J.B.C. (1982) Ins. J. Bio. Macromol., 4, 263 -270. Ey,P.L., Prowse,S.J. and Jenkin,C.R. (1978) Immunochemistry, 15, 429-436. Gahmberg,C.G., Kantor,C., Nortamo,P., Kotovuori,P., Prieto,J. and Patarroyo,M. (1988) In Lernmark,A., Dyrberg,T., Terenius,L. and Hokfelt,B. (eds), Molecular Mimicry in Health and Disease. Elsevier, Amsterdam, pp. 105-121. Garidpy,J. and Hodges,R.S. (1983) FEBS Lett., 160, 1-6. Gulino,D., Boudignon,C., Zhang,L., Concord,E., Rabiet,M.-J. and Marguerie,G. (1992) J. Biol. Chem., 267, 1001-1007. Herzberg,O. and James,M.N.G. (1988) J. Mol. Biol., 203, 761-779. Hildreth,J.E.K., Gotch,F.M., Hildreth,P.D.K. and McMichael,A.J. (1983) Eur. J. Immunol., 13, 202 -208. Humphries,M.J. (1990) J. Cell Sci., 97, 585-592. Hynes,R.O. (1992) Cell, 69, 11-25. Ignatius,M.J., Large,T.H., Houde,M., Tawil,J.W., Barton,A., Esch,F., Carbonetto,S. and Reichardt,L.F. (1990) J. Cell Biol., 111, 709-720. Johnston,S.C., Dustin,M.L., Hibbs,M.L. and Springer,T.A. (1990) J. Immunol., 145, 1181-1187. Kozak,M. (1987) Nucleic Acids Res., 15, 8125-8132. Kraulis,P.J. (1991) J. Appl. Crystallogr., 24, 946-950. Kumar,V.D., Lee,L. and Edwards,F.P. (1990) Biochemistry, 29,
1404-1412. Landis,R.C., Bennett,R.I. and Hogg,N. (1993) J. Cell Biol., 120, 1519-1527. Larson,R.S., Corbi,A.L., Berman,L. and Springer,T. (1989) J. Cell Biol., 108, 703-712. Lewit-Bendley,A., Morera,S., Huber,R. and Bodo,G. (1992) Eur. J. Biochem., 210, 73-77. MacArthur,M.W. and Thornton,J.M. (1991) J. Mol. Biol., 218, 397-412. Marlin,S.D. and Springer,T.A. (1987) Cell, 51, 813-819. Michishita,M., Videm,V. and Arnaout,M.A. (1993) Cell, 72, 857-867. Plumb,J.A., Milroy,R. and Kaye,S.B. (1989) Cancer Res., 49,4435-4440. Ross,L., Hassman,T. and Molony,L. (1992) J. Biol. Chem., 267, 8537-8543. Sekharudu,Y.C. and Sundaralingam,M. (1988) Protein Engng, 2, 139-146. Simmons,D., Makgoba,M.W. and Seed,B. (1988) Nature, 331, 624-627. Smith,J.W. and Cheresh,D.A. (1990) J. Biol. Chem., 265, 2168-2172. Smith,J.W. and Cheresh,D.A. (1991)J. Biol. Chem., 266, 11429-11432. Springer,T. (1990) Nature, 346, 425-434 Staunton,D.E., Marlin,S.D., Stratowa,C., Dustin,M.L. and Springer,T.A. (1988) Cell, 52, 925-933. Staunton,D.E., Dustin,M.L., Erickson,H.P. and Springer,T.A. (1990) Cell, 61, 243-254. Takada,Y. and Hemler,M.E. (1989) J. Cell Biol., 109, 397-407. Tuckwell,D.S., Brass,A. and Humphries,M.J. (1992) Biochem. J., 285, 325 -331. Tufty,R.M. and Kretsinger,R.H. (1975) Science, 187, 167-169. van der Merwe,P.A., Brown,M.H., Davis,S.J. and Barclay,A.N. (1993) EMBO J., 12, 4945 -4954. van Eerd,J.-P. and Takahashi,K. (1976) Biochemistry, 15, 1171-1180. Vyas,N.K., Vyas,M.N. and Quiocho,F.A. (1987) Nature, 327, 635-638. Received on July 28, 1993; revised on February 1, 1994
1798
