Entropically Driven MHC Class I Recognition by Human Inhibitory Receptor Leukocyte Ig-like Receptor...

doi:10.1016/j.jmb.2005.10.057 J. Mol. Biol. (2006) 355, 237–248

Entropically Driven MHC Class I Recognition by HumanInhibitory Receptor Leukocyte Ig-like Receptor B1(LILRB1/ILT2/CD85j)

Mitsunori Shiroishi1†, Kimiko Kuroki1†, Kouhei Tsumoto2

Akiko Yokota2,3, Takashi Sasaki2, Kimie Amano3, Tsukasa Shimojima3

Yasuo Shirakihara3, Linda Rasubala1, P. Anton van der Merwe4

Izumi Kumagai2, Daisuke Kohda1 and Katsumi Maenaka1*

1Division of Structural BiologyMedical Institute ofBioregulationKyushu University3-1-1 Maidashi, Higashi-kuFukuoka 812-8582, Japan

2Department of BiomolecularEngineering, Graduate School ofEngineering, Tohoku UniversityAoba-yama 07, Sendai 980-8579Japan

3Structural Biology CenterNational Institute of Genetics1111 Yata, Mishima, Shizuoka411-8540, Japan

4Sir William Dunn School ofPathology, University of OxfordOxford OX1 3RE, UK

0022-2836/$ - see front matter q 2005 E

† M.S. & K.K. contributed equallyAbbreviations used: MHCI, majo

complex class I; LILR, leukocyte imreceptor; SPR, surface plasmon resoisothermal titration calorimetry; ILTKIR, killer cell Ig-like receptor; TCRHSQC, heteronuclear single quantu

E-mail address of the [email protected]

The human inhibitory receptor, leukocyte immunoglobulin (Ig)-like receptorB1 (also called Ig-like transcript (ILT) 2, CD85j), is broadly expressed onleukocytes. LILRB1 binds to a wide range of major histocompatibilitycomplex class I molecules (MHCIs) and transduces negative signals that can,for example, prevent killing of MHCI-expressing cells. Here we report thekinetic, thermodynamic, NMR and crystallographic analyses of MHCIrecognition by LILRB1. Kinetic studies demonstrated that LILRB1 binds toMHCIs with fast association and dissociation rates, typical of cell–cellrecognition receptors. Thermodynamic analyses showed that LILRB1–MHCIinteractions are entropically driven (KTDSZK9.4wK6.6 kcal molK1) withlow heat capacity changes (DCpZK0.22wK0.10 kcal molK1 KK1). Thecrystal structures of LILRB1 in the different crystal forms exhibited variationin the elbow angle between the two N-terminal Ig-like domains, indicatinginterdomain flexibility. Consistently, NMR analysis provided the directevidence of the conformational changes of LILRB1 upon the MHCI binding.These findings suggest that LILRB1–MHCI interactions, while involvingsome conformational adjustment, are not accompanied by a very largereduction in conformational flexibility at the binding interface. This mode ofbinding is distinct from “induced-fit” binding, which is associated with largereductions in conformational flexibility, and would be suitable for rapidengagement of MHCIs to enable fast monitoring of the expression level ofMHCIs on target cells.

q 2005 Elsevier Ltd. All rights reserved.

Keywords: Leukocyte lg-like receptors; MHC class I molecules; surfaceplasmon resonance; isothermal titration calorimetry; protein–protein inter-actions
*Corresponding author
Introduction

Leukocyte immunoglobulin (Ig)-like receptors(LILRs),1 also known as Ig-like transcripts (ILTs),

lsevier Ltd. All rights reserve

to this work.r histocompatibilitymunoglobulin-likenance; ITC,, Ig-like transcript;, T cell receptor;m coherence.ing author:

or CD85, are a family of immunoreceptors whosegenes are located at the human leukocyte receptorcomplex (chromosome 19q13.4). This locus alsoencodes other immune-related receptors, such askiller cell Ig-like receptors (KIRs), leukocyte-associated Ig-like receptors (LAIRs), NKp46, andthe Fca receptor.1 LILRB1 is broadly expressed; it isfound on monocytes, B cells, dendritic cells, andsubsets of natural killer (NK) and T cells.2–8 LILRB1has four Ig-like extracellular domains and immuno-receptor tyrosine-based inhibitory motifs (ITIMs) inits cytoplasmic tail. ITIMs recruit the proteintyrosine phosphatase SHP-1, which inhibits earlysignaling events triggered by stimulatory receptors.

d.

238 Binding Properties of LILRB1–MHC Interaction

LILRB1 binds to MHCIs, which are widelyexpressed on normal cells, thereby preventingLILRB1-expressing cells from killing normal cells.Conversely, abnormal cells that lack MHCI are notprotected. Indeed the engagement of LILRB1 on NKand T cells has been shown to inhibit the activationsignaling events such as actin reorganization.9

Our previous study10 and a report by Chapmanet al.11 showed that LILRB1 binds to several MHCImolecules (HLA-A,-B,-C and -G) with dissociationconstants (Kd) ranging from 2 mM to 9 mM. Interest-ingly, LILRB1 bound to non-classical MHCI, HLA-G,with a three to fourfold higher affinity than toclassical MHCIs. The mutational analyses of theLILRB1 ectodomain11,12 and the recent report of thecrystal structure of LILRB1 complexed with HLA-A213

showed that the binding site of MHCI includes thea3 domain of the heavy chain and b2-microglobulin(b2m). This is consistent with our previous bio-chemical analysis demonstrating that LILRBscompete with CD8 for MHC binding and thatLILRB1–MHCI interactions are b2m-dependent10

(M.S. et al. unpublished results). The a3 domain isrelatively more conserved among the MHCIs thanthe polymorphic a1-a2 peptide-binding region andb2m is common to all MHCIs. This explains thebroad binding specificity of LILRB1. In contrast, theKIR binding site on MHCs resides at the poly-morphic C-terminal side of the binding peptide inthe a1-a2 region11,14,15 resulting in the allele-specificrecognition by KIRs.

While the LILRB–MHCI interactions have onlybeen subjected to preliminary studies,10,11 now thatthe crystal structure of the LILRB1–HLA-A2 complexhas been solved,13 a more detailed analysis of thethermodynamic and kinetic properties of theseinteractions would greatly enhance our understand-ing of the LILRB–MHC interactions. Here we havestudied the kinetic and thermodynamic properties ofthe interactions between the ligand-binding domainsof LILRB1 and MHCIs using surface plasmonresonance (SPR) and isothermal titration calorimetry(ITC). Furthermore, we performed NMR analysis forthe MHCI binding of LILRB1 and also determined thecrystal structure of LILRB1 in the P21212 crystal form,which is different from that of the previous structuralreport.12 These binding and structural analyses revealhow the LILRB–MHC interactions compare withother cell–cell recognition receptors, and provideinsight into self/none-self discrimination by leuko-cytes.

Results

Thermodynamic analysis of LILRB1–MHCIinteractions

For SPR analysis, the N-terminal 2 domains(D1D2) of LILRB1 (hereafter, designated asLILRB1D1D2) were injected at a flow rate of10 ml/min over the flow cells onto which MHCIs(from 1000 RU to 3000 RU) had been immobilized.

The dissociation constants (Kds) were independentof the amount of immobilized MHCIs (similar Kdswere obtained in the reverse orientation (M.S. et al.,unpublished results)). Thermodynamic data wereobtained by van’t Hoff analysis as follows. The Kd

was determined at different temperatures byequilibrium analysis (Figure 1(a)) and convertedinto the standard state Gibbs energy change (DG).Then the DGs were plotted against temperatures,and fitted with the non-linear van’t Hoff equation(Figure 1(b)). Binding enthalpy (DH), entropy (DS)and heat capacity (DCp) were obtained simul-taneously from this fit. The thermodynamic para-meters of LILRB1–MHCI interactions aresummarized in Table 1. The interactions betweenLILRB1 and MHCIs had binding enthalpies thatwere close to zero or slightly unfavorable at 25 8C(DHZK0.2 to 1.9 kcal molK1), indicating that theseinteractions were entropically driven at 25 8C(Figure 1(b) and Table 1). The heat capacitiesderived from these fittings are small for protein–protein interactions,16 ranging from K0.10 kcalmolK1 KK1 to K0.22 kcal molK1 KK1.

Thermodynamic analysis was also performedusing ITC. Since ITC is technically difficult for theinteraction with low affinity and low endo- orexothermic value, the LILRB1–HLA-G interaction,which showed the highest affinity and enthalpyamong all tested LILRB–MHCs, was analysed.Soluble LILRB1D1D2 (240 mM) was injected intothe thermally shielded sample cell filled with HLA-G1(19.5 mM) solution. The ITC measures the heatabsorbed/generated by the interaction and thusmeasures the binding enthalpy directly. As shownin Figure 1(c) and Table 1, the positive enthalpy(2.4 kcal molK1) was observed for LILRB1–HLA-G1interaction. The resultant binding parameterswere, KdZ1.7 mM, TDSZ9.4 kcal molK1, stoichio-metry (N)Z1.0 at 25 8C. The binding enthalpy isslightly larger than but in reasonable agreementwith the enthalpy change determined by van’tHoff analysis (1.9 kcal molK1). Furthermore, theheat capacity from ITC analysis was also deter-mined to be K0.22 kcal molK1 KK1 by themeasurement of DH at each temperature (15, 20,25, 30, 35 and 40 8C), which is in good agreementwith the value from SPR analysis.

Kinetic analyses of LILRB1–MHCI interactionsby SPR

LILRB1D1D2 was injected into flow cells onwhich MHCI molecules were immobilized. Lowerimmobilization levels (800–1000 RU) and a higherflow rate (50 ml/min) were used than for equili-brium analyses to minimize mass transport effects.Data were collected at the maximum resolutionpossible (10 Hz) to facilitate measurement of veryfast kinetics. As a negative control, biotinylatedOX48 antibody or bovine serum albumin (BSA) wasimmobilized on one flow cell at similar level toMHCIs. The response derived from negativecontrol was subtracted from each response derived

Figure 1. Thermodynamic ana-lyses of LILRB1–MHCI inter-actions. (a) Equilibrium bindinganalysis of LILRB1 to immo-bilized HLA-G1. LILRB1D1D2was injected at the indicated con-centrations through flow cells withHLA-G1 (continuous line) andcontrol (broken line). The sameamount of biotinylated BSA wasimmobilized on a control flowcell. (b) The representative plotsof temperature dependency ofbinding Gibbs energy (DG) deter-mined by SPR experiments. Gibbsenergies at five temperature points(10 8C, 15 8C, 20 8C, 25 8C and30 8C) were obtained from Kd

values from equilibrium bindinganalyses. LILRB1D1D2 wasinjected over the immobilizedHLA-B35 (filled square),HLA-Cw4 (filled circle) and HLA-G1 (filled up triangle). Each plot isfitted with non-linear van’t Hoffequation. Thermodynamic para-meters are summarized in Table 1.(c) Titration calorimetry of theinteraction between theLILRB1D1D2 and HLA-G. Upper,typical calorimetric titration ofLILRB1D1D2 (240 mM) to HLA-G(19.5 mM) at 25 8C. Lower, the fit ofthe model (black line) to theintegrated heats from each injec-tion (black squares). (d) Tempera-

ture dependence of DH of the interaction between LILRB1D1D2 and HLA-G. DH at each temperature (black squares) andthe linear fit (black lines). Thermodynamic parameters are summarized in Table 1.

Table 1. Thermodynamic parameters of the interaction between LILRB1 and MHCIs at 25 8C

(Data from SPR analysis)

naDG

(kcal molK1)DH

(kcal molK1)KTDS

(kcal molK1)DCp

(kcal molK1 KK1)Analyte Immobilized

LILRB1D1D2 HLA-B35b 2 K6.6G0.0 0.6G0.2 K7.2G0.2 K0.10G0.02HLA-Cw4c 3 K6.8G0.2 K0.2G0.1 K6.6G0.1 K0.16G0.06HLA-G1c 4 K7.5G0.3 1.9G0.7 K9.4G0.8 K0.22G0.06

(Data from ITC analysis)LILRB1D1D2 versus HLA-G1b 2 K7.0G0.03 2.4G0.06 K9.4G0.03 K0.22

ReferencesFcgRIIa, IIb hFc1 40 K7.9 to K8.3 K4.4 to K6.4 K1.9 to K3.3 K0.22 to K0.43FcgRIII hFc1 40 K8.0 K15.4 7.4 K0.7KIR2DL3 HLA-Cw7 41 K7.2 K4.1 K3.1 K0.1CD22 CD45 42 K5.1 K10.1 5.0 K0.08NKG2D RaeI 28 K8.6 K5.2 K3.4NKG2D H60 28 K10.5 K23.6 13.1TCR MHC 19,22,36–39 K7.1G0.6 K14.6G5.4 7.1G5.7 K0.62G0.37

For the data from SPR analysis, DG, DH, KTDS, DCp represent binding changes in standard state Gibbs energy, enthalpy, entropy andheat capacity, respectively, calculated with non-linear van’t Hoff equation.

a n, numbers of experiments.b The values are meansGrange derived from two experiments.c The values are meansGstandard deviation derived from multiple experiments (n experiments).

Binding Properties of LILRB1–MHC Interaction 239


from the binding to MHCIs. The global fittinganalysis was performed by simultaneously fittingmono-exponential rate equations (equations (4) and(5) in Materials and Methods) to both the associ-ation and dissociation phases from injections ofseveral concentrations of analyte. Figure 2(a) showsthat the kinetics of the LILRB1–MHCI interactionfits reasonably well to the simple 1:1 Langmuirbinding model, consistent with structural data andprevious binding studies.10–13 There is, however,some small systematic variation in the residuals(Figure 2(a), lower plot). This is commonly seen inSPR experiments17,18 and may be the result ofexperimental error (e.g. mass transport effects) or asmall degree of heterogeneity in the analyte orligand. Since this deviation is small it is unlikely tohave a significant effect on the measured para-meters. The kinetic parameters for the LILRB1–MHCI interactions are summarized in Table 2. Thedissociation constants (Kd) calculated from the rateconstants agreed well with the Kds previouslydetermined by equilibrium binding analyses.10

LILRB1–MHCI interactions have typical associationrate constants (kon; 5.0!105–9.2!105 MK1 sK1) andfast dissociation rate constants (koff; 2.1–5.0 sK1). koff

values obtained separately by fitting only dis-sociation phases were similar to those determinedfrom the global fitting (data not shown).

Activation energies of LILRB1–MHCIinteractions

Binding rate constants generally increase withtemperature. The extent of this increase is ameasure of the amount of thermal energy requiredto cross an energy barrier preventing the associationor dissociation, and is referred to as the activationenergy of association or dissociation (Ea

on or Eaoff).19

In order to obtain the activation energies forLILRB1–MHCI interactions, the kinetic analyseswere performed at a range of temperatures (10, 15,20, 25 and 30 8C). As described above, in eachexperiment, we confirmed that the Kds derivedfrom kinetic analyses were similar to values fromequilibrium analyses at 25 8C. Some of the kons ateach temperature were obtained from the globalfitting, but others were calculated by dividing Kds(from equilibrium analysis) by koffs (from dis-sociation phase fitting only) because at highertemperatures (30 8C) equilibrium was reached tooquickly to allow accurate determination of the konsby analysis of the association phase. The activationenergies of association or dissociation (Eon

a or Eoffa )

were obtained from Arrehnius plots (Figure 2(c)).The small temperature dependency of LILRB1–MHCI association and dissociation resulted in lowEon

a and Eoffa values (Figure 2(b) and Table 3).

Figure 2. Kinetic analysis ofLILRB1 bindings to MHCI. (a)The global fitting kinetic analysesof LILRB1 against immobilizedHLA-G1 are shown as examples.Flow rate was 50 ml/min in thekinetic experiments. Upper,LILRB1D1D2 was injected (filledbar) at the indicated con-centrations over the HLA-G1(w800 RU). Lower, Residual errorsfrom the fits are shown. Kineticparameters are summarized inTable 2. (b) Temperature depen-dencies of dissociation rate.LILRB1D1D2 was injected on theimmobilized HLA-G1. (c) Tem-perature dependencies of asso-ciation (left panel) and dissociation(right panel) rate constants. Arrhe-nius plots of the natural logarithmon and off rates for the LILRB1bindings derived over a range oftemperatures (10 8C–30 8C). Filledsquare, HLA-B35; filled circle,HLA-Cw4; filled up triangle,HLA-G1. The thermodynamicparameters of association and dis-sociation phase are summarized inTable 3.

Table 2. Kinetic parameters of the interactions between LILRB1 and MHCIs at 25 8C

Analyte Immobilized nakon

(!105 MK1sK1) koff (sK1) Kd,kin (mM) Kd,eq (mM)c

LILRB1D1D2 HLA-B35 1 5.0 3.7 7.4 8.8G0.2HLA-Cw4b 2 6.3G0.9 5.0G0.2 8.1G0.9 6.5G0.5HLA-G1d 4 9.2G1.3 2.1G0.4 2.4G0.8 2.0G0.7

Other protein–protein interactionskon

(!105 MK1 sK1) koff (sK1) Kd (mM) References

LILRB1D1D2 UL18 1.4 0.0028 0.0021 11KIR2DL3 HLA-Cw7/DS11 2.1 1.1 5.2 41CD8aa MHC class I R1.0 R18 w 200 43CD22 CD45 R1.5 R18 117 42CD80 CTLA-4 9.4 0.43 0.46 44CD80 CD28 6.6 1.6 2.4 44FcgRIIa,IIb,III hFc1 3.8–4.4 0.31–0.69 0.72–1.9 40TCR Peptide-MHC 0.009–0.2 0.01–0.1 1–90 19,45

a n, numbers of experiments.b The values are meansG range derived from two experiments. Association kinetic constant (kon) and dissociation kinetic constant

(koff) were fitted simultaneously to the response data at several concentrations of analyte. Kd,kins were obtained from global fitting. Kd,eqswere obtained from equilibrium analyses.10

c The values are meansGstandard deviation derived from multiple experiments. Immobilization level of each MHCI molecule to thesensor chip was relatively low (w1000 RU for HLA-B35, w800 RU for HLA-Cw4 and w800 RU for HLA-G1) to prevent the masstransfer effect.

d The values are meansGstandard deviation derived from multiple experiments (n experiments).

Table 3. Activation energies in association and dissociation phase at 25 8C

Activation energies

Analyte Immobilized Eona (kcal molK1) Eoff

a (kcal molK1) DH (kcal molK1)

LILRB1D1D2 HLA-B35 (nZ1) 7.1 5.6 1.5 (0.6)HLA-Cw4 (nZ2)a 4.0G0.3 3.3G0.5 0.7 (K0.2)HLA-G1 (nZ4)† 5.4G0.5 2.2G0.2 3.2 (1.9)

KIR2DL3b HLA-C7-DS11 w9 w13 wK4NKG2Dc MICA w6 w8 wK2TCR (JM22z)d HLA-A2-flu w11 w31 wK20

Values obtained for DH by equilibrium analysis (from Table 1) are shown in parenthesis. n, numbers of experiments. †The values aremeansGstandard deviation derived from multiple experiments (n experiments).

a The values are meansGrange derived from two experiments.b Maenaka et al.41

c McFarland & Strong.30

d Willcox et al.19

Table 4. Data collection and refinement statistics

Data collectionResolution range (A) 37.3–3.0Number of unique reflections 3911 (411)a

Completeness (%) 88.8 (99.8)a

Data redundancy 12.2 (11.9)a

Rmergeb (I) 0.126 (0.344)a

I/sigma (I) 6.7 (1.8)c

RefinementResolution range (A) 37.3–3.0Number of reflections (working set) 3868Number of reflections (test set) 365Number of protein atoms 1356Rcryst

c 0.233 (0.277)d

Rfreec 0.306 (0.386)d

Rmsd of bond lengths (A) 0.012Rmsd of bond angles (deg.) 1.4Average B value (A2) 39.3

a Values for highest resolution shell (3.16–3.0 A).b Rmerge ZSjIKhIij=ShIi, where I is the observed intensity, hIi is

the average measurement values.c Rcryst ZSjjFobsjKjFcalcjj=SjFobsj, where Fobs and Fcalc are the

observed and calculated structure factor amplitudes. Rfree iscalculated from the test set reflections.

d Values for highest resolution shell (3.08–3.0 A).


Crystal structure of LILRB1D1D2 from the P21212crystal

The structure of LILRB1D1D2 from the P21212crystal at 3.0 A resolution was determined bymolecular replacement with the initial model ofLILRB1D1D2 structure from the P41212 crystal.Using the program AMORE,20 the rotational andtranslational search produced the clear solutiononly when the individual Ig-like domain was usedas the search model, because the interdomainangles differed in the two crystal forms (seebelow). It was refined with the 37.3–3.0 A reflectionsusing CNS, and the Rfree and Rwork of the finalmodel were 30.6% and 23.2%, respectively (finalrefinement statistics are summarized in Table 4).The model was well fitted to the electron densitymap (Figure 3(a)). The structure includes the wholeD1D2 domains except for some N and C-terminalresidues (residues 1 and 196–198), some loopregions (29–31, 137–143, 162–163, 165–169) andone part containing helix and sheet (148–153)

Figure 3. Overall structure ofLILRB1D1D2 in the P21212 crystal.(a) Stereo view of the 2Fo–Fc maparound the position 35 (tyrosine) ofLILRB1D1D2. The 2Fo–Fc mapswere contoured at 1.0 s. Carbonatoms are in orange, oxygen atomsin red and nitrogen in blue. (b)Ribbon model of overall structureof LILRB1D1D2 in P21212 crys-tals. The orientation of theLILRB1D1D2 structure is almostthe same as those of (c). Sixtryptophan residues are shownin stick model. Residues 46and 67 located in or close to theMHC class I binding site areshown in cyan. (c) Comparison offree and MHCI-bound states ofLILRB1D1D2 and KIR2DL1; struc-tural flexibility of domain-domainangles. The D1 domain ofLILRB1D1D2 structure deter-mined here from the P21212 crystalis superimposed on that of thepreviously determined P41212structure and that of the HLA-A2complex structure. On the otherhand, the D2 structure of freeKIR2DL1 is superimposed on thatof the HLA-Cw4 complex struc-ture. The two complexes are alsosuperimposed in view of MHCImolecules. Grey and orange,complex structure of LILRB1D1D2(orange) and HLA-A2 (gray);magenta, P41212 structure; red,P21212 structure. Green and blue,

complex structure of KIR2DL1 (blue) and HLA-Cw4 (green); cyan, free KIR2DL1 structure. The calculation ofsuperimposition was done with the program LSQKAB34 in the CCP4 program package32 and shp.35 The Figure wascreated using Bobscript46 and Raster3D.47 (d) Comparison of the 1H–15N HSQC NMR spectra of the 15N-labeledLILRB1D1D2 in the presence (red) and absence (blue) of two equivalents of HLA-Cw7. The low-magnetic-field region ofNMR spectra where the resonances of N1H of tryptophan residues were observed is shown. The amino acid numbers forall six assigned signals are shown. Upon binding to HLA-Cw7, three signals (blue arrows) were changed (two signalsdisappeared and one largely shifted (red arrow)), which corresponded to Trp46, Trp67 and Trp185.


(Figure 3(b)). The domain structures of D1 and D2are almost the same as those of the P41212 structure(the r.m.s.d. for D1 and D2 were 0.27 A (86 Ca

atoms) and 0.39 A (77 Ca atoms), respectively).However, the interdomain angle between domains1 and 2 is more acute than that of the P41212structure by 11 degrees, calculated by the programDymDom21 (Figure 3(c)), indicating flexibility in theinterdomain linkage.

NMR spectroscopic analysis ofLILRB1D1D2–MHCI interaction

In order to directly observe conformationalchanges of the LILRB1 upon the HLA–Cw7complex formation (Kdw7.8 mM), the NMR spectro-scopic analysis of the LILRB1–HLA–Cw7 complexwas performed. All N1H signals of six tryptophanresidues of the free LILRB1D1D2 were observed

at the lower magnetic field region of 1H–15Nheteronuclear single quantum coherence (HSQC)spectra (Figure 3(d)). We successfully assigned allN1H signals shown in Figure 3(d) by comparingwith 1H–15N HSQC spectra of all six Trp mutantsprepared (detail in Materials and Methods). In thecase of 1:2 complex, the signals for three of themwere located in the same positions as those of thefree LILRB1, however, the signals for the other twohad disappeared and the remaining one was largelyshifted. All three unshifted signals were Trp9,Trp170 and Trp171, which were far from theMHCI-binding site on the basis of the crystalstructure of the LILRB1-HLA-A2 complex(Figure 3(b)). Three changed signals correspondedto Trp46, Trp67 and Trp185. While Trp46 and Trp67were in or close to the MHCI-binding site(Figure 3(b), cyan stick models), Trp185 was locatednot in the binding site but in the interdomain


interface. Therefore, this result provided the directevidence for the domain movement upon MHCIbinding.

Discussion

Ligand recognition by immune cell surfacereceptors is an essential first step in discriminationbetween self and non-self by the host defensesystem. Because they are difficult to express andthey bind ligands with low affinities, ligandrecognition by these receptors remains poorlyunderstood. Here we determined the kinetic,thermodynamic and structural characteristics ofthe recognition of MHCIs by LILRB1.The LILRB1negatively regulates the activation of a wide rangeof human leukocytes including natural killer and Tcells, which kill abnormal cells expressing no orlittle MHCIs on the cell surface. In contrast with Tcell receptors (TCRs) and killer cell Ig-like receptors(KIRs), LILRB1 can bind to a broad range ofMHCIs through the highly conserved MHCI a3domain and b2m.

Here, we showed that LILRB1–MHCI inter-actions are entropically driven, and have a smallDCp and fast kinetics, features typical of “rigid-body” interactions,16 which are not accompanied bysignificant conformational rearrangements orreductions in flexibility. However, the presentcrystallographic analysis of LILRB1 reveals somedomain–domain flexibility in the free form, imply-ing the existence of some conformational adjust-ment and/or reduction in the flexibility of LILRB1upon the complex formation. Furthermore, NMRdata also showed that the domain movementoccurred upon the complex formation. Thus thereappears to be a discrepancy between these bindingproperties and the structural features of these

interactions. We will discuss these properties andthis discrepancy in detail below.

Binding properties of LILRB1–MHCI interactions

The van’t Hoff and ITC analyses indicated thatthe LILRB1–MHCI interactions are characterized byfavorable entropic changes, indeed are entropicallydriven. While this is sometimes observed withprotein–protein interactions, it contrasts withtypical examples of antigen–antibody interactionsor TCRs–MHCIs, which are characterized by large,unfavorable entropic changes19 (while the recentreport22 showed that one TCR–MHC interactionshowed entropic and enthalpic-driven thermo-dynamics). This is believed to reflect the substantialconformational adjustment (“induced fit”) requiredwhen TCR binds peptide–MHC complex (pMHC)together with the trapping of solvent molecules atthe TCR/pMHC interface19 (Table 1 and Figure 4).Favorable entropic changes may also reflect a largecontribution of the hydrophobic effects on bindingenergy as a result, for example, of the burial of alarge hydrophobic patch upon complex formation.Based on the crystal structure of the LILRB1–HLA-A2 complex,13 polar and non-polar interfacialareas (DAp and DAnp) were determined using theprogram NACCESS23 to be 762 A2 and 944 A2 ofsolvent exclusion. The ratio of the non-polar surfacearea (DAnp/(DApCDAnp)) is 55%, which is similarto the average non-polar character of protein–protein interfaces, 56%,24 and thus the LILRB1–MHC interactions are not expected to exhibit anunusually large hydrophobic effect.25

The heat capacity changes (DCps) are within therange of values measured for typical protein–protein interactions (K0.33(G0.2) kcal molK1

KK1).16 Spolar and Record26 pointed out that, ifthere are no significant conformational changes

Figure 4. Comparison of thermo-dynamic properties (at 25 8C) ofseveral macromolecular inter-actions. The average values forLILRB1D1D2 are calculated fromthe values of equilibrium analyses(Table 1). The values for antibody-antigen interactions and othernormal protein–protein interactionsare from Stites.16 Data for TCR–peptide–MHC interactions aredepicted as the mean and range ofnine determinations.19,36–39 Errorbars represent standard error of themean.


upon complex formation (i.e. for rigid-body typeinteractions) the expected heat capacity can becalculated from the complex structure using theequation below (equation (1)):

Cp;calc Z ð0:32G0:04ÞDAnpKð0:14G0:04Þ

!DAp ðcal molK1 KK1Þ (1)

Using the above DAp and DAnp the DCp,calc wascalculated to be K0.20 kcal molK1 KK1. This issimilar to the experimentally determined values forLILRB1–MHC interactions (K0.10wK0.22 kcal molK1),consistent with the absence of large conformationaladjustments upon complex formation.

The activation energy of association and dis-sociation (Eon

a and Eoffa ) can be considered to reflect

the number of non-covalent bonds that need to bebroken for association and dissociation, respect-ively. A large Eon

a value for association is indicativeof a requirement for conformational change forbinding. For LILRB1–MHCI interactions Eon

a valuesrange from 4.0 kcal molK1 to 7.1 kcal molK1

(Table 3). These low values are within therange of typical diffusion-controlled reactions(w5 kcal molK1),27,28 and much smaller than inter-actions involving conformational changes(e.g. w11–12 kcal molK1 for TCR–MHC inter-action).19,27 The Eoff

a values were also low(2.2–5.6 kcal molK1), indicating that the number ofinter-molecular bonds that need to be broken todissociate LILRB1 from MHCI is relatively low,consistent with the relatively fast off rate. Takentogether these binding parameters are consistentwith a rigid-body binding interaction, with limitedconformational changes accompanying binding.

Structural properties of LILRB1–MHCIinteraction

The structures of LILRB1D1D2 andLILRB1D1D2–HLA-A2 complex had been deter-mined by X-ray crystallography.12,13,29 Here wedetermined the free form of the LILRB1 structurefrom the P21212 crystal. The interdomain anglebetween domains 1 and 2 is smaller than that of theP41212 structure (as described above, the angledifference between P21212 and P41212 structureswas 11 degrees calculated by the program Dym-Dom21) as shown in Figure 3(c), indicating thatthere is some flexibility between domains 1 and 2 inthe free form. In the complex state, the angle islarger than both free form structures (the angledifferences between free and complex structureswere 19 degrees (P21212) and 14 degrees (P41212)calculated by the program DymDom,21) whichcannot be fitted into MHCI complex formationwithout any structural adjustments (Figure 3(c)).Consistently, NMR analysis of the LILRB1–MHCIcomplex clearly indicated that, upon the complexformation, the conformation changes of Trp185 atthe domain 1–domain 2 interface were observed(Figure 3(d)), providing direct evidence for the

domain movement of LILRB1.Taken together thesestructural data are consistent with the confor-mational adjustment and/or reduction in theflexibility of LILRB1 and possibly MHCI uponcomplex formation.

However, as described above, the bindingproperties of the LILRB1–MHCI interactions aremore consistent with a rigid-body bindingmodel.16,26 Such binding characteristics are alsoobserved in the KIR2DL3–MHCI (HLA–Cw7)interaction (Tables 1 and 2, Figure 3(c)). Thereare several possible explanations for the discre-pancy between the structural and binding data.Firstly, the structural data may give a misleadingimpression of the conformational changes accom-panying binding. While 11–19 degrees of theinterdomain angular differences between freeand complex structures did not seem to be smalland NMR data showed the conformationalchanges of the interdomain interface, this extentof the domain movement may not provide anysignificant entropic loss. There may be onlylimited flexibility in the unbound state (i.e. thereare only a few main conformations in theunbound form, and the dominant conformationmay be the one compatible with binding).Alternatively, the LILRB1–MHCI complex mayretain considerable flexibility, in which casebinding may not lead to a substantial reductionin flexibility. A second explanation may relate tothe contribution of side-chain/side-chain inter-actions to the binding interface. In the LILRB1–HLA-A2 and KIR2DL1–HLA-Cw4 interfaces, side-chain/side-chain interactions of the hydrogenbonds and salt-bridges are dominant (53% and78% of total, respectively) in comparison withthose of the average protein–protein interfaces(36%).24 Because the side-chain/side-chain inter-actions could allow more conformational ensem-bles in the bound states than interactionsinvolving main chain atoms, they may masklarge entropic losses resulting from the interdo-main conformational changes. A third explanationis that solvent effects may be sufficiently large tomask the effects of interdomain conformationalchange on thermodynamics. For example, theremay be an unusually large degree of solventexclusion from the binding interface. This isunlikely however since, as noted above, analysisof the binding interface does not reveal anunusually large hydrophobic surface. Althoughany one or a combination of these three mechan-isms could explain the discrepancy betweenbinding and structural data and further investi-gation will be necessary to reach the conclusion,the first mechanism is likely to be dominant sinceit accounts best for the fast kon and low Eon

a of theLILRB1–MHCI interactions.

The “rigid-adaptation” model proposed byMcFarland and Strong30 showed entropic-dominantthermodynamics with fast kinetics, which weresimilar to those of LILRB1 and KIRs. However,this model did not include such large domain


movements observed here but local side-chain andloop conformation changes, thus the recognitionsystems of LILRB1 and KIRs were distinct.

Functional implication

In the previous studies, LILRB1 was mainlyexpressed on T cells in the early stage of immuneresponse. LILRB1 showed fast kinetic binding toMHCIs. On the other hand, the TCR coreceptor,CD8, which is responsible for TCR signaling andcompetes LILRB1 for MHCI bindings,10 showedlower affinity (Kd w200 mM) with muchfaster dissociation rate (R18 s) than LILRB1(Kd w2–9 mM, kd w2–5 s) (Table 2). Thus, LILRB1can dominantly bind to MHCIs rather than CD8 toquickly transduce the inhibitory signals. Conver-sely TCR–CD8 complexes should overcome theLILRB1 inhibition, suggesting that LILRB1 mayhave pivotal roles to increase the threshold foractivation of T cells in the early immune response.Furthermore, as described above, KIRs that havesimilar binding characteristics to LILRB1 involvethe T cell regulation in the later stages of immuneresponse, such as the production of memory T cells.Therefore, such fast kinetic and preferentiallyentropic-driven binding coupled with domainmovements could be suitable for regulation ofsignal transductions in cell–cell contacts.

Conclusion

Our present kinetic, thermodynamic and struc-tural analysis revealed that the LILRB1–MHCIinteraction did not conform strictly to either therigid-body or the induced-fit type binding mechan-isms. It suggested that the interdomain flexibilityidentified in LILRB1 did not necessarily imposesubstantial kinetic and entropic penalties on bind-ing. These binding characteristics may be advan-tageous for cell surface receptors involved in cell–cell recognition.

Materials and Methods

Preparation of recombinant proteins

The N-terminal two extracellular Ig-like domains(residues 1–197) of LILRB1 (hereafter calledLILRB1D1D2) were prepared using an Escherichia coliexpression system as inclusion bodies. The soluble formsof these proteins were refolded and purified asdescribed.10 Soluble HLA-A*1101 (with peptideAIFQSSMTK), HLA-B*3501 (IPLTEEAEL), HLA-Cw*0401 (QYDDAVYKL) and HLA-G1 (RIIPRHLQL)were refolded from inclusion bodies, purified with gelfiltration and anion exchange chromatography. Sincethese MHCIs have the C-terminal biotin ligase (BirA)recognition sequence (GSLHHILDAQKMVWNHR),these are biotinylated with biotin ligase.10 For crystallo-graphic and NMR analyses, soluble HLA–Cw7 withoutthe biotin tag was prepared as described.10 Site-directed

mutagenesis for Trp mutants of LILRB1D1D2 wasperformed by two-step PCR method.

Surface plasmon resonance (SPR)

Surface plasmon resonance experiments were per-formed using Biacore2000TM (Biacore AB, St. Albans,UK). The biotinylated MHCIs and control proteins(biotinylated OX48 antibody or bovine serum albumin(BSA)) were immobilized on the research grade CM5 chip(BIAcore AB) through covalently coupled streptavidin.LILRB1D1D2, after buffer exchange to HBS-EP (10 mMHepes (pH 7.4), 150 mM NaCl, 3.4 mM EDTA and 0.005%Surfactant P20), were injected over the immobilizedMHCIs. The binding response at each concentration(from 0.2 mM to 60 mM) was calculated by subtractingthe equilibrium response measured in the control flowcell from the response in the MHCIs flow cell. Kineticconstants were derived using the curve fitting facility ofthe BIAevaluation program (version 3.0, BIAcore) to fitrate equations derived from the simple 1:1 Langmuirbinding model (ACB4AB). Other curve fitting wasperformed by Origin version 5 (MicroCal). Affinityconstants (Kds) were derived by Scatchard analysis orby non-linear curve fitting of the standard Langmuirbinding isotherm.

Thermodynamic analyses using SPR

Equilibrium analyses of LILRB1D1D2 were performedat five temperature points (10 8C, 15 8C, 20 8C, 25 8Cand 30 8C). The immobilization levels of HLA moleculeswere from 1000 RU to 3000 RU. The standard stateGibbs energy change upon binding was obtained fromequation (2):

DG Z RT ln Kd (2)

where Kd is dissociation constant expressed in unitsmol/l and R is the gas constant. DG of each data set wereplotted against the temperatures, and fitted with the non-linear van’t Hoff equation (equation (3)):

DG Z DHKTDS CDCpðTK298:15Þ

KDCpT lnðT=298:15Þ (3)

where DH and DS are the binding enthalpy and entropy at298.15 K, respectively, and DCp is the heat capacity, whichis assumed to be temperature-independent.

Kinetic analysis using SPR

Association rate constants (kon) and dissociation rateconstants (koff) were obtained from two strategies. Onewas the global fitting analysis. The following equationsderived from the 1:1 Langmuir binding model weresimultaneously fitted to the raw data for the association(equation (4)) and dissociation (equation (5)) phases atdifferent concentrations of LILRB1.

R Z konCRmax½1KexpfKðkonC CkoffÞðtKt0Þg=

ðkonC CkoffÞCRi

(4)

R Z R0expfKkoffðtKt0Þg (5)

where R is the binding response at t, Rmax is the maximumbinding response, C is the concentration of the analyte, Ri

is the response from bulk effect, and R0 is the response atthe beginning of the dissociation phase. The second


strategy was to obtain the koff by fitting equation (5) todissociation phase and then to calculate the kons usingequation (6):

kon Z koff=Kd (6)

Notably this method was used to determine the kon

values used to obtain the activation enthalpy of asso-ciation (see below). Fitting was performed using BIAevaluation and Origin.

Activation energies of association and dissociationphases

The activation energy of association or dissociation(Eon

a or Eoffa ) was obtained using the Arrhenius equation:

ln k Z ln AKEa=RT (7)

where k is the kon or koff and A is a constant known aspre-exponential factor.

Thermodynamic analysis using ITC

Thermodynamic analysis by ITC was performed usingVP-ITC (Microcal Inc., Northampton, MA). All proteinsused (LILRB1D1D2 and HLA-G1) were dialyzed toHBS-EP buffer. At each injection (total 25 times), 10 ml of240 mM LILRB1D1D2 was put into a sample cell contain-ing about 1.7 ml of 19.5 mM HLA-G1. To determine theheat capacity, the experiment was performed at thetemperatures of 15, 20, 25, 30, 35 and 40 8C. To measurebackground heats, the solution of LILRB1D1D2 wasinjected to a sample cell containing HBS-EP buffer. Alltitration data minus the background heats were analyzedby a fit to the binding isotherm using the Origin version7.0 (Microcal Inc.).

Crystallization and structure determination ofLILRB1D1D2

The free form of the LILRB1D1D2 was crystallized fromthe mixture of 1 ml of the LILRB1D1D2/HLA-Cw7complex protein (10 mg/ml), and 1 ml of 1.6 M sodiumformate, 0.08 M sodium acetate (pH 5.0), 0.2 M GuHCl at25 8C by the hanging-drop method. Data were collected at100 K at SPring-8 beamline BL41XU. These crystalsbelong to the space group P21212 (aZ36.2 A, bZ104.4 A,cZ53.5 A). Data were processed and scaled with the HKLprogram package.31 The structure from the P21212 crystalwas determined by molecular replacement using AmoRein the CCP4 package.32 The LILRB1D1D2 structure fromthe P41212 crystal11 was used as a search probe. One clearsolution was found using E values in the range of15–3.5 A, and these crystals have one LILRB1D1D2 inthe asymmetric unit. The structure of LILRB1D1D2 wasfurther refined with the grouped B-factor and positionalrefinements of CNS33 and Refmac32 using the 37.3–3.0 Areflections. The Rfree and Rwork of the final model were30.6% and 23.3%, respectively. The average B factor was39.3 A2.

NMR spectroscopic analysis of LILRB1D1D2–MHCIinteraction

Uniformly 15N-labeled LILRB1D1D2 was preparedfrom E. coli BL21/pLysS cells carrying an expressionplasmid grown in M9 minimal medium containing 1 g/lof 15NH4Cl and purified by the same method as thenon-labeled protein described above. 15N-labeled

LILRB1D1D2 was dissolved in 95% HBS-EP and 5%2H2O (v/v) (final concentration 2.4 mg/ml). HLA-Cw7was also dissolved in the same buffer (final concentration13.7 mg/ml). For the direct observation of conformationalchanges of the LILRB1 upon the HLA-Cw7 complexformation, a series of 1H–15N HSQC spectra at differentLILRB1: HLA–Cw7 ratios using the mixture of the abovesolutions were measured with a Bruker Avance600spectrometer at 40 8C. In order to assign N1H signals ofindole rings of all tryptophan residues, we prepared allsix Trp mutants (Trp9Tyr, Trp46Phe, Trp67Tyr, Trp170Tyr,Trp171Tyr and Trp185Tyr) and measured 1H–15N HSQCspectra.

Protein Data Bank accession number

The coordinates of the LILRB1D1D2 structure havebeen deposited with the RCSB Protein Data Bank underaccession code 1UFU.

Acknowledgements

We thank H. Hatanaka, E.Y. Jones, V. Braud, S.Rowland-Jones, A. Makadzange, M. Colonna,A. Shiratori, S. Tabata, K. Fukunaga and M.Kajikawa for their advice. We also thank M.Kawamoto, H. Sakai and K. Miura for assistancein data collection at Spring8. This work wassupported in part by the Ministry of Education,Science, Sports, Culture and Technology, Japan, the2000th year Joint Research Project (Soken/K01-4) ofSokendai, the research grants from the Nakajimafoundation, the Uehara foundation and the Kanaefoundation, the NIG Cooperative Research Program(2003~2005), and the JSPS Research Fellowship forYoung Scientist (M.S. and L.R.).

References

1. Martin, A. M., Kulski, J. K., Witt, C., Pontarotti, P. &Christiansen, F. T. (2002). Leukocyte Ig-like receptorcomplex (LRC) in mice and men. Trends Immunol. 23,81–88.

2. Samaridis, J. & Colonna, M. (1997). Cloning of novelimmunoglobulin superfamily receptors expressed onhuman myeloid and lymphoid cells: structuralevidence for new stimulatory and inhibitory path-ways. Eur. J. Immunol. 27, 660–665.

3. Cosman, D., Fanger, N., Borges, L., Kubin, M.,Chin, W., Peterson, L. & Hsu, M. L. (1997). A novelimmunoglobulin superfamily receptor for cellularand viral MHC class I molecules. Immunity, 7,273–282.

4. Borges, L., Hsu, M. L., Fanger, N., Kubin, M. &Cosman, D. (1997). A family of human lymphoid andmyeloid Ig-like receptors, some of which bind toMHC class I molecules. J. Immunol. 159, 5192–5196.

5. Wagtmann, N., Rojo, S., Eichler, E., Mohrenweiser, H.& Long, E. O. (1997). A new human gene complexencoding the killer cell inhibitory receptors andrelated monocyte/macrophage receptors. Curr. Biol.7, 615–618.


6. Colonna, M., Navarro, F., Bellon, T., Llano, M.,Garcia, P., Samaridis, J. et al. (1997). A commoninhibitory receptor for major histocompatibilitycomplex class I molecules on human lymphoid andmyelomonocytic cells. J. Exp. Med. 186, 1809–1818.

7. Colonna, M., Samaridis, J., Cella, M., Angman, L.,Allen, R. L., O’Callaghan, C. A. et al. (1998). Humanmyelomonocytic cells express an inhibitory receptorfor classical and nonclassical MHC class I molecules.J. Immunol. 160, 3096–3100.

8. Fanger, N. A., Cosman, D., Peterson, L., Braddy, S. C.,Maliszewski, C. R. & Borges, L. (1998). The MHC classI binding proteins LIR-1 and LIR-2 inhibit Fc receptor-mediated signaling in monocytes. Eur. J. Immunol. 28,3423–3434.

9. Dietrich, J., Cella, M. & Colonna, M. (2001). Ig-liketranscript 2 (ILT2)/leukocyte Ig-like receptor 1 (LIR1)inhibits TCR signaling and actin cytoskeleton re-organization. J. Immunol. 166, 2514–2521.

10. Shiroishi, M., Tsumoto, K., Amano, K., Shirakihara, Y.,Colonna, M., Braud, V. M. et al. (2003). Humaninhibitory receptors Ig-like transcript 2 (ILT2) andILT4 compete with CD8 for MHC class I binding andbind preferentially to HLA-G. Proc. Natl Acad. Sci.USA, 100, 8856–8861.

11. Chapman, T. L., Heikeman, A. P. & Bjorkman, P. J.(1999). The inhibitory receptor LIR-1 uses a commonbinding interaction to recognize class I MHC mole-cules and the viral homolog UL18. Immunity, 11,603–613.

12. Chapman, T. L., Heikema, A. P., West, A. P., Jr &Bjorkman, P. J. (2000). Crystal structure and ligandbinding properties of the D1D2 region of theinhibitory receptor LIR-1 (ILT2). Immunity, 13,727–736.

13. Willcox, B. E., Thomas, L. M. & Bjorkman, P. J. (2003).Crystal structure of HLA-A2 bound to LIR-1, a hostand viral major histocompatibility complex receptor.Nature Immunol. 4, 913–919.

14. Boyington, J. C., Motyka, S. A., Schuck, P., Brooks,A. G. & Sun, P. D. (2000). Crystal structure of an NKcell immunoglobulin-like receptor in complex with itsclass I MHC ligand. Nature, 405, 537–543.

15. Fan, Q. R., Long, E. O. & Wiley, D. C. (2001). Crystalstructure of the human natural killer cell inhibitoryreceptor KIR2DL1-HLA-Cw4 complex. NatureImmunol. 2, 452–460.

16. Stites, W. E. (1997). Protein–Protein interactions:interface structure, binding thermodynamics, andmutational analysis. Chem. Rev. 97, 1233–1250.

17. Corr, M., Slanetz, A. E., Boyd, L. F., Jelonek, M. T.,Khilko, S., al-Ramadi, B. K. et al. (1994). Tcell receptor-MHC class I peptide interactions: affinity, kinetics,and specificity. Science, 265, 946–949.

18. van der Merwe, P. A. & Barclay, A. N. (1996). Analysisof cell-adhesion molecule interactions using surfaceplasmon resonance. Curr. Opin. Immunol. 8, 257–261.

19. Willcox, B. E., Gao, G. F., Wyer, J. R., Ladbury, J. E.,Bell, J. I., Jakobsen, B. K. & van der Merwe, P. A.(1999). TCR binding to peptide-MHC stabilizes aflexible recognition interface. Immunity, 10, 357–365.

20. Navaza, J. (1994). AmoRe : an automated package formolecular replacement. Acta Crystallog. sect. A, 50,157–163.

21. Hayward, S. & Berendsen, H. J. (1998). Systematicanalysis of domain motions in proteins fromconformational change: new results on citratesynthase and T4 lysozyme. Proteins: Struct. Funct.Genet. 30, 144–154.

22. Boniface, J. J., Reich, Z., Lyons, D. S. & Davi, M. M.(1999). Thermodynamics of T cell receptor binding topeptide-MHC: evidence for a general mechanism ofmolecular scanning. Proc. Natl Acad. Sci. USA, 96,11446–11451.

23. Hubbard, S. J. & Thornton, J. M. (1993). NACCESS,Computer Program, Department of Biochemistry andMolecular Biology, University College London,London.

24. Lo Conte, L., Chothia, C. & Janin, J. (1999). The atomicstructure of protein-protein recognition sites. J. Mol.Biol. 285, 2177–2198.

25. Sturtevant, J. M. (1977). Heat capacity and entropychanges in processes involving proteins. Proc. NatlAcad. Sci. USA, 74, 2236–2240.

26. Spolar, R. S. & Record, M. T., Jr (1994). Coupling oflocal folding to site-specific binding of proteins toDNA. Science, 263, 777–784.

27. Gutfreund, H. (1995). Kinetics for the Life Sciences,Cambridge University Press, Cambridge.

28. O’Callaghan, C. A., Cerwenka, A., Willcox, B. E.,Lanier, L. L. & Bjorkman, P. J. (2001). Molecularcompetition for NKG2D: H60 and RAE1 competeunequally for NKG2D with dominance of H60.Immunity, 15, 201–211.

29. Willcox, B. E., Thomas, L. M., Chapman, T. L.,Heikema, A. P., West, A. P., Jr & Bjorkman, P. J.(2002). Crystal structure of LIR-2 (ILT4) at 1.8 A:differences from LIR-1 (ILT2) in regions implicated inthe binding of the Human Cytomegalovirus class IMHC homolog UL18. BMC Struct. Biol. 2, 6.

30. McFarland, B. J. & Strong, R. K. (2003). Thermo-dynamic analysis of degenerate recognition by theNKG2D immunoreceptor: not induced fit but rigidadaptation. Immunity, 19, 803–812.

31. Otwinowski, Z. & Minor, W. (1997). Processing ofX-ray diffraction data collected in oscillation mode.Methods Enzymol. 276, 307–326.

32. Collaborative Computational Project Number 4(1994). The CCP4 Suite: programs for proteincrystallography. Acta Crystallog. sect. D, 50, 760–763.

33. Brunger, A. T., Adams, P. D., Clore, G. M.,DeLano, W. L., Gros, P., Grosse-Kunstleve, R. W.et al. (1998). Crystallography NMR system: a newsoftware suite for macromolecular structure deter-mination. Acta Crystallog. sect. D, 54, 905–921.

34. Kabsch, W. (1976). A solution for the best rotation torelate two sets of vectors. Acta Crystallog. sect. A, 32,922–923.

35. Stuart, D. I., Levine, M., Muirhead, H. & Stammers,D. K. (1979). Crystal structure of cat muscle pyruvatekinase at a resolution of 2.6 A. J. Mol. Biol. 134,109–142.

36. Anikeeva, N., Lebedeva, T., Krogsgaard, M., Tetin,S. Y., Martinez-Hackert, E., Kalams, S. A. et al. (2003).Distinct molecular mechanisms account for thespecificity of two different T-cell receptors. Biochem-istry, 42, 4709–4716.

37. Garcia, K. C., Radu, C. G., Ho, J., Ober, R. J. & Ward,E. S. (2001). Kinetics and thermodynamics of T cellreceptor–autoantigen interactions in murine experi-mental autoimmune encephalomyelitis. Proc. NatlAcad. Sci. USA, 98, 6818–6823.

38. Lee, J. K., Stewart-Jones, G., Dong, T., Harlos, K., DiGleria, K., Dorrell, L. et al. (2004). Tcell cross-reactivityand conformational changes during TCR engage-ment. J. Exp. Med. 200, 1455–1466.


39. Davis-Harrison, R. L., Armstrong, K. M. & Baker,B. M. (2005). Two different T cell receptors usedifferent thermodynamic strategies to recognize thesame peptide/MHC ligand. J. Mol. Biol. 346, 533–550.

40. Maenaka, K., van der Merwe, P. A., Stuart, D. I.,Jones, E. Y. & Sondermann, P. (2001). The human lowaffinity Fcgamma receptors IIa, IIb, and III bind IgGwith fast kinetics and distinct thermodynamic proper-ties. J. Biol. Chem. 276, 44898–44904.

41. Maenaka, K., Juji, T., Nakayama, T., Wyer, J. R.,Gao, G. F., Maenaka, T. et al. (1999). Killer cellimmunoglobulin receptors and T cell receptors bindpeptide-major histocompatibility complex class I withdistinct thermodynamic and kinetic properties. J. Biol.Chem. 274, 28329–28334.

42. Bakker, T. R., Piperi, C., Davies, E. A. & Merwe, P. A.(2002). Comparison of CD22 binding to native CD45and synthetic oligosaccharide. Eur. J. Immunol. 32,1924–1932.

43. Wyer, J. R., Willcox, B. E., Gao, G. F., Gerth, U. C.,Davis, S. J., Bell, J. I., van der Merwe, P. A. & Jakobsen,B. K. (1999). T cell receptor and coreceptor CD8alphaalpha bind peptide-MHC independently andwith distinct kinetics. Immunity, 10, 219–225.

44. van der Merwe, P. A., Bodian, D. L., Daenke, S.,Linsley, P. & Davis, S. J. (1997). CD80 (B7-1) binds bothCD28 and CTLA-4 with a low affinity and very fastkinetics. J. Exp. Med. 185, 393–403.

45. Ding, Y. H., Baker, B. M., Garboczi, D. N.,Biddison, W. E. & Wiley, D. C. (1999). Four A6-TCR/peptide/HLA-A2 structures that generate verydifferent T cell signals are nearly identical. Immunity,11, 45–56.

46. Esnouf, R. M. (1997). An extensively modified versionof MolScript that includes greatly enhanced coloringcapabilities. J. Mol. Graph. Model. 15, 133–138.

47. Merritt, E. A. & Bacon, D. J. (1997). Raster3Dphotorealistic molecular graphics. Methods Enzymol.277, 505–524.

Edited by J. E. Ladbury

(Received 22 April 2005; received in revised form 11 October 2005; accepted 17 October 2005)Available online 8 November 2005

Entropically Driven MHC Class I Recognition by Human Inhibitory Receptor Leukocyte Ig-like Receptor...

Documents

Transcript of Entropically Driven MHC Class I Recognition by Human Inhibitory Receptor Leukocyte Ig-like Receptor...