energy between the unstable state and thecentre of gravity of the reservoir distribu-tion. In the seemingly more general situationwhere this energy spread is small comparedwith the separation between the unstablestate and the nearest maximum in the reser-voir spectrum, however, the unstable stateshould decay faster as n increases. In this casethe quantum anti-Zeno effect should beobserved because, as n and therefore theenergy spread of the unstable state increases,so does the number of accessible reservoirstates into which transitions can occur.

Judging by typical decay processes andthe nature of the reservoirs used to modelthem, one surmises that the anti-Zeno effectis more common than the Zeno effect1,8,9. Asyet there are no known experiments corrob-orating this idea, but it has already been sug-gested8 that the anti-Zeno effect is cause forconcern in connection with error-correction

Class I molecules of the major histocom-patibility complex (MHC) are normal-ly expressed on the surface of most cells

in the body. They serve two purposes in theimmune system. T cells use MHC moleculesto identify and kill cells that are infected byforeign material, such as viruses. Naturalkiller (NK) cells, meanwhile, use them toidentify and spare the lives of cells that arenormal and healthy1. On page 537 of thisissue2, Boyington and colleagues reveal thestructural nature of a life-saving liaisonbetween a human MHC molecule and itsreceptor on an NK cell.

MHC class I molecules bind intracellularpeptides and present them to the immunesystem for scrutiny. If the MHC moleculespresent foreign peptides — for example,those from a degraded virus protein — T cells are activated and will eventuallydestroy the infected cells. This limits replica-tion and spread of the infection. NK cells alsorely on MHC molecules for decision-mak-ing, but in the opposite way. On recognizingself-MHC molecules on target cells (Fig. 1a,overleaf), the MHC receptors on NK cellsgenerate inhibitory signals3. These cancel anactivating signal initiated previously byother receptors on NK cells that have recog-nized ubiquitously expressed ligands on thetarget cells. If self-MHC molecules are notadequately displayed — as occurs in somecancerous or transplanted cells, for example— the NK receptors cannot deliver theinhibitory signal and will go on to kill the target cell.

schemes for quantum computing. More-over, Fearn and Lamb4 reported “quite thereverse of the Zeno effect” in a computationalstudy involving barrier penetration. Sowatch that pot closely. ■

Peter W. Milonni is in the Theoretical Division, LosAlamos National Laboratory, Los Alamos, NewMexico 87545, USA.e-mail: pwm@lanl.gov1. Kofman, A. G. & Kurizki, G. Nature 405, 546–550 (2000).

2. Itano, W. M., Heinzen, D. J., Bollinger, J. J. & Wineland, D. J.

Phys. Rev. A 41, 2295–2300 (1990).

3. Ballentine, L. E. Phys. Rev. A 43, 5165–5167 (1991).

4. Fearn, H. & Lamb, W. E. Jr Phys. Rev. A 46, 1199–1205 (1992).

5. Kwiat, P., Weinfurter, H., Herzog, T., Zeilinger, A. & Kasevich,

M. in Fundamental Problems in Quantum Theory (eds

Greenberger, D. M. & Zeilinger, A.) 383–393 (New York Acad.

Sci., 1995).

6. Kwiat, P. G. et al. Phys. Rev. Lett. 83, 4725–4728 (1999).

7. Misra, B. & Sudarshan, E. C. G. J. Math. Phys. 18, 756–763

(1977).

8. Lewenstein, M. & Rzazewski, K. Phys. Rev. A 61, 022105-

1–022105-5 (2000).

9. Kaulakys, B. & Gontis, V. Phys. Rev. A 61, 1131–1137 (1997).

news and views

NATURE | VOL 405 | 1 JUNE 2000 | www.nature.com 527

Human NK cells can express MHC class Ireceptors of either the immunoglobulinfamily (the receptors of which are calledkiller-cell immunoglobulin-like inhibitoryreceptors, or KIRs) or the C-type-lectin fam-ily. It is the structure of an immunoglobulin-type receptor, KIR2DL2, in complex with its MHC class I ligand, HLA-Cw3, that isdescribed by Boyington et al.2.

The structure of the extracellular parts of a KIR consists of two globular immuno-globulin-like domains linked by a hingeregion4. In the MHC–receptor complex des-cribed by Boyington et al., the receptor bindswith both of its globular domains in a 1:1 stoichiometry to the MHC molecule, acrossthe MHC’s ‘business end’, consisting of thegroove-containing peptide.Six loops locatedclose to the hinge region contact the MHCmolecule.

Mouse NK cells do not express KIRs.Instead they have inhibitory MHC receptorsfrom the Ly49 subfamily of C-type lectins. Atthe end of last year, the structure of such areceptor, Ly49A, in complex with its MHCclass I ligand, H-2Dd, was described5. Theinteractions between receptor and ligand inthis structure differ completely from thosedescribed by Boyington et al. (Fig. 1b, d), forthe mouse receptor binds at one side of thepeptide-binding groove. At first sight, how-ever, the major ‘footprint’ of the human NK-cell receptor on its MHC ligand seemsremarkably similar to that of the T-cell anti-gen receptor on an MHC ligand6 (Fig. 1c).Both receptors span the two a-helices of the

Immunology

The footprint of a killerKlas Kärre and Gunter Schneider

MHC molecule and the peptide, covering asurface area of about 1,500–1,700 Å. But acloser examination reveals that the bindingsite of the NK-cell receptor is shifted towardsthe carboxy-terminal part of the peptide.The T-cell receptor, by contrast, binds morecentrally. Another difference is that the KIRfootprint is dominated by charged andhydrophilic interactions, whereas the T-cellreceptor relies mainly on hydrophobic, andless on electrostatic, contacts.

The degree of overlap between the twofootprints indicates that the T-cell antigenreceptor and the KIR probably cannot bindto the MHC molecule at the same time. Thisis an important issue, as T cells may, undercertain circumstances, express both an acti-vating MHC receptor and an inhibitory KIR.Might one MHC molecule be used to deliverboth activating and inhibitory signals to thesame T cell? This appears impossible for thehuman KIR studied by Boyington et al. (Fig.1b, c), but might arise for mouse NK cellsbearing the Ly49A receptor (Fig. 1b, d).

The role of the MHC-presented peptidein recognition by the NK-cell receptor ismuch debated. For some receptors, such asmouse Ly49A, there is no influence of thepeptide sequence. But the opposite appearsto be true7 for the group of receptors thatincludes that studied by Boyington et al. Thisdifference is beautifully explained by thethree-dimensional structures of the tworeceptor–ligand complexes2,5 (Fig. 1b, d).The binding site of the murine receptor isdistant from the peptide. But there are directinteractions between the peptide and thehuman receptor. Two positions in the non-americ peptide participate in these interac-tions in the structure — a finding that agreeswell with biochemical binding studies usinga series of peptide variants2.

MHC molecules are polygenic (eachindividual has several genetic loci thatencode MHC molecules) and polymorphic(each locus expresses but one of many possi-ble alleles). Why does one group of KIRs,such as that to which KIR2DL2 belongs,bind to MHC molecules encoded by just oneparticular locus, termed HLA-C for thisgroup? And why do different KIRs withinthis group bind to mutually exclusive sub-groups of allelic products from this locus?This is usually referred to as the allospeci-ficity of the receptor, believed to be impor-tant for the role of NK cells in reactionsbetween donor and host in bone-marrowtransplantation8. The allospecificity of KIRshas been analysed in several studies9, theresults of which are explained and extendedby Boyington et al.’s complex. Of 16 aminoacids involved in contacts with the MHCmolecule, 14 are conserved between the KIRstudied by Boyington et al. and another KIR,both of which react with HLA-C ligands but with different allospecificity. Boyingtonet al. found that one of the two amino acids

© 2000 Macmillan Magazines Ltd

that differ between these KIRs is a partner ina critical hydrogen bond with an amino acidthat differs between the relevant HLA-Csubgroups.

The mouse and human receptor–ligandcrystal structures indicate, unexpectedly,another binding site of the receptor to theMHC molecule, not identical in the two complexes. These sites may be involved in ligand-induced receptor aggregation2 orreceptor interactions with ligands in themembrane of the NK cell itself5. But the biological significance of these interactions is yet to be established.

As ever, questions remain. Why havehumans and mice developed different typesof NK receptor, with analogous function butfrom completely different structural fami-lies? Might it be because the receptors need torecognize distinct parts of the MHC molec-ules (Fig. 1b, d)? And if so, why? Structures ofother receptor–ligand pairs will be requiredbefore we can answer these questions.Genomics studies indicate that these evolu-tionarily recent KIR proteins show consider-able diversity, even at residues that do notcontact MHC molecules10 (P. Parham, per-sonal communication). This diversity isevolving rapidly, even by comparison to the

MHC genes10. Might peptides or other KIRligands from infectious agents represent theevolutionary driving force here? Finally, NK cells sometimes carry variants of KIRproteins that activate killing when they rec-ognize MHC ligands. The function of theactivating receptors is unknown — will theirfootprints on the MHC molecule providethe answer? ■

Klas Kärre is at the Microbiology and TumorBiology Center, Karolinska Institute, Box 280,17177 Stockholm, Sweden.e-mail: klas.karre@mtc.ki.seGunter Schneider is in the Department of MedicalBiochemistry and Biophysics, Karolinska Institute,Box 280, 17177 Stockholm, Sweden.1. Kärre, K., Ljunggren, H. G., Piontek, G. & Kiessling, R. Nature

319, 675–678 (1986).

2. Boyington, J. C., Motyka, S. A., Schuck, P., Brooks, A. G. &

Sun, P. D. Nature 405, 537–543 (2000).

3. Moretta, A. et al. Annu. Rev. Immunol. 14, 619–648 (1996).

4. Fan, Q. R. et al. Nature 389, 96–100 (1997).

5. Tormo, J., Natarajan, K., Margulies, D. H. & Mariuzza, R. A.

Nature 402, 623–631 (2000).

6. Garboczi, D. N. et al. Nature 384, 134–141 (1996).

7. Rajagopalan, S. & Long, E. O. J. Exp. Med. 185, 1523–1528

(1997).

8. Ruggeri, L. et al. Transplantation 94, 333–339 (1999).

9. Mandelboim, O. et al. J. Exp. Med. 184, 913–922 (1996).

10.Wilson, M. J. et al. Proc. Natl Acad. Sci. USA 97, 4778–4783

(2000).

11.http://red.niaid.nih.gov/mhc-complexes.html

news and views

528 NATURE | VOL 405 | 1 JUNE 2000 | www.nature.com

Daedalus

Air currentsMany reactors circulate their moltensodium coolant by an electromagneticpump. A current passes through themolten metal at right angles to an appliedmagnetic field, and the motor effectimpels the sodium at right angles to both.Daedalus reckons it would work withinsulating fluids too. For a voltage appliedacross an insulator shifts its electronstransiently, as the insulator becomespolarized. Remove the voltage, and theywould move back again. In a steadymagnetic field, these two brief opposite‘displacement currents’ would exert apush and then a cancelling pull on thedielectric. But reverse the magnetic fieldin the interval, and the dielectric wouldfeel two pushes in the same direction — atrue motor effect.

So, says Daedalus, put an insulator ina.c. electric and magnetic fields at rightangles, reversing at the same frequencyand the right mutual phase, and it will feela cumulative force. At first he hoped thatthe whole thing could be driven as aresonant circuit, with the insulator as thedielectric of its capacitor, and the inductorproviding the magnetic field; but thephasing comes out wrong. The two fieldswill have to be created and phased byspecial circuitry.

The obvious working fluid for the newpump is air. Fans, propellers, blowers andjet engines are all noisy, complicateddevices. A simple silent air-pump with nomoving parts would be widely welcomed.For maximum thrust, it should work atthe highest feasible frequency — manymegahertz if possible. Daedalus’s firstproduct will be a little cooling blower forcomputers and electronic gadgets. But hesoon hopes to scale it up. Displacement-current pumping could make all sorts ofblowers, compressors, air-conditionersand fans blissfully silent and reliable.

His ultimate goal is a displacement-current aircraft engine. The entire craftwould have to be designed around thisradical new source of thrust. Its fieldswould enclose as large a volume of air aspossible, so that even a gentle induced flowmoved a lot of air. An electromagnetichelicopter with vertical downflow mightfill the bill. The strength and direction ofthrust could be altered at electronic speed,giving the craft a wonderful agility in theair. And its total silence would bewelcomed both by the crew and the long-suffering aerophobic public. David Jones

The Further Inventions of Daedalus ispublished by Oxford University Press.

Figure 1 Life-saving liaisons. a, Natural killer (NK) cells can distinguish normal, healthy cells fromaberrant cells in the body. Activating receptors on the NK cell recognize ligands on most normal cells.Signals from these receptors stimulate the NK cell to kill the target cell. But killing is stopped if killer-cell inhibitory receptors recognize adequate levels on the target cell of major histocompatibilitycomplex (MHC) class I molecules. b, The ‘footprint’ (contact area) made by the human NK receptorKIR2DL2 on the antigen-binding groove of its MHC ligand, HLA-Cw3 (ref. 2), viewed from above. c,d, The footprints of a T-cell antigen receptor6 and of the murine NK receptor Ly49A (ref. 5) are shownfor comparison. MHC is shown in red, MHC-bound peptide in purple, and the NK or T-cell receptorsin green (and cyan). Note a second contact site for Ly49A beneath the antigen-binding groove. (b–d,Reproduced with permission from ref. 11.)

+ –

NK cell

Activatingreceptor

Target cellunder scrutiny

MHC molecule

Human KIR

T-cell receptor

Mouse Ly49A receptor

Inhibitoryreceptor

a b

c

d

© 2000 Macmillan Magazines Ltd