Some 50 years ago, profound insightswere obtained into the nature of themetallic and superconducting states

of electrons in solids. According to Fermi-liquid theory, electrons in ordinary metalscan be viewed as a non-interacting electrongas. Then, as their temperature is lowered,the electrons form Cooper pairs that areresponsible for superconductivity in theBardeen–Cooper–Schrieffer (BCS) theory.On pages 729 and 736 of this issue, Mook et al.1 and Sharma et al.2 announce the resultsof two independent experiments showingthat the conventional theories of metals andsuperconductors are no longer universallyvalid. At the same time, both works areinspired by and add further credibility to analternative picture of electrons in solidsknown as ‘dynamical stripes’.

The first hints of a challenge to the estab-lished paradigm came with the discovery ofthe high-temperature superconductors in1987. Soon after their arrival, suspicionsarose that something completely differentwas going on in these copper oxide super-conductors3. But the conventional theoriesproved flexible and were defended quite suc-cesfully. Mook et al.1 and Sharma et al.2 havenow used different experimental probes(neutron scattering and ion channelling)aimed at dissimilar properties of the copperoxides (spin and ion-lattice fluctuations) todisqualify, once and for all, these theories asan explanation for high-temperature super-conductivity.

The state of electrons in normal metalsshould be understood as an extreme expres-sion of the perpetual motions of quantum

physics. In Fermi-liquid theory, the metallicstate of matter corresponds to a quantum gasof ‘quasiparticles’, which are like electrons inevery way except that they no longer interact.The quasiparticles are fermions and have toobey the Pauli exclusion principle, which for-bids any two fermions from being in the samestate. This forces the electrons into states ofhigh quantum kinetic energy and, remark-ably, the fierce quantum motions wash awaycompletely the inter-actions between elec-trons. Conventional superconductivity is asibling of this state. In the presence of anyresidual attractive interaction the quasiparti-cles form pairs and, because these Cooperpairs are bosons (and so are not constrainedby the Pauli principle), they immediatelycondense into a BCS superconductor.

These quantum gases are very simple andrather featureless. The spin and lattice fluc-tuations in the copper oxides seen by Mooket al. and Sharma et al. reflect a complexity ofbehaviour in the electron system that cannotbe attributed to any gas, even a quantum gasof quasiparticles. A high degree of coopera-tivity is needed to explain these diverse

news and views

714 NATURE | VOL 404 | 13 APRIL 2000 | www.nature.com

High-temperature superconductivity

Stripes defeat the Fermi liquidJ. Zaanen

The external differences between species belie thefact that many of their genes, proteins andintracellular signalling pathways are very similar.So, for example, we can learn much aboutourselves from studying such experimentally usefulanimals as the nematode Caenorhabditis elegans.Elsewhere in this issue (Nature 404, 782–787;2000), however, René H. Medema and colleaguesturn the tables. They describe a new branch of asignalling pathway in mammalian cells that mayalso offer insight into an important developmentalstate in C. elegans.

The story starts with humanity’s fascinationwith extending life. For this reason, long-livedmutants of organisms such as C. elegans havereceived a lot of attention. By identifying the genesmutated in such organisms, we can get some ideaabout the proteins and signalling pathwaysinvolved in longevity. One pathway involves acascade of enzymes known as kinases,culminating in the regulation of DAF-16 — atranscription factor of the so-called Forkheadfamily (see diagram). This pathway also regulatesthe developmentally arrested nematode larval stateknown as ‘dauer’, which occurs as a result ofcrowding and starvation.

At the cellular level, it is possible that, for a cellto stay alive, its death (by ‘apoptosis’) must beactively suppressed. In mammals, a pathway thatrepresses apoptosis contains counterparts of manyproteins from the nematode longevity pathway (seediagram; proteins that are conserved in the twopathways are shown in the same colours). This‘survival’ pathway also ends with inhibition ofmembers of the Forkhead family, resulting in

repression of genes of the death machine.Medema et al. now show that this signalling

pathway has another, possibly more important,function: it puts a brake on the cell-division cycle.The authors find that signalling progresses alongthe now familiar enzymatic cascade, up to thepoint at which Forkhead proteins are regulated.Then, the pathway forks. One line leads tosuppression of cell death, while the other leads tocell-cycle arrest — the key molecular effect herebeing an increase in levels of the protein p27kip1.This appears to occur through increasedexpression of the gene encoding p27kip1, ratherthan through stabilization of the p27kip1 protein.

p27kip1 is a well known molecule: it inhibits theactivity of important regulators of the cell cycle,

called cyclin/cyclin-dependent kinase (CDK)complexes. Increased p27kip1 levels would beexpected to lead to cell-cycle arrest in the periodbefore DNA replication occurs. Such a block isexactly what Medema et al. find when theyactivate Forkhead proteins in mammalian cells.

This new prong of a known signalling pathway may provide an explanation for theproposed role of defects in this pathway in cancer — a condition that is essentiallycharacterized by the undermining of the finelytuned molecular network that ensures controlledcell division and finite cellular lifespan. This story may also have come full circle in explainingwhy, in nematodes, this pathway blocks larvaldevelopment. Bernd Pulverer

Signal transduction

An arresting tale

AKT-1AKT-2

AGE-1(PI(3)K)

DAF-16(Forkhead)

PKB/Akt

(PI(3)K)

Forkhead

DAF-2

Receptor

Survival/growthfactor

Insulin-likeligand

CDK

Cyclin

Apoptosis

Nucleus

Nucleus

Cellcycle

Nematode cell

Longevity

Mammalian cell

p27kip1

© 2000 Macmillan Magazines Ltd

observations, and the best interpretationsare found in the new theory of dynamicalstripes. This increasingly popular theoryrefers to a state that is radically different froma Fermi liquid. In stripes, the electrons havebeen incorporated in complex, quantumfluctuating patterns and, although much isunclear, many now believe that supercon-ductivity and stripes have a deep and pro-found relationship.

Owing to a spectacular series of discover-ies in the past few years (see ref. 4 for areview), there now exists some understand-ing of the counterintuitive nature of this‘complex quantum matter’. In many-bodysystems (so-called quantum fields), quan-tum effects might turn out to be relativelymodest at short-length scales, but increas-ingly important over larger distances. This isunlike few-body systems, where quantummechanics invariably dominates at theshortest distances. Earlier neutron-scatteringexperiments5 suggested that this generalfield-theory mechanism was also at work in the copper oxide superconductors6.Because the long-wavelength quantum fluctuations involve small energy scales, one expects to easily suppress these, drivingthe state to its classical limit. Tranquada andco-workers7 accomplished this feat in 1995,and their frozen version of stripes allowedexperimentalists to have a closer look at thenature of this complex electronic matter.

The static state of stripes is highly unusual(Fig. 1). It consists of antiferromagnetic (andpresumably insulating) domains about ananometre in width, separated by domainwalls on which the charge carriers reside. In the static phase these domain walls, orstripes, condense in a regular pattern. Suchphases were actually predicted theoretical-ly8,9 (triggering the experimental search),and the current consensus is that they arisefrom competition between quantum kineticenergy and electron–electron interactionson the microscopic scale.

How can we understand dynamical (asopposed to static) stripes? This is the stateassociated with superconductivity, in whichthe system of stripes is disordered by long-wavelength quantum fluctuations. Figure 1illustrates a popular view. It gives a snapshotat a fixed (imaginary) time from a quantumMonte Carlo calculation, and the overallquantum state should be viewed as asequence of many of these pictures, in whichthe irregular features vary according to thequantum fluctuations. Yellow and red indi-cate opposite orientations of the stripe antiferromagnet: the background would beuniformly red in the classical limit. Althoughthe stripes are still intact (lines of black dots),they undergo vigorous quantum motions. Infact, several theoretical groups are studyingthese stripe fluctuations in terms of simplestring theories10.

The superconductivity community takessuch pictures increasingly seriously, if onlybecause of their effectiveness in guiding theexperimental work. This does not mean thatthe problem of high-temperature super-conductivity is solved. Over long distancesthe quantum fluctuations get truly out ofhand, destroying the semiclassical picturepresented above. Remarkably, in this regimethere are similarities with a BCS super-conductor, although these might not be asstraightforward as some defenders of theconventional model want to believe. At themoment the relationship between stripes at short distances and BCS-like behaviour at large distances is an enigma. But I am anoptimist and I take this mystery as prime evidence that there will be a glorious theoryof physics waiting at the end of the journey. ■

J. Zaanen is at the Instituut-Lorentz for TheoreticalPhysics, Leiden University, PO Box 9506, NL-2300RA Leiden, The Netherlands.e-mail: jan@lorentz.leidenuniv.nl1. Mook, H. A. et al. Nature 404, 729–731 (2000).2. Sharma, R. P. et al. Nature 404, 736–740 (2000).3. Anderson, P. W. in The Theory of Superconductivity in the High

Tc Cuprates (ed. Treiman, S. B.) (Princeton Univ. Press, 1997).4. Zaanen, J. Science 286, 251–252 (1999).5. Aeppli, G. A. et al. Science 278, 1432–1435 (1998).6. Sachdev, S. Phys. World 12, 33–38 (1999).7. Tranquada, J. M., Sternlieb, B. J., Axe, J. D., Nakamura, Y. &

Uchida, S. Nature 375, 561–563 (1995).8. Zaanen, J. & Gunnarsson, O. Phys. Rev. B 40, 7391–7394 (1989).9. Emery, V. J. & Kivelson, S. A. Physica C 235–240, 189–192

(1994).10.Zaanen, J. Phys. Rev. Lett. 84, 753–756 (2000).

news and views

NATURE | VOL 404 | 13 APRIL 2000 | www.nature.com 715

Figure 1 A snapshot of stripes in action.Fluctuating quantum stripes (lines of black dots)are shown moving through a quantumantiferromagnet (red/yellow background) froman imaginary time slice of a quantum MonteCarlo simulation. In the classical limit thebackground would be, say, uniformly red,corresponding to a simple antiferromagnet. Herethe yellow patches represent spin fluctuations.The first conclusive evidence for the existence ofstripes in high-temperature superconductors isgiven by Mook et al.1 and Sharma et al.2. Aconsequence of this work is that the conventionalpicture of ordinary metals and superconductorscan no longer explain high-temperaturesuperconductivity. (Unpublished figure, O. Y. Osman, J. Tworzydlo and J. Zaanen.)

The transcription of DNA into RNArequires a battery of proteins, the pre-cise complement of which depends on

the circumstances. One group of such pro-teins is the basic helix–loop–helix (bHLH)family of DNA-binding transcription fac-tors. So far, over 250 of these bHLH factorshave been characterized; they regulate geneexpression in various processes such as muscle development, circadian rhythms andcell growth1. In a paper in Chemistry andBiology2, Winston and Gottesfeld use a novelcombinatorial approach to provide newinformation about the bHLH DNA-bindingequipment. Importantly, this approach canbe easily adapted for high-throughput studies of structure–function relationshipsfor any protein.

Basic helix–loop–helix transcription fac-tors were first identified by Baltimore andco-workers3 over ten years ago. The con-served amino acids in the bHLH family arepresent in two a-helices (helix 1 and helix 2,respectively), which are separated by a loopregion of variable length and amino-acid

composition4,5. The bHLH domain of sever-al transcription factors is necessary and sufficient for the required dimerization ofthe protein, and subsequent recognition ofsix specific nucleotides in the target DNAsequence. Residues in the amino terminus of helix 1 provide sequence-specific DNArecognition, whereas dimerization is largelymediated by helix 2 and the carboxy termi-nus of helix 1 (Fig. 1, overleaf).

What Winston and Gottesfeld2 have doneis to use a combinatorial strategy to look atthe role of the loop region in a DrosophilabHLH factor, Deadpan, in sequence-specificDNA recognition. Using solid-phase peptidesynthesis methods and high-resolution massspectrometry, they have determined theminimal loop requirements necessary toretain full DNA-binding activity. Further-more, they have identified a loop residue that has a key part in binding DNA throughelements that lie outside the core-DNArecognition sequence.

The studies of Winston and Gottesfeldinvolved parallel approaches to generate

Functional genomics

Recognizing DNA in the librarySatish K. Nair and Stephen K. Burley

© 2000 Macmillan Magazines Ltd