large tilt of transmembrane helix 6 (TM-VI) and a smaller motion of helix 7 (TM-VII) at the inner surface of the cell membrane. These movements open a cleft that presents the ligand-binding site to the G protein. Scheerer and colleagues’ results broadly match those of earlier biophysical studies8, and agree quanti-tatively with those of work9 that used pairs of ‘spin labels’ — which contain unpaired elec-trons that can be detected spectroscopically — to map out helix movement in activated rhodopsin. On the other hand, the authors’ structure contrasts with a previously reported crystal structure10 of activated rhodopsin (in the absence of a G protein) that differed little from the inactive state. Given the extensive biochemical and biophysical data in support of Scheerer and colleagues’ structure for the activated state, it now seems likely that the pre-viously reported structure10 was modified by the conditions used to stabilize the crystals.

It might seem strange that the activating ligand, all-trans retinal, is not present in Scheerer and colleagues’ opsin complex. But it is well established that active conformations of unliganded receptors, including opsin11, exist in equilibrium with the inactive state. Indeed, 7TM receptors are often observed signalling with high activity through G proteins in the absence of agonist ligands. The crucial point in Scheerer and colleagues’ structure2 is that the observed active form is stabilized by a large excess of the G-protein fragment (the Gα peptide). It is the presence of this peptide that identifies the structure as a true active conformation. What is more surprising is that, in the absence of both all-trans retinal and the Gα peptide, opsin crystallizes in the same conformation as the authors report here (as reported earlier this year12 by the same group).

The X-ray structure and orientation of the bound Gα peptide2 are almost identical to those previously determined by nuclear mag-netic resonance (NMR) studies13 on solutions of a similar peptide interacting with activated rhodopsin. But Scheerer and colleagues’ struc-ture now shows how the peptide is stabilized by interactions with specific amino-acid residues that are exposed by the movement of TM-VI. Of particular note is that the peptide makes direct contact with the most evolutionarily con-served residue in 7TM receptors, an arginine residue at the intracellular end of TM-III.

The mechanism underlying the molecular recognition between opsin and the Gα peptide will probably be a good model for the inter-actions between 7TM receptors and the α-sub-units of G proteins in general. The selectivity of these interactions is largely determined by the five carboxy-terminal residues of the α-subunits14. But it has been unexpectedly difficult to identify the corresponding selec-tivity-determining footprint on the receptors. The structure of Scheerer and colleagues’ com-plex will therefore undoubtedly spur a wave of studies aimed at understanding this recogni-tion process in detail. Moreover, interactions

of 7TM receptors with G proteins, and their subsequent activation processes, are known to involve the recognition of additional struc-tural elements on both the receptor and the G protein (Fig. 1). These processes can now be investigated in structure-based studies, as suggested by Scheerer and co-workers2.

The observed conformation of activated opsin on the intracellular side of the cell mem-brane probably represents the activated state for 7TM receptors in general. But the relatively minor conformational changes that occur in the ligand-binding domain — which lies between the extracellular segments of the transmem-brane helices — is likely to be unique to opsin. In rhodopsin, part of the extracellular region of opsin forms a protein ‘plug’ that fills the entrance to the main ligand-binding pocket, preventing movement of the transmembrane helices (Fig. 1). In contrast, in the previously determined structures of adrenergic recep-tors5–7, the main ligand-binding crevice forms an open funnel towards the extracellular space. This allows ligands to diffuse in and out of the binding site, and could also allow inward movement of the extracellular segments of the transmembrane helices so that they sur-round the ligand. Such movement might aid ligand binding, and form part of a global ‘toggle-switch’ mechanism of receptor activation8. To test this possibility, we will need structures

of 7TM receptors in complex not only with the G protein, but also with small-molecule agonists. ■

Thue W. Schwartz is at the Laboratory for Molecular Pharmacology, University of Copenhagen, Blegdamsvej 3, 2200 Copenhagen, Denmark. Wayne L. Hubbell is at the Jules Stein Eye Institute and the Department of Chemistry and Biochemistry, University of California, Los Angeles, Los Angeles, California 90095, USA. e-mails: tws@sund.ku.dk; hubbellw@jsei.ucla.edu

1. Pierce, K. L., Premont, R. T. & Lefkowitz, R. J. Nature Rev. Mol. Cell Biol. 3, 639–650 (2002).

2. Scheerer, P. et al. Nature 455, 497–502 (2008).3. Schertler, G. F. Curr. Opin. Struct. Biol. 15, 408–415 (2005).4. Rasmussen, S. G. F. et al. Nature 450, 383–387 (2007).5. Rosenbaum, D. M. et al. Science 318, 1266–1273 (2007).6. Cherezov, V. et al. Science 318, 1258–1265 (2007).7. Warne, T. et al. Nature 454, 486–491 (2008).8. Schwartz, T. W., Frimurer, T. M., Holst, B., Rosenkilde,

M. M. & Elling, C. E. Annu. Rev. Pharmacol. Toxicol. 46, 481–519 (2006).

9. Altenbach, C., Kusnetzow, A. K., Ernst, O. P., Hofmann, K. P. & Hubbell, W. L. Proc. Natl Acad. Sci. USA 105, 7439–7444 (2008).

10. Salom, D. et al. Proc. Natl Acad. Sci. USA 103, 16123–16128 (2006).

11. Woodruff, M. L. et al. Nature Genet. 35, 158–164 (2003).12. Park, J. H. et al. Nature 454, 183–187 (2008).13. Koenig, B. W. et al. J. Mol. Biol. 322, 441–461 (2002).14. Conklin, B. R., Farzel, Z., Lustig, K. D., Julius, D.

& Bourne, H. R. Nature 363, 274–276 (1993).

Competing financial interests: declared (see online articlefor details).

SOLID-STATE PHYSICS

New order for magnetismEiji Saitoh

Physicists have come up with an innovative way of manipulating the direction of magnetization in a solid. The approach might be used to make low-power-consumption computer memory devices.

On page 515 of this issue, Chiba et al.1 show how the direction of magnetization in a solid can be controlled electrically. The significance of this achievement lies in the fact that mag-netic data-storage and memory devices, such as hard disks, depend on manipulating the direction of magnetization. But methods for doing this have so far been inefficient.

The fundamental property of a magnet is that, in a magnetic field, it aligns in the direc-tion of the field (Fig. 1a). A magnet consists of atomic-scale magnets, called spins, and if a strong magnetic field is applied to a fixed mag-net, the spins rotate to align with the field. As a result, the magnetization in the magnet — the average of the spin-magnetic moment — takes on the direction of the applied magnetic field (Fig. 1b). Magnetic disks exploit this princi-ple. In a hard disk, for instance, information is stored on a disk-shaped magnet in the form of patterns of local magnetization and, to write information, a pulse of current is applied to a

small electromagnet that scans the disk. But this method is indirect: a magnetic field exists between the current and the manipulated magnetization, a situation that wastes energy.

A promising alternative would be to apply a spin-polarized current directly to the mag-net instead of using current to generate the magnetic field (Fig. 1c). Such a current would exert torque on the magnetization by exchanging spin-angular momentum with it as it passes through the magnet. Exploitation of this phenomenon, called spin torque2, is expected to allow the development of compact magnetic memory devices that can run on low power consumption. The use of multiferroics3, materials in which ferroelectricity and ferro-magnetism coexist, has also been proposed. But direct control of the magnetization direction has not yet been achieved in such systems.

Chiba et al.1 demonstrate a new way of manipulating the direction of magnetization using electric voltage only. The idea behind

474

NATURE|Vol 455|25 September 2008NEWS & VIEWS

���������������� � ���������������������

N

S

M

Ferromagneticsemiconductor

InsulatorMetal

M

Magnet

M

E

N

S

N

S

N

S

N

S

N

S

c d

a b

their experiments is to control the magnetic anisotropy of a magnet and so generate pref-erential directions for the magnetization. This results from an interaction between the electron’s spin and its motion (the spin–orbit interaction), which means that the electron’s energy depends on the direction of magnetiza-tion. Such dependence is in general affected by the electronic structure — that is, the configu-ration and density of electrons in the magnet. The implication is that the magnetic anisotropy can be controlled by regulating the electronic structure.

To do this, Chiba et al. modulated the density of electrons using a metal–insulator–semicon-ductor device involving a semiconductor — a (GaMn)As film — that has ferro magnetic prop-erties at low temperatures (Fig. 1d). The device includes a ‘gate’ electrode isolated electrically from the (GaMn)As film. When a negative voltage is applied to the gate electrode, carriers in the film that have positive charge (electron ‘holes’) are attracted towards the electrode. In contrast, when a positive voltage is applied, the number of electron holes near the gate elec-trode is reduced. This property allows the den-sity of the electron holes, and thus the magnetic anisotropy in the (GaMn)As film beneath the electrode, to be controlled electrically, resulting in a change in the magnetization direction.

To determine the direction of magnetization, Chiba et al.1 used the transport properties of carriers in (GaMn)As, measuring the voltage differences at the ends of the film when an

Figure 1 | Manipulation of magnetism. a, When a magnetic field is applied, a magnet rotates to align with the field. b, The magnetization (M) in the magnet can be rotated by applying an external magnetic field. Today’s magnetic recording technology exploits this property. c, The direction of magnetization in the magnet is changed by applying a spin-polarized current (the spin-torque effect). d, Application of an electric field E rotates the direction of magnetization: Chiba et al.1 demonstrate this effect experimentally with a ferromagnetic semiconductor, (GaMn)As.

electric current flows through it. These voltage differences reflect the relative angle between the current and the magnetization direction. The authors applied a weak external magnetic field to the (GaMn)As film and measured the voltage while changing the direction of the field.

They found that the magnetization direction in the film indeed deviates from the direction of the magnetic field. Moreover, the strength of this deviation varies with the carriers’ density and the voltage applied to the gate electrode. The authors attribute this deviation to the mag-netic anisotropy, and conclude that their result is evidence for voltage-induced modulation of the magnetic anisotropy. From the estimated anisotropy, and in the absence of an external magnetic field, the magnetization direction rotates 10° when the gate voltage changes from –12 volts to about 9 volts. In other words, electric manipulation of the magnetization direction is achieved.

The method proposed by Chiba et al.1 does not require currents flowing in the device, but only the application of voltage. It is thus highly compatible with the existing metal-oxide semi-conductor technology used in microproces-sors and random-access memory devices. The principle involved here — that magnetic aniso-tropy depends on electronic structure — is also applicable to a wide range of materials, such as the surfaces of metallic or metal-oxide ultra-thin films4. Magnetic random-access memory devices made of metallic magnets are now being produced, and it is certainly worth exploring the possibility of electric control of magnetization also in metallic systems. ■

Eiji Saitoh is in the Department of Applied Physics and Physico-Informatics, Keio University, 3-14-1 Hiyoshi, Yokohama 223-8522, Japan.e-mail: saitoheiji@appi.keio.ac.jp

1. Chiba, D. et al. Nature 455, 515–518 (2008). 2. Slonczewski, J. C. J. Magn. Magn. Mater. 159, L1 –L7 (1996). 3. Tokura, Y. Science 312, 1481–1482 (2006).4. Weisheit, M. et al. Science 315, 349–351 (2007).

HEARING

Route to authentic hair cells Mats Ulfendahl

Existing therapies for hearing defects are generally ineffective in severe forms of deafness. A technical feat that generates sound-sensing hair cells in the inner ear of mice might have long-term potential.

Impaired hearing affects the communication abilities of approximately 17% of the US adult population1. Hearing defects of the outer or middle ear are generally treatable. By contrast, defects in the less accessible regions, such as the cochlea (the sound-sensing system of the inner ear) or the auditory nerve, are often permanent and have been considered untreatable. Such impairment, which is usually caused by exces-sive noise, ageing or genetic factors, involves degeneration of the mechanosensory receptors of the hearing organ — the inner and outer hair cells. On page 537 of this issue, Gubbels and colleagues2 demonstrate that gene transfer

to the inner-ear precursor cells of developing mouse embryos can be used to induce the formation of functional hair cells.

For decades, the only therapeutic option for hearing impairment was the use of hear-ing aids. These devices amplify and filter the incoming sound, improving the hearing threshold. But even with their use, hearing is far from normal, as the ear’s receiving elements — the sensory hair cells — still do not func-tion adequately. The introduction of cochlear implants in the 1980s was a big step forward, as they completely bypass the defective sen-sory cells and electrically stimulate the next

475

NATURE|Vol 455|25 September 2008 NEWS & VIEWS

���������������� � ���������������������