the spins under the contact randomize and spin polarization is quenched. A measurement of the difference between the voltage at zero magnetic field and at a large magnetic field provided the contribution of the injected spin polarization to the signal at the ferromagnetic contact3. The strength of the magnetic field required to quench the injected spin polarization pro-vided a measure of the spin coherence time (as well as the distance over which the electrons remain polarized during that time). Control experiments, which included destroying the spin polarization of the current flowing into the silicon (by having ytterbium in the barrier) or modifying the barrier between the contact and the silicon (by adding caesium), led to negligible spin-injection amplitudes. This demonstrates that the carefully designed contact and barrier are crucial for high-efficiency spin injection.

Dash and colleagues achieved successful spin injection in both ‘electron-doped’ and ‘hole-doped’ silicon — the two constituents of complementary metal-oxide-semiconductor (CMOS) technology used in most conventional microelectronic circuitry. One might expect that the resulting spin coherence times measured in doped silicon would correspond to those meas-ured by other techniques. However, the times reported by the authors — 140 picoseconds for electron-doped silicon and 270 picosec-onds for hole-doped silicon — are surprisingly short. By comparison, gallium arsenide, which has a spin–orbit interaction tenfold larger than that of silicon, has a room-temperature spin coherence time only threefold smaller10, about 50 picoseconds. This disconcerting result does not necessarily preclude the use of silicon for spintronic devices, for the distance over which the electrons remain polarized during these times exceeds a couple of hundred nanometres, which is much larger than the expected sizes of devices in modern semiconductor chips. However, it is a surprising result that may require a rethink about the mechanisms of spin decoherence in silicon.

Because of the ubiquitous nature of silicon in modern semiconductor electronics, the demon-stration of semiconductor spintronic function-ality in silicon at room temperature promises to be a major breakthrough. An observation of room-temperature spin transport between two contacts in silicon, in addition to the injection and detection demonstrated in a single con-tact by Dash et al., would be a welcome next step. Initial applications may include using spin injection and detection to enhance the performance of predominantly charge-based devices. However, to dramatically reduce the power consumption of modern electron-ics below the fundamental limit1, additional advances would be required, especially the control of spin orientation by other means than a magnetic field. ■

Michael E. Flatté is in the Department of Physics

and Astronomy, The University of Iowa,

Iowa City, Iowa 52242, USA.

e-mail: michael_flatte@mailaps.org

STRUCTURAL BIOLOGY

Highly charged meetingsAnthony G. Lee

When it comes to proteins and their environments, opposites repel. So how is the highly charged, polar helix of a transmembrane ion channel accommodated by a non-polar membrane? Easily, if the charges are buried.

Early in any biochemistry course, students are told that charged amino acids are not happy in hydrophobic (water-repelling) environments. Because the basic unit of biological membranes — the lipid bilayer — has a hydrophobic core, it follows that the α-helices of membrane-bound proteins should rarely contain charged amino acids. But there are exceptions, of which voltage-gated potassium channels form one class. On page 473 of this issue, Krepkiy et al.1 show that, contrary to textbook teachings, the highly charged α-helix present in these ion channels is fully compatible with a normal lipid bilayer.

Voltage-gated potassium channels are

homotetramers — they assemble from four identical monomers, each of which contains six membrane-spanning α-helices. Two of the helices (S5 and S6) from each monomer come together in the tetrameric structure to form the pore through which potassium ions move across the membrane (Fig. 1). The remaining helices (S1 to S4) form voltage sensors, one for each monomer. A vital role is played by helix S4, which contains four or five positively charged amino-acid residues. It is these positive charges that allow the channel to sense a change in electrical potential across the membrane; subsequent movement of S4 leads to opening of the channel. But how can such a charged

1. Landauer, R. IBM J. Res. Dev. 5, 183–191 (1961).2. Awschalom, D. D., Loss, D. & Samarth, N. (eds)

Semiconductor Spintronics and Quantum Computation (Springer, 2002).

3. Dash, S. P., Sharma, S., Patel, R. S., de Jong, M. P. & Jansen, R. Nature 462, 491–494 (2009).

4. Kikkawa, J. M. et al. Science 277, 1284–1287 (1997).

5. Wolf, S. A. et al. Science 294, 1488–1495 (2001).

6. Awschalom, D. D. & Flatté, M. E. Nature Phys. 3, 153–159 (2007).

7. Lou, X. et al. Nature Phys. 3, 197–202 (2007).8. Appelbaum, I., Huang, B. & Monsma, D. J. Nature 447,

295–298 (2007).9. van ‘t Erve, O. M. J. et al. Appl. Phys. Lett. 91, 212109 (2007).10. Meier, F. & Zachachrenya, B. P. (eds) Optical Orientation

(North-Holland, 1984).

S2

S1

S4

S3

Voltage

sensor

Figure 1 | Structure of a voltage-gated potassium channel. Voltage-gated potassium channels form pores in cell membranes through which potassium ions pass. The channel depicted here is a Kv2.1/Kv1.2 chimaeric channel9, viewed as though looking down on the membrane from the extracellular side (membrane not shown). Three of the four subunits are shown as surface plots in different shades of green, whereas the fourth subunit is shown in cartoon format. Most of the α-helices in the fourth subunit are depicted as red cylinders, but the first five positively charged amino-acid residues in helix S4 are shown in space-filling format. A potassium ion moving through the central pore is shown in purple. The channel has four voltage sensors (each composed of helices S1–S4) that are loosely attached to the central pore. The charged residues on S4 point in towards the other helices of the voltage sensor, and most are located in a water-filled cavity (not shown). Krepkiy and colleagues1 report that this means that the highly charged voltage sensor is fully compatible with the lipid bilayer that surrounds it in the native membrane.

420

NATURE|Vol 462|26 November 2009NEWS & VIEWS

417-423 News and Views NS.indd 420417-423 News and Views NS.indd 420 20/11/09 17:35:3220/11/09 17:35:32

© 2009 Macmillan Publishers Limited. All rights reserved

helix be accommodated in the hydrophobic environment of a lipid bilayer?

Crystal structures of voltage-gated potas-sium channels have shown that the voltage sensors are only loosely attached to the central pore (Fig. 1, overleaf). In the first published structure2, which is recognized to be a distor-tion of the naturally occurring structure, the positive charges in S4 are partly exposed on the outer surface of the voltage sensor and are possibly in contact with the lipid bilayer. To see what effect such an exposed helix might have on a membrane, a computer model was gener-ated of an isolated S4 helix in a lipid bilayer3. The model showed that the helix greatly dis-torts the bilayer, leading to the formation of hydrogen-bonded networks of water and lipid phosphate groups about each charged residue in the helix. The model also showed that the thickness of the bilayer’s hydrophobic core close to the helix falls sharply from its normal value of 27 Å to about 10 Å. Such effects would be very unusual for a protein in a membrane because of the high energetic cost of distort-ing the bilayer4, and because the hydrophobic match between an undistorted lipid bilayer and a membrane protein is usually rather good5.

But does a complete voltage sensor, made up of helices S1 to S4, have the same effect on a lipid bilayer as an isolated S4 helix? The answer turns out to be no. Using neutron-diffraction techniques, Krepkiy and colleagues1 studied the properties of the bilayer surrounding a voltage sensor from a bacterial ion channel. They show that the bilayer remains intact and

that it is only about 3 Å thinner than its normal thickness.

The authors’ neutron-diffraction experi-ments measured the average thickness of the bilayer across the whole membrane plane. It is therefore possible that the bilayer close to the voltage sensor is thinner than the measured average. To find out whether this is the case, Krepkiy et al. performed molecular-dynamics simulations of their system. These models did indeed suggest that the thickness of the bilayer decreases by as much as 9 Å close to the volt-age sensor. This conclusion must be treated with caution, however, because the authors’ simulation also predicted that the unperturbed bilayer is about 4 Å thicker than the experi-mentally determined thickness6 (such models often overestimate the thickness of the bilayer). An earlier, coarse-grained molecular-dynamics simulation suggested that the effects of voltage sensors on bilayer thickness are small7.

The crystal structure of the potassium chan-nel shown in Figure 1 suggests a simple reason for the small effect that potassium-channel volt-age sensors have on membranes: the charges on helix S4 are not actually exposed to the lipid bilayer. Instead, they are buried within the sensor structure, either occupying water-filled cavities or interacting with negatively charged residues in the S2 helix8. To confirm this, the authors performed further experiments that showed that the presence of the voltage sensor results in no measurable change in the amount of water in the lipid bilayer’s core. The sensor is hydrated, however, as revealed by Krepkiy

and colleagues’ nuclear magnetic resonance studies1. The water molecules are probably located in crevices in the sensor, where they can hydrate some of the positively charged resi-dues in S4, ensuring that these residues remain charged and so able to detect changes in poten-tial across the membrane.

Overall, Krepkiy and colleagues’ study is rather comforting — maybe the exceptions to the rules taught to biochemistry students aren’t really exceptions after all. As with other mem-brane-bound proteins5, voltage-gated potas-sium channels must have evolved so that the packing preferences of the helices in the volt-age sensor cause the sensor to adopt a structure that nicely matches that of the surrounding lipid bilayer. In this way, the hydrophobic lipid and the hydrophilic, charged sensor can meet without either having to change very much. ■Anthony G. Lee is in the School of Biological

Sciences, University of Southampton,

Southampton SO16 7PX, UK.

e-mail: agl@soton.ac.uk

1. Krepkiy, D. et al. Nature 462, 473–479 (2009).

2. Jiang, Y. et al. Nature 423, 33–41 (2003).

3. Freites, J. A., Tobias, D. J., von Heijne, G. & White, S. H.

Proc. Natl Acad. Sci. USA 102, 15059–15064 (2005).

4. Marsh, D. Biophys. J. 94, 3996–4013 (2008).

5. Lee, A. G. Biochim. Biophys. Acta 1612, 1–40 (2003).

6. Nagle, J. F. & Tristram-Nagle, S. Biochim. Biophys. Acta

1469, 159–195 (2000).

7. Bond, P. J. & Sansom, M. S. P. Proc. Natl Acad. Sci. USA 104, 2631–2636 (2007).

8. Chakrapani, S., Cuello, L. G., Cortes, D. M. & Perozo, E.

Structure 16, 398–409 (2008).

9. Long, S. B., Tao, X., Campbell, E. B. & MacKinnon, R. Nature

450, 376–382 (2007).

ASTROPHYSICS

Assortment in the GalaxyJudith G. Cohen

Observations of star clusters in the Milky Way defy the view that the constituents of these systems are almost invariably chemically alike. The outlying clusters could be the tattered relics of once larger systems.

In its halo of dark matter, our Galaxy hosts a family of about 150 globular star clusters (GCs). Conventional wisdom holds that they are compact, roughly spherical systems of high stellar density, each containing about 5 million stars held together by gravity. Undergradu-ates are taught that these classic laboratories for studying stellar evolution each contain a single population of stars of uniform age and chemical composition. More than 30 years ago, it became clear that the most luminous of these clusters, ω Centauri, was the exception to the rule: the system contains stars with a range of iron abundances (Fe metallicity) that vary by more than a factor of 30 (refs 1, 2). In this issue, Lee et al.3 (page 480) and Ferraro et al.4 (page 483) report the discovery of two other GCs that harbour stars containing different

proportions of iron and other heavy elements.Variations among the light elements within

individual GCs were also discovered several decades ago. But, unlike heavy elements, light elements can be made during the course of normal stellar evolution in intermediate-mass stars through the fusion of hydrogen at high temperatures5. The resultant ‘ash’ could be mixed into the surfaces of these evolved stars, ejected by gentle winds and then mixed into the gas in the young cluster. A second generation of stars could then be formed, giving rise to the observed variation in light-element content within a GC.

Heavier elements — including calcium, iron and beyond — are mostly produced in stellar explosions known as supernovae. Because material is violently ejected from supernovae

at a very high velocity and the gravitational binding energy of present-day GCs is low, in the current conditions supernova gas ejecta would escape from the cluster. The only way in which such energetic gas, rich in heavy elements, could have been retained would be if the mass of the GC was much higher in the past than is typical today. If we find a GC showing variations in those heavy elements, suspicion naturally arises that it is the remnant of a for-merly accreted small galaxy, as was suggested for ω Centauri. This is not wild speculation; there are indications that the GC called M54 will probably be the only remnant structure from the Sagittarius dwarf galaxy to survive the galaxy’s ongoing violent disruption by the Milky Way.

We now have much better tools with which to search for variations in age and elemental abundance within individual Galactic GCs. These tools operate at a level of accuracy that we could only dream of a decade ago. Lee et al.3 demonstrate definitively that there is a spread in the abundance of calcium within the massive GC M22, which has been a suspect for many years. They find that the population of red-giant branch stars — stars in which the core has ceased to burn hydrogen but the outer shell is still doing so — in the system splits

421

NATURE|Vol 462|26 November 2009 NEWS & VIEWS

417-423 News and Views NS.indd 421417-423 News and Views NS.indd 421 20/11/09 17:35:3420/11/09 17:35:34

© 2009 Macmillan Publishers Limited. All rights reserved