different polarizations without destroyingthe interference11,12. Perhaps the best solu-tion is to use the waves that make up the pulseitself as the clock to measure the delay9. Bymodulating (for example) the size of the barrier, one can distinguish clearly betweenpulses that pass through much faster than the time over which the barrier oscillates and those that pass through so slowly that the barrier oscillates many times before they escape. By changing the frequency of the barrier modulation one can then time the tunnelling process.

Alternatively, the amount of rotation ofspins or dipoles could be used as a clock,where the rotation is caused by the spin13 ordipole14 tunnelling through a region of mag-netic field (the Larmor effect) or electric field(the Faraday effect). As such a clock hasnothing to do with the shape of the risingedge of the wave packet, it might give a moremeaningful measure of the tunnelling time.Experiments of this kind would provide a key to the question of whether superluminaltunnelling is an important development, orjust a misnomer.

On the other hand, in a practical sense,the issue is moot. As a scientist has pointedout, with reference to the commercializationof the tunnelling devices as fast links in optical and microwave transmission, it doesnot matter what we think: people are sellingthem already. The presumption in such commercial ventures is that even short-range, inefficient, repetitive transmissionmay help to provide increased speed in somemanipulations of pulses or particles in futurecomputing or communication. From thepoint of view of physics, however, it seemsthat the class of such applications will benarrow at best. ■

Markus Büttiker is in the Department of TheoreticalPhysics, University of Geneva, 24 quai ErnestAnsermet, CH-1211 Geneva 4, Switzerland.e-mail: buttiker@karystos.unige.chSean Washburn is in the Department of Physics and Astronomy, Applied and Materials Sciences,University of North Carolina, Chapel Hill, North Carolina 27599-3255, USA.e-mail: sean@physics.unc.edu1. Winful, H. G. Phys. Rev. Lett. 90, 023901 (2003).

2. Steinberg, A. M., Kwiat, P. G. & Chiao, R. Y. Phys. Rev. Lett. 71,

708–711 (1993).

3. Nimtz, G., Enders, A. & Spieker, H. J. Phys. I (France) 4,

565–570 (1994).

4. Spielmann, C., Szipöcs, R., Stingl, A. & Krausz, F. Phys. Rev.

Lett. 73, 2308–2311 (1994).

5. Ranfagni, A., Mugnai, D., Fabeni, P. & Pazzi, G. P. Appl. Phys.

Lett. 58, 774–776 (1991).

6. Japha, Y. & Kurizki, G. Phys. Rev. A 53, 586–590 (1996).

7. Sommerfeld, A. Phys. Z. 8, 841–842 (1907).

8. Brillouin, L. Ann. Phys. 44, 203–240 (1914).

9. Büttiker, M. & Landauer, R. Phys. Rev. Lett. 49, 1739–1742

(1982).

10.Büttiker, M. & Thomas, H. Ann. Phys. (Leipz.) 7, 602–617 (1998).

11.Deutch, J. M. & Low, F. E. Ann. Phys. 288, 184–202 (1993).

12.Steinberg, A. M., Kwiat, P. G. & Chiao, R. Y. Fund. Phys. Lett. 7,

223–239 (1994).

13.Büttiker, M. Phys. Rev. B 27, 6178–6188 (1983).

14.Gasparian, V., Ortuño, M., Ruiz, J. & Cuevas, E. Phys. Rev. Lett.

75, 2312–2315 (1995).

It’s often difficult to tell when two people aresiblings — despite their common originsand upbringing, they are usually very

different. Similarly, it is not always obviouswhen distinct cell types have been producedfrom the same precursor cells. For instance,would you have guessed that mammaryglands, sweat glands and hair follicles alldevelop from the same discrete accumu-lation of stem cells resting in the primitive epidermis, the outermost cell layer of anembryo? The first step in the production ofdifferent cell types from these stem cells is theformation of an epithelial bud, a small clusterof cells that bulges out from the epidermis(Fig. 1). On page 317 of this issue, Jamora et al.1 describe how they used hair-follicledevelopment as a model system to show thatdecreased production of the adhesion factorE-cadherin — known for its role in mediatingcell–cell contacts — is key for epithelial-buddevelopment. Moreover, this downregulationof E-cadherin requires the sequential action of two signals, Wnt and noggin, that were previously thought to be unlinked.

During development, the formation ofan epithelial bud is the first evidence thatsome stem cells of the primitive epidermishave been instructed to form a hair follicle,sweat gland or mammary gland. To reachthis new spatial arrangement, epithelial cells must first change the way in which they interact with their neighbours, andbreak their attachments to the underlyingmatrix. Only then can they move about totake up a new rearrangement and overall tissue shape. Cells achieve this by adjustingtheir repertoire of cell–matrix and cell–cell adhesion molecules.

The molecules that instruct epidermalstem cells to form an epithelial bud have long been sought2. For hair follicles, over-whelming evidence points to a role for proteins in the cell nucleus that regulate geneexpression (transcription factors), especiallythe Lef1 transcription complexes2–4. Indeed,forced expression of Lef1 in mice results inthe formation of extra hair follicles, andthese can even develop from tissues that arenormally hairless. By contrast, loss of theLef1 gene results in severely impaired mam-mary gland, tooth and hair development. So what might activate Lef1 in the stem cells and, equally importantly, how does Lef1 then promote follicle formation? Weknow that pathways triggered by the Wntprotein can switch on Lef1 complexes in

other contexts, through stabilization of the Lef1-activating protein �-catenin, but is thisrelevant for follicle development? Jamora et al.1 show that it is. They find that, inresponse to Wnt, Lef1 transcription com-plexes downregulate the expression of the E-cadherin gene in epithelial buds, and soprovide a direct link between Lef1 and thecell’s adhesion machinery. Furthermore, theauthors challenge the view that Lef1 com-plexes always act as transcriptional activatorsin response to Wnt signalling3; they find thatWnt proteins, via �-catenin, instead causeLef1 to act as a transcriptional repressor.

A further twist to this tale is that Wnt sig-nalling alone is not sufficient for Lef1 activa-tion. So what might be the second signal thatacts with Wnt? There are several candidates:hair-follicle development also requires thecoordinated expression of members of vari-ous protein families — the fibroblast growthfactor, bone morphogenetic protein (BMP),Sonic hedgehog and Notch families, as well

272 NATURE |VOL 422 | 20 MARCH 2003 | www.nature.com/nature

Developmental biology

A hairy situationYann Barrandon

A key step in hair-follicle development is the rearrangement of epithelial stemcells. It seems that reduced production of an adhesion protein, through theconcerted action of two signalling pathways, is crucial for this process.

Wnt noggin

β-catenin BMP

Lef1

E-cadherin gene

Decreased cell adhesion

Nucleus

Bud

Figure 1 Making a bud. The formation of anepithelial bud from the epidermal layer of anembryo (top part of figure) is an early step in the development of mammary glands, hairfollicles and sweat glands. Jamora et al.1 havefound that decreased expression of the adhesionprotein E-cadherin is needed for bud formation.The darker green indicates high E-cadherinexpression (epidermis); the lighter greenindicates low E-cadherin expression (bud). The lower part of the figure shows the signallingpathways that Jamora et al. have found to lead to decreased E-cadherin expression.

news and views

© 2003 Nature Publishing Group

as several transcription factors2. Moreover, hair-follicle development is severely impairedwhen the noggin gene, which encodes a BMPinhibitor, is deleted5. Jamora and colleaguesnow link BMP and Wnt signalling by showingthat activation of the Wnt pathway results in the formation of �-catenin-activated Lef1transcription complexes — and consequentlyin decreased E-cadherin expression — onlywhen combined with the inhibition of BMPsignalling by noggin1.

So is the downregulation of E-cadherinindeed necessary for hair-follicle formation?Cadherins are calcium-dependent proteinsthat mediate cell–cell adhesion, often as partof complexes called adherens junctions, andthey are known to be involved in develop-ment and tissue maintenance6. In the skin, E-cadherin is expressed in the developing epi-dermis (whereas P-cadherin becomes pre-dominant later in the growing hair follicle)7.Previous work has shown that the loss of some intracellular cadherin-binding proteinsseverely impairs hair-follicle formation4,8, butno one had looked at whether E-cadherinitself is required in this process. Jamora et al.now test this and show that the forced expres-sion of E-cadherin in the developing skinresults in increased adhesiveness and abnor-mal epithelial buds. So, the loss of E-cadherin(and, by implication, changes in adherens

junctions) appears to be crucial for epithelialbuds to begin forming. This supports a modelin which downregulation of the E-cadheringene in stem cells that receive noggin and Wntsignals facilitates cell movement, rearrange-ment and bud formation. It is also consistentwith previous observations that the loss of E-cadherin expression or function can lead toenhanced invasive behaviour of other cells6,9.

Together with previous work10, the resultsof Jamora and colleagues provide strong evidence that dynamic changes in the com-position of adherens junctions are importantfor the development of skin appendagessuch as hair follicles. The results may also be relevant to the development of other tissuesand organs from epithelial buds (for example,teeth, lungs and limbs). That said, manyquestions remain unanswered. For instance,the pelage follicles (guard hairs) that occur inmice develop in a Lef1-independent manner.Yet the new results imply that downregu-lation of E-cadherin expression is commonto all developing hair follicles. This hints that other transcription regulators can alsorepress E-cadherin expression in epithelialcells during follicle formation (and we knowof good candidates, such as members of theSnail family11,12). In addition, epithelial budsdo not develop randomly, but follow a precisegeometrical pattern. What governs this?

Another important question is whatdetermines whether an epithelial bud con-tributes to a hair follicle, sweat gland ormammary gland. Here there are likely to bespecies-specific mechanisms, because weknow, for example, that hair follicles andsweat glands form in mutually exclusiveregions in mice, whereas in humans both celltypes can develop from neighbouring stemcells. But are these fates specified early in budformation or later during bud expansion?The choices an epithelial bud makes areclearly intricate and await investigation. ■

Yann Barrandon is at the School of Life Sciences,Swiss Federal Institute of Technology (EPFL), and the Department of Experimental Surgery, Vaud University Hospital (CHUV), 1011 Lausanne, Switzerland.e-mail: yann.barrandon@epfl.ch 1. Jamora, C., DasGupta, R., Kocieniewski, P. & Fuchs, E. Nature

422, 317–322 (2003).

2. Fuchs, E. & Raghavan, S. Nature Rev. Genet. 3, 199–209 (2002).

3. Eastman, Q. & Grosschedl, R. Curr. Opin. Cell Biol. 11, 233–240

(1999).

4. Huelsken, J. et al. Cell 105, 533–545 (2001).

5. Botchkarev, V. A. et al. Nature Cell Biol. 1, 158–164 (1999).

6. Takeichi, M. Science 251, 1451–1455 (1991).

7. Hirai, Y. et al. Development 105, 271–277 (1989).

8. Vasioukhin, V. et al. Cell 104, 605–617 (2001).

9. Gumbiner, B. M. J. Cell Biol. 148, 399–403 (2000).

10.Perez-Moreno, M., Jamora, C. & Fuchs, E. Cell 112, 535–548

(2003).

11.Cano, A. et al. Nature Cell Biol. 2, 76–83 (2000).

12.Batle, E. et al. Nature Cell Biol. 2, 84–89 (2000).

NATURE | VOL 422 | 20 MARCH 2003 | www.nature.com/nature 273

Meteorology

Getting the wind up

news and views

SEA

WIF

S P

RO

JEC

T; N

ASA

/GO

DD

AR

D S

FC; O

RB

IMA

GE

Once research into global changetook off on a large scale, fewphenomena related to climateescaped measurement. Researchershave dragged unwieldy equipment upmountains in the tropics to drill intothe glaciers found there. And thedeep ocean has been probed byfleets of sophisticated floats. Butwind speeds in the eyewall ofhurricanes, where the storm is at its most intense, have long defiedobservation. On page 279 of thisissue, however, Mark Powell and his colleagues present just such

measurements (Nature 422,279–283; 2003).

The data — 331 vertical windprofiles from 15 tropical cyclones —were obtained by dropwind sondesfrom the Global Positioning System.The sondes are launched fromaircraft at altitudes of 1.5–3 km;they relay their position and measurepressure, temperature and humidityas they fall through the atmosphereand drift with the wind. Tropicalcyclones — such as HurricaneAlberto, which occurred in August2000 and is shown in the satellite

image here — derive much of theirenergy from sea–air exchange. Forforecasting purposes it is essential to have accurate information about relevant parameters such as the exchange of momentum,which partly depends on the dragfactor exerted by the state of the sea surface.

One of Powell and colleagues’findings is that maximum windspeeds occur at an altitude of about500 m. More notably, however, theyfind that one assumption used inpredicting the intensity and

consequences of hurricanes isincorrect. Previously, in the absenceof observations for wind speedsabove 25 m s�1, levels of increasingdrag with increasing wind speedwere extrapolated to high windspeeds. But now it seems that abovehurricane force — about 33 m s�1

— a layer of foam and bubbles frombreaking waves develops thatreduces drag and effectively lets the hurricane glide over the sea. In consequence, air–sea exchange in hurricanes will need to bereassessed. Heike Langenberg

© 2003 Nature Publishing Group