Corticalbone

Bloodstream

Bonemarrow

CD44 ligands

CD44

RollingFirm

adhesion

Endosteal niche

Vascular niche

β1 integrins

LeukaemichaematopoiesisNormal

haematopoiesis

Leukaemic stem cell

Leukaemic progenitor

Mature leukaemic cell

Normal stem cell

Normal progenitor

Migrating leukaemic cell

injection of antibody-treated human cells into the mouse marrow cavity. Such injected cells failed to establish a robust leukaemia even though these cells were not required to traverse the usual engraftment pathways.

Krause et al.2 used a mouse model of chronic myelogenous leukaemia, which is caused by the expression of the Bcr–Abl fusion protein. They report similar — but not identical — results. To examine the role of CD44, the authors expressed Bcr–Abl in mouse HSC/Ps that were genetically deficient in CD44, and transplanted these cells into recipient mice of the same strain as the donors (to avoid rejection). In these studies, the progression of Bcr–Abl-induced, leukaemia-like disease was delayed in recipi-ent mice. In addition, similar to the results of Jin et al.1, antibody treatment of recipients of Bcr–Abl-expressing LSCs also slowed disease progression. However, unlike the studies with human leukaemia cells, the direct injection of antibody-treated LSCs led to disease similar to that in control-treated cells.

The basis for this difference in the two stud-ies is not clear, but it could be related to the specific tumour type (acute versus chronic), model (cross-species versus same-strain trans-plant) or antibody used. This difference may be quite important because it suggests that at least one mechanism noted above — the direct effect of CD44 binding to its ligand on LSC dif-ferentiation — is potentially not occurring in both models. Additional studies are clearly needed.

The different effect of antibody treatment on LSCs and normal HSC/Ps is intriguing

Figure 1 | Putative migration of stem cells between the bloodstream and bone marrow. Normal stem cells require multiple receptors to adhere firmly to the endothelium (the cell layer that lines blood vessels). They then make their way to putative endosteal and vascular stem-cell niches in the bone marrow. The β1 integrins are shown as an example of the adhesion molecules involved in migration and adhesion of normal stem cells to their niches. Jin et al.1 and Krause et al.2 show that leukaemia stem cells instead seem to depend on the expression of CD44 to localize to the bone marrow and initiate a leukaemia-like disease in mouse models.

and important, but the basis for the differ-ence is not yet clearly defined. Work by Jin et al.1 suggests this difference may be because different isoforms of CD44 are expressed by LSCs compared with HSC/Ps. Krause et al.2

suggest instead that the differential effect may relate to the relative level of CD44 expression in the leukaemia-initiating cells. Similar issues have plagued past studies implicating CD44 in the homing and engraftment of normal HSC/Ps into the bone marrow. In either case, the results are not yet definitive. In spite of this continued uncertainty, these reports1,2 provide significant insight into the biol-ogy of leukaemia-initiating cells with stem-cell-like characteristics and normal HSC/Ps with respect to interaction with the haemato-poietic micro environment. Such differences in the behaviour of the normal and leukae-mic cells could well provide new avenues for targeted therapies. ■

David A. Williams is in the Division of Experimental Hematology, Cincinnati Children’s Research Foundation, Cincinnati, Ohio 45229, USA. Jose A. Cancelas is in the Hoxworth Blood Center, University of Cincinnati Medical Center, Cincinnati, Ohio 45267, USA.e-mail: David.Williams@cchmc.org

1. Jin, L., Hope, K. J., Zhai, Q., Smadja-Joffe, F. & Dick, J. E. Nature Med. 12, 1167–1174 (2006).

2. Krause, D. S., Lazarides, K., von Andrian, U. H. & Van Etten, R. A. Nature Med. 12, 1175–1180 (2006).

3. Lord, B. I., Testa, A. G. & Hendry, J. H. Blood 46, 65–72 (1975).

4. Dexter, T. M. et al. J. Cell Physiol. 91, 335–344 (1976).5. Nilsson, S. K., Johnston, H. M. & Coverdale, J. A. Blood 97,

2293–2299 (2001).6. Calvi, L. M. et al. Nature 425, 841–846 (2003).7. Kiel, M. J., Yilmaz, O. H., Iwashita, T., Terhorst, C. &

Morrison, S. J. Cell 121, 1109–1121 (2005).8. Cancelas, J. A. et al. Nature Med. 11, 886–891

(2005).9. Lapidot, T. et al. Nature 367, 645–648 (1994).

SEMICONDUCTOR ELECTRONICS

Organic crystals at largePaul Heremans

Fabricating large-scale semiconducting surfaces for the flexible screens of the future is a bothersome business. A simple technique for growing single-crystal organic semiconductors brings new vision to the field.

On page 913 of this issue, Zhenan Bao and col-leagues (Briseno et al.)1 describe a method for nucleating and growing single-crystal organic semiconductors over electrodes on an arbi-trary substrate, and so producing extended arrays of high-performance electronic transis-tors. Their method represents a step towards practical applications of high-quality, but fragile, single crystals of organic semiconduc-tors, and could show the way to high-perform-ance electronic devices that extend over large and flexible surfaces.

Transistors are minuscule switches that lie at the heart of all electronic devices — com-puters, memory sticks, displays, you name it. At the heart of every transistor is a channel of semiconductor material that bridges two

electrodes. An electrical potential applied at a third electrode, the ‘gate’, controls whether current is allowed over this bridge or not. The most successful semiconductor for such switching applications, omnipresent in today’s electronics, is silicon.

Silicon transistors can be processed on two very different types of substrate. The first is a thin, single-crystal, nearly defect-free silicon substrate called a wafer. Wafers are limited in size, but can be processed in sophisticated ways to bear a mind-boggling density of nanoscale transistors: a wafer 300 millimetres in diam-eter can hold hundreds or thousands of chips, each one containing up to a billion transistors. Wafer-based technologies are continuously scaling to larger transistor counts and density

828

NATURE|Vol 444|14 December 2006NEWS & VIEWS

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

Nature Publishing Group ©2006

per chip, allowing ever increasing computing power and ever higher memory density.

But other applications require transistors to be sparsely distributed over large areas, and can’t afford to be limited by wafer size. A flat-panel display, for example, has transistors distributed over its whole surface that switch the miniature pixels forming the image on and off. Larger screens require the distribution of transistors over larger areas, and using single-crystal semiconductor wafers as substrates is simply not economical in this case. Instead, large-area applications use ‘foreign’, non-semi-conductor substrates, such as metals, ceram-ics or glass, that can be manufactured as large sheets. In the future, substrates might also include flexible and less brittle materials such as plastic foils2.

Fabricating transistors with sufficient per-formance and high yield over large areas on foreign substrates is enormously challenging. Most conventional semiconductors, including silicon, can be forced to grow as defect-free films only if the growth substrate provides a perfect template for their crystal structure. When deposited on an arbitrary substrate, they tend to form amorphous, defect-infested films. Decades of research have been invested in improving the quality of these amorphous semiconductors to levels acceptable for the display industry.

One particular improvement involves recrys-tallizing the amorphous film into a polycrys-talline film consisting of small, single-crystal grains, each with a particular crystal orienta-tion, separated by thin boundary regions. The

properties of polycrystalline materials are less ideal than those of monocrystalline materials, but are in many respects better than those of amorphous materials. But the recrystalliza-tion process requires high temperatures, so, although it can be applied with success to, say, a glass substrate, it is delicate and difficult to scale the process up to an industrial level for substrates such as plastic foil that cannot with-stand elevated temperatures.

Because the transistors in large-area appli-cations are distributed over the surface, a high-quality semiconductor film is in fact not required over the entire area. A viable way around the large-area problem could thus con-sist of providing semiconductor material only at those places that have to bear transistors.

Briseno et al.1 come up with an original solution to achieve precisely this. Their pro-cess hinges on a unique combination of the properties of organic semiconductors and the science of crystal nucleation and growth. First, pure crystals of so-called conjugated molecules — organic molecules that have alternate single and double covalent bonds — are known to be pretty good semiconductors3,4.

Second, because these molecules are bound into a crystal through weak van der Waals forces, rather than through stronger covalent bonds, the energy needed to form crystals from them is moderate. Such crystals can there-fore be formed at low temperatures, between room temperature and a few hundred degrees Celsius, that are compatible with the process-ing of large-area substrates.

Finally, thermodynamics teaches us that

many controllable parameters influence the energy barrier against the nucleation of crys-tals. These can be used to determine the loca-tion of the nucleation and growth of crystals on a surface. A known method for promoting local crystal nucleation is to ‘stamp’ a substrate with an extremely thin layer of molecules, known as a self-assembled monolayer, that is terminated by a chemical group with a particular affin-ity for the atoms or molecules of the desired crystal material5.

Briseno et al.1 use a different method that is less dependent on chemical affinity. They start by stamping an organic layer of signifi-cant roughness onto the substrate. Roughness is known to increase surface energy, and the nucleation and growth of crystals can then be steered to occur selectively on the rougher areas. The authors show that the method can be used to grow many different semiconduct-ing conjugated molecular crystals. These include molecules that are notoriously dif-ficult to grow as films on substrates, such as rubrene6, and some with high vapour pres-sure, such as anthracene, that evaporate spon-taneously in the high-vacuum environments in which thin-film deposition processes are generally conducted. But most importantly for device applications, the authors grow organic semiconductor crystals directly onto metal electrodes — forming a basic transistor — and show that these electrodes are functional after the process.

Considerable work is still needed to turn this technique into a truly viable semiconduc-tor technology. In particular, the control of

Gold, frankincense and myrrh — the three royal gifts of the Christmas story — remain valuable commodities. But as Toon Rijkers et al. report in the Journal of Applied Ecology (43, 1188–1195; 2006), the latter-day story of frankincense is also a tale for our times.

Frankincense is a resin produced by several small trees of the genus Boswellia, which grow in sub-Saharan Africa, from Nigeria to the Horn of Africa, and Arabia and the Himalayan foothills. The resin is exuded naturally from leaves and shoots, but especially from wounds. Its natural function is probably to ward off grazers and fungal attack. But the resin’s fragrance has also made it prized as a perfume, fumigant, flavouring and medicinal compound.

Rijkers et al. looked at frankincense production in Eritrea, where it is used locally for cultural and medicinal

purposes, and is also exported. Wild Boswellia papyrifera trees are tapped by making incisions around the trunk of the tree, starting in mid-September at the end of the summer monsoon. The resin is harvested every three weeks by reopening the incisions, and harvesting continues throughout the dry season. Boswellia trees also produce their flower buds at the end of the wet season, however. Could the constant harvest of resin, commencing at the time of floral induction, be a serious drain on the trees’ carbon resources? Not least because leaves are lost at the beginning of the dry season, so carbon for flowering and resin production must be obtained from storage products.

To investigate the effects of resin-tapping on the tree life-cycle, and especially seed quality, Rijkers et al. checked the seed-germination potential of tapped and untapped

trees. They found that seeds from annually tapped trees had a germination success of only about 20% compared with a figure of 80–90% for untapped trees. Tapped trees also produced fewer inflorescences, fruits and seeds. So tapping is evidently costly to the individual tree.

There has been a severe reduction in the natural regeneration of

Boswellia in Eritrea, to which more intense tapping may be contributing. Actions that would help to reverse this trend include reducing the amount of resin taken per tree, and resting selected trees for a few seasons; the Boswellia populations also need protection from intense grazing.

Here, however, the spectre of the ‘Tragedy of the Commons’ arises: the resources for both frankincense harvesting and grazing are freely available, and so prone to overexploitation unless the common good is taken

into account. Self-regulation in the cause of long-term sustainability is a difficult lesson to learn, as a wider world has found out from the over-exploitation of fisheries. Peter D. MoorePeter D. Moore is in the Department of Biochemistry, King’s College London, Franklin–Wilkins Building, London SE1 9NH, UK.e-mail: peter.moore@kcl.ac.uk

CONSERVATION BIOLOGY

Unkind cuts for incense

C. O

SBO

RNE/

CO

RBIS

829

NATURE|Vol 444|14 December 2006 NEWS & VIEWS

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

Nature Publishing Group ©2006

–120 –80 –40 40 mV0

–120 –80 –40 40 mV0

1

0

Nav1.7

Nav1.4

a

b

2

1

0

0 600 ms

(0.23 mV ms–1)

% o

f pea

k cu

rren

t%

of p

eak

curr

ent

“God…shouts in our pain. It is his megaphone…” — C. S. LewisPain can be useful — when it warns us that a hot object is burning us, for instance. Yet the amelioration of pain is essential to modern medicine, permitting surgery to take place and making many illnesses more bearable. Even so, some pain doesn’t respond to current treat-ments; in particular neuropathic pain, which can occur in the absence of noxious stimuli following injury to the nervous system, and some types of inflammatory pain. In this issue, Cox et al. (page 894)1 describe a rare inher-ited mutation that renders the people carry-ing it unable to feel any pain. The mechanism behind this deficit could aid the search for novel painkillers.

Where does pain come from, and how can we understand it so that we can tame it? Pain-ful stimuli are conveyed in the form of trains of electrical impulses. These impulses travel from nociceptive (pain-signalling) dorsal root ganglion (DRG) neurons originating in the body’s periphery, through ascending spinal pathways, to the brain. Ion channels in the cell membranes of these nociceptive neurons, including several different types of sodium channel, collaborate to produce such nerve impulses, although the relative roles of the different channels are not fully understood.

Ten different genes encode ten versions (isoforms) of the sodium-channel protein that all share a common structure but have different constituent amino-acid sequences.

One of these genes, SCN9A, encodes a sodium channel known as Nav1.7, which is preferen-tially expressed at high levels in two types of neuron: nociceptive DRG neurons and sym-pathetic ganglion neurons, which are part of the involuntary, or ‘autonomic’, nervous sys-tem2. Nav1.7 is deployed at the endings of pain-sensing nerves (the nociceptors), close to areas where the impulse is initiated2. Stimulation of the nociceptor nerve endings produces ‘genera-tor potentials’ — small changes in the voltage across the neuronal membranes. The Nav1.7 channel amplifies these membrane depolariza-tions3 (Fig. 1), and when the membrane poten-tial difference reaches a certain threshold, the neuron fires.

Clues that Nav1.7 is involved in pain came from the observation4 that DRG neurons in animal models of inflammatory pain show increased expression of Nav1.7. Also, mice genetically engineered to lack Nav1.7 spe-cifically in their nociceptors show markedly reduced responses to inflammatory pain5. The painful inherited human neuropathy known as erythromelalgia, in which sufferers expe-rience a severe burning pain in response to mild warmth, is due to mutations6–8 in Nav1.7 that cause excessive channel activity9,10. This suggests that Nav1.7 sets the gain on pain signalling in humans.

Cox et al.1 have now discovered SCN9A mutations that cause a loss of Nav1.7 function in three families from Pakistan. Their observa-tions link loss of Nav1.7 function with a congeni-tal inability to experience pain, adding to the

Figure 1 | Signalling through sodium channels. The graphs show changes in the potential difference across the cell membrane with gradually increasing stimulus amplitude. (The y-axis shows current as a percentage of the maximal response to an abrupt, large stimulus.) a, The Nav1.7 sodium channel is found in the pain-sensing ends of nociceptive neurons. They produce a depolarization in the membrane (blue) in response to small, slow stimuli such as the ramp stimulus (red), causing the neuron to fire when the potential difference reaches a certain threshold. These channels therefore amplify the ‘generator potentials’ that are evoked by sensory stimuli and they set the gain on nociceptors. b, Other sodium channels, such as the Nav1.4 sodium channel in muscle, do not produce these ramp responses. (Modified from ref. 3.)

NEUROBIOLOGY

A channel sets the gain on painStephen G. Waxman

Nerve impulses that convey pain signals to the brain are produced by sodium channels in the neuronal membrane. Studies on people who are unable to feel pain identify one specific sodium channel as essential to the process.

crystal size and orientation will require fur-ther attention, as will the reproducibility and control of the quality of the electrical contacts between the semiconductor and the electrodes. The single-crystal technique might, uniquely, allow devices that are difficult to realize using thin polycrystalline or amorphous films of the same materials to be integrated onto foil. For example, most organic semiconductors are intrinsically ambipolar7, meaning that they allow the movement of both electrons and so-called holes — positive charges equivalent to the absence of an electron. This property has been difficult to exploit with thin organic semiconductor films, and single crystals might open new avenues towards this end.

So far, the family of technologies for use in electronics on flexible foil has numbered

thin-film organic semiconductors, thin-film metal-oxide semiconductors8,9, and the twin silicons, amorphous and polycrystalline, among its members. Bao and colleagues1 have just presented us with a new baby. ■

Paul Heremans is at IMEC, Kapeldreef 75, B3001 Leuven, Belgium.e-mail: heremans@imec.be

1. Briseno, A. L. et al. Nature 444, 913–917 (2006).2. Huitema, H. E. A. et al. Nature 414, 599 (2001).3. de Boer, R. W. I. et al. Phys. Status Solidi A 201, 1302–1331

(2004).4. Pope, M. & Swenberg, C. E. Electronic Processes in Organic

Crystals and Polymers (Oxford Univ. Press, 1999). 5. Aizenberg, J. et al. Nature 398, 495–498 (1999).6. Käfer, D. & Witte, G. Phys. Chem. Chem. Phys. 7, 2850–2853

(2005).7. Chua, L.-L. et al. Nature 434, 194–199 (2005).8. Nomura, K. et al. Nature 432, 488–492 (2004).9. Wang, L. et al. Nature Mater. 5, 893–900 (2006).

evidence indicting this sodium channel as an essential participant in human nociception. The profound insensitivity to pain in members of these families is remarkable because noci-ceptors contain many sodium-channel isoforms, so the remaining isoforms might have been expected to support at least some degree of noci-ception, even though Nav1.7 is out of action.

The Nav1.7 channel is present in some non-nociceptive DRG neurons11, as well as in noci-ceptors. However, the people described by Cox et al. have no apparent deficits in non-noci-ceptive sensory functions, such as the ability to perceive touch, warmth, cold, tickle, pressure, or the position of their limbs (proprioception). This raises the question of whether Nav1.7 channels have different roles in nociceptive and non-nociceptive DRG neurons. The pain insensitivity in these people seems to be more severe than the deficits seen in the mice lacking Nav1.7 in their nociceptors5. So perhaps there are changes in upstream nociceptive centres in the spinal cord or brain of these people that contribute to their insensitivity to pain.

Nav1.7 is present not only in DRG neurons, but also in sympathetic ganglion neurons2. But

831

NATURE|Vol 444|14 December 2006 NEWS & VIEWS

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

Nature Publishing Group ©2006