cumstellar disks and planet formation. Furthermore, its estimated age of 10 mil- lion years provides a strong constraint on disk evolution time scales and fills a sub- stantial gap in the age sequence between previously known 1-My-old T Tauri stars and 50-My-old nearby open clusters. It has been suggested that circumstellar disks evolve from dense, actively accreting structures to sparse, passive remnants within about 10 My (10). During this tran- sition, grains may assemble into planetesi- mals, or the disk may be cleared by plan- ets. The circurnstellar disks of the TW Hy- drae stars exhibit a wide variety, from classical T Tauri accreting disks, to plane- tary debris systems, to systems without measureable disk emission at near-infrared wavelengths implying cleared-out inner disks (11). A spectacular debris disk with a central cavity has been directly imaged around HR 4796A (12, 13). The diverse disk properties suggest that the TW Hy- drae stars are at an age when disks are rapidly evolving through coagulation of dust and dissipation of gas.

If planets have indeed formed around these stars, it may be possible to detect them with large ground-based telescopes. Adaptive optics, a technique that corrects for the blurring of the atmosphere, allows one to search within several astronomical units (AU) of the TW Hydrae stars (an AU is the average distance between Earth and the sun) for planets a few times as massive as Jupiter. Newborn planets are quite warm, and such objects should therefore be sufficiently luminous to be detected at the distance of this stellar group. In other words, we should be able to look for new- born giant planets located at distances from their parent stars similar to those of giant planets in our own solar system. At least one brown dwarf, a "failed star" not massive enough to ignite hydrogen fusion, has already been found in the TW Hydrae Association (14), and searches for objects of even lower mass are under way (15).

The commotion surrounding the TW Hy- drae Association has prompted astronomers to look for other groups like it. The all-sky survey done by the Roentgen Satellite (ROSAT) satellite has been particularly use- ful in identifying isolated young stars through their x-ray emission. Of the recently discovered stellar groups, MBM12 and Eta Chamaeleontis (Eta Cha) are particularly in- teresting. At about 200 light-years, MBM12 is the second-nearest group of young stars after the TW Hydrae Association, containing only 30 to 100 solar masses of gas. It does not appear to be gravitationally bound and may be breaking up on a time scale compa- rable to the sound-crossing time (16). Thus, in a few million years, the young stars in

SCIENCE'S COMPASS

MBM12 may appear as isolated objects not - A -

associated with any cloud material, very similar to how the TW Hydrae stars appear at present. On the basis of ROSAT detec- tions followed by ground-based optical spec- troscopy, Hearty et al. (17)have identified eight low-mass young stars associated with MBM12. Most of them are classical T Tauri stars and are likely to be a younger popula- tion than the TW Hvdrae members. Eta Cha is a cluster of a dozen young stars first iden- tified in x-ray measurements (18). As with the TW Hydrae group, Eta Cha is far from any substantial cloud. Its members are much less dispersed than the TW Hydrae stars and may represent an epoch intermediate be- tween MBMl2 and TW Hydrae Association.

The exploration of these nearby groups of young stars is progressing at a breath- taking pace. In the past few months, tele- scopes in Arizona, Hawaii, Chile, and Aus- tralia were trained on them with a variety of optical, infrared, and radio instruments. Many questions remain, but the prospects they offer for learning about star forma- tion in the solar neighborhood and the ori- gin and diversity of planetary systems en- sure that interest in them will not wane quickly.

P E R S P E C T I V E S : N E U R O B ! 6 L O C Y

References and Notes 1. C. H. Herbig, in Problems of Physics and Evolution of

the Universe, L. V. Mirzoyan, Ed. (Armenian Academy of Science,Yerevan,Armenia, 1978), pp. 171-180.

2. S. M. Rucinski and j. Krautter. Astron. Astrophys. 121, 217 (1983).

3. R. de la Reza et a/., Astrophys. j. 343, L61 (1989); j. Cregorio-Hetem et dl.. Astron. j. 103, 549 (1992).

4. J. H. Kastner et dl., Science 277,67 (1997). 5. R. A. Webb et a/., Astrophys. j. 512. L63 (1999); M. F.

Sterzik et dl.. Astron. Astrophys. 346, L41 (1999). 6. J. R. Stauffer et dl., Astrophys. j. 454, 910 (1995); D.

R. Soderblom et a/., Astrophys j. 498,385 (1998). 7. E. L. N. jensen et dl., Astron. j.116.414 (1998). 8. E. D. Feigelson. Astrophys. 1.468.306 (1996). 9. L. Hartmann, in Star Formation from the Small to the

Large Scale, F. Favata, A. A. Kaas. A. Wilson, Eds. (ESA SP-445, European Space Agency, Noordwijk, Nether- lands, in press): preprint available at arXiv.orglabs1 astro-phl0001125.

10. S. E. Strom et dl., in Protostars and Planets 111, E. H. Levy and J. I. Lunine, Eds. (University of Arizona Press, Tucson, AZ, 1993).

11. R. jayawardhana e t dl., Astrophys. J. 521, L129 (1999); R.Jayawardhana et a/,, Astrophys. j. 520, L41 (1999); R. D. Cehrz et dl., Astrophys. J. 512, L55 (1999).

12. R. Jayawardhana et dl., Astrophys. j. 503, L79 (1998); D. Koerner etal., Astrophys. j. 503, L83 (1998).

13. C. M.Telesco et dl., Astrophys. j. 530,329 (2000). 14. P. J. Lowrance etal., Astrophys. J. 512, L69 (1999). 15. R. Neuhauser e t dl., Astron. Astrophys. 354, L9

(2000). 16. Any dynamical perturbation would travel through

the cloud at the speed of sound. So the sound-cross- ing time is the typical time scale for dynamical evo- lution of the cloud.

17. T. Hearty et dl., Astron. Astrophys. 353, 1044 (2000). 18. E. E Mamajek et dl., Astrophys j. 516, L77 (1999).

Receptors as Kissing Cousins Craeme Milligan

Cellular processes as different as growth factor signaling and tran- scription depend on interactions be-

tween proteins. Given this, it may not be surprising that certain receptors bind to re- lated but different receptors as well as to

each other. Howev- Enhanced online at er, the dogma has www.sciencemag.org/cgi/ been that members content/fu11/288/5463/65 of the seven-trans-

membrane helix G protein-coupled receptor (GPCR) family pair up with their own kind but do not bind to other family members (I). Several re- cent studies, including a report by Rocheville et al. (2) on page 154 of this is- sue, provide evidence that GPCRs can pair up with even rather distantly related rela- tives to form heterodimeric receptors with distinct properties. Rocheville and co- workers show that the dopamine D2 recep- tor and the somatostatin SSTS receptor form heterodimers. Although pharmaco- logically distinct, these two GPCRs are co-

The author is a t the Division of Biochemistry and Molecular Biology, University of Clasgow, Davidson Building. University Avenue, Clasgow. Scotland C12 8QQ, UK. E-mail: ~milligan@bio.&.ac.uk

expressed in striatal and pyramidal neurons of the cortex. If the reported interaction be- tween the 6 and K opioid receptors (3) (closely related GPCR family members) can be considered as a pairing of brother and sister, then the union of the D2 and SSTS receptors is more akin to a marriage between kissing cousins.

These findings have caused something of a shock. This is despite earlier experi- ments in which the coexpression of two mutant (nonfunctional) angiotensin I1 re- ceptors resulted in formation of a homo- dimer that once more could activate signal transduction pathways after binding ligand (4). Clear evidence emerged last year that formation of heterodimers between GABA (y-aminobutyric acid) R1 and R 2 receptors was necessary for a fully functional GABAB receptor (5).

The usual strategy for studying receptor heterodimerization is to coexpress differen- tially tagged forms of the receptors and then to immunoprecipitate them (3). Although standard for elucidating the interactions be- tween cytoplasmic proteins, the highly hy- drophobic nature of GPCRs mandates their

the membrane before im- munoprecipitation. This necessity renders

www.sciencemag.org SCIENCE V O L 2 8 8 7APRIL 2 0 0 0 65

S C I E N C E ' S COMPASS

the approach less than ideal because artifacts may arise from aggregation of GPCRs. Rocheville et al. decided to apply photo- bleaching fluorescence resonance energy transfer (FRET) to see whether GPCR rela- tives were cohabiting. They produced cell lines that stably coexpressed modest levels of the D2 and SSTS receptors (which were each labeled with a specific dye that could be visualized by FRET). Because energy transfer only occurs when donor and accep- tor molecules are in close proximity, FRET offers an ideal way to follow the interactions of receptors in single living cells.

Surprisingly, the investigators found little indication of heterodimerization between

heterodimerization turns out to be common, then the array of GPCR combinations will be truly bewildering. Both biochemical and physiological data had hinted at the interac- tion of D2 and SST5. For example, coex- pression of the two GPCRs resulted in re- ciprocal shifts in ligand binding affinity. Furthermore, although the somatostatin lig- and was unable to block adenylyl cyclase activity in a mutant SSTS receptor (the GPCR signal transduction pathway is acti- vated through inhibition of adenylyl cyclase by Gi proteins), coexpression of the D2 re- ceptor resulted in rescue of somatostatin h c t i o n (2). This suggested that binding of somatostatin to the SST5-D2 heterodimer

Recently, Liu et al. (12) provided strong biochemical and functional evidence for a direct interaction between the D5 dopamine GPCR and a GABAA receptor with a y2 subunit (see the figure). GABAA receptors (activated by the inhibitory neurotransmit- ter GABA) are C1- ion channels composed of subunits from five related protein fami- lies. D 1 and D5 dopamine receptors stimu- late adenylyl cyclase activity (unlike D2 which inhibits it). Their transmembrane re- gions are highly conserved but their car- boxyl terminal tails are not. The investiga- tors studied interactions between combina- tions of fusion proteins composed of the carboxyl terminal tail of the Dl or D5 re- ceutor and the second intracellular loou of

l . - 0 Q.0 thi five GABAA receptor subunits. fhey

NH,+ am- 0 0 e m - a ' m found a high-affinity interaction between the D5 carboxyl terminus and the second 4: 1- \ @

intracellular loop of the GABAA receptor a,.fi. y2 subunit (see the figure). Remarkably, this interaction was only observed afier ex- posure to agonists for both receptors. Stim- ulation of adenylyl cyclase by D5 (but not D 1) was inhibited in a concentration-depen- dent manner by the y2 subunit of GABAA in the Dresence of GABA. This effect could

receptor

Family and friends. Interactions between CPCRs and ion channels. Each CABA, receptor ion channel consists of five polypeptides.The predominant type of CABA, is composed of two a, two p, and 'one y subunits. As y2 is the predominant y isoform, there is the potential for interactions (with a likely 1:l stoichiometry) between CABA, and D5 dopamine receptors in most neurons that coexpress both.The site of contact between the carboxyl-terminal tail of the D5 receptor and the second intracellular loop of the y2 subunit have not been explored but agonists for both recep- tors (dopamine and CABA, respectively) are required for the interaction.

D2 and SSTS receptors unless a neurotrans- mitter agonist for either receptor was pre- sent. The neurotransmitter for either' receu- tor promoted heterodimerization, but the presence of both ligands did not produce an additive or synergistic interaction. Previous studies often demonstrated high levels of GPCR heterodimerization in the absence of agonist ligands and variations'b the ability of ligands to alter this status (5, 6). These earlier findings may reflect both the high- level expression of receptors and their ca- pacity to form nonspecific aggregates in the absence of a membrane environment. Be- cause the D2 and SSTS receptors are known to form homodimers, it would be predicted that each individual agonist would encour- age homodimerization as well as (or, per- haps, rather than) heterodimerization. Al- though not studied by Rocheville et al. (2), it would be interesting to compare the for- mation of homo- and heterodimers between

$ the D2 and SSTS receptors in the presence of either ligand.

t The GPCR family is probably the largest ! in the human genome. If ligand-induced

induced direct activation of the D2 receptor. All of the somatostatin receptors and three of the dopamine receptors (D2, D3, and D4) activate members of the same family of G proteins, which inhibit adenylyl cyclase and regulate ion flow through a group of Ca2+ and K+ channels. No doubt het- erodimers between other somatostatin and dopamine receptors will be identified soon.

Two separate gene families encode GPCRs and ion channel proteins, and, until recently, there was little evidence for interac- tions between such strangers. But, the reper- toire of proteins that interact with GPCRs is expanding dramatically (7). The purpose of some interactions is clear-the interaction of HomerNesl proteins with metabotropic glu- tamate receptors (8) results in release of these GPCRs from the endoplasmic reticu- lum and their expression in the plasma mem- brane; the intracellular trafficking of the PZ- adrenergic receptor is directed by its interac- tion with the phosphoprotein EBP5O (9). But there are other cases where the consequences of interactions between GPCRs and unrelat- ed proteins remain obscure (10, 11).

be t r i s fkred by exchanging the carboxyl terminus of D l for that from D5. As a corollary, in cells coexpressing GABAA (containing a y2 subunit) and D5, GABA- induced ion currents decreased when dopamine was added. As in the Rocheville study (2), direct evidence from earlier ex- periments for the coexpression of the GABA, and D5 receptors in hippocampal neurons provided the impetus to investigate possible interactions between these two classes of GPCRs.

Large-scale protein-protein interaction screens in model systems (13) together with global gene expression profiling afier receptor activation (14) promise the rapid identification of many more unexpected interactions between GPCRs and other proteins (15). The next step will be to work out the functional importance of all these kissing cousins.

References 1. T. E. Hebert and M. Bouvier, Biochem. Cell Biol 76, 1

(1998). 2. M. Rocheville et al.. Science288, 154 (2000). 3. 8.A. Jordan and LA. Devi, Nature 399,697 (1999). 4. C. Monnot et al., j. Biol. Chem. 271,1507 (1996). 5. F. H. Marshall et dl., Trends Phamacol. Sci. 20, 396

(1999). 6. F.-Y. Zeng and J . Wess. j. Biol. Chem. 274, 19487

(1999). 7. R.A. Hall et dl.. j. Cell Biol.145,927 (1999). 8. K.W. Roche et aL, j. Biol. Chem. 274,25953 (1999). 9. T.T. Cao et aL, Nature401.286 (1999).

10. C. Ullmer etal.. FEBS Lett. 424,63 (1998). 11. Z. Xu et dl.. j. 6\01. Chem. 274,21149 (1 999). 12. F. Liu et al., Nature 403,274 (2000). 13. P. Uetz etal., Nature 403.623 (2000). 14. C. J. Roberts et dl., Science 287.873 (2000). 15. N. Lezcano etal., Science 287,1660 (2000).

www.sciencemag.org SCIENCE VOL 288 7APRIL 2000