news and views

NATURE CELL BIOLOGY VOL 5 MARCH 2003 www.nature.com/naturecellbiology 185

Rho GTP-binding proteins (a subgroupof the Ras superfamily of GTPases)control a wide range of essential sig-

nalling pathways in all eukaryotic cells. Inhigher organisms, they orchestrate cellularprocesses as diverse as cell migration, cell-cycle progression and cytokinesis, microbialkilling (through phagocytosis and NADPHoxidase activity) and agonist-regulated genetranscription. At the biochemical level, themultiplicity of pathways and diversity ofresponses mediated by the Rho proteins areexplained both by their action as molecularswitches, capable of integrating and relay-ing biological signals, and their large spec-trum of downstream effectors. Extensivestudies of cytoskeletal dynamics have firm-ly established that Rho, Rac and Cdc42(the prototypical members of the Rhofamily) control distinct morphogenic sig-nalling pathways, all of which are crucialfor cell adhesion and/or motility1.Complex processes such as cell migrationinvolve the dynamic coordination of adhe-sion and motility. In addition, coherentmovement requires tight spatiotemporalregulation of the balance between pro-adhesive and pro-motile signals. In thisrespect, recent evidence indicates that Rhoproteins can influence each other’s activi-ties. In particular, Rac- and Rho-inducedeffects, which correlate with membraneprotrusion and contractility, respectively,antagonize each other in a variety of celltypes. For example, in fibroblasts, activa-tion of Rac has been specifically shown todown-regulate Rho activity, but the mech-anisms involved are poorly understood2. Areport by Nimnual et al.3 on page 236 ofthis issue now sheds new light on thisquestion, describing a fascinating pathwayby which Rac-mediated production ofreactive oxygen species signals to switchoff Rho activity and modify cell shape.

Rho GTP-binding proteins controlessential cellular functions in eukaryotes,including most of the processes that requirecytoskeletal dynamics, from bud-site selec-tion in yeast to animal cell migration. Giventhe key roles that Rho proteins have in cellbiology, it is probably not surprising to dis-cover that their activity is tightly regulated.In common with all Ras family members,Rho GTP-binding proteins function asmolecular switches that are active whenbound to GTP and inactive when bound to

GDP. Upstream signals, often derived fromextracellular ligands, activate guaninenucleotide-exchange factors (GEFs) thatcatalyse the removal of bound GDP, allow-ing GTP loading and thereby activation ofthe Rho proteins. Only in their GTP-boundform do the Rho proteins interact with spe-cific downstream effectors; that is, signallingor structural proteins that mediate their cel-lular effects. Conversely, inactivation of theRho GTPases is mediated by GTPase-acti-vating proteins (GAPs), which stimulatetheir intrinsic GTPase enzymatic activity1.

Pioneering studies in Swiss 3T3 fibrob-lasts have shown that activation of RhoA,Rac1 and Cdc42 — the best-characterizedmembers of the Rho family — mediategrowth-factor-induced remodelling of theactin cytoskeleton into stress fibres, lamel-lipodia and filopodia, respectively. It is nowgenerally accepted that in most cell types,Cdc42 and Rac1 signalling promote the for-mation of membrane protrusions, drivenby the microtubule and actin cytoskeletalnetworks, whereas Rho activity is associat-ed with acto-myosin-based cell contrac-tion1. In complex biological processes suchas cell migration, repeated cycles of protru-sion at the leading edge and retraction at

the trailing end have to be coordinatedwith cell adhesion to allow coherent,directed motility. Therefore, such globalcellular responses require some form ofcrosstalk between the different RhoGTPase signalling pathways. In serum-starved Swiss 3T3 fibroblasts, a hierarchi-cal, linear cascade of activation links Cdc42signalling to activation of Rac and thenRho, resulting in the successive formationof filopodia, ruffles and stress fibres1.However, in other cell systems, the differentRho GTPase pathways antagonize, ratherthan activate, each other. This has been par-ticularly well documented for Rho and Racsignalling, which mediate opposite pheno-types in many cell types (Fig. 1). For exam-ple, in neuronal cells, activation of Rhoresults in neurite retraction and cell round-ing, whereas activation of Rac promotes cellspreading and neurite outgrowth4. Not onlydo Rho- and Rac-induced phenotypes seemmutually exclusive, but Rho- and Rac-mediated pathways also regulate eachother. Expression of active Rac rendersneuronal cells refractory to Rho-inducedsignalling4. Similarly, inhibition of the Rhopathway in Swiss 3T3 or Rat1 fibroblastsinduces Rac-associated phenotypes (small

Rac signalling: a radical viewEmmanuelle Caron

GTP-binding proteins of the Rho family regulate each other’s activities by largely elusive mechanisms. Now,an unexpected signalling pathway has been identified in fibroblasts that links Rac activation to the inhibitionof Rho activity, through the release of oxygen radicals.

Rac Rho

Protrusion

Cell spreadingMembrane rufflingNeurite outgrowth

Epithelioid morphology (cell–cell contact)

Contraction

Cell roundingStress fibres

Neurite retractionFibroblastoid morphology

a b

Figure 1 Antagonism between Rho and Rac activities. Rac and Rho signalling triggeropposite responses in many cell types. a, Activation of Rac is generally associated withmembrane protrusion (membrane ruffling, cell spreading and neurite outgrowth) and thestimulation of cell–cell contact formation in NIH 3T3 cells2. b, In contrast, Rho activationis linked to acto-myosin contractility, stress fibre formation, cell rounding, neuriteretraction and, in NIH 3T3 cells, a fibroblastoid morphology2.

© 2003 Nature Publishing Group

news and views

NATURE CELL BIOLOGY VOL 5 MARCH 2003 www.nature.com/naturecellbiology186

focal contacts, cell spreading and motility)and vice versa, inhibition of Rac signallingpromotes Rho-related phenotypes5,6.

Indeed, the phenotypic antagonism thatexists between the Rac and Rho pathways isreflected by their activation state in variouscell types, where the levels of active Rac areinversely correlated to the levels of activeRho. This is true, for example, in epithelialcells that are growing or undergoing cad-herin-dependent cell–cell contact forma-tion2,7, and in fibroblasts during spreadingonto fibronectin8. In NIH 3T3 cells, activa-tion of Rac (whether transient or sus-tained) markedly reduces Rho-GTP levels.Moreover, Rac-induced down-modulationof Rho activity is independent of the abili-ty of Rac to remodel the actin cytoskeletonor stimulate the JNK kinase pathway, asdemonstrated by the use of Rac effectormutants2. Despite these observations, themechanism involved in Rac-mediated inhi-bition of Rho activity remains largelyunknown, although Src kinase activity andp190Rho-GAP activation have both beenimplicated in the transient inactivation ofRho elicited by integrin ligation6,8.

Theoretically, two mechanisms couldaccount for such a rapid cross-regulation ofsmall GTPase function. First, regulation byone GTPase could target the effector path-way of the second GTPase. Indeed, it hasbeen shown previously that Rac can regu-late Rho signalling via PAK1, a p21-activat-ed kinase and a downstream effector of Rac.PAK inhibits acto-myosin contractility(that is, it opposes Rho signalling) throughits effect on phosphorylation of myosin-IIheavy chain and myosin light chainkinase9,10. An alternative mechanism couldbe where one GTPase directly modulates

the switch mechanism of the target GTPase.In this issue of Nature Cell Biology, an

exciting report describes just such a mech-anism for a comprehensive signallingpathway, linking activation of Rac to theinhibition of Rho activity in fibroblasts3.This pathway operates both in cells overex-pressing the GTP-bound form of Rac and,more physiologically, when endogenousRac is transiently activated after attach-ment of cells to fibronectin. Confirmingprevious results2,6, the authors show thatactivation of Rac results in a three to four-fold decrease in the levels of active, GTP-bound Rho. This process is independent ofthe effector pathways that mediate Rac-induced ruffling and JNK activation andalso requires activation of p190Rho-GAP.Remarkably, Nimnual et al. show thatdown-modulation of Rho activity is medi-ated by reactive oxygen species (ROS) andcontrolled by the insert region of Rac. Thissmall helix (residues 124–135 in Rac) isunique to Rho proteins and is located welloutside the switch I region (the so-called‘effector’ domain), which controls the bind-ing of downstream effectors. Importantly,this insert region has been implicated byseveral research groups in the activation ofa Rac-dependent process that is rarely asso-ciated with morphogenic changes, but isquite familiar to immunologists and micro-biologists — the production of ROS. Inphagocytic leukocytes, Rac activity regu-lates the assembly and activation of themulticomponent, superoxide-generatingNADPH oxidase. Rac is generally thoughtto control oxidase activity through its GTP-dependent ability to interact with p67-phoxand recruit it to the membrane-associatedoxidase complex, which p67-phox can then

activate. However, according to one recentstudy, Rac controls oxidase activity boththrough its insert region, which interactsdirectly with the catalytic flavoprotein andregulates electron flow, and through p67-phox binding11. Homologues of the phago-cytic flavoprotein have been identified innon-phagocytic cells, where they also con-trol ROS production in a Rac- and insert-region-dependent manner11,12. In this newwork, Rac-mediated, insert region-depend-ent ROS production is now convincinglyshown to be both necessary and sufficient todown-modulate Rho activity in fibroblasts3.The signalling pathway described involvesROS-mediated inactivation of the tyrosinephosphatase LMW-PTP (low-molecular-weight protein tyrosine phosphatase)through direct oxidation of reactive cys-teines in the catalytic pocket, which relievesthe inhibition of p190Rho-GAP (whoseactivity is otherwise inhibited by LMW-PTP-mediated dephosphorylation). Thus,inhibition of LMW-PTP activity results inincreased phosphorylation of p190Rho-GAP and hence inactivation of Rho (Fig. 2).

The results presented by Nimnual et al.are important in several respects. First, theyfirmly establish a role for the insert regionin the cellular roles of Rho GTPases, specif-ically in Rac-mediated ROS productionand morphogenic changes. Whether thishelix simply contributes to effector binding(directly or indirectly) or functions as ananchor to stabilize the association of Racwith specific areas of the plasma membranewill have to be investigated. Second, theyprovide a sophisticated mechanism bywhich a GTPase can control anotherGTPase activity. Interestingly, this pathwayinvolves activation of p190Rho-GAP, but isindependent of the Src kinase-dependentmechanism proposed previously6,8. Aspointed out earlier, regulatory crosstalkcould target either the switch mechanism3,6

or a GTPase effector pathway9,10 with a sim-ilar outcome. In this case, it is now clearthat both of these mechanisms can be usedduring Rac-mediated downregulation ofRho signalling. However, many more cross-regulatory mechanisms exist (for example,see ref. 13) and will no doubt be discoveredwithin the Rho subfamily of GTP-bindingproteins, not to mention across distinct Rassub-families, of which little is currentlyknown. Finally, this study stresses thepotentially immense role of ROS in molec-ular cell biology and signalling mecha-nisms. Rac and oxygen radicals have alreadybeen implicated in microbial killing, cellcycling, transformation and the regulationof gene transcription11,12,14,15. Now, withtheir newly established functions in Rac-mediated inactivation of Rho3, cell–celladhesion and cell migration16 and thereforetheir relevance for both fundamental cellbiology and cancer, ROS are likely to attracta lot of attention in the coming years.

Upstreamsignals

Rac-GEF

Rac-GTP

Rac-GDP

ROS(insert region)

LMW-PTP p190Rho-GAP

PAK(effector domain)

Rho-GTP Rho-GDP

ROK Dia

Acto-myosincontractilefilaments

Figure 2 Inhibition of Rho activity by Rac signalling. Activation of Rac stimulates distincteffector pathways that effectively switch off Rho activity in cells. In this model, whichsummarizes results obtained in several independent studies (see text), Rho signalling isefficiently and quickly shut off through the combined inhibition of Rho activation by ROSand of Rho kinase (ROK) activity, downstream of Rho (mediated by PAK, see main text).This mechanism involves two distinct regions in active Rac.

© 2003 Nature Publishing Group

news and views

NATURE CELL BIOLOGY VOL 5 MARCH 2003 www.nature.com/naturecellbiology 187

Emmanuelle Caron is at the Centre for Molecular

Microbiology and Infection & the Department of

Biological Sciences, The Flowers Building, Room

2:41, Armstrong Road, Imperial College London,

London SW7 2AZ, UK

e-mail: e.caron@imperial.ac.uk

1. Etienne-Manneville, S. & Hall, A. Nature 420, 629–635 (2002).

2. Sander, E. E. et al. J. Cell Biol. 147, 1009–1022 (1999).

3. Nimnual, A. S., Taylor, L. J. & Bar-Sagi, D. Nature Cell Biol. 5,

236–241 (2003).

4. van Leeuwen, F. N. et al. J. Cell Biol. 139, 797–807 (1997).

5. Rottner, K., Hall, A. & Small, J. V. Curr. Biol. 9, 640–648 (1999).

6. Arthur, W. T. & Burridge, K. Mol. Biol. Cell 12, 2711–2720

(2001).

7. Noren, N. K., Niessen, C. M., Gumbiner, B. M. & Burridge K. J.

Biol. Chem. 276, 33305–33308 (2001).

8. Arthur, W. T., Petch, L. A. & Burridge, K. Curr. Biol. 10,

719–722 (2000).

9. van Leeuwen, F. N. et al. Nature Cell Biol. 1, 242–248 (1999).

10. Sanders, L. C., Matsumura, F., Bokoch, G. M. & de Lanerolle, P.

Science 283, 2083–2085 (1999).

11. Bokoch, G. M. & Diebold, B. A. Blood 100, 2692–2696 (2002).

12. Joneson, T. & Bar-Sagi, D. J. Biol. Chem. 273, 17991–17994

(1998).

13. Tsuji, T. et al. J. Cell Biol. 157, 819–830 (2002).

14. Suh, Y. A. et al. Nature 401, 79–82 (1999).

15. Kheradmand, F. et al. Science 280, 898–902 (1998).

16. van Wetering, S. et al. J. Cell Sci. 115, 1837–1846 (2002).

The genome is continuously bombarded by ionizing radiation,bathed in a cocktail of nucleic acid modifying compounds andsubjected to errors during its replication. To ensure faithfulmaintenance of genetic information, all organisms have evolvedintricate and highly specialized means to repair this unavoidabledamage. Defects in a number of these pathways result in cancer,premature ageing, developmental defects or even death.

One of the modifications nucleotides are subjected to is alky-lation. Currently, three independent ways to deal with mutagenicalkylation have been elucidated: methylpurine DNA glycosylaseand the base excision repair pathway removes 3-methyladenine,whereas methylguanine DNA methyltransferase specifically mod-ifies O6-methylguanine and 1-methyladenine or 3-methylcyto-sine are subject to oxidative demethylation by AlkB. The lattertwo base modifications are formed in much higher yields in sin-gle- versus double-stranded DNA, and abundantly also in RNA.In fact, RNA is subject to over 60 other forms of post-transcrip-tional modification, some of which have been shown to affectRNA folding and RNA protein interactions, and hence transcrip-tion and translation. Although some of these modifications areclearly of regulatory importance, others are not and would beexpected to render messenger- and other RNA non-functional atbest or to generate potentially detrimental proteins at worst. Untilnow, it had been unclear how the cell deals with chemically aber-rant RNA: is it selectively degraded, left alone or repaired?

Now, Hans Krokan, Erling Seeberg and colleagues (Nature10.1038/01363) report comparative functional analysis of threemembers of the AlkB demethylase family; E.coli AlkB, and thehuman homologues hABH2 and hABH3. Biochemical analysiswith recombinant enzymes demonstrated that all threedemethylated 1-methyladenine and 3-methylcytosine in DNA.However, the three enzymes showed markedly divergent sub-strate preferences: whereas hABH2 was most active on double-stranded DNA, AlkB exhibited a moderate preference for single-stranded substrates and hABH3 showed a strong preference forsingle strands. Strikingly, both types of methylated bases werealso repaired in the context of RNA by AlkB and hABH3. Krokan

and colleagues then went on to show that the three demethylaseswill also function to re-activate methylated bacteriophages whenexpressed in AlkB-deficient E. coli. AlkB and hABH2 re-activatedthe single-stranded DNA phage M13, as well as the double-stranded λ phage. In contrast, and consistent with the in vitrosubstrate preferences, methylated forms of the single-strandedRNA bacteriophage MS2 were re-activated only by AlkB andhABH3, resulting in significantly increased phage survival.

Although it remains unclear if hABH3 can also function inRNA repair in eukaryotic cells, data is presented on the initialcharacterization of their subcellular localization in cultured cells.The enzymes are nuclear, but localize differentially, with hABH2showing some preference for nucleoli and relocating to replica-tion foci during S phase (see figure).

Altogether, these data may hint at a role for hABH2 duringDNA replication, whereas hABH3 and AlkB may be involved inmaintaining mRNA in a demethylated functional state. It isnotable that certain plant RNA viruses also seem to employ AlkBhomologues, lending credence for an involvement of enzymes ofthis family in RNA repair. Although direct support for the occur-rence and functional importance of RNA repair in eukaryoticcells will have to await future studies, this study opens a newchapter in the intricate book of nucleic acid repair.

BERND PULVERER

RNA repaired

Figure 1 Subcellular localization of hABH2. Transient expressionof hABH2–EYFP and ECFP–PCNA in HeLa cells.

NEXT MONTH Focus on Cell PolarityYeast po lar i ty Contrasted and compared

Membrane t ra f f ick ingThe plast ic i ty of epithel ial ce l l s

Par6, PKC and po lar i tyRegulat ing Lethal Giant Lar vae

The Focus on Cell Polarity

issue of Nature Cell Biology will be free

online for the duration of April 2003

www.nature.com/naturecellbiology

© 2003 Nature Publishing Group