The Rho family of small GTPases comprises some 21 genes in humans, encoding at least 23 signalling proteins. Although these proteins control an amazingly diverse range of cellular functions, one general role is in the establishment of polarity and of polarised structures through dynamic regulation of the actin cytoskeleton. This theme is carried through all three eukaryote kingdoms - from bud formation in S. cerevisiae, to pollen tube elongation in Arabidopsis, to the formation of complex structures such as cochlear stereocilia in mammals. Rho GTPases control the polymerisation, branching and bundling of actin, allowing them to regulate the remodelling of the actin cytoskeleton into distinct architectural elements. Spatial and temporal control of these elements allows Rho GTPases to direct complex mechanical processes such as cell motility and phagocytosis (Bishop and Hall, 2000; Hall, 1998).

The best-characterised family members are RhoA, Rac1 and Cdc42 (human nomenclature). Each controls the formation of a distinct cytoskeletal element in mammalian cells; RhoA stimulates the bundling of actin filaments into stress fibres, Rac reorganises actin to produce membrane sheets or lamellipodia, and Cdc42 is associated with the formation of thin, actin-rich surface projections called filopodia. These family members illustrate the high level of conservation of both structure and function through eukaryote evolution. C. elegans and Drosophila have homologues of all three small GTPases, and downstream effector proteins are also conserved. Rac is absent from yeast, but both the Cdc42 and RhoA (Rho1) homologues are present. The coupling of the yeast Rho1 to the Pck1 kinase is seen in worms, flies and humans in the interaction of RhoA with the Pck1-related PRK/PKN kinases.

In humans these three archetypal family members form subgroups of related proteins. RhoA has two highly related homologues: RhoB and C. Examination of the structures of the genes that encode these proteins suggests that RhoC arose from duplication of the RhoA gene, whereas the intronless RhoB gene appears to represent a retrotransposition of RhoA. Similarly, distinct subfamilies of Cdc42-like (TC10 and TCL) and Rac1-like isoforms (Rac2 and Rac3) are present. In many cases members of these subfamilies share subsets of downstream effectors. Thus, the value of these duplications to cell function may come from the differential expression and/or subcellular localisation of subfamily members.

The conservation of RhoA, Rac and Cdc42 through evolution is striking but gives a somewhat misleading impression of the development of this signalling family as a whole. Comparison of the Rho GTPase family structure between species shows that the overall picture is one of great plasticity, individual species gaining and losing family members to give rise to unique sets of signalling proteins. For example, S. cerevisiae Rho3, Rho4 and Rho5 have no apparent homologues in the other species shown; nor do their known interacting proteins suggest any obvious parallels. Perhaps the clearest example of Rho family plasticity comes from the expansion of the Rac subfamily in plants. Arabidopsis thaliana has no apparent RhoA homologue, nor any proteins containing RhoA-binding motifs. Neither does this species have any obvious Cdc42 homologue or homologues of Cdc42 or Rac effector proteins such as WASP and PAK. However, Arabidopsis has 11 Rho family GTPases that appear very distantly related to the Rac GTPase; these are generally highly related (between 65-99% identical) to each other, but show only weak sequence similarity to Rho GTPases in other species (see poster). The social amoeba Dictyostelium discoideum has no Cdc42 or RhoA homologue, but does have a group of Rac-like proteins (Rac1a, Rac1b, Rac1c, RacF1 and RacF2) and homologues of WASP and PAK. However, Dictyostelium also has a host of other more esoteric Rho family isoforms that appear to lack equivalents in other species.

One case that clearly supports loss of a family member is that of the RhoBTB proteins. The majority of Rho family members are 21 kDa proteins terminating in a C-terminal CAAX motif that becomes modified by prenylation. In the highly unusual RhoBTBs, this CAAX motif is replaced by an approximately 400-residue extension that includes a BTB (for Broad-Complex, Tramtrack and Bric à Brac) domain. Dictyostelium has a single RhoBTB (RacA), Drosophila has one (RhoBTB) and humans have three. However, this highly distinctive Rho GTPase is clearly absent from C. elegans. Similarly, the Mig-2 GTPase, which controls axon guidance in C. elegans and Drosophila (Mtl), has no apparent homologue in the human genome.

It would seem that gene duplication and divergence has allowed expansion of the Rho GTPase family to continue late into evolution. The overall picture is one where family favourites may be retained and expanded into closely related subfamilies, but these exist side by side with a host of divergent proteins that presumably perform specialised functions. This places important constraints on our interpretation of data from studies of different model organisms. While some pathways are clearly conserved, others are misleadingly not. In yeast the Rho1 activation of Pck1 signals to a well-defined MAP kinase pathway that responds to hyposmotic stress. In mammals, RhoA activates the analogous PRK kinase, but a downstream MAP kinase pathway is apparently absent. Similarly, Wilkins and Insall have pointed out that while Dictyostelium shares many aspects of mammalian cell motility, it does so without RhoA or Cdc42 (Wilkins and Insall, 2001). It seems that often the picture may look highly familiar but the brushstrokes differ.

Why so many different Rho GTPases? Perhaps this reflects the successful design of Ras-related small GTPases. The basic molecular switch mechanism presents a rich interaction surface to signalling partners on activation of the protein. While the switch mechanism itself is highly conserved, the interacting surface can easily be varied, with the consequent potential to create new signalling junctions. We also know from structural studies that different effectors can interact with different regions of this surface, allowing several diverging pathways to be driven by a single family member.

A harder question to explain is the clustering of function of Rho family members. With related Rho GTPases carrying out seemingly highly related functions in cytoskeletal regulation, it is easy to forget that the signalling intermediates are a diverse and structurally unrelated set of proteins. Why should these downstream partners so frequently end up signalling to the actin cytoskeleton?

Finally, of the 21 human Rho GTPases, only RhoA, Rac and Cdc42 have been studied in detail. It seems unlikely that any of the 19 uncharted isoforms could possibly have the amazing breadth of function of the big three. However, it is a racing certainty that each will do something unique and interesting. This is an exciting time to be working on this fascinating family of signalling proteins.

FIG1 

We thank Richard Sessions (University of Bristol) for help with the ribbon diagram of RhoA structure. HM holds a Wellcome Trust Research Career Development Fellowship. MW is supported by a BBSRC Committee Studentship.

Bishop, A. L. and Hall, A. (
2000
). Rho GTPases and their effector proteins.
Biochem. J
.
348
,
241
-255.
Hall, A. (
1998
). Rho GTPases and the actin cytoskeleton.
Science
279
,
509
-514.
Wilkins, A. and Insall, R. H. (
2001
). Small GTPases in Dictyostelium: lessons from a social amoeba.
Trends Genet
.
17
,
41
-48.