Some mechanisms involved in cell morphogenesis, such as membrane vesicle transport, are conserved at least among crown eukaryotes (metazoa, fungi and plants) [1,2], whereas others, such as those involving extracellular structures or the precise roles of different Rho-like GTPases [3], are not. Yet other cellular processes, such as cytokinesis, often recruit conserved proteins to accomplish superficially dissimilar tasks (for example, budding, cleavage or phragmoplast-based cell division of plant cells) [4]. For many morphogenetic mechanisms, the question of evolutionary conservation remains unresolved because available information is limited to one or a few model organisms. For example, this is the case for the molecular mechanisms that ensure the communication between the cytoskeleton and the surface of the cell. However, the recent increase in the data available from a number of genome projects allows wide-ranging searches for homologs of known components of signaling and morphogenetic pathways. The results of such searches can lead both to experimentally testable hypotheses and to general conclusions regarding the evolution of morphogenetic processes.

Formins, also known as formin homology (FH) proteins, are proteins implicated in cellular and organismal morphogenesis of both metazoa and fungi. On the cellular level, they are involved in the establishment and maintenance of cell and/or tissue polarity [5,6], in cytokinesis [4] and in the positioning of the mitotic spindle [7]. They interact directly or indirectly with actin, profilin, Rho-like GTPases [5,6,8,9,10,11], the yeast Spa2 protein and septins [12,13], proteins containing SH3 or WW domains [10,14], dynein and microtubules [7,15,16,17]. The yeast formin homolog encoded by BNI1 is localized to the cell periphery and participates in positioning cortical actin patches towards distinct regions of the plasma membrane [5,13,18]. Some kind of contact with the plasmalemma (in addition to that mediated by a Rho-like GTPase) might therefore be expected, although there is no evidence as yet for such a contact. Furthermore, metazoan formins are believed to be cytoplasmic or nuclear proteins [19,20].

Nothing is known about formin function in plants, although the existence of two Arabidopsis thaliana proteins containing the conserved formin-homology 2 (FH2) domain has been reported recently [6,10]. Given that all known formins represent a well-defined family, this class of proteins may be a good candidate for a systematic genome sequence search. Here, I present the results of such an approach, which has led to the identification of putative plant formin genes, as well as to the finding that the evolutionarily old formin domain may be used in a number of different ways and contexts ('modules' as defined by Hartwell et al. [21]) by recent eukaryotes.

Formins are defined by the presence of two sequence domains-the low-complexity, proline-rich FH1 and the carboxy-terminal FH2 [6,10,22]. A third domain-the amino-terminal FH3 motif-has been characterized biochemically but is rather poorly delimited in sequence terms [23]. Despite a conflicting consensus definition, this motif appears to be identical to the amino-terminal conserved block found in some formins by Wasserman [10]. I have used the L-x-x-G-N-x-M-N (single-letter amino-acid notation; x is any amino acid) motif present in the FH2 domain of most fungal and metazoan formins[10] to search for putative Arabidopsis formin homologs and found eight such inter-related genes (see Materials and methods and Table 1). All of them correspond either to hypothetical open reading frames (ORFs) or to unannotated genomic or cDNA clones, indicating that at least some of them are expressed in vivo. These putative genes and their predicted protein products will be referred to henceforth as AtFORMINs 1 to 8.

Sequence comparison with known formins revealed the presence of genuine FH2 domain in all Arabidopsis formins (Figure 1). However, even the longest predicted proteins, encoded by the AtFORMIN 3, -4 and -5 genes, lack parts of the FH2 region ubiquitously conserved among corresponding genes of fungi and metazoa (Figures 1 and 2), although not necessarily among their protein products, because some formin mRNAs undergo complex splicing [24]. Sequence motifs corresponding to the missing regions were found in all cases within the predicted introns by visual inspection of three-frame translation data. Because the reliability of mRNA structure prediction is limited [25], failure to identify exons correctly may explain the apparent deletion of this region of the FH2 domain. The possibly mispredicted intron encoding subdomain g of AtFORMIN4 is split by a frameshift mutation, however. Although this could reflect a sequencing error, the possibility remains that plant formin homologs have a modular structure within the FH2 domain at the gene level, and that at least some of the FH2-related sequences within predicted introns are vestiges of exons lost by mutation.

Proline-rich regions corresponding to FH1 were identified in all Arabidopsis formins. Surprisingly, there are two such regions in AtFORMINs 2, 6 and 8-a feature not observed in the non-plant formins examined (listed in Materials and methods). Neither motifs corresponding to FH3 nor coiled-coil regions flanking FH1 (common but not ubiquitous in non-plant formins [10]) were found. The structure of FH2, the overall protein size (smaller than most non-plant formins) and the domain layout of Arabidopsis formins therefore show possible plant-specific features (Figure 2). This idea is supported by the topology of an evolutionary tree that consistently places Arabidopsis formins in a branch separate from other members of the formin family (Figure 3).

As in the non-plant formins, the amino-terminal portions of all Arabidopsis formins are divergent, although there is 63% identity between AtFORMINs 1 and 4 in the overlaping parts of their sequences. Analysis of AtFORMIN sequences with SMART [26,27] revealed no previously characterized domains outside the FH2 region. However, putative amino-terminal membrane insertion signals (signal peptides) followed by a segment highly likely to be membrane-spanning and a variable number of possible transmembrane domains were found in AtFORMINs 1, 2, 4, 6 and 8. A possible membrane insertion signal was also identified in AtFORMIN5 by one of the two methods used (see Materials and methods, and Figures 2,4). The length of predicted signal peptides suggests that they may represent membrane anchors rather than secretion signals [28]. A putative transmembrane segment was also found in the apparently amino-terminally truncated sequence of AtFORMIN3. In contrast, no signal peptides were found in 12 fungal and animal formins listed in Materials and methods, although transmembrane-like segments were observed in some. Surprisingly, the putative transmembrane segment lies between the two Pro-rich regions in AtFORMINs 2, 6 and 8. Obviously, only the cytoplasmic one of these two motifs can act as a conventional FH1 domain. Its size ranges from 106 to 423 amino acids, with proline content of 13 to 41% and multiple stretches to five to nine consecutive proline residues. This structure roughly corresponds to that of previously characterized FH1 domains [10]. Interestingly, the FH1 domains of AtFORMINs 2, 7 and 8 are extremely rich in serine (up to 20%) and contain stretches of up to seven consecutive serine residues.

The other proline-rich domain of AtFORMINs 2, 6 and 8 is predicted to be exposed to a non-cytoplasmic compartment. Given that polyproline stretches are characteristic for a class of structural cell-wall proteins known as extensins [29], it is tempting to speculate about a possible role for this domain in communication between formins and structures within the cell wall. Apart from this, few predictions of function can be made on the basis of the sequence data. Although formins are well conserved with respect to their molecular structure, we do not know the extent of their conservation within signaling or structural modules [21]. As the relationships between protein structure, module structure and biological function are far from straightforward [30], we can at present neither prove nor exclude the possibility that plant formins contribute to similar functional modules to their animal and fungal counterparts. The question of whether these proteins have a direct role in cytokinesis, in mitotic spindle localization, or in some other cellular process, possibly involving cytoskeleton rearrangement or cell-surface growth, will have to be answered experimentally.

A systematic search of the available Arabidopsis genomic and cDNA sequences revealed the presence of eight genes encoding proteins that define a novel subfamily of the formin family. At least six out of eight Arabidopsis formins appear to be integral membrane proteins. This indicates a mechanism of membrane localization that may be specific to plants and functionally related to a possible role for formins in the communication between the plant cell and extracellular structures.