Most vertebrates have one ADF and two cofilins; the latter are divided into muscle and non-muscle cofilins [8]. The reported human destrin-2 gene (Genbank U72518) is most likely to be a pseudogene [9]. The frog Xenopus expresses two ADF/cofilins, but these appear to be more closely related to the cofilins than the ADFs; the possibility of a Xenopus ADF cannot presently be excluded, however. If there is no Xenopus ADF, this may indicate that the ADF and cofilin lineages may have diverged in the reptilian common ancestor of birds and mammals. Only one cofilin is found in chicken and this is more similar to the mouse muscle cofilin (96.4% identity) than it is to the mouse non-muscle cofilin (81.3%).

Of the invertebrates, Caenorhabditis elegans has one ADF/cofilin gene, unc-60, which encodes two different proteins, UNC-60A and UNC-60B [10] (see below). Drosophila has one ADF/cofilin gene, twinstar [11]. The first ADF/cofilin sequence to be determined, that of depactin, which was isolated from eggs of the sea star Asterias amurensis [12], was determined by direct amino-acid sequencing of the protein [13], and to date no supporting cDNA or gene sequence is available. Although a putative ADF/cofilin gene from another echinoderm, the sea urchin Strongylocentrotus purpuratus, is available, this sequence does not group with depactin. In fact, depactin is the most divergent member of the group so far discovered (see Figure 1).

A surprising finding is that plants have many more ADF/cofilin genes than animals. Using a limited data set, Mun et al. [3] classified the plant ADF/cofilins into four groups (I-IV); our analysis (Figure 1) supports this classification and we have also subdivided groups I and II into two subgroups and group III into three subgroups. Some indications of a separation of the plant ADF/cofilins along the lines of the major plant groups (gymnosperms, angiosperms, monocots, and dicots) is evident: group I is composed exclusively of dicots (although there is a rice gene similar to Petunia hybrida ADF1 on chromosome 3; GenBank accession number AC084320), whereas group III contains both dicots and monocots. Group II contains dicots, monocots and gymnosperms, and group IV presently includes Zea mays ADF3 and an ADF/cofilin from wheat (some trees placed these ADF/cofilins more closely than in Figure 1). Southern blot analysis [14], probing with the wheat ADF/cofilin, reveals the presence of similar sequences in all the monocots tested, Secale cereale, Avena sativa, Hordeum vulgare, Oryza sativa and Zea mays (the latter sequence is presumably ADF3), whereas the dicots tested, Medicago sativa and Brassica napus, did not hybridize, indicating perhaps that group IV is exclusive to the monocots [14]. It is possible that group II is exclusively pollen-specific and that, within this group, monocots and dicots form subgroups [6,7]. Members of group IIIc (the third subgroup of group III, see Figure 1) have an insert of various lengths between sheet 6 and helix 4 (see Characteristic structural features), for no presently apparent purpose.

The Arabidopsis thaliana genome sequencing project is complete, so it is possible to analyze the full complement of ADF/cofilin genes from this plant. Although Arabidopsis has a genome size only 4% that of humans, it has 12 ADF/cofilin genes (AtADFs). It is not yet clear how many of these are expressed, but cDNAs have been isolated for most [15]. Two pairs of AtADF gene products are very similar (AtADF1 and AtADF4, and AtADF8 and AtADF10), making it likely that their functions may be redundant. The phylogenetic analysis (Figure 1) predicts that AtADF7 and perhaps AtADF8 and AtADF10 are pollen-specific, as maize and lily pollen-specific ADFs fall in the same grouping as these three AtADFs. The ADF genes of Arabidopsis are clustered: AtADF3 and AtADF4 are adjacent on chromosome 5, and a putative ADF gene is followed by AtADF2 and AtADF1 on chromosome 3

Compared with animals and plants, there are relatively few ADF/cofilins characterized from other eukaryotes, which limits our interpretation of the evolution of the ADF/cofilin genes (Table 1, Figure 2b). There is only one ADF/cofilin sequence in the fully sequenced Saccharomyces cerevisiae genome, and there is evidence for a single ADF/cofilin gene (actophorin) in the soil amoeba Acanthamoeba castellanii [16]. It was previously suggested on similar evidence, however, that there was only one cofilin gene in Dictyostelium, but more recently the sequence of another Dictyostelium cofilin-like gene, cofilin-2, has been deposited in GenBank (accession number AB055926) by the same group that cloned cofilin-1. The inclusion of this sequence in our phylogenetic analysis has the effect of removing the cofilin-1 sequence from its present position within the tree to an outlying group with cofilin-2. As the cofilin-2 gene has this effect and because it has not been verified as being an ADF/cofilin member, it has not been included in our analysis. Acanthamoeba actophorin most closely resembles the plant ADF/cofilins of the limited number of phyla included in the study; a kinship between Acanthamoeba and plants is suggested in many (but by no means all) ribosomal DNA analyses.

The coccidians, including the bird parasite Eimeria tenella and the cat and human parasite Toxoplasma gondii, appear to have two ADF/cofilins; only one ADF/cofilin gene has been reported in Toxoplasma gondii [17], but at least two differentially spliced forms are found in expressed sequence tag (EST) databases (GenBank BG658910, BG659044 [18]). (Although the actin-binding function of the Eimeria ADF/cofilin protein has not been published, it is similar to Toxoplasma ADF/cofilin, which is a confirmed ADF/cofilin member in terms of its interaction with actin.) The ADF/cofilin sequence from Cryptosporidium parvum is a puzzle, because being from another protozoan (an apicomplexan), it would be expected to group with T. gondii, but instead, it appears in our analysis to group loosely with the nematode C. elegans. Some trees generated in our analysis do (suggest a relationship between Toxoplasma and Cryptosporidium. More sequences are of course needed to resolve this puzzle. A partial sequence from another apicomplexan, Sarcocystis neurona (GenBank BE636150, not included in our analysis), is related to mammalian cofilins, adding to the confusion. This sequence may have been 'picked up' at some point by horizontal transfer as the parasite moved between hosts.

The intron-exon boundaries often provide information on the ontogeny and evolution of genes. As expected, there are several such boundaries within ADF/cofilin genes, and these are preserved across the phyla. A remarkable tendency for ADF/cofilin genes is for the first amino acid (or the first few) to be encoded by a separate exon (Figure 2b). The human muscle cofilin gene (Clf2) produces two different mRNAs that encode identical polypeptides by the use of two alternative first exons encoding the methionine and upstream untranslated region; these mRNAs presumably differ in their localization and/or stability [19]. The opposite is true for the muscle ADF/cofilin of the nematode C. elegans: two different ADF/cofilin proteins are produced from one gene, although the only exon to be shared is that encoding the initiating methionine. The S. cerevisiae Cof1 gene contains one exon in the region encoding the amino terminus of the protein [20], as does one of the two genes encoding identical proteins in Dictyostelium discoideum. Several ADF/cofilin genes, for example those from Schizosaccharomyces pombe, Entamoeba histolytica and Strongylocentrotus purpuratus), have no introns, but some of these have yet to be shown to be functional genes. Genes that contain no introns are likely to be pseudogenes [21,22], so those ADF/cofilin genes identified solely on the basis of their genomic sequence (such as those from E. histolytica and S. purpuratus) must be verified by cDNA cloning. This rule also appears to hold for human ADF genes; a number of pseudogenes homologous to ADF/cofilin genes lacking introns are suspected (such as those with GenBank accession numbers AC009498 (chromosome 2) and AL132765 (chromosome 20)). As far as can currently be determined, plant ADF/cofilin genes are organized in a similar manner, with an intron following the exon encoding the amino terminus and a conserved intron further 3'. This pattern holds for Arabidopsis and Oryza sativa ADF/cofilin genes.

ADF/cofilins are usually localized in parts of the cell where there is a high turnover of actin filaments, such as the leading edge of moving animal cells [16,31,32,33] and the growing tips of plant cells [26]. The main activity of ADF/cofilins has been found from in vitro experiments to be to increase actin-filament turnover [5,34,35]. They accomplish this by severing actin filaments and increasing the rate at which actin monomers leave the pointed end of actin filaments (see below). The rate at which actin filaments depolymerize is the rate-dependent step in the overall turnover of filaments that comes about as cells move forwards [36]. Cells lacking cofilin have impaired locomotion [37], and those over-expressing cofilins are more motile [38]. The effects are specific to certain types of actin filaments: older filaments (those at the base of leading lamellae) are 'marked' for turnover; the mark arises because they tend to contain more ADP-actin monomers and it is with these that the ADF/cofilins preferentially interact [34,35]. ADF/cofilins are also necessary for cytokinesis, depolymerizing the contractile ring between daughter cells as it contracts. ADF/cofilins localize to the contractile ring [39], and cells lacking ADF/cofilins are defective in cytokinesis [11].

In addition to their role in microfilament recycling, ADF/cofilins are also found in actin-rich, spicule-like rods found in stressed cells, in both the cytoplasm and the nucleus [26,40]. ADF/cofilins are also targeted to the nucleus upon heat shock and chemical stress. It may be that actin is taken into the nucleus in this manner so that a pool of tightly packed actin is protected from denaturation, and is then available after the stress is removed. ADF is known to inhibit actin denaturation, supporting this hypothesis [41].

The localization of ADF/cofilins in plant cells is broadly similar to that in animal and protist cells - they are primarily concentrated in regions rich in dynamic actin structures - but pollen and vegetative ADFs appear to have different properties. Pollen ADF has been seen to bind filamentous (F-) actin in vivo in mature pollen, dehydrated pollen and at adhesions between the tip of the pollen-tube and an adjacent substrate. Taken together with the fact that lily pollen ADF has an inefficient actin-depolymerizing activity, these data suggest that pollen ADFs serve to bind and remodel F-actin structures, presumably in cooperation with other actin-binding proteins [42]. In contrast, given that the maize vegetative ZmADF3 locates to the tip of growing root-hair cells, is not seen to co-localize with F-actin in vivo and has an effective actin-depolymerizing activity, its principal role appears to be to increase the turnover of actin filaments. In root-hair cells, the effect of increased actin dynamics at the hair tip would be to promote root-hair growth [26].

In vertebrates, a single ADF gene is expressed in most tissues [32], and ADF tends to have a reciprocal pattern of expression compared with the cofilins, with either the cofilins (generally) or ADF being more abundant. Both ADF and non-muscle cofilin are abundant in brain, both expressed at very low levels in liver and mature muscle [43]. The pattern of expression for most of the AtADFs has yet to be determined, but AtADF and AtADF4 are expressed in the vascular tissues in the entire plant and AtADF5 is expressed at the tip of the root meristem [15]. Dictyostelium Cofilin-2 is expressed specifically at the aggregation stage of Dictyostelium development.

The ADF/cofilins appear to have multiple functions, and this is reflected in their very complex association with monomeric and filamentous actin. They depolymerize actin filaments during, for example, cytokinesis [11,39], cell locomotion [36,37], and plant-cell elongation [26], in addition to being involved in cellular stress responses [44] and pathological situations [45]. ADF/cofilins are regulated by pH [31,46,47], polyphosphoinositides [16,27,28,31], phosphorylation [48,49,50,51], nucleotides bound by actin [36] and the presence of other actin-binding proteins [52,53,54]. They are, so far, unique among the actin-binding protein families in that they alter the twist of the actin filament [55]. ADFs and cofil-ins have very similar properties in vitro, but are present in varying relative concentrations in cells and, where they appear in the same cell as each other, interesting differences in behavior have been noted [31]. Rather surprisingly, the distribution of ADF but not cofilin is modulated by intracellular pH in mouse cells.

The two ADF/cofilins encoded by the C. elegans unc-60 gene, UNC-60A and UNC-60B [10], have distinct actin-binding properties, but understanding this is further complicated by the discovery that UNC-60B behaves differently with respect to its interaction with rabbit muscle actin and actin from the nematode itself [56]. The most dramatic difference is that UNC-60A binds much more weakly to F-actin at pH 7.0 than does UNC-60B [10]. The carboxy-terminal domain of UNC-60B is essential in F-actin binding and it has been postulated to constitute a second actin-binding site [57]. The actin-binding properties of Drosophila Twinstar [11] have not yet been characterized.

The mechanism by which actin filaments are depolymerized by ADF/cofilins has been controversial and the details are still far from clear. Filaments are depolymerized by severing and by an increase in the rate at which actin monomers fall off the pointed end of the actin filament. Phosphorylation is a principal regulator of ADF/cofilin function: ADF/cofilins are phosphorylated on an amino-terminal serine (Ser3 in human non-muscle cofilin) by LIM kinases 1 and 2, TESK 1 [58] and TESK 2 [59], and maize ADF3 is phosphorylated by a calmodulin-like domain protein kinase [50,51]. Phosphorylation by all these kinases prevents ADF/cofilins from binding actin (Figure 4).

Many ADF/cofilins, including vertebrate ADF and cofilins [28], Acanthamoeba actophorin [16], Zea mays ADF3 [27], and Saccharomyces cerevisiae Cof1, have been found to bind PIP₂ and, to a lesser extent, phosphatidylinositol-4-phosphate. Some of the actin-binding interfaces of ADF/cofilins partially overlap with the binding site of PIP₂ [30], explaining why PIP₂ dissociates the actin-ADF/cofilin complex. In turn, ADF/cofilins reciprocally affect the metabolism of the polyphosphoinositides. Vertebrate cofilins [29] inhibit the hydrolysis of PIP₂ by phospholipase C, as does Zea mays ADF3 [27]. Binding of ADF/cofilins by PIP₂, and perhaps by ion channels, may help to localize ADF/cofilins to the membrane, where they function to increase actin-filament turnover as well as to modulate PIP₂ metabolism.

Both Acanthamoeba actophorin [60] and sea star depactin [61] have been reported not to be pH-sensitive, although they are in other respects typical ADF/cofilins. No obvious relationship between sequence and pH sensitivity is yet apparent, and pH dependence has been reported for many ADF/cofilins, including vertebrate ADF [41,47] and cofilins [46], Arabidopsis thaliana ADF1 [34], Zea mays ADF3 [27], and the ADF/cofilins of Saccharomyces cerevisiae [20], Petunia hybrida [3], Triticum aestivum [14], and the acomplexan Toxoplasma gondii [17].