This analysis identified 269 Arabidopsis proteins containing at least one copy of the WD motif. The vast majority of these (237) contained four or more recognizable copies of the motif. We classified these 237 proteins into 143 distinct families, 49 of which contained more than one Arabidopsis member. Approximately 113 of these families or individual proteins showed clear homology with WDR proteins from yeast, fly, and/or human (Table 1 [see Additional file 1]). Where conservation was found, it often extended across all of these organisms, suggesting that many of these proteins are components of basic cellular mechanisms.

The Arabidopsis proteome is apparently lacking counterparts of several WDR proteins that have been extensively studied in other eukaryotes and might have been expected to be conserved. For example, we found no protein related to the cell death initiator Dark (fly)/Apaf-1 (human). This protein is the central scaffolding of the apoptosome, a protein complex that activates specific cellular proteases in response to death signals 13. Many parallels exist between animal and plant apoptotic pathways, and many other components of the animal pathways have been strongly conserved in plants 14. Arabidopsis also appears to lack a protein closely related to the intermediate chain of the microtubule motor protein dynein, involved in transporting cellular cargo along microtubules. In mammalian cytoplasmic dynein, the intermediate chain plays a crucial scaffolding role, mediating interactions among the heavy chain and other dynein subunits 15. Arabidopsis was previously hypothesized to lack the dynein heavy chain, based on nearly-completed genomic sequence 16. It was suggested that if Arabidopsis did lack functional dynein, this could be compensated for by the relative variety of carboxyl-terminal motor domain kinesins in this species 17.

In contrast to these apparently lacking proteins, we found several proteins that were not expected. One example is a protein very closely related to Notchless (Nle), a fly protein that binds to the intracellular domain of the developmental signal receptor Notch and modulates its activity 18 (Fig. 1). Arabidopsis lacks a recognizable Notch, and other components of the associated signaling pathways appear to be absent 19. In addition, we found two proteins strongly related to the transcription-coupled, DNA-repair (TCR) proteins Rad28 (yeast) and Csa (human), even though plants are not known to undergo TCR (Fig. 1).

Also notable were several cases where Arabidopsis apparently did not participate in the expansion of gene families seen in the other eukaryotes. One example is a conserved component of the transcription factor TFIID, represented by the human TAFII-100 protein. This protein interacts directly with at least three other components of TFIID, and thus probably serves as a scaffolding for construction of the complex 1011. Multiple paralogs of this protein exist in fly, worm, and human (Table 1 [see Additional file 1] and not shown); in flies a form designated Cannonball (Can) appears to operate outside of basal transcription as a key regulator of spermatogenesis 20. One possibility is that the paralogs within each species act as a interchangeable components of the general transcription machinery to mediate expression of developmentally regulated target genes 21. The presence of only a single form of this protein in the Arabidopsis proteome suggests that this potential means of expanding the transcriptional repertoire has not evolved in plants. Another example of an evolutionarily stagnant family is Gβ, which exists as only a single form in Arabidopsis (Table 1 [see Additional file 1]; below). In mammals, each of the heterotrimeric G protein subunits, as well as the GPCRs, are encoded by multigene families, and combinatorial interaction among the proteins are believed to modulate much of the diversity of response to extracellular signals. The restriction of this gene family in Arabidopsis would suggest that, if a 'typical' heterotrimeric G protein does exist, it would likely lack the functional complexity seen in mammals. This scenario would be similar to that in yeast, where only a single heterotrimeric G protein, incorporating the Gβ protein Ste4, has a specialized role in transducing mating type signals from heptahelical mating-factor receptors 22.

In contrast, some other Arabidopsis WDR proteins show relatively expanded gene families compared with the other eukaryotes studied. One of the largest Arabidopsis WDR families, consisting of nine members, is orthologous to the conserved Cdc20/Fizzy class of cell cycle regulators including yeast Cdc20 and Cdh1 (Table 1 [see Additional file 1]). These proteins activate the anaphase promoting complex (APC) ubiquitin ligase, which targets downstream cell cycle regulators for proteolysis 23, potentially by mediating interaction of the APC complex with target proteins. Mutation in Cdc20 or Cdh1 affect distinct aspects of the cell cycle, and Cdc20 and Cdh1 coimmunoprecipitate with distinct APC target proteins 2425, indicating that these proteins have non-overlapping functions. One explanation for the expansion of this family in Arabidopsis is that the several distinct proteins each specify distinct targets for the APC. Another example of an expanded gene family is the MSI1/RbAp48 group of chromatin-related proteins, which includes five members in Arabidopsis (below), but is represented by only a single form in flies (Table 1 [see Additional file 1]).

Several examples were seen where Arabidopsis WDR proteins have used elements from the inceptive 'molecular toolbox' in original ways. One example is the pleiotropic developmental regulator LEUNIG (LUG) 26. LUG contains seven, carboxyl-terminal WD motif repeats, internal polyglutamine tracts, and an extended motif termed the single-stranded DNA-binding-protein (SSDP) motif (27, Fig. 1 and not shown). The SSDP motif was defined in a small family of animal proteins including chicken SSDP, which binds to a single-stranded, polypyrimidine region of the α2(I) collagen promoter 28. SSDP-like proteins function in transcriptional complexes with LIM homeodomain proteins and LIM-domain-binding proteins (Ldbs) to regulate specific embryonic developmental processes 29. The arrangement of the SSDP motif with carboxyl-terminal WD repeats appears to be unique to LUG and its orthologs from other plants (not shown). However, the juxtapositioning of polyglutamine tracts with carboxyl-terminal WD repeats, while diverging from the domain structure seen in the SSDPs, resembles that seen in the yeast transcriptional corepressor Tup1 and a related corepressor from fly, Groucho 27. This, in conjunction with the observation that loss of LUG activity leads to ectopic expression of a floral regulatory gene 30 has led to the speculation that LUG acts as a Tup1/Groucho-like transcriptional corepressor 27. Intriguingly, it was recently shown that LUG functions in floral development together with SEUSS (SEU), a protein related to the mammalian LIM-domain-binding protein, Ldb1 26, suggesting the existence of a LUG-SEU transcriptional complex analogous to that involving LIM proteins. Collectively, this information suggests that LUG participates in an evolutionarily distinct mechanism of gene regulation incorporating elements of both Tup1/Groucho and Ldbs.

With the exception of LUG, the Arabidopsis WDR proteins that have been functionally characterized are strongly conserved within the WDR regions among yeast, fly and/or human (Table 1 [see Additional file 1]). Most of these proteins have been identified as components of basic cellular machinery in these other eukaryotes, yet have been found to regulate plant-specific processes (Table 2). An interesting question for further consideration is how these proteins have become adapted to their plant roles.

In several cases, the homologous WDR proteins are highly conserved throughout the length of the proteins, and appear to operate in highly analogous mechanisms, with specificity in function conferred by changes in upstream signaling pathways and/or downstream effectors. One case is AGB1, the only clear Arabidopsis ortholog of Gβ 31 (Fig. 1). Loss of AGB1 function leads to developmental pleiotropy including shortened fruits 32 and changes in patterns of cell division in the hypocotyl and root 33. These phenotypes are associated with the derepression of genes that are normally turned on by auxin, suggesting a role for AGB1 as a negative regulator of auxin signaling 33. There appears to be one Gα-like protein (GPA1) and two Gγ-like proteins (AGG1 and AGG2) in the Arabidopsis proteome 31, and molecular modeling and yeast two-hybrid studies of potential interactions among AGB1, GPA1 and AGG1 are not inconsistent with the possibility that these could form a heterotrimeric protein 3334. In addition, both AGG1 and AGG2 contain domains expected to recruit AGB1 to membranes 34, and GPA1 has been demonstrated to bind GTP(γ)S 35. These findings lead to the prediction that AGB1 participates in a prototypical heterotrimeric G protein. However, the Arabidopsis proteome does not contain obvious heptahelical receptors with which a heterotrimeric G protein might interact 31. One possibility is that the AGB1-containing G protein might be unlinked from a receptor. In animals, several receptor-independent activators of heterotrimeric G proteins are known, including the Ras-related protein Ags1 36, and AGB1 might function in concert with any of the many Ras-related proteins in Arabidopsis. Alternatively, the Arabidopsis G protein may have evolved an interaction with a distinct type of receptor. Intriguingly, genetic experiments suggested that AGB1 may act in a common mechanism of fruit development with ERECTA (ER), a leucine-rich receptor kinase-like (LRR-RLK) protein, and based on this it was suggested that the two proteins could be functionally associated 32. Another link between a potential plant heterotrimeric G protein and LRR-RLKs was demonstrated by the observation that gpa1 mutants share the same level of insensitivity to GAs during seed germination as mutants for the LRR-RLK protein BRI1 33.

An example of downstream functional divergence is shown by CONSTITUTIVELY PHOTOMORPHOGENIC 1 (COP1). In dark-grown plants, this WDR protein acts as a key repressor of photomorphogenesis; in light, COP1 becomes inactivated, at least partly through gradual nuclear exclusion 37. The COP1 protein contains a carboxyl-terminal WDR domain, a RING Zn-finger domain near the amino terminus, and a predicted coiled-coil domain (Fig. 1). The WD domains mediate its interaction with, and negative regulation of, HY5 and HYH, closely related bZIP transcription factors 3839. This negative regulation potentially results from the ability of COP1 to target these proteins for ubiquitin-dependent proteolysis, because COP1 has been shown to function in this manner to downregulate a distinct regulatory protein in vivo 40. Humans and mice have a recognizable COP1 homolog, although yeast or flies do not (41; Table 1 [see Additional file 1], Fig. 1). Similar to the relationship between Arabidopsis COP1 and HY5 or HYH, human Cop1 binds to and negatively regulates Jun (and other related bZIP transcription factors) 42. The mammalian proteins conserve the important structural features of Arabidopsis COP1, including the RING finger, coiled-coil domain, and WDRs. When expressed in plant cells, mammalian Cop1 displays the light-specific nucleo-cytoplasmic partitioning characteristic of Arabidopsis COP1, but fails to rescue the cop1 mutant phenotype 41. Thus it appears that certain 'upstream' mechanisms regulating COP1 activity are conserved, but that downstream effectors of COP1 are not. This is consistent with the observation that components of the COP9 signalosome, which is required for COP1 localization, are conserved in mammals 43.

FASCIATA2 (FAS2) and the Arabidopsis family of Msi1-related proteins provide more examples of highly conserved WDR proteins involved in plant-specific processes through specificity in downstream targets. The FAS2 protein is the single Arabidopsis homolog of one of the three subunits (Cac1, Cac2 and another WDR protein, Msi1) of chromatin assembly factor 1 (CAF-1) 44, which participates in chromatin assembly following DNA replication or repair 45. FAS2 physically interacts with the only Arabidopsis ortholog of Cac1 and with AtMSI1, one of five paralogous Msi1-like proteins, and several lines of evidence suggest that Arabidopsis FAS1, FAS2, and at least AtMSI1 comprise a plant CAF-1 46. Mutations in FAS2 lead to meristem disorganization at the shoot and root apex, associated with the aberrant expression of two previously identified genes that organize the meristem, WUSCHEL in the shoot and SCARECROW in the root 44.

The conserved WDR protein subfamily represented by Msi1 and its Arabidopsis orthologs also includes human RbAp48 and fly p55 (Table 1 [see Additional file 1], Fig. 1). These latter proteins are CAF-1 subunits 47, but may serve supplemental functions. For example, both RbAp48 and p55 are found in several chromatin-remodeling complexes 48, and RbAp48 has been found tightly bound to the catalytic core subunit of the human histone deacetylase HDAC1 49. One possibility is that these Msi1-like proteins link chromatin assembly with the maintenance of gene silencing by recruiting activities that instigate the formation of localized heterochromatin 46. It is not unlikely that this is the case in Arabidopsis as well. Loss of AtMSI1 function leads to various developmental defects, at least some of these associated with the ectopic expression of the floral homeotic genes AGAMOUS and APETALA2 50. Loss of the paralogous AtMSI4 (also called FVE) leads to late flowering, associated with loss of transcriptional repression of the MADS-box flowering inhibitor gene FLC (51; J. Martinez-Zapater, pers. comm.). The observation that loss of AtMSI1 or AtMSI4 confer phenotypes that are distinct from those seen in fas2 mutants suggests that these proteins also probably have functions that are not limited to a role in CAF-1. Thus it seems likely that the FAS2 and Msi1 families of proteins are involved in equivalent regulatory mechanisms, with diversity in function due to the identity of the regulated gene(s).

A similar example is FERTILIZATION-INDEPENDENT ENDOSPERM (FIE). FIE is required for repression of endosperm development in the absence of fertilization and is also essential for proper endosperm and embryo development when fertilization occurs 52. The FIE protein has strong similarity to fly Esc and mammalian Eed, identified originally as polycomb-group repressors of homeotic gene expression 53 (Table 1 [see Additional file 1], Fig. 1). Esc and Eed interact directly with homologous SET-domain E(z) proteins in heterogeneous protein complexes also including the zinc-finger protein Su(z)12 or the histone deacetylase Rpd3 5455. Esc-E(z) polycomb complexes have been shown to repress gene activity, in part through their histone methyltransferase and histone deacetylase activities 5657. In a highly analogous manner, FIE binds directly to the SET-domain protein MEA, and genetic experiments suggest that a FIE-MEA pair interacts strongly with the Su(z)12-like protein FIS2 5859.

Mutations in FIE also cause misexpression of floral induction genes in the embryo and seedling, and consequent ectopic production of flower-like structures 60, suggesting that FIE has an additional and unrelated role in repression of the flowering program during the vegetative stage. Interestingly, this embryonic flowering resembles the phenotype of a mutant for a Su(z)12-like protein designated EMF2 61, and potentially FIE and EMF2 are components of a distinct Esc-E(z)-like complex. Although FIE has no apparent paralogs within the Arabidopsis proteome, there are multiple Su(z)12-related and E(z)-related proteins (not shown). Thus there is the potential for FIE to be involved in numerous aspects of growth and development through combinatorial interactions with several distinct complexes. Similarly, Eed, which also appears to be encoded by a single gene, interacts with multiple E(z)-like proteins 62.

Recently it was shown that FIE is required for transcriptional repression of PHERES1, a member the homeotic-function MADS-box gene class 63. It is notable that FIE and its homologs participate in an evolutionarily conserved mechanism to repress homeotic genes, even as the structure of homeotic genes have greatly diverged between plants (i.e., MADS-box) and animals (i.e., homeobox).

In almost all cases, proteins that we classified as homologous between species showed the highest degree of sequence similarity within the WDR regions, in several instances exhibiting little or no related sequence outside of the WDR region. In at least some of these cases, function of the homologous proteins may nevertheless be conserved. For example, the transcriptional corepressor Tup1 is highly variant outside of the WDRs among different strains of yeast, but is functionally interchangeable among them 64. In other cases, sequences outside of the WDR may confer a new functional specificity to the protein even as conserved protein interactions are maintained through the WDRs, potentially adapting basic cellular mechanisms to organism-specific processes. One potential example in Arabidopsis is PRL1, a nuclear protein implicated in the response to several stimuli including glucose, sucrose, cold, and multiple phytohormones 65. PRL1 interacts through its non-conserved amino-terminal domain with yeast Snf1, a conserved protein kinase that plays an important transcriptional role in sugar sensing and response, and with Arabidopsis Snf1-related protein kinases (SnRKs) 66. The carboxyl-terminal WDR domains of PRL1 exhibit very strong conservation with proteins from yeast (Prp46), fly, and human (Plrg1) (Table 1 [see Additional file 1], Fig. 1 and not shown). These non-plant proteins have not been related to Snf1-associated gene regulation, possibly as a result of divergence with PRL1 in the amino-terminal regions of the proteins. Instead, the yeast and human PRL1 homologs were identified as non-snRNP components of the spliceosome 6768, a large pre-mRNA splicing assembly consisting of RNAs, small nuclear ribonucleoproteins (snRNPs) and numerous other proteins. The WDR region of Prp46 and Plrg1 mediate interaction with other spliceosomal subunits that are structurally and functionally conserved in Arabidopsis (68 and not shown), and therefore it would not be surprising if PRL1 is also found to be associated with the spliceosome. A growing body of evidence points to the physical coordination of transcriptional regulation and downstream events including pre-mRNA processing 69, and thus PRL1 could link conserved spliceosomal components with plant-specific transcriptional regulators.

The flowering regulator FY provides another potential example of this type of recruitment of basic cellular processes for plant-specific functions by WDR proteins. In a developmental role closely related to that of AtMSI1/FVE (above), FY promotes flowering by repressing the flowering inhibitor FLC. FY physically interacts with the plant-specific, RNA-binding-motif protein FCA; this interaction is required for a negative autoregulatory mechanism controlling FCA protein accumulation, where full-length FCA protein promotes premature cleavage and polyadenylation of its own transcript 70. This autoregulation is evidently disrupted in the shoot apex of plants undergoing the transition to flowering, leading to increased accumulation of active FCA 71. In accordance with a role in RNA processing, the WDR region of FY exhibits strong homology with that of the aforementioned yeast RNA splicing factor Pfs2 and homologs from higher eukaryotes (Table 1 [see Additional file 1]; Fig. 1). Outside of the WDR region, these proteins diverge, with FY containing an extensive, proline-rich carboxyl-terminal region through which its interaction with FCA is mediated 71. It was suggested that this non-conserved portion of Pfs2/FY proteins in other eukaryotes might similarly link core RNA-processing machinery with regulatory proteins 71.

In our classification of Arabidopsis WDRs into gene families, we also noted several instances where proteins designated as paralogs diverged significantly outside of the WDR region. These proteins could link a common cellular mechanism with alternative or additional upstream regulators or downstream effectors, potentially in a temporal or spatially restricted manner. One notable case is SPA1, a nuclear-localized, light-dependent repressor of photomorphogenesis proposed to link the phytochrome A-specific branch of light signaling to a COP1-associated mechanism (Hoecker and Quail, 2001). Within its WDR domain, SPA1 is very closely related to COP1, but outside of this region SPA1 substitutes a SER/THR/TYR-protein-kinase-like domain for the RING-finger domain of COP1. SPA1 and COP1 physically interact through coiled-coil regions of both proteins 72. How the functional relationship between SPA1 and COP1 is related to this molecular linkage is not known, but genetic and biochemical studies suggest that SPA1 may somehow stimulate the ability of residual nuclear COP1 to target proteins for destruction after the onset of illumination 40.