Bioinformatic analyses implicate the collaborating meiotic crossover/chiasma proteins Zip2, Zip3, and Spo22/Zip4 in ubiquitin labeling

Abstract

Zip2 and Zip3 are meiosis-specific proteins that, in collaboration with several partners, act at the sites of crossover-designated, axis-associated recombinational interactions to mediate crossover/chiasma formation. Here, Spo22 (also called Zip4) is identified as a probable functional collaborator of Zip2/3. The molecular roles of Zip2, Zip3, and Spo22/Zip4 are unknown. All three proteins are part of a small evolutionary cohort comprising similar homologs in four related yeasts. Zip3 is shown to contain a RING finger whose structural features most closely match those of known ubiquitin E3s. Further, Zip3 exhibits major domainal homologies to Rad18, a known DNA-binding ubiquitin E3. Also described is an approach to the identification and mapping of repeated protein sequence motifs, Alignment Based Repeat Annotation (ABRA), that we have developed. When ABRA is applied to Zip2 and Spo22/Zip4, they emerge as a 14-blade WD40-like repeat protein and a 22-unit tetratricopeptide repeat protein, respectively. WD40 repeats of Cdc20, Cdh1, and Cdc16 and tetratricopeptide repeats of Cdc16, Cdc23, and Cdc27, all components of the anaphase-promoting complex, are also analyzed. These and other findings suggest that Zip2, Zip3, and Zip4 act together to mediate a process that involves Zip3-mediated ubiquitin labeling, potentially as a unique type of ubiquitin-conjugating complex.

A prominent feature of meiosis is a programmed progression of local interhomolog interactions that culminates in formation of DNA crossovers and their cytological counterparts, chiasmata. Recombination is initiated by the occurrence of a large number of programmed double-strand breaks (DSBs) (1). Only a few of these DSBs mature into crossovers/chiasmata. The remaining majority interact with a homolog partner but ultimately are resolved without any accompanying exchange, i.e., as “noncrossovers.”

The decision as to which recombinational interactions will mature as crossovers and which will mature as noncrossovers appears to be made at late leptotene of meiotic prophase and to trigger chromosomal progression through the ensuing “leptotene/zygotene” transition (2). At the time this decision is made, DSBs have occurred and have probably begun to interact with their partner duplexes; concomitantly, at the chromosome structure level, axes have formed and homologs are linked by bridge structures that support corresponding recombination complexes. Once a particular recombinational ensemble has been designated to become a crossover, further progression occurs at both levels. At the DNA level, the nascent DSB/partner interaction progresses to an intermediate involving stable exchange of strands between the interacting duplexes at one end of the break (the “single-end invasion” or SEI) (3). At the structural level, bridges disappear, accompanied by nucleation of the synaptonemal complex (SC), which then polymerizes outward from these sites. Fully formed SC is a close-packed array of transverse filaments that links chromosome axes along their entire lengths (reviewed in ref. 4).

Molecular genetic studies have shown that DSB-containing recombinational interaction complexes designated for future maturation into crossovers are subject to a specifically programmed progression block, which is then overcome to permit formation of SEIs and nucleation of SCs. Progression through this block requires the coordinate action of a particular set of proteins, the ZMM group, which then also mediates the ensuing DNA and structural outcomes (2). Members of the ZMM group identified thus far are Mer3, Msh4/5, Zip1, Zip2, and Zip3 (2). The entire ZMM ensemble likely acts directly at the sites of developing crossovers/chiasmata, at this and later stages, as indicated by coimmunoprecipitation, two-hybrid interactions, and immunocytological studies (5, 6).

Biochemical functions can be assigned for only some of the ZMM proteins: Mer3 is a DNA helicase, Msh4/5 is a meiosis-specific MutS homolog that does not mediate mismatch repair, and Zip1 is the transverse filament component of the SC (reviewed in refs. 2 and 7). There is almost no information regarding the molecular functions of Zip2 or Zip3.

Here we identify Spo22 as an additional member of the ZMM group. Spo22 has also been studied by the Roeder laboratory, who have renamed it Zip4 (G. S. Roeder, personal communication). We also present bioinformatic analyses that, in combination with other information, lead to the proposal that Zip2, Zip3, and Spo22/Zip4 are components of a meiosis-specific complex for protein labeling. Available evidence points to ubiquitin as the ligated moiety, although SUMO (small ubiquitin-like modifier) cannot be rigorously excluded as a possibility. The presented analyses depend in part upon Alignment Based Repeat Annotation (ABRA), an approach to the identification and mapping of repeated sequence motifs within a set of related but evolutionarily diverged proteins. ABRA is described and then applied to Zip2, Spo22/Zip4, and structurally related components of the anaphase-promoting complex (APC). A model for Zip2/3/4 as a unique ubiquitin ligation complex is discussed.

Materials and Methods

A transposon insertion library (8) was screened by using the homothallic ste7-ts strain NKY3643 (HO/HO, MATa/MATα, ste7-1/ste7-1, ura3/ura3, leu2/leu2, ade2::LK/ade2::LK) by the methodology of McKee and Kleckner (9): A population of haploid cells was transformed and plated under selective conditions that also permit self-diploidization. Resulting diploid colonies were screened for spore formation at 23°C and 33°C by replica plating to sporulation medium and visualization of UV fluorescence (2). Fig. 1A used spore clones of homothallic strain NKY3644, which is NKY3643 (above) also carrying spo11(Y135F)-HA-URA3/SPO11, spo22::Hyg4MX/SPO22. Fig. 1 B–D used NKY3639 (wild type) and an isogenic derivative carrying a complete deletion of SPO22 (NKY3645) exactly as described (2).

Fig. 1.

Characterization of spo22Δ/zip4Δ. (A) Spore formation as visualized by UV fluorescence of cell patches replica plated onto sporulation medium. spo22Δ shows temperature- and SPO11-dependent failure in spore formation. (B–D) Events of meiosis in cultures undergoing synchronous meiosis, exactly as previously described (2). (B) Completion of the first meiotic division (MI) as indicated by appearance of two chromosome masses. (C and D) Appearance of crossover (CR) and noncrossover (NCR) recombination products at the HIS4LEU2 hot spot. Gels in C are quantitated in D, which also includes data for a previously analyzed zmm mutant (zip2Δ) taken from ref. 2.

Results

Identification of Spo22 as a Member of the ZMM Group. Mutants lacking a ZMM protein exhibit specific temperature-modulated meiotic phenotypes. At higher temperatures: reduced formation of crossover recombination products (≈15% of the wild-type level), normal formation of noncrossover recombination products, and Spo11-dependent arrest at meiotic prophase, with failure to form spores. At lower temperatures: crossover levels are reduced, noncrossover levels are increased, and exit from meiotic prophase and ensuing spore formation occur efficiently but with a delay (2). Additional ZMM genes were sought by introducing a library of transposon insertion mutations into our standard strain background and screening for insertions that conferred a temperature-dependent defect in spore formation. (Materials and Methods) One such insertion was shown by genomic sequencing to lie in a previously identified (10) meiosis-specific gene, SPO22. A strain carrying a complete deletion null mutation of this gene, spo22Δ, was constructed and found to exhibit the entire array of diagnostic zmm mutant phenotypes (Fig. 1). Spo22 also emerged recently in another mutant screen and was named Zip4 (above).

Saccharomyces cerevisiae Zip2, Zip3, and Spo22/Zip4 Have Homologs in Kluyveromyces lactis, Candida glabrata, and Candida albicans. A psi-blast search for Zip3-related sequences in the nonredundant protein database (accessed at www.ncbi.nlm.nih.gov; cutoff threshold E = 0.005) was carried out by using the sequence from S. cerevisiae as the query (11). This search effectively converges after the second iteration upon retrieving obvious homologs from K. lactis, C. glabrata, and C. albicans (partial sequence) plus other sequences exhibiting less strong homology. An analogous search for Zip2s converges regularly after the second iteration, in this case yielding only homologous sequences from the same three yeasts identified as having related Zip3s. A search for Spo22/Zip4s identified homologs from these same three organisms in the first iteration (E < 3e^–31); however, this search is perpetuated by the inclusion of functionally unrelated tetratricopeptide repeat (TPR)-containing proteins (below).

Zip3s Contain a RING Finger Motif. Pfam database searches are designed to detect known protein sequence motifs by trained hidden Markov models (12). Pfam analysis of Zip3s gives no significant information (accessed through http://pfam.wustl.edu/hmmsearch.shtml; E = 1.0, a default that will report ≈1 false positive hit per query sequence). However, a psi-blast search with amino acid residues 1–91 of S. cerevisiae Zip3, when carried to the third iteration, identified divergent sequences with conserved potential metal ligand positions. When these sequences were analyzed by Pfam, they emerged as statistically significant (e.g., Arabidopsis thaliana CAB46003, E = 5.8e^–06) binuclear Zn RING (really interesting new gene) fingers.

All functionally characterized RING fingers described to date are E3 ubiquitin ligase modules, meaning that they transfer ubiquitin from its activated form on an E2 molecule to a cognate target protein. MIZ (Msx-interacting zinc finger) domains are functionally analogous to ubiquitin E3s except that they can mediate the transfer of SUMO from Ubc9 onto a target molecule. These and other types of evolutionarily related Zn finger motifs (e.g., PHD and FYVE) are distinguishable from one another by specific ligand position signatures (available at the Pfam server noted above). In the RING ubiquitin E3s analyzed thus far, the canonical ligand signature is LXXLX_nLXLXXLXXLX_nLXXL. The Zip3s of S. cerevisiae, K. lactis, and C. glabrata all contain exactly this arrangement as illustrated by comparison with two known RING ubiquitin E3s, Rad18 and Apc11 (Fig. 2A). This is also the case with the likely Zip3 homolog in C. elegans (below). In contrast, we have examined over 100 MIZ-SUMO E3 domains, and none exhibits the ligand pattern observed in ubiquitin E3s and Zip3s, consistent with its annotation as a different type of domain. Most, including experimentally verified MIZ-SUMO E3s from S. cerevisiae (Mms21) and Arabidopsis (Siz1, At5g60410), do not even contain the same number of potential His or Cys ligands (13, 14). These indicators strongly suggest that Zip3s are ubiquitin E3s, although SUMO E3 ligase activity cannot be formally excluded without experimentation.

Fig. 2.

Informatic analysis of Zip3s. (A) The RING fingers of Zip3s compared with those of Rad18s and Apc11s: Sc, S. cerevisiae; Cg, C. glabrata; Kl, K. lactis. The presumed zinc ligands are highlighted. (B) A second domain of homology between Zip3s and Rad18s. Bars are helices and arrows above the sequences are extended strands as determined by Jpred. Arrows below the sequences delimit the region of Rad18 that is necessary and sufficient for heterodimerization with the E2 Rad6 (see text). The boxed segment of ScZip3 (LKSD) was determined by SUMOplot (www.abgent.com/doc/sumoplot) to be a high probability SUMOylation target (91/100). This feature is present only in the S. cerevisiae sequence. (C) Comparison of the domain architectures of ScZip3 and ScRad18. The diagram is approximately to scale.

Zip3s Share N- and C-Terminal Domainal Homology with Rad18, a DNA Binding E3 That Promotes Mitotic DNA Break Repair. Zip3 is a chromosome-associated protein (5, 6). The presence of an N-terminal RING finger occurring in a chromosomal protein is reminiscent of another, more fully characterized S. cerevisiae protein, Rad18. Rad18 is an ubiquitin-ligating E3 that heterodimerizes via its C terminus with the ubiquitin E2 enzyme Rad6 to promote postreplicative DNA repair by monoubiquitinylating DNA pol30 (PCNA) (15, and references cited therein). To explore potential relationships between these two proteins, we performed statistical t-coffee analysis of Zip3 in relation to Rad18 (16). Two regions of high homology emerge: (i) the N-terminal RING finger domain and (ii) a stretch of ≈70 C-terminal amino acids. Secondary structure analysis (Jpred; www.compbio.dundee.ac.uk/∼www-jpred/) further reveals structural homology throughout the C termini of these proteins and permits final manual refinement of the t-coffee alignments (Fig. 2B). Analogous homologies also occur between the Zip3s and Rad18s of K. lactis and C. glabrata (not shown). In Rad18s, the C-terminal Zip3-related region comprises the Rad6 (E2) association interface; moreover, a segment composed of two short strands followed by an amphipathic helix (Fig. 2B) is necessary and sufficient for Rad18/Rad6 heterodimerization (17). In Zip3s, this domain would presumptively perform an analogous function, with Rad6 or another as-yet-to-be-identified E2. Comparative domain architecture diagrams are provided in Fig. 2C.

Development of ABRA and Its Application to Zip2s. Pfam database searches using any of the four retrieved Zip2 sequences as the query fail to identify statistically significant (E = 1.0) domain signatures; automated t-coffee alignment analysis finds only ≈3% identity among the set of putative orthologs, and none of the tertiary structure algorithms linked to the Polish metaserver (http://bioinfo.pl/meta) were able to model full-length Zip2s with any reasonable degree of statistical significance (18). However, RADAR (rapid automatic detection and alignment of repeats) analysis (www.ebi.ac.uk/Radar/) suggested the presence, in all four sequences, of repetitive elements that are between 35 and 50 aa in length (19). We could not identify these segments as corresponding to any known repeat motif by psi-blast, smart, Pfam, or other profile search algorithms. However, visual inspection of the sequences indicated the presence of periodic aromatic amino acid residues. Moreover, automated t-coffee analysis revealed the presence of a single particularly conserved aliphatic-K-hydrophilic-L-X-W-D/N motif near the center of the alignment. These two sequence features are characteristic of WD40-like repeats.

To further explore this suggestion, we developed an approach that is generally applicable to the identification of any type of repeated structural motifs in a set of homologous but significantly divergent proteins: ABRA. ABRA analysis begins with a basic multiple sequence alignment, generated by t-coffee (Fig. 3 Top). This basic alignment is then optimized manually by taking into account three parameters: (i) visual inspection of a set of all pairwise alignments with respect to sequence identities; (ii) amino acid composition/position analysis (e.g., aliphatic hydrophobic in position 1, H-bond acceptor in position 2, etc.); and (iii) secondary structure predictions, producing a refined alignment (Fig. 3, Middle). In this refined alignment, structural elements and repeating segments among various sequences are often immediately visible (e.g., Fig. 3 Bottom). However, if they are not, the RADAR algorithm can be useful by nucleating the repeat array in various positions by identifying repetitive elements in individual sequences that are then inferred in all sequences in the alignment. Structural motifs are then delimited by the fewest number of secondary structure elements (e.g., extended strands, α-helices) required to complete one iteration of the sequence repeat. These motifs are then aligned to one another for further repeat delimitation refinement. Once a repeating structural motif is identified by this approach, its specific nature is inferred by comparison with known structural repeat motifs.

Fig. 3.

ABRA methodology.

With Zip2s, ABRA analysis reveals the presence of 14 reiterated elements. Moreover, the lengths and the particular array of structural and sequence features of these elements are most similar to those of WD40 repeat units (Fig. 4A and Figs. 6 and 7, which are published as supporting information on the PNAS web site). A canonical WD40 unit comprises a set of four antiparallel β-strands with a WD-like dipeptide at its C terminus, although identified units often vary greatly in sequence and the loops that connect consecutive strands can vary greatly in length. The repeat units identified in Zip2s all comprise four β-strands (by averaged consensus) and some, but not all, contain the diagnostic C-terminal dipeptide. These units are aligned to one another in Fig. 4A. Interestingly, whereas the metaserver gives no significant matches with full-length Zip2s, the central region containing the initially observed aliphatic-K-hydrophilic-L-X-W-D/N motif is matched to a murine all-β Ig structure, PDB ID code 1A7Q (18).

Fig. 4.

Application of ABRA to β-repeat proteins. (A) Results of ABRA analysis of Zip2, Cdc20, and Cdh1 from S. cerevisiae. The bottom set of sequences is a Conserved Domain Database-derived structure based alignment of random WD40 repeats, shown for comparison. Zip2 emerges as a 14-blade WD40-like repeat protein. (B) Phylogenetic analysis of the repeats found in Zip2. The tree was constructed by using TreeTop. The numbers represent the percentage of times each branch topology was found during the bootstrap analysis. (C) Structural examples of 7-blade (Ski8, PDB ID code 1SQ9, Upper) and 14-blade (Unc78, PDB ID code 1NR0, Lower) WD40 proteins.

To further confirm the identification of 14 repeat units, we subjected the corresponding alignment to phylogenetic analysis by TreeTop (www.genebee.msu.su/services/phtree_reduced.html) (Fig. 4B). In many cases, consecutive repeats can be related to one another at the terminal branches of a simple phylogenetic tree, consistent with the paradigm that repeat proteins have arisen by successive tandem duplications. These results provide strong evidence that the identified repeat units, which were defined and aligned to one another by sequence and secondary structure elements, are in fact nonrandomly related. Interestingly, each unit has a tendency to be particularly similar to an adjacent unit, with defined pairs comprising units [2, 3], [4, 5], [9, 10], [11, 12], and [13, 14]. This pattern could mean that single unit was duplicated first, followed by successive tandem duplications of the resulting pair of units.

Structural studies have shown that WD40 repeat units are arranged into a β-propeller. The last strand of the each repeat unit interacts with the first three strands of the next repeat unit to form a single blade, and the propeller is closed by a blade formed by analogous interaction between the last strand of the last unit and the first three strands of the first unit (e.g., for Ski8, Fig. 4C) (20). By extension, Zip2 comprises 14 WD40-like repeats and thus could be structurally analogous to Unc78/Aip1, whose 14 WD40 repeats form a linked pair of 7-blade propellers (21).

ABRA Defines APC Subunits Cdc20 and Cdh1 as 7-Blade WD40 Repeat Proteins. Cdc20 and Cdh1 are APC components and have previously been identified as containing WD40 repeats. However, the precise numbers and limits of the repeat units have not been defined, and refined alignments among proteins in the two corresponding families have not been made. ABRA analysis reveals that the WD40 domains of Cdc20 and Cdh1 are each composed of 7 units (Fig. 4A and Fig. 8, which is published as supporting information on the PNAS web site). A publicly available structure-based alignment (www.ncbi.nlm.nih.gov/Structure/cdd/cddsrv.cgi?uid=cd00200) reveals a subset of similar WD40 repeat units identified by ABRA in Zip2, Cdc20, and Cdh1 (Fig. 4A). Both approaches identify repeating structural elements of approximately the same length and sequence but with substantial tolerance for sequence degeneracy and high sequence variability of interunit loops. WD40 domains of Cdc20 and Cdh1 thus emerge as presumptive 7-blade propellers structurally analogous to Ski8 (above).

ABRA Reveals Spo22/Zip4s as TPR Proteins. Despite the obvious relationships among the four most homologous Spo22/Zip4 sequences (above), an initial t-coffee alignment showed only 4.6% identity among them. Such disparities are frequently a hallmark of the presence of a reiterated sequence motif. Correspondingly, RADAR analysis of Spo22/Zip4s indicates the presence of repetitive elements. Further, psi-blast finds statistically significant homology to a number of TPR-containing proteins (E < 0.001) after only the second iteration. Given these clues, we analyzed these sequences by ABRA. ABRA identifies 22 helix–loop–helix repetitive elements in the Spo22/Zip4s of S. cerevisiae and C. glabrata, 21 such elements in C. albicans and 18 such elements in K. lactis (Fig. 5A and Fig. 9, which is published as supporting information on the PNAS web site). The latter two sequences differ from the first two by apparent deletions at their N termini and thus might reflect gene annotation/sequencing errors rather than true sequence differences.

Fig. 5.

Application of ABRA to α-repeat proteins. (A) Results of ABRA analysis of Spo22 from S. cerevisiae. Spo22 emerges as a TPR protein. (B) Structural example of a six-unit TPR polypeptide (Sec17, PDB ID code 1QQE). Rainbow colors roughly delimit successive TPR units.

Three distinct types of helical repeat motifs have been described: HEAT, ARM, and TPR. ARM repeats are three-helix motifs and HEAT and TPR repeats are two-helix motifs. TPRs are distinguishable from HEAT repeats by the absence of a specific defining bend near the middle of the N-terminal helix. Because only one of the repeat elements identified among the analyzed Spo22/Zip4s is predicted to contain such a bend, these elements are most closely related to TPRs (as psi-blast had suggested).

ABRA Defines 15-Unit TPR Arrays in APC Components Cdc16, Cdc23, and Cdc27. For further comparison, we applied ABRA to three of the relevant APC components whose TPRs, although presumptively present, remained unspecified: Cdc16, Cdc23, and Cdc27. Only sequences from the four organisms of the Zip2/3/Spo22/Zip4 cohort were analyzed. Cdc16s, Cdc23s, and Cdc27s all emerge as 15-unit TPR proteins (Figs. 10 and 11, which are published as supporting information on the PNAS web site). These TPR arrays are expected to be similar in general conformation to structurally analyzed arrays, e.g., the six-unit array from the vesicular transport protein Sec17 (Fig. 5B) (22).

Discussion

Identification of an Additional ZMM Protein. Zip2 and Zip3 are formally defined as being part of a single functional unit, the ZMM group (2). Further, the two proteins interact physically and, at later stages of recombination, colocalize in prominent foci that mark the sites of developing crossovers (5, 6). We show here that spo22/zip4 mutants confer the same diagnostic phenotypes as zip2, zip3, and other zmm mutants. These results, as well as cytological studies (G. S. Roeder, personal communication), define Zip4 as an additional member of the ZMM group.

An Evolutionarily Related Cohort of Zip2, Zip3, and Spo22/Zip4 Proteins. Proteins having strong amino acid sequence homology to Zip2, Zip3, or Spo22/Zip4 of budding yeast S. cerevisiae occur only in a small cohort of closely related organisms: K. lactis, C. glabrata, and C. albicans.

It remains possible that divergent homologs of these proteins exist in other organisms. For Zip3, a psi-blast search using the first half of the sequence as a query yielded a number of below-threshold hits, e.g., in Drosophila melanogaster (AAF48955/AAN14131/AAF45708), Drosophila pseudoobscura (EAL27394/EAL32410), Caenorhabditis elegans (CAB00037), Neurospora crassa (EAA26717), and Gibberella zeae (EAA74630). All of these proteins have RING fingers at their N termini and are predicted by psort (www.psort.org/) to be nuclear (23). They are significantly variable in size (≈140 to >1,100 aa), but the differences come primarily from low-complexity elements, also detectable in bona fide Zip3s, which may represent variable-sized linkers between major globular domains. Further, the putative Caenorhabditis elegans gene was previously identified as a Zip3 candidate and phenotypic analysis suggests that this gene is required for completion of recombination, analogous to yeast Zip3 (24). For Zip2 and Spo22/Zip4, reasonable candidates for more-distantly related homologs are not suggested by protein blast search algorithms with the exception of potential Spo22/Zip4 candidates in N. crassa and Cryptococcus neoformans (CAD37031, E = 1e^–04 and EAL19170, E = 3e^–04, respectively). Identification of distantly related Zip2 and Zip4 homologs is particularly difficult due to absence of any diagnostic position-dependent catalytic modules and the presence of structural repeats that share little sequence identity (3–5%) even among close homologs.

Zip3 Has Motifs Found in Ubiquitin RING E3s and Is Homologous to Rad18. Zip3 is identified here as a RING E3 whose most probable substrate is ubiquitin. Further, Zip3 emerges as homologous to Rad18, a known DNA-binding E3, which works in combination with Rad6, a ubiquitin E2. Zip3 might exert its effects by means of Rad6 and/or some other partner. Interestingly, ubiquitin-mediated and SUMO-mediated protein modifications are often intimately functionally linked (e.g., ref. 15). Moreover, we also find that S. cerevisiae Zip3 contains a predicted high probability SUMOylation target consensus (Fig. 2A). Thus, complex functional relationships among Zip3, ubiquitination, and SUMOylation can be anticipated (T.-F. Wang, personal communication).

ABRA: A General Approach to Identification and Assignment of Repeated Motifs in Biological Sequences. With ABRA, a standard alignment is optimized manually by taking into account sequence identities, amino acid properties and position, and predicted secondary structure. Consecutive elements are marked only on the basis of their repetitive nature, without respect to any defined sequence profile. This approach is particularly powerful if used in combination with an algorithm such as RADAR that can nucleate an array by locating cryptic repetitive elements in different locations in individual sequences; additional repeat segments can then be inferred in all sequences at the corresponding positions of the alignment (e.g., as for Zip2s, above). Overall, this method produces an alignment-based structural model, where divergent sequences coalesce to yield a single averaged structure prediction. ABRA therefore differs significantly from the more commonly known structure-based alignment methods in which a known structure is imposed on all available sequences.

Do Zip2, Zip3, and Spo22/Zip4 Comprise a Ubiquitin Ligation Machine? Zip3 appears to be a ubiquitin E3 (above). Additional observations suggest that Zip2 and Zip4 may also be involved in ubiquitin ligation and, if so, in a somewhat unique way.

Two of the major types of ubiquitin ligation complexes are the SCF and the APC. Both contain a cullin scaffolding subunit, Cdc53 and Apc2, respectively. A search of publicly available S. cerevisiae two-hybrid screen databases (www.yeastgenome.org) reveals that Zip2 is a Cdc53-interacting protein. When Cdc53 is presented as “bait” in these screens, there are more hits found with Zip2 (five) than with any other protein, including anticipated hits such as the F-box adaptor protein Skp1 (four); a protein implicated in cullin neddylation [Dcn1 (four)]; the F-box proteins Cdc4 (one), Met-30 (one), and Mdm30 (four); and the ubiquitin E2, Cdc34 (one) (25). Zip2-Cdc53 interactions are also identified in two-hybrid screens when Zip2 is presented as bait. Although not definitive, these findings suggest that Zip2–Cdc53 is a bona fide interaction. By extension these findings further suggest that Zip2 and Zip3 may interact, with Cdc53, in a ubiquitin ligation complex.

No two-hybrid interactions were identified between Cdc53 and Zip3, even though ubiquitin E3 proteins often interact directly with cullin. However, many protein/protein interactions detected by biochemical methods escape detection in two-hybrid screens, including the biochemically well documented association between yeast Cdc53 and the ubiquitin E3 Roc1/Hrt1/Rbx1 (26).

Ligation of ubiquitin to a target molecule can have at least two functional consequences. The most widely studied effect is that polyubiquitin chain formation can target labeled molecules for degradation by the 20S proteosome. A commonly known example of this is the APC-mediated destruction of mitotic cyclins. However, more recently it has become apparent that both mono- and polyubiquitinylation can instead serve as reversible signaling devices, functionally analogous to phosphorylation. In some instances, ubiquitin and SUMO compete for the same lysine residue on a given protein and are functionally antagonistic to one another. This is an increasingly observed mechanism in chromosome-based activities, including transcription and DNA repair. For example, monoubiquitinylation of a certain lysine residue on PCNA (proliferating cell nuclear antigen) by Rad18, followed by Rad5-mediated ubiquitin chain extension, results in promoting polymerase activity to conduct error-free DNA repair. SUMOylation of the same residue inhibits this process and results in the recruitment of the Srs2 helicase that disrupts Rad51 nucleoprotein filaments, presumably to block recombination during S phase (15, 27). If the ZIP complex acts in protein degradation, then the roles of Zip2 and Spo22/Zip4 are likely analogous to those of the WD40 and TPR proteins of the APC (Cdc20/Cdh1 and Cdc16/Cdc23/Cdc27), namely, to serve as shuttle components to bring the substrate(s) molecules into proximity of the Zip3 E3. However, if the ZIP complex acts to enhance or repress the activity of nearby molecules by ubiquitin signaling, the roles of Zip2 and Spo22/Zip4 are less clear, and they may serve simply to help localize Zip3 to recombination complexes.

Zip2/3/4 as a Locally Acting Ensemble. Most straightforwardly, effects of Zip2/3/4 on crossing-over and chiasma formation at the leptotene/zygotene transition would arise by targeted ubiquitination of chromatin, recombinosome components and/or axial proteins locally at the sites of progressing recombinational interactions. Importantly, also, the phenotypes thus far described for zmm mutants all reflect a failure to properly execute the leptotene/zygotene transition. However, the cytologically observable ensembles (immunostaining foci) of Zip2 and Zip3 proteins appear concomitantly with this transition and are then found all along the SC during pachytene. This pattern suggests that immunostaining foci arise as a consequence of the leptotene/zygotene transition (2). Further, and more importantly, the corresponding proteins may have additional functions at later step(s) of chiasma formation, presumptively again acting locally to mediate changes at the chromatin/recombinosome/axis levels.