Identification of a novel family of ubiquitin-conjugating enzymes with distinct amino-terminal extensions.

The ubiquitin/proteasome system is the main eukaryotic nonlysosomal protein degradation system. Substrate selectivity of this pathway is thought to be mediated in part by members of a large family of ubiquitin-conjugating (E2) enzymes, which catalyze the covalent attachment of ubiquitin to proteolytic substrates. E2 enzymes have a conserved 150-residue so-called UBC domain, which harbors the cysteine residue required for enzyme-ubiquitin thioester formation. Some E2 enzymes possess additional carboxyl-terminal extensions that are involved in substrate specificity and intracellular localization of the enzyme. Here we describe a novel family of E2 enzymes from higher eukaryotes (Drosophila, mouse, and man) that have amino-terminal extensions but lack carboxyl-terminal extensions. We have identified four different variants of these enzymes that have virtually identical UBC domains (94% identity) but differ in their amino-terminal extensions. In yeast, these enzymes can partially complement mutants deficient in the UBC4 E2 enzyme. This indicates that members of this novel E2 family may operate in UBC4-related proteolytic pathways.

In eukaryotes, selective protein degradation is largely mediated by the ubiquitin/proteasome system (for reviews, see Refs. [1][2][3][4][5]. Degradation by this system was recently found to be instrumental in a variety of cellular functions such as DNA repair, cell cycle progression, signal transduction, transcription, and antigen presentation. Known substrates of this pathway include transcription factors (MAT␣2, GCN4, c-Jun, p53, NF-B), protein kinases (Mos), cyclins, inhibitors of cyclin-dependent kinases (SIC1, p27), and subunits of trimeric G proteins (for review, see Refs. [1][2][3][4][5]. Moreover, the ubiquitin/proteasome system also eliminates abnormal proteins, e.g. misfolded, mislocalized, or misassembled proteins. Substrate recognition by this pathway involves a specialized recognition and targeting apparatus, the ubiquitin-conjugating system, which operates spatially detached from the proteasome. Proteins recognized by this system are earmarked by the covalent attachment of ubiquitin, a small and highly stable protein. In most cases, ubiquitination involves the formation of multiubiquitin chains attached to the substrate that are subsequently recognized by a specific receptor of the (26 S) proteasome. Proteins bound to the receptor are then probably unfolded and translocated into the central cavity of the proteasome where they are degraded to small polypeptides. Ubiquitin chains are released from substrates and recycled to single ubiquitin moieties (for review, see Ref. 5).
Ubiquitin conjugation involves a reaction cascade (1,3,6,7). Initially, ubiquitin-activating (E1) 1 enzyme hydrolyses ATP and forms a thioester bond between itself and ubiquitin. Ubiquitin is then passed on to ubiquitin-conjugating (E2) enzymes and often subsequently to ubiquitin ligases (E3). Each step involves the formation of a thioester-linked ubiquitin-enzyme (E1, E2, or E3) intermediate (7). E2 and/or E3 enzymes finally catalyze isopeptide formation between the carboxyl terminus of ubiquitin and ⑀-amino groups of internal lysine residues of target proteins. Both E2 and E3 enzymes exist as protein families, and diverse combinations of E2⅐E3 enzyme complexes are thought to define the substrate specificity of the conjugation system.
In the yeast Saccharomyces cerevisiae, 12 different genes for ubiquitin-conjugating enzymes (UBC genes) have been detected to date (3,6). Genetic studies revealed that the encoded enzymes mediate strikingly diverse functions such as DNA repair (8), sporulation (3,8), cell cycle progression (9, 10), peroxisome biogenesis (11), membrane-protein degradation (12), heat shock resistance (13), and cadmium tolerance (15). One of the most prominent E2 enzymes from yeast is UBC4 (13). A principal function of UBC4 and the highly related UBC5 enzyme appears to be the degradation of abnormal proteins as indicated by the sensitivity of ubc4 ubc5 double mutants to heat shock, canavanine (an arginine analog), and cadmium (13,15). In addition, UBC4/5-mediated proteolysis is important for some regulatory processes. One example is the UBC4/UBC5mediated degradation of the yeast transcription factor MAT␣2 involved in mating type control (16). UBC4/UBC5 homologs have been described from several organisms including Drosophila (UbcD1; Ref. 17), Caenorhabditis elegans (ubc-2; Ref. 18) and man (UbcH5; Ref. 19). The function of these enzymes in these organisms is not known, but, in the cases tested, the respective genes can fully complement yeast ubc4 ubc5 mutants. Interestingly, vertebrate UBC4 homologs can mediate p53 (19) and cyclin (20) ubiquitination in vitro, suggesting that UBC4/5-mediated degradation may be of central regulatory importance.
Here we report the identification of a novel family of ubiq-uitin-conjugating enzymes from higher eukaryotes that are related in function and sequence to yeast UBC4/5. In contrast to UBC4 (and its homologs from higher eukaryotes) these enzymes have amino-terminal extensions. The UBC domain of these enzymes from Drosophila, mouse, and man is exceptionally highly conserved, exhibiting 94% sequence identity over a stretch of 149 amino acid residues. In contrast, the aminoterminal extensions of four different members of this family exhibit little sequence similarity and may define substrate specificity or are involved in the regulation of these enzymes. In each species, probably all four (or possibly more) different members of this UBC subfamily are expressed, suggesting that each one of these enzymes fulfils special tasks.

EXPERIMENTAL PROCEDURES
Bacterial and Yeast Strains, Media-Growth and handling of Escherichia coli were by standard techniques (21). Strain XL1 was used as a host for plasmids. Transformation was carried out using the electroporation procedure. phage was propagated on strain C600hfl. S. cerevisiae cultures were grown in rich (YP) or synthetic (S) media containing either 2% glucose (YPD or SD media), 2% raffinose (YPRaf or SRaf media), or 2% galactose (YPGal or SGal media) as carbon sources. Yeast transformation was carried out using standard protocols (21). The ubc4 ubc5 yeast strain is a haploid MATa derivative of YWO23 (14). Growth at 30 and 37°C was assayed by streaking transformed ubc4 ubc5 mutant cells pregrown on SGal plates at 23°C onto YPGal plates and incubating at either 30 or 37°C for 5 or 7 days, respectively.
Isolation and Sequencing of UBC cDNAs-Plasmid isolation, cloning, and screening methods were according to standard protocols (21). New and the region adjacent to the active cysteine, conserved in many yeast UBC proteins, respectively. PCR reactions were carried out as described previously (17). To clone Drosophila melanogaster UBC genes, primers A and C and genomic D. melanogaster Oregon P2 DNA as template were used. Fragments of mouse UBC genes were amplified with primers B and C and genomic mouse DNA as template. Aliquots of the genomic DNA were digested with EcoRI, SalI, or XbaI, pooled, and 2 g per reaction was used. Reaction products were separated on 6% polyacrylamide gels and DNA from bands of the correct size were eluted. The PCRgenerated fragments were digested with EcoRI and XbaI and subcloned into M13mp18/19 vectors (21). Drosophila cDNA were isolated from a pNB40-based D. melanogaster cDNA library, carrying cDNA inserts from oligo(dT)-primed Drosophila RNA from 0 -24-h-old embryos (22), using radiolabeled PCR-generated fragment as a probes. Positive clones were purified, and the 1.25-kb EcoRI-HindIII-digested cDNA insert was subcloned into the M13mp18/19 vectors. UbcM2 cDNA was isolated by screening a gt10-based cDNA library, carrying cDNA inserts from oligo(dT)-primed mouse RNA from embryonic mouse fibroblasts (kindly provided by Dr. P. Ekblom, Uppsala, Sweden), with the PCR-generated fragments as a radiolabeled probe. Positive clones were purified and EcoRI digested. Due to an internal EcoRI restriction site within the UbcM2 cDNA, we obtained two fragments (700 and 900 bp), which were subsequently ligated to obtain the full-length clone in Bluescript plasmid (Stratagene, San Diego, USA). Cloning of UbcM3 cDNA was done by screening a gt10-based cDNA library, carrying cDNA inserts from random-primed mouse RNA from adult mouse Balb/c brain tissue (Clontech, Palo Alto, USA), with the PCR-generated fragments as radiolabeled probes. Positive clones were purified, and the 1.32-kb EcoRIdigested cDNA insert was subcloned. DNA sequences of the three cDNAs were determined after subcloning inserts of appropriate lengths with the Applied Biosystems model 373 DNA sequencer. For each cDNA, both strands were sequenced completely with sufficient overlaps. DNA and deduced amino acid sequences were compared with data bases (GenBank, EMBL, PIR, SWISS-PROT) using the BLAST algorithm (23). Alignments were carried out by the BoxAlign program (GCG package).
Northern Blot Analysis-Isolation of RNA from D. melanogaster embryos and mouse liver tissue was prepared using standard protocols (24,21). Resolution of RNA in a 1.0% agarose gel following glyoxal modification was carried out as described previously (21). Separated RNA was transferred to GeneScreen (DuPont NEN) and hybridized with corresponding probes. The probe for UbcD2 RNA was generated by subcloning the 315-bp PstI-SspI fragment of the UbcD2 cDNA (nucleotides 319 -634 of Fig. 1) into Bluescript plasmid. After recovering the fragment by PstI-KpnI digestion, eluted DNA was used as a template for the random-primed labeling kit (Boehringer Mannheim) to generate a radiolabeled hybridization probe. Likewise, the probes for UbcM2 and UbcM3 were obtained by subcloning and labeling the 335-bp EcoRI-BamHI fragment of the UbcM2 cDNA (nucleotides 194 -528 of Fig. 1) and the 240-bp HindIII-SspI-digested fragment of the UbcM3 cDNA (nucleotides 127-364 of Fig. 1), respectively. Northern blot hybridization and detection of the hybridized probe were carried out by standard procedures (21).
Western Blot Analysis-Isolation of yeast proteins, generation of polyclonal antibodies and Western blot analysis were performed essentially as described previously (14,21). Antibodies were raised against the E. coli-expressed UBC domain of UbcD2. Due to the unique conservation of this domain, the antibody reacts well with UbcM3 (see Fig.  4B).
Construction of Plasmids-For heterologous expression of the identified cDNAs in the yeast ubc4 ubc5 mutant, the cDNAs were cloned into yeast expression vectors. The 1.1-kb AsuII-EcoRI fragment of UbcD2 was cloned into a derivative of pSEY8, a 2 micron based plasmid carrying the URA3 marker and the UBC1 promoter (provided by J. Jungmann, Heidelberg, Germany). Yeast UBC4 was under the control of its own promoter on pSEY8 (14). The 880-bp SmaI-StuI fragment of UbcM2 was subcloned into Bluescript plasmid and subsequently cloned into pSGal3T, a YEplac112 (Ref. 25) based plasmid carrying the inducible GAL1 promoter and the transcription terminator of the ADH gene (provided by S. Smith, Heidelberg, Germany). UbcM3 and truncated UbcM3 were cloned by PCR using UbcM3 cDNA as template. Amplification of full-length UbcM3 was done with primer 1 (GCTCTAGATA-ATGTCGGATGACGATTCG) and primer 2 (CCCGTCGACCCAGTT-TATGTAGCTG). The 600-bp PCR product was digested with XbaI and SalI, cloned into Bluescript plasmid for sequence determination, and subsequently cloned into vector pSGal3T (see above). Amplification of truncated UbcM3 was achieved with primer 3 (GCTCTAGATAAT-GTCTAGCGCTAAGAGGATCC) and primer 2. Primer 3 generated a new ATG start codon together with a TCT codon encoding serine (a highly conserved second amino acid of yeast UBCs) fused to the codons corresponding to the amino acid sequence SAKRI, corresponding to the residues 4 -8 of the conserved UBC domain. The resulting 460-bp PCR product was digested with XbaI and SalI and cloned into Bluescript plasmid for sequence determination before transfer into vector pSGal3T (see above).

Cloning of a Novel UBC Gene Family From Higher
Eukaryotes-To study the function of the ubiquitin system in higher eukaryotes, we initiated a homology-based screen for UBC genes from organisms, which are amenable to genetic analysis such as Drosophila, C. elegans, and mouse (17,18). Ubiquitin-conjugating enzymes are highly related proteins showing at least 30% amino acid sequence identity between different members of this enzyme family. In particular, sequences within the UBC domain, which harbors the cysteine residue required for ubiquitin-thioester formation, are highly similar. Primer pairs specific for conserved sequences were designed (see "Experimental Procedures") and used for the PCR. From Drosophila and mouse genomic DNA as templates, we were able to amplify fragments of sizes equivalent to the corresponding regions within yeast UBC genes. DNA sequence analysis indicated that the amplified fragments correspond to segments of five different genes encoding novel ubiquitin-conjugating enzymes. Interestingly, the deduced amino acid sequence of one of the PCR fragments isolated from Drosophila and two fragments isolated from mice were virtually identical, suggesting that we have identified three members of a family of highly related UBC genes (see below). Using the cloned five PCR fragments as probes, we isolated the complete cDNAs for the corresponding genes from a plasmid-born Drosophila library and a mouse cDNA library in phage, respectively. We named the three Drosophila genes UbcD1, UbcD2, and UbcD3 and the two mouse genes UbcM2 and UbcM3 (ubiquitin-conjugating enzymes from Drosophila or mice; the numbering reflects the order of identification). As reported previously (17), the UbcD1 enzyme is the structural and functional homolog of yeast UBC4. UbcD3 identified by our screen is identical to bendless, a gene implicated in Drosophila nervous system development characterized while this work was in progress (26,27). We mapped UbcD1, UbcD2, and UbcD3 on the Drosophila genome to positions 88D, 32A/B, and 12D, respectively (data not shown).
A Family of Highly Related Ubiquitin-conjugating Enzymes with Distinct Amino-terminal Extensions-The DNA sequences and the deduced amino acid sequences of the three new genes, UbcD2, UbcM2, and UbcM3 are shown in Fig. 1. The initiator methionines of the three genes were assigned to the first ATG of the cDNAs preceded by stop codons. The open reading frames of UbcD2, UbcM2, and UbcM3 predicted proteins of 232 (24.5 kDa), 207 (22.9 kDa), and 193 (21.3 kDa) residues, respectively. By Northern analysis using total RNA isolated either from Drosophila or mouse, we identified with the respective probes transcripts for UbcD2, UbcM2, and UbcM3 of ϳ1.5, ϳ1.7, and ϳ1.5 kb, respectively (Fig. 2). The sizes fit well to the corresponding cDNAs and indicate that the isolated cDNAs contain the complete open reading frames.
A comparison of the deduced amino acid sequences of UbcD2, UbcM2, and UbcM3 shows that they are highly related (Fig. 3). Unlike previously identified E2 enzymes (3,6), these new enzymes possess amino-terminal extensions in addition to the UBC domain. The UBC domains of the three enzymes are almost identical in sequence (94% identity over 149 amino acid residues; between 72 and 79% at the DNA level). UbcD2 differs from the UbcD2/UbcM2/UbcM3 consensus by 6, UbcM2 by 2, and UbcM3 by 3 residues (Fig. 3). In contrast to the extreme conservation of the UBC domains, the amino-terminal extensions of the three enzymes (designated extensions A, B, and C; Fig. 3) show little sequence similarity among each other and differ in size (Fig. 3). The weak sequence similarity of the extensions is largely restricted to clusters of serine/threonine and basic residues. No significant sequence similarities between the extensions and known sequences in the data bases were found except for short consensus sequences for phosphorylation sites. Further data base searches detected (in addition to a human UbcM2 homolog designated UbcH9) a partial open reading frame from a human cDNA fragment (designated UbcH8), and this represents a probable fourth member of this enzyme family. This partial sequence exhibits a 100% match to the corresponding sequences of the UBC domains of UbcD2, UbcM2, and UbcM3 and an amino-terminal extension (designated extension D) nonidentical, but related to, extension B of  UbcM2 and UbcH9 (Fig. 3). This suggests that the novel E2 enzyme family described here has at least four distinct members.
UbcD2, UbcM2, and UbcM3 Are Structurally and Functionally Related to Yeast UBC4 -Computer aided sequence comparisons of the UBC domains of UbcD2, UbcM2, and UbcM3 with the UBC domains of other known ubiquitin-conjugating enzymes revealed that their closest homologs are yeast UBC4/ UBC5 (an enzyme pair expressed from duplicated genes; Refs. 13 and 28) and their respective homologs from higher eukaryotes, including the Drosophila UbcD1 (17), C. elegans ubc-2 (18), and human UbcH5 (19) gene products. The UBC domains of UbcD2, UbcM2 and UbcM3 share 64% sequence identity with these UBC4-like enzymes, suggesting that the FIG. 3. Sequence similarity between members of the novel UBC-family. Primary sequences of UbcD2, UbcM2, UbcM3, and three additional members of this family designated UbcH8, UbcH9 (translated from partial cDNA sequences, accession numbers Z44894 and H12272) and UbcH6 2 are compared with yeast UBC4 (14). Upper panel, schematic diagram of the primary sequences. The highly conserved UBC domain of the new UBC family is shown as a white box and corresponds to the entire sequence of UBC4 (light gray). The four different amino-terminal extensions are shown as boxes in distinct gray shades and are designated A, B, C, and D extensions. Numbers above the boxes correspond to amino acid residues. Lower left panel, sequence comparison of the amino-terminal extensions of UbcD1, UbcM2, UbcM3, and UbcH8. Residues that are identical in at least two proteins are boxed. Gaps (indicated by dashes) were permitted to optimize alignments. Amino-terminal extensions of the homologs UbcM2 and UbcH9 (extension B) differ by only one residue (K31E) and of the homologs UbcM3 and UbcH6 (extension C) by two residues (G26T/S27N). Lower right panel, sequence comparison of the UBC domains of different family members with UBC4. Unavailable sequence data of the partial UbcH8 clone is indicated by a dotted line. Residues that are identical in at least three proteins are boxed. The sequences were aligned using the BoxAlign program (GCG package). Residue numbers are given on the left. newly identified enzymes may be functionally related to UBC4 (Fig. 3). Yeast UBC4/UBC5 ubiquitin-conjugating enzymes are involved in stress-related functions and in the turnover of regulatory proteins (Ref. 13; see the Introduction). Single mutations in these genes are viable and lead to only moderate mutant phenotypes, but ubc4 ubc5 double mutants are slowly growing and inviable at elevated temperatures (13). As reported previously, UbcD1, the UBC4 homolog from Drosophila (80% identical to yeast UBC4), can rescue the deficiencies of yeast ubc4 ubc5 mutants (17). Complementation of UBC4/ UBC5 functions by UbcD1 was nearly complete, even when the enzyme was expressed from a single copy in the genome (17).
To study the activity of UbcD2, UbcM2 and UbcM3 in yeast, we cloned the respective reading frames into yeast high copy number, 2 micron based expression vectors. When yeast cells were transformed with these plasmids, all three genes could complement the growth deficiency and heat sensitivity of ubc4 ubc5 double mutants (Fig. 4A). Although complementation was only partial as indicated by slight growth defects at 30-and, in particular, at 37°C, UbcD2, UbcM2, and UbcM3 are likely to function in similar proteolysis pathways as UBC4 (or UbcD1). The incomplete complementation of ubc4 ubc5 by these genes may indicate that UbcD2, UbcM2, and UbcM3 only interact with a subset of UBC4's substrates or that they may fail to collaborate with certain components of a UBC4-dependent degradation pathway (e.g. ubiquitin ligases, E3), or both.
In addition to a comparably weak sequence similarity to UBC4/UBC5 (identity of 64 versus 80% for UbcD1, Fig. 5) the three enzymes differ from UBC4 and its homologs by the presence of amino-terminal extensions. These extensions could possibly function as regulatory (either activating or repressing) or interacting domains with specific components of the ubiquitinconjugation system. To test these possibilities, we constructed a derivative of UbcM3 lacking the amino-terminal extension (UbcM3⌬1-47). When yeast cells were transformed with UbcM3 and truncated UbcM3⌬1-47, both gene products were expressed with the expected sizes to similar levels (Fig. 4B). Remarkably, full-length and truncated UbcM3 enzymes complemented the yeast ubc4 ubc5 mutant to a similar extend (Fig.  4A). Thus the amino-terminal extension of UbcM3 (and probably also of UbcD2 and UbcM2) does not modulate the activity of the enzyme to function in UBC4-dependent pathways in yeast.

DISCUSSION
Previously identified ubiquitin-conjugating enzymes (3,6) are small proteins (ϳ14 -32 kDa), which either consist of the UBC domain only (class I E2 enzymes) or they possess additional carboxyl-terminal extensions (class II enzymes). Here we describe a novel family of ubiquitin-conjugating enzymes from higher eukaryotes that have amino-terminal but lack carboxylterminal extensions (designated class III enzymes). We have cloned three members of this class and identified a fourth in the data base. The UBC domain of these novel enzymes is virtually identical, but the amino-terminal extensions show limited sequence similarity. The recent identification of human homologs to UbcM2 (UbcH9; Fig. 3) and UbcM3 (UbcH6; Fig. 3), 2 which are homologous to their respective murine counterparts over their entire lengths (including the extensions), and the extreme conservation of the UBC domains of different members of this family strongly suggest that a homolog of each of these novel four UBCs may be present in each of these species, i.e. Drosophila, mouse, and man. We have unsuccessfully tried to identify yeast homologs to these enzymes using different PCR strategies. Thus we assume this family probably evolved relatively late in evolution and may be unique to multicellular organisms. Intriguingly, these enzymes are among to the most highly conserved proteins of these organisms. The UBC domains of UbcD2 from Drosophila and UbcM2, and UbcM3 from mice share 94% identical amino acid residues (homologs of other UBCs are typically 70 -80% identical in sequence; Refs. 17-19, 29 -32; see Fig. 5). The extreme conservation of UbcD2, UbcM2, and UbcM3 is even more remarkable given the likely possibility that the true homologs, i.e. the enzymes with similar extensions are yet to be identified. Proteins of similar high conservation, e.g. histones or ubiquitin, either have multiple interacting partners or most of their amino acid residues participate in intramolecular contacts. Both types of interactions are thought to prevent evolutionary amino acid sequence drift. We thus assume that the novel UBC enzymes interact with several proteins. Candidates for binding partners are components of the ubiquitin/proteasome system or substrates. Since overexpressed UbcD2, UbcM2, and UbcM3 can partially sup-2 M. Scheffner, personal communication. FIG. 4. Complementation of the yeast ubc4 ubc5 mutant by expression of UbcD2, UbcM2, UbcM3 and an amino-terminal truncated UbcM3. A, growth of yeast ubc4 ubc5 double mutant on YPGal plates at normal growth temperature (30°C) and heat shock temperature (37°C) expressing the following UBCs (in clockwise orientation). Vector as negative control, amino-terminal truncated UbcM3 (pUbcM3⌬1-47), UbcM3 (pUbcM3), UbcM2 (pUbcM2), UbcD2 (pUbcD2), and yeast UBC4 (pUBC4) as positive control. B, Western blot analysis of total yeast proteins from ubc4 ubc5 cells expressing UbcM3 and UbcM3⌬1-47, respectively, with antibodies generated against the conserved UBC domain of UbcD2/UbcM2/ UbcM3 family. Additional protein bands cross-reacting with the antiserum serve as loading control. Size references are given on the right. press UBC4/UBC5 deficiency in yeast, these enzymes and UBC4/UBC5 probably have many substrates in common. However, the presence of multiple, highly conserved extensions of these enzymes suggests that they are likely to carry out specialized functions distinct from those of UBC4. What these functions are is not known at present, but the gene expression pattern of the Drosophila UBCs may provide some clues. Interestingly, UbcD1, the UBC4 homolog, is continuously expressed throughout development consistent with a "housekeeping" function of the encoded enzyme. 3 Transcripts of UbcD3(bendless), another UBC4-related gene (which is actually unable to rescue ubc4 ubc5 mutants; Fig. 5) 3 can also be detected at all developmental stages of Drosophila development. In contrast, UbcD2 appears to be exclusively expressed at postlarval (L3) stages, but in eggs the transcript is supplied maternally. 3 Thus the functions of these class III enzymes may be predominantly restricted to distinct tissues in pupae or adult flies.
The significance of the amino-terminal extensions is currently unclear, but their conservation between species (e.g. the extensions B and C; Fig. 3) indicate that they are probably relevant to their cellular functions. The carboxyl-terminal extensions of class II E2 enzymes are known either to contribute to their substrate specificity (UBC2, UBC3; Refs. 9 and 33-35) or they mediate intracellular localization (UBC6; Ref. 12). The prevalence of putative phosphorylation sites within the extensions of the UbcD2, UbcM2, and UbcM3 enzymes may indicate that the enzymatic activity or a possible interaction with other proteins is possibly controlled by enzyme phosphorylation. Alternatively, these sequences rich in serine, threonine, and basic residues may represent binding sites for specific components of the ubiquitin-conjugating system or proteolytic substrates. Class I ubiquitin-conjugating enzymes have highly conserved three dimensional structures with exposed amino termini (36). This suggests that the highly charged amino-terminal extensions of the class III enzymes described here may fold into separate domains, which are probably readily accessible to interacting partners.