Functional Characterization of ABC10α, an Essential Polypeptide Shared by All Three Forms of Eukaryotic DNA-dependent RNA Polymerases*

ABC10α is a small polypeptide shared by the three yeast RNA polymerases. Homologous polypeptides in higher eukaryotes have a zinc-binding CX 2C … CX 2C motif and a conserved basic C-terminal end. These features are also found in archaeal gene products that may encode an RNA polymerase subunit. The CX 2C … CX 2C motif is partly dispensable, since only its first cysteine is essential for growth, whereas the basic C-terminal end is critical in vivo. A mutant in the latter domain has an RNA polymerase III-specific defect and, in vitro, impairs RNA polymerase III assembly. Polymerase activity was, however, not affected in various faithful transcription assays. The mutant is suppressed by a high gene dosage of the second largest subunit of RNA polymerase III, whereas the homologous subunits of RNA polymerase I and II have aggravating effects. In a two-hybrid assay, ABC10α binds to the C-terminal half of the second largest subunit of RNA polymerase I, in a way that requires the integrity of the CX 2C … CX 2C motif. Thus, ABC10α appears to interact directly with the second largest subunit during polymerase assembly. This interaction is presumably a major rate-limiting step in assembly, since diploid cells containing only one functional gene copy for ABC10α have a partial growth defect.

The existence of three distinct nuclear RNA polymerases is documented in all eukaryotes investigated so far and thus probably emerged at a very early step of eukaryotic evolution (1). Despite this ancient origin, which would have left ample time for duplications to generate homologous but distinct subunits in each polymerase, six identical polypeptides are shared by the three eukaryotic transcription systems. The best characterized is undoubtedly the TATA box-binding protein, which has been known for some time to operate in all three transcription complexes (2)(3)(4)(5). However, earlier studies on Saccharomyces cerevisiae (6,7) had also indicated that the three yeast RNA polymerases share three polypeptides (ABC27, ABC23, and ABC14.5), and two other small common subunits (ABC10␣ and ABC10␤) were identified more recently (8). All of them are essential for growth (9 -11) and, except for ABC27, are interchangeable from yeast to man in vivo (12,13).
Although the existence of common subunits has been known for more than 20 years in yeast, their role in transcription is still an enigma. The lack of a eubacterial counterpart argues against a direct catalytic role (14). However, a subform of yeast pol 1 I lacking ABC23 is catalytically inactive (15). Three of these polypeptides (ABC10␤, ABC23, and ABC27) have distinct homology to proven subunits of the archaeal enzyme (14) and are related to gene products of the African swine fever virus (16), a cytoplasmic DNA virus endowed with its own DNA-dependent RNA polymerase (17). They are therefore present in all non-eubacterial RNA polymerases. The other two common subunits, ABC14.5 and ABC10␣, were first thought to be typically eukaryotic (14), but this may not be true for ABC10␣, which is remotely related to an archaeal gene product encoded by the genome of Pyrococcus horikoshii (18). Since its discovery by Carles et al. (8), ABC10␣ was shown to be an essential polypeptide (19) but has not been further investigated, which prompted us to the present genetic and biochemical characterization of that polypeptide in S. cerevisiae.
Mutagenesis-Random mutagenesis of RPC10 was done by PCR amplification of pGENs-RPC10, using the Taq polymerase in the presence of 20% Me 2 SO and at a final concentration of 0.02 mM dATP. The resulting fragments were subcloned between the BamHI-SalI sites of pGENs-RPC10. 6000 independent plasmids were obtained, pooled, and transformed into the ade2 ade3 strain YLR-02. This strain harbors the 2-m URA3 ADE3 RPC10 plasmid pTSV31-RPC10 and thus forms red colonies in the presence of a limiting concentration of adenine, due to the complementation of ade3 by the wild type ADE3ϩ gene. These colonies eventually form white sectors due to spontaneous plasmid loss. About 8000 transformants were obtained by plating on tryptophan-less medium containing a low amount of adenine (4 g/ml). After 1 night at 30°C and 5 days at 37°C, 17 temperature-sensitive mutants unable to lose pTSV31-RPC10 at 37°C were isolated as non-sectored red colonies. The pGENs-RPC10 plasmid was extracted in each case and reintroduced in YLR-02 to demonstrate that it bears the temperature-sensitive mutant phenotype. RPC10 mutants generated by PCR-dependent sitedirected mutagenesis in pGENs-RPC10 were scored for their ability to complement the rpc10-⌬::HIS3 null allele of strain YLR-02, using a plasmid shuffle assay based on 5-fluoroorotic acid resistance (12). Nterminal HA-tagged forms of ABC10␣ were constructed in pGENs plasmids bearing the wild type and two mutants forms (rpc10-11 and rpc10-30) of RPC10, using PCR primers encoding a single copy of the influenza HA epitope. The viability of these tagged forms was tested by the same plasmid shuffling assay.
Dosage-dependent Suppression-Multicopy suppressors were isolated after transformation of the temperature-sensitive rpc10-30 mutant YLR-06 by a wild type genomic DNA library of S. cerevisiae (strain FL100) born on the multicopy vector pFL44L (27). After 1 night at 30°C, plates were incubated at 37°C and observed for 8 days. Ten fast growing transformants were able to lose the plasmid-borne rpc10-30 allele in a plasmid shuffle assay and thus contained the wild type RPC10 gene, as confirmed by plasmid extraction and restriction anal- ysis. The remaining 11 transformants only partially restored growth at 37°C. Plasmid extraction and retransformation of the original mutant strain confirmed that suppression was due to the plasmid itself. Seven clones harbored the functional copy of the non-essential gene SSD1. In contrast to strain FL100, the YPH499 laboratory strain used in this study has a defective SSD1 allele, which is known to aggravate the temperature-sensitive defect of many RNA polymerase mutants (27). In this case, therefore, suppression reflects a genetic polymorphism between laboratory strains rather than a genuine dosage-dependent effect.
Purification of RNA Polymerase III and in Vitro Transcription-Whole-cell extracts from YLR-HA01 and YLR-HA06 strains were prepared from 6-liter cultures grown at 30°C on YPD to an A 600 of 1. Protein concentration, estimated by the Bradford method, was around 15 mg/ml. Microscale purification of pol III and purification of class III transcription factors were as described previously (28). RNA polymerase activity during purification was monitored by nonspecific transcription assays on a poly(dA-dT) template (29).
Whole Purified pol III were tested for faithful multiple round and single round transcription on the pRS316-SUP4 plasmid as described previously (29).
Western Blotting-Immunoblot analysis was performed by loading normalized amount of protein from each extract on SDS-polyacrylamide gels (6 -15% polyacrylamide). Proteins were electroblotted onto nitrocellulose membranes (Hybond C-Super, Amersham Pharmacia Biotech). Membranes were probed with monoclonal antibodies raised against purified pol III (30). HA-ABC10␣ was revealed using the monoclonal antibody 12CA5, recognizing the HA epitope, with an ECL detection system (Amersham Pharmacia Biotech).

RESULTS
Halving the RPC10 Gene Dosage Has a Rate-limiting Effect on Growth-From previous work on yeast RNA pol I and pol III subunits (Refs. 31 and 32 and references therein), we know that deleting one gene copy in the corresponding diploid strains has no detectable effect on growth and that these gene deletions are therefore fully recessive. However, RPC10 does not follow the common rule, since the rpc10-⌬::HIS3 null allele is incompletely recessive in a heterozygous diploid strain grown on rich medium at 30°C (Fig. 1A). Reintroducing RPC10 by transformation with pFL44-RPC10 restores the wild type level of growth. In liquid YPD medium, the doubling time of the heterozygous diploid was 120 min, instead of 90 min for a wild type control. Thus, halving the RPC10 gene dosage detectably impairs growth, indicating that ABC10␣ is produced in ratelimiting amounts. Such a partial dominance might easily be overlooked, and we therefore re-examined the recessivity of null alleles corresponding to the other common subunits shared between all three polymerases (ABC10␤, ABC14.5, ABC23, and ABC27) or between pol I and pol III (AC19 and AC40). We confirmed that all of them are fully recessive. Thus, the dosagedependent effect of RPC10 is a unique property of the ABC10␣ subunit.
Overexpression of ABC10␣ Corrects a Mutant Defective in the Assembly of Pol III Preinitiation Complex-Since halving the gene dosage of RPC10 suffices to reduce growth, one might expect an increased gene dosage to have instead a stimulatory effect when growth is limited by RNA polymerase availability. This was actually suggested by previous work on tfc3-G349E, a thermosensitive mutant altered in the pol III preinitiation factor TFIIIC and therefore affecting the recruitment of pol III on its cognate template genes. In a search for dosage-depend-ent suppressors, Lefebvre et al. (21) found that this defect is partly corrected in the presence of pFL44L-RPC10, and speculated that the overproduction of ABC10␣ increases the local concentration of RNA pol III and therefore partly overcomes the recruitment defect of tfc3-G349E. The plasmid used in the initial experiment (pFL44L-RPC10) turned out to also harbor DCD1 (encoding deoxycytidylate synthase), and we therefore repeated this experiment with pGENs-RPC10, a multicopy plasmid that only bears RPC10. A somewhat stronger suppression was observed in this case, because RPC10 was expressed from the strong PGK1 promoter (Fig. 1B). Control experiments with the four other common subunits showed that they had no suppression effect. pGENs-RPC10 did not alter the temperature-sensitive phenotype of rpc31-236 and rpc160-112 pol III mutants, which are defective in transcriptional initiation or elongation, respectively (29,33). Thus, an increased gene dosage of RPC10 improves the assembly of the pol III preinitiation complex but appears to have no effect on the catalytic activity of pol III itself.
Variable and Conserved Domains of ABC10␣-The human and fission yeast homologs of ABC10␣ can replace the S. cerevisiae polypeptide in vivo (12,34). Plants also contain an ABC10␣-like polypeptide, as shown by the cloning of a corresponding cDNA in Zea mays (GenBank TM accession number AI 372144), and a hypothetical ABC10␣ polypeptide can also be recognized in the Caenorhabditis elegans genome (cosmid CEF23B2). Thus, this subunit is represented in the main phyla of the eukaryotic crown. It was recently noted that P. horikoshii encodes a polypeptide related to ABC10␣ (18). A computer search shows that this also applies to the Archaeaglobus fulgidus genome (35). Hence, there is a distinct possibility that one FIG. 1. RPC10 has rate-limiting effect on growth. A, growth pattern of heterozygous rpc10-⌬::HIS3/ϩ diploid strain. The diploid strains YLR-09 (RPC10/RPC10) and YLR-10 (RPC10/rpc10⌬) were streaked on YPD and grown for 3 days at 30°C. B, extragenic suppression of tfc3-G349E mutation by overexpression of ABC10␣. A tfc3-G349E mutant strain yOL8 (21) was transformed with various URA3ϩ multicopy plasmids as indicated. Cells from the corresponding transformants were serially diluted, spotted on YPD plates, and incubated for 4 days at 30 and 37°C. pFL44L is the vector without insert, pLR57, pFL44L-RPB8, pFL44L-RPB6, pFL44L-RPB5, and pFL44L-RPB10 contain inserts, respectively, encoding the 138 subunit of TFIIIC (encoded by TFC3) and the common subunits ABC27, ABC23, ABC14.5, and ABC10␤. of the poorly characterized subunits of archaeal RNA polymerase (14) is equivalent to ABC10␣. These archaeal and eukaryotic polypeptides all share an invariant CX 2 CX 10 -15 CX 2 CG motif and a basic C-terminal domain that is strongly conserved in all eukaryotic forms of ABC10␣. In contrast, the N-terminal region is quite variable in length and amino acid composition. Fig. 2 summarizes the sequence conservation of the eukaryotic and archaeal forms of ABC10␣.
Mutational Analysis of RPC10 -To examine the physiological importance of the conserved and variable features of ABC10␣, we tested the growth pattern of 30 mutants generated by site-directed mutagenesis, corresponding to partial deletions of the N-terminal or C-terminal ends or to single-site substitutions of conserved residues. RPC10 was also subjected to random mutagenesis, and the resulting mutant library was screened for conditional, temperature-sensitive alleles in a plasmid-shuffle assay (see "Materials and Methods"). In view of the dosage effect shown by RPC10 (see above), these mutants were constructed on the high copy plasmid pGENs-RPC10 to maximize the expression of the mutant subunit. As expected, transferring these mutants on a centromeric plasmid enhanced their adverse phenotype.
As anticipated from its lack of conservation, the N-terminal end is not essential for growth (Fig. 2). Deletions removing up to the first 10 amino acids (rpc10-1 and rpc10-2) were indistinguishable from the wild type parent at 30°C, with a slight growth defect at 37°C. More surprisingly, the invariant zincbinding motif was also partly dispensable. Its four cysteines contribute quite differently to the biological activity of ABC10␣, and only one of them (Cys-31) is essential for growth. This first became evident when we noticed that 10 of the 17 temperature-sensitive mutants obtained by random mutagenesis correspond to substitutions affecting three of the invariant cysteines (Cys-34, Cys-48, and Cys-51) and was confirmed by the phenotype of conservative Cys-Ser substitutions at each of the invariant cysteines, alone or in combination. Cys-31 is strictly indispensable (rpc10-3 is lethal at all temperatures tested, even at high gene dosage), whereas all other single or multiple substitutions, including a triple Cys-Ser substitution at Cys-34, Cys-48, and Cys-51, retain viability. Moreover, the double Cys-Ser substitution at positions 48 and 51 has a wild type growth phenotype, thus revealing a remarkable functional disymmetry between the two halves of the zinc-binding domain.
In contrast to the N-terminal region or to the zinc-binding motif, the conserved C-terminal end is crucial for the biological activity of ABC10␣, and removing more than its very last amino acid or altering the charge of this domain by amino acid replacements (K58L, K58Q, K58E, and R60E) is either lethal or leads to severe growth defects. Substitutions that change the size of the amino acid but not its charge (K58R and R60K) have instead no effect on growth.
rpc10-30, a Mutant in the C-terminal Basic Domain, Specifically Affects Pol III Transcription in Vivo-Since the zincbinding motif and the basic C-terminal end of ABC10␣ are evolutionarily conserved, we investigated in more detail two mutants that respectively affect these two domains. In both cases, a double substitution was required to achieve temperature sensitivity. rpc10-30 combines two replacements in the basic C-terminal end (R60Y and V65D) that are by themselves phenotypically silent. rpc10-11 is a double mutant (I8N and C48R) where the partial temperature-sensitive defect of C48R replacement is enhanced by the phenotypically silent substitution I8N. Both mutants are lethal when harbored on the centromeric plasmid pCM185-RPC10. At 30°C, they grow with a doubling time of 180 min instead of 120 min for the isogenic wild type strain. When shifted to 37°C, they only begin to affect growth after 8 h and then they become completely arrested. Accordingly, we examined their influence on in vivo RNA pol I-, pol II-, and pol III-dependent transcription within the first 6 h of the temperature shift, i.e. under conditions where the mutant cells continue to grow and divide (Fig. 3, A  and B).
rpc10-30 specifically affects the synthesis of tRNAs, with little or no defect on mRNA accumulation and on the pol I-dependent synthesis of the large species of rRNA. There was also little effect on the pol III-dependent production of 5 S rRNA. This pattern (including the limited effect on 5 S rRNA) is indistinguishable from that observed for all pol III-specific mutants analyzed so far (see Ref. 27 for a discussion of this point). Consistent with a predominant pol III defect, ethidium bromide staining indicated that, even at the permissive temperature, rpc10-30 has a reduced steady-state content in tRNA (Fig. 3C). This pol III-specific defect of rpc10-30 is a distinctive property of that mutant. rpc10-11, which is mutated in the invariant zinc-binding motif (see above), shows a general decrease in pol I-, pol II-, and pol III-dependent transcription. This also applies to the conditional mutant strain YLR-08, where RPC10 is under the control of the doxycycline-repressible promoter pTET. As shown in Fig. 4, preventing the de novo synthesis of ABC10␣ by the presence of doxycycline leads to complete growth arrest 9 h after adding the antibiotic and determined a rapid and parallel decrease of pol I, pol II, and pol III transcription. Hence, all three RNA polymerases are coordinately turned off in cells depleted in ABC10␣, and there is no indication that ABC10␣ is, by itself, more critical to the activity of pol III.
Genetic Interaction between ABC10␣ and the Second Largest Subunits of RNA Polymerases-In view of the pol III-specific defect of rpc10-30, we tried to correct its growth defect by increasing the gene dosage of pol III-specific subunits. Multicopy plasmids encoding various pol III-specific subunits were tested for their ability to restore growth at 37°C in an rpc10-30 mutant strain. Suppression was observed only in the case of the RET1 gene encoding the second largest subunit C128 (Fig. 5A,  left panel). RPA135 and RPB2 (encoding A135 and B150, the second largest subunits of pol I and pol II) instead have a negative effect on growth even at 30°C (Fig. 5A, right panel). These positive (RET1) and negative (RPA135 and RPB2) effects are a distinctive property of rpc10-30 and were not observed in an rpc10-11 mutant nor under conditions where RPC10 is progressively depleted in vivo by the addition of increasing amounts of doxycycline (strain YLR-08).
In summary, rpc10-30 is sensitive to its own gene dosage (since it is lethal when harbored on a centromeric plasmid) and to the gene dosage of the second largest subunits of pol I, II, or III. The latter property suggests that ABC10␣ interacts with these subunits during the assembly of the three RNA polymerases and that rpc10-30 specifically impairs the interaction with C128. By using the two-hybrid assay, we have recently found that ABC10␣ can bind the C-terminal region of A135 (24). Since rpc10-30, unlike rpc10-11, is not detectably affected in pol I-dependent transcription in vivo (Fig. 3A), we wondered if these mutants would differ in their two-hybrid interaction The mutant strains YLR-06 (rpc10-30) and YLR-03 (rpc10-11) were tested at 30°C and at various times after the temperature shift to 37°C (see "Materials and Methods"). The wild type (WT) strain YLR-01 was used as control. 5 g of total RNAs were separated by electrophoresis on a 7 M urea, 6% polyacrylamide gel and autoradiographed for 24 h (25 S and 18 S rRNA) and 4 days (5.8 S, 5 S rRNA, and tRNAs). B, steady-state level of short-lived (DED1) and long-lived (ENO2) mRNA species. 8 g of the same RNA preparations as above were separated by electrophoresis on 1.2% agarose gel and probed by Northern hybridization. C, steady-state levels of pol I and pol III transcripts. 5 g of the same RNA preparations as above were separated by electrophoresis on a 7 M urea, 6% polyacrylamide gel and stained with ethidium bromide .   FIG. 4. In vivo depletion of ABC10␣. A, effect on growth. The two isogenic strains, YLR-08 (carrying RPC10 under the control of the doxycycline-repressible pTET promoter) and YLR-01 (carrying RPC10 under the control of the constitutive pPGK1 promoter), were grown exponentially on casamino acid medium as described under "Materials and Methods." After adding 5 g/ml doxycycline, growth was followed for 12 h (about four doubling times). Strain YLR-08 was labeled with [ 3 H]uracil at the times indicated by arrows. B, pol I-and pol III-dependent transcription of YLR-08 strain. Pol I and pol III activities were determined by in vivo labeling as described in Fig. 3A. C, accumulation of pol II-dependent transcripts in YLR-08 strain. The steady-state level of the DED1 and ENO2 pol II transcripts was determined by Northern hybridization as described in Fig. 3B. with A135. Fig. 5B shows that, consistent with its pol IIIspecific defect, rpc10-30 does not affect this interaction, which is instead completely abolished in an rpc10-11 context. However, we were unable to detect a similar two-hybrid interaction with the equivalent domains of the corresponding subunits of pol II (B150) and pol III (C128).
rpc10 Mutants Are Suppressed by Overexpressing Pab1p and Pbp2p, Two Components of the mRNA Polyadenylation System-In order to isolate further dosage-dependent suppressors, strain YLR-06 (rpc10-30) was transformed by a wild type genomic DNA library prepared from strain FL100 and borne on the multicopy vector pFL44L (27). Eleven dosage-dependent suppressors were isolated at 37°C. Eight of them harbored the putative RNA helicase gene DED1, previously isolated as a suppressor of the pol III-specific mutant rpc31-236 (36) and also suppressing tfc3-G349E (Table II). The remaining three suppressors corresponded to the poly(A)-binding protein gene PAB1 (two clones) and to PBP2, encoding a non-essential protein interacting with Pab1p in a two-hybrid screen (37,38). PAB1 and PBP2 had so far never been isolated as suppressors of RNA polymerase mutants.
Given that Pab1p and its putative partner Pbp2p are components of the polyadenylation system, we wondered whether ABC10␣ may directly affect mRNA polyadenylation, for example by helping yeast pol II to recruit the mRNA cleavage and polyadenylation machinery. Two observations argue against this interpretation. First, rpc10-11 and rpc10-30 show no difference in the length of their total mRNA poly(A) tails in vivo relative to an isogenic wild type control (data not shown). Second, suppression by PBP2 is not specific to rpc10 mutants, as shown in Table II. In fact, the suppression patterns of DED1 and PBP2 are quite similar (Table II). In particular, both suppress the pol III-specific mutant rpc31-236 and are thus not restricted to rpc10 mutants. Both Ded1p and Pab1p stimulate translation initiation (38,39), suggesting that their overexpression may overcome a partial translational defect of rpc31-236 and rpc10 mutants due to the pol III-dependent shortage in initiator tRNA. In agreement with this interpretation, direct measurements of the initiator tRNA Met confirmed that it is substantially reduced in rpc10-30 mutant cells. 2 rpc10-30 Impairs Pol III Accumulation but Has No Detectable Effect on Its Catalytic Properties-The data presented so far strongly suggest that rpc10-30 primarily affects in vivo pol III-dependent transcription. This mutant may directly alter the catalytic properties of the pol III enzyme or interfere with its assembly, or both. To examine these possibilities the two isogenic strains YLR-HA01 (HA-RPC10) and YLR-HA06 (HA-rpc10-30) were tested for their ability to support pol III-dependent transcription, using either whole-cell extracts or extensively purified enzyme preparations. In agreement with its pol III defect observed in vivo, Fig. 6A shows that pol III-dependent transcription was strongly affected in the rpc10-30 extract. This defect was canceled by adding the purified mutant enzyme, suggesting that it results from a shortage in the mutant enzyme.
Partial purification by affinity chromatography on heparin (Fig. 6B) yielded 4-fold less pol III activity from the rpc10-30 mutant. However, when adjusted to the same transcriptional activity and revealed by Western blotting using anti-pol III antibodies (30), the mutant and wild type peak fractions were present in the same amount and thus have the same specific activity. Moreover, the subunit composition and stoichiometry of the mutant enzyme were not different from the wild type control (Fig. 6C). Thus, rpc10-30 has no effect on the specific activity of pol III (as tested on nonspecific template) but strongly impairs its accumulation in mutant cells. This is not due to a shortage in ABC10␣ itself, since the amount of free subunit is actually higher in the mutant cells, which presumably reflects a higher copy number of the corresponding plasmid to compensate the mutant growth defect (Fig. 6D).
The mutant and wild type pol III were further purified to near homogeneity by ion exchange chromatography and tested in a reconstituted faithful transcription assay where pol IIIspecific templates are transcribed in the presence of purified pol III-specific initiation factors TFIIIC and TFIIIB (28). As shown in Fig. 6E, there was no difference between the wild type 2 O. Calvo and M. Tamame, personal communication.  Single cell colonies were obtained by streaking yeast cells on YPD plates at 37°C. Symbols: ϩϩ, growth after 3-4 days; ϩ, growth after 6 -7 days; Ϫ, no growth. Rpb10 -157F is defective in the common subunit ABC10␤ (26). tfc3-G348E and rpc31-236 are pol III-specific mutants (21,33). Other pol I, pol II, and pol III-specific mutants were tested for suppression but did not respond to the three suppressors tested.
and mutant enzymes when assayed on a SUP4 template, and similar results were obtained on tRNA Leu3 gene and on SNR6, coding the U6 small nuclear RNA (data not shown). Moreover, the mutant enzyme shows a wild type response to heat inactivation, indicating that its heteromultimeric structure is therefore not particularly labile (Fig. 6E). Taking advantage of the sequence of the tRNA SUP4 transcript, where the first G is at position ϩ18, we also tested the wild type and mutant pol III in a single round transcription assay that measures the synthesis of the first 17 nucleotides of the transcript (29). Again, no difference was observed, indicating that transcription initiation was not affected (data not shown).
In conclusion, rpc10-30 strongly reduces the amount of pol III present in whole-cell extracts but does not detectably alter its catalytic properties in a variety of transcription assays (nonspecific versus specific templates, single round or multiple round transcription). The subunit structure of the mutant en-zyme was not altered and, despite a marked temperaturesensitive defect in vivo, was not particularly unstable at high temperature. The pol III-specific phenotype of rpc10-30 therefore appears to result primarily from a defective polymerase assembly. DISCUSSION ABC10␣ was initially discovered as a common subunit shared by the three nuclear polymerases of S. cerevisiae (8). Related coding sequences have now been identified in plant and animal genomes, and this subunit is therefore widely distributed among eukaryotes. Moreover, the human and fission yeast forms of ABC10␣ can fully replace the budding yeast subunit in vivo (12,13,34), despite moderately conserved amino acid sequences. ABC10␣ has no homology to any of the eubacterial genomes sequenced so far but is also loosely related to archaeal gene products. Thus, ABC10␣, like other common subunits (14), may pre-date the separation of the eukaryotic and archaeal phyla about 1.8 billion years ago (40).
In addition to their unusual small size (58 amino acids in the case of the human product), the eukaryotic polypeptides and their putative archaeal counterparts share an invariant CX 2 CX 10 -15 CX 2 CG motif and a conserved, positively charged C-terminal domain. The former motif presumably acts as a zinc-chelating domain, since ABC10␣ has zinc-binding properties in vitro (19). Despite its evolutionary invariance, the integrity of this motif is not essential in vivo. However, the ABC10␣ domain likely contributes to the proper conformation of the subunit since a mutation in this region (rpc10-11) disrupts its ability to interact with the second largest subunit of pol I in a two-hybrid assay and leads to a partial defect in all three polymerases. In fact, there is a remarkable functional disparity between the two halves of the zinc-binding domain, since a Cys-Ser substitution at position Cys-34 is lethal, whereas a double substitution at positions 48 and 51 is phenotypically silent. The basic C-terminal domain, on the other hand, is very critical for the biological activity of ABC10␣, and mutations altering its positive charge pattern are lethal or have strong growth defects. Somewhat surprisingly, the rpc10-30 mutation mapping in this domain showed a pol IIIspecific defect in vivo, despite the fact that ABC10␣ is common to all three enzymes.
Pol III-dependent transcription can be faithfully reconstituted in vitro with highly purified preparations of the enzyme and of its cognate transcription factors TFIIIB and TFIIIC (28). Mutational defects in five of the pol III-specific subunits (C11, C31, C34, C128, and C160) were unambiguously shown to alter initiation (36,41), elongation (29,33), or termination (42,43) in vitro. rpc10-30, instead, has no detectable effect on the catalytic activity of the enzyme nor on its ability to initiate and terminate transcription from natural templates but strongly reduces the amount of transcriptionally active pol III in whole-cell extracts. Since the accumulation of free ABC10␣ in whole-cell extracts is not impaired and since the stability of pol III is not affected, rpc10-30 therefore appears to be primarily defective in the assembly of pol III. This is also consistent with its delayed transcriptional response when shifted to the restrictive temperature, which indicates that the functionality of the preassembled enzyme is not affected but that new enzyme molecules are not synthesized de novo at 37°C.
Our genetic data indicate that the second largest subunits of pol I (A135), pol II (B150), and pol III (C128) are the primary partners of ABC10␣ during polymerase assembly. Overproducing C128 partly corrects the growth defect of rpc10-30, suggesting that this particular mutant form fails to recognize C128 during pol III assembly, therefore accounting for its pol IIIspecific defect in vivo. A135 or B150 have instead an aggravat-FIG. 6. Effect of rpc10-30 on RNA polymerase III in vitro. A, pol III-specific transcription of SUP4 template using whole-cell extracts of the wild type and rpc10-30 strains YLR-HA01 and YLR-HA06. B, wild type and rpc10-30 pol III activity of the heparin-hyper D eluted fractions. RNA polymerase activity was determined by counting the acidprecipitable incorporation of [␣-32 P]UTP in the presence of a nonspecific poly(dA-dT) template (see "Materials and Methods"). C, Western blot analysis of the wild type and rpc10-30 pol III of the peak heparin-hyper D fraction 65. The wild type and mutant fractions (indicated as WT and M) were adjusted to have the same specific activities on a poly(dA-dT) template. Subunits were separated on a 6 -15% SDS-polyacrylamide gel. The positions of different pol III subunits are indicated. Note that the HA-ABC10␣-30 mutant protein migrates faster than the corresponding wild type tagged subunit, which may reflect the negative charge introduced by these mutations. D, steady-state amount of ABC10␣ in whole-cell extracts of strains YLR-HA01 (HA-ABC10␣) and YLR-HA06 (HA-ABC10␣-30). E, specific transcription assay of SUP4 template by the purified wild type and mutant pol III. The SUP4 template was transcribed for 50 min at 25°C in the presence of TFIIIB and TFIIIC fractions (see "Materials and Methods"). The enzyme preparations were preincubated for 10 min at 25,30,35,40, and 45°C prior to the transcription assay, to follow the thermic inactivation of the enzyme.
ing effect on rpc10-30, indicating that ABC10␣ competes for a domain present on each of these subunits and that the mutant subunit is therefore titrated by A135 and B150. The target domain is presumably a conserved one but remains to be precisely identified. It might be located on the C-terminal half of the second largest subunits, since a corresponding fragment of A135 binds ABC10␣ in the physiological conditions of the twohybrid assay, in a way that is dependent on the integrity of the zinc-binding motif of ABC10␣. It is worth noting that the interacting fragment overlaps with two highly conserved domains that were shown by affinity labeling data to be part of the pol II active site (44,45).
There is a striking parallel between the properties of ABC10␣ and those recently reported for ABC14.5 except that, in the latter case, the targets are the largest subunits A190, B220, and C160. ABC14.5 binds to a conserved domain of these three subunits in the two-hybrid assay (24) and, when replaced by the Schizosaccharomyces pombe subunit, leads to a pol IIIspecific defect that is partly suppressed by overproducing C160 (46). In this case too, in vitro studies are consistent with an assembly defect and show no effect on the functional properties of pol III. In the same vein, mutant forms of ABC10␤ have a preferential pol I phenotype that can also be explained by an assembly defect (26). The properties of conditional mutants isolated for AC40 (47) and ABC23 (48) also indicate assembly or stability defects but with no evidence of polymerase specificity. Hence, mutational changes in five of the common subunits are associated with assembly or stability defects and, in the case of the pol III-specific defects of ABC14.5 and ABC10␣ mutants, were directly shown to have no effect on enzyme activity in vitro. They therefore play a critical role in maintaining the complex heteromultimeric structure of eukaryotic polymerases but may not directly determine enzyme activity, except by ensuring the correct folding of the two largest subunits into a functional active site.
Given the structural role thus ascribed to common subunits, we wondered whether their rate of de novo synthesis would directly set up the formation of new polymerase molecules during cell growth. To identify which of the seven common subunits shared by pol I and pol III is rate-limiting for growth (and, therefore, for the formation of at least one of these polymerases), we compared isogenic diploid strains containing only one functional copy of the corresponding gene. ABC10␣ was the only subunit where this reduced gene dosage had an adverse effect on growth, suggesting that the formation of new polymerases directly depends on the availability of ABC10␣. The interaction between ABC10␣ and the second largest subunits of polymerases therefore appears to be a major ratelimiting step in polymerase assembly.