Association between 36- and 13.6-kDa alpha-like subunits of Arabidopsis thaliana RNA polymerase II.

Two subunits in RNA polymerase II (e.g. RPB3 and RPB11 in yeast) and two subunits common to RNA polymerases I and III (e.g. AC40 and AC19 in yeast) contain one or two motifs related to the alpha subunit in prokaryotic RNA polymerases. We have sequenced two different cDNAs (AtRPB36a and AtRPB36b), the two corresponding genes from Arabidopsis thaliana that are homologs of yeast RPB3, and an Arabidopsis cDNA (AtRPB13.6) that is a homolog of yeast RPB11. The B36a subunit is the predominant B36 subunit associated with RNA polymerase II purified from Arabidopsis suspension culture cells, and this subunit has a stoichiometry of about 1. Results from protein association assays showed that the B36a and B36b subunits did not associate, but each of these subunits did associate with the B13.6 subunit in vivo and in vitro. Two motifs in the B36b subunit related to the prokaryotic alpha subunit were shown to be required for the in vitro interactions with the B13.6 subunit. Our results suggest that the B36 and B13.6 subunits associate to form heterodimers in Arabidopsis RNA polymerase II like the AC40 and AC19 heterodimers reported for yeast RNA polymerases I and III but unlike the B44 homodimers reported for yeast RNA polymerase II.

. Bacterial RNA polymerase has an ␣ subunit with a stoichiometry of 2, and the core enzyme is composed of ␣ 2 ␤␤Ј (12). The yeast B44 subunit is reported to have a stoichiometry of 2 in RNA polymerase II (13), but the AC40 and AC19 subunits in RNA polymerases I and III have apparent stoichiometries of 1 (3,14). The stoichiometry of the yeast B12.5 subunit has not been reported.
Yeast RNA polymerase II contains a total of 12 subunits, and each of these is encoded by a single copy gene (reviewed in Refs. 3 and 4). All of the RNA polymerase II subunit genes in yeast have been sequenced. Only a limited number of RNA polymerase II subunit genes in other eukaryotes have been cloned and sequenced (4). With the exception of genes encoding the largest subunit of RNA polymerase II in soybean and trypanosomes (15)(16)(17), those RNA polymerase II subunit genes that have been identified in organisms besides yeast are reported to be single copy genes.
Nuclear RNA polymerase subunit-subunit interactions, subunit functions, and assembly pathways are only beginning to be unraveled. For example, the AC40 and AC19 subunits of yeast RNA polymerase I and III have been shown to associate with one another in a yeast two-hybrid system (9). Extragenic suppression of mutations in the AC40 and AC19 subunit genes confirmed the interaction between these two subunits and a third subunit, ABC10␤ (9). Studies on mutations in the three largest subunits of yeast RNA polymerase II indicate that the B44 subunit associates with second largest subunit (i.e. B150 or RPB2), which in turn complexes with the largest subunit (i.e. B220 or RPB1) to facilitate assembly of the enzyme (18).
Here, we report on the cloning and sequencing of genes and/or cDNAs for the 36-kDa (B36a and B36b) and 13.6-kDa (B13.6) subunits in Arabidopsis RNA polymerase II, which are homologs to yeast B44 and B12.5 (i.e. encoded by the RPB3 and RPB11 genes in S. cerevisiae), respectively, to determine the stoichiometry of the B36 subunit in the enzyme and investigate its self-association and its association with the B13.6 subunit.

MATERIALS AND METHODS
Antibody Screening of an Arabidopsis cDNA Library-An Arabidopsis thaliana (ecotype Columbia) cDNA library in YES (19) was used for antibody screening. Approximately 5 ϫ 10 5 plaque-forming units were plated on an E. coli Y1090 lawn at moderate density and blotted onto Immobilon-NC (HATF) membranes (Millipore, Bedford, MA). Replica filters were probed with affinity-purified antibody (20,21) raised against the 40-kDa subunit of cauliflower (Brassica oleraceae) RNA polymerase II (60 ng of IgG/ml in Tris-buffered saline and 1% nonfat dry milk) at room temperature for 24 h. Filters were washed with three changes of Tris-buffered saline and then reacted with goat anti-rabbit IgG conjugated to alkaline phosphatase for 90 min at room temperature. Secondary antibody and reaction with alkaline phosphatase were carried out according to the supplier's instructions (5 Prime 3 3 Prime, Boulder, CO).
Isolation of AtRPB36b and AtRPB13.6 cDNA clones and AtRPB36a and AtRPB36b genes from Arabidopsis-The AtRPB36a cDNA clone insert was used to screen another 5 ϫ 10 5 plaque-forming units of the YES cDNA library (19). Six positive clones were selected, purified, and sequenced. Five of the six clones contained identical sequence (of varying length) to the AtRPB36a cDNA clone, and the remaining clone, AtRPB36b, contained a similar but distinct sequence.
An EST cDNA clone (GenBank accession number Z47635) from an Arabidopsis cell suspension library (22) was identified which had homology to yeast RPB11 (8). The EST sequence was reported as a partial sequence of a full-length cDNA clone. The complete sequence of this cDNA clone was obtained from the EST cDNA clone, which was provided by Dr. Gabriel Phillips (Laboratoire de Biologie Moleculaire de Plantes, CNRS, Strasbourg Cedex, France). We refer to this clone as AtRPB13. 6.
Both AtRPB36a and AtRPB36b cDNA clone inserts were used to screen an A. thaliana (ecotype Columbia) EMBL3 genomic library (provided by Harry Klee, Monsanto Chemical Company, St. Louis, MO). Genomic clones were selected, purified, and mapped with restriction enzymes. Restriction fragments corresponding to genomic fragments of AtRPB36a and AtRPB36b were subcloned into pBluescript (Stratagene, La Jolla, CA) vectors or pMOB (23) and sequenced using a Tn1000 kit (Gold Biotechnology, St. Louis, MO).
DNA Sequence Analysis-Oligonucleotides used for sequencing and cloning procedures were synthesized at the University of Missouri DNA Core Facility. Sequencing was performed manually using Sequenase (U.S Biochemical Corp.) and by automated sequencing using the DyeDeoxy procedure (Applied BioSystems Inc., Foster City, CA). Computer analysis was performed using the BLAST family of programs (24) and the E-mail BLAST server at National Center for Biotechnology Information, the Genetics Computer Group package (Genetics Computer Program, Madison, WI), and IBM Pustell Sequence Analysis software (International Biotechnologies, Inc., New Haven, CT). Protein sequence alignments were done using GAP, BESTFIT, and PILEUP programs from the GCG package with the Gap Weight and Gap Length Weight parameters 3.0 and 0.1, respectively.
Northern blotting was carried out with 2 g of poly(A) ϩ RNA isolated from Arabidopsis suspension culture cells (27). RNA was isolated by a standard protocol (25), denatured with glyoxal and Me 2 SO, subjected to electrophoresis on 1.4% agarose gels, and transferred to a nylon membrane (26). AtRPB36a and AtRPB36b cDNAs were labeled with 32 P using the Prime-a-Gene labeling system (Promega Corp., Madison, WI). A mixed probe was used for hybridization in 6 ϫ SSPE (26), 1% nonfat dry milk, 1% SDS, and 0.5 mg/ml denatured herring sperm DNA at 68°C. Washings were in 2 ϫ SSC and 0.1% SDS for 15 min at 25°C, 0.5 ϫ SSC and 0.1% SDS for 15 min at 25°C, and 0.2 ϫ SSC and 1% SDS for 30 min at 50°C Expression of Cloned cDNAs in E. coli-Small N-terminal portions of open reading frames of AtRPB36a and AtRPB36b were isolated using specific primers, Pfu DNA polymerase (Stratagene), and polymerase chain reaction. These were cloned in-frame to the HisTag pET-16b expression vector (Novagen, Madison, WI). In order to avoid mistakes introduced by polymerase chain reaction, most of the ORFs 1 (with the exception of small N-terminal regions) were cloned by using corresponding restriction fragments from the original cDNAs. The N-terminal regions made by polymerase chain reaction were verified by sequencing. Expression of the clones was induced by the addition of IPTG to midlog cultures of the BL21(DE3) strain of E. coli. After 2-3 h of induction, cells were harvested and sonicated. Fusion proteins were purified from inclusion bodies under denaturing conditions as described by the supplier (Novagen).
Nondenaturing gels were constructed and subjected to electrophoresis at 4°C using the Laemmli system (29) minus SDS.
Epitope Tagging and Immunoprecipitation-A double-stranded oligonucleotide was cloned upstream of the start codon for a given ORF to create an in-frame influenza hemagglutinin (HA) epitope tag. These constructs encoded proteins with a 12-amino acid extension, MGYPY-DVPDYAH (the HA epitope is underlined), at their N terminus. Epitope-tagged and untagged cDNAs were co-translated in vitro (as described above), and the 35 S-labeled in vitro translation products were immunoprecipitated with 12CA5 monoclonal antibodies (Berkeley Antibody Company, Richmond, CA), which were immobilized on Protein A-agarose (Sigma). Five l of immobilized antibody was added to 10 l of in vitro translation mixture and adjusted to 200 l with PBS/Tween (phosphate-buffered saline containing 0.05% Tween 20). Immunoprecipitation was carried out at 4°C for 12 h in a rotator. After incubation, the resin was washed five times with 1 ml of ice-cold PBS/Tween, and immunoprecipitates were eluted from the resin in 50 l of SDS sample buffer (29). Five l samples were resolved on 10 or 15% SDS gels and subjected to autoradiography.
N-terminal and C-terminal Truncations of the B36b Subunit-Truncations in the N-terminal or C-terminal portions of the B36b subunit were made using a series of restriction endonuclease sites. The B36b subunit was chosen for truncations because of some convenient restriction endonuclease sites not found in the B36a subunit cDNA. A chimeric construct was made that consisted of amino acids 249 -319 of the B36b subunit fused to the C terminus of GH2/4 (32). The GH2/4 cDNA encodes a glutathione S-transferase (33).
Purification of Arabidopsis RNA polymerase II and Subunit Analysis-A. thaliana (ecotype Columbia) cell cultures were grown in liquid media (27) as 1-liter cultures in 4-liter flasks with constant agitation on a rotatory shaker at 25°C. Eight days after subculture, cells were harvested on two layers of Miracloth (Calbiochem, La Jolla, CA) using a Buchner funnel. Cells were washed with several volumes of cold distilled H 2 0. Water was removed by vacuum filtration, and cells were frozen in liquid N 2 and stored at Ϫ80°C prior to RNA polymerase II purification.
For purification of RNA polymerase II, 200 g of cells were thawed and suspended in 200 ml of grinding buffer (50 mM Tris-HCl (pH 8.0), 0.1 mM EDTA, 60 mM ammonium sulfate, 0.5 mM dithiothreitol, and 20% (v/v) ethylene glycol) containing 1 mM Pefabloc SM (Boehringer Mannheim) 10 g/ml aprotinin, 1 g/ml pepstatin, 300 g/ml benzamidine, and 10 g/ml leupeptin. All purification steps were carried out at 4°C. Cells were broken by grinding for 2 min using full speed with a Polytron PT20ST and subsequently with 15 30-s bursts and 90-s intermittent periods with a Bead-Beater and 100 g of acid-washed glass beads (425-600 microns; Sigma). The homogenate was filtered through two layers of Miracloth and centrifuged at 10,000 ϫ g for 20 min. The supernatant was collected, and RNA polymerase II was purified by precipitation with and elution from Polymin P, ammonium sulfate precipitation, and chromatography on DEAE cellulose and phosphocellulose as described by Jendrisak and Burgess (34). The phosphocellulose fraction was dialyzed against 20 mM HEPES (pH 7.8), 0.1 mM EDTA, 0.5 mM dithiothreitol, and 50% glycerol, and dialysate was stored frozen at Ϫ80°C. Wheat germ RNA polymerase II was purified using the methods of Jendrisak and Burgess (34) with final chromatography on heparin-Sepharose (21).
The purified RNA polymerase II was judged to be nearly homogeneous on 15% SDS-polyacrylamide gels (29) when compared with purified wheat germ RNA polymerase II. Subunit stoichiometries were determined for the three largest subunits (205 ϩ 175-, 135-, and 36-kDa subunits) in purified Arabidopsis RNA polymerase II using 7.5 and 15% SDS gels. Peak areas for the three largest subunits were measured for Coomassie Blue-stained gels using Image I Software (Universal Image, Corp., Westchester, PA). Quantitation of 35 S incorporation into the three largest subunits was carried out with a Fuji BAS1000 instrument and MacBAS1000 software (Fuji Medical Systems, Stamford, CT).
Nucleotide Sequence Accession Numbers-The nucleotide sequences for the AtRPBC13.6 cDNA clone and the AtRPB36a and AtRPB36b genes reported in this paper are entered in the GenBank nucleotide sequence data base with accession numbers U28048, L34770, and L34771, respectively. The accession number for the Arabidopsis AC14 EST cDNA clone is Z25617.

RESULTS
Cloning the Arabidopsis RPB36 Subunit cDNAs-A polyclonal antibody raised against the 40-kDa subunit of cauliflower RNA polymerase II (21) was affinity-purified on a wheat germ RNA polymerase II Affi-Gel resin (see "Materials and Methods" and Ref. 20). The affinity purified antibody was used to screen an Arabidopsis cDNA expression library as described previously (20). Seventeen positive clones were selected from primary screening, and five were purified to homogeneity. One of the purified clones, AtRPB36a, was sequenced and found to be related in amino acid sequence to yeast RPB3 (10) and human hRPB33 (Ref. 35; Fig. 1). The Arabidopsis cDNA had an insert of 1.3 kb that contained an ORF encoding 319 amino acids with a predicted molecular mass of 35.5 kDa and an estimated pI of 4.4. The Arabidopsis B36a amino acid sequence showed 39% identity to the RPB3 subunit in yeast RNA polymerase II and 44% identity to the hRPB33 subunit in human RNA polymerase II but only 31% identity to the AC40 subunit in yeast RNA polymerases I and III (11).
Southern blot analysis of Arabidopsis genomic DNA suggested that more than one copy of this subunit gene was present in the Arabidopsis genome because a variety of restriction endonucleases produced multiple restriction fragments (of varying intensities) that hybridized to the AtRPB36a cDNA probe ( Fig. 2A). To determine if more than one gene encoded the 36-kDa RNA polymerase II subunit, we rescreened 5 ϫ 10 5 plaque-forming units of the YES cDNA library with the AtRPB36a cDNA and selected six positive clones. Each purified clone was partially sequenced. Five of these were identical in sequence to AtRPB36a with the exception of the position of the poly(A) tail in the 3Ј-untranslated region (data not shown), reflecting heterogeneity in the site selection for poly(A) addition. One of the clones contained a full-length cDNA that was related, but distinct from AtRPB36a. This 1.2-kb cDNA clone, AtRPB36b, contained an ORF encoding 319 amino acids with 88% identity to the amino acid sequence in AtRPB36a and 37% identical to yeast RPB3 (Fig. 1). The predicted pI of the B36b protein was 4.7. Within the ORFs, AtRPB36a and AtRPB36b showed 91% identity in nucleotide sequence, and in the untranslated regions, the two cDNA clones were 82% identical (data not shown). A Northern blot with a mixed AtRPB36a and AtRPB36b probe revealed only one size mRNA of 1.5 kb (Fig.  2B). We have not attempted to quantitate the relative amounts of the individual AtRPB36a and AtRPB36b mRNAs. Identical amino acids that are shared by more than one species are shaded in each subunit. Two domains with homology to the ␣ subunit of E. coli RNA polymerase are shown with a double underline. ␣-motif 1 is the N-terminal "␣ motif " (2,7,9), and ␣-motif 2 is the leucine-rich C-terminal ␣-like motif (7,36). Asterisks indicate the positions of cysteines in the putative metal-binding motifs of B36a and B36b. Positions of N-terminal and C-terminal truncations made in the B36b subunit are indicated with arrows above the sequence alignments.
The ORFs in AtRPB36a and AtRPB36b encode putative metal-binding motifs (i.e. "zinc-fingers"), CX 2 CX 5 CX 2 C, starting at position Cys 90 in B36a (Fig. 1). The motif in B36b differs from that in B36a because the B36b clone contains an N-terminal extension of this motif, CX 2 CX 2 CX 5 CX 2 C. These putative metal-binding motifs differ slightly from those found in the homologous RNA polymerase II subunit in S. cerevisiae (10), Schizosaccharomyces pombe (36), human (35), and Tetrahymena thermophila (7), which are conserved as CXCX 3 CX 2 C. The yeast RPB3 subunit has been reported to bind 65 Zn using a zinc-blotting technique (37), and we have preliminary evidence that the Arabidopsis B36a and B36b subunits bind zinc using the methods of Treich et al. (37). 2 The B36a and B36b subunits contain two motifs related to the prokaryotic RNA polymerase ␣ subunit (Fig. 1). One of these motifs (the more N-terminal) consists of a stretch of amino acids that is referred to as the "␣ motif" (Refs. 7 and 9; reviewed in Refs. 3 and 4). The second ␣-like motif consists of a leucine-rich C-terminal region including amino acids Leu 281 to Leu 298 in B36a and B36b. Both of these ␣-like motifs have been previously identified in S. cerevisiae, S. pombe, Tetrahymena, and human subunit homologs (7,36).
The Genes for AtRPB36a and AtRPB36b-To determine the gene structure and promoter sequences for B36a and B36b subunits, we isolated the genes corresponding to the two AtRPB36 cDNAs. The two cDNA clones were used to screen an Arabidopsis EMBL genomic library. Ten genomic clones were selected, purified, and mapped with restriction enzymes.
Genomic fragments for both AtRPB36a and AtRPB36b were subcloned into plasmids, and 5 kb of AtRPB36a and 3.8 kb of AtRPB36b were sequenced. Both genes were composed of three exons with conserved exon/intron borders (Fig. 3). While each exon showed about 90% nucleotide sequence identity between the two genes, the two introns showed only 76 and 65% identity. The promoters of the AtRPB36 genes showed little conservation in nucleotide sequence except near the TATA box and start site of transcription (i.e. 80% identity). The two genes had only 43% identity over 2 kb upstream of the 5Ј ends of the cDNAs (see GenBank entries). Likewise, no highly conserved sequences were identified when AtRPB35a or AtRPB36b promoters were compared with other Arabidopsis RNA polymerase II subunit promoters (i.e. 205-, 135-, or 19.5-kDa subunit promoters) (15,20,38).
The Subunit Structure and Stoichiometry of the B36 Subunit in RNA Polymerase II Purified from Arabidopsis Suspension Culture Cells-Since both Arabidopsis AtRPB36 genes were expressed and both promoters showed similar patterns of expression in transgenic plants, 2 it was of interest to determine if both B36 subunits could be detected in purified RNA polymerase II from Arabidopsis. To make this determination, Arabidopsis RNA polymerase II was purified to homogeneity from suspension culture cells that were or were not labeled with [ 35 S]methionine. Fig. 4A shows Arabidopsis RNA polymerase II subunit structure from two independent purifications (lanes 1 and 2) compared with wheat germ RNA polymerase II (lane 3). Antibody raised against the 40-kDa subunit of cauliflower RNA polymerase II reacted strongly with the Western blotted B36 subunit in purified Arabidopsis RNA polymerase II but showed only a weak reaction with the homologous subunit (AC42) in purified Arabidopsis RNA polymerase III (data not shown). A Coomassie Blue-stained gel and phosphor image of the 35 Slabeled subunits in the gel are shown in the left and right panels, respectively (Fig. 4A). The subunit composition of Arabidopsis RNA polymerase II is highly similar to other plant RNA polymerase II enzymes, with each enzyme containing 11 subunits resolved in 15% SDS gels (39). To obtain better resolution of the three largest subunits (i.e. 205-, 135-, and 36-kDa) in RNA polymerase II, the enzymes shown in Fig. 4A were subjected to electrophoresis in 7.5% SDS-polyacrylamide gels (Fig. 4B). The largest subunit (205 kDa) showed some degradation to a 175-kDa polypeptide depending on the enzyme preparation. On the 7.5% gels, it should be possible to detect both B36 subunits in the purified enzyme because the B36a subunit migrates more rapidly in SDS-polyacrylamide gels than the B36b subunit. The more rapid migration of the B36a subunit is observed with both His-tagged (Fig. 4B, left panel) and in vitro translated subunits (Fig. 4B, right panel).  subunits or recombinant His-tagged subunits to the third largest subunit of Arabidopsis RNA polymerase II revealed that the B36a subunit was the predominant third largest subunit in the enzyme purified from suspension culture cells. A subunit with the predicted mobility of the B36b subunit was not detected in Coomassie Blue-stained gels (Fig. 4B, left panel) or by autoradiography with 35 S-labeled Arabidopsis RNA polymerase II (Fig. 4B, right panel). From this analysis, it is not clear whether the B36b subunit is a minor component of RNA polymerase II in suspension cells, is associated with a form of the enzyme that fails to purify using the methods employed, is modified in vivo so that it migrates identically to B36a, or fails to associate with the enzyme.
The third largest subunits in S. pombe and S. cerevisiae RNA polymerase II are reported to have a stoichiometry of 2 in the purified enzymes (13,36), while the homologous subunits in plant and animal RNA polymerase II are reported to have a stoichiometry of one (39 -43). To determine the stoichiometry of the B36a subunit in Arabidopsis RNA polymerase II, we measured the peak areas for the 205 ϩ 175-, 135-, and 36-kDa subunits by imaging (Image I Software) 7.5% Coomassie Bluestained gels (Fig. 4B, left panel). This analysis indicated that the stoichiometries of the three largest subunits were 1, 1.1, and 1.3 (i.e. using the largest subunit, 205 ϩ 175, as the base line), for the 205 ϩ 175-, 135-, and 36-kDa subunits, respectively. As a second method for determining stoichiometry of these subunits, we quantitated the 35 S incorporation in the three largest subunits in Arabidopsis RNA polymerase II that had been labeled in vivo with [ 35 S]methionine. Since the amino acid sequences for the 205-(15), 135- (38), and 36-kDa subunits (see Fig. 1) are known, the number of methionines in each subunit could be divided into the 35 S incorporated into each subunit (i.e. determined by phosphor imaging) to determine the subunit stoichiometries. The stoichiometry for each subunit was near unity. The relative stoichiometries obtained for three independent assessments are shown to the right of each subunit in Fig. 4B (right panel). With results from Coomassie Blue staining and 35 S labeling taken together, the best estimate for stoichiometry of the B36 subunit in Arabidopsis RNA polymerase II is 1.
The Arabidopsis B13.6 Subunit-Because the stoichiometry of the B36 subunit appeared to be 1 in purified Arabidopsis RNA polymerase II and because a second subunit with homology to the prokaryotic ␣ subunit of RNA polymerase has been identified in yeast (RPB11; Ref. 8) and human RNA polymerase II (hRPB14; Ref. 44), we searched the data base for a plant homolog to yeast RPB11. An EST from an Arabidopsis cell suspension cDNA library (22) was identified (GenBank accession number Z47635). Several additional AtRPB13.6 cDNA clones were isolated from the Arabidopsis cell suspension cDNA library, and all of these were identical in sequence (with the exception of variable lengths in the untranslated regions). 3 We sequenced a full-length AtRPB13.6 cDNA clone, and the derived amino acid sequence indicated that the Arabidopsis homolog (B13.6) is a 13.6-kDa subunit with pI of 6.5 (Fig. 5). The B13.6 subunit has 41 and 42% amino acid sequence identity with yeast RPB11 and human hRPB14, respectively (Fig.  5). The Arabidopsis B13.6 subunit, like the yeast RPB11 and human hRPB14 subunits, contains the "␣ motif " and a second leucine-rich C-terminal motif that may be related to the ␣ subunit in prokaryotic RNA polymerases (see also Fig. 9). Preliminary evidence, based on Southern analysis and cDNA cloning, indicates that the gene encoding B13.6 is single copy in the Arabidopsis genome. 3 Arabidopsis B36a and B36b Subunits Associate with the B13.6 Subunit in Vivo-The unit stoichiometry of the B36 subunit in Arabidopsis RNA polymerase II and the identification of a second smaller subunit, B13.6, containing the "␣ motif " suggested that these two subunits might associate as heterodimers in the Arabidopsis enzyme. Lalo et al. (9) have previously shown that two yeast subunits, AC40 and AC19, which are common to RNA polymerases I and III, associate in vivo in a yeast two-hybrid system. The yeast AC40 and AC19 subunits contain the "␣ motif " as do the yeast RPB3 (also Arabidopsis B36a and B36b) and RPB11 (also Arabidopsis B13.6) subunits. Two determine if the B36a and B36b subunits associated with themselves, with one another, or with the B13.6 subunit, we tested these subunits in a yeast two-hybrid system.  B36b subunits do not associate as homodimers or heterodimers in the two-hybrid system. On the other hand, the B36a and B36b subunits do associate with the B13.6 subunit. This interaction was specific because the GAL4-B13.6 fusion protein failed to interact with out-of-frame B36a and B36b clones or with an unrelated SNF4 clone (45). Likewise GAL4-B36a and GAL4-B36b fusion proteins failed to interact with an SNF1 clone, while SNF1 and SNF4 did interact in this system (45). These results with the B36 and B13.6 fusion proteins are similar to those found by Lalo et al. (9) for yeast AC40 and AC19 and suggest that B36a-B13.6 and B36b-B13.6 associate as heterodimers in vivo.
In Vitro Protein-Protein Interactions with the B36a, B36b, and B13.6 Subunits-To confirm the interactions among the Arabidopsis B36a, B36b, and B13.6 subunits, we carried out in vitro protein-protein interaction analyses. One of these analyses involved immunoprecipitating (with HA monoclonal antibody) 35 S-labeled in vitro co-translated subunits fused with or without HA epitope-tags. To test whether the B36a and B36b subunits could associate with themselves or with one another in vitro, we used an epitope-tagged B36b subunit. Only the epitope-tagged B36b subunit was immunoprecipitated whether it was translated along with the untagged B36a or the untagged B36b subunit (Fig. 6A). The untagged B36a and untagged B36b subunits could be distinguished from the epitopetagged B36b subunit because of their different mobilities in SDS-polyacrylamide gels. These results indicated that the B36 subunits did not form stable associations with themselves or with one another in vitro. On the other hand, the B36a and B36b subunits were immunoprecipitated when co-translated with an epitope-tagged B13.6 subunit (Fig. 6B). A subunit related to the B36 subunits in Arabidopsis RNA polymerase I and III, AC42 (46), was not immunoprecipitated with epitopetagged B13.6. As an additional control, we used an IAA4/5 cDNA clone (GenBank accession number X68216), to synthesize an in vitro translated epitope-tagged protein that was unrelated to any RNA polymerase subunit. Neither the B36a nor the B36b subunit was immunopreciptated with the epitopetagged IAA4/5 protein. No in vitro translated proteins were immunoprecipitated in the absence of an epitope-tagged sub- FIG. 5. Comparison of amino acid sequences derived from Arabidopsis AtRPB13.6 with homologous subunits in yeast (B12.5 or RPB11) and human (B14). Identical amino acids found in more than one species are shaded in each subunit. The N-terminal "␣ motif" (␣-motif 1) and a leucine-rich C-terminal ␣-like motif (␣-motif 2) with homology to ␣ subunit of E. coli RNA polymerase are shown with a double underline.

TABLE I
Quantitative measurement of GAL1-lacZ transactivation by hybrid GAL4 proteins Relative ␤-galactosidase activity was measured in triplicate using a chemiluminescent assay. Enzymatic activities represent the number of photons counted (ϫ 10 Ϫ3 ) on an ASI 3010 luminometer (Analytical Scientific Instruments, Alameda, CA). The GAL4 DNA binding domain in pAS1 and activation domain in pACTII is described by Durfee et al. (31). The SNF1/SNF4 pair used as a positive control for interaction is a protein kinase and its substrate identified in a yeast two-hybrid screen by Yang et al. (45).  6. B36a, B36b, and B13.6 subunit interactions assayed by immunoprecipitation with epitope-tagged subunits. A, B36a and B36b subunit interactions. Lanes 1, 2, 4, and 5 are autoradiograms of in vitro translated subunits. Lane 1, B36a; lane 2, co-translated B36a and HA epitope-tagged B36b; lane 4, B36b; lane 5, co-translated B36b and HA epitope-tagged B36b. Lanes 3 and 6 are autoradiograms of immunoprecipitates with co-translated subunits (shown in lanes 2 and 5) using an HA epitope-tag and HA monoclonal antibody. In vitro translation products were resolved on 10% SDS-polyacrylamide gels. B, B36a and B36b interactions with HA epitope-tagged B13.6. Odd-numbered lanes are autoradiograms of in vitro translated subunits used in immunoprecipitation assays. Lane 1, co-translated B36a and B13.6; lane 3, co-translated B36b and B13.6; lane 5: co-translated Arabidopsis AC42 and B13.6; lane 7, co-translated B36b and epitope-tagged IAA4/5 (i.e. IAA4/5 is an auxin-induced cDNA from pea and is not related to any RNA polymerase subunit). Even-numbered lanes are autoradiograms of immunoprecipitates of co-translated subunits (shown in odd-numbered lanes). The B13.6 subunit was epitope-tagged in lanes 1-6, and the IAA4/5 polypeptide was epitope-tagged in lanes 7 and 8. In vitro translation products were resolved on 10% Tricine-SDS-polyacrylamide gels. Positions of subunits are indicated adjacent to the autoradiograms. unit or in the absence of HA antibody (data not shown).
A second in vitro approach that showed B36b interaction with B13.6 was obtained by resolving in vitro co-translated subunits by electrophoresis in polyacrylamide gels under nondenaturing conditions. When the B36b subunit was co-translated with the B13.6 subunit, the B36b subunit showed a mobility shift in the gels (Fig. 7). The gel lane (lane 2) containing a band with decreased mobility was excised and subjected to electrophoresis in an SDS-polyacrylamide gel. The SDS gel showed that the mobility shift in the nondenaturing gel was due to association of the B13.6 subunit with the B36b subunit.
The N-terminal and C-terminal Truncations in B36b "␣ Motifs" Prevent Association with B13.6 -The C-terminal region in the Arabidopsis B36 subunits contains a motif related to the prokaryotic RNA polymerase ␣ subunit (see Fig. 1). This motif is rich in leucine, and it has been suggested that this motif in S. pombe RPB3 and homologous subunits in other RNA polymerases consists of a putative leucine-zipper motif that might be involved in subunit-subunit interactions (36). To test whether this C-terminal motif in Arabidopsis B36b might be involved in interactions with B13.6, we tested several C-terminal truncations using the immunoprecipitation assay with in vitro co-translated subunits as described above. Fig. 8A shows a schematic diagram of these truncations, and the precise positions of the truncations are shown in Fig. 1. Truncations at the extreme C terminus of B36b did not prevent association with B13.6 (Fig. 8C, lanes 1-4); however, truncations that deleted a portion of the leucine-rich motif (C288; Fig. 8C, lane 5) or the entire leucine-rich motif (C244; Fig. 8C, lane 6) resulted in B36 subunits that did not associate with B13.6. These results are consistent with the C-terminal motif in B36b being important for subunit-subunit interactions. C-terminal truncations in B36a subunit have not been tested, but this C-terminal region in B36a is strongly conserved with that in B36b (see Fig. 1).
We did not observe interactions between the leucine-rich C-terminal ␣-like motif in B36b and B13.6 when the C-terminal motif of B36b was tested as a fusion protein in isolation from the remainder of the B36b subunit (GST-N249) (Fig. 8C, lane   8). In this case, amino acids 249 -319 in B36b were fused to a GH2/4 protein at its C terminus (32). The GH2/4 cDNA encodes a glutathione S-transferase (33). Failure to observe interaction between the fused B36b C-terminal motif and B13.6 could result if more than one interaction motif or a more extensive interaction motif is required for the stable association of B36b and B13.6. Since the "␣ motifs" in the N-terminal regions of RPB3 and AC40 in yeast appear to be important for subunit interactions and enzyme assembly (9, 18), we made an Nterminal truncation that removed a portion of the "␣ motif " in B36b. This truncated subunit was not immunoprecipitated with epitope-tagged B13.6 in co-translation experiments (Fig.  8C, lane 7). This result is consistent with there being two motifs or one extended motif (i.e. including both ␣-like motifs in B36) involved in the B36 subunit interaction with the B13.6 subunit. While it is possible that the N-terminal and C-terminal truncations in B36b resulted in conformational changes (e.g. incorrect folding of the in vitro translated truncations), which indirectly prevented interaction with B13.6, the truncation experiments suggest that both ␣-like motifs in B36a may be required for interaction with B13.6.
Two ␣-Like Motifs Are Found in both the Arabidopsis B36 and B13.6 Subunits and Homologous Subunits in Other RNA Polymerases-The "␣ motif " has previously been reported to be found in both the large (i.e. 36 -44-kDa) and small (i.e. 12.5-19-kDa) subunits of RNA polymerases I, II, and III (reviewed in Refs. 3 and 4). On the other hand, the leucine-rich C-terminal ␣-like motif has only been reported to be found in the 36 -44-kDa subunits in all three classes of eukaryotic RNA polymerase (7,36). Arabidopsis B36 and B13.6 subunits contain this "␣ motif " (Figs. 1, 5, and 9A). The ␣-like subunits in Arabidopsis RNA polymerases I and III, AC42/43 (a homolog of yeast AC40) and AC14 (a homolog of yeast AC19), also contain the "␣ motif" (Fig. 9A). An alignment of the "␣ motif" for pairs of ␣-like large and small subunits in Arabidopsis, yeast, and human are shown in Fig. 9A.
In addition to the "␣ motif," Arabidopsis B36 subunits (Figs. 1 and 9B) and AC42/43 subunits (Fig. 9B) contain the leucinerich C-terminal ␣-like motif originally pointed out by Martindale (7) and Azuma et al. (36) for S. pombe RPB3, S. cerevisiae RPB3 and AC40, human RPB33, and Tetrahymena CnjC subunits. Inspection of the B13.6 subunit in Arabidopsis RNA polymerase II, the homologous subunits in RNA polymerase II from other organisms, and the related 12.5-19-kDa subunits in RNA polymerases I and III suggests that the leucine-rich Cterminal ␣-like motif is also conserved in these small ␣-like subunits (Fig. 9B). Thus, both the larger 36 -44-kDa ␣-like subunit and the smaller 12.5-19-kDa ␣-like subunit in RNA polymerases I, II, and III contain two motifs with similarity to the prokaryotic RNA polymerase ␣ subunit. One of these domains has been previously referred to as the "␣ motif" (2,4,6,7,9), and the second is a more C-terminal motif that is rich in leucine (7,36). While these two ␣-like motifs are spaced apart from one another in the 36 -44-kDa subunits, they are nearly contiguous in the 12.5-19-kDa subunits (see Figs. 1, 5, and 8A).

DISCUSSION
Our results have shown that Arabidopsis contains two genes that encode the third largest subunit of RNA polymerase II. This differs from other RNA polymerase II genes in Arabidopsis and in most other organisms studied, where the genes are single copy. The two genes are expected to encode the third largest subunit in RNA polymerase II, based on stronger homology to the third largest subunit in yeast and human RNA polymerase II and lesser homology to a related subunit in yeast and mouse RNA polymerases I and III. This is supported by our cloning of two additional cDNAs from Arabidopsis that encode proteins that show stronger homology to the yeast and mouse AC40 subunits in RNA polymerases I and III than to the yeast RPB3 or human hRPB33 subunit in RNA polymerase II (46). Therefore, it appears that Arabidopsis expresses two genes for the 36-kDa subunit in RNA polymerase II and two genes for the 42/43-kDa subunit in RNA polymerase I and III. All four polypeptides encoded by these genes contain the N-terminal "␣ motif " (7) and a leucine-rich C-terminal ␣-like motif (7,36), which are related in amino acid sequence to the prokaryotic RNA polymerase ␣ subunit. Two other cDNAs from Arabidopsis that contain these ␣-like motifs in their ORFs have been cloned. One of these, described above, encodes a protein of 13.6 kDa (B13.6) that is a homolog of the yeast RNA polymerase II B12.5 (RPB11) subunit. The second encodes a protein of 14 kDa, which is a homolog of the yeast RNA polymerase I and III FIG. 9. The N-terminal "␣ motif " and a leucine-rich C-terminal ␣-like motif in B36a, B36b, B13.6, and homologous subunits in other eukaryotic RNA polymerases. A, The N-terminal "␣ motif " (␣ motif 1) in large and small ␣-like subunits in RNA polymerases I, II, and III from Arabidopsis, yeast, and human. B, the leucine-rich Cterminal ␣-like motif (␣ motif 2) in large and small ␣-like subunits in RNA polymerases I, II, and III from Arabidopsis, yeast, and human. Amino acid alignments are shown with conserved amino acids shaded black for identity (the predominant amino acid at each position) and gray for similarity (BOXSHADE program, Kay Hofman, Bioinformatic group, Lausanne, Switzerland). Leucines and isoleucines that predominate at specific positions are indicated at the top. B36a, B36b, B13.6, AC42, AC43, and AC14 are Arabidopsis subunits. Yeast subunits are indicated by a y, and human subunits by an h. EcRpoA is the E. coli ␣ subunit, and TobCt is the tobacco chloroplast ␣-like subunit. The N-terminal amino acid position for the conserved domain in each subunit is indicated to the left of the sequences. AC19 subunit. 4 The different mobilities of the B36 subunits in SDS-polyacrylamide gel electrophoresis allowed us to distinguish the B36a from the B36b subunit. RNA polymerase II purified from Arabidopsis suspension culture cells contains a third largest subunit with an apparent molecular mass of 40 kDa, which migrates identically to in vitro translated B36a in SDS-polyacrylamide gel electrophoresis. If the B36b subunit is associated with purified RNA polymerase II, then it is present in amounts not detectable by Coomassie Blue staining or by autoradiography in 35 S-labeled enzyme. The reason for predominance of the B36a subunit in the purified enzyme is not clear because the promoters for both AtRPB36 genes are active in transgenic tobacco tissues undergoing cell division and in transfected protoplasts from carrot suspension culture cells. 2 Furthermore, based on cDNA cloning, both AtRPB36 genes are expressed in the rapidly dividing Arabidopsis suspension culture cells from which the RNA polymerase II was purified. While we have not quantitated the relative amounts of B36a and B36b mRNAs in suspension culture cells, the fact that six out of seven cDNA B36 cDNA clones isolated were B36a suggests that the B36a mRNA is more abundant than the B36b. Our results on in vivo and in vitro protein-protein interactions suggest that the B36b subunit is not defective in assembly (i.e. at least assembly with the B13.6 subunit). The high conservation in amino acid sequence between B36a and B36b suggests that both subunits should be capable of assembly into RNA polymerase II. It is possible that the B36b subunit is expressed at much lower levels or assembles into RNA polymerase less efficiently than the B36a subunit in cell suspension cultures and that this subunit has simply escaped our detection. The B36b subunit may be more abundant in Arabidopsis RNA polymerase II found in specific tissues or specific developmental stages. It is worth noting that heterogeneity in the size of the third largest subunit in RNA polymerase II has been reported for enzymes purified from wheat and rye embryos (39,42), suggesting that more than one gene encodes this subunit in other plant species.
Based upon the yeast two-hybrid system and immunoprecipitation experiments with epitope-tagged in vitro translated subunits, the B36a and B36b subunits fail to stably associate with themselves or one another but do associate with the B13.6 subunit. Our in vivo results are similar to the in vivo results of Lalo et al. (9), who used the yeast two-hybrid system to show that the yeast RNA polymerase AC40 and AC19 subunits associate with one another as a heterodimer, but that the AC40 subunit fails to homodimerize. Our in vitro protein-protein interaction results confirm the in vivo results. Based upon these in vivo and in vitro protein-protein interaction results and the apparent unit stoichiometry of the Arabidopsis B36 and yeast AC40 subunits, it is possible that heterodimers, Arabidopsis B36/B13.6 and yeast AC40/AC19, are the equivalent of an ␣ 2 homodimer in prokaryotic RNA polymerase. A stoichiometry of 1 for the third largest subunit of other plant and animal RNA polymerase II enzymes has been documented previously, based upon the intensity of subunit staining with Coomassie Blue (39 -43, 47). Unit stoichiometry for the third largest subunit in Arabidopsis RNA polymerase II is supported by the relative intensity of Coomassie Blue staining and by the ratio of 35 S incorporated to the number of methionines in this subunit compared with the two largest subunits. In contrast to our results with Arabidopsis RNA polymerase II, the RPB3 subunits in S. pombe and S. cerevisiae RNA polymerase II are reported to have a stoichiometry of 2 (13,36), and evidence has been presented that is consistent with a S. cerevisiae RNA polymerase II assembly pathway initiating with the homodimerization of RPB3 and subsequent interaction with RPB2 and RPB1 (18). While this proposed assembly pathway for yeast RNA polymerase II resembles that reported for bacterial RNA polymerase (48), there is no direct evidence for the homodimerization of yeast B44 (RPB3) subunits. It remains possible that yeast B44 and B12.5 (RPB11) form heterodimers like Arabidopsis B36 and B13.6. On the other hand, it is possible that the subunit stoichiometries and assembly pathways differ in RNA polymerase II from yeast and plants. Lalo et al. (9) showed that a number of mutations in the "␣ motif " (see Fig. 9) of yeast AC40 were lethal. On the other hand, similar mutations within the "␣ motif " of yeast AC19 produced only minor growth defects. These results suggest that the "␣ motif " in yeast AC40 and AC19 may not be functionally equivalent (i.e. in enzyme assembly or activity). Other results have indicated that mutations in the "␣ motif " of the ␣ subunit of E. coli RNA polymerase and the yeast RNA polymerase II RPB3 subunit produce defects in enzyme assembly (18,49,50). Our results with the Arabidopsis B36b subunit suggest that the N-terminal "␣ motif " and a second ␣-like motif in the C terminus of this subunit may both be important for association with the B13.6 subunit. Similar to the N-terminal "␣ motif," the second C-terminal ␣-like motif appears to be conserved in the larger (i.e. 36 -44-kDa) and smaller (i.e, 12.5-19-kDa) RNA polymerase subunits related to the prokaryotic RNA polymerase ␣-subunit. It is possible that both ␣-like motifs (i.e. the "␣ motif " and the C-terminal leucine-rich motif) may contribute to subunit-subunit interactions and enzyme assembly. Recent results with the ␣ subunit in E. coli RNA polymerase indicate that both ␣ subunit motifs (shown in Fig. 9) that are conserved in eukaryotic ␣-like subunits are important for ␣ dimerization (51,52). It is interesting that the C288 truncation (see Figs. 1, 8, and 9) in the B36b subunit, which is the shortest truncation tested that resulted in a loss of association between the B36b andB13.6 subunits, is located within two amino acids (i.e. see the alignment of leucine-rich C-terminal ␣-like motifs in Fig. 9) of an insertion mutant that renders the E. coli ␣ subunit inactive in dimerization (52). Additional experiments will be required to confirm the importance and specificity (i.e. specificity of AC subunit interactions versus B subunit interactions, specificity of plant subunit interactions versus animal or yeast subunit interactions) of the ␣-like motifs in these subunit interactions.