The C53/C37 Subcomplex of RNA Polymerase III Lies Near the Active Site and Participates in Promoter Opening*

The C53 and C37 subunits of RNA polymerase III (pol III) form a subassembly that is required for efficient termination; pol III lacking this subcomplex displays increased processivity of RNA chain elongation. We show that the C53/C37 subcomplex additionally plays a role in formation of the initiation-ready open promoter complex similar to that of the Brf1 N-terminal zinc ribbon domain. In the absence of C53 and C37, the transcription bubble fails to stably propagate to and beyond the transcriptional start site even when the DNA template is supercoiled. The C53/C37 subcomplex also stimulates the formation of an artificially assembled elongation complex from its component DNA and RNA strands. Protein-RNA and protein-DNA photochemical cross-linking analysis places a segment of C53 close to the RNA 3′ end and transcribed DNA strand at the catalytic center of the pol III elongation complex. We discuss the implications of these findings for the mechanism of transcriptional termination by pol III and propose a structural as well as functional correspondence between the C53/C37 subcomplex and the RNA polymerase II initiation factor TFIIF.

The C53 and C37 subunits of RNA polymerase III (pol III) form a subassembly that is required for efficient termination; pol III lacking this subcomplex displays increased processivity of RNA chain elongation. We show that the C53/C37 subcomplex additionally plays a role in formation of the initiation-ready open promoter complex similar to that of the Brf1 N-terminal zinc ribbon domain. In the absence of C53 and C37, the transcription bubble fails to stably propagate to and beyond the transcriptional start site even when the DNA template is supercoiled. The C53/C37 subcomplex also stimulates the formation of an artificially assembled elongation complex from its component DNA and RNA strands. Protein-RNA and protein-DNA photochemical cross-linking analysis places a segment of C53 close to the RNA 3 end and transcribed DNA strand at the catalytic center of the pol III elongation complex. We discuss the implications of these findings for the mechanism of transcriptional termination by pol III and propose a structural as well as functional correspondence between the C53/C37 subcomplex and the RNA polymerase II initiation factor TFIIF.
RNA polymerase III (pol III) 3 transcribes genes encoding short RNA products that are essential for protein synthesis (e.g. tRNAs, 5 S rRNA), RNA processing (e.g. U6 small nuclear RNA, the RNA subunit of RNase P), protein transport (7SL RNA of the signal recognition particle), and diverse other functions, particularly in higher eukaryotes (1). Pol III is brought to the transcriptional start sites of its genes through interactions with its central 3-subunit initiation factor, TFIIIB (subunits Brf1, TBP, and Bdp1). TFIIIB, in turn, is assembled on DNA upstream of the transcriptional start site in a largely sequenceindependent process by its six-subunit assembly factor, TFIIIC, which binds to two promoter elements (boxA and boxB) located downstream of the transcriptional start site. An additional initiation factor, TFIIIA, serves solely as the adaptor for assembling TFIIIC on 5 S rRNA genes. Vertebrates contain an addi-tional form of TFIIIB in which its paralogue Brf2 replaces Brf1 for transcription of a class of genes that use the SNAP c complex in place of TFIIIC as their TFIIIB-assembly factor (2,3). Once it has been assembled onto DNA, Saccharomyces cerevisiae TFIIIB suffices for recruiting pol III for robust and accurately initiating transcription in vitro (4); at least in yeast, this also appears to be the case in vivo (5). The complexity of pol III contrasts with the apparent simplicity of its core transcription apparatus (the TFIIIB⅐DNA complex and diverse assembly factors). Five pol III subunits have no counterparts in the considerably less complex pol II (with its merely 12 subunits). Three of these pol III subunits (C82, C34, and C31) form a subassembly that interacts with the TFIIIB⅐DNA complex and is required for specifically initiating transcription (6). Two subunits, C53 and C37, form a subassembly that limits the processivity of pol III elongation and contributes to the ability of pol III to terminate transcription after the successive incorporation of five or more UMP residues. The association of C53/C37 with pol III is mediated at least in part through interaction with the Rpb9-and TFIIS-related subunit C11 (7,8).
The initial impetus for the analysis that follows was to better understand the function of the 110-residue C11 subunit of pol III. The intrinsic exoribonuclease (proofreading) activity of pol III (9) requires its C11 subunit (7). The amino acid sequences of the C11 N-terminal and C-terminal zinc binding domains are significantly similar to the counterpart segments of the pol II subunit Rpb9, but the C-terminal zinc binding domain of C11 is even more similar to domain III of TFIIS, the extrinsic exoribonuclease pol II elongation factor. TFIIS domain III contains a zinc binding domain that docks at the entrance to the pol II pore 1 and inserts a ␤-hairpin into pore 1. Two acidic residues at the tip of this ␤-hairpin contribute to coordination of a second metal ion for S n 2-type nucleophilic attack on the RNA phosphodiester bond at the pol II catalytic center (10). The ␤-hairpin and the two acidic residues at its tip are conserved in C11, and the acidic residues are essential for the intrinsic cleavage activity of pol III (8). Cryoelectron microscopy structural analysis of pol III (11) shows mass density compatible with the N-terminal zinc binding domain of C11, occupying a position on the C128 lobe domain of pol III that resembles the location of the N-terminal zinc binding domain of Rpb9 on Rpb2 in the structure of pol II (12). A mass corresponding to the C-terminal zinc binding domain of C11 was not observed in the pol III electron microscopic structure, but it has been suggested that this domain may bind to the pol III pore 1 (10). However, the pol I counterpart to C11, A12.2, which is also required for exonucleolytic retraction and also contains a TFIIS-like C-terminal zinc binding domain and ␤-hairpin with the same two acidic residues at its tip, nevertheless situates both its N-and C-terminal zinc binding domains in Rpb9-manner on the pol I lobe, prompting the suggestion that the pol I A12.2 subunit plays only an indirect role in exonucleolytic retraction (13).
We have approached the question of whether C11 promotes the exoribonucleolytic activity of pol III directly at the catalytic site (as TFIIS does) or indirectly from its lobe binding site (as does the A12.2 pol I subunit) by probing C11 proximity to the catalytic center of the pol III elongation complex using photochemical RNA-protein and DNA-protein cross-linking methods. The searched-for specific cross-linking to C11 could not be detected; we have instead found that a segment of C53 lies near the RNA 3Ј end of the pol III elongation complex. Further analysis indicates that the function of the C53/C37 subcomplex is not confined to transcriptional elongation but that it also facilitates the formation of the open promoter complex, suggesting a possible relationship to TFIIF (also originally identified as a pol II initiation factor and more recently understood to play an important role in transcript elongation).
ABdUMP-containing DNA for analysis of photochemical cross-linking was synthesized as described (14) with bottom strand DNA primer 3Ј ends at register ϩ1 and ϩ9 (supplemental Fig. S5A). 3Ј-Overhanging DNA templates for transcription were assembled by annealing individual DNA strands and purified by native PAGE, with 32 P-labeled markers for the separated strands and duplex DNA. The relatively small quantities of DNA produced in this way were quantified by fluorescence of bound PicoGreen relative to a DNA standard curve. The SUP4 probe for KMnO 4 footprinting was generated by PCR of pTA30 (15) with a 5Ј-32 P-labeled upstream primer.
Recombinant C53, C37, and C11 and yeast pol III⌬ were derived from a library of yeast strains 4 expressing multiply tagged pol III subunits, with a matching set of pET21-derived vectors for expression of each tagged subunit in Escherichia coli. Details of all constructs are available on request. Briefly, two forms of pol III⌬ were purified from S. cerevisiae expressing Schizosaccharomyces pombe C11 from a galactose-induc-ible plasmid promoter in place of the deleted chromosomal copy of S. cerevisiae RPC11 (encoding C11). In one strain the chromosomal RET1 gene (encoding C128) additionally encoded an N-His 6 -FLAG 3 tag, and the chromosomal RPO31 gene (encoding C160) was deleted and expressed as an N-HA 3tagged protein from a galactose-inducible plasmid promoter (pol III⌬N); in the other strain the chromosomal RET1 gene additionally contained N-His 6 and C-Myc 3 tags with the chromosomal RPO31 gene deleted and expressed from a galactoseinducible plasmid promoter as a C-HA 3 -tagged protein (pol III⌬C). S. cerevisiae strains were derived from the heterozygous diploid yeast knock-out collection (Open Biosystems; background BY4743) with the kanamycin resistance gene replacement cassette (KANMX) replaced by homologous recombination with hygromycin B resistance (HYGMX), nourseothricin resistance (NATMX), or URA3 (URA3MX) cassettes (17,18) as required; yeast genetic manipulations followed (19). The coding sequences of all of the genes encoding subunits unique to pol III were cloned as SacII-SmaI fragments into the pRS313-317 series of yeast centromeric plasmids (20) modified to contain the GAL1 promoter (21) and a NcoI-SacI-SmaI-XhoI cassette with epitope tags FLAG 3 , HA 3 , Myc 3 , or V5 2 or no tag inserted between NcoI and SacII or between SmaI and XhoI sites. Apart from these tags, construction added amino acid residues MVPR to the N termini and PG to the C termini of their respective pol III subunits.
Genes encoding pol III subunits were also cloned into pET21b with a related cassette (NdeI-NcoI-SacII-SmaI-XhoI) either with a His 6 tag or a His 6 -TEV protease cleavage tag inserted between NdeI and NcoI; the latter vector was used to generate N-His 6 -TEV-C11 and untagged C11 (upon TEV protease cleavage). C128 cloned into this pET21b-N-His 6 vector with either an N-FLAG 3 or C-Myc 3 tag was used as a PCR template for the addition of ϳ100 bp of N-and C-flanking sequence of RET1 for transformation and recombination of the PCR product into a yeast strain in which RET1 was replaced by URA3MX, and C128 was expressed as an N-FLAG 3 -tagged protein from the galactose-dependent promoter on pRS313(HIS3) described above (genotype: MATa his3⌬1 leu2⌬0 ura3⌬0 lys2⌬0 ret1::URA3MX, pRS313:N-FLAG 3 -C128). Recombinants were isolated based on 1) 5-fluorooratic acid counterselection of the URA3MX marker, 2) galactose-independent growth with glucose as carbon source, 3) restoration of the his Ϫ phenotype upon subsequent loss of the pRS313 vector, and 4) sequencing of chromosomally integrated FLAG-or Myctagged RET1. N-or C-HA-tagged RPO31 (encoding C160) was introduced on pRS313 in tandem with the PCR-derived rpo31::KANMX, with the flanking sequence from the C160 knock-out strain for recombinational replacement of RPO31. S. pombe RPC11 was introduced as an untagged (NcoI-XhoI) insert into the pRS316 (URA3) galactose-dependent promoter vector along with PCR-derived rpc11::NATMX from the C11 knock-out strain.
Pol III⌬N and pol III⌬C were purified from 80 g of cells that were lysed with a BeadBeater (Biospec Products) as a 33% (w/v) slurry in 40 mM NaHepes, pH 7.8, 7 mM MgCl 2 , 5% (v/v) glycerol, 0.1 mM EDTA, 10 mM 2-mercaptoethanol, 0.5 M NaCl, 1 mM benzamidine, 0.5 mM phenylmethylsulfonyl fluoride, 1 g/ml each of leupeptin, pepstatin, and aprotinin. Cell debris was removed by centrifugation at 25,000 ϫ g for 30 min, and the supernatant fluid was subjected to ultracentrifugation at 150,000 ϫ g for 90 min. The supernatant was adjusted to contain 20 mM imidazole, pH 7.0, loaded onto a HisTrapHP column (GE Healthcare) and washed successively with 5, 10, and 5 column volumes of the same buffer containing 20, 30, and 40 mM imidazole, respectively. Pol III eluted in the same buffer with 100 mM imidazole. The transcriptional properties of pol III⌬N and pol III⌬C were indistinguishable, and they were used interchangeably in this work. Quantities of pol III are approximately specified based on multiple-round transcription of the SUP4 tRNA gene in the presence of excess recombinant C53, C37, and C11 relative to known quantities of purified wild type pol III. Recombinant C53, C37, and C11 were produced in E. coli and purified on Ni-NTA-agarose as described elsewhere for Brf1 N -TBPc-Brf1 C (22). Each protein was further purified by size exclusion chromatography on Sephadex G75 in buffer CBϩ500 (described in the next paragraph). Quantities are specified as nM protein.
Assays-Elongation complexes were formed on 3Ј-overhang templates for 30 min at 20°C in 20 -40-l volumes of reaction medium containing 40 mM NaHepes, pH 7. Where indicated, 300 l of buffer CBϩ500 mM NaCl was loaded onto the column before sample application. Transcription reaction mixtures and column buffers for experiments not involving photoactive RNA or DNA contained Tris chloride, pH 8.0, in place of NaHepes and dithiothreitol in place of 2-mercaptoethanol. UV irradiation and nuclease treatment followed (14), except that 0.5 units/l micrococcal nuclease replaced S1 nuclease.
Exonuclease III footprinting was performed by adding an equal volume of buffer CB containing 2 units/l exonuclease III to reaction samples in buffer CBϩ100 containing 1 mM MgCl 2 .
Digestion was stopped after 10 min at 25°C, and samples were processed for gel electrophoresis as described (23). Factor-dependent preinitiation complex formation and transcription was performed in the above reaction buffer containing 100 g/ml bovine serum albumin (with 5 g/ml poly(dG-dC):poly(dG-dC) added for the experiments shown in Fig. 6 and supplemental was added to 20 mM for 30s, the reaction was stopped, and samples were processed as described (25). Multiple-round transcription was allowed to proceed for 30 min by the addition of ATP, CTP, and GTP to 200 M and [␣-32 P]UTP (15 cpm/ fmol) to 25 M. Reactions were stopped, and samples were processed as described (23).
Artificially assembled elongation complexes ( Fig. 5) were formed in 20 l of reaction buffer containing 40 mM NaHepes, pH 7.8, 3 mM 2-mercaptoethanol, 1 mM MgCl 2 , 0.5units/l ribonuclease inhibitor, with 50 mM (panel B) or 70 mM (panel C) NaCl. For the RNA-protein cross-linking analysis of elongation complexes in panel C, 6 pmol each of the 10-mer RNA and transcribed DNA strand were annealed, and 5 fmol of pol III⌬ was added for 10 min at 25°C followed by 12 pmol of nontranscribed DNA strand and incubation was continued for 10 min at 30°C. The same procedure was followed for analysis of transcript elongation in panel B with 1.2 fmol of pol III⌬ and 0.9 pmol of N-His 6 -tagged C53 and C37 and with quantities of RNA and DNA strands specified (in pmol) above each lane. pmol each of N-His 6 -tagged C53 and C37 was added for an additional 5 min at 30°C before UV irradiation and chromatography on Sephacryl S300 as described for Fig. 2B; successively eluting 60-l fractions from the 800-l volume column were analyzed (as indicated at the top of Fig. 5C).

RESULTS
We set out to determine directly whether the C11 pol III subunit, like TFIIS, inserts into the pol III pore 1 (10) and into the catalytic center by seeing whether it was able to form a cross-link to a photoactive nucleotide (4-S-UMP) inserted at or near the 3Ј end of the nascent transcript. This approach has Dual Roles of the Pol III C53/C37 Subcomplex JANUARY 22, 2010 • VOLUME 285 • NUMBER 4

JOURNAL OF BIOLOGICAL CHEMISTRY 2697
been successful for detecting TFIIS at the catalytic center of pol II, albeit at low cross-linking efficiency (1-2% relative to the Rpb1 and Rpb2 pol II subunits) (26). We found that 4-S-UTP was inefficiently incorporated by wild type pol III; weak 4-S-UMP-dependent cross-linking to the C160 and C128 subunits of pol III and to a protein approximately the size of the C53 subunit (ϳ50 kDa) was detected (data not shown). We reasoned that poor incorporation of 4-S-UMP was due to C11mediated proofreading (27) responding to the anomalous base pairing of 4-S-U. Loss of cross-linking had also been observed when both TFIIS and Mg 2ϩ were added to pol II elongation complexes containing 4-S-U near the nascent RNA 3Ј end (26).
In response to these findings, Pol III was purified from S. cerevisiae expressing S. pombe C11 in place of S. cerevisiae C11. This substitution leads to the loss of the C11, C53, and C37 pol III subunits during purification. The enzyme lacking these three subunits, pol III⌬, is defective in transcriptional termination and exoribonucleolytic proofreading. E. coli-expressed C11 restores the intrinsic RNA cleavage activity to pol III⌬, and the recombinant C53/C37 subcomplex restores efficient termination (8). The exonucleolytic retraction and termination deficits of a pol III⌬ preparation and the ability of our recombinant C11, C53, and C37 to restore these functions are documented in supplemental Fig. S1.
The DNA initially chosen for incorporation of a photoactive UMP analogue into the nascent transcript (Fig. 1A) contains a 4-nt 3Ј-overhang at one end for factor-independent transcription (28) initiating with the dinucleotide primer GpA followed in turn by a run of 15 AMP residues, CMP ( 32 P-labeled), and the photoactive UMP analogue. We expected the initial run of weakly base-pairing A residues to result in reiterative slippage during repetitive addition of AMP to the initially synthesized RNA, generating a cluster of transcripts differing in length but with identical 3Ј ends (as previously encountered with E. coli RNA polymerase and also seen to occur with pol III (29,30)). The array of nascent transcripts contained within elongation complexes formed in the absence of UTP is shown in Fig. 1B, lane 1. That their size heterogeneity resulted from slippage was confirmed by walking the entire cluster of bands upon the addition of UTP (lane 2) or UTPϩATP (lane 3; note that most of the complexes in lane 3 extended to C27 instead of A26 due to misincorporation of UMP and the lack of proofreading by pol III⌬ (27) and that ϳ20% of complexes did not chase).
Transcription initiating at 3Ј-overhanging DNA ends can generate RNA:DNA hybrid products (31). This was not observed with the template shown in Fig. 1A; the 40-nt RNA of a 3Ј-OMeGMP-arrested pol III elongation complex (Fig. 1C) was resistant to digestion by RNase H and sensitive to degradation by RNase A, whereas the same elongation complex formed with only the template DNA strand (generating a DNA:RNA hybrid duplex product) was RNase H-sensitive and RNase A-resistant (supplemental Fig. S2).
Unlike 4-S-UMP (or 5-I-UMP), the photoactive analogue ABUMP generated efficient protein cross-linking to C160 and C128 when incorporated near (4 nt upstream of) the nascent RNA 3Ј end (data not shown). However, incorporation of tandem ABUMP residues was inefficient (data not shown) and prevented further incorporation (Fig. 1C, compare lanes 2 and   3). A trace level of the 3Ј-OMeG40-arrested transcript was also observed in the absence of UTP (lane 1) and is interpreted as due either to misincorporation of CMP in place of UMP by error-prone pol III⌬ or to the presence of trace UTP (the natural deamination product of CTP) in the [␣-32 P]CTP. Incorporation of tandem 5-p-azidophenacylthio-CMP or -UMP residues has likewise been shown to be inefficient and to prevent further transcript elongation by E. coli RNA polymerase (32).
We initially examined whether one could detect cross-linking of C53, C37, or C11 to ABUMP incorporated at the RNA 3Ј end of the pol III⌬ elongation complex. Pol III⌬ transcription complexes were chromatographed on Sephacryl S300 to remove unincorporated labeled NTPs and released RNA products (schematically diagrammed in supplemental Fig. S3); when C53 and C37 were added before UV irradiation, a cross-linked (i.e. radioactively tagged) ϳ60-kDa protein appeared in addition to the C160 and C128 subunits of pol III (Fig. 2A, lane 2). This ϳ60-kDa protein was positively identified as C53 by substituting a larger FLAG epitope-tagged version (lane 3). Little or no effect on C53 cross-linking was observed when pol III⌬ elongation complexes were chromatographed on Sephacryl S300 through a zone of 0.5 M NaCl to reduce the potential for nonspecific re-binding of released RNA to pol III during chromatography (supplemental Fig. S3). The addition of C53 and C37 to the pol III⌬ elongation complex and UV irradiation before Sephacryl S300 chromatography demonstrated that the crosslinked C53 is associated with the elongation complex (Fig. 2, panels B and C) as evidenced by the nearly perfect co-elution of cross-linked C160, C128, and C53 (panel B), well removed from excess unbound C53/C37 subcomplex and the DNA template (the profile for C37 is omitted for clarity in panel C). The C11 subunit was also cross-linked to RNA in the Sephacryl S300purified elongation complex ( Fig. 2A, lane 5). However, whereas C53 cross-linking to more S300-included fractions progressively diminished, C11 cross-linking increased (data not shown). Moreover, an experiment (performed in parallel with the analysis in panel B) in which C11, C53, and C37 were added to the elongation complex before UV irradiation and Sephacryl S300 chromatography showed no evidence of cross-linked C11 co-eluting with pol III⌬ (data not shown).
These observations suggested the possibility that the observed C11 cross-linking might merely reflect its direct binding to low levels of released RNA and be unrelated to the structure of the pol III elongation complex. That C53 and C11 bind and cross-link directly to purified ABU-containing RNA is shown in Fig. 3. Both C53 and C11 cross-linked to free RNA (first lane) along with a protein corresponding in size with C37. Single-  Fig. 1A were arrested by incorporation of adjacent ABUMP residues and isolated by chromatography on Sephacryl S300. The peak fraction containing the nearly excluded pol III⌬ elongation complex was distributed into tubes providing 100 nM each of untagged C11 and N-His 6 -C37 (lanes 2 and 3), N-His 6 -C53 (lane 2), N-His 6 -FLAG 3 -C53 (lane 3), or 2.5 M N-His 6 -TEV-C11(lane 5). After a 10-min incubation, samples were UV-irradiated and processed for SDS-PAGE on 8% (lanes 1-3) or 13% (lanes 4 -5) gels as described in under "Experimental Procedures." C160, C128, C53, and C11 cross-linked products are identified at the right side. The weaker unidentified labeled bands are not pol III subunits. B, C53 cross-links to the nascent transcript of the pol III elongation complex. ABUMP-arrested pol III⌬ elongation complexes were formed as in A. N-His 6 -tagged C53 and C37 were added to 60 nM each for 10 min followed by UV irradiation and Sephacryl S300 chromatography; eluting fractions 3-12 are shown. Crosslinked C160, C128, and C53 are identified at the left side. C, the phosphor image intensity of cross-linked C160 ϩ C128 (Ⅺ) and C53 (F) is shown normalized to the peak fraction set to 1.0. The undigested nascent transcript also perfectly co-eluted with C160 and C128 (data not shown). A separate mock reaction containing 50 pmol of N-His 6 -tagged C53 (-), C37 (not shown), and 32 P-labeled DNA (⅐⅐⅐⅐) was chromatographed under identical conditions and quantified by Coomassie staining and scintillation counting, respectively, with the peak fraction set to 1.0. Pol III⌬ elongation complexes were formed with the 3Ј-overhang template shown in supplemental Fig. S4 and purified on Sephacryl S300 as specified for Fig. 2. The contained ABUMP-tagged 32 P-labeled RNA (together with unlabeled DNA template) was isolated and aliquots were distributed into buffer CBϩ100. Non-transcribed DNA template strand or yeast tRNA were added before (for concurrent competition) or 10 min after (for post-competition) the addition of N-His 6 -tagged C53 and C37 to 50 nM (upper panel) or N-His 6 -TEV-C11 to 2.5 M (lower panel), as indicated in the figure. Incubation at 20°C continued for 10 min before UV irradiation and sample processing. The crosslinking efficiency of C53 and C11 relative to the no-competitor control (set to 100) is quantified below each panel. C53, C11, and a protein of size comparable with C37 are identified at the left side. ssDNA, single-stranded DNA.
stranded DNA with the same sequence was ineffectual as a competitor for binding C53 (at 100-fold molar excess over C53) when added with the labeled RNA (second lane), whereas tRNA competed effectively in both simultaneous (third lane) and sequential (fifth lane) addition formats. Poly(A) RNA was an effective competitor also (data not shown), indicating that C53 preferentially binds to RNA. Unlabeled tRNA (third and fifth lanes) and single-stranded DNA (second lane) were effective competitors for UV-mediated cross-linking of labeled RNA to C11. TFIIS likewise contains a DNA and RNA binding domain, but it is masked in the full-length protein (33).
UV-cross-linking of C53 and C11 to free RNA warranted further examination of the specificity of RNA cross-linking to C53 in the transcription complex. We considered two ways in which nonspecific binding of C53 to RNA might generate the appearance of cross-linking to the RNA 3Ј end within the elongation complex (Fig. 2 and supplemental Fig. S3). The first alternative is that co-chromatography of cross-linked C53 with pol III⌬ in Fig. 2B could result from stable binding of the UVcross-linked, nonspecific RNA⅐C53 complex to pol III. This was assessed directly by adding RNA-cross-linked C53 to pol III before Sephacryl S300 chromatography (supplemental Fig.  S4A). Although some cross-linked C53 did co-elute with pol III, most of it emerged from the column in later (included) fractions. These findings strongly argued against cross-linked C53 in the Fig. 2 experiment representing adventitiously bound RNA adduct, as the latter would have had to vastly out-compete RNA-free C53/C37 subcomplex for binding to limiting pol III⌬.
We also considered the possibility that the photoactive nascent RNA residue responsible for C53 cross-linking in the more stringently purified elongation complex (supplemental Fig. S3) is not located at the catalytic center but in the RNA exit channel of a minor fraction of elongation complexes reading through and extending to G40 (Fig. 1C, lanes 1 and 2). We addressed this possibility by modifying the transcription unit shown in Fig. 1A, replacing AAA at bp 24 -26 with TGG (supplemental Fig. S4A). Transcription analysis indicated that in the presence of ABUTP few if any complexes extended an RNA chain to bp 24 and that incorporation of 3Ј-OMeGMP at position 25 insured against further elongation (data not shown). The addition of the C53/ C37 subcomplex to the stringently purified elongation complex formed with this modified template again generated cross-linking of C53 to ABUMP incorporated at the RNA 3Ј end (supplemental Fig. S4B).
C53 was previously shown to cross-link to 5-ABdU placed in the non-transcribed (non-template) DNA strand downstream of the open complex transcription bubble (at bp ϩ11, ϩ16, or ϩ17 relative to the start site of transcription as bp ϩ1 (34, 35)), but cross-linking of C53 to the non-transcribed strand was not detected in the elongation complex at register ϩ13, ϩ3, Ϫ1, and Ϫ6 (relative to the RNA 3Ј end as register ϩ1). In contrast, an ϳ40-kDa pol III subunit cross-linked to the non-transcribed strand only in the elongation complex (to ABdU at register Ϫ6 and to 4-S-dT at register Ϫ2/Ϫ3 (34, 36)). We can now identify this cross-linked subunit as C37, as the alternative AC40 subunit, a homologue of pol II subunit Rpb3, is predicted to be far removed from the DNA binding cleft (11,37).
Because C53 appeared to cross-link to RNA near the catalytic site and because cross-linking to the transcribed DNA strand had not been examined previously, we probed C53 cross-linking to ABdU in the transcribed DNA strand near the catalytic center of our elongation complex. The 3Ј-overhang DNA template shown in supplemental Fig. S3 was modified to allow incorporation of ABdU at register Ϫ1/Ϫ2 or ϩ8/ϩ9 (supplemental Fig. S5A). C53/C37-supplemented Sephacryl S300-purified elongation complexes showed C53 cross-linking at register ϩ8/ϩ9 and Ϫ1/Ϫ2 and C37 cross-linking at register ϩ8/ϩ9 (supplemental Fig. S5B). Analysis of protein-DNA cross-linking of control samples in which ATP was omitted from the NTP mixture used to form elongation complexes (data not shown) indicated that the small fraction of transcription complexes that failed to initiate (as evidenced by electrophoretic mobility shift analysis) might contribute to C53 cross-linking to the Ϫ1/Ϫ2 DNA probe. Modest levels of poly(dI-dC):poly(dI-dC) were found to prevent nonspecific binding and cross-linking of pol III and C53 to the Ϫ1/Ϫ2 probe (supplemental Fig. S5C). Accordingly, for the experiment shown in Fig. 4, elongation (ϩATP) and control nonspecific (no ATP) complexes were formed with pol III⌬ and DNA probe Ϫ1/Ϫ2 followed by treatment with 0.5 M NaCl and poly(dI-dC) before chromatography on Sephacryl S300. Individual fractions were supplemented with additional poly(dI-dC) before the addition of C53/C37 and UV irradiation. C53 was found to cross-link to DNA in fractions containing the elongation complex and in constant proportion to C160 as well as C128 in the trailing excluded peak fractions (fractions 3-5; left panel) but not to the bulk of the DNA probe, which eluted with the more included fractions 7 and 8. This cross-linking required the formation of the stable elongation complex as only trace levels of cross-linking to C160, C128, and C53 were discernible in the corresponding excluded fractions . C53 cross-links to the transcribed DNA strand of the pol III⌬ elongation complex at register ؊1/؊2. Elongation complexes (ϩATP) or mock complexes (no ATP) were formed with the Ϫ1/Ϫ2 photoactive (ABdUMP) and radioactive DNA probe as described under "Experimental Procedures." Poly(dI-dC):poly(dI-dC) (to 40 ng/l) and NaCl (to 0.5 M) were added before chromatography on Sephacryl S300 equilibrated with buffer CBϩ100 supplemented with 7 mM MgCl 2 , with an upper zone of the same buffer containing 0.5 M NaCl. 30-l volume fractions were taken; to each, poly(dI-dC): poly(dI-dC) was added to 20 ng/l followed by 30 nM N-His 6 -C53, 30 nM N-His 6 -C37, and 30 nM N-His 6 -TEV-C11. After incubation for 10 min, samples were UV-irradiated and prepared for analysis, as described under "Experimental Procedures." DNA-cross-linked C160, C128, and C53 are identified at the left side. The asterisks at the right identify cross-linked proteins that are not pol III subunits.
in the absence of productive initiation (the no-ATP control; right panel).
We noted that, whereas cross-linking of C160 and C128 to the Ϫ1/Ϫ2 probe was maintained when the polyanion heparin was added to the elongation complex (cf. Ref. 34), C53 crosslinking was heparin-sensitive. DNA-cross-linking of the C53 subunit of elongation complexes formed with highly purified wild type pol III was also found to be heparin-sensitive (data not shown). The implications of this heparin sensitivity are discussed below.
We next examined the possibility that cross-linking of C53 to DNA and RNA near the catalytic center of the elongation complex might be due to backward translocation of the transcription bubble (backtracking (38)), which shifts the DNA:RNA duplex upstream and extrudes the melted RNA 3Ј end toward or into pore 1, with register ϩ1 of the transcribed strand concomitantly reassociating with the non-transcribed strand. We used exonuclease III (3Ј 3 5Ј) digestion from the downstream DNA end of the elongation complex to locate the downstreamfacing edge of pol III (F) relative to the RNA 3Ј end (C, the catalytic center). For E. coli RNA polymerase, the F Ϫ C separation for non-backtracked elongation complexes is 18 bp, and F Ϫ C Յ 16 bp indicates backtracking (38). We found that the pol III elongation complex formed on a template nearly identical to that used for RNA cross-linking (supplemental Fig. S4A) primarily generated F Ϫ C distances of 18 and 19 bp with weaker exo III arrest sites corresponding to F Ϫ C distances of 20 and 17 bp (supplemental Fig. S6). A background of incomplete exo III digestion products in the absence of pol III limited the ability to exclude the existence of a small fraction of extensively backtracked pol III complexes. We judge that the 18 -19-bp F Ϫ C separation reflects the non-backtracked state of the pol III elongation complex. This is consistent with preferential cross-linking of C128 relative to C160 to ABdUMP incorporated at register Ϫ1/Ϫ2 of the transcribed strand (Fig.  4) as well as the relative C128-C160 cross-linking efficiency for ABUMP incorporated at the RNA 3Ј end (supplemental Fig.  S3), which matches the Rpb2-Rpb1 cross-linking efficiency ratio of the pol II elongation complex with ABdUMP incorporated in the transcribed strand at register Ϫ2 and is readily rationalized on the basis of the structure of the pol II elongation complex (39). Extensive (i.e. 6 -10 bp) backtracking should generate preferential cross-linking to C160 (Refs. 34 and 39; see also probe ϩ8/ϩ9 in supplemental Fig. S5). Thus, in sum, the preceding experiments strongly support the conclusion that C53 is located within cross-linking distance of the 3Ј end of the nascent transcript and the transcribed DNA strand at and near the catalytic center.
The other path taken was to limit the extent of, and potential for, backtracking by forming a pol III elongation complex from its component parts using the assembly procedure of Kireeva et al. (40): 10-nt RNA was annealed to its 51-nt transcribed DNA strand, incubated with pol III⌬, and assembly was completed by the addition of the non-transcribed DNA strand (Fig. 5A). The addition of [␣-32 P]CTP, UTP, and 3Ј-OMeGTP generated labeled 16-mer RNA (Fig. 5B, lane 1). Shorter products appeared to result from misincorporation of UTP in place of CTP, as they did not appear when UTP was absent (forming the 13-mer instead, lane 9) or when the 16-mer was formed by the sequential addition of CTP and UTP (data not shown; note that the addition to misincorporated nucleotides is extremely slow (27)). Curiously, while optimizing the conditions for forming the 16-mer RNA-containing elongation complex (by varying the proportions of the two DNA strands and RNA; compare lane 1 with lanes 2-4), we observed that formation of the 16-mer-RNA elongation complex required the assembly of the non-transcribed DNA strand (although some extension to nt 12 did occur in its absence; data not shown). This may have reflected a role of flanking duplex DNA in stabilizing the initial elongation complex, previously observed for E. coli RNA polymerase (41) but not for pol II (40). Inclusion of the C53/C37 subcomplex in these assemblies consistently stimulated the for-  lanes 1-8); the relative yields of these G16 and C13 products are specified below each lane (the line between lanes 4 and 5 designates removal of intervening lanes). C, elongation complexes were assembled and extended to G16, as above, C53 and C37 was added, and samples were UV-irradiated and chromatographed on Sephacryl S300 as described under "Experimental Procedures." Eluting 60-l-volume fractions 3-7 comprise the excluded and more included parts of the elution profile. RNA-cross-linked C160, C128, and C53 are identified at the left side. The asterisk at the right identifies a cross-linked protein that is not a pol III subunit.

Dual Roles of the Pol III C53/C37 Subcomplex
mation of the 16-mer elongation complex 2-3-fold (Fig. 5B,  compare lanes 5-8 with lanes 1-4). Both subunits were required for the stimulation (data not shown), which could not be duplicated by increasing concentrations of the RNA primer or individual DNA strands (lanes 2-4).
The rationale for examining C53-RNA cross-linking with this artificially assembled elongation complex is based on the assumption that backtracking of this complex with its short "transcript" (15-mer RNA with ABU at its 3Ј end; Fig. 5, panel A) more than 7 bp would shorten the RNA:DNA hybrid at its upstream end and either block further backtracking or lead to RNA release as a consequence of invasion of the upstream end of the DNA:RNA duplex by the non-transcribed DNA strand (40). Although the 13-mer elongation complex was at least transiently resistant to exposure to 0.5 M NaCl (as evidenced by its ability to be subsequently walked to G16), a significant amount (ϳ 1 ⁄ 2) of the elongation complex dissociated during Sephacryl S300 chromatography (data not shown). Two approaches were taken to eliminate the possible complication of C53 cross-linking to released RNA; 1) C53 addition and UV cross-linking preceded S300 chromatography, and 2) ABU was incorporated subsequent to isolation of the 13-mer RNA elongation complex on Sephacryl S300. The outcome in both cases was similar and is shown for the first method in Fig. 5, panel C. C53 was found to cross-link to ABU at the RNA 3Ј end and to co-chromatograph with pol III. The cross-linking efficiency of C53 relative to that of C160 and C128 using either of these approaches was the same (ϳ2%) as that obtained with a 3Ј-overhang template when UV irradiation preceded Sephacryl S300 chromatography (Fig. 2B) but was less than the efficiency observed with 3Ј-overhang templates when ABUMP was incorporated before Sephacryl S300 chromatography and subsequent UV irradiation (e.g. supplemental Fig. S4B; ϳ5% C53 cross-linking efficiency relative to C128). In summary, exonuclease III footprinting did not indicate significant backtracking. An artificially assembled elongation complex that should place limits on the extent of backtracking also did not eliminate C53 cross-linking to the 3Ј end of the nascent transcript. These findings argue strongly against the supposition that C53 proximity to the nascent RNA 3Ј end requires and entirely reflects extensive backtracking of the pol III elongation complex. However, we consistently observed increased C53 cross-linking when NTP substrates were removed before UV irradiation. This observation is compatible with the notion that backtracking (perhaps occurring in a time-dependent process in the absence of NTPs) provides additional sites of C53 proximity to the RNA 3Ј end.
The ability of C53 and C37 to stimulate formation of the elongation complex (Fig. 5) prompted an examination of a possible role for these proteins in initiation factor-dependent formation of open promoter complexes. For this purpose we chose the well studied TFIIIB-and TFIIIC-requiring SUP4 tRNA gene and KMnO 4 footprinting (Fig. 6), which reveals strandseparated T residues in the open promoter. The addition of wild type pol III to the TFIIIB-TFIIIC-SUP4 DNA complex generated reactivity of non-transcribed strand T residues between bp Ϫ2 and Ϫ9 (relative to the start site as ϩ1; lane 1) as previously shown (25). At the temperature of this experiment (25°C), the pol III promoter complex is in dynamic balance between open, partially, and fully closed states (25). The addition of ATP, CTP, UTP, and 3Ј-OMeGTP to allow formation of an arrested elongation complex at bp ϩ18 shifted the location of reactive T residues to bp ϩ6 to ϩ16 (lane 7); residual KMnO 4 reactivity between bp Ϫ9 and Ϫ2 reflected a minor fraction of elongation-incompetent pol III (36). For initiation complexes assembled with pol III⌬, promoter opening was barely detectable above background (lane 3; i.e. shifting the steady state levels of open and closed complexes toward the closed complex), although the formation of arrested elongation complexes was readily apparent (lane 8). The addition of C53, C37, and C11 substantially increased promoter opening (lane 5) and formation of arrested elongation complexes (lane 10), whereas C53 and C37 alone were almost (Ն60%) as effectively stimulatory (lanes 4 and 9). These results clearly indicate that the presence of C53 and C37 facilitates promoter opening.
TFIIIB recruits pol III to the start site of transcription; its Brf1 and Bdp1 subunits also participate in subsequent steps of forming the initiation-ready open promoter complex; Bdp1 in nucleating promoter opening at the upstream end of the transcription bubble and Brf1 in propagating DNA strand separation to melt the start site (16,42,43). Brf1 lacking its N-terminal zinc ribbon motif (N⌬68) or Bdp1 with small internal deletions covering a 67-amino acid segment (e.g. ⌬355-372) generate TFIIIB-pol III-DNA complexes that fail to open on relaxed (linear) DNA templates. Combining both mutant subunits in TFIIIB leads to a loss of ability to transcribe supercoiled DNA (16). We examined whether pol III and pol III⌬ would respond differently to these TFIIIB defects in vitro; that is, whether synthetic phenotypes would materialize in combinations of pol III⌬ and defective TFIIIB. For the analysis, we used plasmid

Dual Roles of the Pol III C53/C37 Subcomplex
(supercoiled) DNA containing a variant of the U6 small nuclear RNA gene (SNR6), which allows TFIIIC-independent, ambidirectional binding of TFIIIB to its strong and nearly symmetric TATA box, and monitored the resulting divergent transcription (U6L and U6R in Fig. 7; schematically diagrammed in supplemental Fig. S7). Lanes 1-4 of Fig. 7A recapitulate the prop-erties of wild type pol III (16), including the substantial transcriptional activity of TFIIIB assembled with the individual mutant Brf1 and Bdp1 subunits and inactivity of TFIIIB containing both mutant subunits. The U6R and U6L transcription units contain T 10 terminators, but the termination defect of pol III⌬ is clearly evident by the shift to the start site-distal end of the T 10 stretch (Fig. 7A, compare lanes 5 and 6 with lanes 1 and  2). The effect of the Bdp1 and Brf1 mutations on transcription with wild type pol III and pol III⌬ is quantified (relative to a sample recovery marker, not shown) and normalized to transcription by each polymerase with wild type TFIIIB in panel B. Although no synthetic effect of these mutations was seen for the pol III⌬/Bdp1(⌬355-372) combination (panel B, compare lanes 2 and 6), the combination of pol III⌬ with TFIIIB containing Brf1(N⌬68) yielded no U6L transcription and only trace levels of U6R transcripts (compare lanes 3 and 7). The addition of C53, C37, and C11 substantially restored transcription (com- pare lanes 8 and 3), whereas C53 and C37 alone partially did so (compare lane 9 with lane 7). The addition of C53, C37, and C11 did not stimulate pol III⌬ transcription when assembled with wild type TFIIIB (data not shown). C11 does not appear to function by stabilizing the interaction of C53/C37 with pol III⌬, as an ϳ4-fold increase in C53 and C37 concentration did not further restore transcription with TFIIIB containing Brf1 N⌬68 (supplemental Fig. S7). No restoration of transcription was achieved by C11 alone (supplemental Fig. S7) nor by combinations of C53 or C37 with C11 (data not shown).
These results indicate that the C53/C37 subcomplex is required for downstream propagation of the transcription bubble during formation of the open promoter complex on supercoiled DNA templates in conjunction with TFIIIB lacking the Brf1 zinc ribbon region. The C11 subunit also appears to contribute in this process.

DISCUSSION
A Functional Relationship of the C53/C37 Subcomplex to TFIIF-On the basis of limited sequence similarity and predicted secondary structure, it has been suggested that the C53/ C37 subcomplex may be related to the pol II elongation factor TFIIF; the specifics of the C53/C37:TFIIF relationship was not provided (13,44), but a potential similarity between the N-terminal dimerization domain (residues 37-118) of the RAP30 subunit of human TFIIF (corresponding to yeast Tfg2) and the C-terminal end of yeast C53 (residues 358 -421) is discernible (HHpred p value 1.5 ϫ 10 Ϫ4 (45)). A stronger case for a relationship of the S. cerevisiae pol I A49/A34.5 subcomplex to TFIIF has also been made (13,44). Although both TFIIF and the C53/C37 subcomplex are involved in RNA chain elongation, they exert opposite effects, with TFIIF increasing (46), whereas C53/C37 decreases (8) processivity.
The principal findings of this work, that the C53/C37 subcomplex plays a role in formation of the open promoter complex and that segments of both C53 and C37 lie within the pol III DNA-binding cleft (with C53 and C37, respectively, proximal to the transcribed and non-transcribed strands of the transcription bubble), are reminiscent of the initiation role of TFIIF as follows. 1) S. cerevisiae TFIIF subunits Tfg1 and Tfg2 (corresponding with human RAP74 and RAP30, respectively) are FIGURE 7. Pol III⌬ places additional demands on TFIIIB for initiation of transcription. A, the TATA box of the SNR6 (U6 small nuclear RNA)-derived plasmid pLY1855 permits ambidirectional binding of TFIIIB and generates divergent transcription (see supplemental Fig. S7). Transcription of supercoiled plasmid DNA directed by TFIIIB assembled with wild type (wt, reference type) and mutant Brf1 and Bdp1 is shown. The Brf1 mutation deleting its N-terminal zinc binding domain (residues 1-68; N⌬) was introduced into the background of Brf1(⌬366 -408) (rt); removing the latter non-conserved 43-residue segment increases the activity of Brf1 in vitro. The Bdp1 mutation removing residues 355-372 (int⌬) was introduced into the wild type background. TFIIIB-promoter complexes were formed before the addition of wild type pol III or pol III⌬, with or without 25 nM C53, C37, and C11 followed in turn by the addition of NTPs for 30 min of multiple-round transcription. Leftward (U6L) and rightward (U6R) transcripts are identified at the left side. Asterisks between lanes 6 and 7 point to the increased length of pol III⌬ transcripts reflecting defective termination that is corrected by C53 and C37 (lanes 8 and 9). TFIIIB assembled with N⌬68Brf1 is essentially inactive for transcription by pol III⌬ (lane 7). B, shown is the quantitative presentation of panel A (with U6L and U6R transcripts in lanes 2 and 3 separately normalized to their counterparts in lane 1 and transcripts in lanes 6 -9 separately normalized to lane 5) showing partial restoration of activity to pol III⌬ by C53 and C37, further improved by C11 (less than half of the sample analyzed in panel A, lane 1, was recovered, but this has been corrected relative to its recovery marker in panel B).
located near the catalytic center of the pol II initiation complex (47)(48)(49)(50). 2) The N-terminal zinc ribbon regions of Brf1, TFIIB, and archaeal TFB probably plays similar post-polymerase recruitment roles in stabilizing the open complex and the initially forming short RNA:DNA hybrid through interaction with their respective B finger motifs (16,47,(51)(52)(53). Just as the C53/ C37 subcomplex is required for transcription of a supercoiled DNA template when the zinc ribbon region of Brf1 is removed (Fig. 7), TFIIF can compensate for TFIIB B-finger mutations that reduce or eliminate transcription in a minimal pol II transcription system consisting of TBP, TFIIB, pol II, and a supercoiled DNA template (54). 3) The C53/C37 subcomplex stimulates the artificial assembly of a pol III elongation complex from its RNA and DNA components (Fig. 5), and TFIIF exerts a comparable effect in stabilizing the incorporation of a 5-nt RNA into a pol II heteroduplex DNA complex containing a 10-bp bubble (55). In addition to these functional similarities, TFIIF and the C53/C37 subcomplex appear to attach to similar sites on the second largest subunits of their respective RNA polymerases, the Rpb2/C128 lobe domains above the DNA binding cleft (11,48), and the stability of these interactions is also dependent on attachment of the respective paralogue Rpb9/C11 subunits to the lobe domain (7,8,56).
The proximity of C53 to the pol III catalytic center suggests that the C53/C37 subcomplex may participate directly in open complex formation and also in potentiating transcriptional pausing and termination. With respect to open complex formation, C53 may interact directly with the transcribed DNA strand, to which it cross-links in the elongation complex (Fig.  4), and thus, stabilize localized DNA strand separation. It has been suggested that stimulation of pausing is coupled to termination because increasing the dwell time of the elongation complex at the terminator increases the probability that the elongation complex will disassemble (8). The simple recognition element for terminating pol III transcription consists of a run of five or more adjacent T residues in the non-transcribed strand (57,58), whereas the strengths of intrinsic pause sites for pol III strongly correlate with increasing numbers of U residues at the nascent transcript 3Ј end (59). Sites of pausing and termination are also sites of enhanced C11 subunit-mediated exoribonucleolytic cleavage (60); both pausing and cleavage are likely to arise from fraying of the RNA 3Ј end due to the instability of rU:dA base pairs. We suggest that the nonspecific RNA binding activity of C53 (Fig. 3) may contribute to pausing and termination by transiently stabilizing the frayed state of RNA 3Ј ends and that longer runs of U at the RNA 3Ј end might allow progressively more extensive C53⅐RNA binding, ultimately contributing to termination by disruption of the RNA:DNA hybrid.
The C53/C37 subcomplex may also indirectly contribute to pausing and termination by altering the conformational flexibility of other segments within the pol III catalytic center. In this regard, it is appropriate to recall the isolation of 24 singlesite mutations in RET1 (encoding C128) that are moderately or strongly defective in termination and the demonstration that some of these mutations cause the corresponding pol IIIs to elongate transcripts more rapidly (61,62), apparently comparable with pol III lacking C53 and C37. Based on the pol II homology-derived structural model of pol III (37), all of the termination-deficient mutations of Shaaban et al. (62) can be mapped to the lobe and fork domains of C128. Eight mutations in the lobe are clustered in the N-terminal half of helix ␣9 and the ␣8-␣9 loop (corresponding to Rpb2 residues 338 -354). This C128 segment lies proximal to (within ϳ25 Å) of the N-terminal zinc ribbon domain of C11 but inclined toward the DNA binding cleft and may constitute a C53/C37 binding site. More than one-half of the fork-domain mutations cluster within the 22-residue segment of C128 that contains fork loop 2 at its center (corresponding to Rpb2 residues 496 -517). In the structure of pol II, the mobile fork loop 2 (Rpb2 residues 501-511) is positioned to define the downstream edge of the transcription bubble in the elongation complex by blocking access of the non-transcribed DNA strand at register ϩ3 to the catalytic site as duplex DNA (63,64). Fork loop 2 moves in concert with DNA at register ϩ3 and maintains its interaction with it, suggesting a role in the unwinding of downstream DNA (65). It has also been proposed that this downstream DNA interaction may contribute to the stability of the paused elongation complex (66).
It has recently been proposed that Rpb9 prevents misincorporation of noncognate NTPs from its lobe-surface location through a direct interaction with the Rpb1 trigger loop that delays closure of the latter (the trigger loop), which in turn affects the sequestration and proper positioning of the incoming NTP for catalysis (67). We suggest that, in a similar manner, the C53/C37 subcomplex may affect elongation and termination through direct or allosteric interactions that alter the mobility of fork loop 2.
Is the C11 Subunit Located at the Pol III Catalytic Center?-The sequence similarity of the C11 and TFIIS C-terminal domains and their similar function of promoting exoribonucleolytic proofreading suggested the possibility that the tip of the C11 ␤ finger might also be located in proximity to the 3Ј end of the nascent transcript. RNA-protein cross-linking analysis was originally undertaken to clarify whether this is the case. However, the C11-RNA cross-linking that we have observed appears not to be associated with the pol III elongation complex but, rather, results from nonspecific binding of C11 to low levels of released RNA ( Figs. 2A and 3). Our attempts to demonstrate a residuum of specific RNA cross-linking (i.e. in the elongation complex) above this nonspecific background proved unsuccessful (data not shown). This negative outcome does not eliminate the possibility that the C11 ␤ finger lies proximal to the RNA 3Ј end of the pol III elongation complex. For example, steric occlusion (by C128, for example) may have prevented access of the RNA-attached photoactive function to C11. The amino acid selectivity of cross-linking by aryl azides is an additional consideration; for example, substituting the azide of ABdU with diazirine revealed the proximity of additional pol III subunits to DNA at certain locations of a pol III promoter complex (68).
Proximity of the C53 Subunit to the Pol III Catalytic Center-Although C53 also binds and cross-links to free RNA containing ABU at its 3Ј end (Fig. 3), our evidence indicates that C53-RNA cross-linking occurs within the elongation complex at or near its catalytic center: 1) C53 cross-linked to ABU at the 3Ј end of the nascent transcript co-chromatographed perfectly with the pol III elongation complex (Figs. 2B and 5C), 2) isolation of pol III⌬ elongation complexes (under conditions stringently removing released RNA) before C53/C37 addition and UV irradiation also resulted in C53-RNA cross-linking (supplemental Figs. S3 and S4B) and 3) C53 was seen to cross-link to the transcribed DNA strand at register Ϫ1/Ϫ2 under conditions that required the formation of a stable elongation complex (Fig. 4). These observations are also consistent with finding that pol III lacking the C53/C37 subcomplex has a promoteropening defect (Figs. 6 and 7) and with proximity of C37 to the nontranscribed strand of the elongation complex at register Ϫ2/Ϫ3 (36).
Heparin Sensitivity of C53-DNA Cross-linking-Observations on the interaction of C53 with the polyanion heparin somewhat complicate this picture and require further comment. Heparin, a single-stranded-DNA/RNA mimic, has been used to distinguish between pol III elongation complexes, which are resistant to heparin challenge, and polymerase, which fails to productively initiate transcription and remains bound at the promoter, which is stripped off by heparin treatment (34). As noted under "Results," cross-linking of C53 to the transcribed DNA strand at register Ϫ1/Ϫ2 (Fig. 4) was seen to be sensitive to heparin, whereas C160 and C128 cross-linking was not. The same heparin sensitivity was also observed for C53 cross-linking to the 3Ј end of the nascent transcript in the elongation complex (data not shown). We believe that these effects of heparin are due to its direct binding to C53 in the heparinstable pol III elongation complex, incidentally resulting in local steric interference that prevents access to the nascent transcript-attached photoactive function. Similar observations have been made with TFIIIB, which simultaneously binds to DNA and heparin (69) and with C53 bound and cross-linked to free RNA, which maintains some ability to bind to pol III (supplemental Fig. S4A).