Absolute Gene Occupancies by RNA Polymerase III, TFIIIB, and TFIIIC in Saccharomyces cerevisiae*

A major limitation of chromatin immunoprecipitation lies in the challenge of measuring the immunoprecipitation effectiveness of different proteins and antibodies and the resultant inability to compare the occupancies of different DNA-binding proteins. Here we present the implementation of a quantitative chromatin immunoprecipitation assay in the RNA polymerase III (pol III) system that allowed us to measure the absolute in vivo occupancy of pol III and its two transcription factors, TFIIIC and TFIIIB, on a subset of pol III genes. The crucial point of our analysis was devising a method that allows the accurate determination of the immunoprecipitation efficiency for each protein. We achieved this by spiking every immunoprecipitation reaction with the formaldehyde cross-linked in vitro counterparts of TFIIIB-, TFIIIC-, and pol III-DNA complexes, measuring the in vitro occupancies of the corresponding factors on a DNA probe and determining probe recovery by quantitative PCR. Analysis of nine pol III-transcribed genes with diverse sequence characteristics showed a very high occupancy by TFIIIB and pol III (pol III occupancy being generally ∼70% of TFIIIB occupancy) and a TFIIIC occupancy that ranged between ∼5 and 25%. Current data suggest that TFIIIC is released during transcription in vitro, and it has been proposed that TFIIIB suffices for pol III recruitment in vivo. Our findings point to the transient nature of the TFIIIC-DNA interaction in vivo, with no significant counter-correlation between pol III and TFIIIC occupancy and instead to a dependence of TFIIIB-DNA and TFIIIC-DNA complex maintenance in vivo on pol III function.

A major limitation of chromatin immunoprecipitation lies in the challenge of measuring the immunoprecipitation effectiveness of different proteins and antibodies and the resultant inability to compare the occupancies of different DNA-binding proteins. Here we present the implementation of a quantitative chromatin immunoprecipitation assay in the RNA polymerase III (pol III) system that allowed us to measure the absolute in vivo occupancy of pol III and its two transcription factors, TFIIIC and TFIIIB, on a subset of pol III genes. The crucial point of our analysis was devising a method that allows the accurate determination of the immunoprecipitation efficiency for each protein. We achieved this by spiking every immunoprecipitation reaction with the formaldehyde cross-linked in vitro counterparts of TFIIIB-, TFIIIC-, and pol III-DNA complexes, measuring the in vitro occupancies of the corresponding factors on a DNA probe and determining probe recovery by quantitative PCR. Analysis of nine pol III-transcribed genes with diverse sequence characteristics showed a very high occupancy by TFIIIB and pol III (pol III occupancy being generally ϳ70% of TFIIIB occupancy) and a TFIIIC occupancy that ranged between ϳ5 and 25%. Current data suggest that TFIIIC is released during transcription in vitro, and it has been proposed that TFIIIB suffices for pol III recruitment in vivo. Our findings point to the transient nature of the TFIIIC-DNA interaction in vivo, with no significant counter-correlation between pol III and TFIIIC occupancy and instead to a dependence of TFIIIB-DNA and TFIIIC-DNA complex maintenance in vivo on pol III function.
The transcriptome of RNA polymerase III (pol III) 3 in haploid Saccharomyces cerevisiae consists of 274 tRNA genes, 100 -200 tandem copies of RDN5 (encoding 5 S rRNA), SNR6 (encoding the spliceosomal U6 snRNA), RPR1 (encoding the RNA component of RNase P), SCR1 (encoding 7SL RNA of the signal recognition particle), SNR52 (encoding a snoRNA involved in 2Ј-O-methylation of rRNA), and RNA170 (function unknown) (recently reviewed in Ref. 1). Transcription of the tRNA genes accounts for 10 -15% of nucleoside triphosphates consumed by nuclear transcription (2); reflecting this high energetic cost, pol III transcription is regulated in response to environmental stress and nutrient availability through multiple signaling pathways that converge on Maf1, the central repressor of pol III transcription (3). The pol III transcription apparatus of S. cerevisiae consists of the 17-subunit polymerase, the monomeric 5 S rRNA transcription-specific TFIIIA, the sixsubunit TFIIIC with its two subdomains A (subunits Tfc1, Tfc4, and Tfc7) and B (subunits Tfc3, Tfc6, and Tfc8), and the three subunits of TFIIIB (Brf1, Bdp1, and the TATA-binding protein, TBP). The genes encoding pol III, TFIIIA, TFIIIC, and TFIIIB are all essential for yeast viability (4,5), and each subunit has a counterpart in humans (6) with the addition of a Brf1 paralogue, Brf2, that replaces Brf1 for transcription of vertebrate class 3 genes containing external promoters (7,8). A growing body of evidence also indicates that vertebrate Maf1 is a major negative regulator of pol III transcription and that, like its S. cerevisiae counterpart, it is hypophosphorylated under repressive (serum-starved) growth conditions, is associated with pol III-transcribed genes, and interacts with both Brf1 (and Brf2) and the largest subunit of pol III (9 -12).
TFIIIC is required for transcription of all pol III-specific genes in S. cerevisiae. With the exception of RDN5 (where TFIIIA serves as a platform for TFIIIC binding), TFIIIC binds through its flexibly linked A and B domains to two promoter elements, boxA and boxB, that are variably separated by ϳ20 -200 bp (but generally 30 -60 bp) of intervening sequence. With one exception, these promoter elements lie within pol III-transcribed sequence (the exception being SNR6, whose boxB lies 120 bp downstream of the transcriptional terminator). Intragenically bound TFIIIC serves two roles: first, it assembles TFIIIB at a site centered ϳ25 bp upstream of the start site of transcription through interaction between the Tfc4 subunit of A and the Brf1 subunit of TFIIIB (reviewed in Ref. 13); second, it prevents the encroachment of repressive chromatin (14,15). The TFIIIB-DNA complex suffices for pol III recruitment during rapidly reiterating rounds of transcription in vitro (16); the low levels of enrichment of TFIIIC subunits relative to TFIIIB subunits in chromatin immunoprecipitation (ChIP) analysis (17)(18)(19) may indicate that, once bound to DNA, TFIIIB also suffices for pol III recruitment in vivo (17).
TFIIIB can bind independently of TFIIIC to TATA boxes that are present upstream of SNR6 and a few tRNA genes in vitro (20,21), but its assembly onto DNA is entirely dependent on TFIIIC in vivo and in the context of chromatin in vitro (22,23). Although the TFIIIB-binding site upstream of pol III-transcribed genes is generally AT-rich (24), TFIIIC can assemble TFIIIB onto an entirely GC-containing sequence (albeit less efficiently) with the resulting TFIIIB-DNA complex maintaining the unique property that characterizes complexes with natural tRNA genes and with the TATA box-containing SNR6: resistance to dissociation by high salt concentrations and by polyanions such as heparin (25). In fact, no dissociation of a fully recombinant TFIIIB-DNA complex is observed after 2 h in the presence of 200 g/ml heparin. 4 This great stability of the TFIIIB-DNA complex suggests that its occupancy on pol III genes should be high in the cell. Although Maf1 inhibits the de novo assembly of TFIIIB onto DNA (26), it has little or no effect on preformed TFIIIB complexes in vivo (17). The first definitive analysis of occupancy of the pol III transcription apparatus was performed at the 100 -200 copies of RDN5 by in vivo footprinting (27); occupancy by TFIIIB was determined to be between 23 and 47%, and pol III occupancy in open promoter complexes was seen to be between 8 and 17%. An elegant quantitative electron microscopic analysis of pol III bound to the RDN5 genes (published while this manuscripts was in review) is consistent with the above footprinting analysis and indicates that 20 -30% of the RDN5 gene copies are actively transcribed (28).
We have been interested in how properties of pol III transcription complexes that have been observed in vitro are manifested in terms of absolute gene occupancies in vivo and have approached this problem with a modified ChIP analysis. ChIP is a powerful methodology for enumerating the components of chromatin, determining their disposition on the genome, and specifying correlations between changes of disposition and composition on the one hand with changes of the functional state of genes on the other. ChIP measurements are currently made on populations of cells, and various statistics-based analytical schemes are available for assessing the enrichment of individual DNA segments in the immunoprecipitate (IP) relative to some selected standard (generally one or more genomic regions where the protein of interest is not expected to bind). Enrichment is the product of two quantities: IP efficiency and occupancy. IP efficiency is antibody-specific; differences of IP efficiency may be encountered when different proteins are tagged with the same epitope, and even with different placements of an epitope tag on a protein (essentially because partners or neighbors of the tagged protein can expose or mask epitopes to varying extent). Thus, deconvoluting IP enrichment to specify site occupancy is an unsolved problem. Although it stands to reason that there must be a general correlation between IP enrichment and gene occupancy, it would be helpful to be able to convert shared belief to experimental quantification.
In the work that is presented, we have spiked formaldehydetreated cell lysates with in vitro assembled and formaldehyde-cross-linked pol III transcription complexes as a means of determining chromatin immunoprecipitation efficiencies to derive protein occupancies of a diverse group of pol III-transcribed genes. Occupancy by TFIIIB of all but one transcriptionally functional pol III gene was near saturation levels (Ն80%) with pol III occupancies on average 60% of TFIIIB occupancy. In contrast, TFIIIC occupancies on average were only 14% of TFIIIB occupancies, indicating that TFIIIC is not directly involved in pol III recruitment in vivo. In an attempt to observe a counter-correlation between TFIIIC occupancy and ongoing transcription by pol III, we introduced a plasmid vector containing a modified SUP4 tRNA gene that allows TFIIIB and TFIIIC to form their promoter complex but greatly reduces the ability of pol III to initiate transcription. Surprisingly, the TFIIIC and TFIIIB occupancies of this gene were greatly diminished and, moreover, to an even greater extent than pol III occupancy. The possible implications of this finding are also discussed.

EXPERIMENTAL PROCEDURES
Yeast Strains and Plasmids-Thirteen copies of the c-Myc epitope were inserted at the C termini of RET1, BDP1, TFC4, TFC1, TFC6, TFC7, and TFC8 in yeast strain BY4727 (ATCC) by homologous recombination of a 13Myc-KanMX PCR product (29) generated with primers specified in supplemental Table S2. TA30pyr was obtained by PCR amplification of plasmid pCJ-TA30 (25) with primer 5Ј-TATAAAGCCGCGGTC-CCTTACTCTTTCTTCtcCtcTTcctTtCTCTCGGTAGCC-3Ј and a T7 promoter primer and cloned into the SacII and EcoRI sites of pCJ-TA30. TA30 and TA30pyr were excised as BamHI fragments and inserted into the BamHI site of pRS316. Plasmid pRS316-TA30-70 -tQ-tL and its pyr variant were constructed as follows: tL(CAA)DR2 was amplified by PCR from genomic DNA (bp Ϫ76 to ϩ205, relative to the start site of transcription as ϩ1) and inserted into the SmaI site of vector pGEM1. tL(CAA)DR2 was excised as an XbaI-EcoRI fragment, and tQ(UGG)L was excised as an XbaI-HindIII fragment from plasmid pPC6 (30), and both fragments were simultaneously ligated between the EcoRI and HindIII sites of pET21b-RpoD(1-571) (31). The two tRNA genes and the rpoD segment were excised with SacII (filled in) and NotI and ligated into the NotI and XbaI (filled in) sites of the above-specified pRS316-TA30 plasmids.
ChIP and Quantitative PCR (qPCR)-Yeast cultures (100 ml) were grown at 30°C in YPD to A 600 of ϳ0.8. Formaldehyde cross-linking and ChIP followed procedure and buffers specified by Strahl-Bolsinger et al. (32), after scaling with respect to culture volume and with the following modifications. Crosslinked and washed cells were resuspended in 600 l of lysis buffer (50 mM Na-Hepes, pH 7.8, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, 1 mM phenylmethylsulfonyl chloride, 1 mM benzamidine, 10 g/ml aprotinin, 1 g/ml leupeptin, 1 g/ml pepstatin) and broken with glass beads for 30 min at 4°C. The whole lysate was sonicated to an average DNA fragment size of ϳ500 bp. Sonicated chromatin, corresponding to 600 g of total protein/ChIP sample, was preincubated with 40 l of protein G-agarose for 1 h at 4°C, and the supernatant fluid was recovered. In vitro assembled and cross-linked protein-DNA complex was added to this material (see under "EMSA" below).
An aliquot was then saved as total input DNA/protein reference, and the remainder was incubated with 10 g of mouse monoclonal anti-Myc antibody overnight at 4°C. Protein G-agarose beads (40 l) were added for 2 h at 4°C. The beads were recovered and washed twice as cited above, and crosslinked protein-DNA complexes were eluted with 2 ϫ 75 l of elution buffer (50 mM Tris-HCl, pH 8, 1% SDS, 10 mM EDTA) for 10 min at 65°C. Cross-linking was reversed by incubation for 6 h at 65°C. An aliquot of this material was saved for Western blot analysis (see below), and the DNA of the remainder was purified on a PCR purification spin column. qPCR was performed by titrating total input DNA and ChIP DNA in the presence of [␣-32 P]dCTP, using the primers listed in supplemental Table S3 (M13 forward and reverse primers were used for amplification of TA30 and TA30pyr in pRS316) and quantified by PAGE and phosphorimaging plate analysis. The spiked complex ChIP recovery was quantified by scintillation counting and by qPCR (with SP6 and T7 promoter primers). Fractional occupancy for each pol III-transcribed gene was determined as (N Ϫ C)/(S Ϫ C), where N, S, and C are the fractions of input DNA that were immunoprecipitated for the gene of interest (N), for the added spiked complex (S), and for the Tfc4 open reading frame-negative control (C), respectively.
Western Blotting-Immunoprecipitation efficiency by Western blot analysis (33) was based on titrations of total input and immunoprecipitated and eluted proteins (after reversal of cross-linking). Anti-Myc antibody was detected with rabbit anti-mouse IgG and 125 I-protein A and quantified by phosphorimaging plate analysis.
For protein-DNA complex formation with Myc-tagged TFIIIC subunits, reactions were supplemented with 100 fmol each of recombinant Brf1, TBP, and Bdp1 (rTFIIIB) (35) and incubated for 40 min at room temperature. Formaldehyde was added to 0.5% (v/v) for 5 min, and cross-linking was terminated by the addition of glycine to 0.36 M. After challenge with 0.2 mg/ml heparin for 2 min, the samples were loaded for electrophoresis on a 4% polyacrylamide-TE gel (34) to resolve and quantify the heparin-resistant TFIIIC-TFIIIB-DNA complex. For in vitro assembled Myc-tagged TFIIIB-containing com-plexes, the BR␣ fraction from a BDP1-Myc cell extract was incubated with probe DNA, 100 fmol of recombinant TBP, and 100 fmol of recombinant Brf1. After a 40-min incubation at room temperature, formaldehyde was added to 0.5% (v/v) for 5 min, and cross-linking was terminated with glycine added to 0.36 M. After challenge with 0.2 mg/ml heparin for 2 min, the samples were subjected to electrophoresis as above to resolve and quantify the heparin-resistant TFIIIC-TFIIIB-DNA and TFIIIB-DNA complexes. For in vitro assembled Myc-tagged pol III-containing complexes, the DEAE Sephadex fraction from a RET1-Myc cell extract was incubated with the DNA probe and 100 fmol of rTFIIIB and highly purified TFIIIC (16). After incubation at room temperature for 40 min, 200 M ATP, 100 M CTP, and 50 M UTP (final concentrations) were added for 4 min to produce stalled elongating TFIIIB-TFIIIC-pol III-DNA complexes. These were challenged with 0.2 mg/ml heparin to strip TFIIIC, formaldehyde was added to 0.5% (v/v), and the reaction was further incubated for 5 min before the addition of glycine to 0.36 M. Reaction mixtures were processed to resolve and quantify heparin-resistant pol III-TFIIIB-DNA complexes as specified above.
The concentration of formaldehyde used to cross-link these in vitro complexes (0.5%) was determined as the amount necessary to obtain a heparin-resistant TFIIIC-TFIIIB-DNA complex and a TFIIIB-DNA complex that is resistant to dissociation by 4 M urea. One-third of every in vitro binding reaction mixture was loaded onto a 4% polyacrylamide-TE gel for quantification of complex formation; two-thirds were added to a whole cell extract prepared from each formaldehyde-treated Myctagged strain and subjected to ChIP, as described above.
Primer Extension-The primer 5Ј-CTCAAGATTTCG-TAGTGATAAATTAA-3Ј was annealed to total RNA extracted from yeast bearing the pRS316-TA30 plasmid and yeast bearing the pRS316-TA30pyr plasmid, described above. Reverse transcription was carried out with avian myeloblastosis virus reverse transcriptase (Promega), actinomycin D, and RNAguard (GE Healthcare) for 1 h at 42°C. The samples were loaded onto an 8% polyacrylamide, 8 M urea gel and quantified by phosphorimaging plate analysis.

RESULTS
The genes encoding the second largest subunit of pol III, C128, the Bdp1 subunit of TFIIIB, and the Tfc4 subunit of TFIIIC were modified to contain 13 copies of the c-Myc epitope tag at their C termini (29) in individual S. cerevisiae strains. Two methods for quantifying the efficiency of chromatin immunoprecipitation were compared. In the first method, Western analysis with anti-Myc antibody and 125 I-protein A was used to compare the efficiency of immunoprecipitation of each tagged subunit from their respective formaldehyde-treated cell lysates. This simple approach does not take into account differences in the precipitation efficiency of TFIIIB, TFIIIC, and pol III bound to DNA in the context of cross-linked chromatin or in their respective unbound (free) states. A second method was devised to directly measure the immunoprecipitation efficiency of formaldehyde-cross-linked TFIIIB-, TFIIIC-, and pol III-DNA complexes with the aid of their in vitro assembled counterparts. Partially purified fractions containing Bdp1-Myc, Tfc4-Myc, or C128-Myc were prepared and used to form TFIIIC-TFIIIB-DNA complexes with and without pol III on a 32 P-labeled modified SUP4 tRNA gene probe. Binding conditions were optimized for the formation of TFIIIB-TFIIIC-DNA complexes containing either Tfc4-Myc or Bdp1-Myc, and TFIIIB-pol III-DNA complexes containing C128-Myc that could be unambiguously identified and quantified following reaction with formaldehyde. Sample experiments showing how this was done are presented in Fig. 1. The BioRex70 fraction BR␣ (34) contains TFIIIC, TFIIIB, and pol III but is limiting for Brf1 and more so for pol III. Lane 3 of Fig. 1A shows the complexes formed with fraction BR␣ containing C128-Myc. The mobilities of the two major bands corresponded with expectation for TFIIIC-DNA (lower band) and TFIIIB-TFIIIC-DNA (upper band) complexes. Consistent with this assignment, the addition of recombinant Brf1 and Bdp1 converted much of the lower complex to the upper complex (lane 4). Optimization trials called for the inclusion of nonspecific carrier DNA in these binding reactions; a SUP4 tRNA gene plasmid that was inactivated for TFIIIC binding by a G56 substitution in its boxB binding site (36) was used for this purpose. Substituting the wild type version of SUP4 in this plasmid converted carrier to competitor DNA, eliminating both major complexes (lane 2) and signifying that both complexes contained TFIIIC. That the upper complex contained complete TFIIIB (i.e. Brf1ϩTBPϩBdp1) was evidenced by its nearly quantitative conversion to a TFIIIB-DNA complex by heparin, which strips TFIIIC from DNA but leaves the TFIIIB complex intact (lane 5) (16). (Heparin also strips DNA-bound TBPϩBrf1 (37).) Only trace levels of Myc-tagged pol III-TFIIIB-TFIIIC-DNA complexes were seen. This BR␣ fraction was subsequently chromatographed on DEAE Sephadex to concentrate pol III (as well as TFIIIC) and to remove the bulk of nonspecifically DNA-binding proteins.
When formaldehyde-treated complexes containing Myc-tagged Tfc4, Bdp1, or C128 were produced, a small portion was retained for EMSA ( Fig. 1, B-D); the remainder was used to spike the corresponding formaldehyde-treated cell lysate. In each case, we endeavored to crosslink and quantify higher order complexes that would more closely reflect those that would assemble in vivo. For lane 3 of Fig. 1B, TFIIIB-TFIIIC-DNA complexes were formed with a DEAE Sephadex fraction containing Tfc4-Myc-tagged TFIIIC supplemented with recombinant TFIIIB and reacted with formaldehyde before stripping with heparin. The TFIIIC-DNA and TFIIIB-TFIIIC-DNA complexes and the smear above the latter complex (containing, in addition, pol III and/or nonspecifically binding proteins) were quantified together as containing TFIIIC. Here, the presence of the TFIIIB-DNA complex is due to TFIIIC release by formaldehyde, which occurs to variable extent, rather than to heparin stripping of un-cross-linked TFIIIC (data not shown). For formation of Bdp1-Myc-tagged TFIIIB-TFIIIC-DNA complexes, the corresponding BR␣ fraction was supplemented with recombinant Brf1 and TBP. Radioactivity in the formaldehyde-treated TFIIIB-TFIIIC-DNA and TFIIIB-DNA complexes (Fig. 1C, lane 2) was combined for quantifying Bdp1-Myc (the sum closely matching radioactivity in the TFIIIB-DNA complex released by heparin in the absence of cross-linking; lane 1).
Although it is possible to resolve the pol III-TFIIIB-TFIIIC-DNA and TFIIIB-TFIIIC-DNA complexes electrophoretically, release of TFIIIC by formaldehyde would generate a pol III-TFIIIB-DNA complex that is not resolved from the TFIIIB-TFIIIC-DNA complex. Accordingly, to be able to quantify formaldehyde cross-linked complexes containing Myc(C128)tagged pol III, we took advantage of the fact that the transcriptelongating complex formed on the SUP4 gene in the absence of GTP (containing a 17-nt nascent transcript) is resistant to dissociation by heparin, as is TFIIIB (Fig. 1D, lane 2) (30); in contrast, heparin strips pre-elongating pol III as well as TFIIIC (lane 1). pol III-TFIIIB-TFIIIC-DNA complexes were formed with recombinant TFIIIB and the DEAE Sephadex fraction containing C128-Myc pol III and TFIIIC; ATP, CTP, and UTP were added; TFIIIC and uninitiated pol III were stripped with heparin prior to formaldehyde treatment. The resulting pol III-TFIIIB-DNA complex was quantified (lanes 3-5).
Determinations of the ChIP efficiency of in vitro assembled complexes introduced into formaldehyde-treated cell lysates on the basis of recovery of 32 P-labeled probe and qPCR were in good agreement ( 32 P/PCR ϭ 1.15 Ϯ 0.24). qPCR-derived values of ChIP efficiency were used for all of the analyses that follow. We next compared the immunoprecipitation efficiency of total Myc-tagged subunit by quantitative Western blotting with that of the spiked complex in the same cell lysates (Table 1). Whereas the IP efficiencies of total Tfc4-Myc and Tfc4-Myc present in a TFIIIB-TFIIIC-DNA complex were comparable, the IP efficiencies of DNA-bound Bdp1-Myc and C128-Myc were 2.0-and 2.7-fold higher, respectively. Although the in vitro cross-linked complexes do not perfectly mimic complexes assembled and cross-linked in vivo, gene occupancy values for TFIIIB and pol III determined on the basis of spiked complex immunoprecipitation efficiency were in line with expectation, whereas occupancy values based on quantitative Westerns far exceeded 100%.
The spiked-complex procedure was used to quantify TFIIIC, TFIIIB, and pol III occupancy on a set of nine pol III-transcribed genes with diverse sequence characteristics that might influence occupancy ( Table 2). For example, tRNA genes tP(AAG)CR and tT(TGT)HR2 have initially transcribed sequences that may diminish transcription and pol III occupancy; consequently, if active transcription across boxB releases TFIIIC, these genes might display increased levels of TFIIIC occupancy. Likewise, the upstream DNA sequence contexts of tRNA genes tI(TAT)DR2 and tW(CCA)GL might enhance and decrease TFIIIB binding, respectively. Two anomalous loci were included because they have been determined (by qPCR) to contain TFIIIC with diminished levels of (ZOD1) or no (ETC8) TFIIIB and pol III. The outcome of this analysis is summarized in Table 3.
TFIIIB occupancy of the five tRNA genes was extremely high (ϳ80 -120%), as was occupancy at SCR1, encoding the 7SL RNA of the signal recognition particle (ϳ90%; Table 3). Even taking into account the margin of error and the above-noted underestimation of the spiked complex IP efficiency by PCR, these values suggest some maintenance of TFIIIB-DNA complexes within the 15-20% of cells undergoing mitosis (38). The occupancy of TFIIIB at the U6 snRNA gene, SNR6, was ϳ50%. It is noteworthy that pol III occupancy of these genes (excluding the anomalous ZOD1 and ETC8 loci) was generally ϳ70% of TFIIIB occupancy. We also note that the two genes with the lowest relative pol III occupancy, tP(AAG)CR and tT(TGT)HR2 (ϳ40%), were selected precisely for that potentiality ( Table 2). TFIIIC occupancy ranged between ϳ5 and 25% on all genes (including the ZOD1 and ETC8 loci). TFIIIC occupancy relative to that of TFIIIB correlated well (ϳ15%, again excluding the ZOD1 and ETC8 loci). The higher relative TFIIIC abundance at tT(TGT)HR2 (ϳ25%) stands out and may be a consequence of the lower relative occupancy of pol III on this gene, but the similar level of pol III relative to TFIIIB at tP(AAG)CR did not correlate with a high occupancy by TFIIIC relative to TFIIIB. The SCR1 gene was partly chosen on the basis of its highly divergent boxB TFIIIC-binding site (15), which includes an A54 mutation that lowers the affinity for TFIIIC more than 40-fold (36). Consistent with the suboptimal SCR1 boxB, the occupancy of TFIIIC relative to TFIIIB was ϳ2-fold lower than for the other genes of the set, with the exception of tW(C-CA)GC, which was also ϳ2-fold lower. Occupancy of TFIIIC, TFIIIB, and pol III at the anomalous loci ZOD1 (zone of disparity 1) and ETC8 (extra TFIIIC 8) correlated well with relative occupancy previously determined by qPCR (18); TFIIIC occupancy at ETC8 was at the low end of the range   for the other eight genes, with negligible levels of TFIIIB and pol III, and TFIIIC occupancy at ZOD1 was at the high end of that range, with TFIIIB occupancy ϳ2-3-fold lower and pol III 5-10-fold lower relative to the other genes. SNR6 was chosen for analysis partly for the placement of its high affinity TFIIIC-binding boxB site ϳ120 bp downstream of the transcriptional terminator. If displacement of TFIIIC by pol III transcription through the intragenic boxB element on all other pol III-transcribed genes is responsible for the observed low TFIIIC/TFIIIB occupancy ratios, one might expect to see a higher relative TFIIIC (Tfc4-Myc) occupancy of this gene. This was not the case (Table 3). We considered the possibility that because Tfc4-Myc is part of the A subcomplex that binds to the start site-proximal boxA promoter element, formaldehyde treatment may fail to covalently link the A subcomplex to DNA indirectly through the boxB-binding B subcomplex. To exclude this eventuality, we repeated the spiked complex occupancy analysis with a yeast strain harboring Tfc6-Myc. Tfc6 ( 91 ) is part of the B subcomplex and cross-links to DNA at the start site-distal side of boxB (39). No significant difference was observed between TFIIIC occupancy based on Tfc6 ( B ) and Tfc4 ( A ) cross-linking at SNR6 and three other loci (Fig. 2).
Because the above attempts to assess the existence of a counter-correlation between pol III and TFIIIC occupancy were inconclusive, we constructed a SUP4 tRNA gene that compromises transcriptional initiation by eliminating purine residues on the nontranscribed strand within 10 bp of the normal start site (TA30pyr; Fig. 3A). This altered gene and its reference type (TA30) were separately inserted into the centromeric vector pRS316 and transformed into the yeast strains expressing Myctagged Tfc4, Bdp1, and C128. A mutant SUP4 tRNA gene lacking a purine within 7 bp of the normal start site for transcription functions poorly in vitro and fails to suppress ochre mutations in vivo (40). The ability of TA30pyr to function for transcription was compared with its parental TA30 construct by primer extension of total yeast RNA with reverse transcriptase using a SUP4 intron-complementary primer with the ochre-suppressing mutation at its 3Ј end. Because pre-tRNA transcripts are processed rapidly, the relative abundance of primary transcripts measured by primer extension with intron-specific primers is an approximate measure of the relative rate of transcription. Lane 1 of Fig. 3B shows a sample primer extension analysis of yeast harboring the reference SUP4 gene on pRS316-TA30. The ϩ13 5Ј end results from 5Ј-end processing, which generally precedes intron removal, because splicing of unprocessed tRNAs occurs at a substantially slower rate (41). Cells harboring pRS316-TA30pyr yielded substantially reduced levels of primer extension products (Fig. 3B, lane 2) with 5Ј ends corresponding to initiation at bp ϩ1 and ϩ2. Surprisingly, a 5Ј end mapping to ϩ18 was also detected with this template, but not the reference SUP4 gene.
Whether this 5Ј end is generated by a new initiation site or by aberrant processing was not established, but the ϩ1, ϩ2, and ϩ18 primer extension products for SUP4-TA30pyr were collectively compared with ϩ1 and ϩ13 primer extension products for SUP4-TA30 in estimating that the relative rate of transcription of the TA30pyr gene was at least 6-fold lower than its TA30 reference counterpart. This is an overestimation of the residual activity of TA30pyr because a low level background of ϩ1 and ϩ2 5Ј ends generated by RNA from cells lacking the pRS316 plasmids was not subtracted. (Presumably this background is due to pre-tRNA Tyr that contains the identical intron but lacks the ochre suppressor mutation.) The spiked complex procedure was used to quantify occupancy of these plasmid-borne  SUP4 genes (left half of Fig. 3C). TFIIIB occupancy of the SUP4-TA30 gene was at least 2-fold lower relative to the other analyzed tRNA genes (Table 3). This may be a consequence of the exceptionally GC-rich upstream DNA sequence context of SUP4-TA30 with only a partial TATA box for nucleation of TFIIIB-DNA complex formation (Fig. 3A). The occupancy of TFIIIC and pol III relative to TFIIIB (0.23 and 0.37, respectively) was within the range observed for other tRNA genes. Strikingly, the effect of the pyr variation on occupancy was the opposite of what had been anticipated: TFIIIC occupancy was reduced ϳ10-fold (to less than 1%) and TFIIIB occupancy was reduced ϳ6-fold, whereas pol III occupancy was reduced only ϳ3-fold.
Because the dispersed chromosomal tRNA genes cluster and co-localize in the nucleolus or at its periphery (42), we considered the possibility that the greatly reduced occupancy of the TA30pyr gene by all components of the pol III transcription apparatus could partly reflect a peculiarity of localization of a plasmid containing only a single tRNA gene. To assess this possibility, two additional tRNA genes, tQ(UGG)L and tL(CAA)C, were inserted into each plasmid, separated from SUP4-TA30 by a ϳ1600-bp spacer to prevent co-immunoprecipitation of SUP4 with complexes assembled on the other two tRNA genes. The presence of the additional tRNA genes did not mitigate the deficiency of TA30pyr transcription complexes (right half of Fig. 3B). The implications of the loss of TFIIIC and TFIIIB because of an 8-bp change in the initially transcribed segment of SUP4 are discussed below.
pol III binds to the preformed TFIIIB-promoter complex and executes the initial round of transcription slowly relative to reinitiated subsequent rounds of transcription. This process of rapid reinitiation, termed facilitated recycling, occurs without release of pol III from the transcription unit, suggesting that pol III is directly handed off from the terminator back to the TFIIIB-DNA complex. TFIIIB suffices for facilitated recycling on short transcription units, but TFIIIC is additionally required on longer units, such as the 520-bp SCR1 (43). Curiously, a prior qPCR analysis with tiled amplicons indicated relatively uniform TFIIIC occupancy throughout the SCR1 gene in exponentially growing cells (19) and TFIIIB occupancy of the SCR1 terminator region at 20 -25% of TATA box occupancy.
These results encouraged us to apply quantitative ChIP at SCR1 using an overlapping set of PCR amplicons (Fig. 4A). TFIIIB and TFIIIC occupancy dropped off significantly within the middle of the gene (Fig. 4B; 20 and 30%, respectively, for probe 3 relative to probe 1) and was quite low at the terminator (less than 5% occupancy for probe 5 relative to probe 1). In contrast, pol III occupancy was maintained at relatively high levels throughout the SCR1 gene but decreased progressively with increasing distance from the start site (ϳ60% occupancy with probe 5 relative to probe 1). The diminution of pol III occupancy in the downstream segment of the transcription unit is consistent with termination and pol III release not representing a rate-limiting step for transcription in the cell. The substantial decrease of TFIIIC and TFIIIB occupancy with probe 4 in respect to probe 1, and the absolute levels of pol III with both probes 1 and 4 suggest more than one pol III elongation complex simultaneously traversing the SCR1 gene. This would con-tribute to the high abundance of 7SL RNA in yeast (0.2% of total RNA (44), or approximately one signal recognition complex for every forty 80 S ribosomes on the basis of their respective cellular RNA content). French et al. (28) have also estimated that SCR1 would need to be simultaneously transcribed by multiple elongation complexes to generate the required abundance of 7SL RNA.

DISCUSSION
We have pursued the quantitative determination of (absolute) occupancy by TFIIIC, TFIIIB, and pol III on a sequencedistinctive group of pol III-transcribed genes for its intrinsic interest and also because our quantification could be used as a standard for interpreting qPCR-derived measurements of ChIP enrichment in terms of site occupancy.
The validity of our approach is contingent on the suitability of in vitro assembled complexes for standardization. The in vitro cross-linked complex is admittedly formed in an environment that does not match cellular chromatin. Although we have deliberately used crude yeast fractions for complex formation, it is necessary to add carrier DNA to these extracts to sequester nonspecifically DNA-binding proteins so that TFIIIC-, TFIIIB-, and pol III-DNA complexes can be identified and quantified. On the other hand, it is unlikely that the additional proteins present in cross-linked yeast chromatin substantially mask the Myc epitope of Bdp1 and C128, because this would imply true occupancies greater than 100% (Fig. 4 and Table 3). We conclude that it is also highly unlikely that the Myc epitope of Tfc4 is differentially masked in the context of yeast chromatin, because one would expect such masking to vary between Myc tags placed on different TFIIIC subunits. On the contrary, no significant difference of TFIIIC occupancy was found for TFIIIC Myc-tagged on the Tfc4 and Tfc6 subunits (Fig. 2). Estimates of TFIIIC occupancy at two tRNA genes based on Western-derived IP efficiencies with yeast harboring Myc-tagged Tfc1, Tfc4, Tfc6, Tfc7, or Tfc8 subunits of TFIIIC were also not substantially different (supplemental Fig. S1). The last result also argues against the possibility that the Myc tag exerts an effect on TFIIIC occupancy in vivo. The low absolute levels of TFIIIC occupancy at all transcriptionally active genes relative to TFIIIB and pol III occupancy (Table 3) strongly support the notion that TFIIIB suffices to recruit pol III for transcription in vivo as it does in vitro.
The exceptionally high occupancy of tRNA genes and SCR1 by TFIIIB and pol III (Table 3) suggests that these genes are set at maximal activity for transcription in yeast growing exponentially in rich medium. Maf1, the global repressor of pol III transcription, is associated with pol III-transcribed genes during repressive conditions; the low level of Maf1 occupancy that is maintained in exponentially growing cells is thought to finetune transcription by attenuation (19,45). Elimination of Maf1 activity by deletion leads to a modest (0 -36%) increase in pol III transcription in exponentially growing cells, analyzed in a small sample of genes (46,47), and is correlated with at best modestly increased pol III and TFIIIB occupancy (19,45). These small effects are consistent with the already high TFIIIB and pol III occupancies determined here. A much greater increase in tRNA gene transcription (ϳ3-fold) was observed in yeast harboring a temperature-sensitive Maf1 mutant (maf1-1) when grown at the restrictive temperature to late log phase in a low phosphate medium (48). If this stimulation of pol III transcription in the Maf1-defective cells reflects absolute levels of active transcription complexes in the corresponding wild type, this would indicate that Maf1-mediated repression under these growth conditions leaves only one-third of individual tRNA genes bearing active transcription complexes.
We have attempted to compare our TFIIIC, TFIIIB, and pol III occupancy determinations with the only qPCR analysis of ChIP enrichment (18) to examine more than one gene also represented in Table 3, along with nine tRNA genes that would provide a basis for scaling PCR-based relative enrichment to spiked complex-normalized occupancies. It should be noted that cells for the prior analysis were grown in synthetic medium, whereas the complex, rich medium YPD was used for this work. This difference in growth conditions may account for the extremely low relative occupancies of TFIIIC, TFIIIB, and pol III at SCR1 in the prior analysis compared with our observations. Roberts et al. (19) also reported relative ChIP enrichment of Bdp1 at SCR1 relative to tK(CUU)G1 to be 0.93 compared with 0.19 for the same genes in Moqtaderi and Struhl (18). Scaling on the basis of the average relative occupancy values for the nine tRNA genes analyzed in the prior work (18) to the average absolute occupancies of the five tRNA genes in Table 3 (the resulting raw occupancy scaling factors for Tfc4, Bdp1, and C34 are 0.22, 0.22, and 0.24, respectively) yields average TFIIIC/TFIIIB and pol III/TFIIIB occupancy ratios in the prior work (18) that are nearly identical with our observations (0.14 and 0.67, respectively, compared with 0.14 and 0.60 for genes listed in Table 3; excluding the ZOD1 and ETC loci in each case). Differences in growth conditions preclude definitive conversion of the prior data to quantitative occupancy, but a tentative assessment is provided in supplemental Table S1.
The notion that low TFIIIC occupancy, implied by lower ChIP enrichment, is a consequence of persistent transcription through the boxA-and boxB-binding sites of TFIIIC stems from the observation that, in the absence of TFIIIB, elongating pol III readily displaces TFIIIC from its binding sites in vitro (49). It has been shown that Maf1-mediated repression of pol III transcription results in decreased pol III occupancy on most of its conjugate genes (45), with a concomitant increase of TFIIIC occupancy at some genes (17,19) but not others (45,50). Genespecific differences in response to transcriptional repression may reflect the transient nature of the increase in TFIIIC occupancy that was seen at the onset of repression by Roberts et al. (17,19). Only Roberts et al. (17) examined the time variable (at three genes). Whether these transients are a common property of all pol III-transcribed genes remains to be determined. The transient nature of the TFIIIC occupancy change also raises the possibility that it may represent something more complex than a simple, direct response to a change of transcriptional traffic across its DNA-binding sites. It is also noteworthy that occupancy of ZOD1, which is essentially inert for transcription (unless activated by nucleosome depletion (18,51)), by TFIIIC and TFIIIB is nearly the same (Table 3), yet the absolute level of TFIIIC occupancy is within the range of actively transcribed genes. Activation of this gene by nucleosome depletion was seen to have little or no effect on relative TFIIIC, TFIIIB, and pol III occupancy (51), yet ZOD1 is as susceptible to interference between ongoing transcription and TFIIIC occupancy as the tRNA genes (because its transcriptional terminator is sufficiently close to boxB to exclude simultaneous occupancy by pol III and the B domain of TFIIIC).
We have attempted to force the issue of the relationship between TFIIIC occupancy and transcription by constructing an artificial SUP4 tRNA gene (TA30pyr) that is unable to initiate transcription normally with a purine nucleotide, generating an at least 6-fold reduction relative to its reference construct, TA30 (Fig. 3B). The analysis had two unexpected outcomes. A minor surprise was finding a TA30pyr-dependent RNA 5Ј end corresponding to bp ϩ18, implying aberrant processing. (The more exciting alternative possibility that, when forced, pol III is capable of scanning a significant stretch of DNA in the downstream direction for an initiating purine is unlikely, because a start at bp ϩ18 was not observed when this template was transcribed in vitro in the absence of tRNA processing; data not shown.) The principal surprise was that occupancy of the TA30pyr gene by TFIIIC and TFIIIB was dramatically reduced relative to TA30 (Fig. 3C). The start siteproximal pyrimidine substitutions in TA30pyr are unlikely to be directly responsible for this effect because they are far removed from the upstream binding site of TFIIIB and from the boxA-binding site of TFIIIC (and they also lie outside the boxA region of its DNase I footprint (16)), and TA30 and TA30pyr were equivalently competent in forming TFIIIB-TFIIIC-DNA and heparin-resistant TFIIB-DNA complexes in vitro when assessed by EMSA (supplemental Fig. S2). The reduced formation of promoter complexes on TA30pyr (relative to TA30) is phenomenologically unlike Maf1-mediated repression, in that TFIIIC and TFIIIB occupancy diminishes more than pol III occupancy. In particular, the reduction of occupancy by the very stable TFIIIB-DNA complex to levels comparable with those of pol III (Fig. 3C) suggests a dependence of TFIIIB-DNA complex maintenance on pol III.
pol III transcribes single-copy genes that are essential for cell viability (e.g. RPR1, SNR6). Although the stability of the TFIIIB-DNA complex is beneficial for robust activity of genes whose transcripts are required in abundance, it could be lethal if TFIIIB is damaged in such a way that the DNA binding capacity is maintained but its role in assembling pol III and/or facilitating open complex formation (52) is compromised. Misdirection of TFIIIB placement by TFIIIC (a phenomenon that is encountered in vitro (53,54)) could also have lethal consequences. Thus, it would not be surprising to find that a mechanism exists for sensing TFIIIB-DNA complexes that are not associated with pol III and targeting them for disassembly. Such a pathway comes into play for yeast pol II transcription, with permanently arrested elongation complexes at sites of DNA damage being targeted for ubiquitylation and degradation (reviewed by Svejstrup (55)).