Purification, Characterization, and Reconstitution of DNA-dependent RNA Polymerases from Caulobacter crescentus*

Cell differentiation in the Caulobacter crescentus cell cycle requires differential gene expression that is regulated primarily at the transcriptional level. Until now, however, a defined in vitro transcription system for the biochemical study of developmentally regulated transcription factors had not been available in this bacterium. We report here the purification of C. crescentus RNA polymerase holoenzymes and resolution of the core RNA polymerase from holoenzymes by chromatography on single-stranded DNA cellulose. The three RNA polymerase holoenzymes Eς54, Eς32, and Eς73 were reconstituted exclusively from purifiedC. crescentus core and sigma factors. Reconstituted Eς54 initiated transcription from the ς54-dependent fljK promoter ofC. crescentus in the presence of the transcription activator FlbD, and active Eς32 specifically initiated transcription from the ς32-dependent promoter of the C. crescentus heat-shock gene dnaK. For reconstitution of the Eς73 holoenzyme, we overexpressed the C. crescentus rpoD gene in Escherichia coliand purified the full-length ς73 protein. The reconstituted Eς73 recognized the ς70-dependent promoters of the E. coli lacUV5 and neo genes, as well as the ς73-dependent housekeeping promoters of theC. crescentus pleC and rsaA genes. The ability of the C. crescentus Eς73 RNA polymerase to recognize E. coliς70-dependent promoters is consistent with relaxed promoter specificity of this holoenzyme previously observed in vivo.

Bacterial RNA polymerases (RNAP) 1 are multi-subunit enzyme complexes that can be purified as the core polymerase (E) and the holoenzyme (E; reviewed in Refs. 4 and 5). The core RNAP, composed of the ␣ 2 , ␤, and ␤Ј subunits, carries out RNA chain elongation, whereas the holoenzyme, which also contains the sigma subunit (), recognizes specific promoter sequences. Multiple sigma factors with unique promoter specificities have been identified in many eubacteria, and the use of alternative sigma subunits is a fundamental mechanism for reprogramming RNAP specificity and controlling complex patterns of gene transcription (reviewed in Refs. 6 -8).
The most extensive study of transcription regulation in C. crescentus has been carried out on the genes in the flagellar gene hierarchy (reviewed in Ref. 9). Flagellum formation requires the temporally controlled transcription of approximately 50 genes (10) that are organized in a regulatory hierarchy containing four classes of genes (I to IV). The Class II genes, which are expressed early in the cell cycle and encode basal body and switch components, contain a unique promoter consensus. Recent results have shown that transcription from the Class II promoters is regulated in vivo by the response regulator CtrA (11). The class II gene products are required, in turn, for transcription of class III and class IV genes that are transcribed from 54 -dependent promoters late in the cell cycle. The 54 factor and transcription activator FlbD, which are encoded by Class II genes rpoN and flbD, are required for transcription of the Class III and IV genes (reviewed in Ref. 9).
Biochemical studies of transcription in vitro have employed either the E 54 holoenzyme reconstituted from purified Escherichia coli components (12,13), the heterologous E 54 holoenzyme reconstituted from the C. crescentus 54 and the E. coli core RNAP (14), or a partially purified C. crescentus RNAP (15). The genes encoding three C. crescentus sigma factor subunits, 32 (rpoH; Refs. 16 and 17), 54 (rpoN; Refs. 14 and 18), and 73 (rpoD; Ref. 19) have been cloned and sequenced. Although sigma factors 54 (14) and 32 (16) have now been overexpressed in E. coli and purified, the lack of a purified core RNAP has prevented the reconstitution of a transcription system exclusively from purified C. crescentus components. In addition, the principal C. crescentus sigma factor, which is required for transcription from the housekeeping promoters (19) and predicted to contain 653 amino acids with a molecular mass of 72,623 Da ( 73 ; 20), had not been isolated.
We report here the first purification and characterization of the C. crescentus core RNAP, as well as the two holoenzymes E 73 and E 32 . We also describe the purification of the C. crescentus principal sigma factor, 73 , after overexpression of rpoD in E. coli. The E 73 , E 54 , and E 32 holoenzymes have been reconstituted exclusively from purified C. crescentus pro-teins, and the transcriptional specificity of these RNAP preparations has been examined. The availability of a defined, reconstituted transcription system will allow detailed analysis of the roles of RNAP and accessory factors in the transcriptional regulation of developmental genes during cell differentiation and division in this bacterium.

MATERIALS AND METHODS
Bacterial Strains, Media, and Materials-E. coli strain DH5␣ was used for propagating plasmids and cultured in ML medium supplemented with ampicillin (100 g/ml) or tetracycline (10 g/ml) as necessary. C. crescentus wild-type strain CB15 (ATCC19089) was used for the purification of RNAP and grown in PYE (peptone yeast extract; Ref. 21 RNA Polymerase Purification-Most procedures of the purification are based on the methods described by Burgess and Jendrisak (22). All steps were carried out at 4°C unless noted otherwise. A block of 100 g of frozen C. crescentus cells were broken into small pieces and placed in a 1-liter Warring Blender with 300 ml of grinding buffer (0.05 M Tris-HCl (pH 7.9), 5% (v/v) glycerol, 2 mM EDTA, 0.1 mM dithiothreitol, 1 mM ␤-mercaptoethanol, 0.233 M NaCl, 23 g/ml phenylmethylsulfonyl fluoride, and 130 g/ml lysozyme). The cells were blended to allow lysis and shearing of the DNA. The sample was diluted with 500 ml of TGED (0.02 M Tris-HCl (pH 7.9), 5% (v/v) glycerol, 0.1 mM EDTA, and 0.1 mM dithiothreitol) ϩ 0.2 M NaCl, blended, and then centrifuged for 30 -40 min at 7000 rpm. The supernatant was collected as crude extract.
Crude extract was precipitated with Polymin P at a final concentration of 0.3% and centrifuged at 7000 rpm in Sorvall to collect the pellet. Proteins were eluted from the pellet with TGED ϩ 1 M NaCl after washing the pellet once with TGED ϩ 0.2 M NaCl. The Polymin P extract was then precipitated with ammonium sulfate at 50% saturation. The pellet obtained after centrifugation was resuspended in TGED and dialyzed twice against TGED ϩ 50 mM NaCl.
Dialyzed sample was subjected to column chromatography as follows. It was first applied to a heparin-agarose column pre-equilibrated with TGED ϩ 50 mM NaCl and eluted with a NaCl gradient of 50 -600 mM in TGED. Fractions were examined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and peak fractions containing RNA polymerase (␤␤Ј and ␣ subunits), as visualized by Coomassie Blue staining (Fig. 1A), pooled and dialyzed as above. Dialyzed peak fractions were further purified on a DEAE-cellulose column by fractionating with a linear NaCl gradient of 50 -600 mM. After dialysis, the pooled peak fractions (Fig. 1B) were finally fractionated on a singlestranded DNA cellulose column also with a linear NaCl gradient of 50 -600 mM.
Two peak fractions, peak 1 and peak 2 (Fig. 1C), were pooled separately and dialyzed against storage buffer (TGED ϩ 0.05 M NaCl with 50% (v/v) glycerol). The dialyzed enzyme preparations were aliquoted and stored at Ϫ80°C for future use.
Isolation and Purification of Sigma Factors from RNA Polymerase Preparation-Samples of RNAP holoenzyme were applied to SDS-PAGE gels (10% acrylamide), and after staining with KCl the gel was sectioned into 8 slices corresponding to the positions of visible proteins. Proteins were eluted from the gel slices, and SDS was removed by acetone precipitation, dissolved in 6 M guanidine HCl, and renatured, as described by Hager and Burgess (23).
Construction of DNA Templates Used for in Vitro Transcription-Plasmid pNEO containing the promoter of neomycin phosphotransferase neo gene was constructed by cloning the HindIII/BglII fragment of transposon Tn5 (24,25) in pUC18 restricted with HindIII and BamHI. The linear fragment obtained by restricting with HindIII and EcoRI was used as a template in the in vitro transcription experiment, and the length of the expected transcript is 84 nucleotides (Fig. 2). The fragment containing the lacUV5 promoter-operator region of E. coli lac gene was cloned from the expression vector pINIIA2 (26) as a 305-bp HinfI/BamHI fragment. The DNA fragment from plasmid pINIIA2 was subcloned into pUC18 to yield plasmid pLAC UV5 . The DNA template was obtained from pLAC UV5 with HindIII and BamHI that should produce a 79-nt transcript ( Fig. 2; 26). Plasmid pAKC8 was constructed as described previously (13). The fljK transcription start site lies 451 bp upstream of the T7 terminator site (Fig. 2). Supercoiled plasmid DNA was used in the one-cycle transcription assay to determine the 54 activity (13). Plasmid pJW012 containing the dnaKP1 32 -dependent promoter has been described previously (16). The SacI/SalI DNA fragment was used as template in the run-off transcription assay to test the 32 activity. Plasmid pSSA41 was a gift of John Smit. The HindIII/ClaI DNA fragment containing the rsaA gene promoter was inserted into HindIII/AccI sites of pUC19. The HindIII/NarI fragment from this plasmid was used as template in the run-off transcription assay to determine the 73 activity. A 600-bp HindIII/BamHI fragment containing the C. crescentus pleC gene promoter was subcloned into HindIII and BamHI sites of pBluescript to yield plasmid pJW014. The 636-bp HindIII/SalI fragment from this plasmid was used as a DNA template in the run-off transcription assay to test the 73 activity.
In Vitro Transcription Assays-The formation of open complexes at different promoters from C. crescentus and E. coli was measured in single cycle or run-off transcription assays, as described previously (13,16). The RNA polymerase holoenzymes were reconstituted by the addition of purified C. crescentus sigma factors to the purified C. crescentus core RNA polymerase (final concentration of 1 nmol) and incubated for 10 min at 4°C. Transcripts obtained were fractionated on 7 M urea-PAGE with end-labeled Sau3A fragments of pUC18 as DNA size markers and visualized by autoradiography.
Assay for Core RNAP-The core RNAP activity was determined using poly[d(A-T)] as template based on the method described by Berg et al. (27). Each reaction of a 100-l volume contained 20 l of assay solution (200 mM Tris-HCl (pH 8.0), 50 mM MgCl 2 , and 50 mM ␤-mercaptoethanol), 100 M ATP, 10 M UTP, and 5 Ci of [␣-32 P]UTP. RNA synthesis was initiated by adding RNA polymerase and terminated after 10 min at 37°C with 3 ml of ice-cold 3.5% perchloric acid containing 0.1 M sodium pyrophosphate. The precipitates were collected on Whatman cellulose filter paper (3MM) and washed three times with cold 1 M HCl containing 0.1 M sodium pyrophosphate and finally with cold ethanol. The radioactivity on the dry filter was determined by scintillation counting.
Overexpression and Purification of C. crescentus 73 Protein from E. coli-The cloned rpoD gene was identified and cloned 2 by colony hybridization to its Myxococcus xanthus homologue (28). DNA sequence on analysis confirmed that the open reading frame in this recombinant plasmid was identical to that published for the C. crescentus rpoD gene (20). A SmaI-SstI DNA fragment from plasmid pGIR210, which contains the rpoD gene, was subcloned into the mutagenesis vector pAlter-1 (Promega), and an NdeI restriction site was introduced at the first codon ATG of the rpoD gene open reading frame by site-directed mutagenesis. The 2.7-kilobase pairs NdeI-HindIII DNA fragment was then subcloned into the NdeI and HindIII sites of the expression vector pRSET(A) to yield plasmid pJW41 in which the entire open reading frame of rpoD gene is translationally fused to the first codon of the T7 gene 10. E. coli strain BL21 (DE3) carrying the plasmid pJW41 was used to overproduce 73 protein in the presence of isopropyl-␤-D-thiogalactopyranoside.
The overexpressed 73 protein was purified by a previously described method (29,30). The RpoD protein was not soluble and formed inclusion bodies that were solubilized with 6 M guanidine HCl in TGED buffer. The solubilized protein was renatured by dialysis against the TGED buffer and then further purified by chromatography on a DEAE-cellulose column. This method yielded 73 protein that was greater than 95% pure, as judged by Coomassie Blue staining of SDS-PAGE gels.

RESULTS
Purification of RNA Polymerases-Cellular RNAP from C. crescentus was purified by fractionation of cell extracts with Polymin P, ammonium sulfate precipitation, and chromatography on heparin-agarose (Fig. 1A) and DEAE-cellulose (Fig. 1B), which removed many of the contaminating proteins. Peak fractions from the DEAE-cellulose column containing the RNAP ␤, ␤Ј, and ␣ subunits were pooled and applied to a single-stranded DNA-cellulose column (Fig. 1C). As shown in Fig. 1C, the majority of the RNAP ␤, ␤Ј, and ␣ subunits eluted in two distinct peaks along with several minor proteins. The fractions in peaks 1 and 2 were pooled separately ("Materials and Methods").
Identification of E 73 and E 32 RNAP Holoenzymes-We assayed RNAP activity in peak 1 and peak 2 using E. coli DNA templates containing either the 70 -dependent neo promoter or the lacUV5 promoter ( Fig. 2; "Materials and Methods"). Both of these promoters are recognized in vivo by C. crescentus 3 (see below). A third DNA template (Fig. 2) contained the 32 -dependent, dnaKP1 heat-shock promoter of C. crescentus (31). The RNAP preparation from peak 2 recognized all three promoters, and specific transcripts of the predicted sizes were obtained from each template (Fig. 3, lanes 2, 4, and 6). Thus the peak 2 preparation contained 32 holoenzyme (E 32 ) and 73 holoenzyme (E 73 ) activities. No detectable transcripts were observed when the RNAP preparation from peak 1 was assayed using the same DNA templates (Fig. 3, lanes 1, 3, and 5). These data suggest the possibility that peak 1 contained either inactive RNAP holoenzyme or only core RNAP.
Identification of C. crescentus Core RNAP-We assayed for RNAP core activity directly in an in vitro transcription assay using poly[d(A-T)] as template as described by Berg et al. (27). The results summarized in Table I indicate that both peak 1 and peak 2 contained active core polymerase, although the specific activity was ϳ3-fold higher in the peak 2 RNAP. Interestingly, peak 2 RNAP was also more active on the poly[d(A-T)] template than purified E. coli core polymerase. We next examined if the peak 1 RNAP core preparation could be reconstituted to give active holoenzyme. C. crescentus 54 protein was used in the initial reconstitution experiments because earlier work had demonstrated that E 54 RNAP holoenzyme reconstituted from E. coli RNAP core and purified C. crescentus 54 specifically recognizes 54 -dependent promoters from both E. coli and C. crescentus (14). The reconstituted E 54 holoenzyme also required the activator protein FlbD for initiation of transcription from C. crescentus 54 -dependent promoters, as had been observed both in vitro and in vivo (13). The work described here demonstrates that in the presence of C. crescentus 54 and FlbD, the purified RNAP preparation from peak 1 specifically recognized the 54 -dependent fljK promoter and produces a transcript of the expected size from this template ( Fig. 2 and Fig. 4, lanes 2 and 4). This result confirms that peak 1 fractions contain an active RNAP core enzyme that can be used for assays of sigma factor activity. In the absence of added 54 , however, RNAP in neither peak 1 nor peak 2 recognized the fljK promoter (Fig. 4, lanes 1 and 3), indicating that none of these fractions contained an active E 54 holoenzyme. Consequently, we refer to the peak 1 pool as core RNAP. The peak 2 pool, which appears to contain the E 73 and E 32 holoenzymes (Fig. 3), as well as excess core RNAP (see below), we refer to as peak 2 RNAP.
Isolation and Identification of Sigma Factor Subunits from the RNAP Holoenzyme Preparation-Individual proteins in the peak 2 RNAP preparation ( Fig. 1C; peak 2 pool) were isolated after electrophoresis on preparative SDS-PAGE gels as de- scribed by Hager and Burgess (23). The gel was sectioned into 8 slices corresponding to the positions of bands visualized with KCl, with slice 1 containing bands at the top of the gel. Proteins were eluted from the gel slices, and a portion of each eluted protein sample was then analyzed on a second SDS-PAGE gel (Fig. 5). The peak 2 RNAP preparation contained several proteins in addition to core subunits ␤ and ␤Ј isolated from slice 1 and ␣ isolated from slice 4. Potential sigma factors were the 75-kDa protein in slice 2 and the 34-kDa protein in slice 5. Unidentified proteins A and B of molecular masses ϳ55 and 50 kDa, respectively, were found in slice 3, and a third unknown protein C of ϳ28 kDa was detected in slice 6.
A portion of the proteins eluted from the SDS-PAGE gel slices were renatured (see "Materials and Methods") and combined with C. crescentus core RNAP (peak 1) to determine their ability to direct transcription. Assay of the 34-kDa protein renatured from gel slice 5 (Fig. 5) on the dnaK P1 template produced a run-off transcript of the expected size (62 nt; Fig. 2), which is consistent with the identification of this protein as the 32 factor (16). When renatured proteins from gel slices 1-8

FIG. 2.
Templates used for in vitro transcription assays. Construction of the six DNA templates used in the transcription assays is described under "Materials and Methods." The purified DNA fragments indicated were used in the runoff transcription assays. The supercoiled plasmid containing the fljK promoter was used in one-cycle transcription assays. The predicted sizes of individual transcripts from the templates are indicated in nucleotides (nt). 32 . The activities of holoenzymes in peak 1 RNAP (lanes 1, 3, and 5) and peak 2 RNAP (lanes 2, 4, and 6) were determined in run-off transcription assays with DNA templates containing the 70 -dependent neo (lanes 1 and 2) or lacUV5 (lanes 3 and 4) promoters from E. coli and the 32 -dependent promoter, dnaK P1 (lanes 5 and 6) from C. crescentus, as described previously (16).  4. Identification of core RNA polymerase. The C. crescentus core RNAP activity was determined in in vitro transcription assays by reconstituting E 54 activity from the purified C. crescentus sigma factor 54 and its activator FlbD. The RNAP preparations from pooled peak 1 (lanes 1 and 2) and peak 2 (lanes 3 and 4) were assayed on supercoiled templates containing the 54 -dependent promoter of flagellin gene fljK, as described (13). Assays were carried out with (lanes 2 and 4) or without (lanes 1 and 3) the purified 54 .

TABLE I Core RNA polymerase activities assayed on poly [d(A-T)] templates
were assayed individually in the presence of the peak 1 core RNAP using the neo template, a run-off transcript was detected only in the assay containing proteins from slice 2 (Fig. 6). This transcript was of the size expected (84 nt; Fig. 2) from the 70 -dependent neo promoter. A transcript of the same size was produced by peak 2 RNAP but not by peak 1 core RNAP alone (Fig. 6). No transcriptional activity was detected by proteins eluted from slice 2 when they were assayed with peak 1 core RNAP on DNA templates with 32 -or 54 -dependent promoters. 4 These results indicate that the 75-kDa protein isolated from peak 2 RNAP preparation specifically recognized the E. coli housekeeping promoters and is a functional homologue of the E. coli principal sigma factor, 70 . Micro-sequencing of this protein yielded the amino-terminal sequence (M)NNSSAETE, which is identical to that of the translated DNA sequence of the C. crescentus rpoD gene (20).
Isolation and Purification of the C. crescentus rpoD Gene Product-To further characterize the principal C. crescentus sigma factor identified in the reconstitution experiments (Fig.  6), we overexpressed the rpoD gene in E. coli and purified the full-length 73 protein to near-homogeneity (see "Materials and Methods"; Fig. 7A; lane 3). The C. crescentus rpoD gene has been shown to encode a predicted polypeptide of 653 amino acids with a molecular mass of 72,623 Da and designed as 73 (20).
The size of the overexpressed protein (Fig. 7A, lane 3) is similar to the very faint protein band at ϳ75-kDa in the peak 2 RNAP preparation (Fig. 7A, lane 2) and close to the predicted 72,623-Da size of the rpoD gene product (20). The purified protein was also examined by Western blot analysis. An anti-E. coli 70 antibody cross-reacted with the major protein band at ϳ75 kDa, as well as with several smaller bands that presumably result from proteolysis of RpoD (Fig. 7B, lane 3). The anti-70 antibody can also recognize the protein at ϳ75 kDa present in the peak 2 RNAP (Fig. 7B, lane 2) but failed to recognize any proteins in the peak 1 RNAP preparation (Fig. 7B,  lane 1). These results further support our assignment of peak 1 as core enzyme and the 75-kDa protein present in the peak 2 RNAP as 73 . The 73 present in the peak 2 RNAP (Fig. 7B, lane  2) displays a slightly different mobility from that overproduced from the C. crescentus rpoD gene in E. coli (Fig. 7B, lane 3), perhaps as a result of 73 modification in one of the bacteria.
Functional Analysis of C. crescentus 73 in Vitro-We examined the sigma factor activity of the purified rpoD gene product in reconstitution experiments with core RNAP using in vitro transcription assays (Fig. 8). Purified 73 protein directed transcription from the 70 -dependent promoter of E. coli neo gene in the presence of peak 1 core RNAP (Fig. 8, lane 2), whereas the core enzyme alone did not (Fig. 8, lanes 1). These results and those in Fig. 6 demonstrate that the C. crescentus RpoD protein isolated either from the peak 2 RNAP preparation or E. coli cells overexpressing the cloned the C. crescentus rpoD gene recognizes E. coli 70 -dependent promoters, suggesting that the C. crescentus 73 is a functional homologue of the principal E. coli sigma factor 70 . 4 J. Wu and A. Newton, unpublished observations.  6. Determination of the C. crescentus 73 activity. Individual proteins from the peak 2 RNAP preparation were isolated by SDS-PAGE ( Fig. 5; slices 1-8) and assayed after renaturation for stimulation of transcription in the presence of core RNAP (peak 1 RNAP). The activity of reconstituted E 73 holoenzyme was determined by its ability to recognize and transcribe from the 70 -dependent E. coli neo promote. We next examined the ability of purified 73 to confer transcriptional specificity in the recognition of two promoters, rsaA (32,33) and pleC (34; Fig. 2), that have been used to define the 73 promoter consensus for C. crescentus (19). Both of these C. crescentus promoters have been characterized in vivo (19) and shown to contain a Ϫ35 consensus sequence similar to that in E. coli and a Ϫ10 consensus divergent from that in E. coli. The rsaA and pleC promoters were recognized in the in vitro transcription assays by the reconstituted E 73 holoenzyme (Fig. 8,  lanes 4 and 6). The more efficient transcription from the neo promoter in these experiments (Fig. 8, lane 2) is consistent with measurements of promoter strength in vivo using transcription fusions. ␤-Galactosidase assays of C. crescentus wild-type strains carrying either the neop-lacZ (6876 Miller units) or the rsaAp-lacZ (1545 Miller units) fusions indicates that the neo promoter is 4-to 5-fold stronger than the rsaA promoter under these conditions. 4 The sizes of the rsaA and pleC transcripts observed in vitro (Fig. 8) were those expected from the transcription start sites mapped in vivo for the two genes, i.e. 110 and 62 nt, respectively ( Fig. 2; Ref. 19). Transcription initiation was RpoD-dependent, since the peak 1 core RNAP alone did not recognize either promoter (Fig. 8, lanes 3 and 5). These results confirm that the purified rpoD gene product, 73 , is the principal C. crescentus sigma factor and that it is capable of recognizing the C. crescentus housekeeping gene promoters, as well as E. coli 70 -dependent promoters.

DISCUSSION
Many developmental events in C. crescentus are dependent on differential gene expression regulated at the level of transcription. Unlike Bacillus subtilis, where an extensive cast of alternative sigma factors and regulatory proteins governing sporulation have been identified through genetic and biochemical analysis (reviewed in Ref. 8), a defined in vitro transcription system has not been available in C. crescentus. The purification of C. crescentus RNAP was described in early studies (35,36), but the transcriptional specificity of these enzyme preparations was not characterized. More recently, a partially purified RNAP preparation was used for the study of class II flagellar gene regulation (15). However, this is the first report of the purification and resolution of the holoenzymes and core RNAP and the reconstitution of RNAP holoenzymes exclusively from C. crescentus components.
Heparin-agarose chromatography, which has been used successfully for isolation of RNA polymerase from a number of bacterial species, including E. coli, B. subtilis, B. stearothermophilus, Lactobacillus casei, L. plantarum, and Clostridium pasteurianum (reviewed in Ref. 4), provided a great enrichment of C. crescentus RNAP (Fig. 1A). Chromatography on singlestranded DNA agarose (37) or on phosphocellulose (38) has also been reported to resolve E. coli holoenzyme and core RNAP. In our hands single-stranded DNA-cellulose chromatography was crucial for resolving C. crescentus RNAP into its core and holoenzyme fractions (Fig. 3C). Phosphocellulose and Bio-Rex 70 were not effective in resolving core and holoenzyme, although these earlier attempts were hampered by the lack of a purified sigma factor to assay for core activity. 5 The fact that the first peak from the DNA-cellulose column (peak 1; Fig. 1C) contained active core RNAP and the second peak (peak 2; Fig. 1C) contained E 32 and E 73 holoenzymes, as well as core RNAP (see below), was demonstrated by in vitro transcription assays of the pooled fractions ( Fig. 3; Table I) and the isolation of active 32 and 73 from these RNAP fractions of peak 2 ( Fig. 6; 16). In the presence of purified C. crescentus 54 and its activator protein FlbD, core RNAP from peak 1 recognized the 54 -dependent promoter of the C. crescentus flagellin gene fljK (Fig. 4), as observed previously for a heterologous holoenzyme containing either C. crescentus 54 or E. coli 54 and the E. coli core RNAP (13)(14)(15). Therefore, C. crescentus core RNAP appears to function interchangeably with its E. coli counterpart in this transcription assay.
Our purified peak 2 RNAP preparation displayed activity only on 32 -and 73 -dependent promoters. Possible explanations for the failure to recover active E 54 include (i) an unstable 54 protein that is inactivated during purification and (ii) low affinity of 54 for binding to core RNAP, which results in its dissociation from the core and loss early in protein fractionation. Consistent with the latter possibility are two observations. First, some bacterial RNAP holoenzymes are quite unstable and dissociate early during procedures suitable for isolation of other RNAP holoenzymes (39), and second, fulllength E. coli 54 does not bind to E. coli core RNAP tightly (40).
Three proteins (Fig. 5; bands A, B, and C) of unknown function are also associated with the RNAP holoenzyme fractions. These proteins could represent additional sigma factors, breakdown products of 73 , other RNAP subunits, such as delta or omega, or proteins that fortuitously fractionate with RNAP. Protein C (Fig. 5), like 32 (16) and 73 , has been isolated and subjected to amino acid sequencing, but unlike the latter two proteins, there is no similarity of its amino-terminal amino acid sequence to any sequences deposited in GeneBank. 4 It will be interesting to determine whether any of these three proteins, A, B, and C, are RNAP subunits or accessory proteins that are involved in transcriptional regulation.
The relative amounts of 73 and 32 as visualized by the intensity of staining in SDS-PAGE gels displayed variability depending on the purified preparation examined, but we estimated that pooled peak 2 RNAP contained less than 0.4 mol eq of total factor relative to the core subunits (see Fig. 5). This result suggests that peak 2 RNAP also contains core enzyme. 5 N. Ohta, A. Ninfa, and A. Newton, unpublished observations. The E 73 was reconstituted by mixing the purified C. crescentus peak 1 core RNAP and purified C. crescentus 73 protein in vitro. The E 73 activity was determined by run-off transcription assays on DNA templates containing the 70 -dependent neo promoter from E. coli (lanes 1 and 2) or the pleC (lanes 3 and 4) and rsaA (lanes 5 and 6) housekeeping gene promoters from C. crescentus.
Consistent with the presence of excess core in the peak 2 RNAP fractions was the stimulation of transcription from the 54 -dependent fljK promoter by the addition of purified 54 and FlbD to the peak 2 pool (Fig. 4) and the high activity of this RNAP pool when assayed on the poly[d(A-T)] template (Table I).
The above results raise the question of why the core enzyme elutes from the single-stranded DNA column in two peaks. One possibility is that the more active core enzyme (Table I) binds more tightly to the column and elutes with the holoenzymes in peak 2. Alternatively, the core eluted in peak 2 could represent core that has bound tightly as the holoenzyme from which sigma factors elute at lower salt concentrations than the other polymerases in those complexes. Whatever the explanation of this fractionation, our experiments have depended critically on the isolation of the functional core RNAP. The inability to resolve different RNAP holoenzymes from one another 5 prompted us initially to overexpress and purify C. crescentus factors, including 54 (14), 32 (16), and 73 (Fig. 7) for the reconstitution of specific RNAP holoenzymes reported in this study.
C. crescentus recognizes E. coli 70 -dependent promoters in vivo (20), and our results demonstrate that the purified (Figs. 3 and 6) and reconstituted (Fig. 8) C. crescentus E 73 efficiently recognized the lacUV5 and neo promoters from E. coli, as well as promoters of the C. crescentus housekeeping genes rsaA and pleC (Fig. 8). The Ϫ35 consensus sequence of the C. crescentus biosynthetic and housekeeping gene promoters (19) is similar to the Ϫ35 consensus of E. coli 70 -dependent promoters, but the Ϫ10 sequences from these two bacteria align only poorly. Moreover, the C. crescentus Ϫ10 and Ϫ35 sequences are more closely spaced than in most E. coli promoters (19). These results suggest that the principal C. crescentus sigma factor 73 has less promoter specificity than its E. coli counterpart 70 .
The availability of core RNAP from C. crescentus and the capability of reconstituting active holoenzymes using purified factors will permit biochemical analysis of gene regulation in vitro in great detail. These reagents will also be important in studying the role of RNAP and accessory factors in the temporal and spatial regulation of developmental gene transcription during the cell division and differentiation.