Purification and Characterization of RNA Polymerase from the Cyanobacterium Anabaena 7 120*

A procedure for the purification of RNA polymerase from vegetative cells of the filamentous cyanobacterium Anabaena 7120 is described. Polyethyleneimine precipitation followed by gel filtration and affinity chromatography steps results in >99% purification with 46% yield. The enzyme has a novel core compo-nent of M, = 66,000, designated y, in addition to the typical prokaryotic ,8’,8nz core enzyme. The u subunit has been identified by reconstitution of specific tran- scriptional activity from core enzyme and gel-purified U. In transcription assays, this RNA polymerase initi- ates at a number of Anabaena vegetative cell promoters, as well as from a bacteriophage T4 early pro- moter, but does not initiate at nitrogen fixation (nif) promoters used in heterocysts. The promoter specific-ity of Anabaena RNA polymerase is compared with that of Escherichia coli RNA polymerase. Anabaena

Anabaena is one of the filamentous cyanobacteria that differentiate specialized cells, called heterocysts, at regular intervals along each filament in response to starvation for reduced nitrogen. Heterocysts are the sites of nitrogen fixation; they reduce nitr0ge.n to ammonia and export it in the form of glutamine to neighboring vegetative cells. Differentiation of vegetative cells into heterocysts involves the synthesis of unique glycolipid and polysaccharide components of the heterocyst envelope, the destruction of the oxygen-evolving photosystem I1 apparatus and the carbon-fixing enzymes, and the induction of nitrogenase and the enzymes of the oxidative pentose pathway. Analysis of protein synthesis in differentiating heterocysts has shown that there are many sets of proteins made a t different stages of development (1). For certain genes, which have been cloned, expression has been correlated with the presence or absence of the corresponding mRNA. Thus, a significant aspect of Anabaena heterocyst differentiation is the regulation of transcription of specific genes (2).
Among the genes that have been cloned and sequenced, there are examples of genes that are transcribed only in vegetative cells, genes transcribed only in heterocysts, and genes transcribed in both types of cells. An example of the first are the rbcL and S genes, encoding the large and small subunits of ribulose-bisphosphate carboxylase, respectively. The 3-kilobase transcript of this operon is abundant in ammonia-grown cells but absent from cells starved for nitrogen (3,4). On the other hand, the nitrogen fixation (nif) gene transcripts are found only in cells that are both anaerobic and starved for reduced nitrogen ( 5 ) . Finally, the glnA gene encoding glutamine synthetase is transcribed abundantly under both conditions. The glnA gene has multiple promoters used under different circumstances. In cells growing on ammonia, the promoters used are Escherichia coli-like. In cells starved for reduced nitrogen, a different promoter like that of the nifHDK operon is used instead (see Ref. 6 and Fig. 7).
These observations suggest that at least part of the control of transcription during heterocyst differentiation involves modification of the transcription apparatus. In order to begin to understand that control, we have characterized RNA polymerase from Anabaena vegetative cells grown with ammonia as the source of fixed nitrogen.

EXPERIMENTAL PROCEDURES
RNA Polymerase Assay-To follow activity during purification, the assay contained, in a volume of 35 PI, 50 mM Tris, p H 8.0, 0.05 mM EDTA, 2.5 mM 0-mercaptoethanol, 10 mM MgC12, 0.5 mM CTP, GTP, UTP, 0.125 mM ATP, [8-3H]ATP (5 X lo4 cpm/nmol), and 50 pg/ml chicken erythrocyte DNA. RNA synthesis was initiated by addition of 5 p1 of an enzyme fraction followed by incubation at 37 "C for 10 min. The reaction mixture was spotted onto Whatman GF/C filters and washed three times with 5% trichloroacetic acid. The filters were then dried and radioactivity determined by liquid scintillation counting in a toluene mixture containing 0.5% (w/v) 2,5diphenyloxazole. One unit of RNA polymerase activity corresponds to the incorporation of 1 nmol of AMP in 10 min a t 37 "C.
Anabaena RNA Polymerase Purification-RNA polymerase was purified from ammonia-grown Anabaena 7120 with a final yield of 46% (Table I). Cells were grown to medium density in Kratz and Meyers (7) medium supplemented with 2.5 mM (NH,),SO, under fluorescent lights with stirring and bubbling of 1% CO, in air. The yield from a 15-liter culture was 15-20 g wet weight of cells. All steps of the purification were carried out at 4 "C. The cells were harvested with a Sorvall continuous flow rotor, washed in buffer G (20 mM Tris, pH 8.0, 10 mM MgCI2, 0.1 mM EDTA, 0.1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride) containing 50 mM KCI, resuspended in 40 ml of buffer A (same as buffer G plus 10% glycerol) containing 50 mM KCI, then passed through a cold French pressure cell twice at 20,000 p.s.i. The cell lysate was spun for 90 min a t 120,000 X g, and RNA polymerase was precipitated from the supernatant by addition of PEI' (Sigma) at pH 8.0 to 0.05% (8) and spun for 20 min a t 27,000 x g. The pellet was washed with 25 ml of buffer A containing 400 mM KC1 by resuspension with a Teflon homogenizer and repelleted for 15 min a t 27,000 X g. The washed pellet was dissolved in 10 ml of The abbreviations used are: PEI, polyethyleneimine; BSA, bovine serum albumin; SDS, sodium dodecyl sulfate. buffer A containing 2.0 M KC1, respun, and the PEI was removed by three ammonium sulfate precipitations, each time resuspending the pellet in buffer A containing 2.0 M KC1 and then adding 2.3 volumes of a saturated ammonium sulfate solution. The final pellet was resuspended with 5 ml of buffer A containing 400 mM KC1 and dialyzed against the same buffer.
Gel filtration on a 2.8 X 83-cm Bio-Gel A-1.5m (Bio-Rad) column was carried out in the same buffer. The peak of activity was pooled, precipitated by adding solid ammonium sulfate to 60% saturation, resuspended, and dialyzed against buffer A containing 200 mM KCl. The protein was loaded onto a 1.5 X 7-cm heparin-Sepharose CL-GB (Pharmacia Biotechnology, Inc.) column, washed with 2 column volumes of buffer A containing 200 mM KC1, and then eluted with buffer A containing 300 mM KCI. Pure RNA polymerase was precipitated with 60% saturated ammonium sulfate, resuspended, and dialyzed against buffer A containing 50 mM KC1 and 50% glycerol for storage at -20 "C. Core and holoenzyme were separated on a 1 x 6cm Bio-Rex 70 (Bio-Rad) column. After loading up to 3 mg of protein in buffer A containing 50 mM KC1 and washing with 2 column volumes of the same, core enzyme was eluted with buffer A containing 200 mM KC1 and holoenzyme with buffer A containing 500 mM KC1.
u Subunit Purification-The u subunit was isolated from an SDSpolyacrylamide gel containing 1.25 mg of a Bio-Rex 70 holoenzyme pool and renatured as described (9). In reconstitution assays, core enzyme and u subunit were incubated together for 10 min on ice before addition of DNA template and nucleotides and transfer to 37 "C.
E. coli RNA Polymerase Purification-RNA polymerase from E. coli was purified from 3 g wet weight of cells through the heparin-Sepharose step of the same method described for Anabaena RNA polymerase. This yielded 450 pg of enzyme of >99% purity.
Protein Determination-Protein concentration was estimated with Bio-Rad protein assay reagent, using BSA as standard. Coomassiestained gels were scanned with an LKB 2202 Ultrascan laser densitometer and recorded with a Hewlett-Packard 3390 A integrator to quantitatively compare amounts of B subunit and to measure subunit stoichiometry.
Polyacrylamide Gel Electrophoresis-Proteins were analyzed by SDS electrophoresis in 12.5% acrylamide, 0.1% bisacrylamide running gel with a stacking gel of 5% acrylamide, 0.15% bisacrylamide. Silver staining of gels was performed according to Dion and Pomenti (10). After fixing gels in 10% methanol, 10% acetic acid for 30 min, gels were washed in 5% glutaraldehyde for 30 min and then washed with four changes of water for 20-30 min each. Gels were stained for 15 min with fresh ammoniacal silver made by adding 8.0 ml of 19.4% AgN03, dropwise with stirring, to 43 ml of 0.36% NaOH plus 2.8 ml of concentrated NH,OH; water was added to 200 ml. Gels were washed with two changes of water for 5 min each and then developed with 0.0185% formaldehyde, 0.005% citrate. Color development was stopped with 1% acetic acid before drying between sheets of cellophane. Gels stained with Coomassie Blue were first fixed for 15 min in 5% acetic acid, 10% methanol. Gels were stained for 15 min in the same solution plus 0.1% Coomassie Blue for 15 min and destained with 7% acetic acid. Fixing, staining, and destaining washes were all carried out in a 55 "C water bath (11). RNA synthesis products were separated on 7 M urea-polyacrylamide gels (12).
Glycerol Gradient Centrifugation-Purified Anabaena RNA polymerase was centrifuged at 4 "C through a 5-ml10-30% glycerol gradient made in buffer A containing 500 mM KC1 for 21 h a t 40,000 rpm in a SW 41 rotor. Fractions of 0.25 ml were collected and assayed for RNA polymerase activity and for absorbance at 280 nm. Sedimentation markers of catalase and B-galactosidase were run in parallel gradients.
P h m i d Template Construction-DNA fragments containing promoters of interest were cloned into the plasmid pTE103, which contains the bacteriophage T7 early terminator downstream from a multiple cloning site (13). The 2.3-kilobase EcoRI fragment from pTE55 containing a bacteriophage T4 early promoter (14) was inserted into the unique EcoRI site of pTElO3 to generate the plasmid pT4110. Digestion with HincII was used to determine the orientation of the insert. The 440-base pair EcoRI fragment from pAn503 which contains the promoters for the Anabaena glnA gene (6) was inserted into the same EcoRI site to generate pAn510. The 400-base pair HpaI fragment from pAn602 which contains the promoter for the rbcLS operon (3,15) was cloned into the unique SmaI site of pTE103 to generate pAn610. HinfI digestion was used to determine orientation of the inserts in pAn510 and pAn610 (Fig. 5).
RNA synthesis from circular plasmid templates ( Fig. 6) was performed in an identical manner except that RNA polymerase was allowed to bind to the DNA template for 10 min at 37 "C before the addition of heparin to a final concentration of 100 pg/ml. After a further 10 min at 37 "C, nucleotides were added to the reaction mixture, and the reactions were carried out as described above.

RESULTS
Purification and Determination of Subunit Structure-The procedure used to purify Anabaena RNA polymerase combined the techniques of polyethyleneimine precipitation of DNA protein complexes, gel filtration chromatography, and heparin-Sepharose affinity chromatography to yield a homogeneous enzyme preparation. A high speed spin to clear many of the photosynthetic membranes from the cell lysate was an important step in purification of RNA polymerase from Anabaena. After PEI precipitation, three ammonium sulfate precipitations were necessary to remove the PEI from the 2.0 M KC1 eluate to allow full solubility on resuspension in a lower salt buffer. As shown in Table I, PEI precipitation was a very efficient step with 95% recovery of RNA polymerase in only 6% of the original protein. Gel filtration on Bio-Gel A-1.5m excluded any remaining membrane fragments that eluted in a yellow peak in the void volume. The preparation at this stage contained only one major and a few minor contaminants (Fig. 1, lane 3 ) . Affinity chromatography on heparin-Sepharose yielded a homogeneous enzyme preparation (Fig. 1, lane 4). Washing the heparin-Sepharose column with buffer A containing 2.0 M KC1 immediately before equilibration with buffer A containing 200 mM KC1 resulted in more reproducible elution of the enzyme.
At this point in the preparation, the proteins labeled as p', p, y, U , and O( in Fig. 2 had all co-eluted, in unchanging ratios, with RNA polymerase activity. The proteins D', p, and a were named by analogy to other prokaryotic RNA polymerases. The protein at molecular weight 52,000 was designated u on the basis of its lower stoichiometry. The protein at molecular weight 66,000 was named y. The other protein bands just below the p' and p subunit bands were determined to be proteolysis products of these subunits, because their relative  During this elution, the y protein eluted exactly with the other RNA polymerase core subunits, in agreement with the assertion that y is a part of Anabaena core enzyme (Fig. 2). In run-off transcription assays using a bacteriophage T4 early promoter (see below), the core enzyme lost promoter-specific activity, whereas holoenzyme retained this activity. T o demonstrate that the protein a t M , = 52,000 was responsible for this promoter-specific transcriptional activity, the protein was renatured from an SDS-polyacrylamide gel. Fig. 3 shows gel analysis of core and holoenzyme pools, gel-purified (I, and the result of a holoenzyme reconstitution experiment. The core enzyme and u fractions separately show no transcriptional activity on the bacteriophage T 4 early promoter (Fig. 3B,  lanes 2 and 3), but transcription increases as u is added to the core (lanes [4][5][6][7][8]. Note that the @ (or @') proteolysis band a t M , = 106,000, present in the Bio-Rex 70 holoenzyme fraction, is absent from the core enzyme fraction and thus does not contribute to the RNA polymerase holoenzyme activity reconstituted from that core enzyme fraction and u. The proteins in the Anabaena RNA polymerase complex and their molecular weights were thus determined to be b' = 171,000; @ = 124,000; y = 66,000; u = 52,000; and LY = 41,000.
Chicken erythrocyte DNA was used as a template in all of the assays during purification of the enzyme. Gel analysis of column fractions showed that assayed activity corresponded exactly with elution of the core components. Not until chromatography on Bio-Rex 70 were core and holoenzyme distinguished using this assay, when the core enzyme peak had a lower specific activity (data not shown).
The sedimentation coefficient of the Anabaena enzyme was determined by centrifugation through a glycerol gradient containing 0.5 M KCl, using catalase and @-galactosidase as markers. The observed sedimentation coefficient, 14.6 S, is nearly the same as that of E. coli holoenzyme (14 S) (16). in the plasmid, generating RNA transcripts of a defined length from a circular supercoiled template. The plasmid templates are diagrammed in Fig. 5 . For the three templates shown, the start sites determined by S1 nuclease protection and primer extension on in vivo RNA are indicated, along with the sizes of transcripts expected for each promoter, assuming that transcription terminates at the T7 terminator.

Comparison of Anabaena and E. coli RNA Polymerase
The results of an experiment comparing the transcripts synthesized by Anabaena and E. coli RNA polymerases from the bacteriophage T 4 early promoter, from the glnA promoters, and from the rbcLS promoter are shown in Fig. 6. Lune I contains the transcripts produced from the control plasmid pTE103, which has no promoter in the multiple cloning site. No RNAs less than -1 kilobase are seen. Both enzymes are very active on the bacteriophage promoter cloned into pT4110, synthesizing primarily the predicted 567-nucleotide RNA. Both enzymes also are active on the glnA and rbcLS promoters in pAn510 and pAn610, respectively, although a t a lower level. The RNAs produced from the glnA promoter region correspond to RNA 11, IV, and V detected by S1 nuclease protection, primer extension (6), and run-off transcription experiments' (see Fig. 7). pAn610 produces a 609-nucleotide band consistent with S1 nuclease, primer extension (3), and run-off transcription experiments.' In addition, there are differences in intensity of the RNA bands produced by the two polymerases. Anabaena RNA polymerase synthesizes more RNA V from the &A promoter region, in relation to RNA I1 and IV, than does the E. coli enzyme. E. coli RNA polymerase synthesizes more of the transcripts longer than 1000 nucleotides, especially noticeable in the pTE103 and pAnlO6 reactions.

DISCUSSION
We have purified RNA polymerase from Anabaena and have begun to characterize its structure and activity. We began this work in order to examine biochemically an example of bacterial differentiation, i.e. the differentiation of a subset of ammonia-starved cells into heterocysts. The subunit structure of the Anabaena enzyme is similar to that of other prokaryotes except for the presence of an additional subunit that we have called y. We have prepared antibodies to the various Anabaena RNA polymerase subunits and have determined that y is antigenically unrelated to the other subunits and also that the entire RNA polymerase is immunoprecipitated with antibody to y alone." Thus, although we do not yet * G . J   know its function, y seems to be a genuine Anabaena RNA polymerase subunit, and the holoenzyme has the structure p'pya2u. RNA polymerase has been purified from two other cyanobacteria. We believe that the protein reported as u from Anacystis nidulans (17), which was purified in a 1:1 ratio with the other core subunits and which failed to separate chromatographically from the core, is indeed analogous to the y subunit of Anabaena RNA polymerase. Western blots of A nacystis proteins probed with antisera to the Anabaena subunits confirm this assignment, and indicate a u subunit of molecular weight 50,000.3 Similarly, RNA polymerase from Fremyella diplosiphon (18) contains a protein of molecular weight 72,000 that is probably the y subunit. The molecular weight 91,000 protein present in the published F. diplosiphon preparation is possibly a contaminant that is related to a similarly sized contaminant in our early preparations, which was removed by glycerol gradient centrifugation in 500 mM KCI. The u subunit for the F. diplosiphon RNA polymerase is as yet G . J. Schneider and R. Haselkorn, manuscript in preparation.
unidentified. Table I1 compares the subunit molecular weights of these cyanobacterial RNA polymerases.
The y subunit in the Anabaena core enzyme seems to be a general feature of cyanobacterial RNA polymerases. Only eukaryotes and some of the archaebacteria possess a more complex RNA polymerase. Like the archaebacteria, cyanobacteria are ancient organisms (19); the existence of complex RNA polymerases in both suggests that the simpler bacterial RNA polymerases may have evolved from more complex ancestors (20). We note that the molecular weight of the subunit of the Anabaena enzyme is lower than that of other prokaryotes. It seems possible that the y subunit provides a function that, in E. coli, is due to the / 3 subunit. Put another way, Anabaena p + y would be the equivalent of the E. coli p.
Evolutionarily, the Anabaena @ or y proteins could be related to the E. coli p by interruption of an ancestral gene or by fusion of the two Anabaena genes. Such a situation has apparently occurred in the evolution of archaebacteria (2l), where a single B protein of extreme thermophiles is serologically related to two families of smaller proteins (B' and B") in the methanogens and halobacteria. We cannot rule out the possibility that cleavage of a P-y gene product yields the two mature p and y proteins in Anabaena.
One of the purposes for purifying Anabaena RNA polymerase is to attempt to define a promoter for vegetative cells of Anabaena. Fig. 7 shows the 5"flanking sequences of a number of start sites for transcription by the Anabaena enzyme. For the T4 early transcript, the i n vitro result shown in Fig. 6 is in agreement with the i n vivo and in vitro data of Elliot et al. (14). It is clear that both the E. coli and Anabaena enzymes select the same promoter, whose -10 and -35 sequences, underlined in Fig. 7, are very close to the consensus E. coli promoter: TATAAT in the -10 region, TTGACA in the -35 region, with 17 nucleotides between these two "boxes" (22).
The Anabaena glnA gene seems to have multiple promoters in vivo (6). These are labeled I to IV in Fig. 7. RNA prepared from ammonia-grown cells, i.e. vegetative cells, shows major transcripts beginning a t PII and Plv; the PI11 transcript is minor and the PI transcript variable. RNA prepared from cells induced for nitrogenase by starvation for ammonia under anaerobic conditions is predominantly from PI although P~v is also represented. When the Anabaena glnA gene is expressed in E. coli, the major transcript corresponds to P 1 1 ; no Pill or Plv starts are seen (6).
The in vitro results are consistent with some of the in vivo observations. First, the promoter PI, used in cells fixing nitrogen, is not recognized by either pure enzyme in vitro. This is an expected result because the 5"flanking region shows no homology with the E. coli consensus promoter. The doubleunderlined nucleotides in Fig. 7 are conserved with respect to the corresponding positions 5' to the start of the nifHDK transcript in Anabaena. A modified RNA polymerase, assisted by an activator protein, transcribes nif promoters in other diazotrophs (23), and the nitrogen regulated ( n t r ) glnA promoter in enteric bacteria (24-26). The glnA PI! promoter is recognized by both E. coli and Anabaena RNA polymerase. This corresponds to the in vivo result, indicating that this RNA is indeed a primary transcript and is not produced by cleavage of a longer RNA.
The Plll start which is relatively minor in vivo (6), is not seen in the in vitro experiments shown in Fig. 6. Those experiments used supercoiled intact plasmid DNA as tem-plate. In other in vitro transcription experiments using linear templates,2 the Plll start is seen with both enzymes, in addition to PI1 and Plv. Thus, it seems possible that the complex glnA promoter region is sensitive to the degree of DNA supercoiling as well as to the nitrogen status of the cell.
Plv is an abundantly used Anabaena start both in vitro and in vivo. It was not detected in E. coli in vivo, but is very strong in vitro with the E. coli enzyme. There was some evidence of degradation in the E. coli RNA used by Tumer et al. (Fig. 5 and Ref. 6), so it is possible that the Plv start was missed.
These results are consistent with the hypothesis that the Anabaena enzyme prefers promoters close to the E. coli consensus. Fig. 6 shows a third in vitro transcript from the glnA promoter region. This is the 664-nucleotide RNA shown starting a t Pv in Fig. 7. There is no corresponding RNA in any in uiuo RNA preparation (6). Homology with the E. coli consensus is poor (Fig. 7). We can offer no explanation for the appearance of the transcript in vitro.
The rbcLS promoter produces a single transcription start in vivo (3), shown in Fig. 7. The in vitro transcription experiment shows the same promoter selection in vitro by both enzymes. Although this promoter shows good homology with the E. coli consensus sequence in the -10 region, the fit in the -35 region is poor (Fig. 7).
Comparisons among the different templates are difficult to make. It is clear that the Anabaena enzyme much prefers the T4 promoter (Fig. 6, lane 3 ) . In unpublished experiments, we have found another powerful promoter in cyanophage N-1; this promoter is almost a perfect match with the E. coli consensus. However, with respect to promoter sequences that do not fit the E. coli consensus so closely, it is clear that the Anabaena enzyme's preferences are not the same as the E.
In light of the small number of Anabaena promoters known, and the possibility that some of them may require activators for full utilization, we have not been able to describe a typical Anabaena vegetative cell promoter beyond a resemblance to certain aspects of E. coli-like promoters. As more Anabaena promoters, especially for genes that are likely to be constitu-tively expressed, become known, the structure of Anabaena vegetative cell promoters should be resolved.