ε, a New Subunit of RNA Polymerase Found in Gram-Positive Bacteria

Abstract

RNA polymerase in bacteria is a multisubunit protein complex that is essential for gene expression. We have identified a new subunit of RNA polymerase present in the high-A+T Firmicutes phylum of Gram-positive bacteria and have named it ε. Previously ε had been identified as a small protein (ω₁) that copurified with RNA polymerase. We have solved the structure of ε by X-ray crystallography and show that it is not an ω subunit. Rather, ε bears remarkable similarity to the Gp2 family of phage proteins involved in the inhibition of host cell transcription following infection. Deletion of ε shows no phenotype and has no effect on the transcriptional profile of the cell. Determination of the location of ε within the assembly of RNA polymerase core by single-particle analysis suggests that it binds toward the downstream side of the DNA binding cleft. Due to the structural similarity of ε with Gp2 and the fact they bind similar regions of RNA polymerase, we hypothesize that ε may serve a role in protection from phage infection.

INTRODUCTION

Transcription in bacteria is a tightly regulated and cyclic process carried out by the highly conserved multisubunit enzyme RNA polymerase (RNAP). The majority of work on bacterial RNAP has been performed in the Gram-negative bacterium Escherichia coli. The core E. coli RNAP complex consists of two α subunits and single β, β′, and ω subunits (1). However, there are differences throughout the eubacteria in the subunit composition of RNAP. In the model Gram-positive organism Bacillus subtilis, core RNAP has the subunit composition α₂ββ′δω₁ω₂ (2). The δ subunit is encoded by the gene rpoE and is involved in RNAP promoter specificity and recycling (2–4). The δ subunit appears to be conserved across the Gram-positive bacteria, and despite a high cellular abundance and documented roles in gene expression, deletion strains do not exhibit a strong phenotype (3–5). The occurrence of two ω subunits in Bacillus subtilis at approximately 11 kDa (ω₁) and 9 kDa (ω₂) has been known for some time. The ω₂ subunit is encoded by rpoZ (synonym, yloH) and is similar to the ω subunit present throughout eubacteria (6–8). The ω₁ subunit has been referred to as the second analogous ω subunit simply because it has a size similar to that of the true ω subunit (9–11). This annotation has been accepted despite there being no evidence to suggest that these two proteins are related in function. Using mTRAQ (mass differential tags for relative and absolute quantification) mass spectrometry, it has been shown that both of the ω subunits are approximately equimolar with the β subunit (6). Purification of RNAP shows that both ω subunits remain tightly bound to RNAP even after anion-exchange chromatography (8).

We have determined the structure of ω₁, encoded by the ykzG gene, examined the phenotype generated on its deletion, and determined its location on RNAP. The ykzG gene is present in an operon with the physiologically important RNase RNaseJ1 (RnjA) (12), and the two proteins are transcriptionally and translationally linked. Surprisingly, there is no apparent linkage of the proteins within the cell with respect to any potential intermolecular interaction, and subcellular localization of fluorescent protein fusions indicates that they occupy different subcellular compartments. We propose that this small protein previously referred to as ω₁ is actually a small auxiliary subunit of B. subtilis RNAP, and we subsequently named it ε and the gene encoding it rpoY. Structure analysis indicated that ε is similar to phage T7 Gp2 which inhibits host cell transcription through interaction with RNAP. Mutagenesis studies indicated that both ε and Gp2 interact with RNAP via a shared structural motif. Due to the lack of phenotype on its loss, location within an RNAP complex, and similarity to phage T7 Gp2, we speculate that ε may help protect against phage infection from Gp2-like proteins by occupying their binding sites.

MATERIALS AND METHODS

Strains and growth conditions.

Strains and plasmids used in this study are presented in Table 1. Unless otherwise stated, all strains were grown in LB medium with appropriate antibiotics. Transformation of B. subtilis was based on the methods outlined by Kunst and Rapoport (13). Briefly, B. subtilis strains to be transformed were grown on nutrient agar plates overnight. Colonies were removed from the plates and cultured in Nunc F96 microwell cell culture plates containing 80 μl of MD medium (100 mM K₂HPO₄-KH₂PO₄ at pH 7, 6 mM Na₃ citrate, 300 μM MgSO₄, 50 μg/ml l-tryptophan, 2.5 mg/ml l-aspartate, 11 μg/ml ferric ammonium citrate) supplemented with 0.1% (wt/vol) casein hydrolysate at 37°C with shaking until cells entered stationary phase (A₆₀₀ of 1.2 to 1.5). Eighty microliters of prewarmed MD medium was added to the stationary-phase culture, which was incubated for a further 1 h before 0.8 to 1 μg of DNA was added. The culture was incubated for a further 20 min before 5 μl of 20% (wt/vol) casein hydrolysate was added. The culture was incubated at 37°C for a further 90 min before being plated on nutrient agar plates containing appropriate antibiotics for the selection of transformants.

TABLE 1.

Strains and plasmids used in this study

Strain or plasmid	Genotypea	Reference or source and/or description
E. coli strains
DH5α	λ− ϕ80dlacZΔM15 Δ(lacZYA-argF)U169 recA1 endA1 hsdR17(rK− mK−) supE44 thi-1 gyrA relA1	Gibco BRL
BL21(DE3)	F− ompT hsdSB(rB− mB−) gal dcm (DE3)	53
B834(DE3)	F− ompT gal dcm met(lon) hsdSB(rB− mB−) λDE3	Novagen
B. subtilis strains
BSB1	trpC2 chr::trp	15
EU129	BSB1 chr::cat Pwt rpoC-gfp Pspac ′rpoC	16
EU164	BSB1 chr::erm spc Pspac rpoY-gfp Pxyl rpoC-mCherry cat	This work; BSB1 transformed with pNG622 and pEU14
LK921	BSB1 chr::spc ΔrpoY	This work; rpoY knockout with wild-type rpoY-rnjA promoter
BS489	BSB1 chr::tet ΔrpoY-Pxyl-rnjA	This work; Tetr cassette inserted by double crossover resulting in deletion of rpoY
BS491	BSB1 chr::erm Pwt rpoC-gfp Pspac ′rpoC chr::ΔrpoY tet Pxyl rnjA	This work; BS489 transformed with pEU21
EU337	BSB1 chr::ΔrpoY tet-Pxyl-rnjA amyE::spc Pxyl rpoY-gfp mut1	This work; BS489 transformed with pNG1030
EU344	BSB1 chr::ΔrpoY tet-Pxyl-rnjA amyE::spc Pxyl rpoYF46A,T46A-gfp mut1	This work; BS489 transformed with pNG1037
EU345	BSB1 chr::ΔrpoY tet-Pxyl-rnjA amyE::spc Pxyl rpoY41–50A-gfp mut1	This work; BS489 transformed with pNG1039
EU346	BSB1 chr::ΔrpoY tet-Pxyl-rnjA amyE::spc Pxyl rpoY61–69A-gfp mut1	This work; BS489 transformed with pNG1040
EU347	BSB1 chr::ΔrpoY tet-Pxyl-rnjA amyE::spc Pxyl rpoYE34A-gfp mut1	This work; BS489 transformed with pNG1036
BS509	trpC2 spo0A3 rpoC-His6 spc Pspac rpoY-gfp cam	This work; BS200 (34) transformed with pEU14
Plasmids
pETMCSIII	bla Pϕ10-His6-Tϕ	54
pGEX-5X-3	bla Ptac-GST	GE Healthcare
pSG1154	bla amyE3′ spc Pxyl-′gfp mut1 amyE5′	55
pDG1728	bla erm amyE5′ spoVG-lacZ spc amyE3′	14
pEU14	bla Pspac-′rpoY-gfp mut3-erm	G. Doherty, University of Newcastle, unpublished data
pEU21	bla Pspac-′rpoC-gfp mut3-erm	17
pNG579	bla Pϕ10-rpoY-His9-Tϕ	This work; PCR product of B. subtilis rpoY inserted into NdeI and BamHI sites of pETMCSIII
pNG622	bla cat Pxyl-mCherry	17
pNG689	bla Pϕ10-His6-rpoY′-Tϕ	This work; G. stearothermophilus rpoY PCR product containing a 9-aa N-terminal truncation ligated into NdeI- and EcoRI-digested pETMCSIII
pNG729	bla Pϕ10-His6-rpoY′Leu23Met-Tϕ	This work; G. stearothermophilus rpoY with Leu23 mutated to Met by overlap extension PCR and a 9-aa truncation at C-terminus, ligated into NdeI- and EcoRI-digested pETMCSIII
pNG1030	bla amyE3′ spc Pxyl-rpoY-gfp mut1 amyE5′	This work; B. subtilis rpoY PCR product ligated into KpnI- and EcoRI-digested pSG1154
pNG1036	bla Pϕ10-rpoYE34A-His6 Tϕ	This work; B. subtilis rpoY overlap extension PCR product ligated into KpnI- and EcoRI-digested pSG1154
pNG1037	bla amyE3′ spc Pxyl-rpoYF46A,T48A-gfp mut1 amyE5′	This work; B. subtilis rpoY overlap extension PCR product ligated into KpnI- and EcoRI-digested pSG1154
pNG1039	bla amyE3′ spc Pxyl-rpoY41–50A-gfp mut1 amyE5′	This work; B. subtilis rpoY overlap extension PCR product amplified from pNG984 and ligated into KpnI- and EcoRI-digested pSG1154
pNG1040	bla amyE3′ spc Pxyl-rpoY61–69A-gfp mut1 amyE5′	This work; B. subtilis rpoY overlap extension PCR product amplified from pNG985 and ligated into KpnI- and EcoRI-digested pSG1154
pNG1041	bla Pϕ10-His6-rpoY41–50A-Tϕ	This work; B. subtilis rpoY PCR product amplified from pNG984 ligated into NdeI- and EcoRI-digested pETMCSIII

^abla, ampicillin resistance; P_wt, wild-type promoter; P_spac, IPTG-inducible promoter; P_xyl, xylose-inducible promoter; erm, erythromycin resistance; cat, chloramphenicol resistance; spc, spectinomycin resistance; GST, glutathione S-transferase. Gene subscripts using amino acids should be interpreted according to the following examples: rpoY′_Leu23Met, the Leu-to-Met change at position 23 encoded by rpoY; rpoY_41–50A, mutation of amino acids 41 to 50 to alanines encoded by the rpoY gene.

B. subtilis LK921 (Table 1) with a deletion of rpoY was constructed as follows. The spectinomycin resistance cassette was obtained from pDG1728 (14) by BamHI-EcoRV cleavage and cloned into pGEX-5X-3 (cleaved with BamHI-SmaI). Then, a multiple cloning site (HindIII, EcoRI, PstI; created by annealing oligonucleotides 5′-TCGACCGTGCGGGGCTTCAAATAAGAACTTC-3′ and 5′-TCGAGAAGTTCTTATTTGAAGCCCCGCACGG-3′) was inserted at the BamHI site. The region upstream of rpoY was amplified by PCR with primers 5′-CGGAATTCGCCGTATGTTAGACTCGCTC-3′ and 5′-CGGGATCCCTTGATTTTCAAAGACAACAAAC-3′ and inserted between the BamHI-EcoRI sites of the plasmid. The region downstream of rpoY was amplified with primers 5′-CCGCTCGAGTTAAGGAGGATTTTAGAATGAAATTTGTAAAAAATG-3′ (forward) and 5′-AAGGAAAAAAGCGGCCGCGATTTTCACTGTTTGTGCTG-3′ (reverse) and inserted into the plasmid making use of the XhoI-NotI restriction sites. The forward primer (above) contained a short sequence with a ribosome-binding site (from the tuf gene) to ensure proper translation of the downstream gene (rnjA) encoding RNAseJ1. The whole promoter region of the rpoY-rnjA operon was kept unchanged. The resulting construct was named LK659. B. subtilis BSB1 (15) (Table 1) was transformed with LK659 DNA and selected for spectinomycin resistance to give B. subtilis LK921 (Table 1). The deletion was verified by PCR.

Wild-type B. subtilis rpoY was amplified using primers 5′-CACACACATATGATTTAT-3′ and 5′-GAGGAAGAATTCTAACTCCAA-3′, digested with NdeI and BamHI, and inserted into pETMCSIII to give pNG579 (Table 1). Mutagenized B. subtilis rpoY with alanine substitutions was generated as follows. In the first step, the rpoY gene was amplified by PCR (Expand; Roche) in two parts (separated by 10-amino-acid [aa] gaps that were subsequently substituted with alanines). Primers 5′-GGGCCGGCATATGATTTATAAGGTATTT-3′ (forward) and 5′-CCGCTCGAGTAACTCCAATACTTTAAA-3′ (reverse) were used in combination with primers that were inside the rpoY gene. Internal primers contained 30 nucleotides encoding 10 alanines to mutagenize the 10 respective amino acids at their 5′ ends. In the second step, the two PCR products were mixed in 1×PCR buffer and 25 mM MgCl₂, denatured, and, by decreasing the temperature by 1°C/min (95°C to 35°C), annealed via the polyalanine region. In the third step, DNA polymerase (Expand; Roche) and deoxynucleoside triphosphates (dNTPs; 10 mM each) were added to create double-stranded DNA (dsDNA). The final product was amplified by PCR with the primers 5′-CACACACATATGATTTAT-3′ and 5′-GAGGAAGAATTCTAACTCCAA-3′ and inserted into the NdeI/BamHI sites of pETMCSIII. Wild-type and mutant sequences were subsequently subcloned into pSG1154 (Table 1) prior to transformation into B. subtilis. All clones were confirmed by DNA sequencing.

For the comparison of growth rates, strains were grown in 200 μl of LB medium (minimum of five replicates for each strain) in Nunc F96 microwell cell culture plates with shaking at 100 rpm in a PHERAstar FS (BMG Labtech) at 37°C. Cell density was quantified by measuring the A₆₀₀ every 10 to 12 min for 8 h. Maximum growth rate (doubling time) was determined from the steepest part of the slope during exponential growth.

Fluorescence microscopy.

Cells were grown and imaged as previously described (16, 17).

Phylogenetic work.

All phylogenetic work was performed using information obtained from the National Centre for Biotechnology Information databases (http://www.ncbi.nlm.nih.gov/). Alignments and phylogenetic analysis were carried out using the Clustal W2 and MAFFT algorithms (18, 19).

Protein overproduction and purification.

Geobacillus stearothermophilus rpoY (ε_Gs) with a 9-amino-acid C-terminal truncation was amplified from isolated genomic DNA using the primers 5′-GATCGGCATATGATTTTCAAAGTGTTTTACCA-3′ and 5′-ATCGATGAATTCTCATGATGAAATCTCCAATA-3′, digested with NdeI and BamHI, and inserted into pETMCSIII to give pNG689 (Table 1). In order to produce a protein suitable for structure solution by multiwavelength anomalous dispersion (MAD) through incorporation of Se-Met, the truncated rpoY gene from pNG689 was mutated using the primers 5′-CTTCGATGTACATTGTTTTCGTTTTTTC-3′ and 5′-GAAAAAACGAAAACAATGTACATCGAAG-3′ by overlap extension PCR (20), resulting in the replacement of codon Leu 23 with Met to give pNG729 (Table 1).

For the overproduction of ε_Gs, pNG729 was transformed into the methionine auxotroph E. coli B834(DE3) (Table 1) and cultured in minimal M9 medium with amino acids and vitamin supplements as described previously (21). Cells were grown to an A₆₀₀ of 0.5 to 0.7, and protein overproduction was induced by the addition of isopropyl-β-d-thiogalactopyranoside (IPTG) to 0.5 mM, followed by overnight culture at 18°C. Cells were harvested by centrifugation and stored at −80°C.

Cell pellets were resuspended in imidazole buffer A (20 mM KH₂PO₄, pH 8.0, 500 mM NaCl, 20 mM imidazole), and cells were lysed by passage through an Avestin EmulsiFlex C5 homogenizer, followed by sonication. Following clarification by centrifugation, the lysate was then loaded onto a 1-ml HisTrap HP column (GE Healthcare). Bound protein was washed with 5 ml of imidazole buffer A buffer before being eluted with imidazole buffer B (20 mM KH₂PO₄, pH 8.0, 500 mM NaCl, 500 mM imidazole) in 0.5-ml fractions. If necessary, proteins were further purified by size exclusion chromatography using a HiLoad Superdex 75 16/60GL column (GE Healthcare) equilibrated with 20 mM Tris-HCl, pH 8.0, and 150 mM NaCl. Protein purity and yield were checked by SDS-PAGE. Fractions showing high purity and concentration were pooled and dialyzed into 20 mM HEPES, pH 7.0, 150 mM NaCl, and 1 mM dithiothreitol (DTT).

RNAP with green fluorescent protein (GFP)-tagged ε for use in single-particle analysis was purified from B. subtilis BS509 (Table 1) as previously described (8). The purified sample was dialyzed in 20 mM Tris-HCl, pH 7.8, 150 mM NaCl, 10 mM MgCl₂, 5% (vol/vol) glycerol, and 1 mM DTT.

Protein crystallography and structure determination.

Crystallization screening was carried out at 27°C by sitting-drop vapor diffusion using MRC 96-well crystallization plates and PACT, JCSG, and structure screens (Molecular Dimensions). ε_Gs crystals grew in 100 mM sodium acetate, pH 4.6, and 1.8 M ammonium sulfate.

Crystals were cryo-protected by replacing the lower well with 3.5 M ammonium sulfate and allowing the crystallization drop to equilibrate overnight. Crystals were mounted in cryo-loops and drop frozen in liquid nitrogen. A two-wavelength MAD data set was collected from a single crystal on the MX2 beamline at the Australian synchrotron using the Blu-Ice software (22). The collection wavelengths for peak anomalous signal and a high-energy remote were determined from the anomalous f′ and f″ plots, which were calculated using CHOOCH (23). Reflections used for refinement were collected in-house at the University of New South Wales (UNSW; Sydney, Australia), using a MAR345dtb image plate detector mounted on a Rigaku MicroMax HF007 rotating anode generator with Osmic confocal mirrors using Cu Ka radiation.

Diffraction data were indexed and integrated using iMOSFLM (24) and scaled in Scala (25). After integration, the space group was determined to be P1 21 1 with the cell dimensions of 35.6 Å, 82.81 Å, 83.41 Å, 90°, 92.7°, and 90°, and structure determination was performed using the CCP4 software suite (26). The calculation of phases and initial model building were achieved using the Crank automated structure solution pipeline (27), which used Afro/Crunch2 (28) for substructure detection, Solomon (29) for hand determination and density modification, and, finally, Buccaneer (30) for model building. The initial model was then extended using ARP/wARP and continued manually using COOT (31, 32). Refinement was carried out using phenix.refine in conjunction with COOT. The final model was assessed using the Phenix comprehensive validation tool (31, 33).

Electron microscopy and 3D reconstruction by single-particle analysis.

The purified RNAP-ε-GFP sample was diluted to 0.08 μM in 20 mM Tris-HCl, pH 7.8, 150 mM NaCl, 10 mM MgCl₂, and 1 mM DTT. Four microliters of sample was applied to homemade continuous carbon grids for 30 s before being washed three times with the dilution buffer. The specimens were then stained with a 1% (wt/vol) uranyl formate aqueous solution. The imaging conditions and particle selection were performed as detailed by Yang et al. (34). A total of 260 images were taken, and 10,104 particles were used in three-dimensional (3D) reconstructions.

The EMAN software package, version 1.8 (35), was used for image processing. Individual particles were appended, center aligned, low-pass filtered to 15 Å, and boxed into 72- by 72-pixel images. The previously published RNAP core negative stain electron density map (EMDataBank EMD-1577) (34) was filtered to 30 Å and employed as the initial model for refinement. After 10 rounds of iteration, the model stably converged to ∼24-Å resolution as estimated with the EMAN e/o test.

Transcriptomics.

B. subtilis BSB1 and LK921 were grown to A₆₀₀ 0.7 in CH medium (36) and triplicate samples processed and analyzed as detailed in (15).

Accession numbers.

The final refined structure of ε was deposited in the Protein Data Bank (PDB; http://www.pdb.org/pdb/home/home.do) under accession number 4NJC. The negatively stained molecular RNAP envelope has been deposited in the EMDataBank (EMDB; http://www.emdatabank.org) under accession number EMD-2637.

RESULTS

ε (YkzG) is a core subunit of RNAP.

In previous studies we confirmed that the identity of the low-molecular-weight band copurifying with RNAP from B. subtilis that had been called ω₁ was encoded by the ykzG gene (8). Due to the fact that YkzG could not be separated from other core RNAP subunits using nondenaturing chromatography, we suggested that it was a subunit of the enzyme. These results were subsequently confirmed by Delumeau et al. (6), who isolated transcription complexes from B. subtilis cultures and showed that it was present at similar levels to RNAP core subunits under a range of different growth conditions.

If YkzG is a subunit of core RNAP, it would be expected to colocalize with it in the cell. In order to determine if this was the case, we examined the localization patterns of fluorescent protein fusions of YkzG and RNAP. Strain EU164 carrying chromosomal fusions of GFP to YkzG and mCherry to RpoC (the β′ subunit of RNAP) was grown to mid-exponential phase in LB medium, and fluorescence was observed as detailed in Materials and Methods. Due to the choice of fluorescent protein fusions used in this study, nucleoids could also be counterstained with the DNA stain 4′,6-diamidino-2-phenylindole (DAPI) so that fluorescent protein localization could be correlated with nucleoid morphology as in previous work (37).

As expected, both YkzG (Fig. 1C) and RNAP (Fig. 1B) colocalized with the nucleoid (Fig. 1A and D). The arrows in Fig. 1 indicate the presence of a transcription focus which forms due to the high level of loading of RNAP onto rRNA operons in rapidly growing cells (36, 37). Since YkzG-GFP fluorescence exactly mirrored that of the RpoC-mCherry fusion, this indicates that YkzG is associated with RNAP involved in all classes of transcription (mRNA and stable RNA synthesis), further strengthening the hypothesis that it is an integral subunit of RNAP in B. subtilis. BLAST searches indicated that YkzG was confined to the phylum Firmicutes (see Fig. S1 in the supplemental material). There was also no significant sequence similarity between the true ω subunit encoded by yloH (rpoZ) and ykzG. Therefore, YkzG represents a new subunit of RNAP and has been given the name ε, and the gene encoding it has been named rpoY. Core RNAP from B. subtilis and other firmicutes comprises at least seven subunits and should be represented as α₂ββ′ωδε.

FIG 1.

Colocalization of ε (YkzG) and RNAP. (A) Pseudocolored image of DAPI-stained exponentially growing B. subtilis EU164 (rpoY-gfp rpoC-mCherry) to label nucleoid location and morphology. (B) Pseudocolored image of the RpoC-mCherry fusion illustration of the subcellular location of RNAP. (C) Pseudocolored image of the ε (YkzG)-GFP fusion. (D) Overlay of the three fluorescence images shown in panels A to C. (E) Pseudocolored image of DAPI-stained nucleoids in B. subtilis BS491 (rpoC-gfp ΔrpoY). (F) Pseudocolored image of the RpoC-GFP fusion. (G) Overlay of images from panels E and F. Arrows indicate a transcription focus which represents the site of rRNA synthesis, and a cartoon of a cell containing two nucleoids is shown in panels D and G as a reference. Scale bar, 3 μm.

Examination of sequence and transcriptome data revealed that the B. subtilis rpoY gene is located in a two-gene operon with another gene called rnjA (15). The two genes are always cotranscribed at identical levels and are translationally linked (Fig. 2A). The gene rnjA encodes the RNase RNaseJ1 that is responsible for the maturation of the 5′ end of 16S rRNA and is crucial for proper ribosome assembly (38). Exhaustive searches of available annotated bacterial genomes in the NCBI failed to identify a case where the rpoY gene was not directly followed by rnjA although rnjA is distributed more widely than rpoY and so is often present in its absence. The possibility that the close correlation of the rpoY and rnjA genes was related to the function of the translated proteins was examined. Previously published work examining the localization of an RNaseJ1-GFP fusion showed that it is the same as ribosomes, concentrating toward the poles of the cell and being absent from or in low concentration in the region occupied by the nucleoid (39). This is a completely different localization pattern than we observed for ε (Fig. 1), indicating that the two proteins occupy different subcellular regions and are unlikely to directly/stably interact with each other. In vitro studies also failed to identify any interaction between the two proteins (data not shown).

FIG 2.

Transcription analysis of rpoY. (A) Levels of the rpoY-rnjA (ykzG-rnjA) transcript under 103 different growth conditions (15) taken from the B. subtilis Expression Data Browser (http://genome.jouy.inra.fr/cgi-bin/seb/index.py). A map of the region is shown below, along with an expanded portion showing the location of the ribosome binding site (SD) for rnjA located at the end of the ykzG (rpoY) gene. (B) Transcriptomics data for the rpoY-rnjA (ykzG-rnjA) region in wild-type (BSB1; black) and ΔrpoY (LK921; magenta) strains. Top section, coding strand; bottom, noncoding strand (with respect to rpoY-rnjA). The location and direction of transcription of genes are indicated in the middle section.

Phenotypic characterization of ε.

The rpoY gene was deleted from the B. subtilis chromosome to observe any phenotype associated with ε depletion in the cell. The rpoY gene was removed from BSB1 and replaced with a spectinomycin resistance gene to give strain LK921 (Materials and Methods), and the cellular levels of RNaseJ1 were shown to be identical to those of the parent strain by quantitative Western blotting (C. Condon, personal communication). Strain LK921 grew at growth rates identical to those of the wild type when cultured in a range of liquid media, and the cells appeared morphologically identical to the wild-type cells when examined by phase-contrast microscopy (data not shown). Competition growth experiments between LK921 and the wild type showed that rpoY deletion did not detectably affect the fitness of the knockout. Examination of knockout strain BS491 containing an rpoC-gfp fusion indicated that deletion of ε had no visible effect on RNAP localization within the cell, with transcription foci clearly visible (Fig. 1F, arrow) and all RNAP fluorescence colocalizing with nucleoid signal (Fig. 1E to G). Thus, ε does not appear to have a major influence on general bacterial growth, fitness, or RNAP distribution within the nucleoid.

To determine if the lack of ε in the cell caused changes in the transcription profile of the cell, transcriptomics were carried out comparing exponentially growing strain LK921 and wild-type BSB1 cells (Table 1) as described in Materials and Methods. Since rpoY is expressed at high levels similar to other RNAP subunits under over 100 different growth conditions (15), transcriptomics were performed on mid-exponentially growing cells grown in the defined rich CH medium (36). Remarkably, no significant difference in the transcription profiles between the two strains was observed, other than the loss of expression of rpoY (Fig. 2B). The small number of genes showing a log₂ change of ≥1.0 and a P of ≤0.05 are listed in Table 2. Of the two genes showing increased levels of expression, yhfH encodes a hypothetical protein, whereas nadB is the first gene in the nadBCA operon involved in NAD biosynthesis. Of the genes showing reduced levels of expression, narG is part of the narGHJI operon encoding nitrate reductase, and trpF is part of the trpEDCFBA operon involved in tryptophan biosynthesis. None of the other genes in the trp operon showed altered transcript levels between the wild type and knockout, and previous studies have shown that trpF exhibits highly variable transcript levels (15). Interestingly, adeC transcript levels were also reduced moderately. This gene lies directly downstream from the rpoY-rnjA operon and is transcribed in the opposite direction. The reduced level of expression is not due to lack of termination of transcription of rnjA in the knockout (Fig. 2B) and was observed in only two of the three transcriptome samples (Fig. 2B, magenta lines), suggesting that it is unlikely to be highly significant with respect to ε function.

TABLE 2.

Gene transcript changes in the rpoY-deficient strain LK921

Expression group and gene name	Locus tag	Description	Fold change in expression (log2)	P valuea
Increased expression
yhfH	BSU10230	Hypothetical protein	1.53	0.036
nadB	BSU27870	l-Aspartate oxidase	1.00	0.008
Decreased expression
narG	BSU37280	Nitrate reductase (alpha subunit)	−0.99	0.040
adeC	BSU14520	Adenine deaminase	−1.12	0.019
trpF	BSU22650	Phosphoribosyl anthranilate isomerase	−2.52	0.001
rpoY	BSU14540	RNAP ε subunit	−8.13	0.006

^aTwo-tailed unequal variance Student's t test.

Finally, we conducted a series of in vitro transcription assays that included multiple rounds of transcription (see the supplemental material), promoter binding, open complex formation and decay, promoter escape, and sensitivity of RNAP to the concentration of the initiating NTP (which also indicates the effects on the stability of open complexes). No effect by ε, either positive or negative, in any of these assays could be detected. Overall, we conclude that ε has little effect on global transcript levels in vivo or on transcription initiation in vitro.

Determination of the structure of ε.

In a further attempt to establish a functional role for ε, we undertook studies to determine its structure by X-ray crystallography. In order to increase solubility, a truncation of 9 amino acids at the carboxy-terminal end of the Geobacillus stearothermophilus protein was used in crystallographic studies (see Materials and Methods). In order to obtain experimental phases, leucine 23 was mutated to methionine in order to enable the production of seleno-methonine-substituted protein (Materials and Methods) suitable for the generation of experimental phases that would allow structure solution by multiwavelength anomalous dispersion (MAD) (Table 3), and this protein was subsequently referred to as ε_Gs. The asymmetric unit contained eight monomers, and initial model building proved difficult due to poorly defined density between the β1 and β2 strands in all but one of the monomers. However, one of the monomers (chain A) did contain a limited amount of density into which the backbone could be confidently built (Fig. 3A). This density was of insufficient quality to permit building and refinement of side chains for residues Asp19, Glu20, Arg24, and Asp25, and so these side chains were assigned common rotamers; the poor density fit is reflected in the inflated B-factors of these residues. The resulting structure was a simple β sheet with an α helix running diagonally along the back of the structure forming a ββαβ fold (Fig. 3B).

TABLE 3.

Data collection and refinement statistics for the ε_Gs crystal structure

Statistic	In-house dataa	Synchrotron dataa
Data collection
Space group	P 1 21 1	P 1 21 1
Cell dimensions
a, b, c (Å)	35.65, 82.78, 83.43	36.54, 83.21, 83.43
α, β, γ (°)	90, 92.76, 90	90, 93.28, 90
		Peak	Remote
Wavelength	1.54179	0.97941	0.95369
Resolution (Å)	25.0–2.3 (2.42–2.3)	41.65–2.31 (2.44–2.31)	41.33–2.3 (2.42–2.3)
Observations	158,061 (22,673)	138,779 (22,137)	139,240 (22,352)
No. of unique reflections	21,590 (3,136)	19,757 (3,096)	19,727 (22,352)
Rmerg	0.056 (0.237)	0.097 (0.749)	0.076 (0.489)
Rmeas	0.060 (0.255)	0.135 (0.856)	0.098 (0.566)
Mean I/σ(I)	18.4 (5.9)	11.8 (2.9)	14.2 (3.6)
Completeness (%)	99.92 (100.00)	90.1 (96.8)	91.5 (97.5)
Multiplicity	7.3 (7.2)	7.0 (7.2)	7.1 (7.2)
Anomalous completeness (%)		90.3 (97.0)	90.9 (97.0)
Anomalous multiplicity (%)		3.6 (3.6)	3.6 (3.7)
Wilson B-factors	46.36	35	34.2
Refinement
Resolution (Å)	23.07–2.30 (2.38–2.30)
No. of reflections	21,575 (2,185)
Rwork	0.214 (0.267)
Rfree	0.256 (0.314)
CC*b	0.819 (0.648)
CCwork	0.890 (0.903)
CCfree	0.873 (0.791)
No. of atoms	6,955
Protein	3,506
Ligand/ion
Water	37
B-factors
Overall	61.7
Protein	61.9
Ligand/ion
Water	42.1
RMS deviations
Bond lengths (Å)	0.004
Bond angles (°)	0.72
Ramachandran favored (%)	97
Ramachandran outliers (%)	0.53

^aValues for the highest-resolution shell are shown in parentheses.

^bCC*, true correlation coefficient, as defined by the phenix tool.

FIG 3.

Structure of ε. (A) Electron density as a mesh focused on the loop between β1 and β2 from chain A of the asymmetric unit into which the loop was built. (B) Cartoon representation of the structure of ε (PDB 4NJC) determined by X-ray crystallography to a resolution of 2.3Å. (C) Structure of phage T7 protein Gp2 (PDB 2LMC, chain A) on the left side. The electrostatic potentials of ε and Gp2, scaled between 3 kT/e (blue) and −3 kT/e (red) (with k the Boltzmann constant, T the temperature, and e the elementary charge), are shown mapped onto the protein surfaces on the right. Images were created using PyMol (version 1.3; Schrödinger, LLC) using the APBS plug-in (52).

Searches for structural homologues using PDBeFOLD (http://www.ebi.ac.uk/msd-srv/ssm/ssmstart.html) found four structures with a fold similar to that of ε_Gs but no common functional assignment. These four structures were the Gram-negative T7 phage protein Gp2 (PDB 2WNM) (Fig. 3C), PaaB from Ralstonia eutropha (PDB 3EGR), archaeal ribosomal protein LX (PDB 3J21), and p56 from the Gram-positive bacteriophage ϕ29 (PDB 3ZOQ). PaaB is a small protein with unknown function in the phenyl-acetate catabolism pathway and has not been characterized or linked directly to a cellular function. The LX protein is part of the 50S ribosomal subunit from Haloarcula marismortui, where it interacts nonspecifically with the 23S rRNA (40). p56 is an inhibitor of uracil DNA glycosylase that prevents host cell uracil excision repair (41, 42).

Gp2, which had the highest structural similarity to ε_Gs (root mean square deviation [RMSD] of 1.85) (Fig. 3C), was of particular interest as it is an RNAP binding protein that inhibits the formation of open complexes and initiation of transcription (43, 44). Gp2 binds to a region of RNAP in the DNA binding cleft called the jaw (45, 46) and inhibits transcription initiation by preventing region 1.1 of σ⁷⁰ moving out of the DNA binding cleft to allow promoter DNA entry (45). The interaction between Gp2 and RNAP is predominantly made between the β3 strand and the region of amino acids 1045 to 1189 of β′ (46). We also examined the electrostatic potential on the surface of ε and Gp2 as Gp2 contains a negatively charged strip that is functionally important in the inhibition of host-cell transcription (47). While the negatively charged strip was clearly visible in Gp2 (Fig. 3C, right-hand side red surface), no such charge distribution was visible on the surface of ε (Fig. 3B, right-hand side), supporting the conclusions of the in vitro transcription assays that, unlike Gp2, ε does not inhibit transcription initiation.

Investigation of the ε binding site.

Due to the structural similarity of ε with Gp2, we considered the possibility that ε bound to a similar region of RNAP. A deletion of the jaw region of B. subtilis RNAP (β′ subunit amino acids 963 to 1004) equivalent to that from the E. coli enzyme (amino acids 1045 to 1198) (44) was constructed. Simultaneously, the same deletion was made in plasmid pNG567 that is used for overproduction of recombinant B. subtilis RNAP containing ε (8). Despite multiple attempts, no transformants of jawless GFP-tagged RNAP could be obtained, and no soluble recombinant protein could be obtained, preventing us from determining by either microscopic analysis of live cells or mass spectrophotometric analysis of purified recombinant protein (data not shown) whether ε binds RNAP in the absence of the jaw region.

As an alternative, we utilized negative-stain electron microscopy with single-particle analysis (34) to generate a low-resolution 3D structure in order to determine the approximate location of ε on RNAP. Due to the small size of ε (∼8 kDa), it was not possible to directly visualize ε on the core complex. To overcome the relative size problem, strain BS509 (Table 1) was created which contained rpoC-His6 and rpoY-gfp fusions. The relatively large size of GFP (∼27 kDa) would allow us to assign density to the approximate region of RNAP to which ε binds. Coomassie blue-stained gels and Western blotting with anti-GFP antibodies were used to show that the ε-GFP fusion was present in the RNAP preparation (Fig. 4A, lanes 3 and 5). A total of 10,104 particles were used for a 3D reconstruction of the GFP-tagged complex (Materials and Methods). Class sums (I) and reprojections (II) of RNAP core and the purified RNAP ε-GFP complex are shown in Fig. 4B, and class sums that allowed identification of mass due to GFP are indicated by an asterisk. The final reconstruction converged at a resolution of ∼24Å, and the mass due to GFP can be clearly seen on the downstream side of the enzyme compared with the equivalent RNAP core structure (Fig. 4C and D). The mass due to GFP (Fig. 4D, green circle and rod) was close to the jaw (Fig. 4D, red) and the secondary channel (Fig. 4D, dark gray circle). While this reconstruction does not permit unequivocal determination of the location of ε, it is consistent with ε being able to bind on the downstream side of the DNA binding cleft in a similar region to Gp2 (see below).

FIG 4.

Determination of the structure of ε-GFP RNAP by single-particle analysis. (A) Coomassie blue-stained gel and Western blot. Lanes 1 and 4, molecular mass markers; lane 2, wild-type RNAP; lane 3, RNAP containing an ε-GFP fusion isolated from B. subtilis BS509; lane 5, anti-GFP Western blot of the RNAP containing the ε-GFP fusion. (B) Class sums (I) and reprojections (II) of wild-type core RNAP (RNAP) and RNAP containing the ε-GFP fusion. (D) Face (top) and downstream side (bottom) views of wild-type RNAP (silver) and RNAP containing the ε-GFP fusion (gold), with the mass due to GFP clearly visible on the downstream side. (C) Homology model of a B. subtilis RNAP elongation complex face (top) and downstream side (bottom) with the jaw region shown in red and the location of GFP marked in translucent green. The location of the secondary channel is marked as a dark gray circle in the downstream side view. DNA coding (dark green) and noncoding (orange) strands as well as RNA transcript (blue) are also shown.

Since our results indicate that ε could bind to a similar region of RNAP to Gp2, we explored the possibility that similar regions of the proteins were required for interaction with RNAP. Previous studies have shown that E28, D37, E44, and, in particular, R56 and R58 are required for interaction of Gp2 with the β′ jaw of E. coli RNAP (PDB 2LMC) (46, 48). There is no significant sequence conservation between Gp2 and ε, so amino acids in (approximately) equivalent positions were chosen for mutation. We monitored the localization of ε-GFP fusions within live B. subtilis cells to determine which regions/amino acids were important for interaction with RNAP (Materials and Methods). ε residue F46 and ε T48 are in roughly equivalent positions to the Gp2 R56 and R58 residues which are involved in binding to E. coli RNAP while ε_R34 is approximately equivalent to Gp2 D37 or E44 (Fig. 5A and B). In addition, the 10 amino acids spanning the β3 strand and the final 10 amino acids were also mutated to alanines (ε residues 41 to 50 mutated to alanines [ε_41–50A] and ε_61–69A, respectively) (Fig. 5C). All of the mutants localized to the nucleoids with an identical pattern to the wild-type ε except for ε_41–50A (Fig. 5D to H), indicating that ε R34, F46, T48, and the final 9 amino acids were not essential for interaction with RNAP. Examination of the localization of the ε_41–50A mutant showed that fluorescence was distributed throughout the whole cell (Fig. 5G), suggesting that the 10 amino acids spanning the β3 strand are important for interaction with RNAP. It is possible that changing 10 amino acids to alanine residues caused misfolding of the protein, but we do not believe this is the case as overproduced ε_41–50A behaved identically to wild-type ε during overproduction and purification. Additional mutant proteins with 10 sequential alanines spanning the loop between β1 and β2 (amino acids 11 to 20) and across the β2 strand (amino acids 21 to 30) aggregated in solution and were considered unlikely to form functional GFP fusions in vivo (data not shown). We also created GFP fusions with alanine mutations in the most highly conserved residues (K4, E8, R17, E18, T20, and E45) that cover other regions of ε, but all of them except the K4A mutant showed identical localization to the wild-type protein (data not shown). The K4A mutant was not fluorescent on transformation of B. subtilis, which most likely indicates that the fusion was nonfunctional and was rapidly degraded. Therefore, our data are consistent with ε and Gp2 binding their cognate RNAPs at similar sites via the same regions.

FIG 5.

Determination of the region of ε that interacts with RNAP. (A and B) Structures of Gp2 and ε_Gs, respectively, with amino acids that have been mutated or shown to be involved in the interaction with RNAP highlighted in red. The region of ε mutated to 10 consecutive alanines (residues 41 to 50) is shown in purple. (C) Sequence and structural alignments of Gp2 and ε. Gp2 residues involved in interaction with Escherichia coli RNAP are shown in green. ε residues that were altered are represented the same as in panel B. α, alpha helix; β, β strand. (D to H) Localization of wild-type, ε_F46A,T48A (E), ε_R34A (F), ε_41–50A (G), and ε_61–69A (H) ε-GFP fusions. Left panel, GFP image; middle, DAPI-stained nucleoids (DNA); right, overlay. A cartoon of a cell containing two nucleoids is shown on the right of panel D. Scale bar, 4 μm.

DISCUSSION

We have identified a new subunit of RNAP that is restricted to the medically and industrially important group of high-A+T Gram-positive bacteria (the Firmicutes). While there are highly conserved subunits of RNAP across the kingdoms, within the Firmicutes we now know that there are at least two additional small subunits of RNAP, giving a subunit composition of α₂ββ′δεω in these organisms. The δ, ε, and ω subunits all appear to be present at levels similar to those of the other subunits of RNAP (6, 17), suggesting that they are all core subunit components.

As with δ and ω, ε is not essential for viability and, despite much work, the functional role of the δ and ω subunits is only now being uncovered (2, 4, 49). No change in the transcriptome of B. subtilis could be detected on deletion of ε or any effect of ε in in vitro transcription assays, and additional studies are needed to fully elucidate its functional role in transcription.

Determination of the structure of ε_Gs by X-ray crystallography revealed it to have a ββαβ fold that is present in a diverse range of small proteins of varied/uncharacterized function and that may be important in protein-protein interactions. 3D reconstruction by single-particle analysis indicated that ε is located on the downstream side of RNAP, in the region of the β′ jaw and secondary channel. We do not believe it is likely that ε binds in/around the secondary channel as it is structurally unrelated to proteins such as GreA/B and DksA that are known to bind in that region (50, 51), and we favor the idea that ε binds to the jaw in a similar fashion as Gp2. GreA is present in Gram-positive bacteria, but DksA, which has been implicated in having a role in the stringent response, is not. Transcriptomes of wild-type and Δε strains following induction of the stringent response were identical, indicating that ε has no analogous role to DksA (data not shown). Therefore, we favor the hypothesis that ε binds to/near the β′ jaw in the DNA channel of RNAP in a location equivalent to Gp2 (Fig. 6A). While we could not unequivocally confirm binding to the β′ jaw, interaction with RNAP is mediated via the β3 strand (Fig. 6A and B, green) that is also important in Gp2 binding to RNAP (Fig. 6C, red) as mutation of this region abolished ε localization to the nucleoid.

FIG 6.

Model of ε bound to RNAP. (A) ε docked against the jaw region of an homology model of B. subtilis RNAP using the Gp2-jaw structure (PDB 2LMC) as a guide. (B and C) ε-jaw and Gp2-jaw complexes, respectively. In panel B, ε is shown in red, with the β3 strand thought to be involved in interaction with RNAP shown in green. In panel C, Gp2 is shown in green with the β3 strand known to be involved in interaction with the jaw of E. coli RNAP (pale blue), shown in red.

At this stage the function of ε remains unclear, but one possibility is that it is required for phage protection by blocking access of Gp2-like proteins to RNAP. Such proteins are widely distributed among Gram-negative phage, but due to the small size and level of sequence divergence, it is not possible to ascertain whether they are also present in Gram-positive phage. A comprehensive survey of Bacillus phage infection in ε-positive and knockout strains in the future will help shed light whether ε plays a role in protection from phage infection.