Structural basis of bacterial σ²⁸‐mediated transcription reveals roles of the RNA polymerase zinc‐binding domain

Abstract

In bacteria, σ²⁸ is the flagella‐specific sigma factor that targets RNA polymerase (RNAP) to control the expression of flagella‐related genes involving bacterial motility and chemotaxis. However, the structural mechanism of σ²⁸‐dependent promoter recognition remains uncharacterized. Here, we report cryo‐EM structures of E. coli σ²⁸‐dependent transcribing complexes on a complete flagella‐specific promoter. These structures reveal how σ²⁸‐RNAP recognizes promoter DNA through strong interactions with the −10 element, but weak contacts with the −35 element, to initiate transcription. In addition, we observed a distinct architecture in which the β′ zinc‐binding domain (ZBD) of RNAP stretches out from its canonical position to interact with the upstream non‐template strand. Further in vitro and in vivo assays demonstrate that this interaction has the overall effect of facilitating closed‐to‐open isomerization of the RNAP–promoter complex by compensating for the weak interaction between σ4 and −35 element. This suggests that ZBD relocation may be a general mechanism employed by σ⁷⁰ family factors to enhance transcription from promoters with weak σ4/−35 element interactions.

Cryo‐EM studies of the E. coli σ²⁸‐RNA polymerase complex recognizing a flagellar promoter sequence identifies conformational transitions that promote transcription from a weak promoter.

Introduction

Multi‐subunit DNA‐dependent RNA polymerase (RNAP) is the core enzyme responsible for transcription, the first step of gene expression in cells. In bacteria, the RNAP core enzyme comprises four types of evolutionarily conserved subunits (α₂ββ′ω) (Murakami & Darst, 2003; Borukhov & Nudler, 2008), which is unable to recognize specific promoter sequences alone. Transcription initiation is tightly regulated by sigma factors, in which the core enzyme first binds a specific sigma factor to form a holoenzyme and then recognizes specific promoters. The subsequent step is isomerizing from closed RNAP–promoter complex (RPc) to open RNAP–promoter complex (RPo) with melted base pairs (bp) that is competent to initiate transcription (Feklistov & Darst, 2011; Saecker et al, 2011; Ruff et al, 2015; Browning & Busby, 2016).

On the basis of the structural and functional differences, bacterial sigma factors could be classified into two main families: the primary σ⁷⁰ factor family and the σ⁵⁴ factor family for nitrogen regulation and some stress responses (Feklistov et al, 2014). The σ⁷⁰ family factors are then subdivided into four major groups according to the different compositions of the conserved domains—σ1.1 (σR1.1 region), σ2 (σR1.2, σNCR, and σR2.1–2.4 regions), σ3 (σR3.0 and σR3.1 regions), and σ4 (σR4.1 and σR4.2 regions) (Feklistov et al, 2014; Paget, 2015). Sigma factors in group 1 (σ⁷⁰ in Escherichia coli) contain all the conserved domains, while group 2 sigma factors (σ^S or σ³⁸ in E. coli) lack σNCR region, group 3 sigma factors (σ²⁸ in E. coli, also known as RpoF or FliA) lack σR1.1, σNCR, and σR1.2 regions, and group 4 (extracytoplasmic function, ECF) sigma factors are the most stripped‐down version, possessing only two essential domains σ2 and σ4 (Osterberg et al, 2011; Paget, 2015).

As a representative group 3 sigma factor, σ²⁸ is the most widely distributed alternative sigma factor that controls flagellum biosynthesis in all motile Gram‐negative and Gram‐positive bacteria (Paget, 2015), and is indispensable for motile bacteria to compete with other microorganisms and survive in the adverse conditions like poor nutrition (Zhao et al, 2007). In addition, σ²⁸ has been reported to play a role in cell development in some non‐motile bacteria (Chater et al, 1989; Yu & Tan, 2003). Interestingly, σ²⁸ homologues from distant species may work identically and can successfully restore the motility in E. coli fliA mutant (Chen & Helmann, 1992; Heuner et al, 1997; Studholme & Buck, 2000). Bioinformatic analysis of σ²⁸ promoters suggests that consensus sequences for −35 and −10 elements are TAAAGTTT and GCCGATAA, respectively, which are separated by an 11‐bp spacer (Yu et al, 2006b). This distinguishing promoter feature was confirmed by mutational and biochemical analyses (Yu et al, 2006a; Koo et al, 2009; Hollands et al, 2010), and the first two nucleotides in −10 element were further defined as an extended −10 motif (Koo et al, 2009). Comparing with regulation to σ⁷⁰‐dependent promoters, cyclic AMP receptor protein binds to σ²⁸‐dependent promoters at an atypical location, suggesting that recognition of σ²⁸ to −35 element may be different from σ⁷⁰ (Hollands et al, 2010). Nevertheless, the molecular mechanism of σ²⁸‐dependent transcription initiation remains largely uncharacterized.

The well‐studied transcription initiation complex (TIC) structures of group 1 sigma factors (Murakami et al, 2002b; Zhang et al, 2012; Zuo & Steitz, 2015) and the recent TIC structures of group 2 sigma factor (Liu et al, 2016) and group 4 sigma factors (Li et al, 2019; Lin et al, 2019) have greatly facilitated our understanding of how these sigma factors recognize respective promoter elements and initiate transcription. However, as yet, there is no structure of RNAP holoenzyme or transcription complex for group 3 sigma factor—σ²⁸—to illustrate the mechanism of transcription initiation. In this study, we assembled the intact functional E. coli transcribing complex with the flagella‐specific sigma factor—σ²⁸—and determined the first TIC structures of group 3 factors at around 3.9 Å resolution. The structures show that σ²⁸‐RNAP has strong interaction with promoter −10 element but weak contact with −35 element. Intriguingly, the β′ zinc‐binding domain (ZBD), also previously known as zinc ribbon region (ZNR) (Lane & Darst, 2010), of RNAP stretches out from its canonical position to interact with non‐template strand (NT‐strand) in a distinct architecture, which consequently stabilizes promoter binding to advance transcription initiation by compensating for the weak interaction between σ4 and −35 element. Further in vitro and in vivo tests reveal that ZBD relocation shows an overall effect of facilitating the isomerization from RNAP–promoter closed complex to open complex, and also suggest that the relocation of β′ ZBD is a general mechanism employed by sigma factors with weak σ4/−35 element interactions in transcription initiation. These observations advance our understanding of transcription initiation by RNAP from weakly bound promoters, such as promoters with non‐conserved −35 element, which comprise the majority in bacteria (Ettwiller et al, 2016).

Results

Overall structure of the Escherichia coli σ²⁸‐dependent transcribing complex

To obtain the cryo‐EM structure of the E. coli σ²⁸‐dependent transcribing complex, we assembled the complex with RNAP core enzyme, σ²⁸ factor, the synthetic DNA scaffold (from −39 to +15, positions relative to the transcription start site +1) (Fig 1A). Incubation with nucleotides (ATP and CTP) and Mg²⁺ produced transcription initiation complexes (TICs) with a nascent AAC RNA (Fig EV1A and B).

Figure 1. Cryo‐EM reconstructions of the σ²⁸‐TICs at the two states.

1. Schematic representation of the synthetic promoter DNA scaffold (54‐bp) in the σ²⁸‐TICs. The −10 and the extended −10 elements were annotated by a dashed box according to the previous report (Koo et al, 2009) and shown in color letters based on the observed structure in this study. The predicted −35 element (Yu et al, 2006b) and the observed one were also annotated in the same way.
2. Overviews of the cryo‐EM reconstruction maps of the E. coli σ²⁸‐TICs at the state 1 (left, 3.86 Å resolution) and state 2 (right, 3.91 Å resolution), respectively. The individually colored density maps, created by color zone and split in Chimera and contoured at 1.0 of view value of Chimera, are displayed in transparent surface representation to allow visualization of all the components of the complex.
3. Zoom‐in views of β′ ZBDs in the σ²⁸‐TICs. The split density map for the β subunit was omitted for clear representation.

Figure EV1. Isolation of the σ²⁸‐TIC, cryo‐EM images, and data processing procedure for σ²⁸‐TIC .

• A
  The size‐exclusion chromatography profile of the σ²⁸ ‐TIC is presented. Peak 2 is the target complex, while peaks 1, 3, and 4 are the aggregation form, surplus DNA, and surplus NTP, respectively.
• B
  The SDS–PAGE gel visualized the components and verified the presence of the complex.
• C
  A representative micrograph. The scale bar size is 36.6 nm.
• D
  Flow chart of the cryo‐EM image processing (see Method details).
• E
  Selected 2D classes for the structure 0.
• F
  Gold‐standard Fourier shell correlations (FSCs) of the maps for structures 0–2.
• G, H
  Angular orientation distributions of the particles used in the final reconstruction (left) and FSC curves of half maps and model to map (right) for structure 1 (state‐1 σ²⁸‐TIC) (G) and structure 2 (state‐2 σ²⁸‐TIC) (H).

Source data are available online for this figure.

The cryo‐EM single‐particle reconstruction of the complex showed two different architectures corresponding to two states (Fig 1B and C). The state 1 and state 2 structures were reconstructed at the overall resolutions of 3.86 Å and 3.91 Å, respectively (Fig EV1C–H, and Appendix Table S1). The qualities of map reconstruction evaluated using Mtriage in Phenix showed the resolutions of 4.23 Å (state 1) and 4.27 Å (state 2) according to half‐map Fourier shell correlation (FSC) calculations at the criteria of 0.143 (Fig EV1G and H). Further local resolution maps displayed the resolution ranges for the individual parts of the complex: most parts of RNAP at 3–4.25 Å, β′ ZBD at 3–5.5 Å, and major parts of σ²⁸ and DNA scaffold at 3–5.5 Å, in which the areas involving protein–DNA interactions have better local resolutions, including β′ ZBD and DNA‐binding regions in σ²⁸ (Fig EV2). The local real‐space correlation coefficients (CC) for the ZBD domain calculated at the resolution of 5 Å using Chimera are 0.8872 (6PMI, state 1) and 0.8367 (6PMJ, state 2), respectively. The densities in the cryo‐EM maps were well fitted by the core RNAP, which shows a similar overall architecture to that observed in other TICs (Zuo & Steitz, 2015; Liu et al, 2016, 2017). The σ²⁸ factor has relatively lower sequence identity (22.7%) and similarity (38.9%) to σ⁷⁰ factor (Fig EV3A). However, the folding pattern of σ²⁸ factor in TICs shows high similarity with that of σ³⁸ or σ⁷⁰ throughout the conserved regions, from σR2.1 to σR4.2 regions, indicating the conserved structures of the σ⁷⁰ family in TICs. Most of the side chains of amino acids in σ²⁸ are visible in the density map, suggesting good modeling reliability (Fig EV3B). The structures of the two states also display a well‐ordered density of nucleic acids including the bubble region and the newly synthesized RNA (Fig EV3C and D).

Figure EV2. Local resolution maps for σ²⁸‐TIC .

• A, B
  The overall local resolution map (top) and the local resolution maps for individual parts (bottom) of structure 1 (state‐1 σ²⁸‐TIC) (A) and structure 2 (state‐2 σ²⁸‐TIC) (B). The local resolution map is contoured at 1.0 of view value of Chimera.

Figure EV3. Alignments of the σ²⁸ factor with other factors and the nucleic acid structures in σ²⁸‐TICs.

• A
  The sequence alignment of the σ²⁸ factor with the σ³⁸, σ⁷⁰ factor in E. coli. Sequences are presented with the one‐letter amino acid codes. The second structure (α helices) and the conserved regions are labeled above and below the sequences, respectively. The black triangles are used to denote the promoter‐recognition amino acids of σ²⁸ factor.
• B
  Overall structure of the E. coli σ²⁸ factor in the TICs. The side chains are shown as stick mode, and the transparent density map is contoured at 1.0 of view value of Chimera (left). The E. coli σ²⁸ factor in the σ²⁸‐TIC, the σ³⁸ factor in the σ³⁸‐TIC (PDB ID 5IPL), and the σ⁷⁰ factor in the σ⁷⁰‐TIC (PDB ID 4YLN) are superimposed (right).
• C, D
  The transparent split cryo‐EM maps (contoured at 1.0 of view value of Chimera) for the promoter DNA, the nascent RNA, and the active site Mg ion in the σ²⁸‐TIC at the state 1 (C) and the state 2 (D) are shown, respectively. The red labeled regions on the promoter DNA are the important recognition regions. The color schemes for others are the same as in Fig 1. The right insets are the zoom‐in views of the DNA‐RNA hybrid.

The two structures of σ²⁸‐TIC complexes show high similarity, with a root‐mean‐square deviation (RMSD) of 0.255 Å (Cα aligned). The major difference in two structures lies in the location of β′ ZBD of RNAP. In state 1 structure, ZBD is adjacent to the σ4 domain and does not interact with the upstream DNA; in state 2 structure, ZBD gets close to the upstream DNA and makes contact with the NT‐strand DNA (Fig 1B and C).

Stable promoter recognition on the −10 element but weak on the −35 element

The σ²⁸ factor recognizes the −10 and −35 elements via its σ2 and σ4 domains, respectively, which is similar to other σ⁷⁰ family factors in TIC structures (Bae et al, 2015; Zuo & Steitz, 2015). The strong density on the −10 element suggests a stable recognition here. Five positively charged arginine residues of σ²⁸ have pairs of polar interactions with nucleotides of the template −10 element: R95/−13G, R95/−12C, R98/−12C, R98/−11T, R84/−12C, R34/−11T, and R94/−11T (Fig 2A and B), which explains previous observation that single mutation in R84 or R98 significantly decreased σ²⁸‐dependent transcription (Koo et al, 2009). In addition, several residues in the β and β′ subunits (βN494, βK496, βR470, β′S319, and β′R259) also make contacts with the template strand (T‐strand) −10 element via hydrogen bond or polar interactions (Fig 2A and B), which are similar to those observed in the σ³⁸‐TIC (Liu et al, 2016). The NT‐strand −10 element is contacted by the region of σR2.3, and the path down to the main cleft is the same as that in previously reported TICs (Feklistov & Darst, 2011; Zhang et al, 2012). Multiple residues of the σR2 region contribute to the recognition of the sequence: R58/−13C, R74/−13C, Q73/−12G, Q63/−11A, T69/−9A, and H26/−7C (Fig 2A and B). In addition to recognizing the core −10 element, σ²⁸ also makes contact with the relatively conserved extended −10 motif (Fig 2C) via side chains of R91 (Fig 2A and D), which is consistent with the previous report (Koo et al, 2009). The core −10 element C₋₁₃G₋₁₂ATAAN₋₇ is conserved in promoters of σ²⁸‐dependent flagellar genes (Fig 2C and Appendix Table S2) (Yu et al, 2006a; Zhao et al, 2007; Fitzgerald, 2014), and the major difference with that of σ⁷⁰ is the −13C and −12G, which have been determined to be essential for σ²⁸‐dependent promoters (Yu et al, 2006a; Koo et al, 2009). Consistently, sequence alignment shows that residues with direct contact to these two base pairs in σ²⁸ (R58, Q73, R74, R95, and R98) are different from those in σ⁷⁰ and σ³⁸ (Fig EV3A).

Figure 2. Promoter recognition by the σ²⁸ factor.

1. Summary of protein–nucleic acid interactions. Solid and dashed lines represent polar interactions and hydrogen bonds, respectively. The −35 and −10 elements are colored in red letters. The extended −10 motif is in magenta.
2. Recognition of the promoter −10 element. The RNAP is shown in surface representation: β subunit, cyan; β′ subunit, pink; β′ lid, green; β′ rubber, yellow; and σ²⁸, wheat. The right and left side insets at the bottom part are the zoom‐in views of the interactions on the NT‐strand and T‐strand, respectively.
3. Conserved sequences around the −10 element of flagellar promoters (from −16 to −7 positions) generated by WebLogo (Crooks et al, 2004). Promoter sequences used in this analysis are shown in Appendix Table S2.
4. Recognition of the promoter extended −10 element.
5. Recognition of the promoter −35 element by the σ4 domain.
6. Influence of mutating σ4 on promoter recognition.

Data Information: In (F), quantifications are mean ± SD from three determinations. Statistical analyses were performed using the unpaired Student's t‐test (two‐tailed). **P < 0.01.Source data are available online for this figure.

Apparently, the interactions between σ²⁸ and −35 element are not as extensive as the recognition on −10 element. Being different from previously predicted T_‐34AAAGTTT_‐27 sequence (Yu et al, 2006b), our σ²⁸‐TIC structures show that A_‐36ATAAAG_‐30, which is corresponding to the position of −35 element determined in σ⁷⁰‐TIC structures (Appendix Fig S1), should be the −35 element region recognized by σ²⁸ (Fig 2A and E). Two positively charged residues recognize the phosphate groups of the −35 element: K208/−30C (T‐strand), K208/−31T (T‐strand), and R220/−36A (NT‐strand) (Fig 2A and E). Mutation of the two residues (K208A‐R220A) significantly decreased the transcriptional activity of σ²⁸‐RNAP holoenzyme on tarp (Fig 2F). The number of residues recognizing −35 element in σ²⁸ is obviously less than that in σ⁷⁰ (Campbell et al, 2002), indicating that the interaction might be less extensive than that for σ⁷⁰. The first round of focused classification and refinement also showed that only 33.33% of the particles have clear densities on −35 element region (Fig EV1D), suggesting a weak recognition or a less frequently occupied conformation. In addition, on the basis of the comparison with the σ³⁸‐TIC structure (PDB ID 5IPL) and σ⁷⁰‐TIC structure (PDB ID 4YLN), the position of σ4 helices in our structure is similar to that in the σ³⁸‐TIC, while they are around 2–3 Å farther away from the −35 element than the σ4 helices in σ⁷⁰‐TIC (Appendix Fig S1), indicating that the recognition on the −35 element in the σ²⁸‐TIC is weaker than that in the σ⁷⁰‐TIC.

Recognition of the upstream NT‐strand DNA in the spacer by β′ ZBD

Remarkably, β′ ZBD of RNAP adopts an extended conformation in the state 2 structure compared with that in the state 1 structure (Fig 1C). The β′ ZBD locates at the N‐terminal of β′ subunit, and one zinc ion is coordinated by four cysteines of ZBD (Fig 3A). The β′ ZBD in the state 1 architecture is around 12 Å away from the upstream DNA and makes contact with the σ4 domain; in the state 2 structure, ZBD is dissociated from the σ4 domain with a shift of ~9 Å and reaches to the upstream NT‐strand DNA (Figs 1C, 3A, and Appendix Fig S2A). In the σ³⁸‐TIC structure (PDB ID 5IPL) and σ⁷⁰‐TIC structure (PDB ID 4YLN), β′ ZBDs situate at similar positions as observed in the state 1 structure. Superimposition of them with our state 2 structure also shows ~10 Å distance shift of Zn²⁺ (Appendix Fig S2B and C). The interface on β′ ZBD possesses a patch of positive charges (Appendix Fig S2D). Together with the highly positively charged regions of the σ2, σ3, and σ4 domains (Appendix Fig S2D), we propose that interactions on β′ ZBD might contribute to a more stable promoter binding and more efficient transcription initiation.

Figure 3. Role of β′ ZBD in activating σ²⁸‐dependent transcription.

• A
  Specific interactions between the upstream DNA (−25 and −26 nucleotides) and β′ ZBD.
• B, C
  Effects of deleting β′ ZBD on transcriptional activities of RNAP on different promoters. The fliCp (B) was used as template for σ²⁸‐dependent transcription, and the lacUV5p (C) was used in σ⁷⁰‐dependent transcription. RNA products indicated by solid triangles were quantified, and the signal obtained from 100 nM wild‐type RNAP was normalized to 1 in each test, respectively.
• D
  Transcriptional activities of ZBD‐mutated σ²⁸‐RNAP on fliCp. RNA products indicated by solid triangles were quantified.
• E
  Binding of σ²⁸‐RNAP with fliCp as detected by EMSA (left). The heparin‐resistant RNAP–promoter open complex is indicated as “RPo”, and the heparin‐sensitive RNAP–promoter closed complex is shown as “RPc”. Percentages of RNAP–promoter open complex (RPo) and closed complex (RPc) in EMSA were performed using σ²⁸‐RNAP and fliCp (right).
• F
  DNase I footprinting of the fliCp complex with σ²⁸‐RNAP. Promoter DNA was labeled on the non‐template strand. The promoter region protected by RNAP is indicated by a red box at the bottom panel.

Data Information: In (B–E), data are mean ± SD from three determinations. Individual values of replicates are shown as dots. Statistical analyses were performed using the unpaired Student's t‐test (two‐tailed). *P < 0.05, **P < 0.01.Source data are available online for this figure.

Consistent with our hypothesis, the ZBD‐deleted RNAP (ΔZBD) showed reduced transcriptional activity when reconstituted with σ²⁸ on tarp (Fig EV4A and B) or fliCp promoter (Kundu et al, 1997) (Figs 3B and EV4C), but did not decrease activity with the σ⁷⁰ on the classical lacUV5p (Figs 3C and EV4D), suggesting a specific role of β′ ZBD. Moreover, the ZBD interacts with the NT‐strand mainly through polar interactions: K74/−26C, K74/−25C, and K87/−26C (Fig 3A). Single mutation of either K74 or K87 (named as K74A and K87A, respectively) or both residues (named as K74A–K87A), or mutation of four cysteines in β′ ZBD (named as 4CS) in the RNAP core enzyme, all showed decreased transcriptional efficiency when reconstituted with σ²⁸ (Figs 3D, and EV4E–G), but not with σ⁷⁰ on lacUV5p (Fig EV4H). Additionally, the ZBD‐mutated RNAPs showed decreased promoter binding on fliCp when reconstituted with σ²⁸ (Fig 3E and F), but did not significantly influence promoter binding with the σ⁷⁰ on lacUV5p (Fig EV4I and J). Importantly, the ZBD‐mutated RNAPs formed a higher ratio of unstable heparin‐sensitive “closed complex” and a lower ratio of heparin‐resistant “open complex” for σ²⁸ (Fig 3E). Consistently, DNase I footprinting assay showed that σ²⁸‐WT RNAP protected the fliCp from +16 to −41 positions (Fig 3F), a representative protected region for RPo complex (Chen et al, 2020), but σ²⁸ reconstituted with ΔZBD showed a shorter protected region from +6 to −39 (Fig EV4K), which is characteristic of intermediates between RPc and RPo (Chen et al, 2020). Together, these in vitro data reveal that β′ ZBD significantly contributes to promoter binding of σ²⁸‐RNAP and promotes the formation of heparin‐stable complexes, some of which are RPo.

Figure EV4. Role of β′ ZBD in promoter binding and transcription.

• A–G
  Full image of transcripts from wild‐type (WT) or ZBD‐mutated RNAPs on different promoters. The amount of abortive transcripts showed a similar trend as runoff products for each sample. The tarp (A, B) and fliCp (C) were used as templates for σ²⁸‐dependent transcription, the lacUV5p (D) was used in σ⁷⁰‐dependent transcription, and the fliCp (E) and tarp (F, G) were used as templates for transcription of ZBD point‐mutant RNAPs. The signal obtained from 100 nM wild‐type RNAP was normalized to 1 in each test, respectively. Data are mean ± SD from three determinations. Individual values of replicates are shown as dots. Statistical analyses were performed using the unpaired Student's t‐test (two‐tailed). *P < 0.05, **P < 0.01.
• H
  Effects of deleting β′ ZBD on transcriptional activities of σ⁷⁰‐RNAP on lacUV5p.
• I
  Bindings of σ⁷⁰‐RNAP with lacUV5p as detected by EMSA.
• J
  DNase I footprinting of the lacUV5p complex with σ⁷⁰‐RNAP. The promoter region protected by RNAP is indicated by red box at the bottom panel.
• K
  DNase I footprinting of the fliCp complex with σ²⁸‐RNAP‐ΔZBD. The promoter region protected by σ²⁸‐RNAP‐ΔZBD is indicated by red box, and regions showing reduced protection by σ²⁸‐RNAP‐ΔZBD compared with wild‐type σ²⁸‐RNAP are indicated by red‐dotted boxes.

Source data are available online for this figure.

β′ ZBD contributes to expression of σ²⁸‐dependent genes and bacterial motility

To investigate whether the roles of β′ ZBD revealed in vitro could be validated in vivo, we have tried to construct E. coli strains with mutations in the rpoC gene (encoding the β′ subunit) using a CRISPR/Cas9 system (Fig 4A) (Jiang et al, 2015). In spite of several attempts, we failed to obtain the strain with β′ ZBD deletion (see Discussion). However, we have successfully obtained the mutants with single mutation at β′K74 (named as Ec‐K74A) or β′K87 (named as Ec‐K87A), and with double point mutations (named as Ec‐K74A‐K87A) (Fig 4A). Promoter activities of both tarp and fliCp, but not the native σ⁷⁰‐dependent lac promoter (lacp), were decreased in these mutants (Figs 4B and C, and EV5A).

Figure 4. Importance of β′ ZBD in controlling transcription of σ²⁸‐dependent genes.

• A
  Schematic diagram of constructing chromosomal mutations in rpoC gene in E. coli based on a CRISPR/Cas9 system. A 20‐nt guide RNA targeting a region between K74 and K87 on β′ subunit is shown in purple. Double‐strand DNA fragments, which contains synonymous mutations (yellow background) in sgRNA targeting region to facilitate clone selection and aimed mutations at K74 or K87 or both residues (only example for constructing the Ec‐K74A‐K87A strain is shown), were used as donor templates.
• B, C
  Promoter activities of tarp (B) and fliCp (C) in E. coli wild‐type (Ec‐WT) and its mutants expressing ZBD‐mutated RNAP (Ec‐K74A, Ec‐K87A, and Ec‐K74A‐K87A) shown as mean ± SD from three replicates.
• D
  Volcano plot showing log₂‐fold change of mRNA abundance as a ratio of Ec‐K74A‐K87A and Ec‐WT strains. Blue dots represent genes with significant difference between these two strains. Red dots represent flagellar genes. Data are shown as mean values of three independent colonies.
• E
  Relative mRNA levels of flagellar genes in Ec‐K74A‐K87A strain compared with those in Ec‐WT (normalized to 1 for each gene). Diagram for the role of σ²⁸ (fliA gene) in flagellar regulon hierarchy (Clarke & Sperandio, 2005; Fitzgerald, 2014) is shown in the left panel. Data are mean ± SEM from three replicates.
• F
  Motilities of Ec‐K74A‐K87A strains with over‐expression of wild‐type rpoC gene (Ec‐K74A‐K87A‐C) or carrying the control plasmid (Ec‐K74A‐K87A‐V).
• G
  Activities of fliCp in Ec‐K74A‐K87A‐C and Ec‐K74A‐K87A‐V strains. L‐arabinose was used at final concentration of 0.002%. Data are mean ± SD from three colonies.

Data Information: In (B), (C), (E), and (G), statistical analyses were performed using the unpaired Student's t‐test (two‐tailed). *P < 0.05, **P < 0.01.Source data are available online for this figure.

Figure EV5. Role of β′ ZBD in controlling transcription in vivo .

1. Activities of native lac promoter (lacp) in Ec‐WT and Ec‐K74A‐K87A as measured by β‐galactosidase test. Bacteria were grown to late logarithmic growth phase at 30°C. Individual values of biological replicates (n = 6) are shown as dots, and the mean ± SD values are displayed as error bars. Unpaired Student's t‐test (two‐tailed) was applied in statistical analyses. *P < 0.05.
2. Motilities of E. coli wild‐type and the β′ ZBD‐mutated strains.
3. Promoter activities of fliCp in Ec‐WT strain and Ec‐K74A‐K87A strain carrying the control plasmid (vec) or with over‐expression of FliA (FliA). L‐arabinose was added at final concentration of 0.002%. Data are mean ± SD from three colonies.
4. Motilities of Ec‐K74A‐K87A strains with over‐expression of wild‐type fliA gene (FliA) or carrying the control plasmid (vec).
5. Motilities of Ec‐WT and Ec‐CK strains. Sequences of partial ZBD domain in Ec‐WT and Ec‐CK strains are shown in the left panel. Synonymous mutations introduced in the Ec‐CK strain are indicated in yellow background.
6. Activities of osmYp in Ec‐K74A‐K87A strain with over‐expression of wild‐type rpoC gene (Ec‐K74A‐K87A‐C) or carrying the empty plasmid (Ec‐K74A‐K87A‐V). Data are mean ± SD from three colonies. Statistical analyses were performed using the unpaired Student's t‐test (two‐tailed). **P < 0.01.
7. Growth curve of Ec‐WT and Ec‐K74A‐K87A strains at 37°C. Data are mean ± SD from three colonies.

Source data are available online for this figure.

Transcriptional profiling of Ec‐WT and Ec‐K74A‐K87A strains showed that expression of all flagellar genes was decreased in the Ec‐K74A‐K87A strain (Fig 4D). Consistently, quantitative RT–PCR analysis displayed that transcription of all tested σ²⁸‐dependent flagellar genes (Fig 4E) presented decreased mRNA levels in the Ec‐K74A‐K87A strain. In addition, the Ec‐K74A‐K87A strain appeared almost non‐motile in both swimming and swarming assays (Fig 4F), and mutation of either K87 or K74 also impaired bacterial motility (Fig EV5B). Over‐expression of σ²⁸ (FliA) in the Ec‐K74A‐K87A strain could restore neither the fliCp activity nor the bacterial motility (Fig EV5C and D), but both fliCp activity and bacterial motility can be partially restored by over‐expression of a wild‐type rpoC gene in the Ec‐K74A‐K87A strain (Fig 4F and G). These data suggest that the repressive effect in the Ec‐K74A‐K87A strain is due to the mutation of ZBD but not the repressed expression of σ²⁸. Another constructed strain named Ec‐CK, which carries all synonymous mutations in the rpoC gene of the Ec‐K74A‐K87A strain, behaved similarly as the Ec‐WT strain (Fig EV5E), further indicating that the loss of motility in the Ec‐K74A‐K87A strain is due to the mutations in K74 and K87 residues.

β′ ZBD is generally important for transcription initiated by σ factors without strong σ4/−35 element interactions

Based on the fact that the interactions between σ2 and −10 element are strong but the ones between σ4 and −35 element are weak in σ²⁸‐TIC structures, and that β′ ZBD directly interacts with promoter region close to the −35 element to stabilize the RNAP–promoter complex, we propose that the contribution of β′ ZBD would compensate for the weak interaction between σ4 with −35 element during transcription initiation. To test this hypothesis, we constructed a chimeric sigma factor, named as σ²⁸‐R4m, in which the σ4 region of σ²⁸ was replaced by that of σ⁷⁰ (Fig 5A). Correspondingly, the −35 element of σ²⁸‐dependent tarp promoter was also replaced by σ⁷⁰‐preferred “TTGACA” sequence to obtain a chimeric promoter tarp‐35m (Fig 5A). The chimeric σ²⁸‐R4m efficiently recognized chimeric tarp‐35m, and mutations in β′ ZBD did not significantly affect the transcriptional activity of σ²⁸‐R4m‐type RNAP on tarp‐35m promoter (Fig 5B), supporting the hypothesis that the importance of β′ ZBD in transcription is associated with the strength of interaction between σ4 and −35 element.

Figure 5. β′ ZBD contributes to transcription initiated by σ factors with weak σ4/−35 element interactions.

• A
  Diagram for mutations in σ²⁸ and tarp. Conserved domains in σ factor are shown as numbering. The region 4 of σ²⁸ was replaced by that from σ⁷⁰ to obtain the chimeric σ²⁸‐R4m, and the sequence of −35 element in tarp was replaced by “TTGACA” in the chimeric tarp‐35 m. The proposed strength for interaction between σ4 and −35 in different σ/promoter pairs is indicated by the thickness of arrow.
• B
  Activities of chimeric σ²⁸‐R4m in initiating transcription from tarp‐35m, when reconstituted with wild‐type (WT) or ZBD‐mutated RNAP core enzyme.
• C–E
  Effects of mutating β′ ZBD on transcriptional activities of σ³⁸‐RNAP (C, D) or σ⁷⁰‐RNAP (E) on osmYp.

Data Information: In (B–E), quantifications of RNA products indicated by solid triangles from three determinations are shown as mean ± SD at the bottom panel. Statistical analyses were performed using the unpaired Student's t‐test (two‐tailed). ns, not significant, *P < 0.05, **P < 0.01.Source data are available online for this figure.

In addition to σ²⁸, the σ³⁸ is known to recognize promoters with high sequence variations at the −35 element (Wise et al, 1996; Maciag et al, 2011), suggesting a weak underlying interaction between σ4 and −35 element. Consistent with our hypothesis, mutations in ZBD also decreased the transcriptional efficiency of σ³⁸‐RNAP holoenzyme on a σ³⁸‐dependent osmYp promoter (Fig 5C and D), which contains a conserved −10 element and a non‐conserved −35 element (Fig 5C) (Yim et al, 1994). Consistently, the activity of osmYp was decreased in Ec‐K74A‐K87A strain and partially restored in the Ec‐K74A‐K87A strain with over‐expression of wild‐type rpoC gene (Fig EV5F).

The σ⁷⁰ can also recognize promoters with non‐conserved −35 element, such as osmYp (Colland et al, 2000). Comparing with the lacUV5p, the non‐conserved −35 element in osmYp would decrease the strength of σ4/−35 element interaction for σ⁷⁰‐RNAP (Fig 5E). Mutations in β′ ZBD all decreased the activity of σ⁷⁰‐RNAP holoenzyme on osmYp (Fig 5E), which is different from the results on lacUV5p (Fig EV4D and H). These data indicate that β′ ZBD is also important for σ⁷⁰ to recognize promoters with non‐conserved −35 element. Consistently, the Ec‐K74A‐K87A strain showed the delayed bacterial growth (Fig EV5G), which could not be explained by functional deficiency of σ²⁸ or σ³⁸, since the strains with deletion of fliA gene (encoding the σ²⁸) or rpoS gene (encoding the σ³⁸) in E. coli did not affect bacterial growth (Schellhorn et al, 1998; Wood et al, 2006), although we could not exclude the possibility that this ZBD mutant may also influence bacterial growth through compromising post‐initiation steps of transcription (Sen et al, 2002; King et al, 2004).

Altogether, both in vitro and in vivo data suggest that, independent of the types of sigma factors, strengthening the interaction between σ4 and −35 element would reduce the dependence of β′ ZBD domain in transcription, but transcription from promoters with weak interaction between σ4 and −35 element requires the contribution of β′ ZBD domain for maximal efficiency. Therefore, the ZBD of RNAP functionally augments the σ4 domain of the initiation factor in recognizing the −35 region of a bacterial promoter.

Discussion

In bacterial RNA polymerase, the ZBD locates at the N‐terminal portion of RNAP β′ subunit. ZBD has been implicated in stabilizing the transcription elongation complex (Nudler et al, 1998) and has been shown later that ZBD interacts with product RNA located upstream of the catalytic center and the RNA‐DNA hybrid in the elongation complex to modulate transcription termination (King et al, 2004; Krupp et al, 2019). It was also reported that ZBD participates in the salt‐resistant interaction with double‐stranded DNA and the cysteine mutation renders the elongation complex extremely sensitive to salt and considerably less processive (Nudler et al, 1996). As for the transcription initiation, the mycobacterial TIC with RbpA (a transcription regulator in mycobacteria) shows that the core domain of RbpA makes extensive contacts with β′ ZBD domain (Hubin et al, 2017a), indicating a possible role of ZBD in transcription initiation.

Here, we provide physical model for the role of β′ ZBD in transcription initiation. During the process of transcription initiation, β′ ZBD in core RNAP is first in a canonical position without extension or relocation. The unconstrained σ²⁸ factor binds the core RNAP to form σ²⁸‐RNAP holoenzyme, which then recognizes the specific promoter to form an RPc. The unstable RPc can isomerize into a stable RPo through formation of several intermediate complexes, which can also isomerize back to form an equilibrium (Saecker et al, 2011; Boyaci et al, 2019). The transition from RPc to RPo is stabilized by RNAP–promoter interaction (Saecker et al, 2011), but the interaction between σ4 and −35 element is relatively weak for σ²⁸. Therefore, we propose that the relocation of ZBD results in the direct interaction with the NT‐strand of promoter spacer region, which would have an overall effect of stabilizing the RPo and therefore increase rates of transcription initiation (Fig 6). This is supported by our EMSA data that mutation of ZBD increased the ratio of RPc/RPo (Fig 3E). Since the RPc is unstable, the total amount of RNAP–promoter complexes was decreased for ZBD‐mutated RNAPs as observed in DNase I footprinting assays (Fig 3F). Whether the ZBD would also contribute to other transcriptional steps is still an open question requiring further investigation. Previous studies have suggested that the intrinsic zinc ions in RNAP might play roles in substrate or template binding (Wu & Wu, 1987) and initiation of RNA synthesis (Scrutton et al, 1971). Additionally, Murakami et al also proposed the possibility of the interaction of β′ ZBD with DNA backbone in the spacer between −22 on the T‐strand and −27 on the NT‐strand based on the structure of a RNAP holoenzyme–DNA complex (Murakami et al, 2002a). Our structures and biochemical assays demonstrated the positive role of ZBD in controlling transcription initiation and provided important and direct evidences to support these previous biochemical studies and proposals.

Figure 6. The ZBD relocation mechanism in transcription initiation.

A schematic cartoon model of ZBD relocation mechanism during σ²⁸‐mediated transcription initiation is presented. The ZBD relocation shows an overall effect of facilitating the isomerization of closed RNAP–promoter complex (RPc) to open RNAP–promoter complex (RPo) by contributing to RNAP–promoter interaction. The RPc with the relocated ZBD (state 2) has a higher efficiency in isomerizing to RPo or is more stabilized than the one with the state 1 ZBD, consequently advancing transcription efficiency. The inset is the zoom‐in view of ZBD relocation from state 1 to state 2.

The novel role of β′ ZBD is associated with the strength of interaction between σ4 and −35 element. The β′ ZBD has positive effects in σ²⁸‐, σ³⁸‐, and even in some σ⁷⁰‐directed transcription, but not in σ⁷⁰‐directed transcription on strong promoter, which may be due to a strong σ4/−35 element recognition that is sufficient to stabilize the whole upstream DNA for isomerization and transcription initiation. Promoter sequence analysis based on identified transcriptional start sites in E. coli (as well as in other bacteria) showed that most of promoters contain conserved −10 element but not conserved −35 element (Ettwiller et al, 2016; Hubin et al, 2017b), which suggests that β′ ZBD may be important for transcription of those genes. In support of this hypothesis, the K74A‐K87A mutation in ZBD altered expression of a quarter of genes in E. coli (Fig 4D). These analyses, together with previous reports that the ZBD also contributes to RNAP assembly (Markov et al, 1999) and the transcriptional termination steps (Sen et al, 2002; King et al, 2004), may explain why we could not obtain the E. coli mutant with deletion of β′ ZBD. Previous sequence alignment showed that the β′ ZBD is conserved in 78 prokaryotic‐type β′ subunits (King et al, 2004). Thus, all these suggest that the ZBD relocation mechanism could be widely employed by bacterial RNAPs to enhance transcription initiation from promoters with weak σ4/−35 interactions.

With the nucleic acid‐binding property, the positively charged residues of β′ ZBD make contacts only with the phosphate backbone of the upstream NT‐strand DNA and therefore should have no specific sequence requirement, which is different from sequence‐specific interaction between the β′ zipper domain and promoter spacer DNA (Yuzenkova et al, 2011). After analyzing the E. coli σ⁷⁰‐TIC (PDB ID 4YLN) (Zuo & Steitz, 2015), σ³⁸‐TIC (PDB ID 5IPL) (Liu et al, 2016), and Mycobacterium tuberculosis σ^H‐TIC (PDB ID 5ZX2) (Li et al, 2019) structures, we observed that the β′ ZBD domains have similar polar interactions with the upstream NT‐strand DNA: K74 in σ⁷⁰‐TIC, K87 in σ³⁸‐TIC, and R77 in σ^H‐TIC (corresponding to K87 in E. coli). All these suggest that the two residues (K74 and K87) observed in our σ²⁸‐TIC structures, which are involved in stabilizing the promoter binding, are the common residues among these structures, indicating a general mode of action for β′ ZBD in transcription.

Sigma factors are recruited to form RNAP holoenzyme to recognize specific promoters and regulate transcription initiation. Previous structural studies have mainly focused on molecular mechanisms for transcription from promoters with strong interaction to RNAP (Murakami et al, 2002a; Zhang et al, 2012; Zuo & Steitz, 2015; Liu et al, 2017). In this study, the flagella‐specific σ²⁸‐TIC structures provide key insights into mechanisms of transcription initiation from promoters with weak interaction between σ4 and promoter −35 element, in which the β′ ZBD relocates to make contact with the upstream NT‐strand DNA in the spacer region to compensate for the weak σ4/−35 element interaction and promote the isomerization of RNAP–promoter closed‐to‐open complex. Importantly, our results suggest that ZBD relocation is a general mechanism employed by a broader set of sigma factors/promoters in transcription initiation.

Materials and Methods

Bacterial strains, plasmids, and oligonucleotides

Bacterial strains, plasmids, and oligonucleotides generated and used in this work are listed in Appendix Tables S3 and S4.

Expression and purification of RNAP core enzyme

The E. coli RNAP core enzyme was prepared by co‐expression of genes for RNAP β subunit, C‐terminally His‐tagged β′ subunit, α subunit, and ω subunit using pVS10‐RNAP in E. coli BL21(DE3) as described previously (Zhi et al, 2003; Belogurov et al, 2007; Liu et al, 2016, 2017). Mutations in rpoC gene in pVS10‐RNAP plasmid were introduced by overlap PCR. Mutated fragments were digested by Sbf I and Hind III, which were then inserted into the same digested pVS10‐RNAP plasmid.

Expression and purification of σ factors

For construction of the plasmid pET21a‐Ecoσ²⁸, the E. coli σ²⁸ gene was PCR‐amplified from E. coli genomic DNA and ligated into pET21a vector which contains C‐terminal His₆‐tag using Nde I and Xho I restriction sites. The amplified E. coli σ⁷⁰ gene was inserted into pET28a vector between Nhe I and Hind III restriction sites to obtain pET28a‐EcorpoD, which expresses His₆‐tag at the N‐terminal of σ⁷⁰. The E. coli σ³⁸ gene was also cloned into pET28a using a ClonExpress II One Step Cloning Kit (Vazyme). A fragment expressing His₆‐tagged MBP and a TEV cleavage site was introduced in fusion at the N‐terminal of σ³⁸ gene. All mutations in plasmids were introduced by oligos, and the PCR‐amplified plasmid fragments were then cycled using the ClonExpress II One Step Cloning Kit.

All these constructs were transformed into E. coli BL21(DE3) competent cells. The cells were grown in LB medium with 100 μg/ml ampicillin (pET21a) or 50 μg/ml kanamycin (pET28a) at 37°C to an OD₆₀₀ of 0.6 and induced with 0.5 mM isopropyl‐β‐D‐thiogalactopyranoside (IPTG) to express σ factors at 16°C for 16–18 h. The bacteria were then harvested by centrifugation; resuspended in lysis buffer containing 50 mM Tris–HCl, pH 8.0, 300 mM NaCl, 10 mM imidazole, and 5% (v/v) glycerol; lysed by ultrasonication; and then centrifuged at 80,000 g for 1 h using a T29‐8x50 fixed‐angle rotor (Sorvall LYNX 6000 Superspeed Centrifuge, Thermo Scientific). For σ²⁸ purification, the recombinant protein was purified through 5‐mL HisTrap HP column (GE Healthcare), and the eluted protein was further purified through 5‐mL HiTrap Heparin HP column (GE Healthcare). Finally, the protein was applied to a gel filtration column, 120 ml HiLoad 16/600 Superdex 200 pg (GE Healthcare) in a buffer containing 20 mM Tris–HCl, pH 8.0, and 150 mM NaCl. The peak fractions were pooled and concentrated to around 5 mg/ml. The final sample was aliquoted, flash‐frozen in liquid nitrogen, and stored at −80°C until use. The recombinant σ⁷⁰ was purified in a similar way, but a Superdex 200 Increase 10/300 GL column (GE Healthcare) was used for gel filtration analysis. The recombinant σ³⁸ was firstly passed through 5‐ml HisTrap HP column (GE Healthcare), and the eluted protein was diluted and incubated with purified His₆‐TEV protein expressed by pRK793 plasmid (Kapust et al, 2001). The σ³⁸ protein was then passed through the HisTrap HP column again to remove His₆‐MBP and His₆‐TEV enzymes. Fractions passed through the HisTrap HP column were collected for further purification with the Superdex 200 Increase 10/300 GL column (GE Healthcare).

Assembly and purification of σ²⁸‐TIC

The synthetic DNA scaffold used in assembly is the σ²⁸‐specific tarp promoter with minor changes in the sequence to form the bubble, which corresponds to the promoter region between position −39 and +15 relative to the transcription start site (Fig 1A). The promoter DNA was prepared by annealing the NT‐strand DNA to an equal molar amount of T‐strand DNA. The σ²⁸‐TIC was assembled by directly incubating RNAP core enzyme with a threefold molar amount of purified σ²⁸ protein and the preformed promoter DNA scaffold in buffer A containing 20 mM Tris–HCl, pH 7.5, 50 mM NaCl, and 5 mM MgCl₂ at 37°C for 15 min in the presence of ATP and CTP (0.2 mM each). The reaction mixture was then used for purification through size‐exclusion chromatography with Superdex 200 Increase 10/300 GL column (GE Healthcare) in buffer A to remove extra σ²⁸ protein and nucleic acids.

Cryo‐EM grid preparation and data acquisition

After purification through the size‐exclusion chromatography, 3.5 μl of peak 2 fraction of the purified σ²⁸‐TIC at about 1 μM was applied on Quantifoil R2/2 200‐mesh Cu grids (EM Sciences) glow‐discharged at 15 mA for 60 s. The grid was then blotted for 3 s at 4°C under the condition of 100% chamber humidity and plunge‐frozen in liquid ethane using a Vitrobot Mark IV (FEI). The grids were imaged using a 300 keV Titan Krios microscope (FEI) equipped with a Falcon III direct electron detector (FEI) at the Hormel Institute, University of Minnesota. Data were collected in counted mode with a pixel size of 0.9 Å and a defocus range from −1.6 to −2.6 μm using EPU (FEI). Each micrograph consists of 30 dose‐framed fractions and was recorded with a dose rate of 0.8 e⁻/pixel/s (1 e⁻/Ų/s). Each fraction was exposed for 1 s, resulting in a total exposure time of 30 s and the total dose of 30 e⁻/Ų.

Image processing

Cryo‐EM data were processed using cisTEM (Grant et al, 2018), and the procedure is outlined in Fig EV1. A total of 3,993 movies were collected. Beam‐induced motion and physical drift were corrected followed by dose weighting using the Unblur algorithm (Grant & Grigorieff, 2015). The contrast transfer functions (CTFs) of the summed micrographs were determined using CTFFIND4 (Rohou & Grigorieff, 2015). From the summed images, particles were then automatically picked based on a matched filter blob approach with the parameters: maximum particle radius (80 Å), characteristic particle radius (60 Å), and threshold peak height (1.5, standard deviation above noise); 30 Å of highest resolution was used; and high variance area and areas with abnormal local mean were avoided (Sigworth, 2004). In total, 1,067,697 particles were picked and selected to construct a refinement package with 160 Å of estimated largest dimension/diameter and 324 pixels of box size for particles. 2D classifications (Scheres et al, 2005) were performed using 300‐40/8 Å (start/finish, high‐resolution limit) data and without inputting starting reference. In the first round of 2D classification, 126 of 200 classes were manually chosen by removing poorly populated classes containing unsuitable particles and obvious Einstein from noise to construct a new refinement package of 760,853 particles and subjected to the second round of 2D classification. Then, 46 of 200 classes were selected to construct a new refinement package with 435,521 particles and subjected to ab initio 3D reconstruction (Grigorieff, 2016) to generate an initial 3D model using 20‐8 Å resolution data. The initial 3D model was set as the starting reference for further 3D auto‐refinement. FSC at the criteria of 0.143 resulted in a 3.53 Å resolution for the map outputted from auto‐refinement (Chen et al, 2013). Due to the flexibility of upstream DNA, we performed a focused classification (4 classes) and refinement in 3D manual and local refinement using 300‐8 Å resolution data and a sphere of 40 Å radius centered at the region of σ4. Good classes were obtained after 30 cycles of refinement. The third class (33.33%) was reconstructed with 145,159 particles to generate the architecture at 3.81 Å resolution, with a better density of upstream DNA with clear density on the −35 element. We further performed a second round of focused classification (4 classes) and refinement in 3D manual and local refinement with a sphere of 25 Å radius centered at β′ ZBD. Good classes were obtained after 30 cycles of refinement. The first class (30.20%) and the third class (21.60%) were reconstructed with 43,838 and 31,354 particles to generate two structures both at 3.91 Å resolutions: structure 1 (state 1) and structure 2 (state 2), which then were further refined to 3.86 Å and 3.91 Å resolution, respectively. The other two classes (class 2 and class 4) have poor density on either the σ4 domain or β′ ZBD.

Model building and refinement

The initial models were generated by docking the previous structures of the components in the RNAP core (PDB ID 6B6H) into the individual cryo‐EM density maps outputted from the focused classification and refinement using Chimera (Pettersen et al, 2004) and COOT (Emsley & Cowtan, 2004). The 3.86 Å or 3.91 Å cryo‐EM density maps allowed us to build the σ²⁸ factor by first manually docking and then mutating the structure of σ³⁸ (PDB ID 5IPL) according to the sequence alignment result and the density map, and to dock or build β′ ZBD and the RNA transcript (AAC) at the active site starting from +1 position in COOT. The clear corresponding density allowed us to manually build the promoter DNA scaffold in COOT as well. The ω subunits and β′ rim helices in two structures were also fitted in maps although poor densities suggest the flexibility or low occupancy, while no density allowed us to fit the CTD of the α subunit of RNAP into the map. Although the low occupancy of the ω subunit was observed in the structures, it should have no influence on the normal architecture of RNAP and β′ ZBD relocation since the main part of RNAP architecture is same as previously described and the ω subunit is at least 50 Å apart from β′ ZBD. The intact models were then real‐space‐refined using Phenix. The final maps were put into a P1 unit cell (a = b = c = 291.6 Å; α = β = γ = 90°), and the structural factors were calculated in Phenix (Adams et al, 2010). In the real‐space refinement, minimization global and local grid search were performed with the secondary structure, rotamer, and Ramachandran restraints applied throughout the entire refinement. The final models have good stereochemistry by evaluation in MolProbity (Chen et al, 2010). To calculate the local real‐space map CC for the ZBD domain in Chimera (Pettersen et al, 2004), we first extract the ZBD domain (residues 64–95 of β′ subunit) into a separate pdb file and split the density map to generate an individual local density map of ZBD domain. Next, we set the value of 5 Å resolution (the local resolution of ZBD domain is around 3–5.5 Å) to generate a simulated map from the atomic structure and then fit the simulated map with the split density map of ZBD domain. The local resolution maps were estimated and generated by MonoRes (Vilas et al, 2018). Map reconstruction quality was evaluated by Mtriage in Phenix (Afonine et al, 2018). The split cryo‐EM maps were generated using color zone with 1.5 Å coloring radius in volume viewer of Chimera (Pettersen et al, 2004). The statistics of cryo‐EM data collection, 3D reconstruction, and model refinement are shown in Appendix Table S1. All figures were created using Chimera (Pettersen et al, 2004) or PyMOL (Schrödinger LLC, https://pymol.org/2/).

In vitro transcription assay

In vitro transcription assays were performed as previously described (Hu et al, 2012) with minor modifications. Briefly, promoter fragments (~180 bp in length, shown in Appendix Table S2) were amplified from E. coli genomic DNA and purified by a Gel Extraction Kit (Omega). RNAP holoenzyme was assembled by mixing 100 nM RNAP core and 300 nM σ factor (except those indicated in figures) in transcription buffer (20 mM Tris–HCl, pH 7.9, 50 mM NaCl, 5 mM MgSO₄, 1 mM DTT, 0.1 mM EDTA, 5% glycerol). A promoter fragment (20 nM) was incubated with RNAP holoenzyme at 37°C for 10 min. Transcription was initiated by the addition of 50 μM CTP, UTP, and ATP; 5 μM GTP; and 1 μCi of [α‐³²P]GTP. The reactions were carried out at 37°C for 10 min and then stopped by 1 volume of 95% formamide solution. RNA products were heated at 70°C for 5 min and then analyzed on denaturing (7 M urea) 16% polyacrylamide gel electrophoresis (PAGE).

DNA‐binding analysis

For the electrophoretic mobility shift assays (EMSA), the fluorescein‐labeled promoter fragments were mixed with RNAP holoenzyme as described in the in vitro transcription assay, and reactions were incubated for 10 min at 37°C. Then, 10 μg/ml of poly(dA‐dT) was added and incubated for 5 min at 37°C. Afterward, samples were loaded on 6% native 0.5× TBE‐PAGE. Gels were scanned by Amersham Typhoon scanner (GE Healthcare). Where indicated, 10 μg/ml of heparin was added into samples before being loaded to the PAGE.

For DNase I footprinting analyses, the fluorescein‐labeled promoter fragments were incubated with RNAP in a similar way as described in EMSA. After incubation of RNAP holoenzyme with promoter fragment for 15 min at 37°C, samples were treated with 2 U/ml DNase I (Promega) for 1 min at 37°C. Reactions were stopped by the addition of 10 mM EDTA (pH 8.0) and heated at 70°C for 5 min. Samples were then loaded into an Applied Biosystems 3730xl DNA Analyzer (Hu et al, 2012; Zhu et al, 2017).

E. coli mutant construction

A CRISPR‐Cas9 system (Jiang et al, 2015) was applied to construct the mutants based on E. coli MG1655 strain. Briefly, a small guide RNA (sgRNA) targeting rpoC gene (shown in Fig 4A) was introduced in pTargetF plasmid using the ClonExpress II One Step Cloning Kit to obtain pTargetF‐EcorpoC. The MG1655 strain was first transformed with a Cas9‐expressing plasmid pCas. Afterward, the pCas‐containing MG1655 cells were co‐transformed with the pTargetF‐EcorpoC plasmid and an 1,168 bp donor DNA (containing synonymous mutations in sgRNA targeting region and mutations at K74A, or K87A, or both where needed). Mutants were confirmed by DNA sequencing. The plasmids in these strains were cured as described previously (Jiang et al, 2015). The final mutants were named as Ec‐K74A, Ec‐K87A, and Ec‐K74A‐K87A, respectively. The strain that contains only synonymous mutations was named as Ec‐CK, and the parent strain was renamed as Ec‐WT.

For constructing the complementary strains, wild‐type rpoC gene was amplified from E. coli genomic DNA and cloned into pBAD22 plasmid (Guzman et al, 1995) using the ClonExpress II One Step Cloning Kit. The obtained plasmid was named as pBAD‐EcorpoC and was transformed into Ec‐K74A‐K87A strain to get strains named as Ec‐K74A‐K87A‐C. The Ec‐K74A‐K87A strain transformed with pBAD22 plasmid was named as Ec‐K74A‐K87A‐V. Similar procedure was applied for constructing pBAD‐FliA plasmid.

RNA extraction, RNA‐seq, and qRT–PCR analyses

For analyzing expression of flagellar genes, E. coli MG1655 and its mutants were grown in LB medium at 30°C to late logarithmic growth phase (OD₆₀₀ 0.9–1.0). RNA was extracted using TRIzol reagent (Invitrogen) as described by manufacturer's protocol. For RNA‐seq analysis, the rRNA in extracted RNA was removed by Ribo‐off rRNA Depletion Kit (Vazyme), RNA library was constructed using a NEBNext^® Ultra™ Directional RNA Library Prep Kit for Illumina (NEB) and sequenced by Illumina HiSeq X Ten platform.

For quantitative RT–PCR (qRT–PCR) analysis, RNA was first reverse‐transcribed to cDNA using M‐MLV (Promega). The relative amount of target mRNA was analyzed by quantitative RT–PCR (qRT–PCR) using an iTaq Universal SYBR Green Supermix (Bio‐Rad) following the manufacturer's instructions. Relative transcriptional levels of tested genes were normalized to the 16S rRNA levels.

In vivo promoter activity test and motility assay

In vivo promoter activity was tested using a gfp reporter fusion plasmid named as pGT, which was constructed by integration of three fragments: p15A ori, kanamycin resistance gene, and a promoter‐less gfp fragment. Promoter fragments were cloned into pGT using the ClonExpress II One Step Cloning Kit. The pGT constructs were transformed into E. coli MG1655 and its derivative strains, which were grown at 30°C to late logarithmic growth phase. The fluorescence intensity was quantified by Synergy HT plate reader (BioTek) in 96‐well black plate. The promoter activities are shown as relative fluorescence units (RFU), determined as the fluorescence intensities per OD₆₀₀. The native activity of lac promoter in each strain (cells were grown at 30°C in the presence of 0.2 mM IPTG) was tested by measuring the β‐galactosidase expression level (Hu et al, 2009).

For swimming motility assay, 2 μl of each strain (30°C, late logarithmic growth phase) was stabbed into semi‐solid agar medium (1% tryptone, 0.5% NaCl, 0.3% Difco agar) (Burkart et al, 1998). Plates were incubated at 24°C for 14 h for imaging. For swarming test, 2 μl of each strain was spotted on LB plate containing 0.6% Difco agar and 0.5% glucose (Burkart et al, 1998), which was then incubated at 30°C for 15 h before imaging. L‐arabinose was added at 0.2% into plates for strains carrying pBAD22, pBAD‐EcorpoC, or pBAD‐FliA plasmid.

Quantification and statistical analysis

Quantification, statistical analysis, and validation are implemented in the software packages used for 3D reconstruction and model refinement. RNAs from three replicates of in vitro transcription assays and shifted DNAs in EMSA were quantified by ImageJ software. Statistical analyses were performed using the unpaired Student's t‐test (two‐tailed) between each of two groups. Significant difference in RNA‐seq analysis was computed using R (version 3.2.2).

Author contributions

WS, WZ, YJ, and AS performed molecular cloning and protein sample preparations. WS assembled complexes for structure determination. WS and BL performed cryo‐EM grid preparation, screening, and optimization. BL conducted high‐throughput data collection on Titan Krios. WS and BL performed image processing, map generation, model building and refinement, and structural analysis. WZ and YH performed in vitro transcription and DNA‐binding analyses. BZ and YH constructed E. coli mutants and performed most of in vivo tests. SH, WZ, and YH extracted RNA and performed RNA‐seq and qRT–PCR analyses. WS, YH, and BL wrote the manuscript with contributions from all authors.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Expanded View

Review Process File

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

The EMBO Journal (2020) 39: e104389

Data availability

Coordinates data: Protein Data Bank 6PMI (https://www.rcsb.org/structure/6PMI) (state 1 σ²⁸‐TIC) and 6PMJ (https://www.rcsb.org/structure/6PMJ) (state 2 σ²⁸‐TIC). Cryo‐EM map data: Electron Microscopy Data Bank EMD‐20394 (https://www.ebi.ac.uk/pdbe/entry/emdb/EMD-20394) (state 1 σ²⁸‐TIC) and EMD‐20395 (https://www.ebi.ac.uk/pdbe/entry/emdb/EMD-20395) (state 2 σ²⁸‐TIC). RNA‐seq data: Sequence Read Archive SRP239152 (https://www.ncbi.nlm.nih.gov/sra/SRP239152).