Pseudomonas aeruginosa SutA wedges RNAP lobe domain open to facilitate promoter DNA unwinding

Pseudomonas aeruginosa (Pae) SutA adapts bacteria to hypoxia and nutrition-limited environment during chronic infection by increasing transcription activity of an RNA polymerase (RNAP) holoenzyme comprising the stress-responsive σ factor σS (RNAP-σS). SutA shows no homology to previously characterized RNAP-binding proteins. The structure and mode of action of SutA remain unclear. Here we determined cryo-EM structures of Pae RNAP-σS holoenzyme, Pae RNAP-σS holoenzyme complexed with SutA, and Pae RNAP-σS transcription initiation complex comprising SutA. The structures show SutA pinches RNAP-β protrusion and facilitates promoter unwinding by wedging RNAP-β lobe open. Our results demonstrate that SutA clears an energetic barrier to facilitate promoter unwinding of RNAP-σS holoenzyme.

R ecent studies discovered SutA as a new binding partner of Pseudomonas aeruginosa (Pae) RNAP 1,2 . The expression of SutA is induced under anaerobic survival conditions to help maintaining minimal transcription activity of Pae RNAP. Deletion of SutA causes global gene transcription changes, including substantially decreased transcription of genes encoding ribosomal protein and ribosomal RNA, and severe phenotypic defects, including reduced biofilm formation, virulence, and fitness 1 . Bioinformatic analysis suggests that SutA is present in most species in Alteromonadales and Pseudomonadales orders of Gamma-proteobacteria. SutA has no sequence homology to any other characterized proteins. Previous NMR study and secondary structural prediction suggested that SutA comprises a middle αhelix, a N-terminal D/E-rich region, and a C-terminal disordered tail 2 . Both the middle and N-terminal regions are required for its transcription activation activity on the promoter of ribosomal RNA rrn 2 .
RNAP-β protrusion is likely the primary anchor site of SutA 2 . SutA interacts with both RNAP-σ A and RNAP-σ S holoenzymes and has greater effect on increasing transcription activity of the RNAP-σ S holoenzyme compared with RNAP-σ A holoenzyme 2 . Previous results of cross-linking and FeBABE cleavage assays suggested that SutA first associates with RNAP-σ S , subsequently undergoes conformational change upon promoter dsDNA loading and unwinding, and eventually dissociates upon formation of the RNAP-promoter DNA open complex (RPo) 2 . However, due to the lack of structural information of SutA and SutA-RNAP complexes, the detailed interaction between SutA and RNAP is unknown and the molecular mechanism underlying SutA transcription activation remains elusive.
Promoter unwinding by bacterial RNAP is a multiple-step process [2][3][4][5][6] . First, RNAP recognizes the double-stranded promoter DNA (dsDNA) by making sequence-specific interactions with upstream promoter elements (the UP, −35, and/or extended −10 elements) and presents the downstream dsDNA on top of the main cleft; second, RNAP bends and melts the promoter dsDNA at the −10 element and accommodates the nucleotide at the −11 position of non-template strand DNA (NT-11A) in its cognate pocket; third, promoter unwinding propagates to the downstream of the −11 position and the nucleotides at other key positions (NT-7T, NT-6G, and NT + 2 G) are subsequently recognized and secured in respective protein pockets; lastly, the template ssDNA is loaded into and restrained in the active-center cleft ready for initiation of RNA synthesis 7 .
The process of promoter dsDNA unwinding has to overcome several obstacles inside of RNAP, including: 1) the gate loop of RNAP-β lobe domain and the σ 1.2 domain that interact with each other to seal the upper main cleft 8 , 2) the σ 1.1 domain located at the downstream dsDNA channel [9][10][11] , and 3) the fork loop 2 and switch region 2 motifs that seal the lower main cleft 12 . Previous studies have reported cryo-EM structures of RNAP-promoter DNA intermediate complexes that were trapped by these obstacles during RPo formation, suggesting that they are the ratelimiting factors of promoter unwinding 7,12,13 . A collection of biochemical, structural, and computational data show that widening the main cleft by opening either clamp domain or lobe domain is the primary means to clear the obstacles for protomer loading and unwinding 7,[14][15][16][17] . Not surprisingly, these obstacles are manipulated by macromolecular regulatory proteins, such as bacterial proteins DksA 13,18 , TraR 7 , and phage protein gp2 19 , and small-molecular compounds, such as myxopyronin 12,16,20 , lipiarmycin 21,22 , and ppGpp 11 , to accelerate or hinder the process of RPo formation through lowering or elevating energetic barriers imposed by these obstacles.
Compared with the above-mentioned transcription regulatory factors, SutA is unique in its large portion of disordered regions, highly structural flexibility, and RNAP-σ S holoenzyme preference. To understand the molecular mechanism of SutA activation, here we determined cryo-EM structures of RNAP-σ S holoenzyme at 4.1 Å, SutA-bound RNAP-σ S holoenzyme at 3.1 Å (SutA-RNAPσ S , open lobe) and 3.9 Å (SutA-RNAP-σ S , closed lobe), and SutAbound RNAP-σ S promoter open complex (SutA-σ S -RPo) at 5.8 Å. The structures reveal that SutA pinches RNAP-β protrusion with its RNAP-binding domain and wedges RNAP-β lobe open to facilitate promoter unwinding with its wedge loop and N-terminal D/E-rich region.

Results
SutA-RBD interacts with RNAP-β protrusion. Previous study has indicated that RNAP-β protrusion is likely the anchor site of SutA 2 . We first used a yeast two-hybrid (YTH) assay to map SutA domains responsible for interaction with RNAP-β protrusion. The results confirmed the interaction between the full-length SutA and Pae RNAP-β protrusion previously identified by crosslinking experiments 2 , and further showed the middle region (residues 57-90) of SutA comprising one precited α-helix and one β-strand is mainly responsible for the interaction (Fig. 1a, b). Therefore, we designated the region as the RNAP-binding domain of SutA (SutA-RBD). Although SutA-NTD does not interact with RNAP-β protrusion, it is essential for the transcription activation activity of SutA on RNAP-σ S holoenzyme (Fig. 1c, left panel and Supplementary Fig. 1e) 2 . In contrast, the short C-terminal tail of SutA contributes less to the transcription activation activity (Fig. 1c, left panel).
The cryo-EM structure of RNAP-σ S holoenzyme. As no structural information is available for bacterial RNAP-σ S holoenzyme, we reconstituted Pae RNAP-σ S holoenzyme and determined its structure through the single-particle cryo-EM approach (Supplementary Fig. 1a, b). Focused 3D classification resulted in a single group of RNAP-σ S holoenzyme particles, from which a cryo-EM map of the RNAP-σ S holoenzyme was reconstituted at resolution 4.1 Å (Supplementary Table 2 and Supplementary Fig. 2). The cryo-EM map shows strong continuous signals for all domains of σ S except for σ S 1.1 ( Fig. 2a and Supplementary Fig. 3a). The weak fractionated map signal of σ S 1.1 domain in the main cleft of RNAP suggests its conformational heterogeneity and low occupancy. We compared the binding affinity of Pae RNAP core enzyme towards σ A 1.1 or σ S 1.1 by using a fluorescence polarization assay. The results showed that the σ A 1.1 domain binds RNAP core enzyme with a Kd value of 220 nM, but the interaction between σ S 1.1 domain and RNAP core enzyme is barely detectable under the same experimental condition (Fig. 2c), indicating that σ S 1.1 binds to RNAP core enzyme much weaker compared with σ A 1.1 .
The cryo-EM structure of SutA-RNAP-σ S . To explore how SutA interacts with RNAP, we reconstituted Pae SutA-bound RNAP-σ S complex and determined its cryo-EM structure ( Supplementary  Fig. 1c, d). Focused 3D classification reveals one major population of SutA-bound RNAP-σ S single particles, from which a cryo-EM map was reconstructed at a resolution of 3.1 Å, and one minor population of SutA-bound RNAP-σ S single particles, from which a cryo-EM map was reconstructed at a resolution of 3.9 Å (Supplementary Table 2 and Supplementary Fig. 4). Both maps show unambiguous signals for SutA near the RNAP-β protrusion and for all domains of σ S except for σ S 1.1 . The two maps exhibit different binding modes of SutA and distinct conformations of RNAP-β lobe, a structure module comprising RNAP-β2 and -βSI1 (also named as βi4), and accordingly the two structures were named SutA-RNAP-σ S (open lobe) and SutA-RNAP-σ S (closed lobe) (Fig. 3a-d and Supplementary Fig. 7a, b).
In the structure of SutA-RNAP-σ S (open lobe), the RNAP-β protrusion is embraced by a patch of crescent-shaped map that could be assigned to SutA-RBD ( Fig. 3a-b and Supplementary  Fig. 3b), consistent with our YTH results and cross-linking data in a previous study 2 . The 3.1 Å cryo-EM map exhibits clear feature of an α-helix in the middle of the crescent-shaped map that fits well to the 20-residue predicted helical region (residues 57-76; Fig. 3a, f) 2 . The flanking residues of the middle helix were modeled into both ends of the crescent-shaped map (Supplementary Fig. 3b). One end of the crescent (residues 77-90) travels across the surface of the RNAP-β protrusion and approaches σ 3.1 , four residues (81-84) of which likely form a 4-strand β sheet with RNAP-β protrusion (Fig. 3a, e). Intriguingly. The other end of the crescent (residues 54-57; wedge loop) fits into the narrow gap between RNAP-β protrusion and β lobe and leads the rest part of SutA-NTD to the main cleft ( Fig. 3a and Supplementary Fig. 3b), where it likely adopts an assemble of multiple conformations and thereby does not show distinct map signals (Fig. 3a, g).
SutA-RBD occupies a hydrophobic groove on RNAP-β protrusion. Several hydrophobic residues of SutA (L67, M71, F74, L75, V81, and I84) likely participate in the interaction (Fig. 4a). The polar residues of SutA (K60 and R64) and RNAP-β protrusion (S59 and E73) are in close proximity and likely form polar interaction with each other (Fig. 4b). Protein sequence alignment of SutA of non-redundant bacteria species reveals that the proposed interface residues are conserved (Fig. 4c). To validate the interface observed in the structure, we tested the interaction of wild-type or mutant SutA and RNAP-β protrusion using the YTH approach. The results showed that alanine substitution of the hydrophobic residues (L67, M71, F74, L75, V81, and I84) or polar residues (K60 and R64) of SutA at the interface substantially impaired SutA/β protrusion interactions (Fig. 4d). Moreover, bulky substitution of the two conserved glycine residues (G78 and G79) of SutA at the interface also substantially impaired the interaction (Fig. 4d), confirmed the structure model.
The cryo-EM structure of Pae RNAP-σ S holoenzyme. a The cryo-EM map for Pae RNAP-σ S holoenzyme. RNAP subunits are colored as the color scheme. The σ S 1.1 map is circled. b The structural model for Pae RNAP-σ S holoenzyme (σ S 1.1 is not modeled due to poor cryo-EM map signal around the region). c The results of fluorescence anisotropy assay showed the binding of σ A 1.1 or σ S RNAP-σ S holoenzyme complexes shows differences in RNAP-β lobe conformation and main cleft width. In the SutA-RNAP-σ S (open lobe) structure, the RNAP-β lobe domain swings~10°away from RNAP-β' clamp resulting in~5 Å increase in distance between RNAP-β' clamp and RNAP-β lobe compared with that of RNAP-σ S holoenzyme (Fig. 5a). The gate loop (RNAP-β residues 373-383; corresponding to RNAP-β residues 368-378 in E. coli), which serves as a rate-limiting gate for promoter loading in E. coli and M. tuberculosis RNAP, is opened along with the movement of β lobe (Fig. 5a). The main cleft of SutA-RNAP-σ S (open lobe) can accommodate the downstream dsDNA as shown by structural modelling (Fig. 5b). In sharp contrast, there exists substantial steric clashes between the modeled dsDNA and RNAP in the RNAP-σ S holoenzyme (Fig. 5c). The results suggested that SutA opens RNAP-β lobe and widens the main cleft to allow loading of downstream dsDNA into the main cleft.
As stated in the above section, we also trapped a closed conformation of RNAP-β lobe in SutA-RNAP-σ S (closed lobe). In the structure, the β lobe adopts a slightly more closed conformation than that of RNAP-σ S holoenzyme (Supplementary   Supplementary Fig. 5c, d), suggesting that the invasion of SutA wedge loop induces opening of the β lobe. We further infer that the N-terminal D/E-rich SutA-NTD guides the wedge loop into the wedge position. Electrostatic surface presentation of RNAP shows that the main cleft of RNAP is highly positively charged (Fig. 3g), complementary to the natively charged SutA-NTD (23 D/E residues out of 55 residues in Pae SutA) (Fig. 1a). Therefore, SutA-NTD is likely captured in the main cleft and guides the wedge loop across the gap between RNAP-β protrusion and β lobe. Due to intrinsic flexibility of β lobe, the wedge loop slips in the bottom of the protrusion/lobe gap when the β lobe oscillates to its open conformation.
SutA promotes RPo formation. Because RNAP main cleft has to open to load the double-stranded promoter DNA during RPo formation 17,23-25 , opening RNAP main cleft by SutA thereby is expected to facilitate promoter loading, and thus shift the equilibrium towards RPo formation. To test the hypothesis, we measured the kinetics of RPo formation by a stopped-flow fluorescence assay, in which the Cy3 fluorophore (attached to the +2 position of the non-template strand of promoter DNA) servers as a probe sensing local environment change and increases its fluorescence upon promoter unwinding 17,[23][24][25] . The results showed that Cy3 fluorescence slowly reached a plateau, when RNAP-σ S holoenzyme alone was mixed with promoter DNA, while SutA substantially increased the kinetics of RPo equilibration (Fig. 5f). σ A 1.1 hinders the activation activity of SutA on RNAP-σ A holoenzyme. SutA was initially identified as a transcription activator of RNAP-σ S holoenzyme that functions at slow-growth condition 1 . Later study suggested that SutA is also capable of activating RNAP-σ A holoenzyme (Fig. 1c), albeit with lower activity 2 . Previous reports showed that σ Α 1.1 occupies RNAP main cleft, makes substantial interactions with RNAP-β lobe, and restrains it in the closed conformation 11,26 . To study whether σ A 1.1 hampers opening of RNAP-β lobe and thereby reduces the activity of SutA, we removed the σ A 1.1 region and tested whether SutA activates transcription of RNAP-σ A (Δ1.1) holoenzyme. The in vitro transcription results showed that removal of domain σ A inhibits RPo formation (Fig. 5g) 13,18 . Moreover, SutA increases the transcription activity of RNAP-σ A (Δ1.1) holoenzyme by~50%, close to the extent by which SutA activates RNAP-σ S holoenzyme (Fig. 5g). We propose deletion of σ A 1.1 releases its restrain on RNAP-β lobe and the regained conformational flexibility allows SutA wedge loop to slip into the β protrusion/lobe gap and to lock the β lobe in an open conformation.
The cryo-EM structure of P. aeruginosa SutA-σ S -RPo. To better understand the action of SutA upon RPo formation, we incubated SutA-RNAP-σ S with a duplex rrn promoter DNA (−49 to +21) for 20 s and immediately applied the reaction mixture on grids for cryo-EM data collection (Fig. 6a). Such strategy has been employed in previous studies to obtain structures of intermediate states during RPo formation 7,12,13 . However, we only obtained one major 3D class of single particles, from which a cryo-EM map at 5.8 Å resolution of SutA-bound RPo (SutA-σ S -RPo) was reconstructed (Supplementary Table 2 and Supplementary Fig. 6). The map shows clear density for upstream promoter dsDNA (−35 to −12), 14 bp downstream dsDNA (+4 to +17), and the −10 element (−11 to −10) of non-template ssDNA of the transcription bubble (Fig. 6b, c and Supplementary Fig. 7d). The cryo-EM map also shows clear signals for SutA-RBD that interacts with RNAP-β protrusion as in the structure of SutA-RNAP-σ S (open lobe) (Fig. 6d), but no signal for SutA wedge loop, SutA-NTD, and SutA-CTD (Fig. 6b, c, and Supplementary Fig. 3b), indicating that SutA-RBD remains bound to RNAP upon RPo formation. Structure superimposition of SutA-σ S -RPo and SutA-RNAP-σ S structures suggests that formation of RPo closes both the β' clamp and β lobe of RNAP (Fig. 6e).

Discussion
In this work, we have determined cryo-EM structures of Pae RNAP-σ S holoenzyme, Pae SutA-bound RNAP-σ S holoenzyme, and Pae SutA-σ S -RPo. The structures show that SutA pinches RNAP-β protrusion, wedges β lobe, widens the main cleft, and facilitates promoter loading and unwinding. Our study supports a role of β lobe movement on promoter unwinding 16 . The main cleft is located between two pincers, RNAP-β' clamp and RNAP-β protrusion/lobe. It was long appreciated that RNAP-β protrusion/ lobe is a rigid unit of RNAP, and that RNAP-β' clamp rotates to open or close the main cleft. Such assumption was made based on the fact that most reported bacterial RNAP structures show essentially the same closed conformation of β lobe/protrusion but different conformations of RNAP-β' clamp [27][28][29][30][31][32][33]   All evidence presented here lead to a model for structural mechanism of transcription activation by SutA (Fig. 7). The rrn promoter that contains a G/C-rich discriminator likely follows the "load-melt" pathway during promoter unwinding. SutA wedge loop opens the main cleft and facilitate the rrn promoter loading and unwinding. Loading of promoter DNA likely displaces SutA-NTD in the main cleft, disengages the wedge loop, and resumes the flexibility of lobe domain necessary for subsequent promoter unwinding steps.
Intriguingly, recent cryo-EM structures of TraR/DksA-bound transcription initiation complexes also reported the open state of β lobe (Supplementary Fig. 5e). In contrast, in those complexes, the β lobe is forced open by direct interaction between TraR/ DksA and RNAP-β SI1, the non-conserved insertion extending from RNAP-β lobe 7,13 . Together with these recent reports, our study pinpoints the importance of RNAP-β lobe for promoter unwinding and for regulation by diverse transcription regulators.
In summary, we report the structural mechanism for transcription activation of Pae SutA and propose a new means of transcription regulation. Our study presents structural evidence of intrinsic dynamic nature of RNAP-β lobe and an example that such the feature could be manipulated to affects kinetics of RPo formation by a transcription regulatory protein.

Methods
Plasmid construction. Please see Supplementary Table 1 for the list of plasmids in this study.
SutA was prepared from E. coli BL21(DE3) cells carrying pET-28a-TEV-SutA. The protein expression was induced with 0.4 mM IPTG at 18°C for 20 h at OD 600 of 0.7. The cell pellet was lysed in lysis buffer B (50 mM Tris-HCl, pH 7.7, 500 mM NaCl, 5% (v/v) glycerol, 5 mM β-mercaptoethanol, 0.1 mM PMSF) using an Avestin EmulsiFlex-C3 cell disrupter. The supernatant was loaded on a 5 mL Ni-NTA column that was subsequently washed and eluted with lysis buffer B containing 300 mM imidazole. The eluted fractions were subjected to TEV protease, buffer-exchanged, and reloaded onto a Ni-NTA column to remove impurity. The fractions containing target proteins were concentrated to 4 mg/mL, and stored at −80°C. Pae σ S , σ A and their derivatives were prepared by the same procedure.    Pae RNAP-σ S -holoenzyme. Pae RNAP core enzyme and σ S were incubated in a 1:3 molar ratio for overnight at 4°C and loaded onto a Superdex 200 Increase 10/ 300 GL column (Cytiva) equilibrated in 10 mM HEPES, pH 7.5, 100 mM KCl, 5 mM MgCl 2 , 2 mM DTT. Fractions containing Pae RNAP-σ S -holoenzyme were collected and concentrated to 15 mg/mL.

T TTCGA AA T A A CGCT TGA CGGA ACC C A A CGA T GCT GT AG A A TGCGCGGCCTCTCGGT TGAGACG A AA GC A A AAGCT T T T A TT GCGAACTGCCTT GGTGT GCT A CG T ACA CT T A CGCGCCGGAGAGCC A A CTCTGCTTT G C
Pae SutA-RNAP-σ S . Pae RNAP core enzyme, σ S and SutA were incubated in a 1:3:3 molar ratio for overnight at 4°C and loaded onto a Superdex 200 Increase 10/ 300 GL column (Cytiva) equilibrated in 10 mM HEPES, pH 7.5, 100 mM KCl, 5 mM MgCl 2 , 2 mM DTT. Fractions containing Pae SutA-RNAP-σ S were collected and concentrated to 12 mg/mL.
Cryo-EM structure determination of Pae RNAP-σ S -holoenzyme. The Pae RNAP-σ S holoenzyme complex was freshly prepared as described above and mixed with CHAPSO (Hampton Research, Inc.; final concentration 8 mM) prior to grid preparation. About 3 μL sample was applied onto the glow-discharged C-flat CF-1.2/1.3 400 mesh holey carbon grids (Electron Microscopy Sciences) in the chamber of a Vitrobot Mark IV (FEI; 95% chamber humidity at 10°C). The grids were subsequently plunge-frozen in liquid ethane. The micrographs were collected using Serial EM on a 300 keV Titan Krios (FEI) equipped with a Gatan K2 Summit direct electron detector (pixel size 1.307 Å/ pixel). A total of 1080 images were recorded using the counting mode (exposure, 10 s per 40-frame movie; dose rate, 10.0 electrons/pixel/s; defocus, −1.5 to −2.5 μm). Frames of individual movies were aligned using MotionCor2 34 and CTF estimations were performed using CTFFIND4 35 . 2D classes based on~1000 particles were served as templates for particle picking. The resulting 358,323 particles were subjected to 2D classification in RELION 3.0 with a E. coli RNAP holoenzyme structure (50 Å low-pass-filtered) as the starting reference model for 3D classification (N = 4). One 3D class with distinct shape of RNAP containing 165,721 particles were subjected to 3D classification (N = 3) again, the resulting one class contains 123,317 particles was selected and subjected to 2D classification for generating templates for the second round of particle auto-picking. A total of 489,361 particles were auto-picked by using the 2D references. The resulting particles were subjected to 2D classification in RELION 3.0 by specifying 100 classes. A total of 428,544 particles were selected and subjected to 3D classification using a 40 Ålow-pass-filtered cryo-EM structure of the previous 3D classification. One 3D class with distinct shape of RNAP containing 145,929 particles were subjected to 3D classification (N = 3) again, two classes were combined and used for 3D auto-refinement, CTF-refinement and Bayesian polishing, resulting a 4.69 Å map. To resolve RNAP heterogeneity around β lobe, a soft mask that excluded the RNAP-β protrusion and RNAP-β lobe regions was generated in Chimera and RELION 3.0. The mask was used to make a subtracted particle stack in RELION 3.0. The subtracted particles were applied for masked 3D classification (N = 3, without alignment) resulting in one major 3D class. The 105,629 particles in the 3D class were reverted and processed through iteratively 3D auto-refinement, CTFrefinement, Bayesian polishing and post-processing in RELION 3.0. Gold-standard Fourier-shell-correlation analysis (FSC) indicated nominal resolutions of 4.05 Å for the final mode of RNAP-σ S holoenzyme. E. coli RNAP-σ A holoenzyme model extracted from E. coli RNAP-σ A RPo (PDB: 4YLN) 36 was manually fit into the cryo-EM map using Chimera and refined using Coot and phenix.
Cryo-EM structure determination of Pae SutA-RNAP-σ S . The SutA-RNAP-σ S complex was freshly prepared as described above and mixed with CHAPSO (Hampton Research, Inc.; final concentration 8 mM) prior to grid preparation. About 3 μL mixture was applied onto the glow-discharged Quantifoil R1.2/1.3 300 mesh holey carbon grids (Quantifoil) in the chamber of a Vitrobot Mark IV (FEI; 100% chamber humidity at 22°C), and the grids were subsequently plungefrozen in liquid ethane.
The sample was prepared essentially as above. The micrographs were collected using EPU in the super-resolution counting mode on a 300 keV Titan Krios (FEI) equipped with a Gatan K3 Summit direct electron detector (pixel size 1.10 Å/pixel). A total of 2,805 images were recorded using the counting mode (exposure, 2.67 s per 40-frame movie; dose rate, 22.5 electrons/pixel/s; defocus, −1.2 to −2.2 μm).
The data were processed essentially as above. To resolve RNAP heterogeneity around SutA, the same soft mask that excluded the RNAP-β protrusion and RNAP-β lobe regions was generated in Chimera and RELION 3.0. The mask was used to make a subtracted particle stack in RELION 3.0. The subtracted particles were applied for masked 3D classification (N = 2, without alignment) resulting in two 3D classes with distinct conformations of RNAP-β lobe. The single particles of two 3D classes were separately reverted and processed through 3D auto-refinement and post-processing in RELION  The 138,109 particles were subjected to non-uniform refinement, resulting in a 3.86 Å map. The soft mask that excluded the RNAP-β protrusion /lobe-SI1 regions was generated in Chimera. The mask was used to make a subtracted particle stack and then the subtracted particles were applied for local refinement and sharpening in cryoSPARC v2.15. Gold-standard Fourier-shell-correlation analysis (FSC) indicated nominal resolutions of 4.77 Å. The structure model of Pae RNAP-β protrusion/lobe was refined using phenix. The Pae RNAP-σ S was manually fit into the cryo-EM maps using Chimera and refined using Coot and phenix. The middle helix (residues 57-76) of SutA was first fit into the crescent-shaped density map. The residue register of the helix was determined by density map of bulky residue sidechains and the resulting model agrees to previous BPA cross-linking data 2 . The structural model was subsequently refined in Phenix.
Cryo-EM structure determination of Pae SutA-σ S -RPo. Pae SutA-σ S -RPo (25 μM) was mixed with rrn full duplex promoter DNA (38 μM) with a molar ratio of 1:1.5 for 20 s, and subsequently supplemented with CHAPSO (Hampton Research, Inc., final concentration 8 mM). About 3 μL mixture was applied onto the glow-discharged C-flat CF-1.2/1.3 400 mesh holey carbon grids (Electron Microscopy Sciences) in the chamber of a Vitrobot Mark IV (FEI; 95% chamber humidity at 10°C), and the grids were subsequently plunge-frozen in liquid ethane. The total estimated reaction time is~20 s.
The micrographs were collected using Serial EM on a 300 keV Titan Krios (FEI) equipped with a Gatan K2 Summit direct electron detector (pixel size 1.0 Å/pixel). A total of 2005 images were recorded using the counting mode (exposure, 7.6 s per 38-frame movie; dose rate, 8.0 electrons/pixel/s; defocus, −1.2 to −2.2 μm). The data were processed in a similar procedure as described above except auto-picking particles were subjected multiple rounds of 2D classification to remove junk particles. The 3D class containing 112,959 particles were subjected to 3D autorefinement, CTF-refinement, Bayesian polishing, and post-processing steps in RELION 3.0. Gold-standard Fourier-shell-correlation analysis (FSC) indicated a nominal resolution of 5.77 Å. Pae SutA-RNAP-σ S holoenzyme and promoter DNA from a E. coli RPo (PDB: 7KHB) 13 were manually fit into the cryo-EM map using Chimera. Rigid body and real-space refinement were performed in Coot and Phenix.
Yeast two-hybrid assay. The pGADT7-SutA (or pGADT7-SutA derivatives) and pGBKT7-β protrusion (or pGBKT7-β protrusion derivatives) were transformed into the yeast strain AH109, respectively. The yeast AH109 cells bearing paired plasmids were mated and spotted on agar plates of a stringent selective medium without adenine, histidine, tryptophan, and leucine, or a selective medium without leucine and tryptophan. The growth of yeast colonies were recorded after 5 days. (A2C) (2 mM) and Fluorescein-5-Maleimide (20 mM Thermo Scientific, Inc.) in PBS, was incubated for 2 h at room temperature. The reaction was terminated by addition of DTT (1 mM; final concentration) and the labeled protein was purified by a 5 mL PD-10 desalting column (Biorad, Inc.). The fractions containing labeled protein were pooled and concentrated to 0.5 mg/mL. The Pae σ A 1.1 (S2C) was labelled in a similar procedure as described above. Stopped-flow assay. To monitor promoter unwinding by Pae RNAP-σ S holoenzyme, 60 μL Pae RNAP-σ S holoenzyme (final concentration: 40 nM) or preassembled Pae SutA-RNAP-σ S holoenzyme (final concentration: 40 nM) and 60 μL Cy3-λP R promoter DNA (final concentration: 1 nM; prepared as in 37 ) in buffer (10 mM Tris-HCl, pH 7.5, 200 mM NaCl, 10 mM MgCl 2 , 1 mM DTT) were rapidly mixed and the change of Cy3 fluorescence was monitored in real time by a stopped-flow instrument (SX20, Applied Photophysics Ltd, UK) equipped with a excitation filter (515/9.3 nm) and a long-pass emission filter (570 nm). The data were collected using Pro-Data SX software and were plotted in Prism (GraphPad, Inc.) and the observed rates k obs,1 and k obs,2 were estimated by fitting the data with following equation, F ¼ F 0 þ að1 À e Àk1t Þ þ bð1 À e Àk2t Þ Where F is the fluorescence value at a giving time point, t is the reaction time, F 0 is the fluorescence value at point of t = 0, k 1 and k 2 are the observed rates k obs,1 and k obs,2 , a and b are unrestrained constants.
Measurement of RNAP-β lobe movement. The rotation angle of β lobe and the main-cleft width change of the two states (open and closed β lobe) was measured as in 16 . Briefly, the rotation angle was defined as the angle of Cα atoms of three residues (βQ364 of open lobe, βN455 in closed lobe, and βQ364 in closed lobe; the position of βN455 is unchanged in both structure). The width change of the main cleft in the two states (open and closed β lobe) was defined as change of the distance between Cα atoms of βK383 and σL67.

Data availability
The data that support this study are available from the corresponding authors upon reasonable request.