Determination of RNA polymerase binding surfaces of transcription factors by NMR spectroscopy

In bacteria, RNA polymerase (RNAP), the central enzyme of transcription, is regulated by N-utilization substance (Nus) transcription factors. Several of these factors interact directly, and only transiently, with RNAP to modulate its function. As details of these interactions are largely unknown, we probed the RNAP binding surfaces of Escherichia coli (E. coli) Nus factors by nuclear magnetic resonance (NMR) spectroscopy. Perdeuterated factors with [1H,13C]-labeled methyl groups of Val, Leu, and Ile residues were titrated with protonated RNAP. After verification of this approach with the N-terminal domain (NTD) of NusG and RNAP we determined the RNAP binding site of NusE. It overlaps with the NusE interaction surface for the NusG C-terminal domain, indicating that RNAP and NusG compete for NusE and suggesting possible roles for the NusE:RNAP interaction, e.g. in antitermination and direct transcription:translation coupling. We solved the solution structure of NusA-NTD by NMR spectroscopy, identified its RNAP binding site with the same approach we used for NusG-NTD, and here present a detailed model of the NusA-NTD:RNAP:RNA complex.


Results and Discussion
RNAP interface of NusG-NTD. To identify the RNAP binding surface of transcription factors the methyl groups of Ile (δ 1), Leu (δ 1 or δ 2), and Val (γ 1 or γ 2) residues of the respective, deuterated factor were labeled with [ 1 H, 13 C] ([I,L,V]-labeled transcription factor; for clarity, all protein names without prefix refer to E. coli proteins). The titration of this [I,L,V]-labeled regulator with protonated RNAP was observed by two-dimensional (2D) [ 1 H, 13 C]-methyl transverse relaxation optimized spectroscopy (TROSY). As a test case for the applicability of this method, we asked whether we were able to confirm the RNAP binding surface of NusG-NTD. This surface is known from a crystallographic study of the archaeal Spt4/5 complex with the β ′ clamp domain of RNAP and biochemical experiments on NusG and RfaH, the latter being a paralog of NusG 12,13 .
Upon addition of RNAP, the methyl group signals of [I,L,V]-NusG-NTD decreased in intensity, but not uniformly over all signals (Fig. 2a), likely caused by a combination of several effects. First, a general loss of signal intensity is owed to [  Dark red and light red lines indicate the thresholds for strongly affected (55% of the average relative intensity) and slightly affected (75% of the average relative intensity) methyl groups, respectively. (c) Mapping of affected methyl groups onto the NusG-NTD structure (Protein Data Bank (PDB) ID: 2K06, cartoon representation, grey). Ile, Leu, and Val residues are in stick representation with the carbon atoms of their methyl groups as spheres. Strongly affected methyl groups, dark red; slightly affected methyl groups, light red; unaffected methyl groups, grey; unassigned methyl groups, black. Secondary structure elements and termini are labeled. (d) Mapping of affected residues onto the NusG-NTD structure (surface representation). For graphical illustration of the interaction site the complete amino acid was colored as affected in lieu of the methyl group. Colors are as in (c). Two amino acids on either side of affected Ile/Leu/Val residues are highlighted in yellow unless they were unaffected Ile/ Leu/Val residues. (e) Model of NusG-NTD as in (d) bound to E. coli RNAP (PDB ID: 4KMU). The model is based on the structure of the Pyrococcus furiosus (P. furiosus) Spt4/5 complex bound to the RNAP clamp domain (PDB ID: 3QQC). NusG-NTD was superposed on Spt5 and RNAP β ′ subunit on the clamp domain. As NusG-NTD and RNAP were treated as rigid bodies and no further optimization was carried out some minor clashes occur. β subunit, light blue; β ′ subunit, light green; β ′ CH, dark green; β GL, cyan. Finally, signal intensities can be influenced by chemical exchange processes in the intermediate range of the NMR timescale. Quantitative analysis of signal intensities for the 1:1 complex revealed two patches in the protein structure where signal intensities changed noticeably (Fig. 2b,c). Patch 1 comprises residues in helix α 3 and strands β 1 and β 3, while patch 2 is formed by residues located in helices α 1 and α 2, and these two patches are located at nearly opposite sides of NusG-NTD. No assigned, but unaffected methyl groups were found in either of these patches. This approach provides only information about Ile, Leu, and Val residues, but most likely additional amino acids, especially in the direct vicinity of the affected residues, are involved in the interaction. Thus we graphically extended the representation of patches 1 and 2 by including the two residues preceding and following each affected Ile, Leu, or Val residue, unless they were unaffected Ile, Leu, or Val residues, resulting in two continuous regions (Fig. 2d). In a model of NusG-NTD bound to RNAP based on the crystal structure of the archaeal Spt4/5: β ′ clamp domain complex 12 , residues of patch 1 are in direct proximity of the β ′ CH, indicating that we identified correctly the β ′ CH binding site (Fig. 2e). The NTD of RfaH, an E. coli paralog of NusG, not only interacts with the β ′ CH, but also binds to the β GL via His65, Thr66, and Thr67 which form an HTT motif located at the N-terminus of helix α 2 ( Supplementary Fig. 1) 13 . Although this interaction does not contribute significantly to the overall affinity of RfaH-NTD for RNAP it is essential for the antipausing activity of RfaH 13 . Similarly, structurally homologous residues in NusG-NTD (Ser79-His81) have been proposed to be involved in β GL binding, suggesting that this interaction is a general feature of NusG-like proteins 13 . NusG-NTD patch 2 corresponds to the RfaH region that is in immediate neighborhood of the β GL binding motif suggested for RfaH-NTD ( Supplementary Fig. 1) 13 . Due to the absence of Ile, Leu, and Val residues in the NusG-NTD region that is structurally homologous to the HTT motif in RfaH, no direct information about this region is available in our experiments ( Supplementary Fig. 1). Thus, we conclude that either the β GL binding surface in NusG-NTD differs slightly from the one in RfaH-NTD or that patch 2 constitutes only part of the β GL interaction surface or that residues of patch 2 are indirectly affected as they are located next to the actual binding site.
The clamp domain undergoes structural rearrangements during the transcription cycle, having closed and open conformations, and NusG-NTD/RfaH-NTD is proposed to lock the clamp in a closed state during elongation by making bridging contacts between the β ′ CH and the β GL so that the downstream DNA is completely encircled 13,[30][31][32][33] . Hence, the elongation complex is stabilized and structural rearrangements that occur during pausing are prevented, which, in turn, leads to increased processivity. As we used core RNAP in our experiments the clamp is probably in an open state. Thus our findings indicate that in the absence of nucleic acids NusG-NTD contacts the β ′ CH and β GL either separately or simultaneously, suggesting that the RNAP claw is in a conformation that allows these contacts or that NusG-NTD induces a closed state.
Overall, the binding surfaces identified here are consistent with the previously published interaction sites of NusG-NTD, demonstrating that the present approach may be used to determine the RNAP binding surfaces of transcription factors in solution in a single experiment using intact RNAP and avoiding molecular alteration of the constituents. However, the limited number of NMR probes and their distribution over the structure restricts the structural resolution of the resulting binding site. Although we are not able to distinguish between methyl groups that are directly involved in the molecular interaction from those that are only indirectly affected, the careful interpretation of the surface representation allows us to identify the interaction surface.
RNAP interface of NusE. Transcription factor NusE/S10 not only interacts with RNAP via NusG, but it is also able to bind directly and specifically to the RNAP β subunit during transcription 14,16,20 . The function of this interaction is still unknown. In order to study the molecular details of this interaction we determined the RNAP binding surface of NusE with the same approach as for NusG-NTD. As NusE alone is very unstable and tends to aggregate we used a NusE variant that lacks the ribosome binding loop (NusE Δ ) in complex with NusB for our experiments 34 . The presence of NusB does not influence the NusE Δ :RNAP interaction 20 . For the NMR titration, we labeled the methyl groups of Ile, Leu, and Val residues of NusE Δ in the deuterated NusB:NusE Δ complex with [ 1 H, 13 Upon addition of protonated RNAP, [I,L,V]-NusE Δ methyl group signals decreased in varying proportion (Fig. 3a,b). All highly and slightly affected methyl groups are located in helices α 1 and α 2 as well as strands β 1 and β 4 (Fig. 3c). Inspection of the surface representation and the graphical extension as carried out for NusG-NTD result in a continuous patch (Fig. 3d). As the 7 Ile, 10 Leu, and 7 Val residues of NusE Δ (86 residues overall) are distributed evenly over the sequence and the structure, our definition of the interaction surface is highly reliable. The RNAP binding site is opposite of the NusB:NusE Δ interface and the ribosome integration site, i.e. the NusE Δ :RNAP interaction is not only possible within the context of the NusB:NusE Δ complex, but also when NusE is integrated into the ribosome 35 . NusE could thus simultaneously accommodate the ribosome and the RNAP.
Interestingly, NusE Δ 's binding surface for RNAP strongly overlaps with that for NusG-CTD so that binding of NusE Δ to RNAP and NusG-CTD should be mutually exclusive (Fig. 3e) 14 . Thus we asked whether NusG-CTD and RNAP compete for binding to NusE. We performed a [ 1 H, 15 14 .
These competition experiments support the notion of overlapping binding sites of NusE for NusG-CTD and RNAP, and they show that NusG-CTD can interact with NusE in the presence of RNAP. The complexes NusE:RNAP and NusE:NusG:RNAP via NusG are thus in a delicate equilibrium that can easily be influenced by other regulators such as transcription factors or certain RNA sequences. Overall, formation Dashed black line, average relative signal intensity; dark red and light red lines, thresholds for strongly affected (60% of the average relative intensity) and slightly affected (80% of the average relative intensity) methyl groups, respectively. (c) Mapping of affected methyl groups onto the NusB:NusE Δ complex structure (PDB ID: 3D3B; NusB, purple; NusE Δ , light grey). NusB in surface, NusE Δ in cartoon representation. Ile, Leu, and Val residues in NusE Δ are represented as sticks with the carbon atoms of their methyl groups as spheres. Strongly affected methyl groups, dark red; slightly affected methyl groups, light red; unaffected methyl groups, grey; unassigned methyl groups, black. Secondary structure elements and termini are labeled. (d) Mapping of affected residues onto the NusB:NusE Δ complex structure (surface representation). Colors are as in (c). For graphical illustration of the interaction site the complete amino acid was colored as affected in lieu of the methyl group. Two amino acids on either side of an affected Ile/Leu/ Val residue are highlighted in yellow unless they were unaffected Ile/Leu/Val residues. (e) Structure of the NusB:NusE Δ :NusG-CTD complex. The NusE Δ :NusG-CTD complex (PDB ID: 2KVQ, NusG-CTD in blue cartoon representation) was superposed on the NusB:NusE Δ complex from (d).
Scientific RepoRts | 5:16428 | DOI: 10.1038/srep16428 of the NusE:RNAP complex might play various roles during transcription (Fig. 4d). It might be involved either in transcription:translation coupling as the ribosome could directly contact RNAP via S10, e.g. when the RNA tether is relatively short, or in transcription antitermination where NusB:NusE is part of the antitermination complex 14,16,19 . The amount of free NusE that is not bound to the ribosome is estimated to be very low, but it is essential for transcription antitermination 36 . Thus tethering of NusE or the NusB:NusE complex to RNAP might be an early event in transcription antitermination to increase the local NusE concentration. NusE would remain bound to the TEC until transferred to NusG-CTD during assembly of the antitermination complex. As ribosomal operons comprise a very high density of transcribing RNAPs with high elongation rates 37 , tethering NusE directly to RNAP would ensure fast and efficient transcription antitermination in these operons.

Solution structure of NusA-NTD from E. coli. The six domains comprising transcription factor
NusA associates with RNAP via NusA-NTD, which is necessary and sufficient for the enhancement of pausing during transcription 27 . To determine the solution structure of NusA-NTD by NMR spectroscopy we initially tried a construct containing amino acids Met1-Ile137 carrying an N-terminal His 9 -tag, NusA(1-137). The high degree of heterogeneity in the peak intensities as well as the spectral overlap in the [ 1 H, 15 N]-HSQC spectrum of the [ 15 N]-labeled protein, however, prevented further analysis ( Supplementary Fig. 3). A shorter construct, NusA-NTD Δ , consisting of amino acids Met1-Met125 and a cleavable C-terminal His 6 -tag, led to homogeneous signal intensities with non-overlapping signals in the [ 1 H, 15 N]-HSQC spectra ( Supplementary Fig. 3) and allowed nearly complete backbone and side chain resonance assignment. No resonances were found for residues Asp103, Arg104, Thr106, Thr107, and Gln108. These are located in a flexible loop so that severe line broadening may occur caused by either fast solvent exchange or conformational exchange on the intermediate chemical shift time scale. Structure determination was performed on the basis of 1565 distance and 193 dihedral restraints derived from multiple NMR experiments (Table 1).
To date structures of NusA proteins from different bacteria are available, and although all NusA-NTDs are similar in their overall architecture, they differ in the position of the linker helix ( Supplementary Fig.  5a-f). For NusA-NTD from B. subtilis (BsNusA-NTD), NMR data suggest that this helix occurs in two alternative conformations in solution 28 . However, we have no indication for the presence of multiple conformations of helix α 4 in NusA-NTD Δ . Moreover, unambiguous [ 15 N]-nuclear Overhauser enhancement spectroscopy (NOESY) cross peaks between hydrophobic amino acids could be observed in NMR experiments, demonstrating a direct interaction between helix α 4 and helices α 1 and α 2 in NusA-NTD Δ (Supplementary Fig. 5g)  to the linker helix by only a short loop, this helix might be responsible for the correct positioning of NusA-SKK for RNA binding. Comparing NusA-NTD structures it is striking that MtNusA-NTD and PlNusA-NTD lack the globular head ( Supplementary Fig. 5a-e), which is proposed to interact with the β ′ subunit of RNAP 40 . This might indicate a different mode of action/binding of MtNusA and PlNusA compared to other NusAs.

RNAP interface of NusA-NTD.
NusA-NTD is supposed to bind to RNAP by interacting with the β flap tip helix of the β flap region, which forms the outer wall of the RNA exit channel. To date, available complex models are based on a low-resolution electron microscopy structure, cleavage experiments, targeted amino acid exchanges and NMR experiments using a short β flap construct [26][27][28] . Here we used complete RNAP to determine the RNAP binding site of NusA-NTD Δ by applying the same approach as for NusG-NTD and NusE Δ . Methyl group labeled NusA-NTD Δ ([I,L,V]-NusA-NTD Δ ) was titrated with protonated RNAP leading to a non-uniform decrease of [I,L,V]-NusA-NTD Δ methyl group signals (Fig. 6a). Again, the normalized signal intensity decrease in the 1:1 complex was analyzed to identify highly and slightly affected methyl groups (Fig. 6b). These are located mainly on the concave side of the body and in the acidic head (Fig. 6c). Inspection of the surface representation suggests that the β -sheet on the concave side of NusA-NTD Δ is the center of the interaction surface, although it contains only a Dashed black line, average relative signal intensity; dark red and light red lines, thresholds for strongly affected (65% of the average relative intensity) and slightly affected (85% of the average relative intensity) residues, respectively. (c) Mapping of affected methyl groups onto the NusA-NTD Δ structure. NusA-NTD Δ (grey) in cartoon representation. Ile, Leu, and Val residues are in stick representation with the carbon atoms of their methyl groups as spheres. Strongly affected methyl groups, dark red; slightly affected methyl groups, light red; unaffected methyl groups, grey; unassigned methyl groups, black. (d) Mapping of affected residues onto the NusA-NTD Δ structure (surface representation). For graphical illustration of the interaction site the complete amino acid was colored as affected in lieu of the methyl group. Colors are as in (c). Two amino acids on either side of an affected Ile/Leu/Val residue are highlighted in yellow unless they were unaffected Ile/Leu/Val residues. The positions of Ser29 and Ser53 are marked by black arrows. limited number of Ile, Leu, or Val residues resulting in a low structural resolution (Fig. 6d). Our binding site is in accordance with cleavage experiments using NusA variants NusA(S29C) and NusA(S53C), that indicated that S29 is located in the NusA:RNAP interface, while S53 is at the opposite side of NusA-NTD (Fig. 6d) 27 . Moreover, our results generally agree with mutational analyses showing that the concave side of the β -sheet is involved in NusA-NTD:β flap interaction 28 . Model of the NusA:RNAP complex. NusA has various effects on transcription elongation and termination with the NusA-NTD:RNAP interaction being probably one key step within the regulatory mechanism 27 . NusA-NTD contacts the RNA exit channel by binding to the β flap tip helix of the β flap region, but the resolution of the electron microscopy structure of a NusA-NTD:RNAP complex was too low to unambiguously determine the orientation of NusA-NTD bound to RNAP 26 . Cleavage and crosslinking experiments on the one hand and mutational analyses as well as NMR studies on BsNusA-NTD and a short β flap construct on the other hand lead to two binding models 27,28 .
We used our NMR data to dock NusA-NTD Δ to the β flap tip helix of elongating Thermus thermophilus RNAP (TtRNAP, PDB ID: 2O5I) using HADDOCK 41 (Fig. 7a). In the model most reliable according to HADDOCK, the body of NusA-NTD Δ binds the β flap tip helix via its concave side, which is in accordance with other models 27,28 . The body is oriented towards the RNA exit channel so that the globular head interacts with the β ′ subunit, the latter being in agreement with previous findings that the β ′ subunit might also be involved in NusA-NTD binding 20,40 . This orientation allows a tight interaction with the TtRNAP and is similar to the orientations suggested in earlier models 27,28 , although the absolute position of NusA-NTD Δ strongly depends on the residues chosen as restraints and the position of the β flap tip helix.
Next, we integrated the NusA-SKK domain into the model (Fig. 7b). As the structure of E. coli NusA-SKK is not available and as the position of the linker helix is similar in PlNusA and NusA-NTD Δ , we first used the crystal structure of PlNusA as template. This, however, led to heavy steric clashes of the PlNusA-SKK domain and TtRNAP which could be prevented by rotating the PlNusA-SKK domain away from the TtRNAP, using the 3-4 residues following the linker helix as anchor. Alternatively, the linker helix itself might rotate slightly. Thus, we modeled the position of TmNusA-SKK by superposing TmNusA-NTD (PDB ID: 1L2F) on NusA-NTD Δ , and we added a short piece of RNA from the MtNusA-SKK:RNA complex structure (PDB ID: 2ASB, Fig. 7b). Either way, the NusA-SKK domain can be positioned correctly for RNA binding. As NusA-NTD is necessary and sufficient for enhancing transcriptional pausing and recognizes duplex RNA 27 , exiting RNA might first contact a basic patch on the helical bundle of the NusA-NTD body ( Supplementary Fig. 4), which is in direct vicinity of the RNAP exit channel. The RNA then wraps around the NusA-SKK domain, which, in turn, recognizes specific RNA signals (Fig. 7b) 4,42,43 . Crosslinking experiments showed that the RNA region − 16 to − 23 lies near the NusA-NTD in full-length NusA and that the − 34 to − 40 region of exiting RNA contacts the NusA-KH2 domain 27 , which is consistent with our model. Moreover, the NusA-S1 domain is placed in the vicinity of the β ′ dock domain, being in accordance with a genetically shown NusA-S1:β ′ dock interaction 44 and cleavage experiments using Fe(III)-(S)-2-[4-(2-bromoacetamido)benzyl]ethylenediaminetetraacetic acid (FeBABE) 27 . The position of the C-terminus of NusA-SKK roughly orientates the two NusA-AR domains towards the α -subunits of RNAP and thus localizes NusA-AR2 close to the α -CTD, sterically simplifying a NusA-AR2:α -CTD interaction 4 .
Finally, it has been speculated that reorientation of helix α 4 stabilizes RNA hairpins 28 . However, not only does NusA exhibit large conformational plasticity, but, in addition, the β flap tip helix is also a highly mobile element 28 . During the transcription cycle the flexibility of the β flap tip helix is important for the regulation of the size of the RNA exit channel, of which the β flap forms the outer wall. Thus, we suggest that the orientation of NusA-NTD bound to RNAP as well as the position of helix α 4 may vary, depending on the position of the β flap tip helix. Moreover, this structural flexibility is complemented by the other NusA domains, which are all elastically connected.
Outlook. In this conceptually simple single-experiment approach to identify the RNAP interaction surface of transcription factors with NMR spectroscopy (i) complete RNAP is used, (ii) probes in the transcription factor are directly monitored and, most importantly, (iii) none of the interaction partners needs to be modified. In the future, the method will be refined and used to study these interactions in more detail. Moreover, this approach is very general and can thus be transferred to other systems, with a small binding partner interacting with a supramolecular complex.

Materials and Methods
Cloning. The gene coding for EcNusA-NTD  was cloned into pET19b via BlpI and BamHI.
The resulting E. coli expression vector pET19b_NusA-NTD_1-137 codes for a His 9 tag fused to the N-terminus of NusA-NTD, cleavable by PreScission protease.
Proteins were uniformly labeled with 15 N or 15 N, 13 C by growing E. coli in M9 minimal medium 41,42 with addition of ( 15 NH 4 ) 2 SO 4 (Campro Scientific, Berlin, Germany) or ( 15 NH 4 ) 2 SO 4 and 13 C-D-glucose (Spectra Stable Isotopes, Columbia, MD, USA) as only nitrogen and carbon source. Expression and purification was the same as for proteins produced in LB medium. Methyl group labeling of Ile, Leu and Val residues with [ 1 H, 13 C] in deuterated proteins was performed as described previously 20 . NMR spectroscopy. NMR spectroscopic experiments were conducted on Bruker Avance 600 MHz, 700 MHz and 800 MHz spectrometers, the latter two equipped with cryogenically cooled probes. For resonance assignment of NusA-NTD Δ , standard double and triple resonance through-bond experiments were recorded 47,48 . The protein was in 10 mM potassium phosphate buffer (pH 6.4) containing 50 mM NaCl at 298 K. NMR data were processed using in-house routines (Apodization, Fourier transformation, phase correction and baseline correction) and visualized with NMRView 49 . Distance restraints for structure calculation were derived from [ 15 N]-edited and [ 13 C]-edited NOESY spectra with mixing times of 100-120 ms. NOESY cross peaks were classified according to their relative intensities and converted to distance restraints with the following upper limits: 3.0 Å, strong; 4.0 Å, medium; 5.0 Å, weak; 6.0 Å, very weak. Experimental NOESY spectra were validated semi-quantitatively against back-calculated spectra to confirm the assignment and to avoid bias of upper distance restraints by spin-diffusion. Hydrogen bonds were included for backbone amide protons in regular secondary structure if the amide proton did not show a water exchange cross peak in the [ 15 N]-edited NOESY spectrum. Backbone dihedral restraints were obtained from chemical shift data by using TALOS 50 . Existence of a hydrogen bond was assumed if the acceptor of a slowly exchanging amide proton, based on the absence of a water exchange peak in the [ 15 N]-edited NOESY spectrum, could be identified unambiguously from the results of initial structure calculations. For each hydrogen bond the distance between the amide proton and the acceptor was restrained to less than 2.3 Å and the distance between the amide nitrogen and the acceptor to less than 3.1 Å.
The structure calculation was performed with the program XPLOR-NIH 2.1.2 51 using a three-step simulated annealing protocol with floating assignment of prochiral groups including a conformational database potential 52 . For the final iteration 80 structures were calculated, the 20 structures of lowest energy were accepted and further analyzed with the programs XPLOR-NIH 2.1.2 and PROCHECK-NMR 53 .
TROSY spectra 29 were recorded using [I,L,V]-labeled protein samples (20 μ M) in 25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES, pH 7.5), 50 mM NaCl, 5% (v/v) glycerol, 0.5 mM ethylenediaminetetraacetic acid (EDTA), 10 mM MgCl 2 , 10 μ M ZnCl 2 , 1 mM DTT in 99.9% D 2 O at 298 K. Unlabeled, protonated RNAP in the same buffer was added in two steps (ratios 1:1, 1:2). Non-stereo-specific assignments of methyl groups of NusG-NTD and NusE Δ were taken from previous studies 10, 46 . Signal intensities were normalized by protein concentration and number of scans. As pulse lengths changed less than 1% upon RNAP addition, the influence of these changes on the intensity were neglected. For each titration step the ratio of remaining signal intensities and signal intensities in the spectrum of the free transcription factor were calculated, yielding relative signal intensities. Next, the mean value of all relative intensities in each titration step was determined and experiment-specific thresholds of the mean value were defined. Residues with relative signal intensities below these thresholds were classified as either strongly or slightly affected. Additionally, Leu and Val residues were considered as affected, when at least one of the two signals showed a significant intensity decrease. Only unambiguously assigned signals were used in the analysis.
Proteins  15 N]-HSQC spectra after each titration step. 1D spectra were normalized by protein concentration and number of scans. As pulse lengths changed less than 1% upon RNAP addition, the influence of these changes on the intensity were neglected. Docking of NusA-NTD Δ (model 1) to elongating TtRNAP (PDB ID: 2O5I) was carried out using the HADDOCK webserver 41 . Residues in NusA-NTD Δ that were experimentally determined to be affected by RNAP binding (Leu27, Leu31, Ile43, Val45) were defined as active residues. Solvent exposed residues in the β flap tip helix were chosen as active residues (chain C, residues Arg772, Leu773, Ser776, Ile777). Passive residues were automatically determined by HADDOCK. The coordinates of the β flap tip helix in the docked complex relative to the deposited coordinates of NusA-NTD Δ are shown in Supplementary  Table 1. After docking NusA-NTD Δ to TtRNAP, the position of the NusA-SKK domain was modeled with two alternative procedures. First, PlNusA (PDB ID: 4MTN) was superposed on NusA-NTD Δ (residues G3-D73 of PlNusA; residues Met1-Thr101 of NusA-NTD Δ ). To avoid clashes with TtRNAP the PlNusA-SKK was rotated manually around residues in the linker between PlNusA-NTD and PlNusA-SKK (residues Arg107-Gln109) using PyMOL 54 . In the second approach TmNusA (PDB ID: 1L2F) was superposed on NusA-NTD Δ using residues 1-101. Finally, the MtNusA-SKK:RNA complex (PDB ID: 2ASB, residues Ser108-Gly333 of MtNusA-SKK) was superposed on TmNusA-SKK (residues Glu132-Leu344) to position the RNA. RNA base numbers were estimated.