Structural basis for the function of SuhB as a transcription factor in ribosomal RNA synthesis

Abstract Ribosomal RNA synthesis in Escherichia coli involves a transcription complex, in which RNA polymerase is modified by a signal element on the transcript, Nus factors A, B, E and G, ribosomal protein S4 and inositol mono-phosphatase SuhB. This complex is resistant to ρ-dependent termination and facilitates ribosomal RNA folding, maturation and subunit assembly. The functional contributions of SuhB and their structural bases are presently unclear. We show that SuhB directly binds the RNA signal element and the C-terminal AR2 domain of NusA, and we delineate the atomic basis of the latter interaction by macromolecular crystallography. SuhB recruitment to a ribosomal RNA transcription complex depends on the RNA signal element but not on the NusA AR2 domain. SuhB in turn is required for stable integration of the NusB/E dimer into the complex. In vitro transcription assays revealed that SuhB is crucial for delaying or suppressing ρ-dependent termination, that SuhB also can reduce intrinsic termination, and that SuhB-AR2 contacts contribute to these effects. Together, our results reveal functions of SuhB during ribosomal RNA synthesis and delineate some of the underlying molecular interactions.


INTRODUCTION
Transcription in bacteria is terminated predominantly via two mechanisms (1). Intrinsic termination depends on a stable RNA hairpin followed by a sequence rich in uridines; the hairpin invades the RNA exit tunnel of RNA polymerase (RNAP), while the U-rich stretch forms a weak DNA:RNA hybrid, facilitating termination. In -dependent termination, the hexameric RNA-dependent NTPase, , engages the nascent transcript at C-rich sequences, so-calledutilization (rut) sites, uses its NTP-dependent RNA translocase activity to track down RNAP and, upon encounter, leads to termination. Both modes of termination can be supported or suppressed by transcription factors (2). For example, intrinsic termination can be enhanced by transcription factor N-utilization substance (Nus) A that binds RNAP and stabilizes RNA hairpins in the exit tunnel (3)(4)(5)(6). -dependent termination can be increased by NusG that also binds RNAP via its N-terminal domain (NTD) and contacts via its C-terminal domain (CTD), thereby facilitating clamp-down of on RNA at sub-optimal rut sites (7). Conversely, NusG can also counteract both modes of termination by enhancing RNAP processivity (8,9), while NusA can inhibit -dependent termination by competing for rut sites (10).
As transcription and translation in bacteria are not segregated into different cellular compartments, translation can initiate on mRNAs that are still being transcribed. Indeed, a ribosome trailing RNAP is important for the efficient expression of protein-coding genes, as it hinders from approaching RNAP and thus insulates RNAP from -dependent termination (11). Lack of this effect underlies the principle of translational polarity (12), in which inhibition of translation of an upstream gene in a multi-cistronic mRNA leads to down-regulation of the downstream genes due to premature -dependent transcription termination.
In bacteria, -dependent termination thus presents a potential obstacle for the efficient synthesis of long, noncoding RNAs, such as ribosomal (r) RNAs, which are not translated. Thus, bacteria might require mechanisms that can prevent such premature termination of rRNA synthesis. Indeed, Escherichia coli uses a specialized transcription complex to achieve efficient transcription of rRNA. Initially, Nus factors A, B, E and G (NusE is equivalent to r-protein S10) were recognized as factors participating in this process (13). In vitro reconstitution experiments suggested the presence of additional essential components (14), and other r-proteins, in particular S4, were subsequently identified as some of the missing subunits (15). More recently, the inositol mono-phosphatase, SuhB, has been shown to constitute another key player (16). Together, these molecules are thought to assemble a multi-factorial RNA-protein (RNP) complex on the surface of RNAP in response to RNA signal sequences encoded in the ribosomal DNA leader and spacer regions (13,15) (Figure 1A, C). This complex accompanies RNAP during further transcription elongation, forming an rRNA transcription antitermination complex (rrnTAC) that prevents -dependent termination (15) by mechanisms that are presently unclear. The process is thus referred to as processive rRNA (rrn) anti-termination.
Processive rrn anti-termination is reminiscent of processive anti-termination installed via N proteins of lambdoid phages, which is required for the switch from immediateearly to delayed-early gene expression during the lytic life cycle of the phages (17). N-dependent processive antitermination is also invoked in response to an RNA signal element, the N-utilization (nut) site, encoded in leader regions of the phage genomes, which bears a linear element, boxA, followed by a hairpin structure, boxB ( Figure 1B, D). Recent structural analyses by our group have unraveled the structural basis of N-dependent processive anti-termination (18,19). The N protein together with NusA binds the boxB element of or a consensus nut RNA, while the NusB/E dimer recognizes boxA. N strings the Nus factors together, repositioning them on RNAP and presenting a composite NusA-NusE surface that sequesters the CTD of NusG. It thereby prevents NusA-mediated stabilization of RNA hairpins in the exit tunnel and overrides -supporting functions of NusG. Moreover, a C-terminal region of N traverses the RNAP catalytic cavity, stabilizing the enzyme in a processive conformation and counteracting RNA hairpin invasion of the exit tunnel. N-mediated remodeling of RNAP elements that form part of the RNA exit tunnel and extended guidance of the exiting RNA by repositioned NusA further contribute to hairpin exclusion. N also cooperates with NusG to stabilize upstream DNA and prevent RNAP backtracking.
The leader and spacer regions in E. coli rRNA genes encode RNA signal elements that resemble /consensus nut sites, bearing boxB-like and boxA-like elements in the opposite order, followed by an additional linear boxC sequence ( Figure 1C). As in N-based anti-termination (20), the boxA element serves as a binding site for a hetero-dimer formed by the NusB and NusE subunits (21), and both boxA and NusB are required for counteracting in an in vitro system (14). Similar signal elements are conserved in the rRNA operons of other bacteria, such as Mycobacterium tuberculosis, where boxA and boxC have been shown to be sequestered in stem-loop structures, and binding of NusA to the boxC element is associated with opening of these structures (22). The boxC element is also part of a duplex formed with a complementary region in the spacer between the 16S and 23S rRNA portions, which constitutes a processing site for RNase III-mediated excision of pre-16S rRNA (22)(23)(24). It has therefore been suggested that NusA may support rRNA maturation by presenting the upstream portion of an RNase III cleavage site to the downstream portion (22).
Presently, SuhB is the least understood subunit of the rrnTAC. Based on chromatin immunoprecipitation (ChIP) analyses, it was suggested that SuhB is recruited to the rrnTAC in a boxA-and NusB-dependent manner (16). E. coli SuhB has also been shown to bind RNAP in the form of the holoenzyme (i.e. the ␣ 2 ␤␤' core enzyme in association with a initiation factor) (25) and Pseudomonas aeroguinosa SuhB has been found associated with RNAP in vivo (26). While E. coli SuhB possesses inositolmonophosphatase activity, this activity is not required to alleviate effects associated with a suhB mutant strain (27). On the other hand, SuhB variants that failed to bind RNAP holoenzyme failed to complement a suhB deletion (25), suggesting that SuhB's transcription-related roles could constitute its main functions in the cell. As SuhB is phylogenetically widely conserved, it is likely that these functions are also widespread in bacteria (16).
Here, we delineated molecular interactions, based on which SuhB participates in rRNA synthesis. We show that SuhB directly binds a C-terminal acidic repeat domain of NusA and contacts at least one other region in NusA as well as the nut-like RNA signal element. It thereby facilitates entry of the NusB/E dimer into an rrnTAC. Moreover, we determined crystal structures of SuhB alone and in complex with the main SuhB-binding domain of NusA. Transcription assays revealed that SuhB is the critical subunit that elicits delay or suppression of -dependent termination, that an rrnTAC comprising SuhB can also suppress intrinsic termination and that these activities depend in part on SuhB interacting with the NusA C-terminal acidic repeat. Based on our results, we suggest molecular mechanisms by which SuhB may support anti-termination. Our results also have implications for SuhB and Nus factors promoting rRNA maturation.

Plasmids, DNAs and RNAs
DNA fragments encoding SuhB and S4 were PCRamplified from E. coli (DH5␣) genomic DNA and cloned into the pETM-11 vector (European Molecular Biology Laboratory) via NcoI and HindIII restriction sites. A DNA fragment encoding NusA AR2 (residues 427-495) was PCRamplified from a pETM-NusA plasmid and cloned into pETM-11 via NcoI and HindIII restriction sites. A DNA template for in vitro transcription assays was generated by assembly PCR and cloned into pUC18 vector via XbaI and HindIII restriction sites. All constructs were verified by sequencing (Seqlab). DNAs used for the assembly of transcription complexes were purchased as single-stranded oligonucleotides (Eurofines). RNA constructs were synthesized by in vitro transcription by T7 P266L RNA polymerase (28), using PCR products as templates, and purified as described (19).

Protein production and purification
Full-length SuhB was produced in E. coli BL21(DE3)pLysS overnight at 37 • C in auto-induction medium (29). Cells were harvested by centrifugation and lysed in lysis buffer (50 mM Tris-HCl, pH 7.5, 500 mM NaCl, 1 mM 2mercaptoethanol). All subsequent steps were performed at 4 • C or on ice. Cleared lysate was incubated with Ni 2+ -NTA agarose beads (Macherey-Nagel), beads were washed with lysis buffer supplemented with 20 mM imidazole. Captured protein was eluted with elution buffer (50 mM Tris-HCl, . NusA AR2 was purified via the same protocol as full-length NusA (19). Plasmids encoding SuhB, NusA and NusA AR2 variants were generated by site-directed mutagenesis, and the proteins were produced and purified by the same protocols as used for the wild type (wt) proteins. Other proteins (RNAP, NusA, NusA AR2 , NusB/E, NusG, and 70) were produced and purified as described previously (18,19).

Analytical size exclusion chromatography
Interactions were tested by analytical size exclusion chromatography (SEC). Before loading on a Superdex 200 Increase 3.2/300 column (GE Healthcare), near-stoichiometric amounts of proteins and/or nucleic acids (20 M final concentration for the largest component and 25 M final concentrations for all smaller components) were mixed in running buffer (20 mM HEPES-NaOH, pH 7.5, 50 mM NaCl, 1 mM DTT) and incubated at room temperature for 15 min. SEC was conducted with a flow rate of 40 l/min and 50 l (40 l for runs with RNAP) fractions were collected. Fractions were analyzed on 15% SDS-PAGE gels (11)(12)(13)(14)(15)(16).5% gradient SDS-PAGE gels for runs with RNAP) and 15% 8 M urea-PAGE gels for proteins and nucleic acids, respectively.

Surface plasmon resonance assays
Surface plasmon resonance (SPR) experiments were carried out on Biacore T20 (GE Healthcare) using Sensor Chip NTA (GE Healthcare) at 20 • C. The Sensor Chip NTA was loaded with Ni 2+ by incubating in 0.5 mM NiCl 2 . Subsequently, 3 mM EDTA, pH 8.3, was flowed across the chip to remove unbound Ni 2+ . After equilibrating with SPR buffer (20 mM HEPES-NaOH, pH 7.5, 50 mM NaCl, 1 mM DTT), N-terminally His 6 -tagged NusA AR2 was immobilized on the chip, excess protein was washed away with 45 l of SPR buffer (30 l/min, 90 s). SPR experiments were carried out according to the single cycle-kinetic method, using SuhB variants in SPR buffer at 25, 50, 100, 150 and 200 nM. The same experiments were carried out in parallel in the control channel without prior Ni 2+ loading. Responses of the control channel were subtracted from the responses in the experimental channel and corrected data were analyzed with Biacore T20 software (GE Healthcare).

Size exclusion chromatography/multi-angle light scattering
Size exclusion chromatography/multi-angle light scattering analyses were performed on an HPLC system (Agilent) coupled to a mini DAWN TREOS multi-angle light scattering and RefractoMax 520 refractive index detectors (Wyatt Technology). 60 l (15 nmol) of SuhB were passed over a Superdex 200 increase 10/300 column (GE Healthcare) in 20 mM HEPES, pH 7.5, 50 mM NaCl, 0.02% (w/v) NaN 3 at a flowrate of 0.6 ml/min. Data were analyzed with the ASTRA 6.1 software (Wyatt Technology) using monomeric bovine serum albumin (Sigma-Aldrich) as a reference.

Transcription assays
For in vitro transcription assays, a DNA template containing a T7A1con promotor (T7A1 bearing a consensus -10 element) followed by the anti-termination region from the E. coli rrnG operon, rutA/rutB entry and trpt' termination regions from the trp operon (10) and the tR' intrinsic terminator (without the endogenous zone of opportunity) from the phage genome was designed. Assays were performed in single-round transcription format (30). 100 nM of E. coli RNAP core enzyme and 70 factor, 20 nM template DNA, 10 M ApU, 2 M ATP, GTP and CTP and 2 Ci ␣-[ 32 P]CTP were mixed in 10 l transcription buffer (20 mM Tris-OAc, pH 7.9, 100 mM KOAc, 5 mM Mg(OAc) 2 , 5% (v/v) glycerol, 1 mM DTT) and incubated at 32 • C for 10 min to generate initial transcription complexes with an 11-nucleotide labeled RNA. 200 nM NusA, 1 M NusG, 1 M NusB/E, 500 nM S4 and/or 500 nM SuhB, as well as 500 nM hexameric where indicated, were then added to the reaction and incubated for 5 min at 32 • C. Subsequently, a mixture of all four rNTPs was added to the reaction (final concentrations of 2 mM ATP and CTP, 100 M GTP and UTP). Samples were taken at defined time points, PCI-extracted, isopropanol-precipitated and analyzed via 6% 8 M urea-PAGE. Bands were visualized on a Storm PhosphorImager and quantified with Image-Quant software. Relative read-through of the trpt' region was determined as the percentage of trpt' read-through products relative to all products in a lane, with the value for acting on RNAP alone set to 0 and the values for acting on all other complexes scaled accordingly. Relative read-through of the tR' intrinsic terminator was determined as the percentage of tR' read-through products relative to all products in a lane with the value for RNAP alone set to 0 and the values for all other complexes scaled accordingly.
Diffraction data were collected on beamline 14.2 at the BESSY II storage ring (Berlin, Germany) at 100 K. All data were processed with XDS (31,32). Structures were solved by molecular replacement, using the structure coordinates of NusA AR2 (PDB ID 1WCN) and/or the SuhB R184A variant (PDB ID 2QFL). Structures were refined by alternating rounds of model building in COOT (33) and automated maximum-likelihood restrained refinement in PHENIX (34). Model quality was evaluated with MolProbity (35) and the JCSG validation server (JCSG Quality Control Check v3.1). Structure figures were prepared using PyMOL (36). Data collection and refinement statistics are provided in Table 1.
in which <I h > is the mean intensity of symmetry-equivalent reflections h, I ih is the intensity of a particular observation of h and N is the number of redundant observations of reflection h. c CC 1/2 = (<I 2 > -<I> 2 ) / (<I 2 > -<I> 2 ) + 2 ε , in which 2 ε is the mean error within a half-dataset (45).
. e R free is the same as R work , but calculated on the test set of reflections excluded from refinement. f A.U. -asymmetric unit. g Calculated with MolProbity (35). h Rmsd -root-mean-square deviation from target geometry.

Structure comparisons
Structures were compared by global superposition of complex structures or by superposition of selected subunits in complexes using the Secondary Structure Matching (SSM) algorithm implemented in COOT (33).

SuhB binds directly to NusA and the RNA signal element
Presently it is unknown which other components of the rrnTAC SuhB directly interacts with. To address this question, we conducted analytical size exclusion chromatogra-phy (SEC) analyses using recombinantly produced, purified components. R-protein S4 could not be included in these initial experiments, as in the absence of RNAP, DNA and RNA, it tended to aggregate and precipitate with the other factors. Under the chosen conditions, SuhB did not stably interact with RNAP alone (Figure 2A). Likewise, SuhB did not bind the NusB/E dimer or NusG ( Figure 2B). However, SuhB co-eluted both with NusA and with rrnGnut RNA (residues 1-66; Figure 1C) from the gel filtration column, and these components together eluted earlier than the individual molecules ( Figure  der the chosen conditions. SuhB, NusA and rrnGnut RNA also formed a ternary complex ( Figure 2D). Upon mixing SuhB with all Nus factors and rrnGnut RNA, a complex comprising SuhB, NusA, NusB, NusE and rrnGnut RNA eluted from the gel filtration column ( Figure 2E). A similar complex could not be assembled when phage or consensus nut RNAs (residues 1-36; Figure 1D) were used instead ( Figure 2F). In the latter case, separate ternary SuhB-NusA-nut RNA and NusB-NusEnut RNA complexes formed ( Figure 2F).
NusA is a multi-domain protein that encompasses an RNAP-binding NTD, an array of RNA-binding S1 and two hnRNP K homology domains (KH1 and KH2) and two C-terminal acidic repeat domains (AR1 and AR2; Figure 1E). The latter two domains are not universally conserved in NusA orthologs. To test which region of NusA is bound by SuhB, we recombinantly produced various fragments of NusA and repeated the SEC analyses. SuhB bound to a construct comprising the NusA AR2 domain (residues 427-495; NusA AR2 ; Figure 3A), but failed to interact with a NusA variant lacking this domain (residues 1-426; NusA AR2 ; Figure 3B). These results indicate that the AR2 domain of NusA represents the main contact site of the protein to SuhB.

RrnGnut RNA is required for stable integration of SuhB into transcription complexes
Upon mixing RNAP, DNA with an artificial transcription bubble, rrnGnut RNA that could pair in the transcription bubble, all Nus factors, S4 and SuhB, a stable rrnTAC could be reconstituted ( Figure 4A). S4 was not required for formation of an RNAP-DNA-rrnGnut RNA-NusA/B/E/G-SuhB complex ( Figure 4B). A complex containing all components also still formed when full-length NusA (NusA FL ) was replaced by a variant lacking the NusA AR2 domain (NusA AR2 ; Figure 4C), indicating that the AR2 domain of NusA is dispensable for integration of SuhB into the rrnTAC. In contrast to NusA AR2 , rrnGnut RNA was required for stable integration of SuhB into transcription complexes: SuhB did not join an RNAP-NusA complex ( Figure 4D) or a complex containing RNAP, NusA and DNA but lacking rrnGnut RNA ( Figure 4E); however, it did associate with RNAP-NusA when DNA and rrnGnut RNA were also present ( Figure 4F).
These observations were unexpected, given that SuhB forms a stable complex with NusA and NusA AR2 in iso-lation (see above). However, the NusA AR2 domain can also interact with the C-terminal domain of the RNAP ␣ subunit (␣CTD), as recently visualized in a NusAmodified his-operon hairpin-paused elongation complex (6). The NusA AR2 -␣CTD interaction is thought to allow RNA binding by NusA, as RNA binding in isolated NusA is auto-inhibited by AR1-AR2 folding back onto the S1-KH1-KH2 RNA-binding region in a manner mutually exclusive with the NusA AR2 -␣CTD interaction (37,38). We therefore tested whether binding of SuhB and ␣CTD to NusA AR2 is also mutually exclusive. Indeed, while stable SuhB-NusA AR2 ( Figure 3A) and ␣CTD-NusA AR2 (Figure 3C, middle) complexes formed in analytical SEC upon mixing the respective proteins, a mixture of SuhB, ␣CTD and NusA AR2 gave rise to separate SuhB-NusA AR2 and ␣CTD-NusA AR2 complexes eluting from the column (Figure 3C, bottom). In addition, excess SuhB prevented ␣CTD from associating with NusA ( Figure 3D). We therefore suggest that in transcription complexes lacking rrnGnut RNA, NusA tends to bind via its AR2 domain to the ␣CTD of RNAP, preventing SuhB-AR2 interactions.
; B -fraction bound; B max -maximum fraction bound; X -concentration of SuhB; h -Hill slope; K d -dissociation constant). The K d for the SuhB-rrnGnut RNA interaction is indicated on the right; the K d for the SuhB-boxBA RNA interaction could not be determined (n.d.).

SuhB facilitates recruitment of NusB/E and S4 to transcription complexes
Interestingly, the NusB/E dimer failed to associate with transcription complexes bearing rrnGnut RNA when SuhB was omitted, irrespective of the presence of S4 ( Figure  4G, H). In addition, while a NusA-rrnGnut RNA complex formed with excess material in these SEC runs, NusB/E did not associate with this NusA-rrnGnut RNA complex (Figure 4G,H), suggesting that in the absence of SuhB, NusA competes with NusB/E on rrnGnut RNA. Moreover, in the absence of SuhB, S4 only partially associated with rrnGnut RNA-containing transcription complexes ( Figure 4G).
The apparent competitive binding of NusA and NusB/E to rrnGnut RNA in the absence of SuhB was surprising, due to the above-mentioned auto-inhibition of isolated NusA with respect to /consensus nut RNA binding (37,38). However, isolated NusA clearly did bind to rrnGnut RNA even in the absence of other factors (Figure 4I), indicating that this RNA might have a high enough affinity to overcome NusA auto-inhibition. In double filterbinding assays, SuhB bound rrnGnut RNA with a K d of 3.2 M, while binding to a boxBA RNA, which lacked boxC and the boxA-boxC linker (residues 1-40; Figure 1C), was much weaker (K d could not be determined; Figure 5). This observation suggests that binding of SuhB to the boxA-boxC linker region and/or boxC is important for coordinating the concomitant binding of NusA and NusB/E to rrnGnut RNA. This notion is further supported by the observation that SuhB does not facilitate concomitant assembly of NusA and NusB/E on /consensus nut RNA lacking these regions ( Figure 2F). Based on these results, we suggest that SuhB remodels the NusA-rrnGnut RNA complex or repositions NusA on rrnGnut RNA, by concomitantly interacting with NusA (via the AR2 domain) and the RNA (via the boxA-boxC linker or boxC). Such remodeling/repositioning might facilitate subsequent binding of NusB/E.

Structural basis of the SuhB-NusA AR2 interaction
To begin elucidating the structural basis underlying the transcriptional functions of SuhB, we determined a crystal structure of a SuhB-NusA AR2 complex at 1.65Å resolution (Table 1; Figure 6A). An asymmetric unit of the crystals contained two SuhB molecules and a single copy of NusA AR2 . The structures of the two SuhB monomers are almost identical (root-mean-square deviation [rmsd] 0.49Å for 248 pairs of common C␣ atoms; Figure 6B). The SuhB monomers are folded as an alternating stack of three pairs of ␣ helices (helix pairs I-III) and two ␤ sheets (sheets I and II; Figure 6A). The six-stranded, antiparallel sheet I is sandwiched between helix pairs I and II, the five-stranded, mixed sheet II is sandwiched between helix pairs II and III. The AR2 domain formed a dual helix-hairpin-helix motif with four ␣ helices and one 3 10 helix (order ␣1-␣2-␣3-3 10 -␣4; Figure 6A), as observed before in other molecular contexts (38,39).
The SuhB monomers formed a near-C2-symmetrical dimer ( Figure 6A) that closely resembled the previously observed dimer of an isolated SuhB R184A variant (25) (rmsd 0.66Å for 491 pairs of common C␣ atoms). About 1600Å 2 of combined surface area are buried at the dimer interface. It has been suggested that wild type (wt) SuhB might form monomers, and that the R184A mutation stabilized the dimer state observed in the crystal structure of SuhB R184A (25). We therefore also crystallized wt SuhB alone and determined its structure at 2.25Å resolution (Table 1). Isolated wt SuhB crystallized with one monomer in an asymmetric unit, but a dimer essentially identical to the dimer seen with SuhB R184A (rmsd 0.56Å for 482 pairs of common C␣ atoms) and in the SuhB-NusA AR2 complex (rmsd 0.49 A for 497 pairs of common C␣ atoms) was formed by crystal symmetry ( Figure 6C). Moreover, SEC coupled to multiangle light scattering confirmed that SuhB forms dimers under the chosen conditions in solution (experimental molecular mass: 60.7 kDa; calculated molecular mass for a SuhB dimer: 58.3 kDa).
In the SuhB-NusA AR2 crystal, two symmetry-related copies of NusA AR2 were bound laterally at equivalent surfaces of the two SuhB subunits of the dimer, i.e. below one exposed edge of sheet I and along the exposed lateral face of helix pair II, sheet II and helix pair III of SuhB ( Figure  6A). However, one of the NusA AR2 copies associated with one SuhB subunit via a flat surface formed by its C-terminal ␣3-3 10 -␣4 portion (interaction mode 1; Figures 6A,B and  7A), while the other AR2 domain bound the second SuhB subunit via an edge formed by its ␣1, ␣3 and ␣4 elements (interaction mode 2; Figures 6A, B and 7B).
Interaction modes 1 and 2 bury about 650Å 2 and about 500Å 2 of combined surface area, respectively, showing that interaction mode 1 encompasses a more extended interface  between SuhB and NusA AR2 . Moreover, superimposing the AR2 domain of a ␣CTD-NusA AR2 complex (38) on the AR2 domains in the SuhB-NusA AR2 crystal structure revealed that SuhB bound in interaction mode 1 would preclude ␣CTD-AR2 interactions ( Figure 6D), while interaction mode 2 would allow formation of a ternary SuhB-NusA AR2 -␣CTD complex ( Figure 6E). Thus, only interac-tion mode 1 can explain the mutually exclusive binding of SuhB and ␣CTD to NusA AR2 we observed ( Figure 3C,D). We also exchanged interacting residues in the two observed SuhB-AR2 interfaces by site-directed mutagenesis and assessed the effects of these mutations on complex formation between SuhB and NusA AR2 by analytical SEC and surface plasmon resonance (SPR) analyses. While all protein variants behaved very similarly to the wt proteins during production and purification, equilibrium circular dichroism (CD) spectra and CD-based thermal melting analyses indicated that at least some of the NusA AR2 domain variants may have suffered changes in 3D structure and/or fold stability. Possibly due to these alterations, SEC analyses did not yield conclusive results. In addition, only wt SuhB, SuhB KL251/254AA (interaction mode 1 affected) and SuhB K108A (interaction mode 2 affected) in combination with wt AR2 gave quantifiable results in SPR runs (Figure 7C). The SPR results confirmed formation of a stable complex between wt SuhB and AR2 with an estimated dissociation constant of K d = 94 nM. The analyses were again consistent with interaction mode 1 representing the relevant complex, as SuhB mutations affecting interaction mode 1 led to a reduction of more than three orders of magnitude in the affinity to AR2 (K d = 141 M for SuhB KL251/254AA -AR2), while a SuhB mutation affecting interaction mode 2 only had a mild effect (∼4.5-fold reduction in affinity; K d = 435 nM for SuhB K108A -AR2; Figure 7C). Taken together, the above data suggest that interaction mode 1 represents the SuhB-AR2 interaction mode in solution while interaction mode 2 most likely is a result of crystal packing. Interestingly, while SuhB KL251/254AA consistently failed to stably bind NusA AR2 in SEC as well ( Figure 7D, top), a complex with NusA FL remained partially intact during gel filtration ( Figure 7D, bottom). Therefore, regions beyond the AR2 domain in NusA FL most likely also contribute to stable complex formation with SuhB.

SuhB and NusA AR2 are required for rrn anti-termination
To test the effect of SuhB on transcription, we designed a linear DNA template bearing the T7A1con promoter, followed by a region encoding an rrnGnut site, rutA/rutB entry signals, the trpt' termination region and the tR' intrinsic terminator (Figure 8A), and employed this template in in vitro transcription assays. Production of an initial, radioactively labeled transcript was initiated at the promoter via RNAP holoenzyme by withholding UTP. Subsequently, Nus factors, S4, SuhB and/or were added and transcription was continued by supplying a large excess of all four unlabeled nucleotide tri-phosphates. Samples were taken at 15minute time points and products were visualized by dena- RNA products are identified on both sides of the gel; -term. in trpt' -transcripts terminating in the trpt' region; read-through of trpt' -transcripts extending beyond the trpt' region. Bottom panels -quantification of the data. Relative read-through of trpt' was calculated as the fraction of trpt' read-through transcripts relative to all transcripts in a lane, with the value for acting on RNAP alone set to 0 and the values for all other complexes scaled accordingly. Quantified data represent means ± SD of three independent experiments. In this and the following figures, significance was assessed by Student's unpaired t-test. Significance indicators in this and the following figures: *P < 0.05; **P < 0.01; ***P < 0.001. (C) Long time courses of the same experiments with acting on RNAP alone (left) and on a complex of RNAP, all Nus factors and SuhB (right). Run-off transcripts accumulate in the experiment shown on the right, illustrating anti-activity. turing PAGE and autoradiography. Products observed under these conditions represent the outcome of single-round transcription events.
As expected, addition of to RNAP alone led to increased termination in the trpt' region ( Figure 8B, lanes 1 and 2; note that we quantified relative read-through of the trpt' region). Addition of NusA led to a larger fraction of transcripts extended beyond the trpt' region (Figure 8B, lane 3), consistent with NusA being able to act as a antagonist (10). Further addition of NusG jointly with NusB/E overrode the NusA effect ( Figure 8B, lane 4), most likely due to the NusG CTD aiding RNA engagement by (7) (note that we included NusB/E, although results above suggest that in the absence of SuhB the proteins would not efficiently associate). The pattern of -dependent termination upon subsequent addition of S4 resembled the behavior seen with RNAP-NusA ( Figure 8B, lane 6), indicating that S4 exhibits some -delaying activity, as noted previously (15). However, only when SuhB was included (with or without S4), -dependent termination was significantly delayed or even suppressed ( Figure 8B, lanes 5 and 7). S4 had no significant additional effect in the presence of SuhB ( Figure 8B, lanes 5 and 7), but the SuhB effect was clearly dependent on the presence of all Nus factors ( Figure 8B, lanes 8 and 9).
Long time courses of the experiment showed that in the presence of SuhB and all Nus factors, transcripts that were elongated beyond the trpt' region accumulated monotonously ( Figure 8C). As experiments were performed at high concentrations of ATP and GTP (2 mM and 100 M, respectively), which are the nucleotides that limit elongation at pause sites, the accumulating transcripts most likely resulted from termination rather than long lived elemental/hairpin-stabilized pausing events. While we cannot fully exclude the possibility that some of these transcripts are the result of prolonged stalling/backtracking of RNAP, SuhB in combination with the Nus factors also led to significant accumulation of run-off transcripts ( Figure 8C), confirming that under these conditions SuhB strongly delays and most likely at least partially suppresses -dependent termination. When we employed NusA AR2 instead of NusA FL in the assays, SuhB-mediated delay or suppression of activity was significantly reduced both in the absence ( Figure 8B, lane 10) and presence of S4 ( Figure 8B, lane 12). Likewise, the SuhB KL251/254AA variant, which exhibits strongly reduced affinity to NusA AR2 ( Figure 7C), was significantly less efficient than wt SuhB in delaying or suppressing ( Figure 8B, lanes 11 and 13). Together, the above findings indicate (i) that SuhB is an essential component of the rrnTAC to delay or suppress -dependent termination, (ii) that SuhB requires the Nus factors to elicit its anti-effects and (iii) that a SuhB-NusA AR2 interaction contributes to the anti-activity.
We used the same assay to inspect the influence of the factors on intrinsic termination in the absence of ( Figure  9A, B). This analysis revealed that intrinsic termination by RNAP was significantly enhanced in the presence of NusA ( Figure 9B, lanes 1 and 2; note that we quantified relative read-through of the tR' intrinsic terminator), most likely due to the known ability of NusA to stabilize RNA hairpins in the RNA exit tunnel of RNAP (3)(4)(5)(6). The NusA effect was reduced stepwise by addition of NusG+NusB/E and NusG+NusB/E+S4 ( Figure 9B, lanes 3 and 5). However, only upon addition of SuhB intrinsic termination was significantly suppressed compared to RNAP alone ( Figure 9B,  lanes 4 and 6). Again, the SuhB effect was largely independent of S4 ( Figure 9B, lanes 4 and 6) but strongly dependent on the presence of all Nus factors ( Figure 9B lanes 7 and  8). As for -dependent termination, suppression of intrinsic termination by SuhB depended in part on the presence of a NusA AR2 domain ( Figure 9B, lanes 9 and 11) and on an intact binding site on SuhB for NusA AR2 (Figure 9B, lanes  10 and 12). Thus, in the context of all Nus factors, SuhB is able to override the intrinsic termination-supporting function of NusA, most likely in part due to an interaction with the NusA AR2 domain.

DISCUSSION
The structural details underlying formation of a transcription complex that is specialized for rRNA synthesis in E. coli, the rrnTAC that involves RNAP, a nut-like site on the transcript, NusA, B, E and G, r-protein S4 and SuhB, are presently not known. The nut-like signal element renders the rrnTAC specific for transcription of rRNA operons. As an initial step to characterize this complex on the molecular level, we have investigated the interactions of the SuhB subunit with other rrnTAC components and the functional consequences of the detected interactions for the assembly of a rrnTAC and for the ability of the rrnTAC to counteract -dependent and intrinsic transcription termination.

Inter-dependencies of factors in the assembly and function of an rrnTAC
We find that SuhB stably interacts with the AR2 domain of NusA ( Figures 3A and 7C) and determined a crystal structure of a SuhB-NusA AR2 complex ( Figure 6A). However, in the presence of rrnGnut RNA, the SuhB-NusA AR2 interaction is dispensable for stable integration of SuhB into an rrnTAC ( Figure 4C). Our results suggest that interaction of an RNAP ␣CTD with NusA AR2 might sequester the AR2 domain from SuhB in the absence of rrnGnut RNA and lead to failure of SuhB to integrate into such complexes. By engaging in interactions with multiple components of the rrnTAC, rrnGnut RNA may weaken the ␣CTD-AR2 interaction. Additionally, rrnGnut RNA may increase the local concentration of SuhB so that SuhB can compete with ␣CTD for AR2 binding. Irrespective of the precise mechanism, our data strongly support the view that a SuhB-AR2 interaction ensues in the assembled rrnTAC, as deletion of AR2 from NusA or disruption of the SuhB-AR2 interaction interfered with SuhB-dependent and intrinsic antitermination ( Figures 8B and 9B). ChIP analyses using reporter systems that lacked NusB or boxA (16) and the weakness of the reported SuhB-RNAP holoenzyme interaction (25) suggested that SuhB is incorporated into the rrnTAC in a NusB/boxA-dependent manner (16). However, SuhB deletion also led to a depletion of the NusB ChIP signal (16), leaving open the question of which factor is responsible for stable incorporation of the other. Our results clearly support a functional interplay between SuhB, NusA, the NusB/E dimer and rrnGnut RNA in rrnTAC assembly, and they suggest that SuhB is required for subsequent integration of NusB/E rather than vice versa. Binding of NusA to rrnGnut RNA was incompatible with concomitant binding of NusB/E, but NusB/E binding was possible when SuhB was additionally present ( Figure 2E). NusA might interfere with NusB/E binding by sterically blocking NusB/E or by occupying the boxA element, which is the known interaction region of NusB/E (20,21,40,41). SuhB-NusA contacts may lead to a different conformation of NusA on rrnGnut RNA that no longer blocks NusB/E access to boxA. Alternatively, SuhB, based on its affinity to the boxA-boxC-linker/boxC region of rrnGnut RNA, may guide NusA to this part of the RNA, granting NusB/E access to boxA. Together with a previous crystal structure showing that a S1-KH1-KH2 fragment of M. tuberculosis NusA can bind a boxC-like RNA element (22), our findings are consistent with NusB/E binding to boxA, SuhB to the boxA-boxC linker and NusA to boxC in the final complex.
Most likely, additional interactions of SuhB with other components of an rrnTAC, including RNAP itself (25), ensue in the fully assembled complex. It has been shown that other transcription elongation and anti-termination complexes assemble cooperatively based on many binary interactions among the participating factors, which are not necessarily all stable in isolation (19). Indeed, our interaction studies provided evidence that NusA FL establishes at least one additional contact to SuhB beyond the AR2 domain ( Figure 7D). Moreover, our transcription assays showed that other Nus factors are also required for SuhB to unfold its negative effects on -dependent and intrinsic termination ( Figures 8B and 9B), supporting the idea that more complex, multi-factorial interactions ensue around SuhB in the complete rrnTAC. Finally, our finding that stable binding of S4 to an rrnTAC partly depends on SuhB ( Figure 4G) could be due to direct SuhB-S4 interactions in the full complex.

Molecular basis of transcription anti-termination by an rrnTAC
It has been proposed that the main function of the rrnTAC does not lie in the suppression of premature transcription termination but rather in its support for co-transcriptional rRNA folding, maturation and ribosomal subunit assembly (16,42). However, these conclusions were based on studies with individual nusB or suhB knockout strains, in which residual anti-termination activity based on the other factors may still remain. Our results clearly demonstrate that a transcription complex minimally comprising rrnGnut RNA, all Nus factors and SuhB efficiently counteracts or delays -dependent termination and likewise interferes with intrinsic termination.
We showed that SuhB is responsible to a large part for the ability of an rrnTAC to subdue -dependent as well as intrinsic termination. Deletion of the NusA AR2 domain or disruption of the SuhB-NusA AR2 interaction clearly reduced both types of anti-termination activities ( Figures 8B  and 9B). A recent cryo-electron microscopic structure of a NusA-stabilized his operon hairpin-paused transcription complex suggested that an observed ␣CTD-NusA AR2 contact may contribute to the positioning of NusA on RNAP in a pause hairpin-stabilizing conformation (6). A similar conformation of NusA on RNAP can be expected when it exerts its role in supporting intrinsic termination. Thus, SuhB binding the NusA AR2 domain, which as we showed is mutually exclusive with the ␣CTD-NusA AR2 interaction ( Figure 3C, D), may contribute to a different positioning of NusA on RNAP, such that its ability to stabilize RNA hairpins in the RNA exit tunnel of RNAP is prevented or reduced. We recently showed that a similar principle is at work during N-dependent processive anti-termination, mediated by the N protein (18,19).
How SuhB contributes to anti-activity is presently unclear, but it may again involve similar principles as suggested for N-mediated anti-termination. In the structure of a N-TAC (18,19), the NusG CTD, which aids in engaging rut sites (7), is sequestered by alternative interactions with NusA and NusE. However, additional mechanisms contribute to counteracting in the N-TAC, possibly involving steric hindrance that would prevent from approaching its binding site(s) on RNAP (18,19). Likewise, SuhB in conjunction with the Nus factors might prevent from productively contacting RNAP, as is required fordependent termination (43).

Implications for co-transcriptional rRNA folding, processing and ribosome assembly
A 'delivery' model has been proposed to explain the role of the rrnTAC in co-transcriptional folding and maturation of rRNA (13,42,44). In this model, the RNAP-modifying components of the rrnTAC hold on to the RNA nut-like element encoded in the leader and spacer regions of the rrn operons, thereby keeping following regions of the transcript in close proximity of the RNAP RNA exit tunnel. When 3regions of rRNA domains are synthesized at later stages of transcription and emerge from the exit tunnel, they would thus be presented with 5 -regions that they need to pair with. The findings that NusB and boxA constitute crucial components for the functioning of the rrnTAC (14,16,42) are consistent with this idea, as the NusB/E dimer establishes contacts to boxA. Our data suggest that there might be a more elaborate multi-protein-RNA interaction on the rrnTAC that mediates the delivery mechanism or that contributes in other ways to rRNA maturation, as SuhB and NusA apparently reinforce the grip of the rrnTAC on the transcript leader/spacer region by contacts to the nut-like RNA element (Figure 5), and as SuhB is required for efficient incorporation of the NusB/E dimer in the presence of NusA ( Figure 4G, H).