The crystal structure of the RsbN-σBldN complex from Streptomyces venezuelae defines a new structural class of anti-σ factor.

Streptomyces are filamentous bacteria with a complex developmental life cycle characterized by the formation of spore-forming aerial hyphae. Transcription of the chaplin and rodlin genes, which are essential for aerial hyphae production, is directed by the extracytoplasmic function (ECF) σ factor BldN, which is in turn controlled by an anti-σ factor, RsbN. RsbN shows no sequence similarity to known anti-σ factors and binds and inhibits BldN in an unknown manner. Here we describe the 2.23 Å structure of the RsbN-BldN complex. The structure shows that BldN harbors σ2 and σ4 domains that are individually similar to other ECF σ domains, which bind -10 and -35 promoter regions, respectively. The anti-σ RsbN consists of three helices, with α3 forming a long helix embraced between BldN σ2 and σ4 while RsbN α1-α2 dock against σ4 in a manner that would block -35 DNA binding. RsbN binding also freezes BldN in a conformation inactive for simultaneous -10 and -35 promoter interaction and RNAP binding. Strikingly, RsbN is structurally distinct from previously solved anti-σ proteins. Thus, these data characterize the molecular determinants controlling a central Streptomyces developmental switch and reveal RsbN to be the founding member of a new structural class of anti-σ factor.

INTRODUCTION

Bacterial transcription is driven by an evolutionarily conserved catalytic RNA polymerase (RNAP) that consists of five subunits, α2ββ'ω. This core enzyme must associate with one of several dissociable transcription initiation factors called sigma (σ) factors to elicit promoter-specific transcription (1,2). Transcription of most genes is driven by the primary σ factor (called σ70 in Escherichia coli) but alternative σ factors redirect transcription to smaller regulons of genes, allowing bacteria to mount specific responses to discrete environmental and internal cues (1,3,4). σ factors are modular proteins having up to four conserved globular domains (σ1-4) with distinct functions (5). The −10 and −35 promoter elements are recognized and bound by σ2 and σ4, respectively (1,6–13), whereas the interface that interacts with RNAP is extensive, involving σ2, σ3 and σ4 (1,10,13).

The most abundant and phylogenetically diverse class of σ factors is the extracytoplasmic function (ECF) σ factor subfamily, so-called because more than 80% of these proteins regulate physiological functions in response to stimuli arising outside the cytoplasmic membrane (14–18). These σ factors have also been termed group IV σ factors (3). Structurally, these proteins are the most reduced σ factors, consisting only of σ2 and σ4 connected by a linker (12,14,15,18). It is typical for bacteria to express these σ factors but to hold them in an inactive form until appropriate stimuli trigger their release, permitting a rapid response to the relevant signal. This inhibition is achieved through the action of anti-σ factors, proteins that bind their cognate σ factors and prevent them from interacting with RNAP (1,16–21). Very often ECF anti-σ factors are transmembrane proteins that sequester σ to the lipid bilayer, with their extracytoplasmic domain directly or indirectly receiving the signal that ultimately leads to release of σ on the cytoplasmic side of the membrane (1,16–21).

To date, several ECF σ–anti-σ complexes have been structurally characterized, including the E. coli RseA–σE, Rhodobacter sphaeroides ChrR–σE, Bacillus subtilis RsiW–σW and Mycobacterium tuberculosis RskA–σK complexes (22–25). Interestingly, although the anti-σ factors in these structures show limited sequence similarity, they contain a conserved fold consisting of a three-helix bundle followed by a fourth helix of variable position (22–25). In all these structures, this anti-σ factor domain (ASD) locks the σ factor in a non-optimal position for promoter docking and prevents its interaction with RNAP and DNA. More recent structural studies on the anti-σ factors NepR and CnrY revealed these proteins to be composed of just two helices that act as a clamp to embrace the outer surface of their partner proteins to inhibit their activity (26–28). CnrY binds the σ factor, CnrH, while the NepR protein is involved in a partner switching mechanism wherein it can bind either the σ factor EcfG or the phosphorylated form of the σ factor mimic, PhyR. While a structure of an NepR–EcfG complex has not yet been solved, NepR–PhyR structures suggest this complex mimics the NepR–EcfG complex. In the NepR–PhyR complex, similar to CnrY, NepR forms a two-helix clamp that encircles the surface of PhyR (27,28). Thus, structural studies to date have revealed two ECF anti-σ factor folds.

The structures of σ–anti-σ complexes also show that the individual σ2 and σ4 domains of the ECF σ factors adopt essentially the same fold, which are similar to the corresponding domains of the larger primary σ70 proteins (22–28). The σ4 domains bind the −35 promoter element via a helix-turn-helix (HTH) motif and the σ2 domains use a loop to recognize and melt out the −10 motif of the promoter. Structures of ECF σ domains bound to cognate −10 and −35 sequences reveal that, in general, they dock onto DNA in a similar fashion to the corresponding domains in σ70 proteins (1,6,8,12). However, ECF σ2 domains mediate single base flipping events within their target −10 elements compared to the double base exposure exhibited by primary σ factors. In addition, ECF σ4 domains appear to bind more tightly and specifically to their −35 elements than primary σ factors (1).

ECF σ–anti-σ modules provide a rapid and modular readout of stress responses in bacteria, and they are now recognized as one of the three major pillars of signal transduction in bacteria (16–17). Most bacteria possess multiple ECF σ factors and in general bacteria with large genomes have larger numbers of ECF σ factors. Indeed, members of the genus Streptomyces typically have ∼9 Mb genomes and ∼50 ECF σ factors. These filamentous bacteria have a complex developmental life cycle that ultimately leads to the formation of long chains of exospores. During development, Streptomyces also produce a plethora of secondary metabolites, which serve as important sources of antibiotics and other clinically important agents such as anti-cancer and anti-fungal drugs (29,30). Most ECF σ factors that have been characterized function in environmental or endogenous stress responses, but in Streptomyces the ECF σ factor BldN plays a key role in morphological development (31). Specifically, the bldN gene is a direct target of the master transcriptional regulator, BldD, which sits at the top of the Streptomyces developmental hierarchy (Figure 1). Recent studies showed that c-di-GMP levels control BldD activity by mediating dimerization of two BldD subunits. BldD-c-di-GMP dimers drive repression of the large BldD regulon of sporulation genes, extending vegetative growth and blocking the onset of morphological differentiation (31–33). When c-di-GMP levels drop, the BldD dimer falls apart and dissociates from the DNA, which allows the production of BldN. The BldN protein plays a pivotal role in the onset of development by directing transcription of the rodlin and chaplin genes (31), which encode the proteins required to form an external hydrophobic sheath that permits the reproductive aerial hyphae to escape surface tension and grow into the air (Figure 1). These proteins also likely prevent the desiccation of the aerial hyphae, which undergo a massive synchronous septation event to form chains of exospores for dispersal (34–40).

Figure 1.

The biological role of BldN and RsbN. During vegetative growth, transcription from the bldN and rsbN promoters is repressed by the BldD–(c-di-GMP) complex. At the onset of development, c-di-GMP levels drop and BldD repression is relieved. As a consequence, BldN and RsbN are expressed and RsbN sequesters BldN to the membrane via its cytoplasmic anti-σ domain. In response to an unknown signal, RsbN releases BldN, which directs transcription of the two gene families encoding the chaplins and the rodlins, and these proteins assemble into functional amyloids to form the external hydrophobic sheath that allows the reproductive aerial hyphae to escape surface tension and grow into the air.

A second key direct target of BldD-(c-di-GMP)-mediated repression is the promoter of the rsbN gene (Figure 1), which is a direct target of BldN. We previously showed that RsbN functions as a BldN-specific anti-σ factor, and that deletion of rsbN triggers precocious development, emphasizing the importance of this post-translational level of BldN regulation (31). RsbN is a transmembrane protein, having a single transmembrane helix (residues 118–139) that connects a 117 residue N-terminal cytoplasmic region to a 272 amino acid C-terminal extracytoplasmic domain (Figure 1) (31). RsbN does not show sequence similarity to any known anti-σ factor or, indeed, any known protein. To discover how RsbN binds the key developmental regulator BldN, we determined the crystal structure of the complex between BldN and the cytoplasmic anti-σ factor domain of RsbN. The structure reveals how RsbN inhibits BldN and also identifies RsbN as the founding member of a new structural class of anti-σ factor.

MATERIALS AND METHODS

Purification of RsbN–BldN

We previously showed that full length RsbN interacts directly with BldN (31). Subsequent limited proteolysis analyses revealed that the minimal BldN-interacting region of RsbN consists of residues 1–91. Thus, for structural studies bldN and rsbN(1–91) were cloned into the pETDuet-1 vector for coexpression and purified as a single complex. The resultant complex was >95% pure after a cobalt-nitrilotriacetic acid column chromatography step and was used for crystallization trials. Large crystals of the complex (at a concentration of 80 mg/ml) were obtained via hanging drop vapor diffusion using 30% polyethylene glycol (PEG) 3350, 0.2 M ammonium sulphate as a crystallization reagent. However, the crystals diffracted weakly to 8 Å and were significantly twinned. As previous studies on σ–anti-σ factors showed that the linker between the σ factor domains is flexible, we generated a system in which RsbN(1–91) was expressed with individual BldN domains (domain 2 and domain 4). In this expression system, his-tagged RsbN91 and non-tagged BldN domain 2 (σ2) were expressed from pCOLADuet-1 (pIJ10626), and non-tagged BldN domain 4 (σ4) was expressed from pET15b (pIJ10622) (Supplementary Table S1). For protein expression of this system, both plasmids were transformed into C41(DE3) cells and grown in Luria-Bertani (LB) liquid medium to an OD600 of 0.6 and induced with 1 mM isopropyl β-D-1 thiogalactopyranoside (IPTG) for 3 h at 30 °C. The cells were then suspended in buffer A (25 mM Tris pH 7.5, 5% glycerol, 300 mM NaCl) and disrupted with a microfluidizer. The lysate was loaded onto a cobalt-nitrilotriacetic acid column. The his-RsbN(1–91) bound both non-tagged BldN σ domains and the ternary complex was eluted using increasing concentrations of imidazole in buffer A. For protein used for crystallization trials, an additional size exclusion chromatography (SEC) purification step was included. The protein complex was >97% pure after this step. The purified protein complex from this expression system produced well diffracting crystals that were used for structure determination.

Crystallization and structure determination of the BldN(σ2)–BldN(σ4)–RsbN(1–91) complex

Prior to crystallization, the RsbN his-tag was removed from the RsbN(1–91) that was in complex with BldN(σ2)–BldN(σ4) using a thrombin capture cleavage kit (Qiagen). Following thrombin treatment, the BldN(σ2)–BldN(σ4)–RsbN(1–91) complex was concentrated to 50 mg/ml using 30 kDa cutoff microcon concentrators, which simultaneously concentrated the complex and removed the cleaved his-tag. Crystals were obtained of the concentrated complex using hanging drop vapor diffusion. Specifically, the complex was mixed 1:1 with a crystallization solution consisting of 0.2 M ammonium nitrate, 26% PEG 3500 and placed over a 1 ml well containing the crystallization solution. Crystals were obtained at room temperature (RT) and took one week to grow to maximum size. The crystals were cryo-preserved by a 1–2 s dip in the crystallization solution supplemented with 10–20% glycerol. An x-ray intensity dataset was collected at the Advanced Light Source (ALS) beamline 8.3.1, the data were processed in MOSFLM (41) and scaled using SCALA (42). Molecular replacement was used to solve the structure using a multistep strategy. First, the σ4 domain from E. coli σ70 (PDB ID: 2P7V) (43) was employed as a search model and produced a clear solution in Molrep. This partial model was then used as a static structure to obtain a solution of the σ2 domain from CnrH (PDB ID: 4CXF) (26). The model was subjected to refinement in Phenix to 2.6 Å (44). After several rounds of refinement, rebuilding and replacement of the BldN sequence (45), density for RsbN α3 was evident and built as a polyalanine model. Following several more rounds of refinement, RsbN α1–α2 became clear and the sequence was evident for all residues in the RsbN helices. After several more rounds of rebuilding and solvent addition with concomitant analyses using Molprobity (46) the Rwork/Rfree converged to 19.2/23.5% to 2.23 Å resolution (Table 1). The clashscore and overall score from Molprobity place the structure in the 99% range of structures solved to similar resolutions (46). The final model includes residues 2–80, 114–172 of BldN, RsbN residues 4–21, 39–79 and 72 water molecules.

Table 1.

Data collection and refinement statistics for the RsbN–BldN structure

Structure	RsbN–BldN
PDB code	6C03
Space Group	P21
Cell dimensions
a,b,c (Å)	44.88, 43.58, 58.75
α,β,γ (°)	90, 111.56, 90
Resolution (Å)	34.5–2.23
Total reflections, #	23950
Unique reflections, #	9444
R sym	0.108 (0.208)*
R pim	0.077 (0.190)
CC(1/2)	0.908 (0.889)
I/σI	6.5 (2.9)
Completeness (%)	99.4 (70.1)
Redundancy	2.5 (1.8)
Refinement
Resolution (Å)	34.5–2.23
R work/Rfree (%)	19.2/23.5
RMS deviations
Bond lengths (Å)	0.003
Bond angles (°)	0.678
Ramachandran analyses
Favored (%)	98.4
Disallowed (%)	0.0

*Values in parentheses are for highest resolution shell.

Size exclusion chromatography (SEC) analysis of WT RsbN–BldN

SEC was used to probe the MW of the RsbN–BldN complex with a HiLoad 26/600 Superdex 75 prep grade column. Experiments were performed in a buffer containing 200 mM NaCl, 5% glycerol, 20 mM Tris HCl pH 7.5 and 1 mM β-mercaptoethanol (BME). For SEC analysis, the RsbN–BldN complex was loaded at a concentration of 10 μM.

Biochemical analyses of truncated RsbN–BldN complexes

To further dissect the RsbN–BldN complex, we created a series of constructs to co-express truncated versions of BldN and RsbN, using the pCOLADuet-1 vector backbone (Supplementary Table S1). Plasmids were introduced into BL21(DE3) Rosetta-plysS and the resulting strains were grown in LB liquid medium to an OD600 of 0.6 and induced with 1 mM IPTG for 2–3 h at 30°C. Cell pellets were suspended in buffer A, lysed by homogenization (two passes) using the Avestin Emulsiflex-B15 and the insoluble material was removed by centrifugation. Lysates were loaded onto 1 ml HiTrap columns (GE Healthcare) charged with 100 mM CoCl2 and proteins were eluted using Buffer A + 500 mM imidazole. Representative fractions were run on a 16% Tricine gel (47), designed to resolve low MW proteins, and stained with Coomassie Brilliant Blue. The identity of each eluted protein was confirmed by MALDI-MS.

Analysis of site-directed mutants to probe the RsbN–BldN interaction

To test the effect of replacing key residues that the structure implied would be important for complex formation, we generated RsbN91(L70R) and RsbN91(F14R-L50R) mutants and assessed their abilities to interact with BldN σ2 and BldN σ4, respectively. Specifically, RsbN91(L70R) was co-expressed with BldN σ2 and RsbN91(F14R-L50R) was co-expressed with BldN σ4 (Supplementary Table S1). Proteins were co-expressed in E. coli as above but instead purified using HIS-select spin columns (Sigma H7787) and run on 16% Tricine gels (47) for visualization with Coomassie Brilliant Blue.

To probe the BldN σ4–RsbN α3 interaction, a co-expression system in which the wild-type (WT) full length BldN and a fusion construct consisting of the maltose binding protein (MBP) connected via a linker (SHSGGGSGGGS) to RsbN α3 (herein termed MBP-RsbN α3) were encoded in pETDuet-1 (Supplementary Table S1). For protein expression, the plasmid was transformed into C41(DE3) cells and grown in LB to an OD600 of 0.6 and induced with 1 mM IPTG for 3 h at 28°C. The cells were suspended in buffer A, disrupted with a microfluidizer and the lysate mixed with cobalt-nitrilotriacetic acid resin. Protein containing fractions were then applied (at a concentration of 10 μM) to a HiLoad 26/600 Superdex 75 column in buffer A.

RESULTS

Structure of Streptomyces venezuelae RsbN–BldN complex

We previously showed that Streptomyces venezulae BldN formed a stable complex with the full-length 412 residue RsbN protein (31). Further analysis showed that an N-terminal 91 residue cytoplasmic domain of RsbN (RsbN91) was sufficient to mediate the interaction with BldN. Because BldN shows reduced solubility in the absence of RsbN (31), we constructed co-expression systems for structural studies. Crystals were obtained of full-length BldN in complex with RsbN91 but they did not diffract beyond 8 Å resolution. We speculated that this might be due to the long, 27-residue flexible linker connecting σ2 and σ4 of BldN. The length of this linker is similar to those found in other ECF σ factors (22–26). For example, the linkers in E. coli σE, M. tuberculosis σK, B. subtilis σW and CnrH, which range from 26 to 32 residues in length, are disordered in the σK, σW and CnrH structures in complex with their anti-σ factors and mostly disordered in the structure of σE bound to its anti-σ factor, RseA (22,24,25). Therefore, we generated a three-component co-expression system (‘Materials and Methods’ section) that produced RsbN91 and the σ2 and σ4 domains of BldN as three separate polypeptides. The protein complex purified from this system (herein referred to as RsbN–BldN) produced well diffracting crystals. Data were collected to 2.23 Å resolution, the structure solved by molecular replacement and refined to final Rwork/Rfree values of 19.2/23.5% (‘Materials and Methods’ section and Table 1).

The structure shows that RsbN is composed of three helices that are bound to both the σ2 and σ4 domains of BldN (Figure 2A–D). The three RsbN helices are connected by flexible linkers. This flexibility is underscored by the finding that in the structure the linker between helices 2 and 3 domain swaps, which leads to helices 1–2 contacting a neighboring BldN σ4 region in the crystal (Figure 2B). Domain swapping is an event in which a protein subunit exchanges a domain or part of its structure with an identical partner. This exchange requires a flexible connection between domains and generally occurs at high protein concentrations. The resultant swapped subunits have the identical structures as the unswapped subunit except for the flexible hinge region that connects the exchanging parts. This behaviour is significantly enhanced by certain kinds of hinge sequences in proteins that have been identified as hot-spots for domain swapping (48). The program H-Predictor quantifies for each residue in a protein the propensity to form such a hinge region (48). Strikingly, H-predictor identified the residues PKP in the linker between RsbN helices 2 and 3 as having an extremely high probability to form a hinge region that would be involved in domain swapping (Supplementary Figure S1), exactly reflecting what we observed in the crystal structure. As noted, domain swapping is also promoted by high protein concentration (48) and to obtain crystals, set-ups had to performed using RsbN–BldN at 50 mg/ml. To assess the oligomeric state of the complex in solution, we performed SEC. If the domain-swapped structure exists in solution it would produce an indefinitely large polymer as each RsbN contacts successive BldN subunits (Supplementary Figure S2A). However, if the complex is as shown in Figure 2C, the MW obtained from the SEC experiment should be consistent with a 1:1 complex (with a calculated MW of 33 kDa). The SEC experiment revealed a MW of 39 kDa, supporting that RsbN–BldN exists as a 1:1 complex in solution (Supplementary Figure S2B).

Figure 2.

Crystal structure of the Streptomyces venezuelae RsbN–BldN complex. (A) 2Fo-Fc map (light blue mesh) contoured at 1 σ of the RsbN–BldN structure before addition of RsbN residues. Shown is a region around RsbN α3 (magenta). BldN residues are colored cyan. (B) Overall structure of RsbN–BldN. RsbN is colored magenta, one subunit of BldN is cyan and the other is blue. The asymmetric unit consists of one RsbN and one BldN subunit; however the RsbN molecule undergoes domain swapping leading to interaction of RsbN α3 with one BldN and RsbN α1–α2 with another. Symmetry generation leads to an infinite array of RsbN–BldN complexes. (C) RsbN–BldN structure consistent with solution data showing that one subunit of RsbN contacts one BldN with RsbN α3 encircled by both BldN subunits and RsbN α1–α2 interacting with BldN σ4. (D) Electrostatic surface representation of BldN (shown in same orientation as panel C) bound to RsbN (shown as magenta cartoon). Electronegative and positive regions are colored red and blue, respectively, of the BldN surface. Note the hydrophobic (shown in white) nature of the RsbN interacting regions in BldN.

In the structure, the BldN σ2 domain is composed of helices 1–4, and the σ4 domain is composed of helices 5–8, with helices 7 and 8 forming an HTH motif. Structural homology searches (DALI) with the RsbN–BldN structure revealed that the BldN σ2 and σ4 domains individually show significant structural homology with multiple σ factors. In particular, BldN σ2 and σ4 superimpose on E. coli σE2 and M. tuberculosis σK4 with root mean squared deviations (rmsds) of 1.13 Å and 1.06 Å for 73 and 57 corresponding Cα atoms, respectively. However, no σ factor aligns with both BldN domains simultaneously, indicating that the relative orientation of the two domains in the RsbN–BldN complex is different from those found in other σ factor structures. This is not unexpected as σ2 and σ4 are flexibly attached domains and structures show that they adopt different arrangements relative to each other when bound to anti-σ factors or RNAP (1). In the RsbN–BldN structure there is only one contact between σ2 and σ4, which is between σ2 residue Glu41 and σ4 residue Arg123 (Figures 2D and 3A). Notably, unlike BldN, no structures with any homology to RsbN were detected, using either the domain swapped or non-swapped conformations.

Figure 3.

RsbN–BldN interacting interfaces. (A) Shown is a cartoon of RsbN–BldN (colored as in Figure 2) around RsbN α3. RsbN and BldN residues that interact are shown as sticks and labeled. (B) RsbN–BldN interface involving contacts between RsbN α1–α2 and BldN σ4.

RsbN–BldN interface

The RsbN–BldN interface revealed in the structure is extensive and can be divided into three contact regions. Two of the contact regions are formed by the interaction of RsbN helix 3 (α3) with residues from the BldN σ2 and σ4 domains. The combined interface formed by these interactions buries 2510 Å2 of protein surface from solvent. Indeed, RsbN α3 is fully encased by helices 1–3 from σ2 and helices 5 and 6 from σ4 in these interfaces (Figures 2C–D and 3A). Interactions between RsbN and σ4 are mediated by RsbN residues Val64, Val65 and Ala68, which sit in a hydrophobic pocket composed of residues Leu119 from σ4 helix 5 and Thr134 and Leu138 from σ4 helix 6 (Figures 2D and 3A). In the RsbN–σ2 interface, σ2 contacts the RsbN α3 along its entire length, employing mostly hydrophobic contacts, with a few polar interactions around the edges of the interface. Key hydrophobic contacts are formed between RsbN α3 residues Leu70, Val71, Met74 and Met77 and BldN σ2 residues Val4, Ala12, Phe15, Tyr19, Phe48 and Leu52 (Figure 3A). Polar contacts are mediated between RsbN residues Lys63 and Gln66 and BldN σ2 residues Ser23 and Tyr19, respectively. RsbN residue Arg67 also makes salt bridge interactions with two BldN σ2 acidic residues, Glu11 and Asp20 (Figure 3A).

The third RsbN–BldN interface is formed by interactions between RsbN helices 1 and 2 (α1–α2) and BldN σ4 and buries 1560 Å2. This interface is also primarily hydrophobic and comprised of contacts between RsbN residues that insert into a shallow crevice formed between BldN σ4 helices 5–6 and helix 8 (Figure 3B). A key anchor in this interface is provided by RsbN α2 residue Leu50, which fits into the hydrophobic cavity formed by BldN residues Leu118, Val133, Phe137 and Leu138. σ4 residue Phe137 makes van der Waals contacts with RsbN residues Phe14 and Leu18, which are located on RsbN α1. RsbN Phe14 also stacks between σ4 residues Gln161 and Tyr162 and RsbN residue Ala11 interacts with Tyr162. Finally, σ4 residue Val165 makes hydrophobic interactions with RsbN residues Leu43 and Leu46 (Figure 3B).

Sequence similarity searches indicate that RsbN homologs appear to be found only in Streptomyces. Specifically, each sequenced Streptomyces genome has a single RsbN homolog, always encoded next to the bldN gene. However, even these likely functional RsbN orthologs show a low degree of sequence similarity (Figure 4 and Supplementary Figure S3). Thus, it is notable that the RsbN residues shown to form the three helices that interact with BldN in the RsbN–BldN crystal structure are highly conserved in length and specific RsbN residues that make direct contacts with BldN in the structure are also conserved (Figure 4 and Supplementary Figure S3). In particular Leu70, which makes a central anchoring contact linking RsbN α3 with BldN σ2. Also, RsbN residue Arg67, which makes two salt bridges with BldN acidic residues is completely conserved while RsbN residues located in helices 1, 2 and 3 that make hydrophobic contacts to BldN are highly conserved.

Figure 4.

Sequence alignment of the cytoplasmic anti-σ domains of select Streptomyces RsbN homologs. Helices revealed in the structure are indicated over the sequences. Black asterisks denote residues that are 100% conserved, while red asterisks indicate conserved residues shown to be involved in complex formation in the structure and chosen for mutagenesis experiments. Residues that make hydrophobic, polar and salt bridge interactions with BldN are colored yellow, magenta and cyan, respectively. Residues numbers for Streptomyces venezuelae RsbN are shown above the alignments. Note that the helical regions, while not completely conserved in sequence between homologs, are conserved in length. RsbN homologs were aligned using the program Kalign, provided as a web service by the EBI (https://www.ebi.ac.uk/Tools/msa/kalign).

Regulation of BldN by RsbN

Although there are currently no structures of an RNAP–ECF σ holoenzyme complex, either in the presence or absence of DNA, structures of RNAP with other σ factors bound to promoter DNA have shown that in order for σ2 and σ4 to simultaneously contact the −10 and −35 promoter elements, respectively, the σ factor must adopt an extended structure (1). In this conformation, the σ2 and σ4 domains are separated by more than 80 Å (1,11). In complex with RsbN, BldN assumes a compact conformation with σ2 and σ4 juxtaposed, which would prevent simultaneous contact with −10 and −35 elements. Hence, freezing BldN into a compressed shape is one way in which RsbN inhibits BldN activity. High resolution structures of σ2 and σ4 bound to their cognate −10 and −35 DNA elements have been solved and can be used to analyze the effects of RsbN on DNA binding by BldN (1,6,8). Overlaying BldN σ2 onto E. coli σE2 bound to its −10 element (6) shows that RsbN does not clash with the DNA (Figure 5A). In contrast, superimposition of BldN σ4 onto E. coli σE4 bound to its −35 element shows that binding RsbN would prevent the BldN σ4 domain from binding its −35 element due to a steric clash between RsbN α1 and the DNA (Figure 5B). Transcription initiation by ECF σ factors requires that both σ2 and σ4 bind their cognate −10 and −35 DNA sequences (1,49), and so preventing binding to the −35 element alone would be sufficient to prevent BldN from initiating transcription. Thus, the data indicate that RsbN prevents BldN from functioning as a σ factor through multiple mechanisms. First, RsbN binding locks BldN into a compact shape, preventing it from assuming the conformation needed to bind RNAP and the −10 and −35 promoter elements. Second, RsbN binding directly blocks access of BldN domain σ4 to its −35 promoter element. And third, because RsbN is a transmembrane protein, complex formation with BldN sequesters the σ factor to the membrane.

Figure 5.

BldN inhibition by RsbN. (A) Overlay of BldN σ2 onto that of the Escherichia coli σE2 bound to its cognate −10 sequence (4LUP), indicating that RsbN would not impede BldN σ2 from binding to its −10 promoter element. (B) Superimposition of BldN σ4 onto E. coli σE4 bound to its cognate −35 element (2H27), which shows that RsbN α1 would clash with the DNA, preventing BldN σ4 from binding the −35 promoter element.

RsbN structure defines a new family of anti-σ proteins

Although ECF anti-σ factors display low levels of sequence similarity, the structures of these proteins that have been solved to date reveal two classes of anti-σ factors. The most commonly observed anti-σ factor domain (ASD) is a three-helix bundle fold followed by a short helix, termed the class I ASD (12,18,21–26). This type of ASD can be readily detected by DALI searches, which show that the three-helix bundles of representative class I ECF anti-σ factors RseA, ChrR, RsiW and RskA can be superimposed with rmsds ranging from 2.0 Å to 2.6 Å (Figure 6). The second type of ASD is exemplified by CnrY and NepR, which form a small two helix clamp, called the class II ASD (12,26–28) (Figure 6). Our RsbN–BldN structure indicates that RsbN differs from both these types of ASD, revealing it to be the founding member of a new structural class of ECF anti-σ, herein termed the class III ASD (Figure 6). Sequence similarity searches suggest that RsbN anti-σ factors are widespread in Streptomyces but appear to be limited to these bacteria and as noted, these alignments show that regions corresponding to the three helices in RsbN are conserved while the linker regions are widely variable (Figure 4). The residues that contact BldN are also conserved (Figure 4 and Supplementary Figure S3). These findings suggest that all RsbN homologs are likely to adopt the same fold and interact similarly with their partner BldN proteins.

Figure 6.

RsbN structure reveals a new type of anti-σ factor fold (ASD). Previously characterized anti-σ ASD folds include the class I ASDs as exemplified by the anti-σ factors RsiW, RseA, ChrR and RskA. These ASDs can be superimposed with rsmds of 2.0–2.6 Å. This ASD is composed of a three-helix bundle structure followed by a forth helix. Right, shows cartoon representations for these anti-σ factors (in the same orientation as the ASDs on the left) with their bound σ factor. The class II ASDs include CnrY, which consists of a two-helix clamp that surrounds its σ factor (right). RsbN shows no structural homology to the class I or class II ASDs and thus represents a new ASD class, herein termed class III. Shown to the right is its complex with BldN.

Probing the RsbN–BldN structural model

The RsbN–BldN structure revealed a novel σ–anti-σ complex. To test the structural model, we co-expressed truncated versions of BldN and RsbN to examine interactions between individual domains. First, we co-expressed the N-terminally his-tagged version of the RsbN anti-σ domain (RsbN91) with just σ2 or σ4, and found by cobalt affinity purification that RsbN91 can form a stable complex with σ2 or σ4 in isolation (Figure 7A, lanes 1 and 2). Control experiments showed that non-his-tagged versions of BldN σ2 and σ4 did not interact with the CoCl2 charged Hi-Trap columns (Supplementary Figure S4, lanes 3 and 4). In the structure of the RsbN–BldN complex, RsbN α3 is sandwiched between σ2 and σ4 (Figure 2C and D). To assess its importance, we co-expressed an N-terminally his-tagged version of RsbN α3 alone and found that, by itself, RsbN α3 is sufficient to form a stable complex with BldN (Figure 7A, lane 3). We took this analysis one step further by co-expressing his-tagged RsbN α3 with either σ2 or σ4 in isolation and found that RsbN α3 forms a stable complex with σ2 (Figure 7A, lane 4) but not with σ4 (Supplementary Figure S4, lane 1), showing that the RsbN α3 interaction with σ4 is not enough to form a stable complex without additional contacts from σ2. In addition, we found that RsbN α1–α2 alone could not support complex formation with σ4 (Supplementary Figure S4, lane 2). These studies suggest that the combined interaction interfaces are required for tight complex formation between RsbN and BldN.

Figure 7.

Mutagenesis and biochemical test of the RsbN–BldN structure. (A) Complex formation between his-tagged RsbN91 and σ2 (lane 1), his-tagged RsbN91 and σ4 (lane 2), his-tagged RsbN α3 and BldN (lane 3), and his-tagged RsbN α3 and σ2 (lane 4). Proteins were co-expressed in Escherichia coli, purified over a cobalt affinity column, run on a 16% polyacrylamide tricine SDS gel (47) and stained with Coomassie Brilliant Blue. The identities of all proteins were confirmed by MALDI-MS. Note that his-tagged RsbN α3 is not visible in lanes 3 and 4. (B) Complex formation occurs between his-tagged RsbN91 and σ2 (lane 1) but not between the RsbN91(L70R) variant and σ2 (lane 2). Complex formation also occurs between his-tagged RsbN91 and σ4 (lane 3) but not between the RsbN91(F14R-L50R) variant and σ4 (lane 4). Proteins were co-expressed in E. coli, purified using HIS-select spin columns (Sigma H7787), run on a 16% polyacrylamide tricine SDS gel (47) and stained with Coomassie Brilliant Blue. In both A and B the eluted fractions (E) are shown next to the corresponding clarified supernatants (S/N), which thus represents the soluble material obtained after centrifugation. A clarified supernatant from a strain that does not express either RsbN or BldN is also shown (−), next to the protein ladder (L).

The structural data, combined with our biochemical experiments, suggest that three contact points in the RsbN–BldN structure likely cooperate to form a stable complex. Moreover, they suggest that RsbN α3 forms the lynchpin of the complex between the BldN-interacting region of RsbN and BldN, with α1–α2 maximizing binding via interaction with σ4. To probe the structural model in more detail, we tested the effect of mutating residues that the structure suggests would be central to complex formation. Our data showed that BldN σ4 alone cannot form a stable complex with RsbN α3 (Supplementary Figure S4, lane 1). However, the structure suggests that BldN σ2 and σ4 function together to form a tight complex with RsbN α3. If that is the case, we hypothesized that destabilization of the RsbN α3 interaction with σ4 may disrupt the more stable RsbN α3–σ2 interaction and prevent the overall RsbN α3-BldN interaction. To test this prediction, we generated an MBP–RsbN α3 fusion protein and co-expressed it with WT BldN. As expected, the construct encoding WT MBP–RsbN α3 formed a tight interaction with BldN and eluted from SEC at a MW of 67 kDa, consistent with a 1:1 WT MBP–RsbN α3–BldN complex (the expected MW is 69 kDa). However, when we mutated RsbN α3 residues Val64 and Val65 (which the structure indicates are important in mediating the α3–σ4 interaction) to lysine, the interaction between the MBP–RsbN α3(V64K–V65K) protein and BldN was clearly disrupted, resulting in two peaks; one containing BldN (at a MW of 26 kDa compared to the expected MW of 21 kDa) and one consisting of MBP–RsbN α3(V64K–V65K) at a MW of 47 kDa (compared to the expected MW of the MBP–RsbN α3 protein of 48 kDa) (Supplementary Figure S5). While there was less soluble BldN when expressed with MBP–RsbN α3(V64K–V65K), the soluble BldN present was sufficient to assess complex formation with the mutant MBP–RsbN helix 3 protein. The fact that the mutations in RsbN helix 3 prevented tight complex formation with BldN suggests that the interaction observed between RsbN α3 and BldN σ4 observed in the structure is also found in solution.

We next probed the RsbN α3–BldN σ2 and RsbN α1–α2–BldN σ4 interfaces. The structure suggests that RsbN residue Leu70 plays a central anchoring role in the interaction of RsbN α3 with BldN σ2 (Figure 3A). Leu70 is completely conserved in RsbN homologs and in the structure it inserts into a hydrophobic pocket of BldN σ2. We thus replaced Leu70 in RsbN91 with a large, charged arginine side chain and assessed its ability to form a complex with BldN σ2. Strikingly, whereas WT RsbN91 bound BldN σ2, the mutant RsbN91(L70R) showed no interaction with BldN σ2, supporting our structural information (Figure 7B, lanes 1 and 2). To assess the RsbN α1–α2 interaction with BldN σ4, we substituted RsbN91 residues Phe14 and Leu50, which are highly conserved and in the structure serve as key points of contact with a surface-exposed hydrophobic pocket in σ4. Phe14 and Leu50 were changed to arginine residues and the interaction of the resulting RsbN91 proteins with BldN σ4 was analyzed. Unlike WT RsbN91, the mutant protein failed to form a complex with BldN σ4 (Figure 7B, lanes 3 and 4). BldN σ2 and σ4 regions were found in the clarified lysates when expressed with these RsbN mutants, indicating that the lack of binding with the mutant proteins was not due to their insolubility. Thus, our mutagenesis studies provide strong support for all three interfaces observed in the crystal structure.

DISCUSSION

Streptomyces are a biomedically important group of bacteria and serve as a major source of antibiotics and other clinically important compounds. The generation of these ‘secondary metabolites’ coincides with the onset of differentiation. Here we analyzed a protein complex that plays a central role in controlling this developmental process, the BldN–RsbN complex. BldN is an ECF σ factor whose activity is controlled by an interaction with its cognate anti-σ factor, RsbN, a single-pass membrane protein that has an N-terminal cytoplasmic BldN-binding domain and a large C-terminal extracytoplasmic domain of unknown function (Figure 1) (31). Interestingly, RsbN shows no sequence similarity to any known anti-σ factor or any known protein. To determine the mechanism by which RsbN regulates BldN, here we elucidated the structure of BldN bound to the anti-σ domain of RsbN to 2.23 Å resolution. The RsbN–BldN structure shows that the RsbN anti-σ factor domain contains a fold distinct from other structurally studied anti-σ factors. The structure shows that this RsbN anti-σ domain interacts with BldN at three distinct interfaces; RsbN α3 contacts both the σ2 and σ4 domains of BldN to form two of the interfaces, while RsbN α1–α2 interact with the BldN σ4 domain.

Our mutagenesis studies support that the three interfaces revealed in the RsbN–BldN crystal structure are key for complex formation. These interfaces all involve primarily hydrophobic contacts between BldN and RsbN. Interestingly, a detailed examination of the ECF σ–anti-σ structures solved to date reveals that, while these complexes are distinct in their overall structures, the anti-σ factors all make hydrophobic interactions with the same three σ contact points seen in the RsbN–BldN structure. For example, in all these structures a helical region of the anti-σ factor interacts with a hydrophobic region on the σ2 domain between helices 2 and 3 (using BldN helix numbering) (Figure 8). Interestingly, though all these interactions involve an anti-σ helical region, how the helix interacts in these complexes is variable; α3 from RsbN and ChrR dock with σ2 in an N-terminal to C-terminal orientation, RseA binds σ2 via a C-terminal to N-terminal docking mode and both RskA and RsiW use their α4 helices (not α3) to interact with σ2. The other two hydrophobic contact points between the σ and anti-σ factors involve σ4 regions α5–α6–α8 and α5–α6. In the RsbN–BldN structure the hydrophobic patch in the BldN α5–α6–α8 region forms a complementary interface with RsbN α1–α2. Similarly, in the other ECF σ–anti-σ structures the corresponding region in σ4 is shielded from solvent by helices of the corresponding anti-σ factor. The specific types of interactions in the structures differ but all shield this hydrophobic region of their σ (Figure 8). Finally, the exposed hydrophobic pocket between α5–α6 of BldN σ4, which is complexed with RsbN α3 in the RsbN–BldN structure, is also protected from solvent in the other σ–anti-σ structures but by both the anti-σ factor and regions of σ2 (Figure 8). Thus, these analyses suggest that while σ–anti-σ complexes differ in detail, they appear to share hydrophobic contact points.

Figure 8.

Comparison of σ–anti-σ interfaces. Top row shows comparison of RsbN–BldN with other σ–anti-σ structures with the σ2 domains shown in the same orientation. Boxed is the hydrophobic region of contact between RsbN (magenta) and BldN (cyan). The same region is boxed in the other σ–anti-σ structures, revealing that all share a σ–anti-σ contact interface involving the burial of the α2–α3 hydrophobic patch on σ2. Bottom row shows the σ–anti-σ structures with their σ4 domains in the same orientation. The interface region formed between RsbN α3 and RsbN helices 5–6 is boxed as are the corresponding regions in the other σ–anti-σ complexes. In the latter structures, the hydrophobic patch in the σ4 is similar to the RsbN–BldN structure, covered by anti-σ factor regions and in some cases additional contacts are provided to this σ4 region by σ2. An arrow indicates the contact region between BldN helices α5–α6–α8 with α1–α2 of RsbN. Arrows point to the same region of the σ–anti-σ structures to the right of the RsbN–BldN structure showing that these anti-σ factors also contact the homologous region in their partner σ factors. These comparisons thus underscore a general conservation of hydrophobic contact points shared by the σ–anti-σ complexes despite the differences in the structural details of the interactions.

One result of the multiple σ–anti-σ contacts is that the complexes all assume an overall compact shape. As binding to RNAP requires the σ factor to adopt an extended conformation, this is one facet of the inhibitory role played by anti-σ factors. Anti-σ factor binding in these complexes also directly inhibits binding of the σ factor to RNAP and/or promoter DNA. Therefore, the anti-σ factor must be inactivated to allow σ factor release and transcription of its target genes. The molecular mechanism underlying the release of BldN from RsbN has not been investigated, but by analogy with E. coli RseA–σE (50), Bacillus RsiW–σW (51) and Bacillus RsiV–σV (52), it seems likely that the inactivation of RsbN will involve a process known as regulated intramembrane proteolysis (18,53,54). The signal that triggers RsbN inactivation is also unclear, but given the bulk of RsbN (residues 140–412) lies on the outside of the membrane, that signal will probably be extracytoplasmic. An attractive but speculative possibility is that this signal is the assembly of the hydrophobic sheath itself, which would create a positive feedback loop to reinforce sheath assembly once it has started.

DATA AVAILABILITY

Coordinates and structure factor amplitudes for the RsbN–BldN complex have been deposited with the Protein Data Bank under the Accession code 6C03.

Click here for additional data file.