Diverse and Unified Mechanisms of Transcription Initiation in Bacteria

Abstract

Transcription of DNA is a fundamental process in all cellular organisms. The enzyme responsible for transcription, RNA polymerase (RNAP), is conserved in general architecture and catalytic function across the three domains of life. Diverse mechanisms are used between and within the different branches to regulate transcription initiation. Mechanistic studies of transcription initiation in prokaryotes are especially amenable because the promoter recognition and melting steps are much less complicated than in eukaryotes or archaea. Also, bacteria play critical roles in human health as pathogens and commensals, and the bacterial RNAP is a proven target for antibiotics. Recent biophysical studies of RNAPs and their inhibition, as well as transcription initiation and transcription factors, have illuminated the mechanisms of transcription initiation in phylogenetically diverse bacteria, inspiring this review to examine unifying and diverse themes in this process.

Introduction

Transcription, the act of transcribing DNA into RNA, is one of the key processes in the central dogma of molecular biology and is thus executed in all cellular organisms¹. The enzyme responsible for all cellular transcription, the DNA-dependent RNA polymerase (RNAP), is composed of multiple subunits². The subunit compositions of RNAPs from the three domains of life, Eukarya, Archaea, and Bacteria, vary substantially; yet a core set of subunits share sequence similarity and are conserved in general architecture³. This concordance in structure reflects the universal functions that the enzymes perform. In general, cellular multi-subunit RNAPs locate promoters, sequences that direct the enzyme to the beginning of genes⁴. After promoter binding, the enzyme melts (opens, unwinds) the DNA, catalyzes templated de novo polymerization of ribonucleotides (rNTPs), and then transitions from an initiation complex to an elongation complex until transcription of the gene is completed and the process terminated (Figure 1A). These minimum steps, which are subject to complex regulation, are performed by all cellular DNA-dependent RNAPs.

Figure 1. Overview of the steps of bacterial transcription initiation and RNAP.

(A) Schematic of bacterial transcription initiation. The cartoon shows steps of bacterial transcription initiation, starting from core RNAP to the elongation complex. The core RNAP (colored in grey with the active site Mg²⁺ depicted as a green dot) binds a promoter specificity factor, σ (orange), generating the holoenzyme. Holoenzyme recognizes and melts promoter DNA (blue) to form the open promoter complex. In the presence of rNTP, the initiation complex scrunches and can synthesize many RNA transcripts (red) in a process called abortive transcription. Eventually, the complex produces a transcript length that competes with σ, which in combination with scrunching, causes σ to disassociate from the core RNAP allowing it to escape the promoter. The enzyme then transitions to the elongation complex and completes transcription of the gene. (B) Architecture the active site Mg²⁺, the main channel, the secondary channel, the bridge helix (in cartoon tubes), the β lobe, the β protrusion, and the β’ clamp highlighted. (C) Holoenzyme configuration and recognition of housekeeping promoter. (Left) Holoenzyme is shown as a transparent molecular surface and colored according to the color key with σ, the β flap-tip helix, and the clamp helices shown as cartoon tubes. The regions of holoenzyme that bind the promoter elements are highlighted: αCTDs [pale green] bind the UP element [teal], σD4 binds −35 element [yellow], σD3.2 is buried in the RNAP active site cleft, σD3 interacts with the extended −10 [EXT, blue], σD2 interacts with the −10 [magenta], σNCR is located between σD2 and σD1.2, σD1.2 interacts with the discriminator region [DSR, forest green], and the β subunit of RNAP recognizes the core recognition element [CRE, light blue]. The TSS is shown as a ‘+1’, and the direction of transcription is depicted by the black arrow below the TSS. (Right) Structure of open promoter complex primed for initiation showing the promoter DNA placed into RNAP with an RNA primer.

Although the general transcription cycle and the overall structures of RNAPs are shared, details of initiation, elongation, and termination vary between and within domains of life². This is exemplified by the evolutionary diversity of initiation factors^3,5,6. In bacteria, a primary step for the regulation of gene expression is transcription initiation. In this review, we will first delineate the universal mechanistic framework of transcription initiation in bacteria, including recent structural studies that have shed light on the intricate RNAP motions involved in promoter melting. We will then present newly described specialized mechanisms that regulate initiation, which resulted from recent research that has examined transcriptional systems beyond the historical model bacteria, Escherichia coli and Bacillus subtilis. Research of evolutionary diverse clades, specifically α-Proteobacteria, λ-Proteobacteria, and Actinobacteria, has provided insights into similarities and differences of both the basal and regulated steps of initiation. These recent discoveries have made it abundantly clear that bacterial transcription mechanisms are diverse, causing differences in susceptibility to antibiotics that inhibit transcription. We will review those antimicrobials that have recently been evaluated across clades of bacteria. We hope to impart how studying RNAPs, and their regulators from evolutionary divergent bacteria, inform on universal pathways and yet reveal a vast range of regulatory mechanisms. This review also stresses the essentiality of research on transcriptional systems from pathogens for antibiotic development and optimization.

Universal themes in bacterial transcription

Architecture of RNAP, σ, and promoters

In bacteria, a single RNAP performs all transcription. Bacterial RNAPs consist of an evolutionarily conserved catalytic core with polypeptide subunit composition α2ββ’ω⁷ (Figure 1B). The core enzyme has several modules that are conserved in all cellular RNAPs, and we highlight some of the salient features in Figure 1B. The crystal structures of the bacterial, yeast and archaeal core RNAPs identified features characteristic of all cellular RNAPS^8–10. These include the two mobile pincers that make up the active site cleft for DNA loading. The β’ clamp creates one pincer while the β-lobe and the β-protrusion create the second. The active site is marked by the required catalytic Mg²⁺ and the bridge helix that is important for the addition of nucleotides. The bridge helix divides the active site cleft creating a secondary smaller channel where incoming rNTPS are thought to enter⁸.

Although the catalytic core is capable of RNA polymerization, bacterial RNAPs require an initiation factor, sigma (σ), for promoter-specific DNA binding and unwinding¹¹. All bacteria possess a primary housekeeping σ factor that controls the transcription of essential genes during normal growth conditions. The vast majority of transcription initiation events in bacteria involve RNAP bound to the primary σ^12,13. The domain architecture of primary σ factors consists of three highly conserved domains [σ domain 2 (σD2), σ domain 3 (σD3), σ domain 4 (σD4)], a non-conserved region (σNCR), and a poorly conserved N-terminal domain (σD1.1) that are connected by flexible linkers¹⁴. Binding of σ to core RNAP forms the RNAP holoenzyme (via σD4 interaction with the β flap-tip helix and σD2 interaction with the β’ clamp helices) in which the domains of σ are extended and spatially re-organized to engage with DNA^15–17 (illustrated in Figure 1C, left). This review will focus on transcription initiation by the σ⁷⁰ family of σ factors and will not discuss initiation by σ⁵⁴ family, which was elegantly and recently reviewed¹⁸. The σ⁵⁴ family requires additional activator proteins and initiates transcription using unique mechanisms beyond the scope of the review. σ factors guide the core RNAP to DNA promoters by providing specific contacts with the DNA^14,16. Bacterial promoters that are recognized by the primary σ are generally comprised of two sequence motifs located upstream of the transcription start site: the −35 DNA element (consensus sequence, TTGACA) and the −10 DNA element (consensus sequence, TATAAT)^19,20. Certain promoters also contain additional DNA elements such as the upstream (UP) element²¹, extended −10 element²², the discriminator²³, and/or a core recognition element²⁴.

During promoter binding, the UP-element (located 40–60 bases upstream of the TSS) is recognized by the C-terminal domains of the RNAP α subunits²¹. The −35 element is bound to σD4¹⁴, and unwinding occurs at the start of the −10 element, which interacts with σD2²⁵. If present, the extended −10 element (TG at −14 and −13) is recognized by σD3. The open complex is stabilized by interactions between the discriminator region (between −6 and −3), which contacts σ domain 1.2 (σD1.2)²³. Further stabilizing contacts include the core-recognition element (located at +2), which binds in a pocket of the β subunit²⁴. The combination of these DNA elements and their interactions with the holoenzyme modulates the strength of a promoter by contributing to the initial RNAP binding and subsequent steps leading to the formation and stability of the open promoter complex. Figure 1C (right) highlights key features of promoters and how they interact with the RNAP holoenzyme in the final open complex. Most of these promoter elements and their interactions have been confirmed structurally and functionally for several clades of bacteria, including Thermus species, γ-proteobacteria, and Actinobacteria^26–28. However, genomics studies suggest that the prevalence of each element in different clades may differ considerably, such as the heavier reliance on extended −10 motifs versus −35 elements in gram-positive bacteria^28–30.

Promoter melting

RNAP resembles a crab claw that opens and closes during transcription^31,32. To efficiently initiate promoter melting, RNAP closes transiently to nucleate melting of the −10 element, then opens for DNA loading in the channel³³ (Figure 2). After melting is completed, RNAP has to close to secure the DNA and ensure processive transcription.

Figure 2. Structure-based model of RPo formation highlighting RNAP motions that lead to promoter melting.

Structural models of RNAP were generated from cryo-EM structures of promoter melting intermediates from E. coli (PDB IDs 6PSQ, 6PSS, 6PST, 6OUL)⁴⁷ and M. tuberculosis (PDB ID 6EE8)⁴⁵ with lineage-specific inserts removed for clarity. Proteins are shown as transparent molecular surface, while σD1.1 and promoter DNA are shown as spheres. The Mg²⁺ (shown as a green sphere) marks the active site. The pathway is composed of structures delineating bubble propagation that leads to the open promoter complex.

Until recently, the details on how RNAP unwinds promoter DNA while forming the transcriptionally competent open complex were mostly unknown. Biochemical studies laid the foundation for our understanding of promoter melting and included the following observations: 1) Bacterial holoenzymes use binding free energy to isomerize into the fully melted, transcriptionally-component, open promoter complex (open complex)^34,35. 2) This process is multi-step and characterized by several RNAP-promoter intermediates that have been identified in multiple kinetic studies in E. coli^35–38, and more recently, in M. tuberculosis^39,40. 3) The promoter sequence plays a critical role in the rates of each step leading to the open complex, resulting in differential kinetics between promoters that lead to varying transcriptional output^34,41.

Transcriptional initiation studies in Mycobacteria, Thermus species, and Bacillus subtilis suggest that the kinetics of initiation and the nature of populated stable intermediates differ among the RNAPs from these bacteria^{39,40,42–44}. For example, studies show that on the same promoter, complexes of E. coli RNAP are considerably more stable than M. tuberculosis RNAP^39,40,44. Although the ratios of specific populated intermediates are expected to differ, the RNAP motions and path of DNA melting are expected to conform, given the conservation of the DNA cleft, the consensus architecture of RNAPs, and the structural uniformity of open complexes in different clades of bacteria^26–28,45.

Recent advances in single-particle cryo-electron microscopy (cryo-EM) have led to high-resolution visualization of dynamic macromolecular complexes without the constraints of crystallization, which preferentially captures static conformations. RNAP is a remarkably flexible enzyme, with the pincers moving through a range of 15 Å during the transcription cycle^31,45,46. Cryo-EM allows for classification and extraction of poorly populated structural states not previously amenable to crystallization and has recently been applied to capture a series of stepwise promoter melting intermediates. These approaches include obtaining structures of RNAPs from different organisms and RNAP complexes with promoters that have slow unwinding kinetics or form reversible open complexes. Additional approaches include employing inhibitors and transcription factors that regulate the formation of the open complex to trap melting intermediates.

These structures reveal the clamp dynamics in RNAPs, domain/lineage-specific insertion movements during promoter melting, and the path of DNA during unwinding^45–47. Figure 2 incorporates two structural studies to provide a composite view of RNAP motions that lead to promoter melting. Early promoter melting intermediates of E. coli RNAP were captured with a reversible ribosomal protein promoter using the transcription factor TraR, which increases the population of the intermediates at equilibrium⁴⁷. These structures span the initial recognition of the duplex promoter in a “closed” complex (RPc) to the final open complex. In these structures, the duplex promoter DNA is initially bound as a “closed” complex; then the −10 element is nucleated by β’ clamp closure; followed by bubble propagation leading to an early partially melted intermediate (four to five bases unwound) while σD1.1 still occupies the downstream RNAP channel. A late melting intermediate with a partially melted (eight out of thirteen bases) promoter was observed using the M. tuberculosis RNAP, which forms an unstable and reversible open complex on its ribosomal RNA operon promoter⁴⁵. This intermediate was also observed kinetically on the same promoter³⁹. This structure indicated that unwinding of the downstream part of the transcription bubble (the unwound section of the promoter) occurs within the central RNAP cleft after σD1.1 has been ejected from the downstream RNAP channel. The same study also showed that the antibiotic corallopyronin freezes the enzyme in this intermediate conformation by inhibiting the clamp movements. This suggests that clamp opening is required for late but not early melting.

Post-promoter melting

Once the promoter is fully melted, the DNA template (T-DNA) strand is loaded into the RNAP active site, positioning the transcription start site for templating RNA catalysis^26,27. The enzyme then goes through a series of steps, illustrated in Figure 1A. Briefly, the transcriptionally-competent open complex begins de novo RNA synthesis by catalyzing phosphodiester bonds between incoming rNTPs to form the initial transcribing complex. Some complexes, depending on the promoter, undergo a non-productive cycle, called abortive transcription, where the enzyme cycles between the initial transcribing complex and the open complex, releasing short RNAs^48–50. The “decision” to enter or leave the non-productive cycle to advance into productive transcription depends on the promoter sequence, interactions with the σ factor, and the initial transcribing sequence^51,52. The increased length of the RNA–DNA hybrid induces promoter escape by causing scrunching-distortions in the transcription bubble generated by the transcribing RNAP which remains static while unwinding and pulling the downstream DNA into the active site cleft^53–55; and clashes with σ domain 3.2 (σD3.2)^16,56–60.

Recent biochemical data have added to the nuances of promoter escape, including how pausing after synthesis of a 6-mer RNA affects the branching of productive versus non-productive transcription during initiation⁶¹. Promoter escape is not well characterized in other bacterial systems. However, recent work comparing RNAPs from E. coli and M. tuberculosis have shown that M. tuberculosis RNAP escapes more readily than E. coli RNAP on the same promoter. This study showed that two essential, housekeeping transcription factors not found in E. coli, CarD and RbpA, slow escape kinetics, suggesting promoter escape is a prime step of regulation in this pathogen⁶².

In the productive transcription step, the RNAP “escapes” into the elongating stage, and the σ factor can dissociate from the holoenzyme and bind another core RNAP. The current view is that σ release is not obligatory but, rather, stochastic⁶³. Both σD3.2 and σD4 must be displaced from their original interactions with RNAP as they are in the path of the elongating RNA. However, σ can be retained throughout much of the elongation cycle^64–67, most likely via interaction of σD2 and the clamp helices. This retention has been suggested to serve as a mechanism of regulation for elongation and pausing^63,68,69. Most of the σ cycle paradigm was formulated from studies in E. coli, and it remains to be seen whether these principles apply to other bacteria. We note that the features of σ and RNAP that mediate this cycle are conserved in other bacteria. Indeed, a recent structural study with Thermus thermophilus and M. tuberculosis holoenzymes describes a similar physical mechanism of σ displacement as described for E. coli⁶⁰. Figure 1A encapsulates the pathway from σ association to core RNAP to the elongating step.

Themes in redistributing RNAP holoenzyme populations

Alternative sigma factors

In addition to the primary σ, which is essential for viability during normal growth conditions, almost all bacteria possess additional σs known as alternative σ factors. These factors recognize DNA sequence motifs distinct from the primary housekeeping σ factor. Alternative σ factors allow for specific regulon expression in response to stress, cell density, developmental transitions, and nutritional cues^70–73. The collection of alternative σ factors can vary in number from zero to hundreds between different species. The alternative σs are classified into four general groups based on the presence of ancillary regions. The Group IV σs, also called Extra-Cytoplasmic Function σ factors (ECF σs), contain only the minimal domains, σD2 and σD4, that are sufficient to bind to core RNAP and promoter elements. This group, that represents the most abundant and diverse σs, has been extensively characterized functionally and phylogenetically and is referred to as the third pillar of signal transduction in bacteria (the other two being one- and two-component systems)⁷⁴. ECF σs are involved in a range of functions, including response to oxidative stress, metal homeostasis, virulence, periplasmic/envelope stress, and sporulation.

Thus, the availability of specific holoenzymes can drastically reprogram gene expression providing a powerful way to express discrete regulons under specific conditions. Here we discuss classic and recently described modes of regulating specific holoenzyme populations as a means for fine-tuning gene expression. We review three examples of how bacteria use different means to prevent, enhance, or redistribute specific holoenzyme populations.

What’s new with anti-σ’s?

Anti-σ factors associate with their cognate σ factor and function by preventing the σ factor from interacting with core RNAP, thus preventing assembly of the holoenzyme⁷⁵. The genes encoding the anti-σs are usually found directly downstream of the σ factor they regulate. More than 30 % of the ECF σ factors are regulated by anti-σs^76–78. Multiple structural studies of different groups of σ factors with their cognate anti-σs have revealed the molecular basis of how anti-σs inhibit holoenzyme formation^77,79–87. The structures reveal that anti-σs bind at least one RNAP binding surface of their cognate σ and physically block association with the core RNAP. In addition, the anti-σ often contorts the σ factor in variable configurations such that additional core-binding interfaces are inaccessible.

Although this general inhibitory mechanism is conserved, these factors fall into diverse classes^73,78,88. Many are multi-domain proteins with an N-terminal anti-σ domain fused to one of a wide range of regulatory domains, which is often subject to signal transduction or proteolysis. These regulatory domains often contain transmembrane regions (the majority), metal-binding motifs, short disordered tails, or serine/threonine kinase motifs^76,77. Three anti-σ domain classes have been identified for the ECF σs^76,77,82. Even so, the anti-σ domains are not well conserved neither in sequence nor structure and thus interact with their cognate σ in very different ways (Figure 3A).

Figure 3. Factors that regulate the pool of holoenzymes.

(A) Despite divergent sequence, structure, and phylogenetic origins, anti-σs (red) prevent binding of the alternative σ by occluding the major core binding interface (σD2, green) from the core β’-clamp helices (pink). The β’-clamp helices were modeled by superimposing σD2 from structures of the σ/anti-σ complexes unto the σ^A holoenzyme structure (PDB ID 19LU)¹⁶. Structures shown and numbered are listed with PDB IDs in parentheses as follows: (1) Rhodobacter sphaeroides σ^E/ChrR (PDB ID 2Q1Z)⁷⁶, (2) E. coli σ^E/RseA (PDB ID 1OR7)⁸⁰, (3) P. aeruginosa σ^H/MucA (PDB ID 6IN7)¹⁰⁹, (4) B. subtilis σ^W/RsiW (PDB ID 5WUR)⁸⁵, (5) M. tuberculosis σ^K/RskA (PDB ID 4NQW)⁸⁴, (6) B. Quintana σ^E/NepR (PDB ID 5UXX), (7) Caulobacter vibroides σ mimic PhyR/NepR (PDB ID 3T0Y)⁸⁶, (8) Cupriaviuds metalliiduruns σ^CnrH/CnrY (PDB IDm4CXF)⁷⁷, (9) Sf. venezuelae σ^BldN/RsbN (PDB ID 6DXO)⁸², (10) S. venezuelae σ^WhiG/RsiG (PDB ID 6PFJ)⁸³ and (11) Aquifex aeolicus σ^FliA/FlgM (PDB ID 1SC5)⁸¹. (B) Schematics of the three founding members of σ-tethers (purple), tethering σD2 and σNCR to the core enzyme and or promoter DNA. (C) 6S RNA mimics the promoter architecture of an open complex, sequestering the housekeeping σ holoenzyme. Shown are the cryo-EM structures of 6S/holoenzyme (PDB ID 5VT0)¹⁰⁸ and the crystal structure σ⁷⁰ holoenzyme with promoter DNA (PDB ID 4YLN)²⁷.

However, one common theme emerges when aligning σD2- the most conserved domain of σ factors- from the published structures of σ/anti-σ complexes from distantly related bacteria. Each of the anti-σs employs a single helix to bind to the same surface of their cognate σD2, which would physically prevent interaction with the RNAP β’ clamp helices (Figure 3A). This finding suggests that this interface of σ, although conserved in function and essential for σ factors to interact with core RNAP, is variable enough to be targeted by specific anti-σs. The interaction between σD2 and the β’ clamp helices RNAP is the primary determinant for holoenzyme assembly and stability and would thus serve as an interface for drug targeting. More recently, a new type of anti-σ (anti-σ^RsiG), which requires cyclic di-GMP to bind to the σ factor (σ^WhiG), similarly occludes σD2 binding to the clamp helices⁸³. Our analysis of the catalogue of σ/anti-σ structures reveals that although the sequence, structure, and function of the anti-σs varies widely across bacteria, a common feature of inhibition appears to have been selected. We note that two exceptions to this theme are the phage anti-σ AsiA and the bacterial anti-σ Rsd, both of which inhibit the housekeeping σ in E. coli by binding σD4^88–90.

σ tethers - a new class of σ regulators

In contrast to anti-σs that prevent the binding of σ to core RNAP, recent studies have revealed a new group of factors that promotes the association of σ to core RNAP and/or DNA. These factors have been characterized in diverse species of bacteria, and we introduce the term σ-tethers to describe the mechanism of this emerging group of RNAP regulators (Figure 3B). The first example of a σ-tether is RbpA, an essential housekeeping σ-binding factor that activates transcription in Actinobacteria^91–93. Structural studies of RbpA with the M. tuberculosis holoenzyme and promoter DNA^39,45 illustrated the following: 1) RbpA interacts with the non-conserved region and σD2 of the housekeeping σ via RbpA’s σ-interacting-domain. 2) RbpA also makes contacts with the core RNAP via a second domain called the core binding domain. 3) The linker between these two domains of RbpA contacts the phosphate backbone of promoter DNA and the N-terminal tail contacts the template DNA in the active site. This tethered configuration of RbpA is consistent with its described roles in stabilizing the interactions between σ and core^91,94 and holoenzyme and DNA⁹⁵ (Figure 3B, left).

The second example of a σ-tether is GrcA (conserved in α-proteobacteria) that has been characterized in Caulobacter crescentus. GcrA is a transcription factor that binds to the housekeeping σ factor and actives methylated promoters⁹⁶, a function critical for coupling DNA replication with cell division during cell cycle progression⁹⁷. Two X-ray crystal structures - one of GcrA’s σ-interacting-domain with σD2 and the non-conserved insert of σD2, and the other of GrcA’s DNA binding domain on methylated DNA - revealed that this protein, like RbpA, also functions as a tether between σ2 and promoter DNA⁹⁸ (Figure 3B, middle).

The third σ-tether example, implicated in facilitating holoenzyme assembly in E. coli, is the transcription regulator Crl that binds to the stationary phase σ factor, σ^S. Crl is critical for σ^S to compete with the housekeeping σ factor during the transition from exponential to stationary phase^99,100. Crl activates stationary phase genes in E. coli and plays a role in virulence and infection in many pathogenic γ-proteobacteria like Vibrio cholera, Salmonella typimurium, and Yersinia pestis. The cryo-EM structure of Crl with a σ^S holoenzyme revealed that Crl also tethers σ^S to RNAP via the clamp toe, a domain on the β’ subunit, and there is also evidence that Crl transiently contacts DNA during promoter unwinding^101,102 (Figure 3B, right).

6S RNA - a promoter mimic that modulates the pool of holoenzyme

Bacterial 6S RNA is a noncoding RNA that binds and regulates E. coli RNAP activity¹⁰³. Although 6S RNAs are widely distributed in bacteria^104,105, they have been best characterized in E. coli. E. coli 6S RNA utilizes a unique mechanism in controlling specific holoenzymes availability and thus regulating gene expression (reviewed in¹⁰⁶). The 6S RNA binds to the primary σ, σ⁷⁰, containing holoenzyme, but not holoenzymes containing alternative σ factors; and inhibits the ability of the primary holoenzyme to bind promoter DNA. This sequestration of σ⁷⁰ holoenzyme plays a role in the transcriptional reprogramming during the transitions between exponential and stationary growth phases. The preferential binding of 6S RNA to σ⁷⁰ holoenzyme allows for the increased activity of the stationary phase σ factor, σ^S. Although not conserved in sequence, the secondary structure of 6S RNAs contains a prominent central bubble that is reminiscent of the promoter DNA bubble in the transcription open promoter complex^104,105 and serves as a transcription template during outgrowth from stationary phase¹⁰⁷. Through this mechanism, 6S RNA plays an important role in long-term survival during and escape from stationary phase. Some bacteria have two copies of 6S, and although the roles are less characterized, expression profiles suggest that 6S RNAs may be necessary for virulence and other cellular functions in different bacteria¹⁰⁶.

A cryo-EM structure supported by biochemical assays of E. coli σ⁷⁰ holoenzyme with 6S RNA showed that two amino acids largely determine 6S RNA’s specificity for Eσ⁷⁰ over Eσ^S. The structure also revealed that although 6S RNA maintained A-form C3’-endo sugar puckers, it adopts a widened major groove, giving the RNA an unusual overall architecture that mimics B-form promoter DNA¹⁰⁸. This surprising observation explained how a noncoding RNA serves as a B-form DNA mimetic to regulate transcription by a DNA-dependent RNAP (Figure 3C).

Lineage-specific features and regulation of transcription

Lineage-specific domains in core RNAP

The catalytic core RNAP is conserved in sequence, structure, and fundamental function in all cellular organisms^110,111. In prokaryotes, the β and β’ subunits are highly conserved. However, these shared regions are often separated by spacers of non-conserved lineage-specific inserts that range from 50 to 500 amino acids in size. These inserts are typically independently-folded, located on the surface of the RNAP, and are highly mobile. Figure 4A highlights inserts in the RNAPs of the three phylogenetically distant clades, Thermus (T. aquaticus), Proteobacteria (E. coli), and Actinobacteria (M. tuberculosis), whose structures have been determined.

Figure 4. Diversity in bacterial RNAPs and their regulation.

(A) Structures of T. aquaticus RNAP (PDB ID 4XLR)²⁶, E. coli RNAP (PDB ID 4LK1)¹²², and M. tuberculosis RNAP (PDB ID 6C05)⁴⁶; highlighting the disposition of lineage-specific inserts. RNAPs are shown as molecular surfaces. Lineage-specific inserts are shown as transparent molecular surfaces superimposed with cartoon tubes (T. aquaticus lineage-specific inserts, green; E. coli lineage-specific inserts, blue; M. tuberculosis lineage-specific inserts, magenta). (B) Examples of unconventional transcription factors that regulate transcription initiation. Proteins are shown as a transparent molecular surface. Promoter DNA is shown as cartoons superimposed on a transparent molecular surface. The active site Mg²⁺ is shown as a green sphere. (Left) E. coli RNAP bound to promoter DNA and TraR (PDB ID 6PSV)⁴⁷. TraR (red) is shown as cartoon tubes. Highlighted is the interaction between TraR and the Escherichia coli RNAP lineage-specific inserts: βi4 and β’i6 (colored blue). (Right) M. tuberculosis open promoter complex (PDB ID 6EDT)⁴⁵ bound to RbpA (purple) and CarD (green). Highlighted are the CarD W85 that “wedges” the upstream fork junction and the RbpA N-terminal tail (NTT) that interacts with the T-strand that is loaded in the RNAP active site. β’i1 is also shown and colored magenta.

The specific functions of these inserts are mostly unknown. Based on the structural models of Thermus RNAP, it has been proposed that the β’i2 (Figure 4A, left) stabilizes the binding of σ to the RNAP¹⁷. In E. coli, deletions of the βi4 and β’i6 inserts in RNAP (Figure 4A, middle) lead to temperature sensitivity and affect cell viability, respectively¹¹². Recent work demonstrated that these inserts play essential roles in the transcription regulation by the transcription factor, TraR and likely DksA, which will be discussed in the next section, implicating them in regulation of the stringent response¹¹³. Structures of Mycobacteria RNAPs reveal the β’i1 insert (Figure 4A, right) emerges from the tip of the RNAP clamp module, is coupled to clamp conformations, and interacts with the N-terminus of σ^A, which increases the stability of the open complex^46,114. Besides TraR and DksA, it is unknown if additional transcription factors interact with these lineage-specific inserts leaving their roles in regulating transcription a compelling research area.

Lineage-specific transcription factors

Classical transcription factors bind directly to DNA upstream of the −35 element to either activate transcription by recruiting RNAP to DNA, or inhibit transcription by blocking RNAP binding to DNA¹¹⁵. However, several transcription factors do not directly bind DNA and use unique mechanisms to regulate transcription. Recently, the mechanisms for some of these factors have been revealed by a combination of approaches, including transcriptomics, bioinformatics, biochemistry, and structural biology. These “unconventional” factors include SutA factor found in Pseudomonas¹¹⁶, DksA/ppGpp^117,118 and TraR^113,119 found in Proteobacteria, RbpA^39,92,93 found in Actinobacteria, and the more widespread CarD^120,121 found in one-third of the bacterial genomes. These unconventional factors bind directly to RNAP to exert regulatory effects. Figure 4B highlights two recent RNAP structures from E. coli and M. tuberculosis that depict how these factors regulate transcription initiation.

In E. coli, TraR binds the secondary channel of RNAP and interacts directly with the βi4 and β’i6 lineage-specific inserts (Figure 4B, left), causing conformational changes in the RNAP that lead to differential regulation of certain promoters⁴⁷. In M. tuberculosis, CarD binds the β-protrusion and wedges the upstream DNA fork junction, thereby stabilizing the transcription bubble¹²¹ (Figure 4B, right). RbpA makes interactions with σD2 and tethers σ to the core RNAP³⁹. A recent cryo-EM structure of an M. tuberculosis open promoter complex revealed that the N-terminal tail of RbpA interacts with the T-strand of the promoter DNA near the active site⁴⁵ (Figure 4B, right), which has been shown to influence promoter escape kinetics⁶².

Antibiotics that inhibit transcription initiation

Bacterial RNAP is a proven target for antibiotics. Two clinically employed antibiotics target RNAP: 1) rifampicin is used for first-line treatment of tuberculosis caused by M. tuberculosis and 2) fidaxomicin is another used to combat intestinal infections caused by Clostridium difficile¹²³. Many other RNAP inhibitors have been characterized, and these have been useful as probes of the fundamental mechanisms of transcription. By binding to different regions of RNAP, they block transcription at different steps, providing insight into the roles of these sites in the transcription cycle. Here, we summarize recent work on the mechanisms of RNAP inhibitors that affect transcription initiation. Figure 5 illustrates these inhibitors’ binding sites and the steps of transcription that they affect.

Figure 5. Inhibitors of RNAP.

A) The positions of the antibiotics that bind RNAP. Sorangicin and kanglemycins are not shown as they overlap with rifampicin. The bridge helix (BH) and trigger loop (TL) are drawn in cartoon as indicated. The active site Mg²⁺ is shown for reference. B) The minimal mechanism for transcription initiation is shown, and the steps inhibited by each antibiotic is indicated.

Inhibiting initiation of promoter melting

The required movements of opening and closing of RNAP during promoter unwinding (Figure 2) present opportunities for the regulation of RNAP activity. Two classes of inhibitors have been found to block RNAP closing or opening: fidaxomicin and corallopyronin/myxopyronin, respectively.

Fidaxomicin kills M. tuberculosis¹²⁴, but because of its poor systemic absorption cannot be used to treat tuberculosis in the lungs. Cryo-EM structures of M. tuberculosis RNAP with fidaxomicin revealed that the antibiotic binds in a groove between the clamp domain and the rest of the RNAP, acting as a “doorstop” to prevent clamp closure for the initial recognition and melting of the −10 element^46,125. Remarkably, the essential Actinobacteria-specific general transcription factor RbpA interacts with fidaxomicin and increases the in vivo and in vitro sensitivity to the drug. Thus, fidaxomicin inhibits M. tuberculosis RNAP much better than E. coli RNAP, partly due to the presence of RbpA⁴⁶. These studies emphasize the importance of studying the native RNAPs complexes in drug-design⁴⁶.

Inhibiting downstream promoter melting

In contrast to fidaxomicin, myxopyronin (an RNAP inhibitor structurally related to corallopyronin) closes the RNAP clamp and traps a partially melted promoter bubble^31,126,127. Cryo-EM studies of M. tuberculosis RNAP holoenzyme with corallopyronin and a de novo melted promoter resulted in an intermediate with an 8-nucleotide transcription bubble instead of the full 13-nucleotide bubble⁴⁵. A control experiment without corallopyronin resulted in two structures: an open promoter complex and the same intermediate captured with corallopyronin⁴⁵. This finding suggests that corallopyronin stabilizes a natural intermediate of transcription initiation and inhibits the completion of the open promoter complex.

Inhibiting substrate binding

Kanglemycins are rifamycin congeners, identified through genome mining and metagenomic analysis of soil microbiomes, which inhibit rifampicin-resistant bacteria¹²⁸. Kanglemycins share the same chemical skeleton as rifampicin but contain additional functional groups, including an acid moiety and a sugar moiety. Crystal structures and biochemical assays of Mycobacteria and T. thermophilus RNAPs with kanglemycins show that these inhibitors bind in the same pocket as rifampicin but inhibit transcription at an earlier step. While rifampicin inhibits the transition from the initial transcribing complex to the elongating complex, the kanglemycins prevent binding of the initiating rNTP through clashes with the kanglemycin acid moiety^128,129. This clash was confirmed biochemically as ribonucleosides with monophosphates, but not triphosphates, were capable of initiation¹²⁸. Importantly, the structures showed that the kanglemycins’ sugar moiety, absent in rifampicin, created additional interactions with the RNAP. This observation explained kanglemycins’ potency against rifampicin-resistant RNAPs and informs future approaches to combat drug-resistant mycobacteria.

Pseudouridimycin is another inhibitor of nucleotide substrate binding that is active against Gram-positive and Gram-negative bacteria and not cross-resistant with current antibiotics¹³⁰. It is a nucleoside analog inhibitor that mimics uridine triphosphate. Similarly, another broad-spectrum antibiotic GE23077 also occupies the initiating nucleotide-binding site and prevents rNTP substrate loading¹³¹. Linking this inhibitor with rifamycin covalently enhanced its potency against rifampicin-resistant bacteria¹³¹.

Inhibiting catalysis

Streptolydigin inhibits transcription initiation and elongation by preventing phosphodiester bond catalysis in bacterial RNAP but not eukaryotic RNAPs^132–134. The crystal structures of streptolydigin with T. thermophilus RNAP showed that it binds near the trigger loop and bridge helix. These two dynamic structural elements play a role in the nucleotide addition cycle by changing their conformations^135,136. Hence, streptolydigin was proposed to inhibit transcription by both locking the bridge helix in the straight conformation and unfolding the trigger loop^135,136.

The CBR antimicrobials, discovered through small molecule library screening, are another class of bacterial RNAP inhibitors that inhibit the catalytic function of RNAP¹³⁷. X-ray crystallography studies with E. coli RNAP established that CBR compounds bind near the trigger loop and bridge helix in a region distal to the binding site of streptolydigin and other antibiotics, and hence is not cross-resistant with these RNAP inhibitors^138,139. Through these interactions, the CBR antimicrobials inhibit nucleotide addition and pyrophosphorolysis. The CBRs are narrow-spectrum, only inhibiting RNAPs from gram-negative bacteria. A recent screen revealed a new class of compounds, AAP, that functions similarly to CBRs but selectively inhibit M. tuberculosis RNAP. Structural studies of these compounds with M. tuberculosis RNAP revealed that they also overlap with the CBR-binding site and likely function by the same mechanism¹¹⁴.

Salinamides are compounds proposed to inhibit RNAP by a similar mechanism as the CBR antimicrobials¹⁴⁰. While CBR antimicrobials inhibit Gram-negative bacteria and the AAPs inhibit mycobacteria, salinamides inhibit transcription in both Gram-positive and Gram-negative bacteria^114,140.

Inhibiting productive transcription complexes

Rifampicin is one of the most potent and broad-spectrum antibiotics used against tuberculosis since 1968. However, strains resistant to this antibiotic have arisen, limiting its efficacy and clinical use. Rifampicin sterically blocks the RNA synthesis beyond 2–3 nucleotides by clashing with the elongating transcript^141,142. Another chemically unrelated RNAP inhibitor, sorangicin, shares the same binding site and mechanism of inhibition with rifampicin¹⁴³. Sorangicin inhibits some, but not all, rifampicin-resistant RNAPs hinting at differences in interactions with different mutants. This ability of sorangicin to inhibit some rifampicin-resistant RNAPs was attributed in part to its conformational flexibility relative to rifampicin.

The antibiotics reviewed here target various steps of transcription initiation (Figure 5) and inhibit RNAP by a number of mechanisms. These inhibitors not only differ in their specific mechanisms but also their selectivity against different bacterial RNAPs. Remarkably, in some cases, this selectivity is provided by a transcription factor, rather than RNAP itself. Characterization of the inhibitors provides insight into their mode of action and helps to advance our understanding of the fundamental principles of bacterial transcription.

Conclusion

Structural biology, combined with genetic, single-molecule, and biochemical studies, has revealed new regulatory mechanisms that extend the early paradigms of transcriptional regulators. This review presents established and new research that reveals both universal and lineage-specific mechanisms in the regulation of bacterial transcription initiation. These mechanisms rely on regulatory elements, including parts of the RNAP, transcription factors, and non-coding RNAs. Studies of in vitro bacterial transcription relied on E. coli until very recently. However, partly due to cryo-EM, it is now possible to structurally characterize different bacterial RNAPs. The study of transcription in different lineages of bacteria not only inform us of clade-specific transcription regulation but also contributed to our fundamental understanding of RNAP and its motions as it unwinds DNA. The differences between groups of bacteria are also essential considerations as we seek out new antimicrobials in the age of antibiotic-resistant pathogens, highlighting the importance of studying transcription outside the classical systems and organisms. In vitro characterization of inhibitors against specific pathogenic bacterial RNAPs will provide a platform for the design of diverse inhibitors that specifically target drug-resistant pathogens. In summary, the research we reviewed here has expanded our grasp of transcriptional regulation in bacteria and reveals the ever-expanding range of modulatory mechanisms. This review hopes to convey that we have only just begun to sample the range of diverse systems in bacterial transcription initiation, and we expect similar findings in other steps of transcription.

Box 1. 50 years of studying σ.

Image for 1969 adapted¹¹

Image for 1994 Copyright (1994) National Academy of Sciences, U.S.A.¹⁴⁹

Image for 2014 adapted⁷⁴

The year 2019 marked the 50^th anniversary of the discovery of σ factors. In 1969, the seminal discovery that an extraneous factor was required for the core enzyme to initiate transcription was presented¹¹. Shortly after, the factors capable of inhibiting σ factors (anti-σs) were discovered in bacteriophage T4¹⁴⁴. In 1979, the diversity of σ factors was realized by the discovery that B. subtilis RNAP required alternative σ factors under different environmental conditions^145,146. Between 1988 and 1994, the most abundant and divergent group of alternative σ factors was discovered and classified as the extracytoplasmic function (ECF) σ factors^147–149. Structural studies put into context the extensive genetic and biochemical groundwork, giving details of how σ factors preserve their overall structures yet maintain promoter recognition specificity. This structural era started with the crystal structure of E. coli σD2 in 1996¹⁵⁰, followed by structures of the individual domains of T. aquaticus housekeeping σD4 with the −35 element¹⁴ and σD2 with the −10²⁵, revealing the basis of promoter recognition. The first structures of σ-anti-σ complexes appeared in the early 2000s and explained the basis of their inhibitory activity^79,80. In 2009, a comprehensive genomic analysis of ECF σ factors proposed them as the third pillar of bacterial signal transduction⁷⁴. Now, studies of the ECF σ factors’ promoter specificity and their negative regulation by anti-σs have been applied to synthetic biology¹⁵¹. The crystal structure of σD2 of E. coli σ^E in complex with a −10 promoter element then allowed for comparison of promoter recognition by this divergent class of σs to the housekeeping σs¹⁵². The σ factors have also been structurally characterized in complex with RNAP since 2002¹⁶, both by X-ray crystallography and more recently by cryo-EM, culminating with structures of the ECF σ holoenzymes with DNA^60,153,154.

Box 2. Towards structural elucidations of RNAP initiation complexes.

*Blue text highlights cryo-EM structures and red text highlights X-ray crystallography structures

Image for 1969 adapted¹¹

Image for 1989 adapted¹⁵⁵

In 1969, the composition of the core subunits of RNAP was elucidated⁷. It was not until 1989 that the first structure of RNAP was visualized by 3-D reconstructions of E. coli holoenzyme RNAP taken from EM images of negatively stained 2-D crystals¹⁵⁵. This structure revealed the shape, size, and active site cleft of RNAP. Ten years later, the atomic-resolution structure of Thermus aquaticus RNA polymerase was solved by X-ray crystallography²⁴. Thirty years of biochemical and genetic studies of transcription in bacteria could now be contextualized structurally. The high-resolution structure showed the details of the catalytic site, including the active site Mg²⁺, and identified multiple motifs that were soon identified to be structurally conserved in the eukaryotic RNAPs^6,156. This milestone in transcriptional studies revolutionized the future of the field and set the stage for an impressive number of structures of RNAP complexes to follow. X-ray crystallography of these macromolecules reigned for almost two decades as the method for structural determination of transcriptional complexes. Because it was assumed that the thermophilic enzymes, due to their intrinsic stability, were more amenable to crystallography than their mesophilic homologues, the transcription field limited structural studies of RNAPs to the genetically intractable species, T. aquaticus and T. thermophilus. This approach changed in 2013 when three groups independently crystallized E. coli RNAP in various complexes^157,122,158. The third group of bacteria whose structure of RNAP has been elucidated is Mycobacteria. In 2017, M. tuberculosis and M. smegmatis transcription initiation complexes with antibiotics and the RbpA tether were determined by X-ray crystallography^28,39,114. In the early 2010s, electron microscopy reemerged as a tool to obtain structures of large macromolecular complexes due to advances in electron detectors and software that made high-resolution imaging possible. Given its mass of more than 400 kilodaltons, RNAP was the perfect specimen for the resolution revolution that was occurring in the field of single-particle cryo-EM. In 2016, high-resolution (better than 4 Å) single-particle cryo-EM structures of yeast RNAP transcription initiation complexes were solved¹⁵⁹. These structures were rapidly followed by the cryo-EM structures of various bacterial RNAP complexes^108,160,161. This method was then applied to observe M. tuberculosis RNAP in a range of molecular conformations that the enzyme goes through during promoter unwinding, including the clear capture of a promoter unwinding intermediate^45,46. That work was then followed by additional initiation pathway intermediates with the E. coli RNAP⁴⁷. Single-particle cryo-EM has now made it possible to determine structures of transient steps of transcription initiation. In full circle, the electron microscope is now the leading instrument for determining structures of RNAP macromolecular complexes. The timeline highlights some of the breakthroughs in the structural biology of RNAP, with a focus on the initiation complexes.