Transcription, the first phase of gene expression, is performed by the multi-subunit RNA polymerase (RNAP). Bacterial RNAP is a validated target for clinical antibiotics. Many natural and synthetic compounds are now known to target RNAP, inhibiting various stages of the transcription cycle. However, very few RNAP inhibitors are used clinically. A detailed knowledge of inhibitors and their mechanisms of action (MOA) is vital for the future development of efficacious antibiotics. Moreover, inhibitors of RNAP are often useful tools with which to dissect RNAP function. Here, we review the MOA of antimicrobial transcription inhibitors.
Antibiotic resistant infections represent a growing threat to human life. However, few novel antibiotics have reached the market in recent years. Transcription, the first step of gene expression in all domains of life involves the synthesis of an RNA molecule from a DNA template. In bacteria, transcription is catalysed by a multi-subunit RNA polymerase (RNAP). RNAP performs sequence-dependent transcription in distinct phases of initiation, elongation and termination. During initiation, the catalytic core enzyme, consisting of five subunits (α2ββ′ω) binds to a σ initiation factor to form the holoenzyme [1–3]. Primary σ factors recognise −10 and −35 promoter sequences allowing RNAP to bind to promoter DNA to form a closed promoter complex (RPc) [1–5]. RPc undergoes spontaneous isomerisation to the open promoter complex (RPo) in which double-stranded DNA (dsDNA) is melted around the transcription start site (TSS) and loaded into the active centre [1,2]. Following formation of RPo, RNAP reiteratively produces and releases short RNA transcripts, between 2 and 9 nucleotides in length (abortive products) [6,7]. For transcription to proceed, the enzyme must relinquish contacts with the promoter and dissociate from σ, transitioning to a highly stable and processive elongation complex (EC) [1,8]. ECs transcribe DNA until recognition of a termination sequence or binding to a termination factor .
The overall structure of RNAP resembles that of a crab claw, with two opposing pincer-like structures (β and β′) bordering the main channel [10–12]. The RNAP catalytic site co-ordinates a catalytic Mg2+ situated in the cleft between the pincers [13,14]. The active centre is accessed via several channels. The primary channel accommodates the DNA–RNA hybrid and downstream DNA (dwDNA), the secondary channel provides access for NTP substrates and also accommodates the 3′-OH end of RNA in backtracked ECs while in active ECs, the nascent RNA is extruded from the RNA exit channel [15,16]. The β′ subunit comprises the larger of the two pincers (the clamp). It is a highly mobile domain able to hinge around a flexible region at its base termed the switch region comprised of five discrete elements (SW1–SW5) [17–19]. The clamp is able to swing open to accommodate the DNA within the active centre cleft. Upon binding of DNA, the clamp reverts to a closed conformation to retain melted DNA within the active centre cleft during initiation and elongation [8,20].
The RNAP active centre possesses two distinct binding sites; the ‘i’ site and ‘i + 1’ site. During the nucleotide addition cycle (NAC), RNAP transfers the nucleotidyl moiety from the NTP substrate bound in the ‘i + 1’ site to the 3′-hydroxyl of nascent RNA bound in the ‘i’ site . RNAP then translocates to vacate the ‘i + 1’ site for further NTP addition [21,22]. The NAC is regulated by cyclic opening and closing of the active site through a mobile domain of the β′ subunit, the Trigger Loop (TL) [21,22]. The TL conformation can change between the unfolded (Trigger loop) and folded (Trigger helix) states (Figure 1C). An open active site with an unfolded TL allows translocation of EC and facilitates entry and binding of NTP substrates in the active site. Whereas the folded state of the element, Trigger helix (TH) leads to closed active site and stimulates catalysis. The Bridge Helix (BH), a metastable β′ α-helix with two flexible hinges spanning the RNAP active site makes tight contacts with β-subunit and Trigger helices and works interdependently with TH influencing its formation and therefore the catalysis [23,24]. Another domain of β′, the fork loop (F-loop) is an amino-terminal extension of BH that interacts with Trigger helices and modulates the closed conformation of it (Figure 1B) .
Inhibitors of active centre mobile elements.
A wealth of synthetic and natural compounds inhibits RNAP with varying mechanisms of action (MOA). RNAP inhibitors have been reviewed comprehensively in several publications [19,25–29]. In this mini review, we aim to provide an up to-date summary of the mechanisms by which antibacterial inhibitors impair RNAP function.
Antibiotics blocking nascent RNA extension
Rifamycins (RIFs) are natural product antibiotics belonging to the ansamycin class. They contain a polyketide chain (ansa bridge) spanning a naphthalene moiety from its non-adjacent ends . Since the discovery of naturally occurring RIF-SV, many semisynthetic RIF programmes have aimed at developing compounds with enhanced activity and improved pharmacokinetics . Most modifications have focused on the C-3 (i.e. Rifampicin and Rifapentine), C-4 (i.e. Rifamide) and C-3/C-4 atoms (i.e. Rifaximin, Rifabutin and Rifalazil) of the naphthalene ring [31–36]. Despite extensive synthesis of RIF derivatives, only few have entered into the clinic. Rifampicin (RMP) is the first-line treatment for mycobacterial infections including Tuberculosis (TB) [35–37]. TB, an infection caused by Mycobacterium tuberculosis is the leading cause of death by a single infectious agent worldwide. In addition to RMP there are currently three more RIFs in clinical use: rifaximin (RXM), rifapentine (RPT) and rifabutin (RBT) [33–35].
The antibacterial activity of all RIFs is due to inhibition of DNA-dependent RNA synthesis as a result of strong binding to prokaryotic RNAPs . RMP binds to a pocket (RIF-binding pocket) in β subunit of RNAP within the DNA/RNA channel, 12 Å away from the catalytic Mg2+ ion [39,40]. Binding of RMP to RNAP sterically blocks the formation of the second or third phosphodiester bond and induces the release of short abortive RNAs [40,41]. The steric model alone does not explain the differences in cross resistance of different RIFs . Moreover, depending on the functional group on the C3 and/or C4 of RIFs, they can inhibit the formation of the first (Rifalazil and RBT) or second (RMP and RPT) phosphodiester bond . Artsimovitch and Vassylyev proposed an inhibition model in which instead or in addition to the steric model, RIFs allosterically reduce the affinity of Mg2+ ion to the RNAP active centre . However, all biochemical and structural data strongly support the initial hypothesis that RIFs inhibit RNAP via sterically blocking the path of growing RNA [45,46].
Mutations in rpoB encoding the β subunit of RNAP are the predominant mechanism of resistance to RIFs . Only three mutations, βS531L, βH526Y and βD516 V (Escherichia coli numbering) account for ∼41, 36 and 9% of all clinically isolated RMP-resistant (RMPR) TB, respectively . Depending on the point mutation, the mechanism of resistance to RIFs varies. For example, βS531L alters the RIF-binding pocket to produce a steric clash upon binding of RMP, while βD516V eliminates the bond between RMP and RNAP. As a result, both types of mutations reduce RMP-binding affinity .
RIF-binding affinity is dependent on the spatial conformation of four critical oxygen functionalities at C-1, C-8, C-21 and C-23 (Figure 2A, right). Consequently, almost all modifications of the ansa bridge lead to loss of activity [30,49]. A RIF derivative with a distinct ansa bridge, Kanglemycin A (KglA), has activity higher than RMP against multidrug-resistant TB and also RMPRE. coli RNAPs [50–53]. The higher activity of Kgl compared with RMP against resistant strains and RNAP is due to additional deoxysugar and succinate ansa bridge substituents at C-27 and C-20, respectively (Figure 2A, left) [50,51]. The deoxysugar makes additional non-polar interactions with the β subunit, increasing the binding surface (Figure 2B). Furthermore, the bulky succinate side chain precludes the formation of initial dinucleotides via steric and electrostatic clash with phosphates of the + 1 NTP. This model explains the difference between KglA and RMP inhibition of initial di- or tri- nucleotide synthesis [50,51]. Importantly, a rotational shift of KglA within the binding pocket allows KglA to adopt a novel binding conformation away from the pocket interior (Figure 2C). This rotational shift allows KglA to bind to the most frequent RIFR RNAP, βS531L by preventing a severe clash with the βL531 residue and minimises structural rearrangements of the β region, fork loop 2 . Due to additional binding of the sugar moiety with the RNAP, only an unlikely event of two concurrent mutations in the RIF-binding pocket can lead to very-high-level, >1000×, KglA resistance .
KglA and RMP structures and their binding to RNAP.
Sorangicin (SOR) is a naturally occurring macrolide polyether antibiotic produced in two variants (A and B). SOR possesses a broad spectrum of activity but is more effective against Gram-positive bacteria . A crystal structure of SOR bound to T. aquaticus RNAP  and analysis of mutations conferring resistant to SOR  showed that SOR binds to RNAP in the RIF-binding pocket. SOR specifically inhibits transcription initiation by bacterial RNAP with the same mechanism as that of RMP though the two compounds do not show apparent structural similarities. SOR structure is thought to be more flexible compared with RMP partly due to lack of the naphthyl moiety present in RIFs. This structural flexibility causes SOR to be less sensitive to conformational changes of the RIF-binding pocket and therefore retains its activity against some RMP-resistant mutants. For example, RNAPs with substitutions at βS531 (E. coli numbering) remain relatively susceptible to SOR .
The natural product antibiotic, GE23077 (GE), is a specific inhibitor of bacterial RNAPs including RMPR RNAPs. However, its inability to cross the bacterial membrane results in limited antimicrobial activity [57–59]. GE inhibits transcription initiation by inhibiting the synthesis of 2-nt RNA products. Crystal structure of Thermus thermophilus RNAP holoenzyme in complex with GE shows the inhibitor occupies the RNAP i and i + 1 sites. It makes contacts with the β link region, DII loop and three RNAP Asp residues that co-ordinate catalytic Mg2+. Binding of GE to RNAP sterically blocks the binding of initiating NTPs to the RNAP i and i + 1 site leading to the inhibition of transcription initiation . The different MOA of RIFs and GE, and also the proximity of their binding sites, enables the preparation of a hybrid antibiotic which may reduce the resistance rate and overcome pharmacokinetic variances of the two drugs. Zhang et al. prepared a hybrid of RIF-SV and GE23077 which exhibited activity against WT and resistant RNAPs .
Disrupting holoenzyme assembly
The binding of housekeeping σ factors to RNAP core is essential for the initiation of transcription. The SB-2 series are a group of synthetic compounds identified in a high-throughput screen assessing RNAP holoenzyme assembly in which binding affinity between E. coli RNAP core and σ70 was measured [60,61]. Two ‘hit’ compounds, SB11 and SB15, were shown to specifically prevent σ binding to RNAP core. Both compounds were found to be inhibitors of in vitro transcription by E. coli RNAP. A range of furanyl rhodanine SB derivatives were found to possess antibiotic activity against many Gram-positive bacteria . Interestingly, SB series compounds inhibit σ-independent transcription at poly (dA–dT) DNA, suggesting the binding site is located on RNAP core. Furthermore, SB compounds are able to inhibit transcription after the formation of the holoenzyme suggesting their mechanism of action is allosteric . However, the failure to identify resistance mutations or identify the structure of RNAP complexed with an SB series compound has left the binding sight, as of yet, undetermined.
Nucleoside analogues are compounds structurally similar to NTP substrates that can often compete for the respective binding site. Consequently, they are often selective inhibitors of nucleic acid polymerases . Pseudoridmycin (PUM) is the first specific nucleoside analogue inhibitor of bacterial RNAP (Figure 3). It was identified from a screen of actinobacterial and fungal culture extracts for its ability to inhibit bacterial RNAP (E. coli RNAP) but not a structurally unrelated phage RNAP . The compound exhibits specific antibacterial activity against spectrum of bacteria, including many drug-resistant pathogens . Biochemical observations suggested PUM potently inhibits RNAP by competing with UTP for the i + 1 site (Figure 3A,B). PUM inhibits incorporation of UTP in single and multiple nucleotide addition experiments but fails to prevent incorporation of ATP, GTP and CTP. Additionally, PUM only inhibits transcription at templates that direct the incorporation of UTP. Amino-acid substitutions conferring PUM resistance in E. coli were found at two sites which map to a cluster within the RNAP active centre region overlapping the i + 1 site .
RNAP inhibition by the nucleoside analogue inhibitor Pseudoridmycin (PUM).
Crystal structures of T. thermophillus transcription initiation complexes containing PUM confirm the inhibitor makes direct Watson–Crick interactions with template DNA and interacts with residues of the i + 1 site in a comparable manner to NTPs (Figure 3C). Moreover, the Watson–Crick base pairing interaction with the PUM base moiety and template DNA is only possible at template positions directing UTP incorporation, explaining the specificity of PUM for inhibition of UTP incorporation .
Mobile elements of primary channel
Salinamides comprise a group of structurally related natural antibacterial and anti-inflammatory agents . Salinamides A (SAL), B and F effectively inhibit bacterial RNAPs [66,67]. Biochemical experiments using the E. coli RNAP revealed that SAL inhibits transcription initiation (nucleotide addition) and elongation (nucleotide addition and pyrophosphorolysis — the reverse of phosphodiester bond formation), it does not compete with substrate NTPs and it equally inhibits WT RNAP and RNAP with TL deletion. SAL-resistant mutations within the rpoB or rpoC suggest a possible binding site adjacent to the active centre of RNAP which does not overlap with the binding site of other RNAP inhibitors .
According to the crystal structure of E. coli RNAP holoenzyme in complex with SAL, the inhibitor binds the RNAP BH cap and makes direct interactions with the fork loop, BH N-terminal hinge (BH-HN) and the link region (Figure 1B). Crystal structure of RNAP-SAL also shows that SAL binds to the unbent (straight) state of RNAP BH-HN. Considering that the bent BH-HN conformation is critical for reactions performed by RNAP such as bond formation and pyrophosphate-release, it is proposed that SAL allosterically inhibits transcription by stabilising the RNAP BH-HN unbent state preventing the conformational dynamics required for nucleotide addition .
Despite the biochemical data showing SAL does not need TL for inhibition of nucleotide addition, the crystal structure also indicates that SAL sterically prevents TL folding. Proposedly, this can be a secondary mechanism of inhibition by SAL but it is not essential for its inhibitory activity . Further studies could potentially help to investigate the role of TL in MOA of SAL.
Streptolydigin (STL) is a naturally derived antibiotic comprised of a tetramic acid containing a sugar and streptolol functional groups. It possesses broad spectrum antibacterial activity [68–70]. Early biochemical studies showed that STL inhibits substrate binding, catalysis and translocation of bacterial RNAP .
Two parallel studies used biochemistry, microbiology and structural data to investigate the MOA of STL [72,73]. Based on crystal structures of T. thermophilus RNAP bound to STL and also STL-resistant mutations, STL binds to bacterial RNAP in the vicinity of the active site (Figure 1A,B). The STL streptolol moiety makes direct contacts with two regions of β′ and also the N-terminal end of the BH, while the opposite side of STL occupies a region close to the central portion of BH and TL. The sugar moiety of the tetramic acid makes interactions with downstream DNA and TL (Figure 1B) [72,73]. It is suggested that STL, via trapping a straight BH conformation and trapping a TL open conformation, locks the active site in a conformational state which disfavours substrate loading. Therefore, STL inhibits transcription by blocking the conformational changes that are critical for all three reactions catalysed by RNAP .
CBR703 is the progenitor of a new class of synthetic antibiotics comprising two linked aromatic rings [74,75]. CBRs inhibit transcription by some Gram-positive (excluding Mtb) and Gram-negative bacteria but do not affect human RNAP . Isolated CBR703-resistant mutants and crystal structures of RNAP bound to CBRs suggested that the target of CBR703 is a hydrophobic two-pocket site surrounded by β subunit elements that link fork loop 2, DII loop and also the β′ subunit elements fork loop and BH (Figure 1A,B) [74,76,77].
As CBRs inhibit all catalytic activities of RNAP, but have little or no effect on translocation, it is concluded that CBRs act by inhibition of TL folding. TL folding blockage is a result of CBRs binding to fork loop and/or increasing the coupling between BH and β-subunit that weakens the BH–TH interactions [76–78]. Although CBRs inhibit all catalysis reactions by RNAP, only inhibition of intrinsic hydrolysis is entirely dependent on TL. The CBR inhibition of RNAP nucleotide addition is also partially dependent on TL while pyrophosphorolysis inhibition is TL-independent. Bae et al. speculated that CBR703 inhibits a previously unknown conformational change of the active site only important in catalysis involving triphosphate substrates (nucleotide addition and pyrophosphorolysis). Such conformational changes may potentially be regulated by BH-HN bending .
Recently, a new class of RNAP inhibitors was identified during a high-throughput screening of synthetic compounds. D-AAP1, the prototype of the class inhibits M. tuberculosis RNAP but has poor activity against other bacterial and human RNAPs. Crystal structure of M. tuberculosis RNAP in complex with D-AAP1 and isolated resistant mutants proposed that D-AAP1 binds to a pocket centred on the BH-HN. Because of the similar binding sites of D-AAP1 and CBRs, it is assumed that they inhibit RNAP by the same MOA. The selectivity of the two inhibitors is due to structural differences in M. tuberculosis and Gram-negative bacteria (i.e. E. coli). The three-pocket site in M. tuberculosis accommodates the three ringed D-AAP1 while CBRs (two ringed) fit into the two-pocket site on E. coli RNAP .
Blocking NTP uptake
Microcin J25 (Mcc25) is a cyclic antibiotic peptide synthesised by certain strains of E. coli. Mcc25 is a 21 amino-acid peptide inhibitor of RNAP with an unusual lassoed tail structure [79,80]. It inhibits both transcription initiation and elongation . Saturation mutagenesis experiments indicated all resistance determinants of Mcc25 are located within the secondary channel of RNAP, indicating the location of the putative binding site [82,83]. It has been proposed that Mcc25 inhibits RNAP by blocking the secondary channel and consequently blocking the uptake of NTP substrates into the active centre. Indeed, Mcc25 increases the Km of NTP binding, in agreement with the proposed ‘cork-in-a-bottle’ mechanism . Mcc25 also inhibits RNAP catalysed pyrophosphorolysis. However, some Mcc25 resistance determinants overlap those of the proposed STL-binding pocket on the β subunit, signifying the inhibitory mechanism may be more complex . Additionally, binding of Mcc25 and STL are mutually exclusive suggesting features of their inhibitory mechanisms may be shared. Elucidation of the structure of RNAP complexed with Mcc25 may allow the exact nature of Mcc25's MOA to be identified.
Preventing promoter open complex formation
Fidaxomicin (Fdx) (trade name Dificid), used in the treatment of Clostridium difficille-associated diarrhoea, is known to bind to the RNAP switch region . Structural data obtained by cryo-electron microscopy elucidated the structure of Fdx complexed with M. tuberculosis RNAP (Figure 4A,B). The Fdx-binding site encompasses switches SW2, SW3 and SW4, as well as clamp α helices βa16α1 and β′a4α1. SW1 and SW2 are the main elements that mediate the conformational changes of the clamp (Figure 4C) [84,85]. Fdx makes five critical hydrogen bonds with RNAP, residues βK1303, β′Q94, β′R99, β′248 and β′337 (E. coli numbering) . Substitutions at any one of these residues confer resistance to Fdx. σ-dependent promoter melting and loading of template DNA into the active centre is abolished by Fdx [19,84,86]. In the absence of Fdx, RNAP RPo formation relies upon recognition of −10 and −35 promoter elements by σ factor substructures region 2 (σR2) and region 4 (σR4), respectively. Conserved non-template DNA bases (NT) at the −11 and −7 positions are flipped out of the DNA duplex into protein pockets on σR2 to stabilise RPo [11,87]. It is proposed that, by locking the clamp in an open conformation (Figure 4B), Fdx prevents the correct spatial orientation of σR2 and σR4 required for simultaneous recognition of −10 and −35 core promoter elements. Indeed, RNAP–Fdx complexes bind to upstream promoter elements but are unable to recognise the −10 element and nucleate promoter melting [84,85]. Lin et al. suggest that, in particular, the ‘Trp wedge’ of σR2, responsible for intercalation into the non-template strand at the −12 position to nucleate promoter melting, is unable to contact DNA with the clamp in an open conformation. Furthermore, the σR2 ‘NT −11 pocket’ which binds the flipped out −11 nucleotide of the non-template strand can only engage DNA with the clamp in a closed conformation . Consequently, Fdx–RNAP complexes are unable to recognise the −10 promoter element and transition from RPc to RPo to initiate transcription.
Clamp locking by switch region inhibitors.
Squaramides, myxopyronin, corallopyronin and ripostatin
The natural product transcription inhibitors myxopyronin (Myx), corallopyronin and ripostatin also target the switch region, but bind within an adjacent pocket to that of Fdx and inhibit RNAP through a somewhat different mechanism (Figure 4B) [17,19]. The most extensively characterised, Myx interacts with SW1 and the opposing face of SW2 to lock the RNAP clamp in a closed conformation (Figure 4C) [17,19]. Introduction of mutations in vitro confirmed three crucial Myx-binding determinants all of which when substituted confer resistance to Myx and also to corallopyronin and ripostatin, suggesting they bind within the same pocket [18,19]. Biochemical analysis of Myx indicates the inhibitor does not completely prevent promoter melting or the binding of double-stranded dwDNA. Instead, Myx prevents propagation of promoter melting reaching the TSS, suggesting Myx targets a late initiation intermediate en route to RPo [17–19]. It is significant to note that Myx inhibits transcription at artificially melted promoters, indicating Myx clamp locking may disrupt critical interactions with melted template DNA at the TSS .
Squaramides (SQ) are a group of small synthetic RNAP inhibitors that also target the switch region. Cocrystallisation of SQ with E. coli RNAP show the compounds bind within the same pocket as Myx, but do not possess extensive interactions with SW3 and SW4. Structural data indicated SQ displacement of SW2 would clash with the correct positioning of template DNA at positions +3 and+4, suggesting SQ act to lock the clamp in a closed conformation and sterically occlude correct loading of melted promoter DNA near the TSS .
Antibiotic resistance is an increasing threat and only few new compounds have reached the clinic in recent years. The essential nature of RNAP and its divergence from its eukaryotic counterpart make it an excellent target for development of highly specific antibacterial drugs (i.e. RIFs and Fdx). There are currently more than 10 antibiotics which target RNAP with distinct binding sites. Besides playing a vital role as antibiotics, inhibitors of RNAP are powerful tools for investigations of molecular mechanisms of their target (i.e. STL and CBRs). Moreover, modifications of known inhibitors of RNAP could potentially lead to compounds with enhanced properties. For example, the recently characterised KglA is a starting point for developing advanced RIFs with ansa bridge modifications.
sigma factor region
mechanism of action
closed promoter complex
open promoter complex
Transcription start site
The Authors declare that there are no competing interests associated with the manuscript.
These authors contributed equally to this work.