CarD and RbpA modify the kinetics of initial transcription and slow promoter escape of the Mycobacterium tuberculosis RNA polymerase.

The pathogen Mycobacterium tuberculosis (Mtb), the causative agent of tuberculosis, enacts unique transcriptional regulatory mechanisms when subjected to host-derived stresses. Initiation of transcription by the Mycobacterial RNA polymerase (RNAP) has previously been shown to exhibit different open complex kinetics and stabilities relative to Escherichia coli (Eco) RNAP. However, transcription initiation rates also depend on the kinetics following open complex formation such as initial nucleotide incorporation and subsequent promoter escape. Here, using a real-time fluorescence assay, we present the first in-depth kinetic analysis of initial transcription and promoter escape for the Mtb RNAP. We show that in relation to Eco RNAP, Mtb displays slower initial nucleotide incorporation but faster overall promoter escape kinetics on the Mtb rrnAP3 promoter. Furthermore, in the context of the essential transcription factors CarD and RbpA, Mtb promoter escape is slowed via differential effects on initially transcribing complexes. Finally, based on their ability to increase the rate of open complex formation and decrease the rate of promoter escape, we suggest that CarD and RbpA are capable of activation or repression depending on the rate-limiting step of a given promoter's basal initiation kinetics.

INTRODUCTION

Bacterial transcription initiation is a complex, multi-step process and is the major regulatory determinant for gene expression (1). Transcription is carried out by the RNA polymerase (RNAP) holoenzyme, consisting of the five subunit core RNAP (α2ββ’ω) and a σ subunit involved in specific promoter recognition (2). Extensive studies using Escherichia coli RNAP-σ70 holoenzyme (from here on referred to as Eco holo) have identified numerous kinetic intermediates involved in DNA-dependent RNA synthesis (outlined in Figure 1A). Transcription is initiated when RNAP holoenzyme (R) first binds duplex DNA promoter sequences (P) to form the closed complex (RPc). Binding free energy then drives multiple conformational changes in both the RNAP and DNA that lead to the isomerization of RPc to RPo (open complex), where the duplex DNA is unwound around the transcription start site (TSS), resulting in an ∼13 base-pair (bp) single-stranded DNA bubble. After binding the initiating nucleotide, RNA polymerization begins as nucleotides (nt) complementary to the template DNA strand are incorporated into a growing RNA transcript resulting in an initial transcribing complex (RPitc). In RPitc, upstream contacts between RNAP and the promoter are static while downstream DNA is translocated into the enzyme, resulting in DNA scrunching (i.e. a growth of the DNA bubble) and possible abortive RNA synthesis where RNAP cycles through stages of short transcript production (3,4). Once 9–15 nt are incorporated into the RNA transcript, upstream contacts are broken, σ factor may dissociate, the transcription bubble shrinks, and RNAP fully escapes the promoter. This marks the end of initiation and the beginning of processive elongation (RPe), where full-length transcripts are produced (5–7). From studies of Eco holo, numerous kinetic models have been put forth to describe promoter escape. In the most simplistic scheme, the pathway of promoter escape follows a sequential model where no off-pathway intermediates are generated prior to synthesis of full-length transcripts (Figure 1A – productive synthesis). Alternatively, branched models of promoter escape have been used to account for complexes that are unable to escape the promoter (i.e. moribund complexes) (8,9), backtracked/paused complexes originating from RPitc intermediates (10–12), as well as for cases where abortive initiation is not an on-pathway intermediate to an elongation complex and does not precede full-length transcription (13) (Figure 1A – abortive synthesis).

Figure 1.

Schematic of bacterial transcription initiation and the structure of the Mtb RNAP-σA holoenzyme. (A) A kinetic model of initiation where the promoter DNA (P) containing the template strand (cyan) and non-template strand (purple) with a Cy3-fluorescent label positioned at the +2 position (yellow star) undergoes a fluorescence increase when RNAP (R, grey circle) unwinds the DNA during the isomerization from RPc to RPo. NTP addition results in scrunched RPitc intermediates, where the open DNA bubble is increased in size. Transcription complexes can undergo either abortive cycling (red box) between RPitc and RPo states or productive synthesis (green box). Promoter escape leads to a fluorescence quenching when the DNA strands re-anneal in the vicinity of the Cy3-probe. (B) Mtb RNAP-σA holoenzyme rrnAP3 open promoter complex with CarD and RbpA (PDB: 6EDT) (30). Cryo-EM structure containing the Mtb rrnAP3 promoter (–60 to +30) with a fully open DNA bubble. Cy3-labeling position for use in stopped-flow experiments shown in yellow. Black box indicates a 90o rotation of the region shown in (C) where the predicted interactions of σA, CarD and RbpA with the open DNA bubble are shown. The Mg2+ ion indicates the position of the RNAP active site.

While much progress has been made to determine mechanistic details of Eco promoter escape, the kinetics and mechanism of mycobacterial promoter escape remains largely uncharacterized. To date, our kinetic studies on mycobacterial transcription have only focused on RPo formation. Our results and those of others have indicated that the housekeeping Mycobacterium bovis RNAP-σA holoenzyme (Mbo holo), which differs in only one amino acid residue compared to Mtb RNAP-σA holoenzyme (Mtb holo), and Eco holo demonstrate dramatically different kinetics during the formation of RPo on the Mtb ribosomal RNA (rRNA) rrnAP3 promoter (14,15). In contrast to Eco, which frequently forms a stable RPo, Mycobacterium, Bacillus and Thermus RNAPs have been shown to form an unstable RPo (14–18). Structural studies indicated that mycobacterial RNAPs and RNAPs from other Actinobacteria have unique insertions originating from the β’ and σA subunits that may interfere with promoter DNA entry into the active site (19), providing a possible explanation for the lineage specific deficiency in RPo stability. However, lineage specific structural insertions are common across bacterial core RNAP subunits and don’t always lead to formation of an unstable RPo; the insertions present in E. coli stabilize the open complex (20).

In addition to differences in the RNAP holoenzyme itself, Actinobacteria have unique transcription factors, including CarD and RbpA, which lead to a cooperative stabilization of the rrnAP3 RPo (15,21), and as a result have been characterized as transcriptional activators. Neither CarD or RbpA are present in E. coli (22), suggesting distinct strategies of transcriptional regulation across these bacterial species. carD and rbpA are up-regulated when subjected to bacterial stresses and are essential during growth under nutrient-rich conditions (22–26), suggesting roles in both maintaining cellular homeostasis and modulating Mtb gene expression to drive the pathogenesis of tuberculosis, the current leading cause of death from a single infectious agent. As RNAP is a direct target for first line antibiotics used to treat Mtb infection (27), further mechanistic studies of the unique mycobacterial transcription system is warranted and may lead to novel therapeutic strategies.

How CarD and RbpA interact with different components of the initiation complex has been elegantly elucidated via structural studies. Specifically, structures with σA-containing RNAP holoenzymes reveal that CarD and RbpA interact with both promoter DNA and RNAP (15,19,28–30). CarD consists of an N-terminal RNAP-interacting domain (RID) that binds the β-subunit of RNAP, a C-terminal DNA-binding domain (DBD) that makes non-specific contacts with promoter DNA, and a critical tryptophan residue (W85) that interacts with the upstream edge of the transcription bubble (22,28,30–32) (Figure 1B-C). Mutations in the RID that weaken the association with RNAP, the DBD or W85 can lead to increased sensitivity to antibiotics and decreased RPo stability suggesting that all domains contribute to both biophysical and biological activities of CarD (14,16,31,33,34). RbpA consists of an N-terminal tail (NTT) that contacts RNAP β’ and σ subunits, a core domain (CD) that contacts RNAP β’, and a C-terminal sigma-interacting domain (SID). Separating the CD and SID is a basic linker (BL) region that interacts with promoter DNA just upstream of the −10 element (15,19,29,30,35) (Figure 1B and C). While both the BL and SID are necessary to promote a stable RPoin vitro, removal of the NTT and the CD leads to enhanced RPo stability relative to wild-type RbpA (25). However, in vivo RNA-sequencing analyses have indicated that the NTT and CD are responsible for a wide-ranging activation of gene expression, whereas the BL and SID can either activate or repress transcription depending on the promoter (25).

These RNA-sequencing results suggest that the current model of RbpA being solely defined as a transcriptional activator is not entirely correct. Working under the assumption that RbpA makes the same contacts with the RNAP holoenzyme regardless of which promoter is bound in the initiation complex, we hypothesized that for RbpA to also repress transcription, it must affect multiple kinetic intermediates during transcription initiation (36). For instance, by altering the stability of RPo as well as the rate of promoter escape, transcription factors can exert differential regulatory outcomes (activation or repression) depending upon the rate-limiting kinetic transition for a given promoter (37,38). As RPo stability has been shown to be inversely correlated with escape rates (39), direct measurements of the effects of transcription factors on escape are infrequently obtained despite the fact that the quantitative effect on escape can affect the regulatory outcome of the factor and represents an equally important step in gene regulation. As RbpA increases RPo stability on the rrnAP3 promoter (15,21,25), we hypothesized that the transcriptional repression observed in vivo could be the result of RbpA delaying promoter escape. Similarly, as CarD also increases RPo stability on rrnAP3 (14,15), we predicted that it also may slow promoter escape kinetics.

With the goals of understanding both the differences between Eco and Mtb transcriptional mechanisms and the effect that CarD and RpbA have on the overall rate of transcription initiation, we set out to determine both the basal and regulated rates of promoter escape for Mtb. Here, using a fluorescently labelled promoter construct that is sensitive to changes in DNA conformation (Figure 1A) (39), we measured the kinetics of promoter escape in a real-time stopped-flow assay. These transient kinetic measurements also allowed for detection of RPitc intermediates; states that were previously unable to be observed given the dead-time of manual mixing techniques that have been employed to study Eco promoter escape (39). We show that in relation to Eco, Mtb displays faster escape kinetics on the rrnAP3 promoter, which can be delayed by both RbpA and CarD. By modelling changes in the rate of transcript production based on the factor-dependent effects on the kinetics of RPo formation and promoter escape, we highlight the potential for RbpA and CarD to either up- or down-regulate transcript production depending on the basal rates of a given promoter. This new regulatory model for these factors highlights the possibility that CarD and RbpA can enact both transcriptional activation and repression to benefit Mtb persistence and survival.

MATERIALS AND METHODS

Preparation of fluorescent promoter DNA template

A 150 bp promoter template was generated containing the rrnAP3 promoter labelled with Cy3-NHS (Lumiprobe Corporation) attached to a C6-amine modified thymine on the +2 position of the non-template strand (for sequence, see SI – Extended Methods). For promoter preparation and labelling protocols, see (14).

Protein purification

Plasmids containing the Mtb H37Rv genomic DNA encoding for the different Mtb RNAP holoenzyme subunits were a generous gift from Dr. Mukhodpadhyay (Bose Institute, India) (40). Co-expression and purification of Mtb RNAP-σA holoenzyme was carried out in accordance with Prusa et al., 2018. Mtb RbpA and CarD were expressed and purified as previously described (14,25). Eco RNAP-σ70 holoenzyme (catalogue #M0551S) was purchased from New England BioLabs, Inc.

Preparation of competitors to promote single-turnover conditions

Heparin sodium salt from porcine intestinal mucosa (MilliporeSigma) was dialyzed in transcription buffer (20 mM Tris (pH 8.0), 40 mM NaCl, 75 mM K-glutamate, 10 mM MgCl2, 5 μM ZnCl2, 20 μM EDTA). Heparin concentration was determined by standard curve titration with Azure A (41) and was repeated for each new lot of heparin purchased. Salmon-sperm DNA sheared to a maximum size of 2000 bp (Thermo Fisher Scientific) was buffer exchanged in transcription buffer and concentration was determined by A260.

Stopped-flow fluorescence assays

All experiments were performed at 37°C using a SX-20 stopped-flow spectrophotometer (Applied Photophysics) with a dead time of ∼1 ms and total shot volume of ∼100 μl. Excitation was provided by a 535 nm fixed-wavelength LED light source with a 550 nm short pass cut-off filter (Applied Photophysics) and emission was monitored using a 570 nm long pass cut-off filter (Newport Optics). Accounting for all possible contributions from protein/DNA storage buffers, following equal volume mixing, the final reaction buffer conditions were as follows: 20 mM Tris (pH 8.0), 40 mM NaCl, 75 mM K-glutamate, 10 mM MgCl2, 5 μM ZnCl2, 20 μM EDTA, 5% (v/v) glycerol, 1 mM DTT, and 0.1 mg/ml BSA. For all experiments presented, at least two independent preparations of Mtb holo were used. Within each holoenzyme preparation, multiple technical replicates were combined to measure a ‘shot average’. The shot averages from each holoenzyme preparation were weighted equally in determining the resultant averages and errors. Where applicable, error bars represent the standard error of the mean and statistical significance (P-value < 0.05) was determined using a two-sample t-Test with a 95% confidence interval (Origin Software).

Open complex formation

Prior to data acquisition, Eco holo or Mtb holo ± factors were incubated at 37°C for 10 min. Experiments were conducted by equal volume mixing of 2 nM rrnAP3 promoter Cy3-labelled DNA with 100 nM Eco holo or Mtb holo ± 4 μM RbpA, 2 μM CarD, or 4 μM RbpA and 2 μM CarD. Thus, final concentrations upon equal mixing were 1 nM DNA, 50 nM RNAP, 2 μM RbpA and 1 μM CarD. Data was collected for 250 s with logarithmic sampling over 2500 points. Each protein condition is represented by the average of at least five shots and plotted as fold-change over DNA alone according to the formula: (F – Fo)/Fo, where Fo is the buffer subtracted value for DNA alone and F is the buffer subtracted signal for DNA mixed with protein. t1/2open values were calculated as the time at which it takes to reach half the maximal fluorescence amplitude. For RPo formation experiments conducted in the presence of competitor, reaction conditions are identical as above except that 50 μg/mL salmon-sperm DNA was pre-incubated with 2 nM rrnAP3 promoter Cy3-labelled DNA.

Dissociation from open complex

100 nM Eco holo or Mtb holo ± 4 μM RbpA, 2 μM CarD, or 4 μM RbpA and 2 μM CarD was incubated with 2 nM rrnAP3 promoter Cy3-labelled DNA at 37°C for 10 min. Dissociation of RNAP-promoter bound complexes was measured by subsequent equal volume mixing with various heparin or salmon-sperm DNA concentrations. Real-time traces were collected for 1000 sec on a logarithmic timescale, sampling either 2500 or 5000 points and were normalized to obtain the percentage of the initial signal remaining as a function of time with the formula: [(F – Fo)/(Fstart – Fo)] × 100, where F is the buffer subtracted value for DNA alone, F is the buffer subtracted signal, and Fstart is the buffer subtracted signal at the beginning of the trace. This normalization did not affect the fitted fractional amplitudes or kinetics relative to the un-normalized buffer subtracted data itself (i.e. F, data not shown). Dissociation curves were fit to a sum of exponentials (two and three exponentials for salmon-sperm DNA and heparin experiments, respectively) using ProData Viewer (Applied Photophysics).

Promoter escape of open complexes under ‘single-round’ conditions

Eco holo or Mtb holo ± factors were pre-incubated with rrnAP3 DNA as in dissociation experiments and underwent equal volume mixing with various NTP (Thermo Fisher Scientific, catalogue #R0481) concentrations (50–2000 μM after mixing) and 50 μg/ml salmon-sperm DNA. Each condition is represented by at least a 5 shot average where data was acquired and normalized as in dissociation experiments. The fraction of inactive complexes, defined as those that didn’t escape/dissociate from the DNA, was determined by the final fluorescence value following data normalization (Finactive). At NTP concentrations of 50–100 μM, Mtb holo ± factors and Eco holo were fit to a sum of two and four exponential functions, respectively. At NTP concentrations higher than 100 μM, all traces required a sum of four exponentials for a reasonable fit. t1/2escape values, as determined by the time needed to reach 50% of the signal difference between Fstart and Finactive were calculated for conditions higher than 100 μM NTPs, where dissociation of Mtb holo RPo was minimal relative to escape. The time required to reach the peak fluorescence amplitude following initial mixing (time-to-peak) was calculated by fitting each NTP-dependent promoter escape trace to a fifth order polynomial function from 0.0025 to 3 s (Origin Software). From the fits, the maximal peak amplitude was identified, and its corresponding time was obtained.

NTP subset experiments

NTP subset experiments were performed as in single-turnover promoter escape experiments except that select individual NTPs were added at a 1 mM final concentration to generate the RPitc combinations outlined in Figure 4B. A chain terminating NTP analogue (3’-O-methyl-CTP, TriLink Biotechnologies, catalogue #N-1057) was used to generate at max a 5-mer RNA transcript. Conditions with 3’-O-methyl-CTP added are represented by a three shot average, whereas all other conditions are represented by at least a five shot average. Data was acquired for 3600 s to better obtain an estimate of the end point value for the percentage of signal remaining.

Figure 4.

Transient peak signal present at high NTP concentrations is due to initial transcribing complexes. (A) Comparison of dissociation and promoter escape for Mtb. Mtb holo RPo mixed with 25 μg/ml salmon-sperm DNA, 50 μM NTPs, and 1000 μM NTPs (concentrations following equal volume mixing). Inset shows the percent increase above the starting fluorescence signal plotted from 0.001 to 5 s. (B) NTP-subset experiments for Mtb. Mtb holo RPo mixed with salmon-sperm DNA only (RPo, red) or with salmon-sperm DNA and initiating nucleotide (GTP, RPitc1, blue), nucleotides sufficient for formation of a 4-mer (G/UTP, RPitc4, yellow), nucleotides sufficient for formation of a 5-mer (G/UTP and 3’-O-methyl-CTP, RPitc5, brown), nucleotides sufficient for formation of a 9-mer (G/U/CTP, RPitc9, green) or all four nucleotides to allow escape (RPe, purple). NTP additions yielded a final concentration of 1000 μM for each NTP following equal volume mixing. Inset depicts peak signal plotted from 0.001 to 100 s with the initially transcribing sequence of the non-template strand listed above.

Time-to-peak kinetic simulations

Simulations were performed using KinTeK Global explorer (42). A three state, irreversible model (RPo→RPitc→RPe) was used to model the promoter escape process to assess the time-to-peak and amplitude changes corresponding to an individual net entry (kitc) or exit rate (kescape) from initial transcribing intermediates. RPitc was assigned a 10% higher observable fluorescence value in comparison to the starting RPo signal, whereas RPe lacked an observable fluorescence.

RESULTS

CarD and RbpA increase the rate of Mtb RNAP open complex formation

Use of a fluorescently labelled DNA construct containing the Mtb rrnAP3 promoter (see SI – Extended Methods) permitted us to monitor the events of transcription initiation in real-time (outlined in Figure 1A). Mixing Mtb and Eco holo with rrnAP3 promoter DNA (Materials and Methods) resulted in a rapid increase in the fluorescent signal as the duplex DNA was unwound around the position of the Cy3 fluorophore, indicative of RPo formation (Figure 2). Addition of RNAP was required to observe this fluorescent signal, as only mixing CarD and RbpA against rrnAP3 promoter DNA resulted in no change relative to the DNA alone baseline (Supplementary Figure S1). Under these reaction conditions, which were optimized to saturate the final equilibrium fluorescence fold-change value (see SI – Extended Methods), we were unable to observe factor-dependent effects on RPo stability. However, we were still able to extract kinetic information by calculating the time required to reach half the final fluorescence amplitude (t1/2open). When saturating amounts of CarD (1 μM, t1/2open = 3.2 ± 0.3 s) and RbpA (2 μM, t1/2open = 2.1 ± 0.2 s) were pre-incubated with Mtb holo and mixed with DNA, an increase in the rate of RPo equilibration relative to Mtb holo alone (t1/2open = 8.1 ± 1.4 s) was observed. Additionally, by pre-incubating both CarD and RbpA with Mtb holo (t1/2open = 1.28 ± 0.10 s), similar kinetics to that of Eco holo without the addition of factor (t1/2open = 1.12 ± 0.04 s) were obtained. These kinetic results are consistent with our previous observations obtained with Mbo holo under different reaction conditions (14,21). Furthermore, as the equilibrium fluorescence in these experiments was the same in the presence and absence of factors, direct effects from factor binding on the fluorescent signal can be excluded. This result is consistent with the structure of the factor-bound complexes that show the factors to be located on the other side of the DNA bubble compared to the Cy3-fluorophore (Figure 1C). Lastly, both CarD and RbpA display no kinetic effect on RPo formation when pre-incubated with Eco holo, demonstrating RNAP specificity (Supplementary Figure S2). For MtbCarD, this specificity is consistent with the known lineage-specific determinants of RID/RNAP interactions (43). For MtbRbpA, this specificity has previously been reported using a hybrid RNAP holoenzyme containing Eco core subunits and Mtb σA (44). As both CarD and RbpA exhibit effects on the kinetics of Mtb RPo formation, we hypothesized that they will also have an effect on Mtb promoter escape, given the additional RNAP and DNA contacts that will likely need to be broken for the RNAP to translocate away from the promoter.

Figure 2.

Real-time traces of RPo formation as indicated by the increase in fluorescence relative to rrnAP3 DNA alone. Final concentrations are 50 nM Eco holo or Mtb holo with 1 μM CarD, 2 μM RbpA, or 1 μM CarD and 2 μM RbpA mixed with 1 nM rrnAP3 DNA. An average of five independent shots per condition is plotted as fold-change over DNA alone.

Establishing single-turnover conditions: Heparin strips Mtb initiation complexes

To monitor promoter escape kinetics, we titrated NTPs to pre-formed Mtb RPo, where the initial signal was now what was previously observed as the final equilibrium fluorescence in the RPo formation experiments. In the absence of a competitor (i.e. RNAP trap), multiphasic kinetic behaviour was observed, where more than half of the initial signal remained 1000 s after NTP addition (Supplementary Figure S3). These results indicated potential rebinding to the promoter and subsequent RPo formation by dissociated/escaped RNAPs and/or the presence of moribund complexes. To remove the possibility of promoter rebinding, we evaluated multiple competitors to promote single-turnover conditions. For in vitro studies of E. coli transcription, the polyanion heparin is commonly used to prevent (re)binding of free RNAP to promoter DNA (45). However, in titrating heparin against pre-formed Mtb RPo (Materials and Methods), we were unable to find a heparin concentration at which the dissociation kinetics saturated (see SI – Extended Methods, Figure 3A, Supplementary Figures S4-S5), suggesting that heparin may be actively removing bound complexes (46). Alternatively, the use of a non-specific DNA competitor (salmon-sperm DNA), led to saturable dissociation kinetics (see SI – Extended Methods, Figure 3B, Supplementary Figure S5) and prevented initial binding to DNA quantitatively (see Materials and Methods, Supplementary Figure S6). This data supports that salmon-sperm DNA can be used to ensure single-turnover conditions for the study of NTP-dependent promoter escape of Mtb holo without actively promoting dissociation. For the reasons delineated above, we emphasize that for RNAP-promoter systems where there is an appreciable population of bound, but not open, complexes at equilibrium, heparin should be avoided as a competitor.

Figure 3.

Dissociation from Mtb RPo as a function of heparin and salmon-sperm DNA concentration. 50 nM Mtb holo pre-incubated with 1 nM rrnAP3 DNA titrated against various competitor concentrations (concentrations listed are following equal volume mixing). Data is plotted as percent of signal remaining at a given time. Averages of at least three shots are depicted. Black lines indicate fits to a sum of three exponentials for (A) heparin and a sum of two exponentials for (B) salmon-sperm DNA. Insets depict competitor titrations on a linear time-scale.

Identification of a transient peak signal at high NTP concentrations

As both escape (RPo→RPitc→RPe) and dissociation (RPo→RPc→R+P) will result in a reduction of the fluorescence signal, it becomes essential that these two processes occur on different timescales to allow an accurate assessment of escape kinetics. Mixing Mtb holo rrnAP3 promoter complexes with a low concentration of NTPs (50 μM) under single-turnover conditions yielded quenching of the fluorescent signal over time. However, this trace displayed very similar kinetics to the dissociation alone, no NTP, condition (Figure 4A). When we fit both the 0 and 50 μM NTP traces to a sum of two exponentials (Supplementary Figure S7A), the decay signals were dominated by the slower observed rate, mainly reporting on dissociative events (Supplementary Figure S8). With 100 μM NTPs, the slow observed rate no longer dominated the observed fluorescence amplitude (Supplementary Figure S8), suggesting that under these NTP concentrations promoter escape was now responsible for an appreciable fraction of the signal. Increasing to 1 mM NTPs led to ∼50-fold faster quenching kinetics compared to those observed without NTPs (Figure 4A), indicating that these high NTP concentrations are required for deconvolution of dissociation from promoter escape kinetics for Mtb holo.

In the presence of 1 mM NTPs, we unexpectedly observed a transient increase in fluorescence within the first second followed by the expected decay (Figure 4A, inset). These traces required at least four exponentials to obtain a reasonable fit (Supplementary Figure S7B). This peak was also present in both low (50 μM) and high (1 mM) NTP experiments with Eco holo (Supplementary Figure S9). Previous studies of Eco promoter escape using the same fluorescence readout described here with manual mixing techniques did not report the presence of a transient signal at 100 μM NTPs and were able to fit the escape traces to a sum of two exponentials (39,47). This discrepancy can be explained by the reduced dead-time of our stopped-flow assay (∼1 ms) compared to manual mixing techniques (>2 s). In agreement with these previous studies, only a two exponential fit was required to fit the 50 μM NTP condition starting at 2.5 s, while a four exponential fit was required when starting the fit at 1 ms for both 50 μM and 1 mM concentrations (Supplementary Figure S9). As this increase in the fluorescence becomes more prominent at high NTP concentrations and also occurs on similar timescales previously shown to correspond to short (3–10 mer) RNA synthesis (48) we hypothesized that the transient peak was due to a build-up of initiation intermediates, specifically, a subset of initially transcribing complexes (RPitc).

The transient peak reports on initially transcribing complexes with RNA-DNA hybrids as short as 4 nt

A transient peak observed in the presence of NTPs could be generated by a population of RPitc intermediates or via a further stabilization of RPo. Since initiating nucleotides can stabilize the open DNA bubble (49–51), we asked whether mixing 1 mM GTP (Materials and Methods) with pre-formed Mtb RPo led to an increase in signal. Under these reaction conditions, no fluorescence peak was detected as might be expected given the final equilibrium value in the absence of the initiating nucleotide was already saturated (Figure 2), although GTP binding could be observed via a reduction in the rate of dissociation (Figure 4B). This result showed that the transient peak was not due to an increase of RPo stability. In contrast, when a maximal transcript length of 4 nucleotides (RPitc4) was permitted by including 1 mM G/UTP, an increase in the fluorescence signal was observed (Figure 4B). No increase in fluorescence was observed by adding either G/ATP or G/CTP, indicating that the peak observed with the addition of G/UTP is not due to nucleotide mis-incorporation (Supplementary Figure S10). Addition of G/UTP and 3’-O-methyl-CTP, a NTP-analogue that has been shown to be effective in terminating the growth of the RNA chain in studies of Eco transcription (52) permitted us to monitor a 5-mer RNA product (RPitc5). Relative to RPitc4, RPitc5 displayed a further increase in the magnitude of the peak signal (Figure 4B). When a 9-mer (RPitc9) RNA product was permitted by including 1 mM G/U/CTP, the magnitude of the peak signal was between that observed in the RPitc4 and RPitc5 conditions, suggesting that the maximal fluorescence amplitude corresponds to a population of initial transcribing complexes containing RNA-DNA hybrids between 4 and 9 nt in length (Figure 4B). In RPitc4, RPitc5 and RPitc9 conditions, the same peak entry kinetics as compared to those in the presence of all 4 NTPs (RPe) were observed (Figure 4B Inset). Performing identical NTP subset experiments with Eco holo displayed a similar trend to that of Mtb with respect to the RNA-DNA hybrid lengths required to observe a peak signal (Supplementary Figure S11), indicating that the increase in fluorescence reporting on RPitc intermediates is not specific to Mtb RNAP. Combined, these results indicate a further enhancement of the fluorescence signal (distinct from that of RPo) upon progression through the initial transcribing sequence for initially transcribing complexes with transcripts with as short 4 nt in length. Thus, the stopped-flow traces permit inferences regarding the rates of both initial nucleotide incorporation (through an evaluation of the peak signal) and promoter escape rates (through an evaluation of the fluorescence decay).

Mtb and Eco display different promoter escape kinetics

As rates of promoter escape are often anti-correlated with RPo half-life, it is generally assumed that for a highly stable RPo, the rate of promoter escape becomes limiting (53). As Eco forms a more stable RPo on the rrnAP3 promoter than Mtb (14,15,21), we hypothesized that the rate of escape for Eco will be delayed relative to Mtb. Due to the complexity of the escape traces obtained at the 1 mM NTP condition (i.e. four observed rates (Supplementary Figures S7B, S9B) and inactive complexes that don’t escape the promoter (Materials and Methods)), we chose to use t1/2escape values (i.e. the time it takes for half of the active complexes to escape) to provide a semi-quantitative description of the escape kinetics. This type of analysis has been used previously to determine a single rate describing the overall promoter escape kinetics (kEobs defined as 1/t1/2escape) (46,54,55). We emphasize that here kEobs is an average observed rate that reports on all the combined molecular processes involved in escape, including the rates of initial nucleotide incorporation, and should not be interpreted as an independent rate constant. As expected, Eco (t1/2escape = 13.2 ± 0.3 s) exhibited roughly a 4-fold increase in the t1/2escape value relative to Mtb (t1/2escape = 3.7 ± 0.2 s), indicating slower overall escape kinetics (Figure 5). However, comparing the RPitc4, RPitc5, RPitc9, and RPe conditions, it is clear that entry into RPitc intermediates always occurred on faster timescales for Eco holo (Figure 5, Supplementary Figure S12), indicating faster initial nucleotide incorporation kinetics. As Eco spends more time in the ensemble of RPitc intermediate states than Mtb under NTP conditions that permit escape (Figure 5 Inset), these results indicate that the exit from RPitc intermediates for Eco is more rate-limiting compared to Mtb, leading to the slower kEobs relative to Mtb.

Figure 5.

Eco displays faster initial nucleotide incorporation but slower promoter escape kinetics than Mtb. 50 nM Mtb (green) or Eco holo (orange) pre-incubated with 1 nM rrnAP3 DNA mixed with 25 μg/ml salmon-sperm DNA and 1 mM NTPs. Inset shows the percent increase above the starting fluorescence signal plotted between 0.001 and 5 s.

CarD and RbpA slow the rate of Mtb promoter escape

As both CarD and RbpA act to stabilize Mtb RPo (14,15,21), we hypothesized that they should lead to a reduction in the rate of escape. To assess the effects of CarD and RbpA on the overall promoter escape kinetics, NTP titrations (Materials and Methods) were performed for Mtb holo in the absence and presence of saturating concentrations of each factor individually as well as with both factors combined (Figure 6A, Supplementary Figure S13). Increasing NTP concentrations led to a decrease in the t1/2escape value (Supplementary Figure S14A) and a corresponding increase in the fractional amplitude of the faster decay rate (Supplementary Figure S8) in all cases. The t1/2 escape analysis was limited to conditions of 250 μM NTPs or higher, where the remaining signal was nearly saturated (Supplementary Figure S14B) and the effects of NTP-independent dissociation of RNAP on the observed fluorescence signal were minimal. CarD, RbpA, and both factors combined all led to a decrease in the overall rate of escape. Specifically, in the presence of 1 mM NTPs the factors exhibited 2.9-, 1.4- and 2.4-fold increases in the t1/2escape value respectively (Figure 6B and C). Additionally, large changes in the signal remaining were observed in the presence of CarD and RbpA (Figure 6B Inset, Supplementary Figure S14B). As this remaining signal is dependent upon both the NTP concentration and time of data acquisition (Supplementary Figure S14B-C), we only used conditions of saturating NTPs to ensure measurement of a signal representative of the reaction's true endpoint. This remaining signal was taken as a measurement of the fraction of initiation complexes that were unable to either dissociate or escape the promoter (i.e. the inactive fraction) as CarD and RbpA in the absence of RNAP do not alter the observed fluorescence signal (Supplementary Figure S1). These factor-dependent increases in inactive fraction correlate with factor-dependent increases in RPo stability (21) where CarD, RbpA, and CarD/RbpA combined led to 2-, 1.8- and 2.8-fold increases in inactive fraction respectively compared to Mtb holo alone (Figure 6D).

Figure 6.

CarD and RbpA effects on Mtb promoter escape kinetics and inactive fraction. (A) NTP titrations for Mtb holo. NTP conditions range from 0 to 2 mM after equal volume mixing. Inset presents the transient peak signal as a percentage increase above the initial fluorescence value plotted from 0.001 to 5 s. Increasing the NTP concentration decreases the time-to-peak, as indicated by the dotted black line and arrow. (B) Overlay of escape traces conducted at 1 mM NTPs for Mtb holo ± factors. Inset depicts data on a linear time-scale. (C) Quantification of t1/2escape values and (D) inactive fraction as calculated from the data in (B). Error bars represent standard errors of the mean.

The NTP-dependence of initial nucleotide incorporation as described by a three-state model

NTP titrations for Mtb holo and in the presence of factors yielded traces where the peak time and shape corresponding to RPitc intermediates was dependent upon NTP concentration (Figure 6A Inset, Supplementary Figure S13 Insets). Specifically, the time required to reach the maximal peak amplitude (time-to-peak) decreased and the maximal peak amplitude increased as a function of NTP concentration (Figure 7A and B). To assess the NTP-dependence of the observed peak signal, we performed time-to-peak kinetic simulations using a three-state irreversible model (RPo→RPitc→RPe) where the fluorescence increase is due to the population of molecules in the RPitc state (Materials and Methods). These simulated traces can be used to test hypotheses regarding the NTP dependence of both the entry and exit rates from these intermediate states by comparing trends in the simulated traces to those observed experimentally. The first predictions of this model are that the net rate constant of RPitc entry (i.e. RPo→RPitc, kitc) must be faster than the net rate constant of RPitc exit (i.e. RPitc→RPe, kescape) to observe a peak signal (Supplementary Figure S15), and that increasing the net entry and exit rate constants by equal factors (i.e. both rates depend linearly on [NTP]) leads to a decrease in time-to-peak without a concomitant change in peak amplitude (Supplementary Figure S15B). In contrast, an NTP-dependent increase in the peak amplitude was observed for all conditions (Figure 7B), which could be simulated by either increasing kitc or decreasing kescape while removing the NTP dependence from the other net rate constant (Figure 7C and D, Supplementary Figures S16, S17). Of these models, only increasing kitc led to the experimentally observed decrease in time-to-peak upon increasing [NTP] (Figure 7C, Supplementary Figure S16). Along these lines, we hypothesize that the rate-limiting step in Mtb promoter escape (kescape) is an NTP-independent structural conversion from a late stage RPitc intermediate to RPe, such as σA repositioning or dissociation from the core RNAP (12).

Figure 7.

NTP dependence of the the peak signal. (A) Time-to-peak and (B) corresponding peak amplitudes as a function of NTP concentration for Mtb holo with and with-out factors. Error bars represent standard errors of the mean. (C) Time-to-peak simulations where the net exit rate is fixed while increasing the net entry rate into RPitc and (D) the net entry rate is fixed while decreasing the net exit rate out-of RPitc (see Supplementary Figures S16 and S17). Here, the darkening trend represents larger changes from the starting value and increases in the peak amplitude.

CarD and RbpA have differential effects on RPitc intermediates

To determine possible mechanisms by which CarD and RbpA slow promoter escape, we analysed factor-dependent changes in the peak signal corresponding to RPitc intermediates. In the presence of all NTPs at the 1 mM condition (Figure 8A), CarD had the greatest effect on the peak signal with the slowest time-to-peak (Figure 7A) and the highest peak amplitude (Figure 7B). In the case of RbpA, a small, but significant (P-value = 0.042) increase in the time-to-peak relative to Mtb holo alone was observed (Figure 7A), with no statistically significant decrease in the peak amplitude (Figure 7B). In the presence of both factors, the time-to-peak increase was intermediate to that observed with CarD and RbpA separately (Figure 7A) and no significant change in the peak amplitude relative to Mtb holo alone was observed (Figure 7B). In summary, while the presence of factors always led to an increase in the time-to-peak, only CarD in the absence of RbpA led to an increase in the peak amplitude.

Figure 8.

CarD and RbpA effects on RPitc intermediates. (A) Overlay of peak signal obtained at 1 mM NTPs for Mtb holo ± factors plotted as a percentage increase above the initial fluorescence value. (B) Simulations of factor dependent effects on peak signal. The combined RPo and RPitc signals are plotted as a percent increase above the initial fluorescence signal. Factor dependent effects on kitc and kescape were adjusted according to the experimentally determined fold changes in the time-to-peak relative to Mtb holo alone. CarD (blue) is modeled by a 0.4-fold decrease in the net exit rate, RbpA (red) is modeled by a 0.9-fold decrease in the net entry rate. In the case of both factors, simulations adjusting both the net entry and exit rates as in the individual factor simulations were unable to reproduce the experimentally observed results in the time-to-peak and amplitude (light purple). Simulating a 0.75-fold decrease in both the net entry and exit rates, yielded a similar amplitude and time-to-peak as observed experimentally in the presence of both factors (dark purple).

To explain these trends, we hypothesized that the factors affect the kinetics of RPitc formation and escape differently. Using the three-state model of initial nucleotide incorporation, we asked whether we could replicate the experimentally observed peak signals in the presence of all NTPs and CarD and RbpA (Figure 8A). We found that reducing kescape by the experimental fold-change in time-to-peak in the presence of CarD (0.4x) reproduced the CarD trend and that reducing kitc by the experimental fold-change in time-to-peak in the presence of RbpA (0.9x) reproduced the RbpA data (Figure 8B). While the experimental trace in the presence of both factors is consistent with the slowing of both kitc and kescape, simply changing both net rates using the same fold changes as in the individual conditions for each factor did not replicate the peak signal (Figure 8B). However, decreasing both net rate constants equally (0.75x), yielded a time-to-peak and amplitude as observed experimentally (Figure 8B) suggesting that each factor is able to modulate either directly or indirectly the magnitude of the effect of the other factor. Based on this analysis, we conclude that CarD slows kescape and RbpA slows kitc.

DISCUSSION

Stopped-flow rapid mixing permits detection of RPitc intermediates of bacterial transcription

The experiments described here were designed to measure the rates of promoter escape. Based on our previous use of this assay to study RPo formation, we expected to observe only a fluorescence decay after the addition of NTPs as the polymerase moved downstream, permitting the Cy3-labelled region of the promoter to re-anneal around the transcription start site (TSS). In contrast, the traces exhibited a transient increase in fluorescence, which was detected on timescales not permitted with manual mixing techniques. Experiments with NTP-subsets revealed that the signal increase was due to initially transcribing intermediates with RNA–DNA hybrids as short as 4 nt, with a maximal fluorescence signal between 4 and 9 nt in length for both Mtb and Eco holo (Figure 4B, Supplementary Figure S11). Given the similarities in peak formation for these holoenzymes, we suggest that this technique can be broadly applicable in measuring initial nucleotide incorporation and escape kinetics for any bacterial RNAP. Additionally, as differences in the fluorescence signal were detected between RPitc4 and RPitc5 conditions (Figure 4B, Supplementary Figure S11), this technique permits detection of initial transcribing intermediates with nucleotide resolution at least within this range of RNA lengths. As escape from RPitc intermediates is often rate limiting for transcription initiation (53), measurements of the kinetics and stabilities of these intermediates are of crucial importance in understanding gene expression mechanisms.

A three-state model to explain Mtb RPitc intermediates

To explain the kinetics of RPitc formation and escape for Mtb holo, a sequential, three-state model where kitc (entry into RPitc intermediates) is NTP-dependent and the rate-limiting kescape (exit from RPitc intermediates) is NTP-independent was able to replicate the observed trends in the peak signal as a function of NTP concentration (Figures 6A, 7A–C, Supplementary Figures S13, S16). An NTP-independent rate has been previously proposed from single-molecule studies evaluating the rate of exit from Eco RPitc6 paused-states on a consensus lac promoter, where it was suggested that pausing is not thermodynamically regulated by pre- and post-translocated states but rather is under kinetic control dictated by region 3.2 of Eco σ70 (σ703.2) displacement from the active site (12).

While this sequential model proved useful to describe the entry and exit kinetics of Mtb RPitc intermediates, we discourage the interpretation of the peak signal as a direct reporter of abortive RNA synthesis. As transitions within the ensemble of initial transcribing complexes have been proposed to occur both with and without release of abortive RNAs (12), we only interpret the peak signal as a reporter of RPitc intermediates. Furthermore, we emphasize that this model is an oversimplification when evaluating the entire promoter escape process as it does not predict the phenomenon of promoter-bound inactive fractions. For Mtb holo, the signal remaining after dissociation in the absence of NTPs was nearly identical to that remaining after promoter escape in the presence of 1 mM NTPs (Figure 4, Supplementary Figure S14C), suggesting a mechanism where a subpopulation of unproductive or ‘moribund’ RPo is not able to either dissociate or exhibit NTP-dependent escape on the time-scale of the experiment (9). This complicated kinetic behaviour is consistent with the existence of branched pathways that evolve on distinct time-scales (13). In addition, branched pathways due to inactivating processes like RNAP pausing and/or backtracking may exist during initial rounds of NTP incorporation (10–12), that could arise from the potential pause sequences within the initially transcribed sequence [(T/G) at +3/4 and (C/G) at +8/9] of the rrnAP3 promoter (47). As CarD and RbpA lead to a higher population of non-escaping complexes (Figure 6D), a potential regulatory mechanism resulting in repression, future work will focus on the sequence dependence of inactivation as well as the susceptibility of inactive complexes to cleavage factors. If inactivation of complexes arises from an increase in the backtracked fraction of initiation complexes, one expects this fraction of the inactive complexes to be reduced in the presence of MtbGre, an essential cleavage factor in mycobacteria (56), as is the case with GreB in studies of Eco transcription (10,55,57).

CarD and RbpA effects on RPitc kinetics correlate with the location of their initiation complex contacts

CarD stabilizes RPo by interacting with the upstream edge of the DNA bubble (Figure 1B-C) (28,30). These contacts do not need to be broken during initial nucleotide incorporation where RNAP presumably remains in a fixed position relative to upstream DNA (3,4). This is consistent with our model of CarD slowing kescape while not affecting the NTP-dependent net rate constant, kitc. RbpA also stabilizes RPo by interacting with upstream promoter DNA, however, unlike CarD, RbpA also directly contacts σ (15,19,29,30,35). Specifically, the SID of RbpA contacts region 1.2, the nonconserved region (NCR), and region 2 of σA while the NTT of RbpA interacts with σA3.2 (Figure 1B and C). In E. coli, σ703.2 serves as an important regulatory determinant for abortive initiation as steric clashes within the RNA exit channel can lead to the release of abortive transcripts or to σ dissociation and escape (58–60). Furthermore, Eco σ703.2 can stimulate phosphodiester bond formation via positioning of the template DNA strand (59). As the NTT of RbpA makes direct contacts with σA3.2, we hypothesize that the NTT affects the mobility of the template DNA strand during initial rounds of nucleotide incorporation (29). This is consistent with previous work suggesting an RbpA effect on the selectivity of initiating substrates (61). Thus, the RbpA effects of slowing the NTP-dependent kitc and overall escape kinetics on rrnAP3 measured here seem not to be driven by overcoming a higher energy barrier to escape provided by a RbpA-dependent increase in RPo stability, but rather effects on spatial position of the promoter DNA within the active site.

The kinetic behaviour in the presence of both factors is not predicted from the cumulative effects of each factor alone. Qualitatively, the effect of RbpA on k is potentiated in the presence of CarD and the effect of CarD on kescape is mitigated in the presence of RbpA (Figure 8B). These results are consistent with the observed concomitant binding and cooperativity we observe in the kinetics of RPo formation (21), and further suggests linkage between the factors, perhaps via the induction of different polymerase conformations. A complete mechanistic understanding will require future work with mutant factors to disentangle the domains responsible for factor-factor effects.

RbpA and CarD have the potential to both activate and repress transcription

Prior work has focused on quantifying transcription initiation kinetics on the rrnAP3 promoter up to RPo formation. These experiments indicated that CarD serves to activate transcription by increasing the rate of DNA opening and slowing the rate of bubble collapse, thereby enhancing RPo stability (14–16). RbpA, on the other hand, activates transcription by mainly increasing the rate of DNA opening, with little effect on the rate of closing (15,21,25). However, numerous transcription factors have been shown to affect both RPo formation and promoter escape (37,62). For these factors, the regulatory outcome of factor binding (i.e. activation or repression) cannot be judged solely from effects on RPo stability (53). In addition, RNA-sequencing data for RbpA (25) suggests that while functioning as an activator on a subset of promoters, RbpA is also able to repress transcription on a different promoter subset. This type of behavior has been well-established for the stringent response regulators E. coli DksA and ppGpp where the regulatory outcome depends upon the basal RPo stability of a given promoter (63–65). CarD, RbpA, nor DksA are recruited to initiation complexes via sequence-dependent DNA contacts, suggesting indirect mechanisms for promoter-specific regulation.

We asked if a model based on our knowledge of the kinetics of RPo formation and promoter escape (depicted in Figure 9A and discussed in SI – Extended Methods) would predict the ability of RbpA to either activate or repress transcription. This model assumes that RbpA increases the rate of RPo formation and slows the rate of promoter escape on every promoter and that the resulting regulatory outcome depends on the limiting steps of a particular promoter's basal initiation kinetics. Using the calculated fold-changes in the rate of escape (kEobs as determined by the t1/2escape analysis) and RPo equilibration in the presence of RbpA on the rrnAP3 promoter, we calculated the rate of transcription initiation for promoters with varying basal kinetics using an online resource we have developed (https://github.com/egalburt/transcript-flux-calculator) (36) (SI – Extended Methods). Consistent with our hypothesis, these calculations reveal promoter-dependent differential regulation in transcript production rates (Figure 9B). Here, RbpA is predicted to repress transcription on promoters that are rate-limited at the escape step (i.e. form a stable RPo) and activate transcription on promoters that are rate-limited by RPo formation (i.e. rrnAP3). Even though RbpA only modestly slows escape kinetics, this effect on escape leads to predicted transcriptional repression across 3 orders of magnitude of kopen and kescape basal rates. Furthermore, in comparing the calculated flux ratios to previously obtained RNA-sequencing data (25) showed that that modeled effects of RbpA determined by our in vitro kinetic analyses presented here can account for ∼70% of the predicted regulatory outcome determined in vivo, which could be increased to roughly 90% by modeling the kinetic effects of both RbpA and CarD combined (Supplementary Figure S18). As the presence of both factors on a single initiation complex may better represent physiological conditions and thus our RNA-sequencing data, the observed agreement between our kinetic modeling and the in vivo regulation suggests that that this simple model of regulation driven by a promoter's basal kinetics is sufficient to explain the majority of RbpA-dependent gene regulation. Thus, we propose that the promoter-dependent repression in the RNA-sequencing data is driven by the slowing of escape outweighing the activating effects of increasing the opening rate. Future kinetic studies on a variety of differentially regulated promoters identified in our RNA-sequencing analysis will directly test this hypothesis.

Figure 9.

Differential regulation in flux is dependent upon the basal rates of initiation. (A) Three-state model for calculations of steady-state flux where CarD and RbpA either increase (green arrow) or decrease (red line) specific rate constants. Factor-dependent fold-changes in transcript flux show both activation (green) and repression (red) as a function of the basal opening and escape rates for (B) RbpA and (C) CarD. Promoters that are rate-limited at escape (i.e. fast kopen, slow kescape) are predicted to be repressed, whereas promoters that are rate-limited at opening (i.e. slow kopen, fast kescape) are predicted to be activated.

In the case of CarD, we have previously modeled a regulatory mechanism that assumed CarD slows the rate of escape and the rate of bubble collapse similarly (36). As we now have concrete evidence that CarD slows promoter escape, we used our experimentally derived fold-changes in the rate of escape and RPo equilibration (SI – Extended Methods) and found that like RbpA, CarD also has the potential to both activate and repress transcription in a promoter specific manner (Figure 9C). Promoter specific motifs upstream of the TSS that confer stronger RNAP-promoter interactions, can lead to increased abortive cycling and inhibited full-length transcript synthesis (66,67). On such promoters where RPo stability is high and escape is rate-limiting, our theoretical predictions suggest that CarD and RbpA would reduce the steady-state rate of transcript production. Additionally, CarD and RbpA both increase the inactive fraction of initiation complexes (Figure 6D), which would act to reduce transcript production directly, stochastically, and via a non-steady-state mechanism by blocking RNAP binding to a promoter for a long time relative to the rate of initiation. As a consequence, one might expect this mechanism to lead to fluctuations or bursts in the instantaneous rate of transcription initiation (68) and possibly account for the small fraction of genes predicted to be repressed in vivo by RbpA that fell outside the limits of our in vitro predictions that only accounted for kinetic regulation (Supplementary Figure S18).

DATA AVAILABILITY

Transcript flux calculations can be made using the online resources available at https://github.com/egalburt/transcript-flux-calculator.

Click here for additional data file.