Decision letter: Single-molecule tracking of the transcription cycle by sub-second RNA detection

Abstract

Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods References Decision letter Author response Article and author information Metrics Abstract Transcription is an inherently stochastic, noisy, and multi-step process, in which fluctuations at every step can cause variations in RNA synthesis, and affect physiology and differentiation decisions in otherwise identical cells. However, it has been an experimental challenge to directly link the stochastic events at the promoter to transcript production. Here we established a fast fluorescence in situ hybridization (fastFISH) method that takes advantage of intrinsically unstructured nucleic acid sequences to achieve exceptionally fast rates of specific hybridization (∼10e7 M−1s−1), and allows deterministic detection of single nascent transcripts. Using a prototypical RNA polymerase, we demonstrated the use of fastFISH to measure the kinetic rates of promoter escape, elongation, and termination in one assay at the single-molecule level, at sub-second temporal resolution. The principles of fastFISH design can be used to study stochasticity in gene regulation, to select targets for gene silencing, and to design nucleic acid nanostructures. https://doi.org/10.7554/eLife.01775.001 eLife digest The body produces proteins by transcribing DNA (genes) to make messenger RNA, which is then translated to make a protein. Transcription begins when an enzyme called RNA polymerase binds to the DNA and catalyzes the process by which genetic information from the double helix is copied to a complementary RNA transcript, which subsequently becomes the messenger RNA. Because a living cell usually contains only one or a few copies (alleles) of a given gene, molecular fluctuations play a crucial role in cellular transcription. Therefore, studying transcription kinetics at the level of single molecules may provide critical insights into how cells deal with—or even take advantage of—molecular fluctuations. A number of different single-molecule techniques can be used to follow transcription, but these techniques are often relatively slow compared to transcription in living cells, or they suffer from other problems such as only being able to study one step in the transcription process. Now, Zhang, Revyakin et al. have systematically devised a technique called ‘fastFISH’ that is fast enough to track the production of single RNA molecules directly and instantaneously. FastFISH builds on an existing technique called FISH—short for fluorescence in situ hybridization—in which fluorescent molecules are attached to single strands of DNA or RNA. These single strands pair with specific regions of complementary DNA or RNA molecules, and they can be visualized with a fluorescence microscope. However, conventional FISH is a ‘snap-shot’ technique that is not suitable for making real-time observations under physiological conditions. FastFISH relies on single strands of fluorescently labeled DNA and RNA that bind to complementary strands of DNA or RNA extremely quickly, even under physiological conditions, because they contain only three of the four ‘regular’ nucleotides that make up DNA or RNA. As a proof of principle, Zhang, Revyakin et al. used fastFISH to study the kinetics of transcription by the bacteriophage T7 RNA polymerase and were able to measure multiple stages of the transcription cycle in a single-molecule experimental setup. By allowing each stage of transcription to be tracked in real-time at the level of single-molecules, fastFISH will permit a more in-depth analysis of the factors that regulate how genes are expressed as proteins in our cells. Moreover, the ability to design single-strand probes that bind rapidly to DNA and RNA targets could have many additional applications, including new strategies for more efficient gene silencing. https://doi.org/10.7554/eLife.01775.002 Introduction Transcription is the first, and frequently the most regulated step in the flow of genetic information from DNA to protein. Transcription is a dynamic, multi-step process in which RNA polymerase (RNAP) (i) binds to the promoter to form the closed complex (RPc); (ii) melts the promoter to form the open complex (RPo); (iii) performs several abortive cycles of synthesis and release of short 2–12 nt RNA products as the initial transcribing complex (RPitc); (iv) escapes the promoter; (v) undergoes some promoter-proximal pausing; (vi) forms an elongating complex (RDe) whose processivity can either be interrupted by more pauses or stimulated by trailing RNAPs and elongation factors; and, finally (vii) terminates at the end of the transcription unit (reviewed in DeHaseth et al., 1998; Murakami and Darst, 2003; Cheung and Cramer, 2012). Due to the low copy number of genes in a cell (usually one in prokaryotes and two in eukaryotes), molecular fluctuations at any of the above steps may cause large cell-to-cell variability in the amount of the final RNA transcript produced in populations of otherwise genetically identical cells grown under identical conditions, and thus can affect gene expression, and cell physiology (Ogawa, 1993; Raj and Van Oudenaarden, 2009; Yamanaka, 2009; Gupta et al., 2011; Lionnet and Singer, 2012). Therefore, understanding how molecular fluctuations at different steps of the transcription cycle alter transcriptional outcomes is required to dissect the mechanism of gene regulation. Single-molecule techniques are ideally suited to directly monitor molecular fluctuations in multi-step reactions in real-time, without averaging out their inherent stochasticity (Weiss, 1999), and have provided important insights into the dynamics of transcription, unattainable by conventional ensemble methods (Bai et al., 2006). For instance, single-molecule assays based on optical nanomanipulation have revealed pausing and backtracking of RDe (Neuman et al., 2003; Shaevitz et al., 2003; Galburt et al., 2007) and measured the force and the torque generated by RDe (Wang et al., 1998; Ma et al., 2013). Methods based on magnetic nanomanipulation and single-molecule fluorescence resonance energy transfer have probed conformational changes in DNA and RNAP within RPo, and RPitc (Kapanidis et al., 2006; Revyakin et al., 2006; Tang et al., 2009; Chakraborty et al., 2012; Robb et al., 2013). Finally, single-molecule localization studies have probed the dynamics of the initial promoter search by RNAP (Wang et al., 2012; Friedman et al., 2013). However, most current single-molecule methods focus on only one step of transcription, and are not well suited to relate protein-DNA complex assembly and dynamics during multiple stages of the transcription cycle to RNA production. In addition, most methods do not measure RNA production directly, but rather infer it from changes in DNA conformation or movement of RNAP along the DNA. Most recently, single-molecule tracking of key protein-DNA interactions coupled with detection of the RNA production has been demonstrated in the bacterial (Friedman and Gelles, 2012) and human transcription systems (Revyakin et al., 2012). The former study achieved time resolution for RNA detection on the order of ∼10 s, and thus provided a dynamic, quantitative view of the full transcription cycle of bacterial RNAP. However, time scales of many events in transcription are on the order of ∼1 s, particularly under physiological conditions (for instance, the residence time of transcription factors on DNA (reviewed in Hager et al., 2009), the rate of initiation and promoter clearance by RNAPs (Revyakin et al., 2006; Tang et al., 2009), and the expected time delay between cooperatively elongating RNAPs molecules (Epshtein and Nudler, 2003)). Thus, a real-time method for nascent transcript detection at ∼1 s time scales would significantly enhance our ability to dissect the dynamics of the transcription process, and to correlate stochastic fluctuations at different steps with transcriptional outcomes. Currently, the most sensitive and specific methods for detecting RNA transcripts continue to rely on complementary nucleic acid hybridization. However, oligonucleotide probes typically hybridize several orders of magnitude slower than diffusion under physiological conditions (effective rate constant kon less than 105 M−1 s−1, [Chan et al., 1995 and references therein]) and, as a result, do not permit real-time nascent RNA detection with common single-molecule setups, such as total internal reflection fluorescence (TIRF) microscopy. Here we present a fast fluorescence in situ hybridization method (fastFISH) to detect synthesis of nascent RNA transcripts at sub-second time resolution, at the single-molecule level. The method takes advantage of our finding that single-stranded nucleic acid probes with sequences comprised of just three of the four bases (A, U and C for RNA probes, and A, T, and G for the complementary DNA targets) are intrinsically unstructured and, as a result, hybridize at exceptionally fast rates (∼107 M−1s−1) without compromising sequence specificity. As a proof of concept, we applied fastFISH to probe the production of nascent RNA transcripts by the speedy bacteriophage T7 RNA Polymerase (T7 RNAP) in vitro. Furthermore, by using fastFISH in combination with fluorescently labeled RNAP, we dissected the full T7 RNAP transcription cycle (promoter binding, promoter escape, elongation, and termination). FastFISH can be used to study transcription by multi-subunit prokaryotic and eukaryotic RNA polymerases at the level of stochastic molecular interactions. The rules for generating fastFISH probe-target pairs can also be used for gene silencing, gene profiling and bottom-up assembly of nucleic acid nanostructures (Rothemund, 2006). Results Design and characterization of fastFISH hybridization probes To achieve real-time nascent RNA detection, we set out to find a general rule for designing RNA-probe pairs that hybridize at the fastest possible rates. Hybridization of two single-stranded nucleic acid fragments requires unfolding of the fragments into unstructured coils, which then anneal to form an intermolecular base paired helix (Lima et al., 1992 and references therein). In support of this notion, nucleic acid probes with less stable secondary structures show faster hybridization rates to complementary nucleic acid targets (Lima et al., 1992; Kushon et al., 2001; Wang and Drlica, 2004; Gao et al., 2006; Yilmaz et al., 2006). Likewise, decreasing the length of hairpins in conventional molecular beacons increases their hybridization rates (Tsourkas et al., 2003). Thus, we reasoned that intrinsically unstructured nucleic acid oligonucleotide pairs of optimal lengths should have the fastest hybridization kinetics under physiological conditions without compromising specificity. To systematically examine the propensity of RNA sequences to form base-paired secondary structures, we used the nucleic acid structure prediction tool Mfold (Zuker, 2003) to calculate the free energy of self-folding, ΔG37°C, of randomly selected short RNA sequences (N = 338,417). We chose RNA sequence with length of 19 nt and GC content between 0.4 and 0.6 as representative of a typical oligonucleotide primer. We found that ΔG37°C values were mostly negative and widely distributed (−1.7 ± 1.7 kCal mol−1), indicating that an average random 19-mer RNA sequence is mostly structured (Figure 1A). Next, we closely examined a subset of 19-mer RNA sequences that had positive ΔG37°C (>+2 kCal mol−1, N = 2,768, <1% of the total pool of random sequences), and found that these mostly unstructured sequences were composed predominantly of only three bases A, U, and C (∼84% had only 2 G residues or less, significantly less than the 4.75 G residues expected on average, Figure 1—figure supplement 1). The bias towards the lower G-content in the unstructured sequences was not surprising, because guanine is the most potent in base-paring interactions: it forms a triple hydrogen bond with cytosine and a wobble pair with uracil. We then calculated ΔG37°C for randomly picked 19-mers composed of only A, U, and C (N = 99,777, GC content between 0.4 and 0.6, Figure 1A), and found that these AUC-sequences had mostly positive ΔG37°C, with a much narrower distribution (+2.5 ± 0.6 kCal mol−1, Figure 1A). Analysis of other three-base-derived RNA sequences (AUG, CAG, CUG) indicated that AUC-sequences were unique in their largely positive ΔG37°C (Figure 1A). Calculation of ΔG37°C for random DNA 19-mers (that could be used as complementary probes for RNA targets) indicated that ATG-based and ATC-based DNA 19-mers were also significantly less structured than their four-base counterparts (Figure 1A). Therefore, we hypothesized that the use of the AUC alphabet for RNA targets, and ATG alphabet for DNA probes should allow the fastest annealing reactions under physiological temperature of 37°C. Figure 1 with 3 supplements see all Download asset Open asset Design of fastFISH probe-target pairs. (A) Probability distribution of Mfold-calculated free energies of self-folding of randomly selected, single-stranded 19-mer RNA (left) and DNA (right) oligonucleotides, composed of three or four bases. Results of analysis of three independent sets are shown as ‘+’, ‘x’, and ‘○’. About 100,000 three-letter sequences, and about 300,000 four-letter sequences were analyzed in each set. (B) Lempel–Ziv complexity analysis of three-letter 19-mer oligonucleotides (one set of ∼100,000 AUC sequences), four-letter 19-mer oligonucleotides (one set of ∼300,000 AUGC sequences), and all tiling 19-mers from the exome of the human chromosome 22. (C) Single-molecule measurements of hybridization rates of fastFISH probe-target pairs, and the effect of G-residues in the targets on the rates. Left: schematic of experiment. Cy3-labeled 90-base RNA oligonucleotides containing a single target sequence were immobilized on a glass surface through a biotin moiety at the 3′ end. Atto633-labeled DNA probes were injected into the imaging flow cell, and their hybridization was detected using TIRF/CoSMoS to obtain the probe arrival time Twait. Right: table of RNA target sequences, Mfold-calculated free energies of self-folding of RNA targets (ΔGtarget), DNA probes (ΔGprobe), combined energies of targets and probes, and on-rates calculated based on probe Twait and concentrations. (D) Self-quenching approach to reduce fluorescence background from unbound DNA probes in TIRF imaging of probe-target hybridization. Left: schematic of experiment. A quencher (e.g., Iowa Black FQ) is placed on one end of a DNA probe, and a fluorophore (e.g., Cy3) is placed on the opposite end of the DNA probe. The short persistence length of single-stranded DNA (lo ∼0.8 nm, Smith et al., 1996; Dessinges et al., 2002) enables quenching of Cy3. Upon hybridization to the target, the distance between Cy3 and Iowa Black FQ increases due to the larger lo ∼50 nm of the DNA-RNA duplex, leading to an increase of Cy3 fluorescence. Middle: representative TIRF image and a corresponding three-dimensional plot of target-hybridized F1 probes acquired in the presence of unbound, self-quenched F1 probe at 100 nM. Right: same set of molecules imaged in the presence of unbound, unquenched F1 probe at 100 nM. https://doi.org/10.7554/eLife.01775.003 To ensure that the use of the three-base alphabet did not compromise the specificity of probe-target hybridization, we calculated the textual complexity of randomly picked three-letter 19-mers using the algorithm of Lempel and Ziv (Ziv and Lempel, 1976; Kaspar and Schuster, 1987; Orlov and Potapov, 2004). This algorithm, commonly used in lossless data compression programs, calculates the minimal number of operations required to reconstruct a sequence of symbols by copying and inserting segments of an existing sub-sequence. Thus, nucleic acid sequences that contain repetitive elements would have a lower LZ complexity (and would be less specific in hybridization) (Wright and Church, 2002). We found that three-letter 19-mer sequences had complexity indices of 8.1 ± 0.8 (N = 99,777), while four-letter 19-mers had complexity indices of 9.3 ± 0.8 (N = 338,417, Figure 1B). As a comparative reference, we calculated complexity indices for tiling 19-mers in all exons of human chromosome 22 to be 8.9 ± 1.0 (Figure 1B). Importantly, a significant fraction of three-letter 19-mers (∼31%) had complexity indices of 9 and higher, which matched or exceeded the average complexity of human exons. These calculations indicate that a significant fraction of the random probe sequences composed of only three bases nevertheless retain sequence complexity and specificity that match their physiological, four-base derived counterparts. We applied the AUC rule and the complexity filter to generate two candidate probe-target pairs (referred to as F1 and F2). The calculated ΔG37°C for the F1 and F2 pairs were +1.5 and +2.2 kCal mol−1, respectively, for the AUC-based RNA targets, and +1.0 and +0.9 kcal mol−1 for their ATG-based complementary DNA probes (Figure 1C), suggesting that the pairs are mostly unstructured and likely to be fast-hybridizing. Indeed, the F1 and F2 probes annealed to their surface-immobilized RNA targets at exceptionally fast rates—6 × 106 M−1s−1 and 4 × 106 M−1s−1, respectively, as measured by TIRF-based Colocalization Single-Molecule Spectroscopy (CoSMoS, Friedman et al., 2006) (Figure 1C, Figure 1—figure supplement 2). These on-rates were at least 100-fold faster than typical four-base nucleic acid probes of similar lengths reported in literature (Lima et al., 1992; Kushon et al., 2001; Wang and Drlica, 2004; Friedman et al., 2006; Gao et al., 2006; Yilmaz et al., 2006). Consistent with the AUC rule, introducing back one or more G residues into the F2 RNA target sequence decreased the rate of its hybridization to the complementary DNA probe (10-fold reduction for one G residue and 300-fold reduction for four G residues), which correlated with the progressively lower free energies of self-folding (Figure 1C, Figure 1—figure supplement 2D). We conclude that the target sequences designed using the AUC rule, combined with the complexity filter, achieved our goal of generating superior hybridization kinetics for fast transcript detection. Although the F1 and F2 probes were exceptionally fast, detection of RNA on the sub-second time scales would require the use of fluorescent probe concentrations above 100 nM (Figure 1—figure supplement 2D). Such concentrations are generally incompatible with common single-molecule detection setups such as TIRF microscopy, due to high fluorescence background from the freely diffusing, unbound probe molecules. We overcame this problem by attaching a quencher to the single-stranded probe at the end opposite of the fluorophore (Marras et al., 2002). This self-quenching strategy enabled the use of free probes at concentrations up to 1 μM (Figure 1D, Figure 1—figure supplement 3). We believe that, in the absence of the secondary structure, the self-quenching effect was likely mediated by random polymer motion and/or contact quenching (Johansson et al., 2002; Marras et al., 2002). We refer to our method for fast nucleic acid detection using unstructured, sequence specific, self-quenched fluorescent probes as ‘fastFISH’. Real-time single-molecule detection of transcription with fastFISH To demonstrate that fastFISH can detect nascent transcripts in real time at single-molecule resolution, we used the bacteriophage T7 RNAP. This single-subunit RNAP is an excellent test case for fastFISH: at physiological temperature (37°C) it initiates transcripts at an effective rate of at least 1 s−1 (Martin and Coleman, 1987), and elongates transcripts at ∼240 nt s−1 (Golomb and Chamberlin, 1974; Bonner et al., 1994). As a model template, we used a fluorescently labeled linear DNA fragment containing the consensus promoter for T7 RNAP (Milligan et al., 1987) and the F1 target sequence (positioned between +28 and +46 from the transcription start site, +1), and ending at +295. We surface-immobilized the Cy3-labeled DNA via a biotin moiety attached to the upstream end of the template (position −75) (Figure 2A). In this configuration, RNAP molecules were expected to initiate transcription at the promoter, elongate the nascent RNA in the direction away from the surface, and run off the free, untethered template end at +295, together with their nascent RNA products. Figure 2 Download asset Open asset Real-time single-molecule detection of transcription by T7 RNAP using fastFISH. (A) Schematic of experiment. DNA templates containing a single consensus promoter for T7 RNAP (+1), or a single null mutant promoter (cross), were immobilized on a surface, with the promoter directing transcription towards the free end (+295). The templates contained the F1 fastFISH target sequence downstream from the promoter (from +28 to +46) which was expected to become available for hybridization in the nascent RNA after the RNAP active site reaches position +60. (B) Co-localization analysis of F1 probe-DNA interactions in a representative experiment. Left: wild type promoter. Right: mutant promoter. (C) and (D) Same as panels (A) and (B), but for the F2 target and probe. The templates contained the F2 fastFISH target sequence (from +181 to +199) which was expected to become available for hybridization in the nascent RNA after the RNAP active site reaches position +213. https://doi.org/10.7554/eLife.01775.007 We supplied unlabeled RNAP, NTPs and the fluorescent, self-quenched F1 probe to the imaging flow-cell, and monitored the interactions of the F1 probe with the DNA loci by TIRF/CoSMoS (Friedman et al., 2006; Revyakin et al., 2012). We reasoned that the F1 target sequence in the nascent RNA would become available for hybridization once the active site of elongating RNAP (RDe) reaches position +60 (taking into account the ∼14 bases of nascent RNA protected by RNAP [Huang and Sousa, 2000]). The RDe complex would then remain on the DNA until it runs off at +295. At the expected average RNAP elongation rate of 240 nt s−1, the F1 probe would have a brief ∼1 s window of opportunity to ‘catch’ the nascent transcript at the DNA molecule of origin. Probe hybridization to the RNA product would then be observed as a fluorescent spot co-localizing with the DNA locus. Figure 2B shows the co-localization analysis of probe-DNA interactions in a typical single-molecule fastFISH experiment. This analysis counts all DNA molecules that co-localized with a probe ‘spot’ for more than five cumulative frames (at 0.4 s/frame) during the incubation (∼15 min), and plots the Δx,Δy displacements between the probe and DNA molecules as a two-dimensional histogram (Revyakin et al., 2012). Specific probe-DNA interactions were observed, as indicated by a prominent peak at position (0, 0) of the co-localization histogram. In a typical experiment, 10–30% of DNA templates in a field of view co-localized with a probe. This apparent incomplete template utilization was not due to inefficient probe hybridization (see Figure 4 and ‘Discussion’), but, instead, was likely limited by the accessibility of the DNA template due to surface interference. No probe-DNA interactions were observed in a control experiment with a DNA template containing a null mutant promoter (Raskin et al., 1993), indicating that the co-localization was due to promoter-specific transcription. We also performed real-time single-molecule transcription experiments using the other unstructured probe, F2, and obtained nearly identical DNA-probe co-localization results (Figure 2C–D). Taken together, these findings suggest that fastFISH can detect production of nascent transcripts in real-time. Single-molecule dynamics of T7 RNAP-DNA interactions To measure the efficiency of real-time detection of nascent transcripts by fastFISH, we needed a reference to define the start and end of each productive single-molecule transcription event. We reasoned that monitoring the interactions between RNAP molecules and the DNA templates during promoter binding and run-off can serve this purpose. Therefore, we fluorescently labeled RNAP with Cy5 using HaloTag (Figure 3—figure supplement 1), supplied Cy5-RNAP and NTPs to an imaging flow cell containing immobilized DNA templates, and monitored Cy5-RNAP interactions with the DNA molecules (Figure 3). We observed transient RNAP binding events that lasted between 0.08 and 8 s, and were promoter-specific (Figure 3B–D). The events were comprised of two distinct populations: short-lived events whose dwell times fit well to a single exponential distribution (mean dwell time T0 ∼0.14 s); and a long-lived, bell-shaped population skewed towards long-lived events (peak dwell time T1 ∼1.7 s). The durations of observed interactions were not limited by the lifetime of the Cy5 label before photobleaching (Figure 3—figure supplement 2). We interpret the short-lived, stochastic interactions to be promoter-specific but non-productive RPc and RPo, because no interactions were observed with templates containing a null promoter sequence (at frame rates of 12.5 Hz, Figure 3B) and similar short-lived interactions were observed in the absence of NTPs (single exponential T0 ∼0.3 s, Figure 3D). Figure 3 with 2 supplements see all Download asset Open asset Single-molecule dynamics of T7 RNAP-DNA interactions. (A) Schematic of experiment. (B) Co-localization analysis of RNAP-DNA interactions. Left: null promoter DNA template in the presence of NTPs. Center: wild type promoter DNA template in the presence of NTPs. Right: wild type promoter DNA template in the absence of NTPs. (C) Representative data. Top: video montages of RNAP interactions with template containing consensus promoter for a 5 × 5 pixel region of interest centered at a single, photobleached DNA molecule (1 pixel = 200 nm, imaged at 12.5 Hz). Bottom: fluorescence time traces corresponding to the montages shown on top. Baseline of zero intensity indicates no binding. Left: experiment carried out in the presence of NTPs. Right: experiment carried out in the absence of NTPs. Yellow arrows indicate the first frames of RNAP binding events. (D) Dwell time probability distributions of RNAP-DNA binding events. Left: experiment carried out in the presence of NTP. Fitting to a sum of single exponential and Gaussian functions is shown in blue. Right: experiment carried out in the absence of NTPs. Fitting to a single exponential function is shown in blue. (E) Dependence of the peak dwell time of RNAP-DNA interactions on the length of the transcribed DNA segment: schematic of experiment. DNA templates containing transcribed segments spanning from +1 to +295 (red), +633 (blue), or +910 (black) were separately immobilized, and interactions of labeled RNAP were recorded at 2.5 Hz. (F) Dwell time probability distributions of RNAP-DNA interactions for the three DNA templates shown in (E). N = 749 for the +1…+295 template (red), N = 716 for the +1 … +633 template (blue), and N = 213 for the +1 … 910 DNA template (black). The peak dwell times were calculated by fitting the distributions to a sum of single exponential and Gaussian functions. (G) Plot of the peak dwell time of RNAP-DNA interactions vs the length of the transcribed DNA segment. (H) Decay of intensity of RNAP fluorescence signal during productive RNAP-DNA interactions as an indicator of elongation by RNAP. All RNAP-DNA interactions having dwell times longer than 0.8 s (experiments in F) were post-synchronized by RNAP binding (t = 0, circles) and by RNAP run-off (squares), and weight-averaged plots of RNAP binding and run-off were plotted for DNA templates having transcribed segments of different lengths (red − 295 bp; blue − 633 bp; black − 910 bp). Time offsets between RNAP binding and run-off were set at peak dwell lifetimes, T1, for the respective DNA templates measured in (F). https://doi.org/10.7554/eLife.01775.008 We interpret the long-lived, bell-shaped population of events to be full, productive transcription cycles, because (i) the long-lived events were not observed in the absence of NTPs (Figure 3D); (ii) the peak dwell time of the long-lived events, T1, increased nearly linearly with the length of the transcribed DNA segment (T1 ∼1.7 s, ∼2.7 s, and ∼3.7 s for DNA segments with l = 295 bp, 633 bp, and 910 bp, respectively, Figure 3E–G); and (iii) the RNAP fluorescence signal, on average, decayed towards the end of long-lived events (Figure 3C,H), consistent with RNAP initiating transcription at the promoter located close to the surface (75 bp, or ∼25 nm), and then elongating towards the free, untethered DNA end, down the gradient of the TIRF evanescent field. We conclude that the dwell time distributions of RNAP-DNA interactions define productive transcription events which can be used as a reference to determine the efficiency of nascent RNA detection by fastFISH, and to dissect the full transcription cycle by RNAP. We note that the slope of the plot of long-lived RNAP dwell times vs DNA length, (∼3.2 ± 0.6) × 10−3 s nt−1 (Figure 3G), gives an estimate of RNAP elongation rate of ∼300 nt s−1. The intercept of the plot with the time axis (l = 0) gives an estimate of the net time that RNAP spends on the DNA template without elongating, Tstationary = 0.7 ± 0.3 s, which includes the net duration of promoter opening and abortive cycling (RPc, RPo and RPits), and, possibly, the time RNAP idles at the free DNA end before run-off. Single-molecule dissection of the T7 RNAP transcription cycle with fastFISH To determine the efficiency of nascent RNA detection by fastFISH, and to demonstrate the use of fastFISH in dissecting the kinetics of the full transcription cycle, we simultaneously monitored RNAP-DNA interactions and the production o