Background

Interest in the phosphorylation status of the CTD increased due to recent studies showing its correlation with RNAP2 clustering (Cissé et al., 2013; Cho et al., 2016; Boehning et al., 2018; Pancholi et al., 2021). These findings suggest that RNAP2 clusters form around gene promoters, and early in the transcription cycle, they are enriched in unphosphorylated-RNAP2 and Ser5ph-RNAP2 (Nagashima et al., 2019). RNAP2 phosphorylation distribution along the 1D genome has been extensively studied using immunoprecipitation-based assays (Heidemann et al., 2013; Harlen and Churchman, 2017). However, these techniques are performed using fixed cells and require averaging over a population of cells, limiting temporal resolution and masking heterogeneity (Coulon et al., 2013). Recent advances in fluorescent microscopy and single-molecule tracking (Tokunaga et al., 2008; Chen et al., 2014; Li et al., 2019) have overcome these limitations. Now, it is possible to monitor single RNAP2 dynamics at different locations throughout the genome (Cissé et al., 2013; Steurer et al., 2018) and specific single-copy genes (Cho et al., 2016; Li et al., 2019) pre-marked with MS2 (Tantale et al., 2016; Pichon et al., 2018) or PP7 (Larson et al., 2011; Coulon et al., 2014) RNA stem-loops. When transcribed, MS2 and PP7 tags form hairpin-like secondary structures, which are then bound by their fluorescence-labeled coating proteins, MCP and PCP, respectively. Unfortunately, previous protocols use permanent fluorescent fusion tags to track RNAP2, which cannot distinguish between the crucial RNAP2 phosphorylation states that control transcription.

Fluorescent antibody-based (Fab) probes can detect post-translational modifications to RNAP2 by binding and lighting up specific modifications to the CTD of RNAP2 in vivo (Hayashi-Takanaka et al., 2009; Stasevich et al., 2014; Kimura et al., 2015; Lyon and Stasevich, 2017). However, freely diffusing and unbound Fab probes lead to a high background; hence, this technique was previously limited to large tandem gene arrays (Stasevich et al., 2014) that include multiple copies of the same gene to enhance the signal-to-noise ratio. However, using such a protocol averages over many gene copies, and masks heterogeneity from one gene copy to another. Recently, our lab developed a protocol to measure and predict the spatiotemporal dynamics of RNAP2 phosphorylation and mRNA production throughout the transcription cycle of a single-copy gene. This protocol, which we describe here, combines multicolor single-molecule microscopy, complementary Fabs, and rigorous computational modeling (Forero-Quintero et al., 2021). For our system (Forero-Quintero et al., 2021), we utilized an established HeLa cell line (H-128) expressing a single-copy reporter gene controlled by an HIV-1 promoter, comprising an MS2-tag and its MS2 coating protein tagged with GFP (MCP, blue) (Tantale et al., 2016). Our reporter gene is predominantly active due to persistent stimulation by Tat, leading to a bright MCP-GFP signal, that indicates the location of the transcription site within the nucleus, allowing us to observe fluctuations in mRNA synthesis in real-time (Figure 1a). Our loaded Fabs recognize the CTD of RNAP2 (CTD-RNAP2, conjugated with CF640, red) without or with residue-specific phosphorylation, and specific phosphorylation at the Serine 5 within the CTD of RNAP2 (Ser5ph-RNAP2, conjugated with Cy3, green). This combination of imaging probes makes it possible to observe RNAP2 regions enriched or depleted with Ser5ph within the nucleus (Figure 1a, b). Fab binding and unbinding from their targets occurs rapidly, making it a valuable tool to monitor temporal changes in the phosphorylation status of RNAP2 (Hayashi-Takanaka et al., 2011; Stasevich et al., 2014; Kimura et al., 2015). At the HIV-1 reporter transcription site, we typically observe both Fabs present, but on occasion, the Ser5ph-RNAP2 signal becomes dim or absent despite the presence of CTD-RNAP2 and mRNA signals. Furthermore, all signals occasionally reduce to background levels, evidencing correlated fluctuations of the signals, which are caused by natural bursts and pauses in the transcriptional activity of the reporter gene (Figure 1b). By combining these multicolor elements, we can distinguish three distinct steps of the transcription cycle at the HIV-1 reporter gene: (1) RNAP2 recruitment (marked by Fab targeting CTD-RNAP2), (2) initiation (marked by Fab targeting CTD-RNAP2 and Fab targeting Ser5ph-RNAP2), and (3) elongation (marked by both Fabs and MCP binding to mRNA) (Figure 1a). To add a quantitative interpretation to these observations, we explored many possibilities and identified a simple computational model that matches all data. The optimal model that fits all data, but which avoided overfitting, consists of five basic rate parameters that describe four reactions: (1) the recruitment of RNAP2 in geometrically distributed bursts of average size (β) and frequency (ω), (2) the departure of unsuccessful or lost RNAP2 (k_ab), (3) promoter escape of RNAP2 (k_esc), and (4) a combined rate of transcription completion and RNAP2 release (k_c). Our simple mechanistic mathematical model allows us to estimate each of the above rates with excellent precision, as well as to predict the number of unphosphorylated and/or phosphorylated RNAP2 at each state throughout the transcription cycle at the HIV-1 reporter gene, and to simulate additional experimental features, such as RNAP2 positional densities or response to different transcriptional inhibition mechanisms.

Our integrated technique possesses several advantages: (1) Fab binds endogenous RNAP2, so all RNAP2 present in the cell at a given time has a high probability to be labeled; (2) The fluorescence of the Fabs is naturally amplified by the 52 heptad repeats contained in the CTD of RNAP2; (3) Fab continually binds and unbinds RNAP2, reducing the loss of fluorescence due to photobleaching; (4) Our technology can be employed to measure the RNAP2 dynamics and phosphorylation status at other single-copy genes, provided the location of the gene can be identified without (Gu et al., 2018) or with labeling techniques [e.g., using MS2 or PP7 tags (Larson et al., 2011; Coulon et al., 2014; Tantale et al., 2016; Pichon et al., 2018), or other labeling systems such as ROLEX (Ochiai et al., 2015), ANCHOR (Mariamé et al., 2018), DNA FISH (Takei et al., 2017) or CasFISH (Deng et al., 2015)]; and (5) our modeling approach directly considers the temporal and statistical fluctuations of individual transcription sites, and applies information criteria to identify the simplest model that matches all experimental data.

In this protocol, we describe step-by-step how to fragment and fluorescently label antibodies targeting the CTD of RNAP2, how we put them into human cells using bead-loading (McNeil and Warder, 1987; Hayashi-Takanaka et al., 2009; Stasevich et al., 2014; Cialek et al., 2021), the imaging conditions we employed and how to tune them to get a good signal-to-noise ratio based on the experimental needs, how to quantify signal intensities at the transcription site from 3D imaging movies, and how to infer an appropriate mathematical model to reproduce these data without overfitting.

Figure 1. Schematic of our multicolor system to visualize RNAP2 phosphorylation at the transcription site of a reporter gene. A. The reporter gene contains a 128x MS2 insert (blue) that, when transcribed, produces stem-loops which are then recognized and bound by MCP genetically fused to GFP, thus lighting up the location of the gene within the nucleus, as well as the synthesis of nascent and mature mRNAs. Our Fabs are capable of identifying the CTD of RNAP2, as well as its phosphorylation at serine 5, which were fluorescently labeled with CF640 (red) and Cy3 (green), respectively. By combining these three elements, it is possible to visualize the recruitment (by CTD-RNAP2-CF640), initiation (by Ser5ph-RNAP2-Cy3), and elongation (by MS2-MCP-GFP) steps at the transcription site of a reporter gene. B. Crops from an example cell displaying, from top to bottom, the mRNA, Ser5ph-RNAP2, CTD-RNAP2, and merge signals, at three characteristic time points, in which (left) all three signals are present, (middle) mRNA and CTD-RNAP2 are present, but Ser5ph-RNAP2 is dim or absent, and (right) all signals are absent. These data demonstrate transcription fluctuations and the occurrence of multiple transcription cycles (example cell image adapted from Forero-Quintero et al., 2021).