Real-Time Dynamics of RNA Polymerase II Clustering in Live Human Cells

See allHide authors and affiliations

Science  09 Aug 2013:
Vol. 341, Issue 6146, pp. 664-667
DOI: 10.1126/science.1239053

Pol II Micro Clusters

In higher eukaryotes, messenger RNA (mRNA) synthesis is thought to involve foci of clustered RNA polymerase II (Pol II) called transcription factories. However, clustered Pol II have not been resolved in living cells, raising the debate about their existence in vivo and what role, if any, they play in nuclear organization and regulation of gene expression. Cisse et al. (p. 664, published online 4 July; see the Perspective by Rickman and Bickmore) developed single-molecule in vivo analyses revealing the distribution and dynamics of Pol II clustering in living cells. Pol II clusters were smaller than the diffraction limit (<250 nm). Transient dynamics of the Pol II clusters, and correlation with changes in transcription, pointed to a role in transcription initiation rather than in elongation.


Transcription is reported to be spatially compartmentalized in nuclear transcription factories with clusters of RNA polymerase II (Pol II). However, little is known about when these foci assemble or their relative stability. We developed a quantitative single-cell approach to characterize protein spatiotemporal organization, with single-molecule sensitivity in live eukaryotic cells. We observed that Pol II clusters form transiently, with an average lifetime of 5.1 (± 0.4) seconds, which refutes the notion that they are statically assembled substructures. Stimuli affecting transcription yielded orders-of-magnitude changes in the dynamics of Pol II clusters, which implies that clustering is regulated and plays a role in the cell’s ability to effect rapid response to external signals. Our results suggest that transient crowding of enzymes may aid in rate-limiting steps of gene regulation.

Transcription is believed to take place in the cell nucleus in local RNA polymerase II (Pol II)–enriched foci known as transcription factories. Clusters of Pol II have been observed in fixed cells (1); however, many aspects of transcription factories remain a matter of debate (16). Central questions are whether these foci are stable architectures to which DNA must translocate (7) or if they assemble and disassemble dynamically for transcription regulation in vivo.

Here, we engineered a human osteosarcoma cell line (U2OS) stably expressing the Pol II catalytic subunit (RPB1) labeled (8, 9) with a photoconvertible fluorescent protein, Dendra2, replacing endogenous RPB1 (Fig. 1A and fig. S1). This cell line enabled superresolution imaging—fluorescence-based localization with an accuracy greater than the diffraction limit—of the distribution of Pol II in living cells by means of photoactivation localization microscopy (PALM) (10, 11). As illustrated in Fig. 1B, a nonhomogeneous distribution of Pol II was observed in living cells, which suggested Pol II clustering, comparable to Pol II distributions in fixed cells (fig. S2).

Fig. 1 Live-cell superresolution imaging reveals spatial Pol II clustering.

(A) Preconverted (Dendra2-RPB1 green emission) fluorescence image shows Pol II primarily localized in nucleus [compare (A) and (B)]. (B) Two-dimensional superresolution reconstruction reveals nonhomogeneous distribution of detected Pol II (red). Nuclear contour (white outline) is approximated from preconverted fluorescence in (A). (C) A pair-correlation analysis was implemented as previously described (12) to quantitatively analyze the spatial distribution. Represented is the pair-correlation function computed from the spatial coordinates of the raw PALM detections (black), fitted to a general function (orange) that accounts for contributing factors from the protein clusters and single-molecule stochastic effects as detailed in supplementary text and fig. S3. The corrected spatial correlation function for the protein (green) is decoupled from the fluorophore stochastic contributions (blue). The corrected protein correlation function shows statistically significant clustering, above the theoretical g(r) = 1 (gray dashes) with a fit parameter of rprotein ~ 220 (± 17) nm, distinct from the single-molecule stochastic fit parameter of rstoch ~ 45 (± 1) nm. Errors (in parentheses) represent standard error of the fitted value.

Pair-correlation PALM (pcPALM) analysis was recently used to infer spatial clustering of proteins at the cell membrane, with a correction for possible contributions of single-molecule fluorophore photophysics (12). pcPALM analysis on the PALM distribution of Dendra2–Pol II (fig. S3) yielded a correlation function best fitted by a spatial clustering model, both in living and fixed cells, but not in a control cell line expressing Dendra2 alone, which fitted the simpler model consistent with single-molecule homogeneous distribution (Fig. 1C and fig. S3). The average correlation radius obtained by pcPALM was 220 (± 17) nm in live cells, and 94 (± 4) nm in fixed cells (table S1); this difference can be explained by the observation that temporally distinct clustering events can occur at neighboring loci in vivo (fig. S7). Cluster sizes below the diffraction limit may explain why these structures remained uncharacterized in previous Pol II live-cell studies (4, 5).

PALM images are reconstructed from single detections collected over several minutes of imaging. In live-cell PALM, all the temporal information is lost in the final image. In order to discriminate between a deterministic organization model with static, dedicated Pol II clusters, which necessarily required gene translocation for activation, and a self-organization model with highly dynamic, de novo clustering of Pol II, we analyzed the temporal sequence of detections leading to a particular cluster. We investigated from the time-resolved single-molecule detection data whether the live-cell clusters have discernible and quantifiable temporal signatures indicative of their relative stability. We refer to this approach combining time-correlated detection counting and PALM as time-correlated PALM (tcPALM) (13).

For tcPALM analysis, we selected regions corresponding to individual high-density clusters in live-cell superresolution PALM images. For each selected Pol II cluster, we plotted the single-molecule detection profiles in the form of time series (Fig. 2, A and C) representing the rate of detection of Dendra2–Pol II fluorescence per frame. The detections did not seem uniformly distributed but appeared clustered in time. These temporal clustering events are more evident in the cumulative count of detections, where they appear as large steps (arrows in Fig. 2, B and D) in the representation. For comparison, in a fixed cluster, the cumulative detections from fixed-cell controls under identical conditions (see fig. S4A) exhibited a monotonic slope from the start of the acquisition, indicative of a local concentration of immobile proteins followed by a gradual plateau until the end of acquisition, signature of an exhaustion in detections from the static pool of proteins. By contrast, the live-cell cumulative detections showed a delayed slope onset, which suggested that the clusters dynamically assembled after acquisition started. Moreover, the slope abruptly turned into a plateau, indicative of a sudden disassembly of the cluster. Overall, these observations suggest that Pol II clustering in live cells exhibited transient dynamic signatures clearly distinct from static clusters.

Fig. 2 tcPALM analysis reveals temporal clustering of Pol II in live cells.

(A to D) Representative time-dependent detections from two Pol II clusters in living cells show bursts of temporally correlated, high counts of detections. The cumulative detection profiles (B and D) illustrate dynamic cluster assembly (arrows) and disassembly (plateaus). (E) The distribution of apparent burst lifetimes (τon) is represented with a Gaussian fit. Average τon obtained was 5.1 (± 0.4) s, and the fit mean obtained was 4.2 (± 0.4) s. Errors (in parentheses) represent standard error of the mean. We analyzed 104 clusters from four cells.

To quantify the apparent kinetics of Pol II clustering, we implemented a burst dwell-time analysis. We recorded both a dwell time (τon) corresponding to the apparent lifetime of a selected burst and the count of detections per burst (burst size). The time separating individual bursts (τoff) was often larger than the imaging time (500 s) and, therefore, could not be determined precisely. For the bursts detected, we obtained a distribution of τon with an average lifetime of 5.1 (±0.4) s, and spanning less than 30 s (Fig. 2E). This observation is consistent with our previous study (4) on an artificial gene array, where 99% of Pol II molecules interacted very transiently at the active locus. These results support de novo spatiotemporal clustering of Pol II, which is more consistent with a self-organization model of transcription.

We hypothesized that, if there is any relevance of Pol II clustering to transcription, it should be reflected by systematically changing the cell’s transcriptional state. We therefore measured the cluster dynamics through serum induction, when specific response genes are transcribed at much higher efficiency (1417). Similar to the normal growth case, temporal clustering events were observed after serum stimulation (Fig. 3, A and B). The average τon was 48 (± 9) s (Fig. 3C), an order-of-magnitude increase from normally grown cells. As shown in Fig. 3C, the distribution of τon revealed a variation of bursting lifetimes spanning up to two orders of magnitude. Larger clusters were also observed under serum stimulation (Fig. 3D). Depicted in the examples in Fig. 3, A and B, two clusters exhibiting comparable total detection counts might exhibit 10-fold difference in bursting lifetimes, which suggests that the apparent recruitment rate differs considerably from one locus to another. These observations are consistent with the notion that individual gene loci might exhibit different clustering kinetics. By contrast, Pol II clustering kinetics in serum-deprived cells showed relatively little difference from the kinetics measured in normally grown cells (fig. S5), which indicates that the orders-of-magnitude changes in Fig. 3 are unique to the serum stimulation. In addition, accumulation of Pol II could be observed at the transcription locus of an induction-responsive (β-actin) gene by colocalization analysis with endogenous mRNA imaged by fluorescence in situ hybridization (FISH) (fig. S6); therefore, individual Pol II clusters can be linked to a gene locus with active transcription. These results correlate Pol II clustering dynamics to gene expression regulation at specific loci.

Fig. 3 Pol II cluster dynamics are dependent on transcription induction.

(A and B) Time-dependent detection profiles of individual clusters show several fold higher detection counts (red arrow) during cluster formation (black arrows) with serum induction compared with clusters under normal cell growth (Fig. 2, A and C). (C) Distribution of apparent burst lifetimes (τon) has considerably broadened with an average τon for serum induction of 48 (± 9) s. (D) The probability distribution of burst sizes (detection counts per burst) for serum induction shows broadening with higher counts for burst events that were not present in clusters from normal growth cells (D, inset); the average burst size was 89 (± 14) and 23 (± 1) detection counts for serum-induced and normal growth cells, respectively. Errors (in parentheses) represent standard error of the mean. We analyzed 106 clusters from six cells.

Next, we investigated whether Pol II clustering precedes or succeeds mRNA synthesis (transcription elongation). The inhibition of the positive elongation factor (P-TEFb) impedes productive elongation (4, 6, 18, 19). In chromatin immunoprecipitation (ChIP) experiments, P-TEFb inhibition did not affect polymerases accumulated at the promoter, whereas depletion in the transcribed region of genes was observed (18, 20). We hypothesized that, if clustering reflected an accumulation of the elongating Pol II, then P-TEFb inhibition should substantially reduce clustering. In contrast, upon treatment with the P-TEFb inhibitor flavopiridol (19, 21), Pol II clusters were more pronounced (Fig. 4). These results suggest that Pol II clustering occurs before promoter pause release (20), a regulatory step preceding productive transcription elongation.

Fig. 4 Transcription elongation inhibitory drug treatment suggests Pol II clustering precedes mRNA synthesis.

(A to C) Time-dependent detection profiles of individual clusters after flavopiridol treatment show temporal signature consistent with a steady pool of clusters. (D) The histogram distribution of the total counts of detections per cluster are represented for the flavopiridol-treated (top), normal growth (middle), and serum-induced cells (bottom). CoV denotes the coefficient of variation obtained for the respective data.

After inhibition, 89% of clusters showed a stable signature, akin to fixed-cell static clusters (Fig. 4, A to C), with a slope onset in the cumulative detections from the start of acquisition followed by a gradual plateau. Furthermore, the histogram of the total counts of detections per cluster in flavopiridol-treated cells had a distribution comparable to that of normally grown cells. For both, the coefficient of variation was ~0.6, which suggested that the same populations of gene loci are likely sampled in normal growth and after drug treatment. By contrast, these distributions are distinct from the serum-activated cluster dynamics, which resulted in a coefficient of variation two times larger (Fig. 4D). These results imply that the presence of Pol II clusters might depend on the presence of the preinitiation complex or initiating polymerases but not on elongation. We propose that such dynamic clustering of Pol II may play a role in facilitating and regulating macromolecular complex assembly during the rate-limiting (4, 5, 22, 23) transcription preinitiation and initiation steps (fig. S8).

Protein clustering has been reported in many cellular processes, including DNA replication, DNA repair, mRNA splicing, ribosomal RNA synthesis, and signal transduction. In many cases, macromolecular complex assembly is identified as a rate-limiting step, and detailed kinetic characterization could help elucidate the regulatory mechanisms. The PALM-based approach developed here can be broadly used for other proteins to study cellular processes with high spatial and temporal resolution, in vivo, with single-molecule sensitivity.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S8

Table S1


References and Notes

  1. We note that, in live-cell PALM experiments, only a fraction of the molecules present can be localized; highly mobile or nonphotoactive molecules may be present but not detected. Therefore, tcPALM is not a measure of the total number of molecules present but rather an indication of the relative fluctuations of fluorescent molecules transiently present at a given locus.
  2. Acknowledgments: We acknowledge J. T. Lis, I. Golding, R. Phillips, O. Bensaude, and E. Bertrand for valuable discussions. We thank members of the Darzacq, Dahan, and Howard Hughes Medical Institute–Janelia Farm Transcription Imaging Consortium groups for suggestions. I.I.C. was supported by fellowships from the European Molecular Biology Organization and the Foundation Pierre Gilles de Gennes (FPGG). I.I. acknowledges the Netherlands Organization for Scientific Research and FPGG for financial support. A.S. was supported by La Ligue nationale contre le cancer. L.M. acknowledges support from Centre de Génétique Moléculaire UPR 3404. This work was supported by grants Agence Nationale de la Recherche Pol2Kinetics to X.D. and DYNAFT to X.D. and M.D.; X.D. and M.D. acknowledge the support of Nikon France.
View Abstract

Stay Connected to Science

Navigate This Article