Single cell visualization of transcription kinetics variance of highly mobile identical genes using 3D nanoimaging

Multi-cell biochemical assays and single cell fluorescence measurements revealed that the elongation rate of Polymerase II (PolII) in eukaryotes varies largely across different cell types and genes. However, there is not yet a consensus whether intrinsic factors such as the position, local mobility or the engagement by an active molecular mechanism of a genetic locus could be the determinants of the observed heterogeneity. Here by employing high-speed 3D fluorescence nanoimaging techniques we resolve and track at the single cell level multiple, distinct regions of mRNA synthesis within the model system of a large transgene array. We demonstrate that these regions are active transcription sites that release mRNA molecules in the nucleoplasm. Using fluctuation spectroscopy and the phasor analysis approach we were able to extract the local PolII elongation rate at each site as a function of time. We measured a four-fold variation in the average elongation between identical copies of the same gene measured simultaneously within the same cell, demonstrating a correlation between local transcription kinetics and the movement of the transcription site. Together these observations demonstrate that local factors, such as chromatin local mobility and the microenvironment of the transcription site, are an important source of transcription kinetics variability.

objective is used. The fluorescence is collected through a dichroic short-pass filter and split between two detectors using an additional long-pass dichroic mirror. The signal from the photomultiplier tubes (PMT) is amplified, discriminated (CFD) and the digital output is read out using a fast FCS data acquisition card.

Experimental setup
The excitation laser beam was provided by a MaiTai Ti:Sa (Spectra Physics) laser source, providing pulsed IR excitation tunable in the range between 690 and 1040 nm. The power of the excitation beam is controlled using an AcoustoOpic Modulator (AOM) MT110 (AA Optoelectronics, Orsay, France). The first order beam output from the AOM is beam expanded to approximately 0.5 inches and spatially filtered through a 15 µm pinhole (Thorlabs).
Microscopy experiments were performed on a Zeiss Axiovert 135 microscope frame, modified for orbital particle tracking as illustrated in Supplementary Figure 10. The rear port was fitted with a customized aluminum lens holder to accommodate a short focal distance lens (f=3 mm, a microscope eyepiece) and a collimating f=50 mm lens mounted on a 1 inch cage system (Thorlabs). The two lenses act as an effective beam expander able to overfill the back focal plane (BFP) of the microscope objective (Olympus 60X, 1.2 NA Water). The beam is reflected into the objective BFP after being reflected by a 680 nm short pass dichroic mirror (Semrock). A 1-inch flip mirror (Thorlabs) was inserted after the collimating lens to switch from laser excitation to wide-field excitation using the slightly defocused beam originating from a Xenon Lamp. The voltage to control the Galvo Scanners and the nanopositioner Stage is provided using an ISS 3-axis PCI card. The sample is kept at 37 °C by the use of a PDMI-2 MicroIncubator (Harvard Apparatus), adapted onto the MS2000 stage. The entire data acquisition via the FCS card and the control of the scanning microscope via the 3-axis card is performed using the SimFCS® software, developed by Enrico Gratton at the Laboratory for Fluorescence Dynamics.

MonteCarlo Simulations of Pol II motion along the gene and generation of the fluorescence trajectory
We run simulations of fluorescence trajectories of the MS2-EGFP signal upon a given set of kinetic parameters for PolII elongation on our gene (Supplementary Figure 5). In our case, the use of simulated datasets is necessary to provide an absolute calibration to the phasor plot values.
As discussed in the main text, the phasor plot of the data can tell us that different petals display markedly different elongation kinetics, but cannot provide us with absolute values unless the elongation rate of the PolII were known beforehand. For this reason we had to recur to simulated trajectories for varying values of the elongation rate and the termination rate to calculate the reference phasor points that are displayed in Supplementary Figure 6d  In the current report we followed two fundamental assumptions that were previously employed by 3 in deriving an analytical model for autocorrelation of PolII elongating along MDN1 gene in yeast. First, the mRNA release constant was assumed identical to the effective elongation rate. This is a significant assumption, since a slow release of the mRNA from the gene could be the rate-limiting step. However, as illustrated in Supplementary Figure 7 a- € Which in our hands was not amenable to further simplification.

Analysis of petals intensity carpets:
As discussed in the main text, the fluorescence intensity collected in the EGFP channel of a Second, it is necessary to correct for trends in the fluorescence intensity over long time scales, such as photobleaching or decline in MS2-EGFP signal following transcription inhibition.
However, as the signal intensity decays following perfusion of ActinomycinD (to inhibit transcription), it was necessary to correct the data. A moving average correction, calculated on the carpet of each petal and with a size of 1024 points=256 s was employed in this case.

pCF analysis
pCF was performed, as previously reported, using the SimFCS software, on the intensity carpet collected while performing tracking of a chromatin array. The voltage supplied to the galvanometer mirrors was adjusted in order to obtain a three-lobes shape, the trefoil illustrated in circle in the complex plane is calculated using the S and G coordinates, according to is the fundamental frequency and is determined by the temporal resolution Δt of our measurement and by the time bin used to calculate the ACF, typically 256 s yielding N=1024 .
The angle φ of each phasor component is given by: ). In practice S and G are calculated using a Fast Fourier Transform (FFT) of the Auto Correlation Function and posing:

Equation 4
where k is the discrete frequency index corresponding to the continuous variable ω1 used in the previous formulas. k=1 yields the first harmonic of the measurement, k=2 the second harmonic and so forth.
In being a fit-less approach, phasor analysis requires a calibration to extract quantitative information from the data. We chose to calibrate the kinetic phasor using MonteCarlo generated PolII fluorescence trajectories.
It is worth noticing that the reference phasors depend upon the model and upon the parameters (such as initiation, elongation and termination constants) chosen to run the simulations. Upon changing the model, the absolute position of the experimental phasor would map to a different set of kinetic parameters; however, the observation of a relative difference between any two experimental datasets is not model dependent.
The obvious advantage of this approach is that it provides an immediate and graphical way to compare elongation rates across multiple datasets, capturing relative differences without the constraint of having to fit a model to the experimental data. We should also note here that