Mapping Absolute DNA Density in Cell Nuclei using Single-molecule Localization Microscopy

The present method allows absolute DNA density measurements in cell nuclei to investigate the spatial constraints in chromatin structures that are otherwise only characterized by post-translation histone modifications. Voronoi tessellation combined with SMLM allows absolute density estimations of DNA in terms of base pair per square micrometer. The only apriori knowledge needed is the total DNA content of the measured cell nucleus.

In future experiments, it would be interesting to investigate whether absolute density differences correlate with biological information from other methods, such as immunofluorescence of epigenetic modifications or Hi-C data. Begin by reconstructing the scanned area by tiling the prerecorded single images together. Open CellProfiler and then load the pipeline cellcycleanalysis.cpproj.

In the first step images of the pipeline, drag and drop the images, including the reference image. Next, define a location where the reference images will be stored. Then click on the Start Test Mode icon followed by Run.

Once the histogram of the nuclei in the image to be analyzed is displayed, determine the intensity intervals for the G1, S and G2 phases, and make sure that they are entered in the subsequent filter object steps of the pipeline and that these steps are active. Then click the Run button again to continue with the second half of the pipeline. Look at the three pictures with the overlays representing the different cell cycle stages, G1, S or G2, and select the nuclei in G1 or G2 for further imaging so that the amount of DNA in base pairs of the image nuclei is known.

Relocate the cells on the single molecule localization microscope, and ensure proper blinking of the FPALM process. First, record a 3D stack of the chosen nucleus only using 500 frames per slice. This allows determining the relative amount of DNA in the midsection, which is later imaged with many more frames to reveal its ultrastructure.

Begin the image series acquisition using an exposure time of 50 milliseconds, resulting in a frame rate of approximately 20 frames per second. Take a Z-stack through the nucleus with 200 nanometer step intervals and only 500 frames per light optical section. After the stack was acquired, return to the initial z position and acquire the main data set of 50, 000 frames with the same 50 milliseconds exposure time.

Open the FPALM dataset in ImageJ. To optimize the settings for the detection of blinking signals, click on ThunderSTORM, then Run analysis. Now, choose Camera setup.

Enter the correct pixel size of the image data, the A/D count, quantum efficiency of the used camera, base level, and EM gain, then click OK.Choose the algorithm to determine the localization coordinates by fitting the blinking signals. In the Image filtering section, use the Wavelet filter B-Spline with a B-Spline order 3 and a B-Spline scale 2. Then, set the Approximate localization of molecules Method to Local maximum, set the Peak intensity threshold and Connectivity to 8-neighborhood.

In the Sub-pixel localization of molecules section, as a Method, choose PSF integrated Gaussian, set Fitting radius px to 3, select Fitting method as Maximum likelihood, and set Initial sigma to 1.6. Then click OK to start the detection of the signals and reconstruction of the image. If the acquisition is partitioned into different image stacks, concatenate the localization tables in ThunderSTORM.

To do so, click Import in the popup window. Ensure that the Append to current table option is activated and that the correct starting number is used to avoid overriding localizations in the table. Then select the file path and click OK to import one file after another.

To prevent over-counting signals across consecutive frames, merge the localization data using maximum distance of a 20 nanometer, 1 maximum off frames, and 0 maximum frames per molecule, then click Merge. To measure and correct drift by cross-correlating data subsections, open the Drift correction menu and click the arrows. Add 3 bins and set the Magnification to 5.

Then click Apply, and the window showing the x and y drift will appear. To determine the amount of signal which is proportional to DNA content in each slice of the prerecorded 3D stack with 500 frames for each slice, choose Plugins, then ThunderSTORM, followed by Import/Export, and import the results table for the first slice of the stack. To avoid over-counting signals from other image planes of the 3D stack, signals from outside the 200-nanometer step interval in z between the slices need to be removed.

Press Plot histogram and in the open Distribution dialogue, choose z and press OK.Determine the position of the peak and use the filter field to select the signals with the optical step size of 200 nanometers. Click on Visualization and choose the Histograms option, then click OK.Save the resulting image in a folder in the TIFF format. Open all slices and combine them using Image followed by Stacks and then Images to Stack.

After selecting the whole image area, go to Analyze, click Tools and select the ROI Manager. Click on Add, then click on Analyze, followed by Set Measurements and select Integrated density. Now, choose the More option, select Multi Measure, check Measure all stacks as well as One row per slice.

Click OK, and the results will appear. The sum of the intensities of all slices is proportional to the DNA content of the whole nucleus in the same way as the intensities of individual slices are proportional to their fraction of the whole genome. After opening the results table of the center plane that contains the signals of 50, 000 frames, filter it to a thickness of 100 nanometers.

Create the histogram of the resulting signals as demonstrated earlier, and measure the integrated density of the center section. Convert the large ThunderSTORM localization table containing the ultrastructural information from the center section from the csv format into the ORTE format by running the MATLAB script TS2orte. m, which transforms the localization table into a MATLAB matrix and saves it in the mat format.

Go into the LAND-voronoi folder, and in the coreAlgorithm sub folder, open the file voronoicluster. m and adjust the DNA content, fraction of the observed center section of the whole nucleus fraction DNA and localizations, and the conversion factor according to the used cell type and cell cycle stage for the absolute density calculations. Edit the script that starts the analysis voronoi.m.

Adjust the file path to the orte localization data and the output folder for the result files. Specify the coordinates defining the area in which the tessellation should be done. Define multiple areas to compute within the same input dataset.

After running the script, look for the image showing the absolute DNA densities together with other files containing graphs showing the histogram of the area distribution, and a densities. m file containing the densities of every single calculated Voronoi cell in the output folder. The FPALM measurement consisting of 50, 000 frames resulted in 2.68 times 10 to the power of 6 detected localizations within a 100-nanometer thick midsection.

The localization accuracy of the reconstructed image was 10 nanometers. The large number of localizations allowed the combination of super resolution imaging while showing absolute DNA densities. It was apparent that the measured DNA densities were not uniformly distributed across the nucleus and covered an extensive dynamic range.

Higher magnifications of areas within the nucleus showing larger Voronoi cells indicated low DNA densities, while smaller cells indicated high DNA densities. DNA densities measured in a G1C3H10T half nucleus showed the absolute DNA density distribution within the nucleus of a different type and species. Although the basic organization looked like the early G1 HeLa nucleus, it additionally possessed constitutive clusters of paracentric heterochromatin.

No dramatic architectural changes were observed in a G2 nucleus despite a slightly increased nuclear size. The HeLa nucleus treated with TSA showed remarkable differences with respect to nuclear topography and DNA density. Except for the peripheral heterochromatin, which showed islands of higher DNA density, the rest of the chromatin looked much more homogeneous and decondensed.