TEM images of 50-nm-thin sections were used in the analysis of chromatin packing alterations induced by the DXM treatment for 32 hours. Unlike STEM HAADF imaging mode, the TEM bright-field contrast attenuates following Beer’s law

where I(x, y) is the TEM image intensity distribution, I0 is the incident beam intensity, σ is the absorption coefficient, ρ(x, y) is the density distribution, and t is the section thickness. In our experiment, I0, σ, and t were controlled to be constant for all images, only the chromatin density ρ(x, y) contributes to the final image intensity I(x, y). To obtain the density fluctuation, ρΔ(x, y), we took the negative logarithm of all the TEM images directly and subtracted the mean value. At the same time, the incident beam intensity I0 is canceled out. The 2D ACF was calculated using the Wiener-Khinchin relation as

where F−1 and F are the inverse Fourier and the Fourier transforms, and the ρΔ is the fluctuating part of the chromatin density. To minimize the noise, a rotational average of Bρ(x, y) was taken to obtain the final form of the ACF Bρ(r), representing the correlation of chromatin density as a function of spatial separation r. Notice that mathematically, a fractal structure can be characterized by a power-law ACF, Bρ(r)~rD−3, with D being the fractal dimension. For the chromatin reconstructed by ChromSTEM, the mean ACF Bρ(r) was averaged over the ACFs of each virtual 2D slice and plotted in log-log scale. Linear regression was performed from 50 to 100 nm to obtain the slope p. The chromatin packing scaling D was calculated by 3 + p. Each nucleus was carefully segmented manually in FIJI (49), and the chromatin packing scaling D was calculated through the ACF analysis within the nucleus.