Multi‐Color, Bleaching‐Resistant Super‐Resolution Optical Fluctuation Imaging with Oligonucleotide‐Based Exchangeable Fluorophores

Abstract Super‐resolution optical fluctuation imaging (SOFI) is a super‐resolution microscopy technique that overcomes the diffraction limit by analyzing intensity fluctuations of statistically independent emitters in a time series of images. The final images are background‐free and show confocality and enhanced spatial resolution (super‐resolution). Fluorophore photobleaching, however, is a key limitation for recording long time series of images that will allow for the calculation of higher order SOFI results with correspondingly increased resolution. Here, we demonstrate that photobleaching can be circumvented by using fluorophore labels that reversibly and transiently bind to a target, and which are being replenished from a buffer which serves as a reservoir. Using fluorophore‐labeled short DNA oligonucleotides, we labeled cellular structures with target‐specific antibodies that contain complementary DNA sequences and record the fluctuation events caused by transient emitter binding. We show that this concept bypasses extensive photobleaching and facilitates two‐color imaging of cellular structures with SOFI.


Microscopy setup for SOFI measurements
A home-built microscope setup was used that was previously described in detail [3] . Laser powers of 7 mW and 13 mW (measured at the fiber exit) were used for excitation of AbberiorSTAR 635P and Cy3b conjugated imager strands, respectively. µManager [5] version 1.4.22 was used for camera control and 5000 consecutive frames were acquired at 40 Hz with an EM-gain of 50, a preamplifier gain of 5.3, and a readout rate of 3 MHz.

Microscopy setup for comparison of permanent and exchangeable label
Image acquisition was performed at the N-STORM super-resolution microscopy system (Nikon, Japan) equipped with an oil immersion objective (Apo TIRF, 100x, NA

Super-resolution optical fluorescence imaging
Raw images from the acquired image stack were up-sampled (using lossless interpolation by zero-padding of the frequency-limited Fourier-transformed images) to half the physical pixel size to accommodate the final image pixel size to the SOFIenhanced spatial resolution [6] . Then, for each pixel, we calculated 2 nd and 3 rd order correlation values C2 and C3 over the whole stack of K recorded frames by using the Here, x   denotes the entire part of number x, and I(j) is the difference between the pixel's intensity I(j) in frame j and its mean intensity averaged over all K recorded frames:

Estimation of image resolution
The spatial resolution of diffraction-limited images (mean intensity z-projections) and sub-diffraction images (2 nd and 3 rd order SOFI images) were estimated using Fiji [7] and the plugin of the open source algorithm published by Descloux et al. [8] .

Determination of microtubule diameters
Microtubule diameters in immunolabeled cells were determined by measuring the fullwidth-at-half-maximum (FWHM) of the intensity profile perpendicular to straight filaments using home-written analysis tools in Python 3.7. 50 lines equally spaced and perpendicular to a line matching a microtubule were drawn. FWHM values were extracted from a Gaussian fit to the intensity profile and mean FWHM values were calculated for each condition (mean intensity z-projections, SOFI 2 nd order and SOFI 3 rd order)

Analysis of intensity time traces
Photoinduced fluorophore bleaching was analyzed with open-source program Fiji (1.52a) [7] and home-written routines in Python 3.7. Average image intensities were extracted from movies via ImageJ plugin "Plot Z-axis profile" and saved for further analysis in Python. Intensity traces were normalized to the first frame and the intensities of 40 consecutive frames were averaged. The routine was performed for >20 cells for all conditions and the average intensity values were plotted with the respective standard deviation.

FRC Analysis
FRC analysis was carried out using the NanoJ plugin in Fiji [9] . To create image pairs for FRC, DNA-PAINT movies were split into overlapping sub-sequences using the Fiji plugin "Slice Keeper" (even and odd frames). To generate the diffraction-limited image, these movies were summed, converted to 8 bit and images were scaled 2-fold without interpolation to match the pixel size of the reconstructed SOFI images. Time series were checked for drift and corrected with NanoJ estimate drift plugin prior to generation of sub-sequences. FRC maps were then created using 10 blocks for good sampling of cell and background areas. The average resolution was then determined by averaging FRC values of the different blocks within the cell borders. Three independent measurements were analyzed per structure. As FRC analysis failed for sum images of TOM20-labeled cells in DNA-PAINT measurements (low signal-to-noise ratio due to HILO illumination), image decorrelation analysis was performed in addition on the images derived from the sub-sequences.

Supplementary Tables
Supplementary Table 1 DNA oligonucleotide docking and imager strands used in this study. Supplementary Table 2 Primary antibodies and concentrations used in this study.