Imaging cycler microscopy (ICM) and data processing

An ICM robot (ToposNomos Ltd., Munich, Germany) was used to image random visual fields in frozen skin tissue sections at the dermo-epithelial junction, running robotically controlled, incubation – imaging – bleaching cycles on stage of an inverted epipfluorescence microscope. This can overcome the spectral resolution limit of the fluorescence microscope, as the number of cycles is principally not limited²⁵. Briefly, here we used ICM to run cycles for the co-mapping of 25 cell surface-associated proteins together with histone/cytokeratin as markers for cellular nuclei/epithelial cells for orientation in a cryo-section of an MF skin lesion. Antibodies/affinity reagents for co-mapping had been validated earlier²⁵. Their specificity is listed in Supplementary Table 1. Each cycle applies a specific FITC-conjugated antibody/affinity reagent recognizing and binding to a specific moiety in a preprogrammed sequence of robotic liquid handling on the sample of interest on stage. Together the robotic process involves: removal of an FITC-conjugated antibody/affinity reagent (probe from a probe container), transfer of the probe onto the sample, removal of this probe after incubation, washing the sample repeatedly with buffer, and fluorescence imaging of the corresponding moiety - binding site in situ of the tissue section using a CCD camera, finally repeated buffer incubations while the fluorescence signal is bleached at the excitation wavelength (472 ± 15 nm) and endpoint images after bleaching are recorded, which finalized Cycle 1. Cycle 2 is then initiated with a FITC-conjugated probe with a second binding specificity and so forth. Figure 5 shows the corresponding single-fluorescence signals per cycle of the tissue region displayed in Fig. 1(c) by exemplifying the first fluorescence signals obtained in one optical plane out of 20 imaged in z-direction of the tissue in 200 nm steps (see below). This was followed by a binarization step displaying each signal as absent or present (0/1) by using an expert-based²⁵ and automated approach for threshold detection^44,45 (Fig. 5b). The resulting combinatorial co-map contained 7,161 different topological assemblies (combinatorial molecular phenotypes, CMPs) within the tissue area of interest (Fig. 1b), where each pixel has a dimension of 200 nm × 200 nm (Supplementary Table 2). These distinct biomolecular arrangements (CMPs) together lead to a large combinatorial geometric structure (Fig. 5c, all CMPs) here referred to as binary 2D toponome map, in which distinct CMPs are displayed in different colors. For exemplified CMP annotation within this geometric structure see Fig. 5d. This method and proofs of specificity and selectivity of the mapping procedure has been described earlier^25,32,48, and its use for biomarker detection has been validated experimentally²⁵ and clinically^29,30,38,49.

(a) One optical plane out of 20 co-mapped planes in the z direction is shown to illustrate the fluorescence signals of the indicated biomolecules as in Supplementary Table 1. The visual field corresponds to the visual field shown in Fig. 1c. (b) binarization of the primary signals from (a) displayed parallel to (a). (c) 2D map of the combinatorial molecular phenotypes (CMPs) for all data points computed from the binary data set of (b). (d) two example CMPs from the data points in (b) indicated in red and blue, respectively, corresponding to the colored boxes in (b). See Supplementary Video 4 for the scheme of CMP detection and Supplementary Video 3 for realtime 3D SPIKE interrogation.