Analysis of Fluorescence Microscopy Super-Resolution Data of Protein Assemblies

Abstract

In super-resolution microscopy applications, previously unknown fluorescent signal patterns are recorded. The interpretation of such nanoscale data is often unexpectedly complex, and can be performed by different analysis strategies: 1) empirical statistics of the spatial distribution of intensity values to identify local objects, 2) inverse problem approaches to convert signals into objects based on external data models or other external assumptions, 3) decomposition of spatial signal patterns into spatial modes.Here, we compare and relate two distinct approaches of analysing STED microscopy images of RyR2 clusters (cardiac ryanodine receptor type 2). RyR2 Ca2+ release channels are essential for heart muscle function (excitation-contraction coupling). Yet, the nature of lateral channel organization within super-structural clusters is unknown and important for models of local control mechanisms of RyR2 Ca2+ release activity.We established multi-scale analysis of RyR2 signal patterns employing wavelet analysis. This analysis decomposes the initial image with predefined wavelets into spatial modes identifying dominant scales of signal fluctuations. We tested the sensitivity of this approach for different wavelets with artificial and modified images. Distinct scales represent inter-cluster spacing and intra-cluster patterns, respectively. Furthermore, we compare the spatial mode analysis with object-based approaches. For object-based analysis, RyR2 cluster sub-structures were identified with a multi-step thresholding procedure. After increasing the threshold level step-by-step, the hierarchy of the segmentation output was analysed with logical operators. Accordingly, we identified cluster sub-structures as discrete objects of variable sizes with typical spacings ranging from 78 to 128 nm (IQR range) that we interpret as individual cluster building-blocks.We conclude that identification of common protein cluster building principles in highly variable signal structures as typical for RyR2 clusters benefits from combining object-based approaches with spatial mode analysis.