PySTACHIO: Python Single-molecule TrAcking stoiCHiometry Intensity and simulatiOn, a flexible, extensible, beginner-friendly and optimized program for analysis of single-molecule microscopy data

Abstract

As camera pixel sensors grow larger and faster and optical microscopy techniques become ever more refined, there has been explosions in the quantity of data acquired during routine light microscopy. At the single-molecule level, this analysis involves multiple steps and can quickly become computationally expensive and intractable on ordinary office workstations. Moreover, complex bespoke software can present high activation barriers for new users. Here, we present our recent efforts to redevelop our quantitative single-molecule analysis routines into an optimized and extensible Python program, with both GUI and command-line implementations to facilitate its use on both local machines and remote clusters, and by beginners and advanced users alike. We demonstrate the performance of this code matches our previous MATLAB implementation but at a fraction of the computational cost. We show the code can extract fluorescence intensity values corresponding to single reporter dye molecules and, using these, to estimate molecular stoichiometries and single cell copy numbers of fluorescently labeled biomolecules. It can also evaluate diffusion coefficients for the relatively short single-particle tracking data that is characteristic of time-resolved image stacks. To facilitate benchmarking against other codes, we also include data simulation routines which may trivially be used to compare different analysis programs. Finally, we show that PySTACHIO works also with two-color data and can perform colocalization analysis based on overlap integrals, to infer interactions between differently labelled biomolecules. We hope that by making this freely available for use and modification we can make complex single-molecule analysis of light microscopy data more accessible.

Investigating molecular crowding during cell division in budding yeast with FRET

“…The two sensors are identical except for the fluorescent proteins used, with crGE2.3 making use of mEGFP and mScarletI as donor and acceptor respectively to make use of the greater single-protein intensities compared to mCerulean3 and mCitrine (S. N. Mouton et al, 2020), with the CrGE2.3 thus being suitable for single-molecule tracking. Here we used our redeveloped single molecule tracking code PySTACHIO (Shepherd et al, 2021) which performs two-channel tracking as well as colocalization analysis using overlap integrals alongside straightforward distance cutoffs as described in the Methods section. The localization was used to calculate the normalized fret NFRET which is possible for this FRET pair due to low spectral overlap and cross-excitation.…”

“…Data was analyzed using PySTACHIO (Shepherd et al, 2021) with snr_min_threshold=0.5 and struct_disk_radius=9 in Alternating Laser Excitation (ALEX) mode. We relaxed our usual constraints on trajectory length because ALEX mode with 10 ms exposure allows considerable diffusion between successive captures in one channel and thus trajectory linking is compromized in this single molecule regime.…”

“…We additionally present a new imaging method with combining crowding sensing with mCerulean3-mCitrine FRET sensor and FM4-64, a fluorescent dye that labels endolysosomal lipids (Vida & Emr, 1995), allowing simultaneous visualization of the vacuole membrane. Finally, we present our most recent data in which we have successfully quantified single molecule FRET in vivo using the crGE2.3 crowding sensor and our super-resolution image analysis software suite PySTACHIO (Shepherd, Higgins, et al, 2021a).…”

Combining single-molecule super-resolved localization microscopy with fluorescence polarization imaging to study cellular processes

“…The E. coli aggresome data was imaged until it was fully bleached and in the photoblinking regime and was reconstructed by registering both channels on to each other and summing. This reconstructed single-channel image was then analyzed with our new Python single-molecule tracking code PySTACHIO [49] which plots the integrated foci intensity and finds the peak of a kernel density estimation fit to the intensity distribution. We also checked this against the surfaceimmobilized mGFP data and both were found to give a consistent Isingle value around 130-140 integrated pixel values, equivalent to a quotient of 70 ± 8 (mean ± s.d.)…”

“…With post-processing software written in Python, we spatially register the two polarization detection channels using sub-pixel phase cross correlation-based transformation functions, and find the total integrated pixel intensity for the lateral component of each detected diffraction-limited fluorescent focus in each polarization channel, converting this into a polarization value, on a fluorescent molecule-by-molecule basis. By reconstructing the two-channel image into a single channel, we can also estimate the molecular stoichiometry of in vivo protein complexes by measuring the initial intensity of each fluorescent focus prior to any photobleaching and then normalizing this against the measured total integrated intensity of a single fluorophore [48], denoted here as the Isingle value, using our Python implementation of the fluorescent foci tracking and stoichiometry quantification algorithm [49].…”