Localization-based super-resolution imaging meets high-content screening

Abstract

Single-molecule localization microscopy techniques have proven to be essential tools for quantitatively monitoring biological processes at unprecedented spatial resolution. However, these techniques are very low throughput and are not yet compatible with fully automated, multiparametric cellular assays. This shortcoming is primarily due to the huge amount of data generated during imaging and the lack of software for automation and dedicated data mining. We describe an automated quantitative single-molecule-based super-resolution methodology that operates in standard multiwell plates and uses analysis based on high-content screening and data-mining software. The workflow is compatible with fixed- and live-cell imaging and allows extraction of quantitative data like fluorophore photophysics, protein clustering or dynamic behavior of biomolecules. We demonstrate that the method is compatible with high-content screening using 3D dSTORM and DNA-PAINT based super-resolution microscopy as well as single-particle tracking.

Accelerating single molecule localization microscopy through parallel processing on a high‐performance computing cluster

“…For the high throughput dSTORM reported in Beghin et al . (), the authors argued that postprocessing the SMLM data would lead to unacceptable delays in the workflow and so they acquired and processed their dSTORM data on‐the‐fly, making the use of rapid GPU accelerated analysis software (Kechkar et al ., ). Eight hours were required to image 96 cells in 96 wells and they observed that the buffers used to induce the fluorophore blinking led to degradation of the imaging performance over time (with ∼10 h being a practical limit for a dSTORM experiment).…”

“…Higher throughput automated SMLM has already been demonstrated in pioneering work using PALM (Holden et al ., ), dSTORM and DNA PAINT (Beghin et al ., ) and could be beneficial for many SRM‐based studies but the image data processing presents challenges with respect to long processing times that scale with the imaging throughput achieved. For the high throughput dSTORM reported in Beghin et al .…”

“…This could be addressed by imaging more cells in larger FOV and we note that the use of multimode optical fibres to efficiently deliver 100 s mW of excitation power from low‐cost multimode diode lasers to the sample enables dSTORM of FOV >100 × 100 μm to be routinely acquired (Kwakwa et al ., ). Comparing this to the 20.5 × 20.5 μm FOV reported for (Beghin et al ., ) suggests that high throughput dSTORM can be significantly accelerated if SMLM data are acquired for larger FOV (i.e. containing many more cells), but this would significantly increase the burden of data processing and so would increase the total acquisition time required if the SMLM data are processed on the fly.…”

“…Higher throughput automated SMLM has already been demonstrated in pioneering work using PALM (Holden et al, 2014), dSTORM and DNA PAINT (Beghin et al, 2017) and could be beneficial for many SRM-based studies but the image data processing presents challenges with respect to long processing times that scale with the imaging throughput achieved. For the high throughput dSTORM reported in Beghin et al (2017), the authors argued that postprocessing the SMLM data would lead to unacceptable delays in the workflow and so they acquired and processed their dSTORM data on-the-fly, making the use of rapid GPU accelerated analysis software (Kechkar et al, 2013). Eight hours were required to image 96 cells in 96 wells and they observed that the buffers used to induce the fluorophore blinking led to degradation of the imaging performance over time (with ß10 h being a practical limit for a dSTORM experiment).…”

An Integrated Platform for High-Throughput Nanoscopy

“…Complex biological investigations involving more subtle phenomena, large numbers of imaging targets, or many experimental conditions have required extensive data collection e↵orts spanning months to years [2]. Early automation e↵orts [3][4][5] have helped relieve the tedium of acquiring large numbers of SMS images, but have not increased the speed of either data acquisition or analysis and as a consequence have been most successful in applications imaging small structures in bacteria or yeast where large numbers of cells and structures of interest can be imaged within one field of view (FOV).…”