Poster U13 - Development and Application of a Python-based Image Analysis Pipeline for Phenotypic Screening Reveals Reductions in Calcium Efflux from the Endoplasmic Reticulum Following Bclw Knockout

Introduction: The endoplasmic reticulum (ER) plays an essential role in regulating intracellular calcium levels to achieve a fine balance between amplifying pro-survival calcium signals and initiating degenerative cascades. Monitoring calcium efflux at the single cell level in response to stimulation provides a powerful approach for identifying targets in genetic or drug screens. However, implementation of this approach for high-throughput and high-content screening is often limited by a lack of established image analysis pipelines that automate quantification of timelapse microscopy data from live cells. Challenges frequently faced by experimenters include image segmentation to break fields of view into single cells without introducing sampling bias, extraction of timelapse data for each cell from large datasets, and analysis of time series data to quantify transient properties. Various Python packages have been developed to address each of these challenges individually, but few groups have produced readily available frameworks for using these packages in a unified workflow that can be easily tailored to specific experimenter needs. To address this technological gap, we developed a user informed Python-based image analysis pipeline for quantifying time-dependent changes in single cell fluorescence signal intensity. Here we demonstrate the application of this tool towards identifying phenotypic differences in ATP-evoked calcium efflux following loss of anti-apoptotic Bclw in murine embryonic fibroblasts (MEFs) expressing a previously published genetically encoded ER-targeted calcium indicator.

Materials and

Methods: MEFs were generated from epidermal tissue of single Bclw(GtRosa41) embryos (E12), genotyped for zygosity, single Bclw+/+ (WT) and Bclw-/- (KO) cell lines chosen, immortalized via SV40 transformation, and maintained as distinct cell lines. The presence or absence of BCLw protein was confirmed by western blot analysis of lysate from WT and KO cells. To derive versions of these cell lines stably expressing an ER-targeted genetically encoded calcium indicator GCaMP6s variant (ER-GCaMP6s-150), we purchased custom lentivirus bearing the coding sequence for this construct as cloned from a published AddGene source (de Juan-Sanz et al., 2017) as well as the puromycin resistance gene (VectorBuilder). ER-GCaMP6s-150 expressing cells were selected by puromycin treatment for five days, cell lines expanded and then cryopreserved until the time of experimentation. To measure calcium efflux, cells were cultured on glass coverslips using standard approaches and imaged in a recording chamber (mounted to an inverted widefield fluorescence microscope, 470/525 ex/em), using a 60x (NA1.4) oil immersion lens at 1-second intervals for 5 minutes before, during, and after bath perfusion with 10 uM ATP in calcium-free, divalent ion balanced saline solution (CFES, 320 mOsm, pH 7.4). Timelapse files were imported into ImageJ (NIH) with Bio-Formats or Python with N2reader or Skimage. All analysis code was written in Python (Jupyter Lab) using the packages Cellpose for segmentation and SciPy for signal processing (in addition to the standard set of data science packages) and maintained in a dedicated repository with GitHub.

Results, Conclusions, and Discussions: Following transduction with our custom designed, commercially produced lentivirus containing the ER-GCaMP6s-150 coding sequence from de Juan-Sanz et al., (2017) we found that the genetically encoded calcium indicator (GECI) was expressed at relatively even levels between genotypes, regardless of passage number. To validate GECI localization, we stained live cells with MitoTracker CMXRos and imaged at high resolution (Fig. 1A). Extracellularly applied ATP is known to evoke calcium efflux from ER stores via the purnergic signaling pathway, and we hypothesized that BCLw plays a functional role in modulating the magnitude of calcium release. We developed a calcium flux assay to test whether this puringeric signaling functioned as expected in this system (Fig. 1B), and qualitatively observed predicted decreases in signal intensity following bath perfusion with 10 uM ATP. With this assay we generated a microscopy dataset of timelapse recordings from WT or KO MEFs for testing in our analysis pipeline (Fig. 2A). We found that by taking a maximum projection of the timelapse file we could generate an appropriate seed image for use in Cellpose to segment the field of view into individual ROIs from which the mean pixel intensity would be calculated within each ROI, for every time point, and then represented as a heat map for user review (Fig. 2B). By adding unique identifiers to each ROI we created a point for users to generate exclusion and inclusion criteria. In plotting this timeseries data for each genotype, we observed subtle qualitative differences in their responsiveness to ATP (Fig. 2C). To quantify these phenotypic changes, we wrote analysis code to process percent change signals and identify peaks, global maxima, and calculate the area under the curve (Fig. 3A). We used this code to analyze responses across more than 200 cells per genotype and find that loss of Bclw significantly diminishes responsiveness to stimulation with ATP (Fig. 3B). In conclusion, this pipeline reduces the experimental burden for follow up work to further define the mechanisms by which Bclw contributes to ATP-sensitivity. This resource provides a generalizable framework for researchers wanting to increase yield or content in microscopy based single cell assays.

Acknowledgements (Optional): The authors thank the Core for Imaging Technology & Education (CITE) at Harvard Medical School for their support in imaging assay and analysis pipeline development. This work was supported by NIH grants R01CA205255 to R.A.S. and K00AG078230 to J.A.F., as well as a Bertarelli Foundation award to R.A.S.