RNAiscreener software

rnaiscreener plate montage

Background

Background

 

Large scale studies to investigate mechanisms and regulators of cell-cell contacts have previously been delayed by the absence of quantitative tools specific for epithelial structures and junctional markers. Here we set out to optimize methodology relevant for phenotypic analysis of epithelial structures in RNAi screens to identify new regulators and pathways required for junction stabilization. In spite of extensive progress in image analysis and the availability of increasingly sophisticated tools, automated segmentation and quantification of junctional proteins in a confluent epithelial sheet have been challenging. In particular, high-throughput algorithms have not yet been established to segment F-actin at junctions and thin bundles - parameters with well-established biological relevance for epithelial junctions.

To quantitatively monitor the effectiveness of cell-cell adhesion assembly in an epithelial monolayer, we developed a workflow to quantify the levels of E-cadherin at junctions, junctional actin and peripheral thin bundles.  The workflow is useful to segment any marker that localised at cell-cell contacts.  A parameter for quality control of cell confluence was also validated to automatically eliminate images containing gaps in the monolayer (likely to result in false-negative data). This control ensures that a confluent monolayer was used in all data points in all experiments.

The complete workflow and validation is described in the Supplementary Information paper ‘Erasmus et al (2016) Defining Functional Interactions during biogenesis of epithelial junctions. Nature Commun. In press (DOI: 10.1038/ncomms13542)

rnaiscreener histogram montage

RNAiscreener Software

To support the specific image analysis required to determine cellular phenotype at junctions, we developed a tool called RNAiscreener.This segmentation software takes as input, tif image stacks of 96 well plates from an RNAi screen, and generates three experimental parameters representative of E-cadherin (E-cad), junctional actin (i.e. co-localizing with E-cadherin and named Jun-A) or cytoplasmic actin (i.e. excluding junction and nucleus area; Cyt-A).

Analysis was designed using MBF-ImageJ (http://imagej.nih.gov/ij/) and Metamorph 6.2 (Molecular Devices).  E-cadherin image was thresholded to minimize contribution of cytoplasmic staining (E-cad parameter) A caveat of this parameter is that a positive value may reflect increased cadherin levels at junctions and/or in the cytoplasm. Thresholded E-cadherin image was used as a mask to segment junctional actin pool (Jun-A) from the corresponding image stained with phalloidin (total F-actin). Cytoplasmic actin (Cyt-A) was obtained by subtraction of E-cadherin mask and nucleus image from the total F-actin image.

To quantify the amount of specific proteins at junctions, E-cadherin images were individually thresholded to (i) minimize contribution of cytoplasmic staining, and (ii)maximize junction coverage. The thresholded binary E-cadherin image was dilated and used as a mask to segment the junctional pool from the corresponding F-actin staining image (or any other marker that localises at cell-cell contacts). A global threshold across all experimental conditions was applied to the segmented images to measure the marker localization atjunctions and calculate % thresholded area.

Validation was undertaken using an image set with different levels of disruption of the actin cytoskeleton or E-cadherin at junctions. Algorithms were validated to show a strong correlation between E-cad and Jun-A parameters and between total F-actin with Cyt-A.  There is no correlation with with cell numbers in each sample for any parameters.

Output of the software is an excel spreadsheet with measurements in for each parameter (intensity and %area) and distinct markers in each well processed.

  • Extended protocols for the RNAi screen are available from <coming soon>
  • The image set used for software validation is available to down load from <coming soon>
  • The  RNAiscreener code is written in Java  and is available as a jar file for local installation upon request to Chris Tomlinson