Research Article: Evaluating the pharmacological response in fluorescence microscopy images: The Δm algorithm

Date Published: February 13, 2019

Publisher: Public Library of Science

Author(s): Ana I. Gómez, Marcos Cruz, Juan F. López-Giménez, Konradin Metze.


Current drug discovery procedures require fast and effective quantification of the pharmacological response evoked in living cells by agonist compounds. In the case of G-protein coupled receptors (GPCRs), the efficacy of a particular drug to initiate the endocytosis process is related to the formation of endocytic vesicles or endosomes and their subsequent internalisation within intracellular compartments that can be observed with high spatial and temporal resolution by fluorescence microscopy techniques. Recently, an algorithm has been proposed to evaluate the pharmacological response by estimating the number of endosomes per cell on time series of images. However, the algorithm was limited by the dependence on some manually set parameters and in some cases the quality of the image does not allow a reliable detection of the endosomes. Here we propose a simple, fast and automated image analysis method—the Δm algorithm- to quantify a pharmacological response with data obtained from fluorescence microscopy experiments. This algorithm does not require individual object detection and computes the relative increment of the third order moment in fluorescence microscopy images after filtering with the Laplacian of Gaussian function. It was tested on simulations demonstrating its ability to discriminate different experimental situations according to the number and the fluorescence signal intensity of the simulated endosomes. Finally and in order to validate this methodology with real data, the algorithm was applied to several time-course experiments based on the endocytosis of the mu opioid receptor (MOP) initiated by different agonist compounds. Each drug displayed a different Δm sigmoid time-response curve and statistically significant differences were observed among drugs in terms of efficacy and kinetic parameters.

Partial Text

Recent advances in microscopy technologies have made possible to acquire large numbers of images that require new data analysis methodologies to gain insight on complex biological processes. In this sense, automatic image analysis methods aim to provide quantitative measurements from acquired images with minimal human supervision. They are of greatest interest either for drug discovery processes to quantify biochemical and/or cellular effects produced by a given compound [1] as in other applications such as diagnosis, morphology studies or gene function [2].

Here we test the performance of our algorithm on simulated endosomes added to 17 background images, {I0}, (see Fig 2 and section Application to real experiments for details). The endosomes are added based on the following approximation:

The experiments performed in [10] are used as a proof of concept to test the algorithm. Drugs diluted in physiological saline solution were perfused into the microscope chamber for internalization experiments in real time. Then, images were acquired in an inverted epifluorescence microscope. The initial image stacks consisted of nz = 9 planes of 0.49μm z-step size and nt = 15 at a rate of 1 frame per minute. The 16-bit resulting images had a resolution of 1004 × 1002 pixels (0.13μm pixel size). The maximum intensity z-projection was performed by selecting for each pixel i the maximum value across the nzz-planes. Thus, the stack is reduced to a set of nt time-course images {It}={I0,I1,…,Int−1}. Materials, receptor fusions with fluorescent proteins, generation of stable Flp-In T-REx HEK293 cell lines, cell transfection and living cell epifluorescence microscopy are detailed in [10].

In this report we propose a new algorithm to quantify pharmacological responses in fluorescence microscopy images by calculating the third order moment increment over time after convolution with a Laplacian of Gaussian filter at optimal scale. Receptor endocytosis stimulated by agonist drugs [10] has been used as a proof of concept to validate this methodology.




Leave a Reply

Your email address will not be published.