Florian Sizaire’s research while affiliated with French National Centre for Scientific Research and other places

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (8)


Neural network fast‐classifies biological images through features selecting to power automated microscopy
  • Article

October 2021

·

43 Reads

·

3 Citations

Journal of Microscopy

·

Florian Sizaire

·

Youssef El Habouz

·

[...]

·

Artificial intelligence is nowadays used for cell detection and classification in optical microscopy during post-acquisition analysis. The microscopes are now fully automated and next expected to be smart by making acquisition decisions based on the images. It calls for analysing them on the fly. Biology further imposes training on a reduced dataset due to cost and time to prepare the samples and have the datasets annotated by experts. We propose a real-time image processing that is compliant with these specifications by balancing accurate detection and execution performance. We characterised the images using a generic, high-dimensional feature extractor. We then classified the images using machine learning to understand the contribution of each feature in decision and execution time. We found that the non-linear-classifier random forests outperformed Fisher´s linear discriminant. More importantly, the most discriminant and time-consuming features could be excluded without significant accuracy loss, offering a substantial gain in execution time. It suggests a feature-group redundancy likely related to the biology of the observed cells. We offer a method to select fast and discriminant features. In our assay, a 79.6 2.4% accurate classification of a cell took 68.7 3.5ms (mean SD, 5-fold cross-validation nested in 10 bootstrap repeats), corresponding to 14 cells per second, dispatched into 8 phases of the cell cycle using 12 feature-groups and operating a consumer market ARM-based embedded system. A simple neural network offered similar performances paving the way to faster training and classification, using parallel execution on a general-purpose graphic processing unit. Finally, this strategy is also usable for deep neural networks paving the way to optimising these algorithms for smart microscopy. This article is protected by copyright. All rights reserved


Neural network fast-classifies biological images using features selected after their random-forests-importance to power smart microscopy

November 2020

·

37 Reads

Artificial intelligence is nowadays used for cell detection and classification in optical microscopy, during post-acquisition analysis. The microscopes are now fully automated and next expected to be smart, to make acquisition decisions based on the images. It calls for analysing them on the fly. Biology further imposes training on a reduced dataset due to cost and time to prepare the samples and have the datasets annotated by experts. We propose here a real-time image processing, compliant with these specifications by balancing accurate detection and execution performance. We characterised the images using a generic, high-dimensional feature extractor. We then classified the images using machine learning for the sake of understanding the contribution of each feature in decision and execution time. We found that the non-linear-classifier random forests outperformed Fisher’s linear discriminant. More importantly, the most discriminant and time-consuming features could be excluded without any significant loss in accuracy, offering a substantial gain in execution time. We offer a method to select fast and discriminant features. In our assay, a 79.6 ± 2.4 % accurate classification of a cell took 68.7 ± 3.5 ms (mean ± SD, 5-fold cross-validation nested in 10 bootstrap repeats), corresponding to 14 cells per second, dispatched into 8 phases of the cell cycle using 12 feature-groups and operating a consumer market ARM-based embedded system. Interestingly, a simple neural network offered similar performances paving the way to faster training and classification, using parallel execution on a general-purpose graphic processing unit.


The fast-FLIM microscope can measure FRET efficiency of the AURKA biosensor in fixed cell lines. (A). Model illustrating the mode of action of the fast-FLIM. The computer sends commands to the Inscoper electronic unit, which in turn communicates with the other devices. For each laser pulse, a trigger from the laser is sent to the, intensifier generating a gate of 2.2 ns during which the fluorescence emission is captured. The delay generator is used to position the gate at the desired timing during the fluorescence decay. After intensifying the fluorescence signal of the selected temporal gate, a CCD camera is used to acquire the time-gated image and is sent to the computer. (B). Model illustrating the generation of one fluorescence lifetime image using the AURKA biosensor in U2OS cells. A stack of fives time-gated images are acquired sequentially after controlling the delay generator to position the gate. Each image corresponds to a juxtaposed gate of 2.2 ns to cover the whole fluorescence decay. From the intensity of the five images, the pixel by pixel mean fluorescence lifetime is calculated using a discrete temporal mean of the decay to form the fluorescence lifetime image. (C). The fast-FLIM is suitable to measure FRET efficiency of the AURKA biosensor in fixed cells. (Left panel) On the left, representative fluorescence (GFP channel) of U2OS cells expressing GFP-AURKA or GFP-AURKA-mCherry, synchronized at mitosis, fixed with 4% paraformaldehyde and imaged with a 20× objective (NA = 0.7). On the right, the corresponding fluorescence lifetime images. Pseudocolour scale correspond to the mean lifetime. Scale bar, 30 μm. (Right panel) Corresponding scatter plot of the fluorescence lifetime; Each dot corresponds to the fluorescence lifetime of one cell. n = 20–30 cells per condition from three independent experiments. Blue bars represent median and interquartile. ****P < 0.0001 against the ‘GFP-AURKA’ condition. Statistical test: Wilcoxon-Mann-Whitney test.
Hardware and Software features for the acquisition of a FLIM image. From right to left and from top to bottom: from the Inscoper graphical user interface on the computer, the user set FLIM acquisition sequence which is sent to the Inscoper electronic unit. While computer wait for the acquired images, the Inscoper unit analyses the sequence parameters and set the different devices by communicating to the different microscope devices. Then the electronic unit starts the acquisition of the stack of time-gated images by alternatively setting thee delay generator and triggering the camera. All the camera images where then transferred to the computer and when the stack is completed, the computer calculate and display the FLIM image. In the same time, the Inscoper electronic unit re-start a new FLIM acquisition after setting the different devices if necessary as it was previously planned by the user at the beginning of the acquisition sequence. The process is reproduced as many times as necessary to complete the full sequence.
Multi-well and autofocus modules of the Inscoper user interface for screening applications. (A). Workflow diagram of the screening strategy. (B). Inscoper User Interface. At the top left of the windows is presented the scheme of the 96 multi-well plates. The user determines the wells of interest (where the acquisition will be carried out) by individual well, by line or by column. Then the user determines the number of random points for each well and the minimum distance between each position (as presented in the scheme at the top right). With these information, the interface calculates the coordinates of each point based on the initial calibration of the stage. At the bottom of the windows is presented the autofocus panel which permits to determine the optimal z plane for each xy position. The user determines the interval, the number of steps, the step size and the color channel used for autofocus. The algorithm then calculates the best focus by calculating a sharpness score for each acquired image (see Results section for details). The plan with the highest score is selected. To speed up the execution of the autofocus, it is possible to define a dedicated exposure time different to the FLIM images.
Batch mode analysis by cell segmentation. Analysis of fluorescence images of U2OS expressing GFP-AURKA-mCherry using an ImageJ home-made macro. From top to bottom: the user first chooses the folder containing fluorescence images, and the two destination folders for the calculated lifetime image and for the segmented intensity images with a file containing all the values of the features for each ROI. From the stack of time-gated images, the fluorescence lifetime image is generated using the threshold, size of the gate and step values settled by the user. From the first time-gated fluorescence image, the cells are segmented according to the settings defined by the user. This segmentation is applied on the fluorescence lifetime image and for each ROI (cell) is calculated the mean fluorescence lifetime, the number of pixel (area of the ROI) and the standard deviation between the pixel values. This analysis is automatically applied for all the images coming from the chosen folder.
Multiwell plate screening of AURKA biosensor. (A). (Left panel) Heatmap of the mean fluorescence lifetime values weighted by the area and the standard deviation in each well of a 96-well plate seeded with U2OS cells expressing GFP-AURKA and GFP-AURKA-mCherry. Cells have been synchronized at mitosis and fixed with PFA before being screened with the HCS-FLIM. Pseudocolour scale: lifetime values in ps. (Right panel) Corresponding scatter plot of the fluorescence lifetime values coming from all the wells expressing GFP-AURKA or GFP-AURKA-mCherry. Each dot represents lifetime of one cell and blue bars represent median and interquartile.. ****P < 0.0001 against the GFP-AURKA condition. Statistical test: Wilcoxon-Mann-Whitney test. (B). Scatter plot of fluorescence lifetime values of each cell sorted by well (from A1 to H12). Each individual point represents a single cell. Bars represent median and interquartile. (C). Scatter plot of fluorescence lifetime values of each cell sorted by inhibitor treatment. Each individual point represents a single cell. Bar represent median and interquartile.
Automated screening of AURKA activity based on a genetically encoded FRET biosensor using fluorescence lifetime imaging microscopy
  • Article
  • Publisher preview available

February 2020

·

50 Reads

·

13 Citations

Fluorescence Lifetime Imaging Microscopy (FLIM) is a robust tool to measure Förster Resonance Energy Transfer (FRET) between two fluorescent proteins, mainly when using genetically-encoded FRET biosensors. It is then possible to monitor biological processes such as kinase activity with a good spatiotemporal resolution and accuracy. Therefore, it is of interest to improve this methodology for future high content screening purposes. We here implement a time-gated FLIM microscope that can image and quantify fluorescence lifetime with a higher speed than conventional techniques such as Time-Correlated Single Photon Counting (TCSPC). We then improve our system to perform automatic screen analysis in a 96-well plate format. Moreover, we use a FRET biosensor of AURKA activity, a mitotic kinase involved in several epithelial cancers. Our results show that our system is suitable to measure FRET within our biosensor paving the way to the screening of novel compounds, potentially allowing to find new inhibitors of AURKA activity.

View access options

Développement d’un criblage automatisé de l’activité kinase d’un biosenseur Aurora A par FLIM

October 2019

·

14 Reads

La surexpression d’Aurora A est un marquer majeur de certains cancers épithéliaux. Ce gène code pour la kinase multifonctionnelle Aurora A et son activation est requise pour l’entrée et la progression vers la mitose. Jusqu'à présent, aucun inhibiteur de cet oncogène n'a été approuvé par la FDA et il est donc primordial d'identifier de nouvelles molécules. Notre équipe a développé un biosenseur FRET (Forster Resonance Energy Transfer) pour l’activité kinase d’Aurora A, constitué de la kinase entière flanquée de deux fluorophores, une GFP et une mCherry. Le changement de conformation d’Aurora A lorsqu’elle est activée rapproche les fluorophores et augmente l’efficacité du FRET. Il est ainsi possible de suivre l’activation d’Aurora A dans les cellules vivantes exprimant le biosenseur à des niveaux endogènes. Nous pouvons mesurer le FRET en utilisant la technique de FLIM (Fluorescence Lifetime Imaging Microscopy) grâce à un microscope développé dans l’équipe et appelé fastFLIM. Mes travaux de thèse ont consisté à développer une stratégie de criblage robuste et automatisée en combinant les capacités du fastFLIM et le biosenseur d’activité d’Aurora A. Cette stratégie basée sur une automatisation des acquisitions et de l’analyse de données a permis de cribler une banque de molécules en plaque 96 puits afin de trouver de potentielles inhibiteurs de l’activité kinase d’Aurora A. De plus, j’ai participé à la validation du biosenseur pour un suivi de l’activité kinase dans des cellules vivantes en montrant que les variations de FRET mesurées correspondent bien à l’état de phosphorylation d’Aurora A sur le résidu Thréonine 288, marqueur de son activation. Enfin, j’ai participé à l’élaboration de nouvelles techniques de microscopie pour suivre l’activité du biosenseur. Pour cela, j’ai utilisé un biosenseur de type homoFRET avec l’enjeu de pouvoir utiliser plusieurs biosenseurs dans un contexte multiplex. J’ai aussi utilisé la technique de 2c-FCCS (2-colors Fluorescence Cross Correlation Spectroscopy) sur le biosenseur Aurora A afin de pouvoir mesurer le FRET dans des régions où celui-ci est faiblement exprimant et dont la mesure de durée de vie de fluorescence n’est pas possible par le FLIM. Ainsi, mes travaux de thèse s’inscrivent dans la tendance à développer une microscopie quantitative et autonome avec comme enjeu d’apporter un grande nombre de données phénotypiques.


Optimized FRET Pairs and Quantification Approaches To Detect the Activation of Aurora Kinase A at Mitosis

July 2019

·

72 Reads

·

18 Citations

ACS Sensors

Genetically-encoded Förster’s Resonance Energy Transfer (FRET) biosensors are indispensable tools to sense the spatiotemporal dynamics of signal transduction pathways. Investigating the crosstalk between different signalling pathways is becoming increasingly important to follow cell development and fate programs. To this end, FRET biosensors must be optimised to monitor multiple biochemical activities simultaneously and in single cells. In addition, their sensitivity must be increased to follow their activation even when the abundance of the biosensor is low. We describe here the development of a second generation of Aurora kinase A/AURKA biosensors. First, we adapt the original AURKA biosensor –GFP-AURKA-mCherry– to multiplex FRET by using dark acceptors as ShadowG or ShadowY. Then, we use the novel superYFP acceptor protein to measure FRET by 2-colour Fluorescence Cross-Correlation Spectroscopy, in cytosolic regions where the abundance of AURKA is extremely low and undetectable with the original AURKA biosensor. These results pave the way to the use of FRET biosensors to follow AURKA activation in conjunction with substrate-based activity biosensors. In addition, they open up the possibility of tracking the activation of small pools of AURKA and its interaction with novel substrates, which would otherwise remain undetectable with classical biochemical approaches.


Fig. 2. ShadowG and ShadowY are efficient dark acceptors for mTurquoise2 in the AURKA biosensor. A. Model illustrating the mode of action of the ShadowG-AURKA-mTurquoise2 or the ShadowY-AURKA-mTurquoise2 biosensors. These biosensors switch from an open-to-close conformation upon autophosphorylation of AURKA on Thr288, bringing the donor and the acceptor in vicinity and allowing FRET detection. B. and C. (Upper panels) Representative fluorescence (mTurquoise2 channel) and Lifetime (donor onlybiosensor) images of U2OS cells expressing the indicated constructs and synchronised at mitosis. (Lower panel) Corresponding Lifetime quantification at the mitotic spindle. ShG: ShadowG; ShY: ShadowY; mTurq2: mTurquoise2. Lifetime values for individual cells are represented as black dots in each boxplot. The bar in boxplots represents the median; whiskers extend from the 10th to the 90th percentiles. n=10 cells per condition of one representative experiment (of three). Scale bar: 10 nm. ***P<0.001 against the 'AURKA-mTurquoise2' condition; a P<0.001 compared to the 'ShadowGAURKA-mTurquoise2' condition in B. or a P<0.05 compared to the 'ShadowY-AURKA-mTurquoise2' condition in C.
Fig. 3. The GFP-AURKA-mCherry biosensor does allow FRET detection by 2c-FCCS. Green and red auto-correlation curves, together with the respective cross-correlation curves issued from one representative U2OS cell expressing GFP-AURKA-mCherry (left panel) or GFP-AURKA Lys162MetmCherry (middle panel) and synchronised at mitosis. Measurements were taken in the cytosol; one independent point per condition is shown. G() represents the amplitude of the curves; time is expressed in msec. (Right panel) Ratio of the cross-correlation/green auto-correlation values for GFPAURKA-mCherry (black bar) and GFP-AURKA Lys162Met-mCherry (white bar). Each n represents the average of three independent points per cell at 0.01 msec; n=10 cells per condition of one representative experiment (of three). NS: not significant.
Fig. 4. superYFP coupled to mTurquoise2 ameliorates FRET/FLIM detection and allows FRET detection by 2c-FCCS. A. Model illustrating the mode of action of the superYFP-AURKA-mTurquoise2 biosensor. It switches from an open-to-close conformation upon autophosphorylation of AURKA on Thr288, and allowing FRET measurements. Of note, superYFP is a dimeric acceptor. B. Representative fluorescence (mTurquoise2 channel) and Lifetime (donor only-biosensor) images of U2OS cells expressing the indicated constructs and synchronised at mitosis, together with the corresponding Lifetime quantification at the mitotic spindle. sYFP: superYFP; mTurq2: mTurquoise2.
Optimised FRET pairs and quantification approaches to detect the activation of Aurora kinase A at mitosis

February 2019

·

197 Reads

Genetically-encoded Förster’s Resonance Energy Transfer (FRET) biosensors are indispensable tools to sense the spatiotemporal dynamics of signal transduction pathways. Investigating the crosstalk between different signalling pathways is becoming increasingly important to follow cell development and fate programs. To this end, FRET biosensors must be optimised to monitor multiple biochemical activities simultaneously and in single cells. In addition, their sensitivity must be increased to follow their activation even when the abundance of the biosensor is low. We describe here the development of a second generation of Aurora kinase A/AURKA biosensors. First, we adapt the original AURKA biosensor –GFP-AURKA-mCherry– to multiplex FRET by using dark acceptors as ShadowG or ShadowY. Then, we use the novel superYFP acceptor protein to measure FRET by 2-colour Fluorescence Cross-Correlation Spectroscopy, in cytosolic regions where the abundance of AURKA is extremely low and undetectable with the original AURKA biosensor. These results pave the way to the use of FRET biosensors to follow AURKA activation in conjunction with substrate-based activity biosensors. In addition, they open up the possibility of tracking the activation of small pools of AURKA and its interaction with novel substrates, which would otherwise remain undetectable with classical biochemical approaches.



A FRET biosensor reveals spatiotemporal activation and functions of aurora kinase A in living cells

September 2016

·

284 Reads

·

60 Citations

Overexpression of AURKA is a major hallmark of epithelial cancers. It encodes the multifunctional serine/threonine kinase aurora A, which is activated at metaphase and is required for cell cycle progression; assessing its activation in living cells is mandatory for next-generation drug design. We describe here a Förster’s resonance energy transfer (FRET) biosensor detecting the conformational changes of aurora kinase A induced by its autophosphorylation on Thr288. The biosensor functionally replaces the endogenous kinase in cells and allows the activation of the kinase to be followed throughout the cell cycle. Inhibiting the catalytic activity of the kinase prevents the conformational changes of the biosensor. Using this approach, we discover that aurora kinase A activates during G1 to regulate the stability of microtubules in cooperation with TPX2 and CEP192. These results demonstrate that the aurora kinase A biosensor is a powerful tool to identify new regulatory pathways controlling aurora kinase A activation.

Citations (5)


... However, in general, most samples primarily consist of (lower intensity) background, and our approach is thus widely applicable. In case a sample requires more sophisticated segmentation (e.g. the sample does not contain many objects, or the objects of interest display less discernible features), deep learning applications 11,12 can be implemented in a modular fashion. In our example, we have collected segmentation data of four labels in a single sample. ...

Reference:

Automated STED nanoscopy for high-throughput imaging of cellular structures
Neural network fast‐classifies biological images through features selecting to power automated microscopy
  • Citing Article
  • October 2021

Journal of Microscopy

... To do so and efficiently drive the microscope, we built on the solution sold by the Inscoper company (I.S. Imaging Solution, Inscoper). Of notice, this latter system dissociates the device control from the user interface; it uses a dedicated embedded system to manage the device independently from the user interface, image capture, and storage, tasked to the computer 16,24 . This specialised embedded system works autonomously during the acquisition process using parallel and bidirectional communication with the different microscope devices for optimal speed. ...

Automated screening of AURKA activity based on a genetically encoded FRET biosensor using fluorescence lifetime imaging microscopy

... 9. Incubate the cells until the next day (more than 14 h) to ensure optimal attachment to the imaging support. 10. On the next day, perform cell transfection using a transfection reagent and protocol of choice (see also the key resources table for an example of transfection reagent). ...

Optimized FRET Pairs and Quantification Approaches To Detect the Activation of Aurora Kinase A at Mitosis
  • Citing Article
  • July 2019

ACS Sensors

... FRET can occur when the emission spectrum of the donor fluorophore partially overlaps with the excitation spectrum of the acceptor, and this only when the two fluorescent moieties are in close proximity (less than 10 nm apart) [31]. This phenomenon can be used to monitor many different cellular events including the exploration of protein-protein interactions, the changes in conformation of proteins, and the up-or downregulation of signaling pathways [32,33]. With the recent advances, FRET quantification by fluorescence lifetime imaging microscopy (FLIM) became a very useful method to study molecular activities in living cells [34]. ...

FRET-Based Biosensors: Genetically Encoded Tools to Track Kinase Activity in Living Cells

... Therefore, the conformational change in the protein can be studied by monitoring the laser intensity and threshold, so as to realize the study on spatiotemporal regulation and function of proteins. 59 See the supplementary material for the up-to-date FP lasers' parameters, the qualitative analysis of the redshift between the laser and the fluorescence spectrum, and the threshold energy derivation from the threshold energy density. ...

A FRET biosensor reveals spatiotemporal activation and functions of aurora kinase A in living cells