Ilan Davis (University of Oxford)
The mechanism of embryonic primary axis formation and synaptic plasticity: advanced microscopy methods reveal localized translation of mRNA

 

Elucidating the molecular mechanisms underlying memory and learning is one of the most significant goals of neuroscience and of major importance for understanding and treating neurodegenerative diseases. Our lab has been studying the process of mRNA transport in neurons and their local translation at the synapses, which are thought to play a crucial role in memory and learning. Localized mRNA translation provides a means of regulating protein synthesis rapidly to strengthen and weaken synaptic connection in response to neuronal activation, a process known as synaptic plasticity. Remarkably, the basic molecular mechanisms and machinery responsible for mRNA transport and regulated translation in neurons is still poorly understood, although it is known to be highly conserved and shared with oocytes, in which these processes are responsible for setting up the primary body axes of the embryo prior to fertilization. We are studying the molecular basis of these events in Drosophila, as a model system for the basic processes that govern brain function. While the human brain consists of approximately 10^11 neurons, the fly brain only has about 10^5 neurons and provides a highly tractable and simplified model for studying all the basic processes underpinning the development of the brain as well as synaptic plasticity during memory and learning. First the correct number of neurons have to be specified, each with a unique lineage and cellular identity. This process is government by a precisely orchestrated set of stem cell divisions, self-renewal decisions and differentiation steps. Each neuron then sends out unique patterns of axonal and dendritic extensions, which are regulated by axon path finding mechanisms. Finally, after the exquisite pattern of neuronal connection is established, the strength of synaptic connections is regulated in individual circuits to maintain homeostasis in development and during memory and learning. In my talk, I will describe our discovery of factors common between axis specification in the oocyte and synaptic plasticity in neurons. We have used advanced live cell imaging methods, in conjunction with development of novel image analysis quantitation and visualization approaches to study the kinetics of mRNA motility along microtubules using molecular motors. I will also describe our unpublished work, using machine learning methods to characterize the role of a conserved mRNA binding protein, Syncrip, in regulating the correct number of neurons in the brain, following asymmetric stem cell divisions. I will also describe our unpublished work showing that the same factor plays an important role in synaptic plasticity. My lab’s work is strengthened by the presence of a PhD programme in biomedical imaging (http//:onbicdt.org) and Micron Oxford (http//:micronoxford.com), which I direct. Micron is an interdisciplinary unit funded by a Strategic Award from the Wellcome Trust, to develop and apply advanced imaging approaches to study cellular dynamics. This includes all aspects of advanced microscopy, from specimen preparation and probe development, to bespoke instrument design and creation of novel image analysis and quantitative approaches.

 

                             João Sanches (Technical University of Lisbon)  

Biological modelling and quantification from fluorescence imaging

 

Fluorescence is the basis of several microscope image modalities extensively used in biological and medical research. Biological image processing has been used mainly for restoration and segmentation purposes for morphological analysis, counting and tracking of cells and organelles or simply to improve visual inspection of the biological experiments.

In this presentation, quantitative measures for the characterization of the distribution of tagged molecules within the cell are analyzed in order to predict protein functionally. In this scope two main difficulties need to be addressed:

1) The multiplicative noise that corrupts these images, related with the photon-counting process at the detectors and

2) Geometric compensation to compute distribution measures invariant to cell shape and size variability.

A set of numerical features, extracted from 1D typical profiles, are described to characterize the distribution of the molecules in the intra and inter cellular space. These features should be invariant to the morphological variability of the cells but they should be able to detect trafficking and functional disorders associated with the tagged molecules.

The ultimate goal is to use this bioimaging analyses as a screening method to detect mutations on the coding genes of these molecules that produce dysfunctional molecules that prevent them to accomplish its function. These dysfunctions can be related with severe medical conditions, such as cancer.

 

 

                             Christian Soeller (University of Exeter)  

Biological modelling and quantification from fluorescence imaging

 

For some time we have been using fluorescence imaging techniques to determine structural detail and protein distributions in muscle with the goal to improve our understanding of muscle biophysics. Starting with confocal and multiphoton approaches we developed quantitative techniques to measure sub-resolution structures in cells such as the microscopic transverse tubules. More recently we have employed optical super-resolution techniques based on the localisation of individual fluorescent markers. Localisation microscopy has allowed us to measure the distribution of large proteins such as the cardiac ryanodine receptor quantitatively due to the high resolution and the known density of receptors in a quasi-crystalline packing. In general, however, quantitative imaging is more problematic and the stochastic variability of single molecule localization and photoswitching complicates the interpretation of images which we will illustrate with some examples and simulations. Both the confocal data sets and the super-resolution data help guide the development of detailed mathematical models of cells. We are developing approaches to assist the construction of computer models from our data and allow the computational fusion of data sets from different modalities (such as confocal and electron microscopy). These approaches use spatial statistics to construct model distributions that are statistically compatible with the recorded data.

 

                             Dylan Owen (Kings College London)  

Investigating protein clustering during T-cell activation using PALM and quantitative statistical cluster analysis

 

Super-resolution PALM and dSTORM microscopy generate lists of localised molecular coordinates which can be displayed as 2-dimensional point patterns. For biological applications, it is frequently required to interrogate such data sets to analyse molecular clustering. Here, we demonstrate quantitative cluster analysis of PALM and dSTORM data sets based on Ripley’s K-function. We show that this method can extract a range of cluster parameters such as size, number of molecules per cluster and so on, on both simulated and experimental data. We also demonstrate the technique in 3-dimensions and a 2-colour cross variant to detect the degree of cluster colocalisation in 2-channel PALM and dSTORM data sets. We demonstrate these techniques by examining molecular organisation at the T cell immunological synapse. Here, the generation of microclusters of signalling proteins such as the kinase Lck and the scaffold protein LAT after T cell receptor engagement is critical to T cell signalling efficiency and hence to immune system function. A set of numerical features, extracted from 1D typical profiles, are described to characterize the distribution of the molecules in the intra and inter cellular space. These features should be invariant to the morphological variability of the cells but they should be able to detect trafficking and functional disorders associated with the tagged molecules.

The ultimate goal is to use this bioimaging analyses as a screening method to detect mutations on the coding genes of these molecules that produce dysfunctional molecules that prevent them to accomplish its function. These dysfunctions can be related with severe medical conditions, such as cancer.

 

                              Lunch (3^rd Floor Atrium)

 

                             Susan Cox (Kings College London)  

Accelerating localisation microscopy

 

Localisation microscopy is a powerful tool for imaging structures at a lengthscale of tens of nm, but its utility for live cell imaging is limited by the time it takes to acquire the data needed for a super-resolution image. The acquisition time can be cut by more than two orders of magnitude by using advanced algorithms which can analyse dense data, trading off acquisition and processing time. We have developed two methods which allow different trade-offs to be made.

Modelling the entire localisation microscopy dataset using a Hidden Markov Model allows localisation information to be extracted from extremely dense datasets. This Bayesian analysis of blinking and bleaching (3B) is able to image dynamic processes in live cells at a timescale of a few seconds, though it is very computationally intensive, requiring at least several hours of analysis.

Analysis speed can be improved over 3B by a factor of ten by instead modelling the data using an alternative statistical approach, which automatically determines the number, position, and brightness of fluorescing molecules within a particular image region. This method treats the background noise as a parameter to be found, avoiding the background removal step in 3B or the need to hand set thresholds in more conventional analysis techniques.

Our methods are demonstrated on various live cell systems, including cardiac myocytes and podosomes, showing a resolution of tens of nm with acquisition times down to a second. We also compare our methods to other high density algorithms and discuss the artefacts which can occur during reconstruction of the super-resolution image.

 

                             Jean-Baptiste Sibarita (Interdisciplinary Institute for Neuroscience Bordeaux)  

Resolving molecular organization and dynamics using localization-based super-resolution microscopy

 

Deciphering molecular organization and activity at the molecular level has become possible thanks to the recent advent of super-resolution microscopy. Techniques based on the sequential stochastic photo-conversion of sparse subsets of single fluorophores, i.e. (f)PALM or (d)STORM, rely on the ability of determining the centre of the point spread function (PSF) created by each single point emitter. They have become extremely popular due to their affordability and relatively simple implementation on a conventional microscope. They allow the localization and tracking of a large number of biomolecules with close to molecular accuracy (down to 10 nm in lateral and 40 nm in axial) and millisecond scale temporal resolution. Nevertheless, a major step relies in the image analysis, often time-consuming and not easy to handle by non-specialists.

First, practical implementations enabling the use of such powerful techniques in routine will be detailed. The, as an illustration, a novel organization and dynamic characterization of post-synaptic receptors within live neurons will be presented, revealing that AMPA receptors are highly concentrated in nanodomains, instead of diffusively distributed in the PSD as generally thought.

 

 

                            Ed Cohen (Imperial College London)  

The effect of image registration on the localization accuracy of a single molecule

 

Image registration (the combining of two or more images of the same scene) is an important processing step in fluorescence microscopy, for example in tracking or super-resolution methods. Recent advancements have made it possible to localise a single molecule in a cellular environment and it is therefore important to know the effect that registration has on the accuracy of localizing a single molecule. Assuming an affine transform, control points (CPs) are used to solve a multivariate regression problem. Typically in the biological sciences this is solved with linear least squares. However, with measurement errors existing in the localization of both sets of CPs this is an errors-in-variable problem and linear least squares is inappropriate; the correct method being generalized least squares. To allow for point dependent errors the equivalence of a generalized maximum likelihood and heteroscedastic generalized least squares model is achieved allowing previously published asymptotic results to be extended to image registration.

For a particularly useful model of heteroscedastic noise where covariance matrices are scalar multiples of a known matrix (including the case where covariance matrices are multiples of the identity) we provide closed form solutions to estimators and derive their distribution, thus allowing distributions for the errors involved in localising registered single molecules to be derived in terms of CP numbers and their associated photon counts.

 

                             David Weston (Birkbeck, University of London)  

Analysing Spatial Point Patterns in Nuclear Biology using Aggregate Maps

 

There have been many investigations into identifying possible relationships between the location of bodies within a cell nucleus and their function. Interesting relationships are identified by comparing how different the locations of the bodies are to what is expected if the locations were chosen at random. However, the number of bodies involved is often very low and this has consequences on the effectiveness of quantitative analysis procedures. It becomes increasingly difficult to distinguish between bodies whose locations have been chosen at random and bodies whose locations have a biologically interesting preference, with decreasing number of bodies involved. Therefore a commonly used approach, which is to analyse cells individually, has the potential to overlook interesting structures.

An alternative approach is to aggregate the locations of bodies from multiple cells using simple normalization, but this requires care to choose the appropriate normalization.  It is to address this issue that `Aggregate Maps' has been proposed. An aggregate map for a collection of cells is constructed simply by fusing the images of individual cells using standard methods for image registration.

An investigation, using this methodology, of the spatial preference of nuclear compartments in mammalian fibroblasts will be described.