Shedding light on the molecular interactomes of fast, complex biological processes using multispectral imaging with uncompromised spatiotemporal resolution

Abstract

The hallmarks of life are that it is animate and complex. Advances in live-cell, camera-based fluorescence microscopy have enabled the field to shed light on the spatiotemporal nature of fundamental intracellular processes. However, despite these advances, due to the broad emission spectra of conventional fluorophores only three or fewer fluorescently-labelled structures can be observed in live cells using the majority of microscopes, which is far fewer than the dozens or even hundreds of structures orchestrating biological processes. Moreover, imaging more than one labelled structure introduces a cost in temporal resolution as labels are imaged sequentially in independent channels. These compounding issues drastically limit our understanding of highly-dynamic, complex biological processes, for instance, endosomal receptor sorting. This process occurs within a few seconds and is regulated by tens, or hundreds, of components on the endosomal membrane. Furthermore, these endosomes themselves move on the µm/sec range in 3D within the cell, making this process impossible to mechanistically study, in sufficient molecular and temporal detail, using pre-existing technology. Therefore, to reach a true molecular understanding of fast, complex biological processes, such as endosomal sorting, the interaction of many molecular machineries must be observed simultaneously such that no cost associated with spatiotemporal resolution is introduced by imaging several structures. Multispectral imaging appeared as a promising avenue towards this goal as it tackles issues associated with spectral overlap thereby affording the visualisation of many fluorophores within a sample. However, current methods of acquiring multispectral data either: i) acquire spectral data simultaneously but acquire spatial data sequentially (e.g. point-scanning techniques), or ii) acquire the spatial data simultaneously (e.g. using a camera) but acquire the spectral data sequentially. The ability to acquire multispectral and spatial data in the same snapshot, using camera-based data capture, would facilitate many structures to be observed simultaneously to capture complex biological interactions on fast timescales. Developing technology to facilitate this goal is the overarching aim of this work. In my PhD project, I developed an iterative, spectral unmixing algorithm which can spectrally unmix noisy multispectral datasets such as those deriving from live-cell imaging experiments. This algorithm outperforms conventional spectral unmixing approaches for unmixing low-signal multispectral data thereby affording simultaneous multispectral data capture onto an array of cameras after spectral fractionation of emission light. Towards this goal, I designed and built an eight-channel multispectral acquisition module, which is compatible with any camera-based fluorescence microscope by replacing the native camera. Furthermore, this system was designed to match the spatial resolution of the parent instrument, and offers greater photon-efficiency as emission filters are not required. This system enables the acquisition of up to eight fluorophores simultaneously, whilst the resulting data is spectrally unmixed using the iterative, spectral unmixing algorithm. The multispectral acquisition module was integrated with multiple microscopes, such as a spinning disk confocal microscope and an oblique plane light sheet microscope. These instruments afforded continuous eight-channel timelapses in 2D using the multispectral spinning disk and eight-channel whole cell volumetric acquisition at four volumes/second speed using the multispectral light sheet. Multispectral volumetric light sheet projections were also carried out on this system allowing for eight-channel summed projections to be obtained at the tens of volumes/second timescale. These instruments provided a unique opportunity for studying fast, complex subcellular interactions and the resulting datasets illuminated a plethora of biological insights occurring between dynamic, subcellular compartments. I applied the technology to the biological process it was specifically designed to image, namely endosomal sorting. Through collaborations within the lab and worldwide, I established labelling strategies for the simultaneous visualisation of multiple receptor cargoes within endosomes in living cells at endogenous levels. Ultimately, the unique synergy provided by the new software, hardware and labelling strategies allowed for the direct visualisation of receptor sorting in vivo paving the way for the molecular dissection of this phenomenon. Alas, it must be emphasised that the technological developments presented within this thesis were used to shine light on trafficking within the cell, however, the technology is designed to be general and should enable the study of a great variety of other fast, complex biological processes.