Traditionally CT and OPT require 100’s of projection images, which would normally extend the anaesthetic time required to unfeasible periods for in vivo imaging of zebrafish and so we developed a new “compressive sensing” approach based acquiring only 10’s of projection images and using an iterative computational technique to reconstruct the 3-D images². Compressive sensing OPT thus permits faster image acquisition, which is important for in vivo studies.

For imaging live adult zebrafish, we implemented our novel optical configuration with two multiplexed imaging arms that act together to increase the light collection efficiency and the spatial resolution of the system³. This provides the capability to acquire full 3-D images of adult zebrafish labelled with eGFP and mCherry with a field of view of 3 cm, requiring acquisition times on the order of a minute per spectral channel. The instrument incudes a custom designed 8 window chamber with temperature control to allow further multiplexed image acquisition of transmitted and scattered light as well as fluorescence. It can also be configured for 3-D FLIM.

Besides the challenges associated with generating, maintaining and imaging the zebrafish lines we have also had to address the challenges associated with the data management of this OPT data. To this end we have implemented an OMERO-based solution that stores “corrected raw image data” rather than rendered 3-D stacks since the former are typically ~1 Gb/channel while the latter are >240 Gb/channel. Accordingly, we have developed a semi-automatic registration programme that co-registers the two image stacks from the multiplexed cameras and writes OME-TIFF files that can be automatically uploaded to an OMERO server. We then download OPT image data sets on demand and can reconstruct the 3-D volumetric images “on-the-fly” using fast reconstruction programmes implemented in a GPU incorporated into a desktop work station.  Alternatively, we use the iterative TwIST algorithm to reconstruct the undersampled CS-OPT data sets. This OMERO-based approach lets us share data between colleagues in different buildings/institutions and is probably the most practical means to make data publically accessible.

OPT can be used with transmitted light or with fluorescence and the low illumination intensities associated with the wide-field illumination of OPT present minimal phototoxicity. This enables the same fish us to be repeatedly imaged over time, thereby enabling longitudinal studies. This is important because it can provide more useful data than studies that require fish to be sacrificed at different endpoints and it can reduce the number of fish required for drug testing or other studies.  In this project we imaged transgenic zebrafish lines that have been genetically engineered to suppress pigmentation (making them more transparent) and express fluorescent proteins to provide specific molecular contrast. The figure below shows images acquired over 26 weeks post fertilisation (wpf) from a longitudinal study of the same zebrafish expressing mCherry to label the vasculature.

In this project our key aim was to explore the use of OPT of zebrafish as a new preclinical modality to study disease mechanisms with a view to applications in drug discovery.  To establish and disease model for cancer we generated a zebrafish line expressing mCherryFP labelling vasculature (blood vessels) and eGFP labelling a liver tumour than can be chemically induced. We have shown that we can acquire in vivo 3-D images of the vasculature and tumour in zebrafish aged more than 100 dpf and have undertaken a cross-sectional study where we demonstrated that the 3-D image data we obtain from in vivo OPT correlates with data obtained using standard histopathology techniques.

We have also implemented FLIM combined with OPT in set-ups for imaging larvae and adult zebrafish and. FLIM entails exciting a fluorescent sample with a short (picosecond) optical pulse and using ultrafast time-resovled detectors to measure how fast the resulting fluorescence emission signal decays following excitation.  This “fluorescence lifetime” can tell us about the local molecular state of the fluorescent molecules. In particular, if two proteins of interest are labelled with appropriate fluorescent molecules (fluorophores), then their close proximity can lead to a transfer of energy between the fluorophores that is called “Förster resonant energy transfer” (FRET). The fluorescence lifetime of the molecule from which the energy is transferred (the “donor”) decreases when FRET occurs and this can be exploited as a robust readout of protein interactions. This phenomenon can be utilised in FRET biosensors, which are molecular constructs labelled with both “donor and “acceptor” fluorophores that change their configuration when they bind their target analyte, which can be a signalling molecule. If the change in configuration leads to a change in the distance between the donor and acceptor fluorophores, then mapping the donor fluorescence lifetime can effectively provide a map of the signalling molecules. In this project we have developed FLIM OPT to map the 3-D distribution of signalling molecules in live zebrafish. We have therefore developed a range of transgenic zebrafish lines that genetically express FRET biosensors to read out signalling molecules associated with apoptosis, (cell death), inflammation and cancer.

To date we have made the most progress mapping apoptosis in live zebrafish embryos using a FRET sensor for Caspase 3, which is a signalling molecule associated with apoptosis⁴. We have shown that apoptosis induced by gamma irradiation, can be mapped in zebrafish larvae expressing a Caspase 3 FRET biosensor using in vivo 3-D FLIM.

During this project we also demonstrated a novel approach to 3-D cell tracking during OPT acquisitions with multiplexed image acquisition that essentially entails focussing two imaging channels to the same depth in the sample during an OPT acquisition and localising features in 3-D using triangulation at each time point, i.e. each image acquisition. Thus, while the full OPT acquisition may take 100’s of seconds to acquire, dynamic features such as cells can be tracked with a time resolution equal to the frame rate of the cameras – which can be 100’s frames/second – although in practice integration times have to be sufficiently long to acquire images with adequate signal to noise ratio. The proof of principle was demonstrated using GFP-labelled neutrophils in live zebrafish embryos as illustrated below where the transmission OPT image of a live 2 days post fertilisation (dpf) zebrafish embryo tail is shown overlaid with tracks indicating the trajectories of the neutrophils over ~25 minutes following a tail wound, colour-coded by position and velocity⁵.

To extending the sample size while preserving spatial resolution of OPT, we have explored a focal scanning approach implemented with an electrically tunable lens⁶ as an alternative to multiplexed imaging arms. By rapidly scanning the imaging lens during the acquisition of each projection image, an “in focus” image is recorded throughout the sample while the out of focus background is removed through the usual Fourier back projection reconstruction process.