High content analysis of 3-D cell cultures with multidimensional fluorescence imaging
An Industrial Partnership Award with GlaxoSmithKline PLC funded by the BBSRC (BB/M006786/1)
Professor Paul French(PI), Dr Christopher Dunsby - Photonics Group, Physics Department
Professor Tony Magee - Facility for Light Microscopy (FILM), NHLI
Professor Ed Tate - Chemistry Department
Professor David Carling, Dr Alessandro Sardini - MRC Clinical Science Centre
Dr Martin Rüdiger - GlaxoSmithKline PLC
The response of cells to stimuli depends on cell signalling processes, which are combinations of molecular interactions. Disease is associated with deviations from normal signalling processes and so reading out molecular interactions in cells can help elucidate mechanisms of disease and also provide a means to evaluate potential therapeutic drugs. Many of the signalling molecules in cells are proteins and their interactions are widely studied using microscopy with proteins of interest being labelled with fluorescent molecules ("fluorophores"). For drug discovery, fluorescence imaging of arrays of cells is automated so that the effect of many compounds on cell signalling processes can be "assayed". Fluorophores are "excited" by radiation at a wavelength at which they absorb and the resulting characteristic emission (fluorescence) is recorded using an imaging detector. By labelling different kinds of protein with different fluorophores and comparing the images at their respective emission wavelengths, it is possible to learn about protein interactions by observing when they appear in the same place at the same time. Unfortunately, this "co-localisation" is usually limited by diffraction of light to a resolution of a few 100 nm - much larger than the size of typical signalling proteins (~1-10 nm). It is possible, however, to confirm interactions using Förster Resonant Energy transfer (FRET), which entails labelling the proteins with different fluorophores and observing when energy is transferred between them. This energy transfer can only occur if they are within ~10 nm of each other. The most quantitative way to read out interactions using FRET is through the reduction in the fluorescence lifetime of the "donor" fluorophore as it loses energy due to the energy transfer. Fluorescence lifetime measurements can be made in every pixel and this fluorescence lifetime imaging (FLIM) enables protein interactions to be mapped in space and time. Fluorescence lifetime can also be used to distinguish different molecular species or different states of naturally fluorescent molecules that are involved in regulating the consumption of energy in the cell, which can also be altered by disease.
For cell biology research and for drug discovery, fluorescence-based studies typically involve imaging thin - and therefore transparent - layers of cells. Unfortunately, this is not a physiologically normal environment for cells and they often behave differently compared to when they are in their normal 3-D tissue context. However, it is highly challenging to image cells in native biological tissue and even more so to realise this in a high throughput mode for drug testing. Instead, there is increasing interest in assaying synthetic 3-D cultures comprising many cells that interact with each other, presenting behaviour reminiscent of that in native tissue with greater optically accessibility - although unfortunately they can scatter and absorb light more strongly than thin layers of cells. Such 3-D cell cultures can be arrayed for rapid imaging, however, and we propose to develop an automated platform to provide 3-D images of such cell cultures, utilising FLIM to read out molecular interactions. For this we will optimise the 3-D cell culture and labelling methodologies for fluorescence imaging and will develop and evaluate automated microscopes for FLIM-based assays. For larger samples we will utilise multiphoton excitation, which entails illuminating the fluorophores at twice the wavelength usually required for excitation, such that they need two photons arriving simultaneously. This two photon absorption is intensity-dependent and so can be arranged to occur only in the focal plane of a scanning laser beam such that the emitted photons all originate from a specific depth in the sample. Scanning the focal plane then enables 3-D imaging. The longer wavelength light is less phototoxic and is scattered less by the sample, thus enabling deeper imaging
Progress to date
Growth of 3-D cell cultures
To-date, we have explored three different growth techniques:
- Spheroids cultured in 96-well plates where the base has been coated with agarose to prevent cell adhesion. We have tested this method with HEK293T cells, prostatic adenocarcinoma cells (DU145) and mouse embryonic fibroblasts (NIH3T3) cell lines. NIH3T3 cells were found not to form spheroids. Spheroids were successfully grown with the other two lines. The meniscus formed by the agarose at the bottom of the well enables us to obtain a single spheroid per well. Following growth, spheroids are transferred to a glass-bottomed imaging plate since the agarose coating is optically thicker than the working distance of the microscope objective.
- Spheroids cultured in u-bottomed 96-well plates with bottom coated with a PHEMA hydrogel to prevent cell adhesion. We have tested this method with HEK293T and DU145 cell lines. The u-bottomed plates are not compatible with imaging and so the spheroids are transferred to an imaging plate. If we use multiwell plates with a flat base, this method produces multiple spheroids per well. Single spheroids should be obtained if round-bottomed PHEMA coated 96-well plates are used.
- Spheroids grown in an agarose mould formed using a Microtissues® anti-mould. This method produces multiple spheroids per well, which are transferred manually from the agarose wells into 96-well plates. We have tested this method successfully with HEK293T and DU145 cell lines.
Cell lines for imaging in 3-D cultures
To date we have generated the following cell lines:
- HEK293T-T2AMPKAR – the donor fluorophore of the established AMPKAR FRET biosensor has been replaced by mTurquoise2. The monoexponential decay and higher photostability of mTurquose2 is preferred compared to the eCFP donor in the original AMPKAR biosensor.
- HEK293T-AMPK KO-T2AMPKAR – this construct provides a negative control for the T2AMPKAR sensor following CRISPR knockout of both endogenous AMPKβ subunits that results in the cell failing to form the full trimeric (α, β, γ) complex of AMPK with the remaining subunits being rapidly degraded.
- HEK293T-T2AMPKAR-T391A – this single point mutation of the T2AMPKAR cell line where the threonine residue at 391 is mutated to Alanine, resulting in an inactive form of AMPKAR that cannot be phosphorylated by AMPK, provides a further negative control.
- DU145-T2AMPKAR – these cells also express the new AMPK biosensor
- DU145-mTurquoise2 these cells provide a donor only sample to enable us to study any environmental perturbations to the fluorescence of the donor fluorophore
- HEK293T-FLII12Pglu-700μδ6 – these HEK293T cells express a FRET biosensor reporting intracellular glucose concentration
- DU145-FLII12Pglu-700μδ6 – these DU145 cells express a FRET biosensor reporting intracellular glucose concentration
We are currently in the process of generating stable clones of DU145 cells expressing standard FRET constructs consisting of mTurquoise directly linked to Venus via a range of lengths of short peptide linkers to provide a range of different FRET efficiencies. Expression of these constructs in 2-D and 3-D alongside a biological assay will provide us with a standard sample against which to assess the imaging platform performance.
Automated multiwell plate platform for FLIM FRET of protein interactions in 3-D cell culture
With help from colleagues at EMBL, we have successfully implemented a water immersion objective on a home-built multiwell plate reader with continuous supply of water immersion liquid to the objective to replaces liquid lost during stage translation and due to evaporation. have been able to image fluorescent protein labelled cells up to 40 microns into spheroids with single photon excitation of the mTq2FP donor fluorophore. We have implemented and tested a flexible add-on to our µManager acquisition software that allows ImageJ analysis scripts to be used to provide a pre-find capability. This allows the fields of view for FLIM imaging to be automatically evaluated and selected during the pre-find scan of a 96-well plate.
We have shown that we can acquire FLIM z-stacks of spheroids of HEK cells expressing the T2AMPKAR biosensor – requiring 145 s to image a spheroid with 8 z-planes, of which 23 s is required for sample motion and autofocussing. We therefore expect to be able to read a 96-well plate (with one z-section per well) in 65 minutes, or ~4.5 hrs for a 96 well plate with 8 z-planes per spheroid. Of course the data acquisition time will depend on the brightness of the fluorophores and the number of time-gates acquired. We believe there is still significant scope to optimising the sampling strategy, depending on the precise readout required.
Applications of cell signalling processes underlying disease
We are working to explore the potential of our automated multiwell plate FLIM platform to assay the following cell signalling processes, which are important for a range of diseases.
(a) bromodomain-histone binding and compare inhibitor compound effects between 2-D and 3-D assays
The following constructs have been developed and tested by GSK:
- Venus-BRD4-LH4-mTurquoise - plasmid used for transient expression
- Venus-Brd4-LH4K4R-mTurquoise - control plasmid with mutated bromodomain binding site
We are currently working to generate the required stable cell lines expressing these constructs at Imperial.
(b) multimerisation of the Smo protein in the Hedgehog signalling pathway
We have demonstrated the formation of primary cilia in NIH3T3 cells grown in conventional 2-D culture. Stable expression of SMO tagged with GFP is currently under way to help visualize SMO translocation from the cytosol to the primary cilia upon activation of the Hedgehog signalling pathway.
(c) changes in cellular metabolism using genetically expressed FRET biosensors
We have compared the response of 2-D and 3-D cell cultures using the AMPKAR and FLII12Pglu-700μδ6 glucose biosensors. Results to date comparing HEK293T AMPK activity in 2-D and 3-D cultures exposed to a direct AMPK activator show a qualitatively similar dose response although we have observed larger cell to cell variations in AMPKAR biosensor fluorescence lifetime in 2-D culture compared to 3-D culture. We are currently carrying out further experiments to explore this.
Imaging of glucose concentrations changes in HEK293T spheroids has presented an apparent gradient towards the centre of the spheroids which is currently being investigated further.
We have also undertaken experiments on the effects of co-cultures of DU145 and HS5 cell lines. Initial studies include the influence of co-culture on cell migration and AMPK activity.