The activation of the multifunctional Ser/Thr kinase AURKA is hallmarked by its autophosphorylation on Thr288. Fluorescent probes relying on FRET can discriminate between its inactive and activated states. Here, we illustrate some strategies for probe engineering, together with a rapid FRET protocol to follow the kinase activation throughout mitosis.
Epithelial cancers are often hallmarked by the overexpression of the Ser/Thr kinase Aurora A/AURKA. AURKA is a multifunctional protein that activates upon its autophosphorylation on Thr288. AURKA abundance peaks in mitosis, where it controls the stability and the fidelity of the mitotic spindle, and the overall efficiency of mitosis. Although well characterized at the structural level, a consistent monitoring of the activation of AURKA throughout the cell cycle is lacking. A possible solution consists in using genetically-encoded Förster’s Resonance Energy Transfer (FRET) biosensors to gain insight into the autophosphorylation of AURKA with sufficient spatiotemporal resolution. Here, we describe a protocol to engineer FRET biosensors detecting Thr288 autophosphorylation, and how to follow this modification during mitosis. First, we provide an overview of possible donor/acceptor FRET pairs, and we show possible cloning and insertion methods of AURKA FRET biosensors in mammalian cells. Then, we provide a step-by-step analysis for rapid FRET measurements by fluorescence lifetime imaging microscopy (FLIM) on a custom-built setup. However, this protocol is also applicable to alternative commercial solutions available. We conclude by considering the most appropriate FRET controls for an AURKA-based biosensor, and by highlighting potential future improvements to further increase the sensitivity of this tool.
Aurora kinase A/AURKA is a multifunctional serine/threonine kinase, active throughout the cell cycle and at different subcellular compartments1. Understanding its broad spatiotemporal activation is particularly important in cancer, as AURKA is often overexpressed in epithelial and hematological malignancies, with patients showing poor responsiveness to the currently available therapies2.
Structural studies revealed that AURKA undergoes two steps to convert from an inactive to an active kinase. First, the autophosphorylation of Thr288 changes the conformation of the kinetic pocket of the kinase and activates it3,4,5,6. This step increases the catalytic activity of AURKA in human cells and in Xenopus laevis3,6,7, priming the kinase for full activity. Once activated, the interaction of AURKA with the targeting protein for Xklp2 (TPX2) induces a second conformational change5. This further modification allows AURKA to reach full enzymatic activity towards its substrates in the cell5,8,9,10.
For nearly two decades, insights on the activation and activity of AURKA were obtained mainly through a combination of biochemical approaches. These include the detection of phosphorylated Thr288 in cells or in vivo as a hallmark of AURKA activation, crystallographic analyses, and in vitro or in cellulo kinase assays to probe the activity of AURKA1. However, the spatiotemporal resolution of these approaches is poor or absent, and novel solutions were needed to broaden the knowledge of the dynamics of these two events.
The development of fluorescent probes in the last few years facilitated the monitoring of AURKA in live cells, allowing the following of its activation with greater spatiotemporal resolution. The most specific sensors for AURKA developed so far rely on the FRET principle (Förster’s Resonance Energy Transfer)11 to discriminate between inactive and active AURKA. The first sensor developed was a substrate-based biosensor of AURKA kinase activity. Substrate-based biosensors are constituted by a short aminoacidic sequence targeted by a given kinase for phosphorylation, and inserted within a donor/acceptor FRET pair and a binding domain recognizing the phosphorylated residue, which helps the folding of the biosensor for an efficient FRET process12. In the case of AURKA, a 14-aminoacid fragment of KIF2C targeted by phosphorylation was inserted between a CFP-YFP donor/acceptor pair13. However, this sensor has some major drawbacks. First, the KIF2C sequence used in this probe can be targeted both by AURKA and the closely-related kinase AURKB, thereby decreasing the specificity of this biosensor. Second, the sensor relies on the endogenous kinase for phosphorylation. Therefore, FRET efficiency can be undetectable or not significant if the quantities of the kinase are limiting (e.g., in subcellular compartments or cell cycle phases). To overcome these limitations, a new class of AURKA sensor was created known as “conformational sensors”. In these probes, the full-length sequence of AURKA was inserted within a donor fluorophore at the N-terminus, and an acceptor fluorophore at the C-terminus. Inactive AURKA presents an “open” conformation, which brings the N- and C-termini of the kinase away from each other. With such distance between the two termini (> 10 nm), the donor/acceptor pair are in a non-permissive configuration for FRET. On the contrary, autophosphorylated AURKA adopts a “closed” conformation, with the two protein termini and the two fluorophores in proximity. This was shown to allow FRET between the donor and the acceptor, which can be measured using the variations in the donor lifetime14,15. Such probes present several advantages. First, they are genetically-encoded, and they can be used to replace the endogenous kinase in the cell. Second, they rescue the phenotypes induced by the knockdown of AURKA, indicating that they are functional in the cell. Third, they allow to follow the activation of the kinase at different subcellular compartments and throughout the cell cycle. The probes detected the activation of AURKA at locations where the kinase is known to be activated (i.e. centrosomes and the mitotic spindle), and also participated in discovering the activation of AURKA at the mitochondria16. Last, these sensors allowed high-content screenings based on FRET/FLIM, where the conformational changes of AURKA were used to identify novel pharmacological inhibitors17.
In the present work, we describe a procedure to visualize AURKA activation in cultured cells. First, we will make an insight on potential fluorophore pairs for FRET. The choice of the most suitable donor/acceptor pair will be made according to the available microscope setup, or a particular downstream application as multiplex FRET18,19. Then, we propose a pipeline to explore the behavior of the biosensor(s) chosen in a rapid FRET/FLIM microscope setup. This pipeline will extend from cell culture and synchronization procedures to FLIM acquisition and data analysis. Last, we will discuss the potential advantages of this protocol, as an analogous strategy for biosensor design could be applied to other kinases, and it may also be used with other FRET-based imaging systems.
NOTE: U2OS cells used in this protocol were purchased from American Type Culture Collection (ATCC, HTB-96), and they were tested free from mycoplasma. Step 2.1 to 2.7 should be performed under a laminar flow hood to keep cells and reagents sterile.
1. Choosing the donor/acceptor FRET pair
2. Cell culture, transfection and synchronization
3. FRET/FLIM acquisitions
NOTE: FRET/FLIM acquisitions in this protocol were performed on a custom-built setup, described in35 and equipped with a control solution as in15,17 (Figure 2). The setup is now commercialized by Inscoper and it is made of a spinning-disk microscope with a white laser for pulsed excitation, and a high-rate time-gated intensifier in front of the camera. Temporal gates of 2 ns in a time window of 10 ns are sequentially used to obtain a stack of five time-gated images. These images are then used to calculate the pixel-by-pixel mean fluorescence lifetime according to the following equation: τ = ΣΔti • Ιi/ΣΙi, where Δti corresponds to the delay time of the of the ith gate while I indicates the pixel-by-pixel time-gated intensity image35,36. This method ensures rapid FLIM measurements: no fitting or binning steps are required, and lifetime can be calculated in an online mode, with minimal photon budget. The system also presents a user-friendly software interface. However, the same experiment can be performed under any other commercial microscope setup equipped for FLIM measurements.
4. Calculation of ΔLifetime and comparison of FLIM values among donor/acceptor pairs
We followed the procedure described above to record the autophosphorylation of AURKA on Thr288 using two biosensors with different spectral properties. We compared the initial GFP-AURKa-mCherry probe14 with two biosensors with different spectral properties. These two probes rely on the fluorescent donor mTurquoise2 and on a non-fluorescent acceptor (ShadowG) in one case, or a yellow acceptor (superYFP) in a second case. We then inserted the full-length sequence of AURKA within each donor/acceptor pair. To have a negative control for AURKA activation, two strategies can be pursued. First, the use a small ATP-analog (MLN8237) interferes with the binding of ATP in the kinetic pocket of the kinase and prevents its activation38. Second, the mutation of Lys162 into Met (K162M), creates a kinase-dead version of each biosensor incapable of activating14,15,39. This mutation induces the disruption of a salt bridge normally established between Lys162 and Glu181, which results in a stable opening of the kinetic pocket of the kinase and triggers its overall inactivation40. As a negative control for FRET, we used an acceptor-devoid construct (GFP-AURKA or AURKA-mTurquoise2).
After synchronizing cells in G2/M and releasing them into mitosis, we measured the lifetime of all the transfected constructs at the mitotic spindle (Figure 4). Of note, this structure was considered as a whole, and no ROIs within the spindle were analyzed. We then calculated ΔLifetime for all conditions. As expected, the lifetime of GFP-AURKA or AURKA-mTurquoise2 (the “donor-only” conditions) was close to 0, indicating that the values measured for these constructs fluctuated around the mean value (Figure 4A,4B). Conversely, the ΔLifetime values for GFP-AURKA-mCherry were statistically different from the donor-only condition, with ΔLifetime increasing of ~130 ps (Figure 4A). Similar observations were made for shadowG-AURKA-mTurquoise2 and for superYFP-AURKA-mTurquoise2, with ΔLifetime increasing of ~150 and ~220 ps from the donor-only condition, respectively (Figure 4B,4C). These data can be easily visualized in single cells with a pseudocolor Lookup Table (LUT). In this case, values of ΔLifetime around 0 are pseudocolored yellow, while more significant differences are pseudocolored red/purple. Indeed, the pixel-by-pixel LUT was closer to yellow in cells expressing the donor-only constructs, while it was more in the red/purple spectrum in cells expressing either biosensor (Figure 4A,4B). This was also observed when the GFP-AURKA-mCherry biosensor was treated with the pharmacological inhibitor MLN8237.
We then analyzed the ΔLifetime of kinase-dead biosensors. These constructs showed intermediate ΔLifetime values: ΔLifetime was significantly higher when compared to the donor-only condition (Figure 4B,4C), but it was also significantly lower than their normal counterparts (Figure 4B,4D). The comparisons with cells treated with MLN8237 or expressing kinase-dead biosensors are necessary to estimate whether ΔLifetime variations for each donor/acceptor pair are solely linked to the activation of AURKA. In the case of GFP-AURKA-mCherry, ΔLifetime variations are abolished when an AURKA-specific inhibitor is used. Conversely, ΔLifetime variations are mostly, but not exclusively linked to AURKA activation in the case of shadowG-AURKA-mTurquoise2 and of superYFP-AURKA-mTurquoise2.
Figure 1: GFP (donor) and mCherry (acceptor) excitation and emission spectra.
Spectra were obtained and adapted from the FP base website (https://www.fpbase.org/), and adjusted to a 480 nm-laser excitation. Please click here to view a larger version of this figure.
Figure 2: Image of the experimental workspace.
(1) The control solution; (2) white laser source; (3) CCD camera; (4) microscope setup; (5) external light source/lamp for ocular screening of the sample. Please click here to view a larger version of this figure.
Figure 3: Representative images of the software for FLIM acquisition.
(A, B) (1) excitation and emission parameters for the donor (CFP in A, or GFP in B); (2) exposure time; (3) selection of the objective. Please click here to view a larger version of this figure.
Figure 4: Representative images of AURKA FRET biosensors and their negative controls.
(A) (Micrographs) Fluorescence (green channel) and corresponding pixel-by-pixel ΔLifetime (donor only – biosensor) of U2OS cells expressing GFP-AURKA or GFP-AURKA-mCherry, synchronized at G2/M, released until the bipolar spindle is visible and then treated with DMSO or with MLN8237. ΔLifetime is illustrated with a pseudocolor scale (“Fire” lookup table). (Graph) Corresponding quantification and two-way ANOVA analysis for the indicated conditions. (B) (Micrographs) Fluorescence (cyan channel) and corresponding pixel-by-pixel ΔLifetime (donor only – biosensor) of U2OS cells expressing shadowG-AURKA-mTurquoise2 (upper panel) or superYFP-AURKA-mTurquoise2 (lower panel), synchronized at G2/M and released until the bipolar spindle is visible. ΔLifetime is illustrated with a pseudocolor scale (“Fire” lookup table). (Graph) Corresponding quantification and one-way ANOVA analysis of conditions represented in the above micrographs. (C) (Micrographs) Images of AURKA-mTurquoise2 (upper panel), shadowG-AURKA K162M-mTurquoise2 (middle panel) and superYFP-AURKA K162M-mTurquoise2 acquired and represented as in the micrographs. (Graph) Two-way ANOVA analysis for the indicated transfection conditions. The bar in boxplots represents the median; whiskers extend from the min to the max. n = 10 cells per condition of one representative experiment (of three). Individual values are represented as dots. Scale bar: 10 µm. *P < 0.05, **P < 0.01, ***P < 0.001 against each indicated condition in (A) the “AURKA-mTurquoise2” condition in (B), and against each indicated condition in (C). NS: not significant. Please click here to view a larger version of this figure.
Genetically-encoded FRET biosensors are reliable tools to measure the activation of single proteins or of entire signaling pathways41. Particularly, the AURKA FRET biosensor constitutes a preferential way to explore the activation of the kinase in time and space. However, some elements deserve special attention when designing or optimizing a FRET biosensor, not only in general terms but more specifically for AURKA.
First, the nature and the relative position of the donor/acceptor FRET pair can be adapted for specific functions of this kinase. AURKA is greatly enriched at the mitotic spindle during mitosis, but it is present throughout the cell cycle and at different subcellular locations (e.g., centrosomes, the nucleus, and mitochondria)1,2. If the biosensor is to be used in specific compartments like mitochondria, which can reach acidic pH, the choice of a pH-insensitive donor-acceptor FRET pair as mTurquoise2/shadowG should be made. Moreover, placing the FRET donor at the C-terminus could allow a better visualization of the biosensor at this subcellular compartment, and potentially even optimize FRET detection given that AURKA N-terminus was shown to partially cleave off at mitochondria16,42.
Second, a yet unexplored way of optimizing the AURKA FRET biosensor would require a more careful design of linkers between full-length AURKA and the donor/acceptor pair. Not only the distance between the fluorescent pair, but also the properties of the linker itself were shown to be key factors to improve FRET efficiency43,44,45,46. In this light, increasing the rigidity or the flexibility of the linker could either be detrimental to FRET efficiency, or further improve it.
Third, it is known that the overexpression of AURKA induces mitotic spindle abnormalities in a significant proportion of cells2. It would be interesting to compare ΔLifetime obtained by expressing the same FRET construct under a strong promoter like cytomegalovirus (CMV) – one of the most common promoters found in mammalian expression vectors – or under the AURKA minimal promoter sequence (CTTCCGG)14,47. This promoter was previously shown to rescue monopolar or multipolar spindles arising after the knock-down of the kinase, and its use did not induce cell cycle perturbations per se14,47. Although FLIM is insensitive to protein expression levels and relative concentrations in the cell11, benefiting from a thorough comparison of the two promoters on the same biosensor setup would broaden the understanding of the pool of activated AURKA at any given location. In addition, it would provide novel insights on how AURKA activation may change upon overexpression, which is relevant for epithelial and hematologic cancer paradigms.
Last, the downstream FRET application should also be taken into account. A future perspective in the field of AURKA would be to cumulate the kinase conformational biosensor with a substrate-based biosensor. Analyzing the FRET behavior of two biosensors simultaneously – a process known as multiplex FRET – requires a dark acceptor on the first biosensor to avoid spectral bleed through in the second donor channel. In the context of AURKA, this would open up the exciting new perspective of detecting the activation of the kinase with the first biosensor, and its enzymatic activity towards a given substrate with the second one. Recent developments in multiplexing now allow to cumulate up to three biosensors at a time48. Applying a similar method in the context of AURKA could represent a very promising strategy not only to test the activation-activity interplay of the kinase, but also to explore AURKA signaling cascades with unprecedented spatiotemporal resolution.
In conclusion, FRET/FLIM is a convenient way to deepen the knowledge on protein activity. On one hand, it allows to visualize the localization of a given protein in live cells, thanks to at least one fluorescent moiety. On the other hand, it can unravel protein conformational changes, which could be informative on protein activation and/or activity. Therefore, FRET/FLIM and conformational FRET biosensors have the potential of becoming widespread methods to follow signaling pathways in live cells, and with exquisite spatiotemporal resolution.
We thank the engineers of the Microscopy-Rennes Imaging Center (MRic, BIOSIT, Rennes, France) for advice and help, and particularly X. Pinson for critical reading of the manuscript. MRic is member of the national infrastructure France-BioImaging supported by the French National Research Agency (ANR-10-INBS-04). This work was supported by the Centre National de la Recherche Scientifique (CNRS), the Ligue Contre le Cancer Comités d’Ille et Vilaine, des Côtes d’Armor et du Finistère, and the Association pour la Recherche Contre le Cancer (ARC) to G.B.
Name | Company | Catalog Number | Comments |
Alisertib (MLN8237) | SelleckChem | S1133 | Use at a 250 nM final dilution |
Dulbecco's Modified Eagle Medium (DMEM) | ThermoFischer Scientific Brand: Gibco | 41966052 | High glucose + L-glutamine + Sodium pyruvate |
Fetal Bovine Serum (FBS) | ThermoFischer Scientific Brand: Gibco | 10270106 | |
L15 | ThermoFischer Scientific Brand: Gibco | 21083027 | Leibovitz's L15 medium + L-glutamine, no Phenol red |
LabTek | Nunc | 2515380 | |
Nocodazole | Merck Brand: Sigma-Aldrich | M1404 | Use at a 100 ng/mL final dilution |
Penicillin/Streptomycin | ThermoFischer Scientific Brand: Gibco | 15140122 | Penicillin-Streptomycin 10,000 U/mL (100x) |
Phosphate Buffer Saline (PBS) | ThermoFischer Scientific Brand: Gibco | 14190169 | DPBS, no calcium, no magnesium |
Trypsin/EDTA | ThermoFischer Scientific Brand: Gibco | 25300096 | Trypsin-EDTA 0.05%, Phenol Red (1x) |
This article has been published
Video Coming Soon
ABOUT JoVE
Copyright © 2024 MyJoVE Corporation. All rights reserved