FRET imaging of Rho GTPase activity with red fluorescent protein-based FRET pairs

Abstract

With the development of fluorescent proteins (FPs) and advanced optical microscopy techniques, Förster or fluorescence resonance energy transfer (FRET) has become a powerful tool for real-time non-invasive visualization of a variety of biological processes, including kinase activities, with high spatiotemporal resolution in living cells and organisms. FRET can be detected in appropriately configured microscopes as changes in fluorescence intensity, lifetime, and anisotropy. Here we describe the preparation of samples expressing FP-based FRET sensors for RhoA kinase, intensity- and lifetime-based FRET imaging, and post-imaging data analysis.

1. Introduction

In Förster or fluorescence resonance energy transfer (FRET), a donor fluorophore in an excited state non-radiatively transfers its excitation state energy to a nearby acceptor fluorophore in a ground state via dipole-dipole coupling, leading to fluorescence emission from the acceptor. The efficiency of FRET (E_FRET), defined as the percent of energy transfer from the donor to acceptor fluorophore, is dependent on the inverse sixth power of the distance between donor and acceptor and occurs only over a distance shorter than 10 nm, making it a very sensitive tool for reporting biochemical activities that produce changes in molecular proximity¹. Fluorescent protein (FP)-based FRET biosensors, mainly composed of a donor and red-shifted acceptor FPs and a sensing module, have been developed to monitor a variety of biological processes such as protein-protein interactions, conformational changes of proteins, ion concentrations, and enzyme activities with a high spatiotemporal resolution in intact living cells and organisms². Unlike dye- or nanomaterial-based FRET reporters, FP-based sensors are genetically encoded and can be easily integrated into various cell types and subcellular structures, allowing for long-term tracking of molecular activities in specific cellular populations within an animal or in specific subcellular compartments³.

A key component of designing and optimizing FRET-based biosensors is the selection of donor and acceptor fluorophores. Generally, donors and acceptors should be chosen to maximize the spectral overlap between donor emission and acceptor excitation, minimize the spectral overlap between the respective emission and excitation of the donor and acceptor, and maximize the quantum yield of the donor and absorption coefficient of the acceptor³. Historically, cyan fluorescent protein (CFP) donors and yellow fluorescent protein (YFP) acceptors have been commonly used due to the availability of high quantum yield fluorescent proteins with emission spectra in the 450-550 nm range, derived primarily from Aequorea victoria green fluorescent protein (GFP). However, CFP-YFP FRET pairs exhibit certain disadvantages compared to red-shifted pairs, including fast photobleaching of YFPs, photoconversion of YFPs, phototoxicity from near-ultraviolet excitation, severe spectral cross-talk, and relatively small dynamic range⁴.

By contrast, green fluorescent protein (GFP) donors and red fluorescent protein (RFP) acceptors offer less phototoxicity and greater spectral emission separation. Moreover, GFPs are more photostable than CFPs under both one-photon and two-photon excitation⁵. Additionally, when imaging in cells, the red-shifted excitation and emission in green-red pairs offer higher signal-to-noise due to reduced autofluorescence from flavoproteins⁶ In recent years, development of improved GFPs (mClover3⁷ and mNeonGreen⁸) and RFPs (mRuby3⁷, mRuby4⁹, mScarlet-I¹⁰) have made GFP-RFP FRET increasingly attractive as FRET pairs. Indeed, recently engineered GFP-RFP FRET pairs outperform bright CFP-YFP FRET pairs with respect to dynamic range, the range of E_FRET in which a given FRET reporter operates⁴. In addition, cyan-excitable RFP donors and far-red fluorescent protein acceptors exhibit minimal spectral cross-talk and less phototoxicity. More importantly, they are spectrally compatible with GFP-based sensors, allowing simultaneous imaging of two biological events under single excitation in living cells and organisms.

FRET induces changes in the fluorescence intensity and polarization of donor and acceptor and the fluorescence lifetime of donor, which can be detected in various ways^{3, 11}. Of these, two methods are most commonly used: sensitized emission ratiometric FRET (Ser-FRET) and fluorescence lifetime FRET (FLIM-FRET). In Ser-FRET, the E_FRET is calculated as the ratio of the uncorrected FRET signal from the acceptor (increase) to the signal from directly excited donor (decrease), thus boosting small changes in FRET. In FLIM-FRET, the E_FRET can be achieved by comparing fluorescence lifetimes of the donor in the presence and absence of the acceptor. Compared to Ser-FRET, FLIM-FRET has several advantages and disadvantages. First, FLIM-FRET is more suited for in vivo imaging with intermolecular sensors because the lifetime is independent of donor fluorescence (significantly attenuated in vivo) and the acceptor–to-donor ratio (unknown under co-transfection)^{12, 13}. Second, FLIM-FRET enables true FRET efficiency measurements, which allows for the identification of fractions of molecules involved in FRET¹³. Third, FLIM-FRET can use a less photostable FP as donor and a dark FP (high extinction coefficient and negligible quantum yield) as acceptor, thus potentially increasing multiplexing of FRET pairs¹⁴. Fourth, FLIM-FRET has lower temporal resolution (sec to min) because enough photons need to be collected to achieve lifetime decay kinetics. Fifth, FLIM requires expensive and complicated instrumentation, preventing its wide use in most laboratories. However, any lab with a standard epifluorescence microscope equipped with appropriate filters for a particular FRET pair can perform live-cell Ser-FRET imaging. In summary, Ser-FRET is optimal for intramolecular FRET sensors in living cells while FLIM-FRET is superior for intermolecular FRET sensors in vivo.

Here we outline the use of GFP-RFP FRET to measure RhoA activity in cultured dissociated neurons with Ser-FRET and in organotypic hippocampal slices with FLIM-FRET techniques. For Ser-FRET, we use the intramolecular Raichu-RhoA-CR reporter, in which a GFP-RFP FRET pair, Clover-mRuby2, replaces the original CFP-YFP FRET pair in Raichu-RhoA FRET reporter, which has been used to analyze RhoA activation in dorsal root ganglion neurons¹⁵, hippocampal neurons^{4, 16}, in addition to imaging applications for the cell cycle¹⁷. Raichu-RhoA-CR has been used to study RhoA-dependent axonal growth cone dynamics in hippocampal neurons^{4, 18}. For FLIM-FRET, we use mCyRFP1-RhoA and mMaroon1-Rhotekin-mMaroon1, an intermolecular RhoA FRET sensor that uses cyan-excitable red (mCyRFP1) and far-red (mMaroon1) fluorescent proteins as donor and acceptor fluorophores¹⁴. Both Ser-FRET and FLIM-FRET approaches outlined in this protocol enable live imaging of RhoA activity with high spatiotemporal precision.

2. Materials

2.1. Plasmid

1. pCAGGS-Raichu-RhoA-CR (Addgene #40258)

2. mCyRFP1-RhoA and mMaroon1-Rhotekin-mMaroon1 (Addgene #84358 and 84359)

2.2. Reagents for Ser-FRET imaging

1. Papain (Life Technologies)

2. DNAse I (Life Technologies)

3. Rat neuron nucleofector kit (Lonza)

4. 8-well chamber slides (Nunc)

5. Poly-D-lysine (Sigma) in borate buffer

6. Laminin (BD Biosciences)

7. Plating medium: Neurobasal medium (Life Technologies) supplemented with Glutamax (Thermo Fisher), penicillin/streptomycin, and 5% fetal bovine serum (Life Technologies)

8. Phenol red-free Neurobasal medium supplemented with Glutamax (Thermo Fisher) and B27

9. Preclustered Ephrin-A solution: 10 μg mL⁻¹ of ephrin-A4-Fc and ephrin-A5-Fc (R&D Systems) are separately preclustered with 50 μg mL⁻¹ of goat anti-human IgG (H+L) (Jackson ImmunoResearch) in Hank’s Balanced Salt Solution (Thermo Fisher) for 1 h and then mixed

10. Imaging solution: Hank’s Balanced Salt Solution (Thermo Fisher)

2.3. Reagents for FLIM-FRET imaging

1. Millicell membranes (Millipore)

2. Plate medium: minimal essential medium (Thermo Fisher) supplemented with 20% horse serum (Sigma), 1 mM L-glutamine (Sigma), 1 mM CaCl₂, 2 mM MgSO₄, 12.9 mM D-glucose, 5.2 mM NaHCO₃, 30 mM HEPES, 0.075% ascorbic acid and 1 μg/mL insulin (Sigma)

3. Imaging solution: Mg²⁺-free artificial cerebral spinal fluid (ACSF; 127 mM NaCl, 2.5 mM KCI, 4 mM CaCl₂, 25 mM NaHCO₃, 1.25 mM NaH2PO₄ and 25 mM glucose) containing 1 μM tetrodotoxin (TTX, Sigma) and 4 mM 4-Methoxy-7-nitroindolinyl-caged-L-glutamate (MNI-caged L-glutamate, Tocris) aerated with 95% O₂ and 5% CO₂

4. Biolistic gene transfer system (Bio-Rad) with gold beads

2.4. Fluorescence microscope

1. Inverted epifluorescence microscope (Zeiss Axiovert 200M)

   1. 40 × water-immersion objective (Zeiss 40×1.2-numerical aperture (NA) C-Apochromat lens)
   2. Optical filters (ex, excitation, em, emission, dm, dichroic mirror): ex HQ470/30 nm (Chroma) dm 565 nm, and em 505AELP nm (Omega) for GFP, and ex HQ470/30 nm (Chroma) dm 565 nm, and em BA575IF nm (Olympus) for FRET
   3. Cooled charge-coupled device camera (Hamamatsu Orca-ER)
   4. Environmental chamber with temperature and CO₂ control (Live Cell Instrument, Korea)
   5. Microscopy software (Micromanager 1.4.18 together with ImageJ version 1.49 or later)
2. Custom-built two-photon FLIM setup

   1. 60 × water-immersion objective (Olympus 1.0 NA Apochromat objective)
   2. Optical filters (dm, em): dm 565 nm (Chroma) and em 520/60 nm (Chroma) for GFP, and dm 565 nm (Chroma) and em 620/20 nm (Chroma). for CyRFP
   3. Two photomultiplier tubes (PMTs) with low transfer time spread (H7422-40p, Hamamatsu)
   4. Chameleon Ti:sapphire laser (Coherent) tuned at 720 or 920 nm
   5. Temperature controller (TC-324C, Warner Instrument)

3. Methods

3.1. Ser-FRET imaging

3.1.1. Hippocampal neuron preparation

1. Hippocampal neurons are dissected from embryonic day 18 (E18) rats (Charles River Labs), dissociated with papain and DNaseI.

2. Transfect neurons by electroporation with a rat neuron Nucleofector kit and plated at a density of 30,000 per cm² in 8-well chamber slides. Prior to plating, coat slides by incubating with 0.25 mg mL⁻¹ poly-D-lysine in borate buffer for 12–24 h, washing three times with water for 15 min each, incubating with 18 μg mL⁻¹ laminin in Neurobasal medium for 12–16 h, and washing three times with water for 15 min each.

3. Neurons are plated in Neurobasal medium with Glutamax, penicillin and streptomycin, and 5% FBS.

4. Plating medium is replaced 12 h later with phenol red–free Neurobasal medium with Glutamax and B27.

5. Neurons are imaged 1-2 d after transfection in imaging solution (see 2.2) on an Zeiss Axiovert 200M microscope (see 2.4) at 37°C.

3.1.2. Image acquisition

1. Cells with appropriate expression level are chosen for FRET imaging (see Note 1)

2. Baseline measurements are taken sequentially in FRET donor and acceptor channels every 3 min for 21 min (see Note 2).

3. Add preclustered ephrin-A and IgG dropwise to chambers at a final concentration of 5 μg mL-1 of ephrin-A and 25 μg/mL of IgG (see Note 3). Continue FRET imaging every 3 min for another 21 min (see Note 4)

3.1.3. Image analysis

To quantitate the FRET response from the reporter, we measure the ratios (R) of background-subtracted intracellular green (I_g – I_gb) and red fluorescence (I_r – I_rb) under cyan light excitation in cells of interest and monitor its change over time. Analysis of the FRET response is performed on Fiji/ImageJ, with additional analysis performed on the data processing and visualization software of choice. The FRET response can be represented as the following:

$$\mathit{\left(}{R}_t-R_{\mathit{0}}\mathit{\right)}/R_{\mathit{0}}=\mathit{1}-\mathit{\left[}\mathit{\left(}\mathit{\left(}{I}_{tr}-I_{trb}\mathit{\right)}/\mathit{\left(}{I}_{tg}-I_{tgb}\mathit{\right)}\mathit{\right)}/\mathit{\left(}\mathit{\left(}{I}_{\mathit{0}r}-I_{\mathit{0}rb}\mathit{\right)}/\mathit{\left(}{I}_{\mathit{0}g}-I_{\mathit{0}gb}\mathit{\right)}\mathit{\right)}\mathit{\right]}$$ (1)

where the subscript characters 0 and t represent time 0 and time t after addition of ephrin-A, respectively.

1. Open FRET and GFP image stacks separately on ImageJ.

2. Background subtract both stacks, through either of the methods indicated below:

   1. Built-in rolling ball background subtraction (Process>Background Subtraction…>Rolling Ball Radius = 50.0 pixels
   2. Manual background subtraction:

      1. Draw a region of interest (ROI) in the background area of the field of view, at least 15x15 pixels in size.
      2. Measure the average value of the ROI (Analyze>Measure)
      3. Subtract that value from the entire image (Process>Math>Subtract)
3. To measure the same regions in each cell, we want to use the same ROIs on cells between both stacks. However, the GFP and FRET stacks may not be perfectly aligned depending on the optical alignment between channels in the imaging system. We will first need to align the stacks using one of the methods indicated below (see Note 5):

   1. Manual registration:

      1. Find a reference point in the GFP stack.
      2. Note the x-y coordinates of that reference point.
      3. In the FRET stack, find the same reference point and note the x-y coordinates.
      4. Subtract those coordinates to find the x-y translation required to align each stack.
      5. Align images by translating the FRET stack by the identified x-y distance to match the GFP stack using Image>Transform>Translate…

4. Regions of interest are manually segmented. Using the ROI Manager (Analyze>Tools>ROI Manager…) and the “Polygon selections” option, create ROIs around the desired features to analyze (e.g. neuronal growth cones) on each of the cells in the field of view. Click “Add” on the ROI Manager to save the selected ROI (shortcut: t). This enables retrieval of the ROI in either channel.

5. For each ROI, measure the intensities in each cell across all slices in the stack, for both FRET and GFP stacks. Two simple methods in ImageJ include: (Image>Stacks>Measure Stack…) or (Image>Stacks>Plot Z-axis Profile…), (see Note 6)

6. In your desired analysis software (e.g. Microsoft Excel, MATLAB, R), open the measured values in each channel.

7. Normalize measurements to baseline values in each stack.

8. Divide normalized GFP values by normalized FRET values at each timepoint. This gives the ratiometric FRET response at each timepoint (see Note 7). Imaging results are presented in Figure 1.

Figure 1. Reporting of fast local RhoA activation in neurons with Raichu-RhoA-CR.

(a) Design of the Raichu-RhoA-CR reporter, based on Raichu-RhoA. PKN, protein kinase N. (b) Ephrin-A stimulation locally activated RhoA in a hippocampal growth cone (asterisks) from the first time point after stimulation. Scale bar, 10 μm. (c) Raichu-RhoA-CR acceptor/donor emission ratio change (ΔR/R_AD) graphed as mean ± s.e.m. Peak ratio change (asterisk) was significantly different from baseline by two-tailed t-test (P = 0.019, n = 5 cells). (c) With Raichu-RhoA, peak ratio changes were not statistically different from baseline at any time (P > 0.1, n = 8 cells). (d) With Raichu-RhoA, peak ratio change was not significantly different from baseline (P = 0.468, n = 8 cells). Adapted from Ref. 4.

3.2. FLIM-FRET imaging

3.2.1. Hippocampal slice preparation

1. Hippocampal slices are prepared from postnatal 4- to 6-day-old C57BL/6 mice. 350 μm-thick hippocampal slices were dissected using a tissue chopper and then plated on Millicell membranes in minimal essential medium supplemented with 20% horse serum, 1 mM L-glutamine, 1 mM CaCl2, 2 mM MgSO4, 12.9 mM D-glucose, 5.2 mM NaHCO3, 30 mM HEPES, 0.075% ascorbic acid and 1 μg/mL insulin, with fresh medium replaced every other day.

2. After 7–10 days in culture, neurons are transfected by ballistic gene gun (Bio-Rad) using gold beads (8-12 mg). Bullets are coated with plasmids containing mCyRFP1–RhoA and mMaroon1–Rhotekin–mMaroon1 (20 μg each).

3. Slices are imaged 2–5 days after transfection in imaging solution (see 2.3) on the two-photon FLIM imaging system (see 2.4) at 24–26 °C.

3.2.2. Image acquisition

1. Hippocampal slices are chosen based on presence of CA1 cells positive for RhoA-sensor expression at least two days after transfection. Sensor expression is typically sparse and averages at around 1 cell per slice.

2. Baseline imaging is performed for a single secondary or tertiary dendritic branch of CA1 pyramidal cells every 1 min for 10 min (see Note 8).

3. To activate a single dendritic spine in the dendritic branch, MNI-caged glutamate (Tocris) is uncaged near (~1 μm) the spine of interest by two-photon 720 nm light with a train of 4–6 ms, 2.5–3 mW laser pulses (30 times at 0.5 Hz) (see Note 9).

4. Imaging during uncaging is performed with 2x2 binning at each time-point every 8 seconds, and continued every minute following uncaging for a period of 25-30 minutes (See Note 10).

5. To test the specificity of the sensor response to RhoA activity, we recommend performing control experiments by repeating steps 1-3 with the following changes: a) control imaging solution using 4mM MgCl₂ instead of CaCl₂, and b) using a control sensor that does not target RhoA (e.g. Cdc42 sensor, as per Ref. 14) (see Note 11).

3.2.3. Image analysis

For measurement of fluorescence lifetime, we use a custom script written in MATLAB or C++. The following explanation is intended for code generation and calculation of fluorescence lifetime. Briefly, the fluorescence lifetime decay curve F(t) is fit with a double exponential function. This is used to measure FRET efficiency as it is directly proportional to the lifetime of the donor fluorescent protein, as described in: FRET= 1 − τ_DA/τ_D, where τ_DA is the fluorescence lifetime of the donor in the presence of acceptor, and τ_D is the fluorescent lifetime of the donor alone, τ_D is obtained by fitting the fluorescent lifetime of mCyRFP1-RhoA alone with a mono-exponential convolved with the Gaussian instrument response function:

$$A\mathit{(}t\mathit{)}=A_{\mathit{0}}H\mathit{\left(}t,t_{\mathit{0}},{\tau }_D,{\tau }_G\mathit{\right)}$$ (2)

$$H\left(t,t_0,{\tau }_D,{\tau }_G\right)=½\exp \mathit{\left[}{{\tau }_G}^{\mathit{2}}/\mathit{\left(}2{{\tau }_D}^{\mathit{2}}\mathit{\right)}-\mathit{\left(}t-t_0\mathit{\right)}/{\tau }_D\mathit{\right]}\mathrm{erfc}\mathit{\left[}{{\tau }_G}^2-{\tau }_D\mathit{\left(}t-t_0\mathit{\right)}/\mathit{\left(}2{\tau }_G{\tau }_D\mathit{\right)}\mathit{\right]}$$ (3)

in which τ_G is the width of the Gaussian pulse response function, t₀ is the time offset, and erfc is the complementary error function. A₀ is the initial fluorescence before convolution. The instrument response function represents the characteristic measurement feature for each microscope system. Fitting is performed by calculating weighted residuals, $E(t)={(F(t)-A(t))}^2/F(t)$, by minimizing the error summed over time ${\delta }^2={\text{Σ}}_{t}E(t)$ for fitting parameters t₀, τ_D and τ_G. To analyze RhoA activity using FLIM, we fix the value of τ_D to the fluorescence lifetime obtained from RhoA-mCyRFP1 (3.55 ns) alone. Then we obtain τ_DA by fitting fluorescence lifetime with:

$$A\mathit{(}t\mathit{)}=A_{\mathit{0}}\mathit{\left[}{P}_{D}H\mathit{\left(}t,t_0,{\tau }_D,{\tau }_G\mathit{\right)}+P_{DA}H\mathit{\left(}t,t_0,{\tau }_{DA},{\tau }_G\mathit{\right)}\mathit{\right]}$$ (4)

where P_D and P_DA is the fractional population with the decay time constant of τ_D and τ_DA. It should be noted that reliable fitting requires high binding fraction and number of photons. When this condition is not available, τ_DA = ½ τ_D provides a good approximation. For experiments with small number of photons, we will fix τ_DA and τ_DA, and obtain binding fraction (P_DA).

To generate the fluorescence lifetime images, we calculate the weighted sum of fluorescence decay in each pixel:

$${\tau }_{\text{m}}={\text{Σ}}_{t}tF(t)/{\text{Σ}}_{t}F(t)-t_0$$ (5)

in which t₀ is obtained by a curve fitting to the fluorescence lifetime decay averaged over the whole image.

1. We mark an ROI over the stimulated spine, as well as an ROI over a nearby unstimulated spine in the same image, which can account for non-specific changes in lifetime due to repeated imaging. A background ROI is marked for background subtraction.

2. We calculate and average baseline lifetime and binding fraction for the stimulated spine.

3. We calculate absolute changes in binding fraction during glutamate uncaging, and following stimulation.

4. Using the fluorescent intensity changes, we can compare changes in spine volume indicating structural changes induced by glutamate uncaging. Imaging results are presented in Figure 2.

Figure 2. Measurements of RhoA activity using 2pFLIM.

(a) Representative fluorescent lifetime decay curves of the red-shifted RhoA sensor (RhoA CyRM) expressed in HEK293 cells. Counts of number of photons along with the decay curve double exponential fit allow to quantitatively discriminate activity levels of dominant negative (DN, blue) and constitutive active (CA, red) RhoA mutants. (b) Representative pseudo-colored lifetime images of HEK293 cells showing different basal activity levels of the RhoA sensor with WT, DN and CA versions. Scale bar: 20 μm, (c) Quantification of binding fraction differences of RhoA CyRM sensor showing quantification of differential basal activity of with WT (wild type), DN (dominant negative) and CA (constitutively active) versions. (d) Representative fluorescence lifetime images of RhoA sensor activity in a dendritic branch of CA1 pyramidal neuron in an organotypic hippocampal slice. Images were acquired with 2pFLIM and show the spread of RhoA activity before (−16 seconds) and after (+16 sec, +2 min. +20 min) glutamate uncaging at the dendritic spine indicated with arrowhead. Scale bar: 2 μm. Adapted from Ref. 14.