A New Generation of FRET Sensors for Robust Measurement of Gαi1, Gαi2 and Gαi3 Activation Kinetics in Single Cells

G-protein coupled receptors (GPCRs) can activate a heterotrimeric G-protein complex with subsecond kinetics. Genetically encoded biosensors based on Förster resonance energy transfer (FRET) are ideally suited for the study of such fast signaling events in single living cells. Here we report on the construction and characterization of three FRET biosensors for the measurement of Gαi1, Gαi2 and Gαi3 activation. To enable quantitative long-term imaging of FRET biosensors with high dynamic range, fluorescent proteins with enhanced photophysical properties are required. Therefore, we use the currently brightest and most photostable CFP variant, mTurquoise2, as donor fused to Gαi subunit, and cp173Venus fused to the Gγ2 subunit as acceptor. The Gαi FRET biosensors constructs are expressed together with Gβ1 from a single plasmid, providing preferred relative expression levels with reduced variation in mammalian cells. The Gαi FRET sensors showed a robust response to activation of endogenous or over-expressed alpha-2A-adrenergic receptors, which was inhibited by pertussis toxin. Moreover, we observed activation of the Gαi FRET sensor in single cells upon stimulation of several GPCRs, including the LPA2, M3 and BK2 receptor. Furthermore, we show that the sensors are well suited to extract kinetic parameters from fast measurements in the millisecond time range. This new generation of FRET biosensors for Gαi1, Gαi2 and Gαi3 activation will be valuable for live-cell measurements that probe Gαi activation.


Introduction
The Gα i subclass of heterotrimeric G-proteins consists of 3 members in humans, Gα i1,2,3 encoded by the genes GNAI1, GNAI2, GNAI3 [1] and is activated by a wide range of G-protein coupled receptors. The Gα i family of G-proteins have been implicated in numerous pathologies, from involvement in obesity and diabetes [2], functions in the immune system [3] to their critical roles in several stages of cancer biology [4][5][6][7]. Activation of Gα i is predominantly linked to the inhibition of adenylate cyclases, which leads to decreased cAMP accumulation in cells. However, activation of Gα i has more recently been connected to several other molecular effectors, including PI3K/Akt [8,9], ERK [10] and c-Src [5].
The measurement of Gα i activation is classically performed by measuring the inhibition of forskolin-induced cAMP production. Similar to phosphorylation assays further downstream, such measurements lack spatial resolution, have limited temporal resolution and can be influenced by considerable crosstalk and amplification or desensitization of the signal [11][12][13].
To investigate G-protein activation in a direct way with high spatiotemporal resolution, genetically encoded FRET (Förster Resonance Energy Transfer) or BRET (Bioluminescent Resonance Energy Transfer) biosensors can be employed [14]. These methods are based on the measurement of the non-radiative energy transfer from a donor molecule to an acceptor molecule, which only takes place when donor and acceptor are in close proximity of each other (<10nm). Changes in distance or orientation between the donor and acceptor dipole result in changes in the RET efficiency, which can be quantified.
The RET techniques allow for single cell recordings of the kinetics with millisecond resolution, which can be used to identify cell-to-cell heterogeneity and record pharmacokinetic parameters. Moreover, this approach has the potential to record GPCR activation under physiological conditions in vivo [15].
To perform FRET measurements, a spectrally overlapping donor and acceptor pair is necessary [24], and it was previously shown that the use of brighter fluorescent proteins can improve the sensitivity of FRET biosensor measurements [22,23]. In order to obtain robust FRET measurements that probe Gα i activation, we have made fusions of Gα i1 Gα i2 and Gα i3 with the brightest and most photostable monomeric cyan fluorescent protein (CFP) currently available, mTurquoise2 (mTq2) [25]. As acceptor we have used circular permutated Venus (cpVenus) fused to Gγ 2 , which has previously been used as acceptor in a single plasmid Gα q FRET sensor [26]. We use a single plasmid strategy to facilitate transfection protocols and allow a welldefined donor and acceptor expression ratio in cells [27]. This expression strategy should greatly facilitate the use and reproducibility of the results of these sensors. We present the construction strategy, validation and characterization of this new generation of FRET sensors for the activation of Gα i1 Gα i2 and Gα i3 . These biosensors are very well suited for live cell microscopy and can be used for fast kinetic measurements in the millisecond range, allowing pharmacological drug characterization and determination of on-and off-kinetics for agonists and antagonists at Gα i -coupled GPCRs.

Generation of constructs
The monomeric CFP variant mTurquoise2, the preferred donor in CFP-YFP FRET pairs due to its high quantum yield and photostability [25], was inserted into Gα i1 after the alanine on position 121 in the αB-αC loop. This insertion site that was previously shown to retain nucleotide exchange and GTPase reaction rates comparable to wild-type protein [20]. Gα i1 -mTq2 displays plasma membrane localization when expressed in HeLa cells (Fig 1A). Trial experiments were performed to examine whether Gα i1 -mTq2 is suitable for measuring, by means of FRET, the activation of the heterotrimeric G-protein complex upon GPCR activation. To this end, both Gβ as well as Gγ can be tagged with an acceptor to measure heterotrimeric G-protein activation by FRET [20,21]. We have previously shown that the highest FRET contrast, for Gα q , is obtained with cpVenus-Gγ 2 [26]. Since Gα i has high structural homology to Gα q and the site at which mTurquoise2 is inserted is similar, we decided to employ the same FRET acceptor here. Hence, to introduce a labeled heterotrimeric G-protein complex in cells, we co-expressed Gα i1 -mTq2 together with the FRET acceptor cpVenus-Gγ 2 and untagged Gβ 1 [26]. Stimulation of the co-expressed α 2 adrenergic receptor (α 2 AR) with UK14,304 shows a rapid increase in CFP fluorescence and a concomitant loss of sensitized emission from the YFP channel, reflecting a loss of FRET. The loss of FRET can be interpreted as a dissociation of the heterotrimer or a change in relative conformation of donor and acceptor. To enable robust co-expression of the different components of the multimeric FRET sensor, we introduced the subunits on the same plasmid, as we reported previously for a Gα q sensor ( Fig 1B) [26]. This strategy uses a viral 2A peptide and an IRES sequence to ensure optimal relative levels of the donor (Gα i1 -mTq2) and acceptor (cpVenus-Gγ 2 ) within single cells, while minimizing cell-to-cell expression heterogeneity in the sample. After generating our first variant, we noticed that Gα i1 -mTq2 was mislocalised in the cytoplasm in many cells ( Fig 1C). The Gα i1 subunit is myristoylated and requires a glycine residue immediately following its starting methionine. Detailed inspection of the plasmid sequence revealed an additional starting codon for the Gα i1 -mTq2 upstream of the native start-codon, generated by the IRES sequence ( Fig 1B). We hypothesized that in our first variant of the sensor, most of the Gα i1 -mTq2 protein produced was translated from the upstream methionine in the IRES sequence, which does not result in myristoylated protein. We removed the upstream methionine by whole-vector PCR (see material and methods), which only leaves the native starting codon of Gα i1 -mTq2, followed by a Glycine providing the consensus sequence for myristoylation. Indeed, after transfecting cells with the new plasmid, we observed correctly localized Gα i1 -mTq2 (Fig 1B and 1C). A Gα i2 -sensor and Gα i3sensor were constructed in a similar way. To examine the co-expression of the three subunits (Gα i1 -mTq2, Gβ 1 and cpVenus-Gγ 2 ) from a single plasmid versus three separate plasmids, we quantified the CFP and YFP fluorescence in these two experimental conditions ( Fig 1D). The CFP and YFP fluorescence in the single plasmid strategy transfection had a coefficient of determination r 2 of 0.64, whereas transfections with the three separate plasmids showed a coefficient of determination r 2 of 0.36 between CFP intensity and YFP intensity. In other words, the correlation between CFP and YFP expression is better in the single plasmid configuration, indicating a clear advantage of this design. An additional advantage of this plasmid is the 3:1 protein expression upstream and downstream of the IRES sequence, which has previously been shown to result in a preferred donor (CFP) and acceptor (YFP) expression ratio for an analogous Gα q FRET sensor [27]. Finally, the single plasmid constructs will simplify introduction into primary cells, the generation of stable cell lines or transgenic organisms with Gα i -sensors.

Performance in GPCR activation assays
To test the new Gα i1 biosensor in live cell imaging, we employed a well-characterized GPCR known to couple to Gα i1 , the α 2 adrenergic receptor (α 2 AR). HeLa cells, shown before to contain the α 2 AR endogenously [20], were transfected with the Gα i1 biosensor. Upon addition of 10μM UK14,304 we observed a robust loss of FRET by measuring the ratio between the YFP and CFP fluorescence of the Gα i1 biosensor, which was reversed back to baseline by the expression analysis of the CFP and YFP channels of the cp173Venus-Gγ 2 and Gα i1 -mTurquoise2-Δ9 transfections in HeLa cells. Single plasmid transfection (left) versus the transfection of the separate plasmids (right). The dots depict the CFP and YFP intensity, quantified from individual single cells. The r 2 is the coefficient of determination. Width of the individual images in A and C is 143μm.
doi:10.1371/journal.pone.0146789.g001 addition of 60μM of the α 2 AR antagonist Yohimbine (Fig 2A). Pertussis toxin (PTX) has been shown to inactivate Gα i signaling in cells via ADP-ribosylation of the Gα i subunit [28], which prevents its interaction with GPCRs. The activation of the Gα i1 was completely abolished by overnight incubation with PTX, showing that Gα i1 -mTq2 protein fusion is still PTX-sensitive (Fig 2A). To confirm that the sensor can be used to assay Gα i1 activation of endogenous receptors in primary cells, we repeated this experiment in HUVEC (human umbilical vein endothelial cells). Addition of a well-known stimulant for HUVECs, S1P [29], caused a sustained decrease in FRET ratio of the Gα i1 -sensor (Fig 2B), overnight treatment with PTX completely abolished this response. Next, to investigate how robust the Gα i1 sensor performs on other GPCR activation assays, we tested a variety of GPCRs shown to couple to Gα i . The bradykinin 2B (BK 2B ) receptor [30], lysophosphatidic acid 2 (LPA 2 ) receptor [31,32] and muscarinic acetylcholine 3 (M 3 ) receptor [33,34] were co-transfected with the Gα i1 biosensor in HeLa cells. Upon stimulation with the relevant agonists, all three receptors showed a sustained decrease in FRET ratio of the Gα i1 biosensor (Fig 2C). The M 3 receptor also showed a full recovery back to baseline of the FRET ratio after addition of the antagonist atropine. The M 3 receptor is mainly known for its signaling via Gα q . Still, previous studies have shown Gα i activation via the M 3 receptor [19,[33][34][35][36], fitting with with our observations.
In the control conditions, e.g. absence of over-expressed GPCR, we stimulated HeLa cells with the relevant agonist and antagonist, and we observed only a very minor response on the Gα i1 sensor in the case of LPA stimulation. This is most likely due to the activation of endogenous LPA receptors in HeLa cells [37]. When we co-transfected the β 2 adrenergic receptor (β 2 AR), none of the cells showed Gα i1 activation in response to the agonist and antagonist treatment ( Fig 2C). Of note, β 2 AR is a classical activator of Gα s but switching to Gα i has been reported under certain conditions [38]. Our results fit with the only study that we are aware of that uses similar tools (BRET based sensors) and similar conditions (over-expressed β 2 AR and heterotrimeric G-protein sensors) [19]. Also in that case no activation of Gα i1 was observed by β 2 AR stimulation (and only little activation of Gα i2 and Gα i3 , which was >10-fold lower than activation by the alpha-2C adrenergic receptor, a strong activator of Gα i ).
To verify the performance of the Gα i2 and Gα i3 biosensors, we transfected HeLa cells with their respective plasmids. Similar to Gα i1 biosensor experiment in Fig 2A we observed a robust loss of FRET after addition of 10μM UK14,304, and the signal returned to baseline upon addition of 60μM Yohimbine (Fig 2D). Under these experimental conditions we did not observe substantial differences in the activation kinetics or amplitude of the responses between the three different Gα i subunits. Both Gα i2 -mTq2 and Gα i3 -mTq2 are still sensitive to PTX treatment, as shown by the abolishment of the FRET response after overnight incubation with PTX ( Fig 2D).

Fast kinetic measurements
In order to look at the sub-second kinetics of Gα i1 activation in living cells in more detail, HEK293 cells were co-transfected with the Gα i1 -sensor and the α 2 AR or adenosine A1 receptor, respectively. Using a fast perfusion system for ligand application, single-cell FRET measurements show a rapid loss in FRET ratio of more than 15% after short-term application of 20μM norepinephrine. After ligand washout, the FRET signal returns to baseline levels. This could be reproduced several times without any apparent loss in signal amplitude (Fig 3A). A similar response was observed for the adenosine A1 receptor after short application of the endogenous ligand adenosine (30μM) (Fig 3B).
These fast FRET measurements can be used to estimate the on-kinetics of Gα i1 activation with sub second resolution (Fig 3C), as shown by a close-up of the first stimulation in the  Fig 3A. The curve was fitted to a one component exponential decay function as previously described [39], resulting in an exponential time constant (τ) of 1160ms.
To assess the precise on-rate kinetics of the Gα i1 -sensor, cells were stimulated with saturating ligand concentrations (100μM norepinephrine or 30μM adenosine). Each individual response was fitted to a one component exponential, this resulted in average τ values for α 2 AR of 887ms and for adenosine A1 of 963ms (Fig 3D), corresponding to half-times of 614ms and 668ms respectively. These values are in good agreement with earlier observations for G-protein activation by FRET [21,40,41].

Concluding Remarks
In this manuscript we describe the design, construction and characterization of three new FRET biosensors for the measurement of Gα i1 , Gα i2 , Gα i3 activation. The new sensors contain a Gα subunit fused to the donor fluorophore, mTurquoise2, and the Gγ subunit fused to cp173Venus, as it was previously shown that this combination for a Gα q FRET sensor provides the largest dynamic range [26]. The three subunits of the heterotrimer (Gα i -mTq2, Gβ 1 and cpVenus-Gγ 2 ) were configured on a single plasmid, enabling robust co-expression with a preferred stoichiometry. We show that these sensors are well suited for live cell microscopy and extracting kinetic parameters by single-cell ratiometric FRET imaging. The standardized layout of these FRET biosensors for G-protein activation will improve reliability and reproducibility of experiments within and between laboratories. This is exemplified in this paper by the robust performance of the Gα i1 sensor in three different laboratories, without optimization of the experimental conditions. One limitation of energy transfer based biosensors for heterotrimeric G-proteins is that they depend on overexpression of the heterotrimer, which may affect the natural preference of the GPCR for a certain class of heterotrimeric G-proteins. Tagging the endogenous subunits with fluorescent proteins can potentially alleviate this.
The exquisite sensitivity of these sensors enables the robust detection of Gα i activation in primary cells via endogenous GPCRs. Moreover, these biosensors can be used to directly compare the preferential activation patterns of Gα i1 Gα i2 and Gα i3 between different Gα i coupled GPCRs, which can aid the development of therapeutic strategies targeting Gα i signaling pathways [42].
Human umbilical vein endothelial cells (HUVECs) were purchased from Lonza and cultured at Sanquin Blood Supply (Amsterdam, the Netherlands) on FN-coated dishes in EGM-2 medium, supplemented with singlequots (Lonza, Verviers, Belgium). HUVECs were used at passage number 4 or 5. The Neon transfection system (MPK5000, Invitrogen) and a corresponding Neon transfection kit (Invitrogen) were used as transfection method. A single pulse was generated at 1300 Volt for 30ms to microporate HUVECs with 2μg cDNA, cells were subsequently seeded on FN-coated glass coverslips.

Widefield microscopy
Ratiometric FRET measurements in HeLa cells (results presented in Fig 2A, 2C and 2D) were performed using a wide-field fluorescence microscope (Axiovert 200 M; Carl Zeiss GmbH, Germany) at the University of Amsterdam (Amsterdam, the Netherlands), kept at 37°C, equipped with an oil-immersion objective (Plan-Neo-fluor 40×/1.30; Carl Zeiss GmbH) and a xenon arc lamp with monochromator (Cairn Research, Faversham, Kent, UK). Images were recorded with a cooled charged-coupled device camera (Coolsnap HQ, Roper Scientific, Tucson, AZ, USA). Typical exposure times ranged from 75ms to 150ms, and camera binning was set to 4x4. Fluorophores were excited with 420 nm light (slit width 30nm) and reflected onto the sample by a 455DCLP dichroic mirror and CFP emission was detected with a BP470/30 filter, and YFP emission was detected with a BP535/30 filter by rotating the filter wheel. In the co-expression experiments, YFP was excited with 500nm light (slit width 30nm) and reflected onto the sample by a 515DCXR dichroic and emission was detected with a BP535/30 filter. Acquisitions were corrected for background signal and, for FRET ratio imaging, bleedthrough of CFP emission in the YFP channel (55% of the intensity measured in the CFP channel). Fig 2B), a Zeiss Observer Z1 microscope was used at Sanquin Blood Supply (Amsterdam, the Netherlands) with a 40x NA 1.3 oil immersion objective and an HXP 120 V excitation light source. CFP was excited through a FRET filter cube (Exciter ET 436/20x, and 455 DCLP dichroic mirror (Chroma, Bellows Falls, Vermont, USA)). The emission was directed to an attached dual camera adaptor (Carl Zeiss GmbH, Germany) controlling a 510 DCSP dichroic mirror (Chroma, Bellows Falls, Vermont, USA). Emission wavelengths between 455-510 nm are directed to an emission filter ET 480/40 (Chroma, Bellows Falls, Vermont, USA) and then captured by a Hamamatsu ORCA-R2 camera. Emission wavelength 510 nm and higher are directed to an ET 540/40m emission filter (Ludl Electronic Products, NY, USA) and then captured by a second Hamamatsu ORCA-R2 camera. Image acquisition was performed using Zeiss-Zen 2011 microscope software. All acquisitions were corrected for background signal. Acquisitions were corrected for background signal and bleedthrough of CFP emission in the YFP channel (62% of the intensity measured in the CFP channel).

For the FRET experiments in HUVECs (results presented in
For the rapid kinetic measurements of Gα i1 activation (results presented in Fig 3), imaging was performed on a Zeiss Axiovert 200 inverted microscope at the University of Wuerzburg (Wuerzburg, Germany), equipped with an oil immersion 63x objective lens and a dual-emission photometric system (Till Photonics) as described before [21]. The transfected cells were excited with light from a polychrome IV (Till Photonics). Illumination was set to 40ms out of a total integration time of 100ms. CFP (480 ± 20 nm), YFP (535 ± 15 nm), and FRET ratio (YFP/ CFP) signals were recorded simultaneously (beam splitter DCLP 505 nm) upon excitation at 436 ± 10 nm (beam splitter DCLP 460 nm). Fluorescence signals were detected by photodiodes and digitalized using an analogue-digital converter (Digidata 1440A, Axon Instruments). All data were recorded on a PC running Clampex 10.3 software (Axon Instruments). Resulting individual traces were fit to a one component exponential decay function to extract the exponential time constant, tau [39]. The halftime of activation (t 1/2 ) is defined as τ Ã ln2. In dynamic experiments, cells were stimulated with UK14,304 (10μM), Yohimbine (60μM), Bradykinin (1μM), LPA (1μM), Carbachol (100μM), Atropine (10μM), Isoproterenol (10μM), Propranolol (10μM), S1P (500nM), 20μM or 100μM norepinephrine or 30μM adenosine at the indicated time points. ImageJ (National Institute of Health) was used to analyze the raw microscopy images. Further processing of the data was done in Excel (Microsoft Office) and graphs and statistics were conducted using Graphpad version 6.0 for Mac, GraphPad Software, La Jolla California USA, www.graphpad.com.

Confocal microscopy
HeLa cells transfected with the indicated constructs were imaged using a Nikon A1 confocal microscope equipped with a 60x oil immersion objective (Plan Apochromat VC, NA 1.4). The pinhole size was set to 1 Airy unit (<0.8μm).
Samples were sequentially excited with a 457nm and a 514nm laser line, and reflected onto the sample by a 457/514 dichroic mirror. CFP emission was filtered through a BP482/35 emission filter; YFP emission was filtered through a BP540/30 emission filter. To avoid bleedthrough, images were acquired with sequential line scanning modus. All acquisitions were corrected for background signal.
Supporting Information S1 Data. The compressed file contains all the data that was used in this manuscript. (ZIP)