Spectrally-Resolved Response Properties of the Three Most Advanced FRET Based Fluorescent Protein Voltage Probes

Genetically-encoded optical probes for membrane potential hold the promise of monitoring electrical signaling of electrically active cells such as specific neuronal populations in intact brain tissue. The most advanced class of these probes was generated by molecular fusion of the voltage sensing domain (VSD) of Ci-VSP with a fluorescent protein (FP) pair. We quantitatively compared the three most advanced versions of these probes (two previously reported and one new variant), each involving a spectrally distinct tandem of FPs. Despite these different FP tandems and dissimilarities within the amino acid sequence linking the VSD to the FPs, the amplitude and kinetics of voltage dependent fluorescence changes were surprisingly similar. However, each of these fluorescent probes has specific merits when considering different potential applications.


Introduction
During the last decade, several designs of genetically-encoded optical probes for membrane potential have been explored but only one design, referred to as VSFPs, has been proven to provide a reliable voltage report in mammalian cells so far [1][2][3][4]. These voltage-sensing fluorescent proteins are generated by molecular fusion of a voltage-sensing domain (VSD) with a FRET-based fluorescent protein (FP) pair comprising a donor and an acceptor. Such VSDs are membrane proteins comprising four transmembrane segments S1-S4 with conformational state transitions that are dependent on membrane voltage [5]. In voltage-gated potassium channels (Kv channels), these domains operate the opening and closing of an ion pore. Lately, a homolog to the VSD of Kv channels was found in the Ciona intestinalis voltage-sensitive phosphatase (Ci-VSP) [6]. Interestingly, a single VSD was shown to be functional in Ci-VSP while Kv channels require an assembly of 4 VSD-containing subunits [7]. The self-sufficient nature of the Ci-VSP VSD explains the large improvement between the first generation of VSFPs based on a Kv channel VSD [1] and the second generation (VSFP2s) that uses the VSD from Ci-VSP [2][3]8].
Here, we compared three enhanced variants of the originally reported VSFP2.1 (Fig. 1). The first variant, named VSFP2.3, resulted from linker optimization of VSFP2.1 [1][2]. The second one (VSFP2.4) is composed of a novel yellow and far-red [9] FP pair (mCitrine/mKate2) which is described here for the first time (Fig. S1). The third is the recently reported variant termed Mermaid that involves a FP tandem derived from corals [8].

Results and Discussion
For this study, the three VSFP2.1 variants were expressed in PC12 cells [2]. This expression system has the advantage that in addition to neuron-like membrane properties, the genetic and morphological homogeneity of these cells facilitate quantitative patch-clamp fluorometry.
Voltage clamped cells (35uC) were illuminated with light from a monochromator (425 nm, 480 nm and 460 nm, half width of wavelength (hw) 6 nm, for VSFP2.3, VSFP2.4 and Mermaid, respectively) and, in the first set of experiments, emitted fluorescence was directed via an optical fiber system to a spectrophotometer that acquired the emission spectrum via a back illuminated cooled CCD camera. Emission spectra were recorded during the last 1100 ms of a 1200 ms step to 2100 mV (hyperpolarization) and +40 mV (depolarization) from holding potential (V H ) 270 mV. These recordings allowed us to evaluate the steady-state spectrally-resolved maximal change in fluorescence (DF/F) independent of specific sets of emission filters (Fig. 2). The DF/F values (i.e. dynamic range) for donor and acceptor fluorescence at their emission peak wavelength are summarized in Table 1 along with the corresponding DR/R values. Our analysis did not reveal any differences in the fluorescent signal voltage dependency between the three VSFP2.1 variants (Fig. 3B). Upon depolarization from V H , all three probes exhibited fluorescence signals that could be fitted with two main time constants [3,8,10] that likely correspond to the two known major conformational transitions of the VSDs [5].
The values for these ''on'' time constants were very similar for all three probes. The only striking difference was that the response component with the fast ''on'' time constant contributed to a larger fraction of the total signal in VSFP2.4 as compared to Mermaid (4064% versus 2365% at +60 mV). Accordingly, the initial ''on'' response was faster in VSFP2.4 as compared to Mermaid (asterisk in Fig. 3A). Fluorescence decay upon return to V H (i.e. t off ) was fitted with a single time constant which did not differ between the probes (Fig. 3D).
The DF/F values of FRET-based FP sensors depend on many factors but the conceptually most straightforward are the spectral overlap between donor emission and acceptor absorbance and the acceptor extinction coefficient. Critical structural parameters include the transition dipole baseline orientation and its modulation by the probe activation. These factors are difficult to predict from available structural data since it is likely that the different FPs used within the three VSFP2.1 variants have a different dipole orientation relative to their secondary structure. Indeed, it is well established that single amino acid substitutions linking the components of the fusion proteins can have dramatic effects on  DF values of calcium sensors (see e.g. [11]). Therefore, it was quite surprising to find a relatively modest difference in the response properties among the VSFP2.1 variants. Thus, their voltagedependent donor dequenching was very similar and the observed differences in acceptor modulation can be explained by photophysical differences between the corresponding FP. For instance, a slightly larger modulation of the acceptor fluorescence in Mermaid was anticipated from reduced direct acceptor excitation when  exciting the donor. Although modest, these differences may along with emerging computational approaches [12] drive further enhancements of this class of membrane voltage probes. Furthermore, the absence of difference in response time constants of the three VSFP2.1 variants tested in this study was expected because of the very similar general design of the probes, which therefore leads to tracking of the same conformational changes with different FP tandems [5]. However, the striking difference in the contribution of the two ''on'' time constants demonstrated the importance of probe ''fine tuning'' at the level of single amino acid. Direct comparison of the three variants was made possible using spectrally resolved DF/F measurements. We therefore propose that this method should be used in future to provide definitive evidence of sensor improvements since this measure depends much less on a particular set of filters.
Most conceivable applications of VSFP2.1 type probes for membrane potential will be based on standard imaging which requires the use of band pass filters to select appropriate spectral ranges for donor and acceptor channels. The spectral responses can be multiplied by the transmission spectrum of candidate filters to conveniently predict DF/F values and shot noise characteristics. Our data indicate that each of the three membrane voltage probes has specific merits. VSFP2.3 is based on the most widely used FP pair and hence is suitable for instrumentation with standard optical components. The well balanced absolute fluorescence and dynamic range of donor and acceptor makes Mermaid a good candidate for dual emission (i.e. ratiometric) measurements. Limitations of Mermaid are the relatively low fluorescence quantum yield and high bleaching rate of the used FPs [8] as well as a reduced contribution of the fast ''on'' response component. Another yet to be solved problem with Mermaid is its tendency to form fluorescent aggregates. In experimental configurations where single emission approaches and fast response times are preferred (e.g. based on signal-to-noise consideration), VSFP2.4 (using the YFP channel) is likely the best choice. The far-red channel of this variant could be the best option if green tissue autofluorescence or light absorption by hemoglobin is an issue (e.g. as in in vivo imaging). Furthermore, the spectral properties of VSFP2.4 will facilitate deep tissue imaging using two-photon excitation microscopy.

Molecular Biology
The VSFP2.3 construct was generated as previously described in [3]. VSFP2.4 was obtained by substituting the FP pair in VSFP2C [2] with mCitrine and mKate2 using the NotI and HindIII restriction sites. The R217Q mutation was introduced by sitedirected mutagenesis. Truncated mCitrine (i.e. residues 1 to 232) was fused to mKate2 by overlap-extension PCR using the following set of complementary primers: 59-GCCGGGATCACTCT-CATGGTGAGCGAGCTGATTAAG-39 and 59-CTTAAT-CAGCTCGCTCA CCATGAGAGTGATCCCGGC-39. Mermaid was kindly provided by Dr. Miyawaki (RIKEN BSI, Japan). The coding sequence was amplified using a sense primer comprising a NheI site (59-ATTAGCTAGCGCCACCATGGAGGGATTC-GACGGTTCA-39) and an antisense primer containing a EcoRI site (59-TGGAATTCTTAGGAATGAGCTACTGCATCTTC-TACCTG-39). The amplified PCR fragment was then digested and subcloned in pcDNA3.1(-) vector (Invitrogen). The sequence information presented in Figure 1 was obtained from sequencing the Mermaid DNA and confirmed by Dr. H Tsutsui to be correct.

Data Analysis
The fluorescence and electrophysiological signals were analyzed using Clampfit (Axon Instruments) and Origin software (Origi-nLab, Northhampton, MA, USA). Photobleaching (typically less than 0.1%/s) was corrected by subtraction of a linear fit of the bleaching curve. Fluorescence signals were ''background corrected'' by subtracting offsets measured from regions devoid of cells.
This offset corresponds to reflected and unblocked excitation light and any probe-independent fluorescence within the optical path. Fluorescence transients (F(t)) were fitted with a double exponential function of the form F(t) = F baseline +DFfast ? exp(2t/tfast)+DFslow exp(2t/tslow). Figure S1 Spectral properties of fluorescent proteins used in VSFP2.4. The emission spectrum of the donor (Citrine) and absorption spectrum of the acceptor (mKate2; [9]) are shown in yellow and red, respectively. The spectral overlap between the emission of the donor and absorption of the acceptor is indicated in gray. The emission spectrum of Citrine was obtained from the laboratory webpage of Dr. Roger Tsien (http://www.tsienlab. ucsd.edu/Documents.htm). Found at: doi:10.1371/journal.pone.0004555.s001 (0.66 MB DOC)