Mechanistic Studies of the Genetically Encoded Fluorescent Protein Voltage Probe ArcLight

ArcLight, a genetically encoded fluorescent protein voltage probe with a large ΔF/ΔV, is a fusion between the voltage sensing domain of the Ciona instestinalis voltage sensitive phosphatase and super ecliptic pHluorin carrying a single mutation (A227D in the fluorescent protein). Without this mutation the probe produces only a very small change in fluorescence in response to voltage deflections (∼1%). The large signal afforded by this mutation allows optical detection of action potentials and sub-threshold electrical events in single-trials in vitro and in vivo. However, it is unclear how this single mutation produces a probe with such a large modulation of its fluorescence output with changes in membrane potential. In this study, we identified which residues in super ecliptic pHluorin (vs eGFP) are critical for the ArcLight response, as a similarly constructed probe based on eGFP also exhibits large response amplitude if it carries these critical residues. We found that D147 is responsible for determining the pH sensitivity of the fluorescent protein used in these probes but by itself does not result in a voltage probe with a large signal. We also provide evidence that the voltage dependent signal of ArcLight is not simply sensing environmental pH changes. A two-photon polarization microscopy study showed that ArcLight's response to changes in membrane potential includes a reorientation of the super ecliptic pHluorin. We also explored different changes including modification of linker length, deletion of non-essential amino acids in the super ecliptic pHluorin, adding a farnesylation site, using tandem fluorescent proteins and other pH sensitive fluorescent proteins.


Introduction
ArcLight [1] is a genetically encoded fluorescent voltage probe is based on the voltage sensing domain (VSD) of Ciona instestinalis voltage sensitive phosphatase (Ci-VSP) [2] fused to the fluorescent protein (FP) super ecliptic pHluorin [3] carrying a single point mutation (A227D in the FP). ArcLight responds to voltage changes across the cell membrane of cultured HEK293 cells with large amplitude decreases in fluorescence (230 to 240% DF/F, in response to a 100 mV depolarization; 1). ArcLight allows reliable detection of single action potentials and sub-threshold electric events in cultured hippocampal neurons in single trials [1] and in vivo in Drosophila [4]. It is unclear how this mutation, A227D, produced a probe which has such a large modulation of its fluorescence output in response to changes of membrane potential however it does appear to be transferable to other voltage sensors [5]. We have tested a large number of FPs in conjugation with the Ciona voltage sensitive domain but have not found one with as large modulation as the ArcLight fluorescent protein.
Many other FP voltage sensors are also based on Ci-VSP. These include: VSFP2 [6], Mermaid [7,8], split fluorescent proteins [9], Butterfly [10] and ElectricPK [11]. The Ciona-based FP voltage sensors often show good membrane expression in mammalian cells [1,6,7,8]. Voltage sensitive phosphatase (VSP) from other species were also exploited [5]. ArcLight variants based on chicken VSP showed fast kinetics but decreased response amplitude [5]. ASAP1 is based on chicken VSP and circularly permuted GFP [12]. It also features a fast response. Microbial rhodopsin based voltage sensors [13,14,15] have the most rapid kinetics so far, but the low brightness of these probes limits their use in broader applications. A new approach to combine the rapid kinetics of a fungal rhodopsin with the brightness of genetically engineered protein fluorophores resulted in a probe with fast kinetics and brightness [15,16].

Critical residues required in the super ecliptic pHluorin in ArcLight for producing large signals
Replacing of the super ecliptic pHluorin A227D in ArcLight with eGFP resulted in a probe with little voltage sensitivity [1]. eGFP differs from super ecliptic pHluorin at only nine residues (R80Q, D147S, Q149N, A163V, G175S, F202S, T204Q, T206A, H231L). We identified three residues, D147, F202 and T204, in super ecliptic pHluorin A227D, which are required for producing ArcLight signals [1]. Here, we reverse the process and introduced four critical residues to an eGFP based probe, in hopes of converting its response properties to those of ArcLight ( Figure 1). We found that the new probe produced a signal with voltage sensitivity as large as that of ArcLight. eGFP based probes carrying less than a complete set of four mutations produced signals with greatly reduced amplitude ( Figure 1).
Neither the N149Q nor the A206T mutations are needed for generating a large signal, a result consistent with our previous observations [1].
The A227D mutation does not alter the spectral properties or pH sensitivity of the super ecliptic pHluorin [1]. We sought to determine the effect of the other 3 residues on the spectral properties or pH sensitivity of the fluorescent protein.
Single mutations or a combination of mutations were introduced to the free fluorescent proteins and the emission spectra were measured. We determined the pH sensitivities of these free fluorescent proteins and found that residue D147 is acids that are the same as those in super ecliptic pHluorin A227D. A probe with poor plasma membrane localization is indicated with the Greek letter D. Means with the same letter do not differ significantly (Tukey-Kramer HSD test). eGFP (n510), eGFP-A227D (n55), eGFP-A227D-S147D (n54), eGFP-A227D-S147D-N149Q (n54), eGFP-A227D-S202F-Q204T (n55), eGFP-A227D-S147D-S202F-Q204T (n54), ArcLight-S249 (n510). All the measurements were made in HEK293 cells (See Methods). Insert is the 3D structure of eGFP (PDB file 1EMG). The non-identical amino acids of eGFP and super ecliptic pHluorin A227D that lie on the facing surface of eGFP are colored and numbered. B) Mutations to eGFP and their combinations in making those probes tested in panel A are given in corresponding columns in the responsible for the right shifted pH sensitivity of pHluorin derived fluorescent proteins. D147 alone is sufficient to change the pKa of eGFP (,6.0) to that of super ecliptic pHluorin (,7.1) (Figure 2A and 2B), but it is not sufficient to generate a large voltage sensitive signal (Figure 1). F202 and T204 do not significantly change the pH sensitivities of the fluorescent proteins (Figure 2A and 2B), but are needed for the large signal ( Figure 1).

2.
Is the voltage sensitivity of ArcLight caused by direct sensing of the environmental pH?
The local pH near the plasma membrane could change during the action potential or sub-threshold electric activity, due to movement of protons attracted or repelled by the electric field, or as a result of the ion fluxes that occur during these events. We had already determined [1] that pHluorin alone (without the A227D mutation) does not produce probes with a large DF and yet has the same pH sensitive response curve as the ArcLight FP. We sought to determine if the voltage-dependent fluorescent response of ArcLight is mediated by environmental pH changes or by the conformation changes of the Ciona-based voltage sensor.
First we determined if Ciona-based FP voltage sensors carrying other FPs with base-shifted pH sensitivity curves could result in similar signals as seen with ArcLight. We replaced the fluorescent protein in ArcLight-S249 with either YFP or ratiometric pHluorin [3]. Both fluorescent proteins have pH/fluorescence profiles that are similar to pHluorin [17]. The resulting probes exhibit less than 1.5% changes in response amplitude to a 100 mV depolarization ( Figure 3A-3D). Introducing the A227D mutation reduced the signal amplitude even further ( Figure 3A-3D). Thus not every pH sensitive fluorescent protein can replace super ecliptic pHluorin A227D in generating a large voltage response Next, we compared the F-V curves of three derivatives of ArcLight, ArcLight-249, ArcLight-249 (R217) and ArcLight-A242 (R217), whose voltage sensors are different in their voltage sensitivity. The voltage sensor of ArcLight contains a point mutation, R217Q, which changes its voltage sensitivity to a more physiologically relevant range, with a V 1/2 at around 226 mV [18]. In comparison, the ArcLight-249 and -242 without the R217Q mutation have their V 1/2 at around +82 mV and +86 mV, respectively. We found that these FP voltage sensors are not simply sensitive to a change in pH that results from activity because they have very different voltage sensitivities that correspond to the sensitivities of their own voltage sensors ( Figure 4A and 4B) depending on differences in the S4 transmembrane helix.

Two-photon polarization microscopy study of the FP moiety movement in ArcLight
We used a two-photon polarization microscope [19] to study the orientation and movement of the FP moiety in three similarly constructed probes: Ciona voltage sensitive domain-ecliptic pHluorin (CiVSD-EP), Arclight-S249 and Arclight-Q239. All three probes show similar linear dichroism (differences in absorption of light of distinct linear polarizations), indicative of the fluorophore's long axis being close to parallel to the cell membrane ( Figure 5A). The two-photon F-V curves of these probes ( Figure 5B) are similar to those recorded with singlephoton microscopy [1]. Arclight-Q239 (234%) and Arclight-S249 (220%) showed much larger signal amplitudes than CiVSD-EP (23%) in response to a 100 mV depolarization. However, changes in the dichroic ratio (Dr max /r max ) of the three probes did not correlate with their fluorescence intensity changes in either signal size or sign. The Dr max /r max of CiVSD-EP and Arclight-S249 increased by 3.4% for CiVSD-EP and 2.6% for Arclight-S249 for a 100 mV depolarization, while the Dr max /r max of ArcLight-Q239 decreased by 27.2% ( Figure 5C). The DF/F vs Dr max /r max was linear for each probe ( Figure 5D).
We also measured the dynamics of the fluorescence ( Figure 5E) and dichroic ratio changes ( Figure 5F) with two-photon polarization microscopy. The DF/F changes during a 100 mV depolarization and repolarization were fit with single exponential equations for both ArcLight-S249 (t on 520 ms, t off 5110 ms) and ArcLight-Q239 (t on 538 ms, t off 570 ms). The speed of dichroic ratio change was similar (ArcLight-S249: t on 515 ms, t off 5101 ms; ArcLight-Q239: t on 534 ms, t off 558 ms.). We plotted Dr max /r max against the DF/F of the different time points during the depolarization and repolarization, again, the Dr max /r max vs DF/F appears linear during the voltage transitions ( Figure 5G).

Modification to the linker length of ArcLight
Nineteen linker length derivatives of ArcLight were generated by inserting the super ecliptic pHluorin A227D after each residue between A231 and S249 of the Ciona voltage sensitive phosphatase sequence ( Figure 6A). Three of these derivatives, I233, F234 and Y235, did not express on the plasma membrane in HEK293 cells. The five ArcLight derivatives previously reported, i.e. Q239, M240, K241, A242 and S243 [1], exhibited the largest voltage sensitivity, while probes with greater or shorter linker lengths display a gradual reduction in voltage response ( Figure 6B). The dynamics of all these probes are best fit with double exponential equations during depolarization and repolarization. None of the new linker length modified derivatives had ''on'' response kinetics significantly different than the previously reported five ArcLight derivatives ( Figure 6C and Table 1). However, the time constants (tau) of the fast component during repolarization decreased with shorter linker lengths ( Figure 6D).
The linker sequences of all ArcLight derivatives previously reported include three additional amino acids, glycine, aspartic acid and proline (GDP), translated from a BamHI restriction site introduced during plasmid construction ( Figure 6A). We created a new construct, ArcLight-A242-GDP del , which lacks the three extra amino acids (GDP) and compared its voltage response and dynamics to probes with similar linker lengths. ArcLight-A242-GDP del showed similar  voltage sensitivity as ArcLight-A242 and ArcLight-Q239 ( Figure 6E). However, its response dynamics were slower during both depolarization and repolarization ( Figure 6F and 6G).

Insertion sites
Tau-on-1 (ms) A-on-1 (%) Tau-on-2 (ms) Tau-off-1 (ms) A-off-1 (%) Tau-off-2 (ms) The dynamics of all tested probes of linker modified ArcLight derivatives during depolarization and repolarization were fit with double exponential equations. N-terminus and nine amino acids at the C-terminus can be deleted from the sequence of eGFP without affecting its fluorescent properties [20]. Super ecliptic pHluorin is homologous to eGFP. We determined if deletion of these nonessential amino acids in the super ecliptic pHluorin A227D of ArcLight affects the response properties of the probe. We made derivatives of ArcLight-Q239 with deletion of either the five amino acids from the N-terminus (Q239-ND5) or the  of the depolarization and repolarization response of these two derivatives were not significantly faster, with tau-on-1 around 10 ms and tau-off-1 around 15-20 ms ( Figure 7C and 7D).

Farnesylation of the fluorescent protein in ArcLight
The original ArcLight has the super ecliptic pHluorin anchored to the plasma membrane only by the linker to the Ciona voltage sensor domain. We determined the effect of adding a farnesylation site to the C-terminal of the probe ( Figure 8A). The farnesyl group anchors the fluorescent protein to an additional site of the plasma membrane and therefore may change the response properties of the probe. We found that the sensitivity of the probe with the farnesylation site is reduced by half ( Figure 8B), and the taus of double exponential fitting are also larger than those of ArcLight ( Figure 8C and 8D).

Using tandem fluorescent proteins in ArcLight
One possible mechanism by which ArcLight produces a large response is intermolecular dimerization between FPs in two adjacent ArcLight molecules. Introduction of a tandem FP in a single molecule may result in intramolecular dimerization [21] at the expense of intermolecular dimerization. A new probe, tandem ArcLight, was made with a tandem repeat of super ecliptic pHluorin A227D ( Figure 9A). The linker length and composition were copied from those that are used in tdTomato [22]. The voltage dependent fluorescence response of the tandem ArcLight probe was inverted from a depolarization dependent decrease in fluorescence to an increase and the signal size was greatly reduced to +2% ( Figure 9B). The dynamics of the tandem ArcLight are better fit with single exponential. The time constants (taus) of tandem ArcLight were slower than the slow components of ArcLight ( Figure 9C and 9D).

Molecular biology
Mutations and modification to the original ArcLight probe were introduced by using the QuickChange II XL site-directed mutagenesis kit (Agilent Technologies, INC., CA). All DNA constructs were verified by sequencing using the dyetermination method (W. M. Keck Foundation, Biotechnology Resource Laboratory, Yale University, CT)

Fluorescent protein purification
Expression constructs were generated by inserting PCR amplified fragments of fluorescent proteins' cDNA into the pCR4Blunt TOPO vector (Invitrogen, NY). This procedure introduces a 66His tag to the N-terminus of the fusion proteins to allow affinity purification. Top10 bacteria (Invitrogen, NY) were transformed with the expression constructs and fusion proteins were purified with His-Select Nickel Affinity Gel (Sigma-Aldrich, MO), following the manufacturer's instructions. The purified proteins were concentrated with Amicon Ultra-15 centrifugal filters (MWCO 10,000, Millipore, MA), dialyzed against 100 mM sodium phosphate buffer, pH 7.4 and stored at 4˚C.

Spectrofluorimetry
Fluorescence emission spectra of the fusion proteins were measured with a Horiba Jobin Yvon Fluorolog 3 spectrophotometer (Horiba, Japan). In order to determine pH-dependent fluorescence, purified proteins were diluted to a concentration of 0.36 mM in pH adjusted buffers containing 100 mM NaCl, 1 mM CaCl 2 and 1 mM MgCl 2 . The pH of the buffers was adjusted with MES (for pH 3.5, 4.5 and 5.5), HEPES (for pH 6.5 and 7.5) or Bicine (for pH 8.5 and 9.5) to a final concentration of 25 mM of the buffers. To determine the protein concentration, the protein was denatured in 0.1N of NaOH and absorption at 280 nm was measured using a Beckman Coulter DU 730 uv/vis spectrophotometer (Beckman Coulter, Inc. CA).
Cell culture

Wide field imaging
Whole-cell patch clamped cells were imaged with a Nikon Eclipse E6000FN upright microscope with a water immersion objective, Nikon Fluor 606/1.00 N.A. A MLL-FN-473 nm 50 mW (Changchun New Industries Optoelectronics Tech. Co., Ltd., China) was used as the excitation light source. The laser light was coupled to the microscope by a multi-mode fiber coupler (Siskiyou, OR), a quartz light guide and an Achromatic EPI-Fluorescence Condenser (Till Photonics, NY). The filter cube contains a dichroic mirror 505DCXR and an emission filter HQ510LP (Chroma, Bellows Falls, VT). The fluorescence image was demagnified by an Optem zoom system, A45699 (Qioptiq Inc, NY) and projected onto the 80680 pixel chip of a NeuroCCD-SM camera (RedShirtImaging, LLC) controlled by NeuroPlex software. The images were recorded at a frame rate of 1000 fps.

Data processing
NeuroPlex software was used to view the image sequences and output optical and electrophysiological recordings. The % DF/F was calculated by first subtracting the dark offsets from all frames, then the average of a region of interest in each frame (F) is subtracted from the average of the region taken from one hundred frames prior to the event of interest (F0) and this value is then divided by F0, i.e. % DF/ where a 1 and a 2 are constants, t1 and t2 are time constants in ms, x 0.5 is the membrane potential in mV at half maximal DF/F, and dx is the slope.
The normalized DF/F vs V plot is calculated from the Boltzmann fit:

Statistics
Measurements of each set of data are given as mean ¡ SEM with n equals to the number of independent experiments. For comparison of means between two independent groups, Two-tailed Student t test was performed at significant level (a) of 0.05. Before performing the Student t test, however, the Shapino-Wilk test was performed to determine if the data within each group follow the normal (Gaussian) distribution and equal variance was tested between groups. If variance between two groups is not equal, the Welch's correction was used in the t test. If the data do not follow normal distribution, the Mann-Whitney U nonparametric test was used. For comparison of means between three or more groups, one-way analysis of variance (ANOVA) was performed, followed by Post hoc Tukey-Kramer HSD multiple comparisons. A nonparametric test (Kruskal-Wallis ANOVA) was used when any group of data did not present in a normal distribution. The analyses were performed using the Origin 9.1 software (OriginLab, MA).

Two-photon polarization microscopy (2PPM)
Two-photon polarization microscopy was performed on a laser-scanning microscope iMic (Till Photonics GmbH, Germany), equipped with a Yanus beam scanner (Till Photonics) and a tunable pulsed titanium:sapphire laser (Chameleon Ultra II with GVD compensation, Coherent, CA) operated at 960 nm. The objective was a UApoPlan/IR 660, numerical aperture (NA) 1.2 water-immersion objective lens (Olympus Corp., Japan). We used a long-pass dichroic mirror (Q565LP, Chroma Technology Corp, VT) and an emission filter (520/35, Semrock, NY). Fluorescence was detected by a photomultiplier (R6357, Hamamatsu Photonics, Japan), operated at 700-900 V, providing 16-bit output. A CCD camera Imago QE (Till Photonics) was used for transmitted light imaging.
A rapid polarization modulator RPM-2P (Innovative Bioimaging L.L.C., TX) synchronized with scanning of the microscope was used to alternate the polarization of the excitation beam between horizontal and vertical orientation between acquisition of subsequent pixels. For combined electrophysiology/twophoton polarization microscopy experiments, electrodes were prepared from GC150T-10 borosilicate capillaries (Harvard Apparatus, MA), using a PC-10 puller (Narishige, Japan), and mounted in an MP-225 micromanipulator (Sutter Instrument, CA). Voltage pulses were delivered using the EPC10 amplifier (HEKA Elektronik, Germany) controlled by PatchMaster software (HEKA).
Raw two-photon polarization microscopy images were processed as described previously (Lazar, Bondar et al. 2011) to yield a pair of images acquired with distinct polarizations of the excitation beam. For two-photon polarization microscopy voltage imaging, fluorescence of a manually selected horizontal region (within 5˚of the horizontal direction) of the plasma membrane was used for analysis. After background subtraction, average fluorescence intensities F h and F v , acquired with the horizontal and vertical excitation polarization, respectively, were used to calculate the total fluorescence F (F5F h +F v ) and the dichroic ratio r (r5F h /F v ). At least 10 cells were used for each measurement.

Discussion
We identified four residues, D147, F202, T204 and D227 within the super ecliptic pHluorin, which are critical for the large ArcLight response amplitude [1]. Here we demonstrated that a similarly constructed probe based on eGFP also exhibits large response amplitude if all four critical residues are present ( Figure 1). We found that residue D147 alone is responsible for the base-shifted (relative to eGFP) pH sensitivity of super ecliptic pHluorin (Figure 2), but D147 by itself is not enough to result in a large DF/DV signal when introduced to an eGFP based probe ( Figure 1). These four critical residues are all located in close vicinity and are on one side of the b-barrel surface (Figure 1). This side of the surface is also the dimerization contact surface in avGFP [21]. Further investigation is needed to determine how these mutations coordinate. Intriguingly, the ArcLight probe carrying tandem fluorescent proteins had a greatly reduced response amplitude and a change in response direction (Figure 9). It is possible that the tandem super ecliptic pHluorins in this ArcLight variant form an intramolecular dimer and that this dimerization may inhibit the fluorescence changes of the probe by interfering with intermolecular dimerization or association with the plasma membrane.
Similarly constructed probes carrying pH sensitive fluorescent proteins other than super ecliptic pHluorin A227D, i.e. the YFP and ratiometric pHluorin, did not have large voltage dependent changes in fluorescence (Figure 3). These new probes produced only small response amplitude. Introduction of the A227D mutation only served to further decrease their signal sizes. These results (together with above described result of an eGFP based probe carrying the S147D mutation) show that pH sensitivity of the fluorescent protein alone is not sufficient to produce a large signal. Furthermore, the Ciona protein carrying super ecliptic pHluorin does not exhibit a large response amplitude without introducing the A227D mutation [1]. We also demonstrated that shifting the voltage sensitivity of the Ciona voltage sensor by introducing mutations to the S4 domain also shifts the fluorescence response of ArcLight (Figure 4). The correlation between the voltage sensitivity and the fluorescence changes provides a third line of evidence that ArcLight does not simply sense a pH changes that results from the membrane potential change.
We used two-photon polarization microscopy, an anisotropy method, to examine changes in the FPs orientation with changes in voltage and altered amino acids at the linker domain. Two-photon polarization microscopy allows sensitive observations of changes in orientation of fluorescent moieties with respect to the cell membrane [19]. In two-photon polarization microscopy, molecular orientation can be characterized by the maximum dichroic ratio r max : the ratio of fluorescence intensities (Fh/Fv) observed using two perpendicular polarizations (horizontal and vertical in the frame of the image) of the excitation beam. For a horizontally oriented section of the cell membrane, the r max is largest (theoretically infinite) when the two-photon excitation pseudo-transition dipole moment of the fluorescent moiety is oriented parallel to the plasma membrane, and smallest (theoretically zero) when the pseudo-transition dipole moment is perpendicular to the membrane. We found that the fluorescent protein in all three tested probes undergoes voltage-induced reorientation ( Figure 5). In response to cell membrane depolarization, CiVSD-EP and ArcLight-S249 showed similar, small r max increases. This indicates that these two FPs' pseudo-transition dipole moments rotated slightly towards the plasma membrane. Interestingly, although the two constructs show similar changes in r, ArcLight-S249 exhibits a much larger fluorescence decrease than CiVSD-EP. In contrast to ArcLight-S249 and CiVSD-EP, Arclight-Q239 showed a decrease in r max , indicating that its pseudotransition dipole moment rotated away from the plasma membrane, and the r max change was larger than in CiVSD-EP and ArcLight-S249. Thus, although the three constructs share a decrease in fluorescence intensity upon depolarization, the underlying molecular rearrangements are distinct, implying that the observed changes in fluorescence intensity are not the result of changes in the orientation of an inert b-barrel. However, the changes in r max and in fluorescence intensity closely correlate in temporal dynamics, implying a close mechanistic connection between the FP reorientation and the change in fluorescence intensity.
The response amplitude of ArcLight increases if the fluorescent protein is inserted at positions along the linker closer to the S4 transmembrane domain [1] than in a position more C-terminal. In this study, we scanned the full linker region from A231 to S249 by inserting the super ecliptic pHluorin A227D after each of the linker amino acids. We determined that there is a ''sweet spot'' (Q239 to A242) in the length of this linker outside of which the response amplitude is decreased ( Figure 6). This drop off is precipitous as the linker length is decreased and gradual as it is increased beyond the optimal. Three of these derivatives, I233, F234 and Y235, with the fluorescent protein inserted close to the S4 domain did not traffic well to the plasma membrane. We also found that removing the three extra amino acids, glycine, aspartic acid and proline (GDP) in the linker region of all ArcLight derivatives did not significantly affect the signal size but did slow the response of ArcLight ( Figure 6).
We explored several protein modifications which may change the mobility/ orientation/dimerization of the FP in ArcLight. These included removing nonessential, flexible residues at the N-or C-termini of the super ecliptic pHluorin A227D (Figure 7), adding an anchoring farnesylation site to the C-terminal end of the FP (Figure 8), and using tandem super ecliptic pHluorin A227D (Figure 9). None of these approaches has improved the probes response characteristics. Instead, these modifications tend to decrease the response amplitude and slow down the response kinetics of the probes (Figure 7, 8 and 9). It is not clear why these modification cause deterioration of the ArcLight response properties.
The robust ArcLight fluorescent response may result from the modulatory effect of mutation A227D on the b-barrel structure of the FP, to which deformation may be induced by movement of the S1-S4 domains. The loosening of the b-barrel in turn allows penetration by a H + ion, leading to fluorophore quenching analogous to that observed upon a pH change, and thus to a decrease in fluorescence intensity. However, this hypothesis is not supported by the observation that the fluorescent response is greatly reduced with ArcLight variants featuring obligatory intramolecular dimerization of FPs (tandem FPs, Figure 9) or fixation of the FP position (by farnesylation, Figure 8). Alternatively or concomitantly, ArcLight FPs may undergo a voltage dependent intermolecular dimerization and de-dimerization or association/disassociation with the plasma membrane. The hyperpolarization position of the S4 may allow FPs of two ArcLight subunits to dimerize, producing a fluorescent dimer. If this were true we would expect to see a membrane concentration dependent effect. Upon depolarization and net outward movement of the S4 domain, this dimerization may be disrupted, producing reduced output monomers. In fact, many other FPs are known to show reduced fluorescence output upon monomerization [23]. Alternatively depolarization dependent movements may reversibly alter the FPs association with the plasma membrane.