Linker length and fusion site composition improve the optical signal of genetically encoded fluorescent voltage sensors

Abstract. Several genetically encoded fluorescent sensors of voltage were created by systematically truncating the length of the linker sequence between the voltage-sensing domain and the position of the fluorescent protein, Super Ecliptic A227D. In addition to varying the length, the amino acid composition at the fusion site for the fluorescent protein was modified. Both linker length and amino acid composition affected the size and voltage sensitivity of the optical signal. The truncation mutants revealed a potential structural periodicity with a maximum signal three amino acids from the voltage-sensing domain and another maximum 11 amino acids from the voltage-sensing domain. These results confirm that the linker length and composition can fine tune the size and voltage range of the sensor. The potential periodicity suggests that the orientation of the fluorescent protein could be important for improving the signal size implicating dimerization of the fluorescent protein.


Introduction
In order to faithfully represent neuronal activity, a genetically encoded fluorescent voltage sensor (GeFVS) must respond to changes in membrane potential with a rapid and large fluorescent change. A rapid fluorescent change is required since action potentials only last 1 to 2 ms. A large fluorescent change is needed since the total amount of responsive probe is limited to the plasma membrane.

ArcLight Gives a Large Voltage-Dependent
Optical Signal, while Zahra Gives a Rapid Voltage-Dependent Optical Signal The brightest GeFVS consists of a transmembrane, voltage-sensing domain (VSD) coupled to a cytoplasmic, fluorescent protein (FP) via a linker sequence. Improvements in the speed and size of the optical response to membrane depolarizations have been achieved by modifying different regions of the sensor. ArcLight, a GeFVS with one of the largest optical signals reported, yields a 30 to 40% ΔF∕F∕100 mV depolarization of the plasma membrane. 1,2 This large signal is due to the combination of a mutation to the FP, Super Ecliptic pHlorin A227D (SE 227D) and the optimization of the linker length between the VSD and the FP. While ArcLight gives a large optical signal, it is also slow with taus >20 ms. ArcLight uses the VSD from the Ciona voltage-sensing phosphatase (VSP) gene. Some of the fastest GeFVS published utilize the VSD of the zebrafish VSP (Ref. 3) or the VSD from the chicken VSP. 1,4 Zahra 1 and Zahra 2 are GeFVS containing the cerulean/citrine FP pair that optically reports changes in membrane potential via Förster resonance energy transfer (FRET) with extremely rapid kinetics comparable to voltage-sensing dyes but they have small signal sizes. 3 Both Zahra sensors utilize the VSD of the zebrafish VSP gene. In an effort to improve the kinetics of ArcLight, the Ciona VSD was replaced with the VSD from zebrafish or chicken. 1 These ArcLight derivatives exhibited improved kinetics but weaker signals. The FP fusion site for the zebrafish ArcLight was nearly 20 amino acids upstream of the fusion site for the Zahra constructs. To test the effect of linker length on the voltage-dependent optical signal, the FRET pair in the Zahra constructs was replaced with SE 227D. In this report, we demonstrate that the length of the linker alters the voltage sensitivity of the optical response, resulting in a slight increase in the size of the fluorescent change. A structural periodicity of eight amino acids was also observed in the linker sequence. The fluorescent signal increased when the FP was inserted three amino acids downstream of the VSD and again increased when the FP was inserted 11 amino acids downstream of the VSD. In addition, the amino acid composition at the fusion site of the FP also affected the fluorescent response. These results suggest that the orientation of the FP is important for the voltage-dependent optical signal.

Probe Design
The Zahra 1 SE (227D) and the Zahra 2 SE (227D) were generated by replacing the FRET FPs with SE 227D. The *Address all correspondence to: Bradley J. Baker, E-mail: bradley.baker19@ gmail.com polymerase chain reaction (PCR) was used to introduce an upstream Eco RV site and a downstream Xho I site for the introduction of SE 227D, resulting in the amino acid composition of RYR at the fusion site of the FP. Two-step PCR was used to generate the RYR truncated constructs, A1 through A7. A single-step PCR process was used to introduce a Bam H1 site and systematically increase the linker length between the S4 of the VSD and the FP, resulting in the JE series of constructs that contain the GDP amino acid sequence at the FP fusion site. For altering the amino acid composition at the FP fusion site, a two-step PCR process was employed, enabling the site-directed mutagenesis of the probe. All constructs were verified by DNA sequencing (Cosmogenetech, Republic of Korea).

Arc Lamp and Wide-Field Imaging
During the patch-clamp experiments, the cells were imaged on an IX71 microscope with a 60 × 1.35 numerical aperture oil-immersion lens (Olympus, Japan). Fluorescence excitation was delivered using a 75 W xenon arc lamp (Cairn, England). All constructs utilize the same FP identified in ArcLight, SE (227D). 2 The filter set used was that reported for ArcLight. The excitation filter was FF02-472/30 (Semrock, USA). The emission filter was FF01-496/LP (Semrock, USA). The dichroic was FF495-Di03 (Semrock, USA). The objective C-mount image was demagnified by an Optem zoom system A45699 (Qioptiq LINOS, USA) and projected onto the e2v CCD39 chip of NeuroCCD-SM 80 pixel × 80 pixel camera (RedShirtImaging, LLC, USA). The imaging apparatus was mounted on a Vibraplane Bench Top vibration isolation platform (Kinetic System, USA; Minus K Technology, USA). The mechanical shutter in the incident light patch was mounted on a separate table. Images were recorded at a frame rate of 1 kfps.

Laser Illumination
For neuronal recordings, excitation light was from a 473 nm laser (MLL-FN-473-50 mW; Changchun New Industries Optoelectronics Tech. Co., Ltd., China). The laser light was guided into the microscope by a quartz light guide (Till Photonics, USA) and an achromatic epifluorescence condenser (Till Photonics, USA). The dichroic, emission filter, and image projection are as described above. The frame rate was either 1 or 2 kfps as described in Fig. 6.

Optical-Signal Analysis
Optical-signal recordings were analyzed using Neuroplex (RedshirtImaging, USA) Excel (Microsoft, USA) and Origin 8.6 (OriginLab, USA). The fluorescent traces for constructs expressed in HEK cells were the averages of 24 trials. The fluorescent traces for constructs expressed in neuronal cells were from a single trial. The fluorescent traces are the spatial average of the output of all the pixels receiving light from the cell. For neuronal cells, pixels corresponding to the soma were used to limit the variability of cell size. HEK cells ranged from 30 to 50 μm in length, while neuronal somas ranged from 10 to 15 μm in length. The offline low-pass temporal filtering used to improve the signal-to-noise ratio is indicated in the figure legends. For the kinetics, the optical traces were fitted to a double exponential decay, where t is time in milliseconds, and a single exponential decay, To compare the optical responses that were better fitted to a single exponential decay to those better fitted to a double exponential decay, a weighted tau was calculated as the sum of τ 1 multiplied by the relative amplitude, A1, plus τ 2 multiplied by the relative amplitude, A2, as defined by the following equation: The voltage sensitivity was determined by initially fitting individual cell responses to the Boltzmann equation: where y is −ΔF∕F, and x is membrane potential in mV. A1 is the minimum value, and A2 is the maximum value. x 0 is the membrane potential in mV at half maximal ΔF∕F, and dx is the slope at x 0 . All traces were then normalized such that A1 ¼ 0 and A2 ¼ 1. The trials for each construct were then averaged and refitted. The range of the fluorescent voltage response listed in Fig. 5 was calculated by extrapolating the V 1∕2 value to the normalized maximum and minimum fluorescent change using the slope of the Boltzmann fit at V 1∕2 , where 3 Results

Replacing the Fluorescent FRET Pairs in Zahra 1 and Zahra 2 with the Fluorescent Protein Super Ecliptic A227D Resulted in a Larger Optical Signal with Slower Kinetics
The position of the FP in Zahra was determined by the sequence conservation at the start of the phosphatase domain. Figure 1 shows the alignment of VSP homologs from several different species. Close inspection of this alignment reveals a rather high degree of conservation that is obscured by the Ciona sequence.
Since conservation suggests function and this region is the start of the phosphatase domain, the FP was inserted at position Y194 (numbering based on zebrafish amino acid position). The Y194 position showed the least amount of conservation among VSP homologs and may, therefore, tolerate the FP better than other positions in the linker. These new constructs are referred to as Zahra 1 SE (227D) and Zahra 2 SE (227D). For comparison purposes, a GeFVS utilizing the wild-type zebrafish VSD sequence was also tested. The zebrafish SE (227D) only gives an optical response with depolarization steps of over 150 mV from a holding potential of −70 mV (Fig. 1). Zahra 1 SE (227D) and Zahra 2 SE (227D) give a 3 to 5% ΔF∕F∕100 mV (Fig. 1). Zahra 1 SE (227D) contains a single mutation in the S4 transmembrane domain (R153Q). Zahra 2 SE (227D) has two mutations to the S4 transmembrane domain, the R153Q also present in Zahra 1 and the I165R mutation, which converts the zebrafish sequence in S4 to that of Ciona (Fig. 1). The voltage range of the optical response is quite large for both Zahra 1 SE (227D) and Zahra 2 SE (227D) spanning at least 300 mV from −170 mV to þ130 mV. While having the deleterious effect of diluting the optical signal over a large voltage range, these mutations enable the probes to respond to more physiologically relevant potentials. The signal sizes for Zahra 1 SE (227D) and Zahra 2 SE (227D) are 6-to 10-fold better than their FRET counterparts, which exhibit a 0.5% ΔF∕F∕100 mV depolarization. 3 The speed of the optical signal is slightly slower than the original Zahra constructs with the τ on around 6 ms and the τ off closer to 7 ms.

Truncation of the Linker Length Between the S4
Transmembrane Domain and the Fluorescent Protein Improved the Size of the Optical Signal for Zahra 1 SE (227D) The zebrafish ArcLight derivative gives an 8% ΔF∕F∕ 100 mV, 1 which is almost double the size of the Zahra 1 SE (227D) and Zahra 2 SE (227D) constructs. Since the linker length of the zebrafish ArcLight derivative is shorter than Zahra 1 SE (227D), a series of truncations was performed on Zahra 1 SE (227D) to determine the effect of the linker length on the voltage-dependent optical signal. Seven new constructs designated A1 to A7 are shown in Fig. 2(a). The RYR sequence preceding the FP sequence is due to an Eco RV site used for cloning. As can be seen in Fig. 2(b), the general trend was an increase in the optical signal as the linker length became shorter. This was in part due to the V 1∕2 being shifted to more negative potentials. Another contributing factor for the increase in the signal size was the increase in the slope of the voltage sensitivity for the optical signal. Instead of the optical response ranging over 300 mV, several of these truncated probes exhibited ranges of under 200 mV [ Fig. 2(c)]. Truncation of the linker did not significantly affect the speed of the response with the on and off taus ranging from 5 to 7 ms [ Fig. 2(d)].

Amino Acid Composition at the Fluorescent Protein Fusion Site Affects the Size of the Optical Signal
Another difference between the zebrafish ArcLight derivative and Zahra 1 SE (227D) is the amino acid composition at the FP fusion site. This difference is due to the restriction enzyme used to fuse the FP onto the VSD. ArcLight was constructed using a Bam H1 restriction site, which results in a GDP preceding the methionine of the FP. 1,2 The Zahra constructs used an Eco RV site to preserve the RYR sequence at the homologous position in the Ciona linker 3 (see Fig. 1). To test whether the amino acid composition at the fusion site affected the optical signal, 22 novel constructs were created, which systematically truncated the linker by a single amino acid. In these linker truncations designated JE01 through JE22, a GDP preceded the FP [ Fig. 3(a)]. Note that JE22 contains the same linker sequence as Zahra 1 SE (227D) except that the RYR sequence has been replaced with the GDP sequence at the fusion site. JE05 is equivalent to the zebrafish ArcLight derivative. 1 Many of the truncations (but not all) give a stronger optical signal. Once again, this is partly due to a shift in the voltage sensitivity of the construct [ Fig. 3(c)]. Plotting the signal strength for a 100 mV depolarization step versus linker length revealed a repeating unit consisting of eight amino acids [ Fig. 3(d)]. JE03 gave the largest signal of ∼12% ΔF∕F∕100 mV. Subsequent increases in the linker length resulted in a steady decline of the optical signal to 6% ΔF∕ F∕100 mV for JE10. JE11 jumps to 9% and subsequent amino acid additions again result in a steady decline of the optical signal. The effect of the truncations on speed was marginal with all truncations ranging from 5 to 8 ms [ Fig. 3(e)].

Amino Acid Composition at the FP Fusion Site
Affects the Voltage-Dependent Optical Signal While shortening the linker length improved the signal size, the restriction enzyme used to fuse the FP onto the linker also contributed to differences in the optical signal. Figure 4 compares five linker lengths that differ only at the fusion site. The overlapping traces are shown with the standard error of the mean in faded color. For the longer linker lengths tested, the GDP amino acid sequence at the fusion site yielded a larger signal than the RYR amino acid sequence, while the optical signal for RYR construct plateaued at lower voltage steps. For the shortest linker length comparing the GDP to RYR fusion site, the RYR construct plateaued at a lower voltage step and yielded a slightly stronger optical signal for the 50 mV depolarization step. Comparison of the zebrafish ArcLight derivative (JE05 -GDP fusion sequence) to the RYR fusion sequence did not exhibit a difference.
To determine if a single amino acid in the fusion sequence was responsible for the different voltage-dependent optical signals, 10 new constructs that varied the fusion site amino acid sequence were tested. Since the RYR versus GDP differences were slightly more pronounced in the longer linker constructs, the linker length of the JE18 construct was used to compare the effects of modifying the three amino acids at the FP fusion site (Fig. 5). Mutating the first position of the fusion site from GDP to RDP shifted the V 1∕2 from 27 AE 5 to 3 AE 1 mV. Mutating the middle position from GDP to GYP also shifted the V 1∕2 to 6 AE 3 mV. Changing the final position of GDP to GDR resulted in less of a shift to 19 AE 1 mV. Converting RYR to the GYR fusion sequence had the most dramatic effect on signal size and speed. The GYR fusion sequence gave the largest optical signal for a 100 mV depolarization (13 AE 1% ΔF∕F). This GYR construct was also the slowest. This was due to the presence of a slow component not seen in the other constructs. The GYR fusion sequence was the only construct that exhibited kinetics better fit by a double exponential. The fast component of GYR was 5 AE 1 ms accounting for 51 AE 2% of the total optical amplitude. The slow component for the GYR fusion sequence was 49 AE 2 ms.

Truncated Constructs Are Able to Optically Resolve Action Potentials in Hippocampal Neurons
Truncated constructs were transfected into dissociated hippocampal neurons that were subjected to current clamp to generate action potentials. Figure 6 shows a comparison of zebrafish ArcLight to that of A6 and A7. A 50 mW laser was used to image action potentials as the 75 W xenon arc lamp was unable to optically resolve action potentials from subthreshold depolarizations (data not shown). All three probes are capable of giving rapid optical responses to a train of action potentials in a single trial. Given the speed of the optical signal, the sampling frequency was doubled to 2000 fps. Even though the higher sampling rate decreases the signal-to-noise ratio, A6 is still able to optically resolve action potentials in an unfiltered trace.

Honoring Our Mentor
In 2003, one of the authors (B.J.B.) joined Larry Cohen's lab to generate a transgenic mouse expressing a GeFVS. The goal was to measure the output of mitral cells in the olfactory bulb. That project proved to be about a decade too early as those probes failed to work in mammalian cells. 6 Rather than being discouraged, Larry gathered the laboratories competing to develop these probes and convinced them to work together in order to avoid repeating the same failures over and over. One of those labs solved the trafficking problem by utilizing the newly discovered VSD from the Ciona VSP. 7  In this issue, to honor our mentor and friend, we present our efforts to understand how the linker sequence affects the optical signal of the fast zebrafish probes first developed in the Cohen lab. 3 By replacing the fluorescent proteins in the Zahra probes with SE 227D, a better signal-to-noise was achieved with a slightly slower optical response. Since ArcLight was improved by varying the linker length, we employed the same strategy. None of the truncations vastly improved the speed of the response. However, the size of the optical signal and the voltage range of the fluorescent response were affected by varying the linker length and the amino acid composition of the FP fusion site.

R153Q Mutation in the Zebrafish S4 Domain
Extends the Range of the Voltage-Dependent Optical Signal to 300 mV The Zahra 1 SE (227D) and Zahra 2 SE (227D) probes exhibit a gradual decrease in fluorescence as the membrane potential is varied from −70 to þ130 mV. The wild-type zebrafish VSD fused to SE (227D) results in a probe that requires a strong depolarization of 150 mV to illicit an optical signal but has a much steeper voltage sensitivity than the Zahra probes [ Fig. 1(c)]. The R153Q mutation in S4, therefore, is responsible for dramatically reducing the slope of the voltage sensitivity and expands the range of the fluorescence response to over 300 mV for the Zahra probes. The R153Q mutation exhibited a similar effect on the enzymatic activity of the zebrafish VSP. 8 Unfortunately, this also dilutes the signal since the maximal fluorescent signal has not changed.  [ Fig. 2(c)]. Moving the FP closer to the VSD shifted the V 1∕2 to more negative potentials. Reducing the linker length also steepened the slope of the voltage response.

Orientation of the FP Potentially Affects the Voltage-Dependent Fluorescent Signal
Systematically truncating the linker length revealed a potential structural component [ Fig. 3(d)] in the linker sequence. The fluorescent signal of the truncated probes exhibited a repeated pattern as the linker length was extended from the S4 of the VSD. JE03 is three amino acids downstream of the VSD and gave the largest fluorescent signal for a 100 mV depolarization step. As the linker length was extended one amino acid at a time, the fluorescent signal dissipated from ∼12 to 6% ΔF∕F∕ 100 mV for JE10, which links the FP 10 amino acids downstream of the VSD. JE11 exhibits a slight jump in the fluorescent from 6 to 9%. Subsequent amino acid additions again result in the dissipation of the fluorescent signal. This periodicity and the effects of the amino acid composition at the FP fusion site suggest that the orientation of the FP is important for the voltage-dependent optical signal.

Dimerization of the FP May Contribute to the Voltage-Dependent Fluorescent Change of GeFVS
The mechanism of how the movement of S4 causes the fluorescence change is not well understood. If there is a structural component in the linker affecting the orientation of the FP, then that would implicate a quenching partner. Two potential partners stand out, the plasma membrane or the intermolecular dimerization of the FP to another sensor molecule. It is not easy to understand how the interaction with the plasma membrane would result in the quenching of the fluorescent signal. However, the dimerization of the FP could be the mechanism since a vast majority of the mutations that created the pH-sensitive Super Ecliptic pHlorin reside at the GFP dimerization interface. 9 The A227D in Super Ecliptic, which vastly enhances the voltage-dependent optical signal of ArcLight, is also close to the dimerization interface. 2

Faster Speed of the Zebrafish VSD-Based Sensors Enables the Resolution of Action Potentials Firing at Higher Frequencies
The zebrafish ArcLight and two truncated versions of Zahra 1 SE (227D) were able to optically resolve action potentials from hippocampal neurons. The signal size and kinetics for all three sensors were very similar (Fig. 6). To better resolve an action potential optically, the fluorescent sampling needs to be in the submillisecond range. Figure 6 compares the optical trace from a neuron sampled at 1000 fps to a neuron sampled at 2000 fps.
By reducing the exposure time, the signal-to-noise ratio has been decreased. However, the probe was still capable of resolving single action potentials without the need for offline filtering. Of course, these probes need to have a faster and larger optical response to membrane potential changes. At least significant improvements in the size of the optical signal have been achieved over early optical recordings. 10,11 So we will continue to employ the lessons Larry has taught and continues to teach us to make better probes for imaging the voltage changes that occur during neuronal activity.