SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. The voltage-sensitive fluorescent protein (VSFP) toolbox
  4. Expression of VSFPs in neurons
  5. Optimized optical instrumentation
  6. Specific issues related to in vivo imaging
  7. Outlook
  8. References
  9. Appendix

Over the last decade, researchers in our laboratory have engineered and developed several series of genetically encoded voltage-sensitive fluorescent proteins (VSFPs) by molecular fusion of a voltage-sensing domain operand with different fluorescent reporter proteins. These genetically encoded VSFPs have been shown to provide a reliable optical report of membrane potential from targeted neurons and muscle cells in culture or in living animals. However, these various reporters also exhibit discrepancies in both their voltage-sensing and targeting properties that are essentially related to the intrinsic characteristics of the fluorescent reporter proteins. It is therefore important carefully to select the sensor that is most appropriate for the particular question being investigated experimentally. Here we examine the current state of this subfield of optogenetics, address current limitations and challenges, and discuss what is likely to be feasible in the near future.

In recent years, optical methods in physiology have been complemented by an optogenetic toolbox of genetically encoded, targetable proteins that allow the use of light to either control or report molecular processes of specific cell populations within intact neuronal circuits. Optical reporters generally consist of engineered biosensors derived from the fusion of fluorescent proteins (FPs) to detector proteins that convert physiological signals into changes in fluorescence output. Within this class of biosensors, the development of optical probes for membrane potential has proven to be an especially important and challenging goal.

The voltage-sensitive fluorescent protein (VSFP) toolbox

  1. Top of page
  2. Abstract
  3. The voltage-sensitive fluorescent protein (VSFP) toolbox
  4. Expression of VSFPs in neurons
  5. Optimized optical instrumentation
  6. Specific issues related to in vivo imaging
  7. Outlook
  8. References
  9. Appendix

Several designs have been explored to date for the molecular engineering of voltage-sensitive genetically encoded probes (see Knöpfel et al. 2006 and Perron et al. 2009a for more extensive reviews of earlier work). The first prototypes were based on the insertion of reporter FPs within potassium or sodium channels, wherein a voltage-dependent conformational change within the voltage-sensing domain is associated with a modulation in fluorescence (Siegel & Isacoff, 1997; Sakai et al. 2001; Ataka & Pieribone, 2002). Although these first-generation voltage probes were shown to optically report changes in membrane potential, their application in mammalian systems was limited by poor targeting to the plasma membrane in transfected cells (Baker et al. 2008). Researchers in our laboratory subsequently developed a second generation of voltage-sensitive proteins by replacing the actuator in our VSFP1 probe (Sakai et al. 2001) with the voltage-sensing domain of the self-contained non-ion channel protein Ciona intestinalis voltage-sensor-containing phosphatase (Ci-VSP). These Förster resonance energy transfer (FRET)-based VSFP2 sensors displayed significantly improved targeting to the cell surface (Dimitrov et al. 2007) and reliable responsiveness to membrane potential signalling from targeted neurons in culture, acute brain slices and living mice (Akemann et al. 2010). Like most genetically encoded FRET-based biosensors, the first VSFP2 voltage reporter (VSFP2.1) was derived from cyan- and yellow-emitting variants of Aequorea victoria green fluorescent protein (i.e. Cerulean and Citrine). Optimization of the length of the linkers connecting the actuator to the donor chromophore and the FRET donor–acceptor pair in VSFP2.1 yielded a sensor with improved responsiveness that we named VSFP2.3 (Lundby et al. 2008). Since red-shifted fluorescence provides better spectral separation from the intrinsic green autofluorescence of brain tissue and because long-wavelength light is usually associated with higher penetration depth, we have engineered an additional VSFP2 variant comprising a pair of yellow and far-red emitting FPs (Citrine and mKate2) that we termed VSFP2.4 (Mutoh et al. 2009). Another recently reported VSFP2 variant, Mermaid, was constructed from FPs isolated from corals (mUKG and mKO; Tsutsui et al. 2008). Quantitative comparison of these FRET-based voltage probes in the neuron-like rat pheochromocytoma PC12 cell line revealed rather comparable steady-state, spectrally resolved, maximal voltage-dependent changes in fluorescence with relatively similar activation curves (Mutoh et al. 2009). The VSFP2 fluorescence signals could be fitted with fast and slow on-time constants (Table 1) correlating with the biphasic conformational rearrangement of the voltage-sensing domain of Ci-VSP (Villalba-Galea et al. 2008, 2009). The contribution of the fast on-time constant to the total signal was slightly higher for VSFP2.3 and VSFP2.4 compared with Mermaid (Mutoh et al. 2009).

Table 1.  Summary of on-time response properties of VSFP2 spectral variants (fromMutoh et al. 2009)
VSFP2 variantFast τon at −20 mV (ms)V½ of fast τon (mV)Slow τon at −20 mV (ms)Dynamic range (%)
VSFP2.33.3 ± 0.3−49.5 ± 1.155.9 ± 1.113.3
VSFP2.42.5 ± 0.3−54.2 ± 2.057.3 ± 1.012.4
Mermaid2.5 ± 0.3−43.6 ± 1.057.3 ± 1.012.9

During our early analysis, we noted that a version of VSFP2.1 with a shorter linker (VSFP2A in Dimitrov et al. 2007) clearly exhibits a fast on-time component in the cyan response that is not observed in the yellow channel, suggesting the presence of a FRET-independent response component (Lundby et al. 2008). We exploited this fast response component in a series of monochromic voltage sensors, named VSFP3s, wherein the voltage-sensing scaffold is coupled to a single reporter FP rather than a FRET pair. Since VSFP3s contain a single FP, it was straightforward to take advantage of FPs with photophysical properties that have been optimized for live-cell fluorescence imaging and generate VSFP3 variants that span the yellow to far-red region of the visible light spectrum (Perron et al. 2009b). Interestingly, these VSFP3 colour variants exhibited different response dynamics and sensitivities upon functional characterization in PC12 cells (Table 2). The reason for these differences is still unclear, as the mechanisms of fluorescence modulation are only partly understood. Nevertheless, VSFP3.1_mOrange2 represents the first single-FP VSFP probe to be benchmarked for optical recording in cultured hippocampal neurons (Perron et al. 2009b). The most important advantages of these voltage reporter proteins are their fast kinetics and their capacity to enable multicolour imaging. However, it is important to choose the probe that is most appropriate for each experiment. For instance, although VSFP3.1_TagRFP displays relatively slow kinetics, its dynamic range is about two times greater than that of VSFP3.1_mOrange2. VSFP3.1_TagRFP could therefore be a good candidate for measuring synaptic potentials or slower membrane potential changes in muscle cells.

Table 2.  Summary of on-time response properties of VSFP3 spectral variants (fromPerron et al. 2009b)
VSFP3 variantFast τon at −20 mV (ms)V½ of fast τon (mV)Slow τon at –20 mV (ms)Dynamic range (%)Fast τon contribution*
  1. *The contribution of the fast on-time response component (fast τon) to the dynamic range of the fluorescence response is given by the value in parentheses.

Cerulean1.8 ± 0.3−34.7 ± 3.677.5 ± 341.9(69%)
Citrine2.2 ± 0.2−21.2 ± 2.0 81 ± 211.6 (51.9%)
mOrange23.8 ± 0.3−34.4 ± 3.547.3 ± 4.72.9(34%)
TagRFP3.2 ± 0.6−54.3 ± 1.760.9 ± 9.43.5 (17.5%)
mKate2n.a.  n.a.   54.3 ± 15.71.3(<1%)

Expression of VSFPs in neurons

  1. Top of page
  2. Abstract
  3. The voltage-sensitive fluorescent protein (VSFP) toolbox
  4. Expression of VSFPs in neurons
  5. Optimized optical instrumentation
  6. Specific issues related to in vivo imaging
  7. Outlook
  8. References
  9. Appendix

Although short-term expression is usually sufficient for most studies in cultured cells, expression over many days is practically mandatory for in vivo or acute ex vivo experiments, and it is interesting to note that the shift from transient expression in transfected cell lines to long-term expression in neurons has turned out to be more complicated than originally anticipated. We initially encountered issues related to low expression levels or cell death of transfected neurons. As expression levels mostly depend on the strength of the promoter and copy number of the integrated DNA in transfected cells, we first used the strong cytomegalovirus (CMV) promoter for expression in primary hippocampal cultures. However, the CMV promoter proved unsuitable for sustained expression, and significant cell damage was observed as early as 48 h following transfection. At 72 h post-transfection, only weakly expressing cells were found to have survived. After exploring different promoters, we found that neuronal expression of VSFP2s and VSFP3s was best driven by the CMV early enhancer/modified chicken β-actin (CAG) hybrid promoter (Niwa et al. 1991). This CAG-based expression vector also contains a subregion of the bovine papilloma virus genome that is known to allow high levels of constitutive expression of inserted genes.

As reported by Akemann et al. (2010), lipofection of primary hippocampal neurons and in utero electroporation of mouse embryos with plasmids for CAG-driven expression of VSFP2.3 resulted in strongly expressing and yet healthy neurons. More specifically, high expression levels were observed in cultured cells for up to 7–14 days post-transfection, and in acute brain slices collected on postnatal day 20–60 from electroporated mice. Most importantly, expressing neurons were electrophysiologically healthy, as indicated by their ability to generate spontaneous and evoked action potentials comparable to those observed with untransfected cells. Somewhat surprisingly, many other VSFP variants that we tested in the same conditions turned out to be more problematic upon long-term expression in neurons. As illustrated in Fig. 1, although all of the constructs generated fluorescence in cortical pyramidal neurons from postnatal day 19 electroporated mice, we observed noticeable differences in their expression pattern. Most notably, VSFP2.3 showed especially efficient targeting to the plasma membrane, while VSFP2.42 (an optimized VSFP2.4 variant) exhibited a tendency to form intracellular fluorescent aggregates. Mermaid was also targeted in part to the plasma membrane, although membrane-associated fluorescence was overwhelmed by intracellular fluorescent clusters. Similar aggregates were also observed in neurons expressing VSFP3.1_TagRFP and VSFP3.1_mKate2, as shown in Fig. 2. Such punctate structures were previously reported in cell somata and processes in a series of transgenic mouse lines expressing various reef coral FPs during early postnatal weeks, and were shown to increase substantially in number with the age of the animal (Hirrlinger et al. 2005). In contrast, intracellular fluorescent clusters were not detected in mice expressing A. victoria-derived FPs, even at older ages (Nolte et al. 2001). As the effectiveness of membrane potential indicators largely depends on the efficiency of targeting of the fluorescent probes to the plasma membrane, we are currently applying combinational molecular trafficking strategies to increase the sensitivity of our biosensors (Gradinaru et al. 2010). Such modifications include the introduction of Golgi trafficking signals and additional endoplasmic reticulum (ER) export motifs to favour transport along the secretory pathway to the cell surface. Since anthozoan-derived fluorescent proteins (e.g. mKate2) have a tendency to form dimers that are likely to contribute to aggregation/misfolding and retention in the ER, we also intend to improve the monomerizing properties of our mKate-based voltage sensor (VSFP2.42) by replacing key residues within hydrophobic patches on the surface of the fluorescent protein via a site-directed evolutionary approach.

image

Figure 1. Expression of Förster resonance energy transfer-based VSFP2s in adult mouse cortex The first two columns show macroscopic fluorescence images of VSFP2.3, VSFP2.42 and Mermaid in brains from postnatal day 22–24 electroporated mice. Fluorescence from donor (left) and acceptor chromophores (right) is shown. The third column shows whole-field fluorescence images of VSFP2.3, VSFP2.42 and Mermaid in acute brain slices. The fourth column shows confocal images of a VSFP2-expressing cell in layer 2/3. The fifth column shows confocal images of a VSFP2-expressing cell body at higher magnification. Scale bars represent (from left to right): 2 mm, 2 mm, 100 μm, 10 μm and 10 μm.

Download figure to PowerPoint

image

Figure 2. Expression of single-fluorescent protein-based VSFP3s in adult mouse cortex The first two columns show macroscopic fluorescence images of VSFP3.1_Cerulean, VSFP3.1_TagRFP and VSFP3.1_mKate2 in brains from postnatal day 19–30 electroporated mice. The third column shows whole-field fluorescence images of VSFP3.1_Cerulean, VSFP3.1_TagRFP and VSFP3.1_mKate2 in acute brain slices. The fourth column shows confocal images of a VSFP3-expressing cell in layer 2/3. The fifth column shows confocal images of a VSFP3-expressing cell body at higher magnification. Scale bars represent (from left to right): 2 mm, 2 mm, 100 μm, 10 μm and 10 μm.

Download figure to PowerPoint

Optimized optical instrumentation

  1. Top of page
  2. Abstract
  3. The voltage-sensitive fluorescent protein (VSFP) toolbox
  4. Expression of VSFPs in neurons
  5. Optimized optical instrumentation
  6. Specific issues related to in vivo imaging
  7. Outlook
  8. References
  9. Appendix

The size of the optical signals obtained with VSFPs is proportional to the sensitivity of the probe. Comparing sensitivities between different probes and different experimental systems can lead to confusion and even controversy (see Tsutsui et al. 2008versusMutoh et al. 2009). It is therefore important to understand the difference between the biophysical sensitivity of a voltage probe and the apparent sensitivity observed when the probe is applied in the biological system of choice. The biophysical sensitivity is measured in close-to-ideal conditions, where fluorescence is sampled exclusively from a lipid membrane that is under voltage-clamp control. In such conditions, the sensitivity of advanced VSFPs compares to that of the best classical voltage-sensitive dyes, with changes in fluorescence exceeding 10% of baseline fluorescence (ΔF/F) per 100 mV change in membrane potential (Tsutsui et al. 2008; Mutoh et al. 2009). For ratiometric measurements using VSFP2s, the change in fluorescence ratio (ΔR/R) is roughly the sum of the ΔF/F values for each colour channel. The apparent sensitivity in specific biological experiments depends on the fraction of the photons that is emitted from probe molecules associated with the membrane of interest relative to the total number of collected photons. Typical sources for large amounts of additional photons are the membranes of other stained cells, fluorescence from probe molecules that are not associated with the plasma membrane and tissue autofluorescence. The apparent sensitivity equals the biophysical sensitivity multiplied by this selectivity-of-collection factor (Knöpfel et al. 2006). For instance, in intact brain tissue the selectivity-of-collection factor for VSFP2.3 in typical experiments is between 0.3 (when sampling from single cells in brain slices) and 0.03 (for macroscopic imaging in vivo); exact values depend on the fraction of labelled cells within the sampled volume that are recruited during a specific response. In addition, the magnitude of the physiological signal scales with the size of the voltage change. Based on these various factors, it is important to realize that in conditions typical for experiments in a physiology laboratory, a biophysical sensitivity value of 10% per 100 mV scales down to physiologically relevant VSFP signals that are at least one or two orders of magnitude smaller (for brain slices and in vivo experiments, respectively; Akemann et al. 2010).

A good probe should have a large ΔF/F value, but the more practically relevant signal scale relates to the level of noise associated with the signal. The minimal achievable noise in optical recordings is determined by the stochastic nature of photon emission (following a Poisson distribution, where noise =inline imageumber of photons). Thus, in order to achieve a signal-to-noise ratio (SNR) greater than 1 for a signal of 0.1%ΔF/F, the noise amplitude needs to be minimized to 0.1% of baseline fluorescence, which in turn requires the detection of n= 106 photons per sampling time interval.

Minimal frame rates of 100–1000 Hz are usually required to resolve the time course of physiologically interesting voltage transients. This requirement, along with the SNR considerations described above, constrains the best choice of the light detector. A useful charge-coupled device (CCD) or complementary metal-oxide-semiconductor (CMOS) detector must not only allow for fast imaging but also needs to be able to sample large numbers of photons (specified as ‘well depth’). In the imaging conditions discussed here, the charge read-out noise in the detector, although important, is a frequently overrated specification parameter. Read-out noise increases with the square root of the read-out speed, and can be greater than 100 electrons per pixel in high-speed CCDs (Pawley, 2006). It is important, however, to realize from the above calculations that detector noise is typically overwhelmed by the shot noise associated with even the minimal number of detected photons required, for instance, to resolve 0.1%ΔF/F signals. If a sufficient signal-to-noise ratio can only be achieved by massive sweep averaging, detector noise can become limiting; in such cases, on-chip multiplication of photoelectrons by electron-multiplier-type CCDs (EM-CCDs) can significantly improve the SNR of the averaged image sequence. However, because of the noise generated by the electron-multiplier process, the effective quantum yield for photoelectrons (>80% in back-illuminated CCDs) is reduced with high electron-multiplier gain (Pawley, 2006), and this trade-off should be accepted only if photon noise is indeed below read-out noise; that is, if signals cannot be resolved in single sweeps after optimizing the amount of photons that can be detected.

The need to collect large numbers of photons from membranes of interest establishes a scenario in which both optical components and experimental conditions have to be optimized. This entails high levels of probe expression (taking into account that probes tend to lose membrane targeting specificity when massively overexpressed), sufficient output power of the excitation light source (taking into account the maximally tolerable rate of probe photo-bleaching), minimization of fluorescence background, optimal efficiency of the optical pathway for the collection of emitted photons, and high quantum yield for the photodetector (for details, see Akemann et al. 2009). Another important but frequently underrated consideration is the selection of high-quality optical filters and beam splitters, which need to be carefully optimized for the spectral response properties of the particular VSFP probe being used (Mutoh et al. 2009).

Specific issues related to in vivo imaging

  1. Top of page
  2. Abstract
  3. The voltage-sensitive fluorescent protein (VSFP) toolbox
  4. Expression of VSFPs in neurons
  5. Optimized optical instrumentation
  6. Specific issues related to in vivo imaging
  7. Outlook
  8. References
  9. Appendix

At present, in vivo VSFP imaging has only been established for superficial structures that can be accessed without major structural damage to the brain by using conventional epifluorescence optics. The imaging of deeper brain structures adds new technical challenges that will need to be addressed in the future.

In in vivo epifluorescence imaging conditions, SNRs continue to be constrained by photon quantum statistics, but additional noise sources may dominate. Particularly troublesome are the mechanical noise caused by heart beat and breathing, haemodynamic volume fluctuations and changes in haemoglobin oxygenation. Conventional in vivo methods, such as organic dye imaging and electrophysiological techniques, require craniotomies in order to stain and record from neurons. However, craniotomy decreases optical imaging sensitivity by amplifying mechanical noise and creates the risk of generating an inflammatory response during surgery, particularly in small animals and during long-term recordings. Optogenetic VSFP probes can alternatively be imaged through a thinned but otherwise intact skull, which offers a considerable technical advantage for strictly non-invasive imaging of cortical electrical brain activity without craniotomy. Maintaining the intracranial pressure stabilizes the brain and helps considerably in minimizing motion-induced imaging artifacts. In parallel, heart beat modulation of probe fluorescence is effectively eliminated with FRET-based VSFP2 probes, as the ratio of donor and acceptor fluorescence excludes voltage-independent intensity fluctuations of the same polarity. Thus, by removing correlated noise, ratiometric VSFP imaging yields high-sensitivity and low-noise recordings of in vivo cortical electrical activity. Indeed, by using the FRET-based probes VSFP2.3 and VSFP2.42, we have reliably observed cortical responses to natural sensory stimulation with single trial resolution in adult mice that were electroporated in utero (Akemann et al. 2010). An alternative strategy to entirely eliminate haemodynamic and heart beat-related signals in in vivo VSFP recordings builds on VSFP variants derived from far-red-shifted FPs that lack spectral overlap with haemoglobin oxygenation states. These variants will further expand the toolbox of VSFP probes designed for in vivo experimentation.

Outlook

  1. Top of page
  2. Abstract
  3. The voltage-sensitive fluorescent protein (VSFP) toolbox
  4. Expression of VSFPs in neurons
  5. Optimized optical instrumentation
  6. Specific issues related to in vivo imaging
  7. Outlook
  8. References
  9. Appendix

Work over the last 10 years has yielded an array of established optogenetic probes for non-invasive optical recording of electrical activity from genetically defined neurons both in vitro and in vivo. We have now reached a state where we can begin using this methodology in experiments designed to address physiological questions that were previously intractable using classical electrophysiological methods. Genetically encoded VSFPs benefit from the wide range of genetic techniques that are available for the selective targeting of specific cell types and locations in the nervous system. This includes cell-specific, activity-dependent and inducible promoters, conditional expression via the Cre/lox system and methods for gene transfer using viral carriers, trans-synaptic tracing carriers and transgenic animals (Wickersham et al. 2007; Boldogkoi et al. 2009). This established optogenetic toolbox will bring us and other researchers closer to the goal of being able to optically interrogate specific subsets of neurons regarding their participation in circuit operations during defined behaviours. Even after seeing a tremendous leap in benchmark performance of VSFP probes during the last few years, there remains ample room for further development, and perhaps the greatest general interest lies in the derivation of new colour variants for the simultaneous probing of different neuronal populations without probe cross-talk, as well as the pairing of such reporters with optogenetic tools that enable the stimulation or silencing of specific components of neuronal networks.

References

  1. Top of page
  2. Abstract
  3. The voltage-sensitive fluorescent protein (VSFP) toolbox
  4. Expression of VSFPs in neurons
  5. Optimized optical instrumentation
  6. Specific issues related to in vivo imaging
  7. Outlook
  8. References
  9. Appendix

Appendix

  1. Top of page
  2. Abstract
  3. The voltage-sensitive fluorescent protein (VSFP) toolbox
  4. Expression of VSFPs in neurons
  5. Optimized optical instrumentation
  6. Specific issues related to in vivo imaging
  7. Outlook
  8. References
  9. Appendix

Acknowledgements

We thank all members of the Knöpfel laboratory for discussions and support. We would specifically like to thank Reiko Yoshida, Nisha Jose and Jenny Koivumaa for in utero electroporation of VSFP plasmids. Work reviewed here was funded by grants from RIKEN BSI (T.K.), the RIKEN BSI director's fund (T.K.), a grant from the Ministry of Education, Culture, Sports, Science and Technology (MEXT; H.M.), the JSPS (Japanese Society for the Promotion of Science)–CIHR (Canadian Institutes of Health Research) postdoctoral fellowship programme (A.P.) and a grant-in-aid for JSPS fellows (A.P.).