Spatial and Temporal Considerations of Optogenetic Tools in an All-Optical Single-Beam Experiment

Abstract

All-optical experiments promise neuroscientists an unprecedented possibility to manipulate and measure neuronal circuits with single-cell resolution. They rely on highly fine-tuned microscopes with complex optical designs. Of similar importance are genetically encoded optical actuators and indicators that also have to be optimized for such experiments. A particular challenge in these experiments is the detection of natural firing patterns via genetically encoded indicators while avoiding optical cross-activation of neurons that are photon-sensitized to allow optical replay of these patterns. Most optogenetic tools are sensitive in a broad spectral range within the visible spectrum, which impedes artifact-free read-and-write access to neuronal circuits. Nonetheless, carefully matching biophysical properties of actuators and indicators can permit unambiguous excitation with a single wavelength in a so-called single-beam all-optical experiment.

In this chapter, we evaluate the current understanding of these biological probes and describe the possibilities and limitations of those tools in the context of the all-optical single-beam experiment. Furthermore, we review new insights into the photophysical properties of actuators, and propose a new strategy for a single-beam two-photon excitation experiment to monitor activity minimizing cross-activation with the actuators. Finally, we will highlight aspects for future developments of these tools.

1 Introduction

Over the last few years, the development of advanced optical stimulation technologies in combination with novel optogenetic probes foretold a bright future for the interrogation of brain function in behaving animals. In particular, optically imprinting naturalistic firing patterns onto neuronal circuits while simultaneously obtaining activity readouts at a single-cell level is considered the holy grail of a holistic understanding of neuronal circuits [1,2,3]. In this chapter, we refer to such an experiment as an all-optical experiment and offer a review on the possibilities and pitfalls of two-photon stimulation of photon-sensitized single neurons while optically monitoring activity in the entire neuronal circuit. These days such experiments have mostly been performed in specialized laboratories with a long-evolved expertise in advanced two-photon microscopy or optogenetic tool development groups. They are, however, more difficult to implement in common neuroscience laboratories. To allow the execution of these all-optical experiments, both optical technologies and optogenetics probes need improvements in terms of allowing high signal-to-noise ratio (SNR) imaging and obtaining decreasingly flawed readouts. Furthermore, obstacles for most laboratories include the enormous costs of optical systems including two beam lines and optical modulators. Nevertheless, the overall idea for such an approach is straightforward: optical stimulation of single neurons expressing a sensitive blue-absorbing actuator in combination with a highly sensitive red- or IR-light absorbing indicator to report activity of single neurons in a large field of view (FOV) [4,5,6].

For a more extensive introduction into the optical design of these kinds of systems, several implementations have been employed and are discussed within this book (Chaps. 1, 2, 3, 4, and 5) or reviewed elsewhere [7]. Briefly, kilohertz resonance mirrors are commonly implemented for fast imaging, resulting in line scan speeds between 50 and 100 μs. Alternatively, acousto-optic deflectors (AODs) are of an order of magnitude faster and can yield line scan speeds of 10 μs and lower. With electrically tunable or ultrasound lenses and piezo- and acousto-optic elements scanning speeds can be drastically increased and 3D imaging and random-access scanning can be realized [8,9,10]. Furthermore, liquid-crystal-based spatial light modulators (SLMs) are used to extend imaging into the third dimension via computer-generated holography (CGH) [11].

CGH with liquid crystal SLMs is most often used to optically stimulate dozens of target cells in a three-dimensional brain volume on a single-cell level [5, 12,13,14]. Therefore, two strategies for single-cell optical stimulation have emerged in the last decade. The first strategy involves spiraling of the two-photon laser beam on the somata of neurons expressing an actuator. An increasing number of actuators are then activated along the spiral beam path on the surface of the membrane, and their single-channel photocurrents are integrated toward the depolarization of the entire cell, eventually leading to spiking of the neuron [15]. By combining SLM and galvanometric mirrors, such a spiral approach can easily be multiplexed [2]. Yet, latency and spike jitter are rather ill-defined and can vary around 20 ms and from 5 to 20 ms, respectively due to biological factors such as cell morphology, cell excitability, or expression level. Additionally, the temporal spike accuracy also depends on an exact technical implementation in terms of spiraling speed, determination of light powers, and axial resolution for the excitation of the actuator. Furthermore, how efficiently the photocurrent of a single-channel translates into neuronal depolarization depends on the off-kinetics of the actuator in use: slow-closing actuators integrate photocurrent over a longer time and therefore drive spiking more reliably in neurons. Nonetheless, slow-closing actuators increase the probability of multiple spikes during a single spiral scan.

The second approach for single-cell stimulation is the so-called scanless stimulation approach, which can be implemented using holographic stimulations [16, 17]. Here, an SLM is conjugated to the rear focal plane of the objective, and can generate multiple (circular) patterns to stimulate several entire somata simultaneously. Sub-millisecond jitter can be achieved with holographic stimulation [18, 19], and actuators with fast off-kinetics can be used [20, 21]. Such technologies for the precise control of spike timing in many neurons in parallel will be pivotal to probe computational principles of neuronal circuits or even to artificially drive animal behavior without sensory inputs [1, 22].

Thus in summary, a rapidly increasing number of optical technologies have been developed for neuroscientific applications. However, imaging and optical manipulation still face challenges on both technological and tool fronts as the channels for imaging and manipulation are spectrally distinct and therefore require two independent optical pathways with heavy weight and sensitive as well as expensive optical equipment.

1.1 Optogenetic Tools for All-Optical Experiments

Several advances are being made in tool development as new and better indicators and actuators are constantly being designed. However, the most commonly used green-absorbing calcium indicators GCaMP6 through 8 are still superior to their red-absorbing counterparts jRGECO or K-GECO1 [23, 24]. It is believed that only the pairing of red indicators and blue-absorbing actuators will completely prevent cross-interference between imaging and stimulation. Yet, for the more commonly used combination of a blue-light absorbing indicator in combination with a red-absorbing opsin the actuator is always cross-activated during imaging, since retinal isomerization is triggered via the hypsochromic shifted excitation light (shorter wavelength relative to maximum absorption see Fig. 1a). This can elicit electronic transitions to higher electronic states (e.g., S0 → S2), which are less efficient than those to a lower energy state (S0 → S1) [25, 26]. To some extent, lower imaging light powers can mitigate such effects, but deeper imaging and in vivo applications are often antagonistic to low power imaging.

Fig. 1

Spectral combinations of indicators and actuators in an all-optical experiment, and their respective cross-activation. (a) illustrates the overlap between the absorption spectra of GCaMP with the red-light absorbing opsin ChRimson. Here, during hypsochromic imaging of the indicator, higher electronic state transitions in the all-trans retinal can evoke photoactivation of ChRimson. While in (b) the absorption spectra of RCaMP and the action spectra of ChR2 are shown. Here, both spectra are sufficiently separated to avoid cross-activation of the blue-light absorbing opsin via bathochromic imaging, while contamination of RCaMP emission during opsin photoactivation is minimal due to low absorption of the red indicator for shorter wavelengths

Despite ongoing developments toward better red-light absorbing indicators, artifacts from photophysical processes induced by blue light imaging excitation in these red absorbing indicators still constitute a major obstacle for their widespread use in all-optical experiments.

For example, promising red-light absorbing calcium indicators such as jRGECO1a and RCAMP2, which originate from mApple, exhibit a blue-light sensitive protonation state giving rise to increased red fluorescence emission with similar kinetics as calcium-induced changes [23, 24]. Similarly, far-red calcium indicators, excited at approximately 590 nm, such as R-GECO and K-GECO also exhibit blue-light artifacts, albeit with faster kinetics than calcium-induced changes [27].

In the last decade, the rise of voltage indicators has additionally augmented the sensor landscape, however, the aforementioned bottleneck is only further enhanced by commonly used genetically encoded voltage indicators with a high SNR such as ASAP or Marina that also absorb in the blue/green spectral range [28, 29]. Additionally, a somewhat baffling finding is that promising candidates of opsin-based voltage indicators are not accessible via two-photon excitation due to their poorly understood photophysical processes related to the lifespan of the excited state [30]. Moreover, opsin-based indicators have inherited the aforementioned S2 hypsochromic excitation. Thus, although the collection of genetically encoded tools is constantly expanding and improving, an artifact-free and robust combination of indicators and actuators has yet to be established.

2 All-Optical Single-Beam Experiments

A single-beam experiment refers to an approach in which a single laser beam at a given wavelength is multiplexed to simultaneously image and stimulate neuronal circuits. The most straightforward approach to this problem is to target two distinct cell populations with an actuator and indicator. A frequently used approach is to restrict the expression of actuators and indicators to different brain regions (anatomy-defined approach Fig. 2a). Here, cross-activation of the axonal projections in one target region, in which neurons express an indicator, is reduced, since photon-sensitized axons require higher light power to trigger an action potential than photon-sensitized somata [31]. However, such an approach necessitates identification of the light power regime for artifact-free imaging of indicators and carefully performed control experiments. The light power used to evoke the release of a neurotransmitter from photon-sensitized axons depends on several factors such as myelination, directionality of the axons relative to the plane of light, and axonal caliber [32].

Fig. 2

Fundamental approaches for all-optical experiments. In (a) the expression of actuator and indicator are separated through space. In this anatomical approach, full-field one-photon illumination of axons still triggers the activation of actuators and hinders simultaneous imaging of indicators, but low power imaging of the indicator can reduce cross-activation of the axon drastically. Moreover, if expression of the actuator is restricted to the soma, photo-stimulation can be entirely spatially separated, rendering an artifact-free simultaneous all-optical experiment possible. (b) Illustration of a cell-type-defined approach in which different cell types express either an indicator or actuator, so that imaging and optogenetic stimulation are segmented into small subregions containing only specific cell types. The approach in (c) is a variation of (b): while using a distinct expression of indicators and actuators, here, technology to define arbitrary paths allows for the selection of neurons with indicator expression in the entire field of view. In (d), a typical dual-color all-optical experiment is illustrated. Here, both indicator and actuator are expressed in the same cell. In this case, the choice of indicator/actuator combination is crucial. Usually green-absorbing indicators are employed alongside red actuators to minimize optical crosstalk

A related anatomy-defined strategy is often used for in vivo all-optical experiments. Here, a brain region harboring photon-sensitized somata is stimulated to elicit action potentials that propagate orthodromically to a projection region in which neuronal activity is monitored optically [32]. While such an approach considers the natural latencies and branching of axonal collaterals, it also requires restriction of actuator expression to the soma to avoid unwanted activation of axons. Additionally, for an in vivo application in a single beam experiment, the laser beam would have to be split and the resulting beamlets guided independently to two separate brain regions. Overall, both anatomy-defined strategies are limited to the study of axonal innervation in a specific brain region and are highly dependent on anatomical architecture.

For an all-optical experiment within the same region, actuator and indicator can be expressed in distinct cell populations, e.g., excitatory versus inhibitory neurons. In this case, a small imaging region of interest (ROI) containing few cells (e.g., excitatory neurons) devoid of actuator expression can be selected, while a subset of neurons expressing the actuator (e.g., interneurons) are holographically stimulated outside these ROIs (Fig. 2b).

Similar to the aforementioned anatomy-defined approach, a cell-type defined approach also requires strict soma-targeting of both actuator and indicator. Instead of limiting imaging ROIs to two small subsets, fast scanning approaches using AODs or resonant-galvo-galvo configurations allow an imaging path to be defined in which only neurons expressing indicators are scanned (see Fig. 2c). In particular, AOD technology allows one to traverse such an arbitrary pathway in the kHz range, making it ideal for use with fast voltage indicators as well as for calcium imaging [10, 28].

Importantly, anatomically and genetically defined approaches do not enable both activation and imaging from the same set of neurons. Ultimately, only simultaneous expression of both indicator and actuator can grant all read-and-write access to neurons but it is still particularly challenging to prevent cross-activation with a single laser beam.

Dual-color experiments are classically considered to be a bulletproof concept for an all-optical experiment. Yet, compromising on indicator SNRs due to less sensitive red-absorbing indicators make such an approach challenging in terms of detection (Fig. 2d). Even though dual-color experiments usually rely on two different laser beams, they can serve as an interesting framework and introduce valuable parameters to consider in a single-beam experiment.

As a practical example shown in Fig. 3, the imaging beam (longer wavelength than that of actuator absorption) passes over neurons expressing the actuator and indicator with a specific lateral (r_xy) and axial (r_z) dimension. The spacing between consecutive scan lines (d) depends on the resolution of the acquired image (R_scan), and defines how many times (N_soma) the laser passes an idealized soma (quadratic dimension given as D_soma) during a single frame. Based on typical values used in all-optical experiments (shown in Fig. 3), the following relations apply:

Fig. 3

Illustrations of scanning parameters and actuator stimulation patterns in an all-optical dual-color configuration experiment. (a) For a given x-y plane, the scanning mirrors move a laser spot across the field of view (FOV) with a given lateral r_xy and axial resolution r_z [33, 34]. (b) As an example, we assume a system with a 16x objective with NA = 0.8, yielding a FOV of 700 × 700 μm (d_FOV); an 8-kHz resonant scanner, scanning with a resolution of 512 × 512 lines; and a laser with f_repetition = 80 MHz, and a single pulse width of T_pulse = 140 fs, where we scan a single neuron with a dimension D_soma = 30 μm (in teal an idealized neuron, which was approximated to a rectangular scan area (gray shaded area) of 30 × 30 μm for estimating parameters). Scanning the entire FOV takes T_frame with a single line scan of the duration T_line. We define the distance d_on-cell as the distance scanned by the laser on the cell and d_off-cell as the distance the laser scans over non-opsin expressing parts of the FOV. The distance d is defined in y-direction as the distance between two scan lines. Depending on the resolution, the laser spot runs over a neuron expressing an actuator several times per frame. T_{on_cell} is the time the laser spends scanning the cell for a given line. Depending on the position of the neuron within the FOV, the subsequent line can give rise to another illumination period, i.e., cell is located on the border of the FOV so that d_off-cell is very small on one end of FOV. Within a single given frame, a neuron in our example will be exposed to N_soma = 22 lines with each of them having T_{on_cell} < 3 μs

$$ d=\frac{d_{\textrm{FOV}}}{R_{\textrm{scan}}}=1.4\ \mu \textrm{m} $$

(1)

Here, the spacing d between two consecutive lines is 1.4 μm and FOV denotes the field of view with respect to the entire laser scanning range, which is 700 × 700 μm in our example. Considering a 16x objective (0.8 NA) with a scanner speed (ν_scan) of 8 kHz and a resolution of 512 × 512, a laser scan line (dimension r_xy) passes over a 30-μm (rectangular) neuron approximately 22 times:

$$ {N}_{\textrm{soma}}=\frac{D_{\textrm{soma}}}{d}=\frac{D_{\textrm{soma}}{R}_{\textrm{scan}}}{d_{\textrm{FOV}}}=22 $$

(2)

During these 22 passes of the beam area of r_xy over the cell surface, only 20% of the total membrane surface will be illuminated (we assume the idealized neuron to be a square of an area of 900 μm²):

$$ {A}_{\textrm{illuminated}}={N}_{\textrm{soma}}\ {r}_{xy}\ {D}_{\textrm{soma}}=182\ {\mu \textrm{m}}^2 $$

(3)

The scanning frequency defines the dwell time (T_{on_cell}), which denotes the time the laser beam spends scanning a cell during each scan line of the bidirectional movement of the resonant scanner(example):

$$ {T}_{\textrm{on}-\textrm{cell}}={D}_{\textrm{soma}}\ {\left({\nu}_{\textrm{scan}}\ 2\ {d}_{\textrm{FOV}}\right)}^{-1}<3\ \mu \textrm{s} $$

(4)

In our example, the average dwell time per cell and line is roughly 3 μs. However, for a single opsin molecule, dwell times are even shorter (cross-sectional diameter approximately 2 nm). In our example, an opsin that diffuses within the plasma membrane with 0.5 μm/s can be regarded as immobile during the dwell time of 3 μs [35]. Each opsin will be exposed to several light pulses during one scan line, but once activated, an opsin molecule with its millisecond off-kinetic can only be triggered once during a single line, and depending on r_xy and the FOV only once within a single frame acquisition. This is in drastic contrast to indicators that have fluorescent lifetimes of 2–10 ns, and can be excited and emit a single photon with every light pulse (approximately 10 light pulses when sweeping a beam across a single molecule). Therefore, the number of emitted photons indicative for the state an indicator is in can be further increased through prolonging dwell times in the microsecond range, while avoiding increased photocurrent through double-activation of an opsin with its milliseconds photocycle.

To better understand two-photon activation of actuators and indicators, two-photon cross-section (unit: Goeppert-Mayer; GM) data are invaluable to estimate cross-activation. Yet, GM values are rare for opsin-based indicators, though retinal-based actuators generally tend to have higher values than those of fluorescence proteins. As mentioned above, the light sensitivity and dynamic range of genetically encoded indicators are being periodically improved. Current versions allow two-photon imaging at low excitation powers. A recent report demonstrated that GCaMP6 can be imaged using two-photon light powers as low as 50–80 mW (using an 80 MHz Ti:Sapphire laser) and that this did not lead to any significant change in firing rates of neurons expressing a red-shifted opsin (λ_max = 540 nm) [22, 36]. Nevertheless, subthreshold depolarizations were not monitored, but likely induced and potentially biased ongoing basal excitability levels, thereby causing measurement artifacts in the neuronal circuit under investigation. Ultimately, further two-photon optimized indicators together with carefully selected opsins, as discussed below, can potentially enable single-beam all-optical experiments with little to no cross-activation. In particular, this includes a well-controlled expression level of opsins that is just sufficient to reliably trigger spikes holographically, while minimizing cross-activation when line-scanned.

3 Temporal Considerations of Actuator Kinetics in Single-Beam All-Optical Experiments

The possibility to optically probe neuronal systems is tightly linked to the discovery and exploitation of light-gated ion channels [37,38,39]. Channelrhodopsins are the main actuators used in neuroscience. The success of opsin-based actuators triggered an avalanche of protein engineering studies as well as metagenomic exploitation to design opsins for specific experimental settings [40,41,42]. This new hype around opsin research was accelerated by the groundwork laid by earlier studies on light-driven ion pumps (reviewed in [43]). Based on their established spectroscopic, crystallographic, and electrophysiological paradigms, in less than 10 years, causal relationships between channelrhodopsin structure and ion-conducting pathways could be drawn on an atomic level [41].

The sequence of conformational changes and ion movements is described in a photocycle (see Fig. 4). We now know that an aqueous pore is formed between helix 1, 2, 3 and 7, and is guarded by two main gates. The pore opening is initiated via a so-called central gate, which is in close proximity to the retinal chromophore, and pre-opens the conduction pathway and allows influx of water molecules into the pore without conducting ions or protons. Only upon breaching the inner gate, ions and protons can be conducted along their electrochemical gradients.

Fig. 4

Unified photocycle for Channelrhodopsin2 in relation to its electrophysiological parameters [44, 45]. (a) is a schematic depiction of the unifying photocycle where a single two-photon process can either trigger an anti-photocycle from the dark-adapted (DA) ground state or the transition to the light-adapted (LA) state. The LA state thermally relaxes back to the DA state or a second two-photon process can trigger a syn-photocycle. After relaxation to the open state(s), O₁ or O₂ the molecule transitions back to the closed ground state of the respective cycle. (b) Illustration of a two-pulse experiment in which two light pulses are given with increasing delay times (Δt) while monitoring the recovery of peak photocurrents during the second light pulse. The recovery of the peak photocurrent T_rec after varying Δt obeys a monoexponential increase referring to the transition of LA to DA (dotted red line)

The open time of ion channel pores, which has been a subject of mutational studies that led to many ChR variants, can range from milliseconds to seconds [42, 46]. After closing, the opsin returns to its ground state for a new photon excitation. Despite our increasingly detailed understanding of these processes, it is surprising how photocycles can differ among various ChR variants.

A unified photocycle derived from spectroscopic data also integrates a multitude of electrophysiological parameters such as photocurrent profile, on- and off-kinetics, inactivation and recovery, as well as light sensitivity (Fig. 4) [44]. Photocycles can exist in various forms, from a single photocycle consisting of one dark (D), an open (O), and a non-conducting state (P), to a dual photocycle consisting of two conducting states up to highly complex and branched photocycles [47,48,49].

Indispensable for our understanding of branched photocycles has been Raman spectroscopy. This technique can provide a fingerprint of molecules based on their vibrational or rotational states, revealing an additional photo-induced rotation of the retinal chromophore in ChR between two conformational isomers: syn and anti [50]. These conformational isomers refer to the –C15=NH– bond and are distinct from the photon-isomerization of the –C13=C14– cis/trans configurational isomers.

Such a second “light-switch” in an actuator has potentially interesting consequences for all-optical experiments in neuroscientific applications; in the complete dark-adapted state (DA), where all rhodopsin molecules harbor an all-trans/-C15=N-anti isomer, absorption can either cause the molecule to enter into an anti-photocycle (all-trans/-C15=N-anti → 13-cis/-C15=N-anti) or a syn-photocycle (13-cis/-C15=N-syn → all-trans/-C15=N-syn) ending in the light-adapted ground state (LA) [44, 51]. Molecular dynamics simulation supports that the ground state in the syn-cycle is indeed a pre-opened central gate, which can only completely open the ion-conducting pathway through a second absorption process. The LA ground state can thermally convert back to DA in the range of seconds (seen as recovery kinetic τ_rec), a process that can be monitored with a two-pulse experiment (see Fig. 4b).

Despite the debate on the contribution of this syn-photocycle to the overall photocurrent [52], it is evident that such photocycles exist and can be subjected to further design efforts. Potentially, an opsin could be engineered to exploit the dual photon absorption process to initiate an efficient syn photocycle only after dual-photon processes. Such a dual absorption would strongly depend on T_{on_cell} and the likelihood of a second absorption process from LA being triggered to start the syn-photocycle. The probability of such an effect will be very low within the time window of T_{on_cell}. Additionally, the probability of a two-photon absorption process within T_{on_cell} would quadratically decrease with light power. Engineering efforts would also have to be directed toward reducing the conductance in the anti-photocycle, that is, minimizing photocurrent and enhancing conductance in the syn-photocycle to reliably trigger action potentials during holographic stimulation. Holographic stimulation in the order of a millisecond would efficiently trigger a dual absorption process, and therefore trigger the syn-photocycle. After a single absorption event, the opsin molecule then thermally reverts from LA to DA with τ_Rec.

Yet another interesting photophysical light switch is the so-called photoinduced closing of an opsin. As spectroscopists visualized different photo-intermediates through their different absorption bands, neuroscientists exploited these different states by illuminating light at the corresponding wavelength of a particular state during the photocycle, and therefore short-circuiting photo-intermediates directly back to their ground state. The efficiency to photo-induce an off-switch is highest when certain photo-intermediates are stabilized and have long dwell times. This has been shown for bacteriorhodopsin, the prototype of a light-driven pump, here single-point mutations can result in the accumulation of specific M and N photo-intermediates which can be photo-converted to the ground-state [53, 54]. Similarly for ChR2, mutations in the residues C128 or D156 in ChR2 generate a set of mutants in which opsins accumulate in an open state. These so-called step-function opsins can be turned on for hundreds of seconds with a short light pulse (in the millisecond range), thereby initiating the transition from the dark to stabilized open state. Because the absorption band of the open states is red-shifted, they can then quickly be turned off with red light illumination above 550 nm [46]. Therefore, a one-photon full-field background illumination with red light could potentially mitigate cross-activating the opsin during a two-photon all-optical experiment with a single laser imaging beam around 980 nm.

However, such photo-induced back reactions still remain poorly understood, particularly during two-photon absorption. It would be interesting to explore the possibility to excite deep-blue blue-absorbing ChRs such as PsChR or CheRiFF (λ_max ca. 440 nm) with a 980 nm two-photon beam on the bathochromic (longer wavelength excitation relative to maximum absorption) side of their activation spectra. Eventually, cross-activated opsin molecules reaching O would then be back transferred with 980 nm excitation light.

Therefore, the matching of temporal properties of opsin molecules to imaging parameters can be utilized to minimize cross-activation. Here, off-kinetics that only allow a single activation of an opsin molecule during an imaging iteration in combination with indicators with short fluorescent lifetimes are favorable. In particular, T_on, T_off, and T_rec of opsin molecules have been heavily engineered and are ranging from few milliseconds to seconds [6, 55]. In contrast, photo-induced back reactions are not well understood, but could potentially act as an optical master switch that can render opsin molecules photocurrent-effective or ineffective.

4 Spatial Consideration of Optogenetic Tools in a Single-Beam All-Optical Experiment

For an all-optical experiment with a single laser line, a high SNR and single-cell resolution are essential. For stimulation of single-neurons within a large field of view with hundreds of neurons expressing photon-sensitive opsins, undesired off-target activation via their closely passing axons (and dendrites) along the soma of the targeted neuron becomes a challenge. To decrease background fluorescence as well as confine excitation to selected neurons, it is advantageous to localize the expression of actuators to specific compartments. Therefore, restricting the expression of opsins to the soma and dendrites has been a robust strategy.

Figure 5 gives an overview of different genetic targeting strategies along the neuronal cell body. Genetically fusing the opsin to ankyrin-G protein, couples opsin expression to spectrin-actin network, and consequently restricts expression to the somatodendritic region and axon hill. Despite the size of its targeting motif of more than 700 amino acids, the ankyrin-G sequence also targets the dendritic region and hence still gives rise to off-target activation [67]. However, targeting opsins to more defined and small subcellular regions such as the axon initial segment (AIS) has failed so far [56]. Early attempts to deploy targeting motifs found in voltage-gated ion channels to anchored actuators in AIS such as NaV1.2 have been successful in terms of localizing the transgene to the AIS, but the number of opsins molecules was too small to optically induce action potentials [57]. A similar strategy utilizes a shorter targeting motif derived from NaV1.6 and localizes opsins sufficiently to the AIS, but also changes the intrinsic excitability within the host cell itself [58, 59]. Based on these insights, short targeting motifs to prevent expression within the axons are most promising. For example, fusing a targeting motif from kainate receptor subunit 2 to the N-terminus CoChR yielded a soma-oriented actuator (soCoChR) allowing for holographically triggered action potentials with a 1 ms resolution and minimal off-target activation [19]. Similarly, the cytoplasmic C-terminal from the voltage-gated potassium channel Kv2.1 yielded specific targeting of opsins to somata as well as proximal dendrites (stChronos, stCoChR, or stGtaCR2) [20, 21]. Both targeting motifs, NaV1.2 and Kv2.1, have been successfully used for two-photon connectivity mapping; however, direct comparison between motifs remains to be elucidated.

Fig. 5

Overview of cellular localization of different target sequences. The neuron is divided into four segments: axon initial segment, soma, somatodendritic and soma and proximal dendrites. Here, an overview is given over the different molecular targeting strategies which are employed depending on which of these segments are supposed to be expressing the opsin

Clearly, restricting the expression of optical actuators to a fraction of the entire cell membrane leads to smaller photocurrent. Yet, the soma-targeted opsins exhibit larger photocurrents than do wildtype versions, indicating higher membrane expression of opsins when fused to the Kv2.1 motif [20, 60]. Furthermore, apparent off-kinetics of soma-targeted versions in neurons are faster than unmodified versions, since delayed axonal and distal dendritic current contributions are removed from the overall kinetics.

On a similar basis, targeting cytoplasmic indicators to the soma or nucleus can be beneficial. In particular, nucleus-targeted calcium indicators help to segment calcium transients to individual cells [61]. However, onset latencies are prolonged due to slower calcium rise in the nucleus. For single spike events, calcium transients might not propagate into the nucleus and the overall response sensitivity of calcium indicators is reduced in the nucleus.

Figure 5 gives an overview of different targeting sequences and the resulting expression regions. In a recent screening, novel and optimized targeting motifs have been reported [62]: a shorter ankyrin-based motif combined with an ER trafficking signal from Kir2.1 fused to the N-terminus restricts calcium indicators to the somatodendritic region (Fig. 5). In addition, a de novo synthesized coil-coiled peptide that self-assembles into a complex slows down transport out of the soma.

As cellular targeting helps to avoid cross-activation, a foreseeable future breakthrough will be the exact control of the expression level in relation to the membrane/volume ratio. Here, expression systems that encode an auto-feedback that restricts expression to the necessary amount will greatly reduce the effect of cross-activation with indicators [63].

Note 1: Beyond Temporal and Spatial Constraints: Ion Permeability

Not only spectral properties within a photocycle can be exploited for neuroscientific applications. Ion selectivity is a key feature to adapt ChR to specific neuroscientific experiments. As ChRs are intrinsically non-selective cation channels permeable to protons, sodium, potassium, and even calcium, they are not designed per se for the sensitive electrochemical ionic gradient in neurons. With an atomic structure of different ChRs at hand [56, 57], several protein engineering attempts yielded ChR variants with improved ion selectivity (Permeability, P). Particularly interesting for a neuroscientific application are P_Na+/P_H+ ratios. For example, variants such as ChRomeT (A71S/E90A/H114G), Chronos-D139H, or the naturally occurring opsin PsChR exhibit shifted ratios of ten to hundredfold to higher sodium permeability. Further, P_Ca2+/P_H+ ratios have been modified to improve calcium conductance in ChR2 mutants such as ChR2-L132C, ChR2-S63D, or ChR2-L132C-T159C [6, 58, 59]. An increased sodium selective opsin translates photocurrent more directly into membrane depolarization in neurons.

In a branched photocycle, the high-conductance O₁ state exhibits lower proton selectivity than does the low-conductance state O₂, and therefore the initial peak photocurrent carries a larger fraction of sodium in its photocurrent. So far, attempts to modify ion selectivity of distinct open states remain unsuccessful. The structural changes that lead to different ion selectivities in the respective open state are not well understood. As both open states share the same ion-conducting pore, mutational analysis introducing small structural changes will likely influence the ion selectivity of both open states.

Note 2: Beyond Temporal and Spatial Constraints: Spectral Consideration

In an ideal all-optical experiment, actuators and indicators are sufficiently separated without any cross-activation and optical setups are equipped with two independent spectral laser lines. However, in-depth understanding of photophysical processes underlying wavelength-dependent absorption can help design better all-optical single-beam experiments.

So far, only phenomenologically understood are the findings that stationary photocurrents current saturate at lower light powers and can exhibit slightly shifted absorption spectra [64, 65]. Typically, for one-photon imaging peak photocurrents saturate at light powers higher than 15 mW/mm², whereas stationary photocurrents saturate already at less than 5 mW/mm² [55]. However, for optimized opsins with large photocurrents, light intensities can be orders of magnitude lower [40]. Photocurrent profiles at low light powers exhibit a slow increase in amplitude that sometimes completely lacks any peak photocurrent due to lower absorption probability per time. This feature can be utilized in two-photon imaging, where indicators are monitored at low intensities and high scanning frequencies inducing small and skewed transient currents carried by a mix of O₁ and O₂ (see Fig. 3). For 2P optogenetic activation of neurons, high light powers will evoke large peak photocurrents and efficiently trigger action potentials. Therefore, for utilizing such a strategy, ChRs with strong and fast inactivation and strong expression are preferred. Similarly, activation of opsins outside their maximal absorption range can also lead to photocurrent profiles with reduced peak photocurrents and slow on-kinetics [66].

In summary, low power imaging and excitation wavelengths outside the maximal absorption band reduce transient peak photocurrents and therefore minimize cross-activation of opsin-based actuators during two-photon resonance imaging.

5 Summary

To gain cellularly resolved read-and-write access to an entire neuronal circuit, photo-sensitizing proteins, indicators (read), and actuators (write) need to be expressed within a single neuron. Ideally, neuronal activity is monitored in a large field of view while multiple neurons are stimulated in parallel with sub-millisecond latencies. Here, scanless stimulation technologies for multicell excitation, such as holographically multiplexed spiraling or sole holographic stimulation with spots fitting the size of a neuron, are used to excite actuators, whereas indicators are imaged using fast scanning approaches (Fig. 6). Since an ideal combination of spectrally distinct and high-efficiency indicator/actuator pairs remains unavailable and would require expensive setups with multiple laser lines and a highly complex optical design, in this chapter, we review unexplored spatial and temporal photophysical features of spectrally comparable indicator and actuator pairs permitting an all-optical single-beam experiment with minimal cross-activation.

Fig. 6

Comparison of different stimulation approaches for a single-beam experiment. (a) outlines the spiral stimulation approach, while (b) elucidates the holographic approach, both combined with fast scanning indicator imaging. The grey inserts outline the methods’ respective advantages and drawbacks

Indicators are constantly being improved in terms of higher SNR and better imaging properties. However, mutational screening for a new generation of indicators with improved two-photon cross-sections is rarely performed. Within our introduced theoretical framework, we demonstrate that each actuator opsin molecule is only activated once during a laser beam sweep. In contrast, indicators can emit hundreds of photons during the same time window. Therefore, next to improving the two-photon cross-section of indicators to allow for efficient activation, decreasing their fluorescence lifetimes to below 5 ns can help distinguish between indicator and actuator activities. Spatial restriction of indicator expression within the soma can further help prevent background fluorescence arising from the neuropil, thereby decreasing the imaging contrast.

With regard to actuators, we have highlighted the benefits of using moderate or slow off-kinetic opsins for spiraling approaches because they more efficiently integrate single-channel ion conductances towards crossing the spiking threshold. Fast and complete desensitization from peak to stationary photocurrents reduces the probability of multiple action potentials being triggered during a single spiral scan. In this case, high membrane occupancy of actuators is advantageous since only a fraction of the entire membrane is illuminated during a spiral scan.

In contrast, holographic stimulation concurrently activates the entire opsin-packed neuronal membrane with a millisecond illumination. Such an activation eliminates the need to sum over single-channel conductances and favors the use of actuators with fast off-kinetics. To avoid cross-activation via the imaging beam, we suggest choosing blue-shifted opsins relative to the imaging laser beam. Thus, the laser light will only cross-activate actuators with a decreasing red spectral flank, reducing the overall probability of activation. However, holographic stimulation can still efficiently trigger action potentials via such a bathochromic excitation, given high enough light powers.

As the basic photophysical processes of light-gated ion channels are already well understood, we mainly focused on photophysical processes in the context of two-photon excitation. Further, we discussed photo-induced processes beyond the classical all-trans → 13-cis isomerization in channelrhodopsins. Particularly, we describe a second two-photon absorption process based on an anti/syn conformation.

Harnessing any double two-photon absorption to activate opsin molecules in an all-optical experiment would render the light power dependency quadratic for actuators rather than for indicators. Furthermore, just like in the case of indicators, high-throughput mutational analysis for two-photon optimized opsin variants is virtually absent with this being particularly true for opsin-based actuators. Here, screenings toward high two-photon cross-sections or the facilitation of double two-photon absorption processes could enable artifact-free imaging of indicators and actuators with a single-beam line.

As previous years have demonstrated how prolific and diverse opsin evolution has been through natural selection, future years will reveal whether artificial screening toward two-photon optimization will produce a plethora of actuators for all-optical experiments.