Abstract
The analysis of neural circuits has been revolutionized by optogenetic methods. Light-gated chloride-conducting anion channelrhodopsins (ACRs)—recently emerged as powerful neuron inhibitors. For cells or sub-neuronal compartments with high intracellular chloride concentrations, however, a chloride conductance can have instead an activating effect. The recently discovered light-gated, potassium-conducting, kalium channelrhodopsins (KCRs) might serve as an alternative in these situations, with potentially broad application. As yet, KCRs have not been shown to confer potent inhibitory effects in small genetically tractable animals. Here, we evaluated the utility of KCRs to suppress behavior and inhibit neural activity in Drosophila, Caenorhabditis elegans, and zebrafish. In direct comparisons with ACR1, a KCR1 variant with enhanced plasma-membrane trafficking displayed comparable potency, but with improved properties that include reduced toxicity and superior efficacy in putative high-chloride cells. This comparative analysis of behavioral inhibition between chloride- and potassium-selective silencing tools establishes KCRs as next-generation optogenetic inhibitors for in vivo circuit analysis in behaving animals.
Introduction
The ability to manipulate distinct neuronal populations in a spatiotemporally precise manner is invaluable to research into brain function. A key approach that has revolutionized such research is optogenetics, which uses cell type-specific expression with light gating to precisely control neuronal activity1,2,3. While optogenetic activators have already achieved a high level of potency and sophistication4,5,6,7, their inhibitory counterparts are comparatively less well-developed. Despite progressive improvements8,9,10, light-driven inhibitory chloride pumps require high expression levels and strong light intensities11. As such, the discovery of a pair of natural chloride-conducting light-gated ion channels12 represented a major development in inhibitory optogenetics. Isolated from the cryptophyte algae Guillardia theta, these anion channelrhodopsins (ACRs) have proven to be potent and versatile inhibitors of neuronal activity in Drosophila13,14, zebrafish15, mouse16,17,18, ferret19 and Caenorhabditis elegans20,21.
As an anion channel, the light actuation of an ACR is roughly equivalent to opening a chloride conductance12 (Fig. 1A). Because the equilibrium potential of chloride in neurons usually falls below the threshold for action potentials, ACR actuation will typically inhibit firing by hyperpolarizing the cell12,17. The complexities of chloride physiology, however, mean that chloride-based silencing has at least three relevant caveats. First, the active chloride extrusion found in mature neurons is unusual for animal cells, including both non-excitable and excitable cells, which generally have a high intracellular chloride concentration22,23,24,25. In such cells, a chloride conductance would have a depolarizing effect26,27,28,29 and thus, as has been shown in some cases, rather than inhibiting, ACR actuation can cause activation11,29,30. Second, some neuronal compartments (notably axons) have higher steady-state intracellular chloride levels than the soma31,32,33; again here, in such compartments, chloride conductances can be activating5,11,17,18,34,35,36. Third, prolonged chloride conductances can lead to complex secondary effects (including the redistribution of potassium) with diverse impacts on excitability22,37. Careful opsin engineering has reduced ACR activation of axons by targeting opsin expression to the soma17,18; however, even this innovation does not resolve the potentially ambiguous effects of optogenetic chloride channels on membrane potential in other contexts.
Potassium (K+) channels have fundamental roles in setting the resting membrane potential and terminating action potentials38. As such, researchers have long sought a light-actuated K+-selective channel to use as a neuronal inhibitor. To this end, chimeric potassium channels, such as HyLighter39, BLINK140, BLINK241, and PAC-K42, were engineered to be light-responsive. To date, engineered light-actuated K+ channels have relatively slow kinetics, and have not been used in the major invertebrate model systems. Recently, the discovery of genomic sequences from a stramenopile protist led to the identification of two channelrhodopsins that naturally conduct potassium: Hyphochytrium catenoides kalium channelrhodopsin 1 and 2 (HcKCR1 and HcKCR2, Fig. 1A)43. Actuation of these HcKCRs opens K+-selective conductances that, as shown for HcKCR1, can inhibit action potentials in mouse brain slices43. More recently, HcKCR1 has also been used to successfully suppress neuronal activity during virtual-reality behavior in mice44. A third KCR was subsequently identified from the stramenopile Wobblia lunata—termed the Wobblia inhibitory channelrhodopsin (WiChR)—which has even sharper K+ selectivity and inhibits action potentials in anesthetized mouse brain and cardiomyocytes45.
While KCR efficacy has been shown in brain slices and behaving mice, there are large differences between rodent and invertebrate experiments, including transgene expression levels, membrane targeting, and optical accessibility. As such, we aimed to investigate the utility and efficacy of KCRs to inhibit and silence neurons in vivo in Drosophila, C. elegans, and Danio rerio—the major small animal models.
Results
Trafficking signals improve KCR localization to axons
Heterologous opsins can have poor trafficking to the plasma membrane and are retained in internal membranes10. As such, we aimed to identify HcKCR configurations that efficiently localize to neurites. With standard fly transgenic methods, we targeted different KCR fusion proteins (Fig. 1B) to the mushroom bodies (MB)46,47 and then compared their localization. As the MB somata are located in the posterior brain but their axons project to the anterior MB lobes47,48,49, we could use this spatial separation to estimate relative axonal localization. We found that the simplest KCR fusion protein, KCR1 with an enhanced yellow fluorescent protein (eYFP) linked with three alanine residues (KCR1-AAA, Fig. 1B), was equally localized to the MB soma and axons (Fig. 1C, H). This finding is consistent with prior reports showing that HcKCRs have imperfect membrane localization43,45. Replacing AAA with a longer linker (3× GGGGS, KCR-GS, Fig. 1B) slightly worsened anterior/axonal localization (Fig. 1D, H). Adding endoplasmic reticulum export and Golgi trafficking (ET) motifs (Fig. 1B)10,50,51,52,53, produced KCR-ET variants with superior relative axonal localization (KCR1-ET and KCR2-ET, Fig. 1E, F, and H). Comparing the GFP signal intensity between anterior and posterior brain regions revealed that KCR1-ET, KCR2-ET, and ACR1 (Fig. 1G), were preferentially localized to axons (Fig. 1H). Expressing these opsin variants in cultured mouse neuroblastoma (N2a) cells54 confirmed this improvement, revealing a predominantly intracellular localization of KCR1-GS and increased relative membrane localization of KCR1-ET and KCR2-ET (Fig. 1I).
KCR1 actuation effectively impairs locomotor behavior
Having established the KCR-fusion expression patterns, we next targeted three of the KCR variants and, as a benchmark, ACR1 in Drosophila neurons and tested climbing ability during actuation (Fig. 2A). We used two different drivers: OK371-Gal4 driving expression in motor neurons and elav-Gal4 driving expression in all neurons55,56. Light actuation had large effects on climbing in all test lines. In OK371 flies, the strongest effectors were KCR1-ET and ACR1 (Δheight = −37.9 mm and −37.5 mm, respectively; Fig. 2B–E). KCR2-ET was noticeably weaker than the others (Δheight = −22.6 mm). In elav-Gal4 flies, ACR1 gave the most profound paralysis, while KCR1-ET and KCR1-GS had similarly robust, if incomplete, effects on climbing (Δheight = −53.7, −38.5, and −43.1 mm, respectively, Fig. 2F–H). The results indicate that the blue-light sensitive KCR2 is a weak inhibitor in flies; this may be due to the poor transmission of blue light through the adult fly cuticle57. For this reason, we focused our efforts in Drosophila on the green-light sensitive KCR1 fusions going forward. Indeed, we saw that KCR1 is a potent optogenetic inhibitor, with different effects with the two drivers: inferior to ACR1 in elav-Gal4, but comparable to ACR1 in OK371-Gal4 cells. Despite their differences in cell-surface localization, we observed no major difference in climbing impairment between KCR1-ET and KCR1-GS in this assay.
To generalize the climbing effects to a different motor assay, we also tested the KCR1 lines in the OK371-Gal4 motor neurons in a horizontal walking assay (Fig. 2I, J). In controls, light elicited substantial increases in walking speed when comparing light and dark epochs (Fig. 2J, S3). In all test lines, exposure to light resulted in marked declines in walking speed: the light-elicited locomotion reductions were ranked KCR1-GS > KCR1-ET > ACR1; Δspeed = −0.79, −0.45, and −0.33, respectively. In addition to the larger relative speed reductions compared to ACR1 (which were partly due to faster dark-epoch walking in the KCR flies), both KCR lines exhibited near-complete suppression of locomotion: actuated walking speed = 0.5, 0.2, and 0.2 mm/s for ACR1, KCR1-ET, and KCR1-GS, respectively. Along with impaired walking, actuation of either ACR1 or the KCRs in OK371 motor neurons also induced limb twitching. In ACR1, this did not occur at 44 µW/mm2 (Fig. S3 and Supplementary Video SV4), and in both ACR1 and KCR1 lines, resolved during longer exposure (Fig. S3 and Supplementary Video SV1).
Taken together, these findings show that KCR1 actuation in OK371 neurons suppresses both climbing and horizontal walking, and thus effectively inhibits Drosophila motor neuron function. The KCR1 transgenes have comparable performance to ACR1.
Gustation-dependent feeding and olfactory memory are inhibited by KCR1-ET
We next examined whether KCR1 can be used to inhibit sensory systems. The gustatory receptor Gr64f is expressed by a small cluster of sweet-sensing neurons (Fig. 3A)58; inhibiting Gr64f cells with Gr64f-Gal4 and the potassium rectifying channel Kir2.1 can reduce feeding59. As such, we expressed KCR1-ET and ACR1 in these cells to test their ability to attenuate feeding, as analyzed using an automated assay (Espresso, Fig. 3B)60,61. Specifically, 24 h-starved flies were allowed to feed for 2 h and were illuminated for the initial 30 min of each hour (Fig. 3C). While illumination had no effect on consumption in controls, both ACR1 and KCR1-ET flies consumed less food during light-on epochs. The feed-volume reductions in ACR1 and KCR1-ET flies were similar (Δvolume = −0.34 µl and −0.30 µl, respectively; Fig. 3D). Thus, KCR1-ET effectively inhibits at least one class of primary-sensory neurons in flies to the same extent as ACR1.
To test KCR1-ET in higher-order sensory neurons, we examined the intrinsic neurons of the MB, which are required for associative olfactory memory62. We expressed KCR1-ET or ACR1 in the MB with MB247-Gal4 and subjected the flies to an aversive Pavlovian conditioning paradigm (Fig. 3E)13. During actuation, memory was strongly impaired in both ACR1 and KCR1-ET lines, with conditioned odor preference reduced to near-indifference (∆PI = −0.31, −0.38, respectively, Fig. 3F, G). During a retest without light, conditioned avoidance was intact thereby attributing the defective recall to channelrhodopsin actuation. These results confirm that ACR1 and KCR1-ET have comparable efficacy for inhibiting sensory neurons.
Spontaneous action potentials are strongly inhibited by KCR1
Thus far, we have seen that KCR1 and ACR1 have similar inhibitory effects, except for the elav-Gal4 climbing experiment, where ACR1 outperformed KCR1. To investigate this performance difference further, we performed electrophysiological recordings from abdominal nerves in fly larvae with elav-Gal4 driving KCR1-GS, KCR1-ET, or ACR1. Green light actuation for 30 s produced a strong suppression of spiking for all three genotypes (Fig. 4A, B). These data reveal that all three opsins allowed some residual spiking in some nerves; in all ACR1 recordings there was a rapid (~1 s) and nearly complete inhibition, while in some of the KCR1 recordings, there was a 10–15 s lag before complete inhibition in most nerves (Fig. 4B, C)….