Optogenetic Long-Term Manipulation of Behavior and Animal Development

Channelrhodopsin-2 (ChR2) is widely used for rapid photodepolarization of neurons, yet, as it requires high-intensity blue light for activation, it is not suited for long-term in vivo applications, e.g. for manipulations of behavior, or photoactivation of neurons during development. We used “slow” ChR2 variants with mutations in the C128 residue, that exhibit delayed off-kinetics and increased light sensitivity in Caenorhabditis elegans. Following a 1 s light pulse, we could photodepolarize neurons and muscles for minutes (and with repeated brief stimulation, up to days) with low-intensity light. Photoactivation of ChR2(C128S) in command interneurons elicited long-lasting alterations in locomotion. Finally, we could optically induce profound changes in animal development: Long-term photoactivation of ASJ neurons, which regulate larval growth, bypassed the constitutive entry into the “dauer” larval state in daf-11 mutants. These lack a guanylyl cyclase, which possibly renders ASJ neurons hyperpolarized. Furthermore, photostimulated ASJ neurons could acutely trigger dauer-exit. Thus, slow ChR2s can be employed to long-term photoactivate behavior and to trigger alternative animal development.


Introduction
ChR2 is a light-driven cation channel that enables fast photodepolarization of excitable cells in culture and in live animals ranging from Caenorhabditis elegans to primates [1-6]. However, for long-term photodepolarization, e.g. to influence learning or neuron-controlled alternative developmental pathways, ChR2 is not suited: As it requires continuous illumination with blue light of high intensity ($1 mW/mm 2 ) to keep the channel open a) phototoxicity may arise and b) intrinsic phototactic reactions of animals can occur that interfere with the studied behavior. These limitations may be overcome by the recently described ChR2(C128X) mutants [7][8][9]. Compared to wild type ChR2 (t off = 11.9 ms), mutations of C128 to T, A, or S significantly delay the closing of the channel in the dark (t off = 2 s, 56 s, and 106 s, respectively; [7]). As the open photointermediate P520 accumulates, light of reduced intensity suffices for efficient channel-opening. Once in the open state, C128X mutants can be photoinactivated using green-yellow light, thus they are also termed ''step function opsins''.
C. elegans is a genetic model for studies of neurobiology and development, among other areas of biology. Its nervous system is mapped down to the individual synapse [10], and its neurons form simple functional units, similar to elementary network units found in higher animals [11]. C. elegans exhibits stereotypic behaviors, e.g. escape reflexes in response to particular sensory inputs, and, depending on external conditions, alternative developmental pathways. In a favorable environment, the nematode develops through four larval stages into adult animals [12], while under harsh conditions, reproductive development is bypassed and animals enter a long-lived ''dauer''-state after larval stage L2 [13][14][15]. Dauer larvae exhibit specialized morphology and metabolism, allowing them to survive harsh conditions for several months [13,[16][17][18]. Importantly, harsh or beneficial conditions are detected by sensory neurons that prevent or instruct entry into, or exit from, the dauer-state [19,20].
We characterized slow ChR2 variants for prolonged photoactivation of excitable cells in C. elegans. ChR2(C128X) could photodepolarize body wall muscle (BWM) cells, cholinergic and GABAergic motorneurons for several minutes following a 1 s light pulse. As in other systems, the open state could be terminated by yellow light. Continuous activation of the locomotion command interneurons evoked long-lasting behavioral alterations. Lastly, we could alter the genetically predisposed development of C. elegans by long-term photodepolarization of ASJ sensory neurons, to either prevent the constitutive dauer-entry in daf-11 mutant animals, or to achieve an exit from the dauer-state.

Results
Slow ChR2 mutants enable long-term activation of muscles with low light intensity First, we expressed three slow ChR2 variants (C128T, A, and S), in BWMs (Fig. S1a), as well as ChR2(H134R), which has slightly delayed kinetics and thus larger steady-state conductance than wild type ChR2 [4]. Concomitant depolarization of all BWMs induces a uniform contraction, causing a decrease in body length; thus, body length is a measure for extent and persistence of BWM depolarization [4]. The proteins localized mostly to the plasma membrane, with variable amounts of intracellular, sometimes aggregated protein. We analyzed expression levels based on fluorescence in individual muscle cells, which showed strong differences: ChR2(C128T and S) both expressed better than ChR2(H134R), while ChR2(C128A) expression was low, showing mosaicisms (data not shown). However, as the proteins aggregated to a variable extent and we could not specifically determine cell surface expression levels, these findings did not allow us to predict which protein may be best suited for long-term applications in C. elegans. Thus, to assess this based on function, we monitored the capability of the ChR2 variants to depolarize BWMs. Animals (grown in presence of all-trans retinal -ATR) were illuminated for 1 s with blue light (450-490 nm; 0.69 mW/mm 2 ), and the body length was deduced from videos [5] (Fig. 1a, b, and Video S1). ChR2(H134R) induced a ,12% contraction, and animals returned to initial length ,1 s after light-off. ChR2(C128T, A, and S) induced comparable contraction amplitudes, however, relaxation of the body wall was largely delayed, occurring after 5 s, 3 min, and .5 min, for ChR2 C128T, C128A, and C128S, respectively. Measuring ChR2(C128S)-mediated inward photocurrents confirmed the largely delayed channel-closing (Fig. 1c, and see below).
We determined the lowest light intensity (0.001-2.2 mW/mm 2 ) sufficient to achieve saturating ChR2-dependent effects. Contractions .8% were evoked by ChR2(H134R) in BWMs when at least 0.08 mW/mm 2 light were applied. In contrast, photoactivation of ChR2(C128S) with as low as 0.01 mW/mm 2 still evoked full contractions (Fig. 1d), that were also prolonged for several minutes (Fig. S1b, c), thus establishing ChR2(C128S) as a powerful tool for prolonged depolarization of excitable cells under minimal light-invasive conditions. As daylight already caused marked contraction and uncoordinated locomotion, ChR2(C128S) animals should be kept in the dark and handled under low intensity red light to prevent unwanted photoactivation.
ChR2(C128S) can be repeatedly ''switched'' on and off with blue an yellow light As in other systems, we could photo-switch slow ChR2 variants from the open state to the closed dark-state, by using yellow light [7]. We applied alternating blue (450-490 nm; 1 s; 0.01 mW/ mm 2 ) and yellow (565-595 nm; 1 s; 4.4 or 2.5 mW/mm 2 ) pulses, each followed by an 8 s dark period, and monitored the body length of animals expressing either ChR2(H134R) or ChR2(C128S) in BWMs ( Fig. 1e and Video S1). In H134R animals, low intensity blue light induced ,2% contraction during illumination, while yellow light had no effect. In contrast, in C128S animals, blue light induced a continuous contraction of ,10% that was completely abolished by the yellow light pulse (Fig. 1e), thus allowing full temporal control over ChR2(C128S) induced depolarization.
Long-term stimulation of ChR2(C128S) leads to a partial reduction of function Potential applications of ChR2(C128S) could be to keep neurons depolarized for hours to days to affect processes like learning or even developmental pathways. However, Schoenenberger et al. (2009) found a progressive inactivation of ChR2(C128A), when in the open state, and in response to repeated stimuli. A fraction of molecules appeared to transition into an ill-defined, non-activatable ''lost state'', from which they recovered very slowly. We thus assayed for how long ChR2(C128S) may be continuously activated. ChR2(C128S) in muscles was photoactivated for up to 1 day using two different protocols: a) 1 s blue light every 2 min; or b) 1 s blue, 5 s dark, then 2 s yellow, etc., the latter one to actively prevent loss of ChR2 to inactive states. At 0, 30, 120, 360 minutes and 1 day, animals were given a blue test pulse for full activation, followed by a yellow pulse for inactivation, and the relaxation was measured (Fig. 1f). After 30 min, effects were reduced from initially 7.661.4% to 4.561.0%, after 120 min, and still after 1 day, they were down to 1.460.6%. Thus, long-term depolarization via ChR2(C128S) may cause only ,18% of maximal effects and should be considered when using ChR2(C128S) in the range of several hours or days. Nevertheless, depending on the cell type, this remaining functionality for even 24 h may be sufficient for long-term activation of the particular cell and its potential downstream targets.

Slow ChR2 mutants allow activating motorneurons in C. elegans
We next tested the applicability of slow ChR2 mutants in neurons. We expressed ChR2(C128S) in cholinergic motorneurons, that cause muscle contraction when photostimulated [5]. As photostimulation of cholinergic neurons causes a coiling phenotype, due to concomitant GABA signaling, we analyzed the effects of ChR2 activation in unc-49(e407) mutants that lack the muscular GABA A R. A 1 s, low-intensity (0.01 mW/mm 2 ) light pulse caused prolonged contractions of ,10%, which were longlasting (several minutes; Fig. 2a, d, and Video S2), and could not be achieved using ChR2(H134R). Also for cholinergic neurons, we could photo-switch ChR2(C128S) from the open state to the closed dark-state, using yellow light, and this could be repeated up to 10 times, with no obvious loss of activity (Figs. 2b and S2). As for muscles, the lowest light intensity sufficient to achieve saturating ChR2-dependent effects in cholinergic neurons was 0.01 mW/mm 2 (Fig. 2c). At this light intensity, ChR2(H134R) appeared to evoke contractions more efficiently in cholinergic neurons than in BWMs (compare Figs. 1d and 2c). This may be due to cell-type specific differences in the environment of the channel, affecting its properties, or because contractions evoked by ChR2 in cholinergic motorneurons are effected by ACh release and postsynaptic nAChRs, which may be more efficient, than directly by photocurrents within BWMs. Similar experiments with ChR2(C128S) and ChR2(H134R) in GABAergic motorneurons (evoking body relaxation; [5]) showed qualitatively comparable results (Fig. S3), emphasizing the utility of ChR2(C128S) in several neuron types. Thus, ChR2(C128S) can be used to mimic prolonged synaptic transmission at the neuromuscular junction.

Long-term alteration of behavior in locomotion command interneurons
We next analyzed whether behavior can be altered in the longterm by depolarizing command interneurons, which regulate certain aspects of locomotion (Fig. 3a), particularly the direction and likely also the speed of movement: AVB and PVC neurons trigger forward, whereas AVA and AVD mediate backward locomotion. Each cell type mutually inhibits the opposite type, thus they form a bi-stable switch that fluctuates between backward and forward states. Sensory neuron input alters this balance by depolarizing one command neuron type; in undisturbed animals, forward command neurons dominate, and worms crawl mostly forward, interrupted by brief backward episodes (,2-4 times min 21 ; [21]). Concomitant activation of all command neurons thus perturbs normal activity and affects locomotion.
We used the glr-1 promoter to express ChR2(C128S) and ChR2(H134R) in all command neurons and some additional neurons [22] (Fig. S4). The ratio of backward movement was assayed for three consecutive 1 min periods, and calculated for the whole period. A 1 s blue light pulse (2.1 mW/mm 2 ) was applied after the first minute to activate ChR2 variants. After the second minute, yellow light (1 s; 6.1 mW/mm 2 ) was presented for inactivation. Non-transgenic animals (wild type) did not exhibit any significant change in the proportion of backward movement (Fig. 3b). For ,40% of ChR2(H134R) expressing animals, a 2 mW/mm 2 , presented at t = 0 s; reduction of body length was measured for ChR2(H134R) directly after light off (t = 1 s), for ChR2(C128S), due to the slower onset, 2 s after light off (t = 3 s). (e) Relative body length of worms while alternating 1 s blue (0.01 mW/mm 2 ; 450-490 nm) or 1 s yellow (4.4 mW/mm 2 ; 565-595 nm) light pulses were presented, indicated by blue and yellow bars. (f) Long-term activity of ChR2(C128S). 1 s blue light pulses (0.05 mW/mm 2 ; 470 nm) were presented either every 120 s or every 8 s (in the latter case, followed by a 5 s dark period and a 2 s yellow light pulse (0.04 mW/mm 2 ; 590 nm)). At the indicated times, animals (n = 9-15) were presented a blue test pulse (2.1 mW/mm 2 ; 450-490 nm), followed by a yellow pulse (6.1 mW/mm 2 ; 565-595 nm) for inactivation and the resulting relaxation was measured. n = number of animals; error bars are s.e.m.; ***: p,0.001, **: p,0.005, *: p,0.01. doi:10.1371/journal.pone.0018766.g001 reversal was observed right after the blue light photoactivation. However, as these effects were very brief, i.e. restricted to the time of illumination, they did not become evident in the analyzed oneminute proportion of backward locomotion. In contrast, ,71% of ChR2(C128S) animals reversed upon photoactivation and thereafter crawled backwards for extended periods, often even until the inactivating yellow pulse. The proportion of backward movement increased from 19.464.2% to 76.864.4% during the second minute, which was completely reversed after the inactivating yellow pulse (26.365.3%; Fig. 3b and Video S3). Thus, command neurons can be optically manipulated in the long-term, to evoke downstream effects across several synapses, emphasizing the utility of ChR2(C128S) in prolonged manipulation of neuronal networks and, as a result, behavior.

Long-term photo-activation of ASJ neurons during animal development
Lastly, we asked whether ChR2(C128S) could sufficiently activate neurons at a timescale of hours to days, e.g. cells relevant for animal development, under low light conditions, to prevent phototoxic effects. We thus turned to neurons that affect larval development. Depending on external signals, C. elegans larvae either develop to adulthood, or enter the dauer-state (Fig. 4a). In a simplistic view, but based on results from several studies, favorable external signals are sensed by ADF, ASG, and ASI neurons to prevent dauer-entry and to commit the worm to reproductive development [19,23]. Contrary, ASJ neurons, which can release insulin and possibly other signals to prevent dauerentry, may rather sense unfavorable cues like pheromones, and thus likely become inhibited [20]. Additionally, ASJ may be involved in dauer-exit, i.e. when conditions become favorable again, by releasing molecules that promote dauer-exit [19]. Unfavorable environmental signals appear to inhibit the guanylyl cyclase DAF-11 (which generates cGMP to activate the cGMPgated cation-channel TAX-2,-4), thus likely keeping ASJ in a resting or even hyperpolarized state and initiating dauer-arrest [20,[24][25][26] (Fig. 4a). daf-11(m84) mutants display a constitutive dauer-phenotype (daf-c): most larvae become dauers even under favorable conditions [24]. While additional mechanisms affecting dauer larval development need to be considered, (photo-)depolarization of ASJ neurons, at the right time during development, may nonetheless provide a means to prevent dauer-entry and to promote dauer-exit, particularly in daf-11(m84) mutants.
To allow cell-specific expression of ChR2(C128S) in ASJ neurons, we used the trx-1b promoter, which, however, expresses in ASJ cells only in the context of the genomic locus including the trx-1b coding sequences and introns [27]. We thus needed to fuse ChR2(C128S) to the C-terminus of TRX-1B. Expression of TRX-1B::ChR2(C128S) in ASJ was observed after hatching and through all developmental stages, including the dauer stage (Fig. S5). To verify that the ChR2 portion of the TRX-1B::ChR2(C128S) fusion protein is functional, we expressed it also in body wall muscle cells, where it could photo-evoke muscle contractions, albeit to a lesser extent than ChR2(C128S) alone (Fig. S1a, d). We also (over)expressed only TRX-1B in ASJ cells, to exclude lightindependent effects of TRX-1B on dauer-entry or -exit.
Finally, we investigated whether photoactivation of ASJ could also promote dauer-exit. daf-11 mutants, optionally expressing TRX-1B::ChR2(C128S) or TRX-1B alone in ASJ, were grown in the dark, in the presence or absence of ATR. Without lightstimulation, animals became dauers, and were only then placed into light, to potentially evoke acute dauer rescue. The fraction of animals developing to adults was scored over 4 days: 2-4% of daf-11(m84) dauers and daf-11(m84) dauers expressing TRX-1B recovered every 24 h, independent of ATR and light (Fig. 4c); yet, within the first 24 h of illumination, 16.462.1% of m84 dauers expressing TRX-1B::ChR2(C128S) in ASJ recovered if ATR was added prior to dauer-entry. At later times, the fraction of adults increased as slowly as for non-transgenic daf-11(m84) mutants. Thus, depolarizing ASJ indeed partially induces dauer-exit. However, as this was rather inefficient, additional cells may be required to trigger dauer-exit effectively. Alternatively, as dauers do not feed, ATR may have decayed after the time of dauer entry. We thus analyzed at distinct times after removal from ATR, to what extent ChR2(H134R) and ChR2(C128S) remained functional in muscle cells, by analyzing photo-evoked contractions (Fig. S6). While ChR2(H134R) remained fully functional for 24 h, and showed half maximal activity even after 72 h, ChR2(C128S) was surprisingly susceptible to ATR-deprivation: Already after 4 h, functionality was reduced by ,46%, and essentially non-detectable after 48 h (0.560.3% contraction; Fig.  S6). Importantly, ChR2(C128S) remained fully functional in the presence of ATR even after 72 h (8.760.6% contraction; Fig.  S6). In sum, ChR2(C128S) can be used to alter animal development when expressed in neurons that make developmental decisions, as these can be long-term depolarized using low light intensity.
Photoactivating command interneurons evoked long-term behavioral changes. Zheng et al. (1999) reported a largely increased frequency of ,40 reversals min 21 (the ''lurcher'' phenotype) after expressing constitutively active GLR-1(A687T) AMPARs in command neurons using the same glr-1 promoter fragment that we used. Thus, upon permanent strong depolarization, neither forward nor backward command neurons gain dominance, in line with mutual inhibition between the two neuron types. However, we observed a reversal right after photoactivation, often persisting for the whole minute, until yellow light closed ChR2(C128S). Regardless of photoactivation, animals exhibited ,3-4 reversals per minute. How may these opposing results be explained? Unlike ChR2(C128S), GLR-1(A687T) is expressed in its ''native environment'' (however, in the same cells as ChR2(C128S), as the same promoter was used in both studies), with a likely single-channel conductance in the low pS range, i.e. significantly higher than ChR2 (,40fS; [28]). Thus, stronger depolarization caused by GLR-1(A/T) might account for different behaviors seen in both experiments. Alternatively, GLR-1(A/T) causes depolarization of command neurons from its earliest expression, thus adaptation may occur, evoking different behaviors, while ChR2(C128S) is acutely induced by light during the  [15,19,20,[24][25][26]33]). Sensory neurons, like ASJ, mediate entry into, exit from, or repression of the dauer-state, in response to environmental signals. The molecular mechanisms in ASJ (lower panel, modified; [26]) are depicted. When depolarized, ASJ releases signals causing dauer-repression and dauer-exit. ASJ is depolarized via cGMP-gated TAX-2/-4 channels. Dauer-pheromones and possibly environmental signals (i.e. absence of food, high temperature) inhibit the guanylyl cyclase DAF-11, thus causing dauer-entry by blocking release of ASJ signals. experiment. To test this possibility, we photoactivated ChR2(C128S) in command neurons during development and until adulthood, but we found no emerging lurcher phenotype (data not shown). However, as ChR2(C128S) activity dropped to ,18% during long-term experiments (Fig. 1f), we may not achieve a long-lasting depolarization to the same extent as the GLR-1(A/T) channel did [21].
Photoactivation of TRX-1B::ChR2(C128S)::YFP in ASJ sufficed to depolarize these neurons for hours, allowing effective dauer-rescue of daf-11(m84) mutants. We verified that these effects are specific for the ChR2(C128S) portion of the fusion protein, and that TRX-1B alone had no effects; furthermore, as reported previously, mutation of trx-1 caused neither daf-c nor dafd phenotypes [27]. The photo-evoked dauer rescue was 76.462.7%, however, it was not 100%: This may indicate that ChR2(C128S)-induced depolarization was insufficient in some animals, or that additional cellular mechanisms affect dauer-entry in daf-11 mutants, which could not be overcome by ASJ photodepolarization; clearly the dauer developmental pathway involves many more cells expressing DAF-11 (e.g. ASI, ASK, AWB, and AWC) than just ASJ, and complex signaling pathways that may only inefficiently be triggered via simple depolarization of ASJ neurons. ASJ also promoted dauer-exit in a small, but significant fraction of animals during a 24 h photoactivation period (16.462.1%). Possibly, additional neurons, not photoactivated, need to cooperate with ASJ to promote dauer-exit more effectively; yet, more likely, ASJ was insufficiently depolarized due to progressive ChR2(C128S) inactivation, and due to the observed susceptibility of ChR2(C128S) to ATR-deprivation. Nevertheless, to prevent dauer-entry, ASJ is highly efficient on its own.
ChR2(C128S) has some critical properties that should be considered when designing experiments. One is the dependence on the continuous presence of ATR, the other is the partial inactivation by repeated or premature activation. For example, this makes ChR2(C128S) ill-suited for electrophysiological measurements in C. elegans, which require dissection of the animals under intense white light, which appears to render a majority of ChR2(C128S) to decay to ''lost'' states (hence the small photocurrents measured in Fig. 1c).
Nonetheless, ChR2(C128S) complements present optogenetic tools and expands their field of application, conceivably also in other animal models. Additional developmental pathways may now be probed, e.g. the likely activity-dependent polarity changes of C. elegans DD motorneurons during development [29]. Also adaptation or even associative learning within sensory circuits, which involves long-term neuronal activation [30], may be subjected to optogenetic manipulation using ChR2(C128S).

Behavioral experiments
Young adult animals were transferred to 5.5 cm dishes containing 4 ml nematode growth medium (NGM). Using an Axiovert 40 CFL microscope (Zeiss) with 106 magnification, 50 W mercury lamp, and computer-controlled shutter (Sutter Instruments), animals were illuminated with 450-490 nm blue light for ChR2 photoactivation and with 565-595 nm yellow light for ChR2 photoinactivation. Intensity was adjusted using neutral density filters (AHF Analysentechnik). For long-term photoactivation and inactivation, LEDs, blue (470 nm; 0.05 mW/mm 2 ; Luxeon) or yellow (590 nm; 0.04 mW/mm 2 ; Rapp Optoelectronic), respectively, were used. For body length measurements, videos were recorded (Powershot G5 or G9 digital cameras, Canon). Frames were extracted and either processed using a custom ImageJ script [32] or analyzed with a custom script for Matlab (The MathWorks) [5]. Unless described differently, animals were kept in complete darkness until execution of experiments to avoid unwanted photoactivation of ChR2. To avoid coiling induced by prolonged depolarization of cholinergic neurons [5], we analyzed effects of ChR2 activation in unc-49(e407) mutants, lacking GABA A Rs. Body length was normalized to the last second before illumination. Images yielding incorrect values for body length (e.g. coiling animals) were ignored. To monitor effects on dauer-entry, the following strains were cultivated for at least three days in the dark: daf-11(m84), N2, ZX852, ZX884, and ZX1034. Then, young adults were placed on seeded plates, optionally supplemented with ATR while plates were exposed to continuous illumination of two 18 W neon bulbs for three days (blue light intensity: 0.12 mW/mm 2 at the NGM agar surface). Animals were allowed to lay eggs for 10-12 h, and then removed. The fraction of adults and dauers (grown with or without ATR) was scored. To analyze dauer-exit, the following strains were cultivated on seeded plates with or without ATR for at least two days in the dark (to enlarge the fraction of dauer-animals): daf-11(m84), ZX884, and ZX1035. Dauer animals were then transferred to fresh plates, optionally supplemented with ATR and incubated under constant illumination (two 18 W neon bulbs; 0.12 mW/mm 2 blue light intensity). The fraction of adults was then scored daily.

Fluorescence microscopy and Electrophysiology
Expression of ChR2::YFP was analyzed on an Axiovert 200 microscope (Zeiss) with filterset F41-028 (AHF Analysentechnik) and 100 W mercury lamp. Images were captured with an AxioCam MRm camera (Zeiss). Expression in command interneurons and ASJ was analyzed on a Zeiss LSM 510 confocal microscope. Recordings from BWMs were conducted as previously described [5].

Statistics
Data are given as means6s.e.m. Significance between datasets is given as P-value after two-tailed Student's t-test.