(C) PLOS One [1]. This unaltered content originally appeared in journals.plosone.org.
Licensed under Creative Commons Attribution (CC BY) license.
url:https://journals.plos.org/plosone/s/licenses-and-copyright

------------



A high-throughput method to deliver targeted optogenetic stimulation to moving C. elegans populations

['Mochi Liu', 'Department Of Physics', 'Princeton University', 'Princeton', 'New Jersey', 'United States Of America', 'Sandeep Kumar', 'Princeton Neuroscience Institute', 'Anuj K. Sharma', 'Andrew M. Leifer']

Date: 2022-02 

We present a high-throughput optogenetic illumination system capable of simultaneous closed-loop light delivery to specified targets in populations of moving Caenorhabditis elegans. The instrument addresses 3 technical challenges: It delivers targeted illumination to specified regions of the animal’s body such as its head or tail; it automatically delivers stimuli triggered upon the animal’s behavior; and it achieves high throughput by targeting many animals simultaneously. The instrument was used to optogenetically probe the animal’s behavioral response to competing mechanosensory stimuli in the the anterior and posterior gentle touch receptor neurons. Responses to more than 43,418 stimulus events from a range of anterior–posterior intensity combinations were measured. The animal’s probability of sprinting forward in response to a mechanosensory stimulus depended on both the anterior and posterior stimulation intensity, while the probability of reversing depended primarily on the anterior stimulation intensity. We also probed the animal’s response to mechanosensory stimulation during the onset of turning, a relatively rare behavioral event, by delivering stimuli automatically when the animal began to turn. Using this closed-loop approach, over 9,700 stimulus events were delivered during turning onset at a rate of 9.2 events per worm hour, a greater than 25-fold increase in throughput compared to previous investigations. These measurements validate with greater statistical power previous findings that turning acts to gate mechanosensory evoked reversals. Compared to previous approaches, the current system offers targeted optogenetic stimulation to specific body regions or behaviors with many fold increases in throughput to better constrain quantitative models of sensorimotor processing.

Funding: Research reported in this work was supported by the Simons Foundation ( https://www.simonsfoundation.org/ ) under award SCGB 543003 to AML; and by the National Science Foundation ( https://www.nsf.gov/ ), through an NSF CAREER Award to AML (IOS-1845137) and through the Center for the Physics of Biological Function (PHY-1734030); and by the National Institute of Neurological Disorders and Stroke ( https://www.ninds.nih.gov/ ) of the National Institutes of Health under New Innovator award number DP2-NS116768 to AML. Strains from this work are being distributed by the C. elegans Genetics Center (CGC), which is funded by the NIH Office of Research Infrastructure Programs (P40 OD010440). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data Availability: Datasets from all recordings in this work are publicly available from the IEEE DataPorts repository DOI: 10.21227/t6b0-bc36 , https://dx.doi.org/10.21227/t6b0-bc36 . Machine readable numerical values associated with figures in this work are also provided in Excel spreadsheets listed in the Supporting Information section.

Copyright: © 2022 Liu et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Here, we demonstrate a closed-loop real-time targeting system for C. elegans that tackles all 3 challenges: targeting, closed-loop triggering, and throughput. The system delivers targeted illumination to specified parts of the animal’s body, stimulus delivery can be be triggered automatically on behavior, and the system achieves a high throughput by tracking and independently delivering targeted stimuli to populations of animals simultaneously. We apply this system to the C. elegans mechanosensory circuit [ 52 , 53 ] to characterize how competing stimuli in anterior and posterior mechanosensory neurons are integrated by downstream circuity. We also revisit our prior observation that turning behavior alters the animals likelihood of responding to mechanosensory stimuli [ 48 ]. We deliver closed-loop stimulation triggered to the onset of turning to obtain a dataset with 2 orders of magnitude more stimulus events compared to that investigation. We use these measurements to validate our initial observation with greater statistical power.

A recent system for zebrafish demonstrated the potential of closed-loop illumination in a multianimal setting [ 51 ]. That work probed the zebrafish optomotor response in high throughput by presenting patterned closed-loop visual stimuli, such as moving gratings, to many different animals simultaneously. That work also had an optogenetic component, but it relied on a low-resolution LED array that was restricted to open-loop full-field illumination and therefore could not target regions within animals or even target individual animals.

Targeted and closed-loop illumination systems probe one animal at a time [ 8 , 26 , 31 , 40 , 41 ] or, at most, 2 [ 42 ]. This low throughput poses challenges for acquiring datasets with enough statistical power to constrain quantitative models of neural computation. To increase throughput, a separate body of work simultaneously measures behavior of populations of many animals in an arena, in order to amass thousands of animal hours of recordings [ 43 – 47 ]. Delivering spatially uniform optogenetic perturbations to such populations has helped constrain quantitative models of sensorimotor processing of chemotaxis and mechanosensation [ 9 , 10 , 48 ]. Because the entire body of every animal is illuminated, this approach relies entirely on genetics for targeting. Recent work has used stochastic spatially varying illumination patterns in open loop from a projector [ 49 ] or cellphone [ 50 ] to probe the spatial dependence of optogenetic stimulation. But because these methods are open loop, they cannot target stimulation specifically to any animal or body part. Instead, they rely on after the fact inspection of where their stimuli landed, decreasing throughput.

Closed-loop approaches are further needed to deliver a perturbation timed to a specific behavior. For example, delivering a perturbation triggered to the swing of an animal’s head has informed our understanding of neural mechanisms underlying thermotaxis [ 32 ] and chemotaxis [ 8 ]. The use of closed-loop stimulation triggered on behavior in C. elegans [ 31 , 33 ], Drosophila [ 34 ], and mammals [ 35 , 36 ] is part of a broader trend in systems neuroscience toward more efficient exploration of the vast space of possible neural perturbations [ 37 , 38 ], especially during ethologically relevant naturalistic behaviors [ 39 ].

To stimulate only desired cells, the expression of optogenetic proteins is typically restricted to specific cells or cell types [ 21 – 23 ]. If cell-specific genetic drivers are not readily available, then genetic specificity can be complemented with optical targeting. Patterned light from a projector, for example, can be used to illuminate only a subset of the cells expressing the optogenetic protein [ 24 , 25 ]. For targeting behaving animals, real-time processing is also needed to track the animal and dynamically update the illumination pattern based on its movements [ 26 – 31 ].

Optogenetic investigations of neural circuits underlying behavior confront 3 technical challenges: the first is to deliver stimulation targeted only to the desired neuron or neurons; the second is to deliver the stimulus at the correct time in order to probe the circuit in a relevant state or behavioral context; and the third is to efficiently acquire enough observations of stimulus and response to draw meaningful conclusions. Existing methods each address some of these challenges, but not all 3.

How sensory signals are transformed into motor outputs is a fundamental question in systems neuroscience [ 1 ]. Optogenetics [ 2 , 3 ], coupled with automated measures of behavior [ 4 , 5 ], has emerged as a useful tool for probing sensorimotor processing, especially in small model organisms [ 6 ]. In optically transparent animals, such as Caenorhabditis elegans and Drosophila, such optogenetic manipulations can be performed noninvasively by illuminating an animal expressing light-gated ion channels [ 7 ]. Optogenetically perturbing neural activity and observing behavior has been widely used to study specific neural circuits, such as those involved in chemotaxis [ 8 – 11 ] (reviewed for Drosophila in [ 12 ]), olfaction [ 13 ], learning and memory [ 14 , 15 ], and locomotion and escape [ 16 – 19 ], to name just a few examples. In Drosophila, high-throughput optogenetic delivery to behaving animals has been used to screen libraries of neural cell types and map out previously unknown relations between neurons and behavior [ 20 ].

Results

We developed a closed-loop targeted delivery system to illuminate specified regions, such as the head or tail, in populations of C. elegans expressing the optogenetic protein Chrimson in the 6 gentle touch receptor neurons (AVM, ALML, ALMR, PVM, PLML, and PLMR) under a mec-4 promoter as the animals crawled on agar in a 9-cm dish. The system used red light (peak 630 nm) from a custom color projector made of an optical engine (Anhua M5NP) driven by a separate control board, described in Materials and methods. Animal behavior was simultaneously recorded from a CMOS camera (Fig 1A and 1B). Dark field illumination from a ring of infrared LEDs (peak emission at 850 nm) was used to observe animal behavior because this avoided exciting Chrimson. Optical filters allowed the infrared illuminated behavior images to reach the camera, but blocked red or blue light coming from the projector. Green light from the projector was used to present visual timestamps and other spatiotemoporal calibration information to the camera. Custom real-time computer vision software monitored the pose and behavior of each animal and generated patterned illumination restricted to only specified regions of the animal, such as its head or tail, optionally triggered by the animal’s behavior (Fig 1C, S1 Video).

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Closed-loop targeted optogenetic delivery system. (a) Schematic of system. A projector simultaneously delivers patterned targeted optogenetic stimulation to multiple worms on an agar plate in real time. (b) Photograph of instrument. (c) Image from experiment shows animals expressing Chrimson in touch receptor neurons (AML470) as they crawl on agar. Corresponding video is shown in S1 Video. Tracked paths are shown in yellow. Green and white dots in the center relate to a visual time stamping system and are excluded from analysis. Inset shows details of an animal receiving optogenetic stimulation targeted to its head and tail (0.5-mm diameter stimuli). The 2 white circle in the inset show the targeted location of the stimulus. Red shading shows area where stimulus was delivered. https://doi.org/10.1371/journal.pbio.3001524.g001

Integration of conflicting anterior and posterior mechanosensory signals Mechanosensory neurons act as inputs to downstream interneurons and motor neurons that translate mechanosensory signals into a behavioral response [19,53,56,62,63]. A growing body of evidence suggests that downstream circuitry relies on the magnitude of signals in both the anterior and posterior mechanosensory neurons to determine the behavioral response. For example, a plate tap activates both anterior and posterior mechanosensory neurons and usually elicits a reversal response [45,48,54,56]. But the animal’s response to tap can be biased toward one response or another by selectively killing specific touch neurons via laser ablation [56]. For example, if ALMR alone is ablated, the animal is more balanced in its response and is just as likely to respond to a tap by reversing as it is by sprinting. If both ALML and ALMR are ablated, the animal will then sprint the majority of the time [56]. Competing anterior–posterior optogenetic stimulation of the mechanosensory neurons also influences the animal’s behavioral response. For example, a higher intensity optogenetic stimulus to the anterior touch neurons is needed to evoke a reversal when the posterior touch neurons are also stimulated, compared to when no posterior stimulus is present [27]. To systematically characterize how anterior and posterior mechanosensory signals are integrated, we inspected the animal’s probability of transitioning into reverse, forward, pause/slow, or sprint behavior states in response to 25 different combinations of head and tail light intensity stimuli (Fig 3A–3D). These data are a superset of those shown in Fig 2. Here, 43,418 stimulus events are shown, corresponding to all 25 different conditions. Behavior is defined such that the animal always occupies 1 of 4 states: reverse, pause/slow, forward, or sprint, so that for a given stimulus condition, the 4 probabilities necessarily sum to one (Fig 3E). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 3. Behavioral response to competing stimulation of anterior and posterior mechanosensory neurons. Various combinations of light intensity was delivered to the head and tail of worms expressing Chrimson in gentle touch mechanosensory neurons (strain AML470, n = 43,418 stimulus events total, supserset of data shown in Fig 2). (a) Probability of transitioning into reverse, (b) pause/slow, (c) forward, and (d) sprint behaviors are shown individually (e) and all together as pie charts. (f) The gradient of the plane of best fit is shown as a vector for each behavior. Fitting was performed using methods of least squares. Error-bars are 95% confidence intervals. Numerical values are listed in S2 Data. https://doi.org/10.1371/journal.pbio.3001524.g003 To explore the dependence of the probability of evoking a given behavior on the anterior and posterior stimulus intensity, we fit planes to the probability surfaces and computed the gradient (Fig 3F). Fitting was performed using the method of least squares. The head and tail components of the gradient of the plane serve as a simplified linear approximation of the probability landscape and can provide a succinct estimate of how the probability depends on either head or tail stimulus illumination intensity. For example, the probability of reversal depends strongly on the head stimulus intensity, as indicated by the large positive head component of the gradient. The probability of reversal also depends slightly on the tail stimulus intensity, consistent with [27], but we note this dependence was small and that the 95% confidence intervals of this dependence spanned the zero value. Of the 4 behaviors tested, only sprint behavior depended on both head and tail stimulation intensity such that the 95% confidence intervals of both components of their gradient excluded the value zero. Sprints occurred with highest probability at the highest tail illumination intensity when head illumination was zero. As head illumination increased, the probability of a sprint rapidly decreased. This is captured quantitatively by the gradient plotted in Fig 3F, which shows that the ratio of the dependence of the sprint probability on tail versus head stimulation is roughly 1:−4. One interpretation is that head induced reversals are less likely to be counteracted by a tail stimulation, than tail induced sprints are to be counteracted by head stimulation. Taken together, we conclude that anterior and posterior mechanosensory signals are combined to determine the relative likelihood of different behavioral responses. Sprinting behavior is sensitive to both stimuli in a −4 to 1 ratio, while other behaviors such as reversals depend overwhelmingly more on one stimulus (head stimulation) than the other (tail stimulation). This places constraints on any quantitative models of sensory integration performed by downstream circuitry.

Behavior-triggered stimulation increases throughput when investigating rare behaviors C. elegans’ response to mechanosensory stimulation depends on its behavioral context. The probability that a tap or optogenetic mechanosensory stimulation evokes a reversal is higher when the stimulus occurs as the animal moves forward compared to if the stimulus occurs when the animal is in the midst of a turn, suggesting that the nervous system gates mechanosensory evoked responses during turns [48]. This was first observed in open-loop experiments in which the animal was stimulated irrespective of its behavior. Those experiments relied on identifying, after the fact, stimulus events that arrived by chance during turning. Because turning events are brief and occur infrequently, it can be challenging to observe sufficient numbers of stimuli events delivered during turn onset using such an open-loop approach. For example, in that work, only 15 optogenetic stimulus events for a given stimulus intensity condition landed during turns. The animal’s spontaneous rate of turn initiation varies with the presence of food and other environmental conditions, but it has been reported to be approximately 0.03 Hz [64]. To obtain higher efficiency and throughput at probing the animal’s response to stimulation during turning, we delivered closed-loop stimulation automatically triggered on the onset of the animal’s turning behavior. Full-body red light illumination (1.5-mm diameter) was automatically delivered to animals expressing Chrimson in the gentle touch mechanosensory neurons (strain AML67, same as in [48]) when the real-time tracking software detected that the animal entered a turn. Turn onset was detected by monitoring the ellipse ratio of a binary image of the animal, as described in Materials and methods. A refractory period of 30 seconds was imposed to prevent the same turning event from triggering multiple stimuli and served to set a minimum interstimulus interval. A total of 47 plates of animals were recorded for 30 minutes each over 3 days, and on average, the system simultaneously tracked or stimulated 44.5 ± 20 worms on a plate at any given time (S1 Table). Three different stimulus intensities (0.5, 40, and 80 uW/mm2) and 3 different stimulus durations (1, 3, and 5 seconds) were explored, totaling 22,608 turn-triggered stimuli events delivered to valid worms, of which on postprocessing analysis 9,776 or 43.2% passed our more stringent inclusion criteria for turn onset, worm validity, and track stability (Table 1). To compare the closed- and open-loop approaches, 29 additional plates were stimulated in open loop over the same 3-day period. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Table 1. Comparison of open- and closed-loop approaches for studying the animal’s response to stimulation during a turn. https://doi.org/10.1371/journal.pbio.3001524.t001 We compared the probability of reversing in response to closed-loop stimuli delivered during turn onset against the probability of reversing in response to open-loop stimuli delivered during forward locomotion (Fig 4A). For this analysis, we considered only stimuli of 3-second duration and either 80 uW/mm2 or 0.5 uW/mm2 (control) illumination intensity. A total of 2,999 stimulus events of this duration and intensities were delivered during turn onset and met our inclusion criteria. Consistent with previous reports, the animal was significantly more likely to initiate a reversal in response to stimuli delivered during forward locomotion than during turning. We repeated this experiment in strain AML470. That strain was also statistically significantly more likely to reverse in response to stimuli delivered during forward locomotion than during turning, in agreement with our prior findings. Interestingly, the effect was less striking in this strain compared to AML67 even though animals were overall more responsive. One possible interpretation is that turning-induced inhibition is in competition with mechanosensory signals. Because AML470 is apparently more sensitive to stimulation than AML67, it may also be more likely to generate stronger mechanosensory signals that overcome turning-induced inhibition, consistent with findings in Fig 4. By using a high-throughput closed-loop approach, we achieved larger sample size (2,999 events for a single strain, intensity, and control; compared to 15 [48]), confirmed previous findings, and refined our understanding of how turning-induced inhibition may compete with mechanosensory signals. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 4. Probability of reversing in response to a mechanosensory stimulus is higher for stimuli that arrive during forward locomotion than for stimuli that arrive during turning. Response to whole body optogenetic illumination of gentle touch mechanosensory neurons is shown for stimuli that arrive during either forward or turning behaviors for 2 strains of nominally identical genotypes, (a) AML67 and (b) AML470. Stimuli delivered during turns are from closed-loop optogenetic experiments, while stimuli delivered during forward locomotion are from open-loop experiments. Three seconds of 80 uW/mm2 illumination was delivered in the experiment condition. Only 0.5 uW/mm2 was delivered for control condition. Error bars are 95% confidence interval calculated via 10,000 bootstraps. Z-test was used to calculate significance. *** indicates p<0.001. p-Value for AML67 control group is 0.549. p-Value for AML470 control group is 0.026. The number of stimulus events for each condition (from left-most bar to right-most bar) are 5,968, 1,551, 5,971, and 1,448 for AML67 and 2,501, 1,543, 2,676, and 1,438 for AML470. Machine-readable numerical values are listed in S3 and S4 Data. https://doi.org/10.1371/journal.pbio.3001524.g004 Both throughput and efficiency are relevant for studying stimulus response during turning. Throughput refers to the number of stimuli delivered during turns per time. High throughput is needed to generate a large enough sample size in a reasonable enough amount time to draw statistically significant conclusions. Efficiency, or yield, refers to the fraction of delivered stimuli that were indeed delivered during turns. A high efficiency, or yield, is desired to avoid unwanted stimulation of the animal, which can lead to unnecessary habituation (S5 Fig). We compared the throughput and efficiency of stimulating during turn onset with closed-loop stimulation to an open-loop approach on the same instrument using our same analysis pipeline and inclusion criteria (Table 1). Again, we considered only stimuli delivered within a small 0.33-second window corresponding to our definition of the onset of turns and applied in postprocessing the same stringent inclusion criteria to both open-loop and closed-loop stimuli. Closed-loop stimulation achieved a throughput of 9.2 turn onset–associated stimulation events per worm hour, an order of magnitude greater than the 0.5 events per worm hour in open loop stimulation. Crucially, closed-loop stimulation was also more efficient, achieving a yield of 43.2%, more than 50-fold higher than the 0.7% open-loop yield. We reach similar conclusions by comparing to previous open-loop optogenetic experiments from [48] that had a longer interstimulus interval. Compared to that work, this system achieved a 25-fold increase in throughput and a more than 50-fold increase in yield. Taken together, by delivering stimuli triggered on turns in a closed-loop fashion, we achieved higher throughput and efficiency than open-loop approaches.

[END]

[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001524

(C) Plos One. "Accelerating the publication of peer-reviewed science."
Licensed under Creative Commons Attribution (CC BY 4.0)
URL: https://creativecommons.org/licenses/by/4.0/


via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/