SkinnerTrax: high-throughput behavior-dependent optogenetic stimulation ofDrosophila

Abstract

While red-shifted channelrhodopsin has been shown to be highly effective in activating CNS neurons in freely moving Drosophila, there were no existing high-throughput tools for closed-loop, behavior-dependent optogenetic stimulation of Drosophila. Here, we present SkinnerTrax to fill this void. SkinnerTrax stimulates individual flies promptly in response to their being at specific positions or performing specific actions. Importantly, SkinnerTrax was designed for and achieves significant throughput with simple and inexpensive components.peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/080614 doi: bioRxiv preprint first posted online Feb. 1, 2017; 3 Main textArtificially stimulating or inhibiting neurons is an important tool for unraveling the neural mechanisms that control animal behavior. A widely used method of stimulation is optogenetics, where channelrhodopsin is expressed in the neurons of interest, allowing them to be activated with high temporal resolution using light pulses. In Drosophila, targeted expression of channelrhodopsin is made easy by techniques like the Gal4-UAS system, and no surgery is needed to enable light delivery when using red-shifted channelrhodopsin 1, 2 as red light penetrates Drosophila sufficiently well. Several recent reports 3-6 have demonstrated the effectiveness of this approach in assessing whether activating specific neurons is sufficient to drive specific behaviors. A more sophisticated application of optogenetics, however, is to stimulate neurons of interest with a closed-loop system in real-time response to the fly's positions and actions. Such application is critical for elucidating the roles of specific neurons in several important behavioral tasks such as operant learning.Existing systems that allow such behavior-dependent stimulation have not been designed for high throughput. Two recent advanced systems, FlyMAD 7 and ALTOMS 8, 9 , target a laser to flies in real-time using mirror galvanometers, which in principle allows sharing the laser between several flies, as suggested in the discussion section of Bath et al. 7 But the scalability of sharing the laser is limited (Supplementary Note 1), and laser-targeting systems require expensive components and complex setups. Thus, for a typical lab laser-targeting systems likely make achieving high throughput prohibitively difficult.Here, we present SkinnerTrax, a system for high-throughput behavior-dependent optogenetic stimulation of Drosophila. Importantly, SkinnerTrax was designed for and achieves peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/080614 doi: bioRxiv preprint first posted online Feb. 1, 2017; 4 significant throughput with simple and inexpensive components. SkinnerTrax can deliver red light of different intensities and durations to individual flies p...

Learning a Spatial Task by Trial and Error in Drosophila

“…Further, flies' seeking optogenetic stimulation of their 0273-GAL4 neurons was substantially stronger than their seeking optogenetic stimulation of several groups of neurons that are known to mediate sugar reward or drive self-stimulation [9,12,13] ( Figure 1G). Lastly, flies still strongly pursued light, even if it simultaneously stimulated both the 0273-GAL4 neurons and the avoidance-promoting bitter-sensing taste neurons [14,15] (Figure 1G).…”

“…Conceptually, SkinnerTrax [15] has three custom-built components: (1) a tracker that processes the videos from multiple cameras and detects the positions of flies in real-time, (2) software that implements the rule as to which positions or actions to reward, and (3) a multi-channel light source that can deliver light pulses according to the software's instruction (Figure 2A). Both the tracker and reward rule components of SkinnerTrax are written in pure Python, but care was taken to perform all compute-intensive processing in fast libraries (primarily OpenCV), allowing our Python code to run at essentially the speed of efficient native code.…”

“…To calculate average heatmaps for several flies recorded by different cameras, we used the position of the arena floor in the frame (the positions varied significantly by camera) and the scaling factor for the floor (the floor sizes in the frames varied a little by camera). Both floor position and scaling factor are determined automatically by SkinnerTrax via template matching [15]. We overlaid some of the heatmaps on grayscale versions of the arena backgrounds to, e.g., make flies' use of spatial features of the arena easy to see.…”

Visually Guided Behavior and Optogenetically Induced Learning in Head-Fixed Flies Exploring a Virtual Landscape

“…Direct activation of sensory and dopaminergic neurons has also been used to replace reward or punishment in odor conditioning assays [34][35][36][37]. Further, freely moving flies were shown to avoid areas in which their bitter taste receptors were optogenetically activated, although it was not clear if flies could acquire a conditioned place preference as a result of such an experience [38]. Inspired by such studies, we developed behavioral paradigms to study adaptive, visually guided behavior in moderately complex 2D VR environments.…”

On the adaptive behavior of head-fixed flies navigating in two-dimensional, visual virtual reality

“…Direct activation of sensory and dopaminergic neurons has also been used to replace reward or punishment in odor conditioning assays [35-38]. Further, in a recent report, freely moving flies were shown to avoid areas in which their bitter taste receptors were optogenetically activated, although it was not clear if flies could acquire a conditioned place preference as a result of such experience [39]. Inspired by such studies we developed behavioral paradigms to study adaptive, visually-guided navigation in moderately complex 2D VR environments.…”