1. Neuroscience
Download icon

Closed-loop optogenetic activation of peripheral or central neurons modulates feeding in freely moving Drosophila

Research Advance
  • Cited 0
  • Views 1,231
  • Annotations
Cite this article as: eLife 2019;8:e45636 doi: 10.7554/eLife.45636

Abstract

Manipulating feeding circuits in freely moving animals is challenging, in part because the timing of sensory inputs is affected by the animal’s behavior. To address this challenge in Drosophila, we developed the Sip-Triggered Optogenetic Behavior Enclosure (‘STROBE’). The STROBE is a closed-looped system for real-time optogenetic activation of feeding flies, designed to evoke neural excitation coincident with food contact. We previously demonstrated the STROBE’s utility in probing the valence of fly sensory neurons (Jaeger et al., 2018). Here we provide a thorough characterization of the STROBE system, demonstrate that STROBE-driven behavior is modified by hunger and the presence of taste ligands, and find that mushroom body dopaminergic input neurons and their respective post-synaptic partners drive opposing feeding behaviors following activation. Together, these results establish the STROBE as a new tool for dissecting fly feeding circuits and suggest a role for mushroom body circuits in processing naïve taste responses.

https://doi.org/10.7554/eLife.45636.001

Introduction

Drosophila melanogaster has emerged as a leading model for understanding sensory processing related to food approach, avoidance, and consumption behaviors. However, although the gustatory system is recognized as mediating a critical final checkpoint in determining food suitability, much remains to be learned about the neural circuits that process taste information in the fly brain.

Like mammals, flies detect several taste modalities, each of which promotes food acceptance or rejection (Liman et al., 2014; Marella et al., 2006; Yarmolinsky et al., 2009). Taste compounds activate gustatory receptor neurons (GRNs) localized on the fly’s proboscis, legs, wings, and ovipositor (Scott, 2018). Among the different classes of GRNs present, cells expressing the Gustatory Receptor Gr64f respond to sweet compounds and induce strong acceptance behavior. Conversely, GRNs labeled by Gr66a respond to bitter compounds and evoke avoidance (Dahanukar et al., 2001; Dahanukar et al., 2007; Jiao et al., 2008; Kwon et al., 2014; Kwon et al., 2011; Marella et al., 2006; Thorne et al., 2004; Wang et al., 2004). GRNs connect directly or indirectly to the subesophageal zone (SEZ) of the fly brain (Ito et al., 2014; Rajashekhar and Singh, 1994; Scott, 2018; Stocker and Schorderet, 1981). Taste processing in the SEZ involves local modulatory interneurons (Chu et al., 2014; Pool et al., 2014), second-order neurons projecting locally or to other brain regions (Kain and Dahanukar, 2015; Kim et al., 2017; Yapici et al., 2016), motor neurons driving feeding subprograms (Gordon and Scott, 2009; Hampel et al., 2011; Manzo et al., 2012; Rajashekhar and Singh, 1994), and command neurons driving the complete feeding program (Flood et al., 2013).

Taste processing is not only involved in acute feeding events, but also in the formation of associative memories, which are aversive following exposure to bitter taste (Masek et al., 2015; Kirkhart and Scott, 2015) or positive following sugar consumption (Tempel et al., 1983). Memory formation occurs mainly in a central brain structure called the mushroom body (MB), composed of ~2000 Kenyon cells per hemisphere (Heisenberg et al., 1985). The MBs receive sensory information that is assigned a positive or negative output valence via coincident input from dopaminergic neurons (DANs) (Perisse et al., 2013; Waddell, 2010). Little is known about how taste information is relayed to the MBs, but taste projection neurons (TPNs) connected to bitter GRNs indirectly drive activation of the paired posterior lateral cluster 1 (PPL1) DANs (Kim et al., 2017). PPL1 neurons signal punishment to MBs and are required for aversive taste memory formation (Aso et al., 2012; Aso et al., 2010; Claridge-Chang et al., 2009; Kim et al., 2017; Kirkhart and Scott, 2015; Masek et al., 2015). Conversely, the protocerebrum anterior medial (PAM) cluster of DANs signals rewarding information and is involved in the formation of appetitive memories (Burke et al., 2012; Huetteroth et al., 2015; Liu et al., 2012; Yamagata et al., 2015). Although they have well-established roles in memory formation, PPL1 and PAM involvement in feeding has not been extensively investigated.

Kenyon cells and DANs make connections to specific mushroom body output neurons (MBONs) within discrete compartments of the MBs. MBONs project to protocerebral integration centers and are required for memory formation and retrieval (Aso et al., 2014a; Aso et al., 2014b; Aso et al., 2014b; Bouzaiane et al., 2015; Felsenberg et al., 2017; Ichinose et al., 2015; Masek et al., 2015; Owald et al., 2015; Perisse et al., 2016; Plaçais et al., 2013; Séjourné et al., 2011; Takemura et al., 2017; Tanaka et al., 2008). An emerging model is that DAN/MBON pairs innervating a specific MB compartment produce behavioral responses of opposing valence, and that KC-MBON synapses in that compartment are depressed upon DAN activation (Cohn et al., 2015; Felsenberg et al., 2017; Perisse et al., 2016; Séjourné et al., 2011; Takemura et al., 2017). While MBONs are known to modulate innate behaviors such as taste sensitivity (Masek et al., 2015) and food seeking behavior (Tsao et al., 2018), the possible contribution of MB input and output circuits to feeding behavior remains unclear.

Manipulating neural activity is a powerful method for assessing neural circuit function. Silencing neuron populations in freely behaving flies, which forces the neurons into a chronic ‘off’ state to mimic a situation where the fly never encounters an activating stimulus, is a straightforward way to determine their necessity in feeding. (Fischler et al., 2007; Gordon and Scott, 2009; LeDue et al., 2015; LeDue et al., 2016; Mann et al., 2013; Marella et al., 2012; Pool et al., 2014). However, gain-of-function experiments for feeding and taste, or any other actively sensed stimulus, are more complicated. Behaviors produced by forcing a neuron into a stimulus- and behavior-independent ‘on’ state can be difficult to interpret. The possible exception is activation of a neuron that elicits a stereotyped motor program, but even these situations are more easily interpreted in a tethered fly where the effect of a single activation can be monitored (Chen and Dahanukar, 2017; Flood et al., 2013; Gordon and Scott, 2009; Marella et al., 2006; Masek et al., 2015). To effectively probe the sufficiency of neuron activation during feeding events, it would be ideal to temporally couple activation with feeding.

Recently, three new systems for closed-loop optogenetic control of feeding flies have emerged: the optoPAD and STROBE, developed as additions to the FlyPAD system; and the optoFLIC, built on the FLIC platform (Jaeger et al., 2018; May et al., 2019; Moreira et al., 2019; Ro et al., 2014; Steck et al., 2018). Here, we provide a more extensive characterization of the STROBE and its utility. We demonstrate that coincident activation of sweet GRNs with feeding on agar drives appetitive behavior, and bitter GRN activation elicits aversion. These effects are modulated by starvation and can be inhibited by the presence of chemical taste ligands of the same modality. We also show that activation of central feeding circuit neurons produces repetitive, uncontrolled food interactions, demonstrating the STROBE’s efficacy in manipulating both peripheral and central neurons. We then establish that activation of PPL1 neurons negatively impacts feeding, while activating PAM neurons promotes it. Finally, in agreement with mushroom body circuit models, activating MBONs drives feeding responses in opposition to the DANs from the same MB compartment.

Results

The STROBE triggers light activation temporally coupled with food interactions

The FlyPAD produces capacitance signals that reflect a fly’s interaction with food in either of two sensors (or ‘channels’) in a small arena (Figure 1A, Figure 1—figure supplement 1A) (Itskov et al., 2014). When a fly physically bridges the two electrodes of a sensor by standing on one electrode and making contact with food sitting on the other, it produces a rise in capacitance. This signal, which is acquired at 100 Hz, is then decoded post hoc by an algorithm designed to identify sipping events. We designed the STROBE to track the raw capacitance signal in real-time and trigger lighting within the arena during sips (Figure 1B). To achieve this, we built arena attachments that consist of a lighting PCB carrying two LEDs of desired colors positioned above the channels of a FlyPAD arena. Each PCB is surrounded by a lightproof housing to isolate the arenas from other light sources (Figure 1—figure supplement 1B–C).

Figure 1 with 2 supplements see all
The STROBE setup.

(A) Concept of the FlyPAD: The interaction between the fly’s proboscis and the food is detected as a change in capacitance between two electrodes: electrode 1, on which the fly stands, and electrode 2, on which the food is placed. (B) Concept of the STROBE: when the fly is not interacting with the food, no change of capacitance is detected and the LED is OFF (left); when the fly sips, changes in capacitance turn the LED ON (right). (C) Flowchart of the STROBE signal processing algorithm. (D) Example of capacitance changes during a feeding bout, and the associated sips called by the FlyPAD (blue) and STROBE lighting events (red). (E) Comparison of the sip numbers called by the FlyPAD algorithm and LED illuminations triggered by STROBE algorithm. Sips/illuminations were counted in 1 min bins across a 1 hr experiment for 10 different channels (five arenas). Bins with neither sips detected by the FlyPAD algorithm nor illuminations triggered by the STROBE algorithm were excluded from analysis, as these were deemed times when the fly was not interacting with the food.

https://doi.org/10.7554/eLife.45636.002

In order to trigger optical stimulation with short latency upon sip initiation, we designed an algorithm that applies a running minima filter to the capacitance signal to detect when a fly is feeding. If the capacitance surpasses a set threshold above the minimum value recorded within the preceding 10 cycles (100 ms), the LED is turned on and remains illuminated until the threshold is no longer exceeded. By definition, this means that elevated plateaus of capacitance that last longer than 100 ms will produce a lighting response only within the first 100 ms. Because this algorithm is run on a field-programmable gate array (FPGA), the signal to lighting transition times are theoretically on the order of tens of milliseconds, providing a rapid response following the initiation of a sip. We confirmed the short latency of activation with a wire standing in for the fly’s proboscis. Based on video captured at 178 frames per second, the latency of LED activation following a touch was 37 ± 17 ms (Figure 1—figure supplement 2). This latency is short enough to ensure LED triggering during sips, which generally last longer than 100 ms (Itskov et al., 2014). Indeed, video of flies feeding in the STROBE, shot at 60 frames per second, consistently showed light activation in the same frame as the fly’s proboscis fully extending onto the food (Figure 1—figure supplement 2 and Video 1).

Video 1
This movie shows an individual fly feeding on the light-triggering food in a STROBE arena.
https://doi.org/10.7554/eLife.45636.006

The STROBE records the state of the lighting activation system (on/off) and transmits this information through USB to the PC, where it is received and interpreted by a custom end-user program. This program displays the capacitance signals from each fly arena in real-time, as well as its lighting state. It also counts the number of LED activations over the course of the experiment (Figure 1C; Figure 1—figure supplement 1D–F). To confirm that the STROBE algorithm triggers the LED during sips detected by the original post hoc FlyPAD algorithm, we first used both algorithms to analyze the capacitance signal from a short (~11 s) feeding bout (Figure 1D). Visually, this showed that each time a sip is detected with the FlyPAD algorithm, the STROBE algorithm triggers the LED at a similar time. However, we also noted that the STROBE algorithm triggers more LED activations than the number of sips called by the FlyPAD algorithm. We confirmed these observations on a larger scale by examining the correlation between the output of each algorithm in 1 min bins across a full 1 hr experiment (Figure 1E). Here, there is a strong correlation between the two (R2 = 0.963), with the number of LED illuminations triggered by the STROBE algorithm being about 1.4 times the number of sips detected by the FlyPAD algorithm. This increased number is likely the consequence of the FlyPAD algorithm filtering out capacitance changes not adhering to certain criteria of shape and duration (Itskov et al., 2014). Since these parameters are, by definition, unknown at sip onset, the STROBE cannot use them as criteria. Thus, we expect that a fraction of LED activations in the STROBE are actually triggered by more fleeting interactions with the food. Indeed, video of flies in the STROBE confirmed that a subset of leg touches triggered light activation (Video 1). Thus, we defined each LED activation as representing a food ‘interaction’, the majority of which are sips. Since flies detect tastes on multiple body parts, including the legs, even non-sip interactions are likely still relevant to taste processing and feeding initiation.

Activation of GRNs modifies feeding behavior

To validate the utility of the STROBE, we first tested flies expressing CsChrimson, a red-light activated channel, in either sweet or bitter GRNs (Klapoetke et al., 2014). Flies were given the choice between two identical neutral food options (plain agar), one of which triggered light activation. Under these conditions, flies expressing functional CsChrimson in sweet neurons under the control of Gr64f-GAL4 showed a dramatic preference towards feeding on the light-triggering food, while control flies of the same genotype that were not pre-fed all-trans-retinal, and thus carried non-functional CsChrimson, displayed no preference (Figure 2A–E). The number of interactions on the light-triggering side of the chamber was dependent on light intensity, with increasing interactions up 6.5 mW/cm2, above which interaction numbers decreased as a function of intensity. However, preference was relatively stable above a threshold of 1.85 mW/cm2 (Figure 2D–E). Additionally, interactions generally accumulated linearly over time at all intensities, with a stable preference index established within the first 15 min of a one-hour experiment (Figure 2—figure supplement 1). This suggests that neither sensitization nor adaptation occurs during the course of the experiment. Moreover, the flies do not appear to adjust their behavior in response to the perceived mismatch between the sweet taste and lack of energy content of the food source.

Figure 2 with 4 supplements see all
Activation of sweet and bitter sensory neurons drives feeding preferences in the STROBE.

(A) Immunofluorescent detection of UAS-CsChrimson.mVenus driven by Gr64f-GAL4. (B) Schematic illustrating STROBE activation of sweet neurons. (C) Experimental setup: both channels are filled with 1% agar, only one is paired to LED activation. (D–E) Relationship between light intensity and light side preference (D) or interaction numbers (E) for Gr64f > CsChrimson flies pre-fed retinal (blue squares) or not fed retinal (black squares). (F) Expression of UAS-CsChrimson.mVenus driven by Gr66a-GAL4. (G) Schematic illustrating STROBE activation of bitter neurons. (H) Experimental setup: both channels contain plain 1% agar. (I–J) Relationship between light intensity and light side preference (I) or interaction numbers (J) Gr66a > CsChrimson flies pre-fed retinal (green squares) or not fed retinal (black squares). Values represent mean ± SEM. n = 30–37 (D–E) or 19–28 (H–I). Statistical tests: two-way ANOVA and Bonferroni post hoc: ns p>0.05, **p<0.01, ***p<0.001. Colored asterisks represent significance between sips on each side for the retinal group.

https://doi.org/10.7554/eLife.45636.008

As expected, flies expressing functional CsChrimson in bitter sensing neurons under the control of Gr66a-GAL4 strongly avoided neuronal activation in the STROBE by engaging in fewer interactions with the light-triggering food source (Figure 2F–J). Once again the behavioral response was intensity-dependent, with maximum suppression of interactions occurring at the highest intensity tested (16.4 mW/cm2). As with sweet GRN activation, the flies’ preference develops in roughly the first 15 min and remains relatively stable for the rest of the hour (Figure 2—figure supplement 2). Since the light intensities eliciting maximal effect for Gr64f and the Gr66a activation are 6.5 mW/cm2 and 16.4 mW/cm2, respectively, we decided to use the intermediate value of 11.2 mW/cm2 as the intensity for all further experiments. Using a full set of genotypic and non-retinal controls, we confirmed that these conditions produced robust and specific preference behaviors for both sweet and bitter GRN activation (Figure 2—figure supplement 3).

Next, we sought to test whether GRN activation affects feeding per se, rather than simply driving non-ingestive food interactions. We repeated sweet and bitter GRN activation in the STROBE with the addition of blue dye to one food source and red dye to the other. We then calculated a post hoc preference based on the number of individual flies showing blue versus red dye in their abdomen following the assay, and compared this to the preference calculated based on food interactions measured by the STROBE. In all cases, these studies confirmed that the changes in food interactions driven by GRN activation are accompanied by strong effects on food ingestion in the expected direction (Figure 2—figure supplement 4).

Starvation modulates the behavioral impact of GRN activation

Starvation duration has a well-known impact on fly feeding – the longer flies are food deprived, the more they will initiate and sustain feeding on sweet foods (Dus et al., 2013; Dus et al., 2011; Inagaki et al., 2012; Inagaki et al., 2014; Scheiner et al., 2004). To determine if similar effects would manifest in the STROBE, we tested GRN activation after different periods of food deprivation (Figure 3A,B). Consistent with its effect on sugar feeding, starvation increased flies’ preference for the light-triggering food in the STROBE when their sweet neurons expressed functional CsChrimson (Figure 3C). This elevated preference index is driven by a dramatic increase in interaction number (Figure 3D). In contrast to its impact on sweet sensory neurons, starvation had no significant effect on light avoidance mediated by bitter neuron activation under these conditions (Figure 3E,F). Starvation also had little to no effect on the timing of food choice throughout the assay, with all groups establishing their peak preference within the first 10–15 min and largely maintaining it throughout the assay (Figure 3—figure supplements 1 and 2).

Figure 3 with 2 supplements see all
Behavioral impact of GRN activation is modulated by starvation.

(A) Protocol: flies are subjected to increasing period of starvation (12 hr, 24 hr, 48 hr) prior to the STROBE experiment. (B) Experimental setup: both channels contain plain 1% agar. (C–D) The effect of starvation on light side preference (C) and food interaction numbers (D) of flies expressing CsChrimson in sweet neurons. (E–F) The effect of starvation on preference for the light side (E) and interaction numbers (F) of flies expressing CsChrimson in bitter neurons. Values represent mean ± SEM. n = 21–30. Statistical tests: two-way ANOVA with Bonferroni post hoc. ns p>0.05, *p<0.05, **p<0.01, ***p<0.001. Colored asterisks represent significance between sips on each side for the retinal group.

https://doi.org/10.7554/eLife.45636.014

Chemical taste ligands suppress the impact of light-induced attraction and avoidance

We next asked whether the presence of sweet or bitter ligands would interfere with light-driven behavior in the STROBE. For example, if sugar is placed in both food options, will this reduce the salience of sweet GRN activation by light? Indeed, adding increasing concentrations of sucrose to both food options caused dose-dependent inhibition of GR64f > CsChrimson flies’ preference for the light-triggering food (Figure 4A,B). This change is driven by a progressively higher number of interactions on the no-light side, with relatively constant interaction numbers on the light side as sucrose concentration increases (Figure 4—figure supplement 1A,B). On the other hand, the addition of sucrose mildly enhanced the negative preferences driven by STROBE activation of Gr66a bitter neurons (Figure 4C). This effect appears to manifest from the increasing attractiveness of the no-light side coupled with unwavering and near total avoidance of the light-triggering side (Figure 4—figure supplement 1C). The same pattern is mirrored by the addition of the bitter compound denatonium to both sides: dose-dependent inhibition of the aversion shown by Gr66a > CsChrimson flies (Figure 4D,E; Figure 4—figure supplement 1D,E), and little to no effect on the attraction of Gr64f > CsChrimson flies to the light side (Figure 4F; Figure 4—figure supplement 1F). In general, as with other experiments, we observed no substantial change in preference after establishment in the first 10–15 min (Figure 4—figure supplements 25). This suggests that flies are not displaying satiety effects from sugar ingestion or taste-independent effects from the consumption of denatonium.

Figure 4 with 5 supplements see all
Chemical taste ligands suppress impact of light evoked GRN activity.

(A) Experimental setup: both channels contain the same sucrose concentration (1, 10, 100, or 1000 mM) in 1% agar. (B–C) The effect of sucrose concentration on the light side preference of Gr64f > CsChrimson (B) or Gr66a > CsChrimson (C). (D) Experimental setup: both channels contain the same denatonium concentration (0, 0.1, 1, or 10 mM). For Gr66a > CsChrimson activation, both channels also contain 50 mM sucrose. (E–F) The effect of denatonium concentration on the light side preference of Gr66a > CsChrimson (E) or Gr64f > CsChrimson (F). Values represent mean ± SEM. n = 25–51. Statistical tests: two-way ANOVA with Bonferroni post hoc. ns p>0.05; ***p<0.001.

https://doi.org/10.7554/eLife.45636.018

Activation of the ‘feeding-neuron’ drives extreme sipping behavior

Can the STROBE affect feeding behavior through the activation of central neurons, in addition to those in the periphery? Although the precise nature of higher-order taste circuits is still unclear, several neurons have been identified in the SEZ that influence feeding behavior (Chu et al., 2014; Inagaki et al., 2014; Inagaki et al., 2012; Jourjine et al., 2016; Kain and Dahanukar, 2015; LeDue et al., 2016; Marella et al., 2012; Pool et al., 2014; Yapici et al., 2016). One of them, the ‘feeding-neuron’ (Fdg), acts as a command neuron for the proboscis extension response, and shows activity in response to food stimulation only following starvation (Flood et al., 2013). Strikingly, Fdg activation in the STROBE produced an extremely high number of interactions on the light-triggering food, resulting in a nearly complete preference for that side (Figure 5Video 2). Thus, the STROBE can effectively modulate feeding behavior via the activation of either peripheral or central neurons.

Activation of feeding command neurons elicits extreme sipping behavior.

(A) Immunofluorescent detection of Fdg > CsChrimson. (B) Experimental setup: both channels contain either 1% agar or 100 mM sucrose. (C) Cumulative food interaction numbers for the population of Fdg > Chrimson flies over the course of a 1 hr experiment. (D) Total interaction numbers for individual flies. (E–F) Preference index for Fdg > Chrimson flies over the course of a 1 hr experiment averaged (E) or for ten individual flies (F). (G–H) Preference index for Fdg > Chrimson flies after one hour spent in the STROBE with agar (G) or 100 mM sucrose (H) . Values are mean ± SEM. n = 14–33. Statistical tests: t-test. ***p<0.001.

https://doi.org/10.7554/eLife.45636.025
Video 2
This movie shows the full STROBE system.

Each chamber contains a fly expressing CsChrimson under the control of Fdg-GAL4. The flies on the left side have been fed all-trans retinal. The flies on the right have not been fed retinal.

https://doi.org/10.7554/eLife.45636.007

Manipulating mushroom body extrinsic neurons modifies feeding behavior

Given the important role of the MB in assigning valence to stimuli during learning, we next asked whether MB circuits acutely impact attractive and aversive feeding responses. PPL1 DANs signal punishment or aversive information to the MBs, and thus their activation in the STROBE is predicted to drive avoidance of the light-triggering food (Figure 6A) (Aso et al., 2012; Aso et al., 2010; Das et al., 2014; Kirkhart and Scott, 2015; Masek et al., 2015). Flies expressing functional CsChrimson in the α3/α’3 subset of PPL1 dopaminergic neurons (MB308B-GAL4) showed a negative preference towards light when 100 mM sucrose was present in both options, but not if the food was plain agar (Figure 6C; Figure 6—figure supplement 1A–C). Performing the same experiment with activation of a broader subset of PPL1 neurons (MB065B-GAL4) led to stronger avoidance in the presence of 100 mM sucrose or plain agar (Figure 6—figure supplement 1A–C). Interestingly, activation of MBONs post-synaptic to PPL1 (MB093C-GAL4 and MB026B-GAL4) produced strong attraction in either context (Figure 6D; Figure 6—figure supplement 1A–C).

Figure 6 with 3 supplements see all
Manipulation of mushroom body extrinsic neurons modifies feeding behavior.

(A) Model for PPL1 input to the MB and corresponding output: PPL1 neurons signal punishment and are predicted to drive aversive behavior, while corresponding MBONs are predicted to be appetitive. (B) Experimental setup: both channels contain 100 mM sucrose in 1% agar. (C) Light side preference and interactions of flies expressing CsChrimson in PPL1 neurons α3,α’3 (MB308-GAL4). (D) Light side preference and interactions of flies expressing CsChrimson in MBON α3,α’two neurons (MB093C-GAL4). (E) Model for PAM input to the MB and corresponding output: PAM neurons signal reward and are predicted to drive appetitive behavior while corresponding MBONs are predicted to be aversive. (F) Experimental setup: both channels contain 100 mM sucrose in 1% agar. (G) Light side preference and interactions of flies expressing CsChrimson in PAM β2,β’two neurons (MB056B-GAL4). (H) Light side preference and interactions of flies expressing CsChrimson in MBON neurons post-synaptic to PAM β2,β’two neurons (MB011B-GAL4). (I) Light side preference and interactions of flies expressing CsChrimson in PAM γ3 neurons (MB441-GAL4). (J) Light side preference and interactions of flies expressing CsChrimson in MBON neurons post-synaptic to PAM γ3 neurons (MB083C-GAL4). Values are mean ± SEM. n = 18–29. Statistical test: t-test. **p<0.01, ***p<0.001.

https://doi.org/10.7554/eLife.45636.027

The PAM cluster of DANs is generally thought to signal appetitive reward to the MBs (Burke et al., 2012; Huetteroth et al., 2015; Lin et al., 2014; Liu et al., 2012; Yamagata et al., 2015). Following the same principle described above, PAM activation should drive appetitive behavior, while stimulation of MBONs within the same compartment is predicted to elicit aversion (Figure 6E). Indeed, activating the β2m,β’2 p PAM subset (MB056B-GAL4 and MB301B-GAL4) in the STROBE led to attraction (Figure 6F,G; Figure 6—figure supplement 2). On the other hand, activating the corresponding β2m,β’2 p,γ5 MBONs (MB011B-GAL4 and MB210B-GAL4) produced avoidance when sucrose was present (Figure 6H, Figure 6—figure supplement 2).

Another subset of PAMs, targeting the γ3 compartment, was recently shown to encode a negative valence and induce appetitive memories following transient inhibition by the satiety peptide Allatostatin-A (Yamagata et al., 2016). Consistent with these results, activation of PAM γ3 neurons (MB441B-GAL4 and MB195B-GAL4) in the STROBE produced light avoidance (Figure 6I; Figure 6—figure supplement 3), while activating the corresponding β’1,γ3 MBONs (MB083C-GAL4 and MB110C-GAL4) was attractive (Figure 6J; Figure 6—figure supplement 3). Thus, PAMs targeting different MB compartments can be either attractive or aversive in the context of feeding.

Finally, we chose the PPL1 cluster and their corresponding MBONs to examine the impact of MB circuits on feeding in greater detail. First, we asked whether the interaction preferences driven by activating these populations reflected genuine changes in feeding behavior. Strikingly, dye ingestion in the STROBE consistently produced stronger measures of preference than those calculated with interaction numbers, mirroring the effects seen with GRN activation (Figure 7). This was true across a range of stimulus intensities, even when the differences in interaction numbers were quite low. Interestingly, in contrast to sweet GRN stimulation, which elicited peak interaction numbers midway in the series of intensities tested, MB026B-GAL4 MBON activation produced linearly increasing interactions across the range of intensities (Figure 7H). Second, given the important role that PPL1s play in aversive conditioning, we examined whether activation of either population caused a shift in behavior over the course of the one-hour assay. However, as with GRN activation, relative interaction numbers remained quite stable throughout (Figure 7—figure supplements 1 and 2). Third, we tested whether silencing each population with expression of the inward rectifier potassium channel Kir2.1 would affect feeding behavior. After disabling LED activation in the STROBE, we counted flies’ interactions with either a low concentration of sucrose or plain agar (Figure 7—figure supplement 3). This failed to reveal an effect of PPL1 or MBON silencing on food choice, indicating that, while activation of these populations is sufficient to affect food choice, they are not necessary in the particular task tested.

Figure 7 with 3 supplements see all
Activation of mushroom body circuits drives ingestion behavior.

(A) Schematic of the PPL1 driver used. (B) Experimental setup: one channel contains 1% agar with blue dye, and the other has 1% agar with red dye. Fly abdomen color is determined following the experiment. (C) Food interactions from flies expressing CsChrimson in PPL1 DANs in the STROBE with increasing light intensity. Green lines indicate flies pre-fed retinal, and black/gray lines are non-retinal controls. (D–E) Light side preference indices calculated using interactions (D) or dye ingestion (E). (F) Schematic of the MBON driver used. (G) Experimental setup: one channel contains 1% agar with blue dye, and the other has 1% agar with red dye. Fly abdomen color is determined following the experiment. (H) Food interactions from flies expressing CsChrimson under control of MB026B-GAL4 in the STROBE with increasing light intensity. Blue lines indicate flies pre-fed retinal, and black/gray lines indicate non-retinal controls. (I–J) Light side preference indices calculated using interactions (I) or dye ingestion (J). Values are mean ± SEM. n = 19–30 for (C–E) and 14–27 for (H–J). Statistical test: two-way ANOVA with Bonferroni post hoc. ns p>0.05, *p<0.05, **p<0.01, ***p<0.001. Colored asterisks represent significance between sips on each side for the retinal group.

https://doi.org/10.7554/eLife.45636.032

Discussion

Leveraging real-time data from the FlyPAD, we built the STROBE to tightly couple LED lighting with sipping and other food interactions, thereby allowing us to optogenetically excite specific neurons during feeding. The primary advantage of the STROBE over existing systems for neural activation during fly feeding is its temporal resolution, which provides two key benefits. First, acutely activating neurons while the fly is choosing to interact with one of two available food sources allows us to explore the impact of neural activation on food selection in a way that is impossible with chronic activation. Second, by tightly coupling stimulation with food interaction events, light-driven activity from the STROBE should more closely mimic the temporal dynamics of taste input. Conceptually, these advantages are similar to those achieved by expression of the mammalian TRPV1 in taste sensory neurons and lacing food with capsaicin (Caterina et al., 1997; Chen and Dahanukar, 2017; Marella et al., 2006). Importantly, however, the STROBE allows activation of either peripheral or central neurons.

The implementation of interaction detection and light triggering on an FPGA allows the STROBE to trigger LED activation with minimal latency, well within the time frame of a single sip. Thus, neural excitation is tightly locked to the onset and offset of food interactions, providing the means to manipulate circuits during active feeding. Our decision to implement an algorithm that terminates illumination during capacitance plateaus exceeding 100 ms has both advantages and disadvantages. The main advantage is avoiding the LED becoming ‘stuck’ in the on state during shifts in baseline capacitance that could result in constant illumination for seconds or even minutes during which the fly may not be interacting with the food. A possible disadvantage is LED illumination that does not last the entirety of an interaction. This could produce insufficient activation to mimic specific properties of long sips. It could also result in suboptimal light-mediated silencing, although the efficacy of neural inhibition in the STROBE has not been tested.

These temporal qualities of light triggering are the primary difference between the STROBE and another recently described optogenetic FlyPAD, termed ‘optoPAD’ (Moreira et al., 2019; Steck et al., 2018). OptoPAD carries out sip detection and light control on a USB-connected computer. The benefits of this strategy are the ability to implement a more complex feeding detection algorithm, and more flexible control of the lighting response timing (Moreira et al., 2019). However, the tradeoff is longer and more variable latencies of LED activation. Moreover, the set illumination period further decouples the timing of illumination from the fly’s behavior. Each of these systems may have specific advantages, depending on the application. While they have not been directly compared, it is likely that tight temporal coupling of the STROBE to food contact will be more useful for studying the effects of acutely activating core taste and feeding circuit neurons, while the longer, adjustable, light pulse from the optogenetic FlyPAD may be better for silencing neurons or activating reinforcement circuits.

Interestingly, similar closed-loop optogenetic paradigms have recently been developed for rodents. Lick-triggered blue light stimulation of taste receptor cells or circuits in the amygdala is sufficient to drive appetitive and aversive taste behaviors in mice (Wang et al., 2018; Zocchi et al., 2017). Thus, the same principle is able to reveal important insight into consummatory behaviors in multiple animals.

Although optogenetic neuronal activation is artificial, light-driven behavior in the STROBE shows some important properties that mimic natural feeding. For example, the number of food interactions evoked by stimulation of sweet sensory neurons increased in response to starvation, demonstrating a dependence on internal state that mirrors what is seen for sugar feeding (Dus et al., 2013; Dus et al., 2011; Inagaki et al., 2014; Inagaki et al., 2012; Scheiner et al., 2004). One curious observation is that the number of interactions driven by activation of sweet neurons peaks at a sub-maximal stimulus intensity, with fewer interactions observed at higher levels of illumination. Compensatory decreases in feeding are known to occur at high food concentrations, but this is likely from consumption of nutrients that are not present in the STROBE experiment (Deshpande et al., 2014). Since we observe no evidence of sensory-specific satiety over the time course of our experiments, we suspect that suppression of interactions may occur either from non-physiological neuron responses at high intensities or an interaction between visual cues from the increased light intensity and the sweet taste-mediated attraction. Regardless of the mechanism, the non-linear intensity-dependence of sweet neuron activation contrasts with that of bitter neurons and the MB neurons tested, which all show the largest behavioral responses at the highest light intensity. This demonstrates the important practical point that different neuron types can produce maximal effects at different stimulus intensities. Thus, it will be important for anyone using the STROBE or a similar optogenetic system to carefully calibrate the light intensity for their specific neurons of interest with the goal of matching either the neurons’ physiological responses, or the animals’ behavioral responses, to natural stimuli.

We also showed that the behavioral impact of sweet and bitter GRN activation in the STROBE could be abolished by the presence of natural taste ligands. Interestingly, this property did not generally hold true for attraction mediated by PAM or appetitive MBON activation, which was typically similar in the presence or absence of sugar. This may suggest that sweet taste input and PAM or MBON activation drive attraction via parallel circuits, producing an additive effect when both are present. Or perhaps suppression would be observed at higher sucrose concentrations or lower light intensities. It is also notable that flies preferred 1 M sucrose alone over 1 M sucrose coupled to optogenetic activation of sweet GRNs. We suspect that optogenetic activation of sweet GRNs in the STROBE plateaus below the excitation achieved with 1 M sucrose, and somehow prevents further activation by very high sugar concentrations. We favor this interpretation over the alternative that the combination of 1 M sucrose with optogenetic activation of sweet neurons becomes ‘too sweet’; however, both remain formally possible.

One interesting question is whether the valence of GRN activation in the STROBE is mediated by hedonics or effects on the feeding program itself. For example, sweet neuron activation is thought to carry appetitive hedonics, and therefore the flies may continue feeding because consequent light activation of Gr64f neurons is somehow pleasurable. On the other hand, these neurons also initiate feeding (and conversely, activation of Gr66a neurons terminates it). Thus, it is possible that each food contact evokes light-driven activation of a subsequent contact, and so on, creating a positive feedback loop. This is undoubtedly true of Fdg neuron activation, which is known to initiate a complete feeding sequence, likely downstream of any hedonic effects (Flood et al., 2013). Flies appear to become ‘trapped’ in a feeding loop until the end of the experiment, suggested by the very high number of evoked food interactions (see Figure 5).

Activation of MB input and output neurons also modulates feeding in the STROBE. PPL1 stimulation in the STROBE produces avoidance, while appetitive PAM stimulation produces attraction, consistent with the established valence of each population in memory formation and a previously reported role for PAMs in foraging behavior (Aso et al., 2012; Burke et al., 2012; Aso et al., 2010; Das et al., 2014; Huetteroth et al., 2015; Kirkhart and Scott, 2015; Landayan et al., 2018Lin et al., 2014; Liu et al., 2012; Masek et al., 2015; Yamagata et al., 2015). Moreover, MBON activation drives feeding behavior in the opposite direction to activation of DANs from the same compartment. This relationship supports the current model that DAN activity depresses KC to MBON synapses in their respective compartments (Cohn et al., 2015; Felsenberg et al., 2017; Perisse et al., 2016; Séjourné et al., 2011; Takemura et al., 2017). Interestingly, not all PAM neurons convey a positive signal upon activation. PAM γ3 neurons are excited by electric shocks (Cohn et al., 2015) and inhibited by sucrose stimulation (Cohn et al., 2015; Yamagata et al., 2016). Our findings that PAM γ3 activation drives aversive feeding behavior is consistent with these neurons signaling negative valence to the MBs. Finally, it is worth noting that, although silencing experiments failed to reveal any requirement for PPL1s and their post-synaptic MBONs in the feeding paradigm used, it is possible that effects would be seen with silencing of broader neuron populations.

Although the mechanisms by which DANs affect feeding behavior remain unclear, MBONs can modulate innate behavior such as taste sensitivity (Masek et al., 2015), naïve response to odors (Owald et al., 2015), place preference (Aso et al., 2014b), and food seeking behavior (Tsao et al., 2018). Could the modulation of feeding by DAN activation result from learning? This seems unlikely with our current experimental design, as both food options were always identical, and thus there would be no predictive cues to associate with appetitive or aversive DAN stimulation. We think it is more likely that the same reward or punishment signals that underlie memory formation also acutely modify feeding behavior. However, the possibility of pairing circuit activation with specific food cues may offer a new paradigm for studying food memories, and neuronal activation via self-administration opens new avenues for the study of operant conditioning and addiction.

Materials and methods

STROBE system

Request a detailed protocol

The STROBE system consists of a field programmable gate array (FPGA) controller attached to a multiplexor board, adaptor boards, fly arenas equipped with capacitive sensors and lighting circuits. The hardware, with the exception of the lighting circuit units, is based on the FlyPAD design (Itskov et al., 2014). Each fly arena is paired with a lighting circuit and an opaque curtain (to prevent interference from external light). This pair will be referred to as a fly chamber unit. The entire system accommodates 16 fly chamber units (16 fly arenas and 16 lighting circuits), through eight adaptor boards. The FPGA used is a Terasic DEV0-Nano mounted onto a custom-made multiplexor board.

The multiplexor board is one of the intermediate connection components between the fly chambers and the FPGA controller. The multiplexor board has eight 10-pin ports each of which facilitate communications between two fly chambers and the FPGA controller. The board also has a FTDI module allowing data transfer over serial communications with a computer. The other intermediate connection component is the adaptor board which connects on one side to the multiplexor board via a 10-pin line, and splits the 10-pin line from the multiplexor board into four 10-pin ports which connects to two fly arenas and two lighting circuits. The fly arena consists of two annulus shaped capacitive sensors and a Capacitance to Digital Converter (CDC) chip (AD7150BRMZ) that the main multiplexer board communicates with to initiate and collect data (and ultimately to stop collecting data). The CDC interprets and converts capacitance data from the two sensors on the fly-arena to a digital signal for the FPGA to process (Itskov et al., 2014).

The lighting circuit consists of a two-pin connector to receive power from an external power supply, a 10-pin connector to receive signals from the FPGA controller via the intermediate components, a 617 nm light emitting diode (LUXEON Rebel LED – 127lm @ 700mA; Luxeon Star LEDs #LXM2-PH01-0060), two power resistors (TE Connectivity Passive Product SMW24R7JT) for LED current protection, and two metal oxide semiconductor field effect transistors (MOSFETs; from Infineon Technologies, Neubiberg, Germany, IRLML0060TRPBF) allowing for voltage signal switching of the LEDs.

When a fly performs a sip and triggers a high signal on a capacitive sensor, the CDC chip on the fly arena propagates a signal via the multiplexor to the FPGA controller. The FPGA processes the capacitive sensor signal, decides a legitimate food interaction took place, and sends a high signal through the multiplexor to the MOSFET of the lighting circuit. The MOSFET then switches its lighting circuit on, allowing current to flow and turning on the monocolor LED positioned directly above the capacitive sensor. The process for determining a legitimate food interaction is described next.

In order to trigger optical stimulation with short latency upon food contact, we designed a running minima filter that operates in real-time to detect when a fly is feeding. We implemented this filter by modifying the state machine on the FPGA. When a fly feeds, its contact with the capacitance plate generates a ‘step’, or rising edge in the capacitance signal. Our filter determines the minimum signal value in the last 100 ms and checks whether the current signal value exceeds that minimum by a set threshold. This threshold (100 capacitance units) was selected to be large enough to discriminate rising edges of capacitance representing true food interactions from noise, but small enough to not miss true interactions. If this condition is true, the filter will prompt the lighting activation system to activate the LED (or keep it on if it is already on).

By design, this means that the control system will send a signal to deactivate the lighting upon the falling edge of the capacitance signal, or if the capacitance signal has plateaued for 100 ms, whichever comes sooner. At this point, a low signal is sent to the MOSFET which pinches off the current flowing through the lighting circuit, turning off the light. The signal to lighting response transition times are on the order of tens of milliseconds, providing a nearly instantaneous response.

After each lighting decision (on/off/no change), the system will then automatically record the state of the lighting activation system (on/off) and transmit this information through USB to the computer, where it is received and interpreted by a custom end-user program (built using Qt framework in C++) which can display and record both the activation state and signal measured by the STROBE system for each channel of every fly arena, in real-time.

All STROBE design materials are available as a supplemental download.

All STROBE software is available for download from Github:

FPGA code: https://github.com/rcwchan/STROBE-fpga (copy archived at https://github.com/elifesciences-publications/STROBE-fpga).

All other code: https://github.com/rcwchan/STROBE_software/ (copy archived at https://github.com/elifesciences-publications/STROBE_software).

Latency measurements

Request a detailed protocol

A wire was attached to the outer electrode of the light-triggering channel, and agar was placed on the inner electrode as one would for a normal experiment. Video was captured at 178 frames/s, which corresponds to 5.6 ms per frame. Eight individual touches were analyzed by identifying the last frame where the wire was clearly not touching the agar. The subsequent frame was taken as the time of touch. The number of frames between the touch frame and the frame where LED activation is observed were then counted and multiplied by 5.6 ms to generate the latency of activation for that touch. We consider this a maximal estimate of latency, since the optics of the camera made close proximity and touch difficult to discriminate in some frames, in which case it was assumed that touch was occurring.

Fly strains

Request a detailed protocol

Fly stocks were raised on standard food at 25°C and 60% relative humidity under a 12:12 hr light:dark cycle. For neuronal activation experiments we used the 20XUAS-IVS-CsChrimson.mVenus (in attP40 insertion site) from the Bloomington Drosophila Stock Center (stock number: 55135). For neuronal silencing experiments we used UAS-Kir2.1 (Baines et al., 2001). Specific GRN expression was driven using Gr64f-GAL4 (Dahanukar et al., 2007) and Gr66a-GAL4 (Wang et al., 2004). GMR81E10-Gal4 was used for expression in Fdg neurons (Jenett et al., 2012; Pool et al., 2014). All MB split-GAL4 lines (MB011B-GAL4; MB026B-GAL4; MB056B-GAL4; MB065B-GAL4; MB083C-GAL4; MB093C-GAL4; MB10C-GAL4; MB195B-GAL4; MB210B-GAL4; MB308B-GAL4; MB310B-GAL4; MB441B-GAL4) were described in a previous study (Aso et al., 2014b) and obtained directly from Janelia. The expression patterns of the lines from the Flylight collections are available from the Flylight project websites.

Fly preparation and STROBE experiments

Request a detailed protocol

All experiments were performed with female flies to reduce variability, given that sex differences were not a subject of investigation. After eclosion, flies were kept for several days in fresh vials containing standard medium, and were then transferred at 25°C into vials covered with aluminum foil containing 1 ml standard medium (control flies) or 1 ml standard medium mixed with 1 mM of all-trans-retinal (retinal flies) for 2 days. Flies were then subjected to a 24 hr fasting period where they were transferred to covered vials containing 1 ml of 1% agar (control flies) or 1 ml of 1% agar mixed with 1 mM of all-trans-retinal (retinal flies).

For the starvation curve experiment (Figure 3), flies were transferred into vials containing 1 ml of standard medium ±all trans-retinal for 24 hr (fed group); or 1 ml of 1% agar ±all trans-retinal for 12-24-48 hours.

All flies were 5–9 days old at the time of the assay, and experiments were performed between 10:00 am and 5:00 pm. Both channels of STROBE chambers were loaded with 4 μl of 1% agar with or without sucrose (0, 1, 10, 100, 1000 mM) or denatonium (0, 0.1, 1, 10 mM). For aversive assays using denatonium, 50 mM sucrose was also added to increase food interactions.

Acquisition on the STROBE software was started and then single flies were transferred into each arena by mouth aspiration. Experiments were run for 60 min, and the preference index for each fly was calculated as: (interactions with Food 1 – interactions with Food 2)/(interactions with Food 1 + interactions with from Food 2). The red LED is always associated to the left side (Food 1). For temporal curves, data are pooled within 1 s time-period.

Sucrose, denatonium, agar and all-trans-retinal were obtained from Sigma-Aldrich.

For experiments done in Figure 2, light intensity used are 0, 0.12, 1.85, 6.56, 11.26 and 16.44 mW/cm2. All the other experiments were performed with a light intensity of 11.2 mW/cm2.

Dye feeding assay

Request a detailed protocol

Both channels of the STROBE chambers were loaded with 4 μl of 1% agar, which contained either 0.125 mg/ml blue (Erioglaucine, FD and C Blue#1) or 0.5 mg/ml red (Amaranth, FD and C Red#2) dye. Half the replicates for each experiment were done with the dyes swapped to control for any dye preference. After the experiment, flies were frozen and scored for abdomen color. Preference Index (PI) was scored as « 1 » for color associated to the light channel ; « −1 » for color associated to the no-light channel ; and « 0 » for both colors.

Immunohistochemistry

Request a detailed protocol

Brain immunofluorescence was carried out as described previously (Chu et al., 2014). Primary antibodies used were chicken anti-GFP (1:1000, Abcam #13970) and mouse anti-brp (1:50, DSHB #nc82). Secondary antibodies used were goat anti-chicken Alexa 488 (1:200, Abcam #150169) and goat anti-mouse Alexa 568 (1:200, Thermo Fisher Scientific #A11004).

All images were acquired using a Leica SP5 II Confocal microscope with a 25x water immersion objective. All images were taken sequentially with a z-stack step size at 2 μm, a line average of 2, line-scanning speed of 200 Hz, and a resolution of 1024 × 1024 pixels.

Statistical analysis

Request a detailed protocol

Statistical tests were performed using GraphPad Prism six software. Descriptions and results of each test are provided in the figure legends. Sample sizes are indicated in the figure legends.

Sample sizes were determined prior to experimentation based on the variance and effect sizes seen in prior experiments of similar types. All experimental conditions were run in parallel and therefore have the same or similar sample sizes. All replicates were biological replicates using different individual flies. Data for behavioral experiments were performed with flies from at least two independent crosses. There was one condition where data were excluded, which were determined prior to experimentation and applied uniformly throughout: the data from individual flies were removed if the fly did not pass a set minimum threshold of interactions (15), or the data showed hallmarks of a technical malfunction (rare).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78

Decision letter

  1. Mani Ramaswami
    Reviewing Editor; Trinity College Dublin, Ireland
  2. Catherine Dulac
    Senior Editor; Harvard University, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Closed-looped optogenetic activation of peripheral and central neurons modulates feeding in freely moving Drosophila" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by Mani Ramaswami as Reviewing Editor and Catherine Dulac as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This manuscript describes the technical features and utility of STROBE, a closed-loop stimulation system, which improves on a version of FlyPAD described by Steck et al. (2018). STROBE allows light delivery – and therefore optogenetic activation of neurons – to be temporally coupled to flies' physical interaction with agarose substrates and has good throughput. The new system has tighter coupling of the onset of the light stimulus with feeding initiation, and so may claim greater relevance to the activity of neurons directly activated by the act of feeding or tasting. By quantifying how frequently flies triggered the light pulses, one can quantify how activation of specific groups of CsChrimson-expressing neurons modifies flies' interactions with specific agar substrates. The system will be useful for the field of feeding behavior, as studying how neuronal activation can change due to taste and how it can control behavioral outputs is critical to our understanding of neural circuits and maladaptive feeding relating to human disease. The manuscript's findings that the STROBE is an important new tool for acutely probing feeding behavior questions with manipulation of peripheral taste and central associative neurons is reasonably well supported by a few important proof-of-principle experiments with STROBE such as showing that flies with CsChrimson expressed in their sweet neurons significantly increase their interaction with a sugarless agar that was coupled to light. In addition, they also use STROBE to provide some new biology by demonstrating a potential role of specific dopaminergic neurons (DANs) and mushroom body output neurons (MBONs) in regulating feeding. This is interesting because these specific DA neurons are generally known for mediating reinforcement signals (US) during learning as opposed to regulating innate behaviors.

In general, the technique appears useful but the impact of the manuscript would be improved by both technical clarifications and by some experiments using it to provide increased biological insight.

Essential revisions:

1) The authors assess the accuracy of STROBE – i.e., whether LEDs are turned on when and only when flies sip – by comparing STROBE LED activations to sips detected by FlyPAD. While an independent analysis would be better, this seems acceptable assuming the sip detection in FlyPAD was done well. It is possible however, that for the main comparison between STROBE and FlyPAD (Figure 1E), bins with no sips called by either algorithm were excluded from analysis (according to the figure legend). This may hide errors STROBE may make compared to FlyPAD and so needs to be analyzed in more detail.

2) According to the algorithm shown in Figure 1C and its description in the sixth paragraph of the subsection “STROBE System”, STROBE is designed to turn off the LED after just 100 ms during longer sips. This is also shown in Figure 1D at the end of the trace, where FlyPAD detects a long sip for which only a fraction is illuminated by STROBE. While the manuscript claims this is "by design," a discussion of the impact on the experiments is required. A related issue is that in some figures the y axis was labeled "sips" and in some "interactions." Are the two terms are used interchangeably? The writing should be clarified here.

3) While it is interesting to see that optogenetic activation of punishment-mediating and reward-mediating DANs or specific MBONs impacted fly-substrate interactions in a manner that is consistent with known properties of these neurons, two additional experiments are needed to ascertain that these neurons indeed play a role in regulating feeding. First, one should probably directly quantify food consumption triggered by STROBE activation of these neurons. Second, one should silence these neurons and assess the impact of such silencing on food consumption. (Silencing experiments need not be done with STROBE of course). This is because activation of these neurons may not affect feeding per se; rather, flies may simply seek to increase or decrease activation of these neurons by staying around the light-coupled agarose or moving away from it (as suggested by Aso et al.).

4) Why was only one intensity used to stimulate neurons following the initial studies of intensity-dependent GRN activation? Sweet GRNs, as appetitive neurons, have a U-shaped (bi-directional) response curve to increasing intensity stimuli (Figure 2D, E), while aversive neurons (bitter GRNs) have a linear negative correlation with intensity (Figure 2I, J). This clearly indicates a fundamental difference in how appetitive and aversive stimuli are processed – and may indicate that sweetness intensities above a certain concentration may have aversive side effects, so it would be more appropriate to choose different stimulus intensities for appetitive vs. aversive stimuli, instead of using one stimulus for both. Furthermore, greater intensities may affect actual neuronal function in different ways, depending on the neuronal population (or the circuits they engage). Given the absence of direct measurements of light-evoked neuronal activity and since central-brain neuron populations likely have their own response dynamics distinct from those of the peripheral systems, it seem crucial to test different intensities of light stimulation for these circuits. In the absence of this experiment, even if the behavior moves in the expected directions, there is no reason to believe that the conditions used reflect a biologically relevant stimulus (i.e., consider different patterns of activation, such as slow tonic firing, or bursting, or no APs at all but only gradual changes in membrane potential.) As the intensity chosen was higher than the "optimal" intensity for sweet-GRN responses (and so produced a lesser preference despite greater stimulation), could a lower intensity have had a different effect in Figure 4B? Particularly, since many in the field may base their stimulation protocols on this manuscript after publication, a more detailed and careful evaluation of stimulus intensities for the DA+ neurons and MBONs, together with a discussion of the limitations of this new analysis, will be very useful.

5) The interpretation that increasing concentrations of sucrose to both wells induces an enhancement of aversion (see text for Figure 4C) for flies with bitter-GRN activation is not entirely convincing. An alternative explanation is that the flies become more attracted to the other well – this might be true especially if the light stimulus "traps" a neuron in a particular firing rate regardless of additional stimuli. Increasing only the sucrose concentration in the light-coupled well would clarify this point (as would some analysis of the neuron activation due to the light stimulus if this were possible.)

6) All experiments were done within an hour, and preference was described for the full hour. However, pairing activation only of sweet taste neurons with a starved fly's interaction with a non-caloric substance like 1% agar likely only has very acute effects. How does preference change over the course of the hour? Does the fly learn that there is no energy associated with the "sweet taste" of the light-coupled agar? If so, is this abolished with the DAN/MBON manipulations, or does it persist? A more in-depth temporal analysis of the feeding behaviors would improve the manuscript, especially since the FlyPAD records all the data.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Closed-loop optogenetic activation of peripheral or central neurons modulates feeding in freely moving Drosophila" for further consideration at eLife. Your revised article has been favorably evaluated by Catherine Dulac as the Senior Editor, Mani Ramaswami as the Reviewing Editor and two reviewers.

The reviewers acknowledge that the manuscript has been improved by extensive revisions. Chiefly among these are: a dyed-food consumption experiment which linked actual food intake with STROBE-measured interactions, and an experiment where the PPL1 neurons were deactivated using Kir2.1, to check their necessity in inducing feeding, and not simply light activation.

1) However, some recommended revisions were not addressed in the new manuscript: in particular to point #4 which was meant to guide the experiments toward more physiologically relevant neuronal firing, and thus more interpretable results. The authors performed a curve with PPL1 and MB026B but not with Gr64f. Addressing the reasons behind the U-shaped curve would have been more informative than the experiments conducted. If the authors deem is sufficient to carry out the experiments with the 6.5 mW/cm2 regardless of the GAL4 used, in the least they should provide a paragraph discussing the possibility that this stimulation isn't physiologically significant for the circuits they test (especially for those where ephys data are available) and that in any study using the STROBE, OPTOPAD or OPTOFLIC, all closed loop systems, should first calculate their own light intensity curve to more faithfully mimic a physiological relevant stimulation and outcome. Given that this will be an important resource for those working with sensory and reward circuits, addressing this point is paramount and the authors should encourage the readers to optimize the intensity to match the neurons' fire rate and to use 6.5 as a starting rather than finishing point.

https://doi.org/10.7554/eLife.45636.041

Author response

Essential revisions:

1) The authors assess the accuracy of STROBE – i.e., whether LEDs are turned on when and only when flies sip – by comparing STROBE LED activations to sips detected by FlyPAD. While an independent analysis would be better, this seems acceptable assuming the sip detection in FlyPAD was done well. It is possible however, that for the main comparison between STROBE and FlyPAD (Figure 1E), bins with no sips called by either algorithm were excluded from analysis (according to the figure legend). This may hide errors STROBE may make compared to FlyPAD and so needs to be analyzed in more detail.

We apologize for our ambiguous and possibly misleading wording. We excluded bins only if there were 0 sips/interactions called by both algorithms. These bins were interpreted as time periods when the fly was not interacting with the food on that channel, and therefore were not informative in determining the relationship between sip numbers called by each algorithm. The large number of these bins would have artificially increased the R-squared value, and thus overestimated the correlation between the algorithms. We have corrected the figure legend wording for clarity.

2) According to the algorithm shown in Figure 1C and its description in the sixth paragraph of the subsection “STROBE System”, STROBE is designed to turn off the LED after just 100 ms during longer sips. This is also shown in Figure 1D at the end of the trace, where FlyPAD detects a long sip for which only a fraction is illuminated by STROBE. While the manuscript claims this is "by design," a discussion of the impact on the experiments is required. A related issue is that in some figures the y axis was labeled "sips" and in some "interactions." Are the two terms are used interchangeably? The writing should be clarified here.

We have now described this property of the STROBE design more explicitly in the Results section, and added a passage on possible implications to the Discussion. We have also corrected all the figures to use the term “interactions” when referring to data from the STROBE. This was our intention in the original submission, but an old version of the figures was used in error.

3) While it is interesting to see that optogenetic activation of punishment-mediating and reward-mediating DANs or specific MBONs impacted fly-substrate interactions in a manner that is consistent with known properties of these neurons, two additional experiments are needed to ascertain that these neurons indeed play a role in regulating feeding. First, one should probably directly quantify food consumption triggered by STROBE activation of these neurons. Second, one should silence these neurons and assess the impact of such silencing on food consumption. (Silencing experiments need not be done with STROBE of course). This is because activation of these neurons may not affect feeding per se; rather, flies may simply seek to increase or decrease activation of these neurons by staying around the light-coupled agarose or moving away from it (as suggested by Aso et al.).

We have performed these experiments. First, we validated a dye-based consumption assay in the STROBE using GRN activation and saw that sweet or bitter neuron activation had the expected effects on consumption (Figure 2—figure supplement 4). We then used this assay to show that activation of PPL1 or its postsynaptic MBONs also affect consumption as predicted by the interaction numbers measured by the STROBE (Figure 7). In fact, we saw larger effects on our consumption metric than on interaction numbers in all cases.

We also performed silencing experiments on the same pair of PPL1 and MBON drivers. In this case we expressed Kir2.1 in each neuron set and measured interaction numbers and dye ingestion in the STROBE with no light stimulation (Figure 7—figure supplement 3). Flies were given a choice between 10 mM sucrose and plain agar. In this paradigm we saw no effects of PPL1 or MBON silencing on feeding, suggesting that activation of these circuits is sufficient to drive changes in feeding, but not necessary for a simple food choice.

4) Why was only one intensity used to stimulate neurons following the initial studies of intensity-dependent GRN activation? Sweet GRNs, as appetitive neurons, have a U-shaped (bi-directional) response curve to increasing intensity stimuli (Figure 2D, E), while aversive neurons (bitter GRNs) have a linear negative correlation with intensity (Figure 2I, J). This clearly indicates a fundamental difference in how appetitive and aversive stimuli are processed – and may indicate that sweetness intensities above a certain concentration may have aversive side effects, so it would be more appropriate to choose different stimulus intensities for appetitive vs. aversive stimuli, instead of using one stimulus for both. Furthermore, greater intensities may affect actual neuronal function in different ways, depending on the neuronal population (or the circuits they engage). Given the absence of direct measurements of light-evoked neuronal activity and since central-brain neuron populations likely have their own response dynamics distinct from those of the peripheral systems, it seem crucial to test different intensities of light stimulation for these circuits. In the absence of this experiment, even if the behavior moves in the expected directions, there is no reason to believe that the conditions used reflect a biologically relevant stimulus (i.e., consider different patterns of activation, such as slow tonic firing, or bursting, or no APs at all but only gradual changes in membrane potential.) As the intensity chosen was higher than the "optimal" intensity for sweet-GRN responses (and so produced a lesser preference despite greater stimulation), could a lower intensity have had a different effect in Figure 4B? Particularly, since many in the field may base their stimulation protocols on this manuscript after publication, a more detailed and careful evaluation of stimulus intensities for the DA+ neurons and MBONs, together with a discussion of the limitations of this new analysis, will be very useful.

We agree that manipulating stimulus parameters in different experiments could generate interesting results, in particular for the higher-order circuits in the mushroom body. However, we are also subject to practical limitations and must consider the cost-benefit ratio for additional experiments – for example, performing an experiment on one genotype under one condition across five light intensities requires about 25 2-hour (16 fly) experiments. Thus, performing intensity curves across five different conditions (e.g. different sucrose intensities in Figure 4B) would take 125 experiments, which would be well over a month of full-time work from one person. Given that the most likely outcome is that the curve simply shifts slightly in either direction by changing the intensities, this seems impractical.

However, we did perform intensity curves on two example mushroom body populations – the PPL1s and their postsynaptic MBONs (MB026B-Gal4). This data is now shown in Figure 7, which demonstrates that, in both cases, the effect size increases with increasing light intensities. Interestingly, unlike sweet GRN stimulation, which displays the noted U-shaped response, activation of these attractive MBONs produced linearly increasing interaction numbers across the range of intensities tested. Also, as we now point out in the text, we actually see a more noticeable effect on ingestion than we see on interaction numbers for both neuron populations tested, demonstrating that these circuits do indeed robustly affect feeding (as discussed above).

5) The interpretation that increasing concentrations of sucrose to both wells induces an enhancement of aversion (see text for Figure 4C) for flies with bitter-GRN activation is not entirely convincing. An alternative explanation is that the flies become more attracted to the other well – this might be true especially if the light stimulus "traps" a neuron in a particular firing rate regardless of additional stimuli. Increasing only the sucrose concentration in the light-coupled well would clarify this point (as would some analysis of the neuron activation due to the light stimulus if this were possible.)

We completely agree with this interpretation, and it was always our intention to suggest that it is the relative difference in attractiveness that is critical, and this is primarily driven by the increasing appetitiveness of the non-light side. We have reworded this passage to be more clear. Increasing the sugar concentration on the light-triggering channel only could show us the concentration of sucrose (if there is one) that is able to overcome the aversiveness of bitter GRN stimulation. However, given that light intensity is also a variable (and the experiment number limitations discussed above), this experiment is unlikely to reveal any fundamental insight into the system (or biology). We were more interested in examining the masking of GRN stimulation by taste ligands of the same modality.

6) All experiments were done within an hour, and preference was described for the full hour. However, pairing activation only of sweet taste neurons with a starved fly's interaction with a non-caloric substance like 1% agar likely only has very acute effects. How does preference change over the course of the hour? Does the fly learn that there is no energy associated with the "sweet taste" of the light-coupled agar? If so, is this abolished with the DAN/MBON manipulations, or does it persist? A more in-depth temporal analysis of the feeding behaviors would improve the manuscript, especially since the FlyPAD records all the data.

Detailed temporal data for GRN and DAN/MBON activation is now included in the supplements for Figures 2, 3, 4, and 7. As we discuss in the text, we generally see very stable relative interaction numbers (and therefore preference indices) after they are established in the first 15 minutes. This is true for peripheral and central neuron activation, and does not provide any evidence for sensitization, adaptation, or learning affecting the results we see over the course of a 1-hour experiment. This is not to say that learning does not (or cannot) occur in the STROBE, but this is an interesting avenue where, as we mention at the end of the Discussion, the STROBE could provide valuable insight in the future.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

The reviewers acknowledge that the manuscript has been improved by extensive revisions. Chiefly among these are: a dyed-food consumption experiment which linked actual food intake with STROBE-measured interactions, and an experiment where the PPL1 neurons were deactivated using Kir2.1, to check their necessity in inducing feeding, and not simply light activation.

1) However, some recommended revisions were not addressed in the new manuscript: in particular to point #4 which was meant to guide the experiments toward more physiologically relevant neuronal firing, and thus more interpretable results. The authors performed a curve with PPL1 and MB026B but not with Gr64f. Addressing the reasons behind the U-shaped curve would have been more informative than the experiments conducted. If the authors deem is sufficient to carry out the experiments with the 6.5 mW/cm2 regardless of the GAL4 used, in the least they should provide a paragraph discussing the possibility that this stimulation isn't physiologically significant for the circuits they test (especially for those where ephys data are available) and that in any study using the STROBE, OPTOPAD or OPTOFLIC, all closed loop systems, should first calculate their own light intensity curve to more faithfully mimic a physiological relevant stimulation and outcome. Given that this will be an important resource for those working with sensory and reward circuits, addressing this point is paramount and the authors should encourage the readers to optimize the intensity to match the neurons' fire rate and to use 6.5 as a starting rather than finishing point.

We agree that the U-shaped curve for Gr64f activation is notable and may reflect interesting biology that is worth pursuing. However, we also believe that truly understanding this phenomenon is not a straightforward prospect. The effect could be from non-physiological neuron responses at high intensities; a form of sensory-induced satiety; an interaction between the “virtual” sweet taste and visual responses to the flashing lights; or a host of other factors that are likely difficult to tease apart. The intention of this experiment was to illustrate that intensity was an important variable in the response, and the reviewer is correct that we could have highlighted this more clearly in discussing both the Gr64f and DAN/MBON intensity curves. We have now described the Gr64f result in more detail and added a passage to the Discussion (Discussion, fifth paragraph) that considers possible interpretations of the Gr64f result and emphasizes the general importance of calibrating light intensity to mimic physiological and/or behavioral responses to natural stimuli.

https://doi.org/10.7554/eLife.45636.042

Article and author information

Author details

  1. Pierre-Yves Musso

    Department of Zoology, Life Sciences Institute, University of British Columbia, Vancouver, Canada
    Contribution
    Conceptualization, Formal analysis, Investigation, Visualization, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  2. Pierre Junca

    Department of Zoology, Life Sciences Institute, University of British Columbia, Vancouver, Canada
    Contribution
    Formal analysis, Investigation, Visualization, Writing—review and editing
    Competing interests
    No competing interests declared
  3. Meghan Jelen

    Department of Zoology, Life Sciences Institute, University of British Columbia, Vancouver, Canada
    Contribution
    Investigation, Writing—review and editing
    Competing interests
    No competing interests declared
  4. Damian Feldman-Kiss

    Department of Zoology, Life Sciences Institute, University of British Columbia, Vancouver, Canada
    Contribution
    Investigation, Writing—review and editing
    Competing interests
    No competing interests declared
  5. Han Zhang

    Department of Physics and Astronomy, University of British Columbia, Vancouver, Canada
    Contribution
    Methodology, Writing—review and editing, Designed and built STROBE system with conceptual input from PYM and MDG, Wrote first draft of STROBE methods section
    Competing interests
    No competing interests declared
  6. Rachel CW Chan

    Department of Physics and Astronomy, University of British Columbia, Vancouver, Canada
    Present address
    Department of Computer Science, University of Toronto, Toronto, Canada
    Contribution
    Software, Methodology, Writing—review and editing, Designed and built STROBE system with conceptual input from PYM and MDG, Wrote first draft of STROBE methods section
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1009-6379
  7. Michael D Gordon

    Department of Zoology, Life Sciences Institute, University of British Columbia, Vancouver, Canada
    Contribution
    Conceptualization, Supervision, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    gordon@zoology.ubc.ca
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5440-986X

Funding

Natural Sciences and Engineering Research Council of Canada (RGPIN-2016-03857)

  • Michael D Gordon

Natural Sciences and Engineering Research Council of Canada (RGPAS-49246-16)

  • Michael D Gordon

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Linda Au carrying out part of a screen of MBON activation in the STROBE, and Bonnie Chu for the fly drawing used in Figure 2. We also thank Carlos Ribeiro and Pavel Itskov for kindly providing their original FlyPAD VHDL code, which was instrumental in developing the STROBE system. This work was funded by Natural Sciences and Engineering Research Council (NSERC) grants RGPIN-2016–03857 and RGPAS 492846–16. MDG is a CIHR New Investigator and a Michael Smith Foundation for Health Research Scholar.

Senior Editor

  1. Catherine Dulac, Harvard University, United States

Reviewing Editor

  1. Mani Ramaswami, Trinity College Dublin, Ireland

Publication history

  1. Received: January 31, 2019
  2. Accepted: July 18, 2019
  3. Accepted Manuscript published: July 19, 2019 (version 1)
  4. Version of Record published: July 31, 2019 (version 2)

Copyright

© 2019, Musso et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,231
    Page views
  • 252
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Neuroscience
    2. Stem Cells and Regenerative Medicine
    Isabelle Maria Blomfield et al.
    Research Article Updated
    1. Neuroscience
    Lingxi Lu et al.
    Research Article