- Email: [email protected]
Functional Connectomics: How Maggots Make Up Their Minds
Current Biology
Dispatches 4. Robinson, C., Steinberg, D.K., Anderson, T.R., Arı´stegui, J., Carlson, C.A., Frost, J.R., Ghiglione, J.-F., Herna´ndez-Leo´nd, S., Jackson, G.A., Koppelmann, R., et al. (2010). Mesopelagic zone ecology and biogeochemistry – a synthesis. Deep Sea Res. II 57, 1504–1518. 5. Proud, R., Cox, M.J., and Brierley, A.S. (2017). Biogeography of the global ocean’s mesopelagic zone. Curr. Biol. 27, 113–119. 6. ESRI. (2016). Ecological Marine Units. Accessed at http://www.esri.com/ecological-marine-units on 15th November 2016.
7. Costello, M.J., Cheung, A., and De Hauwere, N. (2010). Topography statistics for the surface and seabed area, volume, depth and slope, of the world’s seas, oceans and countries. Environ. Sci. Technol. 44, 8821– 8828.
10. Basher, Z., and Costello, M.J. (2016). The past, present and future distribution of a deep-sea shrimp in the Southern Ocean. PeerJ. 4, e1713.
foue¨t, S., Costello, M.J., Rast, M., and 8. Andre Sathyendranath, S. (2008). Earth observations for marine and coastal biodiversity. Remote Sens. Environ. 112, 3297–3299.
11. Saeedi, H., Basher, Z., and Costello, M.J. (2016). Modelling present and future global distributions of razor clams (Bivalvia: Solenidae). Helgoland Mar. Res., in press. DOI 10.1186/s10152-016-0477-4.
9. Thresher, R.E., Koslow, J.A., Morison, A.K., and Smith, D.C. (2007). Depth-mediated reversal of the effects of climate change on long-term growth rates of exploited marine fish. Proc. Natl. Acad. Sci. USA 104, 7461–7465.
12. Chaudhary, C., Saeedi, H., and Costello, M.J. (2016). Bimodality of latitudinal gradients in marine species richness. Trends Ecol. Evol. 31, 670–676.
Functional Connectomics: How Maggots Make Up Their Minds William B. Kristan, Jr. Neurobiology Section, Department of Biological Sciences, University of California at San Diego, La Jolla, CA 92093-0357, USA Correspondence: [email protected] http://dx.doi.org/10.1016/j.cub.2016.11.031
How Drosophila larvae select one behavior or a sequence of behaviors, and then persist in the final one, has been addressed by a powerful combination of electron-microscopy reconstruction of neuronal connections, genetic manipulations, electrophysiology, and neuronal modeling. Surprisingly, reciprocal inhibitory synaptic connections are major players in choosing, sequencing and maintaining behaviors. The future is here! For the past decade, I have been telling classes of graduate students that their careers will see huge advances in how circuits of neurons produce behaviors, largely because new technologies are providing approaches for working out these circuits that previous generations of neuroscientists could only dream about. These new technologies include molecular genetics, virology, dyes for recording, activating, and silencing neurons [1], electron microscopy reconstruction to produce cellular-level ‘connectomes’ [2], and computational approaches for both data analysis and modeling [3]. A new paper by Jovanic et al. [4] is an excellent example of how these new techniques can be used in a powerful way to figure out how neurons communicate with one another to produce and coordinate behaviors. This study uses larval Drosophila (aka maggots) which, at the developmental stage used, is a segmented worm composed of a head, a telson (tail), and
eleven intermediate segments — three thoracic and eight abdominal — each with a ganglion composed of fewer than 1000 neurons that connects to adjacent ganglia via connectives. Segments move mainly by longitudinal muscle contractions, with subsequent elongations produced by a high internal pressure. Each intermediate segment can bend, by contracting the longitudinal muscles on one side, or shorten, by contracting all the longitudinal muscles at once. In addition, the larva can crawl by producing a sequence of ventral shortenings that move along the body, pushing the animal forward. Under the experimental conditions used by Jovanic et al. [4], unperturbed larvae crawl continuously; in response to a mild mechanical stimulus, provided by air current directed at them, the larva stops crawling and either bends (randomly left or right) or shortens (called ‘hunching’ in this paper). If the stimulus is maintained, the larva returns to crawling after several seconds, but the authors focused on the
R38 Current Biology 27, R19–R41, January 9, 2017 ª 2017 Published by Elsevier Ltd.
initial response, either bending or hunching. A previous study [5] had shown that a particular type of mechanosensory ending, known as the chordotonal organ, was responsible for producing both bending and hunching. Which of the two behaviors was elicited was probabilistic, even when the same set of chordotonal organ neurons were repeatedly activated optogenetically by expressing the dye CsChrimson [6] in a subset of chordotonal organ neurons. To work out the circuitry of the neurons involved in choosing between bending and hunching, the authors tested the contributions of a set of interneurons called ‘Basin Projection Neurons’ — named for the shape of their dendritic trees and the fact that they had axons that projected out of the ganglion containing the cell body — henceforth called B neurons. These neurons had been previously shown to receive strong chordotonal organ input [7]. They were originally identified by behaviorally
Current Biology
Dispatches screening many fly lines each of which has a small number of neurons stably labeled, then identifying these neurons in a serial electron microscropy reconstruction of the larval nervous system [8]. In each segment there are four bilateral pairs of B neurons, B1– B4, that elicit bending and hunching, of which two, B1 and B2, were studied in detail (Figure 1). The targets of these axons are not known, but it is known that activating B1 alone produced hunching or bending, probabilistically, whereas activating B2, either alone or in combination with B1, produced only bending, implying that hunching is inhibited by B2. These activation experiments show that B1 and B2 are sufficient to produce the two behaviors, and inactivation experiments showed that these neurons are also necessary for chordotonal organ activity to produce the behaviors. Simultaneous electrophysiological recording from B1 and B2 showed that B1 always received excitation in response to mechanical stimuli, but that B2 was sometimes excited and other times inhibited, both between animals as well as in response to repeated stimulation within the same animal. Using picrotoxin to block inhibitory inputs produced by the neurotransmitter GABA changed the responses to the B neuron tested (B1), showing that the mechanical stimuli — in addition to their direct excitatory connections — also activated inhibitory input onto the B neurons. To find the inhibitory pathways onto the B neurons, the investigators once again turned to the electron microscopy reconstruction of the entire larval nervous system, looking for neurons that received input from the chordotonal organ sensory neurons and made output synaptic contacts onto B neurons. They identified four distinct categories of neurons, each category having multiple members (two or three pairs per segment), which they called local neurons because all their processes are within a single ganglion. These local neurons bound an antibody to GABA, so they were taken to be inhibitory (iLNs). All the iLNs connect to both B neurons, but one subset (iLNa) has more synaptic contact sites onto B2, whereas the other subset (iLNb) makes more contacts with B1 (Figure 1). In addition, iLNa and iLNb make many synaptic
+
B1
Hunch -
iLNb Ha
Ch
Hb iLNa
+ Bend B2
+
Excitation Feedforward inhibition Feedback inhibition Projection Current Biology
Figure 1. Circuitry that influences whether a larval Drosophila hunches or bends. Circles represent the four different cell types under study: Ch, chordotonal organ neurons; iLNa,b, inhibitory local neurons; Ha,b, ‘Handle’ neurons; and B1,2, ‘Basin’ neurons. Types of connections are indicated by the key and the strengths of connections (strong, medium, and weak) are indicated by the thickness of the lines. The strengths were inferred from the number of synaptic sites found in the connectome. (Adapted from Figure 3A in [4].)
contacts with one another, but none with iLNs of the same class. The anatomical data, therefore, strongly suggest that the iLNs provide feedforward inhibition onto the B neurons and make reciprocal inhibitory connections to each other. An additional distinct class of iLN, called ‘Handle’ (H) neurons, make reciprocal connections with the iLNs, and again there are two types of these neurons, based upon the number of contacts made with the two iLN classes (Figure 1). The H neurons also receive contacts from both the chordotonal organ and B neurons, which were shown electrophysiologically to be excitatory. The fact that H neurons connect only to the iLNs, and they are likely to be inhibitory (from GABA antibody staining),
suggests that the H neurons can influence the behavioral output only indirectly, by adjusting the inhibitory input experienced by the iLNs. The H neurons, therefore, constitute a feedback inhibitory pathway. To determine whether the identified network is capable of choosing between the two behaviors — hunching and bending — the authors constructed a computational model. They used the connections shown in Figure 1, with the inhibitory synaptic strengths set by the number of synaptic contacts found in the connectome, and they varied the connection strengths of the chordotonal organ neurons onto the ILNs in different iterations of the model. They found clusters of combinations of input strengths that produce B1 and B2
Current Biology 27, R19–R41, January 9, 2017 R39
Current Biology
Dispatches activations appropriate for hunching, bending and, when the connection strengths were too low, no response (usually, intact animals that are neither bending or hunching are crawling, so ‘no response’ in the model may represent ‘crawling’ in a real animal). By activating the two iLNs at different intensities, the model produces bending, hunching, and a hunch-to-bend sequence, a behavior also seen in real animals. It appears that the hunch-to-bend sequencing fell out of the interactions among the inhibitory neurons; it was not specifically a target of the modeling study. These modeling experiments constitute reality tests: could the system function in the way hypothesized? This is an important step in analysing any complex system, and nested reciprocally inhibitory networks certainly qualify as being complex systems [9]. Even more useful, however, is a model that makes testable predictions, suggesting experiments that can be performed on the system. Javonic et al. [4] used the model in this way, too — in spades. They used a variety of combinations of experimental techniques — optogenetic, eletrophysiological, and behavioral— that produced results that matched five different responses of the circuit that were predicted by the modeling studies. These results give substantial confidence that the circuitry found by the electron microscopy connectome contributes significantly to a fly’s choice between two distinct and relevant behaviors, hunching and bending. So, how does this system actually work? We can get some idea, in broad brushstrokes, by starting with the excitatory pathway, then adding in the two inhibitory pathways, as shown by the different colors in Figure 1. If the network consisted of just the chordotonal organ mechanosensory neurons and the B neurons (the inputs and outputs of the network, indicated by green and black pathways), hunching would occur whenever appropriate chordotonal organ neurons were activated that excite B1 and B2 together, and bending would occur when the activation of B2 was so low that it could not produce its inhibitory effect that would produce hunching. Adding in the feedforward inhibitory circuit (the iLN connections, indicated by the red pathway) effectively produces a switch: if
one of the iLNs is even a little more strongly activated than the other one, the more weakly activated iLN will completely shut down. For instance, if the chordotonal organ input excites iLNb a little more than it excites iLNa, the iLNa receives more inhibition from iLNb than it gives back to iLNb, so the activity in iLNa decreases and iLNb increases. The increased iLNb activity further inhibits iLNa. This cycle repeats until iLNb is fully on and iLNa is fully turned off. The consequence is that the activity in B1 is decreased (because the strong inhibition from iLNb is activated) and the activity in B2 is large (because it has lost a major part of its inhibition, from iLNa). This condition produces a bend. On the other hand, if iLNa receives more excitation from chordotonal organ neurons than does iLNb, the opposite effect is seen, which produces bending. Adding the feedback inhibition provided by the H cells (violet pathway in Figure 1) adds multiple pathways for ‘disinhibiting’ (inhibiting the inhibition onto) the feedforward inhibitory pathway. This disinhibition can prolong whatever level of activity is present in the B cells, which would functionally lock in whichever behavior is being generated. Note that there are four weak connections that do not fit into the general functional scheme. For instance, iLNb weakly inhibits B2, a connection that opposes the strong disinhibitory pathway from iLNb to B2 via iLNa. The authors suggest that these pathways may serve to allow the circuit to escape from a locked-in state. Alternatively, these connections may be part of a circuit-wide gain control system. As well as being a tour de force of modern technologies applied to a timely systems neurobiological question, this new study is also interesting because the circuit features that it describes are being found (or at least proposed) in other systems. The selection of behaviors by different combinations of the same two pathways (to Hunch and to Bend) is the simplest form of a combinatorial code, in which the effect of one cell (in this case, B1) is qualitatively different depending upon whether another cell (B2) is also active. Such a combinatorial code for selecting between behaviors has been found in leeches [10]. In addition, the type of reciprocal inhibition seen in the feedforward pathway in this study —
R40 Current Biology 27, R19–R41, January 9, 2017
between neurons that elicit two different behaviors — was initially proposed by the ethologist Niko Tinbergen [11], based entirely upon observing natural behaviors of fish and birds. Electrophysiological evidence for inhibitory interactions between competing behavioral pathways has been found in animals such as mollusks [12], leeches [13], and monkeys [14]. Finally, the disinhibition seen in the feedback circuitry is similar to that seen in the basal ganglia of vertebrates [15], as well as in pathways that hold the position of the eyes after they move to a new position [16]. This parcellation of recognizable functions into distinct sub-circuits suggests that the same connectivity pattern (for example, reciprocal inhibition) can be used as a module to perform different functions (for example, behavioral selection or prolonging any ongoing behavior), depending upon where in the network the module is plugged in [17]. Giving clear examples of this kind of systems-level reasoning may ultimately be the most important contribution of the impressive piece of research by Jovanic et al. [4]. REFERENCES 1. Lerner, T.N., Ye, L., and Deisseroth, K. (2016). Communication in neural circuits: Tools, opportunities, and challenges. Cell 164, 1136–1150. 2. Titze, B., and Genoud, C. (2016). Volume scanning electron microscopy for imaging biological ultrastructure. Biol. Cell. 108, 307–323. 3. Bouchard, K.E., Aimone, J.B., Chun, M., Dean, T., Denker, M., Diesmann, M., Donofrio, D.D., Frank, L.M., Kasthuri, N., Koch, C., et al. (2016). High-performance computing in Neuroscience for data-driven discovery, integration, and dissemination. Neuron 92, 628–631. 4. Jovanic, T., Schneider-Mizell, C.M., Shao, M., Masson, J.B., Denisov, G., Fetter, R.D., Mensh, B.D., Truman, J.W., Cardona, A., and Zlatic, M. (2016). Competitive disinhibition mediates behavioral choice and sequences in Drosophila. Cell 167, 1–13. 5. Ohyama, T., Jovanic, T., Denisov, G., Dang, T.C., Hoffmann, D., Kerr, R.A., and Zlatic, M. (2013). High-throughput analysis of stimulusevoked behaviors in Drosophila larva reveals multiple modality-specific escape strategies. PLoS One 8, e71706. 6. Klapoetke, N.C., Murata, Y., Kim, S.S., Pulver, S.R., Birdsey-Benson, A., Cho, Y.K., Morimoto, T.K., Chuong, A.S., Carpenter, E.J., Tian, Z., et al. (2014). Independent optical excitation of distinct neural populations. Nat. Methods 11, 338–346.
Current Biology
Dispatches 7. Ohyama, T., Schneider-Mizell, C.M., Fetter, R.D., Aleman, J.V., Franconville, R., RiveraAlba, M., Mensh, B.D., Branson, K.M., Simpson, J.H., Truman, J.W., et al. (2015). A multilevel multimodal circuit enhances action selection in Drosophila. Nature 520, 633–639. 8. Schneider-Mizell, C.M., Gerhard, S., Longair, M., Kazimiers, T., Li, F., Zwart, M.F., Champion, A., Midgley, F.M., Fetter, R.D., Saalfeld, S., and Cardona, A. (2016). Quantitative neuroanatomy for connectomics in Drosophila. eLife 5, e12059. 9. Wojcik, J., Schwabedal, J., Clewley, R., and Shilnikov, A.L. (2014). Key bifurcations of bursting polyrhythms in 3-cell central pattern generators. PLoS One 9, e92918.
10. Kristan, W.B., and Gillette, R. (2007). Decisionmaking in small neuronal networks. In Invertebrate Neurobiology, G. North, and R. Greenspan, eds. (New York: Cold Spring Harbor Laboratory Press), pp. 533–554.
14. Wang, X.J. (2008). Decision making in recurrent neuronal circuits. Neuron 60, 215–234.
11. Tinbergen, N. (1951). The Study of Instinct (Oxford, Clarendon Press).
15. Mink, J.W. (1996). The basal ganglia: focused selection and inhibition of competing motor programs. Prog. Neurobiol. 50, 381–425.
12. Pirger, Z., Crossley, M., La´szlo´, Z., Naskar, S., Kemenes, G., O’Shea, M., Benjamin, P.R., and Kemenes, I. (2014). Interneuronal mechanism for Tinbergen’s hierarchical model of behavioral choice. Curr. Biol. 24, 2018–2024.
16. Aksay, E., Olasagasti, I., Mensh, B.D., Baker, R., Goldman, M.S., and Tank, D.W. (2007). Functional dissection of circuitry in a neural integrator. Nat. Neurosci. 10, 494–504.
13. Gaudry, Q., and Kristan, W.B., Jr. (2009). Behavioral choice by presynaptic inhibition of tactile sensory terminals. Nat. Neurosci. 12, 1450–1457.
17. Mysore, S.P., and Knudsen, E.I. (2012). Reciprocal inhibition of inhibition: a circuit motif for flexible categorization in stimulus selection. Neuron 73, 193–205.
Current Biology 27, R19–R41, January 9, 2017 R41
Dispatches 4. Robinson, C., Steinberg, D.K., Anderson, T.R., Arı´stegui, J., Carlson, C.A., Frost, J.R., Ghiglione, J.-F., Herna´ndez-Leo´nd, S., Jackson, G.A., Koppelmann, R., et al. (2010). Mesopelagic zone ecology and biogeochemistry – a synthesis. Deep Sea Res. II 57, 1504–1518. 5. Proud, R., Cox, M.J., and Brierley, A.S. (2017). Biogeography of the global ocean’s mesopelagic zone. Curr. Biol. 27, 113–119. 6. ESRI. (2016). Ecological Marine Units. Accessed at http://www.esri.com/ecological-marine-units on 15th November 2016.
7. Costello, M.J., Cheung, A., and De Hauwere, N. (2010). Topography statistics for the surface and seabed area, volume, depth and slope, of the world’s seas, oceans and countries. Environ. Sci. Technol. 44, 8821– 8828.
10. Basher, Z., and Costello, M.J. (2016). The past, present and future distribution of a deep-sea shrimp in the Southern Ocean. PeerJ. 4, e1713.
foue¨t, S., Costello, M.J., Rast, M., and 8. Andre Sathyendranath, S. (2008). Earth observations for marine and coastal biodiversity. Remote Sens. Environ. 112, 3297–3299.
11. Saeedi, H., Basher, Z., and Costello, M.J. (2016). Modelling present and future global distributions of razor clams (Bivalvia: Solenidae). Helgoland Mar. Res., in press. DOI 10.1186/s10152-016-0477-4.
9. Thresher, R.E., Koslow, J.A., Morison, A.K., and Smith, D.C. (2007). Depth-mediated reversal of the effects of climate change on long-term growth rates of exploited marine fish. Proc. Natl. Acad. Sci. USA 104, 7461–7465.
12. Chaudhary, C., Saeedi, H., and Costello, M.J. (2016). Bimodality of latitudinal gradients in marine species richness. Trends Ecol. Evol. 31, 670–676.
Functional Connectomics: How Maggots Make Up Their Minds William B. Kristan, Jr. Neurobiology Section, Department of Biological Sciences, University of California at San Diego, La Jolla, CA 92093-0357, USA Correspondence: [email protected] http://dx.doi.org/10.1016/j.cub.2016.11.031
How Drosophila larvae select one behavior or a sequence of behaviors, and then persist in the final one, has been addressed by a powerful combination of electron-microscopy reconstruction of neuronal connections, genetic manipulations, electrophysiology, and neuronal modeling. Surprisingly, reciprocal inhibitory synaptic connections are major players in choosing, sequencing and maintaining behaviors. The future is here! For the past decade, I have been telling classes of graduate students that their careers will see huge advances in how circuits of neurons produce behaviors, largely because new technologies are providing approaches for working out these circuits that previous generations of neuroscientists could only dream about. These new technologies include molecular genetics, virology, dyes for recording, activating, and silencing neurons [1], electron microscopy reconstruction to produce cellular-level ‘connectomes’ [2], and computational approaches for both data analysis and modeling [3]. A new paper by Jovanic et al. [4] is an excellent example of how these new techniques can be used in a powerful way to figure out how neurons communicate with one another to produce and coordinate behaviors. This study uses larval Drosophila (aka maggots) which, at the developmental stage used, is a segmented worm composed of a head, a telson (tail), and
eleven intermediate segments — three thoracic and eight abdominal — each with a ganglion composed of fewer than 1000 neurons that connects to adjacent ganglia via connectives. Segments move mainly by longitudinal muscle contractions, with subsequent elongations produced by a high internal pressure. Each intermediate segment can bend, by contracting the longitudinal muscles on one side, or shorten, by contracting all the longitudinal muscles at once. In addition, the larva can crawl by producing a sequence of ventral shortenings that move along the body, pushing the animal forward. Under the experimental conditions used by Jovanic et al. [4], unperturbed larvae crawl continuously; in response to a mild mechanical stimulus, provided by air current directed at them, the larva stops crawling and either bends (randomly left or right) or shortens (called ‘hunching’ in this paper). If the stimulus is maintained, the larva returns to crawling after several seconds, but the authors focused on the
R38 Current Biology 27, R19–R41, January 9, 2017 ª 2017 Published by Elsevier Ltd.
initial response, either bending or hunching. A previous study [5] had shown that a particular type of mechanosensory ending, known as the chordotonal organ, was responsible for producing both bending and hunching. Which of the two behaviors was elicited was probabilistic, even when the same set of chordotonal organ neurons were repeatedly activated optogenetically by expressing the dye CsChrimson [6] in a subset of chordotonal organ neurons. To work out the circuitry of the neurons involved in choosing between bending and hunching, the authors tested the contributions of a set of interneurons called ‘Basin Projection Neurons’ — named for the shape of their dendritic trees and the fact that they had axons that projected out of the ganglion containing the cell body — henceforth called B neurons. These neurons had been previously shown to receive strong chordotonal organ input [7]. They were originally identified by behaviorally
Current Biology
Dispatches screening many fly lines each of which has a small number of neurons stably labeled, then identifying these neurons in a serial electron microscropy reconstruction of the larval nervous system [8]. In each segment there are four bilateral pairs of B neurons, B1– B4, that elicit bending and hunching, of which two, B1 and B2, were studied in detail (Figure 1). The targets of these axons are not known, but it is known that activating B1 alone produced hunching or bending, probabilistically, whereas activating B2, either alone or in combination with B1, produced only bending, implying that hunching is inhibited by B2. These activation experiments show that B1 and B2 are sufficient to produce the two behaviors, and inactivation experiments showed that these neurons are also necessary for chordotonal organ activity to produce the behaviors. Simultaneous electrophysiological recording from B1 and B2 showed that B1 always received excitation in response to mechanical stimuli, but that B2 was sometimes excited and other times inhibited, both between animals as well as in response to repeated stimulation within the same animal. Using picrotoxin to block inhibitory inputs produced by the neurotransmitter GABA changed the responses to the B neuron tested (B1), showing that the mechanical stimuli — in addition to their direct excitatory connections — also activated inhibitory input onto the B neurons. To find the inhibitory pathways onto the B neurons, the investigators once again turned to the electron microscopy reconstruction of the entire larval nervous system, looking for neurons that received input from the chordotonal organ sensory neurons and made output synaptic contacts onto B neurons. They identified four distinct categories of neurons, each category having multiple members (two or three pairs per segment), which they called local neurons because all their processes are within a single ganglion. These local neurons bound an antibody to GABA, so they were taken to be inhibitory (iLNs). All the iLNs connect to both B neurons, but one subset (iLNa) has more synaptic contact sites onto B2, whereas the other subset (iLNb) makes more contacts with B1 (Figure 1). In addition, iLNa and iLNb make many synaptic
+
B1
Hunch -
iLNb Ha
Ch
Hb iLNa
+ Bend B2
+
Excitation Feedforward inhibition Feedback inhibition Projection Current Biology
Figure 1. Circuitry that influences whether a larval Drosophila hunches or bends. Circles represent the four different cell types under study: Ch, chordotonal organ neurons; iLNa,b, inhibitory local neurons; Ha,b, ‘Handle’ neurons; and B1,2, ‘Basin’ neurons. Types of connections are indicated by the key and the strengths of connections (strong, medium, and weak) are indicated by the thickness of the lines. The strengths were inferred from the number of synaptic sites found in the connectome. (Adapted from Figure 3A in [4].)
contacts with one another, but none with iLNs of the same class. The anatomical data, therefore, strongly suggest that the iLNs provide feedforward inhibition onto the B neurons and make reciprocal inhibitory connections to each other. An additional distinct class of iLN, called ‘Handle’ (H) neurons, make reciprocal connections with the iLNs, and again there are two types of these neurons, based upon the number of contacts made with the two iLN classes (Figure 1). The H neurons also receive contacts from both the chordotonal organ and B neurons, which were shown electrophysiologically to be excitatory. The fact that H neurons connect only to the iLNs, and they are likely to be inhibitory (from GABA antibody staining),
suggests that the H neurons can influence the behavioral output only indirectly, by adjusting the inhibitory input experienced by the iLNs. The H neurons, therefore, constitute a feedback inhibitory pathway. To determine whether the identified network is capable of choosing between the two behaviors — hunching and bending — the authors constructed a computational model. They used the connections shown in Figure 1, with the inhibitory synaptic strengths set by the number of synaptic contacts found in the connectome, and they varied the connection strengths of the chordotonal organ neurons onto the ILNs in different iterations of the model. They found clusters of combinations of input strengths that produce B1 and B2
Current Biology 27, R19–R41, January 9, 2017 R39
Current Biology
Dispatches activations appropriate for hunching, bending and, when the connection strengths were too low, no response (usually, intact animals that are neither bending or hunching are crawling, so ‘no response’ in the model may represent ‘crawling’ in a real animal). By activating the two iLNs at different intensities, the model produces bending, hunching, and a hunch-to-bend sequence, a behavior also seen in real animals. It appears that the hunch-to-bend sequencing fell out of the interactions among the inhibitory neurons; it was not specifically a target of the modeling study. These modeling experiments constitute reality tests: could the system function in the way hypothesized? This is an important step in analysing any complex system, and nested reciprocally inhibitory networks certainly qualify as being complex systems [9]. Even more useful, however, is a model that makes testable predictions, suggesting experiments that can be performed on the system. Javonic et al. [4] used the model in this way, too — in spades. They used a variety of combinations of experimental techniques — optogenetic, eletrophysiological, and behavioral— that produced results that matched five different responses of the circuit that were predicted by the modeling studies. These results give substantial confidence that the circuitry found by the electron microscopy connectome contributes significantly to a fly’s choice between two distinct and relevant behaviors, hunching and bending. So, how does this system actually work? We can get some idea, in broad brushstrokes, by starting with the excitatory pathway, then adding in the two inhibitory pathways, as shown by the different colors in Figure 1. If the network consisted of just the chordotonal organ mechanosensory neurons and the B neurons (the inputs and outputs of the network, indicated by green and black pathways), hunching would occur whenever appropriate chordotonal organ neurons were activated that excite B1 and B2 together, and bending would occur when the activation of B2 was so low that it could not produce its inhibitory effect that would produce hunching. Adding in the feedforward inhibitory circuit (the iLN connections, indicated by the red pathway) effectively produces a switch: if
one of the iLNs is even a little more strongly activated than the other one, the more weakly activated iLN will completely shut down. For instance, if the chordotonal organ input excites iLNb a little more than it excites iLNa, the iLNa receives more inhibition from iLNb than it gives back to iLNb, so the activity in iLNa decreases and iLNb increases. The increased iLNb activity further inhibits iLNa. This cycle repeats until iLNb is fully on and iLNa is fully turned off. The consequence is that the activity in B1 is decreased (because the strong inhibition from iLNb is activated) and the activity in B2 is large (because it has lost a major part of its inhibition, from iLNa). This condition produces a bend. On the other hand, if iLNa receives more excitation from chordotonal organ neurons than does iLNb, the opposite effect is seen, which produces bending. Adding the feedback inhibition provided by the H cells (violet pathway in Figure 1) adds multiple pathways for ‘disinhibiting’ (inhibiting the inhibition onto) the feedforward inhibitory pathway. This disinhibition can prolong whatever level of activity is present in the B cells, which would functionally lock in whichever behavior is being generated. Note that there are four weak connections that do not fit into the general functional scheme. For instance, iLNb weakly inhibits B2, a connection that opposes the strong disinhibitory pathway from iLNb to B2 via iLNa. The authors suggest that these pathways may serve to allow the circuit to escape from a locked-in state. Alternatively, these connections may be part of a circuit-wide gain control system. As well as being a tour de force of modern technologies applied to a timely systems neurobiological question, this new study is also interesting because the circuit features that it describes are being found (or at least proposed) in other systems. The selection of behaviors by different combinations of the same two pathways (to Hunch and to Bend) is the simplest form of a combinatorial code, in which the effect of one cell (in this case, B1) is qualitatively different depending upon whether another cell (B2) is also active. Such a combinatorial code for selecting between behaviors has been found in leeches [10]. In addition, the type of reciprocal inhibition seen in the feedforward pathway in this study —
R40 Current Biology 27, R19–R41, January 9, 2017
between neurons that elicit two different behaviors — was initially proposed by the ethologist Niko Tinbergen [11], based entirely upon observing natural behaviors of fish and birds. Electrophysiological evidence for inhibitory interactions between competing behavioral pathways has been found in animals such as mollusks [12], leeches [13], and monkeys [14]. Finally, the disinhibition seen in the feedback circuitry is similar to that seen in the basal ganglia of vertebrates [15], as well as in pathways that hold the position of the eyes after they move to a new position [16]. This parcellation of recognizable functions into distinct sub-circuits suggests that the same connectivity pattern (for example, reciprocal inhibition) can be used as a module to perform different functions (for example, behavioral selection or prolonging any ongoing behavior), depending upon where in the network the module is plugged in [17]. Giving clear examples of this kind of systems-level reasoning may ultimately be the most important contribution of the impressive piece of research by Jovanic et al. [4]. REFERENCES 1. Lerner, T.N., Ye, L., and Deisseroth, K. (2016). Communication in neural circuits: Tools, opportunities, and challenges. Cell 164, 1136–1150. 2. Titze, B., and Genoud, C. (2016). Volume scanning electron microscopy for imaging biological ultrastructure. Biol. Cell. 108, 307–323. 3. Bouchard, K.E., Aimone, J.B., Chun, M., Dean, T., Denker, M., Diesmann, M., Donofrio, D.D., Frank, L.M., Kasthuri, N., Koch, C., et al. (2016). High-performance computing in Neuroscience for data-driven discovery, integration, and dissemination. Neuron 92, 628–631. 4. Jovanic, T., Schneider-Mizell, C.M., Shao, M., Masson, J.B., Denisov, G., Fetter, R.D., Mensh, B.D., Truman, J.W., Cardona, A., and Zlatic, M. (2016). Competitive disinhibition mediates behavioral choice and sequences in Drosophila. Cell 167, 1–13. 5. Ohyama, T., Jovanic, T., Denisov, G., Dang, T.C., Hoffmann, D., Kerr, R.A., and Zlatic, M. (2013). High-throughput analysis of stimulusevoked behaviors in Drosophila larva reveals multiple modality-specific escape strategies. PLoS One 8, e71706. 6. Klapoetke, N.C., Murata, Y., Kim, S.S., Pulver, S.R., Birdsey-Benson, A., Cho, Y.K., Morimoto, T.K., Chuong, A.S., Carpenter, E.J., Tian, Z., et al. (2014). Independent optical excitation of distinct neural populations. Nat. Methods 11, 338–346.
Current Biology
Dispatches 7. Ohyama, T., Schneider-Mizell, C.M., Fetter, R.D., Aleman, J.V., Franconville, R., RiveraAlba, M., Mensh, B.D., Branson, K.M., Simpson, J.H., Truman, J.W., et al. (2015). A multilevel multimodal circuit enhances action selection in Drosophila. Nature 520, 633–639. 8. Schneider-Mizell, C.M., Gerhard, S., Longair, M., Kazimiers, T., Li, F., Zwart, M.F., Champion, A., Midgley, F.M., Fetter, R.D., Saalfeld, S., and Cardona, A. (2016). Quantitative neuroanatomy for connectomics in Drosophila. eLife 5, e12059. 9. Wojcik, J., Schwabedal, J., Clewley, R., and Shilnikov, A.L. (2014). Key bifurcations of bursting polyrhythms in 3-cell central pattern generators. PLoS One 9, e92918.
10. Kristan, W.B., and Gillette, R. (2007). Decisionmaking in small neuronal networks. In Invertebrate Neurobiology, G. North, and R. Greenspan, eds. (New York: Cold Spring Harbor Laboratory Press), pp. 533–554.
14. Wang, X.J. (2008). Decision making in recurrent neuronal circuits. Neuron 60, 215–234.
11. Tinbergen, N. (1951). The Study of Instinct (Oxford, Clarendon Press).
15. Mink, J.W. (1996). The basal ganglia: focused selection and inhibition of competing motor programs. Prog. Neurobiol. 50, 381–425.
12. Pirger, Z., Crossley, M., La´szlo´, Z., Naskar, S., Kemenes, G., O’Shea, M., Benjamin, P.R., and Kemenes, I. (2014). Interneuronal mechanism for Tinbergen’s hierarchical model of behavioral choice. Curr. Biol. 24, 2018–2024.
16. Aksay, E., Olasagasti, I., Mensh, B.D., Baker, R., Goldman, M.S., and Tank, D.W. (2007). Functional dissection of circuitry in a neural integrator. Nat. Neurosci. 10, 494–504.
13. Gaudry, Q., and Kristan, W.B., Jr. (2009). Behavioral choice by presynaptic inhibition of tactile sensory terminals. Nat. Neurosci. 12, 1450–1457.
17. Mysore, S.P., and Knudsen, E.I. (2012). Reciprocal inhibition of inhibition: a circuit motif for flexible categorization in stimulus selection. Neuron 73, 193–205.
Current Biology 27, R19–R41, January 9, 2017 R41