- Email: [email protected]
An Elegant Circuit for Balancing Risk and Reward
Neuron
Previews explain VOR learning highlights the importance of understanding the function(s) of the additional sites and mechanisms of plasticity that are now well established experimentally. Suvrathan et al. (2016) also raise a host of questions for future studies of cerebellum beyond VOR learning. Is LTD in the vermis (Suvrathan et al., 2016 only tested short-term plasticity) also sharply timing dependent with different Purkinje cells tuned to different parallel fiberclimbing fiber intervals? What are the molecular mechanisms underlying the observed timing-dependent plasticity rules? Is the optimal delay for inducing associative depression in a given Purkinje cell fixed, or can it be shaped by experience or early development? Using modern viral and genetic tracing methods, it may be possible to directly test the provocative speculation that the optimal delay for associative depression in a Purkinje cell is matched to the delay of the climbing fiber error signal it receives. Finally, from a theoretical point of view there are two particularly interesting aspects of the findings of Suvrathan et al. (2016). One is the possibility that spike-
timing-dependent plasticity (STDP) windows can be shifted without an increase in their width so that learning proceeds based on pre- and postsynaptic spikes occurring within a narrow set of delays. Second is the possibility that there is diversity in the delay of such windows at the level of individual neurons of the same type within a brain region. Apart from providing an elegant solution to the specific problem of learning based on delayed error signals, we can think of the plasticity rules described by Suvrathan et al. (2016) as implementing the more general computation of learning based on the cross-correlation of signals at a non-zero delay. Standard STDP rules tuned to very brief intervals between pre- and postsynaptic spikes have been studied extensively by experimentalists and theorists (Abbott and Nelson, 2000). The findings of Suvrathan et al. (2016) should prompt further consideration of the potential uses of cross-correlational plasticity. For example, such crosscorrelational plasticity might help bridge the timescales of STDP with timescales relevant to behavior (Drew and Abbott, 2006) and may allow single neurons to
extract more complex features of their inputs (Clopath et al., 2008). REFERENCES Abbott, L.F., and Nelson, S.B. (2000). Nat. Neurosci. 3 (Suppl ), 1178–1183. Boyden, E.S., Katoh, A., and Raymond, J.L. (2004). Annu. Rev. Neurosci. 27, 581–609. Chen, C., and Thompson, R.F. (1995). Learn. Mem. 2, 185–198. Clopath, C., Longtin, A., and Gerstner, W. (2008). Adv. Neural Inf. Process. Syst. 20, 321–328. Drew, P.J., and Abbott, L.F. (2006). Proc. Natl. Acad. Sci. USA 103, 8876–8881. Gao, Z., van Beugen, B.J., and De Zeeuw, C.I. (2012). Nat. Rev. Neurosci. 13, 619–635. Ito, M., and Kano, M. (1982). Neurosci. Lett. 33, 253–258. Jo¨rntell, H., and Hansel, C. (2006). Neuron 52, 227–238. Raymond, J.L., and Lisberger, S.G. (1998). J. Neurosci. 18, 9112–9129. Safo, P., and Regehr, W.G. (2008). Neuropharmacology 54, 213–218. Suvrathan, A., Payne, H.L., and Raymond, J.L. (2016). Neuron 92, this issue, 959–967.
An Elegant Circuit for Balancing Risk and Reward Zhaoyu Li,1,2 Adam J. Iliff,1,2 and X.Z. Shawn Xu1,* 1Life
Sciences Institute, Department of Molecular & Integrative Physiology, University of Michigan, Ann Arbor, MI 48109, USA author *Correspondence: [email protected] http://dx.doi.org/10.1016/j.neuron.2016.11.041 2Co-first
Animals constantly encounter conflicting cues in natural environments. To survive and thrive, they must make appropriate behavioral decisions. In this issue, Ghosh et al. (2016) identified a neural circuit underlying multisensory threat-reward decision making using an elegant C. elegans model. To survive the diverse and changing environment, animals must evaluate potential threats and rewards to make complex decisions. To thrive, animals also have to balance threat tolerance and the potential benefits of finding food. Understanding the circuit and molecular mechanisms that underlie a complex decision-making process is challenging in a highly complex
organism with a large nervous system. Perhaps even more challenging is elucidating the mechanisms of behavioral plasticity that modulate intrinsic decision-making circuits. Thus, being able to address these issues in a simpler nervous system has the potential to uncover novel mechanisms that could guide studies in higher systems. The relatively simple and well-an-
notated nervous system of the nematode C. elegans has proven to be an ideal model system for deciphering the functional circuits underlying behavior. In more recent years, C. elegans sensory integration has emerged as a major focus of research. In this issue, Nitabach and colleagues report a novel top-down neural circuit underlying multimodal sensory integration and how
Neuron 92, December 7, 2016 ª 2016 Elsevier Inc. 933
Neuron
Previews changes in the internal state such as hunger modify this process to alter behavioral choices (Ghosh et al., 2016).
motic barrier. Without tyramine, ASH and RIM activities gradually decline, and evenTYRA-2 TYRA-2 tually the balanced activity ASH AWA ASH AWA levels permit the worm to A Top-Down Circuit for escape from the fructose ring Multisensory Decision to approach the food odor source (Figure 1). Taken Making RIM RIM together, these results sugThe investigators developed an assay to map the imporgest that positive tyraminergic feedback from RIM to tant events that go into the ASH forms a top-down circuit decision to cross an aversive Threat averse Threat tolerant barrier to get to the source of that regulates threat-reward decision making (Figure 1). an attractive food-related Since RIM expresses the cue (Figure 1), by modifying a paradigm previously develneuropeptide PDF-2, they oped by Ishihara et al. examined hyperosmolarity avoidance in pdf-2 null (2002). Diacetyl is a chemical AƩractant mutant worms. pdf-2 null mufound in food, and the odor is attractive to C. elegans. tants exhibited increased threat tolerance, similar to However, hyperosmolarity is Repellent barrier Repellent barrier tdc-1 and tyra-2 mutants. an aversive stimulus to worms, and when surrounded RIM also expresses PDFR-1, Figure 1. Schematic of a Top-Down Circuit for Multisensory the cognate receptor of the by a ring of 2M fructose, Decision Making Internal hunger state modulates the circuit and increases threat tolerance, secreted neuropeptide PDFworms avoid the high osmopermitting worms to cross the hyperosmotic repellent barrier (fructose) to larity and do not exit the ring 2. To determine whether PDF approach the attractant (diacetyl odor). (Figure 1). Therefore, worms signaling functions in RIM as an autocrine loop, they exmust balance the drive for food with the threat of hyperosmolarity- bach and colleagues screened mutants pressed a membrane-tethered version of lacking tyramine receptors and found PDF-2 in RIM to activate PDFR-1 and induced desiccation and death. In this simple paradigm, diacetyl is de- that tyra-2 mutant worms exhibit defects found a complete rescue of the decision tected by an olfactory neuron AWA, and in threat-reward decision making. To balance. Thus, RIM feedback tyraminerfructose is perceived by a nociceptive identify the functional site of tyra-2, Ghosh gic signaling and autocrine PDF-2 neuron ASH (Figure 1). Sensory activities et al. (2016) performed neuron-specific signaling together promote RIM-ASH of these neurons are further processed rescuing experiments. Surprisingly, they top-down positive feedback, which dein the interneuron layer. The authors found found the expression of tyra-2 gene in creases threat tolerance (Figure 1). that the interneuron RIM plays an impor- ASH neuron can fully restore the defects tant role in threat-reward decision making in tyra-2 null mutant worms, suggesting Internal State Modulates Decision (Figure 1). RIM functions in the locomotion that RIM generates a feedback tyramine Making circuitry, but unlike other locomotion signal to regulate threat-reward decision Another major discovery of Ghosh et al. neurons which either promote or inhibit making. Calcium imaging experiments re- (2016) is that the internal physiological movement in one direction, RIM has vealed that tyramine modulates ASH ac- state modulates decision-making circuit been reported to promote both forward tivity, as application of exogenous tyra- and thereby reshapes the behavioral and reverse motion (Kato et al., 2015; Pig- mine increased ASH calcium responses output. They found that after 1 hr of food gott et al., 2011). Since RIM releases the toward high-osmolarity fructose. In order deprivation, worms are more likely to neurotransmitter tyramine (Alkema et al., to investigate whether the neural activity exit the hyperosmotic barrier, indicating 2005), they tested whether tyramine is dynamics would support RIM to execute that the internal physiological state moduinvolved in threat-reward decision making threat-reward decision making by means lates this decision. The authors hypotheby examining tdc-1 null mutant worms. of ‘‘top-down’’ control, the authors devel- sized that food deprivation suppresses tdc-1 encodes tyrosine decarboxylase, oped a computational model of this RIM activity and thus the RIM-ASH posiwhich is required for tyramine biogenesis circuit. As worms move on the assay tive feedback circuit. To test this idea, (Alkema et al., 2005). The investigators plate, they encounter alternating levels they blocked the RIM-ASH feedback usfound that tdc-1 null mutants show of hyperosmolarity and diacetyl. In silico, ing tyra-2 mutation and found that tyra-2 increased exiting, and this defect can be this leads to oscillations in AWA, ASH, mutant worms fail to increase exiting after rescued by expressing tdc-1 gene just in and RIM neural activity. With tyramine 1 hr of food deprivation, suggesting that RIM, suggesting that tyramine release signaling intact, the activity of ASH and TYRA-2 positive feedback signaling is from RIM decreases threat tolerance. To RIM overtakes the activity of AWA, and required for internal hunger state modulaidentify its downstream signaling, Nita- the worm remains inside the hyperos- tion of threat tolerance. The investigators
934 Neuron 92, December 7, 2016
Fed
Starved
Neuron
Previews suggested that the internal physiological state, including hunger, reshapes threatreward decision making by modifying this neural circuit. Other factors have previously been reported to modulate decision making. Bendesky et al. discovered that endogenous catecholamines act on TYRA-3 in sensory neurons to modulate an exploration-exploitation decision (Bendesky et al., 2011). Feeding state is also reported to modulate ASH nociception through the interaction of dopaminergic and neuropeptide signaling pathways (Ezcurra et al., 2016). These studies indicate that circuits are not necessarily hard-wired networks, but rather the internal physiological state may endow circuits with more functional dimensionality. An interesting parallel can be drawn between these studies and those in humans and other mammals. For example, top-down circuits are involved in accurate perception and multisensory integration in mammals, suggesting that neural networks in different species employ similar circuit principles to process decision making (Manita et al., 2015). This exciting study also raises some interesting questions. For example, with increasing lengths of food deprivation, the threat tolerance increases. As tyra-2 is mainly responsible for the regulation
of short-term (1 hr) food deprivation, it would be interesting to identify other genes responsible for the regulation of longerperiod food deprivation. The present study mainly focuses on the interneuron RIM. Previous work reported that the first layer interneurons AIA and AIY are also involved in a related decision-making process (i.e., diacetyl-copper choice assay), which involves the same set of sensory neurons AWA and ASH (Shinkai et al., 2011). Could these interneurons also employ topdown circuitry to integrate multisensory signaling? As ASH is a polymodal nociceptor, it would also be interesting to determine whether hunger-induced modification of the RIM-ASH feedback circuit would extend to other ASH-cued sensory behavior, such as alkaline pH avoidance and copper avoidance (Ishihara et al., 2002; Wang et al., 2016). The current study from Ghosh et al. (2016) presents an elegant case of harnessing the power of a simple model organism to identify fundamental neural and genetic mechanisms underlying decision making at the single neuron resolution.
ACKNOWLEDGMENTS A.J.I. is supported by a T32 training grant (T32DC00011) and an NRSA fellowship grant (F32DC015381) from the NIDCD. Research in the
Xu lab is supported by grants from the NIH (R01GM083241; R01AG048072).
REFERENCES Alkema, M.J., Hunter-Ensor, M., Ringstad, N., and Horvitz, H.R. (2005). Neuron 46, 247–260. Bendesky, A., Tsunozaki, M., Rockman, M.V., Kruglyak, L., and Bargmann, C.I. (2011). Nature 472, 313–318. Ezcurra, M., Walker, D.S., Beets, I., Swoboda, P., and Schafer, W.R. (2016). J. Neurosci. 36, 3157– 3169. Ghosh, D.D., Sanders, T., Hong, S., McCurdy, L.Y., Chase, D.L., Cohen, N., Koelle, M.R., and Nitabach, M.N. (2016). Neuron 92, this issue, 1049–1062. Ishihara, T., Iino, Y., Mohri, A., Mori, I., GengyoAndo, K., Mitani, S., and Katsura, I. (2002). Cell 109, 639–649. Kato, S., Kaplan, H.S., Schro¨del, T., Skora, S., Lindsay, T.H., Yemini, E., Lockery, S., and Zimmer, M. (2015). Cell 163, 656–669. Manita, S., Suzuki, T., Homma, C., Matsumoto, T., Odagawa, M., Yamada, K., Ota, K., Matsubara, C., Inutsuka, A., Sato, M., et al. (2015). Neuron 86, 1304–1316. Piggott, B.J., Liu, J., Feng, Z., Wescott, S.A., and Xu, X.Z.S. (2011). Cell 147, 922–933. Shinkai, Y., Yamamoto, Y., Fujiwara, M., Tabata, T., Murayama, T., Hirotsu, T., Ikeda, D.D., Tsunozaki, M., Iino, Y., Bargmann, C.I., et al. (2011). J. Neurosci. 31, 3007–3015. Wang, X., Li, G., Liu, J., Liu, J., and Xu, X.Z. (2016). Neuron 91, 146–154.
Plasticity to the Rescue Tatyana O. Sharpee1,* 1Computational Neurobiology Laboratory, Salk Institute for Biological Studies, La Jolla, CA 92037, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.neuron.2016.11.042
The balance between excitatory and inhibitory inputs is critical for the proper functioning of neural circuits. Landau and colleagues show that, in the presence of cell-type-specific connectivity, this balance is difficult to achieve without either synaptic plasticity or spike-frequency adaptation to fine-tune the connection strengths. How neural circuits achieve reliable computation with unreliable components has long been appreciated as a key and unique property of biological computation (von Neumann, 1956). Metabolic resources limit the attainable accuracy of in-
dividual components (Faisal et al., 2008), and how these resources are distributed across the stages of neural processing so as to achieve the best overall performance is an open question. This general argument leads to several corollaries.
The first corollary is that it is not worth improving the accuracy of one part of the circuit if later processing stages will not take advantage of that improvement. In other words, metabolic resources need to be distributed in a ‘‘balanced’’
Neuron 92, December 7, 2016 ª 2016 Published by Elsevier Inc. 935
Previews explain VOR learning highlights the importance of understanding the function(s) of the additional sites and mechanisms of plasticity that are now well established experimentally. Suvrathan et al. (2016) also raise a host of questions for future studies of cerebellum beyond VOR learning. Is LTD in the vermis (Suvrathan et al., 2016 only tested short-term plasticity) also sharply timing dependent with different Purkinje cells tuned to different parallel fiberclimbing fiber intervals? What are the molecular mechanisms underlying the observed timing-dependent plasticity rules? Is the optimal delay for inducing associative depression in a given Purkinje cell fixed, or can it be shaped by experience or early development? Using modern viral and genetic tracing methods, it may be possible to directly test the provocative speculation that the optimal delay for associative depression in a Purkinje cell is matched to the delay of the climbing fiber error signal it receives. Finally, from a theoretical point of view there are two particularly interesting aspects of the findings of Suvrathan et al. (2016). One is the possibility that spike-
timing-dependent plasticity (STDP) windows can be shifted without an increase in their width so that learning proceeds based on pre- and postsynaptic spikes occurring within a narrow set of delays. Second is the possibility that there is diversity in the delay of such windows at the level of individual neurons of the same type within a brain region. Apart from providing an elegant solution to the specific problem of learning based on delayed error signals, we can think of the plasticity rules described by Suvrathan et al. (2016) as implementing the more general computation of learning based on the cross-correlation of signals at a non-zero delay. Standard STDP rules tuned to very brief intervals between pre- and postsynaptic spikes have been studied extensively by experimentalists and theorists (Abbott and Nelson, 2000). The findings of Suvrathan et al. (2016) should prompt further consideration of the potential uses of cross-correlational plasticity. For example, such crosscorrelational plasticity might help bridge the timescales of STDP with timescales relevant to behavior (Drew and Abbott, 2006) and may allow single neurons to
extract more complex features of their inputs (Clopath et al., 2008). REFERENCES Abbott, L.F., and Nelson, S.B. (2000). Nat. Neurosci. 3 (Suppl ), 1178–1183. Boyden, E.S., Katoh, A., and Raymond, J.L. (2004). Annu. Rev. Neurosci. 27, 581–609. Chen, C., and Thompson, R.F. (1995). Learn. Mem. 2, 185–198. Clopath, C., Longtin, A., and Gerstner, W. (2008). Adv. Neural Inf. Process. Syst. 20, 321–328. Drew, P.J., and Abbott, L.F. (2006). Proc. Natl. Acad. Sci. USA 103, 8876–8881. Gao, Z., van Beugen, B.J., and De Zeeuw, C.I. (2012). Nat. Rev. Neurosci. 13, 619–635. Ito, M., and Kano, M. (1982). Neurosci. Lett. 33, 253–258. Jo¨rntell, H., and Hansel, C. (2006). Neuron 52, 227–238. Raymond, J.L., and Lisberger, S.G. (1998). J. Neurosci. 18, 9112–9129. Safo, P., and Regehr, W.G. (2008). Neuropharmacology 54, 213–218. Suvrathan, A., Payne, H.L., and Raymond, J.L. (2016). Neuron 92, this issue, 959–967.
An Elegant Circuit for Balancing Risk and Reward Zhaoyu Li,1,2 Adam J. Iliff,1,2 and X.Z. Shawn Xu1,* 1Life
Sciences Institute, Department of Molecular & Integrative Physiology, University of Michigan, Ann Arbor, MI 48109, USA author *Correspondence: [email protected] http://dx.doi.org/10.1016/j.neuron.2016.11.041 2Co-first
Animals constantly encounter conflicting cues in natural environments. To survive and thrive, they must make appropriate behavioral decisions. In this issue, Ghosh et al. (2016) identified a neural circuit underlying multisensory threat-reward decision making using an elegant C. elegans model. To survive the diverse and changing environment, animals must evaluate potential threats and rewards to make complex decisions. To thrive, animals also have to balance threat tolerance and the potential benefits of finding food. Understanding the circuit and molecular mechanisms that underlie a complex decision-making process is challenging in a highly complex
organism with a large nervous system. Perhaps even more challenging is elucidating the mechanisms of behavioral plasticity that modulate intrinsic decision-making circuits. Thus, being able to address these issues in a simpler nervous system has the potential to uncover novel mechanisms that could guide studies in higher systems. The relatively simple and well-an-
notated nervous system of the nematode C. elegans has proven to be an ideal model system for deciphering the functional circuits underlying behavior. In more recent years, C. elegans sensory integration has emerged as a major focus of research. In this issue, Nitabach and colleagues report a novel top-down neural circuit underlying multimodal sensory integration and how
Neuron 92, December 7, 2016 ª 2016 Elsevier Inc. 933
Neuron
Previews changes in the internal state such as hunger modify this process to alter behavioral choices (Ghosh et al., 2016).
motic barrier. Without tyramine, ASH and RIM activities gradually decline, and evenTYRA-2 TYRA-2 tually the balanced activity ASH AWA ASH AWA levels permit the worm to A Top-Down Circuit for escape from the fructose ring Multisensory Decision to approach the food odor source (Figure 1). Taken Making RIM RIM together, these results sugThe investigators developed an assay to map the imporgest that positive tyraminergic feedback from RIM to tant events that go into the ASH forms a top-down circuit decision to cross an aversive Threat averse Threat tolerant barrier to get to the source of that regulates threat-reward decision making (Figure 1). an attractive food-related Since RIM expresses the cue (Figure 1), by modifying a paradigm previously develneuropeptide PDF-2, they oped by Ishihara et al. examined hyperosmolarity avoidance in pdf-2 null (2002). Diacetyl is a chemical AƩractant mutant worms. pdf-2 null mufound in food, and the odor is attractive to C. elegans. tants exhibited increased threat tolerance, similar to However, hyperosmolarity is Repellent barrier Repellent barrier tdc-1 and tyra-2 mutants. an aversive stimulus to worms, and when surrounded RIM also expresses PDFR-1, Figure 1. Schematic of a Top-Down Circuit for Multisensory the cognate receptor of the by a ring of 2M fructose, Decision Making Internal hunger state modulates the circuit and increases threat tolerance, secreted neuropeptide PDFworms avoid the high osmopermitting worms to cross the hyperosmotic repellent barrier (fructose) to larity and do not exit the ring 2. To determine whether PDF approach the attractant (diacetyl odor). (Figure 1). Therefore, worms signaling functions in RIM as an autocrine loop, they exmust balance the drive for food with the threat of hyperosmolarity- bach and colleagues screened mutants pressed a membrane-tethered version of lacking tyramine receptors and found PDF-2 in RIM to activate PDFR-1 and induced desiccation and death. In this simple paradigm, diacetyl is de- that tyra-2 mutant worms exhibit defects found a complete rescue of the decision tected by an olfactory neuron AWA, and in threat-reward decision making. To balance. Thus, RIM feedback tyraminerfructose is perceived by a nociceptive identify the functional site of tyra-2, Ghosh gic signaling and autocrine PDF-2 neuron ASH (Figure 1). Sensory activities et al. (2016) performed neuron-specific signaling together promote RIM-ASH of these neurons are further processed rescuing experiments. Surprisingly, they top-down positive feedback, which dein the interneuron layer. The authors found found the expression of tyra-2 gene in creases threat tolerance (Figure 1). that the interneuron RIM plays an impor- ASH neuron can fully restore the defects tant role in threat-reward decision making in tyra-2 null mutant worms, suggesting Internal State Modulates Decision (Figure 1). RIM functions in the locomotion that RIM generates a feedback tyramine Making circuitry, but unlike other locomotion signal to regulate threat-reward decision Another major discovery of Ghosh et al. neurons which either promote or inhibit making. Calcium imaging experiments re- (2016) is that the internal physiological movement in one direction, RIM has vealed that tyramine modulates ASH ac- state modulates decision-making circuit been reported to promote both forward tivity, as application of exogenous tyra- and thereby reshapes the behavioral and reverse motion (Kato et al., 2015; Pig- mine increased ASH calcium responses output. They found that after 1 hr of food gott et al., 2011). Since RIM releases the toward high-osmolarity fructose. In order deprivation, worms are more likely to neurotransmitter tyramine (Alkema et al., to investigate whether the neural activity exit the hyperosmotic barrier, indicating 2005), they tested whether tyramine is dynamics would support RIM to execute that the internal physiological state moduinvolved in threat-reward decision making threat-reward decision making by means lates this decision. The authors hypotheby examining tdc-1 null mutant worms. of ‘‘top-down’’ control, the authors devel- sized that food deprivation suppresses tdc-1 encodes tyrosine decarboxylase, oped a computational model of this RIM activity and thus the RIM-ASH posiwhich is required for tyramine biogenesis circuit. As worms move on the assay tive feedback circuit. To test this idea, (Alkema et al., 2005). The investigators plate, they encounter alternating levels they blocked the RIM-ASH feedback usfound that tdc-1 null mutants show of hyperosmolarity and diacetyl. In silico, ing tyra-2 mutation and found that tyra-2 increased exiting, and this defect can be this leads to oscillations in AWA, ASH, mutant worms fail to increase exiting after rescued by expressing tdc-1 gene just in and RIM neural activity. With tyramine 1 hr of food deprivation, suggesting that RIM, suggesting that tyramine release signaling intact, the activity of ASH and TYRA-2 positive feedback signaling is from RIM decreases threat tolerance. To RIM overtakes the activity of AWA, and required for internal hunger state modulaidentify its downstream signaling, Nita- the worm remains inside the hyperos- tion of threat tolerance. The investigators
934 Neuron 92, December 7, 2016
Fed
Starved
Neuron
Previews suggested that the internal physiological state, including hunger, reshapes threatreward decision making by modifying this neural circuit. Other factors have previously been reported to modulate decision making. Bendesky et al. discovered that endogenous catecholamines act on TYRA-3 in sensory neurons to modulate an exploration-exploitation decision (Bendesky et al., 2011). Feeding state is also reported to modulate ASH nociception through the interaction of dopaminergic and neuropeptide signaling pathways (Ezcurra et al., 2016). These studies indicate that circuits are not necessarily hard-wired networks, but rather the internal physiological state may endow circuits with more functional dimensionality. An interesting parallel can be drawn between these studies and those in humans and other mammals. For example, top-down circuits are involved in accurate perception and multisensory integration in mammals, suggesting that neural networks in different species employ similar circuit principles to process decision making (Manita et al., 2015). This exciting study also raises some interesting questions. For example, with increasing lengths of food deprivation, the threat tolerance increases. As tyra-2 is mainly responsible for the regulation
of short-term (1 hr) food deprivation, it would be interesting to identify other genes responsible for the regulation of longerperiod food deprivation. The present study mainly focuses on the interneuron RIM. Previous work reported that the first layer interneurons AIA and AIY are also involved in a related decision-making process (i.e., diacetyl-copper choice assay), which involves the same set of sensory neurons AWA and ASH (Shinkai et al., 2011). Could these interneurons also employ topdown circuitry to integrate multisensory signaling? As ASH is a polymodal nociceptor, it would also be interesting to determine whether hunger-induced modification of the RIM-ASH feedback circuit would extend to other ASH-cued sensory behavior, such as alkaline pH avoidance and copper avoidance (Ishihara et al., 2002; Wang et al., 2016). The current study from Ghosh et al. (2016) presents an elegant case of harnessing the power of a simple model organism to identify fundamental neural and genetic mechanisms underlying decision making at the single neuron resolution.
ACKNOWLEDGMENTS A.J.I. is supported by a T32 training grant (T32DC00011) and an NRSA fellowship grant (F32DC015381) from the NIDCD. Research in the
Xu lab is supported by grants from the NIH (R01GM083241; R01AG048072).
REFERENCES Alkema, M.J., Hunter-Ensor, M., Ringstad, N., and Horvitz, H.R. (2005). Neuron 46, 247–260. Bendesky, A., Tsunozaki, M., Rockman, M.V., Kruglyak, L., and Bargmann, C.I. (2011). Nature 472, 313–318. Ezcurra, M., Walker, D.S., Beets, I., Swoboda, P., and Schafer, W.R. (2016). J. Neurosci. 36, 3157– 3169. Ghosh, D.D., Sanders, T., Hong, S., McCurdy, L.Y., Chase, D.L., Cohen, N., Koelle, M.R., and Nitabach, M.N. (2016). Neuron 92, this issue, 1049–1062. Ishihara, T., Iino, Y., Mohri, A., Mori, I., GengyoAndo, K., Mitani, S., and Katsura, I. (2002). Cell 109, 639–649. Kato, S., Kaplan, H.S., Schro¨del, T., Skora, S., Lindsay, T.H., Yemini, E., Lockery, S., and Zimmer, M. (2015). Cell 163, 656–669. Manita, S., Suzuki, T., Homma, C., Matsumoto, T., Odagawa, M., Yamada, K., Ota, K., Matsubara, C., Inutsuka, A., Sato, M., et al. (2015). Neuron 86, 1304–1316. Piggott, B.J., Liu, J., Feng, Z., Wescott, S.A., and Xu, X.Z.S. (2011). Cell 147, 922–933. Shinkai, Y., Yamamoto, Y., Fujiwara, M., Tabata, T., Murayama, T., Hirotsu, T., Ikeda, D.D., Tsunozaki, M., Iino, Y., Bargmann, C.I., et al. (2011). J. Neurosci. 31, 3007–3015. Wang, X., Li, G., Liu, J., Liu, J., and Xu, X.Z. (2016). Neuron 91, 146–154.
Plasticity to the Rescue Tatyana O. Sharpee1,* 1Computational Neurobiology Laboratory, Salk Institute for Biological Studies, La Jolla, CA 92037, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.neuron.2016.11.042
The balance between excitatory and inhibitory inputs is critical for the proper functioning of neural circuits. Landau and colleagues show that, in the presence of cell-type-specific connectivity, this balance is difficult to achieve without either synaptic plasticity or spike-frequency adaptation to fine-tune the connection strengths. How neural circuits achieve reliable computation with unreliable components has long been appreciated as a key and unique property of biological computation (von Neumann, 1956). Metabolic resources limit the attainable accuracy of in-
dividual components (Faisal et al., 2008), and how these resources are distributed across the stages of neural processing so as to achieve the best overall performance is an open question. This general argument leads to several corollaries.
The first corollary is that it is not worth improving the accuracy of one part of the circuit if later processing stages will not take advantage of that improvement. In other words, metabolic resources need to be distributed in a ‘‘balanced’’
Neuron 92, December 7, 2016 ª 2016 Published by Elsevier Inc. 935