Hierarchical architecture of dopaminergic circuits enables second-order conditioning in Drosophila

Curation statements for this article:
  • Curated by eLife

    eLife logo

    Evaluation Summary:

    Second order conditioning is a higher form of learning where a previously conditioned stimulus (e.g. odor A by food) is used to condition the perception of another stimulus (e.g. odor B by odor A). Yamada et al. have used the fly to identify a neural circuit in the insect mushroom body underpinning second order conditioning. This work elegantly combines neural circuit mapping, electrophysiology and modeling to put forward a mechanistic model for this highly conserved form of learning.

    This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.

This article has been Reviewed by the following groups

Read the full article See related articles

Abstract

Dopaminergic neurons with distinct projection patterns and physiological properties compose memory subsystems in a brain. However, it is poorly understood whether or how they interact during complex learning. Here, we identify a feedforward circuit formed between dopamine subsystems and show that it is essential for second-order conditioning, an ethologically important form of higher-order associative learning. The Drosophila mushroom body comprises a series of dopaminergic compartments, each of which exhibits distinct memory dynamics. We find that a slow and stable memory compartment can serve as an effective ‘teacher’ by instructing other faster and transient memory compartments via a single key interneuron, which we identify by connectome analysis and neurotransmitter prediction. This excitatory interneuron acquires enhanced response to reward-predicting odor after first-order conditioning and, upon activation, evokes dopamine release in the ‘student’ compartments. These hierarchical connections between dopamine subsystems explain distinct properties of first- and second-order memory long known by behavioral psychologists.

Article activity feed

  1. Author Response

    Reviewer #3 (Public Review):

    Yamada et al utilizes the full strength of Drosophila neural circuit approaches to investigate second-order conditioning. The new insights into the mechanisms of how a learned cue can act as reinforcement are relevant beyond the fly field and have the potential to spark broad interest. The main conclusions of the authors are justified and the experiments, to my understanding, are well done.

    Some minor aspects must be addressed. To avoid misunderstandings a clear distinction should be made between those experiments using real world sugar and those using artificial activation of dopamine neurons as reward. For example, the proposed teacher - student model is mostly based on the work established with artificial activation.

    We split Figure 1 and made two separate figures. The new Figure 1 displays experiments with only real sugar or optogenetic activation of sugar receptor neurons (new data), whereas the new Figure 2 shows mostly experiments with direct DAN activations. This new figure arrangement makes a clear distinction between experiments with sugar and DAN activation, and allows readers to compare them more easily. We also modified the second paragraph of the discussion to clarify this point.

    To emphasize the generality of the model, it might help to provide some further evidence using real world sugar approaches, especially since the only known sugar-reward driven plasticity is reported in the student (g5b`2a) but not the teacher compartments. In this line, it would be useful to extend the functional interference used during the sugar experiments beyond the a1 compartment.

    In response to the reviewer’s comment, we added new data in Figure 2D to show that blocking PAM-DANs in γ4, γ5 and β′2a compartments impairs second-order conditioning following odor-sugar first-order conditioning. In contrast to blocking α1 DANs, blocking those non-α1 PAM-DANs did not impair one-day first-order memory (Figure 2D), which further strengthens our model of differential requirement of compartments for first-order and second-order memory formation.

    We think transient blocks of individual DAN cell types during second-order conditioning after odor-sugar conditioning will be informative to map second-order memories to specific compartments in naturalistic settings. For the reasons detailed above, however, we will need to develop a new way of transient purturbation for that.

    We would also point out that, to our knowledge, sugar-reward-driven plasticity has not been fully demonstrated in MBON-γ5β′2a. Owald et al., 2015 Neuron (10.1016/j.neuron.2015.03.025) showed a reduced CS+ odor response after odor-sugar conditioning in MBON-b′2mp (their Fig 3). However, they could not investigate the plasticity of MBON-γ5β′2a because the magnitude of odor response was too low (their Figure S3).

    Further, general statements about the compartments, for example for g5 and a1, might need adjustment since the tools used, the respective driver lines, often don't label all dopamine neurons in one specific compartment. In fact, functional heterogeneity among dopamine neurons innervating the g5 compartment have been recently established (sugar-reward, extinction) and might apply here.

    To clarify the point that we are manipulating a subset of DANs in each compartment, we added “cell count” information in Figure 2A. Also, we made Figure 4-figure supplement 2 to show which subtypes of DANs are connected with SMP108.

    Lastly, I would like to recommend that the authors discuss alternative feedback pathways that might serve similar or parallel functions.

    Despite these minor points, the study is impressive.

    Figure 4C shows several candidate interneurons that may have similar functions as SMP108. For instance, CRE011 may acquire enhanced response to reward-predicting odor as an outcome of reduced inhibition from MBON-γ5β′2a, and send excitatory inputs to DANs.

    In Figure 4-figure supplement 3, we made additional scatter plots to visualize other outlier cell types in terms of their connectivity with MBONs and DANs.

  2. Evaluation Summary:

    Second order conditioning is a higher form of learning where a previously conditioned stimulus (e.g. odor A by food) is used to condition the perception of another stimulus (e.g. odor B by odor A). Yamada et al. have used the fly to identify a neural circuit in the insect mushroom body underpinning second order conditioning. This work elegantly combines neural circuit mapping, electrophysiology and modeling to put forward a mechanistic model for this highly conserved form of learning.

    This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.

  3. Reviewer #1 (Public Review):

    Using a combination of behavioral screening, optogenetics, electrophysiology, connectomics and computational modelling, the authors identify a circuit in the fly mushroom body spanning at least two MB compartments.

    While the MB compartment alpha1 is involved in first order appetitive memory formation, at least one other compartment is modified to establish second order appetitive memory (gamma5, beta'2). The authors show that first order memory is very stable and long lasting. By contrast, second order memory decays within 24 h.
    Based on their behavioral data, the authors propose a circuit where the output neuron (MBON) of alpha1 modulates the dopaminergic neuron in the gamma5/beta'2 compartments.

    Electrophysiological recordings and EM connectomics indicate that learning-induced reduction of MBON-alpha1's output enhances the response of DANs innervating the other compartment. Interestingly, no direct connection could be identified between the MBON and the DANs across these compartments. Nevertheless, the authors find a connection over 2 synapses and a 'hub' neuron (SMP108) that connects them. Optogenetic modulation of SMP108 induces a similar memory with characteristics of second-order memory as observed through second order conditioning itself.

    Based on these data, the authors conclude that MBON-alpha1, an inhibitory neuron, is repressed after first order conditioning, in turn leading to an enhanced response to the conditioned odor of SMP108 and the DAN innervating gamma5 and beta'2. This enhanced response triggers a repression in MBONs of this second compartment and thereby induces second order appetitive memory.

    In my opinion, the experiments are strong and of high quality. The electrophysiology is a very convincing addition to the behavioral experiments. The EM data is helpful, but as appears to be the case frequently, also confusing since it suggests multiple, and no direct, routes (possibly redundant) between the first and the second involved MB compartment. This is possibly a weakness of the study, but it emphasizes the biological circumstance that (most of the time) things are more complex than we'd expect them to be given the paradigm that's being studied. The modelling is a nice addition, perhaps not strictly necessary, that helps to dissect this connection complexity.

  4. Reviewer #2 (Public Review):

    The presented manuscript by Yamada and colleagues identifies a circuit mechanism underlying second-order conditioning in the mushroom bodies (MB) of Drosophila. The authors use an impressive set of techniques, ranging from precise genetic manipulations and behavioral analyses to neurophysiological methods and EM connectomics. The newly devised behavioural protocols allow for second-order conditioning in Drosophila and the identification of the mushroom body (MB) compartments involved. The authors furthermore not only identify a novel class of neurons involved in second-order conditioning, but also work out which route plastic changes take along mushroom body output and dopaminergic neurons. Strikingly, similar to motifs previously shown to, for instance, mediate hunger gating or memory extinction, this novel pathway connects different MB compartments, adding a new computational motif to the MB map. This paper is an important contribution for understanding general principles of circuit logic and plasticity underlying neural computations in biological systems at high resolution.

  5. Reviewer #3 (Public Review):

    Yamada et al utilizes the full strength of Drosophila neural circuit approaches to investigate second-order conditioning. The new insights into the mechanisms of how a learned cue can act as reinforcement are relevant beyond the fly field and have the potential to spark broad interest. The main conclusions of the authors are justified and the experiments, to my understanding, are well done.

    Some minor aspects must be addressed. To avoid misunderstandings a clear distinction should be made between those experiments using real world sugar and those using artificial activation of dopamine neurons as reward. For example, the proposed teacher - student model is mostly based on the work established with artificial activation. To emphasize the generality of the model, it might help to provide some further evidence using real world sugar approaches, especially since the only known sugar-reward driven plasticity is reported in the student (g5b`2a) but not the teacher compartments. In this line, it would be useful to extend the functional interference used during the sugar experiments beyond the a1 compartment. Further, general statements about the compartments, for example for g5 and a1, might need adjustment since the tools used, the respective driver lines, often don't label all dopamine neurons in one specific compartment. In fact, functional heterogeneity among dopamine neurons innervating the g5 compartment have been recently established (sugar-reward, extinction) and might apply here. Lastly, I would like to recommend that the authors discuss alternative feedback pathways that might serve similar or parallel functions.

    Despite these minor points, the study is impressive.