by Georgia Dempsey
To find food and avoid danger in changing environments, animals benefit from learning to associate certain cues, such as odours, with pleasant or unpleasant experiences. Memories can be of positive or negative valence, reinforced by either rewarding or punitive stimuli. For example, the work of Pavlov demonstrated how animals can be conditioned to link previously neutral stimuli (e.g., a bell) with a rewarding event (e.g., food presentation) (Pavlov, 1927). Through repeated food presentation after ringing the bell, the animal learned to link the conditioned stimulus (bell) and the unconditioned stimulus (natural food reward). This associative learning is also observed in fruit flies. In a standard learning protocol, flies are trained to associate an odour with a shock (aversive/punishment) or sugar (appetitive/reward). They are then presented with the previously paired odour and a neutral odour. When presented with this binary choice, flies choose to approach or avoid the conditioned odour, depending on the valence during training. Currently, how this information is coded in biological neural networks and utilised to guide behaviour is poorly understood.
In the fly brain, a structure known as the mushroom body (MB) plays a central role in associative learning. Here – similar to mammalian brains – dopaminergic neurons (DANs) release the neurotransmitter dopamine, which acts as a teaching signal, in response to reward or punishment. They thereby modulate the strength of connections between Kenyon cells (KCs), which carry information about odours and other sensory cues, and sets of MB output neurons (MBONs), which drive different behaviours.
KCs consist of three main classes (α/β, α′/β′ and γ KCs) which process sensory (olfactory, visual and thermo/hygrosensory) input alongside other newly emerging subtypes (Tanaka et al., 2008, Marin et al., 2020). Each typical MBON has dendrites along a subset of the parallel axons of ~2000 Kenyon Cells (Aso et al. 2014), making up the mushroom body lobes (α, β, α′, β′ and γ). Anatomically distinct DANs innervate 15 unique zones, or compartments, of the mushroom body lobes, contacting synapses between different populations of KCs and MBONs (Fig.1). Compartments innervated by the protocerebral anterior medial (PAM) DAN cluster are widely associated with reinforcing reward signalling such as sugar learning, whereas the protocerebral posterior lateral 1 (PPL1) cluster DANs are associated with reinforcing punishment signals, such as electric shock or bitter taste. Interestingly, behavioral experiments show that a single neuron of the PPL1 cluster innervating the γ1 compartment of the MB – the PPL1-γ1pedc DAN – is involved with reinforcement of short term negative memory, but also with regulating motivational state-dependent long-term positive memory processes. Moreover, PAM neurons innervating the γ5 compartment of the MB (the PAM-γ5 DANs) can, in addition to their well known role in reinforcing short term sugar reward memory, write positive memories when an “expected punishment” does not occur (Felsenberg et al. 2018, for review Cognini et al. 2017). We wondered how this compartmental multifunctionality is implemented within relatively confined DAN populations or even single neurons.
In Otto et al., 2020, we used a recent whole-brain electron microscopy volume (Zheng et al., 2018) to reconstruct the fine anatomy of three DAN types: PAM-γ5, PAM-β´2a, and PPL1-γ1pedc DANs (Fig. 2). Through this, we identified previously uncharacterised anatomical subtypes within the PAM population of each compartment. In addition, we observed that the single PPL1 neuron has a very diverse quadripartite dendrite. Such anatomical subdivision in DANs innervating single compartments may give rise to the functionally-relevant diversity of DANs, enabling parallel processing during associative learning.
To study functional heterogeneity, we also reconstructed ~821 DAN input neurons putatively providing the information on unconditioned stimuli to DAN dendrites, using CATMAID (Saalfeld et al., 2009). By performing hierarchical clustering, we found morphological cell types, based on key anatomical features, such as soma tract and primary neurite (Fig. 4). This showed that while there are shared inputs across PAM DAN subtypes, many such subtypes receive specialised sets of inputs, further distinguishing them from others in the same compartment. To relate subtypes with memory formation in live flies, we first used computational tools (www.natverse.org) to match the same neurons across databases of genetic tools that allow manipulation of the neurons during learning assays. This way, after matching our reconstructed neurons to a collection of transgenic lines (GAL4 and split-GAL4), we were able to manipulate the neurons’ function in the aforementioned learning protocol. This allowed us to correlate behaviour and respective DAN connectivity for some of the input neuron clusters.
In the DAN input neuron population, certain neurons specifically connect to subtypes of the rewarding PAM population. Others, surprisingly, connect to specific parts of the PPL1-γ1pedc dendrite. This suggests that activation of specific upstream neurons would lead to particular processing of fine tuned positive or negative memories. For example, one input neuron group from the gustatory information processing center, that we showed to be involved with sugar reward learning, was connected to a subtype of γ5 DANs (γ5-uc) which we also showed to contribute the teaching signal for the corresponding positive memory. In contrast, another subtype of PAM-γ5 DANs (γ5-fb) manipulating plasticity of KC to MBON synapses in the γ5 compartment – and thus the same MBON – is the only one receiving significant feedback input from previous experience, which leads to activity during the ‘omission of punishment’. We also showed that the γ5-fb subtype, but not the γ5-uc subtype, is necessary to form the associated positive memory. Finally, we found that both subtypes innervate a defined area of the γ5 compartment MBONs dendrite, suggesting that these PAM-γ5 DANs modulate specific synapses between KCs and the cognate MBON. This specificity of wiring could potentially enable memory modification in parallel, via different information streams on the same output neuron. In contrast, the aversive PPL1-γ1pedc receives a large proportion of its input into different areas of its quadripartite dendrite, suggesting that these inputs are computed differently, enabling the neuron to function in different modes.
Other DAN input information streams originate from diverse sensory processing centers like the innate valence computing lateral horn or from different MBONs as direct feedback from previous experience. Others still can relay indirect feedback from MBONs, or integrate a wide range of information from other brain areas representing a network for integration of information on internal state, environmental cues and previous experience to write and update memories.
Thus, these connectomic analyses reveal additional complexity of dopaminergic circuitry both between and within MB compartments. By determining connectivity of DANs in the fly, we can increase our understanding of how neural circuitry can integrate different sensory information with cues relating to previous experience and internal state in parallel, resulting in the pivotal fine tuning of memories and a broad behavioural repertoire in response.
References:
Aso, Y., Hattori, D., Yu, Y., Johnston, R.M., Iyer, N.A., Ngo, T.-T., Dionne, H., Abbott, L., Axel, R., Tanimoto, H. and Rubin, G.M., 2014. The neuronal architecture of the mushroom body provides a logic for associative learning. eLife, [online] 3.
Griffith, L.C., 2014. A big picture of a small brain. eLife, [online] 3.
Felsenberg, J., Jacob, P.F., Walker, T., Barnstedt, O., Edmondson-Stait, A.J., Pleijzier, M.W., Otto, N., Schlegel, P., Sharifi, N., Perisse, E., Smith, C.S., Lauritzen, J.S., Costa, M., Jefferis, G.S.X.E., Bock, D.D., Waddell, S. 2018. Integration of Parallel Opposing Memories Underlies Memory Extinction. Cell, 175, 1-14.
Marin, E.C., Büld, L., Theiss, M., Sarkissian, T., Roberts, R.J.V., Turnbull, R., Tamimi, I.F.M., Pleijzier, M.W., Laursen, W.J., Drummond, N., Schlegel, P., Bates, A.S., Li, F., Landgraf, M., Costa, M., Bock, D.D., Garrity, P.A. and Jefferis, G.S.X.E., 2020. Connectomics Analysis Reveals First-, Second-, and Third-Order Thermosensory and Hygrosensory Neurons in the Adult Drosophila Brain. Current Biology, 30(16), pp.3167-3182.e4.
Otto, N., Pleijzier, M.W., Morgan, I.C., Edmondson-Stait, A.J., Heinz, K.J., Stark, I., Dempsey, G., Ito, M., Kapoor, I., Hsu, J., Schlegel, P.M., Bates, A.S., Feng, L., Costa, M., Ito, K., Bock, D.D., Rubin, G.M., Jefferis, G.S.X.E. and Waddell, S., 2020. Input Connectivity Reveals Additional Heterogeneity of Dopaminergic Reinforcement in Drosophila. Current Biology, 30(16), pp.3200-3211.e8.
Pavlov, I.P., 2010. Conditioned reflexes: An investigation of the physiological activity of the cerebral cortex. Annals of neurosciences, [online] 17(3).
Saalfeld, S., Cardona, A., Hartenstein, V. and Tomancak, P., 2009. CATMAID: collaborative annotation toolkit for massive amounts of image data. Bioinformatics, 25(15), pp.1984–1986.
Tanaka, N.K., Tanimoto, H. and Ito, K., 2008. Neuronal assemblies of the Drosophila mushroom body. The Journal of Comparative Neurology, 508(5), pp.711–755.
Zheng, Z., Lauritzen, J.S., Perlman, E., Robinson, C.G., Nichols, M., Milkie, D., Torrens, O., Price, J., Fisher, C.B., Sharifi, N., Calle-Schuler, S.A., Kmecova, L., Ali, I.J., Karsh, B., Trautman, E.T., Bogovic, J.A., Hanslovsky, P., Jefferis, G.S.X.E., Kazhdan, M., Khairy, K., Saalfeld, S., Fetter, R.D. and Bock, D.D., 2018. A Complete Electron Microscopy Volume of the Brain of Adult Drosophila melanogaster. Cell, 174(3), pp.730-743.e22.