Article activity feed

  1. Reviewer #2 (Public Review):

    Zhao et al. investigated how a single trial of aversive conditioning could produce a "merged" long-term memory (mLTM). This mLTM is composed of two negatively associated memories: CS+ (odor paired with electric shock) and CS- (odor unpaired), which can be experimentally achieved by the presentation of a third novel odor at the time of testing (memory retrieval). Through a series of behavioral experiments, they determined that both CS+ and CS- LTM depends on protein synthesis. This was supported by cycloheximide feeding prior to aversive conditioning or flies experiencing a cold shock anesthesia after training. Next, they found that mLTM is derived from the same memory component. The re-presentation of either the CS+ or CS- odor at some point before retrieval extinguished mLTM. The authors also show that mLTM forms regardless of odor exposure sequence, but rather does not occur when the temporal interval between CS+ and CS- during training is extended to 20 minutes. They next determined the neural circuit supporting mLTM. Blocking synaptic output from all PPL1 dopamine neurons (DAN) during training impaired expression of mLTM. Downstream of PPL1 DAN, inhibiting synaptic release from the mushroom body neurons (MBN) similarly blunted mLTM which was further mapped to the axons of α2sc MBN. The plasticity was then ascribed to the α2sc mushroom body output neurons (MBON); blocking α2sc MBON behaviorally impaired mLTM. Lastly, the authors showed that the odor-evoked responses to the CS+ and CS- odors in α2sc MBON were significantly depressed when compared to the novel odor. Overall, they propose that the PPL1 DAN: α2sc MBN: α2sc MBON circuit is responsible for generating mLTM.


    The key conclusions stem from a series of behavioural experiments that display a consistent and reproducible phenotype. The data presentation and manuscript text are simple, direct and easy to follow. The interesting observations may potentially garner interest to address how animals incorporate different types of strategies to adapt to their environments when they encounter a threat once or multiple times.


    Although the manuscript describes several intriguing observations, they are outweighed by a number of weaknesses that substantially limit the impact of the manuscript. The data supporting the main conclusions are thin, experimental approaches are not rigorous, and some writing sections are incomplete.

    1. The observation that presentation of a novel third odor leads to mLTM after only a single session of aversive conditioning is intriguing. Authors describe in their methods using three odors for their experiments (as CS+, CS- or novel), but did not alternate/rotate the different combination pairings used as the "novel" one. A panel of odors as "novel", not listed in the manuscript, should be tested which will strengthen the larger conceptual framework and impact. In addition, the authors should perform at least a subset of the experiment using air during testing rather than a 3rd odor.

    2. The authors show that the contiguity of CS+ and CS- is critical, and that a 20 min interval leads to no mLTM. What is the maximum temporal interval that supports the formation of mLTM?

    3. The authors claim that PPL1 DAN during training are key for mLTM (Figure 3A). The GAL4 line used was TH-GAL4 whose expression pattern (Figure S1A) extends beyond the PPL1 cluster. The more specific TH-D'-GAL4 is suitable and needed to rule out other dopamine clusters labeled by the much broader TH-GAL4 line. Additional split-GAL4 lines can be used to fine tune the PPL1 subpopulations that are important for their proposed circuit. Related to this issue, Figure S1 is useless to the reader for deciphering the expression pattern of the Gal4 lines used given the poor resolution. A better general option would be to simply reference papers/websites that have high resolution images of the expression patterns for those lines used widely and provide high resolution images in manuscripts for only those lines that have not been exhaustively described before.

    4. The authors mapped the importance of α2sc MBN for the retrieval of mLTM (Figure 3B). This observation could be strengthened by incorporating additional GAL4 lines that drive expression in α2sc MBN (R28H05-GAL4 or NP3061-GAL4). Inhibiting α2sc MBN via optogenetics (UAS-eNPHR3: inhibitory halorhodopsin) may further support the behavioral phenotype observed, which can also be applied to the notion above using TH-D'-GAL4.

    5. The authors claim that α2sc MBON as the last part of their circuit. This is a massive jump of a conclusion directly from the α2sc MBN side of the proposed pathway. There are six MBON (α3, α2sc, α2pα3p, α1, α2α'2a, α1>α) that are downstream from the α2sc MBN. The authors need to rule out the other neurons before directly claiming α2sc MBON only as the main player. Moreover, the R71D08-GAL4 line (supported by the expression pattern in Figure S1F, and cartoons in Figure 4) drives expression in other MBON, and again the authors should use more specific lines that are available.

    6. More experimentation and discussion regarding the differences between single trial conditioning to form mLTM and spaced conditioning to form complementary LTM, is required. The authors contrast/merge their behavioral results with those published by Jacob et al (2020). The authors should reproduce the essence of those found by Jacob et al and publish them in this paper. Replication of experimental results across labs is very important, especially for behavioral outcomes and when models are constructed using results obtained by other investigators. The authors allude to the two pairs of DAN that project to α2sc MBN for this plasticity, but did not specifically mention those DAN (lines 220-221) nor elaborate on this speculation.

    Was this evaluation helpful?
  2. Reviewer #1 (Public Review):

    Using a third odor for memory test, the authors found that single-trial conditioning produces a protein synthesis-dependent LTM leading the avoidance to both CS+ and CS- for more than 7 days. By acutely blocking neurotransmissions from target neurons, they showed that the merged LTM requires outputs from TH-positive dopaminergic neurons during training, and from αβ Kenyon cells and α2sc mushroom body output neurons during testing. Several lines of evidence support their claim that the long-term avoidances of CS+ and CS- after single-trial conditioning are based on the same memory component: 1, Five independent disruptions of the system produced the same level changes of avoidance from CS+ and CS-; 2, Re-exposure to either CS+ or CS- alone abolished both CS+ avoidance and CS- avoidance; 3, The same PPL1 DANs, αβ KCs, and α2sc MBONs involved in both avoidances; and 4, Similar responses of functional depression to CS+ and CS- occurred in the same α2sc MBONs. Based on these results, the authors suggest that animals can develop distinct memory strategies for occasional and repeated threatening experiences.

    While the entire manuscript is well-written and the data are well presented, the conclusion has several weaknesses. First, memories for CS+ and CS- are clearly distinct from each other at least during the first 24 hours after training (Figure 1B). During this period, CS+ memory shows a gradual decline similar to conventional LTM. Oppositely, CS- memory shows a gradual increase. When these flies were tested immediately after training, they seemed to approach CS-, a result similar to the previous study using a similar third-odor test showed that multiple-trial conditioning induces "two independent LTMs of opposite valence for avoiding CS+ and approaching CS-". How does "the approaching CS- memory" turn into "the avoiding CS- memory"? Or, "the avoiding CS- memory" is derived independently, similar to the anesthesia-resistant memory?

    If initial behavior responses to the CS+ and CS- memories are opposite and thus separated, where does each of them occur? Where and how are they merged after one day? The authors claimed that both CS+ and CS- memories require the same PPL1 DANs, αβ KCs, and α2sc MBONs. But, they manipulate PPL1 DANs with TH-Gal4 that expressed in many other dopaminergic neurons. Similarly, the αβ KCs include at least three major cell populations, each has hundreds of neuron with distinct functions. Thus, while both CS+ and CS- memories require the same family of neurons, it remains uncertain whether these events actually occur in the same neurons.

    In general, the identified merged LTM is original and brings a new concept to the field of learning and memory. The finding suggests that two forms of protein-synthesis-dependent LTM induced respectively after occasional or repetitive experiences are encoded in the brain by different neuronal mechanisms. It appears that merged LTM after single-trial training remembers the event with limited details and repetitive spaced trials of training then adds more details to the classical LTM.

    Was this evaluation helpful?
  3. Evaluation Summary:

    This work reveals a novel form of Drosophila long-term memory (LTM) that is of potential interest to most neuroscientists working on various animals. While classical protein-synthesis-dependent LTM forms only after repetitive spaced trials of olfactory conditioning, the authors discovered that flies also form a "blurred" or "vague" protein-synthesis-dependent LTM which distinguishes the experienced two odors from the third naive odor after single-trial training. This merged LTM lacking the event details likely occurs in most animals since long-lasting memory of occasional threatening experiences for future escape behavior is crucial for survival.

    (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. Reviewer #1 agreed to share their name with the authors.)

    Was this evaluation helpful?