Memory: When neurons split the load
Memories are composed of the molecular and cellular traces that an event leaves in the nervous system. In turn, these neuronal changes enable the brain to weave together different features of the experience – for example, its outcome – with certain properties of the environment at the time.
Even the small worm Caenorhabditis elegans, a tractable and well-studied model organism with 302 neurons, can form such associations. Through conditioning, these animals can ‘learn’ to prefer a stimulus – for instance a smell – that is associated with food being present. This requires neurons to encode information so that an experience (e.g. smelling a specific odor) is correctly linked to valence (whether the situation was positive or negative, depending on the presence or absence of food).
Previous studies have already implicated specific genes and neurons in these processes, for example (Jin et al., 2016; Tomioka et al., 2006). However, this reductionist framework cannot fully capture how different aspects of a memory, such as experience and valence, are represented amongst an entire network of neurons. In C. elegans, it is possible to identify many of the neurons in these networks, and to record their activity simultaneously at single-cell resolution. This offers a unique opportunity to directly measure the features of memory traces during perception. Now, in eLife, Alon Zaslaver and colleagues at the Hebrew University of Jerusalem – including Chrisitian Pritz as first author – report the results of an extensive series of experiments which examined how olfactory memory modulates neuronal responses in this model organism (Pritz et al., 2023).
First, the team trained groups of worms to associate a conditioning odor, butanone (diluted in a solvent) with the presence or the absence of food (appetitive vs. aversive conditioning; Colbert and Bargmann, 1995; Kauffman et al., 2010). The protocol was adapted for the animals to form either short- or long-term memories of these associations.
A choice assay experiment then confirmed that in both short- and long-term conditions, appetitive and aversive conditioning respectively increased and decreased the worms’ preference for butanone over another smell (diacetyl). Two control groups were also tested: naive animals that had not been experimented on, and worms that had been through a ‘mock’ training identical to the one received during conditioning, but in the absence of butanone (only the solvent was present).
This experimental design allowed Pritz et al. to systematically isolate and investigate the different factors that influence behavior and neuronal activity. For instance, comparing mock-treated and naive individuals helped to capture the impact of experimental parameters other than smell and valence, such as the worms experiencing starvation.
Next, Pritz et al. used calcium imaging to record the activity of the same set of 24 sensory neurons in conditioned, naive and mock-treated animals exposed to butanone or diacetyl. This revealed that, for these classes of cells, the modulation of neuronal activity in response to the odors was mainly taking place for short- rather than long-term memories. Overall, a large proportion of the sensory neurons studied showed fine changes in activity following conditioning, with a few neuron classes exhibiting a stronger response. Detailed analyses highlighted that each class could encode one or several features of the memories, such as the presence of the odor, valence or a specific aspect of the training process. Mock treatment also impacted the activity of a large proportion of sensory neurons, shedding light on how parameters such as starvation can affect neuronal responses. Overall, these results suggest that the neuronal changes associated with short-term memories are distributed across multiple types of sensory neurons, rather than one class being solely dedicated to capturing a specific element of the response.
To further explore this possibility, Pritz et al. developed machine learning algorithms that could predict the type of conditioning the worms received based on their neuronal responses. The models made better predictions if information from more neuron types (up to five) was provided. Principal components analysis, which helps to pinpoint patterns in large datasets, further supported the idea that different task parameters (conditioning odor, valence, and starvation experience) create distinct activity profiles across the sensory circuit.
Next, Pritz et al. demonstrated that modulation of the sensory neurons also impacted the interneurons that they projected onto, and which relay sensory information to the rest of the nervous system (Figure 1). Three classes of interneurons were examined: while one of them mainly responded to the mock training, the others showed conditioning-specific responses. Unlike sensory neurons, however, all three interneurons could encode both short and long-term memories. While these initial findings are intriguing, additional work on larger datasets is probably needed to confirm whether short- versus long-term memory processes are generally allocated to specific classes of cells.
Finally, Pritz et al. used statistical modelling to examine how various sensory neurons shaped the activity of the AIY interneuron, which receives most of its inputs from these cells. The results suggest that AIY modulation was provided by different combinations of sensory neurons depending on the type of training: changes in AIY activity were driven by a single class of neurons after appetitive conditioning, but by a complex circuit of several neuronal classes after aversive conditioning. The memory of different treatment experiences is therefore retained in variable degrees of distribution (the number of classes of neurons involved); whether this could be underpinned by complex changes in the strength of the connections between sensory neurons and interneurons is an exciting hypothesis for future studies.
In conclusion, the work by Pritz et al. adds to existing evidence showing that neuronal signals are distributed across the nervous system in a wide range of organisms – from reactions to stimuli and movement control in C. elegans, to memory in animals with larger brains (Lin et al., 2023; Kato et al., 2015; Owald and Waddell, 2015; Tonegawa et al., 2015). As research on C. elegans can now also examine larger neuronal networks, it should provide new insights into how the nervous system computes and yields behavior in this and other animals. Advanced machine learning algorithms could help in this effort, as they are uniquely placed to ‘decode’ the signals embedded in large neural activity datasets – for example, which odor a worm is smelling. However, the way that algorithms process that information does not necessarily match the underlying neuronal mechanisms and biological processes accurately. Addressing these problems will require further developing computational and experimental approaches alongside one another.
References
-
Olfactory learning skews mushroom body output pathways to steer behavioral choice in DrosophilaCurrent Opinion in Neurobiology 35:178–184.https://doi.org/10.1016/j.conb.2015.10.002
Article and author information
Author details
Publication history
Copyright
© 2023, Lev and Zimmer
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 787
- views
-
- 57
- downloads
-
- 0
- citations
Views, downloads and citations are aggregated across all versions of this paper published by eLife.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading
-
- Computational and Systems Biology
To help maximize the impact of scientific journal articles, authors must ensure that article figures are accessible to people with color-vision deficiencies (CVDs), which affect up to 8% of males and 0.5% of females. We evaluated images published in biology- and medicine-oriented research articles between 2012 and 2022. Most included at least one color contrast that could be problematic for people with deuteranopia (‘deuteranopes’), the most common form of CVD. However, spatial distances and within-image labels frequently mitigated potential problems. Initially, we reviewed 4964 images from eLife, comparing each against a simulated version that approximated how it might appear to deuteranopes. We identified 636 (12.8%) images that we determined would be difficult for deuteranopes to interpret. Our findings suggest that the frequency of this problem has decreased over time and that articles from cell-oriented disciplines were most often problematic. We used machine learning to automate the identification of problematic images. For a hold-out test set from eLife (n=879), a convolutional neural network classified the images with an area under the precision-recall curve of 0.75. The same network classified images from PubMed Central (n=1191) with an area under the precision-recall curve of 0.39. We created a Web application (https://bioapps.byu.edu/colorblind_image_tester); users can upload images, view simulated versions, and obtain predictions. Our findings shed new light on the frequency and nature of scientific images that may be problematic for deuteranopes and motivate additional efforts to increase accessibility.
-
- Computational and Systems Biology
The force developed by actively lengthened muscle depends on different structures across different scales of lengthening. For small perturbations, the active response of muscle is well captured by a linear-time-invariant (LTI) system: a stiff spring in parallel with a light damper. The force response of muscle to longer stretches is better represented by a compliant spring that can fix its end when activated. Experimental work has shown that the stiffness and damping (impedance) of muscle in response to small perturbations is of fundamental importance to motor learning and mechanical stability, while the huge forces developed during long active stretches are critical for simulating and predicting injury. Outside of motor learning and injury, muscle is actively lengthened as a part of nearly all terrestrial locomotion. Despite the functional importance of impedance and active lengthening, no single muscle model has all these mechanical properties. In this work, we present the viscoelastic-crossbridge active-titin (VEXAT) model that can replicate the response of muscle to length changes great and small. To evaluate the VEXAT model, we compare its response to biological muscle by simulating experiments that measure the impedance of muscle, and the forces developed during long active stretches. In addition, we have also compared the responses of the VEXAT model to a popular Hill-type muscle model. The VEXAT model more accurately captures the impedance of biological muscle and its responses to long active stretches than a Hill-type model and can still reproduce the force-velocity and force-length relations of muscle. While the comparison between the VEXAT model and biological muscle is favorable, there are some phenomena that can be improved: the low frequency phase response of the model, and a mechanism to support passive force enhancement.