Summary: This work builds on work the previous year from Kunitomo et al. in Yuichi Iino's lab, where they found that C. elegans not only chemotaxes towards salt (a fact published at least as early as 1973), but specifically learns to associate a particular salt concentration with food or the absence thereof.
In this paper, the fundamental phenomenon they seek to explain is one in which C. elegans, cultured at X mM salt, will chemotax back towards the X mM when moved to a different salt concentration within a plate containing a spatial gradient of salt concentrations. For example, a worm cultured at 50mM salt and then moved to 25mM will attempt to climb back up the gradient ('positive chemotaxis'), but if placed at 75mM, will attempt to move down the gradient ('negative chemotaxis')
They focus on several specific neurons they expect are important:
- the ASER/ASEL pair, which are known to sense salt and behave as OFF- and ON- sensing gradient detectors
- and several interneurons known to control motion -
- AVA (the backwards 'command' neuron),
They begin with ablations of the sensory neuron ASER and ASEL, and quickly decide that ASER ablation reduces the chemotactic ability but ASEL doesn't. I find the data presented here somewhat unconvincing - it seems that killing ASEL may also have an effect, but ASEL is no longer investigated for the rest of the paper.
They next argue that chemotaxis towards the salt concentration setpoint is via klinotaxis (biased random walk) as opposed to klinokinesis (modifying the direction during runs, also known as weathervaning). They show convincingly that run duration is longer when going towards the setpoint, and also that turning angles are biased towards the setpoint, but changes in direction during runs are not biased towards the setpoint (again, not klinokinesis/weathervaning).
They next study the calcium dynamics of the ASER neuron in worms grown at 50mM salt, and then immobilized in a microfluidic device and exposed to precisely-controlled salt gradients. In particular, they consider both increasing and decreasing salt gradients, centered at 25mM and 75 mM salt (4 conditions in total). In all conditions, there is massive animal-to-animal variability, yet they claim "These results indicate that both set point memory and gradient perception can be inferred from ASER activity patterns." This claim I don't understand. It seems to me in order to prove this, you would have to show that if I give you a single ASER activity trace, that you could tell me both the setpoint and the current gradient. However, they show no such ability to decode these traces, and by eye, I'm pretty sure you couldn't. I'm thus confused about this point, since it seems pretty central to the paper (and the title).
If you're going to claim A encodes information about B, shouldn't you have to show that given A, you can decode it to learn about B?
They then repeat the calcium dynamics study with a tracking microscope, so that the worms can move freely, instead of being immobilized in a microfluidic device. They show that at reversals ASER dynamics correspond to whether the worm is moving up or down the gradient (as has been shown previously!).
Again however, comes another part I don't understand. They claim that in these traces, "Sharp, discrete calcium transients appear when navigating above the setpoint. Signatures of both perception and memory can be inferred from ASER activity patterns in freely moving worms." But I don't see these sharp, discrete calcium transients! Strangely, the data below and above the setpoints are plotted on different time axes (30s vs 100s), so the data above the setpoint appear more jagged. But this can't be the 'sharp discrete calcium transients' they refer to, because its just an artifact of using different time axes! Again, I don't see any evidence that one could decode whether one is above or below the setpoint from these dynamics.
They then move onto studying AIB and AIY. According to figure 1, AIB is the output node of a feedforward loop, where the ASE neurons are the inputs and drive AIY and and AIB directly, and AIY drives AIB (through AIZ). However, most diagrams like this omit a lot of edges so I don't know correct it really is to call this a feedforward loop.
They show that AIB calcium activity almost always (22/28) rises within 10s following reversals, regardless of gradient orientation. They then show that on average it drops gradually, by 50% within 15 seconds after the reversal ends.
They then claim "AIB activity patterns represent motor output." I might amend this to "AIB activity patterns contain information regarding motor output," because given the AIB activity for a single trial, I think I would be hard-pressed to tell you about motor output! Additionally, no information is given regarding AIB's activity during runs or at any other time than reversals. Simply seeing A and B at the same time does not mean A is encoding B, or even that A and B are correlated!
Now onto AIY. Strangely, they don't show the individual traces for AIY like they did for AIB, so all we've got is the average calcium dynamics centered around reversals. AIY's average calcium signal drops by half within 10 seconds of reversal. During runs, it is higher when the worm is moving at >80um/s. However, for a single worm, the correlation between speed and AIY calcium signal is weak (no R2 is given, but presumably below 0.5, even for a flexible model).
Strangely, AIY and AIB measurements were done differently - one uses ratiometric imaging (GCaMP : mCherry) and one uses just GCaMP fluorescence. No explanation seems to be given. Presumably this just because the experiments were done at different times and with different strains constructed for different purposes, but are there differences in SNR? One thing we don't see at all in these measurements is controls of GFP-only strains. How much technical noise is there in these calcium traces?
At this point, the authors revisit the ablation experiments and ablate the interneurons AIY, AIZ, AIB, and some combinations thereof. They find that ablating AIB alone, or ablating specific pairings (which all include AIB) affect chemotactic ability. However, they never completely eliminate the chemotactic ability, they just reduce it. Similar results are obtained by genetically inactivating specific cells by cell-specific expression of a constitutively active potassium channel (twk-18(gf)) which hyperpolarizes the cell. Additionally, they note that using genetics to inactivate neurons as opposed to laser ablation gives them access to far more animals, enabling much higher statistical power.
Questions I still have:
- Why didn't they show the ability to decode any of the neural traces they recorded?
- If ASER and ASEL are on the left and right side, and worms lie either on their left or right, might it not make a difference which side is oriented towards the agar? Perhaps this is addressed somewhere and I missed it, but it seems like a fundamental question here. Perhaps this is the source of a lot of variability in their ablation experiments?
- Genetic silencing using a constitutively active hyperpolarizing channel could lead to compensatory mechanisms, since the neuron is presumably silent throughout much of development. Why not use a halorhodopsin and show the effects of immediate inactivation as opposed to the cumulative effects of inactivation over the entire life of the animal?