A spatial code in the dorsal lateral geniculate nucleus
Vincent Hok, Pierre-Yves Jacob, Pierrick Bordiga, Bruno Truchet, Bruno Poucet, Etienne Save
Version posted on Biorxiv: 11/19/18
In the current work, Hok et al. report the existence of spatial receptive fields in the dLGN of rats that are similar in nature to HPC place fields. The authors demonstrate that dLGN place fields, which make up approximately 30% of extracellularly recorded dLGN neurons, are less stable and more variable (on a pass by pass basis) than their HPC counterparts. Further, the authors show theta oscillations in the dLGN local field potential and the presence of individual dLGN neurons that have theta rhythmic spiking. The authors report that, unlike CA3 neurons, most dLGN neurons do not exhibit theta phase precession suggesting that the dLGN spatial code does not carry sequential information about the animal’s current, past, or future trajectories. Finally, Hok and colleagues show that a sub-population of dLGN neurons can be modulated by visual stimulation but that this population does not overlap with neurons that exhibit place fields. Interestingly, dLGN place fields are insensitive to the presence or absence of visual information (as determined by similarity in responses in recordings from light vs. dark conditions) but remap when the color of the environment is altered (from all white to all black).
By and large the results are robust and the reported findings will be of interest to researchers considering interactions between different forms of neural spatial representations and visual information. Further, the observation of spatial receptive fields in the dLGN raises important questions about fundamentals of visual processing. We enjoyed the manuscript and have the following suggestions for the authors.
• There is little discussion about dLGN visual sensitivity in the rat despite there being extensive literature on the subject. There is known retinotopy in dLGN (see for example, Montero et al. 1968). How, if at all, does this relate to the spatial response properties of dLGN neurons? For example, is there any correspondence between neurons that would be sensitive to the lower half of the visual field and remapping in the light vs. dark arena conditions? A similar question could be raised regarding the object sensitive sub-population. The authors should attempt to address the relationship between their observations and visual sensitivity of dLGN in more detail.
• The object sensitive sub-population is intriguing, but the details concerning this finding are ambiguous. It is mentioned briefly in the main manuscript but the percentage of neurons exhibiting this tendency is unclear, even when examining the extended data concerning this observation. As a result, the reader is left a bit distracted by this information, especially when considering that object sensitive neurons are not place cells. The authors should consider presenting this component of the manuscript in greater detail.
• The visual stimulation experiment is posed as a control for determining that dLGN place cells were indeed recorded in a visually sensitive area (dLGN), but this experiment was not conducted for CA3 neurons. How can we know for certain that the visual stimulation protocol wouldn’t have elicited modulation in CA3 neurons as well, thus rendering this component of the study inconclusive?
• Tissue can often become displaced for extended periods of time when moving tetrodes along the D/V axis. How confident are the authors in the correspondence between their tetrode turning records and final tetrode placement? The waveform related analyses are pretty convincing and seems to cluster nicely into two groups, but waveform shape is not a definitive metric for determining the location of a recorded neuron. Is it possible to record from dLGN without moving electrodes through the hippocampus first? It would be helpful if the authors could provide more histology so readers can examine for themselves how displaced CA3 tissue was.
• It seems that LFPs were primarily referenced to another tetrode. Instead, LFPs should be referenced against an electrode outside of either the hippocampus or dLGN such as a skull screw above the cerebellum. Because of the current referencing scheme and many other factors, theta oscillations recorded in the dLGN could be explained by volume conduction yet this is not considered. These issues may be of great importance to the theta phase precession analysis. Did the authors attempt to look at phase precession for dLGN place cells against both HPC LFPs as well as dLGN LFPs? If not, this would be an important analysis to consider.
• No power spectral density plots are shown for theta oscillations in CA3 and dLGN. It would be interesting to see if the dLGN theta peak in the power spectrum is broader compared to CA3, which might explain the higher theta frequency and would also affect phase-related analyses. Additionally, a consideration of speed related modulation of theta dynamics should likely be included.
• Table 1 should include a #/% of cells which were “object specific”
• Re Overdispersion analysis (line 460): A linearly scaled Poisson distribution will itself be poisson distributed, not normal, with the approximation being more valid at higher firing rates. The Z value described here will therefore be biased to higher values for lower firing rate neurons. Because of this, it is unclear if the results in lines 90-92 are artificially strong or weak. If thalamic neurons have a higher firing rate than CA3 neurons, they may have an even higher difference in dispersion than reported. A consideration of percentiles rather than standard deviations would likely be revealing.
• Quantification of the object rotation sub-experiment (lines 144-147) should be included in the main manuscript text.
• More details on the properties of theta modulation in dLGN should be presented. What percentage of dLGN neurons had peaks in the theta frequency range in the FFT of their spike train autocorrelations (i.e. how many were significantly theta rhythmic)? What percentage of dLGN neurons were theta phase locked and was there a bias to a particular theta phase? Was theta oscillatory activity for single dLGN neurons modulated by running speed?
• Was a velocity filter used when creating spatial firing rate maps? In a related point, did the authors examine potential differences in mobile vs. immobile activation for dLGN vs. HPC?
• Perhaps we missed it, but what size was the smoothing kernel for 2D ratemaps (in cms)?
• Did the authors examine the spatial distribution of dLGN place fields and, if so, were they uniformly distributed? It would be interesting to know if the fields clustered near objects for example.
• The explanation of the “field index” metric is difficult to understand in the methods and should be clarified if possible.
Figure 2: The color-coded pass index trajectory plots are not very informative because the dots are all layered on top of each other. Authors could do one or multiple of the following to enhance interpretation: make the dots smaller, make the dots transparent, make the color map sequential rather than qualitative.
It would also be helpful to provide an “n” of the number of place cells in each region in Fig 2g.
Figure 3: The close up of a single pass in the upper right corner of sub-plots 3b and 3c are confusing. The legend could be revised to clarify what we are looking at. In general, we are a bit unclear as to why the close up of a single pass is necessary. Are we supposed to see rhythmic activity in the spikes? The close up of a single pass is not useful for this because the rat may not be moving at a constant velocity. It may be the case that the spike train above is sufficient.
- BU NeuroPreprint JC