Molecular identification of wide-field amacrine cells in mouse retina that encode stimulus orientation

Reviewer #1 (Public Review):

Summary:

This manuscript from Park et al examines the molecular, anatomical and functional properties of a subset of wide-field amacrine cell (WAC) types in mouse retina. More than 60 mouse amacrine cell types have been identified by single-cell transcriptomic studies (Yan et al., 2020, PMID: 32457074), but the functions of most of these are unknown and WACs are particularly understudied. The authors use intersectional genetics to target a subset of mouse WACs that co-express Bhlhe22 and the kappa opioid receptor (referred to as B/K WACs). They used electrophysiological and anatomical approaches to determine how WACs contribute to neural computations in the retina.

Strengths:

Overall, the paper presents a technically impressive set of experiments that build strong evidence for the presence of at least 3 discrete WAC types in the B/K transgenic line. These cells vary with respect to their morphology, dendritic stratification, response polarity (On vs Off) and resting membrane potentials. All types have long, monostratified dendrites and appear to lack axons. Electrophysiological recordings establish that these WACs are non-spiking, while calcium imaging revealed orientation selectivity in dendritic segments with tuning that correlates strongly with dendritic orientation. The authors go on to use optogenetics to show that WACs provide strong GABA-A receptor mediated inhibitory input to OFF and ON alpha sustained RGCs. This connectivity is further substantiated, at least for the OFF sustained alpha RGCs, by connectomic analyses from serial block face EM volumes. The use of the APEX2 reporter system to label the B/K cells in one of the EM volumes is particularly nice, making identification of the B/K WACs unambiguous. The conclusions are largely well supported by the experimental data. The study provides novel insights into the structure and function of specific WACs that will provide a foundation for further studies investigating the role of these amacrine cells in retinal circuits.

Weaknesses:

A limitation of the study is that the B/K WAC types described here could not be aligned to specific transcriptomic identities. The authors show more than 15 GABA expressing ACs express Bhlhe22 in the transcriptomic dataset, but it is unclear which of these also express the kappa opioid receptor (Opkr1).

The optogenetic evidence suggests that WACs provide GABA-A receptor mediated inhibitory input to both the sustained OFF and ON alpha RGCs. However, at least in the examples shown, there appears to be a dramatic difference in the timecourse of the rising phase of the inhibitory inputs to these two cell types, with the inputs to the ON sustained alpha RGCs appearing slower than those in the OFF sustained and OFF transient alpha RGCs. This apparent temporal difference was accompanied by a relatively lower sensitivity to light stimulation for the ON sustained cells. The slow timecourse seems unexpected for a direct GABA-A mediated synaptic connections between the WACs and ON alpha sustained cells. Moreover, since the connectomic analyses do not examine inputs to ON RGC types, the direct synaptic connection between B/K WACs and On alpha RGC is less well substantiated.

Reviewer #2 (Public Review):

Summary:

An important frontier in research on the mammalian retina is to understand the role of inhibitory amacrine cells in visual processing. These cell types have been found to play roles in tuning the output of the retina to specific visual features like motion and orientation. These cell types are understudied for two main reasons. First, there are many types of them-over 60 types in the mouse--, and second, they are quite unconventional as far as neurons go, as they have dendrites but often lack axons. The manuscript "Molecular identification of wide-field amacrine cells in mouse retina that encode stimulus orientation" by Park et al. provides a characterization of two (or possibly more) cell types within the amacrine cell class. Specifically, they characterize types of widefield amacrine cells (WACs), which they have gained genetic access to using an intersectional transgenic mouse strategy (Bhlhe22 x KOR). The authors used a broad range of experiments to characterize these WACs' anatomical properties, their stimulus tuning, and their wiring within the retina to their postsynaptic partners. These experiments include anatomy, electrophysiology, calcium imaging, and electron microscopy.

Strengths:

Overall, the manuscript presents strong evidence that the Bhlhe22 x KOR WACs represent multiple WAC types in the retina and that these cell types are orientation tuned. The most exciting finding is that their orientation tuning is correlated with the physical orientations of the dendrites, which suggests that this anatomical feature supports the tuning in these cells.

Weaknesses:

(1) The one common thought about widefield amacrine cells (WACs) is that these are spiking cells, which allows them to transmit signals along their long dendrites. The authors state that "none of the recorded cells fired conventional action potentials (spikes)." (p.7) This is a surprising finding, which leads to an interesting question: how do these cells integrate information from their presynaptic partners to generate the orientation tuning observed without the ability to conduct over long distances? However, the authors have not fully ruled out that the cells do spike.
For instance, one possibility is that spiking requires a specific stimulus and the authors did not play that stimulus during their recordings. Most somatic recordings did not result in very large depolarizations, and the cell could still be below threshold. Depolarizing the cell to attempt to evoke spikes directly could be used to explore this possibility. A second possibility is that the dendrites spike, but these spikes are attenuated at the soma. Direct injections of current into the cells to evoke such spikes could be used to observe whether dendritic spiking occurs. A third possibility is that some important machinery for spiking is being washed out by the whole cell recordings. Cell attached recordings could be used to assess whether spiking occurs in an intact cell. The authors may wish to address these possibilities experimentally, but at least should qualify their statement about spiking in these cells and discuss these possibilities.

(2) It was unclear in this paper how many cell types are present in the intersectional cross. I think the paper would be stronger if they clarified that. For instance, in Fig. 1B: the authors show Bhlhe22 expression in amacrine cells from a previous study. They should also show the expression of the other gene they used in their intersectional strategy, the Kappa Opioid receptor (Oprk1), which is available in the same dataset. Another piece of analysis that could help would be clearer quantification of the anatomical features of the cells. For instance, the cells shown in Fig. 2 A2 vs. B2 have clear differences in number of dendrites and the relative angles of the dendrites. The On cells appear to have more dendrites evenly spread around the soma, while the Off cells appear to have more clumping along a line. Is this the case for all the cells recorded, or just these examples? The authors should present some population-level quantification.

(3) In Fig. 4E, the preferred orientation of calcium responses and physical orientation of the dendrites appears to clump around specific orientations. The Methods don't mention if the retinas were aligned to the body axis during the dissection. Is this clumping real, or is this an artifact of the analysis? If there are specific preferred orientations to these WAC cell types, that would be important to discuss in the paper - for instance how this relates to the preferred direction in the direction selectivity system or how it might relate to the function of these cells for behavior.

Reviewer #3 (Public Review):

Summary:

Amacrine cells are a heterogeneous collection of retinal interneurons. Most are inhibitory, and like inhibitory neurons in other neural circuits, strongly shape retinal function. With a few exceptions, the role of amacrine cells in retinal signaling is poorly understood. This paper introduces an approach to study a set of wide-field amacrine cells that extend processes over large regions of the retina.

Strengths:

A substantial strength of the paper is the combination of genetic manipulations, electrophysiology, optogenetics and electron microscopy used to study these cells. As a result of that broad set of techniques, the results cover many properties of how the cells work and provide a nice overview. The paper is also (with a few exceptions below) clearly presented and the experiments look to be carefully executed with clean results.

Weaknesses:

My largest concern with the paper is that overall the results provided an initial view of an interesting set of issues about the function of these cells, but the interesting initial results are not pursued in more depth.

Spatial spread of signals in neurites
An immediate question about axonless WACs is the extent of spread of signals along their processes, and hence whether they act as a collection of independent or semi-independent elements. This bears directly on interpretation of the responses to oriented stimuli for example. Did you do any experiments that might provide additional information about this issue? For example, if you stimulate one of the WACs peripherally do you see a strong modulation of the somatic voltage? Or in the imaging experiments, if you mask a region of the processes so that it is not receiving a stimulus, do you see responses "leak" into that occluded region from surrounding stimulated regions?

Orientation tuning and connectivity
The most developed functional results in the paper relate to the sensitivity of the WAC processes to oriented stimuli. Interpretation of these results depends on a few factors. First is the spread of signals in the WAC processes - as noted above. Second is connectivity. The paper shows that the B/K WAC activity increases inhibitory input to Off-delta and On-alpha ganglion cells. These cells, as noted in the paper, are not orientation tuned. But the orientation tuned ganglion cells stratify in a similar location within the IPL, and hence are situated in an appropriate place to receive input from the B/K WACs. Did you focus exclusively on connections to the Off-delta and On-alpha cells (along with the Off-alpha) or did you look at any other ganglion cell types? This should at least get discussed in more detail.

In several places it is unclear whether the paper intends to be a methods paper or a basic research paper. One example is the last sentence of the abstract. If it is intended to be a basic research paper (which is my overall impression) then I suggest shifting the emphasis in some of those key locations towards results and away from methods.