Axon arrival times and physical occupancy establish visual projection neuron integration on developing dendrites in the Drosophila optic glomeruli

Reviewer #1 (Public Review):

Summary:

This is a strong paper that sets the foundation for future work that will explore the innervation of the giant fiber, allowing experiments that will link molecular/developmental mechanisms to circuit function at a level of resolution that has not previously been possible. In the course of this work the investigators discover an axon-axon competition that reflects the order of innervation of the target. In addition, a host of reagents are developed that will be of wide use in dissecting this system.

Strengths:

(1) The developmental, functional and connectomic characterization of the wiring pattern to be dissected is impressively thorough and quantitative.
(2) The reagents that the authors establish will be foundational to subsequent effort.
(3) The discovery that axon-axon competition is involved in patterning this system, and might combined with innervation order to give a deterministic outcome is an interesting one (and might be useful to address variation in cell number (see below)!

Weaknesses:

(1) In my opinion, the authors miss an opportunity to leverage their connectomics characterization somewhat more. That is, from characterization of the connectomes of two flies, the authors describe substantial variation in the number of pre-synaptic cells providing inputs (for example, in FAFB, there are 55 LC4 cells, while in the hemibrain, there are 71 - almost 30 percent more), yet the number of total synapses provided by each class of cell types is remarkably stereotyped 2442 synapses versus 2290 synapses). And the ratio of LC4 to LPLC2 synapses is even more stereotyped... As this kind of stereotypy would be consistent with the authors competition model, but inconsistent with a model in which each cell makes a similar number of synapses (which would be the model from the periphery of the visual system), the authors should comment a bit more on what they see. Perhaps the wiring model the authors advocate for compensates for what appears to be quite significant variation in the numbers of LC neurons?

(2) I appreciate how the authors pivoted to interpreting their results using Kir2.1 to reflect the effects of cell ablation. However, I worry that since the mechanism behind Kir2.1 mediated ablation is unknown, there could be other effects associated with this perturbation, creating indirect effects that alter LPLC2 cells somehow. I would therefore ask that the authors repeat these experiments with a more standard cell ablation strategy (such as a light gated caspase, or ricin). More crucially, the author's model that arrival order is functionally important would be greatly strengthened if they did the reciprocal ablation of LPLC2 and asked what happens to LC4. One could easily imagine a model in which these two cell types mutually compete for real estate, after an initial bias is set by arrival order.

Reviewer #2 (Public Review):

Summary:

The authors investigate axonal and synapse development in two distinct visual feature-encoding neurons (VPN), LC4 and LPLC2. They first show that they occupy distinct regions on the GF dendrites, and likely arrive sequentially. Analysis of the VPNs' morphology throughout development, and synaptic gene and protein expression data reveals the temporal order of maturation. Functional analysis then shows that LPLC2 occupancy of the GF dendrites is constrained by LC4 presence.

Strengths:

The authors investigate an interesting and very timely topic, which will help to understand how neurons coordinate their development. The manuscript is very well written, and data are of high quality, that generally support the conclusions drawn (but see some comments for Fig. 2 below). A thorough descriptive analysis of the LC4/LPLC2 to GF connectivity is followed by some functional assessment showing that one neuron's occupancy of the GF dendrite depend on another.
The manuscripts uses versatile methods to look at membrane contact, gene and protein expression (using scRNAseq data and state-of-the art genetic tools) and functional neuronal properties. I find it especially interesting and elegant how the authors combine their findings to highlight the temporal trajectory of development in this system.

Weaknesses:

After reading the summary, I was expecting a more comprehensive analysis of many VPNs, and their developmental relationships. For a better reflection of the data, the summary could state that the authors investigate *two* visual projection neurons (VPNs) and that ablation *of one cell type of VPNs* results in the expansion of the remaining VPN territory.

The manuscript is falling a bit short of putting the results into the context of what is known about synaptic partner choice/competition between different neurons during neuronal or even visual system development. Lots of work has been done in the peripheral the visual system, from the Hiesinger lab and others. Both the introduction and the discussion section should elaborate on this.

The one thing that the manuscript does not unambiguously show is when the connections between LC4 and LPLC2 become functional.

Figure 2:
Figure 2A-C: I found the text related to that figure hard to follow, especially when talking about filopodia. Overall, life imaging would probably clarify at which time point there really are dynamic filopodia. For this study, high magnification images of what the authors define as filopodia would certainly help.
L137ff: This section talks about filopodia between 24-48 hAPF, but only 36h APF is shown in A, where one could see filopodia. The other time points are shown in B and C, but number of filopodia is not quantified.
L143: "filopodia were still present, but visibly shorter": This is hard to see, and again, not quantified.
L144f: "from 72h APF to eclosion, the volume of GF dendrites significantly decreased": this is not actually quantified, comparisons are only done to 24, 36 and 48 h APF.
Furthermore, 72h APF is not shown here, but in Figure 2D, so either show here, or call this figure panel already?

Figure 2D/E: to strengthen the point that LC4 and LPLC2 arrive sequentially, it would help to show all time points analyzed in Figure D/E.

L208: "significant increase ... from 60h APF to 72h APF": according to the figure caption, this comparison is marked by "+" but there is no + in the figure itself.

Figure 3:
A key point of the manuscript is the sequential arrival of different VPN classes. So then why is the scRNAseq analysis in Figure 3 shown pooled across VPNs? Certainly, the reader at this point is interested in temporal differences in gene expression. The class-specific data are somewhat hidden in Supp. Fig. 9, and actually do not show temporal differences. This finding should be presented in the main data.

L438: "silencing LC4 by expressing Kir2.1... reduced the GF response": Is this claim backed by some quantification?

Figure 4K: Do the control data have error bars, which are just too small to see? And what is tested against what? Is blue vs. black quantified as well? What do red, blue, and black asterisks indicate? Please clarify in figure caption.

Optogenetics is mentioned in methods (in "fly rearing", in the genotypes, and there is an extra "Optogenetics" section in methods), but no such data are shown in the manuscripts. (If the authors have those data, it would be great to know when the VPN>GF connections become functional!)

Methods:

Antibody concentrations are not given anywhere and will be useful information for the reader

Could the authors please give more details on the re-analysis of the scRNAseq dataset? How did you identify cell type clusters in there, for example?

L785 and L794: I am curious. Why is it informative to mention what was *not* done?

Custom-written analysis code is mentioned in a few places. Is this code publicly available?

Reviewer #3 (Public Review):

Summary:

In this work, MacFarland et.al. show that difference in the time of contact between axons of LC4 and LPLC2 visual projection neurons (VPNs) in the optic glomeruli and dendrites of large descending neuron, the giant fiber (GF) shapes the differential connectivity between these neurons.

Strengths:

The authors analyzed the development of a well-known circuit between GF dendrites and LC4 andLPLC2 axons using different approaches. Additionally, they developed an ex-vivo patch clamping technique to show, together with correlative RNA-sequencing data, that contact site restriction is not dependent on neuronal activity. Based on this study, the connectivity pattern between GF and the adjacent different sets of VPNs now provides a very interesting model to investigate developmental programs that lead to synaptic specificity.

Weaknesses:

Following are the concerns that significantly impact the veracity of conclusions drawn based on the data provided.

(1) All the data related to the activity of VPNs and GF and how this activity is related to the connectivity and/or maintaining and stabilizing this connectivity is correlative. The expression profiles of synaptic molecules (only at RNA level) over time or the appearance of pre and post synaptic proteins or the spontaneous spike patterns in GF do not show the role of activity in synapse specificity program. Synaptic molecules have been previously shown to be present at presynaptic sites without being involved in activity (Chen et al., 2014, Jin et al., 2018). To show whether activity is indeed not required for connectivity for either of the cell types (LC4 and LPLC2), they should silence each and also both cell types as early as possible (with the LC4 driver that does not ablate them) and then quantify the contacts with GF. In the same vein, the authors should knock down components of the synaptic machinery as early as possible to show directly the effect on 1) contact formation and 2) contact stabilization. For example, authors state in the lines 267-269 "VPN cholinergic machinery arrives too late to contribute to the initial targeting and localization of VPN axons on GF dendrites. Cholinergic activity instead is likely to participate in VPN and GF synapse refinement and stabilization." This statement would only be valid if the authors knock down the cholinergic machinery and find the contact numbers unchanged in the early stages but significantly different in later stages in comparison to the controls. Furthermore, authors only show increase in the VAChT and ChAT in the presynaptic cells but do not show if the cholinergic receptor AChRs are even expressed in GF cells or at what point they are expressed. Without these receptor expression, cholinergic system might not even be involved in the process. Also, there might be other neurotransmitter systems involved. Authors should at least check if other neurotransmitter systems are expressed in these cells, both pre-and post-synaptic.
Line 371-374: "In the later stages of development, the frequency of synaptic events increase as gap junction proteins are downregulated and cholinergic presynaptic machinery is upregulated to enhance and stabilize synapses with intended synaptic partners while refining unintended contacts". The authors did not show the activity they observed in GF is due to the contacts they make with LC4s and LPLC2s. The functionality of these contacts can be shown by silencing the LC4s and LPLC2s and then doing the patch clamping in GF to see a decrease in the activity. Further, the authors did not show that the reduction in contacts are only by refining "unintended" contacts. There is no evidence that can support this statement.

(2) In the LC4 ablation experiments, authors claim that LC4_4 split Gal4 line is expressed around 18APF, prior to GF LC4 initial contact (Line 387). However, authors do not show the time point of first contact between GF dendrites and LC4 cells. In Fig. 2 the first time point shown is at P36, where there is already significant overlap between GF dendrites and LC4 axons. Authors should show the very first time point where they see any, even if minimal, overlap and/or contact between GFs and LC4s. Once the LC4s are ablated, is the increase in the colocalization between GF and LPLC2 due to LPLC2s increasing their contact numbers or due to them not decreasing the maximum contact numbers that the authors observed at P72 (Fig 2G)? In other words, once the LC4s are ablated, what would the new graph for temporal contact numbers for LPLC2 look like and how it would compare to Fig2G?

(3) If the developmental stages for different lines match, that would be more helpful for comparison. Also, as the authors analyzed expression every 12 hours from 0APF, the panel should also contain earlier time points (e.g. P0, P12) for all lines. This is critical to understand at what point the axons of LC4, LPLC2 and LPLC1 reach their position. From the scale bar in Supp Fig.4, it seems LC4 axons have already reached final position at P24 and there is no extension between P24 and P60. Do the authors know at what point LC4 axons start extending and reach the final position? If the LC4 and LPLC2 arbors are already separated medio-laterally even before GF dendrites extend towards them, it would explain why GF dendrites extending from medial region of the brain would encounter LC4 axons first and LPLC2 axons later, just based on their localization in space.
Further to this point, the authors show in the section two of the paper that it is the GF dendrites that extend, elaborate and refine during the phase the authors analyzed and the authors do not show any morphological change in the axons of the VPNs. Therefore, the title of the paper is 'axon arrival times and physical occupancy establish visual projection neuron integration on developing dendrites in the Drosophila optic glomeruli' is slightly misguided.

(4) In the absence of LC4s, does the LPLC1 and GF colocalization increase or do they still stay disconnected?

(5) Does the absence of LC4s have any effect on GF arbor complexity? Does the graph in Fig 2B and C change? Can the increase in colocalization between LPLC2 and GF be at least partially due to the expansion of GF dendritic volume?

(6) Why is there a segregation in the medial-lateral axis but not in the dorso-ventral axis? Wouldn't the same segregation mechanism be in play in both axes? Also, the authors should clarify if this reduction in dorsal-ventral distribution is because dorso-ventral expansion of GF dendrites beyond the LC4 and LPLC2 axons? Theoretically that would seem to make the LC4s move more ventrally and LPLC2 move more dorsally in comparison to the total arbor.

(7) Why the LPLC2 medial connections are regarded as "mistargeting" in the heading of Supplemental Figure 1? Both in EM data and in some of the confocal datasets, these connections are observed. What is the criteria to label a connection "mistargeting" if it is observed, albeit occasionally, both in EM and confocal datasets?

(8) In Line 126-127, authors state that "we sought to determine how the precise VPN localization along GF dendrites arises across development". However, based in EM and microscopic data, there is considerable variability in the contact numbers and distribution. With such variability present, how can the localization be termed "precise"? Authors should clarify.

Author Response:

We thank the editors for their assessment of our manuscript. We appreciate the reviewers’ thoughtful comments and plan to incorporate their feedback into a revised manuscript. We agree that incorporating an additional, more common ablation tool would be highly complementary to our Kir2.1 ablation studies. We also agree that images across timepoints should be expanded for contact analyses, connectomics data can be better leveraged, additional quantifications can be performed as suggested by the reviewers to better support claims, and that the introduction and discussion can be revised to better position our work in the context of previous studies. We also strongly agree that providing data on receptor RNA and protein expression in the GF across timepoints would be extremely informative, however we have found acquiring these data, at the necessary resolution, would require new approaches and tools that may be outside the scope of the project.