Postsynaptic cell type and synaptic distance do not determine efficiency of monosynaptic rabies virus spread measured at synaptic resolution

  1. Maribel Patiño
  2. Willian N Lagos
  3. Neelakshi S Patne
  4. Paula A Miyazaki
  5. Sai Krishna Bhamidipati
  6. Forrest Collman
  7. Edward M Callaway  Is a corresponding author
  1. Systems Neurobiology Laboratories, The Salk Institute for Biological Studies, United States
  2. Neuroscience Graduate Program, University of California, San Diego, United States
  3. Medical Scientist Training Program, University of California, San Diego, United States
  4. Allen Institute for Brain Science, United States

Abstract

Retrograde monosynaptic tracing using glycoprotein-deleted rabies virus is an important component of the toolkit for investigation of neural circuit structure and connectivity. It allows for the identification of first-order presynaptic connections to cell populations of interest across both the central and peripheral nervous system, helping to decipher the complex connectivity patterns of neural networks that give rise to brain function. Despite its utility, the factors that influence the probability of transsynaptic rabies spread are not well understood. While it is well established that expression levels of rabies glycoprotein used to trans-complement G-deleted rabies can result in large changes in numbers of inputs labeled per starter cell (convergence index [CI]), it is not known how typical values of CI relate to the proportions of synaptic contacts or input neurons labeled. And it is not known whether inputs to different cell types, or synaptic contacts that are more proximal or distal to the cell body, are labeled with different probabilities. Here, we use a new rabies virus construct that allows for the simultaneous labeling of pre- and postsynaptic specializations to quantify the proportion of synaptic contacts labeled in mouse primary visual cortex. We demonstrate that with typical conditions about 40% of first-order presynaptic excitatory synapses to cortical excitatory and inhibitory neurons are labeled. We show that using matched tracing conditions there are similar proportions of labeled contacts onto L4 excitatory pyramidal, somatostatin (Sst) inhibitory, and vasoactive intestinal peptide (Vip) starter cell types. Furthermore, we find no difference in the proportions of labeled excitatory contacts onto postsynaptic sites at different subcellular locations.

Editor's evaluation

This study provides valuable new information on using monosynaptic rabies virus for trans-synaptic tracing, a widely used tool to map synaptic inputs to specific types of neurons in the mammalian brain. The authors provide convincing quantitative information about the efficiency for the rabies virus to cross synapses of visual cortical neurons in the mouse. They further demonstrate that the cortical cell type and location of synapse in the postsynaptic cell do not appear to affect the efficiency for the rabies virus to cross synapses.

https://doi.org/10.7554/eLife.89297.sa0

Introduction

Monosynaptic rabies tracing using glycoprotein (G)-deleted rabies virus (RVdG) is a powerful tool for the study of neural circuit connectivity. This method enables scientists to label, genetically manipulate, or monitor the activity of brain-wide monosynaptic inputs to cell populations of interest. Since its introduction (Wickersham et al., 2007a; Wickersham et al., 2007b) it has been widely used for the identification of presynaptic inputs to single neurons (Marshel et al., 2010; Rancz et al., 2011; Rossi et al., 2020; Wertz et al., 2015), projection-defined neurons (Cruz-Martín et al., 2014; Levine et al., 2014), adult-born neurons (Deshpande et al., 2013; Garcia et al., 2014), transplanted neurons (Doerr et al., 2017; Grealish et al., 2015), hPSC-derived organoid neurons (Andersen et al., 2020; Miura et al., 2020), and genetically defined excitatory neurons (DeNardo et al., 2015; Kim et al., 2015), inhibitory neurons (Miyamichi et al., 2013; Wall et al., 2016), and non-neuronal cell types (Clark et al., 2021; Mount et al., 2019). In addition to being used for the identification of inputs, the incorporation of Ca2+ indicators and light-activated opsins into rabies reagents (Osakada et al., 2011) allows rabies tracing experiments to probe the relationship between function and connectivity (Rossi et al., 2020; Tian et al., 2016; Wertz et al., 2015; Wester et al., 2019). The ability of genetically modified RVdG to selectively spread retrogradely between synaptically connected cells allows for the identification of presynaptic partners regardless of their distance from one another and has led to novel insights throughout the nervous system.

Despite the utility and widespread use of monosynaptic rabies tracing to study neural connectivity, there is uncertainty about the efficiency of transsynaptic spread from starter cells to input neurons and what factors influence the probability of spread. Although studies quantifying inputs to single neurons have provided some insight into the efficiency of spread (Marshel et al., 2010; Miyamichi et al., 2011; Rancz et al., 2011; Wertz et al., 2015) results vary widely across experimental conditions and no direct measurements of spread efficiency are available. Furthermore, using the convergence index (CI), defined as the number of rabies-labeled input neurons divided by the number of starter cells, recent studies have shown that the number of input neurons labeled per starter cell can be changed by as much as 10-fold by modifying the rabies glycoprotein (Kim et al., 2016), replacing the rabies virus strain (Reardon et al., 2016), or adjusting the level of glycoprotein expression (Lavin et al., 2020). Despite allowing quantitative comparisons between different reagents and conditions, these studies relied on methods that result in large animal to animal variability, require tedious counting across many animals, and can display high variability depending on the number of starter neurons (Tran-Van-Minh et al., 2022). Most importantly, CI measurements fail to quantify what proportion of all inputs are labeled. Additionally, the use of CI fails to address questions of the influence of biological factors on rabies spread efficiency, such as distance of synapses to the starter cell soma or differences of spread from different starter cell types.

In this study we examine the efficiency of RVdG retrograde spread from starter cell to input neurons at the synaptic level. We designed a new genetically modified rabies virus that labels presynaptic terminals with synaptophysin-RFP (SynPhRFP) and excitatory postsynaptic densities with postsynaptic density-95-GFP (PSD95GFP). Because more than 99% of PSD95 postsynaptic puncta colocalize with a presynaptic terminal (Micheva et al., 2010), this construct allows us to quantify the proportion of excitatory synapses on a starter cell that have their corresponding input neuron labeled with rabies virus (defined here as synaptic fraction [SF]). We find that with the particular reagents and conditions that we used, rabies retrogradely labels about 35–40% of excitatory synaptic contacts onto each of the excitatory and inhibitory cell populations we tested. Furthermore, we found that SF does not vary by the proximity or distance of synapses to the starter cell soma or by the type of neuronal dendrite. Finally, we model the factors that influence the relationship between SF and the fraction of input neurons labeled (input fraction [IF]) and show that under plausible conditions for our results in mouse visual cortex the IF/SF ratio is about 0.75, so an SF of 40% corresponds to an IF of 30% of input neurons labeled. This analysis also allows an estimate of the probability of rabies spread across a single synaptic contact (unitary synaptic efficiency, U) which we estimate to be about 0.28 (28%) when SF = 0.4. Overall, this study provides insight into some long-standing questions about the efficiency of monosynaptic rabies tracing. We further discuss that desired tracing efficiency varies depending on experimental aims, the likely relationships between SF and proportion of input neurons that are labeled, and that efficiency was not maximized with the experimental design used here.

Results

RVdG-PSD95GFP-SynPhRFP viral construct allows simultaneous fluorescent labeling of pre- and postsynaptic specializations

To investigate the transsynaptic spread of rabies virus from starter cells to input neurons at synaptic resolution, we developed a high-throughput and precise method to directly identify synapses labeled by transsynaptic spread. We created a new deletion-mutant rabies virus construct that expresses two synaptic fusion protein transgenes from the rabies G locus (Figure 1A), modeled after existing rabies constructs known to label multiple subcellular compartments of neurons (Wickersham et al., 2013). One gene encodes for a fusion protein of the excitatory postsynaptic marker, postsynaptic density 95 (PSD-95), and enhanced green fluorescent protein (eGFP). The other encodes for a fusion protein of the presynaptic marker synaptophysin (SynPh) and the bright and photostable red fluorescent protein TagRFP-T (Shaner et al., 2008). We chose the synaptophysin-TagRFP-T (SynPhRFP) fusion protein as it has previously been shown to result in a punctate red fluorescent pattern that colocalizes with varicosities on axons when expressed from the rabies genome (Wickersham et al., 2013). Injection of SAD-B19 EnvA+ RVdG-PSD95GFP-SynPhRFP in primary visual cortex (V1) of Sim1Cre mice expressing TVA and oG in Cre+ layer 5 neurons (derived from adeno-associated viruses [AAV] helper viruses, see Materials and methods) resulted in strong punctate green and red fluorescent labeling (Figure 1B), with green labeling prominent on dendrites and dendritic spines (Figure 2E and F). Super-resolution Airyscan imaging (see Materials and methods) revealed that GFP labeling colocalized with a subset of the puncta labeled with antibody staining against endogenous PSD-95 (Figure 1C). In addition to synaptic labeling there were typically nuclear aggregates of both red and green fluorescent protein in infected neurons (Figure 1B).

RVdG-PSD95GFP-SynPhRFP allows simultaneous fluorescent labeling of pre- and postsynaptic densities.

(A) Schematic of RVdG-PSD95GFP-SynPhRFP viral construct design. Two transgenes were inserted into the G locus of the rabies genome. One encodes a presynaptically targeted fluorescent fusion protein, synaptophysin TagRFP-T (SynPhRFP), and the other a postsynaptically targeted fluorescent fusion protein, PSD-95 eGFP (PSD95GFP). (B) Coronal sections of Sim1Cre mouse expressing TVA and oG in V1 infected with EnvA+ RVdG-PSD95GFP-SynPhRFP imaged at 20× with confocal microscopy. Top row shows neurons expressing both PSD95GFP and SynPhRFP fusion proteins. Bottom row shows a zoomed in max intensity projection reconstructed image of the region enclosed by the dashed square in the top row. White arrows point to PSD-95 puncta, blue arrows to synaptophysin puncta, and gray arrows to large non-specific nuclear fluorescent aggregates. Scale bars represent 100 µm (top row) or 10 µm (bottom row). (C) Airyscan super-resolution max intensity projection reconstructed images taken at 63× showing colocalization of PSD95GFP fusion protein expressed from the rabies genome with anti-PSD95 antibody staining in magenta. Scale bar = 1 µm.

Figure 1—source data 1

Plasmid sequence for RVdG-PSD95GFP-SynPhRFP construct.

Plasmid sequence for a new genetically modified rabies virus that labels presynaptic terminals with synaptophysin-RFP (SynPhRFP) and excitatory postsynaptic densities with postsynaptic density-95-GFP (PSD95GFP).

https://cdn.elifesciences.org/articles/89297/elife-89297-fig1-data1-v1.zip
Experimental design for control experiments omitting oG.

(A) Schematic illustration of experimental design and timeline for control experiments without glycoprotein. Nr5a1Cre mice were injected in V1 with a mixture of AAV-CIAO-TVA and AAV-FLEX-smFP-myc. Three weeks later EnvA+ RVdG-PSD95GFP-SynPhRFP was injected into the same site and allowed to express for 7 days. (B) Representative image of V1 injection site, obtained using widefield fluorescence microscopy at 10×. Scale bar = 200 mm. (C) Max intensity projection reconstruction of images obtained using Airyscan super-resolution imaging at 63×. Top, example image of rabies-infected neuron labeled with smFP_myc, without glycoprotein. Bottom, zoomed in image of boxed region in top image, illustrating PSD-95 puncta colocalized with cytoplasmic smFP_myc. Scale bar = 20 mm (top) and scale bar = 2 mm (bottom).

Efficiency of rabies transsynaptic spread across excitatory synapses onto excitatory starter neurons

To quantify the proportion of synaptic contacts labeled with rabies virus, we injected a mixture of three Cre-dependent helper AAVs into the primary visual cortex (V1) of Nr5a1-Cre mice to target initial rabies infection to a sparse population of L4 excitatory neurons (Harris et al., 2014; Figure 3A, B). (1) AAV8-nef-AO-66/71-TVA950 (AAV-CIAO-TVA) expresses the TVA receptor for the avian sarcoma leukosis virus envelope protein, EnvA, which is necessary for entry of pseudotyped EnvA+ RVdG into Cre+ cells. However, recombinant-independent off-target leak expression of transgenes is common in recombinase-dependent DIO and FLEX AAV constructs. Because even miniscule quantities of TVA leak expression is sufficient for pseudotyped rabies to enter off-target cells and confound results (Callaway and Luo, 2015), it was necessary to minimize leak expression of TVA. We therefore used the novel cross-over insensitive ATG-out (CIAO) AAV construct, which has been shown to nearly eliminate leak expression and provide reliable and targeted transgene expression (Fischer et al., 2019), to express TVA in our population of interest. CIAO constructs have the ATG codon placed outside of the loxp mutant pairs loxp66/71 sites to ensure the gene coding region is out of frame with the ATG start signal in the absence of Cre-mediated recombination. Experiments in which AAV-CIAO-TVA and EnvA+ RVdG were injected into V1 of Cre-negative mice resulted in 15.83±11.33 rabies-infected neurons (mean ± SEM, n=3 mice) indicating low levels of Cre-independent TVA expression. (2) AAV8-hSyn-FLEX-H2BBFP-oG expresses optimized rabies glycoprotein (oG) (Kim et al., 2016), which allows for trans-complementation in EnvA+ RVdG-infected neurons, also termed starter cells, allowing the rabies to spread retrogradely into presynaptically connected inputs. Co-expression of nuclear mTagBFP2 (BFP), a brighter and more photostable blue fluorescent protein (Subach et al., 2008), and oG allows for the unambiguous identification of starter cells (Figure 3C and E), defined as any cell that expresses both oG and rabies transgenes. (3) Because H2BBFP only labels the nucleus of starter cells, we also used AAV-CAG-FLEX-smFP_myc to generate cytoplasmic labeling, allowing dendrites of starter cells to be traced. This AAV expresses spaghetti monster fluorescent protein (smFP), which consists of dark non-fluorescent GFP fused to 10 Myc epitope tags that can be combined with anti-Myc antibody staining to attain bright fluorescent labeling of subcellular structures (Viswanathan et al., 2015). Infection of L4 excitatory neurons with AAV-CAG-FLEX-smFP_myc resulted in strong dendritic labeling that improved the ability to accurately trace apical and basal dendrites of starter cells (Figure 3C, E, and F). Three weeks after injection of a mixture of the three AAV helper viruses, EnvA+ RVdG-PSD95GFP-SynPhRFP was injected into the same location in V1 and allowed to express for 7 days. As expected from transsynaptic spread, we observed that many thalamic dLGN long-distance inputs to L4 excitatory neurons were reliably labeled with RVdG-PSD95GFP-SynPhRFP (Figure 3D).

Efficiency of transsynaptic spread from excitatory L4 Nr5a1+ starter cells to excitatory inputs.

(A) Schematic illustration of experimental design and timeline for monosynaptic rabies tracing. Nr5a1Cre mice were injected in V1 with a mixture of AAV-CIAO-TVA, AAV-FLEX-H2BBFP-oG, and AAV-FLEX-smFP_myc. Three weeks later EnvA+ RVdG-PSD95GFP-SynPhRFP was injected into the same site and allowed to express for 7 days. (B) Schematic of rabies retrograde spread efficiency quantification paradigm. Starter neurons are distinguished from input neurons based on expression of nuclear BFP from AAV-FLEX-H2BBFP-oG in addition to fusion proteins from EnvA+ RVdG-PSD95GFP-SynPhRFP. Starter neurons expressing smFP_myc are used for synaptic quantification to allow tracing of distal dendrites. Synaptic fraction is measured by quantifying the proportion of postsynaptic densities (PSD95GFP) on the starter neuron apposed with rabies-labeled presynaptic terminals (SynPhRFP). (C) Representative example images of V1 injection site, obtained using widefield fluorescence microscopy at ×10 magnification. Scale bar = 200 µm. (D) Coronal section example images obtained using widefield fluorescence microscopy at 10× showing long-range monosynaptic input neurons in dorsal lateral geniculate nucleus (dLGN) to Nr5a1+ L4 neurons in V1 when using the new RVdG construct (top). No retrograde spread is observed when glycoprotein is omitted, see Figure 2 for additional information. Insets are zoomed in images of dashed box regions. Scale bar represents 1 mm in hemisection image or 200 mm in inset. (E) Max intensity projection reconstruction of images obtained using Airyscan super-resolution imaging at 63×. Left, example image of starter neuron (H2BBFP+,PSD95GFP+, and SynPhRFP+) labeled with smFP_myc. Middle, example image of the distal domain of an apical dendrite of a starter neuron. Right, example image of the proximal domain of a basal dendrite. Scale bar = 20 µm (all three). (F) Spatial resolution using Airyscan imaging is sufficient to quantify rabies transsynaptic spread at the synaptic level. Zoomed in max intensity projection reconstructed image of boxed region in (E) right, illustrating PSD-95 puncta colocalized with cytoplasmic smFP_myc. Top, yellow boxed region highlights a spine with PSD-95 puncta without an apposed rabies-labeled presynaptic density. White boxed region highlights a spine with PSD-95 puncta with an apposed rabies-labeled presynaptic density. Middle row, zoomed in max intensity projection reconstructed images of yellow boxed region and bottom rows are zoomed in max intensity projection reconstructed images of white boxed region. Top, scale bar = 2 mm and middle and bottom scale bar = 0.5 mm. (G) Percent of postsynaptic densities (PSD95GFP) on Nr5a1+ starter cells apposed with rabies-labeled presynaptic terminals (SynPhRFP). Left, quantification of colocalization at baseline (no glycoprotein) due to L4 to L4 connections compared to colocalization from transsynaptic spread (with glycoprotein). Middle, colocalization on the distal vs proximal domains of apical dendrites. Right, colocalization on apical vs basal dendrites. Values are reported as mean ± SEM. Statistics were calculated from the Wilcoxon rank-sum test for non-parametric comparisons. Individual data points (circles) indicate values for each neuron. n (number of neurons) = 5 and 9 and N (number of mice) = 2 and 3 for no oG and oG groups respectively. p-value > 0.05 = not significant (ns).

To quantify the proportion of excitatory synaptic contacts labeled on starter neurons (defined as SF), we first identified starter neurons that also expressed smFP using widefield fluorescence microscopy. Select areas of starter neuron dendrites that could be traced back to their parent cell bodies were imaged using Airyscan super-resolution microscopy (Huff et al., 2017) to increase the resolution and signal-to-noise and allow visualization of single synaptic puncta (Figure 3F). To assess the transsynaptic spreading efficiency, we quantified the proportion of starter cell excitatory postsynaptic specializations, labeled with PSD95GFP, that were directly apposed with rabies-labeled presynaptic terminals, labeled with SynPhRFP (Figure 3F). Because L4 excitatory neurons connect to one another, the direct connections between starter cells can generate co-labeling of pre- and postsynaptic specializations independent from transsynaptic spread. Co-labeling independent of transynaptic spread might also result from direct rabies infection of the small numbers of non-starter neurons infected with rabies due to leak expression of TVA (see above). It was therefore important to begin by quantifying how much of this ‘background label’ is present under conditions in which similar numbers of L4 neurons are directly infected with EnvA+ RVdG-PSD95GFP-SynPhRFP, but there is no transsynaptic spread. To quantify this we calculated the proportion of PSD95GFP puncta apposed to SynPhRFP in experiments that omitted oG (no AAV8-hSyn-FLEX-H2BBFP-oG, Figure 2A, B, and C). The omission of oG prevents retrograde spread of rabies virus and labeling of input neurons (Figure 3D), therefore observed SynPhRFP colocalization with PSD95GFP must be a result of L4 to L4 excitatory starter neuron connections and/or infection via TVA leak expression. We found that neurons in which oG was used for trans-complementation displayed significantly higher proportions of PSD95GFP puncta colocalized with SynPhRFP compared to neurons in which oG was omitted (50.61 ± 3.92% vs 8.86 ± 1.79% mean ± SEM respectively, Wilcoxon rank-sum test, p=0.001, n=9 neurons across 3 mice and n=5 neurons across 2 mice; Figure 3G). Within experimental condition groups, puncta colocalization did not vary significantly across neurons from distinct experimental animals. The difference between these values (50.61–8.86%) yields an estimate of 42% of postsynaptic densities whose presynaptic partners are transsynaptically labeled (SF = 0.42).

To determine how SF is related to distance from the cell body we sampled from different portions of the dendritic arbors of L4 starter cells. Sampled areas within 75 µm of the soma were classified as proximal dendritic domains and areas within 75 µm of the pial surface were considered distal domains. These distal domains are estimated to be about 250–350 µm from the soma. We found no significant differences between SF at proximal regions of apical dendrites compared to the distal regions as determined by PSD95GFP and SynPhRFP puncta colocalization (49.14 ± 3.49% vs 57 ± 5.43% respectively, Wilcoxon rank-sum test, p=0.34; n=9 neurons across 3 mice, Figure 3G). We also compared SF at basal dendrites versus apical dendrites and observed no significant difference (44.52 ± 3.96% vs 53.05 ± 3.99% respectively, Wilcoxon rank-sum test, p=0.22; n=9 neurons across 3 mice and n=6 neurons across 3 mice, Figure 3G). Note that a fraction of the observed synaptic labeling might occur through indirect pathways rather than direct spread of rabies virus at the observed synapse (see below). This fraction is small relative to the variability in our measured SF at distal versus proximal or apical versus basal dendrites.

Efficiency of rabies transsynaptic spread across excitatory synapses onto inhibitory starter neurons

To assess possible differences in efficiency of rabies retrograde spread for different starter cell types, we conducted monosynaptic rabies tracing using RVdG-PSD95GFP-SynPhRFP as described above, but using mouse lines that express Cre recombinase in two distinct classes of inhibitory neurons. We used the knock-in SstCre and VipCre lines to target initial infection to either somatostatin (Sst)-expressing or vasoactive intestinal peptide (Vip)-expressing inhibitory neurons (Figure 4A). We selected Sst and Vip interneurons as both exhibit very low levels of recurrent connections compared to parvalbumin-expressing neurons, which connect extensively to each other (Campagnola et al., 2022; Pfeffer et al., 2013). We found no significant difference between the percent of PSD95GFP puncta apposed to SynPhRFP on Sst dendrites compared to Vip dendrites (36.13 ± 1.73% vs 38.38± 1.22% respectively, Wilcoxon rank-sum test, p=0.44, n=9 neurons across 3 mice per line; Figure 4F). These values are comparable to the SF for labeling inputs to L4 starter cells (42%, see above). We also assessed whether SF varied based on distance from the cell body. For Sst starter cells, colocalization of PSD95GFP with SynPhRFP on proximal dendritic domains (within 75 µm of the cell body) did not differ significantly from distal dendritic domains (within 75 µm of the pial surface) (37.50 ± 1.93% vs 33.85 ± 2.43% respectively, Wilcoxon rank-sum test, p=0.55, Figure 4F). We observed similar results for Vip starter cells, with no difference between SF at proximal versus distal domains (39.80 ± 1.90% vs 36.122 ± 1.14% respectively, Wilcoxon rank-sum test, p=0.11, Figure 4F).

Efficiency of transsynaptic spread from inhibitory starter cells to excitatory inputs.

(A) Representative example images of V1 injection site for SstCre mouse line (top) and VipCre mouse line (bottom), obtained using widefield fluorescence microscopy at 10×. Scale bar = 200 µm. (BE) Max intensity projection reconstruction of images obtained using Airyscan super-resolution imaging at 63×. Example image of Vip+ (B) and Sst+ (D) starter neurons (H2BBFP+, PSD95GFP+, and SynPhRFP+) labeled with smFP_myc. Scale bar = 20 µm. (C and E) Zoomed in images of boxed regions in (B and D) respectively, illustrating PSD-95 puncta colocalized with cytoplasmic smFP_myc. Top, yellow boxed region highlights a spine with PSD-95 puncta without an apposed rabies-labeled presynaptic density. White boxed region highlights a spine with PSD-95 puncta with an apposed rabies-labeled presynaptic density. Middle row, zoomed in images of yellow boxed region and bottom rows are zoomed in images of white boxed region. Top, scale bar = 2 µm and middle and bottom scale bar = 0.5 µm. (F) Left, percent of postsynaptic densities (PSD95GFP) on Vip+ and Sst+ starter cells apposed with rabies-labeled presynaptic terminals (SynPhRFP). Middle, colocalization on the distal vs proximal domains of Sst+ dendrites. Right, colocalization on the distal vs proximal domains of Vip+ dendrites. Values are reported as mean ± SEM. Statistics were calculated from Wilcoxon rank-sum test for non-parametric comparisons. Individual data points (circles) indicate values for each neuron. n (number of neurons) = 9 and N (number of mice) = 3. p-value > 0.05 = not significant (ns).

Relationships between SF, IF, and U

Because prior studies have found large variability in the number of rabies-labeled input cells per starter cell (CI), there is great interest in understanding what fraction of all input neurons are labeled (IF). But for the great majority of circuits, the actual numbers of inputs to each neuron is unknown, so it is not possible to calculate IF from CI. Below, along with detailed derivation in Materials and methods, we provide a formal analysis and quantitative estimates of the relationships between SF and IF for cortical starter neurons. Overall, these analyses indicate that the SF data we have presented here provide a good estimate and important new insight into IF under the tracing conditions that we have used. But it is important to consider several factors that make SF likely to be somewhat larger than IF. Our analyses (Figures 56, and more detailed considerations in Materials and methods) indicate that for the cortical circuits we studied, our observed SF value of around 40% likely corresponds to an IF of about 30%. We further estimate that when SF is 40% for a cortical neuron, the probability of rabies spread across a single synaptic contact (unitary synaptic efficiency, U) is likely to be about 28%. For most applications, this spreading efficiency is likely to be near optimal because it can allow for detection of weak versus strong input sources (see Discussion); but there is clearly room for increasing spread if the experimental application or question calls for labeling closer to all input neurons.

Quantification from electron micrographic reconstructions of: (A) shared inputs between layer 4 excitatory neurons and (B) the number of synapses per excitatory neuron to excitatory neuron connection.

(A) The percentage of inputs to a layer 4 excitatory neuron that provide common input to another layer 4 excitatory neuron (percentage of neurons connected) at increasing distances between layer 4 neurons. Based on identification of 17,883 presynaptic inputs to layer 4 cells. (B) Distribution of the numbers of synapses per connection for 76,678 excitatory neuron to excitatory neuron connections identified in electron micrographic reconstructions of a volume from adult mouse visual cortex. Y-axis is plotted on a log scale. Average number of synapses per connection = 1.0965 ± 0.3726 (STD). The overwhelming majority of connections involve only one synaptic contact.

Relationships between synaptic fraction (SF), input fraction (IF), and unitary synaptic efficiency (U) depend on synapses per input neuron (M) and proportions of neurons with differences in M.

(A and C) SF against the ratio (R) of neurons with M values of 1 versus either 3 (A) or 2 (C) for different values of U. (B and D) Ratio of IF versus SF against the R of neurons with M values of 1 versus either 3 (B) or 2 (D) for different values of U.

What is the expected relationship between SF and IF and what are the factors that make SF likely to be larger than IF? We have defined SF as the proportion of synaptic contacts that are labeled on a given starter neuron. It is important to note that SF will typically not be the same as the proportion of input neurons that are labeled (IF). In Materials and methods along with Figures 56, we formally consider factors that are likely to impact the relationship between SF and IF. We consider these relationships in the context of cortical circuits, as we have studied here, but the factors considered can also be applied to other brain areas with different circuit organization.

There are two main features of cortical circuit organization that influence the SF to IF relationship. These are: (1) the extent to which different starter neurons share common inputs; and (2) the distribution of the numbers of synaptic contacts provided per input neuron onto a particular starter neuron. Both of these factors have small impacts for our studies of cortical circuits using the Nr5a1Cre line to define starter cells. The first of these, shared common inputs, is small in cortex due to the sparse connectivity and the sparsity of starter cells in our experiments, but at most might cause SF to be about 20% more than IF (see Materials and methods for detailed consideration). The second factor is more complex, but our analyses show that for cortical circuits it is simplified by the fact that the great majority of connections are mediated by a single synaptic contact (Figure 5). As detailed in Materials and methods along with Figures 56, overall IF is likely to be at least 78% of SF (ratio of IF/SF = 0.78). With IF/SF = 0.78 our measured SF of 0.4 corresponds to 31% of input neurons labeled (IF = 0.31) and the probability of rabies virus spread across a single synaptic contact is estimated as 0.28 (U=0.28).

Discussion

Genetically modified RVdG has proven instrumental in deciphering the intricate connectivity patterns of neural networks. Here, we investigated the efficiency with which RVdG spreads retrogradely across excitatory synapses and report, for the first time, direct measurements of the proportion of starter neurons’ postsynaptic densities whose presynaptic inputs are labeled (SF) using the monosynaptic rabies tracing system. Our results demonstrate that with the particular AAV helper viruses, AAV titers, and EnvA+ RVdG titers and strain we used, about 40% of excitatory synaptic contacts onto the examined cell types were labeled. Further considerations detailed in the Results and Materials and methods indicate that the measured SF of 40% likely corresponds to at least 30% of input neurons labeled (IF = 0.30) and a probability of rabies spread across individual synapses (unitary efficiency, U) of 28%.

Of note, we found that efficiency of spread did not vary depending on the distance of synapses to the cell body or across dendrite types. We were concerned that more distal synapses might be less likely to be labeled than proximal synapses because antibody staining for rabies particles shows that during the first 1–3 days post-infection, particles are concentrated around the cell body; but they are not detected more distally until later time points (Nassi and Callaway, 2006). In the experiments described here, efficiency of spread was quantified at 7 days post-infection. It therefore appears that any initial bias for spread at proximal synapses might be eliminated by use of longer survival times. Since we only compared efficiency at one typical time point, care should be taken in interpreting results at shorter times post-infection. The same factors that we have considered for their impact on the relationships between measured SF, IF, and U are also important to consider with respect to the spatial locations of labeled synapses. This is because synaptic labeling resulting from either shared inputs or multiple synaptic contacts from a single input neuron are not the result of the spread of rabies across a synaptic contact. Our analyses show that the proportions of labeled synapses arising from these mechanisms is small relative to the variability and proportions of observed SF at distal versus proximal synapses (Figures 3G and 4F).

Previous studies quantifying the numbers of rabies-labeled input cells per starter cell (CI) have identified several different factors that can increase or decrease the efficiency of transsynaptic spread. While there are few direct comparisons with matched titers of rabies viruses and helper viruses, both the strain of rabies virus and the strain and variant of rabies glycoprotein are likely to be influential. Importantly, when a particular rabies strain and glycoprotein are used, the CI appears to depend most strongly on the levels of G expressed in the starter cells (Callaway and Luo, 2015). Studies suggest that maximizing G expression, either through the use of strong promoters (Miyamichi et al., 2013) or optimized helper virus concentration (Lavin et al., 2020), increases spread efficiency. Importantly, the use of 2A linker elements has been implicated in reduced expression of G (Wall et al., 2010; Watabe-Uchida et al., 2012). Thus, the use of separate AAVs, with one expressing glycoprotein independently from other genes, is recommended to improve efficiency.

A second factor likely to influence efficiency is the number of rabies particles entering a starter cell (Callaway and Luo, 2015). The variability between individual cells that we sampled (Figures 2G and 3C) is likely to reflect these differences. For example, there is expected variability in the numbers of oG-expressing AAV copies and EnvA+ RVdG particles entering each starter cell. Despite the variability that we observe between individual cells, under matched conditions these differences converge to similar means when averaged across even modest numbers of cells (typically 9 cells in our sample) that are far lower than the number of total starter cells in a typical animal. It is noteworthy that it would be straightforward to increase the SF well above the 35–40% we observed by simply increasing the levels of oG expression. For example, our AAV helper viruses were designed to allow unambiguous detection of starter cells for accurate quantification, and we therefore utilized a 2A linker element to express both nuclear BFP and oG from a single helper virus. For experiments where unambiguous starter cell detection is not crucial, oG expression can be increased by expression as a single protein from an alternate helper virus. Additionally, oG expression can be further increased through use of an expression amplification system such as TRE/tTA (Lavin et al., 2020).

These observations suggest that the efficiency of rabies tracing can be manipulated over a wide range depending on experimental goals. But counterintuitively, the ideal rate of spread is probably far less than 100% for most experiments. This is because the spread of rabies is likely to depend on the numbers of synaptic contacts as well as the size of the postsynaptic density, both of which are related to functional connection strength (Holler et al., 2021; Murthy et al., 2001). Thus, if a reasonably small fraction of inputs were labeled, the numbers of labeled presynaptic cells of different types would be expected to scale with the strength of connections from each cell type. But if rabies were to spread across 100% of synaptic contacts, this sensitivity would be lost; every input cell would be labeled regardless of connection strength. On the other hand, if efficiency were exceedingly low one would expect more variability between animals and the potential to fail to detect inputs from weak connections. As considered in our formal modeling of the relationships between proportion of synaptic contacts labeled, proportion of inputs labeled, and efficiency of rabies spread across individual synaptic contacts, a value of 40% of contacts labeled (SF = 0.4) likely corresponds to about 30% of inputs labeled (IF = 0.30) and a unitary synaptic efficiency of about 28% (U=0.28). We therefore suggest that values in the range of 35–50% of synaptic contacts labeled are likely to be near optimal in providing both large numbers of labeled input neurons and some degree of sensitivity to connection strength (Henrich et al., 2020). The equations that we derived (see Materials and methods) can also be used to provide insight into the relationships between numbers of synaptic contacts per input neuron and the relative probability that it will be labeled in comparison to other neurons with a different number of synaptic contacts. These equations can be used to assess circuit configurations that are likely to be different in other brain areas than for the cortical circuits on which we have focused.

Considering the finding that rabies labels less than half of synaptic inputs, it is likely that differential spread of rabies virus leads to differences in the probability of distinct input types being labeled. Although we found evidence that the subcellular location of synaptic contacts of input cells onto the dendrites of starter cells does not affect spread efficiency, it remains unknown what leads to the observed incomplete labeling. It is possible that this is affected by the number of synaptic contacts a presynaptic neuron makes onto the starter cell, differences in uptake receptors on the axon of distinct input cell types, or differences in postsynaptic density size (Callaway and Luo, 2015). Indeed, one limitation of our experimental design is that our light microscopy images do not allow accurate measurement of postsynaptic density size, which would require electron microscopy. Furthermore, it is important to note that we only examined spread efficiency at excitatory synapses labeled with the excitatory postsynaptic density marker PSD-95 and did not quantify the efficiency of spread to inhibitory presynaptic inputs. It should also be noted that the locations and numbers of postsynaptic sites were assessed based on expression of PSD-95-GFP from the rabies virus. If this overexpression were to somehow result in new puncta that do not have a presynaptic input this could result in an underestimate of proportion of actual PSDs labeled.

Overall, we report the efficiency of spread achieved in multiple conditions consisting of distinct starter neuron types. Additionally, the described approach can be used by other researchers to test efficiency of spread across different rabies reagents, distinct input cell types, and starter neuron types not tested in this study.

Materials and methods

Mouse transgenic lines

Request a detailed protocol

All experimental procedures were approved by the Salk Institute Animal Care and Use Committee (IACUC). All the animals were handled according to the approved IACUC protocol 13-00005. C57BL/6J mice were used as wild-type. GENSAT BAC transgenic Nr5a1-Cre (Jackson Laboratory stock # 006364), knock-in Som-IRES-Cre (Jackson Laboratory stock #013044), knock-in Vip-IRES-Cre (Jackson Laboratory stock #028578), and GENSAT BAC transgenic Sim1_KJ18-cre mice have been previously described (Gerfen et al., 2013; Harris et al., 2014; Taniguchi et al., 2011). Transgenic mice were maintained on C57BL/6J backgrounds. Mice were housed with a 12 hr light and 12 hr dark cycle and ad libitum access to food and water. Both male and female mice were used for experiments.

Virus preparation

Request a detailed protocol

The following AAVs were produced by the Salk GT3 Viral Core: AAV8-hSyn-FLEX-H2BmTagBFP2-oG (3.57E+13 GC/ml) and AAV8-nef-AO-66/71-TVA950 (5.25E+13 GC/mL). AAV1-CAG.FLEX.GFPsm_myc.WPRE.SV40 (1.12E+13 GC/ml) was purchased from Addgene. EnvA+ RVdG-5PSD95eGFP-SynPhRFP (1.55E+08 IU/ml) was produced by the Salk GT3 Viral Core. AAVs and rabies virus can be purchased from the Salk Viral Vector Core.

Animal surgery for virus injection

Request a detailed protocol

For rabies transsynaptic spread efficiency experiments, mice received AAV helper injections at postnatal day (P) 50. Mice were initially anesthetized with 2% isoflurane and maintained at 1.5% isoflurane after placement on a stereotax (David Kopf Instruments, Model 940 series) for surgery and stereotaxic injections. A small craniotomy was made with a mounted drill over the primary visual cortex of the left hemisphere using the following coordinates: 3.4 mm posterior and 2.6 mm lateral relative to bregma. For transsynaptic tracing experiments AAV8-hSyn-FLEX-H2BmTagBFP2-oG, AAV8-nef-AO-66/71-TVA950, and AAV1-CAG.FLEX.GFPsm_myc.WPRE.SV40 were mixed at a ratio of 1:40, 1:100, and 1:2 to final concentrations of 8.92E+11, 5.25E+11, and 5.6E+12 GC/ml respectively. For control experiments with no oG AAV8-nef-AO-66/71-TVA950 and AAV1-CAG.FLEX.GFPsm_myc.WPRE.SV40 were mixed to a titer matched concentration of 5.25E+11 and 5.6E+12 GC/ml. 100 nl of mixture was injected into the center of V1 0.5–0.7 mm ventral from the pia using a pulled glass pipette with a tip size of 30 µm connected to a 1 ml syringe with 18 G tubing adapter and tubing. To prevent backflow, the pipette was left in the brain for 5 min after injection. Three weeks after AAV helper injection, 150 nl of EnvA+ RVdG-5PSD95eGFP-SynPhRFP was injected into the same site in V1. After recovery, mice were given water with ibuprofen (30 mg/kg) and housed for 7 days to allow for transsynaptic rabies spread and fluorescent protein expression.

Histology

Request a detailed protocol

Seven days after rabies injection, brains were harvested after transcardial perfusion using PBS followed by 4% paraformaldehyde (PFA). Brains were dissected out from skulls and post-fixed with 2% PFA and 15% sucrose in PBS at 4°C for 16–20 hr, then immersed in 30% sucrose in PBS at 4°C before sectioning. 50 µm coronal brain sections were prepared using a freezing microtome. Free-floating sections were incubated at 4°C for 16 hr with rabbit anti-Myc (1:500, C3956; Sigma-Aldrich) primary antibody in PBS/0.5% normal donkey serum/0.1% Triton X-100, followed by donkey anti-rabbit conjugated to Alexa 647 (1:500, A-21206, Thermo Fisher) at room temperature for 2–3 hr. Immunostained tissue sections were mounted on slides with polyvinyl alcohol mounting medium containing DABCO and allowed to air-dry overnight.

For characterization of the population of Cre-expressing neurons in Nr5a1Cre mice, these animals were crossed with AI14 reporter mice (Jackson Laboratory stock #007914) to yield TdTomato expression in all Cre-expressing neurons (Wang et al., 2023). Three adult animals (73–193 days of age) were perfused and brains removed, and sectioned as described above. Sections were processed as follows: antigen retrieval was performed using 10 nM sodium citrate buffer and then sections were immunostained with rabbit anti-NeuN (1:1000, ab177487, Abcam), mouse anti-GAD67 (1:1000, MAB5406, MilliporeSigma), rat anti-RFP (1:1000, 5F8-100, Chromotek) followed by Alexa 488 donkey anti-rabbit (1:1000, A21207, Thermo Fisher), Alexa 647 donkey anti-mouse (1:500, A31571, Thermo Fisher), and CF568 donkey anti-rat (1:500, 89138-546, Biotium), secondaries. Sections were first imaged on an Olympus BX63 microscope using a 10×/0.4 NA objective (Olympus) to identify the borders of V1 and layer 4. Sections for cell counting were then imaged on a Zeiss LSM880 confocal microscope using a 20×/0.8 NA objective. A stitched z-stack image was taken so that the entirety of V1 in one of the hemispheres of that section could be used for quantification. Within these regions the x,y,z coordinates of all RFP-positive (TdTomato) cells were marked and then these coordinates were used to calculate nearest-neighbor distances and the numbers of labeled neurons located with 50 µm of each labeled neuron. Across 774 labeled neurons whose positions were marked in 7 sections from 3 animals, we find that both the mean and median numbers of labeled neurons located within 50 µm of any given labeled neuron of interest is 7. Additional analyses also confirmed that only GAD-negative excitatory neurons were labeled.

Image acquisition and analysis

Request a detailed protocol

Individual sections were first scanned at ×10 magnification using an Olympus BX63 widefield fluorescent microscope to detect starter neurons triple positive for H2BBFP-oG, smFP_myc, and RVdG-PSD95GFP-SynPhRFP. Individual starter neurons were then imaged using a Zeiss LSM 880 Airyscan FAST Microscope with a Plan-Apochromat 63×/1.4 Oil DIC M27 objective. First, neurons were confirmed to be starter neurons by checking for expression of nuclear BFP. Z-stacks of images were acquired with a step interval of 500 nm for select dendritic domains. Dendrites immediately next to the soma and spanning 75 µm away were defined as proximal dendrites and representative images were acquired. Using expression of smFP_myc, dendrites were traced from the soma up to the pial surface. Dendritic regions spanning 75 µm from the pial surface were acquired and designated as distal domains. Images were processed and analyzed using NIH ImageJ software (FIJI). Airyscan super-resolution imaging can resolve structures ~120 nm apart in the lateral dimension and ~350 nm apart in the axial dimension. The synaptic cleft size, or space between the pre- and postsynaptic density, ranges from 20 to 30 nm, making it outside of the range of the resolution limit. Therefore, we required partial overlap, or colocalization, between pre- and postsynaptic markers and did not allow for any distance between the two markers to consider them a synapse. For synaptic counting, images were first analyzed with only the far-red (smFP_myc) and green channel visible to count all PSD95GFP-labeled postsynaptic densities colocalized with the far-red labeled dendritic region. After postsynaptic counts for that region were quantified, the red channel was turned on to quantify the number of presynaptic densities colocalized with postsynaptic densities. Due to the abundance of synapses across a 50 µm coronal section, puncta were counted manually one z-plane at a time. Images presented in manuscript are max intensity projection reconstructions to allow for 2D visualization of 3D structures. Wilcoxon rank-sum test for non-parametric comparisons was used for statistical analysis. For all figures: ***, p<0.001; ns = not significant, p>0.05.

EM analysis methods

Request a detailed protocol

Analysis of synaptic connectivity statistics from structural electron microscopy was done on the publicly available dataset of mouse visual cortex from the MicRonS project (Bae et al., 2023). To assess the number of synapses per excitatory connection, neurons were surveyed from the collection of cells that were annotated as having ‘clean’ or ‘extended’ axons, which reflect the set of axons which have been removed of false mergers. We included only neurons which had more than 100 synaptic outputs and were excitatory. For each excitatory cell, between 65% and 90% of its synapses could be mapped to a post-synaptic soma (Bae et al., 2023). Most post-synaptic somas have been cell typed in the dataset based on a model trained on soma and nucleus features (Elabbady et al., 2022), using labels provided by a combination of human labeling and unsupervised clustering (Schneider-Mizell et al., 2023). Therefore, we were able to calculate the average number of synapses per excitatory connection as a function of cell type for many cell-type interactions.

Further, for assessing shared connections onto layer 4 cells specifically, we calculated the distance from each post-synaptic soma to others layer 4 somas in the volume, and then measured which of those had connections from the same pre-synaptic axon. We then calculated the mean connected fraction as a function of distance between the two post-synaptic somas. All data analysis was done in python using the CAVEclient to access the dataset (Dorkenwald et al., 2023) and numpy (Harris et al., 2020), pandas (McKinney, 2010; team, T. pandas development, 2023), and scipy (Virtanen et al., 2020) to calculate these values.

Calculations of SF and IF

We conducted formal analyses of how several different factors in the organization of cortical circuits might influence the relationship between the proportion of postsynaptic specializations on a given starter neuron that have rabies-labeled presynaptic terminals (SF), the probability that rabies spreads across a single synaptic contact (U), and the proportion of input neurons to the starter neuron that are labeled (IF). We consider two different influences that are both expected to result in measured values of SF that are larger than IF. These are: (1) the extent to which different starter neurons share common inputs; and (2) the distribution of the numbers of synaptic contacts provided per input neuron onto a particular starter neuron.

The potential impact of common inputs

Request a detailed protocol

Shared common inputs are expected to result in an SF value that is larger than the input proportion labeled (SF>IF). This is because the divergent connections of an input neuron can result in the labeling of a synaptic contact on a starter neuron even if rabies virus never spreads from the starter neuron at that particular synaptic contact. Specifically, rabies can spread from a different starter neuron resulting in expression of synRFP in the divergent input neuron. That synRFP could then result in labeling of a synaptic contact without any spread of rabies virus across the labeled contact. The rate of synaptic labeling on a given starter cell by this indirect route depends on three factors: the numbers and spatial distribution of other starter cells; the proportion of common inputs that are shared between starter cells at different distances; and the probability that inputs to a starter cell are labeled (IF).

We directly assessed the maximum possible numbers of other layer 4 starter cells and their spatial distributions by analyzing the density and distribution of Cre-expressing neurons in Nr5a1Cre mice. And we analyzed a 3D reconstructed EM volume to directly measure the probability that a neuron presynaptic to a layer 4 excitatory neuron also provides a common input to a neighboring layer 4 neuron (Figure 5A).

To calculate the proportion of the measured value of SF that arises from common input we sum together the contributions of all neurons providing shared input and then multiply by the estimated proportion of those neurons that were labeled (IF). Using an estimated IF of 0.3 (IF = 0.75 SF from common input and multiple contacts combined, see below), we obtain the following proportions of the measured value of SF that arise from common inputs at different distances from the subject neuron. At 0–50 µm the proportion of shared input is 5% (Figure 5A) and there are 7, nr5a1 neurons. If the rate at which inputs are labeled (IF) is 0.3 (see above and below) then 30% of the 5% shared inputs = 1.5% more labeled synapses can potentially be added indirectly by a second nearby starter neuron. But if 30% of the neurons providing input to the cell of interest were already labeled directly, then only 70% of the potential additions, or 1%, will be new additions. As additional nearby starter cells are incorporated into the network and their inputs become labeled, each new cell will add a progressively smaller percentage of new indirect inputs because the portion of previously unlabeled synapses will decrease; but for small numbers of input neurons, 1% remains a close approximation. For 7 other starter cells located within 50 µm of the subject neuron the total maximum contribution (assuming all nr5a1 neurons were rabies-infected starters) would be 7% of the measured SF value. At 50–100 µm there are 13 additional nr5a1 neurons and the rate of shared input is 3.3%, yielding a total contribution of 9%. At 100–200 µm there are an additional 28 potential starter neurons and the rate of shared input is 1%, yielding a total contribution of 6%. Contributions beyond 200 µm are negligible (less than 1%) due to the very low rate of common input (~0.1%). Altogether, the maximum possible proportion of labeled synapses that result from shared input is about 0.22 (0.07+0.09+0.06). Thus, this factor would be expected to result in SF being at most 22% greater than IF (IF = 0.78 SF). Since the proportion of nr5a1 neurons that are directly infected with rabies and serve as starters is almost certainly less than 1 and probably closer to 0.5, this is an overestimate. If only half of the nr5a1 neurons are starter cells then the proportion of labeled synapses resulting from shared input would be only 0.11 and IF/SF = 0.89.

The impact of the distribution of the numbers of synaptic contacts provided per input neuron onto a particular starter neuron

Request a detailed protocol

We next consider how the numbers of synaptic contacts provided by each input neuron onto a given starter neuron will influence the relationship between SF and IF. Overall our consideration of this issue indicates that this will have only a small effect on our measures of excitatory inputs to cortical neurons because more than 90% of such connections involve only a single synaptic contact (see below). Regardless of the actual configuration of inputs, the maximum value of IF will be SF (SF = IF), but for most configurations SF is larger than IF (SF>IF; details below). It is informative to first consider scenarios in which SF = IF because this is not entirely intuitive. The most intuitive case is a configuration in which each input neuron invariably provides just one synaptic contact to the starter neuron. Here, it is straightforward to infer that SF and IF are equal because the only way to label an input neuron is for rabies to spread from the single synaptic contact and then for SynRFP to spread to and label that same synaptic contact. But for cases in which some or all input neurons provide more than one synaptic contact onto a starter neuron, the possible scenarios are less intuitive. Generally, such a configuration will cause IF to be less than SF (but not always). For example, rabies virus might spread from one of the contacts provided by an input neuron and this would then result in production of synRFP that labels all of the contacts provided by that input neuron. Taken at face value, one would expect that this would cause SF to be larger than IF; this is usually, but not always the case. In an extreme scenario, suppose that each input neuron has multiple synaptic contacts, but the number of synaptic contacts is equal for each neuron. In this case, SF will be equal to IF. For example, suppose a starter neuron has a total of 1000 synaptic contacts (S) that arise from 200 input neurons (N) each providing exactly 5 contacts. Even if rabies virus spreads from just 1 of the 5 contacts provided by an input neuron and then labels all 5 contacts, the proportion of postsynaptic specializations labeled (SF) will always be equal to the proportion of input neurons labeled (IF), regardless of how efficiently rabies virus spreads between neurons. If N*=number of labeled input neurons and S*=number of labeled synaptic contacts, then IF = N*/200; SF = S*/1000; S*/N*=5; therefore SF = IF. This relationship holds for all cases of S/N provided that all input neurons have the same S/N.

The actual configuration of cortical circuits is likely more complex and the precise features are not known for all input sources. However, available data based on a reconstructed EM volume (see EM analysis methods) indicate that in adult mouse visual cortex, it is rare for a local connection from an excitatory neuron onto another cortical neuron to involve more than one synaptic contact. Figure 5B plots the distribution of the numbers of synaptic contacts per connection for 76,678 excitatory neuron to excitatory neuron connections, involving 503 pre-synaptic cells and 26,079 post-synaptic cells (note the log scale of the Y-axis in Figure 5B). Out of 76,678 connections, only about 6000 (~8%) involve more than one synapse and only about 1000 (~1.3%) involve 3 or more synapses. The average number of synapses per connection is 1.0965 (std = 0.3726). Note that these values are much lower than observed in previously available data based on connections that were first identified by electrophysiology in brain slices from 12- to 23-day-old rats, followed by EM reconstruction of the labeled neurons (Feldmeyer et al., 1999; Feldmeyer et al., 2002; Feldmeyer et al., 2006). We consider the newer, much more extensive, and relatively unbiased data from adult mouse visual cortex to be more relevant to our experiments in adult mouse visual cortex. Nevertheless, we consider the potential impact of larger values since the equations we derive and the examples that we illustrate can be informative for future consideration of labeling in other circuits and cell types.

Here, we describe how the distribution of input ratios (S/N=synapses per input neuron) influences the relationship between SF and IF. We refer to S/N as multiplicity, or M. For simplicity, we consider only cases with two different populations of input neurons (a and b), each with different numbers of synapses per neuron (Ma and Mb). With this scenario the relationship between SF and IF depends on both the proportion of input neurons in the a and b populations (ratio, R=Na/Nb) and on the probability that rabies virus spreads across a single synaptic contact (unitary synaptic efficiency, U). Below, we provide equations and their derivation that allow for calculation of both SF and IF given: (1) the ratio (R=Na/Nb) of the numbers of neurons (N) in population a (Na) and in population b (Nb); (2) the multiplicity of synaptic contacts per neuron (M) for population a (Ma) and population b (Mb); and (3) the unitary synaptic efficiency (U).

At an extreme, there might be just 1 S/N for one population (a, Ma = 1) and 3 S/N for another population (b, Mb = 3). Figure 6A and B plots the calculated values of SF and IF/SF, respectively, for different ratios (R) of Na/Nb and for different levels of unitary synaptic efficiency (U) when Ma = 1 and Mb = 3. Under these conditions and with a worst-case estimate of U=0.22 (our measured SF of 0.4 suggests actual U is closer to 0.28, see below), the smallest possible IF/SF value is ~0.8. If these were the actual values, then our measured SF of 0.4 would correspond to an IF value of 0.32, or 32% of all inputs labeled.

We consider a scenario with Ma = 1 and Mb = 2 to provide a closer simulation of the actual configuration of cortical circuits since values of 1 and 2 synapses per neuron account for nearly 99% of all connections (Figure 5B). Figure 6C and D plots the calculated values of SF and IF/SF, respectively, for different ratios (R) of Na/Nb and for different levels of unitary synaptic efficiency (U) when Ma = 1 and Mb = 2. Here, we see that the smallest IF/SF is now calculated when R is about 2 (Figure 6B) and that with R=2.0 our measured SF values of around 0.4 are typically calculated when U~=0.28 (Figure 6A). With U=0.28 and R=2.0, IF/SF in this scenario is about 0.91 (Figure 6B). Finally, if we consider that based on the actual measurements for cortical circuits the value for R is at least 10 (Figure 5B), our measured SF value of 0.4 corresponds to U~=0.35 (Figure 6C). With U=0.35 and R=10, IF/SF is about 0.95 (Figure 6D). Therefore, under this ‘most plausible’ scenario considering the impact of multiple synaptic contacts, our measured SF of 0.4 suggests that about 38% of input neurons are labeled (IF = 0.38) and that the probability of rabies virus spread across a single synaptic contact is 35% (U=0.35).

These estimates of IF and U are made independently from the effects of common input, which, as detailed above, are expected to result in IF being at most 78% of the measured value for SF. If we incorporate an additional maximum impact due to multiple synaptic contacts based on IF/SF = 0.95, then we obtain an overall estimate of 0.78×0.95 = 0.74=IF/SF. With measured values of SF = 0.4 we estimate IF ~=0.3. Calculations above indicate that the impact of multiple synapses per contact will result in IF that is about 8% larger than U. Therefore, we estimate that, at a minimum, U=0.28.

Derivation of equations relating SF to IF

Request a detailed protocol

We considered conditions in which there are two different populations of neurons, a and b, that each have different numbers of synaptic contacts (S) onto the starter neuron per input neuron (synapses per neuron, S/N, or multiplicity M). We assume that when there are multiple synaptic contacts from an input neuron to a starter neuron, the probability that rabies virus spreads across a synaptic contact (U) is the same for all contacts from both input populations and is independent for each contact. Further, if rabies spreads across one of an input neuron’s contacts, the synRFP expressed from rabies virus in that neuron spreads to and labels all the synaptic terminals from that input neuron.

As shown in the equations and derivation below, the values for SF and IF depend on: (1) the ratio (R) of the numbers of neurons (N) in population a (Na) and population b (Nb); (2) the numbers of synaptic contacts per input neuron (multiplicity, M) for population a (Ma) and population b (Mb), and the probability that rabies virus will spread across an individual synaptic contact (U).

  • N=number of presynaptic neurons; Ntot=Na+Nb

  • S=number of synaptic contacts on the postsynaptic neuron; Stot=Sa+Sb

  • R=ratio of the numbers of presynaptic neurons; R=NaNb ; Na=RNb

  • M=number of synaptic contacts made by each presynaptic neuron; Ma or Mb

  • U=unitary synaptic efficiency, probability of rabies spread across a single synaptic contact

  • p=probability of labeling a presynaptic neuron; pa or pb

  • N*=number of labeled presynaptic neurons; Na or Nb

  • S*=number of labeled synaptic contacts; Sa or Sb

px=1(1U)Mx
Nx=Nxpx
Na=RNbpa
Sx*=Nx**Mx
IF=NtotNtot= Na+NbNa+Nb=(Napa)+(Nbpb)Na+Nb=(NbRpa)+(Nbpb)(NbR)+Nb=(Rpa)+pbR+1
Stot=Sa+Sb=(NaMa)+(NbMb)=(NbRpaMa)+(NbpbMb)=Nb[(RpaMa)+(pbMb)]
Stot=Sa+Sb=Na*Ma+Nb+Mb=R*Nb*Ma+Nb+Mb=Nb*[(Ma*R)+Mb]
SF= StotStot= Nb[(RpaMa)+(pbMb)]Nb[(MaR)+Mb]=(RpaMa)+(pbMb)(MaR)+Mb

Statistics and reproducibility

Request a detailed protocol

No statistical methods were used to pre-determine sample sizes; rather sample sizes were established based upon similar studies in the literature. Stereotaxic injections were repeated in three biological replicates for each mouse line, and both male and female mice were used. All statistical tests were performed using R version 4.1.1 and details of individual tests are described in figure legends. All data are reported as the mean ± SEM, except synapses per connections which is mean ± std.

Data availability

All data generated or analyzed in this study are included in figures and Source data 1.

References

Decision letter

  1. Liqun Luo
    Reviewing Editor; Howard Hughes Medical Institute, Stanford University, United States
  2. Lu Chen
    Senior Editor; Stanford University, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting the paper "Quantification of Monosynaptic Rabies Tracing Efficiency" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen a Senior Editor. The reviewers have opted to remain anonymous.

We are sorry to say that, after consultation with the reviewers, we have decided that this work will not be considered further for publication by eLife.

Comments from all three reviewers are attached. The reviewers also had a robust discussion session after all reviews were submitted. While all three reviewers appreciate the value of your work, the consensus is that the advance is not sufficient for a full research article. In particular, the key measurement of "labeling efficiency at synaptic level" is very easily confused by readers as "the efficiency with which rabies virus crosses synapses", which is a more valuable measure that is less dependent on specific connectivity architectures of specific circuits.

We regret that we cannot consider the current manuscript further. However, if you are able to thoroughly address reviewers' critiques with further work, you are welcome to submit a revised manuscript to eLife as a new submission. We will try our best to recruit the same reviewers.

Reviewer #1 (Recommendations for the authors):

G-deleted rabies virus has been widely used to trace monosynaptic inputs to specific populations of neurons in the mammalian brain since its first report by Callaway and colleagues in 2007. Despite its wide use, many fundamental properties of the method remain to be characterized. In this short report, Patino et al. addressed two important questions: (1) what fraction of synapses does the rabies virus cross from a starter cell? (2) does this vary for synapses located at different distances to the soma? Answers to these questions should help researchers to interpret their data and thus will be of general interest to the neural circuit community.

To answer these questions, the authors produced a new rabies virus that expresses both a presynaptic marker tagged with RFP and a postsynaptic marker (for excitatory neurons) tagged with GFP, and simply counted within starter cells the fraction of GFP+ puncta that have an adjacent RFP punctum. Since most RFP puncta are a result of the trans-synaptic transfer of rabies virus (the authors did a nice control of skipping G in one experiment), the authors can "quantify efficiency of spread at the synaptic level". The authors found about 35-40% of GFP+ puncta have RFP puncta next to them, and interestingly this number does not change whether these quantifications were performed close to the cell body or far from the cell body (thus answering question #2). However, the current version has a number of issues that should be addressed.

1. While the authors conclude properly that under their experimental conditions, "35% to 40% of excitatory inputs to the examined cell types were labeled," many readers would mistakenly think that "rabies virus crosses 35-40% of synapses of the starter cells" from reading this paper. The latter is unlikely true for two reasons. First, if starter cell 1 transfers rabies virus to presynaptic neuron 1, which not only connects with starter cell 1 but also starter cell 2, then a pair of green/red puncta between starter cell 2 and presynaptic neuron 1 could result even if rabies virus never crossed the synapse between starter cell 2 and presynaptic neuron 1. Second, if starter cell 1 forms 5 synapses with presynaptic neuron 1, then it is possible that the rabies virus just crossed 1 of the 5 synapses (there may be some data supporting 5-10 synapses between synaptic pairs of cortical excitatory neurons), and the other 4 pairs of puncta are incidental. Thus, the actual fraction of synapses that the rabies virus crosses is likely substantially below 35-40%. While the first caveat can be overcome by doing experiments from single starter cells, the second caveat cannot be avoided. It is possible that the authors could estimate -based on cortical connectivity data (luckily V1 is one of the best studied in this regard, including serial EM reconstruction)-"what fraction of synapses does the rabies virus cross," which is a more fundamental question likely generalizable across cell types than "what fraction of excitatory inputs are labeled," which will vary greatly across circuits because of specific connection properties. At the very least, the authors should distinguish these two questions clearly, and state why they can differ, so that readers won't come away being misled.

2. I am confused by Figure 2F, one of the most important panels of the entire paper that provides examples of GFP/RFP co-staining at high resolution. GFP marks postsynaptic density and thus should colocalize with the dendritic marker (magenta) and RFP marks presynaptic terminals of presynaptic neurons and should be adjacent but outside the magenta-positive area. But the third row (the positive example) shows the exact opposite. Red is within the magenta area, whereas green is adjacent to the magenta area. Can the authors clarify?

3. This brings an important issue of setting objective criteria for synaptic pairs: how much overlap, what's the maximum/minimum distance, etc, as this is the basis for the entire study. Indeed, what counts as a synaptic pair based on pre and postsynaptic markers in light microscopy is a debated issue. The authors are helped here by the sparse labeling of starter cells and presynaptic partners; still, I highly recommend that the authors state clearly the criteria in Methods.

4. The negative control of omitting oG in Figure 2 is very nice; I hope that the authors will do the same for tracing inputs to the two types of inhibitory neurons. These inhibitory neurons are also known to form gap junctions and it will be interesting to see whether the rabies virus would jump through gap junctions without oG, even if they are not known to form substantial chemical synapses between them.

Reviewer #2 (Recommendations for the authors):

Patiño et al. developed a new rabies viral vector that contains both PSD95GFP (an excitatory postsynaptic marker) and SynPhRFP (a general presynaptic marker) to quantify what proportion of all excitatory presynaptic contacts on a starter cell are labeled by rabies virus-mediated retrograde monosynaptic tracing. They defined the labeling efficiency as the proportion of PSD95GFP puncta that are directly opposed to SynPhRFP puncta, and compared the labeling efficiency values among different dendritic positions (i.e. apical, proximal, and basal dendrites) or different neuronal types (i.e. L4 excitatory neurons, Sst+ interneurons, and Vip+ interneurons). They find no significant difference in the labeling efficiency depending on subcellular locations and neuronal types. They show that about 40% of first-order presynaptic excitatory contacts on a starter cell are labeled in their experimental conditions with careful control experiments. Overall, the experimental design is sound although additional validation is necessary. This work demonstrates "the labeling efficiency at synaptic levels" of rabies virus-mediated input tracing. However, there are several concerns in relation to their claim and the significance of this analysis.

1) The authors assume that all PSD95GFP puncta receive presynaptic contacts and the total number of PSD95GFP represents the total number of synapses on starter cells. This should be validated by examining the proportion of PSD95GFP puncta that associate with endogenous excitatory presynaptic markers such as VGLUT1.

2) The authors appear to use "the efficiency of spread" as a phrase equivalent to "the labeling efficiency". However, these are different; the efficiency of spread should be defined as the proportion of synapses through which rabies virus particles are actually retrogradely transported. Assuming that one input neuron makes five presynaptic contacts onto a starter cell, even if rabies virus particles pass through only one synapse among them, all five presynaptic contacts should be labeled. Thus, labeling efficiency depends on the efficiency of spread and the number of presynaptic contacts per input neuron. The efficiency of spread is equivalent to the labeling efficiency only when one input neuron forms one presynaptic contact with a starter cell.

3) There remain a number of questions about rabies virus-mediated input tracing, part of which are discussed by the authors. What percentage of all (excitatory or inhibitory) input cells are labeled? How large should the size of postsynaptic density be to allow for the retrograde spread of rabies virus particles? How many synapses does an input neuron form onto a starter cell? The number of synapses an input neuron forms depends on the subcellular locations of a starter cell or types of a starter cell. What presynaptic receptors are required for the transsynaptic spread of rabies viruses? Addressing these questions should be important for understanding the mechanism underlying rabies virus spread as well as more detailed connection properties of neural circuits. Unfortunately, the present study falls short of providing biological implications beyond "the labeling efficiency at synaptic levels".

1) In Figure 3, higher magnification images representing PSD95GFP puncta that associate with SypPhRFP puncta should be included.

2) It would be easier for readers to understand how the cross-over insensitive ATG-out (CIAO) AAV construct works if the authors could provide a concise explanation in the text.

Reviewer #3 (Recommendations for the authors):

The efficiency of rabies virus-mediated retrograde trans-synaptic tracing, one of the prevailing methods for deciphering cell-type specific neural connectivities, has long been in the debate yet has not definitively been addressed. To directly determine the efficiency, the authors took a simple and elegant strategy to visualize the excitatory post-synaptic structures with GFP-fused PSD95 and pre-synaptic structures with TagRFP-T-fused synaptophysin. They genetically modified the rabies genome to allow the simultaneous targetings of pre- and post-synaptic fluorescent markers to both the post-synaptic starter cells and their pre-synaptic partners. By high-resolution confocal microscopy, they showed along a dendrite of the starter post-synaptic neurons, the excitatory post-synaptic structures were visualized with GFP+ puncta, some of which were tightly paired with the RFP+ pre-synaptic puncta. After subtracting some background labeling caused regardless of trans-synaptic labeling, the authors revealed that, under their experimental conditions, approximately 42% of the post-synaptic puncta were coupled with the trans-synaptically labeled pre-synaptic puncta, hence providing the labeling efficiency of rabies-tracing. Importantly, the efficiency of rabies tracing was similar between excitatory and inhibitory starter cells, and also grossly invariant the along proximal vs. distal dendrites of the starter cells. These data provide useful basic properties of the rabies tracing method. However, the main conclusion could contain an overestimation because not all the labeled pre-synaptic structures were necessarily visualized from the starter cell of analysis. Overall, the present study at least estimated an upper limit of rabies tracing efficiency, which if carefully discussed should be useful for the community.

1) The conclusion can be accepted without modification if the total number of starter cells in their experimental condition was near one. However, the authors simply generated starter cells by co-infection of AAVs in a cell-type specific Cre driver mouse line, in which usually hundreds of starter cells are generated in the injection site. Given that the labeled pre-synaptic neurons can connect to multiple post-synaptic partners located nearby, not all the labeled pre-synaptic structures found along the labeled dendrites of a given starter cell were necessarily visualized from the exact starter cell of analysis. This property of neural circuits likely led to an overestimation of rabies-tracing efficiency reported in the present study. To characterize if this overestimation happened, the authors can first analyze the number of starter cells in their samples and then examine a correlation between the number of starter cells and the fraction of GFP/RFP coupling. Ideally, increasing the number of animal samples with a smaller number of total starter cells could provide a better estimate. Overall, the present study at least estimated an upper limit of rabies tracing efficiency, which if carefully discussed should be useful for the community.

2) Because GFP+ post-synaptic puncta were highly dense in the field of analysis, it is important to establish that the puncta detected along the Myc+ dendrites were mostly derived from the cell of analysis, not from adjacent uncharacterized cells. This can simply be achieved if the authors could analyze the density of GFP+ puncta along the Myc-positive but oG-negative dendrites as a negative control.

3) For reproducibility, it is important to report more details of viral constructions, in particular, RVdG-5PSD95eGFP-SynPhRFP, and describe resource availability. Because not all non-US researchers obtain rabies stock from the Salk Viral Core, depositing the corresponding plasmids (and most importantly the sequence information) to a more generally accessible distributor such as Addgene would be appreciated.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Detection of monosynaptic rabies spread at synaptic resolution reveals factors influencing tracing efficiency" for further consideration by eLife. Your revised article has been evaluated by Lu Chen (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

1) Please streamline the descriptions in the Results section "Relationships between…" (see comments from Reviewers #1 and #2 below). Please consider using new nomenclatures suggested by Reviewer #2 (all three reviewers agree on this during our discussion).

2) Please address Reviewer #1's critique #1 regarding the 4% probability of shared common inputs only applies to a pair of starter cells using suggested analysis and potentially new data.

3) Please address Reviewer #3's critique regarding the match between the title and data shown in the paper.

4) Please address all other critiques as much as you can.

Reviewer #1 (Recommendations for the authors):

The efficiency of rabies virus (RV)-mediated retrograde trans-synaptic tracing, a prominent technique for unraveling cell-type specific neural connectivities, has not yet been definitively addressed. To assess the labeling efficiency at the synaptic level and examine the relationship between the synaptic and cellular labeling, the authors employed rabies-encoding GFP-fused PSD95 and TagRFP-fused synaptophysin to simultaneously label both post-synaptic structures on the starter cells and their corresponding pre-synaptic counterparts. Following the subtraction of background labeling unrelated to trans-synaptic labeling, the authors revealed that, under their experimental conditions, approximately 40% of the post-synaptic puncta were coupled with the trans-synaptically labeled pre-synaptic puncta, thereby providing the labeling efficiency at the synapse level (synaptic efficiency, SE). Importantly, SE values were comparable between excitatory and inhibitory starter cells, as well as consistent along the proximal and distal dendrites of the starter cells. Lastly, the authors offer a theoretical framework to interpret SE and to potentially establish a connection with the cellular labeling efficiency of pre-synaptic neurons (Input probability, IP). While the manuscript encompasses significant technical and theoretical advancements, I noticed two major points that warrant careful consideration in interpreting the author's conclusions.

1. To analyze the associations between SE and IP, the authors appropriately introduced two factors. First, multiple starter neurons can share common input neurons, implying that the pre-synaptic structure of an observed starter neuron may be non-autonomously labeled from other starter cells. To characterize this effect, the authors conducted a thorough analysis of the existing literature and provided an estimated probability of 4% for two randomly selected cortical starter cells to share a common pre-synaptic neuron. Based on this estimate, they concluded that "this factor would be expected to result in SE being about 4% greater than IP". However, this conclusion holds only when analyzing samples with two starter cells. In reality, SE values are likely to be influenced by the number of starter cells (Ns) because, in a typical experiment involving Ns = 200-500 and given a 4% probability of shared common inputs in any pair of starter cells, nearly all labeled pre-synaptic neurons of the observed starter cell can be non-autonomously labeled from at least one of the other starter cells.

2. The authors also appropriately introduced the second point: multiple pre-synaptic structures provided by an input neuron can all be labeled if the RV passes through any one of them. I appreciate the author's attempt to suggest that this factor may have only a modest influence on the relationship between SE and IP (Figure 5). However, I am concerned that this factor has a more pronounced impact on the interpretation of the observed spatial organization of the pre-synaptic structures. Even in an extreme scenario where the RV passes through only the proximal (P) or distal (D) synapse, assuming that the pre-synaptic neurons evenly provide pre-synaptic structures along the PD axis, the result would be the comparable SE values between P and D synapses. The manuscript would benefit from a more meticulous discussion of this issue.

3. Considering my major point #1 above, I believe the manuscript would gain from the inclusion of an estimation of SE pertaining to a single starter cell, which is unlikely to be 0.4. The authors could begin by analyzing the number of starter cells in their samples and then perform a regression analysis to estimate the SE of a single starter cell. Increasing the number of animal samples with a smaller number of starter cells could contribute to obtaining a more accurate estimate. Furthermore, it would be advantageous to simulate the relationships between the number of starter cells (Ns) and SE, using assumed circuit structures and the efficiency of RV spread across a single synaptic contact (referred to as unitary efficiency, U).

4. Although I appreciate the authors' endeavors to estimate the relationships between SE and IP, the relevant text could benefit from a more concise writing style, aligning with the rest of the manuscript. For eLife readers with diverse backgrounds, the inclusion of schematic diagrams would be instructive.

Reviewer #2 (Recommendations for the authors):

G-deleted rabies virus has been widely used to trace monosynaptic inputs to specific populations of neurons in the mammalian brain since its first report by Callaway and colleagues in 2007. Despite its wide use, many fundamental properties of the method remain to be characterized. In this short report, Patino et al. addressed two important questions: (1) what the fraction of synapses does the rabies virus cross from a starter cell? (2) does this vary for synapses located at different distances to the soma? Answers to these questions should help researchers to interpret their data and thus will be of general interest to the neural circuit community.

To answer these questions, the authors produced a new rabies virus that expresses both a presynaptic marker tagged with RFP and a postsynaptic marker (for excitatory neurons) tagged with GFP, and simply counted within starter cells the fraction of GFP+ puncta that have an adjacent RFP punctum. Since most RFP puncta are a result of trans-synaptic transfer of rabies virus (the authors did a nice control of skipping G in one experiment), the authors can "quantify efficiency of spread at the synaptic level". The authors found about 35-40% of GFP+ puncta have RFP puncta next to them, and interestingly this number does not change whether these quantifications were performed close to the cell body or far from the cell body (thus answering question #2).

I appreciate the length the authors went to address my critiques of the original manuscript on the issue of distinguishing (1) fraction of synapses that are co-labeled in the starter cell, (2) fraction of input neurons that are labeled by rabies virus, and (3) average efficiency of rabies virus crossing an individual synapses. This will clarify to the readers what the authors have done and what kind of conclusions they can take away. However, I have several remaining critiques to the above issue that should be addressed by further textual revision.

1) I don't think the terms the authors used to describe the above three terms are the clearest. The authors used "synaptic efficiency" to describe (1), but it can easily be confused with (3). I recommend synapse proportion (SP) to parallel with input proportion (IP), or perhaps a bit better synapse fraction (SF) vs. input fraction (IF). Then (3) can be changed to "Unitary synapse efficiency (U)". These terms may be widely adopted in future publications on rabies tracing, so careful consideration is warranted.

2) The textual description in the Results section "Relationships between…" is often repetitive and needs to be streamlined. Specifically, I don't like the description of the results in the first paragraph ahead of all the analyses (and the authors have done these repetitively). For example, most of the first paragraph should be trimmed-just give the definition of the terms and move on to the actual "formal analysis." Summarize the results at the end.

3) The results of the formal analysis is based on the cortical circuit. It's important to emphasize that in circuits with denser connectivity, the relationship will be very different. As such, when these properties are discussed in other parts of the manuscript (abstract, introduction, discussion, etc.), please make sure to include the qualification phrase that these inferences were made from cortical circuits (e.g., last paragraph of Introduction).

Reviewer #3 (Recommendations for the authors):

1) Prior studies have quantified the proportion of PSD95 labeled postsynaptic densities that are opposed by presynaptic terminals that are also labeled with various presynaptic markers, including synaptophysin and VGLUT1 (Neuron. 2010 Nov 18; 68(4): 639-653. doi: 10.1016/j.neuron.2010.09.024). Essentially every PSD95 labeled terminal has a labeled presynaptic contact. Synaptophysin is a reliable marker of presynaptic terminals but VGLUT1 is not. We have modified the text in the introduction which now reads: "We designed a new genetically modified rabies virus that labels presynaptic terminals with synaptophysin-RFP (SynPhRFP) and excitatory postsynaptic densities with postsynaptic density-95-GFP (PSD95GFP). Because more than 99% of PSD95 postsynaptic puncta co-localize with a presynaptic terminal (Micheva et al., 2010), this construct allows us to quantify the proportion of excitatory synapses on a starter cell that have their corresponding input neuron labeled with rabies virus (defined here as synaptic efficiency [SE])."

Comment: I understand that "endogenous" PSD95 is a reliable postsynaptic marker in vivo. However, considering this is a new reagent and rabies viruses highly express transgenes, it is fair to experimentally show the colocalization rate of PSD95GFP expressed from rabies viruses and endogenous presynaptic markers (e.g. synaptophysin).

2) We thank the reviewer for this excellent point. We have been very careful not to equate efficiency of spread to labeling efficiency. As described in the above response to a similar comment from reviewer #1 we have added a new Results section, "Relationships between synaptic efficiency, input proportion, and unitary efficiency". Here the value that we formally define as unitary efficiency (U) is equivalent to what the reviewer terms "efficiency of spread", while our measured values of synaptic efficiency (SE) correspond to what the reviewer describes as "labeling efficiency". We show that there are complex relationships between the proportion of labeled presynaptic terminals (SE), efficiency of spread across individual synapses (U), the distributions of numbers of synaptic contacts between connected neuron pairs, and the proportion of input neurons labeled (IP).

Comment: I appreciate the authors' effort to estimate IP and U from SE and that the authors clearly discriminate SE and the spread efficiency in the text. The results are informative to some extent but not definitive because they rely on several assumptions such as "U is identical across synapses" and "there are two populations of input neurons that make the distinct numbers of presynaptic contacts per neuron onto starter cells", which are likely different from real situations. I am afraid that this analysis is not strong enough to claim the utility of the novel rabies vectors.

3) We would like to thank the reviewer for these thought-provoking questions. This is a rather disparate set of comments each of which relates to very different properties of either cortical circuit organization or rabies tracing. Some of these are known for limited aspects of cortical circuits while others remain unknown. With respect to rabies virus spread, some of these are known or can be inferred from published observations.

Below we further address each of the reviewer's specific questions.

Comment: There seems to be a misunderstanding of my previous review comment. I just listed examples of key questions about neural circuit organization and the mechanisms for the rabies virus spread, whose answers should bring important biological insights. I did not mean to request the authors to address all these questions.

4) We disagree with the stated characterization of our results: "Addressing these questions should be important for understanding the mechanism underlying rabies virus spread as well as more detailed connection properties of neural circuits. Unfortunately, the present study falls short of providing biological implications beyond "the labeling efficiency at synaptic levels". First, it was not a goal of our study to understand the mechanisms underlying the spread of rabies virus nor did we intend to reveal detailed connection properties of neural circuits. As stated in our Introduction, we were interested in identifying factors that influence the efficiency of transynaptic spread of rabies and we designed our experiments to address two such factors. These are the distance of synaptic contacts from the cell body and the type of postsynaptic neuron. Our data show that the proportions of labeled synaptic contacts are similar regardless of distance from the cell body and regardless of the postsynaptic cell types tested. These are quite important factors and given that the overall efficiency of spread can easily be increased or decreased by changing glycoprotein expression levels, these measures are more important than the overall measured efficiency in our experiments.

Comment: The title of the manuscript is now "Detection of monosynaptic rabies spread at synaptic resolution reveals factors influencing tracing efficiency". However, major achievements/findings of this study are that new rabies virus vectors that enable visualization of input synapses onto starter cells and estimation of the fraction of labeled presynaptic contacts are developed and that the proportions of labeled synaptic contacts are similar regardless of distance from the cell body and regardless of the postsynaptic cell types tested, as the authors state above. I am afraid that the authors do not provide "factors that influence (alter) tracing efficiency", which makes me feel that the title does not match with the data. If the scope of the manuscript is neither to understand the mechanisms underlying the spread of rabies virus nor to reveal detailed connection properties of neural circuits, can the authors clarify what kind of important questions could be addressed using this novel tool (the utility of the tool)?

https://doi.org/10.7554/eLife.89297.sa1

Author response

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

Comments from all three reviewers are attached. The reviewers also had a robust discussion

Reviewer #1 (Recommendations for the authors):

G-deleted rabies virus has been widely used to trace monosynaptic inputs to specific populations of neurons in the mammalian brain since its first report by Callaway and colleagues in 2007. Despite its wide use, many fundamental properties of the method remain to be characterized. In this short report, Patino et al. addressed two important questions: (1) what fraction of synapses does the rabies virus cross from a starter cell? (2) does this vary for synapses located at different distances to the soma? Answers to these questions should help researchers to interpret their data and thus will be of general interest to the neural circuit community.

To answer these questions, the authors produced a new rabies virus that expresses both a presynaptic marker tagged with RFP and a postsynaptic marker (for excitatory neurons) tagged with GFP, and simply counted within starter cells the fraction of GFP+ puncta that have an adjacent RFP punctum. Since most RFP puncta are a result of the trans-synaptic transfer of rabies virus (the authors did a nice control of skipping G in one experiment), the authors can "quantify efficiency of spread at the synaptic level". The authors found about 35-40% of GFP+ puncta have RFP puncta next to them, and interestingly this number does not change whether these quantifications were performed close to the cell body or far from the cell body (thus answering question #2). However, the current version has a number of issues that should be addressed.

1. While the authors conclude properly that under their experimental conditions, "35% to 40% of excitatory inputs to the examined cell types were labeled," many readers would mistakenly think that "rabies virus crosses 35-40% of synapses of the starter cells" from reading this paper. The latter is unlikely true for two reasons. First, if starter cell 1 transfers rabies virus to presynaptic neuron 1, which not only connects with starter cell 1 but also starter cell 2, then a pair of green/red puncta between starter cell 2 and presynaptic neuron 1 could result even if rabies virus never crossed the synapse between starter cell 2 and presynaptic neuron 1. Second, if starter cell 1 forms 5 synapses with presynaptic neuron 1, then it is possible that the rabies virus just crossed 1 of the 5 synapses (there may be some data supporting 5-10 synapses between synaptic pairs of cortical excitatory neurons), and the other 4 pairs of puncta are incidental. Thus, the actual fraction of synapses that the rabies virus crosses is likely substantially below 35-40%. While the first caveat can be overcome by doing experiments from single starter cells, the second caveat cannot be avoided. It is possible that the authors could estimate -based on cortical connectivity data (luckily V1 is one of the best studied in this regard, including serial EM reconstruction)-"what fraction of synapses does the rabies virus cross," which is a more fundamental question likely generalizable across cell types than "what fraction of excitatory inputs are labeled," which will vary greatly across circuits because of specific connection properties. At the very least, the authors should distinguish these two questions clearly, and state why they can differ, so that readers won't come away being misled.

We thank the reviewer for these excellent points. We have added a new section “Relationships between synaptic efficiency, input proportion, and unitary efficiency” to the results, which also includes a new figure (Figure 5). In this section we formally consider and quantitatively assess the impacts of both of the factors mentioned above (shared inputs and input neurons with multiple synaptic contact). This section along with a new methods section (“Calculations of synaptic efficiency (SE) and input proportion (IP)”) describes the derivation of formal equations that allow an evaluation of how the proportions of labeled presynaptic terminals in our experiments (SE) relate to the proportion of input neurons (IP) labeled and the probability (U) that rabies virus spreads across a single synaptic contact (unitary efficiency). Consideration of known properties of cortical circuits informs the range of variables that are input to the derived equations and the resulting values under different conditions are plotted in the figures. Based on these calculations we infer that our observed value of about 0.4 for SE likely corresponds to about 32% of input neurons labeled and a probability of spread across a single contact of 25%. Further, under unlikely, “worst-case scenario” conditions, percentage of input neurons labeled would be about 24%.

We believe that the addition of this material, as well as slight modifications in wording throughout the manuscript, will help to assure that readers will not incorrectly interpret our results and “mistakenly think that "rabies virus crosses 35-40% of synapses of the starter cells" from reading this paper.” For example, the introduction now summarize this part of the results (which is too long to fit in the Abstract) as: “Finally, we model the factors that influence the relationship between SE and the proportion of input neurons labeled (Input Probability, IP) and show that under plausible conditions the IP/SE ratio is about 0.75, so an SE of 40% corresponds to an IP of 30% of input neurons labeled. This analysis also allows an estimate of the probability of rabies spread across a single synaptic contact (unitary efficiency, U) which we estimate to be about 0.25 (25%) when SE=0.4.”

2. I am confused by Figure 2F, one of the most important panels of the entire paper that provides examples of GFP/RFP co-staining at high resolution. GFP marks postsynaptic density and thus should colocalize with the dendritic marker (magenta) and RFP marks presynaptic terminals of presynaptic neurons and should be adjacent but outside the magenta-positive area. But the third row (the positive example) shows the exact opposite. Red is within the magenta area, whereas green is adjacent to the magenta area. Can the authors clarify?

Regarding GFP and smFP_myc colocalization we would like to highlight that while not perfectly aligned the GFP puncta is indeed within the magenta positive region. We have provided zoomed in images below to highlight this point. Additionally, we have added additional images of synapse examples to figure 3 which further highlight this point.

Regarding overlap of the RFP presynaptic terminal with the magenta cytoplasmic marker, we want to highlight two points. One, the example images are a max intensity projection (MIP) reconstruction of multiple z-planes to enable 2D visualization of 3D structures. Therefore, structures in distinct z-planes may appear to co-localize in the flattened 2D image, for example if the presynaptic axon is above or below the spine. We have edited the manuscript to highlight this critical point by adding the following sentence to the methods sections “Images presented in manuscript are max intensity projection reconstructions to allow for 2D visualization of 3D structures'' and by adding “max intensity projection reconstruction” when describing example images. Of note, MIP was only done for manuscript visualization purposes and counting analysis was done one z-plane at a time as stated in the methods section. Second, although the resolution offered with Airyscan imaging is improved compared to confocal microscopy, it is not without resolution limits. It is therefore likely that structures may not perfectly align as expected, particularly in the axial dimension which has the poorest resolution. Indeed the size of a synaptic cleft (20-40nm) is outside of the resolution limit. Therefore, if the input axon is synapsing perpendicular to the x-y plane it would appear within the magenta region. Finally, we would like to point out that the synapse to the right of the yellow boxed region in Figure 2.F top panel as an example of “RFP should be adjacent but outside the magenta-positive area”. While some synapses do adhere to these expectations, for the reasons discussed above, not all do.

Author response image 1

3. This brings an important issue of setting objective criteria for synaptic pairs: how much overlap, what's the maximum/minimum distance, etc, as this is the basis for the entire study. Indeed, what counts as a synaptic pair based on pre and postsynaptic markers in light microscopy is a debated issue. The authors are helped here by the sparse labeling of starter cells and presynaptic partners; still, I highly recommend that the authors state clearly the criteria in Methods.

We have added the following description to the methods section to address this point:

“Airyscan super-resolution imaging can resolve structures ~120 nm apart in the lateral dimension and ~350 nm apart in the axial dimension. The synaptic cleft size, or space between the pre and postsynaptic density, ranges from 20-30 nm, making it outside of the range of the resolution limit. Therefore, we required partial overlap, or colocalization, between pre and postsynaptic markers and did not allow for any distance between the two markers to consider them a synapse.”

4. The negative control of omitting oG in Figure 2 is very nice; I hope that the authors will do the same for tracing inputs to the two types of inhibitory neurons. These inhibitory neurons are also known to form gap junctions and it will be interesting to see whether the rabies virus would jump through gap junctions without oG, even if they are not known to form substantial chemical synapses between them.

Prior studies (Tang et al. 1999) have shown that rabies virus propagation occurs at chemical synapses but not via gap junctions. Therefore it is not expected that rabies would spread across gap junctions without oG. In addition, all of the components of the rabies virus and its genome are far larger than the small molecules that can spread through gap junctions. Furthermore as stated in the manuscript, Sst and Vip both exhibit very low levels of recurrent chemical synaptic connections, unlike L4 excitatory neurons, and therefore we did not find it necessary to perform similar controls as negligible numbers of direct connections between starter neurons are expected. Indeed, Sst and Vip neurons were chosen for experiments over Pvalb neurons, due to this unique property.

Reviewer #2 (Recommendations for the authors):

Patiño et al. developed a new rabies viral vector that contains both PSD95GFP (an excitatory postsynaptic marker) and SynPhRFP (a general presynaptic marker) to quantify what proportion of all excitatory presynaptic contacts on a starter cell are labeled by rabies virus-mediated retrograde monosynaptic tracing. They defined the labeling efficiency as the proportion of PSD95GFP puncta that are directly opposed to SynPhRFP puncta, and compared the labeling efficiency values among different dendritic positions (i.e. apical, proximal, and basal dendrites) or different neuronal types (i.e. L4 excitatory neurons, Sst+ interneurons, and Vip+ interneurons). They find no significant difference in the labeling efficiency depending on subcellular locations and neuronal types. They show that about 40% of first-order presynaptic excitatory contacts on a starter cell are labeled in their experimental conditions with careful control experiments. Overall, the experimental design is sound although additional validation is necessary. This work demonstrates "the labeling efficiency at synaptic levels" of rabies virus-mediated input tracing. However, there are several concerns in relation to their claim and the significance of this analysis.

1) The authors assume that all PSD95GFP puncta receive presynaptic contacts and the total number of PSD95GFP represents the total number of synapses on starter cells. This should be validated by examining the proportion of PSD95GFP puncta that associate with endogenous excitatory presynaptic markers such as VGLUT1.

Prior studies have quantified the proportion of PSD95 labeled postsynaptic densities that are apposed by presynaptic terminals that are also labeled with various presynaptic markers, including synaptophysin and VGLUT1 (Neuron. 2010 Nov 18; 68(4): 639–653. doi: 10.1016/j.neuron.2010.09.024). Essentially every PSD95 labeled terminal has a labeled presynaptic contact. Synaptophysin is a reliable marker of presynaptic terminals but VGLUT1 is not. We have modified the text in the introduction which now reads: “We designed a new genetically modified rabies virus that labels presynaptic terminals with synaptophysin-RFP (SynPhRFP) and excitatory postsynaptic densities with postsynaptic density-95-GFP (PSD95GFP). Because more than 99% of PSD95 postsynaptic puncta co-localize with a presynaptic terminal (Micheva et al., 2010), this construct allows us to quantify the proportion of excitatory synapses on a starter cell that have their corresponding input neuron labeled with rabies virus (defined here as synaptic efficiency [SE]).”

2) The authors appear to use "the efficiency of spread" as a phrase equivalent to "the labeling efficiency". However, these are different; the efficiency of spread should be defined as the proportion of synapses through which rabies virus particles are actually retrogradely transported. Assuming that one input neuron makes five presynaptic contacts onto a starter cell, even if rabies virus particles pass through only one synapse among them, all five presynaptic contacts should be labeled. Thus, labeling efficiency depends on the efficiency of spread and the number of presynaptic contacts per input neuron. The efficiency of spread is equivalent to the labeling efficiency only when one input neuron forms one presynaptic contact with a starter cell.

We thank the reviewer for this excellent point. We have been very careful not to equate efficiency of spread to labeling efficiency. As described in the above response to a similar comment from reviewer #1 we have added a new Results section, “Relationships between synaptic efficiency, input proportion, and unitary efficiency”. Here the value that we formally define as unitary efficiency (U) is equivalent to what the reviewer terms “efficiency of spread”, while our measured values of synaptic efficiency (SE) correspond to what the reviewer describes as “labeling efficiency”. We show that there are complex relationships between the proportion of labeled presynaptic terminals (SE), efficiency of spread across individual synapses (U), the distributions of numbers of synaptic contacts between connected neuron pairs, and the proportion of input neurons labeled (IP).

3) There remain a number of questions about rabies virus-mediated input tracing, part of which are discussed by the authors.

We would like to thank the reviewer for these thought-provoking questions. This is a rather disparate set of comments each of which relates to very different properties of either cortical circuit organization or rabies tracing. Some of these are known for limited aspects of cortical circuits while others remain unknown. With respect to rabies virus spread, some of these are known or can be inferred from published observations.

We disagree with the stated characterization of our results: “Addressing these questions should be important for understanding the mechanism underlying rabies virus spread as well as more detailed connection properties of neural circuits. Unfortunately, the present study falls short of providing biological implications beyond "the labeling efficiency at synaptic levels". First, it was not a goal of our study to understand the mechanisms underlying the spread of rabies virus nor did we intend to reveal detailed connection properties of neural circuits. As stated in our Introduction, we were interested in identifying factors that influence the efficiency of transynaptic spread of rabies and we designed our experiments to address two such factors. These are the distance of synaptic contacts from the cell body and the type of postsynaptic neuron. Our data show that the proportions of labeled synaptic contacts are similar regardless of distance from the cell body and regardless of the postsynaptic cell types tested. These are quite important factors and given that the overall efficiency of spread can easily be increased or decreased by changing glycoprotein expression levels, these measures are more important than the overall measured efficiency in our experiments.

Below we further address each of the reviewer’s specific questions.

What percentage of all (excitatory or inhibitory) input cells are labeled?

We quantified the spread of rabies virus across excitatory connections and not inhibitory connections. As described in the Discussion: “Furthermore, it is important to note that we only examined spread efficiency at excitatory synapses labeled with the excitatory postsynaptic density marker PSD-95, and did not quantify the efficiency of spread to inhibitory presynaptic inputs.”

We are able to estimate the percentage of excitatory input cells labeled based on our measured values of proportion of PSD95 puncta that are labeled. These estimates depend on many factors that are only partially known, but as detailed in the new Results section, we are able to estimate within a reasonably tight range.

How large should the size of postsynaptic density be to allow for the retrograde spread of rabies virus particles? The number of synapses an input neuron forms depends on the subcellular locations of a starter cell or types of a starter cell. What presynaptic receptors are required for the transsynaptic spread of rabies viruses? Addressing these questions should be important for understanding the mechanism underlying rabies virus spread as well as more detailed connection properties of neural circuits. Unfortunately, the present study falls short of providing biological implications beyond "the labeling efficiency at synaptic levels".

We combine these questions because the answers are interrelated. Despite extensive efforts over several decades to identify receptors for rabies virus uptake, there are not any known molecules that are required for the spread of rabies virus across synaptic contacts. Rabies virus shares this property with many other enveloped viruses such as Vesicular Stomatitis Virus. All available evidence indicates that the only requirement for spread is the presence of a synaptic contact, as rabies does not spread between neurons in vivo in the absence of synaptic contacts. This is a unique property of rabies, likely due to the lack of hetero-dimeric G-proteins that would allow both stabilization of the particle at the cell membrane and subsequent uptake. Instead the homo-dimeric rabies glycoprotein cannot be stabilized at the cell membrane and thus is susceptible to uptake only at synaptic contacts where the viral particle is larger than the synaptic cleft and is immediately stabilized at the presynaptic membrane when budding from a postsynaptic specialization. It is beyond the scope of this manuscript to include details about the mechanisms of rabies spread, about which we conduct no experiments. If only synaptic contact is required, it follows that the probability of rabies virus spread depends on the size of the synaptic contact, and since contact size is related to functional strength, it would follow that the probability of spread is also likely to be related to functional strength. Further, the probability that an input neuron would be labeled should also depend on the number of synaptic contacts it provides. This is an assumption of the formal model of spread and neuron labeling that we have added in the new Results section. We also consider these factors in our Discussion.

It would be of interest to experimentally measure the relationship between synapse size and probability of labeling. For example, this might be accomplished by combining rabies labeling with electron microscopy, which is the most sensitive and accurate method to study questions relevant to synaptic size. This would require the development of novel rabies constructs that are compatible with EM. The Discussion now states: “Indeed, one limitation of our experimental design is that our light microscopy images do not allow accurate measurement of postsynaptic density size, which would require electron microscopy.”

How many synapses does an input neuron form onto a starter cell?

There is not a singular answer to this question because for known connections the number varies depending on the pre- and postsynaptic cell types and there is also diversity within those populations. In the Results section we now consider known values for the small numbers of excitatory cortical connections onto excitatory cortical neurons that have been quantified and assess how those impact the relationships between our measured values of proportions of postsynaptic sites that are labeled and other values of interest.

1) In Figure 3, higher magnification images representing PSD95GFP puncta that associate with SypPhRFP puncta should be included.

Representative higher magnification images have been added to Figure 3.

2) It would be easier for readers to understand how the cross-over insensitive ATG-out (CIAO) AAV construct works if the authors could provide a concise explanation in the text.

We have added the following to the main text.

“CIAO constructs have the ATG codon placed outside of the loxp mutant pairs loxp66/71 sites to ensure the gene coding region is out of frame with the ATG start signal in the absence of Cre-mediated recombination”

Reviewer #3 (Recommendations for the authors):

The efficiency of rabies virus-mediated retrograde trans-synaptic tracing, one of the prevailing methods for deciphering cell-type specific neural connectivities, has long been in the debate yet has not definitively been addressed. To directly determine the efficiency, the authors took a simple and elegant strategy to visualize the excitatory post-synaptic structures with GFP-fused PSD95 and pre-synaptic structures with TagRFP-T-fused synaptophysin. They genetically modified the rabies genome to allow the simultaneous targetings of pre- and post-synaptic fluorescent markers to both the post-synaptic starter cells and their pre-synaptic partners. By high-resolution confocal microscopy, they showed along a dendrite of the starter post-synaptic neurons, the excitatory post-synaptic structures were visualized with GFP+ puncta, some of which were tightly paired with the RFP+ pre-synaptic puncta. After subtracting some background labeling caused regardless of trans-synaptic labeling, the authors revealed that, under their experimental conditions, approximately 42% of the post-synaptic puncta were coupled with the trans-synaptically labeled pre-synaptic puncta, hence providing the labeling efficiency of rabies-tracing. Importantly, the efficiency of rabies tracing was similar between excitatory and inhibitory starter cells, and also grossly invariant the along proximal vs. distal dendrites of the starter cells. These data provide useful basic properties of the rabies tracing method. However, the main conclusion could contain an overestimation because not all the labeled pre-synaptic structures were necessarily visualized from the starter cell of analysis. Overall, the present study at least estimated an upper limit of rabies tracing efficiency, which if carefully discussed should be useful for the community.

We thank the reviewer for these comments

We would like to clarify that our results do not estimate “an upper limit of rabies tracing efficiency”. Instead, for the particular reagents and tracing conditions used, our measures of the proportion of postsynaptic densities with labeled presynaptic terminals provides an upper limit on the proportion of input neurons labeled as well as an upper limit on the probability that rabies spreads across a single synaptic contact. But, as has been shown in prior publications, these values can be readily increased or decreased by changing tracing conditions, such as the titer of rabies virus used, the strain of rabies virus used, the expression levels or strain of rabies glycoprotein, or survival time after rabies injections. These issues are addressed in the manuscript Introduction and Discussion.

1) The conclusion can be accepted without modification if the total number of starter cells in their experimental condition was near one. However, the authors simply generated starter cells by co-infection of AAVs in a cell-type specific Cre driver mouse line, in which usually hundreds of starter cells are generated in the injection site. Given that the labeled pre-synaptic neurons can connect to multiple post-synaptic partners located nearby, not all the labeled pre-synaptic structures found along the labeled dendrites of a given starter cell were necessarily visualized from the exact starter cell of analysis. This property of neural circuits likely led to an overestimation of rabies-tracing efficiency reported in the present study. To characterize if this overestimation happened, the authors can first analyze the number of starter cells in their samples and then examine a correlation between the number of starter cells and the fraction of GFP/RFP coupling. Ideally, increasing the number of animal samples with a smaller number of total starter cells could provide a better estimate. Overall, the present study at least estimated an upper limit of rabies tracing efficiency, which if carefully discussed should be useful for the community.

We thank the reviewer for highlighting the importance of addressing limitations of distinguishing between labeled presynaptic terminals that are the result of direct rabies infection versus indirect routes. As described in more detail in response to the other two reviews (above), we have added a new Results section “Relationships between synaptic efficiency, input proportion, and unitary efficiency” to formally, and quantitatively assess these issues. We note that the approach suggested by the reviewer would still fail to reveal precise values for efficiency of spread across a synaptic contact or the proportion of input neurons labeled. For example, even if there were only one starter cell and every labeled input cell were counted, the expected fraction of labeled postsynaptic densities would still depend on unknown values, including the total number of presynaptic neurons and the numbers of synaptic contacts provided by the various input neurons. The reasons for this should be apparent from the new Results section.

2) Because GFP+ post-synaptic puncta were highly dense in the field of analysis, it is important to establish that the puncta detected along the Myc+ dendrites were mostly derived from the cell of analysis, not from adjacent uncharacterized cells. This can simply be achieved if the authors could analyze the density of GFP+ puncta along the Myc-positive but oG-negative dendrites as a negative control.

We apologize to the reviewer for being unclear about the analysis methods. To clarify, due to the abundance of synapses across a 50m coronal section, GFP+ post-synaptic puncta were counted manually one z-plane at a time (with each z-step spanning 0.5 m) and confirmed to be colocalized with the Myc+ dendrite from the starter cell being analyzed. We have added the following text to the methods section to clarify this point. “Due to the abundance of synapses across a 50m coronal section, puncta were counted manually one z-plane at a time. Images presented in manuscript are max intensity projection reconstructions to allow for 2D visualization of 3D structures.” While puncta from uncharacterized input cells could be near the dendrite being analyzed they would not be expected to colocalize with the Myc+ dendrite when being counted one z-plane at a time. To address the second point, analyzing the density of GFP+ puncta along the Myc-positive but oG-negative dendrites would not serve as a negative control as these neurons are also infected with rabies and express their own GFP+ puncta; in fact such neurons were analyzed in control experiments that omitted the AAV-FLEX-H2BBFP-oG virus (Figure 2).

3) For reproducibility, it is important to report more details of viral constructions, in particular, RVdG-5PSD95eGFP-SynPhRFP, and describe resource availability. Because not all non-US researchers obtain rabies stock from the Salk Viral Core, depositing the corresponding plasmids (and most importantly the sequence information) to a more generally accessible distributor such as Addgene would be appreciated.

In response to the reviewer's concerns about reproducibility, we have begun the process to deposit the plasmid and sequence information for the RVdG-5PSD95eGFP-SynPhRFP viral construct at Addgene to make it more widely available to researchers. The plasmid sequence has also been included as a Supplementary file in the manuscript.

[Editors’ note: what follows is the authors’ response to the second round of review.]

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

1) Please streamline the descriptions in the Results section "Relationships between…" (see comments from Reviewers #1 and #2 below). Please consider using new nomenclatures suggested by Reviewer #2 (all three reviewers agree on this during our discussion).

We thank the reviewers and editor for these suggestions, which we believe have led to a more accessible and clearer description of our findings.

We have now moved most of the text that was in the "Relationships between…" Results section to the Materials and methods section. The remaining text in that section is now a streamlined summary, as suggested. Note that much of the related text that is now in the Methods section has been modified, both in response to issue #2 (below) and due to availability of new data on the numbers of synapses made per connection for excitatory cortical circuits as well as the rate of shared common input.

We have adopted new nomenclature suggested by Reviewer #2. Synaptic Efficiency (SE) is replaced by Synaptic Fraction (SF). Input proportion (IP) is replaced by Input Fraction (IF). Unitary Efficiency (U) is now defined with the term Unitary Synaptic Efficiency (U).

2) Please address Reviewer #1's critique #1 regarding the 4% probability of shared common inputs only applies to a pair of starter cells using suggested analysis and potentially new data.

We thank reviewer #1 for correctly pointing out that the proportions of synapses labeled due to shared common input will depend not only on the amount of shared input, but also on the numbers of starter cells that potentially receive common input. To address this issue we have collected and analyzed new data. We have analyzed data from the MiCRONS electron microscopy (EM) dataset to directly assess that rate with which layer 4 neurons in adult mouse visual cortex share common input and how this relates to the distance between neurons. And we have also collected new data and analyzed the density and spatial distributions of Creexpressing neurons that can potentially serve as starter cells in our experiments using the Nr5a1-Cre mouse line.

We used these two datasets to allow quantification of both the numbers of starter cells that might share common input and the proportion of our observed synaptic labeling that might be due to indirect labeling from shared inputs rather than direct spread of rabies across the labeled synaptic contact. These results and detailed analysis can be found in the revised Methods section. (As noted above, this description is in the Methods section in order to streamline the description in the Results, as suggested.)

3) Please address Reviewer #3's critique regarding the match between the title and data shown in the paper.

We have modified the title, which is now: “Postsynaptic cell type and synaptic distance do not determine efficiency of monosynaptic rabies virus spread measured at synaptic resolution”.

There appears to have been some confusion about what exactly are the “factors” that we were referring to in the previous title: “Detection of monosynaptic rabies spread at synaptic resolution reveals factors influencing tracing efficiency.” The reviewer stated: “I am afraid that the authors do not provide "factors that influence (alter) tracing efficiency", But in the immediately preceding sentence the reviewer lists exactly the same factors that we were referring to in the former title. “However, major achievements/findings of this study are that new rabies virus vectors that enable visualization of input synapses onto starter cells and estimation of the fraction of labeled presynaptic contacts are developed and that the proportions of labeled synaptic contacts are similar regardless of distance from the cell body and regardless of the postsynaptic cell types tested …”

We believe that the new title now clearly identifies the “factors” – postsynaptic cell type and synaptic distance – which our experiments have evaluated with respect to their influences on the efficiency of rabies labeling.

4) Please address all other critiques as much as you can.

We address all other critiques in the context of the reviewer’s comments below.

Reviewer #1 (Recommendations for the authors):

The efficiency of rabies virus (RV)-mediated retrograde trans-synaptic tracing, a prominent technique for unraveling cell-type specific neural connectivities, has not yet been definitively addressed. To assess the labeling efficiency at the synaptic level and examine the relationship between the synaptic and cellular labeling, the authors employed rabies-encoding GFP-fused PSD95 and TagRFP-fused synaptophysin to simultaneously label both post-synaptic structures on the starter cells and their corresponding pre-synaptic counterparts. Following the subtraction of background labeling unrelated to trans-synaptic labeling, the authors revealed that, under their experimental conditions, approximately 40% of the post-synaptic puncta were coupled with the trans-synaptically labeled pre-synaptic puncta, thereby providing the labeling efficiency at the synapse level (synaptic efficiency, SE). Importantly, SE values were comparable between excitatory and inhibitory starter cells, as well as consistent along the proximal and distal dendrites of the starter cells. Lastly, the authors offer a theoretical framework to interpret SE and to potentially establish a connection with the cellular labeling efficiency of pre-synaptic neurons (Input probability, IP). While the manuscript encompasses significant technical and theoretical advancements, I noticed two major points that warrant careful consideration in interpreting the author's conclusions.

We thank the reviewer for noting the significance of the advancements provided by our work.

1. To analyze the associations between SE and IP, the authors appropriately introduced two factors. First, multiple starter neurons can share common input neurons, implying that the pre-synaptic structure of an observed starter neuron may be non-autonomously labeled from other starter cells. To characterize this effect, the authors conducted a thorough analysis of the existing literature and provided an estimated probability of 4% for two randomly selected cortical starter cells to share a common pre-synaptic neuron. Based on this estimate, they concluded that "this factor would be expected to result in SE being about 4% greater than IP". However, this conclusion holds only when analyzing samples with two starter cells. In reality, SE values are likely to be influenced by the number of starter cells (Ns) because, in a typical experiment involving Ns = 200-500 and given a 4% probability of shared common inputs in any pair of starter cells, nearly all labeled pre-synaptic neurons of the observed starter cell can be non-autonomously labeled from at least one of the other starter cells.

We thank the reviewer for pointing out the importance of considering the numbers of starter neurons that can potentially provide common input, which we failed to address in our consideration. If the numbers of such neurons were in fact 200-500 and each of those were to provide 4% common inputs this would indeed be a major problem. Such numbers would probably cause nearly every synapse to be labeled, far more than the ~40% that we observed. The actual numbers of starter cells are likely to be much less than 200, but more importantly the probability of sharing common input is highly dependent on the distance between starter neurons. This is because the rate of connections in cortex falls off quite quickly with distance. Thus, a consideration of the spacing between potential starter cells is also required.

To allow us to incorporate a consideration of not only the numbers of starter cells but also their distances from each other, we conducted a quantitative analysis of the distributions of Creexpressing neurons in the nr5a1-Cre mice. We crossed the nr5a1-Cre mice with tdTomato reporter mice and then marked the locations of every TdTomato-expressing neuron. We then analyzed their spatial distributions and incorporated these data into a new estimate of the potential impact of common inputs on the measured fraction of synapses labeled for each starter neuron (Synaptic Fraction, SF). We also gained access to new data and analyses of EM data from the MicRonS data set which densely reconstructs connections within a volume of adult mouse visual cortex. These data allowed us to directly assess the proportion of shared input between layer 4 excitatory neurons and how that changes with distance between layer 4 cells. We believe this allows for a far more accurate measure than the data that we previously used, which was based on recordings from neonatal brain slices where connectivity might differ from than in the adult.

A detailed description of these data, analyses and methods are found in the revised Methods section. We have also added an author who conducted these analyses and generated the related figures. The new analysis indicates that, at most, about 22% of the measured SF is due to shared input. Note that this is likely to be an overestimate because it assumes that every Cre-expressing neuron is a starter cell, but in our experience using similar reagents only about half of Cre+ neurons become starter cells. For our measured SF values of 40% this corresponds to 31% of input neurons being labeled (IF=0.31).

2. The authors also appropriately introduced the second point: multiple pre-synaptic structures provided by an input neuron can all be labeled if the RV passes through any one of them. I appreciate the author's attempt to suggest that this factor may have only a modest influence on the relationship between SE and IP (Figure 5). However, I am concerned that this factor has a more pronounced impact on the interpretation of the observed spatial organization of the pre-synaptic structures. Even in an extreme scenario where the RV passes through only the proximal (P) or distal (D) synapse, assuming that the pre-synaptic neurons evenly provide pre-synaptic structures along the PD axis, the result would be the comparable SE values between P and D synapses. The manuscript would benefit from a more meticulous discussion of this issue.

We thank the reviewer for pointing out that multiple contacts from a given starter neuron could theoretically influence the observed locations of labeled synapses independently from the location of the synapses across which the rabies virus had spread.

The potential magnitude of such an effect depends on the proportion of connected neurons that actually make multiple synaptic contacts onto a starter cells. As noted above, we have added new data based on EM reconstructions and analysis of nearly 77,000 adult cortical connections. These data show that the overwhelming majority of cortical connections involve only a single synaptic contact. This has resulted in far lower estimates of the impact of multiple contacts relative to estimates based on data from neonatal brain slices.

To directly address this issue we have added the following text to the Discussion:

“The same factors that we have considered for their impact on the relationships between measured SF, IF, and U are also important to consider with respect to the spatial locations of labeled synapses. This is because synaptic labeling resulting from either shared inputs or multiple synaptic contacts from a single input neuron are not the result of the spread of rabies across a synaptic contact. Our analyses show that the proportions of labeled synapses arising from these mechanisms is small relative to the variability and proportions of observed SF at distal versus proximal synapses (Figure 3G, 4F).”

We have also added new text to the Results pointing out that a portion of the synaptic locations measured could be due to indirect labeling and might not reflect the locations where the virus spread. The Results now state: “We found no significant differences between synaptic fraction at proximal regions of apical dendrites compared to the distal regions as determined by PSD95GFP and SynPhRFP puncta colocalization (49.14 ± 3.49% versus 57 ± 5.43% respectively, Wilcoxon rank-sum test, p = 0.34; n = 9 neurons across 3 mice, Figure 3G). We also compared synaptic fraction at basal dendrites versus apical dendrites and observed no significant difference (44.52 ± 3.96% versus 53.05 ± 3.99% respectively, Wilcoxon rank-sum test, p = 0.22; n = 9 neurons across 3 mice and n = 6 neurons across 3 mice, Figure 3G). Note that a fraction of the observed synaptic labeling might occur through indirect pathways rather than direct spread of rabies virus at the observed synapse (see Below). This fraction is small relative to the variability in our measured synaptic fraction at distal versus proximal or apical versus basal dendrites.”

3. Considering my major point #1 above, I believe the manuscript would gain from the inclusion of an estimation of SE pertaining to a single starter cell, which is unlikely to be 0.4. The authors could begin by analyzing the number of starter cells in their samples and then perform a regression analysis to estimate the SE of a single starter cell. Increasing the number of animal samples with a smaller number of starter cells could contribute to obtaining a more accurate estimate. Furthermore, it would be advantageous to simulate the relationships between the number of starter cells (Ns) and SE, using assumed circuit structures and the efficiency of RV spread across a single synaptic contact (referred to as unitary efficiency, U).

We thank the reviewer for this suggestion. While we agree that the assessment suggested by the reviewer could theoretically provide some useful insight, such studies would in practice present technical difficulties that would likely make the results uninterpretable. It has been well established that the efficiency of rabies tracing is highly dependent on both the expression levels of rabies glycoprotein and the numbers of rabies viral particles initially entering the starter cells (See JNS Review by Luo and Callaway). To reduce the numbers of starter cells, it would be necessary to reduce the titers of AAV helper virus used or to reduce the titer of the RV used. Either of these changes would result in reduced tracing efficiency and skew the results from regression analysis. Other alternatives, such as single cell electroporation of starter cells, would also result in changes to glycoprotein expression or rabies infection that would preclude direct comparisons across experiments with different numbers of starter cells.

We note that the equations that we have derived for relationships between U and SF (formerly SE) show that this is independent of the number of starter cells. The numbers of starter cells only effects SF with respect to common inputs, not the configuration of “assumed circuit structures”. We also expect that the reviewer’s impression about the expected SF that might be observed for a single starter cell will change in view of the new data we have provided about: (1) proportion of shared inputs between layer 4 neurons and dependence on distance; (2) distributions of Cre-expressing layer 4 neurons in nr5a1-Cre mice; and (3) numbers of synaptic contacts per connected pair.

4. Although I appreciate the authors' endeavors to estimate the relationships between SE and IP, the relevant text could benefit from a more concise writing style, aligning with the rest of the manuscript. For eLife readers with diverse backgrounds, the inclusion of schematic diagrams would be instructive.

We have replaced the text in the Results with a much more concise description and moved details to Methods. In view of the reduced complexity of the likely circuit configurations indicated by the new EM data we have opted not to add a schematic diagram.

Reviewer #2 (Recommendations for the authors):

G-deleted rabies virus has been widely used to trace monosynaptic inputs to specific populations of neurons in the mammalian brain since its first report by Callaway and colleagues in 2007. Despite its wide use, many fundamental properties of the method remain to be characterized. In this short report, Patino et al. addressed two important questions: (1) what the fraction of synapses does the rabies virus cross from a starter cell? (2) does this vary for synapses located at different distances to the soma? Answers to these questions should help researchers to interpret their data and thus will be of general interest to the neural circuit community.

To answer these questions, the authors produced a new rabies virus that expresses both a presynaptic marker tagged with RFP and a postsynaptic marker (for excitatory neurons) tagged with GFP, and simply counted within starter cells the fraction of GFP+ puncta that have an adjacent RFP punctum. Since most RFP puncta are a result of trans-synaptic transfer of rabies virus (the authors did a nice control of skipping G in one experiment), the authors can "quantify efficiency of spread at the synaptic level". The authors found about 35-40% of GFP+ puncta have RFP puncta next to them, and interestingly this number does not change whether these quantifications were performed close to the cell body or far from the cell body (thus answering question #2).

I appreciate the length the authors went to address my critiques of the original manuscript on the issue of distinguishing (1) fraction of synapses that are co-labeled in the starter cell, (2) fraction of input neurons that are labeled by rabies virus, and (3) average efficiency of rabies virus crossing an individual synapses. This will clarify to the readers what the authors have done and what kind of conclusions they can take away. However, I have several remaining critiques to the above issue that should be addressed by further textual revision.

1) I don't think the terms the authors used to describe the above three terms are the clearest. The authors used "synaptic efficiency" to describe (1), but it can easily be confused with (3). I recommend synapse proportion (SP) to parallel with input proportion (IP), or perhaps a bit better synapse fraction (SF) vs. input fraction (IF). Then (3) can be changed to "Unitary synapse efficiency (U)". These terms may be widely adopted in future publications on rabies tracing, so careful consideration is warranted.

We thank the reviewer for this excellent suggestion. We have adopted the suggested nomenclature.

2) The textual description in the Results section "Relationships between…" is often repetitive and needs to be streamlined. Specifically, I don't like the description of the results in the first paragraph ahead of all the analyses (and the authors have done these repetitively). For example, most of the first paragraph should be trimmed-just give the definition of the terms and move on to the actual "formal analysis." Summarize the results at the end.

We have replaced the text in the Results section with a more concise description and reserved details for the Methods section.

3) The results of the formal analysis is based on the cortical circuit. It's important to emphasize that in circuits with denser connectivity, the relationship will be very different. As such, when these properties are discussed in other parts of the manuscript (abstract, introduction, discussion, etc.), please make sure to include the qualification phrase that these inferences were made from cortical circuits (e.g., last paragraph of Introduction).

We have made the suggested edits in all relevant cases that we have found. The last paragraph of the Introduction now explicitly refers to “our results in mouse visual cortex.” And in the discussion: “These equations can be used to assess circuit configurations that are likely to be different in other brain areas than for the cortical circuits on which we have focused.”

Reviewer #3 (Recommendations for the authors):

1) Prior studies have quantified the proportion of PSD95 labeled postsynaptic densities that are opposed by presynaptic terminals that are also labeled with various presynaptic markers, including synaptophysin and VGLUT1 (Neuron. 2010 Nov 18; 68(4): 639-653. doi: 10.1016/j.neuron.2010.09.024). Essentially every PSD95 labeled terminal has a labeled presynaptic contact. Synaptophysin is a reliable marker of presynaptic terminals but VGLUT1 is not. We have modified the text in the introduction which now reads: "We designed a new genetically modified rabies virus that labels presynaptic terminals with synaptophysin-RFP (SynPhRFP) and excitatory postsynaptic densities with postsynaptic density-95-GFP (PSD95GFP). Because more than 99% of PSD95 postsynaptic puncta co-localize with a presynaptic terminal (Micheva et al., 2010), this construct allows us to quantify the proportion of excitatory synapses on a starter cell that have their corresponding input neuron labeled with rabies virus (defined here as synaptic efficiency [SE])."

Comment: I understand that "endogenous" PSD95 is a reliable postsynaptic marker in vivo. However, considering this is a new reagent and rabies viruses highly express transgenes, it is fair to experimentally show the colocalization rate of PSD95GFP expressed from rabies viruses and endogenous presynaptic markers (e.g. synaptophysin).

The text quoted above was in response to a previous comment from this reviewer which stated: “The authors assume that all PSD95GFP puncta receive presynaptic contacts and the total number of PSD95GFP represents the total number of synapses on starter cells. This should be validated by examining the proportion of PSD95GFP puncta that associate with endogenous excitatory presynaptic markers such as VGLUT1.”

Our response to that comment was intended to clarify that VGLUT1 would not be a good choice, since it is not a reliable marker of presynaptic terminals. The text that we added to the Introduction emphasizes the fact that it has already been established that essentially all PSD95 puncta do in fact receive presynaptic contacts. Furthermore, we show in Figure 1 that the PSD95GFP puncta on rabies infected neurons colocalize with PSD95 antibody staining, precluding the possibility that the PSD95GFP from the rabies genome might somehow fail to label some PSDs. The reviewer’s new comments seem to be suggesting that perhaps through some unknown mechanism, the over-expression of PSD95GFP will result in the formation of new puncta that do not have a presynaptic input. Although this seems very unlikely, if this were to happen it would result in an underestimate of the rate at which rabies virus spreads. We have added new text to the discussion that states: “It should also be noted that the locations and numbers of postsynaptic sites was assessed based on expression of PSD-95-GFP from the rabies virus. If this overexpression were to somehow result in new puncta that do not have a presynaptic input this could result in an underestimate of proportion of actual PSDs labeled.”

2) We thank the reviewer for this excellent point. We have been very careful not to equate efficiency of spread to labeling efficiency. As described in the above response to a similar comment from reviewer #1 we have added a new Results section, "Relationships between synaptic efficiency, input proportion, and unitary efficiency". Here the value that we formally define as unitary efficiency (U) is equivalent to what the reviewer terms "efficiency of spread", while our measured values of synaptic efficiency (SE) correspond to what the reviewer describes as "labeling efficiency". We show that there are complex relationships between the proportion of labeled presynaptic terminals (SE), efficiency of spread across individual synapses (U), the distributions of numbers of synaptic contacts between connected neuron pairs, and the proportion of input neurons labeled (IP).

Comment: I appreciate the authors' effort to estimate IP and U from SE and that the authors clearly discriminate SE and the spread efficiency in the text. The results are informative to some extent but not definitive because they rely on several assumptions such as "U is identical across synapses" and "there are two populations of input neurons that make the distinct numbers of presynaptic contacts per neuron onto starter cells", which are likely different from real situations. I am afraid that this analysis is not strong enough to claim the utility of the novel rabies vectors.

We would like to once again reinforce the fact that our primary aim in these experiments was not to determine what proportion of synapses are labeled but rather to determine whether the probability of labeling synapses depended on dendritic location or postsynaptic cell type. We show that these factors do not have a large impact. The results show that we were able to compare labeling at distal versus proximal sites and onto different cell types, as intended.

We have also included extensive discussion about the expected impact of synapse size and the likelihood that this is related to the efficiency of spread across a single synapse. We do not claim the “U is identical across synapses”, but this is a simplifying assumption in our model that allows the impact of the distributions of synapses per connection to be evaluated independently from potential complications related to synapse size.

We have also provided new data on the distribution of the numbers of synapses formed between connected excitatory cortical neurons. These data show that 92% of the connections involve only one synapse and that 99% of connections have only 1 or 2 synapses. Thus, 99% of connections can be accounted for by a simulation that considers just two populations of input neurons, one population with a single synapse (92% of connections) and a second population with two synapses per connection. This model is considered in Figures 6C and D. In Figures 6A and B, we also consider an extreme model (one population with one synapse and the other with three synapses) which shows that the impact of these parameters remains small even if the population of connections with three synapses were 10-fold larger than observed experimentally.

3) We would like to thank the reviewer for these thought-provoking questions. This is a rather disparate set of comments each of which relates to very different properties of either cortical circuit organization or rabies tracing. Some of these are known for limited aspects of cortical circuits while others remain unknown. With respect to rabies virus spread, some of these are known or can be inferred from published observations.

Below we further address each of the reviewer's specific questions.

Comment: There seems to be a misunderstanding of my previous review comment. I just listed examples of key questions about neural circuit organization and the mechanisms for the rabies virus spread, whose answers should bring important biological insights. I did not mean to request the authors to address all these questions.

4) We disagree with the stated characterization of our results: "Addressing these questions should be important for understanding the mechanism underlying rabies virus spread as well as more detailed connection properties of neural circuits. Unfortunately, the present study falls short of providing biological implications beyond "the labeling efficiency at synaptic levels". First, it was not a goal of our study to understand the mechanisms underlying the spread of rabies virus nor did we intend to reveal detailed connection properties of neural circuits. As stated in our Introduction, we were interested in identifying factors that influence the efficiency of transynaptic spread of rabies and we designed our experiments to address two such factors. These are the distance of synaptic contacts from the cell body and the type of postsynaptic neuron. Our data show that the proportions of labeled synaptic contacts are similar regardless of distance from the cell body and regardless of the postsynaptic cell types tested. These are quite important factors and given that the overall efficiency of spread can easily be increased or decreased by changing glycoprotein expression levels, these measures are more important than the overall measured efficiency in our experiments.

Comment: The title of the manuscript is now "Detection of monosynaptic rabies spread at synaptic resolution reveals factors influencing tracing efficiency". However, major achievements/findings of this study are that new rabies virus vectors that enable visualization of input synapses onto starter cells and estimation of the fraction of labeled presynaptic contacts are developed and that the proportions of labeled synaptic contacts are similar regardless of distance from the cell body and regardless of the postsynaptic cell types tested, as the authors state above. I am afraid that the authors do not provide "factors that influence (alter) tracing efficiency", which makes me feel that the title does not match with the data. If the scope of the manuscript is neither to understand the mechanisms underlying the spread of rabies virus nor to reveal detailed connection properties of neural circuits, can the authors clarify what kind of important questions could be addressed using this novel tool (the utility of the tool)?

We have modified the title, which is now: “Postsynaptic cell type and synaptic distance do not determine efficiency of monosynaptic rabies virus spread measured at synaptic resolution”.

There appears to have been some confusion about what exactly are the “factors” that we were referring to in the previous title: “Detection of monosynaptic rabies spread at synaptic resolution reveals factors influencing tracing efficiency.” The reviewer stated: “I am afraid that the authors do not provide "factors that influence (alter) tracing efficiency", But in the immediately preceding sentence the reviewer lists exactly the same factors that we were referring to in the former title. “However, major achievements/findings of this study are that new rabies virus vectors that enable visualization of input synapses onto starter cells and estimation of the fraction of labeled presynaptic contacts are developed and that the proportions of labeled synaptic contacts are similar regardless of distance from the cell body and regardless of the postsynaptic cell types tested …”

We believe that the new title now clearly identifies the “factors” – postsynaptic cell type and synaptic distance – which our experiments have evaluated with respect to their influences on the efficiency of rabies labeling.

https://doi.org/10.7554/eLife.89297.sa2

Article and author information

Author details

  1. Maribel Patiño

    1. Systems Neurobiology Laboratories, The Salk Institute for Biological Studies, La Jolla, United States
    2. Neuroscience Graduate Program, University of California, San Diego, La Jolla, United States
    3. Medical Scientist Training Program, University of California, San Diego, La Jolla, United States
    Contribution
    Conceptualization, Formal analysis, Supervision, Funding acquisition, Investigation, Methodology, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0820-3315
  2. Willian N Lagos

    Systems Neurobiology Laboratories, The Salk Institute for Biological Studies, La Jolla, United States
    Contribution
    Investigation, Writing – review and editing
    Competing interests
    No competing interests declared
  3. Neelakshi S Patne

    Systems Neurobiology Laboratories, The Salk Institute for Biological Studies, La Jolla, United States
    Contribution
    Investigation, Writing – review and editing
    Competing interests
    No competing interests declared
  4. Paula A Miyazaki

    Systems Neurobiology Laboratories, The Salk Institute for Biological Studies, La Jolla, United States
    Contribution
    Investigation, Writing – review and editing
    Competing interests
    No competing interests declared
  5. Sai Krishna Bhamidipati

    Systems Neurobiology Laboratories, The Salk Institute for Biological Studies, La Jolla, United States
    Contribution
    Formal analysis, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3498-3730
  6. Forrest Collman

    Allen Institute for Brain Science, Seattle, United States
    Contribution
    Resources, Formal analysis, Investigation, Visualization, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0280-7022
  7. Edward M Callaway

    Systems Neurobiology Laboratories, The Salk Institute for Biological Studies, La Jolla, United States
    Contribution
    Conceptualization, Resources, Formal analysis, Funding acquisition, Methodology, Project administration, Writing – review and editing
    For correspondence
    callaway@salk.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6366-5267

Funding

National Science Foundation (DBI-1707261)

  • Edward M Callaway

National Eye Institute (R01 EY022577)

  • Edward M Callaway

Paul and Daisy Soros Fellowships for New Americans

  • Maribel Patiño

National Institute of General Medical Sciences (T32 GM007198)

  • Maribel Patiño

National Science Foundation (2014862)

  • Forrest Collman

National Institute of Neurological Disorders and Stroke (U24NS120053)

  • Forrest Collman

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank all Callaway Lab members for discussion. We thank Helen Wang and Oyshi Dey for generating and sharing data related to Nr5a1 cell counts. We also thank the Salk Viral Vector and Biophotonics Core staff members. This work was supported by NSF grant DBI-1707261 and NIH Diversity Supplement to R01 EY022577 (EMC). MP was supported by the Paul and Daisy Soros Fellowship for New Americans and NIH T32 GM007198. Forrest Collman acknowledges support from NSF NeuroNex 2 award 2014862 and the National Institute of Neurological Disorders and Stroke of the National Institutes of Health under Award Number U24NS120053.

Ethics

All work involving live animals was approved by the Salk Institute Animal Care and Use Committee (IACUC). All the animals were handled according to the approved IACUC protocol 13-00005.

Senior Editor

  1. Lu Chen, Stanford University, United States

Reviewing Editor

  1. Liqun Luo, Howard Hughes Medical Institute, Stanford University, United States

Version history

  1. Preprint posted: September 1, 2022 (view preprint)
  2. Received: May 11, 2023
  3. Accepted: November 19, 2023
  4. Version of Record published: December 14, 2023 (version 1)

Copyright

© 2023, Patiño et al.

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

  • 473
    Page views
  • 57
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

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)

  1. Maribel Patiño
  2. Willian N Lagos
  3. Neelakshi S Patne
  4. Paula A Miyazaki
  5. Sai Krishna Bhamidipati
  6. Forrest Collman
  7. Edward M Callaway
(2023)
Postsynaptic cell type and synaptic distance do not determine efficiency of monosynaptic rabies virus spread measured at synaptic resolution
eLife 12:e89297.
https://doi.org/10.7554/eLife.89297

Share this article

https://doi.org/10.7554/eLife.89297

Further reading

    1. Neuroscience
    Eyal Y Kimchi, Anthony Burgos-Robles ... Kay M Tye
    Research Article

    Basal forebrain cholinergic neurons modulate how organisms process and respond to environmental stimuli through impacts on arousal, attention, and memory. It is unknown, however, whether basal forebrain cholinergic neurons are directly involved in conditioned behavior, independent of secondary roles in the processing of external stimuli. Using fluorescent imaging, we found that cholinergic neurons are active during behavioral responding for a reward – even prior to reward delivery and in the absence of discrete stimuli. Photostimulation of basal forebrain cholinergic neurons, or their terminals in the basolateral amygdala (BLA), selectively promoted conditioned responding (licking), but not unconditioned behavior nor innate motor outputs. In vivo electrophysiological recordings during cholinergic photostimulation revealed reward-contingency-dependent suppression of BLA neural activity, but not prefrontal cortex. Finally, ex vivo experiments demonstrated that photostimulation of cholinergic terminals suppressed BLA projection neuron activity via monosynaptic muscarinic receptor signaling, while also facilitating firing in BLA GABAergic interneurons. Taken together, we show that the neural and behavioral effects of basal forebrain cholinergic activation are modulated by reward contingency in a target-specific manner.

    1. Neuroscience
    Olgerta Asko, Alejandro Omar Blenkmann ... Anne-Kristin Solbakk
    Research Article Updated

    Orbitofrontal cortex (OFC) is classically linked to inhibitory control, emotion regulation, and reward processing. Recent perspectives propose that the OFC also generates predictions about perceptual events, actions, and their outcomes. We tested the role of the OFC in detecting violations of prediction at two levels of abstraction (i.e., hierarchical predictive processing) by studying the event-related potentials (ERPs) of patients with focal OFC lesions (n = 12) and healthy controls (n = 14) while they detected deviant sequences of tones in a local–global paradigm. The structural regularities of the tones were controlled at two hierarchical levels by rules defined at a local (i.e., between tones within sequences) and at a global (i.e., between sequences) level. In OFC patients, ERPs elicited by standard tones were unaffected at both local and global levels compared to controls. However, patients showed an attenuated mismatch negativity (MMN) and P3a to local prediction violation, as well as a diminished MMN followed by a delayed P3a to the combined local and global level prediction violation. The subsequent P3b component to conditions involving violations of prediction at the level of global rules was preserved in the OFC group. Comparable effects were absent in patients with lesions restricted to the lateral PFC, which lends a degree of anatomical specificity to the altered predictive processing resulting from OFC lesion. Overall, the altered magnitudes and time courses of MMN/P3a responses after lesions to the OFC indicate that the neural correlates of detection of auditory regularity violation are impacted at two hierarchical levels of rule abstraction.