Author Response: Rabies virus-based barcoded neuroanatomy resolved by single-cell RNA and in situ sequencing Article Swipe

Full text Figures and data Side by side Abstract eLife assessment eLife digest Introduction Results Discussion Materials and methods Appendix 1 Appendix 2 Data availability References Peer review Author response Article and author information Metrics Abstract Mapping the connectivity of diverse neuronal types provides the foundation for understanding the structure and function of neural circuits. High-throughput and low-cost neuroanatomical techniques based on RNA barcode sequencing have the potential to map circuits at cellular resolution and a brain-wide scale, but existing Sindbis virus-based techniques can only map long-range projections using anterograde tracing approaches. Rabies virus can complement anterograde tracing approaches by enabling either retrograde labeling of projection neurons or monosynaptic tracing of direct inputs to genetically targeted postsynaptic neurons. However, barcoded rabies virus has so far been only used to map non-neuronal cellular interactions in vivo and synaptic connectivity of cultured neurons. Here we combine barcoded rabies virus with single-cell and in situ sequencing to perform retrograde labeling and transsynaptic labeling in the mouse brain. We sequenced 96 retrogradely labeled cells and 295 transsynaptically labeled cells using single-cell RNA-seq, and 4130 retrogradely labeled cells and 2914 transsynaptically labeled cells in situ. We found that the transcriptomic identities of rabies virus-infected cells can be robustly identified using both single-cell RNA-seq and in situ sequencing. By associating gene expression with connectivity inferred from barcode sequencing, we distinguished long-range projecting cortical cell types from multiple cortical areas and identified cell types with converging or diverging synaptic connectivity. Combining in situ sequencing with barcoded rabies virus complements existing sequencing-based neuroanatomical techniques and provides a potential path for mapping synaptic connectivity of neuronal types at scale. eLife assessment This study presents an important tool for tracking the connectivity of neurons in mouse and potentially other mammals using a combined approach of barcoded rabies virus libraries and spatial transcriptomics. The data supporting the technique are convincing, the validation against known anatomical knowledge is rigorous, and the authors advance the techniques by combing them in vivo. Overall, this is a very good paper describing a technique for tracking neural circuits. https://doi.org/10.7554/eLife.87866.3.sa0 About eLife assessments eLife digest In the brain, messages are relayed from one cell to the next through intricate networks of axons and dendrites that physically interact at junctions known as synapses. Mapping out this synaptic connectivity – that is, exactly which neurons are connected via synapses – remains a major challenge. Monosynaptic tracing is a powerful approach that allows neuroscientists to explore neural networks by harnessing viruses which spread between neurons via synapses. and in particular the rabies virus. This pathogen travels exclusively from ‘postsynaptic’ to ‘presynaptic’ neurons – from the cell that receives a message at a synapse, back to the one that sends it. A modified variant of the rabies virus can therefore be used to reveal the presynaptic cells connecting to a population of neurons in which it has been originally introduced. However, this method does not allow scientists to identify the exact postsynaptic neuron that each presynaptic cell is connected to. One way to bypass this issue is to combine monosynaptic tracing with RNA barcoding to create distinct versions of the modified rabies virus, which are then introduced into separate populations of neurons. Tracking the spread of each version allows neuroscientists to spot exactly which presynaptic cells signal to each postsynaptic neuron. So far, this approach has been used to examine synaptic connectivity in neurons grown in the laboratory, but it remains difficult to apply it to neurons in the brain. In response, Zhang, Jin et al. aimed to demonstrate how monosynaptic tracing that relies on barcoded rabies viruses could be used to dissect neural networks in the mouse brain. First, they confirmed that it was possible to accurately detect which version of the virus had spread to presynaptic neurons using both in situ and single-cell RNA sequencing. Next, they described how this information could be analysed to build models of potential neural networks, and what type of additional experiments are required for this work. Finally, they used the approach to identify neurons that tend to connect to the same postsynaptic cells and then investigated what these have in common, showing how the technique enables a finer understanding of neural circuits. Overall, the work by Zhang, Jin et al. provides a comprehensive review of the requirements and limitations associated with monosynaptic tracing experiments based on barcoded rabies viruses, as well as how the approach could be optimized in the future. This information will be of broad interest to scientists interested in mapping neural networks in the brain. Introduction The connectivity of diverse types of neurons constrains the flow and processing of information in neural circuits. Neuroanatomical techniques based on microscopy rely on sparse labeling combined with optical tracing to achieve single-cell resolution (Winnubst et al., 2019; Peng et al., 2021; Gao et al., 2022). However, these approaches are usually labor-intensive and difficult to scale up. Furthermore, because many of these approaches rely on specialized imaging platforms, they are difficult to combine with transcriptomic interrogation for defining cell types. Thus, developing high-throughput neuroanatomical techniques that can map neuronal connectivity and associate connectivity with gene expression at single-cell resolution could help generate new insights into the organization of neural circuits that are difficult to gain using conventional neuroanatomical techniques. RNA barcoding-based neuroanatomical techniques are a new approach that dramatically improves the throughput of single-neuron projection mapping (Kebschull et al., 2016; Chen et al., 2019; Sun et al., 2021). In these techniques, Sindbis virus is used to express random RNA sequences, or RNA ‘barcodes’, to uniquely label each neuron. The barcoded virus RNAs replicate and distribute throughout the somata and axons; therefore, matching barcodes in the axons to those in the somata reveal the axonal projections of individual neurons. Because many barcode molecules can be sequenced in parallel, barcoding-based neuroanatomical approaches can determine the long-range projections of tens of thousands of neurons in single animals. These barcoding-based techniques have been applied to various brain regions (Kebschull et al., 2016; Han et al., 2018; Chen et al., 2019; Gergues et al., 2020; Klingler et al., 2021; Mathis et al., 2021; Muñoz-Castañeda et al., 2021; Sun et al., 2021; Chen et al., 2022) and have validated and extended findings by previous studies using imaging-based tracing techniques. Despite the transformative power of barcoding-based neuroanatomical techniques, the Sindbis virus-based approaches can only reveal information about the axons (i.e. projection mapping) of labeled neurons with cell bodies at the injection sites, because these vectors do not spread between cells. These approaches thus cannot obtain information about the neurons’ synaptic partners. A potentially powerful extension of this approach that could allow mapping of connections, instead of just projections, would be to use rabies virus. Rabies virus naturally spreads between synaptically-connected neurons in the retrograde direction. In particular, the use of deletion-mutant rabies viruses (e.g. ‘∆G’ viruses, with their glycoprotein gene G deleted) to perform transsynaptic labeling has become common in neuroscience, as it allows ‘monosynaptic tracing’ (Wickersham et al., 2007b; Wall et al., 2010), or the viral labeling of neurons directly presynaptic to a targeted starting population of neurons. In the standard monosynaptic tracing paradigm, a targeted starting population of neurons is first transduced by ‘helper’ adeno-associated viruses (AAVs), causing them to express the rabies virus gene G (to complement the G-deleted recombinant rabies) and an avian cell surface protein, TVA. TVA is the receptor for an avian retrovirus (avian sarcoma and leukosis virus subgroup A, or ASLV-A) that is unable to infect mammalian cells (Young et al., 1993). Subsequently, a ∆G rabies virus that is packaged with the ASLV-A envelope protein (EnvA) is injected at the same location to selectively infect the TVA-expressing cells. Due to the expression of the rabies viral gene G in trans, the ∆G rabies virus replicates and spreads from these ‘source cells’ to neurons directly presynaptic to them. In a second approach, ∆G rabies virus that is packaged with its own glycoprotein can be used to retrogradely label neurons with projections to a target area (Wickersham et al., 2007a). This approach, in contrast to the monosynaptic tracing paradigm, reveals only the axonal projections of neurons, but not their synaptic connectivity. RNA barcoding can be combined with both rabies virus-based tracing approaches to drastically improve the throughput at which projections or connectivity can be interrogated at cellular resolution. Retrograde labeling from multiple locations using several preparations of glycoprotein-deleted rabies virus (RVΔG), each carrying different barcodes, would allow multiplexed retrograde tracing, that is the association of neurons with their projections to many different injection sites within single brains (Figure 1A). This approach is conceptually similar to that used in a recent study in which barcoded AAV was used to perform multiplexed retrograde labeling (Zhao et al., 2022), but rabies virus potentially has different tropism that can broaden the use of barcoded retrograde labeling approaches (Chatterjee et al., 2018). Multiplexed retrograde tracing approaches complement anterograde tracing techniques: Because only neurons that are labeled at the injection sites are mapped in anterograde tracing, it is difficult to precisely estimate how the density of projection neurons varies across large brain regions; in contrast, retrograde labeling approaches label neurons with projections to the injection site regardless of which brain region they are in. As anatomical borders between brain areas (e.g. cortical areas) usually correspond to distinct changes in cytoarchitecture (Brodmann, 1909; Vogt and Vogt, 1919; Von Bonin, 1947) and transcriptomically defined neuronal types (Chen et al., 2023), retrograde labeling is particularly valuable for understanding how long-range connectivity is associated with cell types and anatomical boundaries. Figure 1 Download asset Open asset Models of multiplexed retrograde labeling and monosynaptic tracing using barcoded rabies virus. (A) In multiplexed retrograde labeling, rabies viruses carrying different barcodes are injected into different brain regions, and retrogradely labeled neurons can be distinguished based on the barcodes they carry. Numbers and colors indicate different barcodes injected into each area. (B) In multiplexed monosynaptic tracing, potential source cells are first labeled by helper AAV viruses expressing TVA and rabies glycoprotein (yellow cells, left). These cells can be infected by barcoded rabies virus (barcodes are indicated by numbers, middle). These source cells can then pass the barcodes to presynaptic neurons (numbered gray cells, right). Both rabies barcodes and endogenous mRNAs can be read out to infer cell type connectivity. However, if multiple source cells share the same barcode, they may obscure single-cell connectivity mapping and must be filtered out (crossed out cells, right). Alternatively, combining the transsynaptic labeling approach with RNA barcoding could allow sequencing-based readouts of synaptically-connected networks of neurons, providing information about the synaptic partners of neurons, rather than only their patterns of axonal projections. Specifically, simply using the standard monosynaptic tracing system but using a barcoded pool of ΔG rabies virus could allow mapping of synaptic connectivity between many cells at cellular resolution (multiplexed transsynaptic labeling; Figure 1B). Although other techniques, such as electron microscopy (Bae et al., 2021; Androvic et al., 2022; Zheng et al., 2022; Schneider-Mizell et al., 2023) and multi-patch experiments (Campagnola et al., 2022), can also read out synaptic connectivity, these approaches are difficult to scale up to many neurons across large areas of the brain. Thus, achieving barcoded transsynaptic tracing using rabies virus could potentially transform the scale at which synaptic connectivity of transcriptomic types of neurons can be interrogated at cellular resolution. Using barcoded rabies virus-based approaches to associate connectivity or projections of neurons with their gene expression requires overcoming several technical and conceptual challenges. First, rabies virus alters the gene expression of infected cells (Patiño et al., 2022), which may obscure the transcriptomic signatures of neurons. It is unclear how changes in gene expression affect the ability to distinguish fine-grained transcriptomic types and whether cell typing is still possible using spatial transcriptomic approaches, such as in situ sequencing, that interrogate a targeted panel of genes. Second, conventional single-cell sequencing is costly and labor intensive. Thus, sequencing throughput will likely limit the multiplexing advantages associated with barcoding strategies. Third, in a barcoded transsynaptic labeling experiment, synaptic connectivity between individual source cells and their presynaptic partners can only be inferred from networks of barcode-sharing neurons with a single source cell (Figure 1B). A barcode, however, may be found in multiple or no source cells, which would prohibit connectivity mapping at the single-neuron level. Finally, accurately assessing the distribution of barcodes in a transsynaptic labeling experiment requires sampling most barcodes and/or source cells in an experiment. This is difficult to achieve using single-cell sequencing approaches that require tissue dissociation, because loss of cells during dissociation is inevitable. These challenges are fundamental barriers for using barcoded transsynaptic labeling to infer synaptic connectivity at cellular resolution regardless of the choice of virus. Solving these problems will not only allow barcoded rabies virus-based connectivity mapping, but also provide a foundation for potential future techniques based on a wide range of transsynaptic viruses, such as herpes simplex virus (Ugolini et al., 1989; Xiong et al., 2022; Fischer et al., 2023), pseudorabies virus (Martin and Dolivo, 1983), vesicular stomatitis virus (Beier et al., 2013), and yellow fever virus (Li et al., 2021). Because in situ sequencing (Ke et al., 2013; Chen et al., 2023) can read out both endogenous mRNAs and random RNA barcodes (Chen et al., 2019; Sun et al., 2021) with high throughput and low cost and does not rely on tissue dissociation (Chen et al., 2019; Sun et al., 2021), it is uniquely suited to overcome the challenges associated with barcoded rabies tracing. Here we adapt single-cell RNA-seq and an in situ sequencing approach based on Barcoded Anatomy Resolved by Sequencing (BARseq) (Chen et al., 2019; Sun et al., 2021) to map connectivity of transcriptomic types of neurons using barcoded rabies-based retrograde labeling and transsynaptic labeling. We examine the effect of rabies virus infection on the gene expression and clearly identify transcriptomic identities of rabies-labeled neurons. We then explore conceptually how connectivity can be inferred from rabies barcodes in a trans-synaptic labeling experiment. Finally, we perform scRNA-seq and in situ sequencing on neurons that are transsynaptically labeled by rabies virus to identify pairs of cell types that show preferences to synapse onto the same post-synaptic neurons. Results Identifying transcriptomic types of retrogradely labeled neurons by barcoded rabies virus To assess whether we can robustly identify transcriptomically defined neuronal types in a multiplexed retrograde labeling experiment, we performed two-plex retrograde labeling using two libraries of barcoded rabies virus coated with the native rabies glycoprotein (Figure 2A; Materials and methods M4). In addition to encoding the red fluorescent protein mCherry (Shaner et al., 2004), the two viral libraries contained 20-nt barcode cassettes located in the 3’UTR of the rabies nucleoprotein mRNA, which allowed high-level expression of the barcodes (Conzelmann, 1998). In addition, one of the libraries contained a 22-nt exogenous sequence (the 10 x Genomics ‘Chromium capture sequence 2’, referred to below as CCS) next to the barcode cassette. We sequenced the two barcoded libraries using Illumina next-generation sequencing and identified at least 8552 and 13,211 barcodes, respectively (see Materials and methods M8 and Materials and methods M9; Figure 2B). We did not find barcodes that were present in both libraries. Thus, the two libraries could be distinguished both by their barcode sequences and the presence or absence of the CCS. Figure 2 with 3 supplements see all Download asset Open asset High-quality cell typing of rabies virus-barcoded neurons using single-cell RNA-seq. (A) Illustration of the design of barcoded rabies virus libraries. (B) Barcode distribution in the CCS and non-CCS libraries. The y axis indicates the count of barcodes, which are sorted in descending order. The rank of the barcode is shown on the x axis. (C) An outline of the experiments. The two libraries are injected into VISal and LGd. After a week, VISp is dissected, and the mCherry-expressing cells are FACS-enriched and processed for single-cell RNA-seq. (D) Histogram showing the cluster mapping confidence of rabies-labeled cells from this study and AAV-labeled cells from Graybuck et al., 2021. (E) The expression of select marker gene in non-infected cells from Tasic et al., 2018 and in rabies-infected cells of matching cell types. Dot size indicates the proportion of cells with non-zero marker expression, and colors indicate relative gene expression levels scaled per row. (F) UMAP plot of gene expression patterns of cells infected with rabies virus (black) overlaid on non-infected cells from Tasic et al., 2018. The non-infected cells are color-coded by cluster identities and cluster names are indicated. (G) The expression of rabies-encoded genes in the sequenced cells. Columns indicate cells and rows indicate genes. Colors indicate the log transformed count for each rabies-encoded gene after scaling by the sum of all reads that mapped to viral constructs multiplied by 10,000. The bar on top indicates donor animals. (H) Example barcode sequences in three sequenced cells. Letter heights indicate probabilities at each position. Gray boxes indicate the barcode region. (I) Distribution of cell types of retrogradely labeled cells. Colors indicate cell types and match those in (F), and dot size indicates the number of cells. VISp, primary visual cortex; VISal, anterolateral visual cortex; LGd, dorsal lateral geniculate nucleus. We injected the two libraries into the dorsal lateral geniculate nucleus (LGd) and anterolateral visual cortex (VISal), respectively, in two animals (Figure 2C). After 7 days, we dissected out the primary visual cortex (VISp), dissociated the neurons, and FACS-isolated 48 mCherry-expressing neurons from each animal for scRNA-seq using SMART-seq v4 (see Materials and methods M6). scRNA-seq data from rabies virus-infected cells had comparable quality to non-infected cells (Figure 2—figure supplement 1A). Of 96 mCherry-expressing cells sequenced, 75 neurons were of high quality (with >100,000 total reads,>1,000 detected genes, and odds ratios of GC dinucleotides <0.5; see Materials and methods M7 for details). We then mapped each neuron to reference scRNA-seq data (Tasic et al., 2018) by comparing a cell’s marker gene expression with the marker gene expression of the reference cell types (Gouwens et al., 2020; Graybuck et al., 2021) (Materials and methods M7). Briefly, we first selected marker genes that distinguished each cluster in the reference taxonomy tree, and then performed 100 rounds of correlation analysis between a given single cell transcriptome to be mapped and the reference taxonomy tree using 80% of the marker panel selected at random in each round. Cells that were assigned to the same cluster in ≥70 of 100 rounds (mapping confidence >0.7) and with a mapping correlation >0.6 were considered to be mapped to the cluster with high quality. A total of 54 cells were mapped to the reference taxonomy tree with high quality [mapping confidence >0.7 (5 cells removed) and mapping correlation >0.6 (an additional 16 cells removed); Figure 2D; Figure 2—figure supplement 1B; Table 1; Materials and methods M7]; these cells were used for downstream analysis. Most cells that were removed by the mapping correlation thresholds were either ‘Microglia Siglech’ or ‘PVM Mrc1’, which tend to have lower gene counts compared to neurons (Tasic et al., 2018). Consistent with previous studies (Prosniak et al., 2001; Zhao et al., 2011; Huang and Sabatini, 2020; Patiño et al., 2022), some immune response-related genes were up-regulated in rabies virus-infected cells compared to non-infected cells (Figure 2—figure supplement 2A, B). We also noticed that some inhibitory cell types showed higher expression of activity-related genes, such as Baz1, Fosl2, and Jun compared to non-infected cells (Figure 2—figure supplement 3). the expression patterns of cell type were comparable to the reference scRNA-seq (Figure and Thus, transcriptomic types of rabies infected neurons can be robustly read out using Table 1 of cells in the transsynaptic tracing and retrograde tracing experiments. genes and GC confidence correlation Rabies were detected in all cells, the 54 mapped cells (Figure Consistent with their known expression levels (Conzelmann, for the rabies and were than the for mCherry the rabies and the large Consistent with the of rabies out of 54 neurons had sequencing reads the barcode region of the nucleoprotein (Figure Table the neurons, 48 neurons one barcode and two cells a Of these barcodes, had the CCS sequence at the region. The barcode sequences were with known barcodes in the barcode all barcodes known barcodes in the CCS and all barcodes that did not a CCS sequence known barcodes in the non-CCS In the neurons and neurons also known as to the but not the neurons to the cortex and the but not the and et al., 2019; Peng et al., 2021). Consistent with the known connectivity of cortical cell types and Tasic et al., neurons, 3 neurons, and 1 neurons to the LGd, and not in contrast, neurons and 1 neuron to VISal, but not (Figure We also a number of inhibitory neurons labeled by the VISal but not the injection (Figure This labeling was because inhibitory neurons at the of VISp and VISal could potentially be labeled by the VISal injection and dissected for single-cell RNA-seq. on these we estimate that the in the transcriptomic types of retrogradely labeled neurons is between and (see Materials and methods Thus, multiplexed retrograde tracing using barcoded rabies virus known projection patterns of cortical neuronal types. In situ sequencing transcriptomic types of neurons scRNA-seq provides a of the transcriptomic of rabies-labeled neurons, but this approach has several First, a large of neurons is during single-cell dissociation to sequencing. Because neuronal types in their during dissociation, this in the of neurons that are sequenced (Tasic et al., 2016; Tasic et al., 2018). Second, a of tissue dissociation is that the locations of neurons are which can obscure potential spatial organization of neuronal connectivity. Finally, the low throughput and high cost of scRNA-seq limit the scale at which the retrogradely labeled neurons can be To overcome these we next used in situ sequencing to both gene expression and rabies barcodes in retrogradely labeled neurons. in situ sequencing approach is based on a high-throughput technique that can determine both long-range projections of neurons and their gene expression by in situ sequencing of both endogenous gene expression and barcodes in the of Sindbis virus (Chen et al., 2019; Sun et al., 2021). We have shown that in situ sequencing can distinguish transcriptomic types of cortical neurons with high transcriptomic resolution in non-infected mouse brains (Chen et al., To adapt to rabies barcodes (Figure we used that were to the region that was to the barcodes on the to the rabies In addition to the for the rabies barcodes, we also used random to the endogenous After in we used to target both the regions of the barcodes and marker genes which we used to distinguish cortical neuron types (Chen et al., We then the all and performed as in standard experiments et al., 2021). We performed rounds of Illumina sequencing in situ to read out cell type marker genes, one of to detect three high-level cell type and and rounds of sequencing in situ to read out the rabies We the cells using et al., 2021) and gene using (Chen et al., 2018; Chen et al., to obtain single-cell gene expression, as described (Chen et al., To barcode we in the first of barcode sequencing using both an and a We then the in all sequencing across all to barcode sequences (Chen et al., 2018; Chen et al., and assigned barcode to cells. were to the et al., using and et al., 2019; Materials and methods Figure 3 Download asset Open asset In situ sequencing transcriptomic types of neurons infected with barcoded rabies virus. (A) Illustration of and approach for in situ sequencing of rabies (B) Illustration of the experiments. The two libraries were injected in VISal and and from the area were that to the two injection sites would be labeled with different of barcodes, as indicated by different (C) of a by a during the The sequencing for genes and barcodes in the area are shown on the The gene or that to each is indicated. 3 for a of genes sequenced cells shown on a UMAP plot of their gene expression patterns (D) or on a Colors indicate cluster as shown in the (F) matching between and from scRNA-seq (Tasic et al., 2018). Dot size and colors indicate the that neurons from a cluster are assigned to a reference scRNA-seq (G) The number of cells with the indicated number of barcodes per cell (H) The count of the primary barcode and the count of the second most barcode in each barcoded Cells the are considered as than one barcode per (I) of the most