High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling
Curation statements for this article:-
Curated by eLife
eLife assessment
This manuscript describes a valuable new circuit mapping and profiling technique called Multiplexed projEction neuRons retrograde barcodE (MERGEseq) that combines transcriptome and projectome data at a single-cell resolution. The authors provide solid evidence that MERGEseq can be used to identify projection targets and cell type/layer/transcriptome differences of projection neurons in the mouse prefrontal cortex, and validation experiments are rigorous. While this report is a proof-of-principle that MERGEseq is useful for circuit mapping and profiling and many potential details will influence conclusions, this technique could easily be adapted to other regions with known projection targets and adds to a growing arsenal of combinatorial circuit mapping and profiling tools.
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
Abstract
Deciphering patterns of connectivity between neurons in the brain is a critical step toward understanding brain function. Imaging-based neuroanatomical tracing identifies area-to-area or sparse neuron-to-neuron connectivity patterns, but with limited throughput. Barcode-based connectomics maps large numbers of single-neuron projections, but remains a challenge for jointly analyzing single-cell transcriptomics. Here, we established a rAAV2-retro barcode-based multiplexed tracing method that simultaneously characterizes the projectome and transcriptome at the single neuron level. We uncovered dedicated and collateral projection patterns of ventromedial prefrontal cortex (vmPFC) neurons to five downstream targets and found that projection-defined vmPFC neurons are molecularly heterogeneous. We identified transcriptional signatures of projection-specific vmPFC neurons, and verified Pou3f1 as a marker gene enriched in neurons projecting to the lateral hypothalamus, denoting a distinct subset with collateral projections to both dorsomedial striatum and lateral hypothalamus. In summary, we have developed a new multiplexed technique whose paired connectome and gene expression data can help reveal organizational principles that form neural circuits and process information.
Article activity feed
-
-
Author Response
Reviewer #1 (Public Review):
With MERGEseq, the authors sought to develop a scalable and accessible method for getting both projectome and transcriptome information at the single-cell level from multiple projection targets within a single animal. MERGEseq uses a retro rAAV2 to deliver a 15-nucleotide barcode driven by a CAG promoter with co-expression of eGFP to enrich barcoded cells using FACS. Injection of this rAAV2 in distinct regions (with each injection region distinguished by a unique barcode that is specific to the virus used) allows retrograde trafficking and expression of the barcodes in cells that project to the injected region. In this manuscript, rAAVs harboring 5 unique barcodes were stereotactically delivered to 5 targets of the mouse: dorsomedial striatum (DMS), mediodorsal thalamic nucleus (MD), basal …
Author Response
Reviewer #1 (Public Review):
With MERGEseq, the authors sought to develop a scalable and accessible method for getting both projectome and transcriptome information at the single-cell level from multiple projection targets within a single animal. MERGEseq uses a retro rAAV2 to deliver a 15-nucleotide barcode driven by a CAG promoter with co-expression of eGFP to enrich barcoded cells using FACS. Injection of this rAAV2 in distinct regions (with each injection region distinguished by a unique barcode that is specific to the virus used) allows retrograde trafficking and expression of the barcodes in cells that project to the injected region. In this manuscript, rAAVs harboring 5 unique barcodes were stereotactically delivered to 5 targets of the mouse: dorsomedial striatum (DMS), mediodorsal thalamic nucleus (MD), basal amygdala (BLA), lateral hypothalamus (LH), and agranular insular cortex (AI). After a 6-week period to allow for viral transduction and expression, the ventromedial prefrontal cortex (vmPFC) was harvested for scRNAseq. vmPFC scRNAseq data were validated against previously published PFC datasets, demonstrating that MERGEseq does not disrupt transcript expression and identifies the same principal cell types as annotated in previous studies. Importantly, MERGEseq enabled the identification of cell types in the vmPFC that project to distinct areas, with separation occurring largely based on cell type and cortical layer. The application of stringent criteria for barcode index determination is rigorous and improves confidence that barcoded cells are correctly identified. The observation that all barcoded cells were excitatory is consistent with prior work, although it is not clear if viral tropism contributes to this in some way. In a parallel experiment, FAC-sorted cells (vmPFC cells expressing EGFP) were isolated as a comparison. Notably, EGFP+ cells were exclusively excitatory neurons, consistent with literature showing PFC projection neurons are excitatory. Next, barcode analysis was combined with transcriptional identification of neuronal subtypes to define general projection patterns and single-cell projection patterns, which were validated by the DMS and MD in situ using retrograde tracing in combination with RNA FISH. MERGEseq data were also used to identify transcriptional differences between neurons with dedicated and bifurcated projections. DMS+LH and DMS+MD projecting neurons had distinct transcriptional profiles, unlike cells with other targets. RNA FISH for marker gene Pou3f and retrograde tracing from DMS+LH projecting cells demonstrate enrichment of this gene in this projection population. Finally, machine-learning was used to predict projection targets based on transcriptional profiles. In this dataset, 50 highly variable genes (HVGs) were optimal for predicting projection patterns, though this might vary in different circuits. Overall, the results of this manuscript are well presented and include rigorous validation for select vmPFC targets with in situ techniques. The application of unique barcodes for retro-AAV delivery is an accessible tool that other labs can implement to study other brain circuits.
Ultimately, MERGEseq is a subtle conceptual advancement over VECTORseq (retro-AAV delivered transgenes rather than barcodes, in combination with scRNAseq) that offers higher confidence in the described projectome diversity in comparison. The use of a retrograde AAV inherently limits the number of projection areas that can be assessed, a weakness compared to anterograde approaches such as MAPseq/BARseq. However, BARseq demands more time and resources; further, the use of the highly toxic Sindbis virus limits the application of this technique. This manuscript builds upon previous work by utilizing machine learning to predict projection targets. BARseq2 could be used to rigorously validate predicted projectomes and gain single-cell information regarding target neurons. Overall, MERGEseq is an accessible technique that can be used across many animal models and serve as an important starting point to define circuits at the single-cell level.
We thank reviewer for the comprehensive review. We are grateful for reviewer’s recognition of the conceptual advancement of MERGE-seq and the rigorous criteria we applied for projection barcode determination. We have revised the Introduction to highlight advancements in our method. We also discussed the balance of transcriptomic comprehensiveness against spatial resolution in the revised Discussion. Reviewer’s comments have been invaluable in enhancing the clarity and depth of our manuscript.
Reviewer #2 (Public Review):
Investigating the relationship between transcriptomic profiles, their axonal projection and collateralization patterns will help define neuronal cell types in the mammalian central nervous system. The study by Xu et al. combined multiple retrograde viruses with barcodes and single-cell RNA-sequencing (MERGE-seq) to determine the projection and collateralization patterns of transcriptomically defined ventral medial prefrontal cortex (vmPFC) projection neurons. They found a complex relationship: the same transcriptomically defined cell types project to multiple target regions, and the same target region receives input from multiple transcriptomic types of vmPFC neurons. Further, collateralization patterns of vmPFC to the five target regions they investigated are highly non-random.
While many of the biological conclusions are not surprising given recent studies on the collateralization patterns of vmPFC neurons using single neuron tracing and other methods that integrate transcriptomics and projections, MERGE-seq provides validation, at the single cell level, collateralization patterns of individual vmPFC neurons, and thus offer new and valuable information over what has been published. The method can also be used to study collateralization patterns of other neuron types.
Some of the conclusions the authors draw depend on the efficiency of retrograde labeling, which was not determined. Without quantitative information on retrograde labeling efficiency, and unless such efficiency is close to 100%, these conclusions are likely misleading.
We thank reviewer for recognizing the contributions of our MERGE-seq technique in advancing the understanding of projection patterns of vmPFC neurons. We concur that while our conclusions align with previous findings, our single-cell level analysis provides additional depth to the existing knowledge of the field. We acknowledge the challenge to quantify retrograde labeling efficiency to draw quantitive conclusions based on our findings. Alternatively, we have used fMOST-based single-neuron tracing data and analysis to validate our projection patterns and ensure the robustness of our conclusions in the revised manuscript. We also more explicitly clarified the limitations of the quantitive conclusion drawn from MERGE-seq in the revised Discussion. The insights of reviewer are greatly appreciated and will inform the improvement of our research methodology.
Reviewer #3 (Public Review):
This manuscript describes a multiplexed approach for the identification of transcriptional features of neurons projecting to specific target areas at the single-cell level. This approach, called MERGE-seq, begins with multiplexed retrograde tracing by injecting distinctly barcoded rAAV-retro viruses into different target areas. The transcriptomes and barcoding of neurons in the source area are then characterized by single-cell RNA sequencing (scRNAseq) on the 10xGenomics platform. The projection targets of barcoded neurons in the source area can be inferred by matching the detected barcodes to the barcode sequences to of rAAV-retro viruses injected into the target areas.
The authors validated their approach by injecting five rAAV-retro GFP viruses, each encoding a different barcode, into five known targets of the ventromedial prefrontal cortex (vmPFC). The transcriptomes and barcoding of vmPFC neurons were then analyzed by scRNA-seq with or without enrichment of retrogradely labeled neurons based on GFP fluorescence. The authors confirmed the previously described heterogeneity of vmPFC neurons. In addition, they showed that most transcriptionally defined cell types project to multiple targets and that the five targets received projections from multiple transcriptomic types. The authors further characterized the transcriptomic features of barcoded vmPFC neurons with different projection patterns and defined Pou3f1 as a marker gene of neurons extending collateral branches to the dorsomedial striatum and lateral hypothalamus.
Overall, the results of the manuscript are convincing: the transcriptomic vmPFC cell types defined by scRNAseq in this study appear to correlate well with previous studies, the bifurcated projection patterns inferred by barcoding are validated using dual-color retro-AAV tracing, and marker genes for projection-specific cell subclasses are validated in retrogradely labeled vmPFC using RNA FISH for marker detection.
The concept of combining retrograde tracing and scRNAseq is not new. Previous studies have applied recombinase-expressing viruses capable of retrograde labeling, such as CAV, rabies virus, and AAV2-Retro, to retrogradely label and induce the expression of fluorescence markers in projection neurons, therefore facilitating enrichment and analysis of neurons projecting to a specific target. Multiplexed analysis can be achieved with the combination of different reporter viruses or viruses expressing different recombinases and appropriate reporter mouse lines. The advantages of MERGE-seq include that no transgenic lines are required and that it could be applied at even higher levels of multiplexity.
We thank reviewer for the insightful review of our manuscript and the recognition of the advantages of MERGE-seq. We appreciate reviewer acknowledged the robust validation of the method through dual-color retro-AAV tracing and RNA FISH, and the confirmation of previous findings on vmPFC neuronal heterogeneity and collateral projection patterns. We provided additional joint analysis with fMOST-based single-neuron projectome data (Gao et al., 2022, Nature Neuroscience) to further validate the projection patterns (>= 3 targets) that cannot be easily validated with dual-color retro-AAV tracing.
However, previously existing datasets that have already profiled this region with scRNAseq have not been utilized to their full extent. Therefore, for the proper context with prior literature, bioinformatic integration of these scRNAseq and prior scRNAseq data is needed.
Moreover, robust detection of barcodes in neurons labeled by barcoded AAV-retro viruses remains a challenge. The authors should clearly discuss the difficulties with barcode detection in this approach, as well as discuss potential solutions, which are important for others interested in its approach.
While this study is limited to the five known targets of vmPFC, the results suggest that MERGE-seq is a valuable tool that could be used in the future to characterize projection targets and transcriptomes of neurons in a multiplexed manner. As MERGE-seq uses AAVs to deliver barcodes, this method has the potential for application in model organisms for which transgenic lines are not available. Further improvements in experimental design and data analysis should be considered when applying MERGE-seq to poorly characterized source areas or with increased multiplexity of target areas.
In summary, this is a valuable approach, but the authors should clearly provide the context for their study within the existing literature, transparently discuss the limitations of MERGE-seq, as well as suggest improvements for the future.
We appreciate your positive assessment of MERGE-seq as a valuable approach with future potential. As recommended, we have performed integration analysis with existing vmPFC scRNA-seq studies, including Bhattacherjee et al., 2019, Lui et al., 2021, Yao at al., 2021, and specifically recently published MERFISH data of PFC (Bhattacherjee et al., 2023).
In the revised Discussion, we have transparently addressed the current limitations of MERGE-seq, including imperfect retrograde labeling efficiency, variable barcode recovery rates and cell loss during dissociation. We also addressed the challenges in detecting and recovering projection barcodes and suggested potential solutions such as using FAC-sorted EGFP-negative cells for control and applying single-molecule FISH techniques. We sincerely appreciate reviewer’s rigorous and insightful feedback, which has substantially strengthened our manuscript.
-
eLife assessment
This manuscript describes a valuable new circuit mapping and profiling technique called Multiplexed projEction neuRons retrograde barcodE (MERGEseq) that combines transcriptome and projectome data at a single-cell resolution. The authors provide solid evidence that MERGEseq can be used to identify projection targets and cell type/layer/transcriptome differences of projection neurons in the mouse prefrontal cortex, and validation experiments are rigorous. While this report is a proof-of-principle that MERGEseq is useful for circuit mapping and profiling and many potential details will influence conclusions, this technique could easily be adapted to other regions with known projection targets and adds to a growing arsenal of combinatorial circuit mapping and profiling tools.
-
Reviewer #1 (Public Review):
With MERGEseq, the authors sought to develop a scalable and accessible method for getting both projectome and transcriptome information at the single-cell level from multiple projection targets within a single animal. MERGEseq uses a retro rAAV2 to deliver a 15-nucleotide barcode driven by a CAG promoter with co-expression of eGFP to enrich barcoded cells using FACS. Injection of this rAAV2 in distinct regions (with each injection region distinguished by a unique barcode that is specific to the virus used) allows retrograde trafficking and expression of the barcodes in cells that project to the injected region. In this manuscript, rAAVs harboring 5 unique barcodes were stereotactically delivered to 5 targets of the mouse: dorsomedial striatum (DMS), mediodorsal thalamic nucleus (MD), basal amygdala (BLA), …
Reviewer #1 (Public Review):
With MERGEseq, the authors sought to develop a scalable and accessible method for getting both projectome and transcriptome information at the single-cell level from multiple projection targets within a single animal. MERGEseq uses a retro rAAV2 to deliver a 15-nucleotide barcode driven by a CAG promoter with co-expression of eGFP to enrich barcoded cells using FACS. Injection of this rAAV2 in distinct regions (with each injection region distinguished by a unique barcode that is specific to the virus used) allows retrograde trafficking and expression of the barcodes in cells that project to the injected region. In this manuscript, rAAVs harboring 5 unique barcodes were stereotactically delivered to 5 targets of the mouse: dorsomedial striatum (DMS), mediodorsal thalamic nucleus (MD), basal amygdala (BLA), lateral hypothalamus (LH), and agranular insular cortex (AI). After a 6-week period to allow for viral transduction and expression, the ventromedial prefrontal cortex (vmPFC) was harvested for scRNAseq. vmPFC scRNAseq data were validated against previously published PFC datasets, demonstrating that MERGEseq does not disrupt transcript expression and identifies the same principal cell types as annotated in previous studies. Importantly, MERGEseq enabled the identification of cell types in the vmPFC that project to distinct areas, with separation occurring largely based on cell type and cortical layer. The application of stringent criteria for barcode index determination is rigorous and improves confidence that barcoded cells are correctly identified. The observation that all barcoded cells were excitatory is consistent with prior work, although it is not clear if viral tropism contributes to this in some way. In a parallel experiment, FAC-sorted cells (vmPFC cells expressing EGFP) were isolated as a comparison. Notably, EGFP+ cells were exclusively excitatory neurons, consistent with literature showing PFC projection neurons are excitatory. Next, barcode analysis was combined with transcriptional identification of neuronal subtypes to define general projection patterns and single-cell projection patterns, which were validated by the DMS and MD in situ using retrograde tracing in combination with RNA FISH. MERGEseq data were also used to identify transcriptional differences between neurons with dedicated and bifurcated projections. DMS+LH and DMS+MD projecting neurons had distinct transcriptional profiles, unlike cells with other targets. RNA FISH for marker gene Pou3f and retrograde tracing from DMS+LH projecting cells demonstrate enrichment of this gene in this projection population. Finally, machine-learning was used to predict projection targets based on transcriptional profiles. In this dataset, 50 highly variable genes (HVGs) were optimal for predicting projection patterns, though this might vary in different circuits. Overall, the results of this manuscript are well presented and include rigorous validation for select vmPFC targets with in situ techniques. The application of unique barcodes for retro-AAV delivery is an accessible tool that other labs can implement to study other brain circuits.
Ultimately, MERGEseq is a subtle conceptual advancement over VECTORseq (retro-AAV delivered transgenes rather than barcodes, in combination with scRNAseq) that offers higher confidence in the described projectome diversity in comparison. The use of a retrograde AAV inherently limits the number of projection areas that can be assessed, a weakness compared to anterograde approaches such as MAPseq/BARseq. However, BARseq demands more time and resources; further, the use of the highly toxic Sindbis virus limits the application of this technique. This manuscript builds upon previous work by utilizing machine learning to predict projection targets. BARseq2 could be used to rigorously validate predicted projectomes and gain single-cell information regarding target neurons. Overall, MERGEseq is an accessible technique that can be used across many animal models and serve as an important starting point to define circuits at the single-cell level.
-
Reviewer #2 (Public Review):
Investigating the relationship between transcriptomic profiles, their axonal projection and collateralization patterns will help define neuronal cell types in the mammalian central nervous system. The study by Xu et al. combined multiple retrograde viruses with barcodes and single-cell RNA-sequencing (MERGE-seq) to determine the projection and collateralization patterns of transcriptomically defined ventral medial prefrontal cortex (vmPFC) projection neurons. They found a complex relationship: the same transcriptomically defined cell types project to multiple target regions, and the same target region receives input from multiple transcriptomic types of vmPFC neurons. Further, collateralization patterns of vmPFC to the five target regions they investigated are highly non-random.
While many of the biological …
Reviewer #2 (Public Review):
Investigating the relationship between transcriptomic profiles, their axonal projection and collateralization patterns will help define neuronal cell types in the mammalian central nervous system. The study by Xu et al. combined multiple retrograde viruses with barcodes and single-cell RNA-sequencing (MERGE-seq) to determine the projection and collateralization patterns of transcriptomically defined ventral medial prefrontal cortex (vmPFC) projection neurons. They found a complex relationship: the same transcriptomically defined cell types project to multiple target regions, and the same target region receives input from multiple transcriptomic types of vmPFC neurons. Further, collateralization patterns of vmPFC to the five target regions they investigated are highly non-random.
While many of the biological conclusions are not surprising given recent studies on the collateralization patterns of vmPFC neurons using single neuron tracing and other methods that integrate transcriptomics and projections, MERGE-seq provides validation, at the single cell level, collateralization patterns of individual vmPFC neurons, and thus offer new and valuable information over what has been published. The method can also be used to study collateralization patterns of other neuron types.
Some of the conclusions the authors draw depend on the efficiency of retrograde labeling, which was not determined. Without quantitative information on retrograde labeling efficiency, and unless such efficiency is close to 100%, these conclusions are likely misleading.
-
Reviewer #3 (Public Review):
This manuscript describes a multiplexed approach for the identification of transcriptional features of neurons projecting to specific target areas at the single-cell level. This approach, called MERGE-seq, begins with multiplexed retrograde tracing by injecting distinctly barcoded rAAV-retro viruses into different target areas. The transcriptomes and barcoding of neurons in the source area are then characterized by single-cell RNA sequencing (scRNAseq) on the 10xGenomics platform. The projection targets of barcoded neurons in the source area can be inferred by matching the detected barcodes to the barcode sequences to of rAAV-retro viruses injected into the target areas.
The authors validated their approach by injecting five rAAV-retro GFP viruses, each encoding a different barcode, into five known targets …
Reviewer #3 (Public Review):
This manuscript describes a multiplexed approach for the identification of transcriptional features of neurons projecting to specific target areas at the single-cell level. This approach, called MERGE-seq, begins with multiplexed retrograde tracing by injecting distinctly barcoded rAAV-retro viruses into different target areas. The transcriptomes and barcoding of neurons in the source area are then characterized by single-cell RNA sequencing (scRNAseq) on the 10xGenomics platform. The projection targets of barcoded neurons in the source area can be inferred by matching the detected barcodes to the barcode sequences to of rAAV-retro viruses injected into the target areas.
The authors validated their approach by injecting five rAAV-retro GFP viruses, each encoding a different barcode, into five known targets of the ventromedial prefrontal cortex (vmPFC). The transcriptomes and barcoding of vmPFC neurons were then analyzed by scRNA-seq with or without enrichment of retrogradely labeled neurons based on GFP fluorescence. The authors confirmed the previously described heterogeneity of vmPFC neurons. In addition, they showed that most transcriptionally defined cell types project to multiple targets and that the five targets received projections from multiple transcriptomic types. The authors further characterized the transcriptomic features of barcoded vmPFC neurons with different projection patterns and defined Pou3f1 as a marker gene of neurons extending collateral branches to the dorsomedial striatum and lateral hypothalamus.
Overall, the results of the manuscript are convincing: the transcriptomic vmPFC cell types defined by scRNAseq in this study appear to correlate well with previous studies, the bifurcated projection patterns inferred by barcoding are validated using dual-color retro-AAV tracing, and marker genes for projection-specific cell subclasses are validated in retrogradely labeled vmPFC using RNA FISH for marker detection.
The concept of combining retrograde tracing and scRNAseq is not new. Previous studies have applied recombinase-expressing viruses capable of retrograde labeling, such as CAV, rabies virus, and AAV2-Retro, to retrogradely label and induce the expression of fluorescence markers in projection neurons, therefore facilitating enrichment and analysis of neurons projecting to a specific target. Multiplexed analysis can be achieved with the combination of different reporter viruses or viruses expressing different recombinases and appropriate reporter mouse lines. The advantages of MERGE-seq include that no transgenic lines are required and that it could be applied at even higher levels of multiplexity.
However, previously existing datasets that have already profiled this region with scRNAseq have not been utilized to their full extent. Therefore, for the proper context with prior literature, bioinformatic integration of these scRNAseq and prior scRNAseq data is needed.
Moreover, robust detection of barcodes in neurons labeled by barcoded AAV-retro viruses remains a challenge. The authors should clearly discuss the difficulties with barcode detection in this approach, as well as discuss potential solutions, which are important for others interested in its approach.
While this study is limited to the five known targets of vmPFC, the results suggest that MERGE-seq is a valuable tool that could be used in the future to characterize projection targets and transcriptomes of neurons in a multiplexed manner. As MERGE-seq uses AAVs to deliver barcodes, this method has the potential for application in model organisms for which transgenic lines are not available. Further improvements in experimental design and data analysis should be considered when applying MERGE-seq to poorly characterized source areas or with increased multiplexity of target areas.
In summary, this is a valuable approach, but the authors should clearly provide the context for their study within the existing literature, transparently discuss the limitations of MERGE-seq, as well as suggest improvements for the future.
-
-