PMCID
stringclasses 24
values | Title
stringclasses 24
values | Sentences
stringlengths 2
40.7k
|
|---|---|---|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Syt6 is one of the DEGs of MD-projecting neurons (Figure 4B), and is the marker gene of L6-Syt6 cluster.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
By retrograde labeling of MD-projecting neurons and Syt6 FISH experiment, we found that about 51.6% ± 16.9% Syt6 neurons project to MD.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Further statistical analysis showed that, among Syt6 MD-projecting (Syt6EGFP) neurons, 84.2% ± 14.8% were located in layer 6 while 15.8% ± 14.8% were located in layer 5 (Figure 4C), similar to the pattern obtained in our MERGE-seq analysis (Figure 2F).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
These results are in accordance with single-neuron projectomic and transcriptomic analysis of MERGE-seq, indicating that MERGE-seq can faithfully reveal the transcriptomic features of projection-specific neurons.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Axons of projection neurons, including vmPFC neurons, have highly complex collaterals, which could regulate information processing and neural response properties at the microcircuit level (Gagnon and Parent, 2014; Gao et al., 2022; Rockland, 2019).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
However, the molecular features of neurons sending collateral projections remain elusive.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
MERGE-seq provides an opportunity to explore.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Here, we identified DEGs for neurons with dedicated and collateral projection pattern (Figure 5A).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Next, we asked whether there was transcriptional difference between neurons with dedicated projection to A and neurons with bifurcated projection to A and B. DEGs were rare in comparisons between projection patterns A/B vs. A, or A/B vs. B in all of groups we tested, except for the DMS + LH group and DMS + MD group (Figure 5B and C,Figure 5—figure supplement 1).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We found that DMS + LH projection neurons were transcriptionally distinct to DMS but similar to LH, and DMS + MD neurons were transcriptionally distinct to DMS but similar to MD (Figure 5B and C,Figure 5—figure supplement 1).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Specifically, we identified a set of genes which differentially expressed in DMS + LH projection neurons (such as Pou3f1, Igfbp4, and Gprc5b) or DMS + MD projection neurons (such as Rprm, Crym, Hs3st4 and Bc1).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Interestingly, Pou3f1 is marker gene of L5-Bcl6 neurons (layer 5 neuron subtype), representing one of the two distinct neuron subtypes within the DMS + LH projection neuronal population (Figure 3G).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We next verified the specific gene expression in DMS + LH projection neurons by using RNA FISH in combination with dual-color retrovirus labeling assay (Figure 5D).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We found that the expression of Pou3f1 was mainly distributed in layer 5, where Pou3f1 was specifically expressed in dual-color labeled DMS + LH projecting neurons (white arrowheads, Figure 5E) and LH projecting neurons (white arrows, Figure 5E), but not DMS projecting neurons (blue arrows, Figure 5E).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Quantification analysis showed that, among Pou3f1 neurons, there are 55.7% ± 10.4% DMS + LH-projecting (Pou3f1EGFPtdT) neurons, 31.6% ± 13.1% dedicated LH-projecting (Pou3f1EGFPtdT) neurons, and 8.89% ± 2.38% dedicated DMS-projecting (Pou3f1EGFPtdT) neurons (Figure 5G).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We additionally discovered that 3.79% ± 2.91% of Pou3f1 neurons did not project to either DMS or LH (Pou3f1EGFPtdT) (yellow arrowheads, Figure 5F, Figure 5G).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
These results are consistent with our observation based on MERGE-seq data (Figure 3G). (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A) Heatmap showing scaled expression of calculated DEGs based on 10 projection patterns.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Top 10 DEGs ordered by average log2 fold change of each pattern were selected. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
B) Volcano plot showing genes differentially expressed in the DMS + LH-bifurcated projection pattern compared to the DMS-dedicated projection pattern. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
C) Track plots showing normalized data of the selected DEGs in DMS-dedicated, LH-dedicated, and DMS + LH-bifurcated projection pattern. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
D–F) Examining Pou3f1 and DMS + LH-bifurcated projection pattern using RNA FISH and immunostaining of dual-color traced retrograde labeled neurons.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Virus injection scheme was the same as in Figure 3—figure supplement 1.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Scale bars, 200 µm. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
E, F) Enlarged view of dotted box in (D).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Arrow heads indicate Pou3f1EGFPtdTomato neurons, white arrows indicate Pou3f1EGFPtdTomato neurons, blue arrows indicate Pou3f1EGFPtdTomato neurons, and yellow arrowheads indicate Pou3f1EGFPtdTomato neurons.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Scale bars, 50 µm. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
G) Quantification of (D).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
N=3 mice, Data are presented as mean ± SD.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
In (A), 1,853 barcoded neurons (top 10 frequent projection patterns) were represented.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
In (C), 805 barcoded neurons (projection pattern DMS + LH = 35, LH = 176, DMS = 594) were represented.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Figure 5—figure supplement 1.DEGs between dedicated projection neurons versus bifurcated neurons.(A) Volcano plots of DEGs calculated between A and A/B projection patterns.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
See also Figure 5.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Assigned DEGs (red dots) were determined using threshold: Log2 fold change = 0.5, p value cutoff=10. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
B) Track plots showing normalized data of the selected DEGs (DMS versus DMS + MD projection) in DMS-dedicated, MD-dedicated, and DMS + MD-bifurcated vmPFC neurons. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A) Volcano plots of DEGs calculated between A and A/B projection patterns.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
See also Figure 5.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Assigned DEGs (red dots) were determined using threshold: Log2 fold change = 0.5, p value cutoff=10. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
B) Track plots showing normalized data of the selected DEGs (DMS versus DMS + MD projection) in DMS-dedicated, MD-dedicated, and DMS + MD-bifurcated vmPFC neurons.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Together, by MERGE-seq analysis and experimental validation, we uncovered that Pou3f1 predominantly marks neurons projecting to the LH, denoting a distinct subset with collateral projections to both DMS and LH.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Although many efforts have been made to correlate gene expression with neuronal circuit connectivity (Huang et al., 2020; Sorensen et al., 2015; Sun et al., 2021), the lack of a shared coordinate system for two modalities or limited genes examined reduces the prediction precision.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
MERGE-seq overcomes these challenges by acquiring high-throughput gene expression and projection pattern in the same neuron (Figure 6A).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
To evaluate potential relationships between the transcriptome and projectome, we used a probabilistic classifier, Naïve Bayes classifier, to predict binary projection patterns for each projection target based on transcription profiles.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
First, we encoded binary projection labels for each target region, encompassing both barcoded and non-barcoded projections, and subsequently trained a separate set of models for each of the five targets: AI, DMS, BLA, LH, and MD (see Materials and methods).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Subsequently, we conducted a systematic evaluation of the impact of varying numbers of highly variable genes (HVGs), ranging from 2 to 5000, on model performance.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
This analysis revealed that employing the top 50 HVGs for modeling yielded the the highest F1 score (a harmonic mean of precision and recall), area under the curve (AUC), and a comparatively high prediction accuracy (see Materials and methods, Figure 6—figure supplement 1A).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Next, we chose top 50 HVGs as features to build the model.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
As a control model, we chose 50 randomly chosen genes.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Five projection targets models were independently trained by splitting cells into training (70%) and test dataset (30%).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Using top 50 HVGs also gave rise to significantly better model performance in regarding to prediction accuracy, AUC and F1 score, compared to using randomly chosen 50 genes (Figure 6B).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We also performed 100 iterations randomly sampling 1000 cells and swapping barcoded with non-barcoded labels, which substantially decreased model predictive performance across various evaluation metrics (see Materials and methods, Figure 6—figure supplement 1B).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
This outcome underscores the critical importance of label accuracy for the predictive capabilities of the model, suggesting the authenticity of current barcoded cells labels despite potential false positives from stringent UMI thresholding.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Altogether, these results suggest that the top 50 HVGs are more informative for predicting and decoding projection patterns. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A) Schematics of machine learning modeling steps. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
B) Prediction accuracy (left panel), AUC score (middle panel) and F1 score (right panel) of top HVGs and random chosen equal number of genes for modeling building.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A total of 100 trials have been performed by randomly sampling 1000 cells from 8210 cells.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Top 50 HVGs or 50 randomly chosen genes were used as features per trial.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Comparisons were made between models built by the HVGs group and random genes group for each projection target.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
The displayed p value was computed using a two-sided Wilcoxon test.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Data are the mean ± SD. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
C, E) SHAP summary plots of DMS and MD showing important features (genes) with feature effects.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For each model, non-barcoded cells were encoded to class 0 and barcoded cells were encoded to class 1.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Models were built using top 50 HVGs. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
D, F) Normalized expression of the most important genes with positive feature effects in Naïve Bayes modeling of DMS (D) or MD (F) and normalized expression of barcode 1 representing DMS-projecting (E) or barcode 2 representing MD-projecting (F) on PC1 and PC2 embeddings.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Note that bottom panel of (D, F) is identical to DMS and MD barcode expression in Figure 3—figure supplement 1M.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
In (D, F), 1853 barcoded neurons (top 10 frequent projection patterns) were represented.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
In (C, E), For calculating SHAP values, both the training and testing datasets were subsampled to include 1500 cells each.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Figure 6—figure supplement 1.SHAP summary plots of Naïve Bayes models.(A) Prediction accuracy (left panel), AUC score (middle panel) and F1 score (right panel) by tuning number of HVGs used for naïve bayes modeling building.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A total of 100 trials have been performed by randomly sampling 1000 cells from 8210 cells and calculating top HVGs per trial. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
B) Prediction accuracy (left panel), AUC score (middle panel) and F1 score (right panel) of original top 50 HVGs and after swapping barcoded/non-barcoded cell labels for modeling building.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A total of 100 trials have been performed by randomly sampling 1000 cells from 8210 cells.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
In each trial, barcoded/non-barcoded cell labels were swapped.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
The number of swapped cells depends on the minimum number of barcoded cells or non-barcoded cells.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Models were built with the top 50 HVGs using original labels (Original) or labels swapped (Swapping) for comparison for each projection target.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
The displayed p value was computed using a two-sided Wilcoxon test.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Data are the mean ± SD. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
C) SHAP summary plots of AI, BLA, and LH showing important features (genes) with feature effects.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For each model, non-barcoded cells were encoded to class 0 and barcoded cells were encoded to class 1. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A) Prediction accuracy (left panel), AUC score (middle panel) and F1 score (right panel) by tuning number of HVGs used for naïve bayes modeling building.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A total of 100 trials have been performed by randomly sampling 1000 cells from 8210 cells and calculating top HVGs per trial. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
B) Prediction accuracy (left panel), AUC score (middle panel) and F1 score (right panel) of original top 50 HVGs and after swapping barcoded/non-barcoded cell labels for modeling building.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
A total of 100 trials have been performed by randomly sampling 1000 cells from 8210 cells.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
In each trial, barcoded/non-barcoded cell labels were swapped.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
The number of swapped cells depends on the minimum number of barcoded cells or non-barcoded cells.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Models were built with the top 50 HVGs using original labels (Original) or labels swapped (Swapping) for comparison for each projection target.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
The displayed p value was computed using a two-sided Wilcoxon test.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Data are the mean ± SD. (
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
C) SHAP summary plots of AI, BLA, and LH showing important features (genes) with feature effects.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
For each model, non-barcoded cells were encoded to class 0 and barcoded cells were encoded to class 1.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
To interpret the important genes contributing to a certain projection pattern, we used a game-theoretic approach to explain the output of HVGs-based Naïve Bayes models (Lundberg et al., 2020; Figure 6A).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
We used top 50 HVGs to build Naïve Bayes model and summarized effects of HVGs in SHAP (SHapley Additive exPlanations) values for each projection pattern (see Materials and methods; Figure 6C–F, Figure 6—figure supplement 1C).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
As examples, Nptxr gene was the top positive predictors for DMS projection, suggesting that a cell that expresses high levels of Nptxr has a higher probability of projecting to DMS.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Similarly, Rprm was the top positive predictors for MD projection.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
By examining top effective genes (features) on PC embeddings of the projection matrix, we found that the expression pattern of these positive predictors mostly overlaps with projection barcode distribution (Figure 6D and F).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
These results mathematically establish the relationship between gene expression and structural connectivity, indicating the predictive power of a specific gene cluster for projection properties of vmPFC neurons.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Given the complexity of brain circuits, neuronal subtypes must be characterized from multiple viewpoints (Zeng, 2022).
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Information including neuronal projection patterns (i.e. region-to-region connectivity), physiological properties, gene expression, and how they encode information in behavioral paradigms, are essential to understand functional brain circuits.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Therefore, it is inevitably difficult to acquire a complete picture of brain circuits when only one analytic modality is considered.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
In this study, we have developed a multiplexed barcoding method that is integrated with scRNA-seq, enabling simultaneous transcriptome and projectome analyses.
|
PMC10914349
|
High-throughput mapping of single-neuron projection and molecular features by retrograde barcoded labeling.
|
Retrograde AAVs are injected into multiple target regions simultaneously, thereby labeling projection neurons within the brain region of interest and facilitating their transcriptional analysis.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.