Effect of Treatment on the immune landscape in young and old mice
B Cell Subset Analysis and Differential Expression
09/18/25
- 1 Study Overview
- 2 B Cell Population Analysis
- 3 Differential Expression Analysis
1 Study Overview
Principal Investigator:
Bioinformatics Analysis: Dr. Heather Kates (hkates@ufl.edu)
1.1 Experimental Design
This analysis examines changes in B cell population composition and gene expression related to Treatment and to age comparing young and aged mice lymph node and tumor tissues.
Sample Groups:
- AL: Aged mice, Lymph node samples
- AT: Aged mice, Tumor samples
- YL: Young mice, Lymph node samples
- YT: Young mice, Tumor samples
Treatment Conditions:
- Vehicle control (3 biological replicates per group)
- Treatment (3 biological replicates per group)
1.2 Data Generation and Processing
Sequencing: 10X Genomics single-cell RNA sequencing with cell hashing and VDJ-B/T sequencing was performed on all samples.
Analysis Pipeline: Raw sequencing data was processed through Cell Ranger Multi, followed by standard Seurat-based single-cell analysis including quality control, normalization, clustering, and cell type annotation. All sample groups were integrated using Harmony batch correction.
Input Dataset: The integrated dataset contains 6.241210^{4} high-quality cells across all conditions, with 26914 BCR-positive B cells identified through VDJ sequencing.
1.3 Data Access
Complete datasets, intermediate objects, and reproducible code are available at: https://data.rc.ufl.edu/secure/cancercenter-dept/zhangw/GE-8124/
Contact (hkates@ufl.edu) for access credentials.
2 B Cell Population Analysis
2.1 Identifying B Cells Using VDJ Data
B cells were identified using BCR (B cell receptor) information from the VDJ-B sequencing data. This approach is more specific than using gene expression markers alone, as it captures cells that have successfully undergone V(D)J recombination.
## Original dataset: 62412 cells## B cells identified: 26914 cells## B cell percentage: 43.12 %2.2 Clustering B Cell Subpopulations
Because B cells represent a heterogeneous population with distinct functional states, this subset was re-analyzed to identify B cell sub-populations with better resolution than in the full dataset including re-normalization, dimensionality reduction, and clustering.
Multiple clustering resolutions (0.05-0.8) were tested, yielding 4-17 clusters. Resolution 0.1 was selected for producing 6 biologically interpretable clusters that capture major B cell functional states without over-clustering.
b_cells <- NormalizeData(b_cells)
b_cells <- FindVariableFeatures(b_cells, selection.method = "vst", nfeatures = 2000)
b_cells <- ScaleData(b_cells)
b_cells <- RunPCA(b_cells, features = VariableFeatures(object = b_cells))
b_cells <- RunHarmony(b_cells, group.by.vars = "sample_id")
b_cells <- RunUMAP(b_cells, reduction = "harmony", dims = 1:30)
resolutions <- c(0.05, 0.1, 0.2, 0.3, 0.5, 0.8)
b_cells <- FindNeighbors(b_cells, reduction = "harmony", dims = 1:30)
# Store clustering results
clustering_results <- data.frame(
Resolution = numeric(),
Number_of_Clusters = numeric()
)
invisible(capture.output({
for(res in resolutions) {
b_cells <- FindClusters(b_cells, resolution = res)
n_clusters <- length(unique(b_cells[[paste0("RNA_snn_res.", res)]][,1]))
clustering_results <- rbind(clustering_results,
data.frame(Resolution = res,
Number_of_Clusters = n_clusters))
}
}))
Idents(b_cells) <- b_cells$RNA_snn_res.0.1
b_cells$bcell_clusters <- Idents(b_cells)
# Display results table
knitr::kable(clustering_results,
col.names = c("Resolution", "Number of Clusters"),
caption = "B Cell Clustering Results at Different Resolutions")| Resolution | Number of Clusters |
|---|---|
| 0.05 | 4 |
| 0.10 | 6 |
| 0.20 | 10 |
| 0.30 | 12 |
| 0.50 | 14 |
| 0.80 | 17 |
2.3 B Cell Cluster Visualization
The plots show the identified B cell subpopulations and how they are distributed across experimental conditions.
2.4 B Cell Subpopulation Characterization
To identify distinct B cell functional states, differential expression analysis was performed to find cluster-specific marker genes. These markers were then used to annotate clusters into biologically meaningful B cell subtypes.
| Cluster | Top 3 Markers | Log2FC Range |
|---|---|---|
| 0 | Vpreb3, Fcer2a, Chchd10 | 0.7-0.73 |
| 1 | Ifit3, Usp18, Ifit2 | 4.41-6.12 |
| 2 | Cd80, Tbc1d9, Ctla4 | 4.73-5.07 |
| 3 | Themis, Icos, Gm2682 | 8.63-8.73 |
| 4 | Camk1d, St6galnac3, Zfp407 | 1.99-2.79 |
| 5 | Tnfsf13, Ifitm6, Ifitm1 | 10.33-10.55 |
2.5 Annotation Strategy: From Automated to Manual Curation
Initial attempts using automated reference-based annotation tools (SingleR, Azimuth) and predefined gene signature matching yielded inconsistent results with poor cluster resolution. Therefore, a manual curation approach was adopted using cluster-specific top markers and established B cell subtype signatures.
2.6 Manual Marker Visualization
Key B cell subtype markers were visualized to assess cluster identity based on spatial expression patterns in the UMAP embedding.
2.6.1 Final Manual Annotation
Based on the spatial expression patterns of cluster-specific markers and known B cell biology, clusters were manually annotated into functionally distinct subtypes.
| Cluster | Manual Annotation | Cell Count | Percentage |
|---|---|---|---|
| 0 | Transitional B cells | 24292 | 90.3 |
| 1 | Interferon-stimulated B cells | 1046 | 3.9 |
| 2 | Activated B cells | 678 | 2.5 |
| 3 | Alternative activated B cells | 449 | 1.7 |
| 4 | Immature B cells | 269 | 1.0 |
| 5 | Type I IFN B cells | 180 | 0.7 |
2.7 Download Loupe Browser File
A .cloupe file was created from the Seurat object of subsetted re-analyzed B-cells to enable interactive visualization and explorattion in the 10X loupe browser. Download re-clustered annotated B-cells
3 Differential Expression Analysis
3.1 Experimental Comparisons
Differential expression analysis was performed using the pseudobulk methodol, which aggregates cells within each sample before statistical testing. This approach properly accounts for biological replication by converting single-cell data to pseudobulk format by aggregating expression within each sample. This approach treats biological samples (not individual cells) as statistical units.and reduces false positives compared to cell-level testing.
The six comparisons are:
- Treatment effects in aged lymph node: Treatment vs Vehicle in aged mice (lymph node)
- Treatment effects in aged tumor: Treatment vs Vehicle in aged mice (tumor)
- Treatment effects in young lymph node: Treatment vs Vehicle in young mice (lymph node)
- Treatment effects in young tumor: Treatment vs Vehicle in young mice (tumor)
- Age effects in lymph node: Young vs Aged (vehicle-treated, lymph node)
- Age effects in tumor: Young vs Aged (vehicle-treated, tumor)
## Expression matrix dimensions: 21262 132## Pseudobulk samples: 1323.2 Statistical Analysis with DESeq2
DESeq2 was used for differential expression testing on pseudobulk data. This method properly models the negative binomial distribution of RNA-seq count data and controls for multiple testing across genes.
Sample Size Requirements: DESeq2 requires sufficient biological replicates to estimate gene-wise variance accurately. A minimum of 4 total samples are required per comparison, which allows for the missing YT_Vehicle_2 sample while maintaining statistical power.
## Completed 6 of 6 comparisons