Technical details
This page provides the minimal technical details required to interpret the gene expression data and the figures provided on this data portal. For complete experimental details, please refer to the original manuscript (Wu et al., Immunity, in press).
Cells and conditions
The gene expression data provided in this portal is from six mouse T lymphocyte subsets:
- Flow sorted spleen CD4+CD44lo cells: these are referred to as naive CD4 T cells
- Flow sorted spleen CD4+CD44high cells: these are referred to as memory CD4 T cells
- Flow sorted spleen CD8+CD44lo cells: these are referred to as naive CD8 T cells
- Flow sorted spleen CD8+CD44high cells: these are referred to as naive CD8 T cells
- Flow sorted thymus CD4+ single positive cells: these are referred to CD4 single positive thymocytes (CD4SP)
- Flow sorted thymus CD8+ single positive cells: these are referred to CD8 single positive thymocytes (CD8SP)
Four biological replicate samples of each of the six T cell subsets were collected from WT (C57BL/6) and homozygous Thunder mutant (thu/thu) mice for microarray analysis, giving 48 microarray samples in total.
Given that the thymic single positive cells (CD4SP and CD8SP) generally had distinct gene expression profiles compared to the splenic T cell subsets, these conditions were not included in the exon level heatmap visualizations or in the analysis for genes with differential expression between conditions. Data from CD4SP and CD8SP subsets are plotted alongside the splenic T cell subsets in the gene-level bar plots, however.
Arrays and normalization
RNA samples were prepared from the 48 T cell samples described above and hybridized to Affymetrix Mouse Exon 1.0 ST Arrays. For details on RNA preparation, QC, or hybridization, please refer to the original manuscript. The microarrays were normalized at the exon level using PLIER in the Affymetrix Power Tools software package. Gene-level normalization of the microarrays was carried out using RMA and the
BrainArray ENTREZG Version 11 custom CDF file.
Gene-level barplots
Gene-level expression information for individual genes is visualized using three gene-level barplots:
- (TOP): The top figure plots the average log2(expression) value in each of the conditions. This plot shows all gene expression variation (genotype dependent, memory status dependent, T cell subset dependent) and may be difficult to interpret.
- (MIDDLE): The middle plot shows the average log2(fold change) comparing Thunder mutant (thu/thu) T cells to WT T cells. Values greater than zero indicate increased expression in the Thunder mutant cells compared to WT, while values less than zero indicate decreased expression in Thunder mutant cells. Fold changes are computed by comparing Thunder mutant gene expression in each T cell subset to WT gene expression in the same T cell subset (ie Thu CD4 navie vs. WT CD4 naive, Thu CD8 memory vs. WT CD8 memory, etc.) and thus makes consistent Thunder-dependent expression patterns easy to identify. All gene expression variation due to memory status or T cell subset is cancelled out.
- (BOTTOM): The bottom plot shows the average log2(fold change) comparing memory (CD44high) T cells to navie (CD44lo) T cells. Values greater than zero indicate increased expression in the memory cells compared to naive, while values less than zero indicate decreased expression in memory cells. Fold changes are computed by comparing memory cell gene expression in each T cell subset to naive cell gene expression in the same T cell subset (ie memory WT CD4 vs. naive WT CD4 naive, memory Thu CD8 vs. naive Thu CD8, etc.) and thus makes consistent memory status-dependent expression patterns easy to identify. All gene expression variation due to genotype or T cell subset is cancelled out.
For all barplots, the bar heights give the mean log2(expression level) (or log2(Fold change)) computed using all four biological replicates. The error bars represent +/- the standard deviation computed from all four biological replicates for each condition. Expression in WT T cells is given by bars with horizontal stripes, expression in Thunder mutant (thu/thu) T cells is given by bars with vertical stripes. Expression in CD4 T cells is given by blue shaded bars, expression in CD8 T cells is given by red shaded bars. Expression in single positive thymocytes is given by the lightest bars with few stripes. Expression in naive (CD44lo) splenic T cells is given by bars with intermediate shading. Expression in memory (CD44high) splenic T cells is given by bars with the darkest shading and most stripes.
Exon-level heatmaps
Exon-level heatmaps are a convenient way for visualizing complex exon-level gene expression information across multiple conditions, and integrating it with known transcript structure information. In the exon-level heatmaps provided in this data portal, each row represents a different exon-specific probeset from the microarray. Each column represents a different biological replicate for the different T cell subsets investigated. Exons are arranged 5' (top) to 3' (bottom). Probesets that interrogate the same exon are aligned 5' to 3' according to the sequences they target. Colored boxes (black, red, or green) contain the log2(fold changes) compared to the relevant reference for that exon (different references discussed below). Log2(fold changes) that are less than 0.5 and greater than -0.5 are colored black. Log2(fold changes) that are greater than 0.5 and less than 2.0 are scaled red according to their magnitude, with log2(fold changes) that are greater than 2.0 being saturated red. Log2(fold changes) that are less than -0.5 and greater than -2.0 are scaled green according to their magnitude, with log2(fold changes) that are less than -2.0 being saturated green. White boxes give raw log2(expression) values in the relevant reference condition. The right hand side of the heatmaps gives Ensembl transcript structure information. Each blue column corresponds to a different Ensembl transcript for the gene of interest. Blue boxes indicate that the transcript includes the particular exon, while empty boxes indicate the absence of the exon from the transcript. Coupling transcript structure information with exon-level expression information makes it possible to quickly determine whether if the observed patterns of alternative exon expression are consistent with known transcript structures. Ptprc (CD45)
is an example - the three exons that have enhanced expression in the Thunder mutant compared to WT correspond to three exons that are spliced out of 2 out of 5 known Ptprc transcripts.
Exon-level expression information for individual genes is visualized using three exon-level heatmaps that use different references for computing fold changes:
- Exon expression change, compared with exon average: In this heatmap, log2(fold changes) for each exon are computed by subtracting off the overall average expression value for that exon across all conditions. Conditions that are colored green have lower than average expression for that exon, while conditions colored red have higher than average exon expression. The fold changes thus retain all gene expression variation (genotype dependent, memory status dependent, T cell subset dependent) and may be difficult to interpret.
- Exon expression change in Thunder mutant: In this heatmap, log2(fold changes) for each exon are computed by subtracting the median expression value for that exon in WT cells from a particular cell type from the expression values for that exon in Thunder mutant cells from the same cell type. All WT exon expression levels are thus set as references in this visualization. Conditions that are colored red indicate increased expression of the exon in the Thunder mutant cells compared to WT, while conditions colored green indicate decreased expression of the exon in Thunder mutant cells. This visualization makes consistent Thunder-dependent exon expression patterns easy to identify. All gene expression variation due to memory status or T cell subset is cancelled out in the computation of the fold changes.
- Exon expression change: Memory vs. Naive: In this heatmap, log2(fold changes) for each exon are computed by subtracting the median expression value for that exon in naive cells from a particular cell type from the expression values for that exon in memory cells from the same cell type. All naive expression levels are thus set as references in this visualization. Conditions that are colored red indicate increased expression of the exon in memory cells compared to naive, while conditions colored green indicate decreased expression of the exon in memory cells. This visualization makes consistent memory status-dependent exon expression patterns easy to identify. All gene expression variation due to genotype or T cell subset is cancelled out in the computation of the fold changes.
Identification of interesting genes at the gene level
Complete details are provided on the interesting genes page .
|
