Transcriptomes of the B and T lineages compared by multiplatform microarray profiling

Abstract

T and B lymphocytes are developmentally and functionally related cells of the immune system, representing the two major branches of adaptive immunity. Although originating from a common precursor, they play very different roles: T cells contribute to and drive cell-mediated immunity, whereas B cells secrete Abs. Because of their functional importance and well-characterized differentiation pathways, T and B lymphocytes are ideal cell types with which to understand how functional differences are encoded at the transcriptional level. Although there has been a great deal of interest in defining regulatory factors that distinguish T and B cells, a truly genomewide view of the transcriptional differences between these two cells types has not yet been taken. To obtain a more global perspective of the transcriptional differences underlying T and B cells, we exploited the statistical power of combinatorial profiling on different microarray platforms, and the breadth of the Immunological Genome Project gene expression database, to generate robust differential signatures. We find that differential expression in T and B cells is pervasive, with the majority of transcripts showing statistically significant differences. These distinguishing characteristics are acquired gradually, through all stages of B and T differentiation. In contrast, very few T versus B signature genes are uniquely expressed in these lineages, but are shared throughout immune cells.

Figure 1. Defining T vs. B differential signatures

A): RNA preparations from CD4+ cells and CD19 B cells were profiled on Affymetrix and Illumina whole-genome microarrays, and the T vs. B FoldChange was calculated for the same genes on both microarrays. B): Consensus T vs. B cell expression ratios were calculated by combining information from four different microarray platforms, and a false-discovery rate on these FoldChange values was estimated by repeated randomization of the datasets, testing how often the FoldChange observed for a given gene could be observed by chance. The threshold FoldChange values which reached statistical significance were estimated at <0.88 and > 1.11, for a genome-wide p=0.05 C): Datasets from several populations of mature T cells (whole CD3⁺CD4⁺ splenocytes, naïve CD4⁺ and CD8⁺ cells from spleen and LN, CD44^hi CD4⁺ and CD8⁺ splenocytes) and B cells (whole CD19⁺ splenocytes, mature bone marrow “Fraction F” cells, T3 splenic subset, follicular B from spleen and peritoneal cavity, marginal zone B), all profiled on the Affymetrix MuGeneST1.0 platform, were analyzed in combination to generate consensus measures of differential expression. The aggregate T vs. B expression ratios are plotted against the Student's t-test p-value. “Top 100” signature genes for B and T are outlined. D): Comparison of T/B FoldChanges determined from the multiplatform data (black dots) or from the combined ImmGen datasets (grey dots).

Figure 2. The transcripts that most distinguish T and B cells are expressed throughout immune cells

Heat-map representations of the expression of the “Top 100” T cell signature genes across the immune cell populations contained in the ImmGen database. Genes are arranged by hierarchical clustering.

Figure 3. The transcripts that most distinguish T and B cells...continued

Heat-map representations of the expression of the “Top 100” B cell signature genes across the immune cell populations contained in the ImmGen database. Genes are arranged by hierarchical clustering.

Figure 4. The transcripts that most distinguish T and B cells are acquired, or lost, in stages throughout differentiation

Heat-map representations of the expression of the “Top 100” T cell of B cell genes during T cell differentiation in the thymus (A,C) or during B cell differentiation in the bone marrow (B,D). Cell-types have been arranged according to their sequence during differentiation and genes were clustered using an ordering algorithm according to the stage at which they are expressed. E): Population plot in which cell-types have been positioned according to their “T-ness” and “B-ness”, defined from the aggregate expression values of genes most differentially expressed in mature B and T cells (see Materials and Methods).

Figure 5. Partial sharing of co-regulated gene clusters within T cell differentiation and outside the T cell lineage

To determine which transcripts exhibit coordinated expression, as a reflection of possible shared regulatory mechanisms, pair-wise correlation coefficients were calculated for all transcripts of the “Top 200” T cell signature genes, within all ImmGen datasets except for T and B cells (“nonT/nonB”) or within the T cell differentiation datasets. As a reference, the same coefficients were calculated on a set of 2000 transcripts picked at random. A): distribution of the correlation coefficients; note that there is a very significant skewing of the distribution of correlation coefficients between T signature genes in the T-differentiation datagroup (top left), far less marked within the nonT/nonB datagroup (top right). B): Scatter plot comparison of all pair-wise correlations between T signature genes within the nonT/nonB (X-axis) or T-differentiation (Y-axis) datagroups; to avoid artifacts due to the different sizes and composition of the nonT/nonB and T-differentiation datasets, the primary correlation coefficients were transformed to a z-score by reference to the mean and standard deviation of the correlation coefficients for the randomly picked reference gene-set. Note that the majority of transcript pairs that show strong correlation within the T-differentiation datagroup (z-score > 2) show no correlation within the nonT/nonB populations (z-scores distributed around 0), although there is a distinct “shoulder” of gene pairs that do show some correlation across both conditions (top right of the plot). C): A k-means clustering algorithm was used to partition T-signature genes into distinct clusters based on their correlation within the T-differentiation datagroup. Transcript levels for representative clusters are shown as a heat-map for the nonT/nonB (left) and T-differentiation (right) datagroups. A few clusters showed consistent expression across both datagroups (e.g. Cluster 1, top, primarily reflecting shared expression with NK cells), while many were only co-regulated within the T-differentiation datagroup.