T lymphocyte activation is central to the immune response, and much is known about the associated gene expression and regulation (6). Earlier work demonstrated that APA is regulated in activated T cells (4, 7, 8). In order to provide a better understanding of how APA is used in an active gene expression program, we performed a global analysis of alternative 3′ UTR isoforms during T cell activation. We developed a method for probe-level alternative transcript analysis (PLATA) that uses variation in hybridization of individual oligonucleotide probes on Affymetrix Mouse Exon 1.0 ST microarrays. This platform was used to compare transcripts in resting primary murine T cells expressing CD4 (CD4⁺) to cells stimulated through the T cell antigen receptor (TCR) for 6 or 48 hours (Figs. S1-S4, Tables S1-S4).

APA occurs in both a splicing-independent form (multiple PAS in a terminal exon – here called tandem UTRs) and a splicing-dependent form (mutually exclusive terminal exons - here called 3′ exon switching), diagrammed in Figs. 1A, 1B, respectively (4). These classes were considered separately in stimulated and resting cells. We first compared the relative expression of the common 3′ UTR regions (between stop codon and first or ‘proximal’ PAS) and of the extended 3′ UTR regions (between proximal PAS and second, ‘distal’ PAS) of 1,190 genes with EST-supported tandem UTRs and then compared the relative expression of mutually exclusive 3′ exons for 1,991 genes (Figs. 1A-B, S5-6, Tables S5-6). Patterns of splicing-independent and splicing-dependent APA events in stimulated cells were markedly different. Statistically significant 3′ exon switching events were observed at both early (n = 89; Fig. S6) and late (n = 203; Fig. 1B, Table S6) stages of activation, but statistically significant changes in tandem UTR events were observed only at late stages of activation (n = 99; Figs. 1A, S6). Proximal to distal and distal to proximal switching of 3′ exons were similarly represented (Figs. 1B, S6 insets; 143 versus 149 overall). However, 86% of changes in genes with tandem UTRs occurred in the direction of decreased expression of the extended 3′ UTR region (Fig. 1A inset). Changes in isoform expression estimated by PLATA were confirmed by qRT-PCR (Fig. S4D, Table S7). Genes with significantly reduced relative expression of extended 3′ UTR regions had an average 2.2-fold reduction in expression of extended region probes relative to common region probes. Since probes querying the common 3′ UTR region assess the aggregate expression of both isoforms, this value underestimates changes in APA isoform expression. Increased relative expression of shorter isoforms derived from tandem UTR-containing genes was independently validated by Northern analysis (Fig. S4E). A comprehensive analysis of all genes with tandem UTRs expressed at 48 hours revealed a highly significant (P < 6.6×10^-19, t-test) reduction in extended 3′ UTR expression at late stages of T lymphocyte activation (Fig. 1C, Fig. S6). This reduction was not associated with significant changes in median gene expression (assessed by common region probes), although median expression of genes with highly significant individual events was slightly increased in activated cells (Fig. 1D). To determine whether the apparent global decrease in relative expression of extended UTR isoforms was unique to late stages of murine CD4⁺ T lymphocyte activation, we developed a tandem UTR length index (TLI) that assesses aggregate expression of extended 3′ UTR regions relative to overall gene expression levels (Fig. S7, Table S9).

In our mouse CD4⁺ T cell data, TLI changed very little at 6 hours post-stimulation but showed a marked decrease at 48 hours (Fig. 2A). Analysis of array data from human T lymphocytes stimulated through the TCR revealed that decreased relative expression of distal PAS isoforms in tandem UTR genes is conserved between mouse and human (Fig. 2B). TLI also decreased following stimulation of human B cells with anti-CD40 and interleukin-4, and stimulation of human monocytes with lipopolysaccharide and interferon-gamma (Fig. 2B, Fig. S8A). Thus, a global decrease in relative expression of distal 3′ UTRs occurs following activation of several immune cell types with a variety of activating/proliferative stimuli.

Analysis of a broader panel of human and mouse tissues revealed tissue-biased usage of APA isoforms in agreement with previous studies (3, 5) (Fig. S8B). TLI values from cell lines were consistently lower than for the normal tissue type from which the line was derived (Fig. 2C). Considering that cultured cell lines have been selected for rapid proliferation, and that activation of hematopoietic cells is often associated with a dramatic increase in proliferation rate, these observations suggested that decreased relative expression of distal PAS isoforms might be generally associated with cellular proliferation. To address this hypothesis, a gene expression-based measure of cellular proliferation, the proliferation index (PI) was developed (SOM). Analysis of array data over hundreds of individual samples revealed a strong negative correlation between PI and TLI, indicating that reduced relative expression of distal PAS isoforms is broadly associated with cellular proliferation (Fig. 2D).

To assay how different 3′ UTR isoforms influence protein expression, common and full-length 3′ UTRs from several tandem APA genes were fused to a luciferase reporter. To isolate the effects of the UTR on protein expression from efficiency of polyadenylation, a heterologous SV40 PAS was used to terminate both transcripts. Use of the SV40 PAS in the full-length transcript was ensured by deletion of the sequence encoding the proximal PAS (Fig. S9A). In all cases tested, the full-length UTR reporter yielded significantly lower luciferase activity in stimulated primary T cells than the construct containing only the common UTR region (P < 0.01 by two-tailed t-test; Fig. 3A). Thus, sequences within the extended UTR regions of tandem UTR genes commonly reduced protein output, likely by affecting mRNA stability and/or translation.

The extended 3′ UTR region of one tested gene, Hip2, contains conserved seed matches to miR-21 and miR-155, both of which are expressed in activated mouse T cells (9, 10). Overall Hip2 transcript expression was very similar between naïve and activated T lymphocytes; however upon activation relative expression of the extended UTR region decreased while protein levels increased substantially (Fig. 3B, C). This pattern is consistent with a model in which elimination of the extended region of the Hip2 3′ UTR in activated T cells releases this transcript from negative regulation by mir-21 and mir-155. In luciferase assays, whereas deletion of either miRNA seed match in the full-length construct resulted in a modest increase in expression, elimination of both sites restored luciferase expression to levels indistinguishable from those observed with the shorter UTR (Fig. 3D). We concluded that in activated T lymphocytes the extended region of the Hip2 3′ UTR confers reduced protein expression, and that the mir-21 and mir-155 seed matches within this region contribute to this regulation.

In agreement with previous studies (11), we found that extended UTR regions contained slightly higher numbers of conserved miRNA seed matches than common regions (Fig. 3E), but no single target site was significantly enriched in extended regions. However, the set of genes containing conserved seed matches to proliferation-associated T cell miRNAs (mir-155 and the oncogenic mir-17-92 cluster) exhibited significantly reduced expression of the distal PAS isoform relative to non-targets (P < 0.05; Fig. 3F) in stimulated T cells, where expression of these miRNAs is increased ((9) and Fig. S10)). Furthermore, tandem APA genes whose overall transcript expression increased following activation were more likely to be targeted exclusively in their extended UTR region by miRNAs than tandem APA genes whose expression decreased (Fig. 3G, P = 3.2e-6, t-test). Taken together, these data are consistent with a model in which reduced expression of distal PAS isoforms in genes upregulated following activation results in reduced potential for targeting by proliferation-associated negative regulatory factors.

Comparisons of tissue-specific EST libraries have been used to catalog genes with APA and to identify motifs enriched around the PASs of these genes (3, 5, 11, 12). However, to our knowledge a widespread pattern of directed changes in 3′ UTR isoform expression has not been previously described as part of a gene expression program. The observed pattern of decreased relative expression of longer mRNA isoforms could result from a systematic shift in APA or from preferential destabilization of longer mRNA isoforms. The available evidence tends to support regulated APA as being more important, e.g., a stability-based mechanism would predict a significant decrease in mean expression of genes with altered 3′ UTR isoform expression following activation, but no general decrease was observed (Fig. 1F), and Northern blots (Fig. S4) generally showed increased expression of the shorter isoform, which is more consistent with APA regulation. However, our data do not conclusively distinguish between these two mechanisms.

Stimulation of B lymphocytes, T cells, and monocytes results in increased protein expression of the general polyadenylation factor Cstf-64, which plays a role in increased usage of “weaker” upstream PAS in certain 3′ exon switching events in these cell types (7, 13-16). Since Cstf-64 protein expression is also increased as cultured fibroblasts initiate proliferative programs (13), one could imagine that this factor might function as a “master regulator” of PAS selection controlling all of the events that we observe. However, given that Igμ heavy chain PAS selection can be functionally uncoupled from both proliferation and Cstf-64 protein levels in the B cell model (13), and that tandem UTR events respond very differently from 3′ exon switching events in this study, it is likely that additional factors are involved (17-20).

Supporting Onl