Gene array and protein expression profiles suggest post-transcriptional regulation during CD8+ T cell differentiation.

Peripheral CD8(+) T cells circulate in a quiescent naive state until they are primed by specific antigen and differentiate into effector cells. In the effector state, CD8(+) T cells acquire cytolytic activity and produce increased levels of cytokines such as interferon-gamma. They also exhibit increased T cell receptor sensitivity, decreased CD28 dependence, and become inhibitable by CTLA-4 and other negative regulatory pathways. We hypothesized that one mechanism by which these two states are regulated is via differential expression of specific genes. To this end, basal gene expression profiles of naive and effector 2C TCR transgenic x RAG2(-/-) CD8(+) T cells were analyzed using Affymetrix arrays representing 11,000 genes. Of the 177 differentially expressed known genes, 68 were expressed at higher levels in effector cells, but 109 were more abundant in naive cells, supporting the notion that the naive state is not passive. Expression of genes related to metabolism, actin cytoskeletal dynamics, and effector function increased with priming, whereas expression of putative anti-proliferative genes decreased. Semiquantitative reverse transcription-PCR was utilized as a secondary validation for selected transcripts, and Western blot analysis was used to examine protein expression for molecules of interest. Surprisingly, for 24 genes examined, 12 showed discordant protein versus mRNA expression. In summary, our study indicates that: 1) not only does the expression of some genes in naive CD8(+) T cells become up-regulated upon priming, but the expression of other genes is down-regulated as well and 2) the complexities of T cell differentiation include regulation at the post-transcriptional level.

Upon exit from the thymus, CD4 and CD8 single-positive T cells are thought to enter a naïve state in which they remain until they encounter specific antigen appropriately presented by antigen-presenting cells. The engagement of TCRs 1 by peptide/major histocompatibility complexes in conjunction with costimulatory ligands initiates signaling events leading to IL-2 production, cell cycle progression, and differentiation into effector cells. The differentiation fate of T cells undergoing activation is also influenced by exogenous cytokines. For CD4 ϩ T cells, IL-12 and IL-4 promote T helper 1 and 2 differentiation, respectively (1)(2)(3). Acquisition of these effector phenotypes appears to require cell cycle progression and correlates with epigenetic modification of lineage-specific gene loci (4,5). For CD8 ϩ T cells, although T cytotoxic 1 and 2 phenotypes can be similarly induced, priming with TCR and CD28 ligation in the absence of cytokines is sufficient to promote differentiation into cytolytic effector cells that produce IFN-␥ (6). Acquisition of lytic activity by CD8 ϩ cells also requires cell cycle progression (6), implying a similar requirement for epigenetic modification for effector cytotoxic lymphocyte-specific gene expression.
In addition to the development of cytolytic activity and cytokine-producing capacity, effector CD8 ϩ T cells exhibit several additional functional alterations compared with naïve CD8 ϩ T cells. Effector cells exhibit a lower threshold for TCR-mediated signaling (7,8) and display a decreased dependence on CD28 costimulation (9). Phenotypically they up-regulate expression of CD44, down-regulate expression of CD62L, and also change expression of chemokine receptors thought to be important for homing to specific tissues (10,11). Effector but not naïve CD8 ϩ cells become susceptible to inhibition by CTLA-4 (12,13) and to induction of anergy (14) and appear to become inhibitable by other negative regulatory pathways, such as PD-1 (15,16), and killer inhibitory receptors (17,18). Thus, the primed effector CD8 ϩ phenotype is rather complex and extends beyond effector function per se.
Although the naïve T cell state may have previously been viewed as "quiescent" and relatively inert, two lines of recent data have begun to change that perspective. First, several investigators have demonstrated that signaling through the TCR from self peptide/class I major histocompatibility complexes is necessary to maintain the survival of naïve CD8 ϩ T cells in the periphery (19). Thus, the prevention of apoptotic death in naïve cells is an active process. Second, the transcription factor LKLF has been shown to be preferentially expressed in naïve T cells and down-regulated after activation (20 -22). In mice, post-thymic LKLF-deficient T cells fail to enter a naïve resting state and rapidly die (23), suggesting that the naïve state may need to be actively maintained. * This work was supported by Grant R01 AI47919 and a Burroughs Wellcome Fund Clinical Scientist Award in Translational Research. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental information on each differentially expressed gene as determined by the two gene array screens and sequences of gene-specific primers used to verify differential gene expression by RT We hypothesized that many of the functional distinctions between naïve and effector CD8 ϩ T cells may be explained by differential expression of specific genes. To approach this hypothesis, a gene array strategy was pursued using naïve and 5-day-primed 2C TCR transgenic x RAG2 Ϫ/Ϫ CD8 ϩ T cells. Of 11,000 genes represented, ϳ177 known genes were differentially expressed. Counter to expectations, approximately half of the differentially expressed genes were present at decreased levels in effector cells, supporting the notion that the naïve state is not inert. RNA expression of genes related to metabolism, actin cytoskeletal dynamics, and T cell effector function increased with priming. Semiquantitative RT-PCR was utilized as a secondary validation for selected transcripts, and Western blot analysis was used as a tertiary validation for molecules of interest. However, although differential protein expression correlated with differential RNA expression in some instances, this was not observed for many signaling molecules and transcription factors. Comparable RNA but increased protein levels, or decreased RNA with comparable or increased protein levels, were frequently observed. These results suggest that post-transcriptional events, such as regulated translation, protein stability, and post-translational modification, may be common and are likely to contribute to the control of peripheral T cell differentiation.

EXPERIMENTAL PROCEDURES
Cells-2C/RAG2 Ϫ/Ϫ mice have been described (9). Naïve 2C/RAG2 Ϫ/Ϫ CD8 ϩ T cells were isolated using a negative selection protocol. Spleens from 2C/RAG2 Ϫ/Ϫ mice were macerated and washed once with Dulbecco's modified Eagle's medium containing 10% fetal calf serum. CD8 ϩ T cells were enriched using StemSep TM Enrichment Mixture for murine CD8 ϩ T cells (Stem Cell Technologies, Vancouver, British Columbia, Canada) according to the manufacturer's instructions. To generate primed cells, 10 5 freshly purified 2C/RAG2 Ϫ/Ϫ CD8 ϩ T cells were plated with 5 ϫ 10 5 P815.B7-1 cells pretreated with mitomycin C (50 g/ml/ 10 7 cells for 90 min at 37°C; Sigma) in 1.5 ml of Dulbecco's modified Eagle's medium in each well of a 24-well plate. On the fifth day, cells were harvested and subjected to Ficoll-Hypaque centrifugation to remove cell debris. In some experiments, cells primed for a second 4 -5day period were used as indicated.
Cell Counts-At each day of priming, cells grown in a 24-well plate were collected from three wells and stained with trypan blue. The average number of trypan blue-negative cells was calculated. For days 1 and 2, there was the possibility that remaining live P815.B7-1 cells (H-2 d ) would prevent accurate counting of T cells. Therefore, cells were dually stained with PE-anti-CD8 (BD Biosciences) and fluorescein isothiocyanate-anti-K d (BD Biosciences). After the exclusion of propidium iodide-positive cells, the percentage of CD8-positive and K d -positive cells was determined. After day 1 of priming, no viable K d -positive cells remained in the culture. To obtain the number of live T cells, the percentage of CD8 positive cells was then multiplied by the number of trypan-blue-negative cells. Two days after priming, Ͼ98% of live cells stained positive for CD8 (data not shown).
Fluorescence-activated Cell Sorter Analysis-To confirm the naïve and primed states of the cells, they were stained for cell surface markers CD44 (PE-anti-CD44, BD Biosciences) and CD62L (PE-anti-CD62L, BD Biosciences). Potential nonspecific binding was blocked with anti-F c receptor monoclonal antibody 2.4G2. Single-color analysis was carried out using FACScan TM and CellQuest TM software (BD Biosciences, San Jose, CA). Expression of CD44 and CD62L was verified prior to each experiment.
Cytokine Production by ELISA-For cell stimulation, 96-well flatbottom plates (Costar) were coated with varying amounts of anti-CD3 (2C11) and 1 g/ml anti-CD28 (PV-1) (a gift from Carl June, University of Pennsylvania) overnight at 4°C. Supernatant was collected and assessed by ELISA.
Chromium Release Assay-To assess cytolytic activity, 2 ϫ 10 3 51 Crlabeled targets were plated with 2 ϫ 10 5 naïve or effector CD8 ϩ T cells in a 96-well V-bottom plate (ICN Biomedicals, Costa Mesa, CA). After 4 h of incubation at 37°C, 50 l of supernatant was transferred to a LumaPlate-96 (PerkinElmer Life Sciences) and allowed to dry overnight. Plates were then counted using TopCount-NXT (PerkinElmer Life Sciences).
Preparation of RNA for Gene Chip Array Analysis-Total RNA was isolated from purified T cells. Both naïve and primed CD8 ϩ T cells were first subjected to Ficoll-Hypaque to remove dead cells. Cells remaining in the interface were isolated and washed before being lysed in Trizol ® (Invitrogen). Total RNA was then treated with DNase I (Invitrogen) to minimize contamination from genomic DNA. Further purification was carried out using the RNeasy ® Clean-Up protocol (Qiagen, Valencia, CA). RNA was quantitated using the DU-530 spectrophotometer (Beckman).
Gene Chip Array Analysis-Two replicate experiments were performed using independent batches of total RNA. Biotin-labeled in vitro transcripts for hybridization of Affymetrix oligonucleotide arrays were prepared from 5 g of total RNA according to Affymetrix protocols. Briefly, cDNA was made with the SuperScript TM Choice System (Invitrogen) using 1 l of 110 M oligo(dT)/T7 primer (5Ј-GGCCAGTGAA-TTGTAATACGACTCACTATAGGGAGGCGG-(dT) 24 ). The doublestranded cDNA was extracted with phenol:chloroform:isoamyl alcohol, precipitated with 0.5 volumes of 7.5 M ammonium acetate and 2.5 volumes of 100% ethanol. In vitro transcription with biotin-labeled ribonucleotides was performed with the Enzo BioArray High Yield RNA Transcript Labeling Kit (Enzo Diagnostics, New York, NY). Labeled in vitro transcripts were purified over RNeasy ® mini columns (Qiagen, Valencia, CA) according to the manufacturer's instructions. 12.5 g of purified, labeled transcript was used to hybridize with each Affymetrix array according to the manufacturer's instructions.
The average intensities for each probe set were obtained, and the average intensity of each chip was scaled to the arbitrary value of 300 to allow comparisons among the various chips. After the minimum intensity value was set to 50, the ratio of the average probe set intensities between experiments from effector and the corresponding naïve cells was calculated. If a probe set was called absent in the chips from all the experiments, it was excluded from further calculations. The average of the two ratios was calculated, and genes that expressed greater than 3-fold were manually sorted into functional categories. Genes with absolute expression levels less than 100 in all experiments, as well as ESTs and unknown genes, were excluded from further analysis. Visualization of the intensity and ratio was performed using Spotfire ® Array Explorer TM (Spotfire, Somerville, MA) software. A list of differentially expressed genes represented on the microarray is provided in the Supplemental Information at http://www.jbc.org.
RT-PCR-Gene chip array results were confirmed by semiquantitative RT-PCR. 20 g of total RNA was used for first strand cDNA synthesis. Samples were normalized based on equivalent ␤-actin expression by RT-PCR. For each PCR amplification, 1 l of cDNA was used in a 50-l reaction. To ensure that amplification remained within the linear range, 1:5 serial dilutions of cDNA were made. RT-PCR for ␤-actin (25 cycles) was used as a control for mRNA abundance. For other genes, the number of cycles ranged from 25 to 45. Annealing temperatures varied, since the T m of each primer differed, and amplification was carried out at 72°C. RT-PCR for each gene was performed several times using different batches of cDNA. For sequences of genespecific primers, refer to the Supplemental Information at http://www. jbc.org.

RESULTS
Characteristics of Naive and Effector CD8 ϩ T Cells-We used the 2C TCR transgenic model (24) to study the functional differences between naïve and effector CD8 ϩ T cells. These mice were crossed into the RAG2 Ϫ/Ϫ background to ensure a monoclonal population of CD8 ϩ cells (9). Naïve cells were freshly isolated from the spleens of 4-to 6-week-old mice by negative selection. A subset of purified 2C/RAG2 Ϫ/Ϫ T cells was taken to generate effector CD8 ϩ T cells in vitro through a 5-day stimulation with mitomycin C-treated P815.B7-1 cells. As illustrated in Fig. 1A, this stimulation resulted in vigorous thymidine incorporation that peaked on day 3, with an expansion of viable T cell numbers that lagged behind by 24 h. By day 5, the cells re-entered a resting state, at which time they did not proliferate or spontaneously produce cytokines. At later times, without restimulation, the T cells began to die, presumably due to the lack of viable antigen-presenting cells (data not shown). The primed cells used in this study were taken on the fifth day after stimulation.
Primed effector 2C T cells generated in this manner routinely produced the cytokines IL-2 ( Fig. 2A) and IFN-␥ (Fig. 2B) and acquired antigen-specific cytolytic activity (Fig. 2C). Interestingly, the increased IL-2-producing capability of primed cells was reflected by a shift in the dose response to anti-CD3 mAb, whereas the increased IFN-␥ production was reflected predominantly by an increased magnitude of response. These results suggest that at least two molecular mechanisms may account for the greater cytokine-producing capability of primed effector CD8 ϩ T cells.
Gene Array Analysis-RNA from naïve and effector CD8 ϩ T cells was used to hybridize to Affymetrix Mu11k chips containing ϳ11,000 known mouse genes and ESTs as described under "Experimental Procedures." Fig. 3 shows a dot plot of the expression pattern of individual genes in naïve versus effector CD8 ϩ T cells from the first of two independent experiments. Most genes were not significantly expressed in these T cells and thus clustered near the origin of the plot. Most genes that were expressed in 2C T cells were expressed comparably in both the naive and effector populations and thus fell along a line with a slope of one. However, a number of genes was expressed at greater levels in either the effector population (above the diagonal) or in the naive population (below the diagonal). The expression data from two replicate experiments were averaged, and known full-length genes whose levels in each T cell population differed by 3-fold or greater were defined to be differentially expressed as described under "Experimental Procedures." Approximately 177 known genes met these criteria ( Fig. 4 and the Supplemental Information at http://www.jbc. org). RNA expression of 68 known genes increased with priming, while that of 109 known genes decreased. Although the ratios of RNA expression level for some genes differed between the two experiments, the expression patterns of the vast majority of genes were concordant.
As expected, expression of genes related to T cell effector function and surface receptors associated with activation increased upon priming. Primed effector CD8 ϩ T cells expressed markedly increased levels of CTLA-2␣, Granzyme A, SPI-3, and CCPI genes, all of which encode proteins involved in cytolysis (Fig. 4A). Increased basal expression of cytokine genes was not detected, supporting the notion that the effector T cell population was resting at the time of collection. Genes encoding cell surface receptors known to correlate with naive and effector states changed as expected. The RNA level of CD62L was decreased in effector CD8 ϩ T cells compared with naive cells, whereas the levels of Ly6.2 and CTLA-4 were increased (Fig. 4B).
Of note, mRNA levels of genes involved in generating metabolic potential increased (Fig. 4C). Expression of genes encoding ␣-enolase, pyruvate kinase, triosephosphate isomerase, hexokinase II, 6-phosphofructokinase type C (all participants in the glycolytic pathway) was induced in effector CD8 ϩ T cells. Some housekeeping genes such as Apex, which is involved in base excision repair and also regulates DNA binding activity of transcription factors, and DDOST, which is involved in the transfer of an oligosaccharide onto nascent polypeptides from the endoplasmic reticulum, were down-regulated. Expression of several molecules associated with cytoskeletal regulation was also increased in primed cells. Although the levels of actin transcripts did not change appreciably upon priming, mRNAs for gelsolin, calcyclin, and calpactin were markedly up-regulated (Fig. 4D).
Unexpectedly, more genes encoding signal transduction proteins and transcription factors were down-regulated than were up-regulated upon T cell priming. Among the few up-regulated mRNAs were those encoding the heterotrimeric G ␥ subunit, ERK-1, and molecules involved in inositol phospholipid signaling, such as myoinositol-1 (or -4) monophosphatase (IMP) and an inositol 1,4,5-trisphosphate receptor (p400) (Fig. 4E). However, many genes were found to be expressed at lower levels in primed cells, including those encoding phosphatases (MKP-1, PRL-1, and PP2A) and putative adapters (SOCS-3 and 14 -3-3). Primed effector cells also displayed lower levels of the transcription factor genes Jun B, c-Fos, FOG, Ikaros, c-Jun, Ets-1, and Ets-2. (Fig. 4F). Interestingly, basal expression of the Fos family member Fos B was much increased in effector cells over naïve cells, as was that of GATA-3, a transcription factor required for T helper 2 development.
Expression of genes related to cell cycle regulation was varied, presumably because of a fraction of T cells completing cell cycle progression post-activation. Interestingly, transcripts for HMG-14, a non-histone chromosomal protein that binds nucleosomes, and GST1-Hs, a G 1 to S phase transition protein, were in higher abundance in naïve cells (Fig. 4G). In addition, putative negative regulators of the cell cycle or antiproliferative molecules, such as TSC22, GADD45, MyD116, G0S8, BTG1, and TOB1, were overexpressed in naive cells and decreased following differentiation. These genes are attractive candidates for contributing to the quiescent phenotype of naive T cells.
Confirmation of Gene Array Data by Semiquantitative RT-PCR-A subset of 40 genes of interest was selected to reexamine mRNA expression levels by semiquantitative RT-PCR. Complementary DNA from independent batches of naïve and effector 2C CD8 ϩ T cells were used in these analyses. Naïve and effector cDNA were normalized based on ␤-actin expression as determined by RT-PCR because ␤-actin mRNA and protein expression were similar between naïve and effector cells. Of these 40 genes, 35 gave similar results to those seen by gene array, yielding a concordance rate of 88% (Table I). Thus, results obtained using oligonucleotide arrays are not necessarily identical to those obtained by gene-specific RT-PCR, emphasizing the importance of confirmatory screens.
As stated earlier, our gene array experiments showed higher expression of BTG1 and TOB1, which are attractive candidates for contributing to naive cell quiescence (25,26). These genes are members of the BTG/TOB (recently renamed APRO) (27) family, which consists of at least six family members, any of which could have correlated to the gene chip array oligonucleotides. RNA expression of BTG-1, -2, and -3 and TOB-1 and -2 was examined in our 2C T cells by semiquantitative RT-PCR using gene-specific primers. Interestingly, expression of only TOB1 and TOB2 was found to be down-regulated greater than 5-fold in the primed effector state, making them attractive candidate anti-proliferative genes in naive CD8 ϩ T cells (Fig. 5A).
Based on confirmed down-regulation of TOB1 and TOB2 expression in primed effector cells, RT-PCR analysis was performed to analyze expression of TSC-22 and G0S8/Rgs2 as well (Fig. 5B). TSC-22 was originally described in a screen to identify genes induced by TGF-␤ (28). Expression of TSC-22 has been shown to be down-regulated in transformed versus normal cells (29,30), and introduction of TSC-22 into tumor cells can induce apoptosis (31,32). G0S8 was identified by its overexpression at the G 0 -G 1 transition of the cell cycle (33). As shown in Fig. 5B, expression of both of the genes was diminished in Represented on the gene chip microarrays were 11,000 genes and ESTs. Expression level of a gene was quantified in arbitrary units as determined by the screen. Each gene is represented by a single dot. Genes that were expressed at equivalent levels between naïve and effector 2C CD8 ϩ T cells lie close to the line, whose slope is 1. This dot plot is representative of one of two independent gene array screens.
effector compared with naive T cells, supporting a potential role for TSC-22 and G0S8 in regulating the naive state as well.
Correlating RNA and Protein Expression of a Subset of Genes-Theoretically, post-transcriptional mechanisms could contribute to the regulation of expression of specific gene products. Therefore, we examined expression of selected molecules by Western blot analysis in naive and effector 2C T cells. Expression of additional genes and proteins not represented on FIG. 4. Bar graph representation of gene expression profiles of naïve and effector 2C CD8 ؉ T cells. Expression levels from two independent screens were averaged, and ratios of the mean value are shown. Genes whose expression increased after priming have a positive change in gene expression (red). Genes whose expression decreased after priming have a negative change in gene expression (green). the chips but related to genes of interest were also examined in detail.
The total protein content of primed CD8 ϩ T cells was only ϳ1.5 times that of an equal number of naïve 2C cells (naïve: 198 Ϯ 6.8 g/ml/10 6 cells; effector: 295 Ϯ 20 g/ml/10 6 cells), suggesting that changes in protein levels greater than 1.5-fold on a per-cell basis would be meaningful.
Effector Molecules, Surface Receptors, Metabolic Proteins, and Cytoskeletal Molecules-In accordance with the gene array results, increased expression of granzyme A was seen by RT-PCR in primed 2C cells (data not shown). As an antibody against granzyme A was not available, granzyme B was examined by RT-PCR as well as by Western blot (Fig. 6). As expected, both mRNA and protein for granzyme B were detected at higher levels in primed T cells, consistent with the role of granzymes in granule-mediated cytolysis by effector CD8 ϩ T cells. Expression of CD62L mRNA was confirmed by semiquantitative RT-PCR (Fig. 6). Decreased expression of CD62L protein was also observed by flow cytometry (Fig. 1B). Aldolase C and ␣-enolase expression was also assessed by semiquantitative RT-PCR and Western blotting. Effector cells displayed greater RNA and protein levels of both genes relative to naïve cells. Ca 2ϩ -binding proteins that interact with the cytoskeleton (calcyclin, calpactin, and annexin V) were also significantly increased in effector cells at both the mRNA and the protein level (Fig. 6). Collectively, these results suggest that both gly-colysis and actin cytoskeletal activity may be up-regulated in effector CD8 ϩ T cells, perhaps to meet the needs of specific effector functions.
Signaling Molecules and Transcription Factors-We were interested in signal transduction enzymes that might be differentially expressed in naive and effector T cells as a potential mechanism for the increased TCR sensitivity and decreased CD28 dependence of the latter population. Both Lck and Fyn were expressed comparably at both the mRNA and the protein level in naïve and primed T cells (Fig. 7). Gene array results revealed diminished expression of mRNA encoding 14 -3-3 family members, which are adapter proteins that interact with serine-phosphorylated targets (recently reviewed in Refs. 34, 35). 14 -3-3␤ mRNA was confirmed to be down-regulated by RT-PCR in primed cells. In contrast, protein levels were comparable, suggesting the possibility of post-transcriptional regulation of this molecule.
The increased expression of ERK1 observed by gene array was confirmed by RT-PCR and protein analysis (Fig. 7). This result prompted examination of ERK2 as well as of JNK-pathway enzymes. As shown in Fig. 7, expression of ERK2 and JNK1 protein was comparable in naïve and effector 2C cells, although mRNA appeared to be modestly decreased in the effector cell population. In contrast, JNK2 was comparably expressed at the mRNA level but was substantially up-regulated at the protein level. Similarly, N-Ras, Rac1, and RhoA 1 Y a A subset of genes that were differentially expressed based on two gene array screens was selected for verification by semiquantitative RT-PCR. In each reaction, decreasing amounts of naïve and effector 2C CD8 ϩ T cell cDNA (1:5 serial dilution) were used. RT-PCR was performed on each gene at least three times using distinct batches of cDNA.
b Semiquantitative RT-PCR data concurs or does not concur with gene expression pattern determined by gene array screens.
were all expressed comparably at the mRNA level but were expressed at significantly higher levels in effector cells at the protein level (Fig. 7). The decreased expression of several phosphatase transcripts by gene array analysis was also of interest and was investigated further. MKP-1 is a dual specificity phosphatase that dephosphorylates ERK1/2 (36). The decreased expression of MKP-1 mRNA in primed 2C cells was confirmed by RT-PCR analysis (Fig. 7A). However, very little MKP-1 protein was detected in naïve T cells, whereas substantially greater levels were detected in primed effector cells. Similarly, expression of PP1␣ mRNA was markedly diminished in primed cells, whereas protein levels were similar.
Based on the observation that several enzymes in mitogenactivated protein kinase signaling pathways were up-regulated in effector T cells at the protein level, it was anticipated that basal expression of transcription factor targets also might show distinct patterns of expression. Indeed, basal Fos B expression was approximately equal in primed T cells at the mRNA level, but at the protein level effector cells showed marked up-regulation compared with naïve cells (Fig. 7). In contrast, c-Fos mRNA was decreased in primed cells, with c-Fos protein levels remaining constant. In accordance with the gene array data, mRNA levels of Jun B and Ets-2 were also decreased after priming. However, protein levels were either equivalent or augmented. In summary, the results of these analyses comparing mRNA and protein levels between naive and effector CD8 ϩ T cells suggest that expression of many signaling molecules and transcription factors can be can be regulated at the posttranscriptional level. DISCUSSION The goals of this study were to understand sets of genes that define a naïve or effector CD8 ϩ T cell population and to determine whether differences in gene expression could explain their distinct functional characteristics. Of the ϳ11,000 known genes and ESTs we examined, 177 known genes were expressed at relatively higher levels in either naïve or effector cells. Surprisingly, a greater number of known genes was found to be down-regulated (109 of 177) than up-regulated (68 of 177) upon differentiation into the effector state. This observation counters the notion that naïve cells are inert and supports the concept that the naïve phenotype is actively maintained. The process of maintaining T cell quiescence could be intrinsic to the T cells themselves or could occur by a T cell-extrinsic mechanism.
In contrast to previous gene array studies (21,37,38), we used cells generated from TCR transgenic x RAG2 Ϫ/Ϫ mice, enabling us to examine CD8 ϩ T cells with a single specificity. The use of a clonal T cell population minimizes potential differences in gene expression that can be attributed to variation in activation state. Although effector CD8 ϩ T cells were generated in vitro in response to mitomycin C-treated P815-B7.1 cells, there were no viable P815-B7.1 cells by day 5 of differentiation and every effort was made to eliminate contamination  Fig. 6. C, analysis of protein expression by Western blotting as described in Fig. 6. from other cell types. However, even after further purification by Ficoll-Hypaque centrifugation, some erythrocytes may have been inadvertently included in the preparation of the naïve cells, which could theoretically contribute to certain mRNA species despite lacking nuclei and active transcription. However, potential erythrocyte contamination in the naïve population is unlikely to explain the majority of differentially expressed genes observed.
Despite using equivalent cell numbers, effector CD8 ϩ T cells yielded approximately four times more total RNA than naïve cells. This likely represents a combination of greater ribosomal RNA and to some extent mRNA, although these were not distinguished in our study. However, total protein content in effector cells was increased only 1.5 times over that of naïve cells. This increase in total protein content was proportional to the increase in cell diameter and mean fluorescence of forward scatter (data not shown), suggesting that protein concentration remained comparable.
Upon anti-CD3 and anti-CD28 stimulation, naïve cells produced ample IL-2 but only low amounts of IFN-␥. In contrast, effector cells produced high levels of both IL-2 and IFN-␥. Thus, IFN-␥ was produced at an increased magnitude following differentiation. In contrast, effector cells required 10-fold less anti-CD3 mAb to produce the same level of IL-2 as naïve cells, indicating a shift in the TCR dose response. Thus, the greater IL-2 production seen with effector cells is reflected by increased TCR sensitivity. These results suggest that two molecular mechanisms likely explain the increased functional responsiveness of CD8 ϩ effector cells. For IFN-␥ (and likely for cytolytic activity as well), epigenetic modification of specific gene loci probably contributes to the increased magnitude of gene expression following differentiation. Chromatin remodeling at the IFN-␥ locus has been described in antigen-experienced CD4 ϩ T cells (4,5). However, the mechanism to explain increased TCR sensitivity in effector cells is not understood and could be explained by changes in expression of signaling molecules that are responsive to TCR engagement.
Based on initial gene array data, we were attracted to the hypothesis that down-regulation of protein phosphatases could explain increased TCR sensitivity in differentiated effector cells. PP1␣ and MKP-1 mRNA were all expressed at lower levels following differentiation. However, analysis of protein expression revealed comparable or increased expression of MKP-1 and PP1␣ by Western blotting, demonstrating discordance between mRNA and protein data. These results are consistent with the known property of phosphatase-kinase interactions in regulating stability of the phosphatase (39,40) and imply that mRNA levels of phosphatases do not necessarily predict the quantity of protein present due to post-translational modification. Several transcription factors (c-Fos, JunB, and Ets-2) were also found to be down-regulated in effector compared with naive CD8 ϩ T cells at the mRNA level but were expressed comparably at the protein level. It is conceivable that protein stability is similarly prolonged for these factors due to distinct intermolecular interactions or other post-translational effects.
As controls in our experiments, protein levels were examined for a panel of signaling molecules that were comparably expressed between naive and effector CD8 ϩ T cells. This included JNK2, N-Ras, Rac 1, and Rho A. Surprisingly, despite comparable mRNA levels confirmed by RT-PCR, expression of these molecules was found to be markedly up-regulated at the protein level. For JNK2, our result is consistent with those of the Flavell group who reported that JNK activation was detected late during CD4 ϩ T cell differentiation (41). Our findings un-derscore the complex regulation of expression and activity of specific genes through increased translation, post-translational modification, or protein stability. These results do suggest that increased expression of one or more of these signaling molecules at the protein level could conceivably contribute to the increased TCR sensitivity as seen in the effector state.
We observed that several putative antiproliferative genes were more highly expressed in naïve cells, including TSC-22, TOB1, TOB2, and G0S8. It is possible that the functions of these genes, alone or in combination, may contribute to the quiescence of naïve cells. TSC-22, a gene whose mRNA expression was found to be decreased in our effector cells as well as two other studies comparing naïve and activated lymphocytes (21,22), binds DNA and is thought to be a transcriptional regulator. Target genes of TSC-22 are unknown to date; however, it has been shown that it is a transcriptional repressor alone or in conjunction with a closely related homolog, THG-1 (30). Consistent with our observed down-regulation of TOB1 and TOB2 following differentiation out of the naive state, Tob1 was recently shown to inhibit T cell proliferation and cytokine production (42). G0S8/RGS2 has been shown to be an immediately early response gene induced upon T cell activation (33,43). It is involved in inhibiting G q␣ function, which could interfere with cell surface signaling to PLC␤ and inositol lipid signaling (44). Although peripheral T cells lacking Rgs2 have been shown to exhibit a defect in proliferation (45), whether overexpression of Rgs2 also limits cell cycle progression in lymphocytes has not been examined.
Transcripts of genes involved in metabolism were present at higher levels in effector cells relative to naïve cells. In particular, we observed up-regulation of eight of 10 glycolytic enzymes, suggesting that glycolysis may play a significant role in the function of effector T cells. Preliminary results have revealed that glucose uptake and glycolytic rate are markedly increased in resting effector compared with resting naive CD8 ϩ T cells (data not shown). It has been shown that proliferating cultured rat thymocytes rely heavily on glycolysis rather than aerobic respiration even in the presence of normal O 2 levels (46 -48). It is also conceivable that effector cells utilize glycolysis to avoid byproducts of oxidative phosphorylation, such as reactive oxygen species, which can inhibit proliferation (49) or lead to cellular apoptosis (50).
A final category of genes of interest up-regulated in CD8 ϩ effector cells is that of cytoskeletal molecules. Calcyclin, calpactin, vimentin, and gelsolin were notably expressed at higher levels following differentiation. Preliminary results have indicated that several additional molecules that regulate actin cytoskeletal dynamics, including Vav, Wiscott-Aldrich Syndrome protein, and LIM kinase 1 are also expressed at increased levels in effector cells (data not shown). There is some debate as to whether the role of a dynamic actin cytoskeleton in receptor segregation and immune synapse formation is primarily to support signal transduction amplification or rather to promote directed execution of T cell effector function (51,52). The observed increase in cytoskeletal proteins in effector cells supports the latter hypothesis and should be testable by examining the role of individual cytoskeletal regulators in TCR signaling versus cytolysis.
In summary, our study illustrates the complexity of naïve T cell regulation, emphasizing the need to examine both mRNA and protein levels to make conclusions regarding the expression of particular molecules in a specific cell population. Further studies are necessary to determine the role of naive T cell genes in limiting T cell activation, as well as the function of augmented glycolysis, cytoskeletal dynamics, and signaling proteins in effector T cell activities.