Global identification of genes and pathways regulated by Akt during activation of T helper cells

Antibodies and reagents

Biotin-anti-mCD28 (37.51) and biotin-anti-mCD3ε (145-2C11) were obtained from BD-Biosciences (San Jose, CA). Streptavidin was obtained from Invitrogen (Carlsbad, CA) and Akti1/2 was from EMD Biosciences (San Diego, CA). rhIL-2 was obtained through the NIH AIDS Research and Reference Reagent program, Division of AIDS, NIAID, NIH (Cat.#136 from Hoffman-LaRoche, Inc.).

T cells

The D10 T cell line, a fast-growing variant of the D10.G41 murine Th2 T cell clone^17–19 was maintained in RPMI 1640 media (Mediatech, Manassas, VA), supplemented with 10% heat-inactivated bovine growth serum (BGS; Hyclone, Logan, UT), 0.1 mM nonessential amino acids (Lonza, Walkersville, MD), 2 mM l-glutamine, 50 µM 2-ME, 100 U/ml penicillin, 100 µg/ml streptomycin (Mediatech, Manassas, VA) and 25 IU/ml rhIL-2.

RNA extraction and microarray gene expression profiling

D10 T cells were left untreated or pretreated with 10 µM Akti1/2 for 1 h and then stimulated with biotinylated anti-mCD3/CD28 and streptavidin for 0, 2, 6 and 12 hrs. RNA extraction was performed using a commercially available kit (RNeasy, Qiagen, Frederick, MD) according to the manufacturers’ recommendations. RNA quality was confirmed based on a RNA integrity number >8 by use of the Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, CA). The microarray analysis was performed by Genomics and Proteomics Core Laboratories (GPCL) of the University of Pittsburgh, USA. An Illumina mouse RefSeq8 chip was used. Microarray data have been deposited in the GEO database and are accessible through the GEO series accession number GSE45221.

Statistical analysis of gene expression microarray data

To compare the molecular characteristics between different time points, Automated Efficiency Analysis was first performed using 7 transformation methods, 9 normalization methods and 5 tests for differentially expressed genes²⁰. A global normalization method and the J5_Quantile95_None method were applied on each time point. The differentially expressed genes were identified using caGEDA with a reasonable threshold of J5 for each time point²¹. To survey the spectrum of biological functions within genes, which were differentially expressed between different groups, functional classification of the genes were performed using Pathway Express (http://vortex.cs.wayne.edu; a pathway level Impact Analysis as described by Draghici et al., 2007²²). Pathway Express was designed to provide both statistical and biological significance in the indication of which pathways may be affected by the observed changes in gene expression. The results are summarized as Impact scores and p-values. Pathway-Express orders the affected pathways in the decreasing order of their expected importance for the given condition.

Real-time PCR analysis

Quantitative real-time PCR was performed using the ABI Step One Plus Real-time PCR system (Applied Biosystems, Foster City, CA) as described previously³. Amplification was performed on a cDNA amount equivalent to 25 ng total RNA with 1×SYBR green universal PCR Master mix (Applied Biosystems) containing deoxyribonucleotide triphosphates, MgCl₂, AmpliTaq Gold DNA polymerase, and forward and reverse primers. Specific primers for each gene were purchased from SABiosciences (Qiagen, Frederick, MD). Experimental samples and no-template controls were all run in duplicate. The PCR cycling parameters were: 95°C for 10 min, and 40 cycles of 94°C for 15s, 60°C for 1 min. The amount of cDNA in each sample was calculated by the comparative threshold (Ct) method and expressed as 2exp (Ct) using 18S RNA as an internal control. Statistical significance was determined using the Student’s T test. All statistical tests were performed using GraphPad Prism (GraphPad Prism, San Diego, CA).

Enrichment in transcription factor target gene analysis

TFactS was used to predict the activities of transcription factors in our microarray data²³. The lists of up- and down-regulated genes were compared to a list of curated target gene signatures. The nominal p-value (Pval) represents the risk of a false positive for a single test. Since the list of query genes is systematically compared to each target gene signature, a multi-testing condition is required. The e-value (Eval) represents the expected number of false positives for a given nominal value. It is computed using the formula: Eval=Pval*T, where T is the number of tests.

Identification of genes regulated by Akt signaling in activated CD4⁺ T cells

We previously demonstrated that Akt activity was rapidly inhibited in T cells by addition of the allosteric inhibitor Akti1/2, which inhibited phosphorylation of Akt within one minute, an effect that can last as long as twelve hours³. In the present study, microarray analysis was performed at three different time points after CD3/CD28 stimulation to characterize the effects of Akt inhibition on the T cell gene activation program in helper T cells. Thus, D10 T cells (a murine Th2 T cell line) were pre-incubated with 10 μM Akti1/2 or solvent, then stimulated for 2–12 hours. We chose this concentration of inhibitor for two reasons. First, in our recent study we observed good concordance between results obtained with 10 μM Akti1/2 and those obtained with combined siRNA-mediated knock-down of Akt1 and Akt2³. In addition, although 1–5 μM can substantially inhibit Akt activity in different cell types under acute conditions^24,25, at least one study has demonstrated that a higher concentration (10 μM) of Akti1/2 was required for more significant inhibition of Akt substrate phosphorylation over the course of several hours²⁵. This could be related to the fact that full-length Akt is only inhibited approximately 80% by 1 μM (and 90% by 10 μM) Akti1/2 in in vitro kinase assays, as shown in a kinase profiling study by Cohen and colleagues²⁶. After stimulation, mRNA was isolated, which was converted into labeled cDNA for hybridization to Illumina chips for microarray analysis (Figure 1A). A rough analysis of the genes modulated in our study after six or twelve hours of CD3/CD28 stimulation (using the default settings with the GEO2R tool at the GEO database) revealed that of the top 30 genes in each case, seven were modulated to a nearly identical degree in the presence or absence of 10 μM Akti1/2. Thus, we were reasonably confident that Akti1/2, with this cell type, and at the concentration used in our study, did not have widespread, off-target, effects on gene transcription or cell viability.

Figure 1.

Flowchart of experimental outline (A) and data analysis (B), as described further in the text.

We next determined which genes were differentially modulated after T cell receptor (TCR) stimulation using caGEDA with reasonably chosen thresholds for different time points (2 h, 6 h and 12 h). This methodology allows for the capture of a more complete set of differentially modulated genes, which is less dependent on overall expression levels. Further validation and downstream analysis were then performed to confirm some of the differentially expressed genes and to extract functional information from the dataset (Figure 1B). We identified differentially expressed gene sets that were dependent on Akt among the three different time point groups. We compared the gene expression patterns of cells plus or minus addition of Akti1/2. First, we generated two gene lists for each time point. Gene list one represents the genes that were differentially expressed between the unstimulated and CD3/CD28 stimulated group in the absence of Akti1/2. Gene list two represents the genes that were differentially expressed between the unstimulated and CD3/CD28 group in the presence of Akti1/2. When comparing the two gene lists, three different patterns were observed:

1. Genes significantly modulated by CD3/CD28 alone but not modulated in the presence of Akti1/2 (genes in this category showed Akt–dependent expression after T cell activation; column 1, top, in Data File 1–Data File 3 and Supplementary Figure 1).

2. Genes significantly modulated by CD3/CD28 alone but less strikingly modulated in the presence of Akti1/2 (genes in this category showed some dependence on Akt; columns 1–2, middle, in Data File 1–Data File 3 and Supplementary Figure 2).

3. Genes not modulated by CD3/CD28 alone but significantly modulated in the presence of Akti1/2 (genes in this category displayed Akti1/2-specific expression; column 2, bottom, in Data File 1–Data File 3 and Supplementary Figure 3).

By examining multiple time points after stimulation, we were able to obtain a kinetic picture of gene expression ± Akt inhibition. The 6 h Akti1/2 (+) and Akti1/2 (-) comparison showed the highest number of differentially expressed genes, and there were fewer differentially expressed genes after two or twelve hours of TCR/CD28 stimulation. Among these, only the genes that expressed the most consistent differences (either increased or decreased expression) were selected for further analysis. Genes with no known function were excluded.

Our previous work identified several NF-κB target genes that were dependent on Akt after TCR stimulation in T helper cells, including those encoding the cytokines TNF-α, GM-CSF, and IL-10, among others³. Analysis of the microarray data confirmed the dependency of these genes on Akt activation, which inspired confidence in our results. Moreover, expression of the mRNAs encoding many secreted proteins was also decreased by Akt inhibition, including IL-13, IL-5, IL-3 and IL-4 (Figure 2). The protein products of these genes (except IL-3) were examined in our previous paper³, which confirmed similar decreases after Akt inhibition. Our data agrees with Patra et al’s study⁷, which showed that myr-Akt expression in activated CD4⁺ T cells resulted in increased Il-4 and Il-13 expression. In addition we found that expression of Ltb (encoding lymphotoxin β), Areg (encoding amphiregulin) and genes encoding the chemokines CCL1, CCL3 and CCL4 were also affected by Akt inhibition (Figure 2).

Figure 2. Selected Akt-dependent genes differentially expressed between control (0 h) and 2 h, 6 h and 12 h CD3/CD28 stimulation groups.

Relative levels of expression are represented by the J5 score.

Of note, several cell-surface proteins, including CD69, CD52 and CD82 were among the Akt-dependent genes after T cell activation (Figure 2). The CD82 molecule is a type III integral membrane protein with four transmembrane domains and is part of the tetra-span-transmembrane (TST) family, which also includes CD9, CD37, CD53, CD63 and CD81/TAPA-1²⁷. Interestingly, proteins of this family are involved in cell activation. For example, engagement of CD82 can deliver a co-stimulatory signal, similar to CD28, for full T cell activation, leading to strong IL-2 production²⁸. CD52 is a small glycopeptide molecule and tethered to the outer surface of the plasma membrane by a glycosylphosphatidylinositol (GPI)-anchor²⁹. CD52 crosslinking can also provide a co-stimulatory signal that causes the activation of normal human T lymphocytes³⁰ and induction of CD4⁺ regulatory cells³¹. Since Akt was reported to mimic CD28 co-stimulation in a T cell line by synergizing with TCR-induced signals to increase transcription of the IL-2 promoter⁵, activation of these other co-stimulatory molecules might also function through this pathway.

Interestingly, inhibition of Akt resulted in impaired up-regulation of the genes encoding many ribosomal proteins (Figure 3). These included: Rps6 (a major substrate of ribosomal protein kinases³²); Rps8, Rps9 (reported to be activated in various tumors, including colon cancer³³); Rps10, Rps15, Rps24 (mutations in these gene result in Diamond-Blackfan anemia³⁴); Rpl7a (which interacts with a subclass of nuclear receptors and inhibits their ability to activate transcription³⁵); Rplp1 (important for elongation during protein synthesis³⁶). All these proteins belong to either small or large subunits of ribosomes; changes in their expression may contribute to an increase in protein synthesis to accommodate numerous cellular processes involved in T cell activation, in addition to the possible connections to cancer. Importantly, the expression of 18S ribosomal genes, which was used for our qPCR housekeeping gene, did not differ between our experimental groups.

Figure 3. Akt-dependent ribosomal subunit genes differentially expressed between control (0 h) and 2 h, 6 h and 12 h CD3/CD28 stimulation groups.

Relative levels of expression are represented by the J5 score.

Expression validation of selected genes

In order to validate the microarray data and obtain more quantitative data on specific genes, real-time PCR analysis was performed at different time points after CD3/CD28 stimulation ± Akti1/2. Ier3 (also named IEX-1, immediate early response gene X-1), Il13, Ccl1 and Ccl4 were selected for real-time PCR validation because their expression was greatly enhanced after CD3/CD28 stimulation and decreased in the presence of the Akt inhibitor. Of note, the gene expression of Ccl1 and Il3 was rapidly up-regulated by CD3/CD28 stimulation, while the up-regulation of Ier3 and Ccl4 expression by CD3/CD28 stimulation was delayed. Egr1 was also selected as a control gene, which showed no CD3/CD28 dependent increase in gene expression, but rather a sharp decrease. Thus, Akt inhibition could affect different genes that showed diverse kinetics after CD3/CD28 stimulation. At least for these specific genes, the changes in expression, as assessed by real-time PCR, confirmed the microarray results, giving us confidence in the overall quality of the dataset.

Akt is known to directly phosphorylate FOXO3a (including in T cells) and, once phosphorylated, FOXO3a is excluded from the nucleus and becomes transcriptionally inactive³⁷. One of the best characterized FOXO target genes in T cells is Klf2³⁸. For reasons that are unclear, we did not observe modulation of Klf2 expression after CD3/CD28 stimulation, with or without Akti1/2. Among known or suspected FOXO target genes^38,39, several that did display enhanced expression in our microarray experiment in the presence of Akti1/2 were Ctla4, Gadd45, Cebpb and Klf6. The latter was recently identified as a FOXO target gene, which was confirmed by chromatin immunoprecipitation^40,41. Our own real time PCR analysis confirmed that in Akti1/2-treated samples, there was an up-regulation of Klf6, relative to stimulation with anti-CD3/CD28 alone, which by itself resulted in a sharp decline in the Klf6 message (Figure 4).

Figure 4. Quantitative real-time PCR (qPCR) confirmation of the stimulation-dependent expression of selected Akt-dependent genes.

qPCR was used to validate the gene expression of Ier3, Il13, Klf6, Egr1, Ccl1 and Ccl4. The RNA samples were the same as those used for the microarray. Results are presented as relative mRNA expression, compared to the unstimulated control sample, normalized to 18S RNA expression. ***p<0.001, *p<0.05 compared with the control group.

Functional classification of the gene expression signature for Akt inhibition in activated T cells (pathway analysis of Akti1/2-mediated transcriptome changes)

From our previously published array data, we found that at two, six and twelve hours after CD3/CD28 stimulation. Akti1/2 elicited a wide range of effects on expression of numerous genes, including both over- and under-expressed genes. In the present study, genes that showed changes in expression after two, six and twelve hours of stimulation in the presence of Akti1/2 were largely not overlapping and cannot be combined for subsequent analysis. There were 54, 71 and 58 genes dependent on Akt at the two, six and twelve hour time points, respectively, after CD3/CD28 stimulation (Data File 1–Data File 2 and Supplementary Figure 1).

To determine whether our gene expression signature was enriched in specific subsets of genes with known biological functions, bioinformatic functional classification analysis of the genes that were differentially expressed was carried out as described in the Methods. The functional classifications of gene sets are illustrated in Figure 5. We found that genes involved in ribosome, cytokine-cytokine receptor interaction, antigen receptor signaling pathway, hematopoietic cell lineage and asthma were significantly enriched among genes affected by Akt inhibition in the presence of CD3/CD28 stimulation.

Figure 5. Functional classification of Akt dependent genes.

Classification enrichment was determined using Pathway Express. Significance is indicated as –Log (p-value).

Enrichment of transcription factors in target genes

Co-expressed genes are often regulated by common transcription factors. Therefore, we separately analyzed the genes that showed altered expression at 2, 6 and 12 h time points after CD3/CD28 stimulation using TFactS, an algorithm that used a catalog of well-characterized transcription factor targets to predict the activity of transcription factors based on microarray data²³. Enrichment of transcription factors was found at all the time points tested (Figure 6). These included NF-κB family members (RelA, Rel, NF-κB1), Myc, Jun and C/EBP. At the 2 h time point, Myc, NF-κB family members (Nfkb1, RelA and Rel) and Egr1 were significantly enriched. RelA, Nfkb1 and Rel, along with Myc and Egr1 were also significantly regulated after 6 h of CD3/CD28 stimulation. Moreover, other transcription factors, including Sp1, C/EBPB, Ets1, Jun and Nfam1, were also enriched at the 6 h time point. At 12 h after stimulation, however, fewer transcription factors were regulated, with Myc and NF-κB1 still significantly regulated. In summary, the most significantly enriched transcription factor binding sites in the regulatory regions of Akt-dependent genes after CD3 and CD28 stimulation were those of NF-κB family members and Myc.

Figure 6. Transcription factor binding sites enriched in Akt-dependent genes, as predicted by TFactS.

The list of Akt-dependent genes was submitted to TFactS (sign-less) using default settings. Significance of regulated transcription factors was determined with–log10 (e-value).