The aim of the study was to investigate the activation of human NK cells by IL2 through analyzing the global gene expression at different time points (0, 2, 8 and 24 hours) after culture with the cytokine IL2 at 100 IU/ml. NK cells with the CD56⁺/CD16⁺ and CD3^- phenotype were negatively selected by immunomagnetic beads and re-examined by flow-cytometry to ensure greater than 90% purity (Figure 1). Equivalent amount of total RNA from NK cells derived from 3 or 4 different donors for each time point was pooled, amplified and labeled before duplicate hybridizations was performed on spotted (pre-synthesized oligonucleotide) microarrays (additional file 1). To validate the gene expression data, a different set of experiments with 6 other individual donors was performed. RNA was similarly extracted and pooled from 4 different donors, amplified and labled according to the manufacturer's instruction (GeneChipU133plus2^®, Affymetrix Inc, CA). The expression data generated from microarray experiments were uploaded in BRB-ArrayTool1718 to obtain the functional annotations of the geneID or probe sets and their normalized data were used for further analysis. Of the 54,676 probe sets represented on GeneChipU133plus2^® with 20,585 unique genes, 62 % (~12,750) of the genes were also represented on the spotted microarray. Only pathways showing a similar pattern of expression on the two platforms were selected for further analysis and a few of the genes were also validated by RT-PCR (Figure 5).

The resting NK cell signature from the spotted microarray was defined as differentially expressed genes, that had ≥ 2 fold higher expressions in resting NK cells compared to the lymphoid RNA standard and also ≥ 2 fold higher expression than tonsillar cells and resting CD8⁺ T cells. Thus, 1027 transcripts were included in the resting NK cell signature on the spotted array platform (additional file 2). The NK cell signature derived from the GeneChipU133plus 2^® was based on similar comparisons, with the modification that, tonsil profile was replaced with an universal RNA standard (Stratagene Inc, CA). In this format, 1133 unique transcripts were overexpressed in resting NK cells (additional file 3). Comparison of the NK cell signatures from the two platforms resulted in a common set of 164 genes (Table 1). The genes included inhibitory NK cell surface molecules, activating NK cell surface molecules as well as cytotoxic mediators. Many genes encoding chemokines, cytokines and their receptors and genes involved in secretory functions were also expressed at high levels in resting cells. We also examined a curated resting NK signature from the literature (Signature Data-base) 19 derived from the Affymetrix platform 20. Thirty-three of sixty genes (55%) were found in our Affymetrix derived signature, (additional file 4), demonstrating consistency within the individual platform.

Self organizing maps (SOM) generated from spotted microarray data identified the differentially expressed genes that had at least a 2-fold alteration in expression level between the time points analyzed (additional file 5). The expression of the selected pathways was examined and cross-validated on both platforms. Additional genes on Affymetrix microarray platform (GeneChipU133plus2^®) belonging to functional groups that were determined to be differentially expressed were included for more comprehensive interpretation.

Four different genes of the PI3K family were upregulated as were the three isoforms of AKT kinases and simultaneously there was decreased expression of the AKT target genes (FOXO1A, -3A and p27). Together with the facts that activated AKT promotes cell survival through 1) phosphorylation dependent dissociation of the BAD/BCLX_L complex 48 (Bad upregulated) and 2) activation of NF-κB through phosphorylation of IKK-β 49 (IKKβ upregulated) IL2 induced PI3K resulted in the expression patterns of transcripts that appear to promote survival and proliferation. Interestingly, when CTLs were simulated in the presence of other cells in PBMC, the upregulation of both PI3K and AKT were not detected33 even though IL2 induced a general T cell activation and anti-apoptotic effect illustrating the importance of interactions between effector and bystander cells. Our study was focused on purified NK cells and the effects of bystander cells will not be observed.

NF-κB activation could be mediated by pathways other than IL2 induced PI3K activation, namely the TLR/IL1R pathway, the TNF pathway and possibly a NK specific surface receptor pathway involving BCL10. We have observed good evidence of NF-κB activation through the first two pathways. It is well established that in both T and B cells, BCL10 specifically mediate antigen receptor-induced NF-κB activation 37 In NK cells, BCL10 has been observed in the cytoplasm of normal NK cells, and in the nuclei of tumor cells of nasal NK/T-cell lymphomas 50. Since we observed upregulation of BCL10, NF-κB1 and NF-κB2 upon stimulation of NK cells with IL2, it is possible that cytokine receptor mediated signaling also involve BCL10. The expression pattern of CARD11, a participant in BCL10 induced NF-κB activation in T and B cells 51 also supports this idea. In support of NF-κB activation was the upregulation of NF-κB1, NF-κB2 and the upregulation NF-κB target genes (Figure 4E II). In CD8+T cells, Jin et.al observed increased expression of several mediators of NF-κB pathway, possibly through modulation of TCR signaling (see figure 3 in 33)

PI3K can promote NFAT nuclear accumulation in two ways: by enhancing Ca²⁺ mobilization and calcineurin-dependent NFAT dephosphorylation leading to nuclear import, or by AKT-mediated inactivation of GSK-3 that can phosphorylate NFAT and drive nuclear export 52. The increased expression of positive regulators of intracellular Ca²⁺ release, activators of calcineurin, catalytic calcineurin A subunits and kinases for NFAT nuclear shuttling in activated NK cells suggested both mechanisms to be involved in the NFAT signaling induced by IL2. NFATC1 showed highest expression in resting NK cells, similar to a prior observation in resting T cells, where it is the predominant NFAT protein 53. NFATC1 interacts with GATA3 to maintain the differentiated Th2 phenotype 35. This supports the notion of suppressed expression of Th1 type cytokines in resting NK cells. The profile shifted with the downregulation of GATA3 and upregulation of T-BET on activation.

The early increase in transcription (within the first 2 hours) is likely regulated directly by signaling events – "first wave" induced transcription. Genes that were upregulated only after 8–24 hours are likely to represent activation of "second wave" genes and these were numerous in our study and especially represented genes involved in adhesion, secretory pathway, cytotoxicity and cell cycle control. These "second wave" transcripts are probably under the control of upstream genes or induced by factors such as cytokines or chemokines that are elaborated by the cells at a later time point as illustrated by general upregulation of target genes of STAT1 and NF-κB. Many genes had expression patterns where transcript levels were upregulated at 2 hours, low at 8 hours and then high again at 24 hours (genes important for cytotoxicity (TIA1 and PAF1), cytokine signaling (IL2RA and IL18R1), secretory signaling (SLC3A2, HM13 and DEGS1), cell cycle regulation (MCM6 and CCND2)). The control of transcription of these genes is more complex and may involve feedback inhibition or the induction of inhibitors. Time course experiments are therefore important to gain a more complete picture of the biology and function of the cell of interest.

The effect of IL2 on cytotoxic cells have been reported by several groups 1633395455 showing many similarities as well as differences. For example, CD8⁺ T cells stimulated with 300 IU/ml of IL2 for 4 hours either alone or cocultured with PBMC upregulate IL7, IL13, TNFα and IFNγ whereas the NK cells in our study did not express IL7 or IL13, and the upregulation of TNFα and IFNγ was registered at 8 hours. The cytokine profile of activated NK cells showed marked differences with that observed in activated CD8+T cells by Jin et.al33, with the exception of a few genes (e.g. downregulation of IL8 and CXCR4, and upregulation of IL15 and IFN-γ). A few chemokines e.g.CCL3 and CCL4 were only observed to be upregulated at 24 hours in activated NK cells, whereas in CD8+ T cells, these chemokines are upregulated early (4 hours). On the other hand CCL5, IL16 and CXCR3 were downregulated in isolated CD8+ T cells, but upregulated in NK cells. In general, a detailed comparison is very difficult because of the different systems used and the different time point at which the assays were performed. The expression of many of the genes are highly variable over a 24 hr period so there is a significant degree of uncertainty when compared with results from a particular time point.

In most of the studies, the ethnic composition of the subjects studied was not specified but a systematic comparison of two ethnic groups was performed by Jin and co-workers 33. The latter study is important in highlighting racial differences in immune response and may provide insight into the differences in disease susceptibility and response to immune modulation.

Primary NK cells were isolated from seven healthy donors and each time point contained NK cells from 3 or 4 different donors (additional file 8). The healthy donors provided peripheral blood mononuclear cells (PBMCs) with IRB approval. PBMCs were isolated by Ficoll-Hypaque density gradient centrifugation at room temperature. Non-adherent cells were collected after incubation in nylon wool for 1 hour at 37°C and were mixed with anti CD3, CD14, CD19, CD36 and IgE MicroBeads (MiltenyiBiotec, Auburn, CA) at temperatures between 4–10°C for the negative selection of NK cells. NK cells with the CD56⁺/CD16⁺ and CD3^- phenotype were negatively selected at temperatures between 4–10°C and re-examined by flow-cytometry to ensure purities greater than 90% (Figure 1). Purified NK cells were either treated immediately with Trizol (Invitrogen, Carlsbad, CA) or cultured in RPMI 1640 medium (Life Technologies, New York, NY) supplemented with 10% fetal bovine serum, 200 ug/ml streptomycin and 200 IU/ml penicillin, and IL2 (100 IU/ml) for 2, 8 or 24 hours at 37°C and 5% CO₂, before harvest and storage in Trizol at -80°C till RNA isolation. Each healthy donor was represented in at least 3 different time points and each time point contained NK cells from 3 or 4 different donors. For an independent experiment, NK cells from six new healthy donors were purified and RNA from at least 3 different donors at each time point was pooled and hybridized on GeneChipU133 plus 2^® (Affymetrix Inc, CA). The purity of NK cells was determined by two-color flow cytometry with Alexa488-labeled monoclonal antibody (mAb) against CD3 and Alexa647-labeled mAb against CD56 or CD16. Of the negative selected cells 91% to 98% expressed CD56/CD16 but not CD3 (Figure 1). Cytospins of pre- and post purified PBMCs were stained with Wright-Giemsa stain for some samples to assess the enrichment of large granular lymphocytes.

Total RNA was extracted with Trizol and further purified with RNeasy Mini Columns (Qiagen, Valencia, CA) before aliquots were run in agarose gel electrophoresis and measured by spectrophotometer at 260 and 280 nm to assess the quality of the RNA. For each time point, equal amounts of RNA from at least 3 different healthy donors were pooled prior to one round of RNA amplification using MessageAMP™ aRNA kit (Ambion, Austin, TX) according to the manufacturer's instruction. To minimize biases in RNA amplification only one round amplification was performed and using similar incubations times and 200 ng of total RNA of good quality as template for the reverse transcription reaction 56. The quality and quantity of aRNA was monitored on agarose gel electrophoresis and by spectrophotometer. Typically, 10–20 ug of aRNA was generated from 200 ng of total RNA by one round of amplification and 10 ug of aRNA were employed for hybridization.

aRNA was chemically labeled with a platinum-linked cyanine dye using the MicroMax ASAP RNA labeling kit (PerkinElmer, Boston, MA) as per the manufactures instruction. Briefly, 10 ug of aRNA was incubated at 85°C for 15 minutes with labeling buffer and either Cy5 or Cy3 in a total volume of 20 ul prior to termination of the reaction by cooling on ice and addition of 5 ul of stop buffer. Labeled aRNA was purified on MicroCon 50 columns before the final volume was reduced to 3 ul by vacuum centrifuge (Savant, Farmingdale, NY). aRNA from NK cells was labeled with Cy5 whereas samples from similarly amplified lymphoid RNA reference standard, consisting of RNA from tonsil, thymus, spleen, and cell lines derived from malignant pre-B cells, plasma cells and NK cells 57 was labeled with Cy3. Two-color array analyses were used and determination of expression levels among samples was based on differences of ratios between sample and reference. Other reference RNA, including tonsillar, thymus and resting CD8⁺T was hybridized for comparative analysis, to obtain a signature for NK cells that is distinctive from cytotoxic T cells in addition to being distinctive from a general lymphocyte pool represented by the lymphoid standard.

Microarrays with 60-mer oligonucleotides (ONs) (Compugen, Jamesburg, NJ) printed on lysine-coated glass slides representing 17,260 genes, including reference (control) genes were utilized for hybridization. Labeled aRNA from NK cells and the lymphoid standard was combined in 6 ul of deionized water prior to the addition of poly (dA), human cot-1 DNA and yeast t-RNA at a final concentration of 10 ug each in a total volume of 45 ul containing 4.4 × SSC, 4.1 × Denharts and 50% formamide. Hybridization and washing were performed as described57. Arrays were scanned using an Axon 4000 scanner at 10 um resolution and images were analyzed using Genepix 5.0 software (Axon Instruments, Union City, CA). Raw gene expression data for spotted oligonucleotide arrays are available at 58(accession number .GSE8170)

The isolation of RNA, hybridization and image processing was performed strictly according to the protocol set up by Affymetrix (Affymetrix Inc, CA). Briefly, extracted RNA (at least 2.6 μg used for each time point) was assessed by the Bio-analyzer 2100 (Agilent Inc, CA), and single round amplification protocol (one-cycle target labeling kit, Affymetrix Inc) was utilized for RNA amplification. Antisense biotin-labeled cRNA (10 μg), generated by in vitro transcription were hybridized and biotinylated probes hybridized to the array were detected using pycoerythrin labeled streptavidin. All arrays were scanned using the Gene-chip scanner 3000 (Affymetrix Inc, CA). The Affymetrix GeneChip^® Operating Software (GCOS) was utilized for management, sharing, and analysis of data generated. Other reference RNA was also profiled, namely lymphoid standard RNA 57, universal RNA standard and resting CD8⁺ T cells. The data was analyzed using the same procedure, as above analysis and only those genes were selected which show similar trend of expression with the spotted array data. Raw gene expression data for affymetrix arrays are available at 58(accession number GSE8059)

The fluorescent intensity levels corresponding to each hybridized DNA spot was analyzed with Genepix 5.0 software (Axon Instruments, Union City, CA), and hybridized spots having signal-to-noise ratios below 1.25 and spots with blemishes were flagged and removed prior to further analysis. Local background was subtracted from the individual spots and intensities for each channel were normalized with respect to median intensity from each channel for the entire array. The raw expression data from all the time series were uploaded in BRB-ArrayTools software (Biometric Research Branch-NCI, MA) and global normalization was used to median center the log-ratios on each array in order to adjust for differences in labeling intensities of the Cy3 and Cy5 dyes on different arrays. We excluded genes from the analysis showing minimal variation across time series experiments, and genes whose expression differed by at least 1.25 fold from the median in at least 20% of the arrays were retained. Adjusted data were further filtered to remove genes with opposite ratio values and genes differing more than 3 fold in duplicate analysis of the same NK sample. For SOM, (self organizing maps), the weighted average was calculated for each gene from the duplicate hybridizations according to the formula μ = Σ (x_i/σ_i²)/Σ(1/σ_i²), where each data point x_i is weighted inversely by its own variance σ_i² 59. The weighted average ratio between Cy5 and Cy3 were log₂ transformed and arrays from the time series were centered by resetting the equality parameter (ratio = 1) to the mean of all the experiments before SOM analysis60. SOM, an artificial intelligence algorithm based on unsupervised learning configures output vectors into a topological presentation of the original data. Parameters with similar features are mapped to the same map unit or nearby units in a SOM that allows various visual inspections of the clustering of microarray data (additional file 5).

The pathway analysis was performed using the pathway analytical tool of the BRB-ArrayTools (NCI, NIH-MA) software 61. In this tool, two statistics are computed that summarize the p-values for groups of genes in a pathway; the Fisher (LS) statistic and the Kolmogorov-Smirnov (KS) statistic as described 18 We considered a pathway significant differentially regulated if either significance level was less than 0.001 or at least 5 genes of a pathway are represented on the array.

For the supervised pathway analysis, the genes for selected pathway were downloaded from the Superarray database (SuperarrayInc, MD) and the from the immune signature database 19. The normalized expression of the selected genes was extracted and differentially expressed genes were as those that had at least a 2-fold alteration in expression level between the time points analyzed. Other selected genes were grouped according to their functional characteristics available through OMIM database)62, Entrez Gene 63 and Pubmed 64databases. Similar procedure was used for data analysis on GeneChipU133plus2^® and only those genes were selected for further analysis which show similar trend of expression with the spotted array data.

To confirm the differential mRNA expression identified by microarray assay, an independent NK cell isolation, flow cytometry validation of purity and IL2 stimulation was performed. A total of 13 genes were selected for validation by SYBR Green real time quantitative RT-PCR. In brief, 200 ng RNA was reversely transcribed into cDNA with 200 ng random hexamer using MMLV-RNase H^- reverse transcriptase as per the manufacturer's instructions (Invitrogen, Carlsbad, CA). The human actin, RPL13a and UBC transcripts were used to normalize the expression levels of genes across different time points for comparative analysis. The primers were designed to amplify the cDNA close to the 3' end of the transcript and all the PCR products were less than 200 bp in length. Quantifications were done in triplicate and mean values and standard deviation were calculated for each transcript.