On the eight day in culture, CD8-expressing T cells were enriched by negative selections using magnetic beads before sorting (Miltenyi Biotec, Auburn, CA) on an autoMACS separator. Median purity of CD8 T cells eluted from the columns was 85%. Sorting was then performed by high speed flow-cytometry (FACSVantage SE, BD); a logical gate was applied on SSC and live/dead staining with DAPI to check the viability of sorted cells. The sorting was always done in the Normal-R mode, which optimizes for cell purity, as confirmed by re-analysis of the sorted populations. Sorting was based according to level of tHLA-Flu and CFSE staining of CD8-expressing T cells segregating the tHLA-Flu+/CFSE- proliferating cells from the tHLA-Flu-/CFSE+ non proliferating CTLs. Median purity of Flu-specific and non-Flu-specific CTLs was above 95 % in all experiments (Figure 1A). For the various sorting procedures the following monoclonal antibodies were used: CD8-PE or CD8-FITC, CD3-PerCP (all from BD Biosciences Pharmingen, San Diego, CA). As negative control cells were stained with IgG FITC or PE conjugated, according to the respective antibody's isotype. Cells were analyzed by FACS sort (BD Bioscience) gating them on living CD3-expressing lymphocytes.

Total RNA was isolated with RNeasy minikits (Qiagen, Valencia, CA) and amplified into anti-sense RNA as previously described 1516. First strand cDNA synthesis was accomplished in 1μl SUPERase·In (Ambion, Foster City, CA) and ThermoScript RT (Gibco-Invitrogen, Carlsbad, CA) in 2 μg bovine serum albumin. RNA quality was verified by Agilent technologies (Palo Alto). Anti-sense RNA was labeled with Cy5-dUTP (Amersham, Piscataway, NJ) and co-hybridized with reference pooled normal donor peripheral blood mononuclear cells (PBMC) labeled with Cy3-dUTP to custom made 17K-cDNA array platform (UniGene cluster) printed at the Infectious Disease and Immunogenetics Section, DTM, CC, NIH with a configuration of 32 × 24 × 23 and contained 17,500 elements. Clones used for printing included a combination of the Research Genetics RG_HsKG_031901 8 k clone set and 9,000 clones selected from the RG_Hs_seq_ver_070700 40 k clone set. The 17,500 spots included 12,072 uniquely named genes, 875 duplicated genes and about 4,000 expression sequence tags.

Arrays were scanned on a GenePix 4000 (Axon Instruments) and analyzed using BRB-ArrayTools Version: 3.3, Cluster and Tree View software.

There were extensive differences between quiescent CD8 expressing T cells analyzed ex vivo and their counterparts maintained in in vitro culture in the presence of 300 IU IL-2 and Ag stimulation. An univariate F-test with random variance model identified 4,702 clones out of 16,726 present in the array to be differentially expressed at a p-value < 0.001. The probability of randomly obtaining this number of genes at the selected level of significance (p < 0.005) if there were no real differences among groups was calculated as 0 by multivariate permutation test (Table 1). To assess for possible effects due to the in vitro culture conditions, we maintained PBMCs in culture for 24 hours in the absence of IL-2 and sorted CD8 expressing T cells before transcriptional profiling. This control demonstrated high similarity of in vitro maintained CD8 T cells with the profile of ex vivo analyzed CD8 T cells. Thus, the functional genomics changes observed in in vitro stimulated CD8 T cells are specific to stimulation and not simply due to in vitro culture artifacts. Multiple dimensional scaling based on the global 16,726 cDNA clone data set clearly separated the ex vivo and in vitro quiescent from the in vitro activated populations (Figure 1B).

Analysis of individual experimental conditions against each other demonstrated that the biggest differences in transcriptional patterns were present between the Flu-specific CTLs stimulated in vitro and the quiescent ex vivo CD8 T cells (three way t test) with the least differences noted between the Flu-specific CTLs and the convivial non-Flu-specific CTLs from the same culture.

The transcriptional pattern of in vitro stimulated CTLs whether exposed to Ag (Flu-specific CD8 T cells) or only to IL-2 was similar relative to that of ex vivo isolated or in vitro maintained CD8 T cells. In particular, multiple dimensional scaling based on the complete 16,726 gene data set clearly separated the unstimulated from the stimulated populations. Of 5,859 genes differentially expressed between Flu-specific CTLs and ex vivo CD8 T cells, 3,639 were concordantly expressed by convivial CTLs (Figure 1C). Thus, CTLs maintained in the same culture demonstrate similar transcriptional patterns independent of their exposure to Ag-specific stimulation. Among the genes similarly up-regulated in both subgroups of stimulated T cells were perforin, granzyme A, TNF-α and the IL-2 receptor α chain. Overall, the transcriptional profile of the genes concordantly expressed by in vitro stimulated CTLs was similar to that of our previous reported analysis 13.

The aim of this study was to identify those signatures that are determined by the long term effects of Ag stimulation independent of other co-existing factors that may influence the activation and function of CD8+ T cells. For this reason, we focused our analysis on genes that were differentially expressed between CFSE^high, Flu/tHLA positive CD8 T cells and CSFE^low, Flu/tHLA negative CD8 T cells. An unpaired t test identified 1,727 genes to be differentially expressed between the two populations at a p₂-vlue < 0.001 (Table 1). It should be clarified that several of these genes were concordantly differentially expressed in both subpopulations compared with quiescent ex vivo isolated CD8-expressing T cells. However, the degree in which the expression was altered in the two subsets was sufficiently different to result in significant differences between the two populations. The differences identified were significant according to the multivariate permutation test (p-value = 0). Multiple dimensional scaling analysis based on the complete date set confirmed the separation of the two populations (Figure 1D).

Among the genes differentially expressed by the two populations 644 were up-regulated in Ag-specific CTLs compared to convivial CTLs. The rest (1,083) were down-regulated. The annotations related to biological functions derived through gene ontology suggested that the genes that were predominantly up-regulated in Ag-specific CTLs belong to several categories. Although some categories appeared particularly enriched, they contained a relatively small number of genes, while the categories with the largest absolute number of genes included: cell cycle and cell division (111 genes), response to endogenous stimulus (45 genes) and cytokine production (17 genes). Gene Ontology analysis suggested, therefore, that even seven days after the original stimulus the predominant differences between Ag-exposed Flu-specific CTLs and their culture companions were related to a broader activation of pro-proliferative stimuli, signaling and cytokine production in the former.

Gene Ontology was also applied to identify genes associated with immunological functions; this analysis identified 58 (expected number 52.8; observed over expected ratio = 1.10) up-regulated in the Flu-specific CTLs and 212 out of 988 (expected number 80.3; observed over expected ratio = 2.64) down-regulated relative to the convivial CTLs (Fisher test p₂-value < 0.001).

The immunologically-related genes most expressed by Ag-activated or convivial CTLs are shown in Table 2. The transcript most abundant in Ag-specific CTLs was IFN-γ (Figure 2). Conversely, IFN-γ receptor 2 (β-chain) ranked highest among the immune genes up-regulated bye convivial CTLs, which paralleled the over-expression of interferon-stimulated genes (ISGs) selectively in these cells while ISGs where completely shut off in Ag-activated CTLs. The lack of expression of ISGs by Ag-activated CTLs was associated with lack of expression of the IFN-γ receptor α and β and the IFN-γ receptor accessory factor AF-1. This observation suggested that Ag-activated, terminally differentiated CTLs shelter themselves from the potentially harmful autocrine effects of IFN-γ. This observations supports the lack of responsiveness of Ag-specific T cells to IFN-γ during the expansion phase which is slowly regained during the contraction phase previously reported in an experimental animal model 19. On the other hand, Ag-stimulated CTLs expressed higher levels of CSF2 receptor β chain, (GM-CSF/IIL-5/IL-3 receptor common β-chain, CSFR2B, CD131), chemokine (C-X-C motif) receptor 6 (CXCR6), C-C chemokine receptor 1 (CCR1) and IL-2R α- and β-chains. In addition, Ag-stimulated CTLs strongly down-regulated the chemokine (C-C motif) receptor 7 in accordance with their effector T cell differentiation. Interestingly, the over-expression of CD131 was associated with very high expression of its ligand colony stimulating factor 2 (CSF2, GM-CSF) suggesting that this autocrine loop may play an important role in promoting their survival. Moreover, several cytokines were produced including chemokine (C-C motif) ligand 3 (CCL3, MIP-1α), ligand 4 (CCL4, MIP-1β) and ligand 18 (CCL18, PARC).

Ag-activated CTLs also expressed higher levels of granzyme B and to a lesser degree granzyme A while their convivial counterparts expressed high levels of granzyme K. As previously observed, Perforin, was strongly and equally un-regulated in both populations compared to quiescent CD-expressing T cells studied ex vivo or in vitro 4.

Finally, Ag-activated CTLs expressed higher levels of the co-stimulatory molecules CD80 and CD86.

The high levels of IFN-γ and GMSCF transcript (but not protein) expression together it the up-regulation of CMCSF receptor β chain (CD131) and the down regulation of the IFN-γ receptors I and II by Ag-activated CTLs suggested the intriguing possibility of a bipolar relationship of Ag-activated CTLs with the potential autocrine effects of the two cytokines. It appears that Ag-activated CTLs shield themselves from the potential harmful affects of IFN-γ 20 while accepting the potentially proliferative support of GM-CSF; a growth factor without known pro-apoptotic functions 21. This bipolar behavior may explain the selective survival and expansion of Ag-stimulated CTLs in vitro (an potentially in vivo).

To test the potential of such hypothesis, we analyzed CD131 expression by Flu-specific CTLs. As shown in Figure 3, at day 6 and 12 Flu-specific CTLs consistently express CD131 which is totally absent in the convivial CTLs. IVS of HLA-A*0201-expressing PBMCs with 1 μM Flu M1:58-66 peptide in the presence of IL-2, GMCSF, IFN-γ or a combination of the them with IL-2 demonstrated that while GMCSF (1,000 U/ml) selectively enhances the expansion of Flu-specific CTLs after 12 days in culture when combined to IL-2 (300 IU/ML) administration while IFN-γ (500 U/ml) has no effect consistent with their lack of expression of the IFN receptors (Figure 3C). Not the combination IL-2+GMCSF nor other cytokine combination exerted any effect on the proliferation of convivial CTLs (data not shown).

In conclusion, this preliminary study suggests that, at least during IVS, preferential survival/expansion of Ag-activated CTL may be partially mediated through a bipolar regulation of their sensitivity to the autocrine secretion of cytokines; IFN-γ and GMCSF appear to play a dominant role at this junction as other two cytokines known to be produced by activated CTLs (TNF-α and IL-2) where similarly highly expressed at the transcriptional level by both Ag-activated and convivial CTLs. The confirmation of the selective expression of CD131 on the surface of Ag-activated CTLs and its likely functional association with the selective response of Ag-activated CTLs to exogenous GM-CSF suggests a previously unreported positive feed back autocrine loop that may stimulate CTL growth in response to further Ag stimulation 21. In addition, the positive role that GMCSF may play in the proliferation of CD131-expressing Ag-activated CTLs, may explain the beneficial effects of this cytokine used as vaccine adjuvant, which has been so far attributed exclusively to its role in activating and maturing antigen presenting cells 22. Finally, the selective expression of CD131 by Ag-activated CTLs may qualify this surface marker as a non Ag-specific biomarker of Ag-specific CTL activation. Although extensive in vivo and in vitro validation is required to support such hypotheses, we believe that the novelty and the potential biological implications of these findings warrant a preliminary disclosure.