RNA Interference Screen in Primary Human T Cells Reveals FLT3 as a Modulator of IL-10 Levels

Cytokine production is of major importance in controlling the outcome of the immune response. Among them, IL-10 was first described as the cytokine synthesis inhibitory factor (1) because of its potent regulatory functions. It downmodulates Th1 responses, and has potent anti-inflammatory properties. It notably characterizes T regulatory type 1 (Tr1) cells, a subset of regulatory T cells that attenuate immune responses by secreting large amount of IL-10 that mediates bystander suppression (2). Animal models further reinforce the regulatory functions of IL-10. IL-10–deficient mice develop colitis (3). IL-4 and IL-10 treatments can inhibit the development of type I diabetes in NOD mice (4, 5). Hence, IL-10 is a central cytokine, and the generation of IL-10–producing cells as well as the understanding of the mechanisms controlling IL-10 secretion, have generated a large interest in the community for clinical purposes. A recent cohort of papers demonstrated the role of IL-27 along with TGF-β and IL-6 in inducing IL-10 in murine models (6–9). Human IL-10–producing cells (or Tr1) can be generated in vitro by stimulation with vitamin D3 and dexamethasone (10), or by stimulation with CD46 (11), a complement binding receptor that activates T cells (12). In the mouse model of multiple sclerosis (MS), vitamin D3 strongly inhibits myelin oligodendrocyte peptide-induced experimental autoimmune encephalomyelitis. This is dependent on functional IL-10 and IL-10R, demonstrating the role of IL-10 in preventing experimental autoimmune encephalomyelitis (13). Similarly, in humans, increasing evidence shows a regulatory role of IL-10 in pathologies. In most patients with MS in the relapsing-remitting stage, IL-10 production by CD46-induced Tr1 cells is defective (14–17). Likewise, IL-10 plays a role in asthma. A subgroup of patients with asthma is resistant to glucocorticoid treatment. However, treatment with vitamin D3 and dexamethasone reverses this resistance by inducing IL-10–producing cells (18, 19). Nevertheless, the signaling pathways that lead to IL-10 production in human immune cells are not well understood.

Although there has been effective progress in our understanding of T cell differentiation in animal models, our ability to develop human therapeutics requires a better understanding of human-specific mechanisms that animal models may not be able to recapitulate. However, human studies have been challenging in part because of the lack of tools to carry out causal experiments in primary immune cells.

Systematic loss-of-function genetic screens provide an unbiased approach to deciphering genetic networks across many model organisms. More recently, advances in gene silencing using RNA interference (RNAi) have enabled loss-of-function genetic studies in diverse organisms and cell types (20–22). Using an RNAi-based approach to dissect the molecular pathways that control the balance of cytokines secreted by Th cells, we identified regulators of cytokine production that had not been previously described. Among them, FLT3 was selected for further studies. We show that the FLT3 ligand (FL)-FLT3 receptor pair acts in a negative feedback loop to limit IL-10 levels in response to activation of human T cells. FLT3 (also called FLK2) is part of a class of receptor tyrosine kinases that includes KIT, FMS, and platelet-derived growth factor receptor (23). It is characterized by five extracellular Ig-like domains and a short cytoplasmic tail with kinase activity (24). Human FLT3 (or CD135) was first identified as STK1 (stem cell tyrosine kinase), and is highly homologous to its murine counterpart (25, 26). It is highly expressed in CD34⁺ hematopoietic cells, which defines progenitor cells, and has a role in hematopoiesis. Disruption of FLT3 does not affect the viability of the mice. However, transplantation of these cells failed to reconstitute both T and myeloid cell compartments (27). In this study, we demonstrate a new role of FLT3 in regulating IL-10 production. FLT3 and its ligand, FL, were both induced by T cell activation. Silencing of FLT3 led to increased IL-10 levels, whereas the addition of FL suppressed IL-10 secretion and increased FLT3 surface levels. Moreover, CD46 activation, a known inducer of Tr1 regulatory T cells, upregulated surface FLT3 and secreted FL, which then inhibited IL-10 production in T cells. Hence, FL and FLT3 form a novel regulatory feedback loop that limits IL-10 production in T cells. These results demonstrate the inducible expression of FLT3 in primary human T cells and its novel role in the regulation of IL-10 secretion by activated and/or differentiated T cells. Therefore, by using an RNAi library, we were able to identify new proteins involved in the regulation of cytokine production. Moreover, our studies demonstrate that a lentiviral RNAi library can be readily used to silence genes in human T cells and paves the way for diverse screens to dissect primary human T cell functions.

The complete description of the RNAi Consortium (TRC) lentiviral RNAi library used in this study has been reported (28). Briefly, human genes were targeted with ∼5 short hairpin RNAs (shRNAs) expressed under the control of the U6 Pol III promoter in a lentiviral vector (pLKO.1) that also confers puromycin resistance. Plasmid DNA purification and lentiviral production were performed as described (28) (and see www.broad.mit.edu/rnai/trc/lib).

PBMC were isolated from heparinized venous blood by Ficoll-Hypaque density gradient centrifugation (Pharmacia LKB Biotechnology, Piscataway, NJ) from healthy donors after informed consent. CD4⁺ T cells were then negatively isolated using magnetic beads (CD4 isolation kit II, Miltenyi Biotec, Auburn, CA, purification >90%). T cells were then cultured in 96-well plates precoated with anti-CD3 (OKT3, 2.5 μg/ml), anti-CD28 (3D10, 2.5 μg/ml), in presence of human recombinant IL-2, (Tecin, National Cancer Institute, Frederick, MD) (20 U/ml) for 24 h preinfection. The infection was carried out by centrifugation for 90 min at 2300 rpm at room temperature, in the presence of the viral supernatant and polybrene (8 μg/ml). After removal of the virus, fresh medium was added. Infected cells were then selected by addition of puromycin (2.5 μg/ml) 48 h postinfection, and the cells were expanded in IL-2 for 10 d. Secondary stimulation was then performed in conditions similar to the primary stimulation. Twenty-four hours later, supernatants and cells were harvested. Cell numbers and viability were assessed by flow cytometry. For the primary RNAi screen, approximately eight 96-well plates were infected per experiment (with one well per shRNA), and the full screen consisted of more than six independent experiments using different donors.

For stimulation of noninfected primary T cells, 50,000 CD4⁺T cells were stimulated as mentioned previously, in presence or absence of FL (as indicated, R&D Systems, Minneapolis, MN) for 48–72 h, before analysis by flow cytometry or determination of cytokine production in the supernatants by ELISA.

Proliferation was determined by incorporation of [³H]thymidine, or by flow cytometry by gating on the live cell population. Cytokine production was determined in cell culture supernatants using ELISA specific for human IL-10 (BD Biosciences, San Jose, CA), IL-13, and IFN-γ (Endogen, Cambridge, MA). FL concentrations in supernatants were determined by ELISA (Quantikine ELISA from R&D Systems). FLT3 expression was assessed by flow cytometry using anti-FLT3-PE (BD Pharmingen), or an irrelevant control IgG1-PE. PCR amplification of FLT3 was performed according to (25).

We calculated a Z-score to assess changes in cytokine levels in a single well relative to the mean level of the cytokine across all wells of the same plate. Levels of secreted T cell cytokines were measured using cytokine ELISAs for each 96-well plate well and represented as C = {C₁, C₂, ….C₉₆}, excluding wells for which cell counts were <500 cells. We calculated a score for the cytokine level in a well as: Z_i = (C_i − μ)/σ, in which is the μ and σ are the mean and SD of C, respectively.

The data were analyzed using the Prism software and compared using the Mann-Whitney U test, a nonparametric test that does not assume Gaussian variation. ELISA and proliferation data are the average of triplicate wells.

To identify genetic elements required for primary human T cell functions, we were interested in applying large-scale RNAi libraries to unbiased genetic screens in primary human T cells. To develop such a strategy, we first demonstrated that primary human T cells could be efficiently infected by lentiviruses from the TRC1 RNAi library (28). Primary CD4⁺ T cells isolated from blood were activated with anti-CD3/CD28 Abs, in the presence of IL-2 for 3 d and infected with a GFP-expressing lentivirus. GFP expression was detected in 30% of primary T cells by flow cytometry (Fig. 1A) or fluorescence microscopy (data not shown). Puromycin selection of infected cells resulted in 97% of T cells expressing GFP (Fig. 1A, lower panels). As in previous reports, naive T cells were not appreciably infected with HIV-derived lentivirus (29, 30). Second, to test for RNAi-mediated gene silencing activity in primary T cells, T cells were coinfected with a lentivirus expressing an shRNA targeting GFP, resulting in almost complete knockdown of GFP (Fig. 1A , right panels). These data demonstrate both efficient infection and gene silencing in primary human T cells.

In preparation for larger scale screening, we next optimized infection protocols and restimulation conditions to assess cytokine production in infected T cells. Cells were stimulated as previously discussed, infected with a GFP-expressing virus, allowed to rest in fresh plates for 7–10 d with IL-2, and stimulated again with anti-CD3/CD28 Abs (Fig. 1B). The delay in 7–10 d before secondary stimulation of T cells allowed us to avoid contamination by cytokines produced in response to the primary preinfection stimulation and to reduce levels of pre-existing proteins corresponding to targeted mRNAs because of cell division. As the titer of added lentivirus was different from well to well, we sought to find cell culture conditions for which the phenotypic measurements were not significantly altered by variability in virus titer. We titrated the amount of viruses and seeded T cells, and assessed the viability, infection rate, and IL-10 production resulting from secondary stimulation. Although variability in IL-10 secretion was observed when infecting with different amounts of virus at lower cell seeding densities, an initial culture of 25,000 cells per well minimized this variability (data not shown). By optimizing several conditions, we thus developed a protocol to infect cells and assess cytokine production in a way that was mostly independent of variations in virus titer between distinct library constructs.

We carried out an RNAi screen using a subgenomic library targeting ∼1000 genes, focused on kinases and phosphatases, and containing 3–5 shRNAs per gene. Our goal was to identify regulators of cytokines that reflect key T cell states. This included IFN-γ and IL-13 for Th1 and Th2 differentiated T cells, respectively, and IL-10, a potent anti-inflammatory cytokine produced by a variety of cells, and a hallmark of regulatory Tr1 cells. T cells from each well were stimulated with anti-CD3/CD28 Abs, infected with lentivirus particles expressing a distinct and unique shRNA, selected for infection with puromycin and expanded in IL-2 for 10 d. At 10–14 d postinfection, cells were restimulated with anti-CD3/CD28 Abs in the presence of IL-2. At 24 h postrestimulation, the numbers of viable cells were assessed by flow cytometry, and the levels of IL-10, IL-13, and IFN-γ were determined by ELISA. Cell numbers assessed by flow cytometry and proliferation measured using [³H]thymidine were highly correlated (data not shown). We excluded wells containing <500 viable cells from the analysis. Using a statistical Z-score (see 1Materials and Methods), to quantify the deviation of cytokine levels from the mean of all measurements for that cytokine within the same plate, we selected shRNAs that induced significantly higher or lower levels of IL-10, IL-13, and IFN-γ. The amounts of each cytokine produced postinfection with control shRNAs (against nonmammalian reporter genes) were not significantly altered (Fig. 2A). We selected 55 hit genes for which two or more hairpins induced a similar effect on cytokine production, which were reanalyzed in a secondary screen. The Z-scores for each individual hairpin across independent experiments (including the primary screen and additional tests) using different donors are shown in Supplemental Fig. 1 and Table I. Supporting the selection criteria, we found that ZAP70 knockdown led to reduced IFN-γ production, in agreement with prior studies in patients with ZAP70 deficiencies exhibiting reduced IFN-γ production (31) and that PRKR knockdown increased IFN-γ production which was again in accordance with former studies indicating that PRKR controls IFN-γ translation (Fig. 2A).

The importance of IL-10 in controlling inflammation and autoimmunity and our previous observation of lower IL-10 levels in MS patients (14, 32) led us to focus on a novel regulator of IL-10 identified in our screen. Two shRNAs targeting FLT3, a receptor tyrosine kinase, led to an increase in IL-10 secretion (Fig. 2B), suggesting that FLT3 is a negative regulator of IL-10 secretion on T cell activation. Moreover, using qRT-PCR to assess knockdown of FLT3, we found an inverse correlation between the amount of knockdown and the phenotypic score (Fig. 2C).

Because FLT3 was known to be expressed in CD34⁺ progenitor cells and thymocytes, but its expression in mature T cells was not previously reported (25), we first examined whether FLT3 and its ligand, FLT3-ligand (FL), were expressed in resting and activated human T cells. Using purified human CD4⁺ T cells, we found that although FLT3 expression was undetectable by flow cytometry in resting cells, it was detectable at low but significant levels on stimulation with anti-CD3/CD28 Abs in the presence of IL-2 (Fig. 3A). Moreover, the presence of FLT3 in activated T cells could be also detected by PCR (Fig. 3B). Second, it could be further upregulated by addition of FL to activated T cells (Fig. 3C), suggesting a positive feedback loop between ligand and receptor.

Finally, we hypothesized that the original phenotype observed in FLT3 KD cells (Fig. 2) would have required expression of FL by activated T cells to engage FLT3, as no exogenous FL was added to the culture. The predominant human FL form is a transmembrane protein that can undergo proteolytic cleavage to generate a soluble form of the protein (33). An alternatively spliced FL mRNA, encoding a soluble form of the human FL, has also been identified (33, 34). Consistent with our prediction, we found significantly higher levels of FL in the supernatants of CD3/CD28 stimulated T cells (Fig. 3D). Supporting this observation, FL is known to have a broad RNA expression in human tissues and a high level of expression in PBLs (33, 35), and T cells (34). These observations demonstrate that expression of FLT3 and its ligand, FL, are both induced by engagement of CD3 and CD28 in T cells, and that FLT3 levels could be further increased in response to its ligand in a positive feedback loop.

To directly test whether FLT3 activation by FL results in IL-10 production, we added recombinant human FL at several concentrations to T cell cultures stimulated with anti-CD3/CD28 Abs in the presence of IL-2. The resulting changes in proliferation, IL-10 and IFN-γ production were assessed after 3 d (Fig. 4A). Addition of FL to T cell culture decreased IL-10 production to ∼50% of control levels but had minimal effect on proliferation and IFN-γ production. We next determined whether the strength of TCR stimulation could modulate the inhibition of IL-10 mediated by FL. Increasing the doses of anti-CD3 Abs led to a stronger inhibition of IL-10 by FL (Fig. 4B). To further demonstrate a role of FLT3 in IL-10 production, we next assessed the effect of FL on IL-10 secretion in T cells infected with different shRNA constructs targeting FLT3. Only efficient knockdown of FLT3 in T cells abolished the repression of IL-10 by added FL (Fig. 4C). FL inhibited IL-10 production by T cells infected with a nonsilencing shRNA (no. 3), similarly to T cells infected with GFP-control, whereas it had no effect on T cells infected with a silencing shRNA construct (no. 4). These results are consistent with the hypothesis that FL acts through FLT3 on activated T cells to reduce IL-10 levels.

Stimulation of CD4⁺ T cells by CD3 and CD46 in the presence of IL-2 induces the differentiation of Tr1 cells, regulatory T cells that produce IL-10 (11). To find out whether FLT3 plays a role in the production of IL-10 in Tr1 cells, we first measured FLT3 expression on CD46-induced Tr1 cells and found that FLT3 expression was indeed induced on Tr1 cells by anti-CD3 and anti-CD46 Abs and IL-2 (Fig. 5A). When we next added FL to CD46-differentiated Tr1 cells, we found that IL-10 levels decreased, whereas IFN-γ production and proliferation did not change (Fig. 5B). To assess whether the FL-FLT3 regulatory loop is present in Tr1 cells, we determined if IL-10-producing Tr1 cells could themselves secrete FL. When CD4⁺ T cells were stimulated with CD3 alone or with CD3/CD46 Abs in the presence of IL-2, only the combination of CD46 and CD3 Abs, which led to differentiation of Tr1 cells, could induce significant amounts of FL (Fig. 5C). Lastly, we determined the effects of FLT3 knockdown on IL-10 production by CD46-stimulated T cells. Purified T cells were stimulated with CD3/CD46 Abs in presence of IL-2 and infected with the efficient shRNA lentiviruses targeting FLT3 (constructs number 4 and 6), or with shRNA control. Because CD3/CD46 stimulated cells did not survive long enough for a secondary stimulation (in contrast to CD3/28 stimulated cells), we measured the level of IL-10 production 48 h postinfection and found that FLT3 knockdown led to an increase in IL-10 secretion, whereas IFN-γ levels did not change (Fig. 5D).

A large-scale lentiviral RNAi screen in primary human T cells has allowed us to uncover roles for several new genes in the control of cytokine secretion, including a previously unappreciated function for the cell surface receptor, FLT3, in regulating IL-10 production. Although RNAi screens have proven to effective in elucidating gene functions in T cell lines (36, 37), our studies demonstrate the application of this technology to primary human T cells, thus allowing us to study the normal processes of T cell activation and cytokine production. Future studies can take advantage of this tool to study diverse processes in normal T cells, such as differentiation, activation, migration or HIV infection, as well as in abnormal T cells from patients with immune disorders. Consistent with our observation of enhanced IFN-γ production in knockdown cells, PRKR (IFN-induced, dsRNA activated protein kinase) has previously been shown to inhibit IFN-γ mRNA translation (38). This result (as well as the effect of ZAP-70 silencing) provided a positive control for our screen.

Our main goal was to identify regulators of IL-10 production. IL-10 is a critical regulator of inflammation (39) and it plays an important role in human pathologies, such as asthma (40) and MS (14, 32). We found that knockdown of FLT3 led to an increase in IL-10 secretion. Hence, FLT3 may act a negative regulator of IL-10 secretion. Our findings that FLT3 and FL play a role in mature T cell activation were unexpected. FLT3 was previously described as a hematopoeitic growth factor, not expressed by mature human T cells (25), but expressed by thymocytes and playing a role in thymopoeisis (41). FLT3 deficiencies result in defects in multipotent stem cells and lymphoid differentiation (27). In addition, FL has been shown to synergize with a wide variety of hematopoietic cytokines to stimulate the growth and differentiation of early hematopoietic progenitors (42). Finally, FLT3 mutations are the most frequent genetic aberrations that have been described in acute myeloid leukemia (43). Importantly, FLT3 expression was not detected on monocytes, which sometimes contaminate purified T cells (not shown), in accordance with a recent report (44). Moreover, the staining obtained with EOL-1, a leukemia cell line known to express FLT3, although stronger than on activated T cells, was also low (not shown). The expression of FLT3 on T cell activation and the inhibitory effect of FL on IL-10 secretion were also observed using >98% pure positively selected CD4⁺ T cells (not shown). Our finding that FLT3 and FL are induced in activated T cells and function in a negative feedback loop to dampen IL-10 production was not previously reported. Our results, however, may help explain some observations in published studies. First, consistent with our observation that FLT3 is a repressor of IL-10 production, an increase in IL-10 levels was observed in FLT3-deficient mice with an allogeneic aorta graft (45). Although this effect was attributed to dendritic cells, our results suggest an additional role for T cells. Second, administration of FL has been shown to promote tumor rejection (46, 47), by enhancing NK activity (48) and skewing the response toward a Th1 phenotype, as FL-stimulated DCs secrete high amounts of IL-12 (49). Our data suggest that FL could also play a role in rejecting tumors by reducing IL-10 levels in T cells and thus enhancing immunity.

Importantly, our data also demonstrate that the expression of FLT3 and FL is induced on Tr1 regulatory T cells that produce IL-10. IL-10 levels in CD46-induced Tr1 cells were possible to decrease by addition of FL or to enhance by silencing of FLT3, consistent with a negative feedback role for FLT3 in IL-10 regulation, perhaps acting to terminate the regulatory responses of Tr1 cells (see model presented in Fig. 6). Although the source of FL in our system appears to be the T cells themselves, thus forming a negative feedback loop, other sources may also be active in vivo and may act to modulate regulatory T cells in specific tissues and physiological states (34, 42). Given our previous observation that CD46-induced Tr1 cells are defective in patients with MS (14), it will now be important to test the role of the FLT3 and other IL-10 regulating mechanisms in autoimmune and inflammatory pathologies.

We thank Alice Pannerec and Siobhan Ni Choileain for their excellent technical skills.

Disclosures The authors have no conflicts of interest.