Sphingosine-1-Phosphate Reduces CD4⁺ T-Cell Activation in Type 1 Diabetes Through Regulation of Hypoxia-Inducible Factor Short Isoform I.1 and CD69

Female NOD/LtJ mice (stock no. 001976) were purchased from the Jackson Laboratories and were maintained in a pathogen-free barrier facility. All animal studies were performed following the approved guidelines of the University of Virginia (UVA) Animal Care and Use Committee. Mice were monitored for hyperglycemia by weekly measurement of blood glucose levels using a One Touch Ultra glucometer (LifeScan). Mice were considered diabetic when two consecutive blood glucose levels were >13.8 mmol/l. Mice lacking HIF1αI.1 were generated as described previously (18).

S1P was purchased from Cayman Chemicals. Fluorescein isothiocyanate (FITC)-labeled anti-mouse CD69 (clone H1.2F3), phycoerythrin-labeled anti-mouse CD4 (clone GK 1.5), and isotype control antibodies rat IgG2b-PE and hamster IgG-FITC were all purchased from eBiosciences. Enzyme-linked immunosorbent assay (ELISA) kits for murine interleukin (IL)-17, tumor necrosis factor-α (TNF-α), and IL-10 were purchased from eBiosciences. Mouse CD4 subset columns were obtained from R&D Systems. Anti-CD3 monoclonal antibody (mAb)–coated plates and Th1/Th2 cytokine Cytometric Bead Arrays were purchased from BD Biosciences. HIF1α smart pool siRNA was purchased from Dharmacon. Pertussis toxin was provided by Dr. Erik Hewlett (UVA).

Diabetic and nondiabetic NOD mice were bled retro-orbitally to obtain a prebleed sample. Mice were then injected intravenously with 2 mg/kg body wt SEW2871. After 4 h, blood was collected for measurement of plasma cytokine levels. Cytokines (IL-4, IL-10, and γ-interferon [IFN-γ]) were measured using ELISA.

Spleens were harvested from female diabetic and age-matched nondiabetic NOD mice, and the tissue was gently disrupted in PBS containing 5% heat-inactivated fetal bovine serum (HIFBS) on ice. Splenocytes were passed through a 40-μm nylon cell strainer (BD Biosciences) and collected in PBS. Erythrocytes were removed using lysing buffer (Sigma-Aldrich). CD4⁺ T-cells were isolated with mouse CD4 subset column kit following the manufacturer's protocol. The purity of CD4⁺ T-cells after isolation was >95% as measured by flow cytometry using anti-CD4 antibody. Cells were washed and resuspended in RPMI 1640 containing 10% HIFBS and 1% antibiotic-antimycotic solution. T-cells were activated in 96-well plates coated with 2–10 μg/ml immobilized anti-CD3 mAb in the presence and absence of 500 nmol/l S1P, 1 μmol/l S1P, or 1 μmol/l SEW2871 at 37°C in 5% CO₂ for 16 h. In some studies, cells were incubated in the presence of 100 ng/ml pertussis toxin or 10 μmol/l VPC23019 for 16 h to inhibit S1P1 receptor signaling.

Splenic CD4⁺ T-cells were isolated and activated as described above. IFN-γ, IL-4, IL-17, and IL-10 levels secreted into the media were measured by ELISA of cell supernatants after 16 h of treatment with 500 nmol/l or 1 μmol/l S1P, 1 μmol/l SEW2871, or in some cases 100 ng/ml pertussis toxin or 10 μmol/l VPC23019.

Aortic endothelial cells were isolated and cultured as described previously (21) from nondiabetic NOD mice. Activated CD4⁺ T-cells were labeled with Calcein-AM (Molecular Probes) according to the manufacturer's instructions. Fluorescently labeled CD4⁺ T-cells (50,000 per well) were added to a monolayer of endothelial cells in a 48-well tissue culture dish. After 45 min of incubation at 37°C, unbound T-cells were rinsed off, cells were fixed in 1% glutaraldehyde, and bound T-cells were counted within a 10 × 10 grid using epifluorescence microscopy.

T-cells (0.5 × 10⁶) in 100 μl PBS with 1% HIFBS (fluorescence-activated cell sorting [FACS] buffer) were labeled with 0.25 μg each of FITC-labeled anti-mouse CD69 and PE-labeled anti-mouse CD4 antibodies or FITC- and PE-conjugated isotype control antibodies at 4°C for 30 min. Stained cells were washed with 10-fold excess volume of FACS buffer and resuspended in PBS containing 1% paraformaldehyde. The fluorescence intensity was measured by collecting a minimum of 10,000 events using a BD Biosciences FACS-Calibur dual laser benchtop flow cytometer using CellQuest software (Becton Dickinson) and analyzed using FlowJo software (Treestar, San Carlos, CA).

CD4⁺ T-cells were harvested, and activated using anti-CD3 mAb as described above. Total RNA was isolated using an RNeasy minikit (Qiagen). Quantitative mRNA analysis was performed by real-time PCR using SYBR Green PCR master mix (Bio-Rad) in a Bio-Rad MyIQ icycler. Primer sequences used are listed in Table 1. Data were analyzed and presented as relative expression of mRNA of interest normalized to cyclophylin using the ΔC_t method (22).

CD4⁺ T-cells were isolated from nondiabetic mouse spleens as described above. Four million cells were transfected with 400 pmol HIF1α siRNA or scrambled siRNA by nucleofection using a mouse T-cell nucleofector kit (Amaxa Biosystems). After 2 h, transfected cells were incubated with 1 μmol/l S1P, and the cells were activated using anti-CD3 mAb as described above. After 24 h, medium was collected for IFN-γ ELISA. Cells were used for FACS analysis of CD69 and mRNA expression of IFN-γ, CD69, and HIF1α.

CD4⁺ T-cells were isolated from the spleens of HIF1αI.1^−/− and wild-type mice as described above and were activated in the presence and absence of 1 μmol/l S1P for 16 h. Medium was collected for IFN-γ measurement by ELISA. CD69 surface expression was analyzed by flow cytometry.

Data were analyzed by ANOVA and Fisher's protected least significant difference test using StatView 6.0 software program. Results are expressed as the means ± SE of eight mice per group unless otherwise noted in the figure legends.

Recently, we reported that aortic endothelial cells from diabetic NOD mice were highly activated and that monocyte:endothelial interactions were dramatically increased in diabetic mice (3). In the current study, we explored T-cell activation in this diabetic NOD mouse model. Plasma from hyperglycemic NOD mice and normoglycemic littermate controls was analyzed for Th1/Th2 cytokines (Fig. 1). Plasma IFN-γ was approximately twofold higher in the diabetic mice (15.95 ± 0.19 pg/ml) compared with their nondiabetic counterparts (8.54 ± 0.19 pg/ml). Plasma levels of TNF-α, IL-2, IL-4, and IL-5 were not significantly different between groups. However, diabetic NOD mice showed a 50% reduction in plasma IL-10 levels (P < 0.01).

We tested the hypothesis that S1P would reduce T-cell activation and lymphocyte:endothelial interactions. CD4⁺ T-cells were isolated from spleens of control and diabetic NOD mice using a mouse CD4 subset column, resulting in a population that was about 95% pure CD4⁺ T-cells. The remaining 5% of cells were granulocytes, as evidenced by 7/4⁺ expression (data not shown). Lymphocytes were studied either as naïve cells or after stimulation with CD3 antibody. Naïve T-cells from diabetic mice showed a significant increase in IFN-γ secretion compared with nondiabetic littermates (Fig. 2A). Stimulation of cells with anti-CD3 antibody increased IFN-γ release ∼100-fold in both nondiabetic control and diabetic T-cells (Fig. 2, compare ST bars in B with A). After CD3 stimulation, there was a significant threefold increase in the amount of IFN-γ produced by diabetic T-cells (see ST bar in Fig. 2B) compared with nondiabetic control lymphocytes. S1P at both 500 nmol/l and 1 μmol/l concentrations and the S1P1 receptor-specific agonist SEW2871 at 1 μmol/l had profound impacts in reducing IFN-γ secretion by diabetic T-cells (Fig. 2B). S1P and SEW2871 also reduced IFN-γ secretion by nondiabetic lymphocytes (Fig. 2B). We used S1P at 1 μmol/l for all remaining experiments for comparison with SEW2871 because we used SEW2871 at 1 μmol/l in vitro. The results using SEW2871 strongly suggest that the effects of S1P are through action on the S1P1 receptor. The S1P1 receptor couples solely through Gαi, so to confirm the role of S1P1, CD3 antibody-stimulated lymphocytes from nondiabetic and diabetic NOD mice were incubated with S1P or SEW2871 in the presence of either pertussis toxin to inhibit Gαi coupling of S1P1, or with VPC23019, a specific antagonist of S1P1 (23). As shown in Fig. 2B, both pertussis toxin and VPC23019 inhibit the effects of S1P on downregulation of IFN-γ secretion. Taken together, these data support the notion that S1P acts through the S1P1 receptor on T-cells to modulate IFN-γ secretion. We also examined IL-17 secretion, which is indicative of a Th17 response, and IL-4 and IL-10 secretion, which are indicative of a Th2 response. As shown in Fig. 2C, IL-17 levels were increased approximately twofold in CD3 antibody-stimulated diabetic NOD mouse T-cells compared with antibody-stimulated nondiabetic lymphocytes, and both S1P and SEW2871 significantly reduced IL-17 secretion. In contrast, we observed no changes in either IL-4 or IL-10 expression in the diabetic NOD T-cells (Fig. 2C). S1P and SEW2871 had no effect on secretion of IL-4 or IL-10. These data indicate that both Th17 and Th1 lymphocyte responses are activated in type 1 diabetic mice, and that the S1P-S1P1 receptor axis impacts both types of lymphocyte responses. In addition to inhibiting cytokine secretion, S1P and SEW2871 also decreased the expression of CD69, a marker of T-cell activation. Surface CD69 expression was significantly increased in diabetic lymphocytes compared with nondiabetic control lymphocytes, averaging 72 ± 2 and 61 ± 2%, respectively (Fig. 3). Both S1P and SEW2871 reduced CD69 expression in diabetic lymphocytes to levels equal to or below those of control, with SEW2871 having a dramatic impact on CD69 expression in both control and diabetic lymphocytes (Fig. 3). Both pertussis toxin and the S1P1 receptor antagonist VPC23019 prevented S1P action on CD69 expression, again implicating the S1P-S1P1 axis in regulating lymphocyte activation.

As another measure of T-cell activation, we measured T-cell:endothelial adhesion. Our laboratory routinely uses a monocyte:endothelial adhesion assay that uses primary mouse endothelial cells and monocytes (3,21,24). We modified this method to study T-cell:endothelial adhesion. Interestingly, we observed that diabetic T-cells adhered more readily to the endothelium than nondiabetic control T-cells (Fig. 4). We observed a 50% increase in the number of adherent lymphocytes from diabetic mice compared with control mice. In some cases, CD4⁺ T-cells were pretreated for 4 h with 1 μmol/l S1P or SEW2871 before the lymphocytes were added to the endothelial cells for the adhesion assay. S1P and SEW2871 reduced lymphocyte:endothelial interactions by ∼35% (Fig. 4). S1P also slightly reduced nondiabetic mouse lymphocyte adhesion to endothelium, although this did not reach statistical significance. TNF-α stimulation of EC is shown as a control for maximal adhesion by EC in our assay.

The successful reduction of IFN-γ production by T-cells in vitro by the S1P1 receptor agonist SEW2871 led us to test the effects of SEW2871 on lymphocyte activation in vivo. SEW2871 was intravenously injected into nondiabetic control and diabetic NOD mice. At 4 h after injection, plasma cytokine levels were measured by ELISA. Diabetic mice showed increased plasma IFN-γ levels and reduced IL-10 levels (similar to those shown in Fig. 1). Plasma IL-4 levels did not change (data not shown). SEW2871 decreased plasma IFN-γ levels in the diabetic mice by 50% to levels equivalent to those found in control mice, P < 0.005 (Fig. 5). Furthermore, SEW2871 treatment in vivo significantly increased plasma IL-10 levels in both nondiabetic and diabetic mice (P < 0.01; Fig. 5). Taken together, these experiments in vitro and in vivo using both S1P and SEW2871 strongly suggest that S1P reduces CD4⁺ T-cell activation through action on the lymphocyte S1P1 receptor.

HIF1α is a transcription factor that is induced by both hypoxic and nonhypoxic pathways in T-cells. Two HIF1α mRNA isoforms are transcribed and translated, giving rise to HIF1α I.1 and I.2 proteins. The short isoform HIFI.1 negatively regulates lymphocyte activation (18). Based on these findings, we asked whether S1P impacted HIF1α I.1 expression in T-cells. To examine this, lymphocytes from control nondiabetic littermate and diabetic NOD mice were activated in the absence or presence of either 1 μmol/l S1P or SEW2871. HIF1α I.1 mRNA expression was quantified by quantitative real-time PCR. Interestingly, naïve diabetic NOD lymphocytes expressed 50% less HIF1α I.1 mRNA than naïve nondiabetic lymphocytes (Fig. 6A). Similarly, on TCR-stimulated activation, diabetic T-cells expressed 50% less HIF1αI.1 mRNA compared with nondiabetic T-cells (Fig. 6A). Both S1P and SEW2871 significantly upregulated HIF1α I.1 mRNA in both control and diabetic lymphocytes, suggesting that S1P and S1P1 agonists could be beneficial for reducing T-cell activation through upregulation of HIF1α I.1. Figure 6B shows the expression of HIF1α I.2 mRNA levels. TCR-stimulated control and diabetic lymphocytes had approximately a twofold increase in HIF1α I.2 mRNA levels over naïve cells. However, no significant differences were observed between control and diabetic lymphocytes. Furthermore, neither S1P nor SEW2871 changed expression of HIF1α I.2 (Fig. 6B). These data suggest that action of S1P and SEW2871 through binding to the S1P1 receptor specifically regulates expression of anti-inflammatory HIF1αI.1.

To directly test the contribution of HIF1α I.1 on T-cell activation, we used siRNA to HIF1α. The siRNA was not specific for the HIF1αI.1 isoform in that knockdown of HIF1α I.1 mRNA was 53% and knockdown of HIF1αI.2 was 43% in primary CD4⁺ T-cells from control nondiabetic NOD mice (Fig. 7A). The knockdown of HIF1α corresponded to a significant 2.7-fold increase in IFN-γ production by lymphocytes (Fig. 7B). S1P reduced IFN-γ production in lymphocytes transfected with scrambled siRNA, but not in lymphocytes transfected with HIF1α siRNA (Fig. 7B), suggesting that HIF1α is important for S1P action in lymphocytes. Moreover, knockdown of HIFIα caused a dramatic induction of surface CD69 expression on lymphocytes, illustrating activation of T-cells when HIF1α levels are low (Fig. 7C). As shown in the magenta and light blue lines in Fig. 7C, siRNA to HIF1α increased CD69 expression compared with either scrambled siRNA (dark blue and dark red lines) or untreated nondiabetic lymphocytes (light green and orange lines). An isotype control antibody is used in each case to illustrate specificity of the fluorescent signal. Also shown in Fig. 7C are corresponding dot plots from this representative experiment, as well as a bar graph showing the means ± SE of results using lymphocytes from six mice per condition. These data strongly suggest that HIF1α expression regulates CD69 activation in T-cells.

Although we achieved compelling results using siRNA, the siRNA was not specific for the HIF1αI.1 isoform. To prove that HIF1αI.1 is the signaling molecule for S1P in lymphocytes, we studied T-cells isolated from HIF1α I.1-deficient mice. HIF1αI.1-deficient mice appear normal, and thymocytes and T-cells revealed no phenotypic differences compared with wild-type mice (18). CD4⁺ T-cells were isolated from littermate and HIF1αI.1-deficient mice (18), and we found a significant pro-inflammatory induction of IFN-γ secretion (Fig. 8A) and increased CD69 expression (Fig. 8B) in HIF1αI.1 knockout mice (green line) compared with littermate mice (red line). Importantly, S1P was unable to reduce expression of IFN-γ and CD69 in the absence of HIF1αI.1 (orange line), although S1P was effective in reducing IFN-γ and CD69 in littermate control lymphocytes (blue line) (Fig. 8A and B). We also confirmed these data at the mRNA level, suggesting that HIF1αI.1 is a primary transcription factor regulating expression of both IFN-γ and CD69 in lymphocytes (Fig. 8C). These data show a novel link between S1P and HIF1α I.1 signaling to regulate inflammation. Thus, S1P signaling through S1P1 in T-cells drives an anti-inflammatory phenotype via HIF1αI.1 activation that may be an important therapeutic target for inflammation associated with diabetes.

CD4⁺ T-cells have been shown to contribute to the atherosclerotic burden in mice (25). S1P has been shown to regulate lymphocyte homing to lymph nodes from circulation through action on the lymphocyte S1P1 receptor (12). Studies have suggested that S1P also reduces IFN-γ production by lymphocyte cell lines (26), but studies to clearly define the role of S1P in regulating lymphocyte activation have not been performed to date. In the current study, we examined the role of S1P in reducing T-cell activation in vivo. We show for the first time that S1P reduces T-cell activation in type 1 diabetic mice by reducing IFN-γ and CD69 expression in primary mouse CD4⁺ T-cells. Furthermore, we identified that S1P acts via the lymphocyte S1P1 receptor to upregulate HIF1α isoform I.1, an anti-inflammatory transcription factor that regulates IFN-γ and CD69 expression in primary T-cells. Thus, we have identified a novel link between the transcription factor HIF1α I.1 and the anti-inflammatory lipid S1P in lymphocytes.

We found that T-cells isolated from diabetic NOD mouse spleens were highly activated and secreted approximately twofold more IFN-γ and IL-17 than lymphocytes isolated from control, nondiabetic littermate mice. CD69, a marker of early lymphocyte activation, was highly upregulated in diabetic NOD lymphocytes. Taken together, these data illustrate that CD4⁺ T-cells are “preactivated ” in diabetic NOD mice and are capable of significantly contributing to the inflammatory state in vivo.

We observed significant reductions in plasma IL-10 levels in type 1 diabetic mice (Figs. 1 and 5). In vivo administration of SEW2871, the S1P1 receptor-specific agonist, increased expression of IL-10 (Fig. 5). IL-10 is secreted by CD4⁺CD25⁺ T regulatory cells and by macrophages. As shown in Fig. 2C, we observed no changes in IL-10 secretion in diabetic T-cells, and S1P or SEW2871 did not change secretion of IL-10 by these lymphocytes. We have not yet examined IL-10 secretion in response to S1P or SEW2871 by T-cell subsets, including T regulatory cells. In preliminary studies, we found that both S1P and SEW2871 stimulate macrophage secretion of IL-10 through agonism of the S1P1 receptor on macrophages (data not shown). Increased IL-10 levels are indicative of an anti-inflammatory response, again supporting the notion that the S1P-S1P1 axis promotes an anti-inflammatory phenotype in vivo. Experiments to characterize the role of S1P in T-cell subset activation will be important follow-up studies to this current work.

Lymphocytes express the S1P receptors S1P1 and S1P4. We found that the activation state of CD4⁺ T-cells was reduced by S1P through action on the S1P1 receptor using the S1P1 receptor-specific agonist SEW2871 both in vivo and in vitro. This is somewhat in contrast to a recent study by Goetzl and colleagues (13) that indicated that S1P4 regulated IFN-γ secretion by lymphocytes. This group overexpressed murine S1P4 in a D10G4.1 T-cell line that lacked endogenous S1P receptors. They found that addition of 1 μmol/l S1P to the S1P4-D10G4.1 cells caused a significant reduction in IFN-γ secretion (13). We used a S1P4/S1P1 dual agonist, VPC23153 (27), and found similar reductions in lymphocyte activation, especially at doses >1 μmol/l (data not shown). We attributed these findings to agonism at S1P1 rather than S1P4 because of the dose required for maximal effect on lymphocyte function (data not shown). SEW2871 is specific for S1P1 and has no action on S1P4, and we found that SEW2871 was quite effective in reducing lymphocyte activation, both in vitro and in vivo. Thus, our data suggest that S1P1 is the primary receptor through which S1P regulates lymphocyte activation, although we cannot completely rule out some contribution of S1P4.

The murine HIF1α gene contains two different first exons, termed I.1 and I.2 (15). Expression of HIF1αI.1 and -I.2 is regulated via two distinct promoters, which give rise to these two mRNA isoforms (15). Sitkovsky and colleagues (18) have recently shown that the HIF1αI.1 isoform negatively regulates T-cell activation. In Fig. 7, we used siRNA to HIF1α and found approximately a ∼50% reduction in HIF1αI.1 mRNA and HIF1αI.2 mRNA expression in T-cells. Interestingly, the HIF1α siRNA completely abolished the action of S1P on IFN-γ production (Fig. 7B). This was somewhat surprising in that we observed a total loss of the S1P effect with only a 50% loss of HIF1αI.1. There are two possible explanations for this outcome. First, HIF1αI.2 plays a minor role in regulating T-cell activation as we have shown previously (18). Thus, reductions in expression of both isoforms of HIF1α may contribute to the loss of the S1P effect. Second, antibodies to accurately quantify protein levels of the two isoforms are not available, so we are unsure as to the level of expression of HIF1αI.1 protein in the siRNA studies. It could be that there is very little HIF1αI.1 protein remaining in the cell after HIF1α siRNA treatment. However, we anticipate that S1P primarily acts through the HIF1αI.1 isoform to regulate T-cell activation, in that mice deficient solely in the HIF1αI.1 isoform showed increased T-cell activation, and S1P had no ability to reduce this cell activation (Fig. 8). Levels of HIF1αI.2 are normal in these knockout mice, allowing us to specifically study the I.1 isoform in vivo. Thus, our HIF1αI.1 knockout mouse studies indicate that a primary pathway of regulation of T-cell activation by S1P is through HIF1αI.1.

HIF1α is known for its regulation by hypoxia but can also be induced by nonhypoxic stimuli (28). Karliner et al. (29) observed that 10 μmol/l S1P protected rat neonatal cardiomyocytes from hypoxic cell death in vitro, although the mechanisms for this protection were not identified. Further evidence for a link between S1P and hypoxia comes from a recent study by Pyne and colleagues (30) that illustrates that hypoxia increased sphingosine kinase 1 expression to produce more S1P in pulmonary smooth muscle cells. Thus, S1P synthesis could be a protective mechanism induced by the SMC when exposed to chronic hypoxic conditions. Taken together, our results on lymphocytes and the hypoxia studies of others show a novel link between HIF1α and the protective action of S1P.

In summary, we have found that CD4⁺ T-cells from type 1 diabetic mouse models are “preactivated ” in vivo to release pro-inflammatory cytokines. The anti-inflammatory sphingolipid S1P reduces T-cell activation in diabetic mice in vivo. Action of S1P on lymphocyte activation is primarily through action on the lymphocyte S1P1 receptor through downstream target activation of the anti-inflammatory transcription factor HIF1α I.1. Thus, specific targeting of the S1P1-HIF1αI.1 axis in lymphocytes may be an important regulator of immune responses in chronic diseases, such as diabetes and atherosclerosis.

C.C.H. has received research grants from the Juvenile Diabetes Research Foundation and the National Institutes of Health (HL079621).

We thank Joanne Lannigan and Mike Solga in the UVA Flow Cytometry Core Facility for their technical assistance, Dr. Joel Linden (UVA) for helpful discussions, and Dr. Margaret A. Morris (UVA) for providing some of the NOD mice.