Inhibition of Acid Sphingomyelinase Allows for Selective Targeting of CD4⁺ Conventional versus Foxp3⁺ Regulatory T Cells

In contrast with conventional CD4⁺ and CD8⁺ T cells, CD4⁺ Foxp3⁺ regulatory T cells (Tregs) strongly depend on recognition of self-Ags via their TCR (1), CD28 ligation (2, 3), and IL-2 (4, 5) for survival and also for the maintenance of their suppressive activity. Of these pathways, stimulation of both the TCR (6, 7) and CD28 (7, 8) have been shown to activate sphingolipid catabolizing enzymes, that is, neutral sphingomyelinase (Nsm) and acid sphingomyelinase (Asm; Nsm and Asm in mice and NSM and ASM in humans), respectively. These sphingomyelinases cleave sphingomyelin (SM) into ceramide and phosphorylcholine. Whereas Asm is constitutively expressed in lysosomes and released to the outer leaflet of the cell membrane upon, for example, CD28, CD95, or TNF receptor activation (7–11), the Nsm is located at the inner leaflet of the cell membrane after, for example, TCR ligation (6, 7), where, however, SM is much less abundant than in the outer leaflet (12). Because of their distinct localization within the cell, Asm activity is required for lysosomal integrity (13) and in the cell membrane for the generation of ceramide-rich signaling platforms, vesicle formation, and degranulation (14). By contrast, the Nsm produces ceramide in the inner leaflet of the cell membrane, which then interacts with intracellular signaling molecules (15). One prominent signaling cascade long recognized to be modulated, that is, inhibited, by ceramide formation is the Akt/mTOR axis (12, 16). A recent publication describes that Tregs, in which Akt/mTOR signaling is constitutively inhibited, lose their immunosuppressive properties when ceramide-regulated inhibition of this pathway is abrogated (17). Further, for Tregs it is very likely that Asm activity is crucially involved in the function of their effector molecule CTLA-4, because CTLA-4 removes costimulatory ligands from the surface of APCs by transendocytosis (18, 19). This requires internalization of CTLA-4 and its ligands via clathrin-coated pits, which, studying the transferrin receptor, has been shown to be associated with induction of ASM activity in Jurkat cells (20).

Even though the data on the role of sphingolipid metabolism in Tregs are still very limited, a number of studies have addressed the contribution of SM breakdown to conventional effector T cell function. For CD8⁺ T cells, it has been shown that Asm deficiency impairs the release of IFN-γ and the secretion of cytotoxic granule postinfection of mice with lymphocytic choriomeningitis virus (LCMV) (14). Effector cell differentiation, however, appears to be normal in CD4⁺ Th cells from Asm^−/− mice because they could induce acute graft-versus-host disease with similar kinetics and strength as CD4⁺ T cells from wild-type (wt) mice (21). However, using pharmacological inhibitors of the ASM (i.e., antidepressants), some of which have been in clinical use since 1961 (22), recent studies have shown that ASM activity is required to allow for full T cell activation and Th cell differentiation of human T cells in vitro (23, 24). In vivo, treatment with Asm inhibitors protected laboratory animals from T cell–mediated autoimmunity and acute graft-versus-host disease (25). Moreover, Asm inhibitor treatment also alleviated allergic rhinitis in mice, which was associated with a shift from pathogenic Th17 cells toward protective Tregs as a result of the treatment (26).

Our study follows up on this observation with an in-depth analysis of the activity of Asm and sphingolipid content of mouse Tregs and conventional CD4⁺ T cells (Tconvs). For this we used cells from both wt and Asm^−/− mice, and applied pharmacological inhibitors of the Asm to healthy mice in vivo and to primary mouse leukocytes in vitro. We found that genetic deficiency and pharmacological inhibition of the Asm is associated with an increased Treg frequency and activity. Previously we demonstrated in our CNS infection model with recombinant measles virus (MV) that increasing the number of Tregs with a superagonistic anti-CD28 mAb resulted in a reduced immune response and drastically increased brain infection, whereas depletion of Tregs had the opposite effect; thus, this model can be used as a sensitive indicator system for the efficiency of Tregs in vivo (27–29). Using Asm-deficient (Smpd1^−/−) mice and this CNS infection model, we obtained a massive brain infection that was consistent with an enhanced Treg frequency and/or activity.

Asm^−/− (Smpd1^−/−) C57BL/6 mice (30) were obtained from R. Kolesnik (Memorial Sloan-Kettering Cancer Center, NY). B6.129(Cg)-Foxp3^{tm3(DTR/GFP)Ayr}/J mice were bought from The Jackson Laboratories. Both strains were bred under specific pathogen-free conditions in the animal facility of the Institute for Virology and Immunobiology, Würzburg, and of the Institute for Molecular Biology, Essen. All animal experiments were conducted in accordance with German law and approved by the governments of Unterfranken or Düsseldorf, respectively. For comparisons of different genotypes, littermates were used throughout the study.

Recombinant MV expressing the rodent-adapted hemagglutinin of the strain CAM/RB (CAMH) and the enhanced GFP (eGFP) rMV^EdtagEGFP-CAMH, in this article designated as MV (31, 32), was propagated using Vero cells. For infection, 2-wk-old mice were anesthetized using isoflurane and 20 μl virus suspension containing 2 × 10³ PFU injected intracerebrally into the left hemisphere. Infection is quantified microscopically (Leica DMi8 fluorescence microscope, HCX PL Fluotar L 20×/0.40 objective lens, Leica DFC3000 G camera, and Leica LAS-X software for acquisition of images) by counting eGFP⁺ neurons in 50-μm coronal brain slices through the complete brain (27–29). Adobe Photoshop CS4 was used to adjust for brightness and contrast.

To deplete Tregs in vivo, at ∼12 pm we treated B6.129(Cg)-Foxp3^{tm3(DTR/GFP)Ayr}/J mice and wt littermates i.p. with diphtheria toxin (DT; Calbiochem) in 100 μl of PBS at a dose of 500 ng (days 21 and 22 postinfection) and 250 ng (days 24 and 26 postinfection) per injection. At ∼9 am and 5 pm the mice additionally received amitriptyline (Sigma) i.p. at a dose of 0.37 mg/mouse (∼20 mg/kg body weight) in a total volume of 200 μl of 0.9% NaCl.

C57BL/6 mice were injected twice daily with 0.5 mg amitriptyline (∼20 mg/kg body weight) in 200 μl of 0.9% NaCl i.p. for 2 consecutive days. After a fifth injection on the third day, spleens and lymph nodes were analyzed. Asm^−/− mice (4–12 wk of age) received 8 × 10⁶ to 1 × 10⁷ purified CD4⁺ T cells (purity ∼93%) from Thy1.1⁺/Thy1.2⁺ Asm wt mice (age 17–20 wk) in a total volume of 200 μl of PBS i.v. before initiation of amitriptyline treatment as detailed for wt mice.

The following Abs were used to stain mouse cells: anti-CD4 Alexa Fluor 647, Pacific Blue or brilliant violet 421 (clone GK1.5), anti-CD8β Alexa Fluor 700 (clone YTS156.7.7), anti-CD45 brilliant violet 510 (clone 30-F11), anti–CTLA-4 PE-Cy7 (clone UC10-4F10-11), anti-Thy1.1 FITC (clone OX-7), and anti-Thy1.2 PE-Cy7 (clone 53-2.1) (all from Biolegend, San Diego, CA). Further, anti-CD44 FITC (clone IM7), anti–CTLA-4 PE (clone UC10-4F10-11) and PE hamster IgG_κ isotype control, anti-CD62L PE (clone MEL-14), anti–Ki-67 PE (clone B56) (all from BD Biosciences, Franklin Lakes, NJ), and anti-CD25 PE-eFluor 610 (clone PC61.5) and anti-Foxp3 PerCp-Cy5.5 (clone FJK-16s) (eBioscience, San Diego, CA). For di-4-ANEPPDHQ (ANE) stainings, CD4⁺ CD25⁺ and CD4⁺ CD25⁻ (C57BL/6 mice) (Miltenyi Biotec, Bergisch Gladbach, Germany; see later) or CD4⁺ Foxp3-red fluorescent protein⁺ and CD4⁺ Foxp3-red fluorescent protein⁻ (C57BL/6-Foxp3^tm1Flv/J) (33) cells were incubated with 4 μM of ANE (Invitrogen, Carlsbad, CA) for 30 min at 37°C and were immediately analyzed by FACS measuring emission at 570 and 630 nm.

As described earlier (29), MHC class I (H-2D^b) pentamers presenting the MV hemagglutinin peptide (D^bMV-H_22–30) were purchased from ProImmune (Oxford, U.K.). Cells were washed and stained with 5 μl of pentamer solution diluted in 100 μl of PBS containing 0.1% BSA and 0.02% NaN₃ at 4°C for 30 min. After one washing step the cells were analyzed by flow cytometry. MV-specific cells were gated as CD8⁺ and CD19⁻ lymphocytes to exclude pentamer⁺ CD19⁺ cells.

Conventional T cells and Tregs were isolated by MACS separation using the CD4 isolation kit and anti-CD25 PE with anti-PE beads and LS columns (Miltenyi). The CD4⁺ T cells (1 × 10⁶) were either not (Fig. 2A, 0 min) or were preincubated with 1 μg/ml purified anti-mouse CD28 Ab (clone 37.51; Biolegend) on ice. For stimulation, cells were transferred to a 48-well plate precoated with 25 μg/ml anti-hamster IgG (Dianova, Hamburg, Germany) (1 h, 37°C) and stimulated in fully supplemented RPMI 1640 medium (34) at 37°C for the indicated periods.

Asm activity measurements of T cells were carried out as described previously (7). In brief, 5 × 10⁵ T cells were disrupted by freeze/thawing (methanol/dry ice) in Asm lysis buffer (250 mM of Na-acetate pH 5.2, 1.3 mM of EDTA, 0.2% Na-taurocholate). Nuclei were removed by centrifugation for 5 min at 1600 rpm. Postnuclear homogenate was centrifuged for 1 h at 26,000 rpm in 100 mM of Na-acetate, pH 5.2. Cell membrane pellets were resuspended in 40 μl of Asm lysis buffer and aliquots incubated with 1.35 mM of 6-hexadecanoylamino-4-methylumbelliferyl-phosphorylcholine (Moscerdam Substrates) as an artificial sphingomyelinase substrate at 37°C for 17 h (final volume, 30 ml). Quantification of fluorescence was performed using a fluorescence reader (Safire²; Tecan) with excitation at 404 nm and emission at 460 nm.

For lipid analysis, 1.5 × 10⁶ cells (FACS-sorted Tregs or Tconvs) were dissolved in 500 μl of methanol. Ceramides and SM were extracted and quantified as recently described (35, 36). In brief, lipid extraction was performed using C17-ceramide and C16-d31-SM as internal standards. Sample analysis was carried out by rapid-resolution liquid chromatography–tandem mass spectrometry using a Q-TOF 6530 mass spectrometer (Agilent Technologies, Waldbronn, Germany) operating in the positive electrospray ionization mode.

Tregs were isolated as for the Asm assay. The Treg-enriched CD25⁺ magnetic fraction was sorted (BD FACSAria III) to isolate highly pure CD4⁺ CD25⁺ Tregs. The CD25⁻ nonmagnetic fraction (conventional T cells and APCs) was purified using an LD column and labeled with the proliferation dye eFluor 670 (eBioscience). Tregs from wt and Asm^−/− mice were added to eFluor670⁺ cells to obtain Treg/indicator T cell (Tind) ratios of 1:1, 1:2, 1:4, and 1:8. For stimulation, anti-CD3 mAb (5 μg/ml, clone 145-2C11; BD) was added and cell division was analyzed after 3 d. Percent suppression was calculated as follows: 100 × (division index [Tinds] − division index [Tinds + Tregs])/division index (Tinds). Values <0 were set to zero.

Total lymph node cells were cultured in the presence of 25 IU/ml recombinant human IL-2 (Proleukin; Novartis) ± anti-CD3 mAb (145-2C11; 5 μg/ml). To label (capture) the cell surface–exposed CTLA-4, we incubated cells with anti–CTLA-4 PE or a matching PE isotype control in the culture medium at 37°C for 4 and 24 h. Cells were additionally stained for CD4, CD25, and Foxp3 expression and the median fluorescence intensity of captured CTLA-4 quantified on CD4⁺ CD25^high Foxp3⁺ Tregs. Moreover, we determined total CTLA-4 expression after fixation and permeabilization of cells using Perm/Fix and Perm buffer from eBioscience.

Single splenocyte suspensions were prepared and erythrocytes lysed by hypoosmotic shock. A total of 2 × 10⁵ splenocytes per well (96-well round-bottom plate; Greiner, Kremsmuenster, Austria) was cultured in complete RPMI 1640 medium supplemented with 10% FCS (34) for 3 d in the absence or presence of 0.1 or 1 μM of recombinant human IL-2 (Proleukin; Novartis) and desipramine or amitriptyline (both from Sigma-Aldrich, Munich, Germany). Under similar conditions, MACS-purified CD25⁻ nonmagnetic fraction (conventional T cells and APCs) from wt or Asm^−/− mice were mixed and cultured in the presence of sorted Tregs from wt mice. Further, FACS-sorted CD4⁺ Foxp3⁺ Tregs and CD4⁺ Foxp3⁻ Tconvs [B6.129(Cg)-Foxp3^{tm3(DTR/GFP)Ayr}/J mice] were cultured for 3 d in the presence of 1 μM of IL-2 and the indicated concentrations of amitriptyline. During the final 16 h of the culturing period, proliferation was determined by measuring [³H]thymidine incorporation.

For parametric data, either a two-tailed unpaired Student t test (two groups) or two-way ANOVA (more than two groups) was performed (GraphPad Prism 6). Nonparametric data were analyzed using the Mann–Whitney U test (GraphPad Prism 6). Groups were considered to be significantly different when p < 0.05 (*p < 0.05, **p < 0.01, ***p < 0.001). All summary data are displayed as means and SD unless indicated otherwise.

Because Tregs, in contrast with CD4⁺ Foxp3⁻ CD25⁻ Tconvs, require constant stimulation via CD28 for survival and function even under steady-state conditions, we purified both subpopulations from wt C57BL/6 mice (Fig. 1A) to determine Asm enzymatic activity (Fig. 1B). Asm activity was significantly increased in Tregs as compared with Tconvs both without and upon CD28 stimulation (Fig. 1B).

Analysis of SM and ceramide content showed that in line with published data (30, 37), membranes of T cells from Asm^−/− mice contained 7- to 10-fold more SM, with Tregs containing even more SM than Tconvs of these animals (Fig. 1C). In accordance with their strong Asm activity, Tregs of wt mice contained 5- to 8-fold more total ceramide than the membranes of Tconvs (Fig. 1C). In Tregs of Asm^−/− mice, the ceramide content was not reduced (Fig. 1C), but because of the high amount of SM in these cells, the amount of ceramide generated per SM molecule was lower. There was no bias between Tregs and Tconvs regarding SM and ceramide subspecies with different chain lengths (C16, C18, C20, C22, C24, C24:1; data not shown).

For crucial signaling steps in T cells it is not total lipid content, but the lipid composition and the ordered state of the cell membrane lipids that are decisive (38). Using the membrane dye ANE (39–41), we found that in wt mice the membranes of ∼15% of Tconvs, but 46% of Tregs, had a low lipid order, that is, well incorporated the ANE dyes into their cell membrane (Fig. 1D). In contrast, in Asm^−/− mice, only 5% of Tconvs and 9% of Tregs were low order cells (high ANE staining), which is supposed to facilitate T cell activation (39). When dissecting the Tconv compartment in wt mice, we further observed that Tconvs with an effector/memory-like phenotype (CD44^high) (42, 43) were more like Tregs than phenotypically naive Tconvs because they also contained increased frequencies of cells with low lipid order (Fig. 1D). In contrast with Tregs, the frequency of cells with low lipid order was not so dramatically reduced in Tconvs of Asm^−/− as compared with wt mice (Fig. 1D). Tregs of wt mice thus displayed higher Asm activity and ceramide content than Tconvs. The comparison of wt and Asm^−/− mice further revealed that Tregs from Asm^−/− mice displayed a higher membrane lipid order than their wt counterparts.

The suppressive activity of Tregs depends on the expression of membrane-bound molecules like CTLA-4 or the high-affinity IL-2R (CD25), as well as the release of suppressive cytokines such as IL-10 (44). However, neither CD25 nor IL-10 were reproducibly increased in Tregs of Asm^−/− mice as compared with wt mice, and surface CTLA-4 expression is almost undetectable on murine T cells upon direct FACS analysis ex vivo (C. Hollmann, S. Werner, J. Schneider-Schaulies, and N. Beyersdorf, unpublished observations). Because CTLA-4 is subject to high turnover between intracellular stores and the cell membrane, we used a capture assay to determine the rate of CTLA-4 molecules emerging at the cell surface of Tregs within 4 and 24 h of culture (45). In this study, Tregs from Asm^−/− mice spontaneously transported a significantly higher proportion of the total pool of CTLA-4 molecules to the cell surface within 24 h as compared with wt Tregs (Fig. 2A, 2B). CD3 stimulation increased CTLA-4 trafficking to the cell surface in Tregs from both mouse strains, but also under these conditions more CTLA-4 was captured on the surface of Tregs of Asm^−/− mice than wt mice (Fig. 2B).

To directly assess the inhibitory activity of Tregs from Asm^−/− mice, we performed surrogate in vitro suppression assays. Tregs and Tinds containing APCs were separated by magnetic sorting, Tinds were labeled with the proliferation dye eFluor 670, and then the cells were mixed in defined ratios and stimulated with anti-CD3 mAb in solution in the presence of APCs to induce cell proliferation. Tregs of Asm^−/− mice suppressed the proliferation of Tinds of wt (Fig. 2C, histograms and bottom left graph) and Asm^−/− mice (Fig. 2C, bottom right graph) significantly better than Tregs from wt mice. The increased CTLA-4 turnover of Tregs from Asm^−/− mice thus positively correlated with a higher suppressive activity in vitro as compared with wt controls.

Asm deficiency and the associated increased membrane lipid order of Tregs in mice may not only affect the function of individual Tregs, but might also have an impact on the Treg compartment as a whole. Therefore, we characterized Tregs from Asm^−/− and wt mice by flow cytometry. We found significantly higher frequencies of Tregs among splenic, but not lymph node, CD4⁺ T cells of Asm^−/− in comparison with wt mice (Fig. 3A, 3B). This increase in relative numbers of Tregs in spleens was not accompanied by elevated absolute numbers of Tregs in Asm^−/− as compared with wt animals (Fig. 3B), but rather because of reduced absolute numbers of Tconvs in these animals (although this reduction in absolute Tconv numbers did not reach statistical significance, differences in the percentage of Tregs were statistically significant). Apart from their greater abundance among CD4⁺ T cells in the spleen, Tregs of Asm^−/− mice contained more activated, that is, CD62L⁻ CD44⁺ (46), cells than their counterparts from wt animals (Fig. 3C, 3D). These findings suggest that there were not only higher frequencies of Tregs among CD4⁺ T cells in spleens of Asm-deficient mice, but that a higher percentage of these cells was in a more activated and suppressive state than Tregs from wt mice.

Given our data from Asm^−/− mice, transient pharmacological inhibition of the Asm might be sufficient to induce similar changes in the Treg compartment as observed in the Asm^−/− mice. Indeed, five injections of the Asm inhibitor amitriptyline over 3 d increased Treg frequencies among CD4⁺ T cells in lymph nodes and spleen compared with the control-treated mice (Fig. 4A, 4B). Absolute numbers of Tregs were not changed upon pharmacological Asm inhibition, but absolute numbers of Tconvs were reduced in lymph nodes (p < 0.05) (Fig. 4C). In spleens there was a (statistically not significant) tendency toward a reduction after amitriptyline treatment (Fig. 4C). In addition and in contrast with the Asm^−/− mice, short-term Asm blockade in wt mice increased total CTLA-4 expression by Tregs (Fig. 4D).

To test whether the effects observed upon amitriptyline treatment in vivo were indeed due to Asm inhibition, we transplanted Asm wt CD4⁺ T cells from Thy1.1⁺ Thy1.2⁺ congenic mice into Asm^−/− mice (Thy1.1⁻ Thy1.2⁺) followed by amitriptyline treatment for 3 d. In contrast with wt mice, treatment of Asm^−/− mice neither increased Treg frequencies among endogenous CD4⁺ T cells (Fig. 4E) nor reduced endogenous Tconv numbers (Fig. 4F). Tconv numbers of the transplanted wt CD4⁺ T cells that made up 2–5% of all CD4⁺ T cells in the recipient were, however, reduced (Fig. 4F) as after treatment of wt mice with amitriptyline (Fig. 4C). Treg frequencies among the transferred CD4⁺ T cells were not increased upon amitriptyline treatment (Fig. 4E). We assume that the high Treg frequencies in the Asm^−/− mice and the requirement for sufficient amounts of IL-2 in the microenvironment for Tregs to survive upon Asm inhibition (Fig. 5) are responsible for this discrepancy in the behavior of the transferred Tregs as compared with Tregs in wt mice.

Together these experiments show that short-term Asm blockade by amitriptyline treatment thus was sufficient to increase the frequency of Tregs among CD4⁺ T cells in vivo.

Our in vivo data from the amitriptyline-treated mice suggested that Asm inhibition might have direct toxic effects in Tconvs to which Tregs are relatively resistant. To address this question in more detail, we set up an in vitro system that mirrors the observed increase in Treg frequency among CD4⁺ T cells upon Asm inhibitor treatment in vivo. We thus cultured splenocytes from wt C57BL/6 mice in the presence of IL-2 and increasing concentrations of amitriptyline or desipramine for 3 d. As after in vivo Asm inhibitor treatment, desipramine (and amitriptyline, data not shown) induced a significant increase in the frequency of Tregs among CD4⁺ T cells (Fig. 5A). The increase in Treg frequencies was due to a decrease in the absolute number of Tconvs in these cultures as the absolute numbers of Tregs remained constant (Fig. 5A). Depletion of Tregs before initiation of the cultures revealed that the observed increases in Treg frequencies were not due to a de novo induction of Foxp3 in Tconvs (data not shown).

The comparison of Treg-depleted lymph node cells or purified Tconvs (data not shown) from wt and Asm^−/− mice cultured in the presence of Tregs from wt mice revealed that the reduction in Tconv numbers was indeed a direct effect of desipramine on the Asm-expressing Tconvs and not indirectly mediated by inhibition of the Asm in Tregs (Fig. 5B). These data further demonstrate that desipramine reduced Tconv numbers not through off-target effects, but by blocking Asm activity.

In the absence of IL-2, desipramine decreased the absolute cell numbers of both Tregs and Tconvs, and Treg frequencies among CD4⁺ T cells remained unchanged (Fig. 5C). Apart from providing a survival advantage, the proliferation of Tregs induced by IL-2 was also resistant toward amitriptyline-mediated Asm inhibition up to a concentration of 2.5 μM (Fig. 5D). Together these data indicate that IL-2 selectively protected Tregs, but not Tconvs, from desipramine- or amitriptyline-induced cell death, thus leading to higher Treg frequencies among CD4⁺ T cells.

To demonstrate biological consequences of Asm deficiency in vivo, we infected Asm-deficient, heterozygote, and wt mice intracerebrally with recombinant MV expressing eGFP and a hemagglutinin that enables the infection of mouse neurons (Fig. 6A), which is a sensitive model for the activity of Tregs as previously described (28). The numbers of infected neurons in brains were analyzed in coronal sections during the acute phase of the infection (7 d postinfection [dpi]) and when the infection had entered the persistent phase during which a lower number of neurons remained infected (28 dpi; Fig. 6B, 6C, respectively). Seven dpi ∼10 times more (Fig. 6D) and 28 dpi 30–40 times more (Fig. 6E) neurons were infected in Asm^−/− mice than in control mice. Heterozygote Asm^−/+ mice were infected to a similar extent as wt mice. The massive increase in infection in Asm^−/− mice suggests that their antiviral immune response is not fully functional. This was supported by the observation that less MV-specific D^bMV-H_22–30-pentamer⁺ CD8⁺ T cells were present in spleen, cervical lymph nodes, and the brain of infected Asm^−/− mice in comparison with wt mice (Fig. 6F).

Because amitriptyline treatment of healthy mice recapitulated the phenotype of Asm^−/− mice regarding the Treg/Tconv balance (Fig. 4), we used Foxp3-DTR mice (47) to study the contribution of Tregs to the reduced control of the MV infection in the absence of Asm activity (Fig. 7A). Treatment of Foxp3-DTR mice with DT led to a depletion of Tregs as compared with wt littermate controls (Fig. 7B). As expected in this mouse strain, concomitant with Treg depletion, Tconv activation was increased, reflected by the higher frequencies of CD25⁺ Foxp3⁻ Tconvs in these animals (Fig. 7B). Similar to Asm^−/− mice, treatment of the MV-infected control mice (DTR wt) with amitriptyline (Fig. 7A) increased the number of infected neurons in the brain (Fig. 7C). This increase in viral load required the presence of Tregs because amitriptyline had no effect on the MV infection in mice (DTR y/+) depleted of Tregs (Fig. 7C). In the presence of Tregs, genetic or pharmacologically induced Asm deficiency was thus associated with poor control of the MV in the mouse model.

In this study, we report that Tregs are more resistant toward genetically or pharmacologically induced Asm deficiency than Tconvs. Therefore, therapeutic inhibition of the Asm provides a means to preferentially suppress responses of Tconvs, while not only sparing but probably even activating Tregs, which further potentiates the direct immunosuppressive effects of Asm inhibitors on the Tconvs.

Measuring Asm activity in Tregs and Tconvs, as well as our in-depth analysis of the sphingolipid content and the analysis of membrane lipid order of both populations isolated either from wt or Asm^−/− mice, revealed that the Asm increases the ceramide/SM ratio in Treg versus Tconvs and lowers the membrane lipid order in the Tregs (Fig. 1). Expression of SM synthase 1 (Sgms1 gene), which generates SM from ceramide, is suppressed by Foxp3 (17). This, of course, facilitates accumulation of ceramide in Tregs as opposed to Tconvs. With regard to lipid membrane order, we noted that Tregs and Tconvs with an effector/memory-like phenotype showed similar frequencies of cells with low membrane lipid order. In healthy mice these effector/memory-like Tconvs are mainly cells undergoing homeostatic proliferation (42, 43). Because Tregs are also hyperproliferative in vivo, low membrane lipid order appears to be positively correlated with short half-life in vivo. However, the comparison of wt and Asm^−/− mice suggested that Asm activity is crucial for the low lipid order of Tregs of wt mice, whereas other mechanisms/pathways appear to be operating in effector/memory-like Tconvs. Moreover, the biology of Tregs and effector/ memory-like Tconvs is generally not very close because the proliferation of the effector/memory-like Tconvs is mainly induced by IL-7 (48), whereas Tregs respond to IL-2, TCR, and CD28 signaling (1–5). Moreover, true effector/memory T cells generated on purpose in vitro (23, 24) (mouse T cells; our unpublished observations) or during pathogenic immune responses in vivo (26, 49) are susceptible toward pharmacological inhibition of the Asm, which, as we show in this article, differentiates them from Tregs.

The high membrane order of Tregs in the absence of Asm is supposed to facilitate T cell activation (40). Our results confirmed that Tregs of Asm-deficient mice were in a more activated state and more suppressive than Tregs of wt mice (Fig. 2). This higher suppressive activity was associated with an increased turnover of the major suppressive effector molecule CTLA-4 after culture of splenocytes in the presence of IL-2 only and upon CD3 stimulation. Together this suggests that the increased CTLA-4 turnover we observed in vitro is not secondary to preactivation of the Tregs of Asm^−/− mice in vivo. Rather, the data imply that the normal presence of Asm in lysosomes may slow down the transport of CTLA-4 from its intracellular stores to the plasma membrane.

In vivo, the frequency of Tregs among CD4⁺ T cells, but not absolute cell numbers, was increased in Asm^−/− compared with wt mice. In this study, the increased turnover of CTLA-4 may restrict Treg expansion despite stronger activation of the Tregs in these animals (50). As demonstrated, the increased Treg frequency among CD4⁺ T cells in Asm^−/− mice was due to a decrease in the absolute numbers of Tconvs (Fig. 3). This is in line with the apoptosis-inducing effects of selective serotonin reuptake inhibitors, which, at the dosages required for apoptosis induction, also potently inhibit the ASM (25). However, from these data we could not rule out that the phenotype of the Asm^−/− mice, particularly the differential impact of the Asm deficiency on Tconvs and Tregs, was not due to mechanisms compensating for the gene defect. Therefore, we treated healthy wt C57BL/6 mice with amitriptyline to test whether short-term Asm blockade would also shift the Treg–Tconv balance in favor of the former. This was indeed the case, making it a very important finding of our study (Fig. 4). In agreement with our results, Zhang and coworkers (26) found that repeated administration of desipramine slightly upregulated the frequency of CD4⁺ CD25⁺ Foxp3⁺ Tregs while almost halving the frequency of pathogenic IL-17⁺ cells among CD4⁺ T cells in a BALB/c mouse model of allergic rhinitis. As in the Asm^−/− mice, amitriptyline treatment of wt C57BL/6 mice did not increase absolute numbers of Tregs, but did decrease the numbers of Tconvs. This suggested that, indeed, Tregs were far less sensitive toward Asm deficiency than Tconvs.

To better understand the reasons for the increased frequencies of Tregs among CD4⁺ T cells in Asm^−/− and amitriptyline-treated wt mice, we conducted a series of in vitro experiments. With these we could make out Tconvs as the critical target population for Asm inhibition/deficiency (Fig. 5). Only when Asm activity was blocked in Tconvs did we see an increase in the Treg frequency among CD4⁺ T cells. This relative increase in Tregs was again not due to Treg expansion, but the result of reduced numbers of Tconvs. Subsequently we found that the resistance of the Tregs toward Asm inhibition was brought about by their responsiveness toward IL-2 and that IL-2–induced proliferation of Tregs was also relatively resistant toward Asm inhibition. We thus assume that also in vivo the constant supply of survival signals, that is, via the IL-2R, but maybe also CD28 and/or the TCR, protected Tregs from cell death upon amitriptyline treatment or in Asm^−/− mice. One reason why IL-2 might efficiently rescue Tregs from Asm-induced cell death could be that upon IL-2R signaling Zn²⁺ release from lysosomes is accompanied by STAT5 phosphorylation (51). Although Zn²⁺ release from lysosomes has been shown to induce apoptosis in nonleukocytic cells (52, 53), STAT5 phosphorylation protects myeloid cells from apoptosis (54). Because Tregs also strongly rely on STAT5 signaling for survival and function (55), it is very likely that also in these cells STAT5 phosphorylation is capable of counteracting the proapoptotic effect of Zn²⁺ released from lysosomes into the cytosol.

Based on the predictive value of culturing mouse splenocytes in the presence of IL-2 and desipramine or amitriptyline in vitro for the outcome of amitriptyline treatment of normal mice in vivo, we assume that therapeutic inhibition of ASM in humans also positively affects the Treg compartment. In line with this assumption CD4⁺ CD25⁺ Tregs have been reported to be reduced in patients suffering from major depression (56) and to increase during antidepressant therapy (57). Whether this is a direct effect of the treatment with ASM inhibitors or whether it is secondary to clinical improvement remains to be determined. Because inflammatory processes appear to contribute to the onset of major depression (58) it may well be that Treg activation and relative expansion by ASM inhibitors efficiently dampen inflammation in these patients.

We used the mouse MV infection model to study the impact of Asm deficiency on resistance toward viral infections of the CNS. The poor control of MV infection by Asm^−/− mice is in line with an impaired immune response, but also with in vitro and in vivo investigations revealing that Asm can play a role in bacterial and viral uptake and replication, and that Asm-deficient mice in most cases are more susceptible to infections (59–64). In tissue culture we obtained no hints that MV might be taken up or replicates better in Asm-deficient cells (data not shown). Of particular importance to our study, however, we observed a reduced number of MV-specific CD8⁺ T cells in secondary lymphoid organs and the brains of MV-infected mice (Fig. 6). Interestingly, a diminished capacity of CD8⁺ T cells to exert effector functions has been observed in Asm^−/− mice infected with LCMV (14). However, in contrast with the LCMV model, already clonal expansion of virus-specific CD8⁺ T cells was reduced in brain-draining secondary lymphoid organs, that is, the cervical lymph nodes, of MV-infected Asm^−/− mice (Fig. 6). This indicated to us that mechanisms other than the known defects of CD8⁺ T cells to exert effector functions accounted for their reduced clonal expansion after MV infection in Asm-deficient mice. Indeed, using Foxp3-DTR mice and amitriptyline treatment, we observed that Asm inhibition led to poorer viral control only when Tregs were present in these animals (Fig. 7). Therefore, the shift in the balance of Tconvs and Tregs induced upon Asm inhibition in vivo is of biological significance leading to reduced viral control. Unlike our previous study using DEREG mice for Treg deletion (29), we did not observe enhanced viral clearance in Foxp3-DTR mice (Fig. 7), which we think could be because of differences in the time needed to achieve Treg depletion, the dose of DT applied, and the completeness of Treg depletion achieved with the two different mouse strains.

Treg activation and increasing the Treg/effector T cell ratio has been a therapeutic goal in autoimmune diseases like multiple sclerosis and type 1 diabetes or strongly inflammatory conditions like graft-versus-host disease for many years now. Therefore, we suggest further evaluation of the use of ASM inhibitors for the treatment of these conditions in humans.

We thank Nelli Wolf, Christian Linden (cell sorting facility of the IZKF Würzburg), Anika Grafen, and Isabell Schreck for technical assistance, and Sibylle Schneider-Schaulies for helpful discussions and critical reading of the manuscript.

The authors have no financial conflicts of interest.