Tuning IL-2 signaling by ADP-ribosylation of CD25

Control of immunologic tolerance and homeostasis rely on Foxp3+CD4+CD25+ regulatory T cells (Tregs) that constitutively express the high affinity receptor for Interleukin-2, CD25. Tregs proliferate in response to injections of IL-2/anti-IL-2 antibody complexes or low doses of IL-2. However, little is known about endogenous mechanisms that regulate the sensitivity of CD25 to signaling by IL-2. Here we demonstrate that CD25 is ADP-ribosylated at Arg35 in the IL-2 binding site by ecto-ADP-ribosyltransferase ARTC2.2, a toxin-related GPI-anchored ecto-enzyme. ADP-ribosylation inhibits binding of IL-2 by CD25, IL-2- induced phosphorylation of STAT5, and IL-2-dependent cell proliferation. Our study elucidates an as-yet-unrecognized mechanism to tune IL-2 signaling. This newly found mechanism might thwart Tregs at sites of inflammation and thereby permit a more potent response of activated effector T cells.

I nterleukin 2 (IL-2) activates multiple immune cells, but mounting evidence indicates that its primary function is to support the generation and survival of Tregs that inhibit immune responses and prevent autoimmune diseases [1][2][3][4] . IL-2 signals through high and low affinity cell surface receptors [5][6][7] . CD25, the a-chain of the IL-2 receptor, is constitutively associated with lipid rafts 8 . Assembly of the heterotrimeric high affinity receptor complex is initiated by binding of IL-2 to CD25, followed by recruitment of CD122 and CD132. In the absence of CD25, the b and c chains form a receptor with a 10-100 fold lower affinity for IL-2 9 . This low affinity receptor is expressed by naïve T cells, memory CD8 1 T cells, and NK cells. Both receptors signal via phosphorylation of STAT5 (signal transducer and activator of transcription 5) [10][11][12] . The crystal structure of IL-2 in complex with its receptor revealed the residues of CD25 involved in binding of IL-2 6,7 , including a prominent arginine doublet (R35R36) (Supplementary Fig. 1a, b).
T cell subsets are differentially modulated by IL-2 signaling, depending on the concentration of IL-2 and on the expression of the IL-2 receptor subunits. CD25 is constitutively expressed at high levels by Tregs and IL-2 signaling through CD25 is crucial for the generation, survival and function of these cells [1][2][3][13][14][15] . By efficient consumption of IL-2, Tregs can deprive neighboring T cells of this cytokine 16,17 . Recent studies show that in vivo treatment with low doses of IL-2 or with IL-2 in complex with anti-IL-2 antibodies can suppress immunemediated diseases by inducing the expansion of Tregs [17][18][19][20][21][22] . IL-2-specific antibodies can prolong the serum half-life of IL-2 by inhibiting renal filtration of the low molecular weight cytokine 23 . Intriguingly, different IL-2/anti-IL-2 antibody complexes induce distinct responses in vivo, depending on the epitope of IL-2 bound by the particular antibody: mouse IL-2 in complex with mAb (monoclonal antibody) JES6-1A12 or human IL-2 in complex with mAb 5344 induce the preferential expansion of Tregs, whereas mouse IL-2 in complex with mAb S4B6 or human IL-2 in complex with mAb-602 induce the preferential expansion of CD8 1 cytotoxic T cells and NK cells [23][24][25][26][27][28] . These findings raise the question whether endogenous mechanisms exist that could similarly modulate IL-2 signaling in vivo.
NAD 1 -dependent ADP-ribosylation is one of the posttranslational modifications regulating protein functions 29,30 . NAD 1 (nicotinamide adenine dinucleotide) is released from cells during inflammation and acts as a signaling molecule that alerts immune cells to sites of tissue damage [31][32][33][34] . The NAD 1 -dependent toxin-related ADPribosyltransferase ARTC2.2 is a GPI-anchored enzyme expressed on the cell membrane of naïve T cells and Foxp3 1 CD4 1 CD25 1 regulatory T cells (Tregs) [35][36][37] . ARTC2.2 transfers the ADP-ribose moiety from NAD 1 to arginine residues of other raft-associated membrane proteins 35,38 . For example, ADP-ribosylation of the P2X7 ion channel at R125 induces gating of the ion channel 38 . ADP-ribosylation of cell surface proteins can be monitored by autoradiography or by antibody-based immunoassays ( Supplementary Fig. 1c, d).
Here we uncover ADP-ribosylation of CD25 as a novel mechanism that controls IL-2 signaling following exposure of Tregs to extracellular NAD 1 . Using affinity purification of ADP-ribosylated cell surface proteins and mass spectrometry, we identify CD25 as a major target of ARTC2.2. We show that ARTC2.2-catalyzed, NAD 1dependent ADP-ribosylation of CD25 inhibits IL-2 binding, IL-2-induced phosphorylation of STAT5, and IL-2-dependent cellproliferation. We propose that ADP-ribosylation of CD25 provides an endogenous mechanism to tune IL-2 signaling in response to NAD 1 released from injured cells during inflammation and tissue damage. By inhibiting formation of the high affinity IL-2 receptor naturally expressed by regulatory T cells, this mechanism could allow the preferential expansion of activated effector T cells in inflammatory settings.

Results
Identification of CD25 as a major target of ARTC2.2. The YAC-1 lymphoma cell line constitutively expresses ARTC2.2 and CD25 (Fig. 1a). Incubation of these cells with 32 P-NAD 1 resulted in the labeling of several bands, including a prominent band of 50 kD (Fig. 1b, lane 1, arrowhead). Similar results were obtained when etheno-NAD 1 was used as substrate (Fig. 1d, lane 1, arrowhead). Affinity purification of etheno-ADP-ribosylated proteins from YAC-1 cell lysates (Fig. 1c, d) followed by mass spectrometry analyses (Fig. 1e) identified the protein in this band as CD25. Immunoprecipitation analyses with CD25-specific antibodies verified the identity of the protein in this prominent 50 kD band as CD25, i.e. this band was depleted from cell lysates and was recovered in the immunoprecipitate (Fig. 1b, compare lanes 1, 2, and 4).
ARTC2.2 ADP-ribosylates CD25 at R35 in the IL-2 binding site. The 3D-structure of human IL-2 in complex with its heterotrimeric receptor reveals that all eight arginine residues in the visible portion of the extracellular domain of CD25 are localized on the surface of the protein and thus are potentially accessible for ADP-ribosylation ( Supplementary Fig. 1b, Fig. 2a). In order to test which of these arginines could be ADP-ribosylated by ARTC2.2, we individually substituted each of the arginine residues with lysine by sitedirected mutagenesis. In addition, we purchased a synthetic cDNA construct of CD25, in which all 11 arginines were substituted by lysine (RallK). In this construct, we individually substituted each lysine with arginine, thereby generating variants of CD25 carrying only a single arginine residue. Each mutant was co-transfected into HEK (human embryonic kidney) cells together with ARTC2.2. 40 h post transfection, cell surface expression levels were assessed by flow cytometry using ARTC2.2 and CD25-specific mAbs ( Supplementary  Fig. 2). Parallel aliquots of cells were incubated with 32 P-labeled NAD 1 . Radiolabeling of CD25 and other cell surface proteins was assessed by SDS-PAGE autoradiography (Fig. 2b, c).
The results revealed that each of the single R . K mutants of CD25 still served as a target for ARTC2.2, suggesting that CD25 is ADPribosylated at more than one site (Fig. 2b). Similar results were obtained for mouse CD25 (Supplementary Fig. 3). This is reminiscent of previous reports on the ADP-ribosylation of P2X7 and of defensin 1, both of which are ADP-ribosylated at more than one arginine 38,39 . RallK, the mutant in which all arginines were replaced by lysine, did not incorporate any radiolabel (Fig. 2c, lane 1) despite robust expression on the cell surface ( Supplementary Fig. 2). This finding confirms that ADP-ribosylation of CD25 by ARTC2.2 is arginine-specific. Four of the RallK reversion mutants carrying a single arginine residue were radiolabeled by ARTC2.2, indicating that these sites can, in principle, serve as ADP-ribosylation sites for ARTC2.2 (Fig. 2c, R32, R35, R67, R140). Three of these residues (R32, R35, R140) are conserved in mouse CD25 ( Supplementary  Fig. 4). The fourth residue (R67) is located in a linker between the two sushi domains that is not visible in the 3D structure. The stronger labeling of the R67 revertant indicates that this residue is more accessible for ARTC2.2 than the other revertants. R35 is part of an arginine doublet (R35/R36), as are the primary ADP-ribosylation sites on P2X7 and defensin 1 38,39 . R35 is located within the IL-2 binding interface ( Supplementary Fig. 1), R32 and R140 are located on the edges of the first and second Sushi domains, respectively, and R67 is located on the flexible linker between the two sushi domains that is not visible in the 3D-structure of the IL-2 receptor complex ( Supplementary Fig. 4).
ADP-ribosylation inhibits binding of mAb 7D4 but not mAb PC61. ADP-ribosylation results in the attachment of a bulky residue and imparts a local charge change (from a positively charged arginine to a negatively charged ADP-ribosylarginine) 40 . It is, therefore, conceivable that ADP-ribosylation of CD25 at or near an antibody binding-site could hinder antibody binding. In order to address this question, we incubated mouse splenocytes in the absence or presence of NAD 1 followed by staining with the CD25-specific mAbs PC61 and 7D4 (Fig. 3). The results reveal a NAD 1 dosedependent inhibition of staining by 7D4 but little if any effect of NAD 1 on the binding of PC61 (Fig. 3b), indicating that ADPribosylation affects the epitope recognized by 7D4 but not the epitope recognized by PC61.
In order to assess whether ADP-ribosylation affects IL-2dependent cell proliferation, CTLL-2 and CTLL-2 ARTC2.2 cells were depleted of IL-2 by washing with PBS followed by incubation for 6 hours in IL-2 free medium. Cells were then seeded in 24-well plates in medium containing serial dilutions of IL-2 and cells were cultivated further in the absence or presence of NAD 1 for 3.5 days. Cell numbers were assessed by flow cytometry using Trucount beads (Fig. 4d). The results show that NAD 1 does not affect the proliferation of parental CTLL-2 cells whereas it markedly inhibits the proliferation of CTLL-2 ARTC2.2 cells at low and intermediate doses of IL-2 (below 5 ng/ml). These results suggests that ADP-ribosylation of cell surface proteins interferes with IL-2 binding and with IL-2 signaling.
Naturally occurring Foxp3 1 Tregs constitutively express CD25 and ARTC2.2 37 . DEREG-transgenic mice express a diphtheria-toxin receptor-GFP fusion protein under control of the foxp3 promoter, allowing visualization of living Tregs by flow cytometry 42 . The availability of ARTC2 2/2 DEREG mice 37 permits a direct assessment of the effect of NAD 1 -dependent ADP-ribosylation on Tregs, e.g. by comparing the responses of Tregs from ARTC2 2/2 and wildtype DEREG mice to extracellular NAD 1 (Fig. 5c and Supplementary Fig. 5). Treatment of FACS-sorted Tregs with extracellular NAD 1 inhibited IL-2-induced phosphorylation of STAT5 (Fig. 5c). Similarily, treatment of unsorted splenocytes with NAD 1 inhibited phosphorylation of STAT5 in wildtype Tregs cells but not in ARTC2 2/2 Tregs (Supplementary Fig. 5a-c). We have previously shown that ADPribosylation induces gating of the P2X7 ion channel, resulting in Calcium influx and externalization of phosphatidylserine 38 . In order to determine whether the inhibition of IL-2-induced STAT5 phosphorylation was mediated by P2X7, we analyzed the effects of extracellular NAD 1 also on cells from P2X7 2/2 (ARTC2 1/1 ) DEREG mice ( Supplementary Fig. 5e). The results show that NAD 1 -treatment inhibits STAT5 phoshporylation also in purified Tregs from P2X7 2/2 (ARTC2 1/1 mice), indicating that this effect is independent of P2X7 activation.

Discussion
The results presented here identify CD25, the a chain of the interleukin-2 receptor, as a major target for ADP-ribosylation by the toxinrelated ecto-enzyme ARTC2.2 (Fig. 1) and provide insight into the role of ADP-ribosylation of CD25. ADP-ribosylation assays in HEK cells co-transfected with ARTC2.2 and CD25 mutants identified R35 at the arginine doublet R35R36 as a target site for ADP-ribosylation (Fig. 2). This is reminiscent of murine P2X7 and human defensin 1, both of which contain double arginines as primary sites of ADPribosylation, but are additionally ADP-ribosylated on other arginine residues 38,39 . R35 is localized within the IL-2 binding region ( Supplementary Figs. 1d, 4) 7,43 . Considering that ADP-ribosylation results in the attachment of a bulky residue, converting a positively charged arginine into a negatively charged ADP-ribosylarginine, it is not surprising that ADP-ribosylation at this residue interferes with binding of IL-2 (Fig. 4c). Consistently, ADP-ribosylation of CD25 inhibits IL-2-induced phosphorylation of STAT5 (Fig. 5a, c) as well as IL-2-dependent cell proliferation (Fig. 4f). Of note, ADPribosylation of CD25 blocks binding of mAb 7D4 (Fig. 3), an  antibody that is commonly used for bead-assisted sorting of Tregs. Considering that endogenous NAD 1 can be released during manipulation of cells 37,44 , inadvertent ADP-ribosylation of CD25 prior to staining with 7D4 may hamper bead-assisted sorting of Tregs. Indeed, NAD 1 -mediated ADP-ribosylation of CD25 may explain the reported difficulty to purify Tregs from CD38ko mice using bead-assisted sorting 45 since lack of the major cell surface NADhydrolase in these mice causes exposure of cells to higher levels of extracellular NAD 137,46 . We recently described a simple means to prevent inadvertent ADP-ribosylation of cell surface proteins during preparation of cells from spleen or liver, i.e. systemic injection of an ARTC2.2-blocking nanobody ten minutes prior to sacrifice of mice 47 . Figure 6 illustrates a model whereby ADP-ribosylation of CD25 provides a regulatory mechanism for tuning IL-2 signaling, i.e. from CD25-dependent high affinity signaling to CD25-independent low affinity signaling via CD122/CD132. In a non-inflammatory environment, where the concentration of extracellular NAD 1 is low, consumption of extracellular IL-2 by Tregs would deprive other T cells and NK cells of IL-2 (Fig. 6a) 48 . Similarly, systemic injections of low doses of IL-2 may result in the preferential expansion of Tregs 17,[19][20][21]49 . In an inflammatory environment, where stressed and damaged cells release large amounts of NAD 1 into the extracellular compartment 44,50 , ADP-ribosylation of CD25 would divert IL-2 signaling away from Tregs to NK cells and CD8 1 cytotoxic T cells (Fig. 6b). This mechanism of tuning IL-2 signaling may be mimicked by certain IL-2 antibodies used in vivo (Fig. 6c), i.e. systemic injections of mouse IL-2 in complex with mAbs S4B6 or JES6-5HA or of human IL-2 in complex with Mab-602 cause preferential expansion of NK cells and cytotoxic T cells 22,24,27 . In addition to Tregs that constitutively express CD25, CD25 is also up-regulated by activated T cells following triggering of the TCR (T cell receptor) 51 ( Supplementary Fig. 6). Since TCR triggering induces metalloprotease-mediated shedding of ARTC2.2, activated T cells are rendered resistant to ADP-ribosylation of cell surface proteins 52 . Therefore, ADP-ribosylation of CD25 may be a Treg-specific regulatory mechanism. The finding that some Tregs appear unaffected by NAD 1 treatment (Fig. 5C, Supplementary Fig. 5) could be due to lower expression of ARTC2.2 or CD25 or due to differential localization of these antigens in the plasma membrane, e.g. inside/outside lipid rafts 8,35 . Future studies should address potential differences of distinct subpopulations of Tregs in their sensitivity to NAD 1 -mediated regulation of IL-2 signaling.
In summary, we have described here a previously unknown mechanism for tuning signaling by IL-2 by ADP-ribosylation of CD25. Operation of this mechanism in vivo likely depends on the context in which T cells are exposed to NAD 1 . Thus, at sites where NAD 1 is released in large quantities from damaged cells, such as during a lytic viral infection, ADP-ribosylation of CD25 on Tregs would favor proliferation and function of CD8 1 effector T cells, thereby enhancing pathogen eradication. In contrast, in healthy tissues, where little if any NAD 1 is released, Treg function would not be inhibited by ADP-ribosylation, permitting IL-2 mediated expansion of Tregs and efficient suppression of potentially auto-reactive T cells.

Methods
Mice and cells. ARTC2 2/2 mice 53 and P2X7 2/2 mice 54 were backcrossed to C57BL/6 WT and DEREG mice 42 on the C57BL/6 background for 12 generations and were maintained under specific pathogen-free conditions at the central animal facility of the UKE. All experiments were performed according to state guidelines with approval of the local institutional regulatory committee (registration numbers ORG153 and A10a). The YAC-1 lymphoma cells, kindly provided by Jürgen Löhler, Heinrich Pette Institute, Hamburg, Germany, were cultured in RPMI-1640 supplemented with 10% FCS, 2 mM glutamine, 2 mM sodium pyruvate. The IL-2-dependent CTLL-2 cellline, kindly provided by Marc Pallardy, INSERM U 461, Châtenay-Malabry, France, was maintained in complete medium supplemented with 10 ng/ml human recombinant IL-2 (10000 U/ml, Roche) and 0.0001% 2-mercaptoethanol. CTLL-2 cells were stably transfected with an expression construct for ARTC2.2 (pME.CD8LF-ARTC2.2) 55 . Tregs were FACS-sorted as GFP 1 cells from primary spleen cells of DEREG mice at 4uC. Alternatively, after erythrocyte lysis with Ack lysis buffer (Bio Whittaker) and depletion of B-cells with Dynabead-conjugated sheep anti-mouse IgG (Invitrogen), Tregs were positively selected by magnetic cell sorting using PE conjugated anti-CD25 (7D4) and magnetic bead-conjugated anti-PE antibodies according to the manufacturer's instructions (Miltenyi Biotec). Purity of Tregs was monitored by flow cytometry and was .90%.
Site directed mutagenesis and cell transfections. The coding regions for mouse and human CD25 were PCR-amplified from YAC-1 lymphoma cells and from ConAstimulated PBLs, respectively, and cloned into the pCMV-Sport6 vector (Invitrogen). Arginine residues were substituted with lysine using site-directed mutagenesis. A synthetic cDNA encoding CD25 in which all arginine residues in the extracellular domain were replaced by lysine, was purchased from Geneart (Invitrogen) and cloned into pCMV-Sport6. Expression constructs (2 mg per 10 6 cells) were transfected into CTLL-2 cells by electroporation and into HEK cells with the jetPEI transfection reagent (Polyplus).
Purification of etheno-ADP-ribosylated proteins and mass spectrometry. YAC-1 cells (10 9 cells in 10 ml) were incubated for 15 min at RT with 10 mM etheno-NAD 1 (Sigma) and washed twice in PBS before lysis in 1% Triton-X 100 for 20 min at 4uC. Cell lysates were clarified by high speed centrifugation (20 min 15.000 g) before passage through a 1 ml affinity column containing mAb 1G4 immobilized on Amino-Link Matrix (Pierce). The column was extensively washed with 1% Triton X-100 and PBS, bound proteins were eluted with 1 mM etheno-adenosine in PBS/ 0.1% Triton-X-100. Eluted proteins were size fractionated by SDS-PAGE. Proteins were stained with Coomassie blue and visible bands were cut out. A parallel gel was blotted onto a PVDF membrane and subjected to Western-Blot analyses with etheno-adenosine specific mAb 1G4 and peroxidase-conjugated anti-mouse IgG. A www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 8959 | DOI: 10.1038/srep08959 prominent 50 kD band was excised from the gel. Following reduction of disulfide bonds with DTT (10 mM, 56uC, 30 min) and modification of cysteines with iodoacetamide (55 mM, RT, 20 min), proteins were subjected to in-gel digestion with trypsin (5 ng/ml in 50 mM NH 4 HCO 3 , 37uC, 16 h). Peptides were extracted with 50% acetonitrile/5% formic acid, dried in a vacuum concentrater, redisolved in 5% methanol/5% formic acid desalted on a C18 mZipTip (Millipore), eluted with 1 ml 60% methanol/5% formic acid, and analyzed by nano-electrospray mass spectrometry in a QTOF II instrument (Micromass). MS/MS spectra obtained by collision induced fragmentation after manual precurser selection were evaluated both manually and by the Mascot MS/MS search algorithm version 2.2 (Matrix Sciences, London, UK).
Radio-ADP-ribosylation assays. HEK cells were transiently co-transfected with ARTC2.2 (1 mg/10 6 cells) and CD25 (2 mg/10 6 cells) in solution and plated in parallel aliquots onto 6-well plates pre-coated with poly-L-lysine. 48 h post transfection, one aliquot was used to monitor transfection efficiency by FACS analyses using mAbs directed against ARTC2.2 and CD25. The adherent cells of the other aliquot were gently washed with pre-warmed serum-free X vivo medium (Bio Whittaker) and then incubated in serum free X vivo medium containing 32 P-NAD 1 (2,5 mCi, 0,4 mM) for 20 min at 37uC. Cells were gently washed with pre-warmed X vivo medium and then lysed in PBS, 1% Triton-X100, 1 mM AEBSF (Sigma) for 20 min at 4uC. Insoluble material was pelleted by high-speed centrifugation (15 min 13.000 g) and solubilized membrane proteins were size fractionated on precast SDS-PAGE gels (Invitrogen) (1 3 10 5 cell equivalents/lane). Proteins were detected by Coomassie staining of the gels. For detection of radiolabeled proteins, gels were exposed to an X-ray film at 280uC for 6h-6d.
Immunoprecipitation of CD25. The CD25-specific mAb PC61 was conjugated to amino-link beads according to the manufacturer's instructions (Pierce). After incubation with cell lysates for 60 min at 4uC, beads were washed with PBS, 1% Triton-X 100. Samples were then incubated at 70uC for 15 min and the matrix was pelleted by centrifugation. Proteins eluted from the matrix were analyzed by SDS-PAGE autoradiography as described above.
Cell proliferation assay. In order to deplete CTLL-2 cells from IL-2, cells were washed twice and incubated for 6 h in culture medium without IL-2. Cells were seeded in a 24-well plate (3 3 10 4 cells per well) and cultivated in medium containing serial dilutions of IL-2. A small aliquot of medium containing or lacking NAD 1 (final concentration 100 mM) was added every 12 hours. Cell numbers were determined after four days by flow cytometry using Trucount TM beads (BD).
STAT5 phosphorylation assay. Primary splenocytes and purified Tregs (1 3 10 5 cells/100 ml) were incubated in the absence or presence of 30 mM NAD 1 for 15 min at 4uC before addition of 2.5 U/ml mouse IL-2 (eBioscience) and further incubation for 5 min at 37uC. The cells were then fixed in 2% PFA for 10 min at 37uC and 90% methanol at 220uC before staining with APC-conjugated anti-phospho-STAT5 according to the manufacturer's instructions (BD). Splenocytes were counterstained with anti-CD4 before incubation with IL-2. In some experiments, cells were preincubated for 15 min in the presence of the ARTC2.2-blocking nanobody s116a (50 ng/ml) 41 .