Abstract
ARTC2.2 is a toxin-related, GPI-anchored ADP-ribosyltransferase expressed by murine T cells. In response to NAD+ released from damaged cells during inflammation, ARTC2.2 ADP-ribosylates and thereby gates the P2X7 ion channel. This induces ectodomain shedding of metalloprotease-sensitive cell surface proteins. In this study, we show that ARTC2.2 itself is a target for P2X7-triggered ectodomain shedding. We identify the metalloprotease cleavage site 3 aa upstream of the predicted GPI anchor attachment site of ARTC2.2. Intravenous injection of NAD+ increased the level of enzymatically active ARTC2.2 in serum, indicating that this mechanism is operative also under inflammatory conditions in vivo. Radio–ADP-ribosylation assays reveal that shedding refocuses the target specificity of ARTC2.2 from membrane proteins to secretory proteins. Our results uncover nucleotide-induced membrane-proximal proteolysis as a regulatory mechanism to control the substrate specificity of ARTC2.2.
Introduction
Ectodomain shedding of cell surface proteins is a mechanism that allows the rapid coordinated response of T cells to environmental stimuli (1, 2). A potent physiological stimulus of ectodomain shedding is gating of the P2X7 ion channel by extracellular ATP and by NAD+-dependent ADP-ribosylation (3–6).
P2X7 functions as a T cell surface sensor for extracellular ATP and NAD+ (6–9). These nucleotides are released from cells during inflammation and tissue damage and modulate T cell functions (10–14). Gating of P2X7 induces externalization of phosphatidylserine, shedding of metalloprotease-sensitive membrane proteins, blebbing, and pore formation (5, 15–18). P2X7 is gated either directly by high concentrations of ATP acting as a soluble ligand, or indirectly by NAD+-dependent ADP-ribosylation, a covalent modification catalyzed by ecto–ADP-ribosyltransferase ARTC2.2 (5, 19).
ARTC2.2 is a toxin-related, GPI-anchored ecto–ADP-ribosyltransferase expressed on the surface of murine T cells (20, 21). In response to NAD+ released from cells during inflammation, ARTC2.2 catalyzes arginine-specific ADP-ribosylation of P2X7 and of other cell surface proteins (4, 9). Akin to phosphorylation, ADP-ribosylation is a reversible posttranslational modification that regulates the function of target proteins (22–25).
We previously reported that ARTC2.2 is shed by a metalloprotease following T cell activation (26). We hypothesized that ARTC2.2-mediated activation of P2X7 similarly induces shedding of ARTC2.2 itself, and that the release of ARTC2.2 from the cell surface into the inflammatory environment may alter the substrate specificity of ARTC2.2.
The aim of this study was to elucidate the mechanism and functional consequences of NAD+-induced cleavage of ARTC2.2 from the cell surface. We used FACS and ELISA assays to monitor shedding of ARTC2.2 from T cells and transfected cells, and mass spectrometry of affinity-purified shed ARTC2.2 to identify its cleavage site. We monitored the release of ARTC2.2 as a soluble protein into serum upon injection of NAD+ and the ADP-ribosylation of substrate proteins on the cell membrane and in serum by radioactive ADP-ribosylation assays. Our results show that shedding of ARTC2.2 from the cell surface under inflammatory conditions turns the substrate specificity of ARTC2.2 from a subset of membrane proteins to a subset of secretory proteins.
Materials and Methods
Mice and cells
C57BL/6 mice were obtained from The Jackson Laboratory. ARTC2−/− mice (27) and ARTC2.2-transgenic mice (28) were backcrossed to C57BL/6 wild-type (WT) mice for 12 generations. All animal experiments were performed in accordance with local regulations. Single-cell suspensions were prepared in cold (4°C) RPMI 1640 medium from lymph nodes by gentle dissection and passage through Nitex membrane (125-μm mesh; Tetko). B cells were depleted by absorption to Dynabeads-conjugated sheep anti-mouse IgG (Invitrogen). NAD+ (50 mg/200 μl saline) or saline was injected i.v. 20 min prior to sacrifice (29). Yac-1 lymphoma cells, provided by J. Löhler (Heinrich Pette Institute, Hamburg, Germany), and DC27.10 cells, provided by B. Fleischer (Bernhard Nocht Institute, Hamburg, Germany), were cultured in RPMI 1640 supplemented with 10% FCS, 2 mM glutamine, and 2 mM sodium pyruvate. Expression constructs encoding P2X7, ARTC2.2, and CD62L (Genebank AJ489297, NM_019915.2, BC052681) (5 μg per 106 cells) were transfected into HEK cells using jetPEI transfection reagent (Q-Biogen) (9, 30).
Abs, nanobodies, and inhibitors
Rat mAbs against mouse ARTC2.2 (Nika102, Nika109) and against P2X7 (Hano43, Hano44) and nanobodies against ARTC2.2 (s-14, s+16a) were generated by cDNA immunization of rats and llamas, as described previously (31–34). Etheno-adenosine–specific mAb 1G4 was provided by R. Santella (New York, NY) (35). mAbs were conjugated to Alexa-647 fluorochrome (Molecular Probes), and nanobody s-14 was conjugated to agarose beads (Amino-Link matrix; Pierce), according to the manufacturer's instructions. The following mAbs were purchased from BioLegend or BD Biosciences: anti-CD11a (2D7), CD62L (MEL-14), CD8a (Ly-2), and LFA-1 (M17/4). The metalloprotease inhibitors GW (1 μM IC50), GI (1 μM IC50), and Marimastat (4 μM IC50) were provided by A. Chalaris (Kiel, Germany) (36, 37).
Flow cytometry assays
Yac-1 lymphoma cells, transiently transfected HEK cells, and primary lymph node T cells were stained with fluorochrome-conjugated Abs before or after incubation for 30 min at 4°C or at 37°C in RPMI 1640 medium in the presence of NAD+ or ATP (4, 9). Cells were washed and analyzed by flow cytometry (FACSCalibur or FACS-Canto II, BD, and FlowJo software; Tree Star).
ELISA
The ARTC2.2-specific mAbs Nika102 and Nika109 generated in our laboratory bind to distinct, nonoverlapping epitopes on ARTC2.2. These Abs were used to establish a sandwich ELISA. Nika109 (200 ng/100 μl PBS) was coated as a catcher Ab for soluble ARTC2.2 onto the wells of ELISA Nunclon hibond plates. For detecting bound ARTC2.2, Nika102 was biotinylated using the Amino-Link conjugation, according to the manufacturer's instructions (Pierce). Bound Nika102 was detected with peroxidase-conjugated streptavidin (GE Healthcare). Serial dilutions of purified shed ARTC2.2 were used for generating a standard curve.
Purification of ARTC2.2
HEK cells transiently cotransfected with ARTC2.2 and P2X7 were harvested 5 d posttransfection and treated with 5 mM ATP for 20 min at 37°C (5 × 108 cells in 10 ml). DC27.10 cells stably transfected with ARTC2.2 (30) were incubated with 500 μM ATP for 1 h at 37°C. Cell supernatants were clarified by centrifugation (20 min, 2000 × g), and ARTC2.2 was purified on an affinity column containing immobilized anti-FLAG mAb M2. The column was extensively washed with PBS, 1% Triton X-100, and ARTC2.2 was eluted with 100 mM glycine, 10 mM Tris (pH 2.7). The eluate was neutralized with one-tenth volume of 1 M Tris (pH 9.0), and the buffer was exchanged to PBS by gel filtration (PD-10 columns; Pharmacia). ARTC2.2 was concentrated using centrifugal filters (Millipore; molecular weight cutoff 10 kDa). For deglycosylation before analysis by mass spectrometry (MS), ARTC2.2 was captured with nanobody s-14 (33) immobilized on Amino-Link beads (Pierce) in Eppendorf tubes by gentle agitation for 30 min at room temperature. The beads were washed and incubated overnight with endoglycosidase F (New England Biolabs) in the presence of 1% SDS. Proteins were then size fractionated by SDS-PAGE and stained with Coomassie blue. The band corresponding to ARTC2.2 was excised from the gel and analyzed, as described below.
Liquid chromatography–MS-based identification of the ARTC2.2 cleavage site
After a wash step with water/acetonitrile (50:50, v/v), proteins in the excised gel bands were reduced with 10 mM DTT solution for 60 min at 56°C and washed again with water. Alkylation of cysteine thiol groups was performed by adding 55 mM iodoacetamide solution and incubation for 30 min at room temperature in the dark. Excess reagent was removed by five wash steps with water/acetonitrile (50:50, v/v). Enzymatic digests were carried out overnight at 37°C with 5 μg trypsin or 5 μg GluC in 40 mM ammonium carbonate (pH 8.5). Buffer was prepared by dilution of the salt in 95% H218O.
Peptide separation was performed using a U3000 nano-HPLC system equipped with a Probot microfraction collector (Dionex, Idstein, Germany). A sample volume of 20 μl was loaded on an Acclaim PepMap100 C18 trap column (5 μm, 0.3 × 10 mm) with a flow rate of 30 μl/min 0.1 aqueous trifluoroacetic acid (TFA), 3% acetonitrile (ACN) for 8 min at 30°C column temperature. After this desalting step, peptides were separated using an Acclaim PepMap100 C18 column (5 μm, 75 μm × 150 mm) with a flow rate of 300 nL/min with eluent A (0.05% aqueous TFA) and eluent B (20% deionized water, 0.04% TFA in 80% ACN) via the following linear gradient: 8–39 min, 5–50% B; 39–44 min, 50–95% B; 44–49 min, 95% B; 49–50 min, 95–5% B; 50–69 min, 5% B. Eluting peptides were immediately mixed with MALDI matrix (3 mg/ml α-cyano-4-hydroxycinnamic acid in 70% ACN, 0,1% TFA 30% deionized water, 5 nM Glu-1-fibrinopeptide B) and spotted onto a MALDI target (Opti-TOF liquid chromatography MALDI Insert; AB SCIEX, Darmstadt, Germany) in a ratio of 1:4 (v/v) in 15-s intervals from 15 to 65 min.
All MS and tandem MS (MS/MS) experiments were measured with an AB SCIEX TOF/TOF 5800 mass spectrometer (AB SCIEX, Darmstadt, Germany). MS data were acquired in positive ion mode in the mass range from 800 to 4000 m/z by accumulation of 1000 laser shots and in addition to the default calibration. MS spectra were internally calibrated on Glu-1-fibrinopeptide B spiked into the MALDI matrix. Selection criteria for generating the precursor list for the MS/MS experiments were as follows: minimum signal-to-noise ratio 100, precursor mass tolerance ± 200 ppm, and a maximum of 10 precursors per spot. Fragmentation was performed in CID mode with an energy of 1 kV by the accumulation of 2000 laser shots. Raw data were transcribed in Mascot Generic Format files using TS2Mascot Software (Matrix Sciences, London, U.K.) requiring a minimum signal-to-noise ratio of 10; only monoisotopic signals were extracted. Database analyses were performed against the SwissProt database (Mus musculus, 26 January 2012, 16,345 sequences) using Mascot 2.2.04 with semitrypsin and semi-GluC as enzyme, respectively; two missed cleavages; fixed modifications carbamidomethylation of cysteines; variable modifications of one 18O labeling C-terminal; two 18O labelings C-terminal, deamidation of asparagine and glutamine, deamidation of asparagine and glutamine with 18O, and oxidation of methionine; mass tolerance for precursors ± 50 ppm and for MS/MS fragments 0.5 Da. In addition, the database included the sequence of ARTC2.2 and the two proteases used for the sample digests.
Radio–ADP-ribosylation assays
ARTC2.2 was immunoprecipitated from the supernatant of NAD+- or ATP-treated T cells (5 × 106 cells) with mAb Nika-102 immobilized on Sepharose beads (Amino-Link matrix; Pierce) (20 μl matrix). Following extensive washing with PBS containing 1% Triton X-100, the beads were resuspended in 50 μl PBS containing 1 μM [32P]NAD+ (GE Healthcare) with or without 1 mM agmatin (Sigma-Aldrich) and incubated for 15 min at room temperature. Beads were pelleted by centrifugation. Reaction products in the supernatant were subjected to fractionation by TLC. Radiolabeled products were visualized by exposure of the chromatogram to x-ray film for 4 h at −80°C. For comparative analyses of ADP-ribosylation catalyzed by membrane-bound versus shed ARTC2.2, we used purified T cells from WT, ARTC2−/−, and ARTC2.2-transgenic mice following treatment with 250 μM ATP for 20 min at 37°C. Cells were then incubated with 0.5 μM [32P]NAD+ in the absence or presence of serum proteins for 20 min at 4°C. Cells and soluble proteins in cell supernatants were separated by centrifugation. Cells were washed at 4°C in PBS containing cold NAD+, and cell membrane proteins were then solubilized with 1% Triton X-100 for 20 min at 4°C. Cell lysates were clarified from nuclei and other insoluble proteins by high-speed centrifugation (15 min, 13,000 × g). The ARTC2.2 target proteins CD8, P2X7, and LFA-1 were immunoprecipitated from cell lysates with immobilized Abs, as previously described (5). Radiolabeled cell membrane proteins and radiolabeled proteins in cell supernatants were analyzed by SDS-PAGE autoradiography. To monitor ADP-ribosylation of serum proteins by shed ARTC2.2 in vivo, serum samples obtained from NAD+-treated WT, ARTC2−/−, or ARTC2.2-transgenic mice were incubated with 0.5 μM [32P]NAD for 20 min, and radiolabeled proteins were detected by SDS-PAGE autoradiography.
Results
Extracellular NAD+ induces downmodulation of CD62L and ARTC2.2
Yac-1 lymphoma cells endogenously express ARTC2.2, P2X7, and several other T cell surface markers. Treatment of Yac-1 cells with extracellular NAD+ resulted in downmodulation of CD62L, but not of CD11a (Fig. 1A). Remarkably, exposure to extracellular NAD+ also resulted in downmodulation of the ARTC2.2 ecto-enzyme from the cell surface. Downmodulation of CD62L and ARTC2.2 showed similar dose responses to extracellular NAD+. Note that this downmodulation occurred also when cells were labeled with mAbs against CD62L and ARTC2.2 before treatment of cells with NAD+ (Fig. 1A, dashed lines). This indicates that NAD+ induces release of these proteins into the cell supernatant rather than endocytosis of these proteins with the bound Ab, because the latter would not result in a decreased fluorescence. Similar results were obtained upon treatment of Yac-1 cells with ATP (data not shown).
Treatment of Yac-1 cells with extracellular NAD+ or ATP results in downmodulation of CD62L and ARTC2.2 by a P2X7-activated metalloprotease. (A) Yac-1 cells were stained with fluorochrome-conjugated mAbs against CD11a, CD62L, or ARTC2.2 either before or after incubation with different concentrations of NAD+ for 30 min. Cells were then analyzed by flow cytometry. Results are representative of three independent experiments. (B) Cells were incubated for 20 min with ARTC2.2-specific nanobodies s-14 or s+16a or with P2X7-specific mAbs Hano43 or Hano44 or without Abs (control) before addition of 50 μM NAD+. After further incubation for 30 min, cells were analyzed by flow cytometry. (C) Yac-1 cells were incubated with different concentrations of NAD+ or etheno-NAD+ (eNAD) for 30 min before staining with fluorochrome-conjugated mAbs against etheno-adenosine (eAdo), CD62L, or ARTC2.2 and flow cytometry analyses. (D) Yac-1 cells were incubated for 30 min in the absence or presence of NAD+ and the metalloprotease inhibitors GW or GI before staining with mAbs against CD62L or ARTC2.2 and flow cytometry analyses. Results are representative of two independent experiments.
Treatment of Yac-1 cells with extracellular NAD+ or ATP results in downmodulation of CD62L and ARTC2.2 by a P2X7-activated metalloprotease. (A) Yac-1 cells were stained with fluorochrome-conjugated mAbs against CD11a, CD62L, or ARTC2.2 either before or after incubation with different concentrations of NAD+ for 30 min. Cells were then analyzed by flow cytometry. Results are representative of three independent experiments. (B) Cells were incubated for 20 min with ARTC2.2-specific nanobodies s-14 or s+16a or with P2X7-specific mAbs Hano43 or Hano44 or without Abs (control) before addition of 50 μM NAD+. After further incubation for 30 min, cells were analyzed by flow cytometry. (C) Yac-1 cells were incubated with different concentrations of NAD+ or etheno-NAD+ (eNAD) for 30 min before staining with fluorochrome-conjugated mAbs against etheno-adenosine (eAdo), CD62L, or ARTC2.2 and flow cytometry analyses. (D) Yac-1 cells were incubated for 30 min in the absence or presence of NAD+ and the metalloprotease inhibitors GW or GI before staining with mAbs against CD62L or ARTC2.2 and flow cytometry analyses. Results are representative of two independent experiments.
NAD+-induced downmodulation of ARTC2.2 is mediated by ADP-ribosylation of P2X7
Previous studies have shown that gating of the P2X7 ion channel by extracellular ATP or by NAD+-dependent ADP-ribosylation induces ectodomain shedding of CD62L (4, 5). To assess whether the observed downmodulation of ARTC2.2 is also mediated by ARTC2.2-catalyzed ADP-ribosylation of P2X7, we tested whether available Abs against ARTC2.2 or P2X7 could block this phenomenon (Fig. 1B). Indeed, the known blockers of ARTC2.2-catalyzed ADP-ribosylation of P2X7, nanobody s+16a and mAb Hano43 (5, 33), inhibited NAD+-induced downmodulation of CD62L and ARTC2.2 (Fig. 1B). In contrast, the nonblocking nanobody s-14 and mAb Hano44 showed no effects. These results indicate that NAD+-induced downmodulation of ARTC2.2 is mediated by ADP-ribosylation of P2X7.
To further corroborate the role of P2X7 activation by NAD+-dependent ADP-ribosylation as a trigger for ectodomain shedding of ARTC2.2, we used the NAD+-analog etheno-NAD+, which carries an additional etheno-ring on the adenosine moiety (35). Etheno-NAD+ is an efficient substrate of ARTC2.2, and etheno–ADP-ribosylation of cell surface proteins can be monitored by flow cytometry with the etheno-adenosine–specific mAb 1G4 (35). Etheno–ADP-ribosylation of P2X7, however, does not gate the ion channel (9). Consistently, treatment of Yac-1 cells with etheno-NAD+ did not lead to a downmodulation of the staining of CD62L or ARTC2.2, but did result in staining with etheno-adenosine–specific mAb 1G4 (Fig. 1C). These results strengthen the conclusion that downmodulation of ARTC2.2 depends on ADP-ribosylation of P2X7. The results also imply that this downmodulation is not caused by obstruction of Ab binding due to ADP-ribosylation of the Ab-binding epitope (38, 39).
NAD+-induced downmodulation of ARTC2.2 is mediated by a P2X7-activated metalloprotease
Gating of P2X7 is known to induce ectodomain shedding by the metalloproteases a disintegrin and metalloprotease (ADAM)10 and ADAM17 (3, 5, 16, 40). To test the role of these metalloproteases, we treated Yac-1 cells with NAD+ in the presence of GI, a preferential inhibitor of ADAM10, or with GW, which inhibits ADAM10 and ADAM17 (36). The results show that treatment with GW, but not GI, inhibits downmodulation of ARTC2.2 (Fig. 1D), consistent with release of ARTC2.2 by ADAM17-catalyzed ectodomain shedding.
Recovery of ARTC2.2 from the supernatant of nucleotide-treated cells
To further corroborate that loss of ARTC2.2 from the cell surface is mediated by ectodomain shedding, we attempted to recover ARTC2.2 from cell supernatants in sufficient amounts for analysis by mass spectrometry. To this end, we cotransfected HEK cells with FLAG-tagged ARTC2.2 and P2X7 (41) and treated these cells for 15 min with NAD+ and ATP. The cell supernatant was passed over an anti-FLAG affinity column. Bound proteins were eluted, deglycosylated with endoglycosidase F, and analyzed by SDS-PAGE. The results show recovery of a single band corresponding to the predicted size of 29 kDa of the ectodomain of ARTC2.2 (Fig. 2A). The Coomassie-stained protein was cut from the gel and confirmed as ARTC2.2 by MS analyses (Fig. 2B).
Nucleotide-induced cleavage of ARTC2.2 occurs at F239/S240. (A) ARTC2.2 was purified from the supernatant of nucleotide-treated HEK cells. ARTC2.2 was captured on nanobody-coated beads and deglycosylated with endoglycosidase F (EndoF). Proteins were size fractionated by SDS-PAGE and visualized by Coomassie staining. (B) The band corresponding to ARTC2.2 was subjected to in-gel protease digestion in the presence of 18O-labeled water before MS analyses. Peptides identified by MS in trypsin or Glu-C digests are shown in red and blue, respectively. The N-terminal signal peptide and the C-terminal GPI anchor sequence (shaded in gray) are cleaved cotranslationally and are not contained in shed ARTC2.2. Cysteines are highlighted in yellow, and disulfide bonds connecting these residues are indicated by lowercase letters. The glycosylation sites at N59 and N229 are highlighted in green. The engineered FLAG tag is highlighted in magenta. Numbering of amino acid residues was adjusted to that of native ARTC2.2 (i.e., without signal peptide and FLAG tag). The metalloprotease cleavage site identified by MS analyses is marked by a box; amino acids F239 and S240 at the cleavage site are in bold. (C and D) ARTC2.2 was purified from the supernatant of nucleotide-treated HEK cells cotransfected with ARTC2.2 and P2X7 and deglycosylated with endoglycosidase F. A band corresponding to ARTC2.2 visualized by SDS-PAGE and Coomassie staining was cut from the gel. Proteolytic digestion of the band was performed with trypsin or GluC in 18O-labeled water to label newly generated C termini by 18O. The C terminus previously generated by NAD-induced ectodomain shedding of ARTC2.2 lacks this modification. Peptides were analyzed by MALDI MS and MS2. The MS and MS2 spectra of the C-terminal peptides of ARTC2.2 from the trypsin (C) and the GluC digest (D) are shown. Insets show the isotope patterns of the precursor ions. Scores denote the MASCOT peptide scores.
Nucleotide-induced cleavage of ARTC2.2 occurs at F239/S240. (A) ARTC2.2 was purified from the supernatant of nucleotide-treated HEK cells. ARTC2.2 was captured on nanobody-coated beads and deglycosylated with endoglycosidase F (EndoF). Proteins were size fractionated by SDS-PAGE and visualized by Coomassie staining. (B) The band corresponding to ARTC2.2 was subjected to in-gel protease digestion in the presence of 18O-labeled water before MS analyses. Peptides identified by MS in trypsin or Glu-C digests are shown in red and blue, respectively. The N-terminal signal peptide and the C-terminal GPI anchor sequence (shaded in gray) are cleaved cotranslationally and are not contained in shed ARTC2.2. Cysteines are highlighted in yellow, and disulfide bonds connecting these residues are indicated by lowercase letters. The glycosylation sites at N59 and N229 are highlighted in green. The engineered FLAG tag is highlighted in magenta. Numbering of amino acid residues was adjusted to that of native ARTC2.2 (i.e., without signal peptide and FLAG tag). The metalloprotease cleavage site identified by MS analyses is marked by a box; amino acids F239 and S240 at the cleavage site are in bold. (C and D) ARTC2.2 was purified from the supernatant of nucleotide-treated HEK cells cotransfected with ARTC2.2 and P2X7 and deglycosylated with endoglycosidase F. A band corresponding to ARTC2.2 visualized by SDS-PAGE and Coomassie staining was cut from the gel. Proteolytic digestion of the band was performed with trypsin or GluC in 18O-labeled water to label newly generated C termini by 18O. The C terminus previously generated by NAD-induced ectodomain shedding of ARTC2.2 lacks this modification. Peptides were analyzed by MALDI MS and MS2. The MS and MS2 spectra of the C-terminal peptides of ARTC2.2 from the trypsin (C) and the GluC digest (D) are shown. Insets show the isotope patterns of the precursor ions. Scores denote the MASCOT peptide scores.
ARTC2.2 is cleaved at F239-S240, 3 aa upstream of the GPI anchor
To identify the C terminus of shed ARTC2.2, representing the potential cleavage site of ADAM17, we performed liquid chromatography–MS and MS/MS experiments following in-gel digestion with trypsin or endoproteinase Glu-C in 95% labeled H218O. All newly formed C termini then incorporate one or two 18O isotopes, leading to a mass increase that allows discrimination between C termini formed during in-gel digest from the one generated during metalloprotease-mediated ectodomain shedding. In the tryptic digest of ARTC2.2, we identified the peptide S222NFNCFYNGSAQTVNIDF239 that is preceded by a lysine residue; and in the Glu-C-digest, the peptide R219KKSNFNCFYNGSAQTVNIDF239 that is preceded by a glutamate residue (Fig. 2C). The finding that these peptides lack 18O isotopes provides proof for F239-S240 as the cleavage site of ARTC2.2 during ectodomain shedding (Fig. 2D). Comparison of this site with other known proteolytic cleavage sites reveals that the cleavage site of ARTC2.2 is compatible with cleavage by ADAM17 (Table I). Note that residues N59 and N229 were identified in deamidated form (data not shown). Deamidation of Asn to Asp occurs during deglycosylation by endoglycosidase F, confirming N-linked glycosylation of N59 and N229.
P4 . | P3 . | P2 . | P1 . | — . | P1′ . | P2′ . | P3′ . | P4′ . | . |
---|---|---|---|---|---|---|---|---|---|
L | A | A | A | — | V | V | S | S | Consensus |
10 | 12 | 10 | 15 | — | 20 | 14 | 15 | 10 | |
N | I | D | F | — | S | I | S | * | ARTC2.2 |
4 | 2 | 2 | 2 | — | 8 | 4 | 15 | 0 | |
G | P | Q | R | — | F | S | G | A | RANKL |
2 | 4 | 9 | 7 | — | 3 | 5 | 6 | 4 |
P4 . | P3 . | P2 . | P1 . | — . | P1′ . | P2′ . | P3′ . | P4′ . | . |
---|---|---|---|---|---|---|---|---|---|
L | A | A | A | — | V | V | S | S | Consensus |
10 | 12 | 10 | 15 | — | 20 | 14 | 15 | 10 | |
N | I | D | F | — | S | I | S | * | ARTC2.2 |
4 | 2 | 2 | 2 | — | 8 | 4 | 15 | 0 | |
G | P | Q | R | — | F | S | G | A | RANKL |
2 | 4 | 9 | 7 | — | 3 | 5 | 6 | 4 |
Letters indicate amino acids found in 60 known target proteins of ADAM17 upstream (P4–P1) and downstream (P1′–P4′) of the cleavage site. Numbers indicate the number of target proteins (not counting ARTC2.2) containing the indicated amino acid. Long dashes (—) between P1 and P1′ indicate the cleavage site. The asterisk (*) indicates the GPI-anchor attachment site of ARTC2.2. Data were obtained from MEROPS, the database of proteolytic enzymes, their substrates and inhibitors (http://merops.sanger.ac.uk/) (54).
RANKL, TNF superfamily member receptor activator of NF-κB ligand.
NAD+ released during harvesting of cells induces shedding of ARTC2.2
We have shown that mechanical manipulations during trypsinization of HEK cells or during preparation of primary cells from lymph nodes or spleen cause release of NAD+ from cells in sufficient quantity to gate P2X7 via ADP-ribosylation (5, 9). Although GPI-anchored ARTC2.2 efficiently ADP-ribosylates P2X7 even at 4°C, gating of P2X7 requires temperatures above 21°C (5). Indeed, harvesting of HEK cells at 37°C caused shedding of ARTC2.2 even without addition of any exogenous NAD+ (Fig. 3A). Consistently, adding the ARTC2.2 blocking nanobody s+16a during cell preparation inhibited shedding of ARTC2.2 (Fig. 3B). These results imply that NAD+ released during harvesting of cells induces shedding of ARTC2.2.
NAD+ released during harvesting of cells induces shedding of enzymatically active ARTC2.2. (A and B) HEK cells were cotransfected with cDNA expression vectors for ARTC2.2 and P2X7. (A) Twenty-four hours posttransfection, cells were harvested by trypsinization at 4°C. Washed cells were incubated for 30 min at 4°C or at 37°C and then stained with mAbs directed against ARTC2.2 or P2X7. (B) Twenty-four hours posttransfection, cells were harvested by trypsinization at 4°C in the absence or presence of the ARTC2.2-blocking nanobody s+16a. Cells were incubated further for 30 min at 37°C, stained with mAbs directed against ARTC2.2 or P2X7, and analyzed by flow cytometry. Results are representative of three independent experiments. (C) T cells were prepared from lymph nodes of ARTC2−/− (Ca, Cd, Cg, Cj), WT (Cb, Ce, Ch, Ck), and ARTC2.2-transgenic (Cc, Cf, Ci, Cl) mice by depletion of B cells at 4°C. Purified T cells were incubated for 30 min either at 4°C (Ca–Cc) or at 37°C in the absence (Cd–Cf) or presence of exogenously added 25 μM NAD+ (Cg–Ci) or 250 μM ATP (Cj–Cl). Cells were stained with mAbs directed against CD62L and ARTC2.2 and analyzed by flow cytometry. (D) Purified T cells from ARTC2.2-transgenic mice were incubated as in (Ca)–(Cc) and (Cd)–(Cf) for 30 min at 4°C (lanes 3 and 4) or at 37°C (lanes 5 and 6). Cells were pelleted by centrifugation, and membrane proteins were solubilized in 1% Triton X-100. ARTC2.2 was immunoprecipitated from cell lysates (c) and cell supernatants (s) with immobilized ARTC2.2-specific mAb. Bead-bound ARTC2.2 was incubated for 20 min in the presence of [32P]NAD+ and the arginine analog agmatin. Reaction products were analyzed by thin-layer chromatography, and [32P]ADP-ribosylagmatine was detected by autoradiography. Solvent (co) and recombinant ARTC2.2 were used as negative and positive controls, respectively. Results are representative of two independent experiments. (E) Purified T cells from ARTC2.2-transgenic (Tg) and ARTC2−/− (−/−) mice were incubated for 30 min in the presence of ATP. Cells were pelleted by centrifugation, and cell supernatants were incubated for 20 min in the presence of serum proteins and [32P]NAD. Proteins were size fractionated by SDS-PAGE and visualized by Coomassie staining. Incorporated radioactivity was detected by autoradiography.
NAD+ released during harvesting of cells induces shedding of enzymatically active ARTC2.2. (A and B) HEK cells were cotransfected with cDNA expression vectors for ARTC2.2 and P2X7. (A) Twenty-four hours posttransfection, cells were harvested by trypsinization at 4°C. Washed cells were incubated for 30 min at 4°C or at 37°C and then stained with mAbs directed against ARTC2.2 or P2X7. (B) Twenty-four hours posttransfection, cells were harvested by trypsinization at 4°C in the absence or presence of the ARTC2.2-blocking nanobody s+16a. Cells were incubated further for 30 min at 37°C, stained with mAbs directed against ARTC2.2 or P2X7, and analyzed by flow cytometry. Results are representative of three independent experiments. (C) T cells were prepared from lymph nodes of ARTC2−/− (Ca, Cd, Cg, Cj), WT (Cb, Ce, Ch, Ck), and ARTC2.2-transgenic (Cc, Cf, Ci, Cl) mice by depletion of B cells at 4°C. Purified T cells were incubated for 30 min either at 4°C (Ca–Cc) or at 37°C in the absence (Cd–Cf) or presence of exogenously added 25 μM NAD+ (Cg–Ci) or 250 μM ATP (Cj–Cl). Cells were stained with mAbs directed against CD62L and ARTC2.2 and analyzed by flow cytometry. (D) Purified T cells from ARTC2.2-transgenic mice were incubated as in (Ca)–(Cc) and (Cd)–(Cf) for 30 min at 4°C (lanes 3 and 4) or at 37°C (lanes 5 and 6). Cells were pelleted by centrifugation, and membrane proteins were solubilized in 1% Triton X-100. ARTC2.2 was immunoprecipitated from cell lysates (c) and cell supernatants (s) with immobilized ARTC2.2-specific mAb. Bead-bound ARTC2.2 was incubated for 20 min in the presence of [32P]NAD+ and the arginine analog agmatin. Reaction products were analyzed by thin-layer chromatography, and [32P]ADP-ribosylagmatine was detected by autoradiography. Solvent (co) and recombinant ARTC2.2 were used as negative and positive controls, respectively. Results are representative of two independent experiments. (E) Purified T cells from ARTC2.2-transgenic (Tg) and ARTC2−/− (−/−) mice were incubated for 30 min in the presence of ATP. Cells were pelleted by centrifugation, and cell supernatants were incubated for 20 min in the presence of serum proteins and [32P]NAD. Proteins were size fractionated by SDS-PAGE and visualized by Coomassie staining. Incorporated radioactivity was detected by autoradiography.
To determine whether ADP-ribosylation of P2X7 on primary T cells could similarly induce ectodomain shedding of ARTC2.2, we performed comparative analyses of T cells obtained from WT, ARTC2−/−, or ARTC2.2-overexpressing transgenic mice (27, 28) (Fig. 3C). Indeed, a substantial fraction of ARTC2.2-transgenic T cells shed CD62L and ARTC2.2 when incubated for 30 min at 37°C, but not at 4°C (Fig. 3Cc, 3Cf). Treatment of ARTC2.2-transgenic T cells with NAD+ or ATP enhanced shedding of ARTC2.2 and CD62L (Fig. 3Ci, 3Cl). Shedding of ARTC2.2 occurs also on WT T cells treated with NAD+ or ATP (Fig. 3Ce, 3Ch, 3Ck), but, owing to the low expression levels of ARTC2.2, this phenomenon is less evident on WT T cells than on ARTC2.2-transgenic T cells. Consistently, NAD+ treatment of T cells from ARTC2−/− mice did not induce shedding of CD62L (Fig. 3Cg), but, as predicted, gating of P2X7 with ATP results in shedding in all three strains of mice (Fig. 3Cj).
ARTC2.2 is shed from T cells as an active enzyme
To determine whether ARTC2.2 is shed in enzymatically active format from T cells in response to ADP-ribosylation of P2X7, we used radioactive ADP-ribosylation assays (Fig. 3D) (21). T cells from ARTC2.2-transgenic were incubated at 4°C or 37°C, and ARTC2.2 was immunoprecipitated from cell supernatants and from solubilized membrane proteins. Immunoprecipitates were incubated with 32P-labeled NAD+ and the arginine analog agmatine, and ADP-ribosylation of agmatine was monitored by TLC and autoradiography (21). The results show strong enzyme activity in the membrane protein fraction, but little, if any, in the supernatants of T cells incubated at 4°C. Conversely, strong enzyme activity was found in supernatants of T cells incubated at 37°C, suggesting that ARTC2.2 is shed from T cells as an active enzyme in response to NAD+ released during cell preparation.
To further verify that the ADP-ribosylation activity shed by nucleotide-treated T cells corresponds to ARTC2.2, we performed comparative analyses of cell supernatants obtained from ARTC2−/− or ARTC2.2-transgenic T cells upon treatment with ATP for 20 min in the presence of serum proteins of ARTC2−/− mice. Cell supernatants were incubated for 20 min with 32P-labeled NAD+, and radiolabel incorporated into soluble proteins was detected by SDS-PAGE and autoradiography (Fig. 3E). The results show radiolabeling of several distinct protein bands in the serum containing supernatants of ARTC2.2-transgenic T cells, but not of ARTC2−/− T cells (Fig. 3E).
The major T cell subsets are sensitive to P2X7-dependent shedding of ARTC2.2
To determine whether T cell subsets differ in their sensitivity to P2X7-dependent ectodomain shedding of ARTC2.2, we performed comparative flow cytometric analyses of ATP-treated T cells from WT, ARTC2−/−, P2X7−/−, and ARTC2.2-transgenic mice (Fig. 4A, 4B). The results show that all major T cell subsets, including naive and regulatory Th cells (CD25−/CD4+ and CD25+/CD4+) as well as naive and memory cytotoxic T cells (CD44high/CD8+ and CD44low/CD8+), express ARTC2.2 and are sensitive to ATP-induced shedding of ARTC2.2. Among WT T cells, shedding of ARTC2.2 is most apparent on the CD8+/CD44high subpopulation. Importantly, T cells from P2X7−/− mice do not shed ARTC2.2 in response to treatment with ATP, confirming that nucleotide-induced shedding of ARTC2.2 is entirely dependent on P2X7. As in case of Yac-1 lymphoma cells, GW inhibited shedding of ARTC2.2 more potently than the ADAM10-specific inhibitor GI. The broad metalloprotease inhibitor Marimastat also effectively blocked shedding of ARTC2.2 (Fig. 4C).
The major T cell subsets are sensitive to P2X7-dependent shedding of ARTC2.2. (A and B) Primary splenic T cells of WT, ARTC2.2-transgenic, ARTC2−/−, and P2X7−/− mice were incubated for 20 min in the absence or presence of ATP. Cells were stained with mAbs directed against CD4, CD8, CD25, CD44, and ARTC2.2 and analyzed by flow cytometry. (C) Primary splenic T cells of CD8+ WT T cell mice were incubated for 20 min in the absence or presence of ATP and the metalloprotease inhibitors GW, GI, or Marimastat (each at 5 μM). Cells were stained with mAbs directed against CD8 and ARTC2.2 and analyzed by flow cytometry.
The major T cell subsets are sensitive to P2X7-dependent shedding of ARTC2.2. (A and B) Primary splenic T cells of WT, ARTC2.2-transgenic, ARTC2−/−, and P2X7−/− mice were incubated for 20 min in the absence or presence of ATP. Cells were stained with mAbs directed against CD4, CD8, CD25, CD44, and ARTC2.2 and analyzed by flow cytometry. (C) Primary splenic T cells of CD8+ WT T cell mice were incubated for 20 min in the absence or presence of ATP and the metalloprotease inhibitors GW, GI, or Marimastat (each at 5 μM). Cells were stained with mAbs directed against CD8 and ARTC2.2 and analyzed by flow cytometry.
Shedding shifts the substrate specificity of ARTC2.2 from membrane proteins to secretory proteins
To determine whether shed ARTC2.2 retains the capacity to ADP-ribosylate membrane proteins, we performed comparative radio–ADP-ribosylation analyses of WT, ARTC2−/−, and ARTC2.2-transgenic T cells following treatment of cells for 20 min with ATP in the presence of serum proteins (Fig. 5). The results show that ATP-induced shedding of ARTC2.2 results in drastically reduced radiolabeling of cell surface proteins (Fig. 5A), including each of the known target proteins CD8, P2X7, and LFA-1 (Fig. 5B). Conversely, ATP treatment resulted in markedly enhanced radiolabeling of secretory proteins in the cell supernatant of ATP-treated ARTC2.2-transgenic T cells (Fig. 5C, lane 4 versus lane 3).
Shedding shifts the substrate specificity of ARTC2.2 from membrane proteins to secretory proteins. (A) T cells of WT mice were incubated for 20 min without ATP (−) (lanes 1 and 4) or in the presence of ATP (+) (lanes 2 and 5) to induce ectodomain shedding. [32P]NAD was then added, and cells were further incubated for 20 min. T cells from ART2−/− mice were used as controls (lanes 3 and 6). Triton X-100–solubilized proteins were size fractionated by SDS-PAGE and were visualized by Coomassie staining. Radiolabeled proteins were detected by autoradiography. (B) The ARTC2.2 target proteins CD8 (lanes 1 and 2), P2X7 (lanes 3 and 4), and LFA-1 (lanes 5 and 6) were precipitated from lysates of untreated (−) and ATP-treated (+) T cells. Incorporated radioactivity was detected by SDS-PAGE autoradiography. (C) T cells of ARTC2.2-transgenic mice were incubated for 20 min at 4°C without ATP (−) (lanes 1 and 3) or in the presence of ATP at 37°C (+) (lanes 2 and 4) to induce ectodomain shedding of ARTC2.2. [32P]NAD was added, and cells were further incubated for 20 min. Cells and cell supernatants were separated by centrifugation. Triton X-100–solubilized membrane proteins (lanes 1 and 2) and proteins in the cell supernatants (lanes 3 and 4) were analyzed by SDS-PAGE autoradiography.
Shedding shifts the substrate specificity of ARTC2.2 from membrane proteins to secretory proteins. (A) T cells of WT mice were incubated for 20 min without ATP (−) (lanes 1 and 4) or in the presence of ATP (+) (lanes 2 and 5) to induce ectodomain shedding. [32P]NAD was then added, and cells were further incubated for 20 min. T cells from ART2−/− mice were used as controls (lanes 3 and 6). Triton X-100–solubilized proteins were size fractionated by SDS-PAGE and were visualized by Coomassie staining. Radiolabeled proteins were detected by autoradiography. (B) The ARTC2.2 target proteins CD8 (lanes 1 and 2), P2X7 (lanes 3 and 4), and LFA-1 (lanes 5 and 6) were precipitated from lysates of untreated (−) and ATP-treated (+) T cells. Incorporated radioactivity was detected by SDS-PAGE autoradiography. (C) T cells of ARTC2.2-transgenic mice were incubated for 20 min at 4°C without ATP (−) (lanes 1 and 3) or in the presence of ATP at 37°C (+) (lanes 2 and 4) to induce ectodomain shedding of ARTC2.2. [32P]NAD was added, and cells were further incubated for 20 min. Cells and cell supernatants were separated by centrifugation. Triton X-100–solubilized membrane proteins (lanes 1 and 2) and proteins in the cell supernatants (lanes 3 and 4) were analyzed by SDS-PAGE autoradiography.
It is difficult to ascertain whether the residual radiolabeling of cell surface proteins in ATP-treated WT T cells (Fig. 5A, lane 5) and ARTC2.2-transgenic T cells (Fig. 5C, lane 2) is due to radiolabeling in cis by residual cell surface ARTC2.2 or to trans-radiolabeling of cell surface proteins by shed ARTC2.2. To determine whether shed ARTC2.2 can ADP-ribosylate cell surface proteins, we used T cells from ARTC2−/− mice, which express the known target proteins of ARTC2.2, but lack cell surface ART activity (27). In accordance with previous findings, T cells from WT mice displayed prominent (etheno)-ADP-ribosylation of cell membrane proteins (35) (Fig. 6A, right panel). In contrast, T cells from ARTC2−/− mice did not incorporate any etheno-ADP-ribose into cell surface proteins (Fig. 6A, left panel), not even in the presence of large amounts of shed ARTC2.2 (0.5 μg per 106 cells) (Fig. 6A, middle panel). Moreover, a large fraction of WT T cells externalized phosphatidylserine and became permeable to propidium iodide when treated with NAD+ (Fig. 6B, right panel). In contrast, ARTC2−/− T cells were completely resistant to NAD+-induced cell death (Fig. 6B, left panel), even in the presence of large amounts of shed ARTC2.2 (Fig. 6B, middle panel). Taken together, these results indicate that shed ARTC2.2 loses the capacity to ADP-ribosylate membrane proteins and instead acquires the capacity to ADP-ribosylate secretory proteins.
Shed ARTC2.2 does not ADP-ribosylate cell surface proteins and cannot induce NAD-induced cell death of ARTC2−/− cells. T cells from ARTC2−/− mice were incubated for 20 min in the absence (−/−) or presence of shed ARTC2.2 (−/− +sARTC2) with either etheno-NAD+ (A) or NAD+ (B). T cells from WT mice without exogenously added ARTC2.2 (WT) were used as controls. (A) Etheno–ADP-ribosylated cell surface proteins were detected by flow cytometry using etheno-adenosine–specific mAb 1G4. (B) NAD+-induced cell death was monitored by flow cytometry using annexin V to detect externalized phosphatidylserine and propidium iodide to detect permeabilization of the cell membrane. Results are representative of two independent experiments.
Shed ARTC2.2 does not ADP-ribosylate cell surface proteins and cannot induce NAD-induced cell death of ARTC2−/− cells. T cells from ARTC2−/− mice were incubated for 20 min in the absence (−/−) or presence of shed ARTC2.2 (−/− +sARTC2) with either etheno-NAD+ (A) or NAD+ (B). T cells from WT mice without exogenously added ARTC2.2 (WT) were used as controls. (A) Etheno–ADP-ribosylated cell surface proteins were detected by flow cytometry using etheno-adenosine–specific mAb 1G4. (B) NAD+-induced cell death was monitored by flow cytometry using annexin V to detect externalized phosphatidylserine and propidium iodide to detect permeabilization of the cell membrane. Results are representative of two independent experiments.
ARTC2.2 is shed in vivo in response to NAD+ exposure
To determine whether ARTC2.2 is shed also in vivo in response to activators of the ARTC2/P2X7 signaling pathway like inflammation or NAD+ exposure, we used systemic injection of NAD+ as a convenient tool to mimic systemic inflammation (29, 42). To quantify ARTC2.2 in biological fluids, we established a sensitive sandwich ELISA that permits the detection of shed ARTC2.2 with a detection limit of 1 ng/ml (Fig. 7A). Under steady state conditions, the serum level of soluble ARTC2.2 in WT mice was <1 ng/ml (Fig. 7B), but reached 60 ng/ml in the serum of ARTC2.2-transgenic mice. Systemic injection of NAD+ led to a >10-fold increase in the level of serum ARTC2.2 (up to 50 ng/ml in WT mice, and up to 750 ng/ml in ARTC2.2 transgenic mice) (Fig. 7B). Radio–ADP-ribosylation assays revealed that NAD+-induced shedding of ARTC2.2 resulted in prominent ADP-ribosylation of serum proteins in WT and ARTC2.2-transgenic mice, but not in ARTC−/− mice (Fig. 7C). Fig. 8 illustrates the proteolytic cleavage site and other posttranslational modifications of ARTC2.2.
Intravenous injection of NAD+ induces shedding of ARTC2.2 in vivo. (A) ELISA plates coated with ARTC2.2-specific mAb Nika109 were incubated with serial dilution of purified shed ARTC2.2. Bound ARTC2.2 was detected with biotinylated Nika102 and peroxidase-conjugated streptavidin. (B and C) WT, ARTC2−/−, and ARTC2.2-transgenic mice received i.v. injections of NAD+ or saline 20 min before sacrifice. (B) Soluble ARTC2.2 in serum was quantified by ELISA. Data shown are means ± SD. **p < 0.01, ***p < 0.001 (Student t test). (C) Serum samples were incubated for 15 min with [32P]NAD+, and radiolabeled proteins were visualized by SDS-PAGE autoradiography.
Intravenous injection of NAD+ induces shedding of ARTC2.2 in vivo. (A) ELISA plates coated with ARTC2.2-specific mAb Nika109 were incubated with serial dilution of purified shed ARTC2.2. Bound ARTC2.2 was detected with biotinylated Nika102 and peroxidase-conjugated streptavidin. (B and C) WT, ARTC2−/−, and ARTC2.2-transgenic mice received i.v. injections of NAD+ or saline 20 min before sacrifice. (B) Soluble ARTC2.2 in serum was quantified by ELISA. Data shown are means ± SD. **p < 0.01, ***p < 0.001 (Student t test). (C) Serum samples were incubated for 15 min with [32P]NAD+, and radiolabeled proteins were visualized by SDS-PAGE autoradiography.
Schematic diagram of the posttranslational modifications of ARTC2.2. ARTC2.2 (modeled on rat ARTC2a, pdb code 1og3) is represented as a surface diagram (left) or as a ribbon diagram (right), and bound NAD+ as a stick model in cyan. F239—identified in this study as the C-terminal amino acid generated during shedding of ARTC2.2—and flanking residues are shown in single-letter code. Glycosylation sites confirmed in this study at N59 and N229 are indicated by green forks. S242 is the predicted GPI anchor attachment.
Schematic diagram of the posttranslational modifications of ARTC2.2. ARTC2.2 (modeled on rat ARTC2a, pdb code 1og3) is represented as a surface diagram (left) or as a ribbon diagram (right), and bound NAD+ as a stick model in cyan. F239—identified in this study as the C-terminal amino acid generated during shedding of ARTC2.2—and flanking residues are shown in single-letter code. Glycosylation sites confirmed in this study at N59 and N229 are indicated by green forks. S242 is the predicted GPI anchor attachment.
Discussion
We found that activation of the P2X7 ion channel on T cells by ADP-ribosylation or by ATP induces ectodomain shedding of ARTC2.2, redirecting its target specificity from membrane proteins to secretory proteins. Remarkably, exposure of T cells to extracellular NAD+ induces an ARTC2.2-dependent cascade of reactions that results in the shedding of its own ectodomain from the cell surface by membrane-proximal proteolytic cleavage.
Shedding of ARTC2.2 is inhibited by GW, but not by GI, consistent with the notion that ADAM17 is responsible for cleavage of ARTC2.2 (Figs. 1D, 4C) (40). The results of our MS/MS analyses show that cleavage of ARTC2.2 occurs close to the cell membrane, that is, at F239-S240, 3 aa upstream of the predicted GPI anchor attachment site (Fig. 2). Membrane-proximal cleavage is consistent with cleavage by ADAM17, which tends to cleave its targets close to their transmembrane region (43–45). Furthermore, comparison of the ARTC2.2 amino acid residues flanking F239-S240 with those of other known ADAM17 target proteins, for example, receptor activator for NF-κB ligand, reveals that the ARTC2.2 flanking sequence is compatible with cleavage by ADAM17 (Table I). These results, however, do not rule out the possibility that another metalloprotease may contribute to nucleotide-induced shedding of ARTC2.2.
ARTC2.2 is a promiscuous ADP-ribosyltransferase that can ADP-ribosylate several structurally unrelated cell surface proteins, including P2X7, LFA-1, and CD8 (9, 38, 39). By virtue of its GPI anchor attachment, ARTC2.2 constitutively localizes in lipid rafts (30). Localization in lipid rafts restricts the substrate specificity of ARTC2.2 to proteins constitutively or transiently associated with lipid rafts (30). It has been proposed that ADAM17 also transiently associates with lipid rafts, and that ADAM17 preferentially cleaves raft-associated target proteins (46). Gating of P2X7 is known to dramatically perturb the plasma membrane, causing rapid externalization of phosphatidylserine and membrane blebbing (15, 47, 48). It is conceivable that raft association and P2X7-induced membrane perturbations both contribute to the rapid ectodomain shedding of ARTC2.2.
The results presented in this work show that shedding of ARTC2.2 dramatically alters its substrate profile: comparative ADP-ribosylation assays with membrane-bound and shed ARTC2.2 show that GPI-anchored ARTC2.2 predominantly ADP-ribosylates membrane proteins, whereas shed ARTC2.2 ADP-ribosylates distinct secretory proteins, but not membrane proteins (Fig. 5). Interestingly, shed ARTC2.2 could not ADP-ribosylate cell surface proteins on ARTC2−/− T cells and did not mediate NAD+-induced cell death of these cells (Fig. 6). These findings imply that shifting the substrate specificity of ARTC2.2 by metalloprotease-induced shedding protects the cells from a biologic outcome of inflammation (NAD+-induced cell death).
On the basis of these findings, we propose that proteolytic release of ARTC2.2 from the cell membrane provides a mechanism to rapidly refocus its target profile, that is, from membrane proteins to secretory proteins. Considering that NAD+ is released during inflammation (11) and can trigger shedding of ARTC2 from T cells in vitro (Fig. 3) and in vivo (Fig. 7), ADP-ribosylation of cytokines or other secretory proteins by shed ARTC2.2 may provide a new mechanism to modulate inflammatory reactions.
A model of the ARTC2.2 structure—based on the crystal structure of rat ARTC2 in complex with NAD+ (49)—and a schematic diagram of its posttranslational modifications are shown in Fig. 8. Note that the proteolytic cleavage site occurs 13 aa downstream of C226, which in turn is engaged in an intrachain disulfide bond conserved in the ARTC family (49, 50). The amino acids between C226 and the GPI anchor presumably form a flexible stalklike structure. The results of our MS analyses confirm that N229, three residues downstream of C226 and 10 aa upstream of the cleavage site, carries an N-linked oligosaccharide side chain (Fig. 8). Interestingly, the stalk region between the conserved cysteine residue and the GPI anchor is extended in other members of the ARTC family, that is, ARTC1 and ARTC3, by short stretches of amino acids encoded in small exons (50). Thus, it is conceivable that regulated proteolysis similarly controls the target specificity of other members of the ARTC family. Regulated proteolysis may also account for the previously reported release of soluble ADP-ribosyltransferase activity from activated granulocytes and heterophils (22, 51) and the ADP-ribosylation of secretory proteins such as platelet-derived growth factor-B, human neutrophil peptide 1, and tuftsin (51–53).
In summary, our results uncover a hitherto unrecognized mechanism to redirect the substrate specificity of ARTC2.2 from raft-associated membrane proteins to secretory proteins. Membrane-proximal proteolytic cleavage evidently provides a mechanism to rapidly convert ARTC2.2 from a cell surface enzyme into a secretory signaling protein.
Acknowledgements
We thank Fabienne Seyfried, Joanna Schmid, and Gudrun Dubberke for excellent technical assistance. We thank Athena Chalaris (Kiel, Germany) for providing GW, GI, and Marimastat, and Regina Santella (New York, NY) for providing the 1G4 Ab. We thank Drs. B. Fleischer, E. Tolosa, and H. W. Mittrücker for critical reading of the manuscript.
Footnotes
This work was supported by Deutsche Forschungsgemeinschaft Grants 310/6 (to F.H. and F.K.-N.), SFB877-A5 (to F.K.-N.), and SFB877-Z2 (to A.T.); by a stipend from the Integrated Research Training Group of Sonderforschungsbereich 877 (to M.M.); and by a Deutsche Forschungsgemeinschaft Cluster of Excellence “Inflammation at Interfaces” grant (to A.T.).
References
Disclosures
F.K.-N. and F.H. receive royalties from sales of Abs developed in their laboratories via MediGate GmbH, a 100% subsidiary of the University Medical Center (Hamburg, Germany).