NAD⁺ Released during Inflammation Participates in T Cell Homeostasis by Inducing ART2-Mediated Death of Naive T Cells In Vivo¹

Mammalian mono-ADP-ribosyltransferases (ART)³ constitute a family of ectoenzymes structurally related to bacterial toxins catalyzing the transfer of the ADP-ribose group from NAD⁺ onto amino acid residues of target proteins (1, 2, 3). ART1–ART4 paralogs are GPI-anchored in the outer leaflet of the plasma membrane and are thus responsible for unique posttranslational modifications of proteins in the extracellular compartment. Like phosphorylation mediated by protein kinases, ART-mediated modification of proteins by ADP-ribosylation leads to structural changes that can either inhibit or stimulate target protein functions. ART activities in vitro can be split into those relevant to trans-ADP-ribosylation of free soluble targets such as cytokines and growth factors or proteins produced by neighboring cells and those relevant to cis-ADP-ribosylation of proteins present on the surface of the ART-expressing cells (3). Although ADP-ribosylation of proteins can easily be shown in vitro (2, 4, 5, 6), its reality and role in vivo are still poorly documented (7, 8).

In the mouse, two of the six ART paralogs, namely, ART2.1 and ART2.2, are expressed on the surface of most mature peripheral T lymphocytes (1, 9). It was first shown that a GPI-anchored ART is expressed on the surface of mouse cytotoxic T cell lines and that incubation of cells with NAD⁺ leads to the inhibition of both T cell proliferation and cytotoxic activity (6). This inhibition was associated with ADP-ribosylation of several membrane proteins such as LFA-1, CD8, CD27, CD43, CD44, or CD45 (10, 11, 12). Recently, the association of ART2 with lipid rafts has been shown to focus ART2 on specific targets (13). It was postulated that ADP-ribosylation of coreceptors inhibits TCR-signaling by altering receptor association as well as cell contacts and T cell trafficking (10, 12). We subsequently showed that incubation with NAD⁺ leads to the rapid induction of ART-dependent T cell death in vitro detectable at NAD concentrations as low as 1 μM, a phenomenon that we proposed to call NAD⁺-induced T cell death (NICD) for NAD⁺-induced cell death (4, 14). We also discovered that NICD results from the activation of the P2X7 ATP receptor, which is another target of murine ART2 (14).

P2X7 belongs to the P2X family of ATP-gated ion channels expressed on different cell types in the immune system including T lymphocytes (15, 16). P2X7 activation triggers calcium flux, shedding of CD62L, phosphatidylserine (PS) exposure, opening of a large nonselective membrane pore and ultimately cell death by apoptosis (17, 18, 19). Incubation of mouse T lymphocytes with micromolar NAD⁺ concentrations induces all these effects, whereas P2X7 activation by ATP requires millimolar concentrations (14, 17). NICD depends on ART2 and P2X7 as assessed by the resistance observed with T cells from ART2-deficient or P2X7-deficient mice (14, 20, 21). Moreover, C57BL/6 T cells, which harbor a P451L natural mutation in the C-terminal intracytoplasmic domain of P2X7 that severely impairs the response to ATP (22), are relatively resistant also to NICD in vitro, further supporting the conclusion that P2X7 is required for NAD⁺-induced T cell apoptosis.

The origin of the ecto-NAD⁺ ART-substrate is still poorly understood. The intracellular NAD⁺ concentration is in the range of 1 mM, whereas the plasma concentration is in the range of 0.1 μM, below the Km of ARTs (5, 23, 24). It is speculated that high amounts of ecto-NAD⁺ can be released during tissue injury as a consequence of cell lysis. However, nucleotides such as ATP or NAD⁺ also seem to be released by nonlytic processes under various physiological conditions including hypoxia, inflammation, and mechanical or chemical activation (25, 26, 27). Connexin 43 hemichannels, for instance, mediate transmembrane NAD⁺ fluxes in intact cells (25). Hence, high local NAD⁺ concentrations may be reached under various physiological and pathophysiological situations which would permit ADP-ribosylation of membrane proteins on ART-expressing neighboring cells. The fate of this NAD⁺ is controlled by ecto-NADases. As recently shown, the ecto-NADase CD38 expressed on B lymphocytes, competes with ARTs for the NAD⁺ substrate, controls its availability and therefore the level of ART2-catalyzed ADP-ribosylation on the surface of mouse T lymphocytes (28). This regulation is likely to occur in vivo in secondary lymphoid tissues where T and B lymphocytes are in close contact.

The aim of the present study was to evaluate NAD⁺ release during inflammation and to explore its consequence on T cell populations in vivo. We used a model of inflammation induced by polyacrylamide beads to investigate NAD⁺ release in exudates during the course of an acute inflammatory response. We then explored the effect of NAD⁺ on the different T cell subsets in lymph nodes drained by the inflammatory site or after systemic injection of NAD⁺. Our results show that high amounts of NAD⁺ are released during acute inflammation. They also demonstrate that NAD⁺ can induce strong T cell depletion in vivo and that cells surviving NAD⁺ treatment are enriched in activated/memory T lymphocytes. We finally show that NAD⁺ treatment in vivo affects Ab production to T dependent Ags providing evidence that NICD participates in T cell homeostasis in the course of the immune response.

BALB/c mice were obtained from the Centre d’Elevage Janvier (Le Genest Saint Isle, France). ART2-deficient mice were generated by standard homologous recombination procedures and backcrossed for 12 generations onto the BALB/c background as described elsewhere (20). CD38-deficient mice were gifts from Dr. F. Lund (Trudeau Institute, New York, NY) (29). All mice were maintained under standard conditions in the animal quarter of the Institute Jacques Monod (University Paris 7, Paris, France). Experiments were performed according to European regulations.

ATP, NAD⁺, propidium iodide (PI), Con A, and Neurospora crassa NAD⁺ glycohydrolase (NADase) were obtained from Sigma-Aldrich. PE- and FITC-conjugated anti-B220, anti-CD3, anti-CD4, anti-CD8, anti-CD25, anti-CD44, anti-CD69, anti-CD62L, anti-CD11b, anti-Gr1, and Annexin V^FITC were from BD Pharmingen. ART2.2-specific Nika 102 mAb (9) was purified from hybridoma supernatant by affinity chromatography on protein G-Sepharose (Pharmacia) and conjugated to FITC (Sigma-Aldrich). P2X7-specific Hano44 mAb was obtained by genetic immunization of rats and fusion of spleen cells with Sp2/0 myeloma cells as described previously (9, 30). Abs were purified from hybridoma supernatants and conjugated to Alexa Fluor 488 (Molecular Probes) according to the manufacturer’s instructions. SRBC were purchased from Eurobio.

Single-cell suspensions from lymph nodes were prepared and processed by flow cytometry on a FACSCalibur instrument using the CellQuest-Pro program (BD Biosciences) as described elsewhere (4, 9). B cells were depleted by magnetic cell separation using anti-B220 Abs (BD Pharmingen) conjugated to Adembeads prepared according to the manufacturer’s instructions (Ademtech). Purity of T cells was always >95% as verified by FACS analysis using PE-conjugated anti-B220 and FITC-conjugated anti-CD3. Following treatment with NAD⁺ or ATP for 30 min at 37°C, cells were washed in RPMI 1640 medium (Invitrogen Life Technologies) and were resuspended in annexin V binding buffer (BD Pharmingen). They were stained for 20 min on ice with FITC-conjugated annexin V (1 μg/ml) and propidium iodide (10 μg/ml).

Inflammation was induced as previously described (31). Briefly, animals were shaved and 800 μl of a sterile 67% suspension (53 mg dry weight/ml) of Biogel P100 (Bio-Rad) in PBS was injected s.c. into the bottom of the back. At given times, pouches were incised on sacrificed animals. The Biogel mass was collected under a dissecting microscope with a spatula and resuspended in 0.5 ml of PBS/20 U/ml heparin. Tubes were weighed before and after addition of the beads to determine the Biogel mass recovered. After vigorous shaking to extract cells and exudate components from the gel, beads were allowed to sediment and supernatants were collected. Cells were then recovered by centrifugation and cell-free exudates were immediately used for NAD⁺ measurement or frozen and kept at −20°C. Total cells in exudates were counted on a hemocytometer. Cell subpopulations were differentially evaluated by flow cytometry after staining with PE-conjugated anti-Gr-1 and FITC-conjugated anti-CD11b.

NAD⁺ concentrations were determined by a cyclic enzymatic assay as described previously (23). The first step of the cycle is an enzymatic reaction catalyzed by the NAD⁺-dependent alcohol dehydrogenase (ADH). The next steps are nonenzymatic reactions. NADH₂ produced during the enzymatic step is oxidized by phenazyne ethosulfate (PES) which, once reduced, is itself oxidized by thiazolyl blue (MTT). Oxidized NAD⁺ and PES are used in the next cycle. Absorbance of the reduced form of MTT is measured at 570 nm in a two-chamber spectrophotometer. A “premix” solution containing 3.2 ml of sodium-1.5 M (pH 7.8) bicine (N,N-bis(2-hydroxyethyl)glycine) buffer, 2.4 ml of 10 M ethanol, 2.0 ml of 10 mM MTT, 1.0 ml of 0.2 M EDTA, 0.4 ml of 100 mg/ml BSA, and 1.0 ml of H₂O was prepared. Measurements were performed by mixing in the dark 0.3 ml of premix solution with 0.1 ml of 40 mM PES, 0.1 ml of 1 mg/ml ADH in sodium-bicine buffer (0.1 M; pH 8.8), and 0.1 ml of a standard NAD⁺ solution or sample. OD was followed over a period of 3–5 min. NAD⁺ concentration was deduced from standard curves. Determination of the initial speed of the cyclic reaction, for each known concentration of NAD⁺ allowed us draw a standard curve: V = f[C], which then permits accurate calculation of NAD⁺ concentration in samples. We routinely used initiation by NAD⁺ for inflammatory exudates (colorless in our model) but initiation by ADH for plasma. All reagents were purchased from Sigma-Aldrich.

Mice were immunized i.p. with 0.2 ml of a 10% sheep erythrocyte suspension (Eurobio) in PBS (32). Serum samples were collected at appropriate times and hemagglutination titers were determined by serial dilutions of 50 μl of serum in PBS with 25 μl of a 0.4% SRBC suspension.

Results are expressed as the mean ± SD. Significance was assessed by Student’s t test with p at least inferior to 0.05 indicating statistical significance.

To evaluate whether NAD⁺ is released during inflammation and to determine its concentration in inflammatory exudates, we took advantage of the experimental model of acute inflammation induced by s.c. injection of Biogel polyacrylamide beads (31, 33). In this model, surgical recovery of the beads and their suspension in a defined volume of PBS gives access to the number of cells and to the real concentration of NAD⁺ trapped per unit of gel mass recovered. This method, which is not traumatic for cells, allows the determination of free NAD⁺ concentration in inflammatory exudates. Wild-type BALB/c mice were injected with Biogel and sacrificed at various time points to follow the kinetics of NAD⁺ concentration and cellular infiltration during the course of the inflammatory reaction (Fig. 1,B). Cells recovered in the gel were counted and NAD⁺ concentration was determined by an enzymatic assay. In this model, maximal cell infiltration is achieved after 24 h and up to 48 h and primarily correspond to neutrophils as assessed by Gr-1/CD11b staining (Fig. 1,A), which are then progressively replaced by macrophages (31). Remarkably, free NAD⁺ was detected in exudates. The concentration peaked at 12 h after induction of inflammation, a time when the cellular response was still building up (Fig. 1 B). This suggests that NAD⁺ is actively released from cells as an early response to inflammatory stimuli as opposed to being passively set free as a result of cell lysis or neutrophil apoptosis.

The influence of CD38 and ART2 in this process was explored by performing similar experiments in CD38 and ART2-deficient BALB/c mice. Twenty-four hours after treatment with Biogel, the number of neutrophils was lower in exudates from CD38^−/− mice compared with wild type (Fig. 2,A). This is consistent with the established role of CD38 in neutrophil chemotaxis and homing in response to inflammatory stimuli (34). At peak, NAD⁺ concentration reached >10 μM in exudates from CD38^−/− BALB/c mice but <4 μM in their wild-type or ART2^−/− counterparts (Fig. 2,B). This indicated that the CD38 NADase participates in the clearance of NAD⁺ at the inflammatory site. Yet, NAD⁺ release presented the same kinetics as in wild-type animals with a peak at 12 h preceding massive neutrophil influx in the pouch (Fig. 2,C). NAD⁺ release in exudates had no effect on the basal plasma level, which remained unchanged throughout the inflammatory reaction (Fig. 2 B and data not shown). Higher NAD⁺ concentrations were found in plasma of untreated (data not shown) and Biogel-treated CD38^−/− mice compared with wild-type animals. This may either reflect elevated steady-state NAD⁺ levels in these animals or have resulted from NAD⁺ released by hemolysis occurring during blood collection and preparation.

In vitro, micromolar concentrations of NAD⁺ are sufficient to induce rapid apoptosis of mouse T cells which can be detected by annexin V/PI staining (4, 14). We, therefore, next tested whether NAD⁺ present in inflammatory exudates could induce cell death in vitro. As shown in Fig. 3 A, exudates collected after 12 h from CD38^−/− BALB/c mice induced significant apoptosis of purified lymph node T cells. The effect was limited due to the dilution of exudates that was necessary to extract the NAD⁺ from the Biogel, leading to a working concentration in the range of 1.5 μM. The notion that NAD⁺ in inflammatory exudates caused PS flashing is supported by the finding that PS flashing was inhibited by pretreatment of exudates with exogenous NADase. Moreover, inflammatory exudates did not induce any detectable annexin V/PI staining of ART2-deficient T cells (data not shown).

These results prompted us to evaluate the effect of Biogel-induced inflammation on T cells in lymph nodes from wild-type, CD38^−/−, and ART2^−/− BALB/c mice. One day after injection of Biogel beads s.c., in the bottom of the back, draining (inguinal and para-aortic) and nondraining (axillary) nodes were separately collected. Cells were counted and analyzed by flow cytometry. A significant diminution of >20% in the number of CD3⁺ T cells was observed in the draining nodes of wild-type-treated mice but not in the nondraining axillary nodes of the same animals (Fig. 3,B). A more severe reduction of >50% was observed in the draining nodes from CD38^−/− mice but no effect of inflammation was observed in ART2^−/− mice treated under the same conditions (Fig. 3 B). Together, these results demonstrated that NAD⁺ released during inflammation can reach the draining nodes and can induce T cell depletion in vivo. Consistent with previous reports (28), CD38 evidently partially protects T cells from the ART2-dependent deleterious effect of NAD⁺.

In CD38^−/− mice, T cell depletion in the draining nodes was accompanied by a relative increase in the fraction of CD62L^low T cells (Fig. 4). This could reflect the resistance of CD62L^low T cells to NAD⁺-induced death and/or shedding of CD62L by surviving cells as a consequence of P2X7 activation (14). When 10 mg of NAD⁺ was mixed with Biogel, a more pronounced T cell depletion and CD62L shedding was observed in draining nodes but also in nondraining ones (Fig. 4). This confirmed that NAD⁺ released at local inflamed sites can reach draining lymph nodes (see Fig. 4, Biogel, middle panels) and even nondraining lymph nodes through lymphatic and blood circulations provided that the NAD⁺ amount is high enough (see Fig. 4, Biogel + NAD, right panels).

To further explore the effect of NAD⁺ on T cell populations in vivo, BALB/c mice were injected i.v. with the maximal tolerated dose of either ATP (6 mg) or NAD⁺ (10 mg) in 0.2 ml of PBS. After 24 h, spleen and lymph nodes were recovered and the size of the T cell population was evaluated by CD3/B220 staining in flow cytometry. A significant reduction in the percentage of CD3⁺ cells was observed in lymphoid organs from NAD⁺-treated mice, whereas ATP treatment did not have any significant effects (Fig. 5,A). The differential effects of ATP and NAD⁺ were consistent with the higher concentration of ATP compared with NAD⁺ required to induce T cell death by P2X7 activation in vitro (14) and with the short half-life of ATP in body fluids due to widely distributed ecto-ATPases (35, 36). Reduction of the percentage of CD3⁺ cells was accompanied by a 40% drop in the absolute T cell count in secondary lymphoid tissues (Fig. 5,B). As expected, T cell depletion was more dramatic in CD38^−/− mice compared with wild-type BALB/c mice, particularly in lymph nodes (Fig. 5,C), whereas NAD⁺ treatment had no effect on T cells in their ART2^−/− counterparts (Fig. 5,D). Evidently, NAD⁺-induced T cell depletion in vivo is ART2 dependent, in accord with previous observations in vitro (14, 20). Results in CD38^−/− mice are in agreement with our previous report demonstrating the antagonistic effect of the CD38 NADase on ART2-mediated T cell death (28). In CD38^−/− BALB/c mice, T cell depletion was clearly detectable as early as 3 h after NAD⁺ injection and reached its maximum at >70% after 1 day (Fig. 5 E). NAD⁺ treatment had little if any effects on B cell counts in spleen and lymph nodes. The effect on T lymphocytes was long-lasting and mice had not fully recovered normal T cell numbers, 1 wk after NAD⁺ treatment. These observations strongly suggested that NAD⁺ induces rapid T cell death in vivo as already described in vitro (4).

The effect of treatment with NAD⁺ in vivo on different T cell subsets was followed by examining cell phenotypes in lymph nodes 24 h after the injection of NAD⁺ into CD38^−/− BALB/c mice. The results indicate, first, that CD4⁺ cells appear to be more sensitive to NICD than their CD8⁺ counterparts (Fig. 6, A and B). Indeed, 78 ± 11% of CD4⁺ cells and 52 ± 21% of CD8⁺ cells were affected in the lymph nodes of these animals 24 h after the injection of NAD⁺ (p = 0.003, n = 10). This could be accounted for by a difference in ART2 and/or P2X7 density on the plasma membrane. In accord with a previous report, we show that CD8⁺ T cells express a higher level of ART2.2 on their surface than CD4⁺ cells as assessed by staining with ART2.2-specific Nika102 mAb (Ref. 9) and Fig. 6,C). Yet, CD8⁺ T cells display a lower P2X7 receptor density than CD4⁺ cells as assessed by P2X7-specific Hano44 mAb (Fig. 6 C). This suggests that susceptibility to NICD is more influenced by the density of P2X7 than of ART2 on the cell surface.

Second, cells surviving NAD⁺ treatment in vivo show striking phenotypic differences to cells of sham-treated mice. Indeed, increased percentages of CD44^high and CD69^high along with a decreased percentage of CD62L^high T lymphocytes were observed among cells resistant to NICD in vivo (Fig. 6,D). These markers characterize the pool of recently activated/memory T cells. Analysis of cell numbers in vivo reveals that NAD⁺ treatment has no direct effect on the size of the CD44^high T cell population and only a slight effect on the size of the CD69^high population (Fig. 6 E). Yet, both populations seem to expand 10 days after treatment, a time at which the peripheral T cell pool is being replenished. Our results therefore indicate that NAD⁺ treatment has a strong effect on naive T cells and that activated/memory T lymphocytes are resistant to NICD in vivo.

Analysis of ART2.2 expression after NAD⁺ treatment, using Nika102 mAb, reveals that a significant fraction of the CD4⁺ cells and most of the CD8⁺ cells resistant to NAD⁺ in vivo do not express ART2.2 on their surface (Fig. 7,A). Since activated T cells shed ART2 from their surface (37), this along with their CD44, CD62L, and CD69 phenotypes supports the conclusion that endogenously activated or memory cells are less sensitive to the deleterious effect of NAD⁺. Moreover, CD4⁺ and CD8⁺ cells resistant to NICD in vivo are also poorly stained with Hano44 mAb compared with cells from untreated mice (Fig. 7,A), consistent with the notion that sensitivity to NICD is greater for cells expressing high levels of P2X7. Correlatively, we show that activation of murine T cells by Con A (Fig. 7 B) or by anti-CD3 Abs (data not shown) leads to a down modulation of both P2X7 and ART2.2 on the plasma membrane.

The influence of NAD⁺ on the immune response was tested in wild-type, ART2^−/−, and CD38^−/− BALB/c mice after immunization with SRBC. This Ag was selected for several reasons. First, SRBC Ab response is T dependent and quickly elicits detectable primary and secondary serum Ab titers. Second, response to SRBC does not require the use of adjuvants, which are inflammatory and delay Ag delivery. In a first series of experiments, mice were treated with 10 mg of NAD⁺ i.v. at the time of immunization with 2 × 10⁸ SRBC in saline i.p. Mice were bled after 6 days and primary agglutination titers were determined. A much lower Ab titer was observed in CD38^−/− mice treated with NAD⁺ vs untreated mice, whereas NAD⁺ treatment did not show any detectable effect on Ab induction in ART2^−/− mice (Fig. 8 A). This is consistent with the dramatic T cell depletion induced by NAD⁺ in CD38^−/− animals and the resistance of T cells from ART2^−/− mice to NICD in vivo. Conceivably, elimination of most naive T cells by NAD⁺ in CD38^−/− mice affects the recruitment of Ag-specific cells.

A second group of mice was treated with NAD⁺ on day 4 after priming, a time when the response was still building up. They were then boosted with SRBC on day 14 and the secondary response was tested on day 20. Agglutination titers were compared with those in mice treated with NAD⁺ on day 0 and to untreated mice. Results in Fig. 8 B show that treatment with NAD⁺ on day 4 after priming did not block the secondary SRBC response in CD38^−/− mice or even slightly increased it in comparison to untreated CD38^−/− mice. Notice that the secondary response in mice treated with NAD⁺ on day 0 was equivalent to the primary one in untreated animals. These results further support the notion that NICD affects naive T cells, whereas activated T cells escape the deleterious effect of NAD⁺.

Remarkably, different results where obtained in wild-type BALB/c mice which are less sensitive to NICD than their CD38^−/− counterparts. Indeed, treatment with NAD⁺ in these mice led to a significant improvement of both primary and secondary responses compared with controls (Fig. 8). This suggests that NAD⁺-induced partial T cell depletion of naive T cells, which likely corresponds to physiological situations in vivo, may facilitate the expansion of Ag-primed T cells. Note that NAD⁺ treatment at day 14, i.e., immediately before the secondary antigenic challenge, had no effect on the secondary response in wild-type or in CD38^−/− BALB/c mice (Fig. 8 B). This result is consistent with the idea that recently primed T cells responsible for the secondary response are refractory to NICD.

The discovery of mono-ART in vertebrates has opened a new field in cell communication by introducing new ectoenzymes catalyzing posttranslational modifications of proteins using NAD⁺ in the extracellular compartment (3). ART1 and ART2 are presently the two best-characterized paralogs (1), but their role in vivo remains elusive. In mice, ART2 is expressed on most peripheral T cells and remarkably induces P2X7 activation by ADP-ribosylation leading to cell death in vitro (4, 14). Yet, the origin of endogenous extracellular NAD⁺ is still a matter of debate, in which both lytic and nonlytic processes can be envisioned (25, 26, 27). Experiments reported herein provide the first evidence that a significant amount of NAD⁺ is present in exudates during the early phase of the inflammatory response. Inflammation induced by polyacrylamide beads is Ag free and thus reflects tissue reaction to a foreign bioincompatible material. NAD⁺ release precedes massive neutrophil influx into the inflammatory site. Hence, NAD⁺ release may not be conditioned by massive infiltration of neutrophils. Oxidative stress induced by activated neutrophils, as well as hypoxia resulting from the pressure of the Biogel mass under the skin, may contribute to the liberation of NAD⁺ by neighboring cells through a nonlytic process. Importantly, the high local NAD⁺ concentration in exudates, which can reach >10 μM, has no effect on the plasma NAD⁺ level. Local NAD⁺ release at an inflammatory site is thus unlikely to have a global effect on T cell populations in the whole body. Yet, NAD⁺ concentrations in the range of 3 μM were found in the plasma of CD38^−/− mice vs <0.3 μM in their wild-type counterparts. Whether such a high concentration reflects the physiological level in these mice or results from NAD⁺ released in the absence of the major NAD⁺-hydrolase CD38 during blood preparation remains questionable. Micromolar NAD⁺ concentrations are sufficient to induce T cell death in vitro (4) and a NAD⁺ concentration in exudates is sufficient to induce detectable NICD in vitro. Yet, nonmanipulated CD38^−/− mice have T cell numbers and subsets similar to those in wild-type animals. It seems thus likely that the high NAD⁺ plasma level in CD38^−/− animals corresponds to an experimental artifact. More importantly, NAD⁺ released at a local inflammatory site induces a significant T cell depletion in the draining but not in the nondraining nodes from wild-type or highly sensitive CD38^−/− BALB/c mice developing inflammation to Biogel. This depletion is accompanied by CD62L shedding on a significant fraction of remaining T cells, an early event associated with P2X7 activation (18). This, in combination with T cell resistance in ART2^−/− mice, despite NAD⁺ release at the inflammatory site, provides the first evidence that NICD can occur under physiopathological situations in vivo.

Consistently, injection of NAD⁺, the ART2 substrate, into normal mice expressing both ART2 and P2X7 induces massive T cell depletion detectable within a few hours following injection. In support of our findings, accumulation of apoptotic T cells in the liver has been described in the C57BL/6 mouse strain within 12 h after NAD⁺ injection (7), although these mice harbor a partially deficient P2X7 receptor (22). In our experiments, the depletion is aggravated particularly in CD38^−/− mice where it affects >80% of the T cell population in peripheral lymphoid tissues. No effect of treatment on the thymus was observed (data not shown), in agreement with the absence of ART2 on thymocytes (38). The high sensitivity of CD38^−/− vs wild-type BALB/c mice to the deleterious effect of NAD⁺ is consistent with our previous observation indicating that the CD38 NADase, mainly expressed on B lymphocytes, controls the level of ART2-mediated ADP-ribosylation of T cell surface proteins and, therefore, T cell apoptosis in unfractionated peripheral lymphoid populations (28). The resistance of T cells from ART2^−/− BALB/c mice clearly establishes that T cell depletion after NAD⁺ treatment results from ART2- and therefore P2X7-dependent NICD.

The dramatic effect of NAD⁺ injection in vivo not only affects the number of T lymphocytes in peripheral organs from CD38^−/− BALB/c mice but also profoundly influences the distribution of T cell subpopulations. The fraction of CD44^high, CD62L^low and CD69^high T cells is strongly enlarged, suggesting that cells resistant to NICD in vivo are activated/memory T cells. Three arguments support this conclusion. First, T cell activation leads to the shedding of ART2 by the TACE metalloproteinase and renders cells resistant to NICD (37, 38). Evidently, most of the CD4⁺ and CD8⁺ T cells present in lymph nodes from NAD⁺-treated CD38^−/− BALB/c mice do not express ART2.2 on their surface. Then, most of the CD4⁺ and CD8⁺ cells from NAD⁺-treated mice have a low P2X7 density on their surface compared with controls. We directly show that T cell activation induces a down-modulation of P2X7 expression and resistance to NICD. This is in agreement with the low ability of CD44^high activated/memory CD4⁺ T cells to open the P2X7 large pore permeable to ethidium bromide in the presence of ATP (39). Finally, treatment with NAD⁺ at the time of priming with SRBC inhibits the primary response in CD38^−/− BALB/c mice, but treatment on day 4 after priming has no effect or even stimulates the subsequent response to antigenic challenge (Fig. 8). This demonstrates that primed T cells escape the deleterious effect of NAD⁺ administration. Altogether, these results demonstrate for the first time that NAD⁺ can modulate the pool of naive peripheral T cells in vivo, suggesting that endogenous sources of NAD⁺ may exert a similar effect.

One important aspect of ART2 biology concerns its role in the regulation of the immune response. Previous experiments have shown that ART2 ADP-ribosylates several surface proteins including LFA-1, CD8, CD27, CD43, CD44, or CD45 (10, 11, 12). This clearly inhibits cell contact and T cell trafficking (12). ART2-mediated ADP-ribosylation of membrane proteins would thus have an inhibitory effect on the expansion of the T cell response. These conclusions were drawn from experiments performed in C57BL/6 mice, which express ART2 at a very high density on their T lymphocytes (9) but have an impaired P2X7 receptor (22). Conversely, BALB/c T cells have a lower ART2 density but display a functional P2X7 receptor. Our present results indicate that in the BALB/c genetic background, CD4⁺ cells, which are ART2^lowP2X7^high, are more prone to NICD in vivo than CD8⁺ cells which are ART2^highP2X7^low compared to the CD4^+. This suggests that susceptibility to NICD is governed by P2X7 more than by ART2 density.

Why should NAD⁺ set free in the inflammatory phase of an infection kill naive T cells during the early development of the adaptive immune response? This situation, as illustrated by our experiments in CD38^−/− BALB/c mice treated with NAD⁺ at the time of immunization, could severely impair the response. The same experiments, however, show that the situation is different in a normal BALB/c mouse. Priming with NAD⁺ does not impair the response to Ag but even leads to a weak but significant increase of the Ab titers. This is also verified during the secondary response when mice are treated with NAD⁺ 4 days after priming. In these mice, naive T cells are protected from ART2-mediated P2X7 activation and NICD by CD38 expressed on B lymphocytes. Hence, during the early phase of the immune response, while activated T cells down-modulate ART2 and P2X7 expression on their surface, elimination of part of the naive T cells could give space for the expansion of the Ag-primed T cell population and thus increase the response. Alternatively, CD4⁺Foxp3⁺ regulatory T cells may be highly sensitive to NICD, as recently suggested in C57BL/6 mice (21). Elimination of regulatory T cells, which exert a negative effect, could contribute toward improving the response. However, the extent to which the effects of extracellular NAD⁺ on the development of the immune response can be attributed to ART-mediated NICD remains questionable. CD38 is not only a NADase but also an ADP-ribose cyclase. Cyclic ADP-ribose plays an important role in the response of neutrophils and professional APCs to chemotactic signals (34, 40). Consequently, CD38^−/− mice have a diminished Ab response to T-dependent Ags (29). The reduced response to SRBC of CD38^−/− vs wild-type BALB/c mice (Fig. 8 A) is in fair agreement with this conclusion. Combined with NICD, a defect in Ag presentation may thus explain the dramatic impairment of the primary SRBC response in CD38^−/− mice treated with a high dose of NAD⁺. Similarly, in CD38^+/+ BALB/c mice, NAD⁺ injection might increase cyclic ADP-ribose production and improve APC activation and trafficking, compensating the deleterious effect of NICD or even improving the response. The observation that NAD⁺ administration does not enhance the primary response in ART2^−/− mice, however, makes it likely that this effect is primarily due to ART2- mediated NICD.

Taken together, our results demonstrate that NAD⁺ is released during the early phase of inflammation and illustrate its role in immune regulation in vivo. They also stress that ART2 and CD38 are two important, but not necessarily opposed partners in this regulation. Evidently, NICD is a phenomenon which can occur in vivo. Yet further investigation is needed to define its main target and its role in the regulation of T cell homeostasis and the control of autoimmune diseases in normal mice.

The authors have no financial conflict of interest.