CD56^brightCD16⁻ NK Cells Produce Adenosine through a CD38-Mediated Pathway and Act as Regulatory Cells Inhibiting Autologous CD4⁺ T Cell Proliferation

Natural killer cells, which were originally characterized as cytotoxic cells, are a lymphocyte population belonging to the innate immune system, able to spontaneously lyse cancer cells and virus-infected cells (1). In the last two decades, the role of these cells in the control of the adaptive immune response, in particular through the production of proinflammatory and anti-inflammatory cytokines, has been characterized (2, 3).

In humans, two major subsets of NK cells can be defined according to their expression of CD16 (FcγRIIIA, low-affinity receptor for the Fc portion of Ig G) and of CD56 (an adhesion molecule that mediates homotypic interactions) (4). The main subset (90% of total NK cells) in the peripheral blood (PB) is represented by CD16⁺CD56^dim NK cells, which display a high natural and Ab-dependent cytotoxicity and express killer Ig-like receptors and Ig-like transcript 2. In contrast, CD16⁻CD56^bright NK cells are poorly represented in the PB (10%), but they are the main NK cell subset in secondary lymphoid organs, where they make up 75–95% of total NK cells (5). These cells display low cytotoxicity and secrete a variety of cytokines upon stimulation. Recent studies have demonstrated that CD16⁻CD56^bright NK cells are a more immature subset of NK cells, developing toward the CD16⁺CD56^dim phenotype. Yu et al. (6) have demonstrated that the surface expression of CD94 defines an intermediate stage between the two major subsets, represented by CD94^highCD16⁺CD56^dim NK cells, that are on their way to developing into terminally differentiated CD94^lowCD16⁺CD56^dim NK cells.

Recent studies suggest a possible regulatory role for CD16⁻CD56^bright NK cells (7–9). These cells release anti-inflammatory cytokines in response to different stimuli. Moreover, in patients with multiple sclerosis, treatment with IFN-β reduces inflammation in relapsing patients through an increase in the release of anti-inflammatory cytokines, which is associated with an increase of PB CD16⁻CD56^bright NK cells (10). Similarly, treatment of multiple sclerosis patients with daclizumab, a humanized anti–IL-2Rα Ab, led to the expansion of PB CD16⁻CD56^bright NK cells, which was correlated with decreased T cell survival and better prognosis (11).

Recently, Laroni et al. (12) have demonstrated that CD16⁻CD56^bright NK cells suppressed autologous CD4⁺ T cell proliferation in a contact-dependent manner, and that the inhibition increased upon IL-27 treatment. However, the mechanisms whereby CD16⁻CD56^bright NK cells act as regulatory cells have not yet been identified.

Regulation of the immune responses may be achieved through the expression and/or release of different molecules by immunoregulatory cells. Among these, extracellular adenosine (ADO) is responsible for the control of the immune response in physiological and pathological conditions through the interaction with four different G protein–coupled receptors (ADO receptors [ADOR] A1, A2a, A2b and A3) that are expressed by T (13–15) and B (16–18) lymphocytes and NK cells (18–21). ADO is a purine nucleoside produced as the final product of a complex ectoenzyme network. This network is composed of surface molecules with an extracellular catalytic domain, including 1) ADP ribosyl-cyclases (CD38, CD157) (22), 2) ectonucleotide pyrophosphatase/phosphodiesterase-1 (CD203a/PC-1) (23), 3) ectonucleoside triphosphate diphosphohydrolase 1 (CD39) (24), and 4) ecto-5′-nucleotidase (CD73) (24).

The working hypothesis of this initial work is that the expression of selected ectoenzymes by CD16⁻CD56^bright NK cells makes this subset able to produce ADO and support the regulatory functions of this cell subset. Thus, in this study we have analyzed the expression and function of a panel of ectoenzymes in CD16⁻CD56^bright NK cells from healthy donors (HD) and from patients affected by juvenile idiopathic arthritis (JIA), demonstrating, to our knowledge for the first time, that the regulatory function of CD16⁻CD56^bright NK cells is at least in part related to ADO production mediated by an ectoenzymatic network and might be altered during autoimmune/inflammatory diseases.

This study was approved by the Ethics Committee of the G. Gaslini Institute, Genoa, Italy. Buffy coat preparations were obtained from five HD. Surgically removed tonsils from three patients and paired PB/synovial fluid (SF) samples from five JIA patients were obtained following informed consent of patients’ parents or legal guardians. All JIA individuals were classified as limited oligoarticular JIA according to International League of Associations for Rheumatology Durban criteria (25). All samples were collected from untreated patients at diagnosis. PB and pleural effusion (PEF) samples were obtained from patients with pachypleuritis at the Ospedale di Circolo of Varese, following informed consent, with local Ethics Committee approval.

Mononuclear cells (MNC) were isolated by Ficoll-Hypaque density gradients (Sigma-Aldrich, St. Louis, MO). Whole PEF cells were obtained by a single step of centrifugation. In some experiments, CD16⁻CD56^bright NK cells and CD4⁺ T cells were isolated using a CD16⁻CD56⁺ NK cell isolation kit and a CD4⁺ cell isolation kit, respectively (Miltenyi Biotec, Bergisch Gladbach, Germany), following the manufacturer’s protocol. CD16⁻CD56^bright NK cells were cultured overnight in the presence of 5 ng/ml IL-15 before use.

The expression of ectoenzymes was evaluated on total PB or SF MNC from HD and JIA patients, gating on CD3⁻CD16⁻CD56^bright or CD3⁻CD16⁺CD56^dim NK cells, using the following Abs: anti-CD3 allophycocyanin (Becton Dickinson, Franklin Lakes, NJ), anti-CD16 PE (Beckman Coulter, Brea, CA), and anti-CD56 PC7 (Beckman Coulter). Ectoenzyme expression was analyzed using the following mAbs generated in our laboratory and FITC-conjugated by AcZon (Bologna, Italy): anti-CD38 (clone IB4), anti-CD73 (clone CD75), anti-CD57 (clone TB01), anti-CD157 (clone SY/11B5), anti-CD203a/PC-1 (clone 3E8, provided by J. Goding). CD39 expression was analyzed using anti-CD39 FITC mAb (Beckman Coulter). FITC-conjugated irrelevant isotype-matched mAbs were purchased from Beckman Coulter. In some experiments, the expression of ectoenzymes was evaluated on CD16⁻CD56^bright NK cells that were purified as described above.

The expression of ADOR was evaluated on total MNC from JIA patients’ PB or SF, gating on CD3⁻CD16⁻CD56^bright NK cells, using the following purified Abs: rabbit polyclonal anti-ADORA1 (LifeSpan BioSciences), rabbit polyclonal anti-ADORA2a (Thermo Scientific), and goat polyclonal anti-ADORA2b (Santa Cruz Biotechnology). FITC-conjugated goat-anti rabbit Ig (Abcam) and swine anti-goat Ig (Life Technologies) were used as secondary reagents.

Cells were run on a Gallios cytometer and analyzed using Kaluza software (Beckman Coulter). Data were expressed as mean relative fluorescence intensity (MRFI), obtained as follows: mean fluorescence intensity obtained with specific mAb/mean fluorescence intensity obtained with irrelevant isotype-matched mAb.

CD16⁻CD56^bright NK cells (10⁵ or 2 × 10⁵ cells/well) were cultured in RPMI 1640/10% FBS at 37°C and 5% CO₂ in round-bottom 96-well plates (Costar Corning) in the presence or absence of AMP, ADPR, ATP, or NAD⁺ (20 or 50 μM). In some experiments, CD16⁻CD56^bright NK cells were treated for 30 min with erythro-9-(2-hydroxy-3-nonyl)adenine (EHNA; inhibitor of ADO deaminase, 100 μM), NaN₃, or POM-1 (inhibitors of CD39, Sigma-Aldrich, 100 μM) before being cultured with the substrates. Supernatants were collected after 5, 30, or 45 min and acetonitrile was immediately added at a 1:2 ratio at 4°C to stabilize ADO. Samples were then centrifuged at 12,000 rpm and supernatants were collected and stored at −80°C until use. The presence of ADO, AMP, ATP, and NAD⁺ was investigated by HPLC.

Chromatographic analysis was performed with an HPLC System (Beckman Gold 126/166NM, Beckman Coulter) equipped with a reverse-phase column (Hamilton C18, 5 μm; 250 × 4.5 mm). Separation of nucleotides and nucleosides was performed using a mobile-phase buffer (0.125 M citric acid and 0.025 M KH₂PO₄ [pH 5.1]) with 8% acetonitrile (Sigma-Aldrich) during 10 min at a flow rate of 0.8 ml/min. UV absorption spectra were measured at 254 nm. HPLC-grade standards used to calibrate the signals were dissolved in AIM V serum-free medium (Invitrogen, Paisley, U.K.), pH 7.4, 0.2 μm sterile-filtered, and injected in a buffer volume of 20 μl. The retention times (in minutes) of standards were: AMP, 2.15; NAD, 2.8; ADPR, 3.2; and ADO; 5.56. Peak integration was performed using Karat software (Beckman Coulter).

Acetonitrile-treated CD16⁻CD56^bright NK cell supernatants (see above) were evaporated by SpeedVac, reconstituted in mobile-phase buffer, and assayed by HPLC.

The qualitative identity of HPLC peaks was confirmed by comigration of known reference standards. The presence of ADO was also confirmed by spiking standard (50 μM ADO), followed by chromatography. Quantitative measurements were inferred by comparing the peak area of samples with calibration curves for peak areas of each standard compound. Product concentrations were expressed as pmol/30 min/no. of cells (1 × 10⁵ or 2 × 10⁵ cells). Interpolation of data have been obtained using a Microsoft Excel algorithm.

Autologous CD4⁺ T cell proliferation was assessed by CFSE dilution assay. Briefly, cells were stained with CFSE (Invitrogen, 1 μg/ml, 15 min at 37°C), washed, and then cultured in RPMI 1640/10% FBS at 37°C and 5% CO₂, alone or in the presence of anti-CD3/anti-CD28 mAb-coated beads (T cell activation/expansion kit, Miltenyi Biotec). Stimulated CD4⁺ T cells were cultured in the presence or absence of autologous CD16⁻CD56^bright NK cells (NK/CD4 ratio ranged from 1:1 to 1:16). In some experiments, CD16⁻CD56^bright NK cells were treated for 30 min with the following specific inhibitors before being cultured with autologous CD4⁺ T cells (at 1:1 NK/CD4 ratio): kuromanin (inhibitor of CD38, 10 μM, Sigma-Aldrich), POM-1, dipyridamole (inhibitor of nucleoside transporter, 50 μM, Sigma-Aldrich), EHNA, α-β-meADP (inhibitor of CD73, 300 μM, Sigma-Aldrich), and β-γ-meATP (inhibitor of CD203a/PC-1, 300 μM, Sigma-Aldrich).

After 5 d, cells were harvested and washed and then stained with anti-CD4 PE mAb (Beckman Coulter). After additional washes, cells were run on Gallios cytometer, and CFSE dilution was analyzed by gating on CD4⁺ cells, using Kaluza software (Beckman Coulter).

Statistical analysis was performed using Prism 5.03 software (GraphPad Software). Gaussian distributions of data were analyzed using a Kolmogorov–Smirnov test. The Student t test or Mann–Whitney U test was used to compare data, depending on data distribution. Data from paired SF/PB samples were analyzed using a paired t test. A p value <0.05 was considered to be statistically significant.

The expression of ectoenzymes was first analyzed by flow cytometry on PB MNC from four HD, gating on CD16⁻CD56^bright or CD16⁺CD56^dim NK cells (as shown in Fig. 1A). Additionally, we have evaluated the expression of CD57, a marker of NK cells with an unrelated ectoenzyme activity (β-1,3-glucuronyltransferase). As shown in Fig. 1B, the expression of CD38 was similar in the two subsets (mean ± SD: 39.45 ± 10.21 versus 40.44 ± 16.44), whereas the expression of CD73 (86.74 ± 5.42 versus 41.24 ± 13.74, p < 0.05), CD39 (2.27 ± 0.36 versus 1.45 ± 0.12, p < 0.05), CD157 (4.93 ± 0.34 versus 3.16 ± 0.67, p < 0.05), and CD203a/PC-1 (3.29 ± 0.75 versus 1.31 ± 0.18, p < 0.05) was higher in CD16⁻CD56^bright than CD16⁺CD56^dim NK cells. Conversely, the expression of CD57 (131.2 ± 54.88 versus 1.62 ± 0.87, p < 0.05) was higher in CD16⁺CD56^dim than in CD16⁻CD56^bright NK cells. A representative experiment is shown in Fig. 1A. The expression of ectoenzymes was then evaluated on purified PB CD16⁻CD56^bright NK cells from three normal donors, obtaining similar results (Fig. 2A, mean ± SD: CD57, 1.6 ± 0.06; CD157, 5.27 ± 0.51; CD203a/PC-1, 4.04 ± 1.43; CD38, 552 ± 23.56; CD73, 69.33 ± 21.01; CD39, 17.18 ± 0.37). A representative staining is shown in Fig. 2B.

Finally, ectoenzyme expression was evaluated on CD16⁻CD56^bright NK cells purfied from human tonsil. The pattern of expression was similar to PB CD16⁻CD56^bright NK cells (Fig. 2C, mean ± SD: CD57, 1.61 ± 0.05; CD157, 12.09 ± 0.45; CD203a/PC-1, 4.04 ± 0.5; CD38, 53.7 ± 0.45; CD73, 236.2 ± 3.28; CD39, 7.71 ± 0.5).

We asked whether CD16⁻CD56^bright NK cells could produce ADO from different substrates. As shown in Fig. 3A, CD16⁻CD56^bright NK cells produced high amounts of ADO when cultured in the presence of AMP (234 ± 10 pmol ADO/30 min/10⁵ cells) and NAD⁺ (153 ± 160.3 pmol ADO/30 min/10⁵cells), whereas low amounts of ADO were detected using ATP as substrate (4.09 ± 0.12 pmol ADO/30 min/10⁵ cells). Pretreatment of NK cells with EHNA, a specific inhibitor of ADO deaminase, increased the production of ADO from AMP (262.9 ± 33.57 pmol ADO/30 min/10⁵cells), NAD⁺ (188.5 ± 157.4 pmol ADO/30 min/10⁵ cells), and significantly from ATP (43.75 ± 50.54 pmol ADO/30 min/10⁵ cells, p = 0.02). Taken together, these results demonstrate that PB CD16⁻CD56^bright NK cells efficiently produce ADO from AMP and NAD⁺, indicating that CD38, CD203a/PC-1, and CD73 are functional.

The immunosuppressive potential of CD16⁻CD56^bright NK cells was next assessed by cocolture with autologous CD4⁺ T cells at different NK/CD4 ratios in the presence of anti-CD3/anti-CD28–coated beads. As shown in Fig. 3B, the percentage of proliferating CD4⁺ T cells (percentage of proliferating cells ± SD: medium alone, 80.55 ± 10.74) was significantly decreased in the presence of CD16⁻CD56^bright NK cells at 1:1 (51.49 ± 19.53, p < 0.0001), 1:2 (61.51 ± 19.98, p = 0.004) and 1:4 (66.02 ± 18.87, p = 0.015) but not 1:8 (74.24 ± 14.27) or 1:16 (75.87 ± 11.72) NK/CD4 ratios. A representative experiment is shown in Fig. 3C.

To test the impact of each ectoenzyme on the regulatory function of CD16⁻CD56^bright NK cells, the latter cells were pretreated with specific inhibitors of CD38 (kuromanin), CD39 (POM-1), CD73 (α-β-meADP), CD203a/PC-1 (β-γ-meATP). Additionally, specific inhibitors of ADO deaminase (EHNA) and ADO transporter NT (dipyridamole) were tested. As shown in Fig. 3D, inhibition of CD4⁺ T cell proliferation by CD16⁻CD56^bright NK cells at 1:1 CD4/NK ratio (percentage of proliferating cells ± SD: medium alone, 79.57 ± 11.72; CD4 plus NK, 45.91 ± 18.53; p = 0.001) was significantly reverted in the presence of kuromanin (61 ± 25.39, p = 0.05), but not in the presence of POM-1 (34.25 ± 12.88), α-β-meADP (37.61 ± 17.94), β-γ-meATP (40.81 ± 22.56), or EHNA (31.07 ± 29.05). Conversely, NK cell–mediated inhibition was increased in the presence of dipyridamole (18.89 ± 8.24, p = 0.001).

These observations suggest that the function of CD38 is crucial for the regulatory activity of CD16⁻CD56^bright NK cells. Moreover, the inhibition of the nucleoside transporter, which transports ADO across the plasma membrane (thus depleting it from the microenvironment), increases NK cell–mediated inhibition of CD4⁺ T cell proliferation, thus indicating that ADO is responsible for such inhibition.

To test whether the expression and function of ectoenzymes on CD16⁻CD56^bright NK cells might be altered during autoimmune/inflammatory diseases, we analyzed the expression of ectoenzymes on NK cells from PB and SF obtained from JIA patients.

As shown in Fig. 4A, CD16⁻CD56^bright NK cells are significantly enriched in SF, as compared with paired PB samples (percentage of MNC ± SD: PB, 0.23 ± 0.07; SF, 2.75 ± 1.2; p = 0.001; percentage of CD3⁻CD56⁺ NK cells ± SD: PB, 6.5 ± 0.72; SF, 73.55 ± 2; p = 0.0003). Conversely, CD16⁺CD56^dim NK cells are more enriched in PB than in SF (percentage of cells ± SD: PB, 6.89 ± 2.3; SF, 0.72 ± 0.28; p = 0.0004; percentage of CD3⁻CD56⁺ NK cells ± SD: PB, 52.9 ± 11.3; SF, 11.58 ± 1; p = 0.01).

Next, the expression of ectoenzymes was evaluated on MNC from PB or SF, gating on CD16⁻CD56^bright NK cells (as shown in Supplemental Fig. 1). As shown in Fig. 4B, the expression of CD57 (mean ± SD: PB, 1.42 ± 0.19; SF, 1.22 ± 0.08), CD157 (mean ± SD: PB, 9.42 ± 4.7; SF, 4.88 ± 1.8), CD203a/PC-1 (mean ± SD: PB, 5.21 ± 1.51; SF, 4.47 ± 2), and CD39 (mean ± SD: PB, 2.88 ± 2.86; SF, 1.28 ± 0.2) was similar between PB and SF CD16⁻CD56^bright NK cells. Conversely, the expression of CD38 (mean ± SD: PB, 77.5 ± 34.49; SF, 28.09 ± 8.1; p = 0.02) and CD73 (mean ± SD: PB, 173.8 ± 73.57; SF, 83.12 ± 29; p = 0.04) was significantly lower in SF than in PB CD16⁻CD56^bright NK cells. A representative experiment is shown in Supplemental Fig. 1.

To test whether the lower expression of CD38 and CD73 may affect the regulatory activity of CD16⁻CD56^bright NK cells, the latter cells were cocultured with autologous stimulated CD4⁺ T cells. As shown in Fig. 4C, SF CD16⁻CD56^bright NK cells failed to inhibit autologous CD4⁺ T cell proliferation at 1:1 CD4/NK ratio (percentage of proliferating cells ± SD: CD4⁺ T cells alone, 94.12 ± 6.21; CD4/NK 1:1, 97.1 ± 2.48). In contrast, CD16⁻CD56^bright NK cells from inflammatory PEF strongly inhibited autologous CD4⁺ T cell proliferation at 1:1 CD4/NK ratio (percentage of proliferating cells ± SD: CD4⁺ T cells alone, 79.82 ± 1.91; CD4/NK 1:1, 26.12 ± 5.83, p = 0.05).

Fig. 4D shows a representative experiment performed with CD16⁻CD56^bright NK cells from HD PB, JIA patient SF, and inflammatory PEF cultured with autologous CD4⁺ T cells at different CD4/NK ratios. These data suggested that regulatory function of CD16⁻CD56^bright NK cells was increased during inflammation, whereas such function was impaired during autoimmune diseases.

We next evaluated whether the loss of regulatory activity in SF CD16⁻CD56^bright NK cells might be related to a different ADO production in these cells. ADO production from different substrates was evaluated at different time points using CD16⁻CD56^bright NK cells from PB or SF of JIA patients. As shown in Fig. 5A, ADO production profile was similar in PB and SF NK cells using as substrate AMP (upper left panel) or ADPR (upper right panel). ADO production from AMP showed a peak at 5 min in both cell populations and then decreased, whereas a linear increase of ADO production from ADPR was detected.

A different behavior between PB and SF NK cells was observed using ATP as substrate, because SF NK cells displayed a peak of ADO production at 5 min, whereas PB NK cells showed a very low ADO production at 5 min, and then a linear increase after 30 min. Conversely, ADO production from NAD⁺ in PB NK cells displayed a peak after 5 min and then increased again after 30 min, whereas in SF NK cells displayed a peak after 30 min and then rapidly decreased.

Taken together, these results suggest that CD73 and CD203a/PC-1 activities is similar in PB and SF NK cells. On the contrary, the activities run by CD38 and CD39 appears different. However, the ADO levels produced by both NK cell populations are high.

To analyze the impact of the classical (CD39 and CD73) and alternative (CD203a/PC-1 and CD73) ectoenzymatic pathway on ADO production by CD56^bright CD16⁻ NK cells, we tested the consumption of ATP and the production of AMP in the presence of NaN₃ and POM-1, two CD39 inhibitors. As shown in Fig. 5B, ATP consumption and AMP production were only partially inhibited in the presence of NaN₃ (percentage of inhibition, mean ± SD: 35.6 ± 3.11 and 23.83 ± 3.58, respectively) or POM-1 (percentage of inhibition, mean ± SD: 37.25 ± 12.06 and 41.47 ± 2.07, respectively). These results indicate that CD203a/PC-1 is more efficient than CD39 in the conversion of ATP in AMP, the substrate used by CD73 for ADO generation by CD56^brightCD16⁻ NK cells.

Finally, to test whether the lack of inhibitory activity of SF CD56^brightCD16⁻ NK cells might be related to a different expression of ADOR on such cells, we analyzed the expression of ADORA1, ADORA2a, and ADORA2b in PB and SF NK cells from JIA patients. As shown in Fig. 5C, the expression of such receptors was very low in these cell populations. However, PB NK cells display a significantly higher expression of ADORA1 (percentage of positive cells ± SD: 13.4 ± 7.8 versus 1.3 ± 1.3; p = 0.01) and ADORA2a (percentage of positive cells ± SD: 7 ± 4.3 versus 1.4 ± 0.7; p = 0.05) than SF NK cells. Conversely, the expression of ADORA2b was higher in SF NK cells than in PB NK cells (percentage of positive cells ± SD: 6 ± 3.9 versus 1.9 ± 2.3; p = 0.05).

Considerable evidence suggests that CD16⁻CD56^bright NK cells may be involved in the control of the immune responses, mainly through the production of anti-inflammatory molecules (7–11). Moreover, a correlation between the clinical success of anti-inflammatory treatments and the expansion of CD16⁻CD56^bright NK cells has been described in patients with different autoimmune and inflammatory diseases (7, 9–11). Recently, novel immunoregulatory properties have been ascribed to this NK cell subset, including the ability to suppress autologous CD4⁺ T cell proliferation (12). However, the body of information available concerning the mechanisms involved in the “regulatory” properties of CD16⁻CD56^bright NK cells is quite limited.

The original findings of the present study are that CD16⁺CD56^dim and CD16⁻CD56^bright NK cells display different patterns of ectoenzyme expression, which in turn may reflect their different roles in controlling the immune response. In detail, CD16⁺CD56^dim NK cells virtually lack CD203a/PC-1 and CD39, and they express lower levels of CD73. CD203a/PC-1 is a key molecule in the alternative ADO generation pathway, because it may convert ADPR (generated by CD38 from NAD⁺) or ATP to AMP (26, 27). CD39 is involved in a classical pathway independent of CD203a/PC-1, because the former molecule generates AMP from ATP and ADP (28). Both pathways converge to CD73, which converts AMP to ADO.

Consistent with data obtained by Laroni et al. (12) we found that CD16⁻CD56^bright NK cells suppress autologous CD4⁺ T cell proliferation at 1:1 NK/CD4 ratio. Notably, our data demonstrated that CD16⁻CD56^bright NK cells significantly inhibited CD4⁺ T cell proliferation also at lower NK/CD4 ratios (1:2 and 1:4), which are closer to the physiological figures that can be achieved in secondary lymphoid organs or inflamed tissues.

We have demonstrated that CD16⁻CD56^bright NK cells are able to secrete high amounts of ADO from different substrates. Among these, NAD⁺ is converted to AMP more efficiently than ATP, suggesting that CD38, rather than CD203a/PC-1, is the key molecule for ADO production in this context. This conclusion was reinforced by the finding that NK cell–mediated inhibition of autologous CD4⁺ T cell proliferation was dampened by kuromanin, a specific inhibitor of CD38. Such effect was not achieved using POM-1 or β-γ-meATP (specific inhibitors of CD39 and CD203a/PC-1, respectively). These results were not surprising, because we have demonstrated that ADO can be produced by CD16⁻CD56^bright NK cells using either the canonical or the alternative pathway. Moreover, β-γ-meATP may also activate NK cells, either by interacting with P₂X receptor instead of inhibiting CD203a/PC-1 enzymatic activity (29), or inducing cAMP formation. Surprisingly α-β-meADP, a specific inhibitor of CD73, did not restore T cell proliferation. It is conceivable that, in our experimental conditions, the inhibition of CD73 was not total. Another possible explanation is the presence of alkaline phosphatase, which may act as a surrogate for the activity of CD73, leading to ADO production in the presence of CD73 inhibition (4).

CD16⁻CD56^bright NK cells may function as regulatory cells in physiological and/or pathological conditions (such as cancer or immune deficiencies), and this feature may be dampened in autoimmune/inflammatory settings. This hypothesis was supported by the finding that CD16⁻CD56^bright NK cells in PEF exert a potent regulatory activity. In contrast, CD16⁻CD56^bright NK cells infiltrating the synovial tissue of patients with JIA displayed a lower expression of CD38 and CD73 (the key molecules for ADO production) than did their PB counterparts, and they failed to inhibit autologous CD4⁺ T cell proliferation. The lower expression of CD38 in synovial NK cells is apparently unrelated to a cleavage operated by metalloproteases, as confirmed by the absence of soluble CD38 in SF (data not shown). Despite lower CD38/CD73 expression, SF NK cells produced high amounts of ADO from different substrates. However, the kinetics of ADO production from ATP or NAD⁺ in PB was different from SF NK cells. NAD⁺ was converted more rapidly in ADO by PB NK cells, thus reflecting that CD38 activity in PB is higher than that of SF NK cells. This feature may explain the defective immunosuppressive activity featured by SF NK cells. Another possible explanation might be related to the different expression of ADORA1, ADORA2a, and ADORA2b between PB and SF NK cells. Beavis et al. (30) have demonstrated that ADORA2a and ADORA2b display different signaling pathways and differential activity on NK cells. It is tempting to speculate that the higher expression of ADORA2b detected on SF NK cells may lead to an autocrine consumption of ADO produced by the latter cells, and it may cause a functional impairment, rendering them unable to inhibit autologous CD4⁺ T cell proliferation.

In conclusion, this study has delineated a novel immunoregulatory function for CD16⁻CD56^bright NK cells. Such function is related to the production of ADO by a complex network of ectoenzymes, a process in which CD38 plays a pivotal role, and may be altered during autoimmune/inflammatory diseases. These results may pave the way to characterization of the function of these cells in different pathological conditions, where the regulation of the normal immune response is altered.

We thank Camilla Valentino for excellent secretarial assistance.

The authors have no financial conflicts of interest.