Specific Requirements for Vγ9Vδ2 T Cell Stimulation by a Natural Adenylated Phosphoantigen¹

The majority of human peripheral T cells of the γδ type express a TCR comprising Vγ9 and Vδ2 variable regions. These usually represent 1–10% of circulating T lymphocytes, expand, and release inflammatory cytokines in response to diverse infectious agents such as mycobacteria, bacteria, and parasites (1). They also display cytotoxicity toward diverse tumor cells including laboratory tumor lines (Daudi Burkitt’s lymphoma, RPMI-8226 myeloma) and multiple cancer lines from patients with, among others, lung, prostate, liver, or kidney cancers (2, 3, 4).

Metabolites that stimulate Vγ9Vδ2 T cells in a TCR-dependent manner have been purified from mycobacteria and called phosphoantigens because they consist in a short alkyl chain and a terminal pyrophosphate group (5, 6, 7). The most powerful natural Ag reported thus far is hydroxyldimethylallyl pyrophosphate (HDMAPP),⁴ a metabolite produced by microbial organisms through the deoxylulose pathway, an alternative bacterial route to the isoprenoid precursor isopentenyl pyrophosphate (IPP; Refs. 8, 9, 10). Pyrophosphorylated Ags can directly activate Vγ9Vδ2 T cells and do not absolutely require presentation by specialized APCs or display on MHC or CD1 Ags (7, 11). Nucleotidic derivatives of phosphoantigens have been purified from mycobacteria and have been found to stimulate Vγ9Vδ2 cells (5, 6). Multiple synthetic Ags analogous to pyrophosphorylated compounds as well as nucleotidic variants have been studied, and some of them proved to be powerful stimulants of Vγ9Vδ2 T cells (12, 13, 14).

The precise mechanism underlying phosphoantigen recognition is still unclear. Structure-activity correlation studies of synthetic phosphoantigens indicated that, although short alkyls such as ethyl pyrophosphate display some activity, optimal structures contain five carbon atoms and one double bond (15). Belmant et al. (16) reported that the β-phosphoester bond in pyrophosphorylated Ags must be cleavable, and a recent study by Boedec et al. (17) reported the strong stereospecificity of structural alkyl variants. Whether this reflects constraints for binding to the TCR or to a putative presentation molecule is unknown.

TCR gene transfer experiments indicate that phosphoantigen recognition by Vγ9Vδ2 T cells is TCR dependent (11). Nevertheless, a direct interaction between the TCR and phosphoantigens is only indirectly supported by structure-function correlations (18, 19, 20) and molecular interaction models (21). In these studies, amino acids in the γ-chain CDR3 were found to be essential for phosphoantigen recognition.

The presence of APCs improves direct Vγ9Vδ2 T cell responses to phosphoantigens (22), but their exact contribution is only partially characterized. Analyses with strong pyrophosphorylated agonist Ags such as bromohydrin pyrophosphate, or HDMAPP (22, 23, 24) have shown that to some extent these Ags can be captured by APCs and confer stable stimulatory activity, indicating that they can be somehow displayed on the cell surface as on stimulatory tumors. A mechanism for phosphoantigen presentation is, however, supported by recent experiments showing binding of soluble Vγ9Vδ2 tetramers on APCs pulsed with HDMAPP. This revealed a direct and Ag-specific recognition of the membrane of pulsed cells which was dependent on a trypsin-sensitive structure (24, 25).

Tumor cells such as Daudi Burkitt’s lymphoma produce weak phosphoantigens through the mevalonate pathway. The major stimulatory compound, IPP, is overproduced in some tumors following hyperactivation of the mevalonate pathway (26, 27). IPP can also accumulate in cells after treatment with aminobisphosphonates through inhibition of the mevalonate pathway enzyme farnesyl pyrophosphate synthase, resulting in an increased sensitivity to Vγ9Vδ2 T cell cytolytic activity. Aminobisphosphonates also induce intracellular accumulation of 1-(adenosin-5′-yl) 3-(3-methylbut-3-enyl)triphosphoric diester (ApppI), an adenine nucleotide derivative of IPP in macrophages (28). Because the structure of this compound is reminiscent of bacterial phosphoantigens such as TubAg3/4 and other analogous synthetic nucleotide derivatives with stimulatory activity, we have studied its biological activity on Vγ9Vδ2 T cells (6). This revealed unique properties for this nucleotidic phosphoantigen, making this type of compound of particular interest for in vivo administration. We show evidence that nucleotidic Ags are naturally produced in tumor cells and can play a major role in their stimulatory activity toward Vγ9Vδ2 lymphocytes.

Anti-TCRVδ2-FITC (IMMU360) and CD3-PE (UCHT1) were from Beckman-Coulter, anti-IFN-γ-PE (25273) was from R&D systems, and anti-CD107a-PE (H4A3) was from BD Pharmingen. Apyrase, Crotalus adamanteus venom nucleotide pyrophosphatase (NPP) and other reagents were from Sigma-Aldrich except isopentenyl pyrophosphate (Isoprenoids) and ApppI (see below). Apyrase and NPP were used in cellular and biochemical assays at concentrations of 0.2 and 0.02 U/ml, respectively.

The procedure was adapted from Ryu and Scott (29). Briefly, 110 mg (0.2 mmol) of ATP (sodium salt) were converted into the corresponding tetrabutylammonium salt and then dissolved in dry acetonitrile (2 ml). To this mixture a solution of isopentenyl tosylate (1 Eq, 48 mg) was added in dry acetonitrile (2 ml). The reaction mixture was stirred at room temperature overnight and concentrated. The oily residue was purified on a DEAE-Sephadex column using a gradient of triethylammonium bicarbonate buffer (pH 6.9, 0.2–1 M). The fractions of interest were evaporated under vacuum, and passed through an ion-exchange Dowex (Na⁺ form) column to give ApppI as sodium salt with a final yield of 45%. ApppI was stored at −20°C. Aliquots were reconstituted in molecular biology grade water and stored at −80°C. In some preparations, HPLC analysis of reconstituted product revealed the presence of 5–10% ADP (presumably from spontaneous degradation). ADP does not affect responses of Vγ9Vδ2 cells in stimulations. This was, however, checked additionally in control experiments in which this contaminating product was removed by apyrase treatment. To this end, ApppI was pretreated for 1 h with 0.2 U/ml apyrase, and the enzyme was then inactivated by heating at 100°C for 10 min.

Cell culture reagents were from Invitrogen except human serum (PAA Laboratories) and IL-2 (Sanofi-Synthelabo). Tumor cell lines were obtained from American Type Culture Collection and grown in RPMI 1640-Glutamax supplemented with 10% heat-inactivated FCS, sodium pyruvate, and penicillin/streptomycin or in Hybridoma SFM medium supplemented with sodium pyruvate, penicillin/streptomycin, and l-glutamine (complete SFM) for pulsing experiments given that pulsing was less efficient in the presence of FCS. In this case, cells were washed and incubated with the indicated concentration of IPP or ApppI for 2 or 16 h, washed again four times, pelleted by centrifugation, and used in T cell stimulation assays.

For polyclonal Vγ9Vδ2 T cell lines establishment, PBMCs isolated from buffy coats from healthy donors (Etablissement Français du Sang) were stimulated with 5 μM phosphoantigen (IPP or ApppI) in RPMI 1640-Glutamax medium supplemented with 10% heat-inactivated human serum, sodium pyruvate, and penicillin/streptomycin. Twenty-four hours after stimulation, IL-2 was added at a final concentration of 200 U/ml. On day 5, IL-2 concentration was raised to 400 U/ml. Cells were passed every 2–3 days, maintained at a concentration of 6 × 10⁵/ml, and frozen after 22–24 days of culture. Cell lines contained >94% Vδ2⁺ cells. Upon thawing, lymphocytes were left to recover overnight in culture medium without IL-2 before assessment of Ag responses.

PBMCs from healthy donors were sorted for CD14⁺ cells using a MACS column (Miltenyi Biotec). DCs were obtained by culturing CD14⁺ cells in six-well plates (3 × 10⁶ cells/ml) for 7 days in AIM-V serum-free medium (Invitrogen) supplemented with GM-CSF (100 ng/ml; Novartis) and IL-13 (50 ng/ml). Fresh IL-13 was added again after 4 days of culture. The phenotype of immature DCs (iDC) was monitored as described previously (30) and was CD1a⁺MHC-I⁺MHC-II⁺CD64⁻CD83⁻CD80^lowCD86^low. For induction of maturation, iDCs were cultured for 48 h in the presence of 100 ng/ml LPS and checked for maturation markers (CD83 and CD86). In some experiments, DCs were either pulsed with pamidronate-IPP-ApppI (10 μM) for 14–16 h or kept untreated. After pulsing, cells were washed three times with RPMI 1640 and used as APCs in Vγ9Vδ2 T cell activation assay using a T cell-APC ratio of 1:1.

PBMCs were stimulated as described for polyclonal Vγ9Vδ2 T cell line establishment, collected at the indicated times, and stained for 30 min at 4°C with anti-TCRVδ2-FITC and anti-CD3-PE Abs (1/25 and 1/50 final dilutions, respectively) in PBS-5% FCS (FACS medium). Data were acquired using a FACScan cytometer (BD Biosciences).

Polyclonal Vγ9Vδ2 T cell lines were stimulated for 5 h in 96-well U-bottom plates with the indicated Ags after a brief centrifugation to increase cell-cell contact. Brefeldin A (10 μg/ml) was added 1 h after the initiation of culture. Cells were then washed and stained with anti-TCRVδ2-FITC Ab (1/25 final dilution) in FACS medium, fixed for 15 min with PBS containing 2% paraformaldehyde (w/v) at room temperature, and then permeabilized with saponin buffer (PBS, 10 mM HEPES, 5% FCS, 0.1% saponin) for 15 min at room temperature. Cells were then stained with anti-IFN-γ-PE Ab (1/100 final dilution) in saponin buffer for 30 min at 4°C and washed twice in FACS medium. Data were acquired and analyzed as for proliferation assays. Results are expressed as the mean of triplicate microwell cultures ± SD.

Experiments were performed as previously described (31) with minor modifications. Briefly, polyclonal Vγ9Vδ2 T cell lines (10⁵ cells) were stimulated in 96-well U-bottom plates with the indicated Ags, Ab-coated beads, or tumor cell lines (1:1 ratio) for 5 h after a brief centrifugation, in the presence of the anti-CD107a-PE Ab (1/25 final dilution) in a final volume of 100 μl. Brefeldin A was omitted. Cells were washed twice in FACS medium and stained for TCR-Vδ2. Data were acquired and analyzed as for proliferation assays. Results are expressed as the mean of triplicate microwell cultures ± SD.

K562 were either pulsed with ApppI (20 μM) or kept untreated for 16–18 h, washed twice, stained with hexidium iodide (Molecular Probes; 5 μM), and mixed with Vγ9Vδ2 T cells (1:1 ratio), centrifuged at 1200 rpm for 1 min, and incubated at 37°C for 30 min (APCs). Cells were then resuspended gently and layered onto poly-l-lysine-coated eight-well slides. After 10 min, 0.02% NaN₃ was added, and slides were kept at room temperature for a further 15 min. The wells were washed once with PBS containing 5% FBS and 0.02% NaN₃. Finally, cells were stained with anti-δ2-FITC, and the slides were mounted and observed under a Zeiss LSM 510 (Zeiss) confocal microscope with a 63 Plan-Apochromat objective (1.4 oil), electronic zoom 3.

For NPP and apyrase sensitivity assays, reaction mixes containing nucleotidic compounds were adjusted to 200 μl with water and products were separated by HPLC (System Gold; Beckman Coulter) on a PL-SAX anion-exchange column (1000 Å, 5 μm, 50 × 4.6 mm; Polymer Labs Varian) using water (solution A) and 1 M (NH₄)₂HPO₄, pH 3.5 (solution B) and the following elution gradient: 0–1 min, 0% solution B; 1–11 min, 0–30% solution B; 11–13 min, 100% solution B; 13–20 min, 100% solution A. Eluted nucleotides were detected by monitoring UV optical density at 260 nm. To evaluate phosphatase sensitivity, ApppI and ATP (100 μM final concentration) were incubated for 1 h at 37°C in 20 μl of RPMI in the presence of 0.2 U/ml apyrase or 0.02 U/ml nucleotide pyrophosphatase before analysis.

For the assessment of intracellular Ags, Daudi or K562 cells (3 × 10⁸ in 250 ml) were washed in PBS, pelleted, and resuspended in 600 μl of ice-cold H₂O plus 600 μl of ice-cold acetonitrile. Cell lysis was completed using a FastPrep-24 tissue homogenizer and lysing matrix D ceramic spheres (MP Biochemicals) set at 4m/s, with 4 pulses of 15 s. The cell lysate was ultracentrifuged (2 × 10⁵ g, 15 min) and the supernatant (cytosolic fraction) was lyophilized and redissolved in 200 μl of H₂O. This was loaded on an anion exchange cationic column (Amersham Life Sciences, type Q) and bound material was eluted (2 ml/min) using H₂O (solution A) and ammonium acetate 1 M, pH 6.5 (solution B), with the following gradient: 0–1 min, 0% solution B; 1–17 min, 0–100% solution B; 17–20 min, 100% solution B; 20–25 min, 0% solution B. Fractions were collected every 0.5 min, lyophilized, and redissolved in 100 μl of H₂O. Ten microliters of each fraction were then tested for stimulatory activity on Vγ9Vδ2 T cell microcultures.

Mass spectrometry was performed using an LCQ ion-trap mass spectrometer (Finnigan MAT). Nano electrospray analysis was performed using a nanospray ESI source (Protein Analysis). Palladium- and gold-coated glass capillaries (Proxeon Biosystems) were filled with 3 μl of analyte solution and positioned using a stereomicroscope at a distance of 1 mm to the opening of the heated transfer capillary kept at a temperature of 150°C. A nebulizer gas was not necessary in this spray mode. The nanospray needle voltage was set to 700 V. Samples were analyzed in the negative ion mode. ApppI (m/z 574) was fragmented using the following parameters: parent ion isolation width, 3 m/z; collision energy, 30%; activation Qz, 0.25; activation time, 30 ms. MS³ analysis of the ApppI fragment at m/z 408 was performed using the following parameters: parent ion isolation width, 3 m/z; collision energy, 33%; activation Qz, 0.25. Analysis was performed with the Xcalibur 1.2 software (Thermo Electron).

The adenylated derivative of IPP will be referred to as ApppI. In this compound, the isopentenyl moiety is linked to the γ-phosphate of ATP instead of being linked to a pyrophosphate group (Fig. 1). When added in a PBMC culture, ApppI induced the selective expansion of Vγ9Vδ2 T cells, and the culture reached homogeneity (∼94%) in 3 wk, as is the case for most phosphoantigens (Fig. 2,A). Although IPP and ApppI both induce proliferation and intracellular IFN-γ production, ApppI required a 5- to 10-fold higher molar concentration for a similar effect on Vγ9Vδ2 T cell expansion from PBMCs (Fig. 2,B). The adenylated derivative was also less bioactive than IPP for the stimulation of lymphokine production by Vγ9Vδ2 T cells, with half-maximum responses obtained at ∼30 μM instead of ∼1 μM for IPP (Fig. 2,C). Both compounds induced the cytotoxicity machinery as evidenced by the surface expression of the lysosomal marker CD107a, although IPP induced again a stronger stimulation (Fig. 2,D). Addition of ApppI to IPP slightly increased stimulation, indicating that ApppI does not compete with IPP and has no toxic activity on responder cells (Fig. 2 E).

The phosphate groups of IPP are sensitive to terminal phosphatases such as alkaline phosphatase (5, 14) or apyrase. Conversely, ApppI is expected to be resistant to degradation by the same enzymes. Indeed, treatment of ApppI with apyrase did not change its HPLC profile, whereas ATP used as a control for enzyme activity is completely hydrolyzed in similar conditions (Fig. 3,A). Strikingly, when added in stimulation assays, apyrase abolished the response to ApppI in the CD107 expression assay (Fig. 3 B) as well as in cytokine production assays (data not shown). A toxic effect of apyrase on lymphocyte responses is excluded because this enzyme does not influence the responses induced by anti-CD3 stimulation. This suggests that ApppI has no stimulatory activity per se and requires conversion into a compound in which the phosphate is terminal, most likely IPP.

The structure of ApppI indicates that it could be hydrolyzed to give IPP plus AMP through cleavage of the αβ phosphoester bond by a nucleotide pyrophosphatase activity (32). ApppI was thus treated with C. adamanteus venom NPP and subsequently analyzed by HPLC. The profile of NPP-treated ApppI indicates the release of a nucleotidic moiety that elutes like AMP but not like ADP and is compatible with the expected cleavage of ApppI between the α and β phosphates relative to the adenosinyl group (Fig. 4,A). The release of IPP following NPP treatment was further checked by mass spectrometry detection of IPP in a fraction eluting with the same retention time as ADP (data not shown). We then evaluated the effect of NPP addition in the culture medium of ApppI-stimulated Vγ9Vδ2 T cells. NPP addition to the culture medium strongly increased the stimulatory activity of ApppI, reaching a stimulation level close to that of the same molar concentration of IPP. Because the enzyme was deprived of any effect in the absence of Ag, this was not due to the presence of nucleotidic Ags in the cultures or to a nonspecific stimulatory activity of the enzyme (Fig. 4,B). NPPs are sensitive to the pyrophosphatase inhibitor pyridoxal phosphate-6-azophenyl-2′,4′-disulfonic acid (PPADS; Ref. 33). PPADS abolished the weak spontaneous response to ApppI in the absence of exogenous NPP, confirming that this response is due to partial conversion of ApppI into IPP in the culture medium. PPADS had no effect on the CD3-mediated stimulation, excluding a nonspecific toxicity of this compound (Fig. 4 C). Altogether, these data indicate that ApppI has no significant intrinsic stimulatory activity and requires conversion into IPP for the stimulation of Vγ9Vδ2 T cells in the absence of APCs.

It is largely documented that Vγ9Vδ2 T cell stimulation is increased by the presence of bystander APCs (22). However, although strong agonist pyrophosphorylated Ags can be pulsed on APCs (23, 24), this is generally considered as poorly efficient even on DCs (34, 35). When RPMI-8226 cells, which are weak stimulatory tumors, were added in cultures of Vγ9Vδ2 T cells and phosphoantigens, we observed a synergistic activation which was more pronounced with ApppI than with IPP (Fig. 5,A). This prompted us to compare the pulsing efficiency of these two Ags on different APCs. We first compared the ability of IPP and ApppI to be pulsed on tumor lines that are not professional APCs. K562 and RPMI-8226 myeloma cell lines have weak spontaneous stimulatory activity, whereas Daudi has a strong spontaneous stimulatory activity for Vγ9Vδ2 T cells. All three cell lines were incubated overnight with both Ags, extensively washed, and used for stimulation. Although pulsing with both Ags increased the stimulatory activity of tumors, pulsing with ApppI was again more efficient than with IPP, and ApppI-pulsed tumors were stronger stimulators than cells pulsed with the same molar concentration of IPP (Fig. 5 B).

These results could be explained by an effect of ApppI similar to aminobisphosphonate blockade of farnesyl pyrophosphate synthase. If this was the case, an ApppI effect would be abolished by simultaneous treatment of tumors with lovastatin, preventing the accumulation of endogenous IPP (26). Fig. 5 C shows that lovastatin treatment abolishes Daudi spontaneous stimulatory potential and K562 sensitization with pamidronate but has no effect on K562 sensitization with ApppI. Thus ApppI does not act as an inhibitor of farnesyl pyrophosphate synthase.

To determine whether the acquisition of stimulatory activity by tumor cells required an active process, we studied the effect of temperature and cytoskeleton integrity on pulse efficiency. We used RPMI-8226 cells and short pulsing conditions (2 h) which confer stable stimulatory activity without cell alteration when varying the temperature of pulse or with drug addition during pulsing. Pulsing was abrogated when performed at 4°C or on cells that were previously treated with the actin polymerization inhibitor cytochalasin D (Fig. 5,D). This indicates that pulsing is not due simply to passive adsorption, requires an active mechanism, and involves cytoskeletal components, indicative of an active process. Finally, a specific TCR engagement following ApppI recognition on K562 cells is substantiated by a massive TCR aggregation on the T cell surface (Fig. 5 E) as well as subsequent TCR down-modulation by FACS (not shown).

Because the direct stimulation of Vγ9Vδ2 T cells by ApppI can be modulated by addition of NPP and apyrase, it was important to check the effect of such treatments on the stimulation by pulsed cells. These enzymes were thus added in T cell stimulation assays using either Daudi cells or RPMI-8226 cells pulsed with pamidronate or ApppI (Fig. 5 F). In all cases, stimulation with tumor cells was completely resistant to treatment with apyrase and NPP whereas direct simulation was sensitive to phosphatases. Thus, the Ag was not recognized in soluble form, was somehow processed to become resistant to these enzymes, and was stably presented on tumor surfaces. This also indicates that the Vγ9Vδ2 T cell stimulation by Ag-pulsed tumors and their direct stimulation by soluble IPP or ApppI result from different mechanisms and different antigenic forms.

DCs have previously been shown to potentiate phosphoantigen responses (36). Based on above observations we compared the pulsing efficiency of IPP and ApppI on monocyte-derived DCs. Because iDCs are more prone to capture exogenous Ags than mature DCs (mDC) we also analyzed the ability of DC to be pulsed with IPP and ApppI and to stimulate Vγ9Vδ2 cells following maturation by 24-h LPS treatment (Fig. 6). Maturation of DC was confirmed by the increase of CD83 and CD86 surface expression. After incubation with IPP or ApppI (10 μM) and extensive washing to eliminate passively adsorbed Ags, both iDC and mDC populations were able to stimulate Vγ9Vδ2 lymphocytes. ApppI was, however, more efficiently pulsed than IPP, and the stimulatory activity of ApppI-pulsed DC reached that of the same population treated with pamidronate (50 μM) instead of phosphoantigen. The use of mature DC was associated with a much lower Vγ9Vδ2 lymphocyte response, suggesting that Ags are captured less efficiently after DC maturation. Thus, ApppI is similarly captured with higher efficiency than IPP by professional and nonprofessional APCs.

Above results indicate that ApppI-pulsed cells are quite similar to spontaneously stimulating tumors and that ApppI is more prone than IPP to be captured and displayed in a stimulatory form by tumors. This suggests that nucleotidic derivatives of phosphoantigens could be produced intracellularly by tumors as well as IPP. To check for the presence of nucleotidic stimulatory compounds in Daudi cells, cytosolic extracts were prepared and separated by ion exchange HPLC. After elution of bound material, fractions were collected, treated or not with apyrase to remove nonnucleotidic phosphoantigens, and tested for their stimulatory activity on Vγ9Vδ2 cells in the presence or absence of NPP. In the absence of enzyme treatment, the major peak of activity eluted at 9.75 min, corresponding to the elution time of purified IPP (not shown). This activity was severely decreased following apyrase treatment, in agreement with the presence of Ags with terminal phosphates in this fraction. After apyrase treatment, stimulatory activity was barely detectable in the absence of NPP. However, in the presence of this enzyme, several fractions were active, suggesting the presence of nucleotidic phosphoantigens. A major peak of activity eluted with a retention time close to that of purified ApppI (10.5–11 min), and the resulting activity exceeded that of the apyrase-sensitive fraction (Fig. 7). This fraction was subsequently analyzed by electrospray ionization ion trap mass spectrometry.

Synthetic ApppI gives a m/z 574 pseudomolecular ion ([M-H]⁻). As previously described (28), its fragmentation produces major ions at m/z 408 and m/z 227). To increase the mass spectrometry signature of ApppI, MS³ analysis was performed on the major ion with m/z 408 which could be further fragmented and produced an ion with m/z 273 and the pyrophosphate-related ion with m/z 159 (Fig 8 A). The HPLC fraction containing the putative natural nucleotidic phosphoantigen was then analyzed using identical mass spectrometry parameters. An ion with m/z 574 was present at trace levels (not shown). After fragmentation, this gave the characteristic ions of ApppI in both the MS² and the MS³ spectra (Fig. 8 B). Thus, although other stimulatory compounds may also be present in cell extracts, ApppI is produced intracellularly by Daudi cells. In our experimental conditions, IPP could not be unequivocally identified in the corresponding elution fraction by mass spectrometry (not shown).

A similar experiment was performed on extracts from the K562 cell line which does not efficiently activate Vγ9Vδ2 cells. An ion with the corresponding m/z 574 was detected and fragmented to identify ion fragments characteristic of ApppI. The m/z 227 ion could not be detected. An ion with a mass of m/z 408 was detected at trace level. However, its fragmentation released an unidentified compound (m/z 326) and no pyrophosphate (m/z 179). Thus, if present, ApppI content is below the detection level in K562 cells. As a control, a similar analysis of pamidronate-stimulated K562 extracts unequivocally revealed the presence of ApppI (supplemental Fig. 1).⁵

When studying the inhibitory effect of aminobisphosphonates, Monkkonnen et al. (28, 37) found that these compounds induced the intracellular accumulation of ApppI which presents proapoptotic properties on macrophages and osteoclasts. They also provided evidence that this nucleotidic metabolite arose in cells following farnesyl pyrophosphate synthase inhibition and IPP accumulation and a tRNA synthase activity was proposed to be responsible for IPP to ApppI conversion. The possibility that ApppI is somehow involved in the stimulatory effect of tumors on Vγ9Vδ2 lymphocytes was, however, never investigated before, and although many nucleotidic derivatives of phosphoantigens have been studied, ApppI has never been tested.

We show here that ApppI efficiently stimulates Vγ9Vδ2 T cells when added in PBMC cultures and is thus comparable with other phosphoantigens. However, ApppI has no intrinsic stimulatory activity on Vγ9Vδ2 T cells in the absence of APCs, requiring cleavage of the αβ phosphoester bond and IPP release for activity. This is apparently at odds with previous reports indicating that nucleotide addition had little effect on phosphoantigen activity. However, although it is not excluded that isopentenyl pyrophosphate derivatives display specific properties, in previous reports, the activity of nucleotidic Ags was evaluated in the presence of APCs (12, 14).

ApppI can be captured by professional or nonprofessional APCs and confers them a stimulatory activity. Strikingly, this is more efficient than pulsing with an equivalent molar concentration of IPP. Once captured by cells, the phosphoantigenic activity is stably expressed on the cell surface, and this display is strongly impaired when cells are kept at 4°C or when cytoskeletal components are altered by cytochalasin D. After exposition to ApppI, cells behave like spontaneously stimulatory tumors or aminobisphosphonate-treated cells, and their stimulatory activity is not only stable on the cell surface but also resistant to phosphatases which normally degrade pyrophosphorylated Ags (such as apyrase) or nucleotidic derivatives (such as NPP), indicating Ag modification or protection. Altogether, these data suggest the existence of a specific receptor and transport mechanism for nucleotide derivatives. Is tumor cell stimulatory activity due to pyrophosphates or nucleotidic Ags? HPLC analysis of cytosolic extracts reveals that nucleotidic Ags are present in stimulatory tumors. Their characterization is based on the fact that the stimulatory activity requires NPP addition for activation of Vγ9Vδ2 cells in the absence of APCs.

Another peak of activity was detected in the same biological extracts in the absence of apyrase or NPP treatment. As this was sensitive to apyrase, it is probably related to the presence of stimulatory pyrophosphate Ags including IPP. This is in agreement with the data of Gober et al. (26), although we could not undoubtedly identify IPP in our samples. This may be due to differences in the sample preparation or sensitivity of mass spectrometry equipment.

Thus, both nucleotidic and nonnucleotidic Ags may be responsible for tumor-stimulatory activity. ApppI ultimately requires rehydrolysis into IPP, and the latter might be the only biologically relevant antigenic form for Vγ9Vδ2 T cell stimulation. However, we propose that IPP to ApppI conversion has two biologically important consequences: 1) ApppI represents an inactive storage form of phosphoantigen that may bind to carrier proteins and may be protected from degradation. Thus, storage of Ags as nucleotidic derivatives might prolong their half-life and immunogenicity or allow their intracellular transportation. Moreover, the activity revealed by NPP treatment of cell extracts is higher than that of apyrase-sensitive fractions indicating that this antigenic material is not marginal; 2) even if ApppI does not present any direct stimulatory activity, it is more readily captured and displayed as a stimulatory Ag by cells. After sensitization with ApppI, tumors acquire a stimulatory activity that is resistant to phosphatases and in this respect are similar to naturally stimulating tumors.

On the basis of these observations, we propose that tumoral phosphoantigens are in fact transported and displayed on the cell surface of tumors in the form of nucleotidic derivatives and are stably associated with a presentation structure. Such presentation mechanism is supported by classical demonstrations of cell-cell contact requirement (34, 38) and the recent demonstration that soluble recombinant TCR constructs bind the tumor cell surface after exposition of cells to HDMAPP (24). Although the Ags used in the latter quoted experiments were not nucleotidic derivatives, it is quite possible that part of the Ag has been converted into a nucleotidic derivative and exposed in this form on the cell surface, either covalently bound or strongly associated to a presentation structure.

It remains that ApppI is not active directly and requires final conversion to IPP to activate T cells. However, this process could conceivably be achieved by cell surface-associated and tightly regulated NPPs (39). Thus, nucleotidic derivatives of phosphoantigens may represent important intermediates in a possible phosphoantigen-processing mechanism, although direct evidence of phosphoantigens on cell membranes is still actively searched for.

We thank F. L'Faqihi and V. Duplan-Eche (technical platform of cytometry of IFR150) for their technical assistance in cytometry.

The authors have no financial conflict of interest.