Escherichia coli produces phosphoantigens activating human gamma delta T cells.

Human Vgamma9delta2 T lymphocytes are suggested to play an important role in the immune response to various microbial pathogens. In contrast to alphabeta T cells, gammadelta T lymphocytes recognize small, non-protein, phosphate-bearing antigens (phosphoantigens) in a major histocompatibility complex-independent manner. Four different phosphoantigens termed TUBag1 to TUBag4 with a common 3-formyl-1-butyl-pyrophosphate moiety and isopentenyl-pyrophosphate have been isolated and identified from mycobacteria. However, natural occurring gammadelta T cell ligands from other bacterial species were not characterized so far. Here, we describe the structural identification of the two compounds responsible for the gammadelta T cell-stimulating capacity of Escherichia coli as similar to the mycobacterial phosphoantigens 3-formyl-1-butyl-pyrophosphate and its M(r) 275 homologue TUBag2. In addition, E. coli phosphoantigens exert bioactivities on gammadelta T cells with similar potencies to the mycobacterial phosphoantigens at 5-15 nm concentration. Furthermore, our results clearly prove that the deoxyxylulose 5-phophate pathway (also referred to as Rohmer metabolic route of isoprenoid biosynthesis) is essential for the biosynthesis of the phosphoantigens in E. coli. Because this pathway is absent from human cells, it proves an ideal target for focusing efficiently the antimicrobial selectivity of human gammadelta T lymphocytes.

Human V␥9V␦2 T lymphocytes are considered to play an important role in the immune response to various microbial pathogens. In contrast to ␣␤ T cells, ␥␦ T lymphocytes recognize small, non-protein, phosphate-bearing antigens (phosphoantigens) in a major histocompatibility complex-independent manner. These phosphoantigens are produced by different Grampositive and Gram-negative bacteria as well as by eukaryote parasites (1). Mycobacterium tuberculosis produces four different phosphoantigens termed TUBag1 1 to TUBag4 wherein TUBag1 and TUBag2 are phosphatase-sensitive pyrophosphate monoesters, whereas TUBag3 and TUBag4 are nucleotide conjugates of TUBag1 (2). These compounds are active for stimulation of human ␥␦ cells at 5-15 nM concentration and have been found not only in M. tuberculosis but also in most other mycobacterial species. Although complete identification of their structure was recently achieved by elucidation of their common 3-formyl-1-butyl-pyrophosphate (3fbPP) moiety (3), isopentenyl pyrophosphate (IPP) had been characterized as another mycobacterial phosphoantigen from Mycobacterium smegmatis (4). IPP is a ubiquitous precursor for vitamins, steroids, and ubiquinones in all living organisms. However, in the case of Escherichia coli we found that the ␥␦ T cell-stimulating capacity of IPP did not correlate with its amount in bacterial extracts (5). A different set of non-phosphorylated microbial metabolites stimulating human ␥␦ T cells is composed of alkylamines, although their selective bioactivity is most frequently observed within millimolar concentration ranges (6).
Until recently, biosynthesis of IPP has been thought to rely solely upon the "classical" mevalonate pathway of isoprenoid biosynthesis. However, recent studies on the isoprenoid biosynthesis in bacteria, algae, and plants have led to the discovery of a second metabolic route, the so-called deoxyxylulose 5-phosphate (DOXP) or Rohmer pathway (7), which appears to be absent in all animals including man. Key metabolites of this pathway, e.g. deoxyxylulose 5-phosphate, 2-C-methylerythritol 4-phosphate, 4-diphosphocytidyl-2-C-methylerythritol or 2-Cmethylerythritol cyclopyrophosphate (8 -10) share a common carbon skeleton with the mycobacterial TUBag ligands (3). Furthermore, it was demonstrated that the presence of the Rohmer pathway in different bacterial species correlates with the ability of the metabolites to stimulate ␥␦ T cells (5). These observations support the concept that ␥␦ T cells discriminate between self and non-self through the recognition of specific biosynthetic routes and, therefore, contribute to the specific immune response to bacteria and parasites (11).
Although the presence of phosphoantigens was evidenced in Plasmodium falciparum (12), Francisella tularensis (13), Vis-* This work was supported by Interdisziplinä res Zentrum fü r klinische Forschung der Universität Wü rzburg Grant 01KS9603 (to J. F., S. E., and M.W.) and the Fonds der Chemischen Industrie (to M. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. cum album (14,15), and Brucella suis (16), few non-mycobacterial phosphoantigens have been fully characterized so far. In the current study we focused on isolating and characterizing ␥␦ T cell stimulating ligands from E. coli. We describe here the structural identification of these E. coli antigens as similar to the mycobacterial phosphoantigens. Furthermore, the relation of these bacterial antigens to the Rohmer metabolic pathway of isoprenoid biosynthesis is demonstrated by use of a selective inhibitor.
Chromatographic Procedures-The E. coli extracts were fractionated on a 20 ϫ 200-cm Q-Sepharose anion-exchange column (Amersham Biosciences, Inc.) eluted with a 50 -500 mM ammonium acetate gradient. The active fractions were collected and further purified on a 10 ϫ 17-cm diol column (Amersham Biosciences, Inc.) with an acetonitrile/ water gradient from 10 to 100% water. After removal of acetonitrile, the active fraction was desalted by gel chromatography using a Bio-Gel P-2 column (Bio-Rad) and 200 mM ammonium acetate as eluent. The final step was an anion-exchange separation on a HigLoad 26/10 Q-Sepharose high performance column (Amersham Biosciences, Inc.) with a 50 mM-1 M ammonium acetate gradient. The active fraction was concentrated by lyophilization. The antigen content of the resulting solution was greatly enhanced by this procedure. Each chromatographic step was monitored by testing the collected fractions in the ␥␦ proliferation assay.
The bioactive fraction isolated by the above procedure was then finally separated by HPAEC using a Dionex DX500 (Dionex, Sunnyvale, CA) system run with the Peaknet 5.1 chromatography software. A CD20 conductimeter was used downstream of an anion self-regenerating suppressor (ASRS-ultra, Dionex), which was set in external water mode. A diode array detector (Spectra System UV 6000LP Thermo Separation Products) scanning from 200 to 500 nm was coupled downstream online with the conductimeter. The HPAEC column was a Dionex Ion Pac AS11 column (4 ϫ 250 mm) equipped with a guard column (4 ϫ 50 mm). The 2 ml/min effluent rate from HPAEC was directed to a fraction collector. The eluent sequence comprised a binary solvent system involving 0.1 N NaOH (solvent A, prepared from sodium hydroxide 50% (Fisher) and 18 megaohm-cm deionized water (B) with a flow rate of 2 ml/min. The eluting sequence was as follows: 0 -2 min isocratic 5% A in 95% B; 2-9 min linear gradient 5-15% A in B; 9-15 min linear gradient 15-35% A in B followed by a final step of 15 min under the initial conditions.
Liquid Chromatography-MS-HPLC-MS was carried out as described (19). In brief, chromatography was performed on a Nucleodex ␤-OH HPLC column, 250 ϫ 4 mm, 5 m from Macherey-Nagel (Dü ren, Germany) with a binary gradient by an Applied Biosystems 140b pump and a rheodyne 7725 injection valve equipped with a 5-l sample loop. Solvent A was 5 mM ammonium acetate; solvent B was acetonitrile. Injection volume was 5 l, and flow rate was 0.7 ml/min. The following gradient was applied for separation: 0 min 80% B, 1 min 80% B, 11 min 50% B, 15 min 50% B. Using a post-column T-splitter, column eluate was split 1:27 between mass spectrometer and collection outlet.
HPLC-ESI-MS analysis was carried out on a triple stage quadrupole TSQ 7000 mass spectrometer with an ESI interface (Finnigan MAT, Bremen, Germany). Data acquisition and evaluation were conducted on a personal DECstation 500/33 (Digital Equipment, Unterföhring, Germany) and ICIS 8.1 software (Finnigan MAT). Nitrogen served as the sheath gas, and argon served as the collision gas. The following ESI parameters were used: negative mode, capillary voltage 4 kV, capillary temperature 200°C, sheath gas (N 2 ) 60 p.s.i., scan m/z 120 to 500 in Nanospray MS-The bioactive HPAEC fractions were collected and analyzed by nano-ESI-ITMS with an LCQ ion-trap mass spectrometer (20 -22) (Finnigan MAT, San Jose, CA). A commercial nanospray ESI source (23, 24) (The Protein Analysis Co., Odense, Denmark) was used with palladium and gold-coated glass capillaries (The Protein Analysis Co), which were positioned using a stereomicroscope at 1 mm from the entrance hole of the heated transfer capillary. The glass capillaries were filled with 5 l of sample. Nebulizer gas was not necessary in this spray mode. The nanospray needle voltage was set to 700 V. The heated transfer capillary was kept at a temperature of 150°C. Samples were analyzed in the negative ion mode, with maximum ion collection time set at 800 ms and two microscans summed per scan. The full scan mass range was m/z 60 -500. For MS 2 experiments (tandem mass spectrometry), ions were isolated with 2-Da windows, and the collision energy was 25 eV. The software used was Navigator 1.2 from ThermoQuest (San Jose, CA).
Cell Lines-␥␦ cell lines were obtained as follows. Lymphocytes were isolated from peripheral blood obtained from healthy donors on Ficoll-Paque PLUS (Amersham Biosciences, Inc.) and washed three times before use. Polyclonal V␥9V␦2 T cell lines were obtained by 15 days of culture of PBMC (10 6 /ml) in complete RPMI 1640 medium with Glutamax-I (Life Technologies, Inc.), supplemented with 10% AB human serum, 25 mM Hepes, 1 mM sodium pyruvate, 100 units/ml penicillin G, and 100 g/ml streptomycin in the presence of purified 3-formyl-1butylpyrophosphate (10 nM (3)) plus 100 units/ml interleukin-2 (Sanofi-Synthélabo). Expansion of V␥9V␦2 T cells was followed by cytometric analysis as described below, and only cell lines having more than 95% TCR V␦2-positive cells were used for subsequent experiments.
Lymphocyte Proliferation-The ␥␦ T cell proliferation assay was described previously (25). Briefly, cells from patients with more than 90% B cells in the mononuclear cell fraction in peripheral blood diagnosed as having a leukemic B cell lymphoma of low malignancy (B-CLL) were used as bystander cells. Cells from B cell lymphoma and mononuclear cells from healthy donors (PBMC) were isolated from heparinized peripheral blood by Ficoll-Hypaque density centrifugation (Amersham Biosciences, Inc.). Cells were washed three times with Hanks' buffer supplemented with 1% heat-inactivated fetal calf serum and cryopreserved in liquid nitrogen until use. 5 ϫ 10 4 PBMC were cultivated in 96-well round-bottom microtiter plates (Nunc, Wiesbaden, Germany) together with 1 ϫ 10 5 B-CLL in 100 l of RPMI 1640 medium/well supplemented with 2 mM glutamine, 100 g/ml penicillin-streptomycin (all purchased from Life Technologies, Inc.), 10% pooled human AB serum, and 100 units/ml interleukin 2. Cells incubated with medium alone served as negative control, and cells incubated with IPP (Sigma) served as positive control.
Phenotype Analysis by Flow Cytometry-Cells were harvested on day 3 (for measurement of CD25 expression) or on day 7 (for proliferation assay) and double-stained with fluorescein isothiocyanate-or phycoerythrin-conjugated monoclonal CD3, CD25, or TCR pan-␥␦ antibodies, respectively (Coulter Immunotech, Krefeld, Germany). 5 ϫ 10 3 cells from each sample were analyzed using a FACScan supported with Cellquest as acquisition and data analysis software (Becton Dickinson, Heidelberg, Germany). The lymphocytes were gated using forward/ sideward scatter gating.
Measurement of TNF-␣ Secretion-TNF-␣ released by activated ␥␦ T cells was measured by a bioassay using the TNF-␣-sensitive cells WEHI-13VAR (ATCC CRL-2148) as follows. 10 4 ␥␦ T cells/well were incubated with the 10 l of the tested fractions in 100 l of culture medium plus 25 units of interleukin 2/well for 24 h at 37°C. 50 l of the culture supernatant were then added to 50 l of WEHI cells plated at 3 ϫ 10 4 cells/well in culture medium plus actinomycin D (2 g/ml) and LiCl (40 mM) and incubated 20 h at 37°C. Viability of WEHI cell was then measured by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay; 50 l of MTT (Sigma; 2.5 mg/ml in phosphatebuffered saline)/well were added, and after 4 h of incubation at 37°C, 50 l of solubilization buffer (SDS 20%, dimethylformamide 66%, pH 4.7) were added, and absorbance (570 nm) was measured. For each sample tested in triplicate, TNF-␣ release was then deduced from a calibration curve obtained using human recombinant TNF-␣ (PeproTech, Inc, Rocky Hill, NJ).

E. coli Produces Specific
Antigen-activating Human ␥␦ T Cells-To assess the ␥␦ T cell stimulatory capacity of E. coli, bacteria were grown to late-log phase, disintegrated in a French press, ultrafiltered (Ͻ3 kDa), and the extracts were subjected to a ␥␦ T cell proliferation assay in presence of exogenous interleukin 2. E. coli low molecular weight preparations (LMWPs) induced activation (CD25 expression) as well as outgrowth of ␥␦ T cells (Fig. 1A). The antigenic activity could be totally abrogated by treatment with alkaline phosphatase (Fig.  1B). Therefore, the ␥␦ T cell-stimulating bioactivity from E. coli LMWPs is solely due to phosphoantigens. However, in contrast to the nucleotidic mycobacterial antigens TUBag3 and TUBag4, which are only susceptible to a combined treatment with nucleotide pyrophosphatase and alkaline phosphatase, the bacterial antigen(s) under study seems to bear a terminal phosphate moiety essential for antigenic activity such as IPP, 3fbPP, or synthetic bioactive alkylpyrophosphates (4) and phosphorylated carbohydrates (26). In E. coli ATCC 11303, the possibility of a mixture of nucleotidic conjugates together with phosphomonoesters was ruled out by sequential treatment of the LMWPs with alkaline phosphatase and nucleotide pyrophosphatase. As shown in Fig. 1B, the ␥␦ T cell stimulatory activity in E. coli ATCC 11303 is sensitive to alkaline phosphatase alone, demonstrating the sole presence of an antigen composed of a terminal (pyro) phosphate group as in IPP or 3fbPP.
To determine whether the ubiquitous IPP was responsible for the observed stimulation of ␥␦ T lymphocytes by E. coli, we selectively monitored this compound in LMWPs by LC-MS. For this aim, a high performance liquid chromatography was coupled online to an electrospray ionization tandem mass spectrometry (HPLC-ESI-MS 2 ) through a post-column T-splitter. By analyzing LMWPs with by this device, we attempted to correlate the detection of IPP (m/z 245 for [M-H] Ϫ in negative mode) with its bioactivity on ␥␦ T cells. Although LMWP bioactivity was well detected in these experiments ( Fig. 2A), we could not detect relevant IPP ion species in parallel (not shown). By comparing to an LMWP sample spiked with 2 mM synthetic IPP (Rt 10.05 min, Fig. 2B), the ␥␦ T cell-activating component appeared chromatographically close to, but separated, from IPP. Therefore, the bioactive LMWP component is distinct from IPP.
To selectively trace in the E. coli LMWPs any compound carrying a terminal phosphate, we applied a phosphate-specific precursor ion scan. Because we could not detect any unknown phosphate-bearing compounds co-eluting with the ␥␦ T cell stimulatory activity (not shown), the bioactive antigen concentration in LMWPs could be estimated at 10 nM or lower. This range of bioactivity is about 100 times superior to that of IPP in similar bioassays (data not shown) but appeared similar to these reported for the mycobacterial phosphoantigens (2, 3, 27, 28). It has been shown recently that 3fbPP (formerly referred to as TUBag1) has a molecular mass of 262 atomic mass units and yields a pyrophosphate fragment (m/z 159) upon tandem mass spectrometry (3). To probe whether a related compound was present in LMWPs, we conducted a selected reaction-monitoring experiment using this fragmentation pattern and detected a compound co-eluting with the ␥␦ T cell stimulatory activity (data not shown). Thus, the active compound in E. coli could be related to 3fbPP. However, an analogous selected reactionmonitoring experiment scanning for the nucleotidic derivative 3fb-PPP-uridine (formerly TUBag3, selected reaction monitoring m/z 567/261) achieved no signal (not shown). Together these observations confirmed the aforementioned sensitivity of the LMWP bioactivity to degradation by alkaline phosphatase alone.

E. coli LMWP Antigens for ␥␦ T Cells Are Identical to the Mycobacterial Phosphoantigens 3fbPP and 3-Formyl-1-pentyl-PP (3fpePP)-Because evidence for small amounts of 3fbPP
in LMWPs was obtained from the above MS experiments, further chromatographic separation and enrichment of the bioactive component from LMWPs was needed. We undertook this by HPAEC with online detection by conductimetry, UV, and final collection of the fractions (every 3 s) to enable bioactivitybased detection. This was done through measurement of the TNF-␣ released by a ␥␦ T cell line cultured 24 h with 1/10 of the final volume from the collected fractions concentrated 10-fold (Fig. 3A). Although the analyzed fraction appeared to comprise several well identified organic and inorganic anions such as acetate, chlorine, and a major peak of N-acetylmannosaminuronic-1-phosphate (14 mM in the sample), only a minor amount of remaining IPP was detected by conductimetry. However, the bioactivity for ␥␦ T cells was separated into two consecutive fractions that eluted before IPP, as already observed in the former HPLC (see Fig. 2). These fractions were pooled, concen-trated 50-fold in water, and analyzed by nanospray MS in the negative mode (Fig. 3B). Among several other ion species, only two signals, namely m/z 261 and m/z 275, corresponded to phosphorylated structures as revealed by phosphate losses in their subsequent MS 2 spectra. These signals could identify 3-formyl-1-butylpyrophosphate, formerly discovered in mycobacteria and TUBag2 (3), if the observed signals corresponded to their pseudomolecular species in negative mode. This assignment was then confirmed by MS 2 of these parent ions, which matched the reported spectra of pseudomolecular, fully protonated forms of the natural 3fbPP (M r 262, formerly referred to as TUBag1) and TUBag2 antigens (M r 276) (3). Such acidic forms readily fragment in negative mode when using collision energies of around 18 -25 eV. These fragmentation spectra harbor prominent pyrophosphate (m/z 159 and m/z 177)-and phosphate (m/z 97 and m/z 79)-derived signals in addition to a product ion corresponding to each respective anhydro derivative of the pseudomolecular ion (Fig. 4). The TUBag2 molecule had been detected in former studies about mycobacterial phosphoantigens. With the same spectroscopic properties as, but a higher (ϩ14) molecular mass than 3fbPP, it corresponds to its longer chain homologue. We tentatively position this -CH 2 -on C3-C4 from the 3fbPP structure and propose the structure of 3fpePP for this molecule. Based on their biological activity, molecular mass, and the results of the MS 2 experiments, the E. coli phosphoantigens present in the LMWP sample are 3fbPP and 3fpePP, identical to the mycobacterial TUBag1 and TUBag2. E. coli Pathway of Phosphoantigen Biosynthesis-We proposed earlier that the ␥␦ T cell-stimulating antigens of E. coli and of other bacteria might be derived from the DOXP pathway (5). Thus we tested several known metabolites of this biosynthetic route for their capacity to elicit a ␥␦ T cell response. DOXP, 2-C-methylerythritol 4-phosphate (2CMEP), 4-diphosphocytidyl-2-C-methylerythritol, 4-diphsophocytidyl-2-C-methylerythritol 5-phosphate, and 2-C-methylerythrtiol cyclopyrophosphate did not exert any ␥␦ T cell stimulatory activity in both activation and proliferation assays (Fig. 5A). Therefore a direct action of these upstream intermediates of the Rohmer pathway on ␥␦ T cell activation could be excluded.
However, our previous experiments show that supplementation of E. coli cultures with deoxyxylulose enhances their ␥␦ T cell-stimulating potency (5), suggesting that compounds derived from the DOXP pathway are responsible for ␥␦ T cell activation. To further probe the relation between the Rohmer pathway and production of phosphoantigens, we cultured E. coli in the presence of the dephosphorylated 2-C-methylerythritol (2CME) metabolite from the DOXP pathway. Bacterial growth rates and the ␥␦ T cell stimulatory capacity of LMWPs from 2CME and control cultures were compared. No differences in bacterial growth rate and cell numbers were observed between the cultures with or without 2CME. Unlikely to the experiments using deoxyxylulose as a supplement, the bioactivity of 2CME LMWPs was not enhanced when compared with control (data not shown). Thus dephosphorylated 2CME is neither an obligate requirement for E. coli cell growth nor for phosphoantigen production.
In the Rohmer pathway, 2CMEP is formed from DOXP by the action of an essential enzyme, the DOXP reductoisomerase, which can be selectively targeted by the phosphonate inhibitor fosmidomycin (29). Thus E. coli were grown in culture broth with and without this inhibitor, and both growth rates and phosphoantigen production were measured. Although fosmidomycin concentrations as high as 20 g/ml did not make any difference in cell growth rates, the LMWPs of E. coli cultured in the presence of fosmidomycin were totally devoid of activity on ␥␦ T cells (Fig. 5B). This effect was not due to inhibition of ␥␦ T cell proliferation by the antibiotic itself (data not shown).

DISCUSSION
␥␦ T cells are suggested to play a role as the first line of defense against various viral, bacterial, and parasitical pathogens. They recognize soluble phosphoantigens from different Gram-negative and Gram-positive bacteria as well as certain parasites (for review, see Refs. 30 and 31). To better understand the way of action of this still enigmatic T cell population, it is important to characterize these antigens. So far only five phosphoantigens, 3fbPP, TUBag2, 3fb-PPP-uridine, 3fb-PPPthymidine, and IPP had been identified as natural occurring ␥␦ T cell ligands (2)(3)(4)28). These compounds were isolated and identified from several mycobacterial species (32), and they are presumably common to the entire mycobacterial genus. Yet, IPP represents a ubiquitous metabolic precursor of isoprenoids and prenylated proteins.
In the present study, we have isolated the two compounds responsible for the ␥␦ T cell-stimulating capacity of E. coli and proved that these phosphoantigens are identical to the mycobacterial metabolite 3-formyl-1-butyl-PP and its M r 275 homologue TUBag2. By treating E. coli LMWPs with either alkaline phosphatase or nucleotide pyrophosphatase alone or with both enzymes sequentially, we demonstrated that no nucleotidic phosphoantigens such as the mycobacterial 3fb-PPP-uridine or 3fb-PPP-thymidine were present (Fig. 1B). Thus, the ␥␦ T cell-stimulating capacity of E. coli was entirely due to one or several compound(s) bearing a terminal phosphate, which therefore are phosphoantigens. Two lines of evidence demonstrated that these phosphoantigens do not comprise IPP. On the one hand, the ␥␦ T cell-activating compound(s) was chromatographically separated from IPP, and on the other hand, the E. coli-derived IPP was not detected in quantities sufficient to exert its bioactivity on ␥␦ T cells (Fig. 2), which is known in the micromolar range (4). This finding is in agreement with a previous quantification study showing that the IPP content of E. coli is not high enough to elicit a ␥␦ T cell response (5).
The bioactivities of the E. coli phosphoantigen(s) were estimated in the nanomolar range on the basis of the titration of bioactivity in partially purified E. coli extracts as compared with the phosphate detection threshold of HPLC-ESI-MS 2 . Therefore, these E. coli phosphoantigens exert bioactivities on ␥␦ T cells with similar potencies to the mycobacterial phosphoantigens (2). Furthermore, a selected reaction-monitoring experiment scanning for a typical fragmentation reaction of 3-formyl-1-butyl-pyrophosphate (precursor ion, m/z 261 pseudo molecular ion; daughter ion, pyrophosphate moiety) resulted in a signal co-eluting with the ␥␦ T cell-stimulating activity of the E. coli sample, indicating that the E. coli phosphoantigens could comprise a molecular isobar of 3fbPP.
The enrichment of E. coli phosphoantigens in a bioactive fraction from HPLC-separated LMWPs was undertaken by subsequent HPAEC monitored by bioassay. Here again, although a minute amount of IPP could be detected conductimetrically, this product was separated from the phosphoantigens, which eluted earlier from the column. The direct MS analysis of the pooled bioactive fractions by nanospray MS indicated the presence of both 3fbPP (M r 262) and its longer chain homologue TUBag2 (M r 276). These identifications were then confirmed by the complete MS 2 fragmentation map of these parent ions (Fig. 4). The lower mass compound was 3fbPP, whereas the second phosphoantigen was tentatively identified as 3fpePP. As indicated by the above enzyme assays, no nucleotidic analogue of phosphoantigens was present in these extracts, so E. coli phosphoantigens are only composed of 3fbPP and 3fpePP.
In the third part of this study, we studied the metabolic pathway involved in biosynthesis of these phosphoantigens by E. coli. Like mycobacteria, E. coli synthesize IPP and subsequent isoprenoids by the Rohmer pathway (33,34). In agreement with several other reports, we demonstrate here that this metabolic route (also referred to as DOXP pathway) is responsible for phosphoantigen biosynthesis in E. coli. Because DOXP has also been described as a precursor for the biosynthesis of thiamin and pyridoxol (35,36), the first specific metabolite of the Rohmer pathway is 2-C-methylerythritol 4-phosphate (CMEP, Fig. 6). Hence, we conducted E. coli feeding experiments using 2-C-methylerythritol (CME) as a supplement of the culture medium. Although bacterial culture supplemented with the dephosphorylated deoxyxylulose intermediate enhances the ability of E. coli LMWPs to stimulate ␥␦ T cells (5) (indicated by reaction 1 in Fig. 6), we find here that similar experiments with 2-C-methylerythritol do not cause such effects (data not shown; see reaction 2 in Fig. 6). These observations do not necessarily argue against phosphoantigen production through metabolites of the Rohmer pathway. Others have already reported extremely low incorporation rates of free 2CME as compared with deoxyxylulose in E. coli and in other species. This could be due to missing a specific kinase for 2CME, whereas another kinase able to phosphorylate deoxyxylulose seems to be present in bacteria (Ref. 10; see Fig. 6).
Because it was not possible to proof the involvement of 2CMEP in the biosynthesis of the ␥␦ T cell-stimulating antigen by direct supplementation of 2CME, we attempted to inhibit DOXP reductoisomerase instead. This enzyme catalyzes the formation of 2CMEP from DOXP, probably through involvement of an aldehydic intermediate. The antibiotic fosmidomycin inhibits this step of the DOXP pathway (8). Here we have found that addition of fosmidomycin to E. coli cultures completely abolishes ␥␦ T cell-stimulating activity (Fig. 5B). This clearly proved that formation of 2CMEP and, thus, the Rohmer pathway is essential for the biosynthesis of the phosphoantigens in E. coli. This conclusion, based on biochemical studies, is in perfect agreement with conclusions drawn by others from genetic lines of evidence. In recently published experiments, it is demonstrated that disruption of dxr, a gene coding for DOXP synthase, and of gcpE, a gene of the DOXP pathway with unknown function, abrogates the ability of E. coli extracts to stimulate ␥␦ T cell proliferation (37).
The data presented here allow us to state that the E. coli phosphoantigens 3fbPP and 3fpePP are either authentic me-tabolites of the Rohmer pathway to IPP downstream of 2CMEP or that their formation branches somewhere from this route. Because the bioactive 3fbPP and 3fpePP found in E. coli were already characterized as mycobacterial phosphoantigens, it seems very likely that they were formed by the same way. However, the nucleotide-containing antigens of mycobacteria 3fb-PPP-thymidine and 3fb-PPP-uridine may most likely arise from a side reaction, as no corresponding compounds were found in E. coli so far. Such a side reaction is expected to bind 3fbPP to a nucleotide moiety, most likely as an energetically favorable carrier. Interestingly enough, the E. coli gene ygbB, an ortholog of the microbial gene ygbP involved in terpenoid biosynthesis, was recently characterized as coding for an enzyme of the DOXP pathway that hydrolyzes the nucleotidic conjugate 4-diphosphocytidyl-2C-methyl-D-erythritol-2-phosphate into 2C-methyl-D-erythritol-2,4-cyclodiphosphate and CMP (38). In this regard, it is highly conceivable that the mycobacterial phosphoantigens 3fb-PPP-thymidine and 3fb-PPP-uridine arise from related enzymatic activities encoded by mycobacterial orthologues of some ygb genes missing from the E. coli genome.
A common feature of many microorganisms recognized by ␥␦ T cells is their ability to synthesize isoprenoids via the DOXP pathway. Indeed, this pathway is absent from human or pri-FIG. 6. Biosynthesis of E. coli phosphoantigens. Biosynthesis of isopentenyl pyrophosphate, thiamin, and pyridoxal involve metabolites from the Rohmer pathway that are structurally related to 3fbPP and 3fpePP (tentative structure for this latter). This metabolic route is absent from animal cells but present in bacteria, green algae, higher plants, and apicoplasts from Plasmodium spp parasites. mate cells, which all share the TCR V␥9V␦2-specific reactivity to phosphoantigens. Therefore, the Rohmer metabolic route proves an ideal target for focusing efficiently the antimicrobial selectivity of major histocompatibility complex-unrestricted human T lymphocytes.