P2X7 pre-synaptic receptors in adult rat cerebrocortical nerve terminals: a role in ATP-induced glutamate release

Authors


Address correspondence and reprint requests to Guido Maura, Department of Experimental Medicine, Pharmacology and Toxicology Section, University of Genoa, Viale Cembrano 4, Genoa I-16148, Italy. E-mail: maura@pharmatox.unige.it

Abstract

Although growing evidence suggests that extracellular ATP might play roles in the control of astrocyte/neuron crosstalk in the CNS by acting on P2X7 receptors, it is still unclear whether neuronal functions can be attributed to P2X7 receptors. In the present paper, we investigate the location, pharmacological profile, and function of P2X7 receptors on cerebrocortical nerve terminals freshly prepared from adult rats, by measuring glutamate release and calcium accumulation. The preparation chosen (purified synaptosomes) ensures negligible contamination of non-neuronal cells and allows exposure of ‘nude’ release-regulating pre-synaptic receptors. To confirm the results obtained, we also carried out specific experiments on human embryonic kidney 293 cells which had been stably transfected with rat P2X7 receptors. Together, our findings suggest that (i) P2X7 receptors are present in a subpopulation of adult rat cerebrocortical nerve terminals; (ii) P2X7 receptors are localized on glutamatergic nerve terminals; (iii) P2X7 receptors play a significant role in ATP-evoked glutamate efflux, which involves Ca2+-dependent vesicular release; and (iv) the P2X7 receptor itself constitutes a significant Ca2+-independent mode of exit for glutamate.

Abbreviations used
BAPTA-AM

1,2-bis-(o-aminophenoxy)-ethane-N,N,-N′,N′-tetraacetic acid tetraacetoxy methyl ester

BBG

brilliant blue G

BzATP

2′-3′-O-(benzoylbenzoyl) ATP

CGS 15943

9-chloro-2-(2-furanyl)-[1,2,4]triazolo[1,5-c]quinazolin-5-amine

dl-TBOA

dl-threo-β-benzyloxyaspartate

DPCPX

1,3-dipropyl-8-cyclopentylxanthine

fura-2AM

fura-2 acetoxymethylester

GFAP

glial fibrillary acidic protein

HEK293

human embryonic kidney 293

oATP

oxidized ATP

PBS

phosphate-buffered saline

ROI

regions of interest

SCH 58261

7-(2-phenylethyl)-5-amino-2-(2-furyl)-pyrazolo-[4,3-e]-1,2,4-triazolo[1,5-c]pyrimidine

VDCC

voltage-dependent Ca2+ channel

ATP is a ubiquitous signaling molecule, which is released from glia and neurons. In the CNS of mammals, it may take part in neuron–glia crosstalk by co-operating with glutamate as a neurotransmitter/gliotransmitter (Araque et al. 1999; Fellin and Carmignoto 2004; Fields and Burnstock 2006; Burnstock 2007). It has been suggested that P2X7 receptors are targets for ATP in neuron–glia interplay. Indeed, P2X7 receptors have been reported to sustain calcium signaling and trigger ATP, glutamate, and GABA release from astrocytes (Ballerini et al. 1996; Wang et al. 2002; Duan et al. 2003; Nobile et al. 2003; Suadicani et al. 2006), and stimulate excitatory transmitter release from motor neurons or in the hippocampus (Deuchars et al. 2001; Sperlagh et al. 2002). In addition, pre-synaptic P2X7 receptors coupled with intracellular calcium signaling have been described in nerve terminals from various rat brain regions (Lundy et al. 2002; Miras-Portugal et al. 2003).

Immunohistochemical evidence of P2X7 receptors on central and peripheral nerve terminals has repeatedly been reported (Deuchars et al. 2001; Lundy et al. 2002; Sperlagh et al. 2002; Miras-Portugal et al. 2003; Atkinson et al. 2004; Hervas et al. 2005). However, as antibody-based evidence for pre-synaptic P2X7 receptors is not completely reliable (see Kukley et al. 2004; Sim et al. 2004), multiple experimental criteria need to be rigorously applied in order to attribute neuronal functions to P2X7 receptors (see Anderson and Nedergaard 2006). Furthermore, while most information on neuronal or glial P2X7 receptors comes from studies on primary neonatal cell cultures, there are important differences between neonatal and adult cells.

Calcium influx induced by P2X7 receptors inside nerve terminals of rat cerebral cortex (Lundy et al. 2002) and P2X7 immunoreactivity on glutamatergic terminals (Atkinson et al. 2004) have been described. Furthermore, in cerebrocortical cell cultures, it has been found that release-facilitatory pre-synaptic P2X7 receptors are located on GABAergic neurons (Wirkner et al. 2005). To investigate the possible roles of neuronal P2X7 receptors in adult cerebral cortex, we carried out experiments on cerebrocortical nerve terminals (purified synaptosomes, which ensure low contamination of non-neuronal cells and allow exposure of ‘nude’ pre-synaptic receptors) freshly prepared from adult rats and on human embryonic kidney 293 (HEK293) cells stably transfected with the P2X7 receptor; different techniques were used. Combining the pharmacological analysis of neurotransmitter release with a calcium imaging technique enabled us to locate, as confidently as possible, P2X7 receptors linked to calcium influx and glutamate efflux on rat cerebrocortical nerve terminals. The ATP-evoked glutamate efflux appeared to be related mainly to action upon P2X7 receptors; about 50% of glutamate efflux was dependent on calcium influx through the receptor and about 50% was calcium independent, probably through the receptor itself. The latter results were confirmed by experiments on tritium efflux from HEK293 cells pre-labeled with [3H]d-aspartate.

Materials and methods

Animals

Adult male rats (Sprague–Dawley 200–250 g) were housed at constant temperature (22 ± 1°C) and relative humidity (50%) under a regular light–dark schedule (lights on 7 am–7 pm). Food and water were freely available. Experimental procedures were approved by the Ethics Committee of the Pharmacology and Toxicology Section, Department of Experimental Medicine, in accordance with European legislation (European Communities Directive of 24 November 1986, 86/609/EEC). All efforts were made to minimize the number of animals used and their suffering.

Preparation of purified synaptosomes

After decapitation, the cerebral cortex was rapidly removed and placed in ice-cold medium, and purified synaptosomes were prepared as previously reported (Nakamura et al. 1993; Stigliani et al. 2006). Briefly, the tissue was homogenized in 10 volumes of 0.32 mol/L sucrose, buffered at pH 7.4 with Tris–HCl, using a glass-Teflon tissue grinder (clearance 0.25 mm). The homogenate was centrifuged (5 min, 1000 g at 4°C) to remove nuclei and debris, and the supernatant was gently stratified on a discontinuous Percoll gradient (2%, 6%, 10%, and 20% v/v in Tris-buffered sucrose) and centrifuged at 33 500 g for 5 min. The layer between 10% and 20% Percoll was collected and washed by centrifugation. For release experiments, synaptosomes were then suspended in standard medium with the following composition (mmol/L): NaCl 125, KCl 3, MgSO4 1.2, CaCl2 1.2, NaH2PO4 1.0, and NaHCO3 22 with glucose 10 (gassed with 95%O2–5%CO2, pH 7.4). For immunofluorescence confocal microscopy or [Ca2+]i measurement, synaptosomal pellets containing 2 mg proteins were resuspended in 1 mL HEPES medium (mmol/L: NaCl 128, KCl 2.4, MgSO4 1.2, KH2PO4 1.2, CaCl2 1.0, and HEPES 10 with glucose 10, pH 7.4); 15 μL were deposited onto 20-mm glass coverslips coated with poly-l-lysine or Cell-Tak (BD Biosciences, San Jose, CA, USA) and centrifuged at 1550 g for 3 min.

Synaptosome superfusion experiments

Synaptosomes were incubated (15 min at 37°C) with [3H]d-aspartate (0.03 μmol/L), transferred to parallel superfusion chambers at 37°C and superfused (0.5 mL/min) with standard medium aerated with 95% O2–5% CO2 (Raiteri et al. 1974). Briefly, after 33-min superfusion, superfusate fractions were collected in 3 min samples (from the first fraction, basal1, B1 to B5); after 38 min of superfusion, synaptosomes were exposed (120 s) to ATP, 2′-3′-O-(benzoylbenzoyl) ATP (BzATP), or high K+ (in a subset of experiments, KCl 15 mmol/L substituting an equimolar concentration of NaCl was used to depolarize synaptosomes). To evaluate the effect of oxidized ATP (oATP), synaptosomes were pre-incubated (60 min at 37°C) in medium supplemented with oATP before superfusion with standard medium (from which the antagonist was omitted) was started. The effect of lowering extracellular Mg2+ concentration was assessed in synaptosomes superfused with a medium containing 0.01 mmol/L MgSO4, starting 18 min before addition of the agonist. The effect of Zn2+ was evaluated on the BzATP or ATP responses in a medium containing 0.01 mmol/L MgSO4 (see Sperlagh et al. 2002), by adding ZnSO4 18 min before the agonist. Dependence of drug-evoked tritium release or endogenous glutamate release on external Ca2+ was assessed in synaptosomes superfused with Ca2+-free medium (supplemented with EGTA 0.5 mmol/L) starting 18 min before addition of the agonist; dependence on intracellular Ca2+ was assessed in synaptosomes pre-incubated (30 min) in the presence of the membrane-permeable Ca2+ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,-N′,N′-tetraacetic acid tetraacetoxy methyl ester (BAPTA-AM). The effect of brilliant blue G (BBG) or 9-chloro-2-(2-furanyl)-[1,2,4]triazolo[1,5-c]quinazolin-5-amine (CGS 15943), 7-(2-phenylethyl)-5-amino-2-(2-furyl)-pyrazolo-[4,3-e]-1,2,4-triazolo[1,5-c]pyrimidine (SCH 58261), 1,3-dipropyl-8-cyclopentylxanthine (DPCPX), or dl-threo-β-benzyloxyaspartate (dl-TBOA), niflumic acid, or carbenoxolone was evaluated by adding the drug 8 min before the agonist; voltage-dependent Ca2+ channel (VDCC) blockers were added 5 min before BzATP or high K+. At the end of superfusion, the radioactivity in synaptosomes and superfusate samples was determined by liquid scintillation counting. The amount of endogenous glutamate released in the fractions collected was measured by HPLC (Waters 715 Ultra Wisp, Milford, MA, USA) as previously described (Marcoli et al. 2003). The analytical method involved automatic pre-column derivatization with o-phthalaldehyde, followed by separation on a C18 reverse phase chromatography column (10 cm × 4.6 mm, 3 μm; Chrompack International, Middleburg, The Netherlands) and fluorimetric detection. Homoserine was used as an internal standard. The detection limit was 100 fmol/μL. Protein determinations were carried out according to Bradford (1976). The efflux of radioactivity in each fraction was calculated as a percentage of the total radioactivity present at the onset of the fraction considered (fractional release). The amount of endogenous glutamate released in the fractions was expressed as pmol/mg protein. The mean tritium fractional release or endogenous glutamate in B1 and B2 fractions was taken as the 100% control value for each chamber. Tritium and endogenous glutamate efflux in Bn fractions were evaluated as the percentage variation in tritium fractional release or in endogenous glutamate with respect to the corresponding control value. The drug (or depolarization)-evoked tritium or endogenous glutamate efflux was measured by subtracting the area under the curve of percentage variations in tritium fractional release or in endogenous glutamate release in appropriate control chambers from the area under the curve of the percentage variations in drug-treated chambers (or in chambers supplemented with high K+). In each experiment at least one chamber was used as a control for each condition and was superfused with standard medium or with medium appropriately modified. When possible, drugs were dissolved in distilled water or in physiological medium. Stock solutions of SCH 58261, CGS 15943, or niflumic acid were prepared in dimethylsulfoxide and diluted at least 1 : 1000 in physiological medium; stock BBG solutions were freshly prepared in ethanol and diluted 1 : 3000; dimethylsulfoxide diluted 1 : 1000 or ethanol diluted 1 : 3000 had no effect on tritium or endogenous glutamate release.

Immunofluorescence confocal microscopy

Synaptosomes were fixed with 2%p-formaldehyde, permeabilized with 0.05% Triton X-100 (5 min) and incubated for 60 min with primary antibodies diluted in phosphate-buffered saline (PBS) containing 3% albumin. The following antibodies were used: rabbit anti-glial fibrillary acidic protein (GFAP) (1 : 100), mouse anti-synaptophysin (1 : 5000), rabbit anti-synaptophysin (1 : 300), mouse anti-oligodendrocyte (clone RIP; 1 : 10 000), and mouse anti-integrin-αM (clone OX-42; 1 : 25). After washing with PBS, the preparations were incubated (60 min) with Alexa Fluor 488 or 594 (1 : 1000) labeled secondary antibodies in PBS containing 0.5% albumin. Images were collected by means of an MRC-1024ES laser scanning system (Bio-Rad, Hercules, CA, USA) attached to a Nikon Eclipse TE300 inverted microscope (plan-apochromatic oil immersion objective 60/1.4 numeric aperture, Nikon, Tokyo, Japan). Imaging parameters and excitation/emission filters (488/522 and 594/589 nm) were set by lasersharp 2000 software (Bio-Rad). Fluorescence collection was optimized according to the combination of chosen fluorochromes, and sequential acquisition was performed in order to avoid crosstalk between color channels.

Calcium imaging in single synaptic terminals

The intrasynaptosomal calcium concentration was measured by the fura-2 acetoxymethylester (fura-2AM) microfluorimetric technique on single glued synaptosomes (see Miras-Portugal et al. 1999; Gomez-Villafuertes et al. 2001; Gualix et al. 2003). Synaptosomes adhering to coverslips were loaded with 5 μmol/L fura-2AM (45 min at 37ºC) and then mounted in a microperfusion chamber on the stage of an inverted fluorescence microscope (Nikon TE200) equipped with a dual excitation fluorimetric Ca2+ imaging system (Hamamatsu, Milan, Italy) as described previously (Alloisio et al. 2006). Synaptosomes were excited alternately at 340 and 380 nm with a sampling rate of 0.15 Hz, the resultant emission being collected above at 510 nm by a cooled digital CCD camera (C-4880-91; Hamamatsu Photonics, Sunayama-Cho, Japan) and recorded by means of dedicated software (Aquacosmos; Hamamatsu Photonics). The changes in fluorescence ratio (F340/F380) were determined from selected regions of interest (ROI) covering a single synaptic terminal. Some experiments were also performed in macro ROI including 20–50 terminals to verify the signal quality of single terminals. The ratio F340/F380 was used to indicate the changes in [Ca2+]i. No calibration was performed in view of the many uncertainties related to the dissociation constant of fura-2 in the intracellular milieu (Karaki et al. 1997). However, in some experiments [Ca2+]i was calculated according to Grynkiewicz et al. (1985), by using a Kd of 230 nmol/L for the Ca2+/fura-2 complex. Drugs were superfused at the rate of about 2.5 mL/min with HEPES medium and imaged through a Nikon 100× lens. Each experiment was carried out on at least three independent synaptosomal preparations. A pulse of high K+ (40 mmol/L) was applied at the end of each experiment to verify the viability of synaptosomes.

Stable transfection of HEK293 cells

Cultures of the HEK293 cell line were maintained in Dulbecco’s Modified Eagle’s medium–F12 Ham supplemented with 10% fetal bovine serum and gentamycin/glutamine (5 mg/mL and 200 mmol/L, respectively). The plasmid containing the full-length rat P2X7-green fluorescent protein cDNA in pcDNA3 was kindly provided by Dr Francesco Di Virgilio (Department of Experimental and Diagnostic Medicine, University of Ferrara, Italy). One day before transfection, cells were replaced on plastic dishes (35 mm diameter) in antibiotic-free growth medium. The following day, 4 μg of plasmid DNA were used for HEK293 transfection using cationic liposomes (Lipofectamine 2000) according to the manufacturer’s instruction. About 16 h after transfection, the medium was substituted by one supplemented with 1.5 g/L of G418 sulfate in 100 mmol/L HEPES (Sigma-Aldrich, Milan, Italy).

[3H]d-aspartate efflux from HEK293 cells

The presence of excitatory amino acid transporter 3 (EAAT3) glutamate transporters on native HEK293 cells (Toki et al. 1998) was exploited in order to load these cells with [3H]d-aspartate. Briefly, HEK293 stably expressing rat green fluorescent protein-tagged P2X7 in C-termini and native HEK293 plated on poly-l-lysine-coated 12-well plates and grown to confluence were incubated (20 min at 37°C) in the presence of [3H]d-aspartate (0.06 μmol/L) in 1.5 mL HEPES (mmol/L: NaCl 135, KCl 2.4, MgSO4 1.2, KH2PO4 1.2, CaCl2 1.2, NaHCO3 5, and HEPES 10 with glucose 10, pH 7.4). Uptake specificity was assessed by carrying out parallel uptake procedures in Na+-free HEPES (in which equimolar choline replaced Na+) and in the presence of the glutamate transporter inhibitor dl-TBOA. After uptake, cells were washed six times with HEPES, supplemented with 5 mL of ice-cold water and rapidly frozen; cells were then sonicated, and the radioactivity remaining in each well was determined by liquid scintillation counting. When tritium efflux was evaluated, native HEK293 cells and HEK293 cells stably expressing the P2X7 receptor, loaded with [3H]d-aspartate (0.06 μmol/L in 5 mL HEPES; 30 min at 37°C), were transferred to parallel superfusion chambers at 37°C, stratified on Millipore filters (Millipore Corp., Bedford, MA, USA), and superfused (0.5 mL/min) with standard medium. After 31 min of superfusion, superfusate fractions were collected in 5 min samples (from B1 to Bn) until the end of the experiment. After 40 min of superfusion, cells were exposed to BzATP (5 min) and then reperfused with standard medium. When used, BBG, dl-TBOA, niflumic acid, or carbenoxolone was added 10 min before the agonist. oATP was pre-incubated 60 min before the experiment was started; cells were then superfused with standard medium. At the end of superfusion, the radioactivity of filters on which cells were layered and of superfusate samples was determined by liquid scintillation counting.

The efflux of radioactivity in each fraction was calculated as fractional release and the drug-evoked tritium efflux was measured as described for synaptosome superfusion experiments; in each experiment, two chambers superfused with standard medium were used as controls.

Calculation and statistics

Log concentration–response relationships were obtained by using a four-parameter logistic function fitting routine (Sigma Plot software, Jandel-Scientific, San Rafael, CA, USA). Mean ± SEM of the numbers of experiments (n) are indicated throughout. Significance of the difference was analyzed by anova followed by Student’s t-test, with statistical significance being taken at p < 0.05.

Materials

[3H]d-aspartate was from Amersham Radiochemical Centre (Buckinghamshire, UK); ω-agatoxin IVA, ATP, BAPTA-AM, BzATP, BBG, carbenoxolone, CGS 15943, G418 sulfate, nifedipine, niflumic acid, and oATP were from Sigma-Aldrich (Milan, Italy); fura-2AM was from (Sigma-Aldrich); DPCPX or dl-TBOA were from Tocris Cookson (Bristol, UK); ω-conotoxin GVIA and ω-conotoxin MVIIC were from Alomone Labs Ltd. (Jerusalem, Israel); Lipofectamine 2000 was from Gibco (Invitrogen, S. Giuliano Milanese, Milan, Italy); and poly-l-lysine or Cell-Tack were from BD Biosciences. The compound SCH 58261 was kindly donated by Schering Plough (Milan, Italy). Rabbit anti-GFAP and mouse anti-synaptophysin antibodies were obtained from Sigma Chemical Co. (St Louis, MO, USA). Rabbit anti-synaptophysin, mouse anti-oligodendrocytes (clone RIP), and mouse anti-integrin-αM (clone OX-42) monoclonal antibody were from Chemicon International (Temecula, CA, USA). Donkey anti-mouse Alexa Fluor 488-conjugated and chicken anti-rabbit Alexa Fluor 594-conjugated secondary antibodies were from Molecular Probes Europe (Leiden, The Netherlands).

Results

A subpopulation of rat cerebrocortical purified synaptosomes is endowed with bona fide P2X7 receptors

Purified cerebrocortical synaptosomes were labeled with anti-synaptophysin antibody, a marker of synaptic vesicles that identify synaptic terminals. The synaptosomes proved negative for GFAP, integrin-αM, and RIP, which are markers for astrocytes, microglia, and oligodendrocytes, respectively; this indicates negligible non-neuronal contamination of cerebrocortical nerve terminals (Fig. 1).

Figure 1.

 Confocal microscopy analysis. Immunofluorescence for synaptophysin (a, d, and g) and for the astrocyte marker GFAP (b), the microglia marker integrin-αM (e), or the oligodendrocyte marker RIP (h), and merge image showing negligible contamination of purified synaptosomes by non-neuronal fractions (c, f, and i).

Isolated cerebrocortical synaptosomes and microareas containing 20–50 synaptosomes were tested by means of the calcium imaging technique to investigate their ability to respond to the P2X7-preferential agonist BzATP (Fig. 2a). When synaptosomes were challenged with 100 μmol/L BzATP, an increase in intrasynaptosomal Ca2+ concentration, expressed as F340/F380 ratio, was observed in selected microareas and in single synaptosomes. Figure 2b shows typical BzATP-activated [Ca2+]i responses obtained from a single synaptic terminal (ROI 6) and from a microarea (ROI 20) of the synaptosomal pattern shown in Fig. 2a. The [Ca2+]i response to depolarization (40 mmol/L K+) was used at the end of each experiment to confirm the integrity of the synaptosomal preparation. The microareas, with a better signal–noise ratio, displayed similar [Ca2+]i responses to those of single synaptosomes, indicating that our measurement in single particles is an efficacious method of revealing the Ca2+ signal of a single synaptosome. In the continuous presence of 100 nmol/L BBG, a selective inhibitor of P2X7 receptors (Jiang et al. 2000), BzATP failed to elicit [Ca2+]i signals. This inhibitory effect was reversible, as shown in Fig. 2c (n = 56 determinations from four independent preparations). To quantify the ability of BzATP and 40 mmol/L K+ to increase the Ca2+ level of single synaptosomes, the maximal increases from the basal level in the F340/F380 ratio were evaluated; these are shown in the bar graph in Fig. 2d. The bar graph also shows that the BzATP-induced [Ca2+]i increase was eliminated after pre-incubation with the irreversible P2X7 antagonist, oATP (Murgia et al. 1993; 1 h pre-incubation with 100 μmol/L oATP; n = 67 determinations from four independent preparations). Figure 2e shows the percentage of single synaptosomes responding to 100 μmol/L BzATP (about 50%) and 40 mmol/L K+ (n = 231 determinations from 13 independent preparations). The basal [Ca2+]i level was calculated according to Grynkiewicz et al. (1985) and amounted to 208 ± 56 nmol/L (n = 34 determinations from three independent preparations).

Figure 2.

 Intracellular calcium measurements in single synaptosomes and microareas of synaptosomes. (a) Fluorescence image showing fura-2AM loaded synaptosomes. The regions of interest (ROI) are represented by numbered squares and circles. (b) Typical Ca2+ responses induced by 100 μmol/L BzATP and 40 mmol/L K+ obtained from a single synaptosome (ROI 6) and from a microarea of about 25 synaptosomes (ROI 20). Microareas and single synaptosomes displayed BzATP- and 40 mmol/L K+-induced [Ca2+]i responses that were very similar in kinetics and amplitude. Interestingly, this demonstrates that Ca2+ measurement, despite the bad signal–noise ratio, is possible at the level of single particles. The reactivity to 40 mmol/L K+ confirms the integrity of our synaptosomal preparation. Solid bars indicate the period during which substances were applied. (c) Typical Ca2+ responses induced by BzATP in the presence and, later, in the absence of 1 μmol/L BBG. Note the reversibility of the BBG effect (n = 56 determinations from four independent experiments). (d) The bar graph reports the [Ca2+]i rises above basal levels evoked by 100 μmol/L BzATP and 40 mmol/L K+ in single synaptosomes (n = 231 determinations from 13 independent experiments). The [Ca2+]i responses after 1 h pre-incubation with oATP are also shown. Values of [Ca2+]i increase are expressed as F340/F380 ratio. Data were mean ± SEM; *p < 0.01 with respect to BzATP alone. (e) Percentage of synaptosomes responding to 100 μmol/L BzATP and 40 mmol/L K+.

Activation of receptors exhibiting the P2X7 profile-evoked glutamate efflux from purified rat cerebrocortical synaptosomes

Glutamate release was studied by measuring tritium efflux from synaptosomes pre-labeled with [3H]d-aspartate. In order to rule out the possibility that tritium release was an artifact resulting from the use of [3H]d-aspartate, we monitored the release of endogenous glutamate; the behavior of endogenous glutamate proved to be indistinguishable from that of tritium after labeling with [3H]d-aspartate, thus justifying the use of [3H]d-aspartate. The basal fractional tritium outflow in the first two fractions collected from superfused purified rat cerebrocortical synaptosomes amounted to 0.23 ± 0.01%/min (n = 57); the basal outflow of endogenous glutamate in the first two fractions amounted to 106 ± 7.5 pmol/mg protein/min (n = 35).

Both ATP and BzATP increased tritium efflux from purified synaptosomes in a concentration-dependent manner, BzATP appearing more potent than ATP (Fig. 3b); the time course of ATP- and BzATP-induced tritium or endogenous glutamate efflux in a representative experiment is shown in Fig. 3a. As divalent cations (Zn2+ and Mg2+) differentially affect the agonist activation of different P2X receptor subtypes, we evaluated the influence of external Mg2+ and Zn2+ on the effects exerted by ATP or BzATP. The ATP- or BzATP-evoked tritium release from purified rat cerebrocortical synaptosomes was sensitive to extracellular Mg2+ concentration: for both ATP and BzATP, the concentration–response relationships were shifted to the left when medium containing 0.01 mmol/L Mg2+ was used (instead of 1.2 mmol/L Mg2+; Fig. 3); Zn2+ (15 μmol/L), which potentiates several P2X subtypes but inhibits the P2X7 receptor (Virginio et al. 1997), inhibited the effects of both ATP and BzATP on tritium or endogenous glutamate efflux (Fig. 4a–d).

Figure 3.

 (a) Time course of a representative experiment for BzATP- and ATP-induced release of [3H]d-aspartate and endogenous glutamate. Time course of tritium and endogenous glutamate efflux in the 3 min fractions (from B1 to B5) collected from a control chamber (superfused with standard medium; no drug was added) and a chamber supplemented with BzATP 100 μmol/L or ATP 1 mmol/L in a typical experiment is shown. The horizontal solid line indicates addition of the agonist to the superfusion medium. Tritium and endogenous glutamate contents in the fractions collected were measured as fractional release and as pmol/mg protein, respectively. The mean tritium fractional release or endogenous glutamate in B1 and B2 fractions was taken as the 100% control value for each chamber; the drug-evoked tritium or endogenous glutamate efflux was measured as described in the Materials and methods section. (b) Dependence on external Mg2+ of the BzATP- or ATP-evoked [3H]d-aspartate efflux from superfused cerebrocortical synaptosomes. Log concentration–response curves for BzATP or ATP (120 s) in evoking tritium efflux in standard medium (containing 1.2 mmol/L Mg2+) and in medium containing 0.01 mmol/L Mg2+ are shown; medium containing 0.01 mmol/L Mg2+ was added 18 min before agonists. Data were mean ± SEM values of 4–20 experiments performed in triplicate.

Figure 4.

 Antagonism by Zn2+ of the BzATP- or ATP-evoked [3H]d-aspartate or endogenous glutamate efflux from superfused cerebrocortical synaptosomes. Bars represent percentage increase in tritium (a and c) or endogenous glutamate (b and d) in the presence of the drugs at the concentrations indicated. BzATP (a and b) or ATP (c and d) was added for 120 s during superfusion; medium containing 0.01 mmol/L Mg2+ (supplemented or not with Zn2+ at the concentrations indicated) was added 18 min before agonists. Data were mean ± SEM of three to six independent experiments in triplicate. *p < 0.05 when compared with the effect of the agonist alone.

Pre-incubation with oATP (300 μmol/L) or the addition of BBG (100 nmol/L) greatly reduced the effects of BzATP (100 μmol/L) or ATP (1–3 mmol/L) on the release of tritium or endogenous glutamate from purified rat cerebrocortical synaptosomes (Fig. 5a–d).

Figure 5.

 Antagonism by oATP or brilliant blue G of the BzATP- or ATP-evoked [3H]d-aspartate or endogenous glutamate efflux from superfused cerebrocortical synaptosomes. Bars represent percentage increase in tritium (a and c) or endogenous glutamate (b and d) in the presence of the drugs at the concentrations indicated. BzATP (a and b) or ATP (c and d) was added for 120 s during superfusion; brilliant blue G was added 8 min before the agonist; the irreversible antagonist oATP was pre-incubated for 60 min before superfusion. Data were mean ± SEM of three to nine independent experiments in triplicate. *p < 0.05 when compared with the effect of the agonist alone.

The selective antagonists of adenosine A1 and A2A receptors, DPCPX and SCH 58261, and the non-selective A1, A2, and A3 receptor antagonist CGS 15943 (Klotz 2000) had no effect on the ATP-evoked efflux of tritium (Table 1) or endogenous glutamate (3 mmol/L ATP: +320 ± 25.2% and 3 mmol/L ATP + 1 μmol/L CGS 15943: +333 ± 23.5%; n = 3).

Table 1.   Effect of adenosine receptor antagonists on the ATP-evoked efflux of [3H]d-aspartate from superfused rat cerebrocortical synaptosomes
DrugsPercentage increase (n)
  1. The effect was measured as percentage increase in tritium efflux with respect to the control. ATP was added for 120 s; the antagonists were added 8 min before the agonist. Data were mean ± SEM of n (number in parentheses) experiments in triplicate. DPCPX, 1,3-dipropyl-8-cyclopentylxanthine; SCH 58261, 7-(2-phenylethyl)-5-amino-2-(2-furyl)-pyrazolo-[4,3-e]-1,2,4-triazolo[1,5-c]pyrimidine; 9-chloro-2-(2-furanyl)-[1,2,4]triazolo[1,5-c]quinazolin-5-amine.

ATP (3 mmol/L) 426.3 ± 22.9 (6)
ATP (3 mmol/L) + DPCPX (1 μmol/L) 400.5 ± 34.2 (3)
ATP (3 mmol/L) + SCH 58261 (1 μmol/L) 412.5 ± 38.0 (3)
ATP (3 mmol/L) + DPCPX (1 μmol/L) + SCH 58261 (1 μmol/L)433.7 ± 32.5 (3)
ATP (3 mmol/L) + CGS 15943 (1 μmol/L) 410.3 ± 35.1 (3)

Pre-incubation with oATP or the addition of BBG or adenosine receptor antagonists at the concentrations used did not affect the basal efflux of tritium or endogenous glutamate (data not shown).

Mode of exit of glutamate evoked by activation of receptors exhibiting the P2X7 profile

The BzATP (100 μmol/L)-evoked efflux of [3H]d-aspartate and endogenous glutamate from purified rat cerebrocortical synaptosomes were reduced by 42% and 55%, respectively, when extracellular Ca2+ was removed (Fig. 6a and b). The BzATP (100 μmol/L)-evoked efflux of tritium and endogenous glutamate in the absence of extracellular Ca2+ was greatly reduced by pre-incubation with oATP (300 μmol/L) or the addition of BBG (100 nmol/L; Fig. 6a and b). Chelating intracellular Ca2+ by means of BAPTA-AM (50 μM) approximately halved the BzATP (100 μmol/L)-evoked efflux of tritium or endogenous glutamate; the effects of extracellular Ca2+ removal and intracellular Ca2+ chelation were not additive (Fig. 6a and b). Blockade of VDCCs belonging to the L-type (nifedipine, 10 μmol/L), N-type (ω-conotoxin GVIA, 1 μmol/L), and P/Q-types (ω-agatoxin IVA, 0.1 μmol/L), or the wide-spectrum VDCC blocker ω-conotoxin MVIIC (1 μmol/L), did not affect the BzATP (100 μmol/L)-evoked efflux of tritium or endogenous glutamate (Fig. 6c and d). In a subset of experiments designed as a positive control, the blockade of N- or P/Q-type VDCCs proved to inhibit the depolarization-evoked tritium efflux (15 mmol/L K+: +152 ± 3.7%, n = 4; 15 mmol/L K+ + 1 μmol/L ω-conotoxin GVIA: +71 ± 10.8%, n = 3, p < 0.05 vs. K+; 15 mmol/L K+ + 0.1 μmol/L ω-agatoxin IVA: +85 ± 9.2%, n = 3, p < 0.05 vs. K+).

Figure 6.

 Modes of exit of glutamate evoked by P2X7 receptor activation. BzATP-evoked [3H]d-aspartate or endogenous glutamate efflux from superfused cerebrocortical synaptosomes: Ca2+-dependence (a and b), effects of the VDCC blockers nifedipine, ω-conotoxin GVIA, ω-agatoxin IVA, and ω-conotoxin MVIIC on the efflux in the presence of external Ca2+ (c and d), and effect of dl-TBOA, niflumic acid, or carbenoxolone on the Ca2+-independent efflux (e and f). Bars represent percentage increase in tritium (a, c, and e) or endogenous glutamate (b, d, and f) in the presence of the drugs at the concentrations indicated. BzATP was added for 120 s during superfusion; Ca2+-free EGTA (0.5 mmol/L)-containing medium was added 18 min before the agonist; BBG, dl-TBOA, niflumic acid, or carbenoxolone was added 8 min, and VDCC blockers 5 min, before the agonist. oATP was pre-incubated for 60 min before superfusion. BAPTA-AM was pre-incubated for 30 min before superfusion. Data were mean ± SEM of 3–14 independent experiments in triplicate. *p < 0.05 when compared with BzATP alone in standard medium and #p < 0.05 when compared with BzATP alone in Ca2+-free EGTA (0.5 mmol/L)-containing medium.

The blockade of glutamate transporters by the non-substrate inhibitor dl-TBOA (30 μmol/L; Shimamoto et al. 1998) did not affect the Ca2+-independent BzATP (100 μmol/L)-evoked efflux of endogenous glutamate or tritium (Fig. 6e and f). Carbenoxolone (10–50 μmol/L), a non-selective compound which behaves as an effective inhibitor of pannexin-1 currents (at 1–20 μmol/L) and of gap junction hemichannels by blocking connexin (at 10–500 μmol/L; Bruzzone et al. 2005; Pelegrin and Surprenant 2006) did not affect the Ca2+-independent BzATP (100 μmol/L)-evoked efflux of endogenous glutamate or tritium (Fig. 6e and f). Blockade of the anion channels by niflumic acid (300 μmol/L; Raiteri et al. 2001) did not affect the Ca2+-independent BzATP (100 μmol/L)-evoked efflux of endogenous glutamate (Fig. 6f).

Pre-incubation with BAPTA-AM or exposure to Ca2+-free medium or addition of the VDCC blockers, dl-TBOA, or carbenoxolone at the concentrations used did not affect the basal efflux of tritium or endogenous glutamate; niflumic acid 300 μmol/L did not affect basal endogenous glutamate efflux (data not shown).

P2X7-transfected HEK293 confirm that the P2X7 receptor constitutes a pathway of exit for [3H]d-aspartate

Both native HEK293 cells and HEK293 cells stably expressing the P2X7 receptor were proven to take up [3H]d-aspartate through glutamate transporters: [3H]d-aspartate uptake was almost totally blocked in Na+-free medium (−95%) or by 10 μmol/L of the non-transportable glutamate transporter blocker dl-TBOA (−90%; data not shown). The fractional basal [3H]d-aspartate outflow in the first two fractions collected from HEK293 cells amounted to 0.95 ± 0.12%/min (n = 18) and to 0.78 ± 0.12%/min (n = 4) in HEK293 cells stably expressing P2X7 receptors and in native HEK293 cells, respectively.

2′-3′-O-(benzoylbenzoyl) ATP increased [3H]d-aspartate efflux in HEK293 cells expressing the P2X7 receptor (Fig. 7), while it was ineffective in control native HEK293 cells (data not shown). The BzATP effect was antagonized by pre-incubation with oATP (300 μmol/L) or by BBG (0.1–1 μmol/L; Fig. 7). The BzATP-evoked [3H]d-aspartate efflux was unaffected in conditions in which any possible [3H]d-aspartate exit through the glutamate transporter or the anion channels was prevented by dl-TBOA (10 μmol/L) or niflumic acid (300 μmol/L), respectively (Fig. 7). Carbenoxolone at concentrations (10–50 μmol/L) inhibiting pannexin-1 or connexin hemichannels did not affect the BzATP (30 μmol/L)-evoked [3H]d-aspartate efflux (Fig. 7).

Figure 7.

 BzATP-evoked [3H]d-aspartate efflux from HEK293 cells stably expressing the P2X7 receptor. Antagonism of BzATP effect by pre-treatment with oATP or by BBG, and ineffectiveness of dl-TBOA, niflumic acid, or carbenoxolone on BzATP-evoked tritium efflux. Bars represent percentage increase in tritium efflux in the presence of the drugs at the concentrations indicated. The agonist was added for 5 min; BBG, dl-TBOA, niflumic acid, or carbenoxolone was added 10 min before the agonist. Cells were incubated with oATP for 60 min before the release experiment was started. Data were mean ± SEM of 3–18 independent experiments performed in duplicate. *p < 0.05 when compared with BzATP alone.

dl-Threo-β-benzyloxyaspartate, niflumic acid, or carbenoxolone or pre-incubation with oATP at the concentrations used did not per se significantly affect the basal [3H]d-aspartate outflow from transfected HEK293 cells (data not shown).

Discussion

Our findings suggest that (i) bona fide P2X7 receptors are present on a subpopulation of adult rat cerebrocortical nerve terminals; (ii) P2X7 receptors are localized on glutamatergic nerve terminals; (iii) Ca2+ entry through the P2X7 receptor is linked to vesicular glutamate release; and (iv) the P2X7 receptor itself constitutes a Ca2+-independent mode of exit for glutamate.

Indication of the presence of functional pre-synaptic P2X7 receptors was obtained by using a preparation that provided direct evidence of receptor localization on nerve terminals. Purified preparations of rat brain nerve terminals (synaptosomes) ensure very low contamination of non-neuronal cells (Nakamura et al. 1993; Lundy et al. 2002; see also Gualix et al. 2003). Rat cerebrocortical synaptosomal (and gliosomal) fractions had been characterized in our labs (Stigliani et al. 2006); here, we confirm that the synaptosomal fraction was contaminated to a negligible extent: purified synaptosomes, positive for the neuronal marker synaptophysin, were negative for the glial marker GFAP, the microglia marker integrin-αM or the oligodendrocyte marker RIP. Release monitoring from a superfused synaptosomal monolayer (Raiteri et al. 1974), by minimizing metabolism and removing any released compound prevents indirect effects, avoiding the receptor biophase and enabling ‘nude’ receptors to be exposed; in these experimental conditions, only targets located on glutamatergic nerve terminals (when monitoring glutamate release) are selectively acted upon, allowing a detailed pharmacological characterization of release-regulating pre-synaptic receptors.

Bona fide P2X7 receptors are present on subpopulation(s) of adult rat cerebrocortical nerve terminals

Our results extend previous Ca2+ imaging evidence (Lundy et al. 2002), and allow quantification of the subpopulation of rat cerebrocortical synaptosomes responding to functional testing for P2X7 receptors. The BzATP-induced [Ca2+]i increase appeared to be because of bona fide P2X7 receptor activation, as indicated by the absence of the response after pre-incubation with oATP, which is recognized to irreversibly block the P2X7 receptor (Murgia et al. 1993). Even though oATP pre-treatment can also repress P2X2 homomeric or P2X2/P2X3 heteromeric channels (see in Beigi et al. 2003), the involvement of P2X2 receptors in BzATP-evoked Ca2+ entry is unlikely, as purified rat cerebral cortex nerve terminals do not appear to express the P2X2 receptor subtype (see Lundy et al. 2002). On the other hand, although irreversible P2 receptor-unrelated effects seriously question the reliability of oATP as an unambiguous probe for P2X7 receptor function in immune cells (see Beigi et al. 2003 and the comment by Di Virgilio 2003), oATP blockade of BzATP-induced Ca2+ influx is considered to predominantly reflect antagonism of P2X7 receptors in most cell types (Beigi et al. 2003; Di Virgilio 2003); in our preparation, antagonism by 100 nmol/L BBG (Jiang et al. 2000) confirmed the involvement of P2X7 receptors. Irreversible oATP blockade, and antagonism by 100 nmol/L BBG, of BzATP-evoked immediate ionotropic response (Ca2+ influx) indicated bona fide P2X7 receptor responses in about 50% of the purified adult rat cerebrocortical nerve terminals.

P2X7 receptors are localized on glutamatergic nerve terminals

Evidence that activation of the P2X7 receptor is responsible for glutamate release from rat cerebrocortical synaptosomes emerges from the following experimental findings. First, the order of potency of the agonists, BzATP being more potent than ATP, was expected for the P2X7 receptor (see North and Surprenant 2000). Second, extracellular cations affected receptor activation, as expected for the P2X7 receptor (Di Virgilio 1995; Virginio et al. 1997); BzATP and ATP concentration–response curves shifted to the left when extracellular Mg2+ was decreased and extracellular Zn2+ inhibited the effects of BzATP and ATP. Third, pre-incubation with oATP, an irreversible P2X7 antagonist (Murgia et al. 1993), antagonized both BzATP and ATP. Fourth, BBG, a selective inhibitor of the rat P2X7 receptor in the nanomolar range (Jiang et al. 2000), antagonized both BzATP and ATP. The ineffectiveness of adenosine receptor antagonists excludes the possibility that adenosine A1, A2, or A3 receptors might be involved in the ATP effect as a result of ectonucleotidase activity (see Cunha et al. 1996). Although antagonism by oATP or BBG indicates that receptors exhibiting a P2X7 pharmacological profile play a major role in the glutamate-releasing effect of ATP on rat cerebrocortical glutamatergic terminals, the inability of these antagonists to completely suppress the agonist effect indicates that stimulation of non-P2X7 receptor types is also involved in the glutamate-releasing effect of ATP. Indeed, glutamatergic terminals in other rat brain areas have been found to be endowed with multiple P2X receptor subtypes (Gualix et al. 2003; Shigetomi and Kato 2004; Hervas et al. 2005; Rodrigues et al. 2005).

Ca2+ entry through the P2X7 receptor is linked to vesicular glutamate release

Receptors exhibiting the P2X7 profile appeared to activate glutamate vesicle exocytosis in rat cerebral cortex, the Ca2+-dependent glutamate release amounting to about 50% of total efflux. Interestingly, the labeling of actively recycling vesicles in motor neuron terminals has suggested the occurrence of P2X7-activated vesicle exocytosis (Deuchars et al. 2001). Extracellular Ca2+ removal or intracellular Ca2+ chelation by BAPTA-AM caused comparable, non-additive inhibition of glutamate efflux, indicating that extracellular Ca2+ entry was responsible for P2X7-evoked glutamate release, while the contribution of Ca2+ from intracellular stores appeared negligible.

Vesicular glutamate release in rat brain appears to be primarily dependent on Ca2+ influx through VDCCs of the P-type, N- and Q-types playing minor roles (Vazquez and Sanchez Prieto 1997). Although pre-synaptic P2X7 receptor activation might depolarize the membrane and open VDCCs, we found that ω-agatoxin IVA, ω-conotoxin GVIA, or nifedipine, and the wide-spectrum VDCC blocker ω-conotoxin MVIIC did not affect P2X7-evoked glutamate release. Therefore, P2X7 receptors appear to be the way through which Ca2+ enters cerebrocortical glutamatergic terminals, following receptor activation, and promotes glutamate release, as already found in midbrain synaptosomes (Miras-Portugal et al. 2003). Accordingly, in cerebrocortical nerve terminals, [Ca2+]i accumulation evoked by P2X7 receptors has been reported to be unrelated to VDCC opening (Lundy et al. 2002). In the hippocampus, Ca2+ entry through P2X receptors has been reported to facilitate neurotransmitter release, resulting in a rise in [Ca2+]i proximal to the active zone and triggering vesicle exocytosis; this indicates that pre-synaptic P2X receptors have potential roles as ‘exocytosis-linked Ca2+ channels’ (Khakh et al. 2003; Shigetomi and Kato 2004). We propose that pre-synaptic P2X7 receptors may play a role as action potential-independent Ca2+ channels linked to neurotransmitter exocytosis at the rat cerebrocortical glutamatergic terminals. This would imply receptor location close to the active zone, which remains to be demonstrated.

The P2X7 receptor itself constitutes a mode of exit for [3H]d-aspartate

About 50% of the BzATP-evoked glutamate efflux from rat cerebrocortical glutamatergic terminals was Ca2+-independent. The Ca2+-independent BzATP-evoked glutamate efflux involved activation of P2X7 receptors, as indicated by oATP or BBG antagonism. In-out functioning of glutamate transporters constitutes a Ca2+-independent mode of glutamate exit from nerve terminals; volume-regulated anion channels, which can also open in isotonic conditions, are an alternative Ca2+-independent mode of glutamate efflux (Raiteri et al. 2001). Gap junction hemichannels, which allow Ca2+-independent glutamate release from astrocytes (Ye et al. 2003), have also been reported to mediate glutamate release from neurons (Thompson et al. 2006). The blockade of glutamate transporters by dl-TBOA, of volume-regulated anion channels by niflumic acid or of hemichannels by carbenoxolone (50 μmol/L; see Bruzzone et al. 2005; Pelegrin and Surprenant 2006) did not affect Ca2+-independent neurotransmitter efflux, indicating that these modes of exit are not involved in P2X7-evoked glutamate efflux from nerve terminals.

The possibility remains that P2X7-evoked glutamate efflux might occur through the activated receptor. Indeed, we confirmed the exit of glutamate through P2X7 receptors by assessing the passage of [3H]d-aspartate, in HEK293 cells stably expressing the rat P2X7 receptor. HEK293 cells, which do not natively express P2X receptors, have been widely used for the functional characterization of recombinant P2X7 receptors (see Humphreys et al. 1998; Alloisio et al. 2006). After loading the cells with [3H]d-aspartate, by taking advantage of the presence of dl-TBOA-sensitive glutamate transporters, we assessed the ability of [3H]d-aspartate to exit the cells following P2X7 receptor activation, when other possible ways of exit (glutamate transporters, anion channels, or gap junction hemichannels) were blocked. [3H]d-aspartate efflux from stably transfected HEK293 cells, which is consistent with in-out [3H]d-aspartate transport through the activated P2X7 receptor, suggests that P2X7-evoked Ca2+-independent glutamate efflux from nerve terminals could occur through the receptor itself; a similar mode of glutamate exit from cultured astrocytes has been hypothesized (Duan et al. 2003). Activated P2X7 receptors switch from selective cation channels (allowing Ca2+ influx) to large pores (resulting from receptor arrangement or from accessory protein recruitment). Pannexin-1 proteins, which have been proposed as accessory proteins whose recruitment allows P2X7 receptors to form pores (Pelegrin and Surprenant 2006; Di Virgilio 2007) in non-neuronal cells, are highly expressed in neurons (Bruzzone et al. 2003; Söhl et al. 2005; Barbe et al. 2006). We investigated (in nerve terminals and in HEK293 cells, which endogenously express pannexin-1; Pelegrin and Surprenant 2006) whether or not recruitment of pannexin-1 as an accessory pore-forming mechanism was required in order to allow glutamate/[3H]d-aspartate exit following P2X7 receptor activation. The finding that carbenoxolone, a pannexin-1 inhibitor (at 10 μmol/L; see Bruzzone et al. 2005; Pelegrin and Surprenant 2006), had no effect on P2X7-evoked [3H]d-aspartate efflux in HEK293 cells or on P2X7-evoked [3H]d-aspartate or endogenous glutamate efflux from the nerve terminals indicates that pannexin-1 recruitment is not required.

Conclusions and functional implications

In conclusion, our findings suggest that pre-synaptic P2X7 receptors, which are present on adult rat cerebrocortical nerve terminals, are linked to Ca2+ influx and Ca2+-dependent exocytotic glutamate release, and allow glutamate efflux through the receptor itself. In our experimental conditions, recruitment of pannexin-1 does not appear to be required for glutamate efflux through the P2X7 receptor. As an additional point, we propose that the characteristics of [3H]d-aspartate movement(s) through the P2X7 channel/pore should be suitably investigated in the new and simple model of stably transfected HEK293 cells. In the CNS, resting extracellular ATP concentrations are in the low nanomolar range. Nevertheless, owing to astrocyte stimulation, cell lysis and neuronal activity (see Anderson and Nedergaard 2006; Fields and Burnstock 2006; Sperlagh et al. 2006; Burnstock 2007), ATP could reach local concentrations capable of activating low-affinity (P2X7) receptors in the nearest astrocytes (Guthrie et al. 1999) or nerve terminals (Fields and Burnstock 2006; Sperlagh et al. 2006; Burnstock 2007). By causing calcium overload and glutamate efflux, pre-synaptic P2X7 receptors on glutamatergic nerve terminals might play a pivotal role in the control of cerebrocortical glutamatergic transmission and in neuron-damaging effects (and excitotoxic damage facilitation) in CNS injury (Wang et al. 2004; Sperlagh et al. 2006).

Acknowledgements

This work was supported by an Italian MIUR Network Grant to MN and GM. The financial support of Mariani Foundation of Milan (Grant no. R-07-66) to MN and GM is gratefully acknowledged. The authors thank Maura Agate for her excellent assistance in preparing the manuscript.

Ancillary