Search tips
Search criteria 


Logo of wtpaEurope PMCEurope PMC Funders GroupSubmit a Manuscript
J Neurosci. Author manuscript; available in PMC 2007 November 22.
Published in final edited form as:
PMCID: PMC2092412

Cysteine substitution mutants give structural insight and identify ATP binding and activation sites at P2X receptors


P2X receptors for extracellular ATP are a distinct family of ligand gated cation channels involved in physiological processes ranging from synaptic transmission to muscle contraction. Common ATP binding motifs are absent from P2X receptors and the extent of the agonist binding site is unclear. We used cysteine scanning mutagenesis, radiolabelled 2-azido ATP binding, and methanethiosulfonate (MTS) compounds, to identify amino acid residues involved in ATP binding and gating of the human P2X1 receptor. The pattern of MTSEA-biotinylation was also used to determine the accessibility of substituted cysteine residues and whether this changed on addition of ATP. Analysis of cysteine substituted mutants of the last 44 amino acid residues (S286-I329) in the extracellular loop before the second transmembrane segment showed that N290, F291, R292 and K309 mutants had reduced ATP potency and 2-azido ATP binding. MTS reagents produced further shifts in ATP potency at these residues suggesting that they are directly involved in ATP binding; the effects were dependent on the charge of the MTS reagent at K309C, one explanation for this is that K309 interacts directly with the negatively charged phosphate of ATP. The remainder of the cysteine substitutions had little or no effect on ATP potency. However at the mutants D316C, G321C, A323C, and I328C MTS reagents did not change ATP potency but modified agonist evoked responses suggesting that this region may contribute to the gating of the channel.

Keywords: ATP, P2X receptor, mutagenesis, binding, ion channels


P2X receptors for ATP are ligand gated cation channels and form from the homo- or heterotrimeric assembly of seven different receptor subunits (P2X1-7) to give a range of phenotypes (North, 2002). They are expressed widely in the nervous system and contribute to gut peristalsis (Ren et al., 2003), neuronal hypoxia reflexes (Rong et al., 2003), mechanotransduction (Cockayne et al., 2000), and pain sensation (Cockayne et al., 2000; Souslova et al., 2000; Tsuda et al., 2003; Chessell et al., 2005). P2X receptors do not contain common consensus ATP binding sequences e.g. the Walker motif (Walker et al., 1982) and the structure and organisation of the binding site is unclear. A model incorporating four co-ordinates for ATP action at the P2X1 receptor; with positively charged lysines K68 and K309 co-ordinating binding of the phosphate tail of ATP, and the adenine ring sandwiched between the domains incorporating phenylalanine (F) residues F185, T186 and N290, F291, R292 with R292 co-ordinating the binding of the α-phosphate has been proposed (Vial et al., 2004). This model is supported by mutagenesis data from P2X2 (Jiang et al., 2000; Clyne et al., 2002) and P2X7 receptors (Worthington et al., 2002). In the model the residues associated with ATP binding are at least 18-20 amino acid residues from the TM2 region that forms part of the ion conducting transmembrane pore of the channel. The model predicts that the region between K309 and the second transmembrane segment contributes to the transduction of agonist binding to channel gating. However to substantiate models of the receptor it is also important to consider the contribution of adjacent non-conserved residues. The variation in agonist potency and other pharmacological properties between P2X receptor subtypes indicates that non-conserved amino acid residues contribute to drug action at the receptor and may lead to subunit specific localised changes in the agonist binding environment or channel properties. To address further the role of N290, F291, R292 and K309, adjacent non-conserved amino acid residues and the contribution of the 20 amino acid region before the second transmembrane segment we have undertaken substituted cysteine scanning mutagenesis of the region S286-I329 of the human P2X1 receptor. This approach has been used to study the structure of a range of ion channels and proteins (Akabas et al., 1992) including P2X receptors (Rassendren et al., 1997; Egan et al., 1998; Jiang et al., 2000; Haines et al., 2001; Jiang et al., 2001) and the ATP binding site of Kir6.2 (Trapp et al., 2003). Analysis of effects of P2X1 receptor substituted cysteine mutants on the time-course, ATP potency, binding of radiolabelled 2-azido ATP, regulation by charged methanethiosulfonate (MTS) compounds (Akabas et al., 1992) and determination of the accessibility of residues in a MTSEA-biotinylation assay has allowed us to identify amino acid residues involved in ATP binding, channel properties and information on the secondary structure of this section of the extracellular segment.

Materials and Methods

Site-directed Mutagenesis

Cysteine point mutations for residues S286 – I329 were introduced via the QuikChange™ mutagenesis kit (Stratagene) using a human P2X1 receptor plasmid as the template as described previously (Ennion et al., 2000). Production of the correct mutations and absence of coding errors in the P2X1 mutant constructs was verified by DNA sequencing (Automated ABI Sequencing Service, University of Leicester).

Expression in Xenopus laevis oocytes

Wild type and mutant constructs were transcribed to produce sense strand cRNA (mMessage mMachine™, Ambion, Texas, U.S.A.) as described previously (Ennion et al., 2000). Manually defoliculated stage V Xenopus laevis oocytes were injected with 50 nl (50 ng) of cRNA using an Inject+Matic microinjector (J.Alejandro Gaby, Genéva, Switzerland) and stored at 18 °C in ND96 buffer (96 mM NaCl, 2 mM KCl, 1.8 mM CaCl2, 1 mM MgCl2, 5 mM sodium pyruvate, 5 mM HEPES, pH 7.6). Media was changed daily prior to recording 3-7 days later.

Electrophysiological recordings

Two-electrode voltage clamp recordings (at a holding potential of −60 mV) were carried out on cRNA injected oocytes using a GeneClamp 500B amplifier with a Digidata 1322 analog-to-digital converter and pClamp 8.2 acquisition software (Axon Instruments, USA) as previously described (Ennion et al., 2000). Native oocyte calcium activated chloride currents in response to P2X receptor stimulation were reduced by replacing 1.8 mM CaCl2 with 1.8 mM BaCl2 in the ND96 bath solution. ATP (Mg salt, Sigma) was applied via a U-tube perfusion system. Due to the large size of oocytes it is difficult to apply solutions rapidly. For desensitising responses this results in the true time-course of the response being underestimated due to the relatively slow solution exchange and activation of responses. Experiments on non-desensitising P2X2 receptors expressed in oocytes indicate that 10-90% solution exchange takes ~ 300-500 ms and means the rise times of P2X1 currents (~100 ms for WT) only give an indication of the speed of activation. The slower decay time of ~1s for P2X1 WT currents is less likely to be affected by solution exchange and can be used to discriminate mutants with a change in time-course. ATP (0.01 μM to 10 mM) was applied at 5 minute intervals, using this regime reproducible ATP evoked response were recorded. Individual concentration response curves were fitted with the Hill equation: Y = [(X)H × M] / [(X)H + (EC50)H] where Y is response, X is agonist concentration, H is the Hill coefficient, M is maximum response, and EC50 is the concentration of agonist evoking 50% of the maximum response. pEC50 is the −log10 of the EC50 value. Mutants that had considerably shifted ATP potency were tested with ATP concentrations up to 10 mM. For the calculation of EC50s individual concentration response curves were generated for each experiment and statistical analysis carried out on the pEC50 data generated. In the figures concentration response curves are fitted to the mean normalised data.

Characterisation of the effects of methanethiosulfonate compounds

To study the effect of methanethiosulfonate (MTS) compounds on ATP activation at wild type and cysteine mutants, ATP (~ EC90-95 concentration) was applied and either (2-aminoethyl)methanethiosulfonate hydrobromide (MTSEA) or sodium (2-sulfonatoethyl)methanethiosulfonate (MTSES) (Toronto Research Chemicals, Toronto, Canada) were bath-perfused (for at least 5 minutes; the recovery time required between application to see reproducible responses) prior to co-application with ATP via the U-tube. This procedure should allow MTS reagents to bind to free cyteine residues in either the non activated (absence of ATP) or ligand bound activated state (presence of ATP). MTS reagents (1 mM) were made in ND96 solution immediately prior to use. The effect of MTS reagents on the concentration responses to ATP were investigated by application of ATP in the presence of MTS reagents, or following incubating oocytes overnight in MTS reagents, and the following day ATP concentrations were applied via the U-tube with ND96 bathing solution (no MTS reagents present) the same results were found with either method.

2-AzidoATP and P2X1 receptors; patch clamp recordings from HEK293 cells expressing P2X1 receptors

Human Embryonic Kidney 293 (HEK-293) cells and HEK293 cells subcloned after transfection with the human wild-type P2X1 receptor (P2X1cl-1 cells), were maintained in minimal essential medium with Earle's Salts (with GlutaMAX™ I) supplemented with 10 % fetal bovine serum, 1 % non-essential amino acid (Invitrogen, U.K.) at 37 °C in a humidified atmosphere of 5 % CO2 and 95 % air as described previously (Vial et al., 2004). HEK 293 cells were transiently transfected with mutant P2X1 receptors using LipofectAMINE™2000 Reagent (Invitrogen, UK)/ Opti-MEM (Invitrogen). For patch clamp studies, as a control to identify cells that were efficiently transfected, the pEGFP-N1 (Clontech) vector (10% of total plasmid transfected) was co-transfected with P2X1 plasmids and recordings were made only from fluorescent cells expressing the pEGFP. After 24 h, the cells were plated onto glass coverslips and kept in culture for a maximum of 3 days for use in patch clamp experiments. HEK293 cells expressing WT or mutant P2X1 receptors were incubated in the presence of 1 μM 2-AzidoATP (Affinity Labelling Technologies, Lexington, USA). They were then placed 7.5 cm from a uv light source (Uvitech LF206.MS, Flowgen, Ashby Park, Leicestershire, U.K.) on the 312 nM setting and illuminated three times for 2 minutes with gentle mixing by swirling between exposures. Any effects of uv illumination in the presence of 2-AzidoATP on subsequent ATP evoked P2X1 receptor mediated currents were determined in whole-cell patch-clamp experiments as described previously (Vial & Evans, 2002). In preliminary studies we showed that the UV irradiation protocol, in the presence or absence of ATP (1 μM), had no effect on the peak amplitude of ATP (1 μM) evoked P2X1 receptor currents.

32P 2-AzidoATP radiolabelling of the P2X1 receptor

To assess the effects of mutations on agonist binding we used a 32P 2-AzidoATP binding assay. We used this for WT, F289C, a mutant with no change in ATP potency, and the five mutants with the greatest decrease in ATP potency (G288C, N290C, F291C, R292C and K309C). Five oocytes injected with wild type or mutant receptor cRNA were incubated in the presence of 1 μM 32P 2-AzidoATP (Affinity Labelling Technologies, Lexington, USA). They were then placed 7.5 cm from a uv light source (Uvitech LF206.MS, Flowgen, Ashby Park, Leicestershire, U.K.) on the 312 nM setting and illuminated three times for 2 minutes with gentle mixing by swirling between exposures. Oocytes were washed in ice cold PBS and lysed in 100 μl Buffer H (100 mM NaCl, 20 mM Tris·Cl pH 7.4, 1 % Triton X100 and 10 μl/ml protease inhibitor cocktail (Sigma P8340)) and centrifuged for 2 minutes at 16,000 g. Supernatant (85 μl) was removed and 0.9 μg anti-P2X1 antibody (Alomone, Jerusalem, Israel) added and incubated for an hour on ice (it is unlikely that 2-azido ATP binding to the extracellular segment interferes with anti P2X1 receptor binding as the epitope for this antibody is the final 16 amino acid residues of the intracellular carboxy terminus of the receptor). 30 μl of Protein A sepharose (Amersham Biosciences) (40mg/ml) was added and incubated overnight with rolling at 4°C. The sepharose:Ab:protein complex was centrifuged for 2 minutes and the beads washed four times with Buffer H. After the final wash the sepharose pellet was resuspended in SDS sample buffer and heated to 95 °C for 5 minutes. Eluted P2X1 receptors were loaded onto a 10 % SDS PAGE gel and run at 120 V. Following electrophoresis the gel was dried overnight and exposed to Hyperfilm™ film (Amersham Biosciences, Buckinghamshire, U.K.) with an intensifying screen at −80 °C for a week after which the film was developed. Densitometry analysis was carried out on the resultant films with data corrected by background subtraction calculated for each lane and expressed as percentage of WT. We have assumed that the surface expression of WT and the mutants used in the binding assay are equivalent as (i) there were no differences in the peak amplitude of ATP evoked currents, (ii) no differences in levels of total protein expression (e.g. Figure 2).

Figure 2
Effect of P2X1 cysteine mutants on 32P 2-azido ATP binding

Western Blotting

MTSEA biotin was used to probe the various cysteine mutants to map the extracellular loop region S286-I329 of the P2X1 receptor. MTSEA biotin reacts covalently with sulfhydryl groups provided by accesible cysteine residues. The wild type P2X1 receptor contains no ‘free’ cysteine residues (Ennion and Evans, 2002) and therefore an introduced cysteine would be free to bind MTS compounds if accessible.

Six oocytes injected with wild type or mutant receptor cRNA were preincubated at room temperature for 30 minutes in 0.5ml ND96 solution containing either 15u/ml of Apyrase (Type III, Sigma A7646) or 1mM ATP. After the preincubation period a further 0.5ml of ND96 was added containing either 15u/ml of Apyrase and 2mM MTSEA biotin or 1mM ATP and 2mM MTSEA biotin. Oocytes were incubated for a further 30 minutes after which they were washed twice in ND96 and lysed in Buffer H at 20μl/oocyte as previously described. Lysates were centrifuged at 16,000g for 5 minutes and the cleared lysate removed from the cell debris pellet. 10μl of cleared lysate (1/12th sample) was taken for total P2X1 protein analysis and the remainder was mixed with 30μl of 4% streptavidin agarose beads (Sigma S1638) and rolled overnight at 4°C. Beads were washed four times with Buffer H after which 20 μl of SDS PAGE sample buffer was added to the beads and heated at 95 μC for 5 minutes prior to loading on a 10 % SDS PAGE gel. Proteins were transferred to nitrocellulose, probed with P2X1 primary antibody (1:1000, Alomone, Jerusalem, Israel), secondary goat anti-rabbit antibody (Sigma A6154, 1:2,500 dilution), developed using ECL Plus (Amersham) and exposed to Hyperfilm™ film (Amersham Biosciences, Buckinghamshire, U.K.). Densitometry analysis was carried on scanned images of the resultant films (ImageJ, Analysis of P2X1 wild type and mutant MTSEA biotinylated protein levels and the corresponding total P2X1 protein band were determined (corrected for background gel levels). The level of biotinylation was calculated for apyrase and ATP treatment and corrected for the total P2X1 receptor. For WT this was 0.58 ± 0.09 % and reflects not only the level of biotinylation but also the efficiency of the streptavidin purification steps. A threshold value of biotinylation was set as three times the WT level of biotinylation and data are shown as fold biotinylation relative to this threshold. The level of biotinylation in the presence of ATP was also expressed as a percentage of the biotinylation in the presence of apyrase in paired samples and any effects were determined using a Student's paired two tailed t-test.


All data are shown as mean ± standard error of the mean. Significant differences between WT and mutant current amplitudes, time-courses, agonist potency and MTS reagent modification were calculated by one way analysis of variance followed by Dunnett's test for comparisons of individual mutants against control using the SPSS 12.0 for Windows package. The significance of any changes in ATP potency by MTS compounds, and the effects of ATP on MTSEA-biotinylation at particular mutants were determined with Student's T-test. n corresponds to the number of oocytes tested for electrophysiological data and number of experiments for biochemical studies.


The effect of cysteine scanning mutagenesis on human P2X1 receptor ATP mediated responses

To characterise the role of amino acid residues S286-I329 on ATP action at the P2X1 receptor we generated individual cysteine mutants and determined their effect on ATP potency (Figure 1). Wild type human P2X1 receptors responded to ATP in a concentration dependant manner (EC50 ~ 0.9 μM). Oocyte recordings provide only an approximation of the true EC50 due to the speed of solution exchanges and the non-steady state nature of the desensitising P2X1 receptor current. Cysteine substitution was well tolerated and functional responses were recorded from all of the mutant channels. Cysteine substitution at positions R295 and R305 produced small currents to ATP (10 mM) of 16 ± 4 and 94 ± 36 nA respectively. R295C and R305C were poorly expressed at the cell surface and they were unaffected by reducing agents (data not shown) suggesting the small amplitude of responses did not result from the formation of disulphide bonds, and these mutants were not studied further. The point mutation R295A however had no effect on peak current amplitude or ATP potency (pEC50 6.3 ± 0.1) and we have previously shown that R305A has normal ATP potency (Ennion et al., 2000). The majority (33/44) of cysteine substitution mutants showed no significant change in ATP potency (Table 1). A modest 2-3 fold increase in ATP potency was seen for 2 mutants, L318C and G321C (EC50s ~ 0.3 μM), previously we have shown that the G321A mutation has no effect on ATP potency (Digby et al., 2005). Approximately 5 fold decreases in ATP potency were recorded for F297C (EC50 4.8 μM) and F311C (EC50 6.2 μM), with ~ 10 and ~ 20 fold decreases in potency for G288C (EC50 9.3 μM) and R292C (EC50 15 μM). The largest change in ATP potency was at the mutants N290C, F291C and K309C with a greater than 50 fold decrease in agonist potency (EC50s 64, 44 and 175 μM respectively, table 1 and Figure 1). For the mutants N290C and F291C this reduction in potency was associated with a decrease in Hill slope (1.3 ± 0.1, 0.9 ± 0.1 and 0.7 ± 0.1 for WT and mutants respectively p<0.01) however in contrast the Hill slope was increased for the R292C mutant (1.8 ± 0.2 p<0.05).

Figure 1
Effect of point cysteine P2X1 mutations on ATP potency
Table 1
Summary of the effects of cysteine substitution

A decrease in ATP potency could result from an effect on agonist binding and/or gating of the channel. To address this issue we used a cross-linking assay with 2-azido ATP, this contains a photo-active azido (-N3) group that is chemically inert until activation by UV light when a nitrene group is formed that reacts according to free-radical rules with any nearby part of the receptor. 2-azido ATP is an agonist at the P2X1 receptor and upon irradiation, a covalent linkage between the compound and the active site of the receptor is formed (given the desensitisation of agonist evoked responses in the continued presence of agonist within 1-2s, it is likely that 2-azido ATP cross-links to desensitised receptors). In ongoing studies on P2X1 WT receptors expressed in HEK293 cells we showed that UV irradiation, or incubation with 1 μM 2-azido ATP followed by > 5mins washout, has no effect on the amplitude of ATP evoked currents (data not shown). However UV irradiation in the presence of 1 μM 2-azido ATP followed by > 5 mins washout reduced subsequent ATP currents (1 μM) by ~ 60 % (2277 ± 588 and 770 ± 270 pA for control and UV irradiation in the presence of 2-azido ATP respectively, n=7 for each). This indicates that the 2-azido ATP had cross linked to the receptors, most likely locking them in an agonist bound desensitised state, and blocked subsequent receptor activation. We therefore tested the effects of 1 μM 2-azido ATP on ATP (EC90 concentration) evoked responses at the mutants G288C, N290C, F291C, R292C and K309C. Irradiation in the presence of 2-azido ATP reduced by ~ 55% the ATP response at G288C (2524 ± 405 and 1158 ± 308 pA for control and 2-azido ATP irradiation, n=9 for each, p<0.05) this is a similar reduction as that of the WT and indicates there is little effect on 2-azido cross-linking to the G288C mutant channel (and consistent with the lack of effect of MTS reagents at this mutant, see later). Irradiation in the presence of 2-azido ATP had no effect on the amplitude of subsequent responses to ATP at N290C, F291C, R292C and K309C mutants (data not shown). These results suggest that there is a reduction in cross linking of 2-azido ATP to the mutants N290C, F291C, R292C and K309C and would be consistent with a decrease in agonist binding at these mutants. We therefore sought to test whether this could be measured directly using radiolabelled 32P 2-azido ATP.

Wild type and mutant P2X receptors expressed in Xenopus oocytes were incubated with radiolabelled 32P 2-azido ATP and exposed to UV light. Oocytes were lysed and P2X1 receptors were immuno-precipitated (to remove background non-specific binding), run on a gel and exposed to film. Levels of 2-azido ATP binding were compared with WT for the five mutants with the greatest shifts in ATP potency. The autoradiographs show a single band at ~ 55 kDa corresponding to P2X1 receptors. WT and the mutants had the same level of total P2X1 receptor and peak amplitude currents (Figure 2) indicating that the receptors are expressed at equivalent levels and a reduction in intensity of the band corresponds to a decrease in binding. Radiolabelling at WT and the mutants G288C (10 fold decrease in ATP potency) and F289C (no change in ATP EC50) was detected at similar levels, in contrast binding to the mutants N290C, F291C was reduced by >50% and by ~ 80% for R292C and K309C (Figure 2). The lack of effect of the G288C mutation on radiolabelling suggests that this cysteine substitution reduced ATP potency by an effect on the gating of the receptor (and this is consistent with the lack of modification by MTS reagents compared to adjacent the residues N290C, F291C and R292C – see later). For the mutants N290C, F291C, R292C and K309C there is no simple relationship between the reduction in the amount of 2-azido ATP binding and the decrease in agonist potency at the receptors. The ~ 80% reduction for R292C and K309C suggests that the change in potency at these receptors results predominantly from a decrease in agonist binding. The increase in Hill slope for the R292C mutant also indicates a contribution from effects on conformational changes and gating of the receptor. At N290C and F291C even though these showed a greater decrease in agonist potency than R292C they did not show as great a reduction in radioligand binding. The ~ 40% reduction in 2-azido ATP cross linking and decrease in Hill slope for the mutants N290C and F291C suggests that the reduction in their potency results from an ~ 50% decrease in agonist binding and with the additional decrease in potency due to an effect on channel gating. These results show directly for the first time that mutation of residues N290, F291, R292 and K309 decreased agonist binding at P2X receptors.

Wild type P2X1 receptor channels open rapidly and then desensitise during the continued presence of ATP (time for the peak response to decay by 50% is 986 ± 56 ms for responses to 100 μM ATP, a maximal concentration). The time-course of the response gives a measure of the transduction steps following agonist binding; i.e. channel gating, as does the deactivation step on washout of ATP. For the majority of cysteine substitutions there was no change in the time-course of decay of the response during continued ATP application (maximal ATP concentration). Mutants with a change in time-course could be divided into two categories (i) mutants G288C, N290C and K309C with decreased ATP potency showed a prolongation in the decay of the evoked current (2669 ± 165, 2907 ± 158 and 2198 ± 353 respectively) and (ii) mutants A323C and G324C had no effect on ATP potency but doubled the decay time (2125 ± 281 and 2022 ± 212 respectively) (Figure 3). These results suggest that A323 and G324 contribute to the gating of the P2X1 receptor with mutation of these residues to cysteine leading to prolonged channel opening before entering the desensitised state.

Figure 3
Effect of P2X1 mutations to cysteine on the time course of ATP action

Effect of MTS compounds on ATP responses at cysteine mutants

For P2X receptors the ability to introduce charge using MTS reagents may be a useful tool as (i) the ATP molecule has negative charge associated with the phosphate tail, (ii) in physiological solutions magnesium is complexed with ATP resulting in localised positive charge and (iii) the channel is selective for cations. We have therefore tested the ability of positively charged (2-aminoethyl) methanethiosulfonate hydrobromide (MTSEA) and negatively charged sodium (2-sulfonatoethyl) methanethiosulfonate (MTSES) to modify ATP evoked responses at WT and mutant P2X1 receptors. For these protocols following establishment of reproducible control responses MTS reagents (1 mM) were applied for the 5 minutes between ATP applications as well as being co-applied with ATP (5 minutes is required between ATP applications to allow for recovery from receptor desensitisation). This should allow the effects of MTS reagents at any accessible cysteine residues either in the non-liganded (MTS reagent application in the absence of ATP) or ATP bound receptors (MTS reagent in the presence of ATP) to be determined.

At wild type P2X1 receptors ATP (30 μM) evoked responses were potentiated slightly by positively charged MTSEA (1 mM, 112 ± 2% of the control response)(Figures response)(Figures44 & 5). This effect in unlikely to be at a free cysteine residue in the receptor, as the ten cysteine residues in the extracellular segment are considered to form five disulphide bonds (Clyne et al., 2002; Ennion and Evans, 2002), and is most likely to be a non-specific effect. At the majority of cysteine mutants MTSEA had no effect on the amplitude of responses compared to WT channels (Figure 5). A reduction of the amplitude of ATP responses following MTSEA (1 mM) treatment was observed for the mutants N290C, R292C, G321C and I328C (MTSEA reduced the responses to 4 ± 2%, 42 ± 7%, 41 ± 7% and 41 ± 4% relative to the control ATP response respectively)(Figure respectively)(Figure44 & 5). In contrast at mutants F291C and K309C addition of MTSEA potentiated ATP evoked responses to 164 ± 15% and 168 ± 10% relative to control respectively. In all cases the maximal effects of MTSEA on the amplitude of ATP responses were seen in 5-10 minutes (1-2 ATP applications; ATP was applied at 5 minute intervals).

Figure 4
Effect of MTS compounds on ATP responses of P2X1 cysteine mutants
Figure 5
Summary of effects of MTS compounds on ATP responses of P2X1 cysteine mutants

Negatively charged MTSES (1 mM) had no effect on the amplitude of ATP evoked responses at WT P2X1 receptors (100 ± 3 % of control response) and the majority of individual cysteine substituted mutants (Figures (Figures44 & 5). However ATP responses at the mutants N290C, R292C, K309C, D316C and G321C were all significantly inhibited by 1 mM MTSES (41 ± 4 %, 10 ± 1 %, 6 ± 0.2 %, 69 ± 6 % and 44 ± 6 % relative to the control ATP response respectively). F291C and A323C were both significantly potentiated (129 ± 5 % and 138 ± 2 % relative to control respectively). In all cases the maximal effects of MTSES on the amplitude of ATP responses were seen in 5-10 minutes of MTSES application (ATP was applied at 5 minute intervals).

The modification of some mutant P2X1 receptor responses highlights a contribution of these residues to ATP action at the receptor. The amplitude of P2X1 responses could be regulated by a change in the ATP potency of the receptor and/or an effect on channel gating and ionic permeation. To establish whether a change in ATP potency was involved we determined the effect of MTS reagents on the concentration response curves. For this MTS compounds were present in the superfusion solution and concomitantly applied with ATP. For mutants D316C (pEC50 5.7 ± 0.07 and 5.9 ± 0.22 for control and MTSES respectively), G321C (pEC50 6.5 ± 0.07, 6.8 ± 0.20 and 6.4 ± 0.18 for control, MTSEA and MTSES respectively), A323C (pEC50 5.9 ± 0.17 and 6.1 ± 0.05 for control and MTSES respectively), I328C (pEC50 6.1 ± 0.15 and 6.2 ± 0.13 for control and MTSEA respectively) the MTS compounds had no effect on ATP potency demonstrating that at these mutants the MTS compounds were not acting to interfere with ATP binding to the receptor but having an effect on channel gating or ionic permeation. In contrast at residues N290C, F291C, R292C and K309C, MTS compounds changed the potency of ATP at the receptors. At N290C MTSES and R292C MTSEA decreased ATP potency by ~18 and 5 fold respectively (p<0.005 for pEC50 values) while MTSEA potentiates at F291C by 5 fold (p<0.005). The similar modulatory effects of positively charged MTSEA and negatively charged MTSES at N290C (inhibition), F291C (potentiation) and R292C (inhibition) receptors suggest that they result from the introduction of the bulk of the side groups and that localised charges at these residues do not contribute to ATP action. At K309C the effects of the MTS compounds were dependent on the charge. ATP potency was increased ~ 20 fold following incubation with positively charged MTSEA (the MTSEA modified cysteine residue is of similar size and charge to the lysine residue and has been used to rescue function of a lysine mutation in the intracellular domain of the P2X4 receptor, Fountain and North, 2006)(p<0.0005 for pEC50 values), and responses even to a maximal concentration of ATP (10 mM) were abolished by negatively charged MTSES. This suggests that charge at position K309 is important for mediating the binding of ATP. MTS reagents react irreversibly with cysteine residues and overnight incubation with MTS compounds had the same effects on ATP potency as bath application (Figure 6) and can be used to reduce the amount of costly MTS reagents used.

Figure 6
Effect of MTS compounds on ATP potency at cysteine mutants

MTSEA biotin accessibility at P2X1 wild type and cysteine mutants

MTS compounds have identified substituted cysteine residues that contribute to ATP action at the P2X1 receptor. For residues where MTS compounds have no effect it is unclear if this is as the residue does not contribute to the actions of ATP at the receptor, or whether the residue is inaccessible. To help discriminate between these two possibilities, and gain some experimental evidence for the secondary structure of the extracellular loop, we have determined directly whether MTSEA-biotin reacts with individual cysteine point mutants of the P2X1 receptor. MTSEA biotin molecule is like a ball and chain with an ~ 0.5 × 0.7 nm planar head group and the reactive disulphide is near the end of a flexible chain (~1.7 × 0.5 nm) incorporating the MTSEA (~ 1.0 × 0.5 nm)(estimated using SwissPDBViewer). Residues that react with MTSEA biotin must therefore be accessible through an opening at least 0.5 nm in diameter and greater than 2.2 nm deep widening to at least 0.7 nm (to incorporate the biotin). MTSEA-biotin can be used to differentiate water-exposed cysteines from cysteines buried within a protein structure, because only the water-exposed cysteines react at an appreciable rate.

MTSEA-biotinylation was detected by Western blotting with an anti-P2X1 receptor antibody following isolation of biotinylated proteins with streptavidin agarose beads (Figure 7)(any endogenous ATP released from the oocytes was broken down by incubations in the presence of apyrase. At wild type P2X1 receptors very low levels of biotinylated protein were detected as reported previously (Ennion and Evans, 2002) consistent with the ten cysteine residues in the extracellular loop forming five disulphide bonds (Clyne et al., 2002; Ennion and Evans, 2002). The threshold for positive biotinylation was set at >3 times the WT background level (Figures (Figures77 and and8).8). For the majority of substituted cysteine residues (36/42) MTSEA-biotinylation was detected demonstrating that these substituted cysteine residues are accessible at the receptor surface. For N300C-T302C this corresponds to a site of golgi mediated glycosylation consistent with surface accessibility (Rettinger et al., 2000; Roberts and Evans, 2006)). The degree of biotinylation was dependent on the position of substitution indicating that the local environment or accessibility is variable across this region of the receptor. We also determined whether the level of biotinylation was changed by stimulation with ATP (Figure (Figure77 & 8).

Figure 7
MTSEA biotin labelling of mutant and wild type P2X1 receptors
Figure 8
Summary of MTSEA biotin labelling of P2X1 mutants and wild type receptor and the effects of ATP stimulation

ATP reduced MTSEA-biotinylation for cysteine substituted residues in the regions S286-E299, Y304, F308C-D316C and L318-D320. In these cases (with the exception of N290F291R292 and K309 cysteine mutants) there was little or no effect on ATP potency. One explanation for these changes in MTSEA-biotinylation could be that localised charge in the binding pocket associated with the bound ATP interfered with MTSEA-biotinylation. However when either positively (MTSEA) or negatively charged (MTSES) reagents are applied to these mutants they had no effect on ATP potency (supporting that they do not contribute directly to the agonist binding site) and suggest that at these positions ATP does not directly interfere with MTSEA-biotinylation. An alternative is that the decrease in MTSEA-biotinylation at these residues results from a conformational change in the extracellular loop of the P2X1 receptor associated with agonist binding making the substituted cysteine residues less accessible. A conformational change is also supported by the increase in MTSEA-biotinylation in response to ATP for K325C a residue close to the predicted pore region of the P2X receptor channel. Such a conformational change associated with agonist binding could be similar to the “clam shell” closure of the agonist binding pocket of ionotropic glutamate receptors (Jin et al, 2003). If the decrease in MTSEA-biotinylation does result from a conformational change in the P2X1 receptor this gives us a first glimpse of how agonist binding leads to structural changes in the extracellular ligand binding loop of the receptor. It is tempting to speculate that the lack of change by ATP application to MTSEA-biotinylation in the region N300C-N303C whilst the adjacent regions showed marked decreases suggests that this region could act as a stable point or hinge around which conformation changes occur bringing together or stabilising the proximity of the N290F291R292 and K309 residues that we suggest contribute to the ATP binding site.

Barely above threshold levels of biotinylation were detected for H306C, R314C, G321C, A323C and F326C (either in the presence or absence of ATP) suggesting that there is restricted accessibility of MTSEA-biotin at these positions. Very low levels equivalent to WT were detected for K322C (either in the presence or absence of ATP) indicating that this residue is buried within the protein or MTSEA is unable to bind. Overall the MTSEA-biotinylation results indicate regions of the extracellular loop that are readily accessible on the surface of the protein, can be used to develop models of the extracellular segment of the receptor, and give an insight into conformational changes following receptor activation.


The majority of cysteine mutants of 44 amino acids before the second transmembrane segment had little (4 mutants) or no effect (33 mutants) on ATP potency and were unaffected by MTS reagents. Several of these mutants react with MTSEA-biotin, indicating their accessibility, but MTSEA had no effect on ATP responses (Figure 9). This allows us to exclude an essential role of these residues in mediating ATP action at the P2X1 receptors. We identified residues where cysteine substitution (i) reduced currents and expression at the cell surface (R295C and R305C), however mutants R295A and R305A (Ennion et al., 2000) had no effect on ATP potency indicating that individual positive charges at these residues are not essential (ii) decreased ATP potency, 2-azidoATP binding and modification by MTS reagents (K309C, R292C), (iii) decreased ATP potency, modification by MTS reagents, modest reduction in 2-azidoATP binding and a reduction in Hill slope (N290C and F291C) and (iv) little or no effect on ATP potency and modification of amplitude by MTS reagents (D319C, G321C, A323C and I328C). Mutations in the extracellular domain could have effects on the binding of agonist to the receptor and/or the conformation change following agonist binding that controls activation of the channel (Colquhoun, 1998).

Figure 9
Summary of the effects of cysteine mutants on ATP potency, modification by MTS reagents and MTSEA biotinylation

The ~ 200 fold decrease in ATP potency at K309C is consistent with previous alanine scanning mutagenesis of this residue in P2X1 (Ennion et al., 2000), P2X2 (Jiang et al., 2000) and P2X7 (Worthington et al., 2002) receptors. Several features of this mutation are indicative of an effect on agonist binding. For the 200 fold reduction in potency to result from a gating effect this would require an ~ 40,000 fold reduction in gating constant with two molecules of ATP, and considerably higher if three bind to activate the channel. As there is no reduction in the maximal response gating would have to be highly efficacious at the wild type receptor, which is unlikely given the low probability of single P2X channel openings (Ding and Sachs, 1999). Similarly there is no detectable reduction in the Hill slope that would be expected if there were a change in gating. The ~80% reduction in the amount of radiolabelled 2-azido ATP cross-linked to the mutant K309C receptor further supports a decrease in agonist binding. The rescue of potency with positively charged MTSEA suggests that K309 is involved in binding of the negatively charged phosphate group of the ATP molecule. Finally the binding of MTSEA-biotin to the free cysteine residue introduced in the K309C mutant receptor was reduced by ATP application. Taken together these results support the idea that lysine residue 309 forms part of the ATP binding site of the P2X receptor.

The decreased ATP potency of the cysteine mutations N290C, F291C, and R292C is consistent with previous alanine scanning mutagenesis studies (Ennion et al., 2000; Jiang et al., 2001; Roberts and Evans, 2004; Roberts and Evans, 2006). The decrease in 2-azido ATP cross-linking, and MTSEA-biotinylation by ATP application indicate that these residues contribute to the ATP binding site. At N290C and R292C ATP potency was further decreased by MTSEA or MTSES. In contrast at F291C ATP potency increased towards the WT potency following treatment with MTS reagents. The lack of discrimination between the effects of positive and negatively charged MTS compounds suggests that it is the localised disruption associated with binding of the MTS reagent, and not the charge, that modifies ATP potency. This indicates that it is unlikely that the positive charge of R292 co-ordinates binding of negatively charged phosphate on ATP as previously suggested (Ennion et al., 2000; Vial et al., 2004). The NFR motif is totally conserved in P2X receptors and we have suggested that the aromatic phenylalanine is involved in binding of the adenine ring of ATP (Roberts and Evans, 2004), and this is consistent with residues involved in ATP binding at other proteins (see below). In addition the current results suggest that this motif is involved in agonist induced conformational changes at the P2X1 receptor.

A model of the C terminal half of the extracellular segment of the P2X4 receptor was proposed based on similarity to the tRNA synthetases (Freist et al., 1998) and this has been supported by some mutagenesis studies (Yan et al., 2005). However the residues suggested to form the P2X4 receptor binding site are different from those identified for the P2X1 receptor (Stojilkovic et al, 2005) suggesting that there may be differences in ATP action between P2X receptor subunits.

Analysis of the local ATP binding environment in a number of crystal structures identified a similar pattern of residues involved in ATP binding to P2X1 receptors as at rat synapsin II (Roberts et al., 2006) with an asparagine-aromatic-positive charge (NWK) motif with the aromatic group and arginine co-ordinating the binding of the adeninine and ribose groups of ATP (PDB code li71 (Berman et al., 2000)). An additional aromatic residue in synapsin II was associated with the adeninine ring and this is consistent with our previous work showing F185 mutants decrease ATP potency (Roberts and Evans, 2004). Three positively charged lysine residues are also key components of the ATP binding environment at synapsin II and these could correspond to K68, K70 and K309 that we have proposed to co-ordinate ATP binding (Ennion et al., 2000). Based on the similarity of residues identified in our mutagenesis studies and with those present in the local binding environment of synapsin II we suggest a revised working model of the binding site at the P2X1receptor (Figure 10). This is supported by recent work on P2X1 receptors showing that a disulphide bond can be formed between adjacent subunits of cyteine residues introduced at positions K68 and F291 (Marquez-Klaka et al., 2007) and suggests that the ATP binding site forms between subunits has been shown for the zinc binding site in P2X2 receptors (Nagaya et al., 2005).

Figure 10
Model of ATP action at P2X1 receptors

MTSEA-biotinylation has provided a useful assay to determine the surface accessibility of introduced cysteine residues. The lack (K322C) or threshold levels of MTSEA-biotinylation (S286C, H306C, R314C, G321C, A323C and F326C) suggest that these residues are either inaccessible (K322C) or give restricted access to MTSEA-biotin. In contrast the majority of substituted cysteine residues showed MTSEA-biotinylation with the relative levels giving an indication of relative accessibility. Of the MTSEA-biotinylated cysteine residues only the N290F291R292 motif and K309 cysteine mutants showed changes in ATP potency following MTS application. MTSEA-biotinylation at these residues was also reduced by ATP application. This would be consistent with ATP competing with the MTSEA-biotin for occupation of the agonist binding site and further supports the close association of the NFR motif and K309 with the ATP binding site.

The cysteine mutants D316C, G321C, A323C and I328C were sensitive to MTS reagents. At P2X2 receptors (Rassendren et al., 1997) MTS reagents had no effect on residues equivalent to the conserved residues G321C and A323C suggesting that constraints imposed by non-conserved parts of the receptor contributed to actions of MTS compounds at the P2X1 receptor. At D316C and G321C both MTSEA and MTSES inhibited responses indicating that introduction of bulk and not charge mediated the inhibition. At A323C only negatively charged MTSES had an effect and at I328C only positively charged MTSEA was effective and suggests that the adjacent charged K322 and D327 residues contribute to efficacy of the MTS compounds at these residues. The MTSEA-biotinylation of residues G324C, K325C and D327C-I329C also gives an insight into the minimum dimensions and accessibility of this region of the channel (see above) that has been suggested to contribute to the outer vestibule of the channel pore (Rassendren et al., 1997).

The second transmembrane segment forms part of the P2X receptor channel pore and starts at about residue P330. Individual cysteine mutants in the region D316-I329 had little or no effect on ATP potency, MTS reagents modified responses at D316C, G321C, A323C and I328C mutants with no effect on ATP potency, in addition the mutants A323C and G324C showed a slowing in the time-course of desensitisation (Figure 3). These results suggests that this region of the receptor does not contribute directly to ATP binding and reduction in amplitude of ATP responses at D316C, G321C, and I328C mutants by MTS compounds results from a decrease in ATP efficacy. This reduction in efficacy suggests that these residues are likely to be involved in regulation of ionic permeation or channel gating (Figure 10). This is similar to the cys-loop ligand gated ion channels and glutamate ion channels where the ligand binding site is distant from the transmembrane segments and residues close to the transmembrane segments link agonist binding to channel opening (Czajkowski, 2005) and supports recent work on P2X4 receptors (Yan et al., 2006).

In summary this study has provided new insight into the molecular basis of ATP action at P2X receptors, direct experimental evidence for the ATP binding site, the accessibility of the C terminal portion of the extracellular segment, and aided in the refining of a model of agnist action at this novel family of ATP binding proteins.


This work is supported by the Wellcome Trust. We thank Kelvin Agboh for his contributions to the radiolabelling work, Rebecca Allsopp for helping with some of the Western blotting and Helen Digby for discussion of the synapsin II model of ATP binding.


  • Akabas MH, Stauffer DA, Xu M, Karlin A. Acetylcholine receptor channel structure probed in cysteine-substitution mutants. Science. 1992;258:307–310. [PubMed]
  • Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN, Bourne PE. The Protein Data Bank. Nucleic Acids Res. 2000;28:235–242. [PMC free article] [PubMed]
  • Chessell IP, Hatcher JP, Bountra C, Michel AD, Hughes JP, Green P, Egerton J, Murfin M, Richardson J, Peck WL, Grahames CB, Casula MA, Yiangou Y, Birch R, Anand P, Buell GN. Disruption of the P2X7 purinoceptor gene abolishes chronic inflammatory and neuropathic pain. Pain. 2005;114:386–396. [PubMed]
  • Clyne JD, Wang LF, Hume RI. Mutational analysis of the conserved cysteines of the rat P2X2 purinoceptor. J Neurosci. 2002;22:3873–3880. [PubMed]
  • Cockayne DA, Hamilton SG, Zhu Q-M, Dunn PM, Zhong Y, Novakovic S, Malmberg AB, Cain G, Berson A, Kassotakis L, Hedley L, Lachnit WG, Burnstock G, McMahon SB, Ford APDW. Urinary bladder hyporeflexia and reduced pain-related behaviour in P2X3-deficient mice. Nature. 2000;407:1011–1015. [PubMed]
  • Colquhoun D. Binding, gating, affinity and efficacy: the interpretation of structure- activity relationships for agonists and of the effects of mutating receptors. Br J Pharmacol. 1998;125:924–947. [PMC free article] [PubMed]
  • Czajkowski C. Neurobiology: triggers for channel opening. Nature. 2005;438:167–168. [PubMed]
  • Ding S, Sachs F. Single channel properties of P2X2 purinoceptors. J Gen Physiol. 1999;113:695–720. [PMC free article] [PubMed]
  • Egan TM, Haines WR, Voigt MM. A domain contributing to the ion channel of ATP-gated P2X2 receptors identified by the substituted cycteine accessibitlity method. J Neurosci. 1998;18:2350–2359. [PubMed]
  • Ennion S, Hagan S, Evans RJ. The role of positively charged amino acids in ATP recognition by human P2X1 receptors. J Biol Chem. 2000;275:29361–29367. [PubMed]
  • Ennion SJ, Evans RJ. Conserved cysteine residues in the extracellular loop of the human P2X1 receptor form disulfide bonds and are involved in receptor trafficking to the cell surface. Mol Pharmacol. 2002;61:303–311. [PubMed]
  • Freist W, Verhey JF, Stuhmer W, Gauss DH. ATP binding site of P2X channel proteins: structural similarities with class II aminoacyl-tRNA synthetases. FEBS Lett. 1998;434:61–65. [PubMed]
  • Haines WR, Voigt MM, Migita K, Torres GE, Egan TE. On the contribution of the first transmembrane domain to whole-cell current through an ATP-gated ionotropic P2X receptor. J Neurosci. 2001;21:5885–5892. [PubMed]
  • Jiang L-H, Rassendren F, Surprenant A, North RA. Identification of amino acid residues contributing to the ATP binding site of a purinergic P2X receptor. J Biol Chem. 2000;275:34190–34196. [PubMed]
  • Jiang L-H, Rassendren F, Spelta V, Surprenant A, North RA. Amino acid residues involved in gating identified in the first membrane-spanning domain of the rat P2X2 receptor. J Biol Chem. 2001;276:14902–14908. [PubMed]
  • Jin R, Banke TG, Mayer ML, Traynelis SF, Gouaux E. Structural basis for partial agonist action at ionotropic glutamate receptors. Nat Neurosci. 2003;6:803–810. [PubMed]
  • Marquez-Klaka B, Rettinger J, Bhargava Y, Eisele T, Nicke A. Identification of an intersubunit cross-link between substituted cysteine residues located in the putative ATP binding site of the P2X1 receptor. J Neurosci. 2007;27:1456–1466. [PubMed]
  • McGuffin LJ, Bryson K, Jones DT. The PSIPRED protein structure prediction server. Bioinformatics. 2000;16:404–405. [PubMed]
  • Nagaya N, Tittle RK, Saar N, Dellal SS, Hume RI. An intersubunit Zn2+ binding site in rat P2X2 receptors. J Biol Chem. 2005;280:25982–25993. [PMC free article] [PubMed]
  • North RA. Molecular Physiology of P2X Receptors. Physiol Rev. 2002;82:1013–1067. [PubMed]
  • Rassendren F, Buell G, Newbolt A, North RA, Surprenant A. Identification of amino acid residues contributing to the pore of a P2X receptor. EMBO J. 1997;16:3446–3454. [PubMed]
  • Ren J, Bian X, DeVries M, Schnegelsberg B, Cockayne DA, Ford AP, Galligan JJ. P2X2 subunits contribute to fast synaptic excitation in myenteric neurons of the mouse small intestine. J Physiol. 2003;552:809–821. [PubMed]
  • Rettinger J, Aschrafi A, Schmalzing G. Roles of individual N-glycans for ATP potency and expression of the rat P2X1 receptor. J Biol Chem. 2000;275:33542–33547. [PubMed]
  • Roberts JA, Evans RJ. ATP Binding at Human P2X1 Receptors: contribution of aromatic and basic amino acids revealed using mutagenesis and partial agonists. J Biol Chem. 2004;279:9043–9055. [PubMed]
  • Roberts JA, Evans RJ. Contribution of conserved polar glutamine, asparagine and threonine residues and glycosylation to agonist action at human P2X1 receptors for ATP. J Neurochem. 2006;96:843–852. [PubMed]
  • Rong W, Gourine AV, Cockayne DA, Xiang Z, Ford AP, Spyer KM, Burnstock G. Pivotal Role of Nucleotide P2X2 Receptor Subunit of the ATP-Gated Ion Channel Mediating Ventilatory Responses to Hypoxia. J Neurosci. 2003;23:11315–11321. [PubMed]
  • Souslova V, Cesare P, Ding Y, Akopian AN, Stanfa L, Suzuki R, Carpenter K, Dickenson A, Boyce S, Hill R, Nebenius-Oosthuizen D, Smith AJH, Kidd EJ, Wood JN. Warm-coding deficits and aberrant inflamatory pain in mice lacking P2X3 receptors. Nature. 2000;407:1015–1017. [PubMed]
  • Stojilkovic SS, Tomic M, He ML, Yan Z, Koshimizu TA, Zemkova H. Molecular Dissection of Purinergic P2X Receptor Channels. Ann N Y Acad Sci. 2005;1048:116–130. [PubMed]
  • Trapp S, Haider S, Jones P, Sansom MS, Ashcroft FM. Identification of residues contributing to the ATP binding site of Kir6.2. EMBO J. 2003;22:2903–2912. [PubMed]
  • Tsuda M, Shigemoto-Mogami Y, Koizumi S, Mizokoshi A, Kohsaka S, Salter MW, Inoue K. P2X4 receptors induced in spinal microglia gate tactile allodynia after nerve injury. Nature. 2003;424:778–783. [PubMed]
  • Vial C, Evans RJ. P2X1 receptor-deficient mice establish the native P2X receptor and a P2Y6-like receptor in arteries. Mol Pharmacol. 2002;62:1438–1445. [PubMed]
  • Vial C, Roberts JA, Evans RJ. Molecular properties of ATP-gated P2X receptor ion channels. Trends Pharmacol Sci. 2004;25:487–493. [PubMed]
  • Vial C, Tobin AB, Evans RJ. G-protein coupled receptor regulation of P2X1 receptors does not involve direct channel phosphorylation. Biochem J. 2004;382:101–110. [PubMed]
  • Walker JE, Saraste M, Runswick MJ, Gay NJ. Distantly related sequences in the alpha- and beta-subunits of ATP synthase, myosin, kinases and other ATP requiring enzymes and a common nucleotide binding fold. EMBO J. 1982;1:945–951. [PubMed]
  • Worthington RA, Smart ML, Gu BJ, Williams DA, Petrou S, Wiley JS, Barden JA. Point mutations confer loss of ATP-induced human P2X7 receptor function. FEBS Lett. 2002;512:43–46. [PubMed]
  • Yan Z, Liang Z, Tomic M, Obsil T, Stojilkovic SS. Molecular determinants of the agonist binding domain of a P2X receptor channel. Mol Pharmacol. 2005;67:1078–1088. [PubMed]