Logo of jphysiolThe Journal of Physiology SiteMembershipSubmissionJ Physiol
J Physiol. Dec 1, 1998; 513(Pt 2): 317–330.
PMCID: PMC2231302

Opposite modulation of NMDA receptors by lysophospholipids and arachidonic acid: common features with mechanosensitivity


  1. Two classes of amphiphilic compounds, lysophospholipids and arachidonic acid, have been suggested to produce opposite deformations of the lipid bilayer. We have found that their effects on N-methyl-D-aspartate (NMDA) responses are opposite, and resemble those of mechanical deformations of the plasma membrane.
  2. Lysophospholipids inhibited NMDA responses both in nucleated patches taken from cultured neurons and in cells expressing recombinant NMDA receptors. This inhibition was reversible, voltage independent and stronger at non-saturating doses of agonist. It was not linked to the charge of the polar head, and was not mimicked by lysophosphatidic acid or phosphatidylcholine. In outside-out patches, lysophospholipids reduced the open probability of NMDA-activated channels without changing their single-channel conductance.
  3. The inhibition produced by lysophospholipids occluded that produced by a mechanical compression induced by changes in osmotic or hydrostatic pressure.
  4. The potentiation of NMDA responses by arachidonic acid was observed both in native and recombinant receptors, including those in which the putative ‘fatty acid binding domain’ had been deleted. This suggests that, like lysophospholipids, arachidonic acid alters the NMDA receptor by insertion into the lipid bilayer.
  5. Recombinant receptors in which the cytoplasmic tails had been modified or deleted were still sensitive to mechanical deformation. A linkage to the cytoskeleton is therefore not required for NMDA receptor mechanosensitivity.
  6. The fact that the NMDA responses are depressed similarly by compression and lysophospholipids, and potentiated similarly by stretch and arachidonic acid supports the notion that the modulation of NMDA receptor activity by asymmetrical amphiphilic compounds involves pressure changes transmitted through the lipid bilayer. Compounds with a large hydrophilic head mimic the effects of a compression, and compounds with a small hydrophilic head mimic the effects of stretch.

Classifying ionic channels by the stimulus to which they are most sensitive leads to separate families of ligand-gated, voltage-gated and mechano-sensitive channels. However, it is known that a given channel is often sensitive to more than one stimulus type and that some channels are sensitive to both ligands and voltage, or voltage and stretch. The channels opened by glutamate in receptors of the NMDA family are unusual in that they are actually sensitive to all three types of stimuli. Although they are primarily ligand-gated channels (their opening normally requires two ligands, glutamate and glycine), they are also sensitive to voltage (at least in physiological conditions, in the presence of external Mg2+) and to mechanical forces: at a given voltage and in the presence of fixed concentrations of glutamate and glycine, membrane stretch increases and membrane compression decreases their probability of opening (Paoletti & Ascher, 1994).

Although a functional role of the mechanosensitivity of NMDA channels has not been established, this property provides a possible tool for understanding the conformational changes of the NMDA receptor associated with its opening and closing. The experiments presented here, performed in this general perspective, aimed at characterizing the ‘point(s) of application’ of the mechanical forces which are capable of modulating the NMDA receptor function.

During the last 15 years, as the study of mechanosensitive channels has developed, two main types of hypotheses have been proposed to explain how mechanical forces applied to a cell membrane are transmitted to ion channels. One hypothesis assumes the presence of structures pulling on the hydrophilic structures of the ion channels either extracellularly, as in the case of hair cells where a key role is attributed to the ‘tip-link’ connecting neighbouring stereocilia, or intracellularly, by an elastic link between the channel and the cytoskeleton (see Guharay & Sachs, 1984). The second hypothesis (which does not necessarily exclude the first) assumes that membrane tension is transmitted to the channel via the lipid bilayer. This hypothesis appears unavoidable in the cases where mechanosensitivity is observed for channels incorporated into a lipid bilayer (Opsahl & Webb, 1994; Sukharev et al. 1994; Ismailov et al. 1997), but it has not been submitted to unequivocal tests for most of the mechanosensitive channels recorded in native cell membranes.

The NMDA receptor is connected to cytoskeletal elements which could be candidates for force transmission (Kornau et al. 1997), but the transmission of force through such a link is made somewhat unlikely by the fact that the stretch sensitivity is still observed in excised patches and nucleated patches, i.e. in conditions where the link of the NMDA receptors with the cytoskeleton is probably severed (Paoletti & Ascher, 1994). The involvement of the lipid bilayer was therefore the favoured hypothesis and we decided to evaluate it in more detail. To this purpose we analysed the effects on NMDA receptors of amphipathic compounds that have been found to induce tension or deformation of various artificial and biological membranes by altering the organization of the lipid bilayer.

Among the first observations of mechanical deformations induced by amphipathic compounds were those of Sheetz & Singer (1974), who interpreted the changes in cell shape according to the ‘bilayer couple theory’. This theory assumes that, in a cell at resting potential, charged amphipaths accumulate preferentially in one half of the lipid bilayer (negatively charged compounds in the outer layer, positively charged compounds in the inner layer) and induce membrane curvature in opposite directions (crenation or cup formation). Martinac et al. (1990), used the theory to explain the effects of amphipathic compounds on the stretch-sensitive channels of bacterial membranes. They observed that both cationic and anionic amphipaths mimicked the effect of stretch and activated the channels, and that two amphipaths of opposite charge neutralized each other when added together. This was explained by assuming that both a convex or a concave curvature of the membrane creates a mechanical stress on the channels; if both halves of the bilayer expand, the global curvature of the membrane is not altered. More recently, however, Petrou et al. (1994) analysed the effects of a series of charged lipophilic compounds on smooth muscle stretch-sensitive K+ channels and found that, in contrast with the results discussed above, negatively and positively charged compounds had opposite effects: arachidonic acid and other negatively charged compounds mimicked the effects of stretch and activated the channels, whereas sphingosine (a positively charged compound) had an inhibitory effect. This could still be explained in the frame of the bilayer couple theory, if the channels could differentiate between two opposite curvatures, convex or concave. But other observations are definitely not compatible with the bilayer couple theory. Eddlestone & Ciani (1991) found that KATP channels are sensitive to lysophosphatidylcholine (LPC) and lysophosphatidylinositol (LPI), but not to lysophosphatidylethanolamine (LPE) and lysophosophatidylserine (LPS). Since LPI and LPS are both negatively charged, whereas LPC and LPE are neutral, the charge difference cannot explain the difference in the effects. Furthermore, in the bacterial channels used by Martinac et al. (1990) for their pioneer study, the neutral LPC mimics the effects of charged amphipaths (Markin & Martinac, 1991).

One way to account for the effects not explained by the bilayer couple theory is to consider lipid shape rather than charge. In the 1970s increasing attention was paid to the molecular geometry of lipids, and in particular to the relative size of the polar head and of the hydrophobic tail (e.g. Israelachvili et al. 1977; Cullis & de Kruijff, 1979). Chernomordik et al. (1987) proposed to label as cones compounds with a head larger than the tail (e.g. lysophospholipids; for LPC the chain cross-section is 1.8–2.0 nm and the polar head group has a diameter of around 5.2 nm according to Pascher et al. 1992). Compounds with a head smaller than the tail (like arachidonic acid) were labelled inverted cones, and compounds in which head and tail have similar sizes were called cylinders. LPC and LPI, the two active compounds in the study of Eddlestone & Ciani (1991), have larger head groups than LPE and LPS, which were inactive. In the system used by Martinac et al. (1990) and Markin & Martinac (1991), all the active compounds have the shape of cones. For gramicidin channels incorporated into lipid bilayers, Lundbaek & Andersen (1994) and Andersen et al. (1995) found that lysophospholipids (cones) and cholesterol (inverted cone) had opposite effects. Similarly, detergents like Triton X-100 and β-octyl glucoside (cones) and cholesterol (inverted cone) produced opposite effects on voltage-dependent Ca2+ channels (Lundbaek et al. 1996).

The present study aimed to examine the possibility that a lipid-protein interaction was important for the mechanosensitivity of the NMDA receptor. It was already known that lipophilic compounds like arachidonic acid or docosahexaenoic acid, which have the shape of inverted cones, mimic the effects of stretch (Miller et al. 1992; Nishikawa et al. 1994; Paoletti & Ascher, 1994). To examine the validity of the prediction that lipophilic molecules having the shape of cones induce a membrane deformation opposite to that produced by inverted cones, we analysed the effects of lysophospholipids on NMDA receptors. We found that the effects of lysophospholipids are indeed opposite to those of arachidonic acid. The effects of lysophospholipids mimic (and are not additive with) the effects of a membrane compression. Further support for the involvement of the lipid bilayer in the stretch sensitivity of NMDA receptors was derived from the observation that mutant NMDA receptors lacking part of their link with the cytoskeleton were still stretch sensitive.


Cell culture

Cortical and diencephalic neurons were obtained from 15- to 16-day-old embryos removed from pregnant mice under ether anaesthesia. Neurons were cultured for 2–4 weeks as previously described (Ascher et al. 1988). Human embryonic kidney (HEK 293) cells plated on 35 mm Petri dishes were transfected with 2 μg of cDNAs mixed at a ratio of 1 NR1 : 3 NR2 : 3 GFP (where NR1 and NR2 are NMDA receptor subunits and GFP is Green Fluorescent Protein) using calcium phosphate precipitation (Chen & Okayama, 1987). Positive cells were visualized by fluorescence. For single-channel recordings from Xenopus oocyte patches, NMDA channels were expressed after nuclear injection of the corresponding cDNAs. Frogs were anaesthetized by immersion in an ice bath containing 0.2% MS-222 for 20 min. An insertion was made in the abdominal cavity from where an ovarian lobe was removed. After suture the animal was allowed to recover for at least 24 h in tap water. Each oocyte was injected with 50 nl of a 0.1 μg μl−1 cDNA solution at a ratio of 1 NR1 : 2 NR2A. The oocytes used for single-channel recordings showed glutamate (100 μm glutamate + 100 μm glycine) whole-cell currents of 3–10 μA, as assessed by two-electrode voltage clamp. The vitelline membrane of each oocyte was removed before making patch-clamp recordings. All experiments were carried out in accordance with the guidelines of the Centre National de la Recherche Scientifique animal welfare commitee.


The NMDA subunits and the postsynaptic density protein PSD-95 cDNAs, which were kindly provided by S. Nakanishi (Kyoto, Japan) and M. B. Kennedy (Caltech, CA, USA), respectively, were subcloned into a modified expression vector (pcDNA3; Invitrogen, Leek, Netherlands). Single mutations were introduced by the mutagenesis approach described previously (Kupper et al. 1996).

Recording conditions and data analysis

Recordings were performed in the usual patch-clamp configurations (whole-cell, outside-out, inside-out and cell-attached; Hamill et al. 1981) as well as in nucleated patches (Sather et al. 1992). All experiments were performed at room temperature (18–25°C). Patch pipettes of 4–8 MΩ were filled with a solution containing (mm): 120 CsF, 10 CsCl, 10 EGTA and 10 Hepes unless otherwise specified. CsOH was added to adjust the pH to 7.2. The osmolality was 265 mosmol kg−1. The standard external solution contained (mm): 130 NaCl, 2.8 KCl, 1 CaCl2 and 10 Hepes. NaOH was added to adjust the pH to 7.3. Hyperosmotic solutions were obtained by addition of mannitol to the standard external solution. Drugs were applied to the cell or patch by means of a multibarrel fast-perfusion system. Solutions flowed continuously by gravity from all barrels. Lysophospholipid stocks were prepared as 8 mm suspensions into water. Phosphatidylcholine was prepared from a 4 mm stock in DMSO. Controls in the presence of the same concentration of carrier were performed.

For single-channel recordings in cell-attached and inside-out patches, NMDA (1, 3 or 10 μm) and glycine (100 μm) were added to the Ca2+-free CsF pipette solution. Mg2+ (20 μm) was also added to the pipette solution to identify NMDA receptor openings unequivocally, by their characteristic voltage-dependent flickering. Patches showing openings with no flickering were discarded.

Currents were recorded with a List EPC-7 amplifier (Darmstadt, Germany). The voltage-clamp current was filtered (8-pole Bessel) with a corner frequency of 100 Hz for nucleated patch and whole-cell recordings and 1 kHz for single-channel recordings. Unless otherwise specified, the sampling frequency was 400 Hz for macroscopic currents and 5 kHz for single channel recordings. Data were acquired and analysed by Strathclyde Electrophysiology Software (a gift from Dr J. Dempster, University of Strathclyde, Glasgow, UK) and pCLAMP 6 (Axon Instruments).


Lysophospholipids inhibit NMDA responses in nucleated patches taken from cultured neurons

Lysophospholipids were first tested on nucleated patches from cultured neurons. Responses to brief applications of NMDA (10 μm) in the continuous presence of 100 μm glycine were recorded alternately in the absence and in the presence of the lysophospholipids. Figure 1A illustrates the responses observed in control conditions, 30 s after addition of 2 μm of lysophosphatidylcholine (LPC) and 30 s after its removal (recovery). LPC inhibited NMDA responses and recovery was complete 30 s after returning to the control solution. The figure also illustrates that the inhibition was voltage independent. Lysophosphatidylethanolamine (LPE) and lysophosphatidylinositol (LPI) produced inhibitions similar to those produced by LPC, whereas phosphatidylcholine (PC) and lysophosphatidic acid (LPA) had no effect (Fig. 1B).

Figure 1
Lysophospholipids inhibit NMDA responses

Mg2+ block of NMDA currents was not affected by lysophospholipids. Mg2+ (20 μm) reduced currents recorded at −60 mV by 63.3 ± 0.9 % in control conditions and by 63.5 ± 0.7 % in the presence of 2 μm LPI (n= 4). At −90 mV the inhibition by 20 μm Mg2+ reached 88.8 ± 1.6 % in control conditions and 88.2 ± 2.6 % in the presence of LPI, respectively (n= 2).

In order to compare results from different experiments, LPI was used for most further studies. A LPI dose-response curve was constructed for responses to 10 μm NMDA plus 100 μm glycine recorded at −50 mV in the presence of LPI at concentrations ranging from 0.01 to 5 μm (Fig. 2A). The inhibition increased monotonically, reached 50 % at about 0.8 μm and 76 ± 4 % at 5 μm LPI (n= 5). At concentrations of 10 μm, which are close to the critical micelle concentration (Marsh & King, 1986), LPI induced an irreversible decrease of the responses and patch loss (probably because of mechanical damage by micelle bombardment) and thus no attempt was made to apply higher concentrations. However, the small difference between the effects of 2 and 5 μm LPI (Fig. 2A) suggests that in this range of concentrations we were close to saturation but that the inhibition remained partial.

Figure 2
Concentration dependence of the lysophosphatidylinositol effect

We measured the inhibition produced by a given LPI concentration (2 μm) on the response to two concentrations of NMDA (10 and 300 μm) in the presence of a saturating glycine concentration (100 μm) (Fig. 2B). The inhibition was observed for both NMDA concentrations but decreased from 59.0 ± 4.5 % to 37.8 ± 10.4 % (n= 5) when the NMDA concentration was increased from 10 to 300 μm. This indicates that LPI produces both a reduction of the maximal response and a reduction of the apparent affinity for NMDA.

The lysophospholipid effect was not observed, or was much less marked, on other neurotransmitter receptors present in the same nucleated patches. A LPC concentration of 0.5 μm, which strongly inhibited the currents induced by 10 μm NMDA (46.3 ± 4.7 % residual current, n= 8) produced little or no effect on the currents induced by GABA, AMPA or kainate applied at concentrations slightly below their EC50 values. (Residual currents were 90.8 ± 1.1 % (n= 3), 81.7 ± 2.0 % (n= 3) and 98.1 ± 2.9 % (n= 4), respectively) (Fig. 3).

Figure 3
Receptor selectivity of the lysophospholipid effect

Lysophospholipid inhibition of NMDA responses of recombinant receptors expressed in HEK cells

To evaluate the effects of lysophospholipids on recombinant NMDA receptors composed of known subunits, we expressed various combinations of wild-type and mutated subunits in HEK 293 cells (Fig. 4). In all cells tested in the whole-cell configuration, LPI caused an inhibition similar to that produced in nucleated patches taken from cultured neurons. The residual current recorded after application of LPI at a concentration of 2 μm was 30.8 ± 2.6 % (n= 14) for NR1-1a + NR2A, 32.4 ± 8.2 % (n= 3) for NR1-1b + NR2A, 24.1 ± 1.3 % (n= 2) for NR1-1a + NR2B and 37.9 ± 4.6 (n= 7) for NR1-1a + NR2C. We also tested the effect of LPI on two receptors associating NR2A and a NR1-1a subunit with a mutation in the asparagine 598 position, which is well known for its role in permeation (see Hollmann & Heinemann, 1994) and which has also been found to be critical in the expression of various modulations of NMDA gating (e.g. Burgess et al. 1996; Schneggenburger & Ascher, 1997). In the two mutants used (in which the asparagine 598 of NR1 had been replaced either by an arginine (N598R) or a glutamine (N598Q)), the inhibition produced by LPI was similar (33.9 % and 26.2 %, respectively) to that found in wild-type receptors. When tested on receptors showing glycine-independent desensitization (every tested combination except NR1 + NR2C), the addition of LPI decreased the peak-to-steady state ratio, resulting in stronger inhibitions at the peak than in the steady state.

Figure 4
Lysophosphatidylinositol inhibition of whole-cell NMDA currents

The effect of LPI (2 μm) was tested on the responses to different NMDA concentrations (Fig. 5A). When tested on recombinant receptors formed upon co-expression of NR1 and NR2A subunits, it was found that LPI produced stronger effects on responses to non-saturating doses of agonist (64.1 ± 5.1 % inhibition at 10 μm NMDA, 36.6 ± 7.6 % inhibition at 300 μm NMDA, n= 8). This is in agreement with the observations made in native receptors from cultured neurons shown in Fig. 2B, and was expected from the fact that NMDA receptors from the same cultures show the typical properties of NR1 + NR2A and/or 2B (Paoletti & Ascher, 1994; Paoletti et al. 1995, 1997). In contrast, when applied on (non-desensitizing) NR1 + NR2C receptors, LPI produced a similar degree of inhibition at both low and high concentrations of NMDA (56.8 ± 4.5 % inhibition at 10 μm NMDA and 62.2 ± 5.4 % at 300 μm NMDA, n= 4; Fig. 5A).

Figure 5
Concentration dependence and time course of lysophospholipid effect on NMDA responses

The rate of onset and recovery from the lysophospholipid inhibition could be followed by applying a pulse of LPI on a steady background of NMDA and glycine. The whole-cell record illustrated in Fig. 5B was obtained from a cell transfected with NR1 and NR2A in which desensitization was small. Both the onset and recovery from the inhibition of the NMDA steady-state current by LPI (2 μm) could be fitted by single exponential functions of respective time constants τon= 2.9 ± 0.3 s and τoff= 9.0 ± 1.3 s (n= 6).

The effect of lysophospholipids on excised patches depends on the side of application

Whether the effect of LPI on NMDA responses depended upon the face of the membrane to which it was applied was tested on excised patches (Fig. 6). Bath-applied LPI inhibited NMDA currents in outside-out patches, and the extent of the inhibition was similar to that observed in nucleated patches or whole-cell recordings. Analysis of single-channel openings induced by 10 μm NMDA showed that 2 and 5 μm LPI reduced the open probability to 57.8 ± 6.6 % (n= 4) and 33.3 ± 6.8 % (n= 3) of the control values, respectively. Single-channel conductance was not altered (chord conductances at −50 mV, 49.6 ± 1.2 pS in control conditions and 49.0 ± 1.4 pS in the presence of 2 μm LPI). The results were identical whether the outside-out patches were obtained from neurons, HEK cells or oocytes. In contrast, LPI produced no change in NMDA receptor activity in any of the five inside-out patches tested, even when the concentration was raised to 5 μm (96.9 ± 3.4 % of the control).

Figure 6
Effect of lysophospholipids on excised patches

Occlusion between the inhibition produced by lysophospholipids and that produced by mechanical compression

Paoletti & Ascher (1994) showed that in nucleated patches the NMDA response can be reduced both by suction applied via the recording pipette and by application of a hyperosmotic solution. Inverse effects are observed upon application of a hyposmotic solution or application of a positive pressure in the pipette. All four effects were interpreted on the basis of a mechanical deformation of the membrane: the potentiation of the NMDA response was assumed to result from a stretch imposed on the receptor whereas the inhibitory effects were assumed to reflect the reduction of a ‘resting’ stretch or a mechanical compression. If we now assume that the inhibitory effects of lysophospholipids are due to the fact that, after entering the lipid bilayer they mimic the effects of a compression, one predicts that a strong inhibition produced by the application of lysophospholipids should reduce the inhibitory effect of a suction or of a hyperosmotic solution.

The sensitivity of the NMDA response to a suction applied through the patch pipette was tested on nucleated patches before and after application of LPI at a concentration (5 μm) producing a maximal inhibition (Fig. 7A). Under control conditions (no added lysophospholipid), application of negative pressures of −70 to −100 mbar resulted in a 55.5 ± 5.3 % decrease in the size of the NMDA current (n= 3). During continuous perfusion of 5 μm LPI (which by itself inhibited 86.0 ± 5.2 % of the current), application of the same negative pressure produced only a slight decrease (11.1 ± 10.1 %) of the residual NMDA current.

Figure 7
Occlusion between lysophospholipid inhibition and mechanical inhibition

In another series of nucleated patches, the effect of 2 μm LPI was tested at different extracellular osmolalities (Fig. 7B). LPI inhibition of NMDA currents was reduced from 69.8 ± 4.6 % in an isotonic extracellular solution to 36.4 ± 12.4 % in a hyperosmotic (+100 mm mannitol) solution which by itself reduced NMDA currents by 39.2 ± 2.7 % (n= 4).

These results indicate that there is indeed occlusion between the inhibitions of the NMDA responses produced by mechanical forces and those produced by lysophospholipids, as predicted by the assumption that lysophospholipids induce a change in membrane tension (Lundbaek & Andersen, 1994; and see Discussion).

Arachidonic acid does not exert its potentiating effect on NMDA receptors via its putative fatty acid binding domain

Arachidonic acid (Miller et al. 1992) and docosahexaenoic acid (Nishikawa et al. 1994) both potentiate NMDA responses, and the analogy between these potentiations and that produced by stretch was already noted by Paoletti & Ascher (1994): all three potentiations involve an increased apparent affinity for NMDA, an increase in the maximal response, and an increase in the glycine-insensitive desensitization. It was actually this analogy which was at the origin of the comparison between the effects of amphipathic compounds and of mechanical deformation. The results obtained with lysophospholipids extend the analogy: again the changes concerned apparent affinity, maximal current and desensitization. We further extended the analogy by distinguishing the actions of arachidonic acid when applied to the internal and external faces of the membrane. In outside-out patches or nucleated patches obtained from cultured neurons, arachidonic acid induced a ‘classical’ (Miller et al. 1992) potentiation of the NMDA response (Fig. 8A). This effect is due to an increase in open probability (NPo= 297 ± 29 % of control, n= 4) with no significant changes in single-channel conductance (chord conductances at −50 mV, 51.1 ± 0.3 pS in control conditions and 52.4 ± 0.6 pS in the presence of 20 μm arachidonic acid). Application of the same concentration of arachidonic acid to inside-out patches did not affect channel behaviour in four out of five patches tested. In the fifth one, however, there was a marked increase in the open probability. Overall the results indicate that arachidonic acid as well as lysophospholipids act preferentially when applied in the extracellular solution.

Figure 8
Arachidonic acid effect

The fact that the effect of arachidonic acid is the inverse of that of lysophospholipids can be explained if the two types of compounds insert in the lipid bilayer and produce opposite pressure changes on the channel. However, it has been suggested on the basis of sequence homology that arachidonic acid binds to a putative fatty acid binding domain in NR1, at a position suggesting contact with the extracellular solution rather than with the lipid bilayer (Petrou et al. 1993). We have tested the effect of arachidonic acid on a series of receptors in which the most conserved residues of that domain had been mutated. As shown in Fig. 8B, mutation of the NR1 residues phenylalanine 303 into leucine, arginine 372 into threonine and tyrosine 374 into serine did not affect arachidonic acid potentiation of NMDA responses.

Linkage to the cytoskeleton is not required for NMDA receptor mechanosensitivity

The experiments described above support the hypothesis that the transmission of force onto the NMDA receptor protein involves the lipid bilayer, but does not exclude a role of the cytoskeleton in this transmission. In an attempt to evaluate such a role we expressed in HEK cells different combinations of NMDA receptor subunits which differed by the sequence and the length of their cytoplasmic C-terminal, known to be the link between the NMDA receptor and intracellular proteins (Kornau et al. 1995, 1997; Wyszynski et al. 1997). This involved both various combinations of ‘wild-type’ subunits with different C-terminals and receptors incorporating mutated subunits in which deletions of variable length had severed the NR1 (NR1stop833) or NR2A (NR2Astop1082 and stop1438) C-terminal link with intracellular proteins. Since no detectable currents were observed when co-expressing both C-terminal truncated NR1 and NR2 subunits, we studied receptors where either the NR1 or NR2 C-terminus had been severed. We tested the mechanosensitivity of the various NMDA receptors in outside-out or cell-attached patches by measuring changes in the open probability of channels in response to pressure application through the pipette (see Paoletti & Ascher, 1994) (Fig. 9). All the receptors tested showed stretch sensitivity, regardless of the splice variant of wild-type or mutated NR1 or NR2 subunit expressed. NMDA receptors co-expressed with postsynaptic density protein PSD-95 (Kornau et al. 1995) were also found to be stretch sensitive.

Figure 9
Stretch sensitivity of recombinant NMDA receptors

We also tested the sensitivity to lysophospholipids of mutants lacking parts of the cytoplasmic C-terminus tail. Receptors formed by combining NR1-1a with a mutated NR2A subunit in which a stop codon had been inserted at residue number 844 (NR2AΔC) showed responses similar to those of wild-type receptors (Köhr & Seeburg, 1996). LPI inhibition in these receptors (26.8 ± 3.1 % of the control current at 2 μm LPI, n= 5) was similar to that found in the wild-type.

These results reinforce the conclusion that the strong cytoskeletal link of the NMDA receptor plays little role in the transmission of mechanical forces to the receptor, and is more probably involved in the local organization of receptor aggregates at the synaptic level (see Sheng, 1996; Kornau et al. 1997). However, experiments of pressure application through the patch pipette are not quantitatively comparable from patch to patch and do not allow the evaluation of whether the cytoskeletal link of the NMDA receptor could modulate its mechanosensitivity.


Lysophospholipids inhibit NMDA currents

The inhibition of NMDA currents by lysophospholipids is fully reversible and voltage independent. Lysophospholipids having different polar head groups and different electrical charges show roughly the same capacity to inhibit NMDA receptors, while structurally related compounds such as phosphatidylcholine or lysophosphatidic acid are not active.

The action of lysophospholipids seems to require their insertion into the lipid bilayer. The slow time course of their effect probably reflects the time required for their passage from the extracellular solution into the membrane and back. The apparent affinity of lysophospholipids is relatively high compared with that measured for some of their other effects (on sodium channels, Burnashev et al. 1989; on membrane fusion, Chernomordik et al. 1993), but a precise comparison is difficult, since the scarce information on the partition coefficients of these compounds between aqueous and membrane phases does not allow a precise calculation of the lysophospholipid molar fraction in membranes under our experimental conditions.

In receptors assembled from NR1 and NR2C, the inhibition produced by lysophospholipids was the same at low and high NMDA concentrations, suggesting a non-competitive inhibition in which the affinity of NMDA was not affected. In receptors assembled from NR1 and NR2A, but also in the native receptors of cultured neurons, the reduction of the maximal response was associated with an increase of the EC50 (the inhibition was larger at low concentrations of agonist), and the glycine-insensitive desensitization, which is not seen in NR1 + NR2C, was reduced. It seems difficult to interpret all these effects by a single change in the kinetics of NMDA responses, but it is worth noting that all of them are opposite to those produced by arachidonic acid.

The effects of lysophospholipids and compression are opposite to those of arachidonic acid and stretch

We have provided two main sets of arguments in favour of a key role of the bilayer in the transmission of membrane tension or deformation to the NMDA receptor.

Firstly, the effects of lysophospholipids and compression are similar and tend to decrease the NMDA responses; they are opposite in sign to those of arachidonic acid and stretch, which tend to increase the responses. The parameters altered by the four manipulations (lysophospholipids, arachidonic acid, stretch, compression) are the same: apparent agonist affinity, maximal response, desensitization. The suggestion that the four effects involve a common pathway is reinforced by the fact that there is occlusion between lysophospholipids and compression. The data are compatible with the hypothesis that the lysophospholipids and arachidonic acid both modify membrane tension by inserting into the lipid bilayer (Martinac et al. 1990).

Secondly, deletion of the C-terminal residues interacting with different cytoplasmic elements (Kornau et al. 1995; Wyszynski et al. 1997) does not suppress the modulation of the receptors by membrane tension or lysophospholipids. Although potentially unknown targeting mechanisms could maintain cytoskeletal coupling, this is unlikely to be responsible for mechanosensitivity a long time after excision of the patches.

The suggestion that lipophilic compounds act by binding to an extracellular ‘fatty acid binding site’ on NMDA receptors (Petrou et al. 1993) would not explain the slowness of the onset and offset of the effect of the various lipophilic compounds. For an affinity in the micromolar range as calculated from the dose-response curve, and a diffusion coefficient in the range of 107 to 108 M−1 s−1, a readily accessible site would imply an off rate between 10 and 100 ms. Measured off rates are close to 10 s for lysophospholipids and in the order of minutes for arachidonic acid (see Miller et al. 1992). This can only be explained by introducing a limitation to the accessibility to the ‘site’, such as the partitioning of these compounds between aqueous and lipidic phases. Furthermore, the hypothesis of extracellular binding is made very unlikely by the persistence of the sensitivity to arachidonic acid of receptors in which the putative fatty acid binding site has been eliminated by mutation of its most conserved residues.

As indicated in the introduction, the interpretation of the mechanical effects of amphipathic compounds has in the past used mostly two theories. In the first, amphipaths can enter from both sides of the bilayer, but membrane potential induces an asymmetrical distribution of charged amphipaths across the two hemilayers, leading to either crenation or cup formation (Sheetz & Singer, 1974; Martinac et al. 1990). In the other theory, the amphipaths distribute mainly in one hemilayer, which is deformed in opposite ways by molecules having the shape of cones or inverted cones (Israelachvili et al. 1977; Cullis & de Kruijff, 1979; Lundbaek & Andersen, 1994). Our experiments do not support the bilayer couple theory predictions on the role of charge: negatively charged and neutral lysophospholipids produced similar effects, and negatively charged arachidonic acid produced effects opposite to those of negatively charged LPI. Our data also do not support the hypothesis that amphipathic compounds could enter the bilayer from both sides and accumulate in one half-layer independently of the side on which they had been applied. We provide evidence for a dependence of the effects on the face of application by showing that the effects of LPI and arachidonic acid (with one exception) could not be observed in inside-out patches if the compounds were applied on the internal side of the membrane. This is in agreement with the reported low rate of lysophospholipid flip-flop movement across the membrane (Cullis & de Kruijff, 1979) although we cannot exclude the possibility of different partition coefficients between the aqueous phase and the two faces of the lipid bilayer.

On the other hand, the data are well explained by assuming a key role of the shape of compounds inserted into the outer hemilayer. Lysophospholipids, with their ‘cone’ shape, would tend to compress integral membrane proteins at the polar region of the phospholipid-protein interface. In contrast, membrane tension at this interface is expected to be created by arachidonic and other polyunsaturated fatty acids in which double bonds make the cross-section of the hydrophobic region larger than that of the polar carboxyl head. Thus the opposite effects of lysophospholipids and polyunsaturated fatty acids on NMDA receptors would result from the fact that they tend to produce opposite pressure on the channel at the level of the outer leaflet of the membrane.

Physiological role of NMDA receptor sensitivity to lysophospholipids and arachidonic acid

Lysophospholipids are naturally produced by phospholipases A2 (PLA2), a family containing both cytosolic and secreted enzymes involved in signal transduction (for review, see Dennis, 1997). Similarly, arachidonic acid has been shown to be an important cellular messenger in a wide variety of cell types. It modulates the activity of various types of ion channels and some of its effects have been observed in excised patches, suggesting a direct effect not involving arachidonate metabolism (for review, see Meves, 1994). From a physiological point of view, a number of transmitters and hormones produce both arachidonic acid and lysophospholipids as second messengers and this could be seen as the source of two potentially antagonistic effects on NMDA receptors. However, the fact that the effects of both types of compounds are mostly seen when they are applied on the extracellular face of the membrane makes it unlikely that the PLA2 cytoplasmic activity of a neuron has a marked effect on the NMDA receptors of the same cell. If any, the physiological modulation of NMDA receptors by these compounds is likely to involve either an extracellular production by secreted PLA2 (Lauritzen et al. 1994), or a transmembrane diffusion from the cell producing the compound to the neuron bearing the receptors (as has been proposed for arachidonic acid; see Meves, 1994). The place of synthesis, the differential diffusion across the membranes and the different rates of degradation by other enzymatic activities will determine the differential effects of lysophospholipids and arachidonic acid on NMDA receptors in vivo.


This work was supported by the Centre National de la Recherche Scientifique (URA 1857). M. C. was the recipient of an EMBO long-term fellowship. We thank Boris Barbour and Jacques Neyton for comments on the manuscript and Dani Lévy for culturing the neurons.


  • Andersen OS, Lundbæk JA, Girshman J. Channel function and channel-lipid bilayer interactions. In: Mosekilde E, Mouritsen OG, editors. Dynamical Phenomena in Living Systems. Berlin: Springer; 1995. pp. 131–151.
  • Ascher P, Bregestovski P, Nowak L. N-methyl-D-aspartate-activated channels of mouse central neurones in magnesium free solutions. The Journal of Physiology. 1988;399:207–226. [PMC free article] [PubMed]
  • Burgess MF, Powers J, Traynelis SF. Asparagine 616 in the NR1 subunit controls spermine block, spermine potentiation and pH sensitivity. Society for Neuroscience Abstracts. 1996;22:236.2.
  • Burnashev NA, Udrovinas AI, Fleidervish IA, Rosenshtraukh LV. Ischemic poison lysophosphatidyl-choline modifies heart sodium channels gating introducing long-lasting bursts of openings. Pflügers Archiv. 1989;415:124–126. [PubMed]
  • Chen C, Okayama H. High-efficiency transformation of mammalian cells by plasmid DNA. Molecular and Cellular Biology. 1987;7:2745–2752. [PMC free article] [PubMed]
  • Chernomordik LV, Melikyan GB, Chizmadzhev YA. Biomembrane fusion: a new concept derived from model studies using two interacting planar lipid bilayers. Biochimica et Biophysica Acta. 1987;906:309–352. [PubMed]
  • Chernomordik LV, Vogel SS, Sokoloff A, Onaran HO, Leikina EA, Zimmerberg J. Lysolipids reversibly inhibit Ca2+-, GTP-, and pH-dependent fusion of biological membranes. FEBS Letters. 1993;318:71–76. [PubMed]
  • Cullis PR, De Kruijff B. Lipid polymorphism and the functional roles of lipids in biological membranes. Biochimica et Biophysica Acta. 1979;559:399–420. [PubMed]
  • Dennis EA. The growing phospholipase A2 superfamily of signal transduction enzymes. Trends in Biochemical Sciences. 1997;22:1–2. [PubMed]
  • Eddlestone GT, Ciani S. Lysophospholipids modulate the K(ATP) channel. Biophysical Journal. 1991;59:16a. (abstract)
  • Guharay F, Sachs F. Stretch-activated single ion channel currents in tissue-cultured embryonic chick skeletal muscle. The Journal of Physiology. 1984;352:685–701. [PMC free article] [PubMed]
  • Hamill OP, Marty A, Neher E, Sakmann B, Sigworth FJ. Improved patch-clamp techniques for high-resolution current recording from cells and cell-free membrane patches. Pflügers Archiv. 1981;391:85–100. [PubMed]
  • Hollmann M, Heinemann S. Cloned glutamate receptors. Annual Review of Neuroscience. 1994;17:31–108. [PubMed]
  • Ismailov II, Berdiev BK, Shlyonsky VG, Benos DJ. Mechanosensitivity of an epithelial Na+ channel in planar lipid bilayers: release from Ca2+ block. Biophysical Journal. 1997;72:1182–1192. [PMC free article] [PubMed]
  • Israelachvili JN, Mitchell DJ, Ninham BW. Theory of self-assembly of lipid bilayers and vesicles. Biochimica et Biophysica Acta. 1977;470:185–201. [PubMed]
  • Köhr G, Seeburg PH. Subtype-specific regulation of recombinant NMDA receptor-channels by protein tyrosine kinases of the src family. The Journal of Physiology. 1996;492:445–452. [PMC free article] [PubMed]
  • Kornau H-C, Schenker LT, Kennedy MB, Seeburg PH. Domain interaction between NMDA receptor subunits and the postsynaptic density protein PSD-95. Science. 1995;269:1737–1740. [PubMed]
  • Kornau H-C, Seeburg PH, Kennedy MB. Interaction of ion channels and receptors with PDZ domain proteins. Current Opinion in Neurobiology. 1997;7:368–373. [PubMed]
  • Kupper J, Ascher P, Neyton J. Probing the pore region of recombinant N-methyl-D-aspartate channels using internal and external magnesium block. Proceedings of the National Academy of Sciences of the USA. 1996;93:8648–8653. [PMC free article] [PubMed]
  • Lauritzen I, Heurteaux C, Ladzunski M. Expression of group II phospholipase A2 in rat brain after severe forebrain ischemia and in endotoxic shock. Brain Research. 1994;651:353–356. [PubMed]
  • Lundbæk JA, Andersen OS. Lysophospholipids modulate channel function by altering the mechanical properties of lipid bilayers. Journal of General Physiology. 1994;104:645–673. [PMC free article] [PubMed]
  • Lundbæk JA, Birn P, Girshman J, Hansen AJ, Andersen OS. Membrane stiffness and channel function. Biochemistry. 1996;35:3825–3830. [PubMed]
  • Markin VS, Martinac B. Mechanosensitive ion channels as reporters of bilayer expansion. Biophysical Journal. 1991;60:1120–1127. [PMC free article] [PubMed]
  • Marsh D, King D. Prediction of the micelle concentrations of mono- and di-acyl phospholipids. Chemistry and Physics of Lipids. 1986;42:271–277. [PubMed]
  • Martinac B, Adler J, Kung C. Mechanosensitive ion channels of E. coli activated by amphipaths. Nature. 1990;348:261–263. [PubMed]
  • Meves H. Modulation of ion channels by arachidonic acid. Progress in Neurobiology. 1994;43:175–186. [PubMed]
  • Miller B, Sarantis M, Traynelis SF, Attwell D. Potentiation of NMDA receptor currents by arachidonic acid. Nature. 1992;355:722–725. [PubMed]
  • Nishikawa M, Kimura S, Akaike N. Facilitatory effect of docosahexaenoic acid on N-methyl-D-aspartate response in pyramidal neurones of rat cerebral cortex. The Journal of Physiology. 1994;475:83–93. [PMC free article] [PubMed]
  • Opsahl LR, Webb WW. Transduction of membrane tension by the ion channel alamethicin. Biophysical Journal. 1994;66:71–74. [PMC free article] [PubMed]
  • Paoletti P, Ascher P. Mechanosensitivity of NMDA receptors in cultured mouse central neurons. Neuron. 1994;13:645–655. [PubMed]
  • Paoletti P, Ascher P, Neyton J. High-affinity zinc inhibition of NMDA NR1-NR2A receptors. Journal of Neuroscience. 1997;17:5711–5725. [PubMed]
  • Paoletti P, Neyton J, Ascher P. Glycine-independent and subunit-specific potentiation of NMDA responses by extracellular Mg2+ Neuron. 1995;15:1109–1120. [PubMed]
  • Pascher I, Lundmark M, Nyholm P-G, Sundell S. Crystal structures of membrane lipids. Biochimica et Biophysica Acta. 1992;1113:339–373. [PubMed]
  • Petrou S, Ordway RW, Hamilton JA, Walsh JV, Jr, Singer JJ. Structural requirements for charged lipid molecules to directly increase or suppress K+ channel activity in smooth muscle cells. Journal of General Physiology. 1994;103:471–486. [PMC free article] [PubMed]
  • Petrou S, Ordway RW, Singer JJ, Walsh JV., Jr A putative fatty acid-binding domain of the NMDA receptor. Trends in Biochemical Sciences. 1993;18:41–42. [PubMed]
  • Sather W, Dieudonné S, MacDonald J, Ascher P. Activation and desensitisation of N-methyl-D-aspartate receptors in nucleated outside-out patches from mouse neurons. The Journal of Physiology. 1992;450:643–672. [PMC free article] [PubMed]
  • Schneggenburger R, Ascher P. Coupling between gating and permeation in an NMDA channel pore mutant. Neuron. 1997;18:167–177. [PubMed]
  • Sheetz MP, Singer SJ. Biological membranes as bilayer couples. A molecular mechanism of drug-erythrocyte interactions. Proceedings of the National Academy of Sciences of the USA. 1974;71:4457–4461. [PMC free article] [PubMed]
  • Sheng M. PDZs and receptor/channel clustering: Rounding up the latest suspects. Neuron. 1996;17:575–578. [PubMed]
  • Sukharev SI, Blount P, Martinac B, Blattner FR, Kung C. A large-conductance mechanosensitive channel in E. coli encoded by mscL alone. Nature. 1994;368:265–268. [PubMed]
  • Wyszynski M, Lin J, Rao A, Nigh E, Beggs AH, Craig AM, Sheng M. Competitive binding of α-actinin and calmodulin to the NMDA receptor. Nature. 1997;385:439–442. [PubMed]

Articles from The Journal of Physiology are provided here courtesy of The Physiological Society
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...