Introduction

The reversible phosphorylation of proteins on serine, threonine and tyrosine residues represents a primary post-translational mechanism underlying the regulation of cell functions, including DNA replication, energy metabolism, cell growth and signalling (Thomas and Brugge, 1997). An abundance of data has demonstrated that Src family protein tyrosine kinases (PTKs) may play a central role in the regulation of N-methyl-D-aspartate (NMDA) receptors by multiple signalling pathways (Yu and Salter, 1998; Lu et al., 1999; Huang et al., 2001; Manzerra et al., 2001; Vissel et al., 2001). Recent studies have documented that Src family PTKs may potentiate NMDA receptor functions via both the enhancement of the channel gating and an increase in the number of NMDA receptors on the cell surface (Yu et al., 1997; Grosshans et al., 2001). However, it remains unknown how the activity of Src family PTKs is controlled in the regulation of NMDA receptors.

The Src kinase family is comprised of a total of nine members, at least five of which—Src, Fyn, Lyn, Lck and Yes—are known to be expressed in the central nervous system (CNS) (Thomas and Brugge, 1997; Ali and Salter, 2001). All members of the Src family contain highly homologous regions, including an inhibitory C-terminal sequence, as well as catalytic, SH2 and SH3 domains (Brown and Cooper, 1996). The activity of these PTKs is tightly regulated by the reversible phosphorylation of a C-terminal tyrosine residue (Tyr527 in chicken c-Src) (Cooper et al., 1986; Nada et al., 1991; Brown and Cooper, 1996). Although phosphatases have long been proposed to be responsible for the dephosphorylation of phosphorylated Tyr527, and consequently increase Src family PTK activity (Brown and Cooper, 1996), it has been demonstrated recently via inactivation of the protein tyrosine phosphatase (PTP) gene by homologous recombination (Ponniah et al., 1999; Su et al., 1999) that PTP may act as an endogenous activator of Src family PTKs by dephosphorylating phosphorylated Tyr527 (Zheng et al., 1992; den Hertog et al., 1993).

PTP belongs to the family of receptor-like PTPs (Neel and Tonks, 1997; den Hertog, 1999), which at present includes seven members discovered in mammalian cells (den Hertog, 1999). Most receptor-like PTPs contain two phosphatase domains, a single transmembrane region and a variable extracellular N-terminus (Neel and Tonks, 1997; den Hertog, 1999). PTP is highly expressed in the CNS (Sap et al., 1990; Van Vactor, 1998), but the function of this enzyme in the CNS is still poorly understood. To clarify functions of phosphatases in the CNS, we examined the role of PTP in the regulation of NMDA receptor activity.

Results

PTP complexes with NMDA receptors by directly binding to the PDZ2 domain of PSD95

In order to determine whether PTP is a critical factor in the regulation of NMDA receptors, we first set out to determine whether PTP is an integral component of the NMDA receptor-associated signalling complex in the CNS. We performed co-immunoprecipitation experiments using antibodies recognizing either the NMDA receptor NR1 subunit (which is present in all functional NMDA receptors; Dingledine et al., 1999; Cull-Candy et al., 2001), PTP or Src from membrane protein extracts of adult rat brain solubilized under non-denaturing conditions. We found that antibodies recognizing either the NMDA receptor NR1 subunit, PTP or Src could co-precipitate the other two remaining proteins (Figure 1A−C). In contrast, none of these proteins were detected in immunoprecipitates using non-specific IgG (Figure 1A−C). These data indicate that PTP, Src and the NMDA receptor may associate in the CNS.

Recently, mechanisms underlying the association of Src family PTKs, Src and Fyn, with NMDA receptors have been investigated in depth by other groups (Tezuka et al., 1999; Hajdur et al., 2001). It is found that Src and Fyn may bind directly to the PDZ domain-containing protein, post-synaptic density 95 (PSD95), and thereby associate with NMDA receptors (Tezuka et al., 1999; Hajdur et al., 2001). In order to get an indication as to how the Src family PTK activator PTP associates with NMDA receptors, we examined whether PSD95 may also be involved in the association of PTP with NMDA receptors. For this purpose, we first investigated whether the association of PTP with NMDA receptors was affected by a PSD95 depletion. PSD95 immunoprecipitation was conducted twice to deplete PSD95 from the rat brain lysates, after which very little (<3%) PSD95 was found remaining in the supernatant (Figure 1D). The left blots of Figure 1E show an example of PSD95 depletion experiments. PSD95 antibodies co-precipitated both the NMDA NR1 subunit and PTP (the right lane of the left blots). After PSD95 immunoprecipitation, the supernatant was subjected to immunoprecipitation with anti-NR1 antibodies. The resultant precipitates were loaded in the middle lane of the gel. Compared with the amount of PTP co-precipitated with NR1 antibodies from rat brain lysates without the initial PSD95 immunodepletion (the left lane of the left blots in Figure 1E), there was much less PTP detected in NR1 immunoprecipitates after the PSD95 depletion (the middle lane of the left blots in Figure 1E). As a control, immunoprecipitation performed twice with non-specific IgG produced no significant change in PTP association with NMDA receptors (the right blots of Figure 1E). The fact that the ratio of band intensity of co-precipitated PTP versus that of NR1 subunit protein detected in NR1 immunoprecipitates was significantly reduced by PSD95 depletion (Figure 1F) strongly suggests that PTP may associate mainly with those NMDA receptors that are also associated with PSD95.

It is known that the PDZ1 and PDZ2 domains of PSD95 may bind to the C-terminus of NMDA NR2 subunits (Kornau et al., 1995; Sheng and Sala, 2001). To identify whether PSD95 may act as a linker to recruit PTP into the NMDA receptor complex, we examined whether PSD95 expression is sufficient to enable PTP−NMDA receptor association in non-neuronal cells that normally express neither NMDA receptors nor PSD95. HEK293 and COS7 cells were transfected with cDNAs encoding NMDA NR1-1a, NR2A subunits (which are found in most NMDA receptors of adult animals; Dingledine et al., 1999; Cull-Candy et al., 2001) and/or PSD95. Co-immunoprecipitation experiments were performed on cell lysates using an antibody against PTP. We found that expressed PSD95 was co-precipitated with PTP from the lysate of cells lacking NMDA receptor subunit co-expression (Figure 2A). In contrast, PTP antibodies could not precipitate overexpressed NMDA receptor subunit proteins without PSD95 co-expression (Figure 2A). Moreover, without NMDA NR2A subunit co-expression, PTP failed to be co-precipitated with NMDA NR1-1a subunit, even if PSD95 was co-expressed (Figure 2A). Thus, PTP may associate with a functional NMDA receptor by way of PSD95 association with the NR2A subunit.

The data above led us to infer that PTP may bind to PSD95, leading to its association with NMDA receptors. To verify this hypothesis, we directed our investigation towards determining how PTP interacts with PSD95. cDNAs encoding PSD95 mutants, in which a segment corresponding to either PDZ1, 2, 3 or the SH3 and/or GK domain of PSD95 was deleted (Tezuka et al., 1999), were transfected into HEK293 or COS7 cells. The left blot in Figure 2B shows the expression of these mutants in HEK293 cells by immunoblotting with an antibody against PSD95 residues 1−54 (anti-PSD95₁₋₅₄). We found that all of the expressed PSD95 mutants except for one lacking residues 157−256 (corresponding to the PDZ2 domain of PSD95) could be co-immunoprecipitated by anti-PTP antibodies (Figure 2B). These data suggest that PTP association with PSD95 may be due to the binding of PTP to the PDZ2 domain of PSD95. We substantiated this finding by performing in vitro GST fusion protein precipitation assays. Figure 3A shows the constructs of PTP peptides used in these assays. Figure 3B shows that ³⁵S-labelled peptides produced by in vitro transcription/translation corresponding to the entire cytoplasmic portion (D1 + D2) and the membrane-distal phosphatase domain including the C-terminal (D2), but not the membrane-proximal phosphatase domain (D1) of PTP, could be precipitated by glutathione−agarose beads bound to GST fusion proteins containing the PSD95 PDZ2 domain. In contrast, neither GST alone nor the GST fusion proteins containing PSD95 PDZ1 or the PDZ3 domain could precipitate any of these ³⁵S-labelled peptides (Figure 3B). Thus, the PTP D2 domain appears to be involved in the interaction between PTP and PSD95.

Since there is no typical E(S/T)XV motif at the C-terminus of PTP available for PSD95 PDZ2 domain binding, and recent studies have documented that the PDZ domain of PSD95 may also bind to the internal peptides of proteins lacking the typical E(S/T)XV terminal motif (Hillier et al., 1999; Tezuka et al., 1999; Sheng and Sala, 2001), we sought to identify further the sequence which may underlie the binding between the PTP D2 and PSD95 PDZ2 domains. By using a D2 domain that was truncated from its N- and/or C-terminus, we found that a GST fusion protein containing the amino acids 539−711 of PTP could still bind to the PDZ2 domain of PSD95 in vitro (Figure 3B). Thus, we conclude that PTP binds to PSD95 via the D2−PDZ2 domain interaction, and thereby complexes with NMDA receptors.

In non-neuronal cells, it has been found that Src (Harder et al., 1998; Zheng et al., 2000) and Fyn (Bhandari et al., 1998) may complex with PTP. To identify the role of PTP in the regulation of NMDA receptor activity, we then investigated whether PTP expression changes Src expression or the physical association of this kinase with NMDA receptors. cDNAs encoding NMDA receptor subunits and wild-type PSD95 were transfected into PTP-deficient (PTP^-/-; Su et al., 1999) fibroblasts. Figure 4A shows the results of co-immunoprecipitation experiments using antibodies against NR2A/B subunits on lysates of these cells with or without the re-introduction of PTP by transfection of PTP cDNA into PTP^-/- fibroblasts. Compared with that observed in the fibroblasts without the re-introduction of PTP, it was found that PTP expression did not alter the ability of PSD95 or Src to associate with co-expressed NMDA receptors (Figure 4A). Western blot analysis using antibodies against Src (clone327), or C-terminally tyrosine-phosphorylated Src (Src-pTyr529), showed no significant difference in the total amount of Src in cells with PTP expression versus those cells lacking PTP, while a marked reduction in anti-Src-pTyr529 immunoreactivity, indicative of increased Src activity (Zheng et al., 1992; den Hertog et al., 1993; Ponniah et al., 1999; Su et al., 1999; Zheng et al., 2000), was found in the cells into which PTP was re-introduced (Figure 4B and C).

PTP activity is necessary for initiating and maintaining the regulation of recombinant NMDA receptors by endogenous Src family PTKs in fibroblasts

To determine the role of PTP in the regulation of NMDA receptor functions, we recorded whole-cell currents mediated by the NMDA NR1-1a/NR2A receptor expressed in PTP^-/- fibroblasts with or without the re-introduction of the phosphatase (Figure 5). In all of the patch clamp recording experiments conducted in fibroblasts, cDNA encoding wild-type PSD95 was co-transfected. Whole-cell currents were evoked with L-aspartate or NMDA (250 M) applied through a double-barrel pipette system. The averaged peak and steady-state amplitudes of whole-cell currents recorded in PTP^-/- cells were 476 69 and 404 52 pA, respectively (n = 28, mean SEM). The decay of whole-cell currents during the agonist application was fitted using two exponential components with time constants 185 40 ms (_fast) and 1217 147 ms (_slow), respectively. Compared with the currents recorded in PTP^-/- cells, the re-introduction of PTP into PTP^-/- cells significantly increased the amplitude of currents mediated by the recombinant NMDA receptors (peak, 839 148 pA; steady state, 663 113 pA; n = 38; also see Figure 5A), but did not significantly change the decay time (_fast, 129 14 ms; _slow, 1094 146 ms). To clarify whether the effects of the re-introduction of PTP into PTP^-/- cells on recombinant NMDA receptors resulted from the direct modulation of NMDA receptors by PTP, we transfected cDNAs encoding NMDA NR1-1a, NR2A subunits and wild-type PSD95 into fibroblasts lacking PTKs Src, Fyn and Yes (SYF cells; Klinghoffer et al., 1999) with or without PTP cDNA co-transfection, and recorded currents evoked by L-aspartate or NMDA. The amplitudes of the peak and steady-state whole-cell currents mediated by NMDA receptors expressed in SYF cells with (n = 39) and without (n = 20) overexpression of PTP were 674 66 and 458 55 pA, and 758 95 and 425 83 pA, respectively. No statistically significant difference could be found. Thus, it is implied that the enhancement of NMDA receptor-mediated responses in cells in which PTP was re-introduced may be a result of PTP-dependent activation of Src family PTKs.

Then, we examined the effects of the Src family PTK inhibitor, PP2 (0.5−10 M), on NMDA NR1-1a/NR2A receptors co-expressed in cells in which PTP was re-introduced. An example presenting the effect of PP2 applied to cells with the re-introduction of PTP is shown in Figure 5B. We found that PP2 application significantly reduced NMDA NR1-1a/NR2A receptor-mediated whole-cell currents (Figure 5B and C) without changing the reversal potential of currents recorded (Figure 5B). These data indicate that the PP2-induced reduction of current amplitudes recorded in cells in which PTP was re-introduced results from a decrease in NMDA receptor-mediated whole-cell conductance. The PP2 effect was concentration dependent (Figure 5C), while PP3 (an inactive PP2 isomer) had no effect (Figure 5C). To confirm that the PP2 effects detected in these experiments were produced by the inhibition of Src family PTKs, we examined the effects of PP2 on NMDA NR1-1a/NR2A receptors expressed in SYF cells. We found that PP2 application had no effect on the recombinant NMDA receptors expressed in SYF cells unless Src was re-introduced (Figure 5D). This indicates that the recombinant NMDA NR1-1a/NR2A receptors in cells expressing PTP may be constitutively regulated by endogenous Src, Fyn and/or Yes.

Surprisingly, PP2 (10 M) application did not produce any inhibition of recombinant NMDA receptors expressed in PTP^-/- cells (Figure 5E). To determine whether the lack of PP2-induced inhibition was related to the absence of PTP phosphatase activity, we investigated the effects of PP2 on NMDA receptors expressed in PTP^-/- cells co-transfected with cDNA encoding catalytically inactive PTP (PTP with a cysteine to alanine mutation at residue 433 in the D1 phosphatase domain) (Harder et al., 1998). PP2 had no effect on NMDA receptors expressed in these cells (Figure 5E), suggesting that PTP activity may be necessary for initiating and maintaining the NMDA receptor regulation by Src family PTKs.

To confirm the role of PTP activity in the regulation of NMDA receptors by endogenous PTKs, we examined the effects of another PTK inhibitor, lavendustin A (10 M) (Wang and Salter, 1994; Hemmings, 1997). We found that similarly to PP2, lavendustin A application did not induce any significant reduction of recombinant NMDA receptor-mediated currents recorded from cells without PTP expression or expressing catalytically inactive PTP (the peak and steady-state amplitudes during lavendustin A application were 100 3 and 98 4% of control, n = 8, or 92 6 and 92 4% of control, n = 7). In contrast, lavendustin A significantly reduced NMDA receptor-mediated responses (the peak and steady-state amplitudes were 72 4 and 73 5% of control, n = 7, P < 0.05, Wilcoxon test) without changing the reversal potential of currents in cells in which PTP was re-introduced (data not shown). Lavendustin B, an inactive isomer of lavendustin A, had no such effects (the peak and steady-state amplitudes during lavendustin B application were 97 4 and 87 3% of control, n = 6). These data provide strong support for the concept that PTP may govern the activity of endogenous PTKs in the regulation of NMDA receptors.

Blockade of the PTP catalytic (D1) domain inhibits NMDA receptor activity and turns off the regulation of NMDA receptors by Src family PTKs in hippocampal neurones

PTP, like the majority of receptor-like PTPs, contains two cytoplasmic phosphatase domains, with the D1 domain displaying most of the enzyme's catalytic activity (Lim et al., 1998; Blanchetot and den Hertog, 2000). We found that application of a monoclonal antibody (clone21, 2.5 g/ml, mouse IgM; Transduction Lab, Lexington, KY), raised against the D1 domain of PTP, into fibroblasts expressing PTP significantly inhibited the activity of NMDA receptors expressed in these cells (Figure 6A−C). In contrast, application of the antibody into PTP^-/- fibroblasts, or application of non-specific mouse IgM (an isotype control of the antibody, clone21) into PTP-expressing fibroblasts had no effect on NMDA receptor activity (Figure 6C). Thus, these data demonstrate that the effect of the anti-PTP D1 antibody on NMDA receptors in PTP-expressing cells is produced by specific blockade of PTP.

In order to identify the function of PTP in central neurones, we investigated the effects of clone21 on NMDA receptor activity in cultured hippocampal neurones. We found that application of clone21 (2.5 g/ml) into neurones through recording pipettes reduced NMDA receptor-mediated whole-cell currents (Figure 6D and E). The average current amplitude recorded at 20 min after breakthrough was reduced to 52 5% of the first response (Figure 6H) with no change in the reversal potential (Figure 6F). Application of an equal concentration of non-specific mouse IgM into neurones did not reduce NMDA receptor-mediated currents (Figure 6G and H). Furthermore, we found that the blockade of the D1 phosphatase domain of PTP by pre-loading clone21 into neurones abolished the effect of PP2 on NMDA receptor activity (Figure 6J), while PP2 significantly inhibited NMDA currents in neurones intracellularly pre-loaded with non-specific IgM (Figure 6J). PP2 application inhibited NMDA receptor-mediated whole-cell responses recorded from hippocampal neurones without clone21 pre-application (Figure 6I and J), while application of PP3 produced no effect (Figure 6J). These findings further suggest that endogenous PTP may be an up-regulator of NMDA receptors and govern the regulation of NMDA receptors by Src family PTKs in central neurones.

PTP is a novel up-regulator of synaptic NMDA receptors and plays an important role in the induction of synaptic long-term potentiation (LTP) in CA1 hippocampal neurones

In light of the findings that PTP may bind directly to PSD95 (Figures 2 and 3), and that the blockade of the main catalytic domain of PTP may down-regulate NMDA receptor-mediated channel activity in neurones (Figure 6), it would be logical to expect that PTP may be very important in the regulation of excitatory synaptic transmission, namely the up- or down-regulation of PTP activity may enhance or reduce synaptic responses mediated by glutamate. Previous studies have demonstrated that the cytoplasmic portion, (D1 + D2), of PTP represents the functional domain of the enzyme that regulates the phosphatase activity, protein targeting and interaction (Zheng et al., 1992, 2000; den Hertog et al., 1993; Neel and Tonks, 1997; Harder et al., 1998; den Hertog, 1999). To obtain direct evidence showing the function of PTP in central neurones, we examined the effects of application of GST fusion protein containing (D1 + D2) into neurones on glutamate-mediated synaptic responses.

We recorded miniature excitatory post-synaptic currents (mEPSCs) in cultured hippocampal neurones. Consistent with previous findings (Bekkers and Stevens, 1989; Yu et al., 1997; Yu and Salter, 1998), most of the recorded mEPSCs in hippocampal neurones consisted of two components: one with a fast decay, which could be blocked by bath application of the non-NMDA receptor antagonist DNQX (5 M; data not shown), the other one with a slow decay, which could be blocked by bath application of the NMDA receptor antagonist APV (50 M; Figure 7A). GST−(D1 + D2) was applied into neurones through recording electrodes. We found that the application of (D1 + D2) into hippocampal neurones may potentiate the NMDA, but not non-NMDA (Figure 7A and C), receptor-mediated mEPSC component in a concentration-dependent manner (Figure 7B). No significant changes in glutamate-mediated mEPSCs could be noted when 0.05 M (D1 + D2) was applied into neurones (Figure 7B). With increases in the concentration of the applied protein, the NMDA receptor-mediated mEPSC component was enhanced (Figure 7B). At 15 min after application of 1.8 M (D1 + D2) into neurones, the NMDA receptor-mediated mEPSC component increased to 148 5% of initial values (Figure 7A and B), while no significant changes in the non-NMDA receptor-mediated mEPSC component could be found (Figure 7A and C). Application of GST alone (1.8 M) into neurones did not induce any significant increase in mEPSCs (Figure 7B).

To determine mechanisms underlying the effects of (D1 + D2), we investigated whether the up-regulation of NMDA receptors induced by the application of (D1 + D2) was produced through Src family PTK activity. Consistent with our previous finding that NMDA, but not non-NMDA, receptor-mediated mEPSCs are altered by activation or inhibition of endogenous Src family PTKs (Yu et al., 1997), we found that bath application of the Src family PTK inhibitor, PP2 (10 M), did not significantly change non-NMDA receptor-mediated mEPSCs (83 13% of control, n = 3), but reduced the NMDA receptor-mediated mEPSC component to 64 4% of control (n = 3). Importantly, application of (D1 + D2) into neurones treated with PP2 could not produce any significant increase in NMDA receptor-mediated synaptic responses (Figure 7C). Moreover, we also found that the effect of (D1 + D2) could be prevented by removal of ATP from the intracellular solution (Figure 7C). Thus, application of PTP (D1 + D2) protein into neurones may potentiate NMDA receptor activity via the activation of Src family PTKs. Furthermore, we found that application of GST fusion protein containing only the D1 domain did not produce the same effect as that induced by (D1 + D2) (Figure 7C). Thus, it is implied that the D2 domain interaction may be necessary for the regulation of NMDA receptors by PTP.

Conversely, intracellular application of the anti-PTP D1 antibody, clone21 (2.5 g/ml), through recording pipettes significantly reduced the NMDA receptor-mediated component of mEPSCs (Figure 7D and E). At 15 min after the antibody application, the NMDA mEPSC component was reduced to 64 3% of that recorded at 2 min after breakthrough (Figure 7D and E), while no significant change could be found in the non-NMDA receptor-mediated mEPSC component (Figure 7F). The effect of the antibody on NMDA mEPSCs could be blocked by co-application of 0.05 M (D1 + D2) (Figure 7E), while the application of 0.05 M (D1 + D2) alone into neurones did not significantly alter NMDA receptor-mediated synaptic responses (Figure 7B). As a control, non-specific mouse IgM applied into neurones had no effect (Figure 7E).

The role of endogenous PTP in the regulation of synaptic plasticity was also investigated. Tetanic stimulation-induced long-lasting potentiation of EPSCs of CA1 pyramidal neurones in hippocampal slices of adult rats was recorded. The EPSCs were evoked by stimulating the Schaffer collateral pathway. The effects of antibody clone21 and non-specific IgM applied into post-synaptic neurones were examined in two adjacent neurones recorded simultaneously with the double whole-cell patch recording configuration. One neurone was patched with a recording pipette filled with intracellular solution supplemented with antibody clone21 (2.5 g/ml), and the other was supplemented with an equal amount of non-specific IgM. No difference in the basal EPSCs recorded from neurones applied with clone21 and non-specific IgM was found until 20 min after breakthrough (Figure 7I). The tetanic stimulation delivered at 20 min after breakthrough produced an enhancement of EPSCs in neurones with intracellular application of non-specific IgM that was similar to that observed in neurones without the IgM application (intracellular solution only, data not shown); the peak amplitude of EPSCs at 30 min after tetanic stimulation was 163 10.2% (n = 8) of baseline (Figure 7G and I). Pre-application of antibody clone21 (2.5 g/ml) into post-synaptic neurones significantly reduced the amplitude of the long-lasting increase in EPSCs (Figure 7G and I), but did not significantly affect the membrane depolarization during tetanic stimulation (Figure 7H) or the short-term potentiation of EPSCs immediately after the tetanic stimulation (Figure 7I) when compared with those recorded from neurones applied with non-specific IgM. Conversely, intracellular application of GST fusion protein containing (D1 + D2) (1.8 M) through recording pipettes significantly increased the EPSC amplitude (Figure 7J and K). At 30 min, the EPSC amplitudes were increased to 147 8% (n =7) of that recorded at the first minute after breakthrough (Figure 7J and K). In contrast, application of GST alone (1.8 M) did not have such an effect (Figure 7J and K). Thus, we have demonstrated that PTP may be an important factor involved in the induction of activity-dependent enhancement of synaptic function in the CNS.

Discussion

Previous studies have demonstrated that NMDA receptors are associated with endogenous Src family PTKs (Yu et al., 1997; Tezuka et al., 1999), which regulate NMDA receptor functions (Yu et al., 1997; Lu et al., 1999; Huang et al., 2001; Vissel et al., 2001). In more detailed mechanistic studies, it has been found that Src may form a complex with NMDA receptors via the binding of its SH2 domain to a region N-terminal to the PSD95 PDZ1 domain (Hajdur et al., 2001), while Fyn participates via the binding of its SH2 domain to the PSD95 PDZ3 domain (Tezuka et al., 1999). Both PSD95 PDZ1 and PDZ2 domains may bind to the ESDV motif at the C-terminus of NMDA receptor NR2 subunits (Kornau et al., 1995; Sheng and Sala, 2001). Our present study demonstrated that the Src kinase activator, PTP, may also bind to PSD95. Recently, it was documented that the PDZ domain of PSD95 can bind to the internal peptides of proteins with a -finger structure, but lacking the typical E(S/T)XV terminal motif (Hillier et al., 1999; Sheng and Sala, 2001). Since there is thus far no crystallized PTP D2 domain available to aid in determining the structure of the PTP D2 domain, whether such a -finger structure exists in the PTP D2 domain is unknown. Nonetheless, our study revealed that: (i) the amount of PTP co-immunoprecipitated with NMDA receptors after the depletion of PSD95 from rat brain lysates was significantly reduced; (ii) in heterologous cells, PTP may co-precipitate all of the expressed PSD95 mutants except for one lacking the PDZ2 domain; and (iii) the peptide corresponding to the entire cytoplasmic portion (D1 + D2), the D2 phosphatase domain or the internal part (amino acids 539−711) of the PTP D2 domain, but not the D1 phosphatase domain of PTP, could be precipitated by a GST fusion protein containing the PSD95 PDZ2 domain in vitro. According to these findings, we conclude that PTP may bind to PSD95 through a direct interaction between the internal peptide amino acids 539−711 of the PTP D2 domain and the PSD95 PDZ2 domain. By combining our present findings with those reported by other groups demonstrating the interactions of Src family kinases and PSD95 (Kornau et al., 1995; Tezuka et al., 1999; Sheng and Sala, 2001), we propose a model showing that Src family PTKs, their activator (PTP) and substrate (NMDA receptors) are closely linked together by the scaffold protein, PSD95 (Figure 8).

In this model, the D2 domain of PTP plays a key role in the interaction of the phosphatase with the synaptic scaffold protein, PSD95. The D2 domain in most receptor-like PTPs is found to be involved in protein−protein interactions (Bilwes et al., 1996; Felberg and Johnson, 1998; Majeti et al., 1998; Wallace et al., 1998; Jiang et al., 1999). Through the D2 domain binding, receptor-like PTPs may form homo- or heterodimers, and thereby regulate the enzyme activity (Bilwes et al., 1996; Felberg and Johnson, 1998; Majeti et al., 1998; Wallace et al., 1998; Jiang et al., 1999). Furthermore, a recent study has demonstrated that the D2 domain interaction may regulate the clustering and location of receptor-like PTPs, such as LAR (Serra-Pages et al., 1998). We found that application of the D1 domain only did not produce the same effect as that induced by (D1 + D2) (Figure 7C). Thus, it has been implied that the D2 domain interaction may be necessary for the regulation of synaptic functions by receptor-like PTPs.

Previous studies, which discovered that NMDA receptor activity may be reduced by application of PTK inhibitors (Wang and Salter, 1994; Wang et al., 1996; Yu et al., 1997), indicate that NMDA receptors may be constitutively regulated by endogenous Src family kinases. However, until now, it has remained unclear whether certain structures are required for this regulation. The 'kinase−kinase activator−kinase substrate complex' shown in Figure 8 may represent such a structure required for the initiation and maintenance of the constitutive regulation of NMDA receptors by endogenous Src family PTKs.

Our present study shows that knockout of PTP activity abolished the effects of PTK inhibitors PP2 and lavendustin A on NMDA receptors. This finding strongly suggests that the constitutive regulation of NMDA receptors by PTKs, particularly Src family PTKs, may depend upon the activity of PTP. So far, there is no evidence to support the possibility that PTP may modulate NMDA receptor activity directly. Therefore, PTP may in reality act as an endogenous driver for initiation and maintenance of the constitutive regulation of NMDA receptors by endogenous Src family PTKs. This concept was confirmed further by the observation that blockade of the main catalytic domain of PTP occludes the effect of PP2 on NMDA receptors in cultured hippocampal neurones.

The regulation of NMDA receptor activity by protein kinases and phosphatases has been studied extensively (for reviews see Dingledine et al., 1999; Kennedy, 2000; Ali and Salter, 2001; Greengard, 2001). According to these studies, protein kinases and phosphatases appear to act in opposition in the regulation of NMDA receptor activity, i.e. kinases up-regulate, while phosphatases down-regulate NMDA receptor activity. However, our present study revealed that: (i) PTP-deficient cells exhibited down-regulated whole-cell responses mediated by recombinant NMDA receptors expressed in fibroblasts; (ii) application of PTP protein into hippocampal neurones significantly increased synaptic NMDA receptor activity; and (iii) blockade of the main catalytic domain of endogenous PTP inhibited NMDA receptor activity and LTP induction. Thus, we have demonstrated the first direct evidence indicating that a phosphatase may also up-regulate ligand-gated ion channel functions and excitatory synaptic transmission in the CNS.

LTP is believed to be an important cellular mechanism related to learning and memory processes. In studies of mechanisms of LTP induction, serine/threonine phosphatases have been found to be actively involved in the suppression of the induction and/or maintenance of LTP (for reviews see Kennedy, 2000; Winder and Sweatt, 2001). Functions of PTPs in the regulation of LTP induction have also been investigated recently (Uetani et al., 2000; Pelkey et al., 2002). Uetani et al. (2000) found an enhanced LTP in CA1 and CA3 neurones in PTP-deficient animals, suggesting that PTP may be a tonic down-regulator of LTP induction. Pelkey et al. (2002) reported that STEP, a cytoplasmic PTP, is a down-regulator of NMDA receptors and acts as a tonic brake on LTP induction. In contrast to all of the findings mentioned above, we found that the blockade of endogenous PTP inhibited LTP induction in CA1 neurones, and that application of the PTP functional domain (D1 + D2) into neurones induced an enhancement of synaptic responses. These data suggest that PTP may be a critical factor in the signalling pathway(s) responsible for LTP induction. To date, PTP is the first phosphatase found to be actively involved in the induction of LTP. Thus, further characterization of mechanisms underlying the function of PTP in LTP induction will be essential for the understanding of activity-dependent neuroplasticity in the CNS.

Materials and methods

Immunoprecipitation and western blotting

Brain tissue from adult rats was homogenized in ice-cold homogenizing buffer (320 mM sucrose, 10 mM Tris−HCl pH 7.4, 1 mM NaHCO₃ pH 7.4, 1 mM MgCl₂) supplemented with 1 mM sodium orthovanadate and 1% protease inhibitor cocktail after two washes with ice-cold phosphate-buffered saline (PBS), and subsequently spun at 5000 g for 15 min. The supernatant was collected and spun at 40 000 g (Beckman ultracentrifuge, Palo Alto, CA) for 60 min. Pellets were then dissolved in a lysis buffer (10 mM Tris−HCl pH 9, 150 mM NaCl, 0.5% Triton X-100, 1% sodium deoxycholate, 0.5% SDS, 2 mM EDTA, 1 mM sodium orthovanadate and 1% protease inhibitor cocktail) for subsequent experiments. For collection of cultured cells, attached cells were mechanically dissociated from the dish after two washes with ice-cold PBS, and then spun at 4000 g for 5 min. Cell pellets in the lysis buffer were kept on ice for 1 h, and then sonicated and centrifuged at 15 000 g for 20 min. The supernatant was collected for subsequent experiments. In immunoprecipitation experiments, solubilized proteins (1 mg) were incubated with antibodies (see below) at 4°C and gently shaken overnight. Antibodies used for immunoprecipitation were: anti-NR1 (1 g, mouse IgG; PharMingen, San Diego, CA), anti-NR2A/B (0.5 g, rabbit IgG; Chemicon, Temecula, CA), anti-PTP (1 g, rabbit IgG; F.Jirik, University of British Columbia; Harder et al., 1998), clone21 (1.3 g, mouse IgM; Transduction labs, Lexington, KY), clone327 (0.5 g, mouse IgG; Oncogene, Cambridge, MA) and anti-PSD95 (4 l, mouse IgG; ABR, Golden, CO). The immune complexes were collected with 10 l of protein G−Sepharose beads for 2 h at 4°C. Immunoprecipitates were then washed four times with ice-cold PBS, resuspended in 2 Laemmli sample buffer and boiled for 5 min. These samples were subjected to SDS−PAGE and transferred to a nitrocellulose membrane. The blotting analysis was performed by repeated stripping and successive probing with antibodies: anti-NR1, anti-NR2A/B, anti-PTP, clone327, anti-Src-pTyr529 (rabbit IgG; Biosource, Camarillo, CA), anti-PSD95 and anti-PSD95₁₋₅₄ (rabbit IgG; T.Yamamoto, Tokyo University). All chemicals used were purchased from Sigma, except where indicated.

Cell culture and transfection

For biochemical experiments, HEK293 and COS7 cells, and PTP^-/- fibroblasts (Su et al., 1999) (a gift of Dr J.Sap, New York University) were grown in Dulbecco's modified Eagle's medium (DMEM) (Gibco-BRL, Gaithersburg, MD) supplemented with 10% fetal bovine serum (Gibco-BRL). These cells were transfected with expression vectors (pcDNA3 or pBCMGneo) containing cDNAs encoding NR1-1a (8 g), NR2A (32 g), PTP (1 g), and wild-type (1 g, a gift of Dr D.Bredt) or mutant (1 g; a gift of Dr T.Tezuka and T.Yamamoto) (Tezuka et al., 1999) PSD95 in which amino acids 532−724 [corresponding to the guanylate kinase domain (GK), GK], amino acids 411−724 [including both the GK and SH3 domains, (GK + SH3)], amino acids 255−433 (corresponding to the PDZ3 domain, PDZ3), amino acids 157−256 (corresponding to PDZ2, PDZ2) and amino acids 54−158 (corresponding to PDZ1, PDZ1) of PSD95 were deleted. The calcium phosphate method (Gibco-BRL, Gaithersburg, MD) was used for transfection following the protocol recommended by the manufacturer. Cells transfected with both NR1-1a and NR2A cDNAs were cultured in medium supplemented with the NMDA receptor antagonist APV (500 M) for 48−72 h before lysis.

For electrophysiological recordings, cDNAs encoding the NMDA receptor subunits NR1-1a (0.3 g) and NR2A (1.2 g), PSD95 (0.15 g), green fluorescent protein (GFP; 0.15 g), wild-type PTP (0.15 g), the catalytically inactive mutant of PTP [PTP with a cysteine to alanine mutation at residue 433 (Harder et al., 1998), 0.15 g], pp60^c-Src (0.3 g) or the corresponding empty vectors were transfected into PTP^-/- or SYF fibroblasts (Klinghoffer et al., 1999) (ATCC, Manassas, VA) cultured in 35 mm culture dishes using the lipofectamine method (Gibco-BRL). After 5−12 h, transfected cells were maintained in DMEM supplemented with 10% fetal bovine serum and APV (500 M) for 48 h before recording.

Neurones used for electrophysiological recordings were prepared from cultures of fetal hippocampal tissues of Wistar rats (E18) as previously described (Yu et al., 1997). Hippocampal tissue was then mechanically dissociated by trituration and plated onto 35 mm, collagen-coated culture dishes at a density of <1 10⁶ cells/ml. Hippocampal neurones were used for electrophysiological recordings 12−17 days after plating.

In vitro GST pull-down assay

DNA sequences encoding the entire cytoplasmic portion of PTP (D1 + D2, amino acids 167−793), the membrane-proximal phosphatase domain (D1, amino acids 167−555), the membrane-distal phosphatase domain (D2, amino acids 510−793) and amino acids 539−711 of PTP (Harder et al., 1998) were synthesized using PCR. Plasmid pBCMGSneo containing the full-length cDNA of PTP (a gift from Dr Jirik; Harder et al., 1998) was used as the template.

Primer sequences made for the PCR for D1 + D2, D1, D2 and amino acids 539−711 of PTP were 5'-TTTAAGCTCAGGCCACCTGTGAGGC-3' and 5'-TTTGATATCTTACTTGAAGTTGGCATAATC-3'; 5'-TTTAAGCTCAGGCCACCTGTGAGGC-3' and 5'-TTTGATATCTTAGGTCCCTGGGATTTTGTTG-3'; 5'-TTTAAGCTTGACAAGATGCGGACTGG-3' and 5'-TTTGATATCTTACTTGAAGTTGGCATAATC-3'; and 5'-TTTAAGCTTGACAAGATGCGGACTGG-3' and 5'-TTTGATATCTTACTGCTGCTTCTGCACGGC-3', respectively. All the PTP fragments were confirmed by DNA sequencing and constructed in the HindIII−EcoRV sites in vector pcDNA3 (Invitrogen, Carlsbad, CA) for in vitro transcription and translation.

pGEX2T- PSD95 PDZ1 (amino acids 55−156), PDZ2 (amino acids 158−256) and PDZ3 (amino acids 257−432) plasmids were gifts from Dr T.Yamamoto (Tezuka et al., 1999). The PSD95−GST fusion proteins were produced in Escherichia coli (BL-21 strain; Invitrogen) by isopropyl--D-thiogalactopyranoside (IPTG; 0.1 M) induction and purified with glutathione−Sepharose beads. The bead-bound fusion proteins (30 g) conjugated with the PDZ1, 2 and 3 domains of PSD95 were each mixed with 2 l of in vitro synthesized [³⁵S]methionine-labelled PTP peptides D1 + D2, D1, D2 or amino acids 539−711 (TNT-coupled reticulocyte lysate systems; Promega, Madison, WI) in binding buffer (250 mM NaCl, 50 mM HEPES pH 7.9, 0.5 mM EDTA, 0.1% NP-40, 1 mM dithiothreitol) for 10 min at 4°C. The beads were then washed four times with binding buffer, and bound proteins eluted with elution buffer (20 mM Tris−HCl pH 7.9, 100 mM KCl, 20 mM glutathione) before separation by SDS−PAGE and detection by autoradiography.

Whole-cell recordings in cultured cells

For whole-cell recordings in hippocampal neurones, the cultures were bathed in a standard extracellular solution containing 140 mM NaCl, 5.4 mM KCl, 33 mM glucose, 1.3 mM CaCl₂, 25 mM HEPES, 0.001 mM tetrodotoxin (TTX) and 0.003 mM glycine pH 7.35; osmolarity 310−320 mOsm. The standard extracellular solution containing 0.03 mM glycine was used for recordings in fibroblasts. Recording pipettes were made from thin-walled borosilicate glass capillaries (World Precision Instruments, Sarasota, FL) pulled to a diameter of 1−2 m at the tip with a resistance of 4−7 M, and filled with a standard intracellular solution composed of 140 mM CsCl, 1 mM BAPTA, 10 mM HEPES, 2 mM MgCl₂ and 4 mM K-ATP pH 7.25; osmolarity 300−310 mOsm. All the agents used for testing PTP or Src family PTKs were freshly prepared from high concentration stock solutions and diluted (1:200 or 1:1000) with the standard intra- or extracellular solution for recording experiments.

Whole-cell currents were evoked by application of NMDA receptor agonists L-aspartate or NMDA (250 M). The standard and NMDA receptor agonist-containing extracellular solutions were applied to cells via a double-barrel perfusion pipette made from two square glass capillary tubes attached to a stepper motor (SF-77B perfusion fast-step system; Warner Instrument, Hamden, CT) which was controlled by a computer. The time constant of solution exchange was 5 ms. Recordings were conducted under the voltage clamp condition at a holding potential of -60 mV except where indicated. Whole-cell currents were recorded using Axopatch 200B amplifiers (Axon Instruments, Foster City, CA). Data were filtered at 2 kHz and digitized on-line using Digidata 1200 DAC units (Axon Instruments). On-line data acquisition and off-line analysis were performed using pClamp8 software (Axon Instruments). Current−voltage (I/V) relationships were calculated by the subtraction of currents during the voltage ramp from -60 to +60 mV without L-aspartate or NMDA application from currents during the analogous ramp over the steady state of L-aspartate- or NMDA-evoked responses.

For mEPSC recordings in hippocampal neurones, the cultures were bathed in an extracellular solution containing 100 mM Na₂SO₄, 9 mM Cs₂SO₄, 33 mM glucose, 1.3 mM CaCl₂, 25 mM HEPES, 0.001 mM TTX, 0.003 mM glycine, 0.01 mM bicuculline and 0.01 mM strychnine pH 7.35; osmolarity 310−320 mOsm. Recording pipettes were filled with the standard intracellular solution (see above). mEPSCs were detected semi-automatically off-line and analysed after alignment of the rising edge using Synaptic Toolbench software (Yu et al., 1997; Yu and Salter, 1998). The threshold for detection was approximately -4 pA and was optimized for each cell so as to collect >95% of mEPSCs. Traces with noise artefacts or with more than one mEPSC per 200 ms recording period were not included in the analysis. Detected mEPSCs were averaged during the periods following breakthrough: 0−2 min, 2−5 min, 5−10 min and 10−15 min. Averages were required to contain at least 25 traces. For each neurone tested, charges (Q) mediated by non-NMDA and NMDA receptors were calculated respectively by integrating currents mediated by these receptors, and normalized to that measured during the initial 2 min period (Q₂).

Whole-cell recordings in brain slices

Hippocampal slices (300 m) from 28- to 30-day-old Sprague−Dawley rats were prepared as previously described (Lu et al., 1998). In brief, the slice in the recording chamber was superfused continuously with artificial cerebrospinal fluid (ACSF, 2 ml/min) saturated with 95% O₂/5% CO₂ at 30 1°C. The ACSF contained 124 mM NaCl, 3 mM KCl, 1.25 mM NaH₂PO₄, 2 mM MgCl₂, 2 mM CaCl₂, 26 mM NaHCO₃, 10 mM dextrose and 0.01 mM bicuculline. CA1 pyramidal neurones were visualized with infrared illumination and differential interference contrast. Two adjacent neurones (separated by a distance of <20 m) were selected for double whole-cell patch clamp recordings. Recording electrodes (3−5 M) were filled with intracellular solution containing 142.5 mM Cs-gluconate, 7.5 mM CsCl, 10 mM HEPES, 0.2 mM EGTA, 2 mM Mg-ATP, 0.3 mM GTP, and 5 mM QX-314 pH 7.4; osmolarity 290−300 mOsm. For double-patch clamp recordings, one neurone was patched with an electrode filled with the intracellular solution and additionally with the antibody clone21 or GST−(D1 + D2), and the other electrode with intracellular solution and additionally with an equal amount of non-specific mouse IgM or GST alone. EPSCs were recorded at a holding potential of -70 mV with Axopatch 200B amplifiers. Currents were filtered at 2 kHz with a low-pass filter, and data were digitized at 10 kHz and stored on-line using the pclamp8 system (Axon Instruments). The input and series resistance were monitored using voltage steps (1 mV, 200 ms) at 5 min intervals throughout the duration of the experiment. Series resistance ranged from 11 to 16 M, and input resistance was 266−288 M. For each neurone, if the input resistance changed >20% relative to a 10 min period prior to tetanic stimulation, the neurone was rejected from the statistical analysis. The resting membrane potential of recorded neurones was -64 to -68 mV. Neither the antibody nor GST fusion protein application affected resting membrane potential or input resistance. The Schaffer collateral pathway, 50 m from CA1 pyramidal cell bodies, was stimulated with a bipolar tungsten electrode that delivered test stimuli at a frequency of 0.1 Hz. The tetanic stimulation consisted of two 100 Hz stimuli of 1 s duration with an inter-train interval of 10 s in current clamp mode. The averaged amplitude of EPSCs for the 10 min period before tetanic stimulation was defined as baseline (i.e. 100%).

Acknowledgements

We thank Drs M.W.Salter, Y.-T.Wang, H.Van Tol and L.-Y.Wang for critical comments on the manuscript. We are also grateful to Dr J.Sap for the gifts of PTP^-/- cells and anti-PTP antibody, as well as critical comments on the manuscript, Drs T.Yamamoto and T.Tezuka for the gifts of the PSD95 constructs and anti-PSD95₁₋₅₄ antibody, as well as critical comments on the manuscript, Dr F.R.Jirik for the gifts of the antibodies and plasmids for PTP, and for comments on the manuscript, and Dr D.Bredt for the gift of PSD95 plasmid. This work was supported by grants from Canadian Institutes of Health Research (CIHR) and Ontario Neurotrauma Foundation to X.-M.Y., who is a New Investigator of CIHR.

References

Ali DW and Salter MW (2001) NMDA receptor regulation by Src kinase signalling in excitatory synaptic transmission and plasticity. Curr Opin Neurobiol, 11, 336–342. | Article | PubMed | ISI | ChemPort |

Bekkers JM and Stevens CF (1989) NMDA and non-NMDA receptors are co-localized at individual excitatory synapses in cultured rat hippocampus. Nature, 341, 230–233. | Article | PubMed | ISI | ChemPort |

Bhandari V, Lim KL and Pallen CJ (1998) Physical and functional interactions between receptor-like protein-tyrosine phosphatase and p59fyn. J Biol Chem, 273, 8691–8698. | Article | PubMed | ISI | ChemPort |

Bilwes AM, den Hertog J, Hunter T and Noel JP (1996) Structural basis for inhibition of receptor protein-tyrosine phosphatase- by dimerization. Nature, 382, 555–559. | Article | PubMed | ISI | ChemPort |

Blanchetot C and den Hertog J (2000) Multiple interactions between receptor protein-tyrosine phosphatase (RPTP) and membrane-distal protein-tyrosine phosphatase domains of various RPTPs. J Biol Chem, 275, 12446–12452. | Article | PubMed | ISI | ChemPort |

Brown MT and Cooper JA (1996) Regulation, substrates and functions of src. Biochim Biophys Acta, 1287, 121–149. | Article | PubMed | ISI | ChemPort |

Cooper JA, Gould KL, Cartwright CA and Hunter T (1986) Tyr527 is phosphorylated in pp60c-src: implications for regulation. Science, 231, 1431–1434. | Article |