In contrast to the well established function of TRPV1 receptors in the PNS, their role in the central nervous system (CNS) is not well defined. The presence of TRPV1 receptors in the mammalian brain has been demonstrated using in situ hybridization and reverse transcription polymerase chain reaction (RT-PCR) (Sasamura et al., 1998; Mezey et al., 2000), immunochemical staining methods (Sanchez et al., 2001; Toth et al., 2005; Cristino et al., 2006) and [³H]resiniferatoxin autoradiography comparing wild-type and TRPV1 receptor knockout mice (Roberts et al., 2004). These studies indicate the presence of potentially functional TRPV1 receptors in brain regions including the thalamic and hypothalamic nuclei, the locus coeruleus, periaqueductal grey and cerebellum, cortical and limbic structures including the hippocampus, the caudate putamen and the substantia nigra pars compacta. Nonetheless, the functional significance of TRPV1 receptor expression in the brain remains elusive, although there is evidence that TRPV1 receptors in the CNS are involved in pain modulation and may serve as useful drug targets (Cui et al., 2006). TRPV1 receptor mRNA and protein are expressed in hippocampal neurons (Sasamura et al., 1998; Roberts et al., 2004; Toth et al., 2005; Cristino et al., 2006) including those of the human hippocampus (Mezey et al., 2000), and functional effects of these receptors have been shown using electrophysiological methods (Al-Hayani et al., 2001; Huang et al., 2002; Marsch et al., 2007). A recent study using mice lacking TRPV1 receptors suggests their involvement in anxiety-related behavior and two behavioral measures of hippocampal-dependent learning, conditioned and sensitized fear (Marsch et al., 2007). Moreover, hippocampal long-term potentiation (LTP) was attenuated in the CA1 region of brain slices from TRPV1 knockout mice, indicating alterations in synaptic circuit function in this brain region, although the mechanism remains unknown (Marsch et al., 2007).

TRPV1 receptors in the CNS are less likely than those in the PNS to be activated by heat or low pH, and therefore it has been suggested that other endogenous ligands of this ion channel, such as the endovanilloids mentioned above, are likely activators (Huang et al., 2002; Marinelli et al., 2003; Van Der Stelt and Di Marzo, 2004; De Petrocellis and Di Marzo, 2005; Marsch et al., 2007). Anandamide and NADA are also members of the endocannabinoid family, activating CB1 receptors as well (Zygmunt et al., 1999; Huang et al., 2002), and it remains unclear whether or not any of these ligands are responsible for the TRPV1-mediated physiological and pathological effects in and outside of the CNS (Van Der Stelt and Di Marzo, 2004).

The most commonly observed mechanisms underlying synaptic depression are a decrease in presynaptic neurotransmitter release or a decrease in postsynaptic receptor number or responsiveness (Malenka and Bear, 2004). When synaptic plasticity results from a change in neurotransmitter release, this is generally accompanied by an altered coefficient of variation of the EPSCs (CV), and changes in the paired-pulse ratio (PPR) and synaptic failure rate (del Castillo and Katz, 1954; Malinow and Tsien, 1990; Manabe et al., 1992). Consistent with this interpretation, we observed a decrease in 1/CV² and an increase in the PPR and number of synaptic failures during LTD (Figure 1C, D, E). If LTD at interneuron synapses results from a persistent decrease in presynaptic glutamate release, we would also predict depression of the NMDAR-mediated component as well as the AMPAR-mediated component of the EPSC (isolated in the experiments in Figure 1A, B). We therefore measured the isolated NMDAR-mediated EPSC at +40 mV and found that HFS delivered to the afferents elicited robust LTD of the NMDAR EPSC (Figure 1F, G; NMDAR EPSC amplitudes post-HFS: 64.2 ± 11.1% of control values before HFS; P < 0.001, n = 7). Taken together, these findings indicate that LTD is AMPAR- and NMDAR-independent and results from a persistent decrease in presynaptic glutamate release as monitored by both AMPA and NMDA receptors.

How might high-frequency activation of excitatory afferents trigger LTD at interneuron synapses? Neither NMDARs nor AMPARs are necessary for LTD (Figure 1), but group I metabotropic glutamate receptors (mGluRs) are expressed on these cells (Ferraguti et al., 2004) and will also be activated by the glutamate released during HFS. We found that LTD was entirely blocked in the presence of the selective mGluR1 antagonist, CPCCOEt (25–50 μM; Figure 2A; EPSC amplitudes 10–15 minutes post-HFS in CPCCOEt: 111.4 ± 17.0% of control values before HFS; P < 0.01 compared to control LTD, n = 9). Our findings thus far are reminiscent of other recent examples of LTD in which activation of postsynaptic group I mGluRs produces endocannabinoids (Maejima et al., 2001; Gerdeman et al., 2002; Robbe et al., 2002; Chevaleyre and Castillo, 2003, 2004; Ronesi et al., 2004; Kreitzer and Malenka, 2005; Takahashi and Castillo, 2006). Endocannabinoids can act as retrograde messengers, traveling across the synapse to activate presynaptic CB1 receptors, thereby reducing presynaptic neurotransmitter release (Llano et al., 1991; Pitler and Alger, 1992; Kreitzer and Regehr, 2001; Ohno-Shosaku et al., 2001; Wilson and Nicoll, 2001). To ask whether endocannabinoids might mediate LTD at interneuron synapses, we tested two CB1 receptor antagonists. SR141716A (rimonabant; 2–5 μM) effectively blocked LTD (Figure 2B; EPSC amplitudes 15–20 minutes post-HFS: 110.6 ± 19.3% of control values before HFS; P < 0.01 compared to control LTD, n = 10). However, the selective CB1 receptor antagonist, AM251 (2 μM), did not block LTD in any of the 8 interneurons tested (Figure 2C; EPSC amplitudes 15–20 minutes post-HFS: 48.6 ± 5.3% of control values before HFS, n = 8). To confirm that AM251 indeed blocks CB1 receptors under these experimental conditions, we found in separate experiments that AM251 (2 μM) blocked the synaptic depression of these synapses elicited by the CB1 receptor agonist, WIN 55,212–2 (1 μM) (Figure 2D, E; EPSC amplitudes after 10–15 minutes in WIN 55,212–2 alone: 66.1 ± 7.9% of pre-drug control values; P < 0.001, n = 14; EPSC amplitudes after 10–15 minutes in both WIN 55,212–2 and AM251: 101.7 ± 8.9% of pre-WIN 55,212–2 control values; P < 0.05 compared to WIN55,212–2 depression in the absence of AM251, n = 5). In addition, pre-treatment with WIN 55,212–2 (1 μM) for at least ten minutes did not prevent synaptic depression triggered by high-frequency synaptic stimulation (EPSC amplitudes 15–20 minutes post-HFS: 51.3 ± 7.5% of control values in WIN55,212–2 before HFS; P = 0.25 compared to control LTD, n = 12; data not shown). The block of LTD by SR141716A but not by AM251 was surprising, indicating that CB1 receptors are not necessary for this form of LTD and instead that SR141716A blocks LTD via a CB1 receptor-independent mechanism.

SR141716A may antagonize not only CB1 receptors but also the TRP channel family member, TRPV1 (De Petrocellis et al., 2001). TRPV1 is found in hippocampal neurons (Hajos and Freund, 2002; Roberts et al., 2004; Toth et al., 2005; Cristino et al., 2006; Marsch et al., 2007) and we therefore first tested whether transient application of a TRPV1 agonist mimics LTD induction. The extremely selective TRPV1 agonist capsaicin (1 μM) significantly depressed excitatory synaptic currents in interneurons (Figure 3A, B; EPSC amplitudes after 10–15 minutes in capsaicin: 73.3 ± 4.6% of pre-drug control values; P < 0.001, n = 14). If capsaicin maximally activates signaling mechanisms in common with LTD, HFS following the synaptic depression elicited by capsaicin should not cause any further depression. As predicted, HFS after capsaicin exposure failed to produce further LTD (Figure 3A, B; EPSC amplitudes 15–20 minutes post-HFS: 112.5 ± 21.9% of control values in capsaicin before HFS; P < 0.05 compared to control LTD, n = 10). This result suggests that the processes underlying LTD induction are fully activated by treatment with capsaicin, again supporting the requirement for TRPV1 channels in LTD. Like synaptically-induced LTD, the synaptic depression elicited by capsaicin was accompanied by an increase in synaptic failures, supporting our hypothesis that LTD at interneuron synapses results from a persistent decrease in presynaptic glutamate release (Figure 3C; average synaptic failures pre-capsaicin application: 38.4 ± 3.5%; average synaptic failures in capsaicin: 73.5 ± 7.8%; P < 0.01, n = 5).

We reasoned that if SR141716A blocks LTD by an antagonist action at TRPV1 receptors on hippocampal neurons, then SR141716A should also prevent capsaicin-induced synaptic depression. After pretreatment with SR141716A (2 μM), capsaicin (1 μM) did not depress the synapses (Figure 3D; EPSC amplitudes after 10–15 minutes in capsaicin and in the presence of SR141716A: 102.8 ± 9.2% of pre-capsaicin control values; P < 0.01 compared to capsaicin depression in the absence of SR141716A, n = 6). This finding emphasizes that at this concentration SR141716A cannot be regarded as a selective CB1 receptor antagonist, but instead appears to antagonize capsaicin-sensitive receptors, presumably TRPV1 channels. If, as these data suggest, TRPV1 is necessary for LTD induction at interneuron synapses, then TRPV1 antagonists should interfere with LTD. Both capsazepine (10 μM) and 5′-Iodoresiniferatoxin (100 nM) potently blocked LTD when bath applied prior to HFS (Figure 3E, F; EPSC amplitudes 15–20 minutes post-HFS in capsazepine: 105.6 ± 8.6% of control values before HFS; P < 0.001 compared to control LTD, n = 9; in 5′-Iodoresiniferatoxin: 96.2 ± 13.6% of control values before HFS; P < 0.01 compared to control LTD, n = 7). These data support an essential role for TRPV1 receptors in LTD induction. Once LTD is initiated, TRPV1 channel activity is no longer necessary to maintain synaptic depression since following LTD induction, EPSC amplitudes were not restored to basal values by blocking TRPV1 receptors (10 μM capsazepine was added 10 minutes after HFS; EPSC amplitudes 10–15 minutes after adding capsazepine: 50.1 ± 6.1% of pre-HFS values, n = 6; data not shown).

The pharmacological data presented above are all consistent with an essential role for TRPV1 channels in the induction of LTD. To further test this hypothesis, we asked whether LTD could be elicited in transgenic mice lacking TRPV1 receptors (TRPV1^−/−) (Caterina et al., 2000). LTD was markedly reduced in slices from TRPV1^−/− mice, when compared to LTD in interleaved slices from wild-type control mice (Figure 4A, B; EPSC amplitudes 15–20 minutes post-HFS in TRPV1^−/− mice: 95.8 ± 7.0% of control values before HFS; P < 0.001 compared to control LTD in wild-type mice, n = 9; in C57BL/6 wild-type mice: 52.1 ± 5.2% of control values before HFS, n = 15). While application of capsaicin (1 μM) to slices from wild-type mice elicited synaptic depression, this was not seen in slices from TRPV1^−/− mice, confirming the lack of functional TRPV1 receptors (Figure 4C, D; EPSC amplitudes after 10–15 minutes in capsaicin in TRPV1^−/− mice: 100.7 ± 6.6% of pre-drug control values; P < 0.01 compared to capsaicin response in wild-type mice, n = 8; EPSC amplitudes after 10–15 minutes in capsaicin in C57BL/6 wild-type mice: 50.5 ± 12.1% of pre-drug control values, n = 6). These data complement the pharmacological evidence, and strongly suggest that TRPV1 channels or TRPV1-containing heteromultimeric channels are signaling components required for interneuron LTD.

How is LTD initiated by high-frequency synaptic stimulation? Our data are consistent with a model analogous to that of endocannabinoid-mediated LTD (Chevaleyre et al., 2006), in which activation of mGluR1 produces a lipid retrograde messenger capable of activating TRPV1 receptors located on presynaptic pyramidal cell terminals. Activation of group I mGluRs can produce both endocannabinoids and eicosanoid metabolites of arachidonic acid, and these endogenous messengers effectively activate TRPV1 receptors (Zygmunt et al., 1999; Hwang et al., 2000; Shin et al., 2002). The eicosanoid, 12-(S)-HPETE, is known to be liberated during electrical stimulation of hippocampal slices (Feinmark et al., 2003), and thus we asked whether or not this lipid messenger can mimic LTD at interneuron synapses. Application of 12-(S)-HPETE (100 nM) depressed excitatory synapses on interneurons (Figure 5A; EPSC amplitudes after 10–15 minutes in 12-(S)-HPETE: 40.6 ± 11.7% of pre-drug control values; P < 0.01, n = 8), and subsequent HFS did not produce further LTD (Figure 5B; EPSC amplitudes 15–20 minutes post-HFS: 98.7 ± 15.3% of control values in 12-(S)-HPETE before HFS; P < 0.05 compared to control LTD, n = 6). In addition, the synaptic depression as a result of 12-(S)-HPETE application, like that caused by HFS or capsaicin, was associated with an increase in synaptic failures (Figure 5C; average synaptic failures pre-12-(S)-HPETE application: 35.2 ± 1.9%; average synaptic failures in 12-(S)-HPETE: 89.7 ± 3.0%; P < 0.001, n = 6). 12-(S)-HPETE synthesis from arachidonic acid requires 12-lipoxygenase. To determine whether or not endogenously released 12-(S)-HPETE is responsible for triggering LTD following synaptic stimulation, we attempted to induce LTD using HFS in the presence of baicalein (500 nM), an inhibitor of 12-lipoxygenase. Synaptically-induced LTD was blocked in the presence of baicalein, and in fact in four of ten cells we observed potentiation (greater than 125% of control 20 minutes following the HFS)(Figure 5D; EPSC amplitudes 15–20 minutes post-HFS: 129.6 ± 20.3% of control values before HFS; P < 0.01 compared to control LTD, n = 10). Moreover, the depression caused by 12-(S)-HPETE was prevented by either capsazepine (10 μM) or SR141716A (2 μM) (Figure 5E, F; EPSC amplitudes after 10–15 minutes in 12-(S)-HPETE and in the presence of capsazepine: 103.0 ± 8.9% of pre-12-(S)-HPETE control values; P < 0.01 compared to 12-(S)-HPETE depression in the absence of capsazepine, n = 6; in 12-(S)-HPETE and in the presence of SR141716A: 106.9 ± 5.5% of pre-12-(S)-HPETE control values; P < 0.01 compared to 12-(S)-HPETE depression in the absence of SR141716A, n = 5). These observations demonstrate a similar pharmacological profile for 12-(S)-HPETE and synaptically-triggered LTD. Finally, we found that in slices from TRPV1^−/−mice, 12-(S)-HPETE did not depress synaptic transmission at excitatory synapses on interneurons (Figure 5G, H; EPSC amplitudes after 10–15 minutes in 12-(S)-HPETE in TRPV1^−/− mice: 109.0 ± 6.8% of pre-drug control values; P = 0.33, n = 6). To rule out any possible involvement of the retrograde messenger, nitric oxide (NO), we tested whether or not LTD was affected when nitric oxide synthase (NOS) was inhibited. When 200 μM L-NAME was bath applied 10 minutes prior to HFS, LTD appeared entirely normal, suggesting that NO does not have a role in this form of synaptic plasticity (EPSC amplitudes 15–20 minutes post-HFS: 46.3 ± 10.4% of control values before HFS, n = 5; data not shown). Together, our data strongly suggest that 12-(S)-HPETE acts at TRPV1 receptors to depress synaptic transmission at excitatory synapses onto interneurons, and that 12-(S)-HPETE liberated during HFS is essential for triggering LTD.

Interneurons in stratum radiatum of hippocampal area CA1 receive their major excitatory synaptic inputs from CA3 pyramidal cells but can also receive recurrent collaterals from CA1 pyramidal cells (Freund and Buzsaki, 1996). We next tested whether or not field excitatory postsynaptic potentials (fEPSPs) from synapses between CA3 pyramidal cells and CA1 pyramidal cells also exhibit TRPV1-mediated synaptic depression. Surprisingly, 1 μM capsaicin, a concentration that significantly depressed excitatory synapses on interneurons (Figure 3A, B), did not depress synapses on CA1 pyramidal cells (Figure 6A; fEPSP slopes after 10–15 minutes in capsaicin: 102.8 ± 5.4% of pre-drug control values; P = 0.58, n = 5). Although 10 μM capsaicin depressed synaptic transmission at the CA3–CA1 synapse, as previously reported (Hajos and Freund, 2002), we found that this was often associated with a depression in the presynaptic fiber volley component of the field potential, suggesting a possible confounding effect on presynaptic excitability. Furthermore, we found that 100 nM 12-(S)-HPETE, a concentration that significantly depressed excitatory synapses on area CA1 interneurons (Figure 5A), did not depress synapses between CA3 and CA1 pyramidal cells (Figure 6B; fEPSP slopes after 10–15 minutes in 12-(S)-HPETE: 99.0 ± 3.5% of pre-drug control values; P = 0.83, n = 5). Our data strongly suggest that TRPV1 channel activation does not depress glutamate release at these CA3 excitatory synapses onto CA1 hippocampal pyramidal cells, but potently inhibits excitatory synapses on interneurons in area CA1 stratum radiatum.

We next investigated the involvement of the recorded interneuron in the generation of LTD. We found that intracellular perfusion of recorded interneurons with either GDPβS(250 μM), to block G-protein signaling, or BAPTA, (25–40 mM) to chelate postsynapticCa²⁺, reduced interneuron LTD (Figure 7), a result that can be explained if lipid retrograde messengers required for LTD are largely produced by the interneuron. Furthermore, delivery of the 12-lipoxygenase inhibitor, baicalein (140 nM) into the postsynaptic neuron via the patch pipette also markedly attenuated LTD (Figure 7). Together, these data indicate that the 12-(S)-HPETE necessary for LTD induction is produced in the recorded interneuron, and that G-protein signaling and postsynaptic Ca²⁺ play an important role.

We also report that in the hippocampus at least, SR141716A appears to be insufficiently selective to distinguish CB1 from TRPV1 receptors. In our study, SR141716A blocked LTD, in addition to responses to capsaicin and to 12-(S)-HPETE, whereas the very similar CB1 receptor antagonist, AM251, was ineffective. SR141716A has been shown to attenuate responses to capsaicin in other systems as well, particularly at concentrations above 1 μM (Zygmunt et al., 1999; De Petrocellis et al., 2001). A pharmacological profile similar to what we have observed was reported for the vasorelaxation of small mesenteric blood vessels that was mediated by an endothelial receptor in response to NADA, also blocked by SR141716A but not AM251 (O’Sullivan et al., 2004). Our findings may also relate to previous reports of a vanilloid receptor-like response at hippocampal excitatory synapses (Al-Hayani et al., 2001; Hajos and Freund, 2002). SR141716A (also known as rimonabant or Acomplia) is in wide clinical use outside the United States as an anti-obesity aid (Tucci et al., 2006; Padwal and Majumdar, 2007). A large percentage of patients stop taking this drug as a result of psychiatric side-effects, and our findings suggest the possibility that some of the central effects of rimonabant result from the antagonism of TRPV1 receptors as well as CB1 receptors (Pegorini et al., 2006).

TRPV1 receptors are expressed in hippocampal neurons (Mezey et al., 2000; Szabo et al., 2002; Toth et al., 2005; Cristino et al., 2006) and may be activated in several different ways, including by lipoxygenase derivatives that can be released as a result of group 1 mGluR activation, as we have shown here (Hwang et al., 2000; Sohn et al., 2007). 12-(S)-HPETE is known to be released during field stimulation of hippocampal slices (Feinmark et al., 2003), and our data indicate that 12-(S)-HPETE production is necessary and sufficient for LTD at excitatory interneuron synapses. Our previous study showed that LTD was triggered simultaneously at both activated and non-activated synapses on interneurons, indicating that the LTD is not synapse-specific or activity-dependent (McMahon and Kauer, 1997). The heterosynaptic nature of interneuron LTD may be accounted for by the local spread of 12-(S)-HPETE from interneurons activated during HFS. The most likely source of this eicosanoid is the recorded interneuron itself, based on our data using internally-perfused drugs; when applied intracellularly to the interneuron the Ca²⁺ chelator, BAPTA, the G-protein inhibitor, GDPβS, and the 12-lipoxygenase inhibitor, baicalein, all reduced the number of interneurons exhibiting LTD, suggesting that a Ca²⁺-sensitive process, a GPCR-mediated process and 12-lipoxygenase generation within the interneuron are necessary for LTD. If pyramidal cells, whose processes surround stratum radiatum interneurons, were a significant source of 12-(S)-HPETE following HFS, drugs delivered intracellularly to the recorded interneuron should not block LTD. Instead, in most experiments the intracellularly delivered drugs blocked LTD (Figure 7). However in some interneurons even with BAPTA, GDPβS or baicalein present, LTD of normal magnitude was induced, suggesting that the necessary signaling molecules can also arise elsewhere; we favor the idea that in some cases neighboring interneurons may release sufficient 12-(S)-HPETE to depress synapses, even when postsynaptic processes are blocked in the recorded cell. It is alternatively possible that perhaps the heterogeneity of hippocampal interneurons could also account for these data (Freund and Buzsaki, 1996; Parra et al., 1998).

The simplest model to account for our results is that synaptic stimulation releases glutamate that activates group 1 mGluRs producing 12-(S)-HPETE, which may act as a retrograde messenger (Feinmark et al., 2003). 12-(S)-HPETE in turn may open TRPV1 channels on the presynaptic glutamatergic terminals of CA1 and/or CA3 pyramidal cells that synapse onto interneurons (Figure 8D). How might activation of a Ca²⁺-permeable ion channel lead to persistent synaptic depression? Calcium entry through TRPV1 channels on glutamatergic terminals could initiate a signaling cascade responsible for the persistent downregulation of glutamate release observed during LTD. In dorsal root ganglion neurons, TRPV1 channel opening triggers calcineurin activation, which then rapidly depresses multiple voltage-gated calcium channels (Wu et al., 2005, 2006). Moreover, presynaptic NMDARs are required for spike-timing dependent LTD in neocortical neurons (Sjostrom et al., 2003) and Ca²⁺ arising from presynaptic activity is required for LTD at striatal synapses (Singla et al., 2007), suggesting that presynaptic Ca²⁺ signals are required to initiate these forms of LTD as well. However, in both of these examples, co-active CB1 receptors are also required for LTD, whereas CB1 receptors are not required for LTD at excitatory synapses onto hippocampal interneurons, since AM251 was ineffective in blocking this form of LTD. The model we present is the simplest to account for all of our data; however, while we report here that functional TRPV1 receptors are present on CA3 and CA1 pyramidal cell bodies, TRPV1 receptors are also expressed in glial cell populations (Doly et al., 2004; Kim et al., 2006), so it remains possible that an alternative, more complex signaling pathway is involved.

TRPV1 was first identified as a heat-sensitive ion channel in peripheral sensory neurons (Caterina et al., 1997). The temperature threshold of 43°C for TRPV1 channels (Caterina et al., 1997) is normally outside the brain’s physiological range, but the sensitivity of the channel to heat and other activating stimuli can be modulated by endogenous lipids and by the phosphorylation state of the channel (Vellani et al., 2001; Benham et al., 2003). It is therefore conceivable that during fever TRPV1 channels in the hippocampus may be activated, producing LTD at interneuron synapses. Depression of these synapses is expected to increase the excitability of innervated pyramidal cells. In this regard, it is intriguing that the in vivo treatment of animals with SR141716A after the induction of febrile seizures reduced hyperexcitability in hippocampal area CA1 and prevented the emergence of long-term limbic hyperexcitability (Chen et al., 2007). Our data suggest that the blockade of TRPV1 receptors could contribute to the anticonvulsant effect of SR141716A. The selective depression of excitatory synapses on interneurons but not on CA1 pyramidal cells that we report suggests that TRPV1 receptors are differentially distributed on hippocampal excitatory afferents and offers the potential to target hippocampal inhibitory circuits selectively through TRPV1 receptors.