Roles of distinct glutamate receptors in induction of anti-Hebbian long-term potentiation

LTP has attracted intense interest for over three decades, not least because it remains the leading candidate mechanism for memory encoding. Four important advances provide compelling justification for the substantial effort devoted to this phenomenon. The first advance actually pre-dates the discovery of LTP, and comes from the theoretical argument put forward by Donald Hebb in the 1940s that networks of idealized neurons could store information if the connections between them obeyed a simple rule (Hebb, 1949). This rule can be approximated by the maxim ‘cells that fire together wire together’. The second advance is the discovery that LTP induction in principal cells is essentially ‘Hebbian’, because it is triggered by the conjunction of presynaptic glutamate release with postsynaptic depolarization (Wigstrom & Gustafsson, 1986). The third is the discovery that N-methyl-d-aspartic acid (NMDA) receptors are necessary for LTP induction in principal cells (Collingridge et al. 1983) and the fourth is the finding that some behavioural tests of memory function are profoundly disrupted by NMDA receptor blockers (Morris, 1989). At a mechanistic level, these observations are united by the fact that NMDA receptors themselves act as coincidence detectors, because they are normally blocked by extracellular Mg²⁺ in a voltage-dependent manner (Nowak et al. 1984), and only respond to presynaptic glutamate release when the neuron is depolarized. Although many questions remain, the link from Hebbian NMDA receptor-dependent LTP to memory encoding is one of the most compelling achievements of late 20th century neuroscience.

Nevertheless, LTP is not monolithic, and there are many convincing reports describing other forms of long-lasting activity-dependent plasticity of synaptic transmission, which differ with respect to the type of transmission being altered (GABAergic, for instance – Caillard et al. 1999), or the sign of the change (depression instead of potentiation – Dudek & Bear, 1992). Other forms of plasticity result in qualitative rather than quantitative changes in transmission, such as alterations in the molecular and pharmacological identity of receptors at the synapse (Liu & Cull-Candy, 2000).

We have recently added to this list by describing a strengthening of glutamatergic synaptic transmission that comes about not as a consequence of pairing pre- with postsynaptic activity, but by the conjunction of presynaptic activity with postsynaptic quiescence (Lamsa et al. 2007). More precisely, this form of LTP can be induced by deliberately hyperpolarizing the postsynaptic neuron via a recording pipette at the same time as the presynaptic glutamatergic axons are made to fire. Alternatively, it can be induced without manipulation of the postsynaptic voltage, by evoking brief bursts of activity in presynaptic axons at an intensity that is insufficient to depolarize the postsynaptic neuron by more than a few millivolts from rest. Although the exact dependence of induction on postsynaptic voltage has yet to be mapped out, the striking feature of this form of LTP is that postsynaptic depolarization prevents its induction, and none of the conventional Hebbian induction protocols result in potentiation at the same synapses. We have therefore termed LTP resulting from the conjunction of presynaptic activity with relative postsynaptic hyperpolarization ‘anti-Hebbian’.

We have recently reviewed much of the phenomenology of anti-Hebbian LTP, its likely expression mechanisms, its possible computational significance, and the reasons why (in our opinion) it has been overlooked by others (Kullmann & Lamsa, 2007). Briefly (i) it can be induced in several classes of interneurons in strata oriens and pyramidale by high- or low-frequency stimulation patterns applied to axon collaterals of local pyramidal neurons, as long as the postsynaptic membrane potential is kept negative to the action potential threshold (whether firing itself is the key determinant of plasticity remains to be determined); (ii) it appears to be expressed presynaptically; (iii) it may contribute to shaping the temporal structure of information flow through the hippocampal circuitry; and (iv), it is very difficult to induce in the whole-cell recording configuration of the patch-clamp method (presumably because of rapid wash-out of as-yet undetermined cytosolic substances), and its anti-Hebbian nature depends critically on cytoplasmic polyamines, which are frequently omitted from pipette solutions.

This report focuses instead on the roles of different glutamate receptors in the induction of anti-Hebbian LTP.

NMDA receptors

Of the ionotropic receptors, a role for NMDA receptors can be discounted on three grounds. First, the requirement for postsynaptic hyperpolarization is precisely opposite to the conditions needed to relieve the Mg²⁺ block of these receptors; second, anti-Hebbian LTP can be induced by pairing high-frequency tetanization of local axon collaterals with postsynaptic hyperpolarization in the presence of NMDA receptor blockers; and third, when cells exhibiting anti-Hebbian LTP were re-patched in whole-cell voltage-clamp mode, the NMDA receptor-mediated component of transmission was uniformly small (Lamsa et al. 2007).

Ca²⁺-permeable AMPA receptors

Activation of Ca²⁺-permeable α-amino-3-hydroxy-5-methylisoxazole propionic acid (CP-AMPA) receptors, on the other hand, is extremely likely to play an instructive role in anti-Hebbian LTP. By ‘instructive’ we mean that activation of these postsynaptic receptors is a necessary link between the induction stimulus (pairing of presynaptic activity and relative postsynaptic hyperpolarization) and LTP. This level of implication is analogous to that ascribed to NMDA receptors in Hebbian LTP, and is greater than a ‘permissive’ role, which would simply state that the signalling mechanism needs to be constitutively active to allow another cascade to trigger LTP induction. What is the evidence that CP-AMPA receptors play such a privileged role? First is the fact that the Ca²⁺ permeability and voltage dependence of CP-AMPA receptors make them an obvious candidate: they show pronounced inward rectification because they are blocked by polyamines upon depolarization, reducing Ca²⁺ influx much more than expected from a simple decrease in driving force (Bowie & Mayer, 1995; Donevan & Rogawski, 1995; Kamboj et al. 1995; Koh et al. 1995). Second, in all cases where we have been able to induce anti-Hebbian LTP and subsequently re-patch the interneuron in voltage-clamp mode with a spermine-containing pipette solution, the evoked fast synaptic currents have shown steep inward rectification (Lamsa et al. 2007). Conversely, where anti-Hebbian LTP could not be induced, the AMPA receptor-mediated component has often (although not always) shown a relatively linear current–voltage relation, implying that the AMPA receptors were Ca²⁺ impermeable. Third, blocking AMPA receptors with the broad-spectrum antagonist 2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline-2,3-dione (NBQX) prior to application of an anti-Hebbian pairing protocol uniformly abolished LTP, as evidenced by the equivalent recovery of transmission in test and control pathways following wash-out of NBQX (Fig. 1). Fourth, the voltage dependence of LTP induction was radically altered when rectification of CP-AMPA receptors was removed. We showed this by, initially, identifying interneurons competent to exhibit anti-Hebbian LTP by pairing one pathway with hyperpolarization. Subsequently, we re-patched the cells in whole-cell mode with a polyamine-free pipette solution, and showed that high-frequency stimulation of another pathway led to potentiation when the cell was held at a positive membrane potential. In contrast, if the cell was re-patched with a polyamine-containing solution, the anti-Hebbian voltage dependence of LTP induction in the second pathway was maintained.

Nevertheless, the evidence that activation of CP-AMPA receptors have such a privileged role is not complete. It remains to be shown that their activation is sufficient to induce LTP, for instance by applying exogenous agonists in the absence of presynaptic exocytosis. Moreover, there remains considerable uncertainty about the steepness with which depolarization shuts off Ca²⁺ influx via CP-AMPA receptors. Some of this uncertainty hinges on the natural concentration of polyamines in the cytoplasm, which may even be dynamically regulated (Aizenman et al. 2002). The polyamine sensitivity of CP-AMPA receptors has also recently been shown to depend not only on their subunit composition but also on their interactions with trans-membrane AMPA receptor regulatory proteins (TARPs – Soto et al. 2007).

A potential problem for the hypothesis that rectification of CP-AMPA receptors is necessary and sufficient to make LTP anti-Hebbian comes from the finding that polyamine blockade can be attenuated by repetitive synaptic activity (Rozov & Burnashev, 1999). How then can high-frequency presynaptic tetani, which we showed to be a powerful tool to induce LTP when paired with hyperpolarization, fail to cause substantial Ca²⁺ influx via CP-AMPA receptors when the postsynaptic neuron is depolarized? A possible explanation is that activity-dependent relief of polyamine-dependent rectification itself requires hyperpolarization (Bowie et al. 1998; Rozov et al. 1998). Thus, if the interneuron is already depolarized when presynaptic axons fire, the receptors may remain blocked.

Ultimately it will be important to measure activity-dependent Ca²⁺ influx at different membrane potentials at synapses where anti-Hebbian LTP can be induced. This is far from trivial, because the success rate for evoking LTP rapidly diminishes when recording in whole-cell mode, and so conventional methods to introduce Ca²⁺ indicators and manipulate the membrane potential cannot easily be applied. The precise shape of the dependence of synaptic strengthening on Ca²⁺ entry during pairing also remains to be determined. If it is highly non-linear (as is probably the case for NMDA receptor-dependent plasticity – Perkel et al. 1993), this too could potentially explain why baseline low-frequency activity does not entrain the LTP cascade.

Kainate receptors

The roles of kainate receptors in brain function remain unclear. Although these receptors are expressed abundantly, and depolarize interneurons particularly powerfully, they only seem to respond to glutamate released from excitatory axons when these fire intensely. Several reports describe a small, slow kainate receptor-mediated component to glutamatergic transmission at various synapses in response to brief high-frequency bursts of stimuli (Castillo et al. 1997; Vignes & Collingridge, 1997; Frerking et al. 1998; Mulle et al. 2000). The slow kinetics remain difficult to reconcile with studies of either native or heterologously expressed kainate receptors in response to application of agonists, which show kinetics similar to those of AMPA receptors. Kainate receptors are also present presynaptically, and roles for these receptors in heterosynaptic interactions or even as autoreceptors have been proposed (Kullmann, 2001). Recently, Cossart and co-workers have reported that relatively fast kainate receptor-mediated spontaneous action-potential-independent excitatory signals can be detected in interneurons (Cossart et al. 2002), in particular in oriens-lacunosum moleculare (O-LM) cells (Goldin et al. 2007). These interneurons lie in stratum oriens, receive glutamatergic innervation mostly from local pyramidal cells via dendrites running in strata oriens and alveus, and innervate the apical dendrites of pyramidal cells in stratum lacunosum-moleculare (Blasco-Ibanez & Freund, 1995). They are the most abundant neurons where we have reported anti-Hebbian LTP (Lamsa et al. 2007). Kainate receptors, in common with AMPA receptors, can be Ca²⁺ permeable and inwardly rectifying, depending on their subunit composition and degree of post-transcriptional editing (Bernard et al. 1999; Bleakman, 1999; Lerma et al. 2001). This raises the question whether Ca²⁺-permeable kainate receptors play a role in anti-Hebbian LTP induction. We have recently addressed this by applying a lower concentration of NBQX, which, at 1 μm, is relatively selective for AMPA receptors (Bureau et al. 1999). When the anti-Hebbian LTP induction protocol was applied to one pathway in the presence of NBQX, no potentiation emerged when comparing to an un-conditioned pathway after washing out the blocker (Oren I, Somogyi P, Nissen W, Kullmann DM & Lamsa K, unpublished observations). Conversely, when the protocol was applied in the presence of the highly selective GluR5-subtype-specific kainate receptor blocker UBP-302, LTP was readily elicited. A similar test for the role of GluR6-containing receptors, which have also been reported in interneurons (Mulle et al. 2000; Paternain et al. 2000), is hampered by the lack of specific blockers. Nevertheless, overall the results argue that CP-AMPA, but not kainate, receptors are necessary for anti-Hebbian LTP induction.

Group I metabotropic glutamate receptors

What about group I metabotropic glutamate receptors (mGluR1 and mGluR5)? mGluR1 is abundantly expressed in O-LM cells (Baude et al. 1993), although is not uniformly prominent in other types of interneurons that exhibit anti-Hebbian LTP, some of which also occur in stratum pyramidale (Lujan et al. 1996; Ferraguti et al. 2004). When both mGluR1 and mGluR5 were blocked pharmacologically, anti-Hebbian LTP could not be induced in a small sample of interneurons in stratum oriens with proximal dendrites running in the plane of the strata, and which may have included some O-LM cells (Lamsa et al. 2007). What does this mean for the role of mGluRs in anti-Hebbian LTP induction? It is highly unlikely that they can account for the anti-Hebbian nature of LTP, because, in contrast to CP-AMPA receptors, they are not thought to be voltage dependent. Are they therefore merely permissive? Group I mGluRs tend to occur in a perisynaptic annulus (Lujan et al. 1996), and so their occupancy is generally thought to be enhanced by intense glutamate release such as occurs with high-frequency presynaptic activity. Although pairing postsynaptic hyperpolarization with presynaptic tetanic stimulation is a highly effective anti-Hebbian LTP induction protocol, it can also be induced by pairing hyperpolarization with low-frequency stimulation at 5 Hz (Lamsa et al. 2007). Whether the occupancy of group I mGluRs is elevated substantially above baseline with 5 Hz stimulation is a matter of speculation. Further work is also required to understand whether blocking mGluR1 or mGluR5 individually has other effects on the induction of anti-Hebbian LTP induction.

Group I mGluRs may have a more prominent role in another form of activity-dependent synaptic plasticity that has been studied extensively by Lacaille and co-workers in interneurons in stratum oriens, including O-LM cells. This is induced by repetitive pairing of bursts of high-frequency presynaptic stimulation with brief postsynaptic depolarization, and is therefore arguably Hebbian (Perez et al. 2001). Hebbian LTP in these neurons is independent of NMDA receptors and is prevented by selective blockade of mGluR1. Although this, on its own, only argues for a permissive role of mGluR1, Lacaille and co-workers have gone further to show that exogenous activation of the receptors induces Ca²⁺ elevations that are accentuated by depolarization, through an indirect cascade involving transient receptor potential channels and intracellular Ca²⁺ stores, and which also interact with Ca²⁺ influx via CP-AMPA receptors (Topolnik et al. 2005, 2006).

The relationship between Hebbian mGluR-dependent LTP and anti-Hebbian LTP remains to be elucidated. The finding that Ca²⁺ influx with weak presynaptic stimulation principally evokes Ca²⁺ via CP-AMPA receptors, while strong stimulation engages an mGluR1-linked cascade (Topolnik et al. 2005), suggests that either the anti-Hebbian or the Hebbian form of LTP might be induced in the same cell depending on the pattern of activity in presynaptic axons. Both of these forms of NMDA receptor-independent LTP appear to be expressed presynaptically (Perez et al. 2001; Lapointe et al. 2004; Lamsa et al. 2007), implying that the induction mechanisms converge on the same cascade.

Conclusion

An instructive role for Ca²⁺-permeable AMPA receptors in anti-Hebbian LTP is supported by abundant evidence. Indeed, this form of use-dependent plasticity provides a teleological explanation for why these receptors should have evolved to allow Ca²⁺ influx preferentially at hyperpolarized voltages. Of course, this is not the only role for these receptors, since they have been implicated in other forms of synaptic plasticity (Isaac et al. 2007), as well as contributing to neuronal death in response to various insults (Pellegrini-Giampietro et al. 1997), although the latter is arguably a counter-adaptive phenomenon. Neither NMDA receptors nor kainate receptors appear to be necessary for anti-Hebbian LTP induction. As for group I mGluRs, their roles require further work: they are at least permissive in O-LM cells, but they may have additional roles in this and other forms of LTP in interneurons.