Long-term potentiation in hippocampal oriens interneurons: postsynaptic induction, presynaptic expression and evaluation of candidate retrograde factors

Several types of hippocampal interneurons exhibit a form of long-term potentiation (LTP) that depends on Ca2+-permeable AMPA receptors and group I metabotropic glutamate receptors. Several sources of evidence point to a presynaptic locus of LTP maintenance. The retrograde factor that triggers the expression of LTP remains unidentified. Here, we show that trains of action potentials in putative oriens-lacunosum-moleculare interneurons of the mouse CA1 region can induce long-lasting potentiation of stimulus-evoked excitatory postsynaptic currents that mimics LTP elicited by high-frequency afferent stimulation. We further report that blockers of nitric oxide production or TRPV1 receptors failed to prevent LTP induction. The present results add to the evidence that retrograde signalling underlies N-methyl-d-aspartate (NMDA) receptor-independent LTP in oriens interneurons, mediated by an unidentified factor.


Introduction
Plasticity of glutamatergic synapses onto hippocampal interneurons has been proposed to play several roles, including stabilizing network excitability and preserving the fidelity of spatio-temporal processing in the face of long-term potentiation (LTP) at synapses among excitatory neurons [1][2][3]. LTP and long-term depression (LTD) in the inhibitory circuitry may also affect the input-output relationship of principal cells in qualitatively different ways than that achieved by plasticity restricted to excitatory neurons. Plasticity of inhibition may have further adaptive roles particular to specific types of interneurons [4]. For example, LTP of glutamatergic synapses on somatostatin-positive interneurons located in stratum oriens, which tend to innervate targets in strata radiatum and lacunosum-moleculare, is likely to enhance CA1 inputs from CA3 by disinhibiting Schaffer-collateralassociated interneurons, while concomitantly inhibiting extrahippocampal perforant path inputs to the apical dendrites of CA1 pyramidal neurons [5].
Induction of LTP at glutamatergic synapses on several types of interneurons in stratum oriens is independent of N-methyl-D-aspartate (NMDA) receptors, but requires Ca 2þ influx through Ca 2þ -permeable a-amino-3-hydroxy-5methyl-4-isoxazolepropionic acid (AMPA) receptors as well as activation of group I metabotropic glutamate receptors (mGluRs) [6 -12]. Among these are interneurons with dendrites running parallel to stratum pyramidale that project an axon to stratum lacunosum-moleculare [13]. Oriens-lacunosum-moleculare (O-LM) cells can be further recognized by their regular firing pattern upon injection of depolarizing current, voltage sag upon injection of hyperpolarizing current and expression of somatostatin and mGluR1a [14][15][16][17]. An efficient LTP induction stimulus is to hyperpolarize the postsynaptic membrane potential while stimulus trains are delivered to axon collaterals of pyramidal neurons running in the alveus. This 'anti-Hebbian' LTP protocol has been hypothesized to maximize Ca 2þ influx via rectifying AMPA receptors [8,11], although nicotinic receptors have also been proposed to play a role [18,19]. NMDA receptor-independent LTP in stratum oriens interneurons is associated with a decrease in the rate of failures of stimulus-evoked transmission, a decrease in the paired-pulse ratio (PPR) of excitatory postsynaptic currents or potentials (EPSC/Ps), a decrease in normalized coefficient of variation (equivalently, an increase in the parameter CV 22 ) and an increase in estimated AMPA receptor occupancy [8,10,13]. All these observations strongly argue for a presynaptic locus of expression of LTP, in striking contrast with conventional NMDA receptor-dependent LTP at glutamatergic synapses on pyramidal neurons [20]. A recent report has however suggested that anti-Hebbian LTP in parvalbumin-positive interneurons may also be expressed postsynaptically [21].
The postsynaptic induction requirements and evidence for presynaptic expression of LTP on O-LM cells and other interneurons in stratum oriens imply the existence of a retrograde messenger. Identification of this factor has been hampered by the relative difficulty of eliciting LTP while recording in the whole-cell configuration of the patch-clamp method [8,11]. We therefore first set out to identify recording conditions where LTP could be reliably elicited in whole-cell mode and systematically interleaved control experiments throughout this study whenever pharmacological interventions were applied. We confirm that LTP requires postsynaptic Ca 2þ and show that it can be induced by trains of action potentials elicited in the interneuron alone, implying that multiple sources of Ca 2þ may converge on the induction trigger. We provide further evidence of presynaptic expression, and test two candidate retrograde factors.

Material and methods
(a) Brain slicing Horizontal brain slices were prepared from postnatal day 21-25 mice in accordance with the UK Animals (Scientific Procedures) Act 1986. Slices (300 mm thick) were cut using a vibrating microtome (Leica VT 1200) in an ice-cold solution bubbled with 95% O 2 and 5% CO 2
Interneurons within stratum oriens of CA1 with dendrites running parallel to stratum pyramidale were patch-clamped with 4 -5 MV glass pipettes, to achieve an access resistance of less than 20 MV. Data were acquired using a Multiclamp 700 B amplifier (Molecular Devices), low-pass filtered (4 -5 kHz) and digitized at 10 -20 kHz (National Instruments).
Neurons were held in current clamp mode, and current was injected when needed to maintain the membrane potential between 270 and 275 mV. Passive membrane properties and firing patterns were tested with hyperpolarizing and depolarizing steps. Cells that did not display a voltage sag, deep afterhyperpolarization (AHP) and regular firing pattern typical of O-LM cells (figure 1a) were discarded.
Stimuli were delivered via two concentric bipolar electrodes, connected to constant current isolated stimulators (Digitimer), positioned in the alveus/stratum oriens border, 100-300 mm either side of the recorded interneuron. The stimulus duration was 100 ms and its intensity was adjusted between 20 and 320 mA to elicit an EPSP with peak amplitude of approximately 5 mV. Paired stimuli with an interstimulus interval of 50 ms were alternately delivered via each electrode every 20 s. The LTP induction protocol consisted of 100 stimuli at 100 Hz, delivered twice with a 20 s interval. We tested a second protocol consisting of 20 depolarizing current injections (500 pA, 500 ms, interval 5 s) to elicit trains of action potentials at a frequency of approximately 60 Hz. LTP was studied by measuring the initial EPSP slope (2-4 ms from onset) to restrict attention to monosynaptic transmission [22 -24]. For pooled data, the initial slopes of EPSPs evoked in the test and control pathways were normalized to their baselines and paired t-tests applied, with significance set at p , 0.05. Paired-pulse ratios (PPRs) were calculated in cells which exhibited LTP by averaging approximately 20 consecutive traces and expressed as the ratio of the second to the first EPSP slope. Cells with an action potential on the second pulse were omitted from PPR analysis. Failures of evoked transmission were identified by sight, and when averaged together showed no systematic deviation from the baseline, although were sometimes followed by polysynaptic EPSPs. Spontaneous EPSP frequency and amplitude were analysed using CLAMPFIT v. 10.2 (Axon Instruments) using the first 100 events prior to LTP induction, starting at the beginning of the experiment, and 100 events 20 min after the LTP induction protocol. In both cases, the spontaneous EPSPs were obtained from periods immediately prior to stimulation. Stimulus control, and data acquisition and analysis, were achieved with custom software (National Instruments, LABVIEW) and PCLAMP v. 10 software. Data are presented as mean + s.e.m.

(c) Anatomical analysis
Some interneurons were filled with biocytin during whole-cell recordings. Brain slices were then fixed in 4% paraformaldehyde overnight. After washing in phosphate-buffered saline (PBS), slices were incubated in PBS containing 0.3% triton and 0.1% streptavidin-alexa-488 for 3 h at room temperature, washed in PBS and mounted on coverslips with Vectashield mounting medium. Cells were visualized with an AxioImager microscope (Zeiss) and drawn in Photoshop.

Results (a) Long-term potentiation at glutamatergic synapses onto putative oriens-lacunosum-moleculare interneurons
We recorded from cells with horizontal dendrites in stratum oriens in whole-cell current clamp mode and only continued the experiment if they showed a voltage sag with hyperpolarizing current injection, and a regular firing pattern with a deep AHP upon depolarizing current injection, typical of O-LM cells [8,13] (figure 1a). Of the 12 cells that were filled with biocytin and subsequently reconstructed, nine showed the typical O-LM morphology with an axon projecting into stratum lacunosum-moleculare (figure 1b). No axon was observed in the remaining three reconstructed cells, suggesting that it may have been cut during the slicing procedure. We evoked monosynaptic EPSPs via two stimulation electrodes positioned at either side of the interneuron and recorded a baseline period lasting at least 5 min. Two highfrequency stimulation trains were then delivered via one electrode (100 Hz, 1 s, Â2 separated by 20 s), while stimulation via the other electrode was interrupted. The membrane potential was allowed to vary during the tetanus and the cells fired at an average rate of 34 + 22 Hz (mean + s.d.). In the example cell (figure 1c), the control pathway EPSP slope remained around 0.43 mV ms 21 , while in the tetanized pathway it approximately doubled from 0.83 to 1.6 mV ms 21 (figure 1c).
When repeated in 29 similar experiments, we observed a robust potentiation of approximately 77% in the tetanized pathway ( p , 0.001, paired t-test; figure 2a).

(b) Evidence of presynaptic long-term potentiation expression
In previously unidentified stratum oriens interneurons, LTP has been proposed to have a presynaptic locus of expression [7,8]. LTP in identified O-LM cells is associated with an increase in CV 22 [13], consistent with an increase in glutamate release probability, although this could also be explained by postsynaptic unsilencing of AMPA receptor clusters [25]. We assessed several parameters to determine whether a presynaptic locus of LTP expression holds true for the cells recorded in this study. The failure rate decreased by approximately half in the tetanized pathway ( p , 0.005), with no change in the control pathway ( p ¼ 0.29; figure 2b). LTP was also associated with a decrease in paired-pulse facilitation in the tetanized pathway ( p ¼ 0.002), with no change in the control pathway ( p ¼ 0.15; figure 2c). We also observed an increase in the frequency of spontaneous EPSPs after LTP induction, with a significant decrease in inter-event interval ( p , 0.05; figure 2d,e) and a non-significant trend towards an increase in amplitude ( p ¼ 0.11; figure 2f ). These findings argue strongly for an increase in presynaptic transmitter release.
(c) Long-term potentiation requires postsynaptic Ca 2þ LTP in horizontal oriens interneurons requires both Ca 2þ influx through Ca 2þ -permeable AMPA receptors and activation of group I mGluRs, with evidence for both Ca 2þ release from intracellular stores and Ca 2þ influx through transient receptor potential channels [6,8,[10][11][12][13]. We took advantage of the whole-cell recording mode to chelate intracellular Ca 2þ , by substituting BAPTA (25 mM) for EGTA (0.2 mM). Ca 2þ chelation prevented LTP induction (n ¼ 8, figure 3a), while interleaved control experiments continued to exhibit robust LTP ( p ¼ 0.005, n ¼ 9; figure 3b). High-frequency stimulation was followed by an increase in PPR when Ca 2þ was chelated in postsynaptic neurons ( p ¼ 0.02, n ¼ 8; figure 3c). This is opposite to the pattern seen in control experiments, although the decrease in PPR in this set of experiments did not reach significance ( p ¼ 0.29, n ¼ 9; cf. figure 2c). The increase in PPR was unexpected and may reflect additional presynaptic plasticity unrelated to NMDA receptorindependent LTP. Inclusion of 25 mM BAPTA also prevented the increase in spontaneous EPSP frequency ( p ¼ 0.76, n ¼ 6; figure 3d), which was seen in interleaved controls ( p ¼ 0.05, n ¼ 8; figure 3d).
(d) Long-term potentiation can be induced by postsynaptic action potentials alone We asked whether postsynaptic Ca 2þ influx via voltage-gated channels could induce LTP. EPSPs were evoked as before, but rather than delivering tetanic stimulation, twenty depolarizing currents (500 pA for 100 ms) were injected via the recording rstb.royalsocietypublishing.org Phil. Trans. R. Soc. B 369: 20130133 pipette to elicit action potential trains at approximately 60 Hz. This protocol was followed by a persistent increase in EPSP slope to approximately 150% of control, which lasted for at least 20 min ( p , 0.001, n ¼ 15; figure 4a). The potentiation elicited by action potential trains was associated with a decrease in paired-pulse ratio ( p ¼ 0.004, n ¼ 15; figure 4b) and spontaneous EPSP inter-event interval ( p ¼ 0.015, n ¼ 8; figure 4c). We asked whether the potentiation evoked by action potential trains occluded LTP. After eliciting action potential trains, tetanic stimulation failed to induce a further increase in EPSP slope (n ¼ 11; figure 4d). Similarly, action potential trains failed to increase EPSP slope after tetanus-induced LTP (n ¼ 8; figure 4e).
The two-way occlusion and similarity of effects on PPR and spontaneous EPSP frequency suggest that action potential trains and tetanic LTP converge on a common synaptic plasticity induction mechanism. However, an alternative possibility is that 'washout' of cytoplasmic constituents impaired induction of additional plasticity after a delay. To address this hypothesis, we performed a further set of experiments where high-frequency stimulation was delivered to the control pathway 30 min after inducing LTP in the test pathway. The tetanus-induced increases in EPSP slope in each pathway were not significantly different from one another ( p ¼ 0.92; unpaired t-test, n ¼ 15; figure 4f ), arguing against washout as an alternative explanation for the data summarized in figure 4d,e.

(e) No evidence of involvement of nitric oxide and transient receptor potential vanilloid 1 channel receptors in long-term potentiation
The dependence of LTP induction on postsynaptic Ca 2þ , together with the evidence for a presynaptic locus of expression, suggests that a retrograde message signals from the postsynaptic interneuron to presynaptic glutamatergic boutons to increase glutamate release. A prominent candidate retrograde messenger that has been invoked in other examples of presynaptic LTP is nitric oxide (NO) [26][27][28][29][30][31]. Consistent with NO triggering LTP is the finding that NMDA receptor-independent LTP can be elicited in NO synthase (NOS)-positive ivy cells [12]. Direct evidence of a role of NO in LTP in unidentified stratum oriens interneurons has, moreover, been reported [32]. We incubated slices for at least 4 h in 100 mM NG-nitro-Larginine (L-NNA), an inhibitor of NOS, prior to and during LTP experiments. Tetanus-induced LTP was no different in treated slices ( p , 0.05, n ¼ 8; figure 5a) than in interleaved controls (n ¼ 11; figure 5b).
To examine whether a similar mechanism may be involved in potentiating excitatory synapses onto putative O-LM cells, we applied the TRPV1 receptor antagonist, 100 nM 5 0 -iodoresiniferatoxin, prior to and throughout the induction of LTP. Blocking TRPV1 had no effect on LTP (n ¼ 10; figure 5c) compared with interleaved controls (n ¼ 15; figure 5d).
We thus obtained no evidence to support a role for either NO or TRPV1 receptors in the induction of tetanus-induced LTP.

Discussion
The present results add to the evidence that NMDA receptorindependent LTP in stratum radiatum interneurons depends on postsynaptic Ca 2þ and is expressed through a presynaptic increase in glutamate release probability. Although this argues for a retrograde factor, we have found no positive evidence to support a role for two candidate signalling mechanisms that have been invoked previously. Neither blockade of NOS nor of TRPV1 receptors affected LTP.
We focused on 'horizontal' interneurons with electrical properties typical of positively identified O-LM cells. We   reconstructed only a subset of interneurons but confirmed that their dendritic arborization was confined to stratum oriens, and that they projected an axon to stratum lacunosummoleculare. Because a positive identification was not available for all interneurons, we refer to the cells in this study as putative O-LM interneurons. Importantly, however, we systematically interleaved all experiments where either BAPTA or NOS or TRPV1 blockade were tested with control experiments. Moreover, we used a control pathway to verify that LTP was confined to the tetanized pathway. Another refinement in this study was to record from interneurons in whole-cell mode, with a pipette solution devoid of polyamines. Although prolonged whole-cell recording profoundly diminishes the success rate for induction of LTP, omitting polyamines removes the preferential requirement for hyperpolarization during presynaptic stimulation (anti-Hebbian LTP) [8]. The induction protocol is thus more akin to that used in several other studies that have reported on LTP in stratum oriens interneurons [7,10].
For the purpose of this study, we suggest that the relatively unphysiological induction protocol is justified by the ability to manipulate postsynaptic Ca 2þ chelation and introduce biocytin without re-patching neurons. Surprisingly, trains of postsynaptic action potentials could induce a persistent potentiation of transmission. Although cells fired during the LTP induction protocol, the total number of action potentials induced by postsynaptic current injection was approximately 10-fold higher. This difference may explain why tetanic stimulation-induced LTP was pathway-specific. Although back-propagating action potentials have been invoked in LTP induction in principal cells, this also requires presynaptic glutamate release to activate NMDA receptors [48]. Postsynaptic Ca 2þ influx alone only in principal cells elicits a transient potentiation of transmission [49]. In this study, postsynaptic activity was sufficient to induce a longlasting potentiation in the absence of presynaptic stimulation. However, we cannot rule out the possibility that spontaneous glutamate release during the action potentials was sufficient for minimal activation of Ca 2þ -permeable AMPA receptors and/or group I mGluRs. Postsynaptic hyperpolarization paired with pharmacological group I mGluR activation has also previously been reported to be sufficient to trigger a persistent potentiation in stratum oriens interneurons, which occludes anti-Hebbian LTP [11]. A direct comparison of these results is difficult because this study was carried out using whole-cell patch-clamp recordings. Nevertheless, we tentatively speculate that, in contrast to principal cells, several sources of Ca 2þ can converge on an LTP-inducing cascade at glutamatergic synapses on stratum oriens interneurons. The difference between these synapses and those on pyramidal neurons may to some extent be explained by the biochemical compartmentalization offered by dendritic spines in the latter.
The NOS inhibitor L-NNA had no effect on LTP induction. Although we did not perform additional control experiments to detect positive effects of L-NNA, this lends no support to the hypothesis that NO acts as a retrograde messenger in this system. Previous studies of LTP in principal cells have reported divergent results regarding the role of NO. Differences in species and strains, developmental stage, temperature and stimulus protocols have been invoked to explain some of the discrepancies [50][51][52][53][54]. Nevertheless, it is generally established that NO synthesis in pyramidal neurons is triggered by Ca 2þ influx through NMDA receptors [55,56]. LTP in O-LM cells is NMDA receptor independent [13]. While one study reported that an NOS inhibitor prevented LTP induction in alveus/oriens interneurons [32], this form of plasticity also required NMDA receptors. NMDA receptor-dependent LTP is not prominent in stratum oriens interneurons [8], but has been demonstrated in hippocampal interneurons in strata radiatum and pyramidale [21,24,57]. However, 'Hebbian' NMDA receptor-dependent LTP in interneurons is not associated with changes in PPR or receptor occupancy, arguing against presynaptic expression [21,24]. TRPV1 receptors have been reported in hippocampal neurons with mRNA, immunohistochemistry and radioligand binding [58][59][60][61][62], although data from a genetically altered TRPV1 reporter mouse suggest that TRPV1 receptors are not abundant [63]. The effect of presynaptic TRPV1 activation on synaptic transmission is also unclear, with some studies reporting a depression of synaptic release [33] through downregulation of voltage-gated Ca 2þ channels [64,65], and others reporting a potentiation of presynaptic release [34][35][36]. Although endogenous TRPV1 ligands can be released upon group I mGluR activation [33,40,46,47], and group I mGluR activation is required for NMDA receptor-independent LTP [7,10,11], we observed no evidence for TRPV1 involvement in LTP in this study: the blocker 5 0 -iodoresiniferatoxin, applied prior to and during the induction protocol at a concentration previously reported as effective in similar preparations [33], had no effect on LTP.
In conclusion, LTP induction in putative O-LM cells depends on postsynaptic Ca 2þ . It is associated with changes in PPR, failures of transmission and spontaneous EPSP frequency strongly suggestive of presynaptic expression. However, we found no evidence of a role of either NO or TRPV1 receptors in its induction. The retrograde messenger remains to be identified.