Regulation of Spike Timing-Dependent Plasticity of Olfactory Inputs in Mitral Cells in the Rat Olfactory Bulb

The recent history of activity input onto granule cells (GCs) in the main olfactory bulb can affect the strength of lateral inhibition, which functions to generate contrast enhancement. However, at the plasticity level, it is unknown whether and how the prior modification of lateral inhibition modulates the subsequent induction of long-lasting changes of the excitatory olfactory nerve (ON) inputs to mitral cells (MCs). Here we found that the repetitive stimulation of two distinct excitatory inputs to the GCs induced a persistent modification of lateral inhibition in MCs in opposing directions. This bidirectional modification of inhibitory inputs differentially regulated the subsequent synaptic plasticity of the excitatory ON inputs to the MCs, which was induced by the repetitive pairing of excitatory postsynaptic potentials (EPSPs) with postsynaptic bursts. The regulation of spike timing-dependent plasticity (STDP) was achieved by the regulation of the inter-spike-interval (ISI) of the postsynaptic bursts. This novel form of inhibition-dependent regulation of plasticity may contribute to the encoding or processing of olfactory information in the olfactory bulb.


Introduction
At the first station of central odor processing, the main olfactory bulb (MOB), signal processing is regulated by synaptic interactions between glutamatergic and GABAergic inputs of the mitral cells and tufted cells (M/T cells), which are the major projection neurons. The M/T cells receive both excitatory glutamatergic inputs from the olfactory sensory neurons (OSNs) and inhibitory GABAergic inputs from the interneurons, which predominantly consist of periglomerular cells (PGCs) and GCs [1,2]. The GABAergic inhibitory inputs from the interneurons deliver lateral or recurrent inhibition to the M/T cells to modulate incoming sensory inputs to an optimal level. Because the GCs outnumber the PGCs in the bulb and because only approximately 20% of the PGCs make connections with both presynaptic OSNs and postsynaptic M/T cells [3,4], most of the lateral inhibition is mediated by the GCs [1,5]. The lateral inhibition from the GCs is exerted via reciprocal dendrodendritic synaptic connections with the M/T cells and functions as a contrast enhancer to facilitate the discrimination of correlated ON inputs [6].
Long-term potentiation (LTP) and long-term depression (LTD) are two forms of synaptic plasticity that are considered as the cellular substrates for learning and memory [7,8]. LTP or LTD in the OSN-MC synapses in the main olfactory bulb can be produced by a brief tetanic or low frequency ON stimulation [9,10]. Recently, LTP of the field excitatory postsynaptic potentials (fEPSPs) in the glomerulus was induced by theta burst stimulation (TBS) of the ON [11]. Moreover, Hebbian spike timing-dependent plasticity (STDP), a form of plasticity that strictly requires a temporal correlation between the pre-and postsynaptic responses, was also induced in vivo in b-lobe neurons of the locust mushroom body through the electrical stimulation of Kenyon cells [12] and in the excitatory inputs to the GCs in the main rat olfactory bulb [13]. Despite these elegant investigations in different levels and species, there have been a lack of studies that directly examine STDP in the MCs. In the olfactory bulb, the lateral inhibition driven by the GCs tends to affect the dendritic depolarization induced by the excitatory inputs and either prevents or delays firing in the MCs or decreases the firing rate during stimulus presentation [14,15]. This change in the firing rate begins at the end of the spiking period induced by odor stimulation in vivo [15], indicating that the recent activity history could affect lateral inhibition between the MCs. However, little is known about whether and how prior persistent modification of the inhibition modifies the predisposition for subsequent induction of longlasting changes of the excitatory OSN-MC synapses. This form of synaptic plasticity regulation has not been fully investigated in the olfactory system. Because STDP is a form of synaptic plasticity whose induction is heavily dependent on the timing of the incoming spikes and the frequency of postsynaptic spikes when the spike bursts are induced by the pairing protocol [16], the presence of lateral inhibition onto the postsynaptic MCs may profoundly influence the induction of STDP and the total olfactory bulb output. Moreover, the timing of synaptic inhibition itself may also be regulated [17,18]. Therefore, the lateral inhibition exerted by the GCs plays critical roles in odor information processing and in olfactory learning and memory [5,6,19,20,21,22,23]. The mechanism by which the frequency of the intra-burst spikes is controlled by lateral inhibition is of great importance in understanding how sensory information is encoded and processed.
GCs receive two types of excitatory inputs on their proximal and distal dendrites [13,24]. The reciprocal dendrodendritic synapses with the MCs are the primary source of distal excitatory inputs, which mediate local dendrodendritic inhibition (DDI) [25]. Cortical feedback input is one of the major sources of proximal input and mediates the global top-down modulation of DDI in the olfactory bulb [24]. These two anatomically distinct excitatory inputs display persistent modification with opposing directions when subjected to the same stimulating protocol [13], which suggests that they may differentially exert their influence on the subsequent induction of synaptic plasticity in MC excitatory synapses and thus have different implications in regulating the total olfactory bulb output. Because the extent of granule cell (GC) function in the local versus global output modes can have an important impact on the computational role GC performs in the olfactory bulb circuit [1], the modification of the excitatory ON input by prior plasticity of the distinct excitatory inputs to GCs will profoundly influence the encoding and processing of odor information in the bulb.
In this study, we found that TBS could elicit differential plasticity of two different excitatory inputs to the GCs. Interestingly, the same protocol also induced a persistent modification in the lateral inhibition to the MCs with opposing directions. This bidirectional modification of the inhibitory inputs differentially regulated the predisposition of the subsequent induction of STDP of excitatory ON inputs to the MCs. Further evidence demonstrated that this regulation was achieved by the regulation of the spike frequency within the bursts employed by the pairing protocol for STDP induction. Thus, our results revealed one of the mechanisms by which the frequency of the intra-burst spikes is controlled in the sensory system. Because fine temporal burst structure is proposed to convey stimulus-related information to postsynaptic cells, this novel form of inhibitiondependent regulation of plasticity may contribute to the encoding or processing of olfactory information in the olfactory bulb.

Ethics Statement
The protocols for the animal care and use were approved by the Experimental Animal Ethics Committee at the Nanjing Medical University (permit number 20100582).
Fresh slices were incubated in the chamber with carbogenated ACSF and recovered at 30uC for 30 min and then maintained at room temperature.

Electrophysiological studies
Conventional whole-cell recordings in the current-clamp mode were made with patch pipettes containing (in mM) 140 Kmethylsulfate, 4 NaCl, 10 HEPES, 0.2 EGTA, 4 MgATP, 0.3 Na 3 GTP, and 10 phosphocreatine. The pH was adjusted to 7.4 with KOH. The micropipettes were made from borosilicate glass capillaries (Sutter Instrument Co.) and had resistances in the range of 5-8 MV. The cells were viewed under an upright microscopy (Eclipse E600-FN, Nomarsky, Nikon Corp., Tokyo, Japan) and recorded with an Axopatch-200B amplifier (Molecular Devices, Palo Alto, CA). Moreover, the MCs were identified by their morphology, size and location [23,26]. In the GCs recordings, the cells were selected from the granule cell layer (GCL) based on their small cell-body diameters (,10 mm) [27]. The olfactory bulb slices were perfused with 32uC ACSF that was bubbled continuously with carbogen (95% O 2 /5% CO 2 ). The EPSPs or inhibitory postsynaptic potentials (IPSPs) were recorded in the control ACSF in the absence of any receptor blockers to ensure that both the glutamatergic and GABAergic neurotransmission were intact. The membrane potentials (mV) of the MCs in the current-clamp mode were from 252 to 261 mV. A bipolar stimulating electrode (inside diameter, 25 mm, FHC Co., USA) was placed into the external plexiform layer (EPL) or granule cell layer (GCL), to evoke synaptic responses at the distal or proximal inputs, respectively. The current intensity of the test stimuli (0.01-0.30 mA) was set to produce half-maximal EPSPs (one-peak monosynaptic responses, with amplitudes between 1 and 5 mV). The basal evoked synaptic responses were produced by 100 ms electrical stimulation at 0.05 Hz except during the induction of STDP. The TBS consists of five bursts of five stimulations (intra bursts: 100 Hz; inter-burst 5 Hz) repeated 5 times at 0.1 Hz. The data were low-pass filtered at 2 kHz and acquired at 5-10 kHz. The series resistance was always monitored during the recording for fear that re-sealing of the ruptured membrane would cause changes in both the kinetics and amplitude of the EPSPs. The cells in which the resistance or capacitance deviated by .20% from the initial values were excluded from the analysis. The data were collected with the pClamp9.2 software and analyzed using Clampfit9.2 (Molecular Devices, Palo Alto, CA).
The STDP in the MCs was induced by pairing the EPSPs with postsynaptic spike bursts. The pairings were repeated 60 times at a 0.1 Hz stimulation and the EPSPs were evoked by stimulating the glomeruli that correlated with the recorded MCs. These postsynaptic bursts comprised of three spikes and were elicited by a current injection (intensity: 50-300 pA; duration: 50 ms). The inter-spike interval (ISI) in control was set as 19-21 ms (20.160.2 ms, n = 12; frequency approximately 50 Hz) by adjusting the intensity of the injected current. The time interval Dt was defined as the time between the onset of the compound EPSP and the onset of the action potential (AP) burst (first AP in the burst). The positive time window was defined as the EPSPs that occurred before the postsynaptic bursts, whereas the negative time window referred to the postsynaptic bursts that occurred before the EPSPs. The persistent potentiation or depression was defined as the percentage changes in EPSP or IPSP amplitude during the last 10 min of the recording after the repetitive stimulation. To examine whether direct activation of M/T cells also contribute to the changes in STDP, we usually used one or two single stimuli to EPL or GCL before TBS was delivered and examine whether EPSP can be elicited. In most of our recordings, TBS in EPL or GCL with moderate intensity usually could not directly activate M/T cells. Occasionally we observe EPSPs were induced in MCs. We discarded these cells and did not continue to perform further recording on these cells.
The data are presented as the mean 6 SEM. Paired Student's t tests were applied as statistical tests if not indicated otherwise, and the statistical significance was asserted for p,0.05.Within-group comparisons were performed using a two-tailed t test, and the difference between the groups was compared using ANOVA post hoc comparisons. An ANOVA post hoc LSD test was used when equal variances were assumed. The differences were considered significant when p,0.05 and significance for homogeneity of the variance test was set at 0.1.

Distinct long-term plasticity of excitatory inputs to the GCs
In order to investigate whether persistent modification of the inhibitory inputs regulates the predisposition of the subsequent STDP induction of excitatory ON inputs to the MCs, we need to induce both long-term plasticity of inhibitory inputs and STDP of excitatory ON inputs to the MCs. Since GCs are the main type of interneuron in the olfactory bulb that laterally inhibits the M/T cells, we first tried to induce long-term plasticity of excitatory inputs to the GCs and then determine whether this plasticity could in turn elicit long-term plasticity of inhibitory inputs to MCs. We performed whole-cell patch clamp recordings of evoked EPSPs in the GCs of MOB slices to examine whether long-term plasticity of excitatory inputs onto GCs (excitatory inputsRGCs synapses) could be elicited using a specific induction protocol (Fig. 1A). A TBS protocol (five 100-Hz bursts of 5 shocks, repeated at 5 Hz) was focally delivered to the distal or proximal excitatory inputs to the GCs to induce long-term plasticity of these inputs. The EPSPs were recorded in the current clamp mode without disturbing GABA A receptor-mediated inhibitory synaptic transmission. The bipolar stimulating electrode was placed in either the EPL or GCL within a 200-300 mm distance from the cell body of the MCs to stimulate distal or proximal excitatory inputs, respectively (Fig. 1A,D) [28]. Strowbridge and colleagues had employed a two-photon guided minimal stimulation to selectively activate the two inputs [13]. In order to ensure that we also activated a relatively homogenous population of presynaptic processes when stimulating either of the two sites (i.e., the specificity of the stimulations), we deliberately controlled the intensity of the local stimulation to ensure the independence of the stimulation on distinct sites. The synaptic responses we obtained from these two inputs displayed distinct properties in their current kinetics (rise time, GCL: 19.963.7 ms, EPL: 44.565.2 ms, p,0.001, Fig. S1B), similar to the observations obtained from the Strowbridge lab [13]. At the distal excitatory inputs in EPL that mediated local DDI, TBS elicited LTD of these inputs in 6 out of the 7 cells (78.062.3%, n = 6, p,0.001, paired-samples t test, p-values are between baseline and the last 10 minutes after induction; Fig. 1B, C). In contrast, the same protocol, when delivered to the GCL, produced a LTP of the proximal excitatory inputs in 7 out of the 8 cells (153.264.1%, n = 7, p,0.001, paired-samples t test; Fig. 1E, F). These two forms of synaptic plasticity were not due to a rundown of the EPSPs, because no persistent changes in the EPSP amplitude were observed when the TBS was absent (104.766.3%, n = 6, p.0.05, Fig. S2). In addition, the synaptic plasticity was not associated with any obvious changes in the input resistance or membrane potential (Fig. S3). The resting potentials of the GCs detected in this study (approximately 260 mV) were in the normal range of the resting potential (276 to 254 mV) reported by a previous study [29], indicating that the recorded GCs were healthy. Thus, distinct long-term plasticity of excitatory inputs to the GCs could be induced with identical TBS delivered to EPL or GCL.

Modulation of inhibition onto the MCs by plasticity of the GCs
Lateral inhibition in the olfactory bulb is largely mediated by reciprocal dendrodendritic synaptic connections between the MC lateral dendrites and the dendrites of inhibitory GCs. It is possible that plasticity at the excitatory inputs onto the GCs may alter their intrinsic property and/or driving force onto the MCs and, consequently, modulate the lateral inhibition onto the MCs. This potential modulation in lateral inhibition might contribute to the refinement of encoding or processing of olfactory information in the MCs. Thus, we further investigated whether the differential plasticity of GC excitatory inputs cause differential long-term plasticity of lateral inhibition in the MCs. The evoked IPSPs were recorded in the MCs by extracellularly stimulating the EPL or GCL (GCsRMCs synapses) before and after TBS ( Fig. 2A. 2D). These responses appeared to be mediated by GABA A receptors, as they exhibited a reversed polarity near 270 mV and were blocked by gabazine (10 mM; Fig. S4). We only analyzed the responses from the MCs that showed clear inhibitory responses to single stimulations (see sample traces in Fig. 2B, C). The TBS depressed the IPSPs evoked by the EPL stimulation for at least 30 min (65.762.8%, n = 6; p,0.001; Fig. 2B,C). Moreover, using the same protocol, the TBS delivered on to the GCL potentiated the IPSPs (124.563.3%, n = 7; p,0.001; Fig. 2E,F) [13]. These two forms of synaptic plasticity were not associated with obvious changes in the input resistance or membrane potential (Fig. S5). As a control, no persistent changes in the IPSP amplitude were observed when the TBS was absent (95.968.7%, n = 4; p.0.05; Fig. S6). The resting potentials of the MCs detected in this study (approximately 255 mV) were in the normal range of resting potential (265 to 247 mV) reported by previous studies [29,30], indicating that the recorded MCs were healthy. These results revealed that using an identical protocol, which induced plasticity with opposing directions in GCs when delivered to two excitatory GC inputs, could also elicit distinct long-term plasticity of inhibitory inputs onto the MCs.

STDP of ON inputs
In order to determine whether and how the plasticity of inhibitory inputs mediating lateral inhibition affects STDP of the ON inputs from sensory neurons, we need to further induce STDP in M/T cells. As a Hebbian synaptic learning rule, STDP has been previously demonstrated in various neural circuits, including the olfactory system [48]. Induction protocols for STDP commonly consist of the repetitive pairing of single pre-and postsynaptic spikes at regular intervals. However, neuronal activity in vivo is far more complex [31], with a spectrum of activity level from almost no activity to short bouts of high-frequency spike bursts. Therefore, the pairing of the presynaptic spikes with the postsynaptic bursts sometimes represents a more natural situation in vivo. The lateral inhibition from the GCs may prevent or delay firing or may decrease the firing rate during stimulus presentation in MCs [14,15]. Because STDP is a form of synaptic plasticity whose induction is heavily dependent on the timing of the incoming spikes and the frequency of the postsynaptic spikes [16], the presence of lateral inhibition in the postsynaptic MCs may profoundly influence the STDP induction and the total olfactory bulb output. We initially found that the repetitive pairing of the EPSPs with single spikes at 0.1 Hz failed to induce any persistent changes in the EPSPs of the ON inputs in MCs (100.467.0%, n = 7; p.0.05, Fig. S7A; 108.566.8%, n = 7; p.0.05, Fig. S7). In contrast, the repetitive pairing of the EPSPs with postsynaptic bursts consisting of three postsynaptic APs within a critical time window induced persistent modifications in the EPSPs in MCs (at ON inputsRMCs synapses; Fig. 3). The EPSPs were evoked by stimulating the glomeruli associated with the recorded MCs [32,33] and could be blocked by co-application of NMDA-and AMPA-type glutamate receptors antagonists such as AP5 (50 mM) and NBQX (20 mM), indicating that they were mediated by glutamatergic neurotransmission (Fig. S4). The postsynaptic APs were elicited by a current injection. The basal level of ISI of the postsynaptic bursts was set to 19-21 ms (20.060.3 ms when Dt = +30 ms and 20.160.3 ms when Dt = 250 ms; see Fig. 3E) by adjusting the intensity of the injected current. The Dt was defined as the time between the onset of the EPSP and the onset of the first AP in the burst [16,34]. The pairings were repeated 60 times at 0.1 Hz. We found that the EPSPs potentiated in 6 out of the 8 cells when the Dt was set to +30 ms (143.762.5%, n = 6; p,0.001; Fig. 3B, C), whereas it was depressed when the Dt was set at 250 ms (74.361.3%, n = 6; p,0.001; Fig. 3F, G). These two forms of synaptic plasticity were not associated with obvious changes in the input resistance or membrane potential (Fig. S8). The entire time window for the induction of potentiation and depression of excitatory ON inputs was then determined (Fig. 3D). These results indicate that STDP of excitatory OSN inputs to MCs could be induced by pairing EPSPs with postsynaptic spike bursts within a critical time window.
It has been reported that stimulation of ON induced longlasting depolarizations (LLDs) that lasted as long as several seconds [30,35,36,37]. The LLDs may only occur synchronously in cells whose apical dendrites ramify within the same glomerulus, suggesting that the LLDs involved intraglomerular interactions among the M/T cells. Therefore, glomeruli stimulation in this study might also induce LLDs in the correlated MCs. Indeed, occasionally we observed LLD-like responses in some of our results (Fig. S9). These responses were initiated by a fast and graded monosynaptic input from the OSNs followed by a slower component. To ensure that the EPSPs we detected were mediated by monosynaptic neurotransmission, we discarded those EPSPs with multiple peaks and only used those that displayed single-peak responses. In addition, because ON stimulation with high intensities more frequently produces LLDs, we deliberately controlled the stimulation intensity within moderate level in most of our recordings. Thus, the responses we report here represent monosynaptic inputs to the MCs.

Regulation of plasticity in the ON inputs
After confirming the feasibility of the STDP induction, we further investigated whether and how the prior plasticity of lateral inhibition affects the subsequent STDP of ON inputs from the OSNs. We began to collect whole-cell baseline EPSP data in MCs approximately 10 min after the prior stimulation was finished (Fig. 4A). The baseline EPSP recording in MCs lasted for 10 min and was followed by a STDP induction. Thus, the interval between the end of the prior TBS stimulation and subsequent STDP induction was 20 min (Fig. 4B, C). The prior TBS protocol, which was used to induce LTP or LTD of lateral inhibition when delivered to the proximal or distal inputs to GCs, respectively, did not affect the baseline EPSPs of ON inputs in MCs (OSNRMCs synapses; 106.863.2%, n = 6; p.0.05, Fig. S10B; 96.761.9%, n = 6, p.0.05; Fig. S10D). However, it modified the predisposition for the subsequent spike timing-dependent LTP or LTD induction in MCs. The protocol used in EPL to induce the LTD of lateral inhibition facilitated the subsequent LTP of EPSPs on ON inputs by elevating the magnitude of potentiation (186.062.4%, n = 6; compared with control, p,0.001; Independent-samples t test; Fig. 4B), whereas it suppressed LTD by reducing the magnitude of depression (99.861.2%, n = 6; compared with control, p,0.001; Independent-samples t test; Fig. 4C). In contrast, when the same stimulating protocol was delivered to the proximal inputs onto the GCs to induce LTP of lateral inhibition (Fig. 5A), it suppressed the subsequent LTP of EPSPs on ON inputs by reducing the magnitude of potentiation in MCs (109.162. 2%, n = 6; compared with control p,0.001; Independent-samples t test; Fig. 5B), whereas it facilitated LTD by elevating the magnitude of depression (50.161.5%, n = 6; compared with control, p,0.001; Independent-samples t test; Fig. 5C). These results reveal the bidirectional regulation of STDP by prior TBS of dictinct inputs onto GCs and suggest the changes in STDP may be caused by plasticity originated in GCs.
If the changes in STDP are truly caused by plasticity originated in GCs, then blockade of this plasticity should also suppress the changes in STDP. To test this possibility, we treated the slices with antagonists of AMPA-and NMDA-type glutamate receptor only during TBS to block the glutmatergic neurotransmission. This treatment totally abolished the induction of plasticity in GCs. (EPL 94.062.1%, n = 5; compared with baseline p.0.05; Fig. S11A; GCL 91.663.5%, n = 4; compared with baseline p.0.05; Fig.  S11C). As a result, no subsequent changes in STDP were observed (EPL 132.968.0%, n = 6; compared with control p.0.05; Fig.  S11B; GCL 140.265.3%, n = 7; compared with control p.0.05; ANOVA LSD test; Fig. S11D). Taken together, these data strongly suggest that the prior plasticity history of lateral inhibition driven by the two distinct excitatory inputs to GCs exerted differential regulation on the subsequent plasticity of excitatory inputs from the OSNs.

Changes in the ISI is the causative factor for regulation of plasticity
What is the mechanism that underlies the regulation of plasticity in ON inputs? It has been reported that recent activity history could affect the lateral inhibition between MCs and the lateral inhibition may alter MC spike-timing [15]. In addition, the frequency of postsynaptic spikes within the pairing protocol can affect the induction and magnitude of STDP [16], most likely due to a change in the back propagation of spike trains [38,39]. This raises the possibility that the regulation of STDP exerted by prior plasticity in lateral inhibition was achieved by regulating the ISI or frequency of APs in the burst. Therefore, we further examined in MCs whether and how the prior stimulating protocol affected the subsequent spike patterns within the induction protocol for STDP. Here, the TBS on the EPL or GCL was defined as the priming stimulation and the recordings were performed on the MCs. Interestingly, we found that prior EPL stimulation of distal inputs onto the GCs significantly shortened the ISI of the three-spikeburst at the time-point when the subsequent pairing protocol was delivered (ISI 18.060.3 ms, n = 8; compared with baseline 20.560.2 ms, p,0.001; paired-samples t test; Fig. 6B, E). In contrast, an identical prior stimulating protocol delivered to GCL prolonged the ISI (22.660.5 ms, n = 7, compared with baseline, p,0.001; Fig. 6B, E). As a control, the ISI of the spike burst was kept unchanged when no prior stimulation was delivered (20.960.5 ms n = 6, compared with baseline 20.860.5 ms, p.0.05; Fig. 6B, D, E). These results suggested that the regulation of the subsequent STDP may be correlated with coincident changes in the ISI within the STDP induction protocol.  Although these results suggest that the changes in the ISI and the subsequent regulation of LTP/LTD may be correlated, we still do not know whether the changes in the ISI are the causative factor for the regulation of STDP. It has been reported that partially up-or down-regulating GABA A receptor function with its agonist or antagonist can bidirectionally regulate the ISI within a spike burst [40,41,42]. Therefore, to examine this possibility, we used a selective GABA A receptor agonist or antagonist to determine how it affected the pattern of spikes within the induction protocol. Gabazine (GBZ) and isoguvacine (ISO) are highly selective GABA A receptor antagonists and agonists, respectively. We first investigated the dose-response relationship of these two agents and determined the concentration that matched the ISI regulation exerted by the previous stimulating protocol. We found that a partial blockade of GABA A receptor function with GBZ at 1.5 mM displayed similar effects on the ISI exerted by a prior stimulation at the EPL (Fig. 7A), whereas a submaximal activation of GABA A receptor function with ISO at 3.0 mM produced comparable effects on the ISI exerted by a prior stimulation at the GCL (Fig. 7E). This partial manipulation of GABA A receptor function could mimic, to a certain extent, the prior potentiation or depression of lateral inhibition the MCs received from the TBS. It was reported that the GABA A receptor blockade increased the excitatory MC responses to odors [21], suppressed the adaptation of MC firing rate [15] and disrupted the animal's ability to distinguish between similar odors [43]. However, the concentration of drugs used in this study were largely decreased compared with previous studies [40,44], and we determined that GBZ at this level failed to clearly or persistently affect the cell excitability, manifested by the absence of changes in the membrane potential or synaptic responses ( Fig. S12; n = 4, ANOVA LSD test, p.0.05). Thus, we largely avoided these unwanted side effects. Then we examined how these treatments could affect the induction of the STDP. GBZ (1.5 mM) facilitated spike timing-dependent LTP (tLTP) by increasing the magnitude of potentiation at OSNRMCs synapses (172.364.7%, n = 6, p,0.001, Independent-samples t test; Fig. 7B), but suppressed spike timing-dependent LTD (tLTD) by decreasing the magnitude of depression (93.661.5%, n = 6, p,0.001; Fig. 7C). In contrast, the ISO at 3.0 mM suppressed tLTP induction by decreasing the magnitude of potentiation (92.961.5%, n = 6, p,0.001; Fig. 7F), but facilitated tLTD by increasing the magnitude of depression (58.861.2%, n = 6, p,0.001; Fig. 7G). These results demonstrate that mimicking ISI modification with a GABA A receptor agonist or antagonist could cause a similar modification of the predisposition for subsequent STDP induction. To further ensure that the change in the ISI was the causative factor for plasticity modification, we injected three shorter current steps to the recorded MCs so that each current step was sufficiently strong to trigger a single action potential. When we set the ISI to 17.4 ms and 22.9 ms to mimic the bidirectional changes in the ISI following the prior stimulations, we were able to observe similar changes in the STDP (192.463.2%, n = 6, compared with control: 165.5615.4% P,0.001, Fig. S13B; 97.762.0%, n = 6, compared with control: 78.867.4% P,0.001, Fig. S13C; 95.862.4%, n = 6, compared with control: 165.5615.4% P,0.001, Fig. S13E; 61.361.0%, n = 6, compared with control: 78.867.4% P,0.001, Fig. S13F). These results suggest that the changes in ISI may contribute to the regulation of STDP.
The regulation of the ISI that was present continually after the TBS did not require acute GC stimulation, suggesting that TBS causes a tonic change in inhibition. To further examine whether the regulation of the ISI depend on a tonic change in inhibition that was continuously present after TBS, we also monitored miniature IPSPs (mIPSPs) before and after TBS and found that the TBS at EPL suppressed the mIPSPs amplitude in MCs (control, 1.6860.15 mV, after EPL TBS, 1.3860.09 mV; p,0.01; Fig. S14). These results provided further evidence that the TBS induces tonic and continuous regulation of the ISI and suggest that the changes in the ISI by prior plasticity of lateral inhibition may be the causative factor for the regulation of STDP. Moreover, it also suggests that modification in the GC-to-MC synapse may take a role in changes of inhibition.
If the changes in ISI really contribute to the regulation of STDP, then reversing the changes in ISI by resetting ISI back to control level should also abolish changes in STDP. To this purpose, we either employed pharmacological intervention of ISI with GABA A agonist/antagonist or directly reset the ISI with the three individual APs to get the ISI back to control levels (around 20 ms) after TBS. Following the EPL priming stimulation, which induced LTD of lateral inhibition, we applied 3.0 mM of ISO immediately before and during the pairing stimulation protocol for STDP induction. This treatment elevated the ISI value in MCs, which was already decreased by the prior EPL stimulation, to nopriming control levels (19.660.2 ms, n = 6, compared with control, 20.060.3 ms, p.0.05; Fig. 8E). As a result, the induction with a pairing protocol failed to elicit any persistent changes in the magnitude of spike timing-dependent LTP or LTD at OSNRMCs synapses, similar to controls without prior stimulation (LTP 152.961.9%, n = 6, p.0.05, ANOVA LSD test; LTD 77.961.4%, n = 6, p.0.05; Fig. 8A, B). Similarly, treating with 1.5 mM GBZ following GCL priming stimulation decreased the ISI value, which was already increased by prior GCL stimulation, to no-priming control levels (20.860.4 ms, n = 6, compared with control 20.460.5 ms, p.0.05; Fig. 8E). As a result, the induction with the pairing protocol failed to elicit any persistent changes in the magnitude of LTP or LTD (LTP 141.762.9%, n = 6, p.0.05, ANOVA LSD test; LTD 73.261.4%, n = 6, p.0.05; Fig. 8C, D). Moreover, adjusting the current injection level after prior TBS to convert the ISI back to control levels also reversed the STDP to control levels (147.561.9%, compared with control, p.0.05, Fig.  S15A; 145.261.1%, compared with control, p.0.05, Fig. S15B). Taken together, these results further confirm that the change in the ISI is the causative factor for regulation of plasticity.

Discussion
In the olfactory system, few studies have investigated the regulation of plasticity in olfactory sensory input by prior activity from GCs. This study demonstrates that using an identical TBS stimulation delivered to the distal or proximal inputs could induce persistent modification in the inhibition to MCs with opposing directions. This bidirectional modification of the inhibitory inputs differentially regulated the subsequent synaptic plasticity of excitatory ON inputs to the MCs. Because the modulation of the inhibition onto the MCs could alter the firing rate of spikes [14,15], the observed changes in the ISI caused by the plasticity in the excitatory inputs of GCs could be ascribed to several scenarios. First, the differential modification of drives in GCs to MCs is the simplest explanation for the TBS-mediated persistent changes of mitral cell inhibition. Second, the plasticity in the GC-to-MC synapse may also be involved. Third, the regulation of GC spiking by plasticity of excitatory inputs may relieve the Mg 2+ blockade of NMDA receptors at the dendrodendritic synapses and, thus, dynamically regulate the inhibition of the MCs [13].
The neurons that fire high-frequency bursts of spikes were found in various sensory systems including the olfactory system [14,45]. The firing of the bursts in response to sensory input depends on the intrinsic cellular mechanisms that function with feedback from higher centers to control the discharge properties of these neurons. A growing number of studies indicate the possibility that the bursts possess a distinct function in the transmission of sensory information. It has been observed that the intrinsic membrane properties favored MCs firing at 40 Hz and the MC discharge was also stabilized at a preferred frequency of 40 Hz [46]. In addition, additional bursts of action potentials may be triggered by the synaptic inputs locked to the air inhalations. Moreover, it has been proposed that neuronal oscillations enhanced stimulus discrimination by ensuring AP precision, and the maintenance of AP precision could be compromised at oscillation frequencies higher than 50 Hz [47]. These findings imply that the natural spike frequency in MCs in vivo that permits high levels of optimal stimulus discrimination is approximately 50 Hz. In this study, the STDP in MCs was induced by the pairing of presynaptic EPSPs with realistic 50 Hz postsynaptic bursts within a critical window of tens of milliseconds. The changes in synaptic strength depend on the inter-burst interval rather than the precise timing of the individual spikes [16]. These burst timingdependent plasticity rules may be specifically beneficial for the circuits in which the information relevant for synaptic refinement is contained in the timing of the bursts rather than that of the individual spikes. Given the very precise dependence of the magnitude of STDP on the ISI, a slight change in the ISI in MCs may produce significant implications in the output of the olfactory bulb and, thus, may have a particular physiological relevance in the olfactory system.
In the developing olfactory bulb, the inhibitory synapses distributed along the secondary dendrites of the MCs can dynamically regulate the extent of spike propagation, with a smaller activation of the inhibitory synapses facilitating the spike propagation [48]. The lateral and recurrent inhibitions in the olfactory bulb play distinct roles in shaping the MC spiking pattern, which is critical to odor information processing [49]. Both the two forms of inhibition are thought to be important in odor discrimination and in the generation and synchronization of odorevoked rhythmic MC activity [26,50,51]. Although the inhibition we observed in this study is largely driven by the excitatory inputs on to the GCs and thus may largely represent the lateral inhibition, we could not exclude the potential contribution of recurrent inhibition [15].
The two types of anatomically distinct excitatory inputs on to the proximal and distal dendrites of GCs may exert different functions through distinct mechanisms. The excitatory input onto the distal dendrites in the EPL mediates local dendrodendritic inhibition [25], whereas the excitatory inputs onto the proximal dendrites at least partially, if not completely, are cortical feedback inputs that mediate global top-down modification of the DDI. Using a two-photon guided minimal stimulation in acute rat brain slices, Strowbridge and colleagues positioned an extracellular stimulating electrode near a specific dendritic segment and thus were able to activate relatively homogenous populations of presynaptic processes, based on the kinetic properties of the resulting postsynaptic currents [13,24]. Although we did not use a similar technique, we were able to deliberately control the intensity of the local stimulation to ensure the independence of the stimulation on distinct sites in this study. Moreover, the synaptic responses we obtained from these two inputs also displayed distinct properties in the current kinetics, similar to the findings obtained from Strowbridge's group [13]. Therefore, we could still activate a relatively homogenous population of presynaptic processes when stimulating either of the two sites. Because the extent of GC function in the local versus global output modes could exert a critical effect on the computational role it performs in the olfactory bulb circuit [1], our present findings of the modification of excitatory ON inputs by prior plasticity of the distinct excitatory inputs to GCs reveal an efficient route to regulate the encoding and processing of odor information in the bulb.