Abstract
Principal cells in the olfactory bulb (OB), mitral and tufted cells, receive direct sensory input and generate output signals that are transmitted to downstream cortical targets. Excitatory input from glutamatergic receptor neurons are the primary known sources of rapid excitation to OB principal cells. Principal cells also receive inhibitory input from local GABAergic interneurons in both the glomerular and plexiform layers. Previous work suggests that the functional effect of these inhibitory inputs, including numerous dendrodendritic synapses with GABAergic granule cells, is to reduce firing probability. In this study, we use in vitro patch-clamp recordings to demonstrate that rat (of both sexes) OB mitral cells also can be excited by GABAergic synapses formed outside the glomerular layer. Depolarizing GABAergic responses to focal extracellular stimulation were revealed when fast ionotropic glutamate receptors were blocked, and occurred with short, monosynaptic latencies. These novel synaptic responses were abolished by gabazine, bicuculline, and picrotoxin, three structurally dissimilar GABAA receptor antagonists. The likely location of depolarizing GABAergic input to mitral cells was the proximal axon based on the actions of focally applied gabazine and GABA near this region. Excitatory GABAergic synaptic responses, commonly studied in cortical brain regions, have not been reported previously in OB principal cells. Excitatory GABAergic responses promote action potential firing and provide a mechanism for mitral cells to be excited independently of olfactory sensory input.
SIGNIFICANCE STATEMENT Odor stimuli generate distinctive activity patterns in olfactory bulb neurons through a combination of excitatory and inhibitory synaptic interactions. Most of the excitatory drive to each principal cell is assumed to arise from a highly restricted subset of sensory neurons. This study describes a novel second source of synaptic excitation to principal cells to arise from GABAergic inputs to the proximal axon, a common site of action potential initiation. This new pathway provides a synaptic mechanism to excite OB principal cells that is independent of the canonical excitatory sensory input contained in the glomerular layer.
Introduction
Our sense of smell is processed in a series of brain areas beginning in the olfactory bulb (OB). The interaction between excitatory and inhibitory synaptic potentials allows principal cells in this region to generate distinct spiking patterns in response to different odors (Cang and Isaacson, 2003; Wilson and Laurent, 2005; Fukunaga et al., 2014; Najac et al., 2015). How synaptic integration in mitral cells (MCs) and tufted cells enables olfactory function is a complex and unsolved problem. More is known about the sources of synaptic input to mitral and tufted cells. Excitatory inputs arise primarily from a glutamatergic synaptic drive from a single glomerulus (Ressler et al., 1994; Schoppa and Westbrook, 2001; Murphy et al., 2005) generating excitatory responses that occur only when a sensory stimulus activates that particular olfactory sensory neuron (OSN) subtype.
Mitral and tufted cells receive inhibition primarily from the following three sets of interneurons: axonless GABAergic granule cells (GCs); parvalbumin-positive interneurons in the external plexiform layer (EPL; Kato et al., 2013; Miyamichi et al., 2013) and a diverse set of interneurons located in the glomerular layer that function to regulate orthodromic input strength. The OB contains a large population of GCs that form unique reciprocal dendrodendritic microcircuits with both mitral and tufted cells along their secondary dendrites and somata (Rall et al., 1966; Isaacson and Strowbridge, 1998). Most reports on the synaptic control of principal cells focus on these three sets of inhibitory inputs functioning together with OSN-mediated glomerular excitation.
Despite the large number of bulbar GABAergic interneurons (Price and Powell, 1970; Nagayama et al., 2014), activating large-amplitude IPSPs in MCs has proven surprisingly challenging using electrical stimulation in the GC layer (GCL). In the neocortex and hippocampus, antagonists of inotropic glutamate receptors (iGluRs) proved invaluable for studying inhibitory synaptic function since they block the excitatory component of evoked responses. In the presence of AMPAR and NMDAR blockers, focal electrical stimulation in cortical brain regions evokes isolated GABAergic responses whose amplitude is graded with stimulus intensity (Davies et al., 1990). In the OB, stimulating within the GCL, the layer containing most of the GABAergic interneurons in the OB, often evokes small-amplitude IPSPs (Balu and Strowbridge, 2007; Gao and Strowbridge, 2009), but only rarely large IPSPs (B. Strowbridge, R. Pressler, unpublished observations). Occasionally, GCL stimuli evoke complex biphasic responses even after ionotropic glutamate receptors are blocked (Gao and Strowbridge, 2009, their Fig. 2c,d). The identity of the different synaptic inputs to MCs activated by these GCL stimuli are not known.
The present study was sparked by the unexpected finding that antagonists of GABAA receptors (GABAARs) abolished the anomalous depolarizing response recorded in MCs following GCL stimulation. Our study demonstrates that OB principal cells are excited through GABAA receptors, likely located within the proximal axon, consistent with earlier ultrastructural reports of symmetric (presumed GABAergic) synapses on the MC axon hillock and initial segment (Price and Powell, 1970; Willey, 1973). The hypothesized high intracellular chloride concentration in the axon compartment reverses the normal chloride electrochemical gradient and enables GABAA receptors to depolarize MCs and trigger action potentials (APs). Unlike the spatially restricted region of the glomerular layer where electrical stimulation drives synaptic excitation, stimulation over a wide expanse of the GCL evokes depolarizing GABA responses in OB principal neurons. These results suggest that MC spiking can be triggered by multiple excitatory synaptic inputs through both glomerular and extraglomerular pathways.
Materials and Methods
Animals.
Horizontal OB slices (300 μm thick) were prepared from postnatal day 14 (P14) to P25 or P60 Sprague Dawley rats of either sex anesthetized with ketamine, as described previously (Balu et al., 2007; Pressler and Strowbridge, 2017, 2019). All experiments were conducted in accordance with the guidelines approved by the Case Western Reserve University Animal Care and Use Committee.
Electrophysiology.
Olfactory bulb slices were incubated initially at 30°C for 30 min and then maintained at room temperature until use. During recording sessions, slices were placed in a submersion chamber and perfused with oxygenated artificial CSF (ACSF) at a rate of 1.5 ml/min. Recordings were made between 29 and 31°C. ACSF consisted of the following (in mm): 124 NaCl, 3 KCl, 1.23 NaH2PO4, 1.2 MgSO4, 26 NaHCO3, 10 dextrose, and 2.5 CaCl2, equilibrated with 95% O2/5% CO2. Drugs were added to the submerged recording chamber by changing the external solution source (except for the focal application experiments seen in Figs. 9, 10) Whole-cell patch-clamp recordings were made using AxoPatch 1D amplifiers (Molecular Devices) and borosilicate glass pipettes (3–7 MΩ) under infrared-differential interference contrast video microscopy. Under current-clamp conditions, recording electrodes contained the following (in mm): K-methylsulfate 140, NaCl 4, HEPES 10, EGTA 0.2, MgATP 4, Na3GTP 0.3, and phosphocreatine di-tris 10. We substituted 30 mm KCl for an equimolar amount of K-methylsulfate in the high-Cl– internal solution seen in Figure 8 (containing 39 mm Cl– vs 9 mm in control internal solution). The low-Cl– internal solution (containing 3 mm Cl–) used in that experiment was made by reducing NaCl from 4 to 1 mm and reducing QX-314 chloride from 5 to 2 mm. Alexa Fluor 594 (100–150 μm; Thermo Fisher Scientific) was also included routinely to permit visualization of the recorded neuron using two-photon imaging, as described previously (Pressler and Strowbridge, 2017, 2019). All internal solutions were adjusted to pH 7.3 and 290 mOsm. Intracellular recordings were not corrected for the liquid junction potential.
All drugs were obtained from Tocris Bioscience except where noted. The following drugs were used in this study: DNQX (6,7-dinitroquinoxaline-2,3-dione; 20 μm), APV (aminophosphonovalerate) (50 μm), gabazine [SR 95531; 6-imino-3-(4-methoxyphenyl)−1(6H)-pyridazinebutanoic acid hydrobromide, 10 μm], bicuculline ([R-(R,S)]−6-(5,6,7,8-tetrahydro-6-methyl-1,3-dioxolo[4,5-g]isoquinolin-5-yl)furo[3,4-e]−1,3-benzodioxol-8(6H)-one, 20 μm), DHβE [(2S,13bS)−2-methoxy-2,3,5,6,8,9,10,13-octahydro-1H,12H-benzo[i]pyrano[3,4-g]indolizin-12-one hydrobromide), 2 μm], MLA ([1α,4(S),6β,14α,16β]−20-ethyl-1,6,14,16-tetramethoxy-4-[[[2-(3-methyl-2,5-dioxo-1-pyrrolidinyl)benzoyl]oxy]methyl]aconitane-7,8-diol citrate; 10 μm), atropine (5 μm), tetrodotoxin (TTX; 1 μm), picrotoxin (100 μm), bumetanide [3-(aminosulfonyl)−5-(butylamino)−4-phenoxybenzoic acid; 50 μm], and GABA (100 μm; Sigma-Aldrich).
The intracellular voltage-gated Na+ channel blocker QX-314 (5 mm) was added to the internal solution in experiments recording depolarizing GABA responses (except for the experiment shown in Fig. 1F) to minimize intrinsic voltage gated subthreshold Na+ currents. Ionotropic glutamate receptors also were blocked in experiments recording depolarizing GABA responses using 50 μm d-APV and 20 μm DNQX. In the focal application experiments seen in Figures 9 and 10, gabazine or GABA were dissolved in a pH-buffered solution containing 124 mm NaCl, 3 mm KCl, and 10 mm HEPES, pH 7.4. Single and array tungsten electrodes were manufactured by FHC. The light scattered from the electrode tip during live two-photon imaging was used to verify that one contact of the array stimulation electrode in the glomerulus associated with a recorded MC (see Fig. 4B).
Data analysis and statistics.
Intracellular recordings were low-pass filtered at 5 kHz (model FLA-01 Filter/Amplifier, Cygus Technology) and acquired at 10 kHz using an ITC-18 interface (InstruTech/HEKA) and custom Visual Basic.NET and Python software on a Windows 7-based personal computer. Live two-photon imaging used the Toronado Laser Scanning software system written by B.W.S. (https://github.com/StrowbridgeLab/Toronado-Laser-Scanning). The amplitude of GCL-evoked depolarizing GABA responses described in this report were quantified by calculating the mean membrane potential (Vm) between 45 and 55 ms after the stimulus for single GCL shocks, relative to the prestimulus baseline Vm. For tetanic GCL stimuli, we calculated the mean membrane potential between 5 and 15 ms following the final stimuli in the train, relative to the prestimulus baseline. Summary results are presented as the mean ± SEM t tests, which were used when comparing differences of means between normally distributed datasets; nonparametric tests (Mann–Whitney U rank test or the Wilcoxon signed-rank test) were used for the comparisons between datasets that were not normally distributed by visual inspection. The test used to compare mean values is indicated in the figure legends. Statistical tests were performed in Python using the SciPy library (version 1.4.1).
Results
The focus of this study is identifying the mechanism and location of a novel excitatory synaptic drive to OB MCs evoked by GCL stimulation. Our experimental results are divided into three parts beginning with work defining the synaptic and functional properties of the depolarizing GABA responses in MCs. We then present evidence that the novel depolarizing postsynaptic response arises because of activity of the NKCC1 transporter. Finally, we use two photon-guided focal application of GABA and GABAA receptor antagonists to localize the depolarizing synaptic response to the MC axon.
Depolarizing GABAergic responses in OB mitral cells
Under control conditions, extracellular stimulation in the GCL can evoke a variety of synaptic responses in MCs held near rest, including hyperpolarizing IPSPs (Balu and Strowbridge, 2007) and fast-rising depolarizing responses (Fig. 1A,B). GABAergic responses likely reflect the activation of either dendrodendritic synapses formed between MCs and GCs (Isaacson and Strowbridge, 1998) or other interneurons with processes in the EPL, such as parvalbumin-positive GABAergic cells (Kato et al., 2013; Miyamichi et al., 2013). The local circuit mechanisms responsible for the depolarizing MC responses recruited by GCL stimuli in some experiments are less clear and will be the focus of this report.
Depolarizing GABAergic responses in olfactory mitral cells. A, Diagram of experimental configuration. B, Example responses to GCL stimulation (40 μA) recorded under control conditions (black traces), following the blockade of iGluRs with 20 μm DNQX and 50 μm d-APV (blue traces). Adding 10 μm gabazine to block GABAARs (orange traces) abolished the DNQX + APV-resistant response. C, Summary of the reduction in the GCL-evoked response recorded in iGluR blockers following gabazine application. Mean response amplitudes and N values are indicated inside the bars (and in subsequent plots). Error bars indicate the mean ± SEM p = 0.0017, t = 6.217, df = 4, paired t test. D, Mitral cell response to graded intensity GCL stimulation in DNQX + APV (average of 3–5 trials). Dashed lines indicate baseline Vm. E, Summary plot of amplitudes of individual responses to different intensity GCL stimulation in DNQX + APV. F, Demonstration that GABAergic PSPs evoked by GCL are excitatory. Top traces, Enlargement of individual responses evoked by GCL stimulation in DNQX + APV (blue), which are abolished by gabazine (red). Middle traces, The same GCL stimulation triggers an AP when paired with a brief, just-subthreshold step. The depolarizing step presented alone failed to trigger an AP (tested at the same membrane potential). Bottom traces, Gabazine blocks AP generation by combined GCL stimulation and just-subthreshold depolarizing steps in the same MC. G, Top traces, GABAergic PSPs evoked by increasing duration tetanic GCL stimulation (blue). Blockade of GABAARs with bicuculline (BIC; 20 μm; orange traces) abolishes response to five-shock GCL stimulus. Bottom traces, Superimposed responses from increasing durations of GCL stimulus traces. Vertical lines indicate the magnitude of the response to a single GCL stimulus and the response enhancement recruited by additional GCL stimuli (e.g., Δ1–2 represents the magnitude of the single shock response subtracted from the two-shock response). All responses recorded in DNQX + APV.
Mitral cells express glutamate autoreceptors (Friedman and Strowbridge, 2000; Salin et al., 2001; Lowe, 2003), providing a potential mechanism for stimulus-evoked depolarizations. However, blockade of ionotropic iGluRs with high concentrations of DNQX and d-APV attenuated but did not eliminate the depolarizing response recorded at −60 mV. In five of five experiments similar to those seen in Figure 1B, bath application of the GABAAR antagonist gabazine (10 μm) abolished the remaining depolarizing response in DNQX and APV (Fig. 1B, orange traces); results are summarized in Figure 1C. The response time course remained similar over a threefold variation in amplitude (Fig. 1D) and increased monotonically with stimulus intensity (Fig. 1E), suggesting that these depolarizing responses represent a single type of GABAergic synaptic input.
We next asked whether the depolarizing GABAergic response functioned as an EPSP—a synaptic response that increases the probability of firing. In this experiment, we first determined the just-subthreshold current stimulus intensity for each MC. We then determined whether an AP was triggered when we combined just-subthreshold depolarizing steps with GCL stimulation (Fig. 1F). In three experiments, we alternated testing the same depolarizing current pulse with and without GCL stimulation and found that synaptic stimulation (in DNQX and APV) triggered reliable firing while responses to the depolarizing step alone remained subthreshold, confirming the excitatory effect of the depolarizing GABA input. Bath application of gabazine blocked the ability of GCL stimulation to convert just-subthreshold current steps into suprathreshold responses in all three cells (Fig. 1F, top traces).
The depolarizing GABA postsynaptic potential (PSP) depressed on repeated stimulation. Figure 1G shows overlapped responses to one to five stimuli at 50 Hz with the response increment for each stimulus indicated by vertical lines. Paired stimuli triggered ∼1.5× larger responses than single shocks, indicating that the response to the second shock was smaller than to the first stimulus. Similarly, short trains of three GCL stimuli triggered only slightly larger responses than paired stimuli. The diminishing effectiveness of longer-length GCL stimulus trains suggests that the underlying synaptic response was strongly depressed with repeated stimuli. We observed similar response patterns indicative of short-term depression in all 11 MCs tested systemically.
In many cortical regions, depolarizing GABA synaptic responses are especially prominent in juvenile animals and absent, or very small amplitude, in adults (Mueller et al., 1984; Ben-Ari et al., 1989; Luhmann and Prince, 1991; Yamada et al., 2004). Unlike this pattern, we find clear examples of depolarizing GABA responses in adult rat (P60; Fig. 2A) MCs tested under the same conditions used in our adolescent rat experiments above. The mean amplitude of GCL-evoked depolarizing GABA responses was 0.98 ± 0.20 mV (N = 4 P60 MCs). In all four experiments with P60 MCs, gabazine eliminated most of the GCL-evoked response, confirming that it was mediated by GABAARs (Fig. 2B).
Depolarizing GABA responses evoked in adult rats. A, Responses to GCL stimulus trains recorded in a P60 rat MC. GCL stimulation evoked depolarizing GABA responses that were evident after ionotropic glutamate receptors were blocked with DNQX and APV. Depolarizing GABA responses were blocked by gabazine, paralleling results obtained from adolescent rats. B, Summary plot of reduction in depolarizing GABA response amplitude by gabazine in four P60 MCs. p = 0.034, T = 10.0, Wilcoxon signed-rank test.
We next performed a series of experiments to test whether the depolarizing response reflected synaptic, calcium-dependent GABA release. We first asked whether release events appeared to be all-or-none at low stimulus intensities, indicative of synaptic neurotransmission. Distinct failures and successes were evident in three experiments with low-voltage noise, including the example responses shown in Figure 3A. These recordings demonstrate both failures and successes with unitary amplitudes of ∼0.5 mV at a perithreshold stimulus intensity (20 μA). Increasing the stimulus intensity to 30 μA eliminated failures while decreasing the intensity to 10 μA generated all failures. The kinetics of the successes evoked at perithreshold intensity were similar to larger-amplitude responses evoked at higher stimulus intensities.
Effect of synaptic blockers on depolarizing response evoked by GCL stimulation. A, Example recordings illustrating GCL-evoked responses at perithreshold stimulus intensity (20 μA in this experiment). Over four trials, both failures (blue traces) and successes (orange) were recorded. Two sets of responses to 20 μA stimulation are shown separately to illustrate differences between failures and successes. All responses were recorded following iGluR blockade using DNQX and APV (throughout figure). B, Example traces illustrating the method used to assay response onset latency. The gabazine-sensitive response (bottom trace) was estimated by subtracting control and gabazine responses (top traces). Onset latency was determined visually by measuring the time from the stimulus artifact to the positive inflection in the response (indicated by two vertical dashed lines; 2.4 ms in this example). C, Summary plot of onset latency of the depolarizing GCL-evoked response using the trace subtraction method illustrated in B. D, Switching to ACSF containing reduced Ca2+ ions (0.5 mm) and elevated Mg2+ ions (8 mm) reversibly blocked GCL-evoked responses in MCs. E, Time plot of the experiment shown in D. Following washout into control ACSF (still containing DNQX + APV), the addition of bicuculline abolished the GCL-evoked response. F, GCL-evoked response is unaffected by the blockade of nicotinic receptors using DHβE (2 μm) and MLA (10 μm). GCL-evoked responses are abolished by bicuculline applied at the end of the recording. Example responses before (black trace) and after DHβE + MLA (orange) are shown in the inset. G, Summary of the effects of receptor and synaptic blockers on GCL-evoked responses. Plots reflect response amplitudes normalized to control values. Neither DHβE + MLA nor atropine (5 μm) reduced the GCL-evoked response (p > 0.05, one-sample t test). Both low-Ca/high-Mg ACSF and TTX (1 μm) significantly reduced GCL-evoked responses (low Ca: p = 7.36 × 10–5, t = 14.09, df = 4; TTX: p = 0.0013, t = 19.4, df = 2; both one-sample t tests) as did all three GABAA receptor antagonists (gabazine: p = 4.93 × 10–5, t = 15.6, df = 4; bicuculline: p = 1.78 × 10–8, t = 55.6, df = 5; picrotoxin: 100 μm; p = 4.63 × 10–4, t = 32.8, df = 2; all were one-sample t tests.
Depolarizing responses evoked by GCL stimulation also appeared to have a distinct onset latency that was apparent in gabazine-subtracted responses but often obscured by the stimulus artifact in raw records (Fig. 3B). On average, the onset of the depolarizing GABA response was 2.5 ± 0.1 ms (N = 12; Fig. 3C), which is consistent with synaptic release of GABA.
A synaptically evoked transmitter requires transmembrane Ca2+ entry, which can be competitively blocked by elevating extracellular Mg2+ ions with a concomitant reduction in extracellular Ca2+ ions. We found that switching to an external ACSF solution containing 8 mm Mg2+ and 0.5 mm Ca2+ ions reversibly blocked the depolarizing GABA response, as illustrated in Figure 3D and the time plot in Figure 3E. The application of bicuculline at the conclusion of the experiment abolished the response (Fig. 3E), demonstrating that it was mediated by GABAA receptors. We found a similar blockade of the depolarizing GABA response in five of five experiments using low-Ca2+/high-Mg2+ ACSF. Blockade of fast Na+ channels with tetrodotoxin (1 μm) also blocked depolarizing GABA responses in three of three experiments, demonstrating that the response requires spike-evoked transmitter release.
While our preliminary experiments thus far indicated that the GCL-evoked depolarizing response was mediated by GABAA receptors, we also explored the possibility that ACh release triggered the PSP. We tested antagonists of both muscarinic receptors (atropine, 5 μm) and nicotinic receptors (2 μm DHβE and 10 μm MLA applied together). Neither treatment blocked the depolarizing response to GCL stimulation. Figure 3F illustrates an example of nicotinic receptor antagonists failing to attenuate a depolarizing GCL-evoked response that was subsequently abolished by bicuculline. By contrast, GCL-evoked responses were reliably abolished by three structurally different GABAA receptor antagonists (gabazine, bicuculline, and picrotoxin). Results from this group of pharmacological experiments are summarized in Figure 3G and suggest that the GCL-evoked depolarizing response is mediated by synaptic release of GABA activating GABAA receptors.
Functional significance of depolarizing GABA response
Olfactory bulb principal cells receive strong glutamatergic excitation from one class of OSN inputs whose axons converge within only one or two glomeruli (Ressler et al., 1994; Vassar et al., 1994). We demonstrated the focal nature of glomerular excitation functionally by recording from an MC and used two-photon imaging to verify that the dendritic ramifications at the distal end of the apical dendrite were in the same glomerulus as the left-most contact of an array stimulation electrode (Fig. 4A,B). When stimulating the “on-beam” glomerulus (containing the dendritic tuft), we found a sharp threshold intensity that triggered a large EPSP response. The MC failed to respond following activation of the other stimulating electrode contacts at this same threshold intensity (Fig. 4C). Over five similar experiments, none of the off-beam electrode contacts triggered EPSP responses when tested at the threshold intensity for the on-beam glomerulus (mean response recorded in the nearest off-beam contact, 1.1 ± 1.1% of the on-beam response). These results demonstrate that MCs are excited via glutamatergic input primarily within a narrow region of the glomerular layer corresponding to the glomerulus where the apical dendrite resides.
Depolarizing GABA responses from proximal and distal stimulation sites. A, Diagram of experiment configuration using an array stimulation electrode positioned in the glomerular layer (4 sharpened monopolar tungsten electrodes with 115 μm tip spacing). B, Example two-photon z stack demonstrating the left-most array electrode tip positioned in the same glomerulus as the apical dendritic tuft of the recorded MC (“glom 0,” referred to as “on-beam” stimulation). C, Example responses evoked by each of the four array electrode contacts at the same stimulus intensity (0.875 V, the threshold stimulus intensity for the on-beam contact at glom 0). Just-subthreshold responses from the on-beam electrode presented below glom 0 successful trials. D, Diagram of experiment configuration for recording depolarizing GABA responses. E, Average responses recorded in the same MC following GCL stimuli at two sites within the GCL ∼200 μm apart. F, Comparison of the time course of the EPSP triggered by focal glomerular stimulation and the depolarizing GABA response triggered by GCL stimulation (in different MCs). Ionotropic glutamate receptors were blocked using DNQX and APV in all experiments recording depolarizing GABA responses.
Depolarizing GABA responses were elicited by stimulation over a much broader spatial range than glomerular glutamatergic EPSPs. Over 13 experiments, we assayed depolarizing responses to bipolar stimulation at a constant intensity either near (laterally displaced 50–100 μm from the soma in the GCL; N = 7) or far (300–400 μm displacement; N =6; Fig. 3D) from the recorded MC. In all experiments, the distant stimulation sites were effective in triggering depolarizing GABA responses that could be blocked by gabazine. On average, the depolarizing gabazine-sensitive response evoked by far stimulation sites was half the amplitude of responses triggered by near sites 200 μm closer to the recorded MC [1.52 ± 0.23 mV near GCL stimulation (N = 7) vs 0.77 ± 0.06 mV far GCL stimulation (N = 6); p = 0.0071, t = 2.908, df = 11, t test]. We directly compared near versus far stimulation sites in two MCs, a subset of the N values used in the comparison above. In both experiments, responses to the far stimulation site were >50% of the amplitude of responses triggered at nearby sites. These results demonstrate that the depolarizing GABA input to MCs does not share the similarly narrow spatial origin as glomerular sensory inputs.
We next asked whether depolarizing GABA responses contributed to sensory-evoked excitation of MCs. This question is difficult to answer by blocking ionotropic glutamate receptors as that treatment eliminates both the glomerular synaptic input and local circuit excitation necessary to activate bulbar GABAergic neurons. Instead, we exploited the overlapping rising-phase kinetics of glomerular EPSPs and depolarizing GABA responses (Fig. 4F) to ask whether blocking GABAA receptors would slow and attenuate sensory-evoked MC responses. Most of our experimental manipulations thus far focused on depolarizing GABA responses evoked by the stimulation of GABAergic axons in the GCL (consistent with the short, monosynaptic response latency; Fig. 3B,C). If the depolarizing PSP we identified arises from local bulbar GABAergic interneurons, then those interneurons may be activated by OSN stimulation as well, either monosynaptically or polysynaptically, following the activation of bulbar principal cells.
The experiment depicted in Figure 5A tests whether the activation of intrinsic OB GABAergic interneurons could evoke depolarizing GABAergic PSPs on MCs when ionotropic glutamate receptors were not blocked. As shown in the example MC recording in Figure 5B, OSN stimulation triggered a large, complex depolarizing response that increased in amplitude when GABAARs were blocked with gabazine. This overall increase in amplitude was expected as gabazine blocked most of the feedforward and feedback inhibition that normally follows sensory stimulation (Cang and Isaacson, 2003). However, gabazine had the opposite effect on the initial phase of the response, slowing the onset kinetics, as illustrated in the enlargement shown in Figure 5C. Over five similar experiments, the initial phase of the OSN stimulus response was consistently reduced by gabazine (Fig. 5C, “early” time point), while the later peak response always increased following gabazine in the same experiments (summarized in Fig. 5D). These results suggest that gabazine blocked an early component of the OSN response that helped activate MCs, likely the depolarizing GABA response.
GABAergic PSPs accelerate mitral cell responses to sensory inputs. A, Schematic diagram of experiment configuration. B, Example responses to OSN stimulation before (black trace) and after (orange) blockade of GABAARs with gabazine. C, Enlargement of the two responses shown within the dashed box in B. D, Summary plot of change in OSN-evoked response in gabazine at the two time points indicated by vertical arrows in C (early: mean response between 3 and 5 ms following the stimulus, p = 0.0019, t = 6.0, df = 4; late: between 80 and 90 ms, p = 0.008, t = 4.0, df = 4). Both one-sample t tests. Response amplitudes normalized to control values at early and late time points. E, Responses to OSN stimulation evoked during just-subthreshold depolarizing current steps before (Control, black trace) and after (orange trace) bath application of gabazine. In both cases, OSN stimulation triggered a discharge that was not present in interleaved responses to the current step stimulus by itself (not shown). F, Enlargement of the initial phase of OSN-evoked discharge in control and following gabazine treatment illustrating the increased first spike latency in gabazine. G, Summary plot of the results from six experiments similar to F (p = 0.003, t = 4.6, df = 5, paired t test). H, Enlargement of the initial phase of OSN stimulus responses aligned to the average first spike latency to illustrate increased spike time jitter in gabazine. I, Summary plot of the results of six experiments similar to H (p = 0.014, T = 21, Wilcoxon signed-rank test).
Next, we determined the functional effect that depolarizing GABA responses have on MC excitability by focusing on the early phase of OSN responses, when depolarizing GABA responses are maximal. We held MCs near their firing threshold using long-duration current pulses and then used a single OSN shock to trigger a spike discharge (Fig. 5E). Consistent with the effect on the time course shown in Figure 5C, blocking GABA responses with gabazine delayed AP latency (Fig. 5F,G) and increased AP onset latency jitter (Fig. 5H,I). Both effects were statistically significant and suggest that one component of the excitatory response in MCs from activating sensory neurons is mediated by the same depolarizing GABA response we recorded using GCL stimulation (Figs. 1–4). The depolarizing GABA response appears to shorten spike latency and reduce jitter in MC sensory responses.
Bath application of GABAAR antagonists will impact a wide variety of GABAergic bulbar local circuits. However, attenuating conventional dendritic inhibitory inputs would be expected to have an effect that was the opposite of what we observed (accelerating rather than delaying AP onset timing when hyperpolarizing IPSPs are blocked by gabazine). In principle, the blockade of the extensive local GABAergic circuitry within the glomerular layer (including numerous periglomerular interneurons; Aungst et al., 2003; Kiyokage et al., 2010; Zhou et al., 2020) also contributes to the slowing of the response in gabazine. However, the primary postsynaptic effect triggered by activating those interneurons appears to be inhibitory, not excitatory (Gire and Schoppa, 2009; Shao et al., 2019). Blocking this input would lead to response acceleration, not to the increased delay we observed.
Mechanism of depolarizing GABA response
The response evoked by GCL stimulation in ionotropic glutamate receptor blockers remained depolarizing when tested at a membrane potential ranging from −90 to −40 mV (Fig. 6A). (The depolarized limit in these current-clamp recordings was determined by the increase in membrane potential noise at voltages above −40 mV.) In 12 of 12 cells tested systematically, the PSP always increased in amplitude with membrane hyperpolarization, which is consistent with a GABAergic conductance with a reversal potential of approximately −10 mV (Fig. 6B, inset, mean estimated reversal potential from all experiments). In most experiments, including the one illustrated in Figure 6B, the plot of PSP amplitude versus membrane potential was linear except for responses triggered at −90 mV. We obtained a slightly more hyperpolarized estimate of the GABAergic PSP reversal potential when excluding data recorded at −90 mV [likely because of IH (hyperpolarization-activated cation current) activation; Fig. 6B, dashed line].
Estimated reversal potential of GABAergic response. A, Responses to GCL stimuli recorded in a mitral cell at different membrane potentials. Responses recorded in DNQX + APV (blue traces at left) and following the blockade of GABAAR with 20 μm bicuculine (BIC; orange traces at right). Dashed lines indicate prestimulus membrane potential baseline. B, Left, Summary plot of response amplitudes in the experiment in A. Vertical arrow indicates estimated reversal potential of the GABAergic PSP recorded in DNQX + APV; horizontal bar indicates the range of reversal potential estimates obtained with and without data points at −90 mV. Scatter plot at right summarizes the GABAergic PSP reversal potential estimated from 12 experiments.
While neurons express a wide range of ion transporters to maintain normal ionic gradients, depolarizing GABAergic responses are often associated with high expression of the Na–K–Cl cotransporter (NKCC1; Yamada et al., 2004; Bonalume et al., 2021; Kilb, 2021), a transport protein expressed in both juvenile and adult MCs (Kanaka et al., 2001; Wang et al., 2005). We tested whether bumetanide, a specific blocker of NKCC1 (Isenring and Forbush, 1997), attenuated the depolarizing GABA response in MCs held at −60 mV. As shown in Figure 7A–C, bumetanide reduces the amplitude of the depolarizing GABAergic response evoked by GCL stimulation. Attenuating NKCC1 with bumetanide decreased both the response to the initial GCL stimulus (p = 0.0084, t = 3.95, df = 4, paired t test) and the response to 50 Hz tetanic stimulation (Fig. 7B,C). The relatively rapid time course of attenuation, reaching a new steady-state response amplitude in <10 min, is consistent with previous work using bumetanide in brain slices to block NKCC1 (Dzhala et al., 2005). Bumetanide was effective in attenuating both large- and small-amplitude GABAergic responses (Fig. 7C), suggesting that a similar mechanism likely underlies these responses.
Blockade of NKCC1 cotransporter selectively attenuates depolarizing GABA responses. A, Reduction in depolarizing GABA response evoked by GCL stimulation and recorded at approximately −60 mV by bumetanide (50 μm). Both control and bumetanide responses recorded in DNQX and APV. B, Plot of reduction in the amplitude of the depolarizing GABA response recorded at approximately −60 mV in 5 MCs. C, Summary plot of pre-bumetanide treatment and post-bumetanide treatment depolarizing GABA response amplitudes. Each experiment is represented by a line connecting pretreatment and post-treatment response amplitudes (p = 0.022, T = 15, Wilcoxon signed-rank test). D, Plot of shift in estimated reversal potential of depolarizing GABA response amplitude following bumetanide treatment in one MC. Example responses recorded at −49 and −71 mV shown above the plot. E, Example epochs containing spontaneous hyperpolarizing GABA responses before and after bumetanide treatment. Responses recorded at three different membrane potentials. F, Example plot of hyperpolarizing spontaneous IPSP reversal potentials before and after bumetanide treatment. G, Summary plot of the hyperpolarizing shift in the estimate reversal potential of the GCL stimulus-evoked response with bumetanide in 5 MCs (p = 0.022, T = 15, Wilcoxon signed-rank test). H, Plot of estimated reversal potential calculated from spontaneous IPSPs in the same 5 MCs before and after treatment with bumetanide (−60.3 vs −61.8 mV; p > 0.05, Wilcoxon signed-rank test).
Attenuating NKCC1 reduced but never blocked the depolarizing GABA response; varying the holding potential after bumetanide treatment revealed both depolarizing and hyperpolarizing GABAAR-mediated responses (Fig. 7D). This result suggests that diminishing transmembrane ion transport with bumetanide functioned to shift the chloride gradient (and, therefore, the GABAAR reversal potential) rather than blocking the response itself. As shown in the example recording, bumetanide hyperpolarized the estimated reversal potential of the GABAergic response by 25 mV. The similar slopes of the response amplitude versus membrane potential plots in Figure 7D also suggests that bumetanide did not block the underlying response.
We next asked whether all GABAAR-mediated responses were affected by bumetanide. In addition to depolarizing GABA responses, MCs also received spontaneous hyperpolarizing IPSPs that likely arise from numerous dendrodendritic inhibitory inputs along MC secondary dendrites (Rall et al., 1966; Isaacson and Strowbridge, 1998; Fig. 7E) and are also mediated by GABAAR (Pressler and Strowbridge, 2017). Both evoked depolarizing and spontaneous hyperpolarizing IPSPs could be recorded in the same neurons and at the same approximately −60 mV holding potential. The mean IPSP amplitude at −60 mV and reversal potential of spontaneous GABAergic IPSPs were unaffected by bumetanide (p > 0.05 for both comparisons; Fig. 7F). Figure 7, G and H, summarizes the shift in reversal potential of the evoked depolarizing GABA response with bumetanide and the stability of reversal potential of spontaneous IPSPs recorded in the same five MCs. The specificity of bumetanide for evoked depolarizing GABA responses raises the possibility that depolarizing and hyperpolarizing GABA responses maybe occur in spatially separated subcellular compartments that have different chloride gradients, as shown recently in recent studies on cortical neurons (Weilinger et al., 2020; Rahmati et al., 2021).
We next tested whether the estimated reversal potential of depolarizing GABA response was sensitive to changes in intracellular chloride ion concentration, comparing internal solutions with both higher (39 mm) and lower (3 mm) chloride ion concentrations with our control internal solution that has 9 mm Cl–. The estimated reversal potential for the depolarizing GABA response under control conditions (9 mm Cl–) was −10.4 ± 5.9 mV (N = 12; Fig. 8A, blue symbols) with all estimates between −40 and +25 mV. Increasing intracellular Cl– concentration would be expected to shift the chloride reversal potential in the depolarizing direction by 38 mV but failed to modulate the estimated reversal potential of the GCL evoked response (Fig. 8A, orange symbols). Similarly, decreasing the intracellular chloride ion concentration from 9 to 3 mm would be expected to shift the chloride reversal potential in the hyperpolarizing direction by 29 mV but failed to alter the depolarizing GABA reversal potential (Fig. 8A, green symbols).
Effect of varying intracellular chloride ion concentration on GCL-evoked PSP reversal potential. A, Plots of reversal potentials from example recording in control internal solution containing 9 mm chloride ions (blue symbols), reduced chloride concentration (3 mm; green), and increased chloride concentration (39 mm; orange). Arrows indicate estimated reversal potentials for each experiment. B, Estimated reversal potentials of spontaneous GABAergic IPSPs recorded in the same 3 experiments in A. C, Summary of estimated reversal potential of GCL-evoked responses. No significant differences among the three groups defined in A (p > 0.05, t test). D, Summary of estimated reversal potential of spontaneous IPSPs in the same set of experiments presented in C (3 vs 9 mm: p = 0.0041, t = 3.1, df = 15; 9 vs 39 mm: p = 6.72 × 10–5, t = 5.2, df = 14; both t tests).
While the average estimated reversal potential of GCL evoked responses in MCs was not altered by changing the intracellular chloride concentration, the reversal potentials of spontaneous IPSPs recorded in the same neurons closely followed the shifts predicted by changes in the electrochemical driving force (Fig. 8B). This control experiment demonstrates that the different recording conditions were effective in altering intracellular chloride ion concentrations in MC secondary dendrites.
Results from these experiments are summarized in Figure 8, C and D, and demonstrate that altering the intracellular chloride ion concentration shifted the reversal potential of spontaneous IPSPs in the direction and approximate magnitude expected from changes in the chloride equilibrium potential (Fig. 8D). Surprisingly, the estimated reversal potential of the GCL evoked response recorded in the same neurons was unaffected by increasing or decreasing intracellular chloride ion concentration (Fig. 8C).
We used three approaches to determine where the depolarizing GABAergic PSPs arise on MCs (Fig. 9). First, we compared depolarizing GABAergic responses in MCs with intact and truncated apical dendrites. Mitral cells form complex synaptic interactions with interneurons in the glomerular layer (Aungst et al., 2003; Kiyokage et al., 2010; Zhou et al., 2020), providing a potential substrate for depolarizing PSPs. However, we found that MCs with apical dendrites severed within the EPL (Fig. 9A), lacking the distal apical tufts that contact glomerular layer interneurons, also showed robust depolarizing GABAergic PSPs evoked under the same conditions used previously in this study (Fig. 9C, black trace; N = 5 truncated MCs). We could also evoke depolarizing GABAergic PSPs in neurons with axons severed close to the soma, excluding the distal axon as the source of depolarizing PSPs (N = 4 with axons severed ∼50 μm from the soma). Together, these findings suggest that the depolarizing GABAergic input arises from the cell body or proximal axonal or dendritic regions of the MC.
Depolarizing GABAergic inputs are localized to the proximal axon. A, Two-photon z-stack montages showing an example intact (left) and truncated (right) GC. Border between the EPL and glomerular layer indicated by a dashed white line. Arrow indicates apical dendritic truncation in truncated MC. B, Timeline for focal gabazine application experiment. Asterisks indicate GCL stimulation times (every 10 s). C, Reduction in depolarizing GABAergic PSP response by focal gabazine (50 μm) applied to the axon of the recorded MC. Control response (black trace) evoked by GCL stimulation was reduced when tested 3 s after a single 20 ms gabazine application to the axon (orange trace). Depolarizing GABA response partially recovers at 13 s (blue) and 23 s (red) postapplication. Recorded MC was truncated and had no apical dendritic tuft (same truncated MC shown in A, right). D, Focal gabazine application near the axon of an intact MC also reduces the depolarizing GABAergic PSP. E, Graded attenuation of depolarizing GABAergic PSP by varying duration gabazine applications near the axon. Responses recorded from the same MC and the same gabazine application pipette position. Pressure application duration varied from 10 ms (left panel), 20 ms (middle), and 50 ms (right). Washout traces (green) obtained 1–2 min following gabazine application. F, Diagram of experiment configuration for testing spatial localization of depolarizing GABA response. G, Plot of attenuation of depolarizing GABA response by focal gabazine application (20 ms duration, applied 3 s before GCL stimulation) to the secondary dendrite (orange bar), soma (red; <1: p = 3.3 × 10–4, t = 5.8, df = 7), proximal axon 25 μm from soma (green; <1: p = 7.4 × 10–7, t = 12.6, df = 8, t test), proximal axon 50 μm from soma (blue; <1: p = 3.6 × 10–9, t = 11.6, df = 15). Soma versus Axon25: p = 0.02, t = 2.25, df = 15; soma versus Axon50: p = 0.007, t = 2.67, df = 22; both t tests. H, Similar analysis as in G generated from responses recorded 13 and 23 s following gabazine application in the same experiments. Soma <1: p = 1.3 × 10–4, t = 6.8, df = 7, t test; Axon25 <1: p = 1.8 × 10–5, t = 8.2, df = 8; Axon50 <1: p = 1.0 × 10–7, t = 9.0, df = 15; Soma versus Axon50: p = 0.032, t = 1.96, df = 22; t test.
In the second approach, we positioned a pipette filled with gabazine and a fluorescent dye under live two-photon visualization near the recorded MC to test whether blocking GABAARs in different subcellular regions attenuates the GCL-evoked depolarizing GABA response. In pilot experiments with the gabazine pipette positioned 5–10 μm away from the proximal axon, 20 ms duration pressure pulses ejected enough gabazine to rapidly attenuate the depolarizing GABA response (Fig. 9C–E). Responses recorded 3 s following gabazine application showed the most attenuation; depolarizing GABA responses slowly recovered on subsequent test GCL stimuli evoked every 10 s. Full recovery required 1–2 min intervals between gabazine applications (Fig. 9E, compare black, green traces). The degree of response attenuation was graded with the duration of gabazine pressure application (Fig. 9E, orange traces). Focal gabazine application to the proximal axon region attenuated depolarizing GABA responses in both intact MCs (Fig. 9D) and truncated MCs lacking apical dendritic tufts (Fig. 9C).
We compared the effectiveness of focal gabazine applications to different subcellular regions of the MC (Fig. 9F, experiment configuration; all 20 ms duration applications). Focal gabazine application significantly reduced the depolarizing GABA response when applied to the soma and axon locations but not when applied to the secondary dendrite (p > 0.05). The two axonal regions tested with focal gabazine (25 and 50 μm from the soma) generated significantly greater reduction in depolarizing GABA response amplitude than somatic gabazine applications (Fig. 9G), suggesting that the proximal axon was the site of depolarizing GABA input. We found a slightly greater average reduction at the 50 versus 25 μm axon site, but this difference was not statistically significant (p > 0.05). In control experiments, we found little effect of gabazine ejected in the GCL but not directly on the recorded MC (86.4 ± 3.2% of control response; N = 12).
We found similar patterns of gabazine effectiveness when analyzing the responses recorded as gabazine started to diffuse away from the cell (acquired 13 and 23 s following the gabazine application; Fig. 9H). Although the gabazine blockade of the depolarizing GABA response had started to subside, we still found statistically significant reductions following focal application to the soma and proximal axon, and no significant reduction when gabazine was applied to the secondary dendrite (p > 0.05). As with the results acquired immediately after gabazine application, these responses showed the proximal axon as the most sensitive site tested (significantly greater reduction in response amplitude from gabazine applied to the axon, 50 μm from the soma, than to the soma itself). Together, these results suggest that GCL stimulation depolarizes MCs by activating GABAARs located on the axon.
The previous gabazine focal application experiment suggests that the depolarizing GABA input arises from GABAA receptors located on the proximal axon. In the third experiment, we tested this prediction by applying GABA to sites along the proximal axon visualized with live two-photon imaging. We first verified that focal GABA application at sites along the MC secondary dendrite evokes an inhibitory response that reverses near rest, as previously reported (Xiong and Chen, 2002). Consistent with these reports, we found hyperpolarizing GABA responses recorded at a −50 mV holding potential in nine of nine experiments with dendritic GABA applications (Fig. 10A,B). The average reversal potential in these experiments was −63.1 ± 1.1 mV (N = 5), similar to the reversal potential of spontaneous GABAergic IPSPs shown in Figure 7H (−60.3 mV in control conditions).
Responses to focal GABA application on MC dendrites and axons. A, Two-photon z stack montage illustrating the MC recorded in B in relation to the GABA application pipette. Pressure application solution contained 100 μm GABA and 100 μm Alexa Fluor 594. Images acquired after the recording session was concluded; tip fluorescence diminished during the postrecording period. B, Example responses to focal GABA applied to the secondary dendrite position illustrated in A (10 ms application at 10 psi). C, z-Stack montage illustrating MC recording in D in relation to the pipette containing GABA directed at the MC axon. D, Example responses to axonal GABA applications in the experiment depicted in C. E, Summary plot of estimated reversal potentials from experiments using GABA applications directed at the proximal axon (<75 μm from cell body) and the secondary dendrite (p = 0.0023, t = 5.1, df = 6, t test). F, Depolarizing responses to GABA applied to the proximal axon are blocked by bath application of gabazine. G, Plot of time course of gabazine blockade of axon GABA responses.
We then tested responses to focal GABA applications along the initial 100 μm of the axon in 12 experiments from nine MCs. In ∼40% of those experiments (5 of 12), focal pressure GABA applications evoked depolarizing PSPs recorded at −50 mV (Fig. 10C–E). These responses resembled the depolarizing GABAergic PSPs evoked by GCL stimulation and had estimated reversal potentials near −10 mV (−7.6 ± 14.8 mV; N = 3), similar to the reversal potential of depolarizing GABA responses evoked by GCL stimulation. We abolished the depolarizing response to focal GABA applied to the axon using bath application of gabazine in three of three experiments (Fig. 10F,G), demonstrating that the response required GABAA receptors. We found hyperpolarizing responses to GABA recorded at −50 mV in the other seven axonal application experiments, including in one MC recording that generated both depolarizing and hyperpolarizing GABA responses at different sites along the proximal axon. The reversal potential of axonal hyperpolarizing GABA responses (−57.5 ± 1.2 mV; N = 4) was similar to the reversal potential of dendritic GABA applications and spontaneous IPSPs. There was a small but statistically significant difference in the distance from the soma GABA application sites that yielded hyperpolarizing (average, 44.3 ± 5.2 μm) and depolarizing (65.0 ± 6.1 μm; p = 0.027, t = 2.586, df = 10, t test) responses. While understanding the mechanistic basis of the heterogeneous GABA responses along the axon will require a future study, these results demonstrate that similar depolarizing GABAergic responses can be evoked in MCs using GCL stimulation and focal GABA application.
Discussion
We make three principal conclusions in this study. First, MCs can be excited by synaptic potentials mediated by GABAA receptors. The reversal potential of these inputs is more depolarized than the AP threshold voltage in MCs, enabling depolarizing GABA inputs to drive spiking. Second, the likely target of depolarizing GABA inputs is the proximal axon based on experiments using focal blockade of GABAA receptors. The sensitivity of depolarizing GABA responses to bumetanide suggests that differential expression of ion transporters in the axon may depolarize the chloride equilibrium potential in that subcellular compartment. And third, the depolarizing GABA input represents a novel local circuit pathway that enables MCs to be excited by inputs beyond the spatially restricted glomerulus containing classical OSN-to-MC synapses (Kosaka et al., 2001; Kosaka and Kosaka, 2005).
Mechanism of depolarizing GABA excitation
Multiple lines of evidence suggest that the gabazine-sensitive depolarizing input we find in MCs following GCL stimulation represents a synaptic input, not a recording artifact. The depolarizing GABA response we recorded has a defined onset latency (Fig. 3C) consistent with other types of monosynaptic inputs to CNS neurons. An ∼2 ms latency would not be expected in passive response triggered by extracellular stimulation nor would clearly separable failures and successes at perithreshold stimulus intensities (Fig. 3A). The depolarizing GABA response was sensitive to TTX and displayed short-term depression, suggesting that it required spiking in presynaptic GABAergic processes and providing additional evidence against passive stimulation responses. Our finding that the amplitude of depolarizing GABA responses varies linearly with changes in membrane potential suggests that the response is generated on MCs and is not coupled in MCs through electrotonic connections. Dendritic gap junctions play a central role in synchronizing clusters of MCs (Christie et al., 2005). The large attenuation of responses through gap junctions (Christie et al., 2005; mean coupling coefficient between MC dendrites, 0.04) likely limits the functional role of these nonsynaptic interactions to transmitting depolarizations from large-amplitude action potentials rather than smaller postsynaptic potentials.
To our knowledge, there are no previous reports about depolarizing GABAergic responses in postnatal MCs. Mitral cells form a myriad of complex interactions within the glomerular layer that presumably regulate how sensory inputs activate distal dendritic compartments of the OB principal cells (Schoppa and Westbrook, 2001; Najac et al., 2011; Gire et al., 2012; Liu et al., 2016). The novel synaptic input we identified appears to be distinct from those glomerular circuit connections since it was present in MCs with truncated apical dendrites. The most sensitive site for reducing depolarizing GABAergic responses using focal applications of gabazine was the axon, the subcellular MC compartment most distant from the glomerular layer. One caveat in this experiment is that we applied gabazine to the secondary dendrite on the side opposite to the stimulating electrode. This potentially impaired the receptor blockade in these control experiments, compared with axonal gabazine applications. However, the negative reversal potential we find for dendritic GABA applications and spontaneous IPSPs both provide additional support for the hypothesis that the depolarizing GABA response does not originate in dendrites. We are also unaware of other reports that used the approach we found most reliable for revealing depolarizing GABAergic inputs including focal GCL stimulation, bath application of high concentrations of ionotropic glutamate receptor antagonists to maximally suppress coactivated glutamatergic EPSPs, and intracellular blockade of voltage-gated Na+ channels (with QX-314) to reduce membrane potential noise and to enhance linearity when the membrane was stepped to different potentials. Under these conditions, the vast majority of the GCL stimulation-driven response was blocked by gabazine and other GABAAR antagonists (Figs. 1B,C, 3E–G).
The depolarizing GABA response we report differs from conventional dendritic GABAergic inhibition in the following three principal ways: the presumptive location on the proximal axon based on results from focal gabazine and GABA applications; the slow kinetics of postsynaptic response following GCL stimulation; and the depolarized reversal potential. While many types of principal neurons in CNS brain regions receive responses mediated by axo-axonic synapses, to our knowledge axo-axonic synapses have not been demonstrated in the OB. However, at least two classic ultrastructural studies describe symmetric synapses on the proximal MC axon, including both the axon hillock and initial segment (Price and Powell, 1970; Willey, 1973). Some of these synapses appear to be part of reciprocal synapses with presumptive interneuron processes, perhaps basal dendrites of GCs. Given the variety of GABAergic responses we find on MC axons, understanding of the ultrastructure of synapses in the GCL would benefit from revisiting this topic. We are unaware of previous reports of interneurons resembling chandelier cells, the classic GABAergic cell type associated with cortical axo-axonic synapses (Somogyi, 1977; Peters et al., 1982), in the OB. However axo-axonic connections are generated by interneurons with different morphologies in noncortical regions (Veres et al., 2014) in basolateral amygdala. Our results do not address the identity of the presynaptic neuron that contacts MC axons and also cannot exclude the possibility that the presynaptic GABAergic input arises from neurons outside the OB.
While multiple previous studies explored MC responses to focally applied GABA (Chen et al., 2000; Wang et al., 2005; Gödde et al., 2016), few focused on identifying potential axonal synaptic inputs. One study that did focus on axonal inputs uncaged GABA along the proximal axon (Lowe, 2002) but performed their recordings using a high-intracellular chloride internal solution. Identifying a normally depolarizing axonal input is difficult under those conditions since the reversed chloride electrochemical gradient made all GABAAR-mediated responses depolarizing. Intriguingly, the authors reported inward voltage-clamped currents whose magnitude varied as the uncaging spot was swept along the MC axon, consistent with the heterogeneity we find with both depolarizing and hyperpolarizing axonal responses to focally applied GABA.
The kinetics of the depolarizing GABA response is slower hyperpolarizing GABAergic responses recorded in MCs (Isaacson and Strowbridge, 1998) and GCs (Pressler and Strowbridge, 2006). The slower kinetics could reflect structural features associated with the proximal axonal region that impair electrical access, specific GABAA receptor subunits, or different synaptic arrangements. Intriguingly, the kinetics we find on depolarizing GABA inputs closely resemble GABAA,slow IPSPs recorded in the hippocampus and neocortex (Capogna and Pearce, 2011; Overstreet-Wadiche and McBain, 2015). These inputs are evoked by defined interneuron subtypes, neurogliaform interneurons and Ivy cells, that create dense axon networks near postsynaptic regions. These connections often fail to form typical tight synaptic junctions with postsynaptic processes. The slow kinetics of GABAA,slow responses likely reflect these relatively long distances (several micrometers) between presynaptic and postsynaptic sites, which likely allow the pooling of GABA released from multiple interneuron varicosities (Oláh et al., 2009). Future work will be required to determine whether a similar mechanism operates near MC axons to enable slow GABAAR-mediated responses when GABA is released simultaneously from multiple nearby interneuron processes.
The depolarized reversal potential we find in MC GABA responses likely reflects subcellular differences in the chloride equilibrium potential. The selective change in the reversal potential of depolarizing GABA responses evoked by GCL stimulation bumetanide treatment (Fig. 7) suggests that NKCC1 helps to set depolarized chloride reversal potentials in the MC axon. Previous studies identified NKCC1 as an important mechanism for generating depolarizing GABAergic responses in cortical neurons (Yamada et al., 2004; Brumback and Staley, 2008). The same transporter appears to promote depolarized GABA responses in the axon initial segment of cortical pyramidal cells (Khirug et al., 2008) that are capable of triggering APs (Szabadics et al., 2006). In both the study by Khirug et al. (2008) and the present study, the reversal potential of presumed dendritic GABA responses were unaffected by blocking NKCC1, suggesting that this transporter plays an especially important role in axons.
The hypothesis that spatially localized axonal compartments may exist with depolarizing chloride reversal potentials is counterintuitive, given the presumed ability of chloride ions to diffuse throughout the MC. However, several recent studies have demonstrated localized subcellular regions where chloride reversal potentials are substantially more depolarized than in the somatic region (Weilinger et al., 2020; Rahmati et al., 2021). Both of these reports suggest that distinct chloride electrochemical gradients in subcompartments are maintained by nonuniform expression of several ion transporters, including the NKCC1 cotransporter our studies implicate in regulating chloride ion concentration in MC axons. These studies also highlight the observation that mobile anions, such as chloride ions, coexist with far more numerous nonmobile negative charges associated with intracellular proteins and nucleic acids (Veech et al., 2002; Morawski et al., 2015). This large collection of fixed anions is not uniformly distributed among different intracellular compartments (Gut et al., 2018; Chen et al., 2020). Presumably intracellular concentrations of mobile anions, such as chloride ions, must also be distributed nonuniformly to compensate for heterogeneous distribution of fixed anions (but see Rahmati et al., 2021).
The inability we find of intracellular chloride manipulations to affect the depolarizing GABA response may reflect high expression of ion transporters such as NKCC1 in axonal sites near GABAergic synapses. The small diameter of MC axons also may help restrict ion diffusion into postsynaptic regions. Alternatively, postsynaptic axonal regions may represent functionally isolated nanodomains. Future studies could explore these potential mechanisms by manipulating intracellular chloride concentration while blocking different ion transporters.
Functional roles of depolarizing GABA input
The results presented in this study suggest that principal cells in the OB can be excited through multiple synaptic mechanisms, including both by direct sensory afferents and by GABAergic axonal inputs. The relative strengths of both excitatory circuits is difficult to assess from our present results since most of our experiments used low-intensity stimuli and Na+ spike generation was blocked with QX-314 to lower membrane potential noise at depolarized holding potentials. Relatively weak axonal depolarizations likely function to modulate the threshold voltage required in the soma/dendritic region to trigger APs that propagate to downstream cortical target regions. If depolarizing GABAergic inputs are strong enough to trigger APs in resting MCs, then the origin of MC discharges becomes ambiguous—whether they are triggered by sensory input to apical dendritic synapses or direct axonal depolarization. Direct activation of MCs via axonal GABAergic inputs also may occur only under restricted conditions such as when many presynaptic GABAergic cells are firing synchronously during emergent oscillations.
GABAergic interneurons play a central role in generating emergent network oscillations in the OB, especially at gamma-band frequencies commonly recorded in vivo immediately following odor stimulation (Bressler and Freeman, 1980; Kay, 2014). Previous work demonstrated that pharmacological blockade of GABAA receptors can disrupt olfactory oscillations (Stopfer et al., 1997; Friedman and Strowbridge, 2003; Lagier et al., 2007), potentially linking inhibitory synaptic interactions with the cellular mechanisms responsible for network oscillations in the olfactory system. The present study suggests a second, independent action of GABAA receptor antagonists—attenuating GABAergic excitation of MCs—that may contribute to the ability of these agents to disrupt olfactory oscillations.
If the GABAergic interneurons mediating short-latency MC excitation participate in olfactory oscillations (Wilson and Laurent, 2005; Brea et al., 2009; Fukunaga et al., 2014), their direct connection to the axonal spike generation region in OB principal cells (Chen et al., 1997, 2002; Djurisic et al., 2004; Lorincz and Nusser, 2008) could function to help entrain precisely timed spikes in MCs during oscillation epochs. Precise spiking timing in OB output neurons is critically important in downstream regions when decoding sensory inputs (Schaefer et al., 2006; Fukunaga et al., 2012, 2014; Wilson et al., 2017; Stern et al., 2018), suggesting the potentially large functional role that depolarizing of GABAergic inputs plays in the normal operation of the OB. If this hypothesis is correct, a central prediction of this work is that selective blockade of depolarizing GABAergic inputs should selectively degrade spike coherence in MCs without necessarily abolishing gamma-band oscillations in interneurons.
Footnotes
This work was supported by National Institutes of Health Grant R01-DC-04285 to B.W.S. We thank Drs. Edward Cui and Chris Ford for helpful discussions related to this project.
The authors declare no competing financial interests.
- Correspondence should be addressed to Ben W. Strowbridge at bens{at}case.edu
