Male mice (Kunming, Mus musculus) were used in all experiments (21–30 days old for slice recording, and 8–16 weeks old for other experiments). The number of mice used for each experiment is reported in the Section “Results.” Mice were purchased from the National Rodent Experimental Animal Seed Center (Shanghai, China) and bred in the animal facilities of the Xuzhou Medical University. They were housed in a vivarium with a 12/12 light/dark cycle with lights on at 8:00 a.m. Experiments were performed during the light cycle. Food and water were available ad libitum. All experiments were performed according to protocols approved by the Xuzhou Medical University Institutional Animal Care and Use Committee.

Mice were anesthetized with diethylether and decapitated. The entire brain was quickly immersed in ice-cold artificial cerebrospinal fluid (ACSF; composition in mmol/L: 119 NaCl, 2.5 KCl, 1.3 MgCl₂, 2.5 CaCl₂, 1.25 KH₂PO₄, 1.3 NaHCO₃, and 20 glucose). Osmolarity was 300–320 mOsm and the pH was adjusted to 7.4. ACSF was bubbled continuously with Carbogen (95%O₂/5%CO₂). Coronal brain slices, 300 μm thick, were cut with a Leica VT1000s vibratome (Leica Biosystems, Wetzlar, Germany). Fresh slices were incubated in a chamber with carbogenated ACSF and were incubated at 35°C for 1 h and then at room temperature (26°C) before being transferred to the recording chamber.

Slices were visualized with a 60× objective on an upright microscope (ECLIPSE FN1, Nikon). APC was identified based on its location relative to the lateral olfactory tract (LOT) and the compact density of cells in layer 2. Pyramidal neurons in layer 2 in APC were recorded by patch clamp with an Axopatch-700B amplifier (Molecular Devices). The pyramidal neurons were identified by cell body morphology and location (Suzuki and Bekkers, 2011; Sheridan et al., 2014).

In current-clamp mode, for whole-cell action potential (AP) recording, pipettes were filled with an internal solution containing (in mM): 140 K-methylsulfate, 4 NaCl, 0.2 EGTA, 4 MgATP, 0.3 Na₃GTP, 10 phosphocreatine and 10 HEPES, 310 mOsm, pH 7.3 adjusted with KOH. A step current (200 pA with 500 ms duration) was used to evoke AP firing. The interval of each trial was 60 s and a total of 5 trials were used in both control and insulin conditions. Insulin (human recombinant insulin, Sigma, St Louis, MO, United States) was used at 1 μg⋅ml^-1 (172 nM) (Fadool et al., 2000, 2011; Colley et al., 2004). The concentrations of tetra-ethylammonium (TEA, Sigma, St Louis, MO, United States) and margatoxin (Sigma, St Louis, MO, United States) were 20 mM and 10 nM, respectively. All drugs were dissolved in ACSF and bath perfused over the entire slice. A single cell was recorded from each slice.

In voltage-clamp mode, miniature-excitatory postsynaptic currents (mEPSCs) and LOT-evoked excitatory postsynaptic currents (eEPSCs) were recorded at -60 and -70 mV holding potential respectively while the slices were bathed in ACSF perfusion media including (+)-bicuculline (10 μmol/L, Enzo) to block GABA_A receptor-mediated inhibitory synaptic currents. The pipettes were filled with the internal solution containing (in mM): 105 K-gluconate, 30 KCl, 10 HEPES, 10 phosphocreatine, 4 ATP-Mg, 0.3 GTP-Na, and 0.3 EGTA. Miniature-inhibitory postsynaptic currents (mIPSCs) were recorded at -70 mV holding potential in slices bathed in ACSF perfusion media containing D-(-)-2-amino-5-phosphonopentanoic acid (50 μM, APV; Tocris Bioscience) to block the N-Methyl-D-aspartate (NMDA)-type ionotropic glutamate receptors and 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide (20 μM, NBQX; Tocris Bioscience) to block the α-amino-3-hydroxy-5-methyl-ioxazole-4-propionic acid (AMPA)- type receptors. The pipettes were filled with the internal solution containing (in mM): 135 CsCl, 2 Na₂ATP, 0.2 EGTA, 10 HEPES, 0.3 Na₃GTP, and 10 glucose, 310 mOsm, pH 7.2 adjusted with CsOH. In both mEPSC and mIPSC recordings, APs were eliminated by blockage of Na⁺ channels with tetrodotoxin (TTX, 1 μM; ACROS, Waltham, MA, United States). The mEPSCs and mIPSCs were recorded separately.

The mEPSCs/mIPSCs were recorded for 30 min, and the insulin was applied to the slice at the end of the 5th min. The eEPSCs were recorded for 15 trials in both control and insulin conditions under five different pulses (0.1, 0.15, 0.2, 0.25, and 0.3 mA, respectively) with a bipolar concentric electrode placed on the LOT and the pulses (200 μs) were delivered through a stimulus isolation unit (A.M.P.I). The interval between each trial was 20 s. Data were acquired at 10 kHz and low-pass filtered at 2 kHz. During the recording, the pipettes had a tip resistance of 3–5 MΩ. The series resistance was always monitored during recording to avoid resealing the ruptured membrane, which would cause changes in both the kinetics and amplitude of the mEPSC/mIPSC. Recordings were excluded from analysis when the series resistance or capacitance deviated by more than 15% from initial values. Data were collected and processed with pCLAMP10.2.

We measured spatial expression of IRs and CaMkII alpha (a marker of pyramidal cells) by immunostaining in the APC of the mice (n = 5). The mice were euthanized and then intracardially perfused with 0.9% saline followed by 4% paraformaldehyde in phosphate buffer (0.1 M, pH 7.4). The brain was harvested and post-fixed for 24 h in 4% paraformaldehyde at 4°C and then was cryoprotected with 30% sucrose until it sank. The tissue was sectioned coronally, 40 μm thick, with a cryostat (Leica). For immunohistochemistry, brain cryosections were preincubated for 4h at room temperature with a blocking buffer containing phosphate-buffered saline (PBS) and 0.5% Triton X-100 with 5% normal serum from the host species of the secondary antibodies. The sections were then incubated overnight at 4°C with rabbit anti-IR-Beta primary antibody (1:50; #sc-711, Santa Cruz Biotechnology) and anti-mouse CaMKII alpha primary antibody (1:150; #05-532; Millipore) diluted in the blocking buffer. Then, sections were washed with PBS and incubated for 1.5 h at room temperature with secondary antibodies for immunostaining: Alexa Fluor 488 anti-rabbit (1:500; catalog #A11008; Invitrogen), and Alexa Fluor 555 anti-mouse (1:500; catalog #A31570; Invitrogen). After the final wash, slides were incubated with DAPI for nuclear staining, and were coverslipped with 50% glycerol mounting medium.

Images were obtained by confocal scanning microscopy (Zeiss, LSM710) and were processed via ZEN 2011 (Zeiss). We manually drew 1–2 regions of interest (ROIs) in layer 2 of APC of the same size (diameter = 50 μm) in the different images from 5 mice. The immunostained neurons were counted with the ImageJ using Cell Counter plugin.

Mice were anesthetized with intraperitoneal injection of sodium pentobarbital (90 mg/kg) and were mounted on a stereotaxic apparatus equipped with an electric heating pad. The skin was cut and a small craniotomy was made above targeted areas at stereotaxic coordinates (APC: AP, +2.1 mm; Lateral, 2.0 mm; DV: 4.0 mm; OB: AP, +4.3 mm; Lateral, 1.0 mm; DV: 1.8 mm). A cannula (O.D = 0.48 mm, stainless steel needle, RWD Life Science Company, Shenzhen, China) was implanted into the lateral cerebral ventricle (ICV) for insulin application in vivo.

For viral injection and fiber implant for fiber photometry, the virus (AAV- CaMKII-GCaMP6s, 1 μL, Brainvta, Wuhan, China) was slowly injected (50 nl/min) into the APC through a glass pipette, using a microsyringe pump (The Stoelting Quintessential Injector; Stoelting Co.). The glass pipette was left in place for 10 additional minutes and then slowly withdrawn. Following AAV-CaMKII-GCaMP6s virus injection, an optical fiber (200 μm O.D., 0.37 numerical aperture (NA); NEWDOON) was placed in a ceramic ferrule and inserted toward the APC through the craniotomy. The ceramic ferrule was supported with a stainless steel screw and dental acrylic. A custom-made aluminum head plate was attached to the skull with stainless steel screws and dental cement.

After implant of the fiber, mice were individually housed for at least 10 days for recovery from the surgery and expression of the virus. Fluorescence emission was recorded with a fiber photometry system (Thinkerbiotech, Nanjing, China) using methods similar to previous studies (Guo et al., 2015; Li et al., 2016). Briefly, a laser beam from a 488 nm laser (OBIS 488LS; Coherent) was reflected by a dichroic mirror (MD498; Thorlabs), focused through an objective lens (x10, NA = 0.3; Olympus) and then coupled to an optical commutator (Doric Lenses). An optical fiber (200 mm O.D., NA = 0.37, 1.5 m long) coupled the light between the commutator and the implanted optical fiber. The laser power was adjusted at the tip of the optical fiber to the level of 40–60 μW. The GCaMP6s fluorescence emission was bandpass filtered (MF525-39, Thorlabs) and detected by a photomultiplier tube (R3896, Hamamatsu). An amplifier (C7319, Hamamatsu) was used to convert the photomultiplier tube current output to voltage, which was further filtered through a low-pass filter (35Hz cut-off; Brownlee 440). The analog voltage signals were digitalized at 500 Hz and recorded by fiber photometry software (Thinkerbiotech, Nanjing, China).

Local field potentials were recorded in the APC using stainless steel electrodes (catalog #791000; A-M systems) targeted to the APC stereotaxic coordinate (AP, +2.1 mm; Lateral, 2.0 mm; DV: 4.0 mm) with reference to a skull screw implanted posteriolateral to bregma. Signals were amplified (2000×, A-M systems), filtered at 0.1–300 Hz, and sampled at 2 kHz. LFP signals along with odor stimulation event markers were recorded with a homemade recording system based on a National Intruments sampling card (NI-USB-6009). For both fiber photometry and LFP recordings, insulin was applied via ICV injection, immediately after the recordings in control condition were done. The recordings under insulin condition started about 1 h later. Between the two recording conditions, the mice were free-moving in their home cages.

Awake mice were head-fixed with two horizontal bars (fixed with head-plate by two screws) and were able to maneuver on an air-supported free-floating Styrofoam ball (Thinkerbiotech, Nanjing, China). Different odors were presented by an odor delivery system (Thinkerbiotech, Nanjing, China). Four odorants, isoamyl acetate, 2-heptanone, phenyl acetate and benzaldehyde (Sinopharm Chemical Reagent Co. Shanghai, China) were used. The odors were dissolved in minal oil at 10% dilution. A stream of charcoal-filtered air flowed over the oil, and then was diluted to 1/20 by an olfactometer. The odor stimulation was synchronously controlled with the data acquisition system by a solenoid valve, which was driven by a digital to analog converter. Air or odorized air were delivered to the nose at a constant rate of 1 l/min to eliminate the effect of the airflow. For each odor, 20 trials were presented with inter-trial intervals of 30 s. The duration of each odor presentation was 2 s.

One hour before APC tissue separation, mice received ICV injection from a microliter syringe connected to the needle implanted into the brain (AP: -0.5 mm; ML: 1.0 mm; DV: 3.0 mm) with either NaCl (0.9%) or 14 mU insulin (human, recombinant, expressed in yeast, Sigma–Aldrich) administered in a 2 μL NaCl vehicle over 60 s, using a 10 μL microliter syringe (Aime et al., 2012). Insulin was extracted from APC tissues according to the procedure of Fadool et al. (2000). Insulin level was determined using an insulin enzyme-linked immunosorbent assay (ELISA) kit (Nanjing Jiancheng Biology Engineering Institute, Nanjing, China) according to the manufacturer’s defined protocol.

For the current-evoked APs (Figure 2), firing properties for each neuron were analyzed using Clampfit software (Molecular Devices). The distribution of inter-spike intervals (ISIs) was determined for the 5 trials prior to insulin application and 5 trials midway through the perfusion of insulin. The burst index was calculated as t_l / t_f, where t_f is the time interval between the first and second APs, and t_l is the interval between the last two APs. For a given pyramidal cell, insulin was considered to affect the ISIs of this cell significantly when P < 0.05 by a two-sample Kolmogorov–Smirnov test (Figures 2A–C,E1,E2). The first latencies of a given cell were compared by Student t-test (unpaired, P < 0.05 was considered as significant effect, Figures 2D1,D2). The effect of insulin on latency/ISI across the group of neurons recorded was tested by a paired t-test (Figures 2D3,E3). The correlation between two groups (e.g., Figures 2D4,E4, 5K,L, 6K,L) was tested by liner regression, and the significance was further tested by an F-test.

For the mEPSC and mIPSC, the frequency and amplitude were characterized using Minianalysis (Synaptosoft) software with a bin of 30 s. The rising time and decay time of the mEPSC/mIPSC were characterized using Clampfit software. The data 5 min prior to insulin application (0–5th min, 10 bins) was considered as control, and data for the 5 min following 2 min after insulin application (8th–13th min, 10 bins) was considered as the insulin condition. The frequency/amplitude between the two conditions for each neuron was compared by unpaired t-test (10 bins/samples for each condition, Figures 5C,D, 6C,D). The averaged effect of insulin on frequency/amplitude between the two conditions across the group of neurons recorded (22 for mEPSC and 20 for mIPSC, respectively) was compared by a paired t-test (Figures 5I,J, 6I,J).

For the eEPSC, the amplitude was characterized using Clampfit software. The amplitudes were averaged for all the 3 trials under the same stimulation pulse prior or subsequent to insulin application. For each neuron, the amplitude of eEPSC evoked by 0.3 mA was normalized to 100 (Figure 7B), and the amplitude evoked by other current intensities and all currents intensities under insulin condition were normalized relative to this amplitude. Two-way ANOVA (effect on insulin and the intensities of stimuli) was applied to test whether insulin affects the amplitudes significantly across intensity (Figure 7B).

A custom program written in MATLAB was used to analyze the LFP signals. Raw data of 5 s prior to the onset of odor stimulation were selected as the ongoing activities. Time-frequency transformation was performed based on this 5 s window (Hanning window; FFT size, 2048; frequency resolution, 0.977 Hz), and the spectral power was calculated for each frequency resolution. As in previous studies, LFP signals were divided into different frequency bands (Figure 8): theta (2–12 Hz), beta (15–35 Hz), and gamma (low, 36–65 Hz; high, 66–95Hz). For odor-evoked beta band LFP responses, windows 3 s prior to and 5 s after the onset of odor stimulation were selected (Figure 9). To obtain both high time and frequency resolution, this time course was divided into segments with 1 s duration and 90% overlap. Time-frequency transformation was performed based on the 1 s windows. The spectral power was normalized to dB and calculated for each frequency bandwidth; the power from all frequencies included within the bandwidth was averaged. For each trial, the baseline was normalized to 1, and all the trials for each odor stimulation were averaged based on the normalized data (Figure 9).

Photometry data were exported to MATLAB mat files from the photometry software files for further analysis. The data were segmented based on the onset of odor stimulation within individual trials. We derived the values of fluorescence change (ΔF/F) by calculating (F-F0) /F0, where F0 is the baseline fluorescence signal averaged over a 5-s-long control time window, which was typically set preceding the onset of odor stimulation. ΔF/F values were presented as heatmaps or average plots (Figure 10).

The APC is one of the brain areas that express high levels of insulin receptors in rats (Unger et al., 1989). Here we sought to confirm this finding in mice and asked whether the insulin receptors are expressed by pyramidal neurons. In layer 2 of APC, where the soma of pyramidal neurons are located, we found that a large proportion of the DAPI-positive cells were insulin-receptor immunopositive (Figure 1, 67.4% ± 5.3, counted in 9 ROIs of 5 sections from 5 mice, total number of cells = 531). The insulin-receptor immunopositive signals were distributed in both the cytoplasm and the nucleus. Nuclear translocation of the insulin receptor has been proposed to be involved in mediating insulin’s long term effects (Podlecki et al., 1987). Furthermore, specific staining of the pyramidal neurons identified by immunostaining for CaMKII alpha revealed that most of the insulin receptor immunopositive cells and pyramidal neurons were colocalized (Figure 1). In total, we found 92.2% ± 0.02 of CaMKII alpha positive neurons were insulin receptor immunopositive, and 53.4% ± 0.06 of insulin receptor immunopositive cells were CaMKII alpha positive (counted in 9 ROIs of 5 sections from 5 mice, total number of cells = 200 CamKII positive, 531 DAPI labeled). Thus, these results indicate that the insulin receptors exist in a large percent of pyramidal neurons in the mouse APC.

To test the effects of insulin on pyramidal neuron excitability, we recorded current-evoked APs from pyramidal neurons in APC slices. We observed a substantial decrease elicited by insulin in the first latency of the current-evoked APs. Examples of the shortening of the latency are shown in Figures 2A1,C1. This shortening was observed in most of the neurons recorded (23/38, 60.1%, 9 mice, Figures 2D1,D2), although no effect (3/38, 7.9%) or the opposite effect (12/38, 31.6%) were also found in other neurons (Figures 2D1,D2). On the average, insulin shortened the latencies significantly across the group of neurons recorded [Figure 2D3, 24.1 ± 2.3 ms vs. 14.2 ± 1.8 ms, paired t-test, n = 38 neurons from 9 mice, t₍₃₇₎= 4.64, P < 0.001]. Furthermore, we found a significant linear negative correlation between the first latency under control condition and normalized change in latency (linear regression, n = 38, r = 0.60, P < 0.001, Figure 2D4). This correlation suggests that insulin tends to shorten the first latencies of pyramidal neurons with longer latencies.

In addition, we noticed that insulin application elicited a modest decrease on the ISI of APs for a proportion of the neurons recorded (example shown in Figure 2A). Insulin also increased the ISI of a small number of neurons, and caused no effect on the remaining neurons (examples shown in Figures 2B,C). We found that the ISI of more than half of the neurons recorded was affected by insulin (18/38, 58.1%, 9 mice, Figures 2E1,E2), among which 77.8% (14/18) displayed a decrease in ISI, and 22.2% (4/18) showed an increase (Figures 2E1,E2). On the average, insulin shortened the ISI slightly, but significantly, across the group of neurons recorded [Figure 2E3, 46.7 ± 2.0 ms vs. 43.5 ± 1.8 ms, paired t-test, n = 38 neurons from 9 mice, t₍₃₇₎ = 3.00, P = 0.005]. Finally, we found a significant linear negative correlation between the change in ISI elicited by insulin and the ISI for each neuron (linear regression, n = 38, r = 0.36, P = 0.025, Figure 2E4).

To determine whether the effects observed above were caused by insulin as opposed to other factors such as rundown, we performed a control experiment in 11 pyramidal neurons using inactivated insulin. We found that the inactivated insulin affected neither the ISI [50.7 ± 4.3 ms vs. 50.1 ± 4.3 ms, paired t-test, n = 11 neurons from 3 mice, t₍₁₀₎ = 0.35, P = 0.73] nor latency [22.4 ± 4.6 ms vs. 19.2 ± 4.4 ms, paired t-test, n = 11, t₍₁₀₎= 1.46, P = 0.18].

From Figures 2A1,C1, we also observed an obvious elevation of RMP during insulin (red traces vs. black traces). We compared the RMP under control and during insulin application for all the 38 neurons, and found that insulin elevated the RMP of most of the neurons [Figures 3A,B, -66.5 ± 1.2 mV vs. -58.7 ± 1.3 mV, paired t-test, n = 38 neurons from 9 mice, t₍₃₇₎ = 8.04, P < 0.001]. As a control, we did not find a significant change of RMP [-63.1 ± 2.6 mV vs. -62.7 ± 3.1 mV, paired t-test, n = 11 neurons from 3 mice, t₍₁₀₎= 0.29, P = 0.78] using the inactivated insulin. Furthermore, we didn’t find a significant correlation between the RMP under control condition and normalized ΔRMP induced by insulin (Figure 3C, linear regression, n = 38, r = 0.23, P = 0.17), suggesting this is a common effect for all the neurons.

Next, we asked whether the elevation of the RMP was caused by blocking of potassium channels. We selected tetra-ethylammonium (TEA) and margatoxin, two potassium channel inhibitors known to block different Kv1 and Kv3 potassium channel subtypes (Ishikawa et al., 2003). Kv1.3 is the major contributor to insulin modulation of the activity of mitral cells in the OB (Fadool et al., 2000, 2011; Palouzier-Paulignan et al., 2012). After potassium channels were blocked by TEA, insulin still elevated the RMP significantly [Figures 3D,F, from -62.0 ± 1.7 mV to -58.3 ± 1.8 mV, n = 16 for each group from 6 mice, paired t-test, t₍₁₅₎ = 3.09, P = 0.007]. However, the increase elicited by insulin was significantly decreased [Figure 3H, from 11.5% ± 1.4 to 5.8% ± 2.0,unpaired t-test, n = 38 and 16, t₍₅₂₎ = 2.28, P = 0.027], indicating that TEA partly abolished the effect of insulin. After application of both TEA and margatoxin, we found that the effect of insulin on pyramidal neurons was almost completely abolished [Figures 3E,G, from -60.2 ± 3.2 mV to -59.3 ± 3.1 mV, n = 16 for each group from 6 mice, paired t-test, t₍₁₅₎= 0.99, P = 0.34]. Therefore, these results show that potassium channels, perhaps Kv1.3, are the major contributors to the effect of insulin on the pyramidal neurons in APC.

Since there are two types of pyramidal neurons in layer 2 of APC, SL (semilunar) and SP (superficial pyramidal) (Suzuki and Bekkers, 2006; Tantirigama et al., 2017), we next asked whether the effects of insulin differ between these neurons. As reported in previous study (Suzuki and Bekkers, 2006), the SL and SP could be differentiated by morphology and burst index (see Materials and Methods): the burst index is smaller than 2 for SL and larger than 2 for SP (Figure 4A). For the 38 neurons, 20 were SL and 18 were SP (Figure 4A). The RMP of SP was significantly lower than SL [-63.7 ± 1.6 mV vs. -69.6 ± 1.5 mV, unpaired t-test, n = 20 and 18, t₍₃₆₎ = 2.67, P = 0.01, Figure 4B], which was consistent with previous studies (Suzuki and Bekkers, 2006; Sheridan et al., 2014). However, we did not find any significant difference between SP and SL for the effect of insulin on ISI [20.0% ± 4.1 vs. 12.8% ± 5.0, unpaired t-test, n = 20 and 18, t₍₃₆₎ = 1.12, P = 0.27, Figure 4C], first latency [5.5% ± 2.7 vs. 6.0% ± 3.5, unpaired t-test, t₍₃₆₎ = 0.13, n = 20 and 18, P = 0.90, Figure 4D], or RMP [12.6% ± 1.9 vs. 10.3% ± 2.1, unpaired t-test, n = 20 and 18, t₍₃₆₎ = 0.81, P = 0.42, Figure 4E]. Therefore, it is likely that insulin modulates the response of these two types of cells with same mechanisms and to a similar magnitude.

To further investigate the effects of insulin on synaptic transmission in the pyramidal neurons of APC, we recorded mEPSCs and mIPSCs of the pyramidal neurons in the slices. To ask whether insulin causes any effect on excitatory synaptic component, we recorded the mEPSCs while blocking all inhibitory synapses (Figures 5A,B, see Materials and Methods). We found that insulin significantly changed the frequency and/or amplitude of the mEPSCs in some neurons. Figures 5A–D shows that the frequency [Figure 5C, 0.53 ± 0.05 Hz vs. 0.85 ± 0.06 Hz, unpaired t-test, n = 10 bins for both control and insulin, t₍₁₈₎ = 4.10, P = 0.0007] and amplitude [Figure 5D, 7.8 ± 0.1 pA vs. 8.4 ± 0.2 pA, unpaired t-test, n = 10 bins for both control and insulin, t₍₁₈₎ = 2.32, P = 0.03] of mEPSC of a representative pyramidal neuron were both significantly increased by insulin. We found that the mEPSC frequencies of 50.0% (11/22) of the neurons were significantly changed by addition of insulin, among which 72.7% (8/11) were increased and 27.3% (3/11) were decreased (Figures 5E,G). For mEPSC amplitudes, 36.4% (8/22) neurons were significantly changed, among which 75.0% (6/8) were increased and 25.0% (2/8) were decreased (Figures 5F,H). In general, insulin increased the frequencies of mEPSC significantly [Figure 5I, 1.9 ± 0.3 Hz vs. 2.1 ± 0.3 Hz, paired t-test, n = 22 neurons from 11 mice, t₍₂₁₎ = 2.26, P = 0.035] but caused not significant effect on the amplitude [Figure 5J, 9.9 ± 0.7 pA vs. 10.0 ± 0.6 pA, paired t-test, n = 22 neurons from 11 mice, t₍₂₁₎ = 0.60, P = 0.55] across the group of neurons recorded. The increased effect on the frequency of mEPSC was observed for the application of insulin, but not for inactivated insulin [1.98 ± 0.18 Hz vs. 1.89 ± 0.17 Hz, paired t-test, n = 8 neurons from 3 mice, t₍₇₎ = 1.48, P = 0.18]. We did not find any difference between control and insulin for the rising time [8.60 ± 1.03 ms vs. 9.15 ± 1.28 ms, paired t-test, t₍₂₁₎ = 0.36, n = 20 neurons from 7 mice, P = 0.72] or decay time [20.8 ± 2.05 ms vs. 25.4 ± 1.84 ms, paired t-test, t₍₂₁₎ = 1.76, n = 20 neurons from 7 mice, P = 0.09] of the mIPSCs. Furthermore, we found a significant linear negative correlation between the frequency/amplitude of mEPSC under control condition and normalized ΔFrequency/normalized ΔAmplitude (Figures 5K,L, linear regression, n = 22 neurons from 11 mice, r = 0.46 and 0.61, P = 0.03 and 0.002, respectively). These correlations suggest that insulin tends to increase the frequency/amplitude of mEPSC of the pyramidal neurons with lower frequency/amplitude.

We also recorded mIPSCs while blocking all the excitatory components to test whether there is an effect of insulin on inhibitory synaptic components (Figures 6A,B). We found that insulin significantly changed the frequencies and/or amplitudes of the mIPSCs in some neurons. Figures 6A–D shows the frequency [Figure 6C, 2.1 ± 0.08 Hz vs. 1.2 ± 0.05 Hz, unpaired t-test, n = 10 bins for both control and insulin, t₍₁₈₎ = 9.25, P < 0.001] and amplitude [Figure 6D, 12.7 ± 0.2 pA vs. 11.7 ± 0.3 pA, unpaired t-test, n = 10 bins for both control and insulin, t₍₁₈₎ = 2.76, P = 0.01] of mIPSC of a representatively pyramidal neuron are both significantly decreased by insulin. We found the mIPSCs frequencies of 60.0% (12/20, 9 mice) neurons were significantly changed by insulin, among which 16.7% (2/12) were increased and 83.3% (10/12) were decreased (Figures 6E,G). For mIPSC amplitudes, only 30.0% (6/20) were significantly changed, in which 33.3% (2/6) were increased and 66.7% (4/6) were decreased (Figures 6F,H). In general, insulin decreased the frequencies of mIPSC significantly [Figure 6I, 1.3 ± 0.2 Hz vs. 1.1 ± 0.1 Hz, paired t-test, t₍₁₉₎ = 2.64, n = 20 neurons from 7 mice, P = 0.016] and caused not significant effect on the amplitudes [Figure 6J, 11.9 ± 0.5 pA vs. 12.5 ± 0.7 pA, paired t-test, n = 20 neurons from 7 mice, t₍₁₉₎ = 1.74, P = 0.10] across the group of neurons recorded. These effects were observed for application of insulin, but not for inactivated insulin [1.26 ± 0.07 Hz vs. 1.24 ± 0.06 Hz, paired t-test, n = 11 neurons from 4 mice, t₍₁₀₎ = 1.11, P = 0.29]. We did not find any difference between control and insulin for the rising time [2.92 ± 0.27 ms vs. 2.88 ± 0.28 ms, paired t-test, t₍₁₉₎ = 0.19, n = 20 neurons from 7 mice, P = 0.85] or decay time [13.7 ± 1.62 ms vs. 13.4 ± 1.12 ms, paired t-test, t₍₁₉₎ = 0.26, n = 20 neurons from 7 mice, P = 0.80] of the mIPSCs. Interestingly, we found a significant linear negative correlation between the frequency, but not the amplitude of mIPSC under control condition and normalized Δfrequency/normalized Δamplitude (Figures 6K,L, linear regression, n = 20 neurons from 9 mice, r = 0.64 and 0.17, P = 0.002 and 0.46, for frequency and amplitude, respectively). This correlation suggests that insulin tends to decrease the frequency of mIPSCs of the pyramidal neurons with higher frequencies.

Since the APC receives direct and dense innervation from the olfactory bulb through the LOT, electrical stimulation of the LOT should induce strong EPSCs in the pyramidal neurons of the APC. We asked whether insulin has an effect on the LOT-evoked EPSCs in the APC. We found strong EPSCs in response to electrical stimulation of the LOT (Figure 7A). Insulin increased the amplitudes of the EPSCs significantly [Figure 7B, two-way ANOVA, n = 11 neurons from 4 mice, F_(1,100)= 12.7, P = 0.0006].

Next, we investigated the effects of insulin on the neural activity of the APC in vivo. We recorded the LFP signals from the APC with a single stainless steel electrode, and changed the insulin levels by injection in the ICV (see Materials and Methods, Figure 8A). In a parallel experiment, we tested the insulin level in the APC before and 1h after injection of insulin in the ICV, and found that the insulin level increased significantly after injection [0.24 ± 0.03 ng/mg vs. 0.36 ± 0.03 ng/mg, unpaired t-test, n = 4 groups for both control and insulin, each group contains 3 mice, t₍₆₎ = 2.85, P = 0.029, Figure 8B]. Thus, using this method, we changed insulin levels in the APC.

We compared the LFP signals in awake head-fixed mice (n = 9) before and after insulin injection. As in previous studies (Li et al., 2014, 2015), the raw LFP signals were divided into different frequency bands: theta, 2–12 Hz; beta, 15–35 Hz; low gamma, 36–65 Hz; high gamma, 66–95 Hz (Figure 8C). An example is shown in Figure 8C. Insulin caused no significant effect on theta and beta bands but decreased significantly low and high gamma bands of the LFP signals [Figure 8D, two-way ANOVA, n = 9 mice (11, 22, 31 and 30 frequencies for theta, beta, low gamma and high gamma bands, respectively), F_(1,176) = 2.97, F_(1,352) = 2.55, F_(1,496)= 22.7, and F_(1,480) = 75.2; P = 0.09, 0.11, 2.5^∗e-6, and 6.6^∗e-7 for theta, beta, low gamma and high gamma, respectively]. Therefore, insulin mainly decreased the gamma (both low and high) bands of ongoing LFP signals in awake mice.

We also investigated the effects of insulin on odor-evoked LFP responses in awake mice. Figure 9A shows an example of responses to isoamyl acetate in an awake mouse, with strong beta response. After application of insulin, the isoamyl acetate-induced increase in beta band was decreased (Figures 9A,B1). This phenomenon was also observed for other odorants (e.g., Figures 9B2,B4).

When the effect of insulin on odorant-induced changes in LFP power was surveyed in different mice and odors, we found that insulin decreased the odor-evoked beta response significantly [Figures 9C,D, two-way ANOVA, n = 36, F_(1,64)= 6.8, P = 0.01] across the group of mice recorded. Further analysis revealed that the decreased effect was not significant different among odors [Figure 9C, two way ANOVA, n = 36, F_(3,64) = 1.36, P = 0.26; Figure 9E, one way ANOVA, n = 9 mice for each group, F_(3,32) = 0.39, P = 0.76]. Therefore, our data indicate that insulin primarily decreases the odor-evoked beta responses.

Since mice rely on the respiration / sniffing to sample the odor, the decreased responses may be caused by the change of sampling rate (respiration rate) after application of insulin. We analyzed the theta oscillation of the OB LFP during the odor presentation which is highly correlated with the cycle of respiration / sniff in awake rodents (Rojas-Libano et al., 2014; Li et al., 2015). We found that there was no significant difference between control and insulin conditions [2.74 ± 0.13 Hz vs. 2.56 ± 0.12 Hz, n = 9 mice, paired t-test, t₍₈₎= 1.37, P = 0.21]. Therefore, the decreased odor-evoked responses after insulin application are not due to the change of the odor sampling rate.

The LFP signals reflect the local extracellular current-evoked by neural circuit activity in the APC. To ask whether insulin modulates the odor evoked activity of pyramidal neurons in awake mice, we used fiber photometry to specifically record the calcium signals from pyramidal neurons. We injected the virus AAV-CaMK2-GCaMP6s into APC, and a fiber was implanted after the virus injection (Figure 10A). Ten days later, we found extensive expression of GCaMP6s in layer 2 of the APC (Figure 10B). Odors evoked an increase in GCaMP6s fluorescence in the awake head-fixed mice (Figure 10C, upper row). After insulin application, the response amplitudes were dramatically changed for many odors and/or mice (Figures 10C,E). In general, insulin significantly decreased odor-evoked calcium responses across the group of mice recorded (Figures 10E,F, two-way ANOVA, n = 40 (10 mice with 4 odors), F_(1,72)= 19.8, P = 3.1^∗e-5]. The two-way ANOVA also indicated that the effect on different odor responses was significantly different [Figure 10E, F_(3,72)= 4.4, P = 0.007]. Further one-way ANOVA revealed that the decreased effect caused by insulin on 2-heptanone-evoked response was significantly larger than isoamyl acetate-evoked response [Figure 10G, n = 40 (10 mice with 4 odors), F_(3,36)= 4.4, P = 0.01]. Therefore, consistent with the LFP results, the main effect of insulin is to decrease the odor evoked calcium responses of pyramidal neurons.