A unique olfactory bulb microcircuit driven by neurons expressing the precursor to glucagon-like peptide 1

The presence of large numbers of local interneurons in the olfactory bulb has demonstrated an extensive local signaling process, yet the identification and purpose of olfactory microcircuits is poorly explored. Because the discrimination of odors in a complex environment is highly dependent on the tuning of information by local interneurons, we studied for the first time the role of preproglucagon (PPG) neurons in the granule cell layer of the olfactory bulb. Combining electrophysiological recordings and confocal microscopy, we discovered that the PPG neurons are a population of cells expressing the precursor of glucagon-like peptide 1 and are glutamatergic; able to modulate the firing pattern of the mitral cells (M/TCs). Optogenetic activation of PPG neurons resulted in a mixed excitation and inhibition that created a multiphasic response shaping the M/TCs firing pattern. This suggests that PPG neurons could drive neuromodulation of the olfactory output and change the synaptic map regulating olfactory coding.

To confirm the expression and define the activity of light-activated ChR2, we patch-clamped PPG neurons in the whole-cell configuration as guided by the expression of RFP (Fig. 3c,d). PPG neurons could be differentiated from granule cells (see Table 1) by their discrete biophysical properties that included -high input resistance (850 ± 150 MΩ, n = 15), greater cell capacitance (15.7 ± 0.6 pF), specific resting potential (−75.1 ± 1.6 mV), and the absence of spontaneous firing activity in current-clamp. No spontaneous action potentials were observed in the cell-attached configuration that keeps the intracellular composition unchanged. In response to the injection of current steps ranging from −100 to + 50 pA, we noticed the presence of a sag and a rebound spike, which usually depicts the presence of a hyperpolarization-activated current (I h ) (Fig. 3e). This hypothesis was verified by adding a selective blocker of I h current, ZD7288, to the bath solution, which resulted in the suppression of the sag and a reduction in the rebound spiking.
Next, we investigated the response of PPG neurons to 473 nm light activation of the ChR2, and determined the minimum duration and intensity of light exposure in order to minimize phototoxicity. Voltage-clamp recordings were performed in the presence of the Na + channel blocker, tetrodotoxin (TTX, 1 µM), to isolate ChR2 currents and avoid action-currents generated from depolarization of the neuron. The latency measured from the beginning of the light stimulation (<1 ms) and the neuron's insensitivity to TTX and synaptic blockers confirmed that the currents recorded were intrinsic to the PPG neurons and elicited by ChR2 activation (Table 1). We found that a 5 ms stimulus (Fig. 3f,f ') with the light intensity set to 75% (Fig. 3g,g') was sufficient to obtain a maximum current amplitude (118.5 ± 29.2 pA, Table 1) in most of the PPG neurons. In current-clamp mode, light exposure using these parameters evoked consistent action potentials at 5 and 10 Hz frequencies, although signs of spike adaptation were observed at higher frequencies where suprathreshold depolarizations due to ChR2 currents were observed but failed to elicit action potentials (20 Hz) (Fig. 3h). ppG neurons to M/tcs connection. The connection between PPG neurons and M/TCs was investigated by placing the recording electrode on the M/TCs, while light-stimulating the PPG neurons in the GCL using the Mightex illuminator that allows a precise spatial and temporal light illumination on the brain slice (Fig. 4a). Because M/TCs post-synaptic currents can differ depending on the internal solution 20 , we compared cesium-methane sulfonate (CsMeSO 3 )-vs. potassium gluconate (KGluc)-based internal solutions to record light-elicited, post-synaptic currents generated by PPG neuron activation. The control of holding potential (V hold ) allowed us to isolate either inhibitory-or excitatory-postsynaptic currents (IPSC, V hold = + 0 mV; EPSC, V hold = −70 mV) on the same cell (Fig. 4b,c). Interestingly, we found that a single light pulse on the PPG neurons (5 ms, 75%) was sufficient to elicit both EPSCs and IPSCs in M/TCs (Fig. 4b). As expected, currents recorded using CsMeSO 3 had a larger amplitude than those using potassium-gluconate based solution (KGluc) for both IPSCs and EPSCs (Table 1), due to the inhibition of K + currents, which allowed the conservation of currents elicited distantly from the soma. Postsynaptic events had notably longer onset latencies and time constants in CsMeSO3 than in KGluc based internal solutions. The light-evoked IPSCs had longer latencies than that of the EPSCs for both internal solutions tested (IPSCs = 11.1 ± 0.6 ms vs EPSCs = 6.7 ± 1.1 ms for CsMeSO 3 based internal solution, and IPSCs = 8.6 ± 0.7 vs EPSCs = 3.2 ± 0.3 ms for KGluc solution;  www.nature.com/scientificreports www.nature.com/scientificreports/ pA/ms in CsMeSO 3 internal solution; Table 1) when compared to IPSCs (decay time = 39.4 ± 3.8 ms, decay slope = 3.7 ± 0.7 pA/ms in CsMeSO 3 internal solution; Table 1) revealing a longer excitation than inhibition of the M/TCs. It is noteworthy that some of EPSCs recorded from M/TCs present a very slow decay time >1 s. In addition, latency jitter, i.e. the SD of the latency of the EPSCs recorded were shorter (0.182 ± 0.025 µs) than for IPSCs (1.422 ± 0.926 µs).
We then investigated the nature of the synaptic connection between PPG neurons and M/TCs by applying synaptic blockers to the bath solution ( Fig. 4d-f '). As we anticipated, EPSCs were not affected by the GABA A antagonist, gabazine (GBZ, 10 µM), but were almost abolished after AMPA/Kainate and NMDA glutamate receptor antagonists, NBQX (10 µM) and AP V (50 µM), suggesting a glutamatergic origin (Fig. 4d,d'). Surprisingly and contrarily, IPSCs were totally abolished both following GBZ application (100%, n = 7 cells; Fig. 4f,f ') and also after application of glutamate blockers (94%, n = 7 cells; Fig. 4g,g'), suggesting that the inhibitory input from PPG neurons on the M/TCs had a multi-synaptic origin. This observation was in accordance with the greater latency and jitter of IPSCs compared to that of the EPSCs (Table 1). To determine if the glutamatergic EPSCs were mono-or multi-synaptic in origin, we used a previously-described technique 21 to depolarize ChR2-positive presynaptic termini directly by blocking action potentials using TTX (1 µM) and blocking potassium channels using 4-aminopyridine (4-AP, 100 μM). TTX prevents recurrent activity whereas 4-AP is used to allow more efficient depolarization of axonal and presynaptic membranes, which thereby allows the release of presynaptic vesicles just by the action of ChR2 currents. The partial recovery of light-evoked EPSCs in the absence of any action potentials, and after application of 4-AP (Fig. 4e,e'), demonstrated that the excitatory input of PPG neurons on the M/TCs was monosynaptic. ppG neuron to Gc connection. GCs provide GABAergic inhibitory input to the M/TCs through a rapid feedforward inhibition involving dendrodendritic M/TC to GC glutamatergic stimulation, followed by a release of GABA from the GCs to the M/TCs 22 . Consequently, we investigated if the multi-synaptic IPSCs previously recorded on the M/TCs following light activation of the PPG neurons could originate from the activation of granule cells. Recordings were thus performed directly on GCs following light-stimulation of PPG neurons (Fig. 5a). In a cell-attached configuration that allows the recording of small cells without interfering with their intracellular composition, we found that a single pulse light stimulation of PPG neurons generated action currents in the GCs (Fig. 5b) indicating that this was a stimulating synaptic input to the GCs (n = 7). Recordings made in the whole-cell configuration showed that single light-pulses failed to generate IPSCs but were able to generate light evoked EPSCs in some GCs (6/13) (responses with a latency larger than 20 ms were considered spontaneous events and not light-evoked and were omitted from the analysis (outlier's test), 78.0 ± 5.7 pA, n = 6; Fig. 5c, Table 1). Similarly to the M/TCs, EPSCs were blocked after bath application of the glutamatergic blockers AP V and NBQX (Fig. 5d,d'), however, the latency of the generated EPSCs was significantly longer than those recorded in M/TCs (12.5 ± 1.6 ms in GCs vs 6.7 ± 1.1 ms in M/TCs, Table 1) and they presented an asynchronous pattern (Fig. 5c,d). This asynchronous pattern of light-evoked EPSP suggested that the activation of the GCs was the result of PPG to M/TC monosynaptic activation, which in turn triggered a M/TC-to-GC glutamate release. The latency of light-evoked EPSCs in GCs was not significantly different from the latency of the IPSCs recorded in the MCs (12.5 ± 1.6 ms vs 11.17 ± 0.58 ms, Table 1) consistent with the inhibitory response observed at the M/TC level being the result of feedforward inhibition at the MC/GC synapse as discussed below.
ppG neurons elicit complex response in M/tcs. How does the dual inhibitory/excitatory input of PPG neurons influence M/TCs? To answer this question, we performed current-clamp recordings on M/TCs while stimulating the PPG neurons (Fig. 6a). For this experiment, M/TCs were maintained at their resting potential (~ −65 mV; Table 1). Overall, 65% of M/TCs were found to respond to a single 5 ms, 75% light stimulation of the PPG neurons. Whether the non-responding cells did not receive inputs from PPG neurons, or the input stimulation was too weak to record any change in the firing frequency, or if the connections were lost during the olfactory bulb slice preparation remained unresolved. M/TCs can either show spontaneous bursting firing, or they are silent or have low basal activity 23,24 . In M/TCs that exhibit spontaneous bursting, a single light pulse elicited a triphasic response, consisting of a slight depolarization, followed by a hyperpolarization, and a subsequent rebound excitation (Fig. 6b). This response was directly in agreement with our voltage-clamp observations in Fig. 4, where on average EPSCs displayed shorter latency and slower decay times compared to IPSCs, meaning a longer excitatory input. In other cases, M/TCs had much lower spontaneous firing activity as presented in Fig. 6c. These cells received stronger inhibitory inputs from inhibitory interneurons as depicted by numerous spontaneous inhibitory post synaptic potentials (sIPSPs). M/TCs with a low basal activity responded to the light-activation of PPG neurons hyperpolarization only (Fig. 6c, top trace). Interestingly, addition of the GABA A antagonist GBZ resulted in increased spontaneous firing, due to the inhibition of inhibitory input from interneurons, a suppression of the light-evoked hyperpolarization, followed by an increase in the excitation (Fig. 6c, middle trace). Consistent with our voltage-clamp observations in Fig. 4, when we applied glutamatergic receptor blockers, the light response was completely abolished (Fig. 6c, lower trace).
Because the resting potential of MCs is close to the reversal potential for GABA A -receptors, inhibitory post synaptic potentials can be masked when recording in the current-clamp mode. To overcome this, we replaced the conventional KGluc internal solution with potassium methane sulfonate (KMeSO 3 ) that displaced the chloride reversal potential to −110 mV, lower than the resting potential of M/TCs. KMeSO 3 increased the light-induced hyperpolarization or inhibitory postsynaptic potential (IPSP) but did not reveal any excitatory inputs on M/TCs (Fig. 6d). Similarly to IPSCs, light-induced IPSPs were totally suppressed after bath application of either GBZ or AP V /NBQX (Fig. 6d). www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
The design principles and function of microcircuits are not yet well understood but serve as elementary processing units linking single neurons to global brain function and ultimately behavior 25 . The dSACs remain one of the most understudied cell types in the olfactory bulb neuronal network. Microscopic analyses have estimated a total number of around 13,500 dSACs in the rat olfactory bulb 9,10 which, compared to the MC total population (50,000), suggests that the dSACs and the microcircuit they define is substantial with regards to olfactory bulb processing. Our study provides the first evidence that terminations of some GLP-1 producing neurons in the inner part of the granule cell layer (GCL) are capable of providing a fine tuning of M/TC output by activating both excitatory glutamatergic synapses and an inhibitory GABAergic response.
We show that PPG neurons represent a unique class of interneurons by mediating monosynaptic excitation on MCs. PPG neurons are a particular subclass of olfactory bulb interneurons of the GCL 4 ; more precisely, their slightly larger soma compared to GCs, and their stellar dendrites with dendritic buttons suggest they are dSACs.  www.nature.com/scientificreports www.nature.com/scientificreports/ piriform cortex, drive a direct feedforward inhibition of GCs 27 . Burton and colleagues (2017) extensively studied the interneuron features of GL-dSACs, another class of dSACs located in the internal plexiform layer that sends their projections to the GL 28 . GL-dSACs receive convergent external tufted cell (ETC) excitation, mediate the inhibition of tufted cell apical dendrites, and deliver selective output onto interneurons and principal tufted cells. These authors recorded excitation of periglomerular cells that they attributed to an excitation by GABA on depolarized Clreversal potentials triggering GABA-induced GABA release 28 .
Some differences can be denoted between dSACs described by previous studies and the PPG neurons we described in this study. For example, dSACs characterized by Eyre and colleagues using electron microscopy indicated that dSAC terminals targeted GC dendrites but did not form direct synaptic contacts with M/TCs 9 . It must be noted that all previous studies have been performed in rats, thus the differences we observed in the present study such as the lack of expression of Kv2.1 or the muscarinic acetylcholine receptor 1α may be the result of differences between species. These data add further evidence that PPG neurons, which have a monosynaptic www.nature.com/scientificreports www.nature.com/scientificreports/ connection to M/TCs, are a different population of dSACs compared to those characterized in previous studies. It is possible that they did not sample this connection in their microscopic analysis. Other studies using paired-recordings have shown that dSACs generate spontaneous IPSCs onto GCs 29 . The fact that we did not record spontaneous firing in the PPG neurons, whether in whole-cell or cell-attached mode, further suggests that these neurons are electrophysiologically distinct from those dSAC previously reported 9,28,30 .
Interestingly, although PPG neurons exhibit a dual glutamatergic and GABAergic phenotype with the co-expression of both GAD67 and vGLUT2, we did not find evidence of a direct release of GABA from PPG neurons onto M/TCs or GCs. Such a duality is not unique in the OB, for instance in the glomerular layer, Tatti and collaborators identified a population of GAD67-positive neurons co-expressing the vesicular glutamate transporter 3 (vGLUT3), which displayed electrophysiological properties of glutamatergic neurons 31 . Alternatively spliced or truncated versions of the GAD enzyme that synthesize GABA have been reported during developmental stages, which have yet no known function 32 . Another evidence of a glutamatergic phenotype lays in the lack of expression of the vesicular GABA transporter (vGAT) in the PPG-neurons. Moreover, in other areas of the brain, as in the paraventricular (PVN) and dorsomedial (DMH) hypothalamic nuclei, PPG neurons have been shown to co-express vGLUT2 and GLP-1 at the level of their axon varicosities, suggesting that glutamate is a co-transmitter at synapses within the hypothalamus 33,34 . At the synaptic level, GLP-1R signaling has been shown to modulate AMPA/Kainate glutamate receptor activity by increasing mEPSC frequency and decreasing the paired-pulse ratio in medium spiny neurons of the nucleus accumbens and the ventral tegmental area 35,36 . While we did not observe a significant change in pair-pulse ratio recorded from MCs (data not shown); this is consistent with our previous study that showed that GLP-1 directly increases the evoked spiking activity in MCs through an inhibition of the potassium channel Kv1.3 37 . These collective findings suggest a paracrine role of GLP-1 in the modulation of the olfactory output.
In exploring the intrinsic membrane properties and light-evoked currents in the PPG-MC-GC microcircuit, we encountered some complexities in the kinetics of the responses. Our latency data show that light-activation of PPG neurons evokes glutamate-induced EPSCs in MCs that are significantly faster than the GABA-induced IPSCs in MCs (Table 1). While this would be consistent with a classical dendrodendritic feedforward release of GABA from GCs to MCs, we also noted that the EPSCs in GCs were not significantly different in latency from the IPSCs in MCs. Although one would anticipate the IPSCs in the MCs to have the greatest latency, because they are the last one in this chain of events, it may be difficult to accurately measure them. Feedforward inhibition occurs within the dendrodendritic microcircuit of the MC/GC synapses and we were not able to make direct recordings of local currents within the dendrodendritic connection. Rather whole-cell recordings at the soma of MCs and GCs, respectively, both of which are remote to the synapse, were used to record these events. Therefore, the recorded kinetics reflect when the EPSC and IPSC have travelled to the respective soma. This is dependent on the exact cell morphology and the GC EPSC might reach the GC soma later than the MC IPSC might reach the MC soma due to the different architecture of the cells, despite the EPSC at the dendrodendritic synapse happening before the MC IPSC. An additional factor in accurately resolving the kinetics of activation within the synapses of this microcircuit are the size differences between the MCs and the smaller GC and PPG neurons. Finally, we cannot exclude that fibers of other centrifugal pathways could provide further complexity and lead to generation of inhibitory responses. The observation that some M/TCs could be exclusively inhibited in response to light-evoked PPG activity suggest that further multisynaptic inhibitory routes from PPG to M/TCs in addition to the observed M/TC-GC dendrodendritic M/TC-inhibition, exist.
In mammals, the olfactory bulb is the first stage of olfactory processing and is thereby crucial to shape the olfactory output by M/TCs. Microcircuits involving interneurons in the olfactory bulb have been shown to increase the gain control, filter weak inputs, or temporally shape and synchronize M/TCs spiking patterns, resulting in a reduction of the signal-to-noise ratio of odor representation in complex environments [38][39][40] . Our current study now demonstrates that selective activation of PPG neurons results in a complex tri-phasic response in MT/Cs. This tri-phasic response is uniquely characterized by a brief activation, which establishes a transient activation of GC feedforward inhibition, followed by a rebound long-lasting excitation (Fig. 6b). In MCs not displaying spontaneous activity and receiving more inhibitory inputs, PPG neuron activation was not sufficient to depolarize the M/TCs and resulted solely in a reinforcement of lateral inhibition. Taking together, these data suggest that PPG neurons serve to increase the contrast of the pattern of activation of olfactory output, by increasing the response of the cells with high activity and increasing the inhibition of low spiking M/TCs. Alternatively, in neurons with prominent spontaneous activity, the multiphasic inhibition-rebound excitation may serve the role of temporal control of M/TC spike patterns and in shaping the spike output. Recent work has demonstrated the importance of M/TC-selective lateral inhibition in the representation of odors and the timing of spike generation 38,40,41 . In MCs, lateral inhibition is thought to be mainly mediated at the GL level by periglomerular cells in comparison to the inhibitory inputs generated at the EPL level by GCs 42 ; however, PPG neurons do not promote long-lasting inhibition but rather a brief window of inhibition/excitation allowing a reset of spiking synchronicity of the M/TCs.
At their resting potential, PPG neurons do not show any spontaneous firing activity, and a low rate of sEPSCs, and sIPSCs, hence we can hypothesize that they receive centrifugal projections from the central nervous system. Feedback from the piriform cortex has been shown to regulate global inhibition in the olfactory bulb 27,43 . Boyd and collaborators demonstrated direct excitatory projections from the pyramidal cells in the piriform cortex to the dSACs in the olfactory bulb. Our findings demonstrate a new class of glutamatergic neurons in the GCL of the olfactory bulb. In vitro, optical activation of molecularly-identified, PPG neurons results in a multiphasic response in the M/TCs. This type of time locking response ensures the ability for the PPG neurons to increase the contrast between spiking and silent M/TCs. This important function of the PPG -M/TC -GC microcircuit may refine olfactory perception or allow neuromodulation of the contrast during different nutritional states given the potential co-release of the GLP-1 hormone.

Materials and Methods
ethical approval. All experiments described in this report were approved by the Florida State University Institutional Animal Care and Use Committee (IACUC) under protocol #1733, and were conducted in accordance with the American Veterinary Medicine Association (AVMA), the National Institutes of Health (NIH), and the UK Home Office Regulations under the Scientific Procedures Act 1986. In preparation for olfactory bulb slice electrophysiology, mice were anesthetized with isoflurane (Aerrane; Baxter, Deerfield, IL, USA) using the IACUC-approved drop method and then were killed by decapitation (AVMA Guidelines on Euthanasia, June 2007).
Mouse lines and animal care. Detection of preproglucagon (PPG) neurons expressing a red fluorescent protein (RFP) was achieved by crossing Rosa26-tandem-dimer red fluorescent protein (tdRFP) reporter mice (Gt(ROSA)26Sor tm1Hjf ) 11 with mice expressing Cre recombinase under the control of the preproglucagon promoter (GLU-Cre12 mice, Tg(Gcg-icre)12Fmgb) 12 . For simplification, homozygous progeny resulting of the breeding of GLU-Cre12 and Rosa26 tdRFP mice will be referred as PPG-Cre-RFP in this manuscript. Briefly, GLU-Cre12 mice were created using a construct based on the bacterial artificial chromosome (BAC) RP23-343C17 (Children's Hospital Oakland Research Institute, Oakland, CA, USA) in which the sequence between the proglucagon start codon in exon 2 and the stop codon in exon 6 was replaced by iCre using Red/ET recombination technology 44 (Genebridges, Heidelberg, Germany). PPG-Cre-RFP homozygous offspring were back-crossed into C57BL/6 mice for at least 7 generations. Channelrhodopsin-2 (ChR2) was expressed in PPG neurons by crossing to heterozygosity PPG-Cre-RFP mice with the Ai32 line (B6;129S-Gt(ROSA)26Sor tm32(CAG-COP4*H134R/ EYFP)Hze /J, Stock# 012569, RRID:IMSR_JAX:012569, Jackson Laboratories, Bar Harbor, ME, USA) that contains a floxed allele expressing the fusion protein ChR2(H134R)-EYFP in the presence of the CRE recombinase in PPG neurons 18 . Heterozygote Swiss Webster GAD67-EGFP knock-in mice (CB6-Tg(Gad1-EGFP)G42Zjh/J, RRID:IMSR_JAX:007677) 45 were a generous gift from Dr. Pradeep Bhide (College of Medicine, The Florida State University) and were bred to PPG-Cre-RPF mice to analyze GAD67 expression in the PPG neurons in the GCL. Given that GAD67 mice are kept as heterozygotes, we found an expression of the GFP only in approximately half of the mice resulting from the crossing (4/9 mice).
All mice were housed at the Florida State University vivarium in accordance with the institutional requirements for animal care. All mice used in this study (Mus musculus, C57BL/6, 129 S, and Swiss Webster background strains) were maintained on a standard 12 h/12 h light/dark cycle and were allowed ad libitum access to 5001 Purina Chow (Purina, Richmond, VA, USA) and water. Mice were not treated with different conditions or derived of different experimental genotypes so no blinding was performed to the investigator in assigning mice to studies for electrophysiology or immunocytochemistry. A total number of 55 crosses were established to generate mice for our combined experiments and progeny from these breedings were randomly assigned to a given experiment. Mice were never used for multiple experiments due to different preparation needs between electrophysiology and immunocytochemistry. Mice were of both sexes and ranged from postnatal day 21 to 35 for electrophysiology and 1-2 months for immunocytochemistry. Because our vivarium is set for reverse light phase, mice were sacrificed for experiments between 3-5 hours (h) into the dark phase.
A total of 144 mice were used in our study. A total of 15 mice were excluded from the study that did not have strong ChR2 expression in the PPG-Cre-RFP x Ai32 progeny. A total of 16 mice were excluded from histological analyses that were used to optimize antigenicity in piloting the project.
In some cases, the red fluorescence from the Rosa26tdRFP in PPG neurons was amplified using a rabbit DsRed polyclonal antibody (1:1000, cat#632496, RRID:AB_10013483, Clontech Laboratories Inc., Mountain View, CA, USA). Host-specific Cy3, Cy2, or Alexa 647 secondary antisera were purchased from Jackson ImmunoResearch (West Grove, PA, USA), used per manufacturer's suggested protocols, and applied at a dilution of 1:400 in blocking solution.
Surgeries. To comprehensively examine the phenotype of the GLU-Cre-RFP mice, injections of Cre-dependent adeno-associated virus were made into the GCL (relative to bregma, 0.45, ± 0.75, 0.23 mm). Mice were initially anesthetized in an induction chamber primed with isoflurane and subsequently mounted to the stereotaxic apparatus (Stoelting Co., IL) where they were maintained under anesthetic state [2% isoflurane gas mixed with oxygen (0.5 L/min)]. AAV5 containing pAAV-EF1a-double floxed-hChR2(H134R)-EYFP-WPRE-HGHpA 1 × 10¹³ vg/mL (Addgene, Watertown, MA) were delivered bilaterally using glass 10 µL Hamilton syringes (O.D. ~15-20 μm) (1.5 µL per olfactory bulb) at a rate of 100 nL/min. The analgesic buprenorphine was administered immediately following surgery and at 12 h intervals for one day following surgery (0.05 mg/kg). Mice were sacrificed 15 days post-surgery for anatomical analyses. olfactory bulb slice electrophysiology. A total of 47 mice were used for 165 patch-clamp electrophysiological recordings from olfactory bulb slices. Following isoflurane anesthesia, mice were killed by decapitation (see Ethics Approval). Olfactory bulbs were dissected and coronal sections (275-300 μm) were prepared as previously described 4,47 . The blue light source (Polygon 400, Mightex, Pleasanton, CA, USA) used to stimulate channel rhodopsin at 470 nm was mounted on the side of the microscope to a dichroic mirror located between the binoculars and the objective, allowing the light to be directed toward the slice through the 40 × Zeiss objective. The spatial illumination pattern on the slice was controlled by Dynamic Spatial Illuminator software (Mightex, version 2.0.0) in conjunction with a Zeiss Axiocam digital camera and associated AxioVision software (version 4.8, Carl Zeiss Microimaging, RRID:SCR_002677). The duration and intensity of the light stimuli were set in BioLED source control module software (Mightex, version 1.0.2). The light stimulus was controlled in pClamp 10.5 software (Axon Instruments/Molecular Devices, Sunnyvale, CA, USA) using a digital output channel to connect the digitizer (Digidata 1440 A, Axon Instruments/Molecular Devices) to the BioLED controller.
Membrane voltage and current recordings were generated using pClamp 10.5 software in conjunction with an Axopatch 200B amplifier (Axon Instruments/Molecular Devices, RRID:SCR_011323). The analog signal was filtered at 2 kHz and minimally digitally sampled every 50 μs (20 kHz). Electrodes were fabricated from borosilicate glass (Hilgenberg no. 1405002, Malsfeld, Germany) to a diameter of approximately 2 μm to yield pipette resistances ranging from 4 to 7 MΩ. Positive pressure was retained while navigating through the olfactory bulb laminae until a high resistance seal (Re = 2-10 GΩ) was obtained on a cell in the slice 4,49 . The morphology and biophysical properties of the neurons were used to distinguish M/TCs from GCs 1 . All neurons within the MC layer were designed as M/TCs and were not distinguished from tufted cells. Neurons within the GCL were classified as either GCs or PPG neurons. PPG neurons were identified by excitation of the RFP. We previously biophysically characterized these neurons  as a subclass of neurons in the GCL called deep short-axon cells (dSACs), also named Cajal cells, Blanes cells, or Golgi cells, as previously described by Eyre and collaborators 1-3,9,10 . Cell-attached recordings were made when the resistance was above 1 GΩ with pipettes filled with intracellular solution. The whole-cell configuration was established by applying gentle suction to the lumen of the pipette while monitoring resistance. The pipette capacitance was electrically compensated through the capacitance neutralization circuit of the Axopatch 200B amplifier. Likewise, series resistance compensation was used to electrically reduce the effect of pipette resistance.
Each neuron was sampled for adequate resting potential (less than -55 mV) and proper series resistance (less than 40 MΩ) right after rupturing the patch prior to initiating a series of current-clamp recordings. Voltage-clamp traces were subtracted linearly for leakage conductance. Resting membrane potentials were corrected for a -14 mV junction potential offset. In voltage-clamp, excitatory postsynaptic currents (EPSCs, V h = -70 mV) and inhibitory postsynaptic currents (IPSCs, V h = 0 mV) were recorded in multiple epochs of 1-5 s. The inter-stimulus interval was always set to a minimum of 20-30 s to allow for a full recovery of the activated channels. (2019) 9:15542 | https://doi.org/10.1038/s41598-019-51880-9 www.nature.com/scientificreports www.nature.com/scientificreports/ MA, USA, RRID:SCR_002815), and Igor Pro 6.12 A (WaveMetrics Inc., Portland, OR, RRID:SCR_000325) software with the NeuroMatic version 2 plug-in (written by Jason Rothman, RRID:SCR_004186). Rise and decay time constants of postsynaptic currents were analyzed starting from the middle of the light pulse between 10 and 90% of the event maximal amplitude. All electrophysiological data are available by contacting the D. Fadool Laboratory and are supplied in pClamp v10 acquisition compatible files (*.abf).

Data availability
All data generated or analyzed during this study are included in this published article. Transgenic mice lines were generated by professors Gribble, and Reimann and can be made available through material transfer agreements (MTA) arranged with these authors. The communicating author is happy to arrange shipment of mice to individuals in the USA following completion of the MTA while other co-authors will arrange shipment to other locations.