Layer- and cell type-selective co-transmission by a basal forebrain cholinergic projection to the olfactory bulb

Abstract

Cholinergic neurons in the basal forebrain project heavily to the main olfactory bulb, the first processing station in the olfactory pathway. The projections innervate multiple layers of the main olfactory bulb and strongly influence odor discrimination, detection, and learning. The precise underlying circuitry of this cholinergic input to the main olfactory bulb remains unclear, however. Here, we identify a specific basal forebrain cholinergic projection that innervates select neurons concentrated in the internal plexiform layer of the main olfactory bulb. Optogenetic activation of this projection elicits monosynaptic nicotinic and GABAergic currents in glomerular layer-projecting interneurons. Additionally, we show that the projection co-expresses markers for GABAergic neurotransmission. The data thus implicate neurotransmitter co-transmission in the basal forebrain regulation of this inhibitory olfactory microcircuit.Cholinergic neurons innervate multiple layers in the main olfactory bulb but the precise circuitry of this input is not known. Here the authors show that VGLUT3⁺ cholinergic neurons selectively innervate deep short axon cells in specific layers and elicit robust monosynaptic GABAergic and nicotinic postsynaptic currents.

Fig. 1

HDB cholinergic neurons express tdTomato in VGLUT3^Cre-tdTomato mice. a–c HDB from a P56 VGLUT3^Cre-tdTomato mouse shows considerable overlap between VAChT (cyan) and the tdTomato reporter (yellow). Approximately 46% of VAChT⁺ cells are positive for tdTomato and ~48% of tdTomato⁺ cells express VAChT (108 of 237 cells and 59 of 121 cells, respectively, n = 3 animals). Arrows denote colocalization; arrowheads denote tdTomato without VAChT. Scale bar = a–c 100 µm, (c inset) 20 µm

Fig. 2

VGLUT3⁺ neurons in the HDB preferentially innervate the IPL of the MOB. a Approximately 26% (61 of 238 cells, n = 6 animals) of the HDB neurons that express VAChT (cyan) were infected following virus injection (yellow) in VGLUT3^Cre mice. b, c Injection of ChR2-EYFP virus into the HDB of VGLUT3^Cre mice labels axons preferentially innervating the IPL. Representative images from two mice are shown. d Approximately 56% (47 of 84 cells, n = 2 animals) of the HDB neurons that express VAChT (cyan) were infected following ChR2-EYFP virus injection (yellow) into ChAT^Cre mice. e ChR2-EYFP virus injected into the HDB of ChAT^Cre mice labels axons densely innervating both the GL and IPL. Given that ~46% of cholinergic neurons are VGLUT3⁺ (Fig. 1), the numbers indicate that ~56% of neurons positive for both VGLUT3 and VAChT were infected with the virus, consistent with what is observed in the ChAT^Cre animals. f Quantification of ChR2-EYFP-labeled axons in the two mouse lines shows a peak in the IPL in VGLUT3^Cre mice (yellow) and peaks in the GL and IPL in ChAT^Cre mice (cyan). Scale bars = 100 µm

Fig. 3

VGLUT3⁺ HDB projections to the IPL express VAChT. a ChR2-EYFP-labeled projections to the IPL (green) in a HDB-injected VGLUT3^Cre mouse colocalize with VAChT (blue). b VGLUT3^Cre-tdTomato mouse shows colocalization of VAChT (blue) and tdTomato (red) in the IPL of the MOB. Scale bars = a, b 10 µm; b inset 5 µm

Fig 4

HDB projections to the IPL mediate mixed nicotinic-GABAergic postsynaptic currents in dSACs. a MOB circuitry. b Experimental design. c, d Representative GL-dSAC recording (arrowhead: axon; inset: axon in GL; scale bar: 20 μm) from a VGLUT3^Cre mouse. Optogenetic stimulation evoked EPSCs and IPSCs at negative and positive holding potentials, respectively. NBQX/AP5/Gbz modestly reduced light-evoked EPSCs. Subsequent mecamylamine (Mec) abolished the residual EPSC. Mean response shown; individual trials in lighter color. e, f Same as c, d for a ?-dSAC of a ChAT^Cre mouse. NBQX/AP5 modestly reduced the light-evoked EPSC. Subsequent Mec strongly reduced the residual EPSC, but not IPSC. Subsequent Gbz abolished the residual IPSC (and unclamped component at V _hold = –60 mV). g Summary of EPSC pharmacology. In ChAT^Cre mice (left), Mec reduced EPSC amplitudes to 19.4 ± 8.6% of control values (n = 6; p = 2.9 × 10⁻⁶, two-tailed paired t-test). In VGLUT3^Cre mice (middle), NBQX/AP5/Gbz reduced EPSC amplitudes to 76.2 ± 20.5% (n = 6); subsequent Mec reduced EPSC amplitudes to 15.8 ± 11.8% (n = 5) (p = 7.4 × 10⁻⁸, one-way ANOVA; control vs. NBQX/AP5/Gbz, p = 0.015, control vs. Mec, p = 5.4 × 10⁻⁸, NBQX/AP5/Gbz vs. Mec, p = 5.6 × 10⁻⁶, post hoc Tukey-Kramer). In ChAT^Cre mice (right), NBQX/AP5 insignificantly reduced EPSC amplitudes (p = 0.38), while subsequent Mec and Gbz reduced EPSC amplitudes to 33.0 ± 22.4% (n = 4) and ~ 7.7% (n = 2), respectively (p = 1.2 × 10⁻⁵, one-way ANOVA; control vs. Mec, p = 1.1 × 10⁻⁴, control vs. Gbz, p = 3.9 × 10⁻⁵, NBQX/AP5 vs. Mec, p = 8.7 × 10⁻⁴, NBQX/AP5 vs. Gbz, p = 1.8 × 10⁻⁴, post hoc Tukey-Kramer). h Peak EPSC and IPSC amplitudes. Closed symbols (and error bars): median (and first/third quantiles). EPSC and IPSC amplitudes did not differ between the two mouse lines (EPSC: 57.0 ± 42.7 [n = 13] vs. 69.7 ± 83.4 pA [n = 6], ChAT^Cre vs. VGLUT3^Cre, p = 0.90, two-sided Wilcoxon rank-sum test; IPSC: 65.1 ± 103.3 [n = 8] vs. 124.5 ± 134.3 pA [n = 6], ChAT^Cre vs. VGLUT3^Cre, p = 0.18, two-sided Wilcoxon rank-sum test). ChAT^Cre and VGLUT3^Cre data are therefore pooled for subsequent analyses. i Mean EPSC (n = 18) and IPSC (n = 14) waveforms. Inset: expanded timescale. j EPSC and IPSC latencies did not differ (7.5 ± 1.4 vs. 7.8 ± 1.6 ms; EPSC vs. IPSC; n = 10; p = 0.39, two-tailed paired t-test)

Fig. 5

HDB projections to the IPL innervate GL-dSACs. a–m Post hoc Neurobiotin brightfield image (scale bar: 100 μm), morphological reconstruction, and mean current (V _hold = −60 mV) of a subset of recorded MOB cells exhibiting significant a–i or non-significant j–m responses to brief optogenetic activation (blue line) of BF projections in ChAT^Cre or VGLUT3^Cre mice. Gray lines in reconstructions denote GL/EPL, EPL/MCL, MCL/IPL, and IPL/GCL borders. A subset of cells b, e were recorded with a K⁺-based intracellular solution. A, anterior; D, dorsal. n Mean light-evoked responses (V _hold = −60 mV) across GL-dSACs (n = 6), ?-dSACs (n = 13), GCL-dSACs (n = 3), GCs (n = 9), MCs (n = 11), and TCs (n = 3). o Total number of different MOB cell types exhibiting a significant response (V _hold = −60 mV) to optogenetic activation of HDB projections

Fig. 6

Colocalization of markers for acetylcholine and GABA in the HDB and MOB. a HDB from VGLUT3^Cre-tdTomato mouse shows tdTomato mRNA (green) partially colocalizes with VGAT mRNA (red, arrows). Not all tdTomato colocalizes with VGAT (arrowheads). b HDB from VGLUT3^Cre-tdTomato mouse shows VAChT mRNA (green) partially colocalizes with VGAT mRNA (red, arrows). Not all VAChT colocalizes with VGAT mRNA (arrowheads). c HDB from GAD2^Cre-GCAMP6f mouse shows colocalization of ChAT (red) and GCAMP6f (green, arrows). Arrowhead marks a ChAT⁺ neuron not colocalized with GCaMP6f. d MOB from a VGLUT3^Cre-tdTomato mouse shows colocalization of VGAT (green) and VAChT (blue, arrows). Inset from d is the same as in Fig. 3b but also with VGAT (green). tdTomato, VGAT, and VAChT colocalize (arrows). e MOB from VGLUT3^Cre mouse injected with ChR2-EYFP virus in the HDB shows colocalization of VGAT (red) and VAChT (blue, arrows). VGAT and VAChT colocalize with a ChR2-EYFP-labeled axon (green, arrowhead). Scale bars = a–c 50 µm, d 20 µm, d inset 5 µm, e 10 µm, e inset 10 µm