Characterization of basal forebrain glutamate neurons suggests a role in control of arousal and avoidance behavior

Abstract

The basal forebrain (BF) is involved in arousal, attention, and reward processing but the role of individual BF neuronal subtypes is still being uncovered. Glutamatergic neurons are the least well-understood of the three main BF neurotransmitter phenotypes. Here we analyzed the distribution, size, calcium-binding protein content and projections of the major group of BF glutamatergic neurons expressing the vesicular glutamate transporter subtype 2 (vGluT2) and tested the functional effect of activating them. Mice expressing Cre recombinase under control of the vGluT2 promoter were crossed with a reporter strain expressing the red fluorescent protein, tdTomato, to generate vGluT2-cre-tdTomato mice. Immunohistochemical staining for choline acetyltransferase and a cross with mice expressing green fluorescent protein selectively in GABAergic neurons confirmed cholinergic, GABAergic and vGluT2+ neurons represent distinct BF subpopulations. Subsets of BF vGluT2+ neurons expressed the calcium binding proteins calbindin or calretinin, suggesting multiple subtypes of BF vGluT2+ neurons exist. Anterograde tracing using adeno-associated viral vectors expressing channelrhodopsin2-enhanced yellow fluorescent fusion proteins revealed major projections of BF vGluT2+ neurons to neighboring BF cholinergic and parvalbumin neurons, as well as to extra-BF areas involved in the control of arousal or aversive/rewarding behavior such as the lateral habenula and ventral tegmental area. Optogenetic activation of BF vGluT2 neurons elicited a striking avoidance of the area where stimulation was given, whereas stimulation of BF parvalbumin or cholinergic neurons did not. Together with previous optogenetic findings suggesting an arousal-promoting role, our findings suggest BF vGluT2 neurons play a dual role in promoting wakefulness and avoidance behavior.

INTRODUCTION

The basal forebrain (BF) is an important brain region involved in the control of sleep-wake behavior, attention and reward processes (Detari et al., 1999; Zaborszky and Duque, 2003; Brown et al., 2012; Brown and McKenna, 2015; Lin et al., 2015). The BF contains three main populations of neurons utilizing the neurotransmitters acetylcholine, GABA and glutamate (Gritti et al., 1993, 2006; Hur and Zaborszky, 2005). Of these three neuronal phenotypes, BF glutamatergic neurons are the least well-understood since only recently have tools become available to study these neurons.

The identification of vesicular glutamate transporters (vGluT) as selective markers for glutamate neurons allowed identification of glutamatergic somata in BF and throughout the brain using in situ hybridization (Herzog et al., 2001; Fremeau et al., 2004; Hur and Zaborszky, 2005; El Mestikawy et al., 2011); as well as BF glutamatergic terminals in projection regions using immunohistochemistry (Henny and Jones, 2006; 2008). vGluT1 and vGluT3 are present at only low levels in BF, with vGluT3 located in a subset of cholinergic neurons which lack the p75 neurotrophin receptor and project to the amygdala (Nickerson Poulin et al., 2006). Thus, vGluT2 is the most prevalent isoform within the BF (Fremeau et al., 2001; Herzog et al., 2001; Hur and Zaborszky, 2005). The development of vGluT2-cre mice (Vong et al., 2011) allowed selective targeting and manipulation of these neurons using double-floxed adeno-associated viral (AAV) vectors (Anaclet et al., 2015; Xu et al., 2015; Peng et al., 2020). Furthermore, crossing vGluT2-cre mice with a cre-dependent reporter strain expressing the red fluorescent protein, tdTomato, facilitated identification of vGluT2 neurons for anatomical and in vitro electrophysiological studies (McKenna et al., 2015a, b, 2016; Xu et al., 2015; Yang et al., 2017). Here we used vGluT2-cre-tdTomato mice to precisely characterize the distribution, calcium-binding protein content and projections of BF vGluT2+ neurons. Major subpopulations of BF vGluT2 neurons contained the calcium-binding proteins calbindin or calretinin, suggesting functionally distinct subpopulations.

Analysis of the intra- and extra-BF projections of BF vGluT2+ neurons gives helpful clues to aid us in understanding their functional role. In vitro studies in mice suggested excitatory intra-BF connections to PV and somatostatin-containing GABAergic neurons as well as to cholinergic neurons (Xu et al., 2015). Both retrograde (Hur and Zaborszky, 2005) and anterograde studies (Henny and Jones, 2006; 2008; Do et al., 2016; Agostinelli et al., 2017, 2019) in the rodent suggested that subsets of BF vGluT2+ neurons are projection neurons which target the frontal cortex, lateral hypothalamus and other brain areas. Our anterograde tracing experiments demonstrated projections to brain areas involved in the control of arousal and (negative) reward processing. We therefore directly evaluated the role of BF vGluT2 neurons in place preference/avoidance, and report that optogenetic stimulation experiments in vivo revealed a dramatic avoidance of the area where stimulation was given. Thus, BF vGluT2 neurons play a dual role in promoting wakefulness and avoidance behavior. Early descriptions of these results were presented previously in abstract form (McKenna et al., 2015a, b, 2016).

METHODS

Animals

vGluT2-cre (strain 016963 or congenic strain 028863), PV-cre (congenic strain 017230), ChAT-cre mice (congenic strain 028861) and cre-reporter mice (007905) were purchased from Jackson Labs (Bar Harbor, ME, USA). To identify BF vGluT2+ (glutamatergic) neurons, we crossed vGluT2-cre mice with a Cre-reporter strain expressing a red fluorescent marker (tdTomato; Strain 007905; Jackson Laboratory) to generate vGluT2-cre-tdTomato mice. The selective expression of Cre-driven reporters in vGluT2 neurons in vGluT2-cre mice has been validated in many brain areas (Krenzer et al., 2011; Vong et al., 2011), including BF (Anaclet et al., 2015). To identify BF GABAergic neurons, we used heterozygous GAD67-GFP knock-in mice (Tamamaki et al., 2003; Brown et al., 2008; McKenna et al., 2013) from our in-house colony. Male, heterozygous GAD67-GFP mice on a Swiss-Webster background were crossed with female wild-type Swiss-Webster mice. To test if there was co-localization of tdTomato in GABAergic neurons, we crossed vGluT2-tdTomato animals with GAD67-GFP animals. In these mice, both glutamatergic (red) and GABAergic (green) neurons endogenously fluoresce. Both male and female animals were used for experiments. We did not observe any obvious sex differences in our anatomical or behavioral results. Thus, results from male and female animals were pooled.

Mice were housed under constant temperature and a 12:12 light:dark cycle (7AM:7PM), with food and water available ad libitum. All experiments conformed to U.S. Veterans Administration, Harvard University, and U.S. National Institutes of Health guidelines on the ethical use of animals. All measures were taken to minimize the number of animals used and their suffering and were carried out in accordance with the National Institute of Health Guide for the Care and Use of Laboratory Animals (NIH Publications No. 80–23). Experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the VA Boston Healthcare System.

Target area within the BF

Our target area in the BF for both neuronal characterization and anterograde tracing experiments was the same as in our previous studies (McKenna et al., 2013; Yang et al., 2014). We focused on intermediate areas of BF (substantia innominata (SI), horizontal limb of the diagonal band (HDB), magnocellular preoptic nucleus (MCPO) and ventral pallidum (VP)), where previous studies in the rat (Rye et al., 1984; Gritti et al., 2003; Hur and Zaborszky, 2005; Henny and Jones, 2008) and mouse (Anaclet et al., 2015; Do et al., 2016) found neurons projecting to the neocortex, approximately centered in BF at AP +0.14 mm; ML 1.6 mm; and DV −5.3mm. Our analysis did not include the rostral aspect of BF (medial septum, vertical limb of the diagonal band (MS/DBV)), containing neurons largely projecting to the hippocampus, since previous studies have studied the anatomy and physiology of glutamatergic neurons in this region in detail (Manseau et al., 2005; Huh et al., 2010; Fuhrmann et al., 2015; Leao et al., 2015) or the most caudal aspect of the cholinergic BF which extends into and intermingles with neurons in the globus pallidus and lateral hypothalamus.

General methods for anatomical studies

Mice were deeply anesthetized with sodium pentobarbital (50 mg/ml), exsanguinated with saline and perfused transcardially with a solution of 10 % buffered formalin. Brains were post-fixed for 1 day in 10% formalin, and then transferred to a 30% sucrose solution for 24 hrs at 4ºC. Tissue was cut at a thickness of 40 μm on a freezing microtome and collected into four wells of PBS. Following all immunohistochemistry procedures, tissue was washed and mounted onto chrome-alum gelatin-coated slides, dried, and coverslipped using Vectashield hard set mounting medium (H-1400; Vector Laboratories, Inc., Burlingame, CA).

Immunohistochemistry methods

Prior to all immunohistochemical stains coronal sections from one well were washed with PBS and placed in a blocking solution for one hour (0.5% TX-100 in PBS + 3% normal donkey serum (NDS)). Tissue was twice rinsed with phosphate buffered saline (5 mins per rinse) in between primary and secondary antibody incubations. To identify cholinergic neurons tissue was incubated in rabbit anti-choline acetyltransferase (ChAT, the synthetic enzyme for acetylcholine) for three days at 4ºC (1:150; Cat#AB143, EMD Millipore, Billerica, MA) followed by incubation in a secondary antibody coupled to a blue (donkey anti-rabbit AF350, Cat#A10039, Thermo Fisher Scientific), green (donkey anti-rabbit AF488, Cat#A21206, Thermo Fisher Scientific, Cambridge, MA), or red (donkey anti-rabbit AF594, Cat#A21207, Thermo Fisher Scientific) fluorophore for 3 hrs at room temperature (1:200; RT, ~24.4ºC). To identify calbindin (Calb) neurons, tissue was incubated for 2 days at 4°C in goat anti-calbindin (1:200; Cat#SC-7691; Santa Cruz Biotechnology, Dallas, Texas), followed by 2.5 hrs incubation at RT of secondary antibody coupled to a green fluorophore (1:200; donkey anti-goat AF488, Cat#A11055, Thermo Fisher Scientific). To identify calretinin (Calr) neurons, tissue was incubated for 2 days at 4°C in goat anti-calretinin (1:200; Cat#AB1550, EMD Millipore), followed by 2.5 hrs incubation (RT) of secondary antibody coupled to a green fluorophore (1:200; donkey anti-goat AF488, Cat#A11055, Thermo Fisher Scientific). To identify parvalbumin (PV) neurons, tissue was incubated in sheep anti-PV overnight at 4ºC (1:150; Cat#AF5058, RnD Systems, Minneapolis, Mn) followed by a secondary antibody (1:200) coupled to a green (donkey anti-sheep AF488 Cat#A11015, Thermo Fisher Scientific) or red fluorophore (donkey anti-sheep AF594, Cat#A11016, Thermo Fisher Scientific) for 3 hrs at RT. To amplify EYFP signal for anterograde tracing using AAV5-DIO-ChR2-EYFP, tissue was incubated in mouse polyclonal anti-GFP primary antibody (1:300; Cat#MAB3580, EMD Millipore) for 3 days at 4°C, and secondary antibody (1:500; donkey anti-mouse AF488, Cat#A21202, Thermo Fisher Scientific) overnight (16 hrs) at 4ºC. For depiction of intra-BF projections (Fig. 6), tissue was then processed for either ChAT or PV labeling (red chromophores, as described above).

Validation of primary antibodies used for immunostaining

All primary (Table 1) and secondary (Table 2) antibodies used here have been previously validated and used in peer reviewed publications. Regions expressing ChAT immunoreactivity were similar to those previously reported, including the medial septum/vertical limb of the diagonal band, BF subnuclei as investigated here, striatum, and the pontine tegmental region (Rye et al., 1984; Semba and Fibiger, 1992; Gritti et al., 1993). GFP in GAD67-GFP knock-in mice was expressed throughout many brain regions, similar to that previously reported for GABA/GAD67 immunostaining results, including the cortex, preoptic regions, BF, reticular nucleus of the thalamus, and numerous brainstem regions (Gritti et al., 1993, 2003, 2006; Brown et al., 2008; McKenna et al., 2013). The distribution of Calb and Calr staining was similar to that previously published in the rat and mouse, including the cortex, hippocampus, preoptic regions, BF, and thalamus (Jacobowitz and Winsky, 1991; Rogers and Resibois, 1992; Hof et al., 1999; Zaborszky et al., 1999; Gritti et al., 2003). To the best of our knowledge, none of the antibodies used have been evaluated against specific antigen knock-out or knock-down animals. The ChAT primary antibody has been used to confirm neurotoxic lesion of BF cholinergic cell populations using saporin targeting of the p75 receptor (e.g., Sotthibundhu et al., 2008; Cutuli et al., 2013; Kim et al., 2015). Secondary antibody-only controls were also performed, omitting the above-listed primary antibodies.

Table 1.

Primary antibody characterization.

Antigen	Immunizing Specificity	Host Species	Source/Catalog Number	Dilution	References (Mouse Brain Tissue)	Antibody Registry
Choline acetyltransferase (ChAT)	68-kDa band on western blot assays Source: Manufacturer; Bruce et al., 1995	Rabbit	EMD Millipore AB143	1:150	Hamamoto et al., 2016 Pabst et al., 2016 Chens et al., 2019	AB_2079760
Calbindin (Calb)	28-kDa band on western blot assays Source: Manufacturer	Goat	Santa Cruz SC-7691	1:200	Shin et al., 2018 Ch’ng et al., 2019 Saito et al., 2019	AB_634520
Calretinin (Calr)	29-kDa band on western blot assays Source: Manufacturer; Winsky et al., 1996	Goat	EMD Millipore AB1550	1:200	Suzuki et al., 2015 Srinivasan and Stevens, 2017 Hardv et al., 2018	AB_90764
Parvalbumin (PV)	12-kDa band on western blot assays Source: Manufacturer	Sheep	RnD AF5058	1:150	Iijima et al., 2014 Usoskin et al., 2015 Stephany et al., 2016	AB_2173907
Green Fluorescent Protein (GFP)	27-kDa monomer of 238 amino acids No detectable cross-reactivity with RFP Source: Manufacturer	Mouse	EMD Millipore MAB3580	1:250	Helgager et al., 2013 Jo et al., 2018 Thankachan et al., 2019	AB_94936

Table 2.

Secondary antibody characterization.

Secondary antibody	Vendor	Host Species	Catalog Number	Dilution	Incubation
Donkey anti-goat IgG AF488	Thermo Fisher	Donkey	A11055	1:200	2.5 hr RT
Donkey anti-mouse IgG AF488	Thermo Fisher	Donkey	A21202	1:100	16 hrs 4°C
Donkey anti-rabbit IgG AF350	Thermo Fisher	Donkey	A10039	1:200	3 hrs RT
Donkey anti-rabbit IgG AF488	Thermo Fisher	Donkey	A21206	1:200	3 hrs RT
Donkey anti-rabbit IgG AF594	Thermo Fisher	Donkey	A21207	1:200	3 hrs RT
Donkey anti-sheep IgG AF488	Thermo Fisher	Donkey	A11015	1:200	3 hrs RT
Donkey anti-sheep IgG AF594	Thermo Fisher	Donkey	A11016	1:200	3 hrs RT

AAV5-DIO-ChR2-EYFP injection for anterograde tracing of BF glutamatergic projections.

Viral vectors.

Viral vectors encoding Channelrhodopsin-2 (ChR2) and enhanced yellow fluorescent fusion proteins (AAV5-DIO-ChR2-EYFP) originally generated by the laboratory of Dr. Karl Deisseroth (Stanford University, CA, USA) were purchased from the University of North Carolina vector core facility.

Stereotaxic surgery and viral injection.

Mice were deeply anesthetized with isoflurane (1–3%) and viral injections were performed using a 1 μl Hamilton syringe (Cat#7001KH, Point Style 3, Hamilton Company, Reno, NV, USA), targeting BF (AP +0.14 mm; ML 1.6 mm; DV - 5.3mm). 250 nl of viral vector was injected at a flow rate of 25 nl/min. The needle was left in place for an additional 10 mins to allow virus diffusion in the brain and to avoid backflow along the needle tract. Once viral injections were complete, the needle was removed. Animals were sacrificed following one month of survival post-injection.

Chromophore co-localization quantification/photographic depiction

For each immunolabeling stain (ChAT, PV, Calb, Calr) and vGluT2-Tomato/GAD67-GFP crossed animals, representative BF sections [3 coronal section levels per animal (+0.38 mm (Rostral), +.014 mm (Medial), and −0.10 mm (Caudal) from Bregma)] were quantified using Neurolucida software (Version 8; Microbrightfield, Williston, VT) and a Zeiss Imager.M2 microscope. The perimeter, brain areas, and landmarks of BF were first traced at a low magnification (10X), and neuronal location was then plotted at higher magnification (20X). Neuronal markers were then overlayed onto appropriate schematic templates (Franklin and Paxinos, 2008) and replotted using Adobe Illustrator. The distribution of labeled neurons was determined using a mouse brain atlas (Franklin and Paxinos, 2008). vGluT2-tdTomato neurons were identified by the presence of red (excitation/emission at 590:617 nm) fluorescence in the cytoplasm and nucleus. Neurons with red, blue (excitation:emission 346:442 nm), or green (excitation:emission 488:509 nm) fluorescence were identified by fluorescent signal in the cytoplasm (ChAT, Calb, Calr, or PV ) or the cytoplasm and nucleus (green, GAD67-GFP). In a subset of experiments, ChAT or PV labeled neurons were identified by the presence of red fluorescence (excitation/emission at 590:617 nm) in the cytoplasm. EYFP fluorescence in mice injected with AAV5-DIO-ChR2-EYFP was amplified with anti-GFP antibody and secondary antibody conjugated to a green fluorophore. BF vGluT2 efferent regions (AAV-EYFP anterograde projections) were defined where varicosities of AAV-EYFP-labeled axons were located (see Fig. 5c for a representative example). Long-axis measurement: The long-axis diameter of labeled neurons was measured for each of the 4 BF subnuclei using Neurolucida software. Up to 10 neurons per BF subnuclei (VP, SI, HDB, MCPO) were analyzed, bilaterally, for the 3 representative sections (rostral, intermediate, and caudal) in each animal.

Digital images of fluorescently labeled neurons were captured using a Zeiss Imager.M2 microscope. Co-localization of vGluT2 with Calb or Calr (Figs. 3 and 4) was depicted using 3-D Z-stack photography employing the Zeiss Imager.M2 microscope Apotome2 Structured Illumination system. Confocal photography was performed using a Leica True Confocal Scanning SP8 FALCON system, to depict BF vGluT2 fibers containing ChR2-EYFP apposed to BF cholinergic or PV+ neurons, suggesting intra-BF synaptic contact (Fig. 6). Extra-BF anterograde fiber labeling following AAV5-DIO-ChR2-EYFP injection was evaluated using the Zeiss Imager.M2 microscope.

Statistical Analysis for anatomical experiments

All neuroanatomical density, long-axis diameter, and co-localization measures were analyzed using two-way analysis of variance (ANOVA), to compare differences between BF subnuclei (VP, SI, HDB, MCPO), as well as between the three coronal BF section representations (rostral, medial, and caudal). If a main effect of the ANOVA was significant, a pair-wise comparison between groups was then made, using Tukey’s HSD correction. Statistical analysis utilized JMP software (release Pro 14.2.0), and differences determined to be significant when p < 0.05.

Real-Time Place Preference/Avoidance and Optogenetics

Adult vGluT2-cre, PV-cre or ChAT-cre mice received bilateral injections of 0.7 μl AAV5-DIO-ChR2-EYFP into the BF followed by bilateral implantation of fiber-optic cannulae (200μm, 0.22 NA optical fiber, Doric Lenses at AP 0.0, ML ±1.6, DV −5.4). Experiments began at least 3 weeks after surgery. To test whether optical stimulation affected the preference of the mice for the side of the chamber where optical stimulation was delivered, we used a 3-compartment open chamber (46.5 cm L × 12.7 cm W × 12.5 cm H) consisting of left and right sides (each 17.4 cm L × 12.5 cm H) with different floor patterns and wall colors and a center compartment (11.7 cm L × 12.5 cm H)(Med Associates, Saint Albans, VT). A camera was placed above the chamber to provide real-time tracking of the position of the mouse in the chamber and to record each trial for post-hoc analysis using Ethovision XT 14 software (Noldus, Leesburg, VA). Mice were tethered with optical fibers for 5 mins before the test. During the test, the mice were released from the center compartment and allowed to freely move between the open chambers for 30 mins. Whenever the center of the body of the mouse entered one of the two sides, an Arduino Uno with custom code was triggered via a TTL output pulse to continuously drive optical stimulation at frequencies close to the maximal discharge rate of these neuronal types during wakefulness (bilateral, 473 nm, 20 mW, 10 ms pulses, vGluT2: 20 Hz, ChAT: 10 Hz, PV: 40Hz) (Xu et al., 2015), with cessation of stimulation when the mouse exited the chamber. The side paired with stimulation (STIM) was counterbalanced between mice. The following day the mice were again placed in the chamber for 15 mins without optical stimulation to test if the stimulation resulted in a learned place preference/aversion. Mice were included in the analysis if post-hoc analysis confirmed bilateral transduction targeted to the BF i.e. EYFP-positive neurons and fibers. On this basis, one vGluT2-cre and one ChAT-cre mouse were excluded from the analysis. Paired t-tests were used to analyze the time spent on the stimulated side vs the unstimulated side.

RESULTS

tdTomato is not expressed in cholinergic or GABAergic neurons in the BF of vGluT2-tdTomato mice (Fig. 1)

Figure 1. Neurons expressing the vesicular glutamate transporter subtype 2 (vGluT2) are located throughout the basal forebrain (BF) and do not overlap with cholinergic (a-b) or GABAergic (c-d) neurons.

a) Fluorescent images of glutamatergic (vGluT2-tdTomato, red) and cholinergic (ChAT, green) neurons in the BF of one representative case, demonstrating lack of co-localization of vGluT2-tdTomato (Red) and ChAT (green) neurons. b) Mapping of the non-overlapping location of glutamatergic (vGluT2-tdTomato, red) and cholinergic (ChAT, green) neurons in intermediate regions of BF. c) Fluorescent images of glutamatergic (vGluT2-tdTomato, red) and GABAergic (GAD67-GFP, green) neurons in the BF of one representative case, demonstrating lack of co-localization. d) Mapping of the non-overlapping location of glutamatergic (vGluT2-tdTomato, red) and more numerous GABAergic (GAD67-GFP, green) neurons in BF. e) Tri-color photographic depiction of vGluT2-tdTomato (glutamatergic, red), ChAT (cholinergic, blue) and GAD67-GFP (GABAergic, green) neurons in BF demonstrates three largely separate neuronal populations in the BF. Scale bars = 50 μm for a and c, Low Magnification (Low Mag); 25 μm for a and c, High Magnification (High Mag); 50 μm for e.

To confirm that the fluorescent marker (tdTomato) is not expressed ectopically in BF cholinergic neurons, we performed immunohistochemical staining for the selective marker of cholinergic neurons, ChAT (Fig. 1a, b). ChAT+ and vGluT2-tdTomato neurons were present in all BF subregions. However, none of the ChAT+ cells analyzed in BF (n = 4, 3 sections/animal) co-expressed tdTomato, suggesting that cholinergic neurons do not express vGluT2. To assess co-localization with GABAergic neurons, we crossed vGluT2-tdTomato mice with GAD67-GFP knock-in mice and assessed the extent of co-localization of the red (tdTomato) and green (GFP) fluorescent markers. Again, tdTomato was not expressed in GAD67-GFP (GABAergic) neurons in BF (Fig. 1c), schematically depicted for one representative case in Fig. 1d. These results suggest that neurons with a dual GAD67/vGluT2 phenotype are not present in the BF regions we analyzed. In conclusion, glutamatergic (vGluT2-tdTomato), cholinergic (ChAT), and GABAergic (GAD67-GFP) neurons in BF are three separate neuronal populations in the BF, as illustrated in Figure 1e.

Distribution and long-axis diameter of vGluT2-tdTomato neurons in BF (Fig. 2)

Figure 2. Quantification of the density and size of vGluT2-tdTomato (glutamatergic) neurons in BF subnuclei of vGluT2-tdTomato mice.

Measures of the density (a) and long-axis diameter (b) of vGluT2-tdTomato neurons were determined in the ventral pallidum (VP), substantia innominata (SI), horizontal limb of the diagonal band (HDB), and magnocellular preoptic (MCPO) regions of rostral, medial, and caudal BF sections. a) The relative density of vGluT2-tdTomato neurons was highest in the magnocellular preoptic area (MCPO), followed by HDB and SI in the BF of vGluT2-tdTomato mice. Significantly lower densities were found in the VP compared to other BF subnuclei. b) No significant differences were found between mean long-axis diameters of different subnuclei. Comparisons along the rostrocaudal axis revealed that diameters were slightly increased in caudal sections. c) Histogram illustrating the cell frequency distribution for the long-axis diameters of vGluT2-tdTomato neurons (out of 1543 neurons evaluated across four animals), centered around a mean of 13.4 μm. Rostral representation = +0.38 mm from bregma; Medial representation = +0.14 mm from bregma; Caudal representation = −0.10 mm from bregma.

A precise mapping of the distribution, relative density and size of vGluT2 neurons within the mouse BF is not available. Thus, we first quantified the relative density and size (long-axis diameter) of vGluT2-tdTomato neurons in 3 coronal sections spanning intermediate regions of the BF (rostral, medial and caudal section representations), and across the individual subnuclei (VP, SI, HDB, MCPO) using the same methodology that we previously used to map BF GABAergic and parvalbumin neurons (McKenna et al., 2013). vGluT2-tdTomato neurons were localized throughout BF (Figs. 1, 2; n = 4), with the highest density of vGluT2-tdTomato neurons located in MCPO (56.8 ± 3.5 tdTomato+ neurons/mm2/section; Fig. 2a), followed by the HDB (47.0 ± 3.2) and SI (45.4 ± 5.0) regions. The VP region in the dorsal BF had the lowest densities (25.2 ± 6.4). Significant differences were noted between the four BF subnuclei (F_3,36 = 5.031, P = 0.005), and post-hoc pairwise comparisons indicated significant density differences between VP vs. HDB (t = 3.00, P = 0.022) and vs. MCPO (t = 3.54, P = 0.005). vGluT2-tdTomato neurons were most dense in the medial portion of BF (46.3 ± 5.0), followed by caudal (36.6 ± 8.0) and rostral (31.9 ± 0.6) regions. To summarize, vGluT2-tdTomato neurons were distributed throughout BF subnuclei, with a decrease in density noted in dorsal BF (VP) when compared to the other subnuclei. Thus, we focused on ventral BF regions in medial BF for our subsequent injections of adenoviral vectors expressing channelrhodopsin2-enhanced yellow fluorescent protein (AAV5-DIO-ChR2-EYFP) for anterograde tracing and in vivo optogenetic stimulation experiments.

In marked contrast to BF cholinergic, GABAergic and parvalbumin neurons (McKenna et al., 2013), there were virtually no large diameter (>20 μm) BF vGluT2-tdTomato neurons (Fig. 2b). BF vGluT2-tdTomato neurons were almost all small (<15 μm) or medium sized (15–20 μm) neurons (Fig. 2b and c). No significant differences were noted of the vGluT2-tdTomato neuronal long-axis diameters between BF subnuclei (n = 6; VP, 13.5 ± 0.2 μm, SI, 13.5 ± 0.3; HDB, 13.4 ± 0.3; MCPO, 13.3 ± 0.3). Slight differences were noted across the rostrocaudal axis (F_2,36 = 4.177, P = 0.02; Rostral, 13.3 ± 0.4 μm; Medial, 13.2 ± 0.4 μm; Caudal, 13.8 ± 0.2 μm). Post-hoc analysis revealed that mean values were significantly higher in Caudal vs. Medial sections (t = 2.63, P = 0.029).

The calcium-binding proteins calbindin (Calb) and calretinin (Calr) are expressed in subsets of BF vGluT2-tdTomato neurons

Previous studies in the rat demonstrated the presence of Calb in a subset of BF noncholinergic, non-GABAergic neurons that project to the cortex (Gritti et al., 2003). To determine if Calb labels a subpopulation of BF vGluT2 neurons in the mouse, we performed immunohistochemistry for Calb in vGluT2-tdTomato mice tissue. Indeed, a subpopulation of vGluT2-tdTomato neurons in the BF was Calb (Fig. 3). Photographic depiction (Fig. 3a) and neuronal mapping of one representative case across the rostral-caudal extent of BF (Fig. 3b) demonstrated vGluT2-tdTomato co-localization in a subpopulation of Calb neurons in all subnuclei of BF. Overall, 19.3% of vGluT2-tdTomato neurons also contained Calb in BF (Fig. 3c). Highest levels of co-localization were seen in MCPO (24.8 ± 3.2%), followed by HDB (19.8 ± 3.0%), VP (16.3 ± 1.6%) and SI (16.1 ± 1.2%). Statistical analysis revealed significant differences of co-localization between the BF subnuclei (F_3,36 = 5.366, P = 0.003). Post-hoc analysis confirmed that co-localization was significantly greater in the MCPO when compared to the VP (t = 3.41, P = 0.008) and to SI (t = 3.51, P = 0.006). Significant differences were not found along the rostrocaudal axis.

Figure 3. A subpopulation of BF vGluT2-tdTomato neurons co-express the calcium-binding protein calbindin (Calb).

a) Fluorescent images (Apotome Structured Illumination, z-Stack, 2 μm thick slices) of glutamatergic (vGluT2-tdTomato, red) and Calb (green) neurons in the BF, demonstrating co-localization. Scale bar = 25 μm. T, top example; B, bottom example. b) Mapping of glutamatergic (vGluT2-tdTomato, red) and Calb (green) neurons depicts many neurons which co-express both markers (black) in BF. c) The highest colocalization was observed in MCPO. There were no significant differences across the rostrocaudal axis. d) The diameter of vGluT2-tdTomato neurons that co-labeled with Calb did not significantly differ between BF subnuclei but caudal vGluT2-tdTomato/Calb neurons were larger compared to rostral and medial ssctions. e) Histogram illustrating the cell frequency distribution for the long-axis diameters of vGluT2-tdTomato/Calb neurons (out of 834 neurons evaluated across four animals), centered around a mean of 13.8 μm. Rostral representation = +0.38 mm from bregma; Medial representation = +0.14 mm from bregma; Caudal representation = −0.10 mm from bregma.

Measures of the long-axis diameter of vGluT2-tdTomato neurons co-localized with Calb (Fig. 3d) did not differ between BF subnuclei (n = 4; VP 13.7 ± 0.2 μm, SI 13.9 ± 0.2; HDB 14.1 ± 0.4; MCPO 13.6 ± 0.3), but diameter measures did significantly differ along the rostrocaudal axis (F_2,36 = 4.926, P = 0.012; Rostral 13.4 ± 0.4 μm; Medial 13.7 ± 0.4 μm; Caudal 14.3 ± 0.2 μm), and post-hoc analysis revealed that mean values were significantly higher in the Caudal vs. Rostral (t = 3.07, P = 0.01). The vast majority of vGluT2-tdTomato/Calb neurons were small, indicated in the histogram profile (Fig. 3e), centered around the overall size mean of 13.8 μm.

We next performed immunohistochemistry for another calcium binding protein, Calr, in the BF of vGluT2-tdTomato mice (Fig. 4). Calr was co-expressed in a subpopulation of vGluT2-tdTomato neurons in all subnuclei of BF (Fig. 4a–b). Overall, 24.6% of vGluT2-tdTomato neurons co-localized with Calr in BF (Fig. 4c; n = 4). Highest levels of co-localization were seen in VP (28.2 ± 1.0%), followed by MCPO (23.6 ± 2.4%), SI (22.5 ± 2.3%) and HDB (21.8 ± 1.3%). Statistical analysis revealed significant differences of co-localization between the BF subnuclei (F_3,36 = 2.852, P = 0.049). Post-hoc analysis confirmed significant pairwise differences between VP vs. HDB (t = 2.84, P = 0.034). Significant differences were not noted along the rostrocaudal axis (Rostral 24.0 ± 2.8%; Medial 25.2 ± 0.7; Caudal 24.5 ± 2.7). Measures of the long-axis diameter did not differ between BF subnuclei (Fig. 4d; n = 4; VP, 13.9 ± 0.4 μm; SI, 13.6 ± 0.5; HDB, 13.9 ± 0.5; MCPO, 13.5 ± 0.5) or along the rostro-caudal axis (Rostral 13.8 ± 0.5 μm; Medial, 13.1 ± 0.8; Caudal, 14.0 ± 0.4). The vast majority of vGluT2-tdTomato/Calr neurons in BF were small, as indicated by the histogram profile (Fig. 4e), centered around the overall mean of 13.6 μm.

Figure 4. A subpopulation of BF vGluT2 neurons co-express the calcium-binding protein calretinin (Calr).

a) Fluorescent images (Apotome Structured Illumination, z-Stack, 2 μm thick slices) of glutamatergic (vGluT2-tdTomato, red) and Calr (green) neurons in the BF, demonstrating co-localization. Scale bar = 25 μm. T, top example; B, bottom example. b) Mapping of glutamatergic (vGluT2-tdTomato, red) and Calr (green) neurons depicts many neurons which co-express both markers (black) in BF. c) Significant differences between BF subnuclei were demonstrated, and the highest colocalization was observed in VP. Significant differences between rostrocaudal sections were not seen. d) The diameter of vGluT2-tdTomato neurons that co-labeled with Calr did not significantly differ between BF subnuclei. e) Histogram illustrating the cell frequency distribution for the long-axis diameters of vGluT2-tdTomato/Calr neurons (out of 1029 neurons evaluated across four animals), centered around a mean of 13.6 μm. Rostral representation = +0.38 mm from bregma; Medial representation = +0.14 mm from bregma; Caudal representation = −0.10 mm from bregma.

Finally, we performed staining for the calcium-binding protein parvalbumin (PV). Only 1.69 ± 0.48 % of BF vGluT2-tdTomato neurons colocalized with PV; and of all BF PV neurons, 1.46 ± 0.44 % were vGluT2-tdTomato. Thus, we did not perform more detailed analysis of the subregional density or size of these neurons. We previously found that most BF PV neurons are GABAergic (McKenna et al., 2013).

Anterograde tracing using AAV5-DIO-ChR2-EYFP injections reveals projections of BF vGluT2-tdTomato neurons to neurons/brain areas involved in the control of arousal and the response to aversive and rewarding stimuli

To understand the intra- and extra-BF projections of BF vGluT2+ neurons, we performed anterograde tracing experiments using unilateral injections of AAV5-DIO-ChR2-EYFP into the BF of 6 adult vGluT2-cre-tdTomato mice to specifically express ChR2-EYFP in BF vGluT2 neurons (Fig. 5). An additional three vGluT2-cre mice were used to analyze intra-BF connections to neighboring cholinergic and PV neurons (Fig. 6). Following a survival period of one month, mice were sacrificed, and brain tissue prepared for subsequent analysis. ChR2-EYFP fusion proteins were expressed in the somata of many vGluT2-tdTomato neurons within the BF (Fig. 5a). >80% of cell somata transduced with AAV co-localized with vGluT2-tdTomato signal, confirming selectivity of viral transduction. Transduced neurons and fibers were located in the target region of BF, indicating successful AAV injection targeting (Fig. 5b). Target regions exhibited transduced fibers with varicosities (Fig. 5c). In 4/6 cases transduced neurons were restricted to the BF whereas in 2/6 cases the injections impinged on the neighboring lateral preoptic area (LPO) region (Fig. 5d).

Figure 5. Viral transduction of BF vGluT2 neurons in vGluT2-cre-tdTomato mice.

Successful AAV5-DIO-ChR2-EYFP transduction revealed by fluorescence microscopy in vGluT2-cre-tdTomato mice. a) Virally-transduced neurons (EYFP, green) co-localized (arrow) with vGluT2-tdTomato signal (red) in vGluT2-cre-tdTomato mice. Scale bar = 25μm. b) The transduced (AAV5-DIO-ChR2-EYFP, green) neurons were located in the target BF region between −0.10 to 0.38 mm from Bregma (Franklin and Paxinos, 2008). (c) Higher magnification photomicrograph of an anterogradely-labeled terminal field in the Lateral Hypothalamus following AAV5-DIO-ChR2-EYFP injection into BF. Arrowheads indicate en passant axonal varicosities, and arrows indicate terminaux. (d) Distribution of BF injection sites represented across three rostral-caudal sections of BF. 2 of the 6 cases impinged on the neighboring lateral preoptic area (LPO). Abbreviations: aca, anterior commissure (anterior); acp, anterior commissure (posterior); Cpu, caudate putamen; CTX, cortex; HDB, horizontal limb of the diagonal band; LPO, lateral preoptic area; MCPO, magnocellular preoptic nucleus; MS, medial septum; SI, substantia innominata; Tu, olfactory tubercle; VP, ventral pallidum. Scale bars: (a) 25 μm; (b) 0.5 mm; (c) 10 μm; and (d) 1mm.

Figure 6. The fibers of BF vGluT2 neurons appose neighboring cholinergic and parvalbumin neurons within BF.

(a) AAV5-DIO-ChR2-EYFP transduced BF vGluT2+ neuronal fibers (green) appose BF cholinergic neurons (ChAT+, red) (indicated by arrowheads) in vGluT2-cre mice. (b) Transduced vGluT2+ fibers (green) also appose BF parvalbumin (PV, red) neurons (indicated by arrowheads). Apposition of the BF PV fiber to post-synaptic targets, suggestive of a synaptic contact, was confirmed by the presence of a yellow (merge of red and green)-labeled axonal varicosity (arrowheads) in the orthogonal planes of the x–z (Lower) and y–z (Right) projection images. Scale bars = 20 μm.

We analyzed the fibers of vGluT2 neurons within the transduced area of BF in vGluT2-Cre mice. As shown in Fig. 6, fibers amplified with anti-GFP antibody (green) closely apposed neurons labeled with ChAT (indicating cholinergic phenotype, red, Fig. 6a) and PV (red, Fig. 6b). These findings suggest synaptic connections of vGluT2+ fibers with neighboring cholinergic and PV neurons in BF, which play a major role in BF function including arousal and sleep/wake regulation, as suggested by previous in vitro electrophysiological studies (Yang et al., 2014, 2017; Xu et al., 2015).

Next, we examined the projections of BF vGluT2+ neurons outside the transduced area. In the telencephalon, labeling included minor/moderate projections to the cerebral cortex, particularly the medial prefrontal (taenia tecta, prelimbic, and infralimbic cortices) and anterior cingulate cortices (Fig. 7a), as well as the lateral orbital and endopiriform regions (Fig. 7b). Labeled fibers were strongly evident in rostral regions of the BF including the medial septum and horizontal limb of the diagonal band (Fig. 7c), in the olfactory tubercle, as well as in the lateral septum, particularly in its intermediate aspect. Moderate labeling was observed in the amygdala including the basolateral, basomedial, and central nuclei. Moderate levels of transduced fibers were also seen in extended amygdala regions including the bed nucleus of stria terminalis (Fig. 7d).

Figure 7. BF vGluT2 fibers are evident in numerous brain regions following AAV5-DIO-ChR2-EYFP injection into BF.

Low magnification fluorescent photomicrographs of coronal sections throughout the brain show patterns of labeling produced by AAV5-DIO-ChR2-EYFP injection into BF of vGluT2-cre-tdTomato mice. Moderate/strong EYFP labeled fibers (green) innervate cortical regions including (a) the prelimbic and anterior cingulate cortices; and (b) lateral orbital and endopiriform cortices. (c) the medial septum/diagonal band and lateral septum, and moderate labeling just caudal including (d) the bed nucleus of stria terminalis and e) median preoptic nucleus. Labeling in the hypothalamus included the posterior and lateral hypothalamic regions (f); and strong labeling was noted in the (g) epithalamic lateral habenula. Furthermore, strong fiber labeling was noted in (h) the lateral supramammillary nucleus and neighboring (i) ventral tegmental area of Tsai. See list for abbreviations. Scale bar for (a, d, f-i) 100 μm; (b-c) 1 mm; and (e) 200 μm.

Fibers were seen throughout much of the hypothalamus, including the median preoptic nucleus (Fig. 7e) and lateral preoptic area. We note the possibility that preoptic labeling may reflect transduction of neighboring non-BF vGluT2+ neurons, particularly in the 2 cases in which viral transduction slightly impinged on hypothalamic/preoptic regions (Fig. 5c). Labeling was largely consistent across all six cases, though, including the four cases in which AAV injections were restricted to BF. Other hypothalamic areas with notable labeling were the arcuate nucleus, dorsomedial hypothalamic nucleus, posterior and lateral hypothalamic regions (Fig. 7f).

Many thalamic nuclei were labeled following transduction of BF vGluT2 neurons, especially midline and intralaminar nuclei. Intralaminar nuclei which exhibited labeling included the anterior centromedial and posterior parafascicular nuclei. Moderate labeling was also noted in the midline nucleus reuniens, as well as the paratenial, paraventricular, centromedial, mediodorsal, intermediodorsal, anterodorsal, and ventromedial thalamic nuclei. An absence of labeling in the thalamic reticular nucleus was a notable contrast with the prominent fiber labeling seen in this structure after transduction of BF PV neurons with the same viral vector (Kim et al., 2015; Thankachan et al., 2019). It also contrasts with the previously described projection of basal forebrain cholinergic neurons to the reticular thalamus (Hallanger et al., 1987). Within the epithalamus, a strong labeling of the lateral habenula (Fig. 7g), but not the neighboring medial habenula, was one of the most consistent and striking findings. Projections to the subthalamic zona incerta nucleus were also observed.

BF vGluT2 descending projections to caudal diencephalon, midbrain and brainstem efferent recipients included the supramammillary nucleus (Fig. 7h), substantia nigra pars compacta, ventral tegmental area of Tsai (Fig. 7i), the periaqueductal gray (largely lateral subregions), parabrachial area, nucleus incertus, red nucleus, median and dorsal raphe nuclei, laterodorsal tegmentum, and reticular formation subregions, particularly its isthmic and mesencephalic aspects.

When organized according to functional roles (Fig. 8), our BF vGluT2 circuit mapping revealed numerous efferent projections to brain nuclei that play a crucial role in arousal and vigilance state regulation (Brown et al. 2012), including moderate ascending projections to frontal cortex (cingulate, prefrontal, and orbital cortical regions), moderate/strong projections to forebrain regions including the medial and lateral septum, the lateral and preoptic hypothalamic regions, as well as the supramammillary nuclei (largely its lateral aspect). Minor/moderate projections were also noted to some thalamo-cortical nuclei that are responsible for the cortical activation evident during arousal. Descending projections to arousal-related nuclei included the parabrachial nucleus, select raphe nuclei, tegmental, and reticular formation subregions. Notably, moderate/strong projections were observed in areas involved in hippocampal theta rhythm generation (Vertes and Kocsis, 1997; Buzsaki and Moser, 2013) including the ventral tegmental area, dorsal and medial raphe nuclei of the brainstem, suprammamillary nucleus of the hypothalamus, medial septum and horizontal limb of the diagonal band, anteromedial, anterodorsal and reuniens nuclei of the thalamus, hippocampal formation, and select prefrontal cortex regions.

Figure 8. Schematic depiction of BF vGluT2 projections to brain regions involved in arousal and in the response to aversive and rewarding stimuli.

Top: Strong projections of BF vGluT2 neurons were observed to the following areas involved in arousal regulation and theta rhythm generation: medial and lateral septum, preoptic area, lateral hypothalamus, supramammillary nucleus and dorsal raphe. Less dense projections were observed in frontal cortices, midline thalamus and brainstem arousal nuclei. Bottom: Strong projections of BF vGluT2 neurons were observed in the following areas involved in the response to aversive and rewarding stimuli: Lateral septum, lateral habenula, lateral hypothalamus, ventral tegmental area, amygdala and dorsal raphe. Less dense projections were observed in prefrontal/cingulate cortices, BNST, and PAG.

Beyond arousal and theta rhythm regulatory neural circuits, we noted dense fiber concentrations in nuclei involved in the response to aversive and rewarding stimuli (Fig. 8), most notably the lateral habenula, an area which has been closely linked to negative reward predictions and aversive behavior (Tian & Uchida 2015; Lazaridis et al., 2019), as well as the ventral tegmental area and lateral hypothalamus. Minor/moderate projections were also observed in cortical regions implicated in reward neural systems regulation, including frontal cortical regions (infralimbic, prelimbic, anterior cingulate and orbitofrontal cortices). A number of other reward-related nuclei also exhibited moderate/strong labeling, such as the bed nucleus of stria terminalis, interfascicular nucleus, thalamic lateral habenula and substantia nigra pars compacta. Moderate labeling was also seen in other regions rich in dopaminergic neurons including zona incerta and A11–14.

To evaluate the laterality of BF vGluT2 projections we analyzed brain nuclei that received prominent anterograde projections (Figs 7 and 8). Projections were predominantly ipsilateral although less dense contralateral projections were observed in all nuclei analyzed with the exception of the cingulate and orbitofrontal cortices.

Optogenetic stimulation of BF vGluT2 elicits a marked place aversion

The marked projections of vGluT2 neurons to the lateral habenula and other brain areas involved in reward and aversive behavior led us to investigate the effect of optogenetically activating BF vGluT2 neurons in a place-preference paradigm (Fig. 9). Control experiments targeted two other BF cell types which our immunohistochemical staining experiments indicated are separate from vGluT2-tdTomato neurons; i.e. ChAT and PV neurons. Mice received optogenetic stimulation at a frequency close to their maximal discharge frequency during wakefulness whenever they entered one, randomly chosen side of the chamber. When such stimulation was given bilaterally to vGluT2-cre mice, the mice immediately left that side of the chamber (Video 1, Fig. 9), a dramatic effect which was repeated many times over the course of the 30 min session in all 7 mice with successful bilateral transduction targeted to the BF. One additional vGluT2 mouse, which did not have good bilateral transduction confined to the BF was excluded from the analysis. Notably, this mouse lacked an avoidance response. Overall, vGluT2-cre mice spent only 120 ± 33 s on the stimulated side versus 1158 ± 91 s on the unstimulated side, resulting in a place-preference ratio of 0.12 ± 0.04 (t=8.45, P=0.00015 paired t-test), indicating an extremely strong place aversion. Surprisingly, when mice were returned to the chamber on the following day, in the absence of optogenetic stimulation this place aversion was not maintained (Fig. 9, place-preference ratio of 1.06 ± 0.18, n.s.). Bilateral optogenetic stimulation of cholinergic neurons in ChAT-cre mice (n=6) or PV neurons in PV-cre mice (n=7) did not cause a significant place preference or aversion either on the stimulation day or on the recall day (Fig. 9. ChAT-cre: Video 2, Place preference ratio 1.10 ± 0.15 on the stimulation day, 1.24 ± 0.17 on the recall day; PV-cre: Video 3, Place preference ratio 0.70 ± 0.15 on the stimulation day, 1.06 ± 0.13 on the recall day). In addition to providing a control for our experiments with vGluT2 neurons, these experiments suggest that the effect of BF vGluT2 stimulation in producing a place aversion is not due to local excitatory effects of BF vGluT2 neuronal stimulation on neighboring cholinergic and PV neurons.

Figure 9. Optogenetic stimulation of BF vGluT2 neurons elicits a marked place aversion.

Left: BF vGluT2, cholinergic (ChAT) or parvalbumin (PV) neurons were briefly stimulated at a frequency close to their maximal discharge rate during wakefulness whenever they entered one, randomly chosen side of a three-part chamber, depicted schematically in the top part of the figure. Heatmaps showing the location of the midpoint of the mice (middle part of the figure) clearly show that the mice with stimulation of vGluT2 mice neurons avoid the area where optogenetic stimulation is given, whereas stimulation of ChAT or PV neurons does not. Quantification of these results (bottom) showed that stimulation of vGluT2 neurons results in a significant (p<0.05) place aversion (n = 7) whereas stimulation of cholinergic neurons (n = 6) or PV neurons (n = 7) does not. Right: Returning the mice to the chamber one day later in the absence of optogenetic stimulation revealed that none of the genotypes exhibited a significant place preference/aversion for the side where optical stimulation was given the previous day.

DISCUSSION

The main findings reported in this study are as follows: (i) vGluT2-tdTomato is not expressed in BF cholinergic and GABAergic neurons; (ii) BF vGluT2-tdTomato neurons are almost all small or medium sized neurons and are less dense in VP of BF; (iii) Major subsets (20–25 %) of BF vGluT2-tdTomato neurons co-express Calb or Calr, but rarely PV; (iv) BF vGluT2+neurons project within BF to cholinergic and PV neurons and outside BF to multiple areas involved in regulation of arousal and the response to aversive and rewarding stimuli; and (v) Optogenetic stimulation experiments showed that stimulation of these neurons elicits strong avoidance behavior, whereas stimulation of BF parvalbumin or cholinergic neurons did not. In the subsequent sections we discuss each of these findings and their importance.

vGluT2-cre-tdTomato mice can be used to reliably identify BF glutamatergic neurons.

Identification of glutamatergic neurons has proven challenging due to the lack of specific methods to identify them. Although staining for phosphate-activated glutaminase (Manns et al., 2001; Gritti et al., 2006) or glutamate itself (Ottersen and Storm-Mathisen, 1985) gave some hints about the distribution and function of BF glutamatergic neurons, these immunohistochemical stains are not specific for glutamatergic neurons (Kaneko and Mizuno, 1988; Akiyama et al., 1990; Laake et al., 1999), likely due to the involvement of glutamate in cellular metabolism and as a precursor for GABA synthesis. The identification of the genes for the vesicular glutamate transporters allowed glutamatergic neurons to be unequivocally identified using in situ hybridization for the first time (Fremeau et al., 2001, 2004; Herzog et al., 2001; Hur and Zaborszky, 2005). However, in situ hybridization is not easy to combine with in vitro electrophysiological recordings, and immunohistochemical staining for vGluT2 does not reliably label cell bodies (Henny and Jones, 2006, 2008), hampering the identification of the cellular properties of glutamatergic neurons. In some studies, single cell-PCR for vGluT mRNA has been used to identify glutamatergic neurons (Sotty et al., 2003). However, this technique has a tendency for false-positives and does not allow targeting of putative glutamatergic neurons prior to in vitro recording. Therefore, the development of mice with Cre recombinase expressed under the control of the vGluT2 promoter was a considerable advance (Vong et al., 2011).

In order to use vGluT2-tdTomato mice to characterize BF glutamatergic neurons in vitro and/or for optogenetic or pharmacogenetic experiments in vivo, it is important to verify that ectopic expression does not occur in cholinergic and GABAergic neurons, the other two major subtypes of BF neurons. Our results suggested that there was no or very little ectopic expression in the BF, consistent with results in vGluT2-cre mice in other subcortical regions (Vong et al., 2011) and with experiments using in situ hybridization to confirm selective expression (Anaclet et al, 2015; Xu et al., 2015). In particular, we observed no colocalization with ChAT, a selective marker for cholinergic neurons. Similarly, no colocalization with GFP was observed in vGluT2-tdTomato/GAD67-GFP knock-in mice, with GFP selectively expressed in GABAergic neurons (McKenna et al., 2013). Thus, vGluT2-cre-Tomato mice can be used to study the major subpopulation of BF glutamatergic neurons.

BF vGluT2-tdTomato neurons in the mouse do not label for ChAT, indicating lack of co-expression with cholinergic neurons

Co-localization of glutamate with acetylcholine in BF had been previously suggested (Forloni et al., 1987), including co-release at fiber terminals (Docherty et al., 1987; Allen et al., 2006), although a follow-up investigation suggested a much smaller percentage of co-localization of glutamate and acetylcholine in BF neuronal soma (Szerb and Fine, 1990). Immunohistochemical staining for vGluT2 and the vesicular acetylcholine transporter (VAChT) combined with anterograde tracing of efferent projections emanating from BF in the rat did not reveal any co-localization in varicosities in the cortex or in the lateral hypothalamus, brain regions in which both neurotransmitter markers were expressed (Henny and Jones, 2006; 2008). As reported here, vGluT2-tdTomato neurons had a different distribution and size profile than cholinergic neurons. Our previous in vitro study also suggest that they have distinct intrinsic membrane properties (Yang et al., 2017). Thus, several convergent lines of evidence suggest that cholinergic neurons do not express vGluT2.

A subpopulation of BF cholinergic neurons which largely lack the p75 neurotrophin receptor expresses vGluT3, with highest levels of co-localization in the ventral pallidum (Nickerson Poulin et al., 2006). Using retrograde tracer techniques, this unique subset of BF neurons was shown to project to the basolateral amygdala. ChAT and vGluT3 co-localization was analyzed in a number of areas known to be strongly innervated by the cholinergic component of BF, and co-localization in fibers was noted only in the amygdala, but not other brain regions. Also, vGluT3 was not detected in axonal terminals of BF cholinergic projections to other brain regions outside of the amygdala, including the neocortex and lateral hypothalamus (Henny and Jones, 2006, 2008). Furthermore, electrophysiological studies of the BF cholinergic projection to the amygdala did not reveal a glutamatergic component (Unal et al., 2015). Overall, these findings further suggest that BF cholinergic neurons do not release glutamate as a co-transmitter at their distal targets, although a local autoregulatory role involving vGluT3 cannot be ruled out.

BF vGluT2-tdTomato neurons in the mouse do not co-localize with GAD67-GFP

Once thought infeasible, co-release of glutamate and GABA has now been demonstrated at several synapses in the mammalian nervous system, in particular at entopeduncular and ventral tegmental projections to the habenula (Root et al., 2014; Shabel et al., 2014) and in the anteroventral periventricular nucleus of the hypothalamus (Ottem et al., 2004). Switching of the relative contribution of glutamatergic and GABAergic neurotransmission at these synapses is a newly identified mechanism of neural plasticity which allows a change in the sign of the synapse (excitatory or inhibitory) in response to changing behavioral needs (Shabel et al., 2014). Given these findings, we carefully examined whether there was co-expression of tdTomato and GFP in vGluT2-tdTomato/GAD67-GFP crossed mice. Our results demonstrate that vGluT2 is not co-expressed in GABAergic (GAD67+) neurons in BF. Taken together, our findings suggest that ChAT+, GAD67+ and vGluT2-tdTomato neurons represent three largely non-overlapping neuronal BF subpopulations, as also suggested by findings from other groups (Henny and Jones, 2008; Hassani et al., 2009; Anaclet et al., 2015; Xu et al., 2015). We note, however, that these findings do not exclude the possibility that some cholinergic neurons may co-express GAD65 and release GABA at their cortical targets (Saunders et al., 2015).

vGluT2-tdTomato neurons represent a subpopulation of small to mid-sized neurons scattered throughout BF

vGluT2-tdTomato neurons were located throughout BF, similar to previous in situ mapping studies of the rat BF (Fremeau et al., 2001; Herzog et al., 2001; Lin et al., 2003; Hur and Zaborszky, 2005), as well as investigations employing the vGluT2-IRES-cre mouse model used here (Anaclet et al., 2015; Xu et al., 2015; Do et al., 2016; Tooley et al., 2018). However, a detailed mapping and analysis of vGluT2 neurons in mice had not previously been reported. The density of vGluT2-tdTomato neurons in the VP subnuclei was significantly decreased compared to other BF subregions, similar to previous studies in the rat, where the vGluT2+ neuronal population was noted as “scarce” in VP (Hur and Zaborszky, 2005). Glutamatergic neurons in the rat BF were qualitatively described as small-to-medium sized, similar to the measures of vGluT2-tdTomato long-axis diameters reported here, as well as in our in vitro investigations employing the same vGluT2-tdTomato mouse model (Yang et al., 2017). Overall, vGluT2-tdTomato neurons are smaller in size than mouse BF GABAergic and PV+ neurons (McKenna et al., 2013). Our in vitro and anatomical investigations (McKenna et al., 2013; Yang et al., 2014, 2017) therefore suggest that, in the mouse, the average BF long-axis diameter is largest in cholinergic neurons, followed by GABAergic neurons, and the smallest BF neurons are of the glutamatergic vGluT2+ neuronal phenotype.

Major subsets of BF vGluT2-tdTomato neurons in the mouse express the calcium binding proteins Calb and Calr, but not PV.

The calcium binding proteins Calb, Calr, and PV are commonly used as markers for functionally distinct subsets of GABAergic interneurons in the cortex, hippocampus and striatum (Celio, 1986; Kawaguchi and Kubota, 1993; Kubota et al., 1993). However, this association of calcium binding proteins with GABAergic interneurons does not always hold in subcortical regions. Here we found that major subsets of BF vGluT2-tdTomato neurons express Calb or Calr, but few express PV, which appears to a be a selective marker for a subset of very fast-firing BF GABAergic projection neurons (McKenna et al., 2013) involved in the control of cortical gamma band oscillations (Kim et al., 2015) and arousals from sleep (McKenna et al., 2020).

Calb.

Calb+ neurons have been previously described in the rat BF (Riedel et al., 2002; Gritti et al., 2003), including co-localization with GABAergic and cholinergic populations. One study in the rat proposed that ~40% of BF Calb+ neurons co-localized with phosphate-activated glutaminase (Gritti et al., 2003) but this enzyme is not a selective marker for glutamatergic neurons. Thus, our study is the first to definitively identify a subpopulation of BF glutamatergic (vGluT2+) neurons that co-express Calb.

Calr.

In addition to PV and Calb, Calr has also been previously reported as a calcium binding protein localized in a subset of neurons distributed throughout BF in the rat (Kiss et al., 1997; Riedel et al., 2002; Gritti et al., 2003). Calr+ neurons in the rostral BF, including the medial septum and horizontal limb (both HDB and VDB), were not cholinergic and did not co-localize with PV or Calb (Kiss et al., 1997). Calr was reported to be present in a subset of noncholinergic, non-GABAergic neurons which do not project to the cerebral cortex (Gritti et al., 2003). A significant proportion of Calr+ fibers in the entorhinal cortex were identified as glutamatergic (Wouterlood et al., 2000), although it was not determined if BF was a source of this projection. Overall, growing evidence supports the suggestion that Calr is present in a subpopulation of BF glutamatergic neurons, as we directly demonstrate here.

Our results and previous results in the rat (Gritti et al., 2003) suggest distinct distributions of BF neurons expressing Calb or Calr. We show that subsets of vGluT2 neurons express one of these two calcium-binding proteins. Although these proteins are generally expressed in distinct neuronal populations, we have not performed stains for both markers in vGluT2-cre-tdTomato mice. Thus, it is also possible that there may be vGluT2 neurons which express both Calb and Calr. In the future, it would be interesting to determine the projections of these subsets of neurons and their contribution to the wakefulness promoting and avoidance effects resulting from stimulation of BF vGluT2 neurons.

Intra-BF glutamatergic projections to neighboring PV+ and cholinergic BF neurons

Examination of vGluT2+ fibers/varicosities in BF indicated that vGluT2+ neurons appose local cholinergic and PV+ neurons. Early investigations employing in situ hybridization suggested that vGluT2+ neurons in the neighboring MS/DBV provide input to local PV+ neurons that project to the hippocampus and play a major role in promotion of hippocampal theta rhythmicity (Wu et al., 2003; Hajszan et al., 2004). By combining optogenetics and in vitro patch clamp recordings, a recent study described involvement of local MS/DBV glutamatergic projections to neighboring GABAergic and, to a lesser extent, cholinergic neurons, as part of the intra-septal mechanism responsible for hippocampal theta generation (Robinson et al., 2016). Using the vGluT2-IRES-cre mouse model, in vitro electrophysiology and optogenetic experiments demonstrated local BF glutamatergic excitatory inputs to PV+ and cholinergic neurons (Xu et al., 2015). While definitive evidence of synaptic contacts requires electron microscopic analysis, collectively these light microscopy and in vitro electrophysiology and optogenetic experiments provide evidence for excitatory inputs from BF vGluT2 neurons to local PV and cholinergic neurons. In vivo optogenetic experiments suggested that strong synchronous activation of BF vGluT2 neurons promotes arousal (Xu et al., 2015), which may involve both the intra-BF connections as well as the extra-BF projections to other arousal promoting regions described here and in previous studies (Anaclet et al., 2015; Do et al., 2016; Agostinelli et al., 2017, 2019). BF vGluT2 neurons also regulate the level of the sleep homeostatic modulator, adenosine (Peng et al., 2020). The projections of vGluT2 neurons to other BF neurons shown here are likely to be involved in this regulation since glutamatergic agonists promote adenosine release in vivo (Wigren et al., 2007) and in BF slices in vitro (Sims et al., 2013).

Extra-BF glutamatergic projections to arousal-related brain regions

Our BF vGluT2 specific anterograde mapping revealed fiber labeling in brain regions previously described to play a crucial role in sleep-wake control and cortical activation (Fig. 8). The BF is a key forebrain node in the brain circuitry which promotes wakefulness and promotes high-frequency cortical activity typical of conscious states (Brown et al., 2012). However, the role of individual BF cell-types is still being uncovered. A previous study showed that strong optogenetic stimulation of BF vGluT2 neurons was highly potent in inducing prolonged periods of wakefulness (Xu et al., 2015). Weaker chemogenetic activation of the BF vGluT2 neuronal population was less potent in promoting arousal than activation of GABAergic neurons (Anaclet et al., 2015), possibly due to the relatively hyperpolarized resting membrane potential of BF vGluT2 neurons (Yang et al., 2017) compared to GABAergic neurons (McKenna et al., 2013). In addition to the local effects on other BF subtypes described above, our anterograde tracing experiments here suggest pathways by which strong activation of BF vGluT2 neurons may promote arousal.

Projections to the cortex were relatively sparse/moderate, including labeling in prefrontal, orbital, and endopiriform cortices, consistent with previous tracing studies (Hur and Zaborszky, 2005; Do et al., 2016). In particular, our findings confirm previous retrograde tracer studies that revealed a sparse projection from BF glutamatergic neurons to the medial prefrontal cortex (mPFC) in the rat, in which the glutamatergic component of the total BF-mPFC projection was estimated to be ~5% (Hur and Zaborszky, 2005). Manns et al. (2001) reported a higher estimate of glutamatergic contribution to the BF efferent projection to the entorhinal cortex, employing co-labeling of BF retrograde labeled neurons with immunohistochemical detection of phosphate-activated glutaminase. Again, immunolabeling for this enzyme may also detect GABAergic neuronal processes, contributing to a possible overestimation of BF glutamatergic projections. Considering this evidence together, BF vGluT2 output is a relatively small component of the overall direct BF-cortical projection.

Considering the relatively small direct BF glutamatergic projection to cortex, projections of BF glutamatergic neurons to other subcortical regions may be responsible for the wakefulness-promoting effects associated with optogenetic stimulation of these neurons (Xu et al., 2015). In the forebrain, moderate/strong fiber labeling was seen in the bed nucleus of stria terminalis, which has been recently described as playing a role in cortical activation (Kodani et al., 2017), as well as in the supramammillary region which promotes wakefulness and theta activity (Vertes 2015; Pedersen et al., 2017). Dense fiber labeling was also observed in a key theta-rhythm generating area, the MS/DBV (Vertes and Kocsis, 1997).

Fiber labeling was moderate to strong in the lateral hypothalamic area, where orexin/hypocretin neurons that promote consolidated wakefulness are located, confirming projections reported by other investigators (Henny and Jones, 2006; Agostinelli and Scammell 2017, 2019; Do et al., 2017). Previous work in rats (Henny & Jones 2006) and mice (Agostinelli et al., 2017) showed that BF vGluT2 fiber terminals appose orexin/hypocretin neurons and release glutamate (Agostinelli et al., 2017). Thus, this projection is one likely candidate to explain the increased wakefulness produced by stimulation of BF vGluT2 neurons. Moderate labeling was also observed in numerous thalamic regions, largely localized to select midline and neighboring thalamo-cortical nuclei previously described as essential for cortical activation (anterodorsal, mediodorsal, centromedial, intermediodorsal, and the reuniens/rhomboid nuclei) (Steriade et al., 1993; Van der Werf et al., 2002; Gent et al., 2018; Dolleman-van der Weel et al., 2019).

Although largely implicated in reward system processing recent evidence suggests that the ventral tegmental area of Tsai (VTA) may also play a role in arousal and cortical activation (Eban-Rothschild et al., 2016; Oishi et al., 2017; Sun et al., 2017; Yu et al., 2019) and defensive responses (Barbano et al., 2020). We found particularly strong fiber labeling here in VTA, one of the strongest BF vGluT2+ projections. Sparse/moderate labeling was observed in a number of classic posterior midbrain and brainstem arousal nuclei, including the laterodorsal and pedunculopontine tegmental regions, dorsal and median raphe, and the reticular formation, particularly noted in its isthmic and mesencephalic subregions. Moderate fiber labeling was also evident in the parabrachial nucleus, thought to be a key brainstem nucleus responsible for cortical activation (Fuller et al., 2011).

BF vGluT2+ neurons project to multiple regions involved in processing aversive and rewarding stimuli

As depicted in Figure 8b, our anterograde mapping studies also revealed BF vGluT2 input to many nuclei involved in processing aversive and rewarding stimuli, including VTA. Glutamatergic inputs to both dopaminergic and non-dopaminergic VTA neurons play a key role in regulation of VTA firing (Lammel et al., 2014), suggested by studies employing glutamatergic agonist infusion (Grace and Bunney, 1984; Johnson et al., 1992; Chergui et al., 1993) and regulate self-stimulation and the effects of drugs of abuse, including cocaine and opiates (Panagis and Kastellakis, 2002; Harris and Aston Jones, 2003; Harris et al., 2004). A previous investigation in the rat determined that a small portion of the BF projection to VTA from the VP region of BF is vGluT2+ (Geisler et al., 2007). However, as we show here, VP has the lowest density of BF vGluT2-tdTomato neurons. Thus, other BF subregions may have more prominent roles. Input from VP to dopaminergic, glutamatergic, and GABAergic VTA neurons has been demonstrated (Faget et al., 2016), but the relative importance of the projection to each neuronal class remains to be determined. A recent study concluded that VP glutamatergic input to VTA GABAergic neurons elicits place aversion (Tooley et al., 2018), consistent with our in vivo optogenetic findings of a strong place aversion elicited by stimulation of a larger proportion of BF vGluT2+ neurons. However, it is likely that other projections of BF vGluT2+ neurons also are important in eliciting place aversion.

In addition to the prominent BF vGluT2+ projection to VTA, some of the strongest fiber labeling was evident in the lateral habenula, which plays a crucial role in negative reward predictions (Tian and Uchida, 2015; Baker et al., 2016) and depressive behavior (Proulx et al., 2014; Browne et al., 2018). Recent studies have suggested that VP glutamatergic projections to the lateral habenula and VTA may promote behavioral response to stress or punishment avoidance (Tooley et al., 2018; Wulff et al., 2019), consistent with the response we observed here in vivo.

Besides the VTA and lateral habenula, moderate to strong inputs were observed to other nuclei important in processing aversive and rewarding stimuli (Figure 7b) such as the lateral septum (Zernig and Pinheiro, 2015; Tsanov, 2018), amygdala (Baxter and Murray, 2002; Murray, 2007; Janak and Tye, 2015), lateral hypothalamus (Aston-Jones et al., 2010; Castro et al., 2015; Tyree and de Lecea, 2017) and dorsal raphe (Ikemoto and Bonci, 2014; Luo et al., 2015; Li et al., 2016). Cortical projections to the medial prefrontal and orbitofrontal regions are also consistent with a role in processing of aversive/rewarding stimuli (Weinberger et al, 1988; Lammel et al., 2014; Boot et al., 2017).

Optogenetic activation of BF vGluT2 neurons elicits a striking place avoidance response

The prominent projections of BF vGluT2 neurons to brain areas involved in processing aversive and rewarding stimuli led us to test the effect of activating them in vivo in a place preference paradigm. Bilateral optogenetic activation of these neurons led to a striking and highly consistent avoidance response which was markedly different from the response in control experiments which targeted BF cholinergic or PV neurons. The different responses in cholinergic and PV neurons also suggest that the effect of stimulating BF vGluT2 neurons is not mediated by the local connections within BF. Instead, our anterograde tracing results presented here together with prior literature suggest that the avoidance response elicited by optogenetic stimulation of BF vGluT2 neurons is most likely mediated by a coordinated activation of neurons in the lateral habenula, lateral hypothalamus and VTA (Knowland et al., 2017; Tooley et al., 2018; Laziridis et al., 2019; Patel et al., 2019; Barbano et al., 2020).

The avoidance response we observed is consistent with a growing body of work implicating BF vGluT2 neurons in the response to aversive stimuli and in constraining reward seeking (Faget et al., 2016; Tooley et al., 2018; Patel et al., 2019). Surprisingly, in our experiments this avoidance response was not maintained one day later. Thus, BF vGluT2 neurons are involved in immediate avoidance responses but one 30 min session of optogenetic stimulation was not sufficient to develop a learned avoidance response. In the future it would be of interest to test whether multiple sessions of stimulation would change this result or not. The signals which activate BF vGluT2 neurons to cause avoidance behavior are not fully understood but an interesting recent study identified an excitatory population of BF neurons that are activated by food odor-related stimuli and drive selective avoidance of food with unpleasant odors, in part via projections to the lateral hypothalamus (Patel et al., 2019).

Conclusion

This study provides a comprehensive description of vGluT2+ neurons in the mouse BF. vGluT2-tdTomato neurons are distinct from neighboring cholinergic and GABAergic BF populations. The calcium binding proteins Calb and Calr are co-localized in subsets of BF vGluT2-tdTomato neurons indicating distinct functional subtypes. Our cell-specific anterograde investigation of BF vGluT2+ projections, together with our in vivo optogenetic experiments and previous literature proposing a prominent role in promoting wakefulness (Xu et al., 2015), suggest that these neurons are uniquely positioned to promote arousal and avoidance behavior in response to aversive stimuli. Several psychiatric disorders such as post-traumatic stress disorder, anxiety disorders and alcohol use disorder are associated with both increased arousal and enhanced aversive responses (Sharma et al., 2010; Lind et al., 2020). Thus, in the future it will be of interest to investigate a potential role for BF vGluT2 neurons in animal models of these disorders.

Table of abbreviations

3V

Third ventricle

A11

A11 dopamine cells

A12

A12 dopamine cells

A13

A13 dopamine cells

A14

A14 dopamine cells

AAV

Adeno-associated virus

aca

Anterior commissure, anterior

acp

Anterior commissure, posterior

ADP

Anterodorsal preoptic nucleus

AI

Agranular insular cortex

AMY

Amygdala

BF

Basal forebrain

BNST

Bed nucleus of stria terminalis

CA1

Hippocampal field CA1

Calb

Calbindin

Calr

Calretinin

CPu

Caudate Putamen

cc

Corpus callosum

Cg

Cingulate cortex

Cg1

Cingulate cortex, area 1

ChAT

Choline acetyl transferase

ChR2

Channelrhodopsin 2

Cl

Claustrum

cp

Cerebral peduncle

CTX

Cotex

DBv

Vertical limb of the diagonal band

DEn

Endopiriform nucleus, dorsal

DG

Dentate gyrus

DM

Dorsomedial hypothalamic nucleus

DP

Dorsal peduncular cortex

DR

Dorsal raphe nucleus

EYFP

Enhanced yellow fluorescent protein

f

Fornix

fmi

Forceps minor of the corpus callosum

fr

Fasciculus retroflexus

GAD67

Glutamic acid decarboxylase-67

GFP

Green fluorescent protein

GP

Globus pallidus

HDB

horizontal limb of the diagonal band

ic

Internal capsule

IEn

Endipiriform cortex, intermediate

IF

Interfascicular nucleus

IL

Infralimbic cortex

IMD

Intermediodorsal thalamic nucleus

LH

Lateral hypothalamic area

IPR

Interpeduncular nucleus, rostral

LDT

Laterodorsal tegmental nucleus

LHb

Lateral habenula

LM

Lateral mammillary nucleus

LO

Lateral orbital cortex

LP

Lateral posterior thalamic nucleus

LPO

Lateral preoptic area

LS

Lateral septum

LSUM

Supramammillary nucleus, lateral

M1

Primary motor cortex

M2

Secondary motor cortex

MCPO

Magnocellular preoptic nucleus

MD

Mediodorsal thalamic nucleus

Me

Medial amygdaloid nucleus

MHb

Medial habenula

ml

medial lemniscus

ML

Medial mammillary nucleus, lateral

MM

Medial mammillary nucleus, medial

MnPO

Median preoptic nucleus

MO

Medial orbital cortex

mp

Mammillary peduncle

MPA

Medial preoptic aream

mPFC

Medial prefrontal cortex

MR

Medial raphe nucleus

MS

Medial septum

mt

Mammillothalamic tract

mtg

Mammillotegmental tract

NAc

Nucleus accumbens

ns

Nigrostriatal bundle

opt

Optic tract

PAG

Periaqueductal gray

PB

Parabrachial nucleus

PFC

Prefrontal cortex

Pir

Piriform cortex

pm

Principal mammillary tract

PN

Paranigral nucleus

Po

Posterior thalamic nuclear group

PrL

Prelimbic cortex

PV

Parvalbumin

PVT

Paraventricular thalamic nucleus

RE

Nucleus reuniens

RLi

Rotral linear raphe nucleus

RPC

Red nucleus, parvicellular

SEM

Standard error of the mean

SH

Septohippocampal nucleus

SI

Substantia innominata

SNC

Substantia nigra, compacta

SNR

Substantia nigra, reticulata

SUM

Supramammillary nucleus

TT

Tenia tecta

Tu

Olfactory tubercle

vGluT2

Vesicular glutamate transporter 2

VLPO

Ventrolateral preoptic nucleus

VM

Ventromedial thalamic nucleus

VMH

Ventromedial hypothalamic nucleus

VO

Ventral orbital cortex

VP

Ventral pallidum

VPT

Ventral posterior thalamic nuclei

VTA

Ventral tegmental area of Tsai

VTM

Ventral tuberomammillary nucleus

ZI

Zona incerta