Skip to main content
PMC home page
eLife logoLink to eLife
. 2016 Sep 19;5:e13214. doi: 10.7554/eLife.13214

Cell type-specific long-range connections of basal forebrain circuit

Johnny Phong Do 1,†, Min Xu 1,†, Seung-Hee Lee 1,†,‡, Wei-Cheng Chang 1, Siyu Zhang 1, Shinjae Chung 1, Tyler J Yung 1, Jiang Lan Fan 1, Kazunari Miyamichi 2,§, Liqun Luo 2, Yang Dan 1,*
Editor: Gary L Westbrook3
PMCID: PMC5095704  PMID: 27642784

Abstract

The basal forebrain (BF) plays key roles in multiple brain functions, including sleep-wake regulation, attention, and learning/memory, but the long-range connections mediating these functions remain poorly characterized. Here we performed whole-brain mapping of both inputs and outputs of four BF cell types – cholinergic, glutamatergic, and parvalbumin-positive (PV+) and somatostatin-positive (SOM+) GABAergic neurons – in the mouse brain. Using rabies virus -mediated monosynaptic retrograde tracing to label the inputs and adeno-associated virus to trace axonal projections, we identified numerous brain areas connected to the BF. The inputs to different cell types were qualitatively similar, but the output projections showed marked differences. The connections to glutamatergic and SOM+ neurons were strongly reciprocal, while those to cholinergic and PV+ neurons were more unidirectional. These results reveal the long-range wiring diagram of the BF circuit with highly convergent inputs and divergent outputs and point to both functional commonality and specialization of different BF cell types.

DOI: http://dx.doi.org/10.7554/eLife.13214.001

Research Organism: Mouse

Introduction

The BF has been implicated in a variety of brain functions such as arousal, attention, and plasticity (Bakin and Weinberger, 1996; Brown et al., 2012; Everitt and Robbins, 1997; Froemke et al., 2013; Hasselmo and Sarter, 2011; Jones, 2011; Lin et al., 2015; Saper et al., 2010; Sarter et al., 2001). The dysfunction or loss of BF cholinergic neurons is an important feature of Alzheimer’s disease associated with cognitive impairment (Schliebs and Arendt, 2011; Whitehouse et al., 1982). In addition to forming extensive local synapses (Xu et al., 2015; Yang et al., 2014; Zaborszky and Duque, 2000), BF neurons receive inputs (Asanuma, 1989; Freund and Meskenaite, 1992; Grove, 1988a, Henny and Jones, 2006; Manns et al., 2001; Parent et al., 1988; Rye et al., 1984; Semba et al., 1988; Zaborszky and Cullinan, 1992) and send outputs (Cullinan and Zaborszky, 1991; Grove, 1988b; Jones and Cuello, 1989; Mesulam and Mufson, 1984; Paré and Smith, 1994) to many other brain areas (Steriade and McCarley, 2005; Zaborszky et al., 2012). However, how these long-range connections contribute to BF functions remains unclear.

An important challenge in understanding the function of the BF circuit is its neuronal heterogeneity. There are three major cell types spatially intermingled in the BF: cholinergic, glutamatergic, and GABAergic (Semba, 2000; Zaborszky et al., 2012). Selective lesion or pharmacological manipulation of the cholinergic system is well known to affect multiple brain functions (Wenk, 1997). For example, 192-IgG-saporin-mediated lesion of cholinergic neurons impaired the ability of rats to discriminate between signal and non-signal visual events in an attention task (McGaughy et al., 1996) and disrupted training-induced cortical map reorganization associated with motor learning (Conner et al., 2003). Glutamatergic and GABAergic BF neurons are also likely to serve important functions (Lin et al., 2015). For example, in recent studies the activity of non-cholinergic BF neurons was found to correlate with sustained attention (Hangya et al., 2015) or to encode reward and motivational salience information (Avila and Lin, 2014; Lin and Nicolelis, 2008; Nguyen and Lin, 2014), and optogenetic activation of PV+ GABAergic neurons was shown to regulate cortical gamma oscillations (Kim et al., 2015). In a study on sleep-wake control, cholinergic, glutamatergic, and PV+ neuron activity was found to promote wakefulness, while SOM+ neurons promoted sleep; these four cell types form extensive but highly specific local connections with each other for brain-state regulation (Xu et al., 2015). Thus, to understand the BF circuit function, it is crucial to map its inputs and outputs with cell-type specificity.

Most of the previous studies of BF long-range connections focused on specific regions connected to the BF, making it difficult to assess their whole-brain distribution. Recent advances in virus-assisted circuit tracing (Callaway and Luo, 2015; Huang and Zeng, 2013) and high-throughput imaging have greatly facilitated whole-brain mapping of long-range connectivity in a cell-type-specific manner (Oh et al., 2014; Osten and Margrie, 2013). In this study, we traced the long-range inputs and outputs of four genetically defined BF cell types. While the input distributions were similar across cell types, their output patterns showed striking differences. Our quantitative analysis of the whole-brain distributions of inputs and outputs for each BF cell type can serve as an anatomical blueprint for future studies of inter-regional pathways mediating BF functions.

Results

Four Cre mouse lines were used to target different BF subpopulations for virus-mediated circuit tracing: choline acetyltransferase (ChAT)-Cre for cholinergic neurons, vesicular glutamate transporter 2 (VGLUT2)-Cre for glutamatergic neurons, and PV-Cre and SOM-Cre mice for two subtypes of GABAergic neurons. These four Cre lines have been shown to label largely non-overlapping BF neuron populations with high specificity (Xu et al., 2015).

To identify the long-range inputs to each cell type, we used RV-mediated transsynaptic retrograde tracing, which has been shown to label monosynaptic inputs to selected starter cells with high specificity (Miyamichi et al., 2011; Wall et al., 2013; Watabe-Uchida et al., 2012; Wickersham et al., 2007). First, we expressed avian-specific retroviral receptor (TVA), enhanced green fluorescent protein (eGFP), and rabies glycoprotein (RG) specifically in each cell type by injecting two Cre-inducible AAV vectors (AAV2-EF1α-FLEX-eGFP-2a-TVA and AAV2-EF1α-FLEX-RG) into the BF of ChAT-, VGLUT2-, PV-, or SOM-Cre mice (Figure 1A). The expression of RG was highly cell type specific and not detected in wild-type mice not expressing Cre recombinase (Figure 1—figure supplement 1). Two to three weeks later, we injected a modified RV (rabiesΔG-tdTomato+EnvA) that only infects cells expressing TVA, requires RG to spread retrogradely to presynaptic cells (Figure 1—figure supplement 2), and contains the tdTomato transgene. After histological sectioning and fluorescence imaging, each sample was aligned to a reference atlas (Allen Mouse Brain Atlas, see Materials and methods) to facilitate 3D whole-brain visualization and quantitative comparison across brain samples (Figure 1C). The starter cells (expressing both tdTomato and eGFP) and the transsynaptically labeled presynaptic neurons (expressing tdTomato only) were identified manually, and their locations were registered in the reference atlas (Figure 1—figure supplement 3).

Figure 1. Experimental and analysis procedures for cell-type-specific circuit tracing.

(A) RV-mediated transsynaptic retrograde tracing of BF inputs. Upper panel, viral vectors and injection procedure. Lower panel, fluorescence images of BF in the region of the NDB (red box in coronal diagram) in ChAT-, VGLUT2-, PV-, and SOM-Cre mice. Scale bar, 200 µm. Inset, enlarged view of the region in white box showing starter cells (yellow, expressing both eGFP and tdTomato, indicated by white arrowheads). Scale bar, 50 µm. NDB, diagonal band nucleus; SIB, substantia innominata, basal part; MCPO, magnocellular preoptic nucleus; VP, ventral pallidum; LPO, lateral preoptic area. (B) Viral vector and injection procedure for tracing BF axonal projections. (C) Flow chart showing the main steps in data generation and processing.

DOI: http://dx.doi.org/10.7554/eLife.13214.002

© 2008 Elsevier. All Rights Reserved

Lower panel, brain outline adapted from Figure 32 from The Mouse Brain in Stereotaxic Coordinates, 3rd edition, Franklin, K.B.J. and Paxinos, G.

Figure 1.

Figure 1—figure supplement 1. Cell-type specificity of Cre-dependent rabies glycoprotein expression.

Figure 1—figure supplement 1.

(A) Colocalization of rabies glycoprotein immunostaining with Cre expression (indicated by tdTomato or mCherry reporters) in each of the four Cre lines. White arrowheads indicate cells with colocalization. No rabies glycoprotein expression was detected when injected into wild type mice. (B) Percentage of rabies glycoprotein expressing cells that are positive for tdTomato or mCherry, averaged across brain samples. Error bar, ± standard deviation (91 ChAT cells; 89 VGLUT2 cells; 70 PV cells; 100 SOM cells; n = 2 mice per line).
DOI: http://dx.doi.org/10.7554/eLife.13214.003

Figure 1—figure supplement 2. Control experiments for RV tracing of inputs.

Figure 1—figure supplement 2.

(A) Injection of RV without prior AAV injection resulted in no tdTomato-labeled neurons, indicating dependence of the RV infection on AAV-induced expression of TVA. (B) Injection of AAV2-EF1α-FLEX-eGFP-2a-TVA and AAV2-EF1α-FLEX-RG followed by RV injection in the BF of wild-type mice not expressing Cre led to no eGFP expression, indicating Cre-dependence of the AAV vector. However, tdTomato-labeled neurons were observed at the injection site (radius < 500 μm), most likely due to the leaky expression of a low level of TVA, as previously noted (Miyamichi et al., 2013; Wall et al., 2013). (C) Upper panel, Sagittal view of the experiment shown in B (but a different brain sample), with a tdTomato expression near the injection site but not outside of the exclusion zone. Lower panel, enlarged view of the region in the white rectangle. (D) Sagittal view of brain samples injected with AAV2-EF1α-FLEX-eGFP-2a-TVA followed by RV in the BF of different Cre lines (without AAV2- EF1α-FLEX-RG that enables transsynaptic spread of RV) to determine the spatial extent of the exclusion zone in the RV tracing experiments. After excluding the horizontal limb of the diagonal band of Broca (part of the BF region targeted), we found very few (<30 per brain) labeled cells beyond 850 μm. Subsequent analyses were thus performed only in coronal sections >850 μm from the injection site and outside of the horizontal limb of the diagonal band of Broca.
DOI: http://dx.doi.org/10.7554/eLife.13214.004

Figure 1—figure supplement 3. Heat map distribution of starter cells.

Figure 1—figure supplement 3.

Normalized starter cell density across all samples for each cell type. Each brain slice depicts the density accumulated from an anterior-posterior axis range of 0.24 mm.
DOI: http://dx.doi.org/10.7554/eLife.13214.005

Figure 1—figure supplement 4. The relationship between the numbers of starter cells and input cells.

Figure 1—figure supplement 4.

(A) The total number of starter cells for each brain sample. (B) The convergence index (input cell count/starter cell count) for each brain sample grouped by cell-type.
DOI: http://dx.doi.org/10.7554/eLife.13214.006
Open in a new tab

Brain samples were excluded from the analyses if very few input neurons (<200) were labeled in the whole brain. As noted in previous studies, due to the extremely efficient interaction between TVA and EnvA-pseudotyped rabies virus, the very low-level expression of TVA in non-Cre-expressing cells (not detectable based on fluorescent protein markers) allows the rabies virus to infect and label these cells with tdTomato at the injection site, independent of synaptic connections with starter cells (Beier et al., 2015; Menegas et al., 2015; Miyamichi et al., 2013; Ogawa et al., 2014; Pollak Dorocic et al., 2014; Wall et al., 2013; Watabe-Uchida et al., 2012; Weissbourd et al., 2014). However, this local contamination does not compromise the mapping of long-range inputs because RG (required for transsynaptic spread of RV) is not expressed in any non-Cre-expressing cells at sufficient levels for trans-complementation of rabiesΔG to allow transsynaptic spread of RV (Callaway and Luo, 2015; Miyamichi et al., 2013). To determine the spatial extent of the local contamination, we performed control experiments in the absence of RG and found very few non-specific labeling at >850 μm from the injection center (Figure 1—figure supplement 2D). Thus, presynaptic neurons were counted only in coronal sections outside of this range. While this procedure precludes identification of local inputs, synaptic interactions of the four cell types within the BF have been characterized electrophysiologically in a recent study (Xu et al., 2015). Another technical limitation of the study is that when the brain was removed for histological processing, the olfactory bulb was often damaged, which led to a significant underestimation of labeling (both the input neurons and axon projections) in the olfactory bulb.

We found 900 – 14,631 (median 7002) tdTomato-labeled presynaptic neurons in each brain (n = 17), and the convergence index (ratio between the number of input cells and starter cells) ranged between 4.3 and 77.7 (Figure 1—figure supplement 4). Such variability is comparable to that found in other studies using similar methods (Miyamichi et al., 2011; DeNardo et al., 2015). The presynaptic neurons were predominantly ipsilateral to the starter population (<5% contralateral) but were distributed in a large number of brain areas (Figure 2, Figure 3, Video 1). Since the number of labeled neurons varied across brain samples, and there was no significant difference between the four cell types (P = 0.27, one-way ANOVA), we normalized the data in each area by the total number of labeled neurons in each brain. When the brain was divided into 12 major regions (Figure 3A), the striatum and hypothalamus provided the highest numbers of inputs, while few labeled neurons were found in the medulla or cerebellum (Figure 3A).

Video 1. 3D whole-brain view of RV-labeled inputs to ChAT+, VGLUT2+, PV+ and SOM+ BF neurons.

Download video file (1.8MB, mp4)
DOI: 10.7554/eLife.13214.010
Open in a new tab

Shown are data from four example brains (one for each cell type). Each red dot represents one RV-labeled presynaptic neuron. Blue, coronal sections within 850 µm from the injection site; neurons within this region were excluded from analyses to minimize contamination by the local background (Figure 1—figure supplement 2).

DOI: http://dx.doi.org/10.7554/eLife.13214.010

Figure 2. Inputs to each BF cell type from selected brain regions.

Figure 2.

Open in a new tab

Examples of RV-labeled input neurons to each of the four BF cell types in seven selected brain structures (black box in each coronal diagram). Scale bar, 200 µm. In each coronal diagram, RV-labeled neurons detected in all four brain samples are indicated by red dots. Bottom panel, mean percentage of input neurons in each brain structure for the four BF cell types. Error bar, ± s.e.m. Bar color indicates which of the 12 regions the given brain structure belongs to as depicted in Figure 3. ac, anterior commissure; aq, cerebral aqueduct; BLA, basolateral amygdalar nucleus; DMH, dorsomedial nucleus of the hypothalamus; DR, dorsal nucleus raphe; IPN, interpeduncular nucleus; opt, optic tract; scp, superior cerebellar peduncles; SNr, substantia nigra reticularis; VMH, ventromedial hypothalamic nucleus.

DOI: http://dx.doi.org/10.7554/eLife.13214.007

© 2008 Elsevier. All Rights Reserved

Upper panel, brain outlines adapted from Figures 13, 19, 44, 48, 60, 70, 74, from The Mouse Brain in Stereotaxic Coordinates, 3rd edition, Franklin, K.B.J. and Paxinos, G.

Figure 3. Whole-brain distributions of inputs to the four BF cell types.

Figure 3.

Open in a new tab

(A) Percentages of retrogradely labeled input neurons in 53 brain areas (ChAT, n = 5 mice; VGLUT2, n = 5; PV, n = 3; SOM, n = 4). Brain areas are grouped into 12 generalized, color-coded brain structures. HPF, hippocampal formation. Abbreviations of the 53 brain areas and their percentages of inputs are listed in Figure 3—source data 1. Error bar, ± s.e.m. Since labeled neurons in coronal sections near the injection site were excluded from analysis (see Figure 1—figure supplement 2), inputs from the pallidum are likely to be underestimated. (B) Whole-brain 3D reconstruction of the inputs to the four BF cell types. The blue-shaded area denotes the region excluded for analysis due to potential local contamination (see Figure 1—figure supplement 2).

DOI: http://dx.doi.org/10.7554/eLife.13214.008

Figure 3—source data 1. Distribution of input cells in 53 brain areas for ChAT+, VGLUT2+, PV+, and SOM+ BF neurons.
Shown are the mean ± s.e.m. of the percentage of inputs from each area for individual cell types. Note that within each of the 12 brain structures, there are unnamed sub-regions outside of the 53 areas listed in the table; thus the percentages of inputs in the listed areas do not always add up to the total percentage in the given brain region.
DOI: http://dx.doi.org/10.7554/eLife.13214.009
elife-13214-fig3-data1.xlsx (15KB, xlsx)
DOI: 10.7554/eLife.13214.009

To facilitate data visualization at different levels of detail, we also used an interactive sunburst diagram (adapted from Allen Mouse Brain Atlas, http://www.brain-map.org/api/examples/examples/sunburst/) to represent the whole-brain distribution of inputs to each cell type (http://sleepcircuits.org/bf/). The brain structures are arranged hierarchically from inner to outer circles, and the size of each sector represents the percentage of input from the corresponding structure. The name of each structure and its input percentage can be read out by pointing the cursor, and each region of interest can be expanded with a mouse click.

When the input distribution was analyzed at a finer spatial scale (e.g., the 6th ring of the sunburst plot), the nucleus accumbens (Mesulam and Mufson, 1984; Zaborszky and Cullinan, 1992), lateral hypothalamus (Cullinan and Zaborszky, 1991; Grove, 1988b; Mesulam and Mufson, 1984), and central nucleus of the amygdala (Grove, 1988b; Pare and Smith, 1994) were among the structures containing the highest numbers of input neurons (Figure 2). Interestingly, many close neighbors of these densely labeled structures (e.g., the basolateral nucleus of the amygdala, immediately adjacent to the central nucleus) showed very sparse or no labeling, indicating high spatial specificity of the long-range inputs. On the other hand, the input distributions were qualitatively similar between cell types, although with quantitative differences. For example, glutamatergic neurons received significantly more inputs from the lateral hypothalamus than the other cell types (P = 0.001, VGLUT2 vs. ChAT; P = 0.001, VGLUT2 vs. PV; P = 0.001, VGLUT2 vs. SOM;, one-way ANOVA and post-hoc Tukey’s test), and PV+ neurons received more inputs from the nucleus accumbens (ACB) (P = 0.004, PV vs. ChAT; P = 0.003, PV vs. VGLUT2; P = 0.001, PV vs. SOM; one-way ANOVA and post-hoc Tukey’s test).

To further verify the inputs revealed by RV-mediated retrograde tracing, we optogenetically tested the synaptic connections from the prefrontal cortex (PFC) and ACB (Figure 4). To verify the innervation from PFC to BF cholinergic neurons, we injected AAV (AAV-DJ-CaMKIIα-hChR2-eYFP) expressing the mammalian codon-optimized channelrhodopsin-2 (hChR2) fused with enhanced yellow fluorescent protein (eYFP) in the orbital and agranular insular areas of the PFC (Figure 4—figure supplement 1) in ChAT-eGFP mice and made whole-cell voltage-clamp recordings from eGFP-labeled cholinergic neurons in acute BF slices (Figure 4A). Activating the ChR2-expressing axon terminals with blue light evoked excitatory responses in all recorded BF cholinergic neurons (n = 9, Figure 4B and C), confirming the input revealed with RV tracing. To confirm the innervation from ACB, we injected Cre-inducible AAV (AAV-DJ-EF1α-FLEX-ChR2-eYFP) expressing ChR2-eYFP in ACB of GAD2-Cre mice, made whole-cell current-clamp recordings from unlabeled postsynaptic BF neurons, and used single cell reverse-transcription PCR (RT-PCR) to identify the cell type. We found that all four BF cell types received inhibitory responses from the ACB (Figure 4D and E; ChAT+: 2 out 5 showed significant responses; VGLUT2+: 2/4; PV+: 3/3; SOM+: 4/8), which is consistent with the finding of an electron microscopic double-immunolabeling study performed in rats (Zaborszky and Cullinan, 1992).

Figure 4. Optogenetic characterization of monosynaptic inputs to the BF from PFC and ACB.

(A) Schematic of experiment. ChR2 was expressed in excitatory neurons in the prefrontal cortex of ChAT-eGFP mice by injecting AAV-DJ-CaMKIIα-hChR2-eYFP. Coronal slices of the BF were used for recording experiments. (B) Excitatory postsynaptic potentials recorded from ChAT+ neurons (under whole-cell current clamp) evoked by blue-light activation of the prefrontal cortical axons. Upper, response to a single light pulse (5 ms) in an example ChAT+ neuron; lower, responses to 10 pulses at 10 Hz recorded from a different ChAT+ neuron. (C) Summary of the peak amplitude of the response to a single light pulse. Each circle represents data from one BF ChAT+ neuron (n = 9 neurons from 2 mice). Bar, mean ± s.e.m. (D) Diagram illustrates virus injection site in the ACB and recording site in the BF. AAV-DJ-EF1α-FLEX-ChR2-eYFP was injected into the ACB of GAD2-Cre mice and whole-cell voltage-clamp recordings (clamped at 0 volts) were made from BF neurons. Single-cell gene-expression analysis was performed after each recording session to identify the cell type of each recorded neuron. (E) Example traces of laser-evoked responses in the four BF cell types. (F) Summary of the peak current amplitude of each neuron’s response (ChAT+, n = 5 neurons from 5 mice; VGLUT2+, n = 4 neurons from 4 mice; PV+, n = 3 neurons from 3 mice; SOM+, n = 8 neurons from 4 mice). Gray indicates no significant response.

DOI: http://dx.doi.org/10.7554/eLife.13214.011

© 2008 Elsevier. All Rights Reserved

Panels A and D, brain outlines adapted from Figures 14, 18, 30, from The Mouse Brain in Stereotaxic Coordinates, 3rd edition, Franklin, K.B.J. and Paxinos, G.

Figure 4.

Figure 4—figure supplement 1. Basal forebrain input from the prefrontal cortex.

Figure 4—figure supplement 1.

(A) Example fluorescence image of a coronal section at the virus injection site in the prefrontal cortex (PFC). (B) Example fluorescence image of the PFC axon fibers in the basal forebrain from the same experiment as shown in panel A.
DOI: http://dx.doi.org/10.7554/eLife.13214.012
Open in a new tab

We next mapped the output of each BF cell type. To label the axonal projections, we injected AAV with Cre-dependent expression of mCherry (Figure 1B) into the BF of ChAT-, VGLUT2-, PV- or SOM-Cre mice. Two to three weeks after injection, the brain tissues were processed, images were registered to the reference atlas, and labeled axons were detected (Figure 1C, see Materials and methods). After the injection site (identified by the existence of labeled cell bodies) and locations with known major fiber tracks were excluded, the projection to each brain area was quantified by the number of pixels occupied by the detected axons (Oh et al., 2014) (see Materials and methods).

Parallel to the broad distribution of inputs (Figure 3), we found that each BF cell type also projected to a large number of brain areas (Figure 5, Figure 6, Video 2, http://sleepcircuits.org/bf/, >95% ipsilateral). Among the 12 major brain subdivisions (Figure 6A), the hypothalamus, pallidum, and striatum received the heaviest BF projections (Grove, 1988a; Henny and Jones, 2006), while very few axons were detected in the medulla or cerebellum (Figure 6A). Analysis at finer scales revealed high spatial specificity of the projections. For example, while several cell types projected strongly to the lateral habenula (Figure 5), few axons were detected in the immediately adjacent but anatomically distinct medial habenula (Hikosaka, 2010). In addition to providing extensive inputs to the BF (Figure 2), the lateral hypothalamus was also a major recipient of BF projections (Figure 5), indicating a strong BF-hypothalamus loop that may be important for brain-state regulation (Brown et al., 2012; Jones, 2011; Saper et al., 2010). Importantly, whereas the input distributions were generally similar across BF cell types (Figure 2, Figure 3), the output patterns showed striking differences. For example, compared to the other cell types, the projection from cholinergic neurons was much stronger in the basolateral amygdala, hippocampus, and visual cortex but much weaker in the lateral hypothalamus, lateral habenula, and the ventral tegmental area (Figure 5). The different projection patterns among cell types are also apparent in the 3D whole-brain view (Figure 6B, Figure 6—source data 1).

Video 2. 3D whole-brain view of mCherry-labeled axonal projections from ChAT+, VGLUT2+, PV+ and SOM+ BF neurons.

Download video file (1.7MB, mp4)
DOI: 10.7554/eLife.13214.016
Open in a new tab

Shown are projections averaged across all brain samples of each cell type.

DOI: http://dx.doi.org/10.7554/eLife.13214.016

Figure 5. Axon projections of each BF cell type to selected brain regions.

Figure 5.

Open in a new tab

Examples of axon projections from each of the four BF cell types to seven selected brain structures (black box in each coronal diagram). Scale bar, 250 µm. DMH, dorsomedial nucleus of the hypothalamus; IPN, interpeduncular nucleus; MH, medial habenula; SNr, substantia nigra reticularis; VMH, ventromedial hypothalamic nucleus.

DOI: http://dx.doi.org/10.7554/eLife.13214.013

© 2008 Elsevier. All Rights Reserved

Upper panel, brain outlines adapted from Figures 23, 48, 57, from The Mouse Brain in Stereotaxic Coordinates, 3rd edition, Franklin, K.B.J. and Paxinos, G.

Figure 6. Whole-brain distributions of axonal projections from the four BF cell types.

Figure 6.

Open in a new tab

(A) Percentages of labeled axons in 53 brain areas (ChAT, n = 3 mice; VGLUT2, n = 3; PV, n = 3; SOM, n = 3). Error bar, ± s.e.m. Abbreviations of the 53 brain areas and their percentages of inputs are listed in Figure 6—source data 1. (B) Whole-brain 3D reconstruction of axon projections from each of the four BF cell types. Note that although VGLUT2+ and PV+ neuron projections showed the similar spatial distribution, there were fewer labeled axons from PV+ than VGLUT2+ neurons.

DOI: http://dx.doi.org/10.7554/eLife.13214.014

Figure 6—source data 1. Distribution of axonal projections to 53 brain areas from ChAT+, VGLUT2+, PV+, and SOM+ BF neurons.
Shown are the mean ± s.e.m. of the percentage of projections to each area for individual cell types.
DOI: http://dx.doi.org/10.7554/eLife.13214.015
elife-13214-fig6-data1.xlsx (14.9KB, xlsx)
DOI: 10.7554/eLife.13214.015

To further compare the inputs and outputs between cell types, we averaged the spatial distributions across brain samples of each cell type and computed the correlation coefficient (CC) between cell types. For input distribution, the CCs between all cell types were high (Figure 7A), confirming their overall similarity observed earlier (Figure 2, Figure 3). On the other hand, when we computed the CCs between individual brain samples, we found higher CCs between samples of the same cell type (0.81 ± 0.04, s.e.m.) than of different cell types (0.70 ± 0.02, P = 0.01, t-test; Figure 7—figure supplement 1A). This indicates that despite the overall similarity, there were genuine differences between cell types that were beyond experimental variability.

Figure 7. Comparison of input and output distributions.

(A) Matrix of correlation coefficients (CCs) between input distributions of each pair of cell types. (B) Similar to A, for output distributions. (C) CCs between input and output distributions. All CCs were computed at the spatial scale of the 12 major brain subdivisions (Figure 7—source data 1). (D) Percentage of input vs. percentage of output in each region, for each of the four BF cell types. Filled circles, strongly connected brain regions contributing to the high CCs for glutamatergic and SOM+ neurons and low CCs for cholinergic and PV+ neurons in C.

DOI: http://dx.doi.org/10.7554/eLife.13214.017

Figure 7—source data 1. Distribution of BF input and output from 12 major brain subdivisions across cell-type.
Shown are the mean ± s.e.m. of the inputs (A) and outputs (B) for each color-coded brain region, for individual cell types.
DOI: http://dx.doi.org/10.7554/eLife.13214.018
elife-13214-fig7-data1.xlsx (12.3KB, xlsx)
DOI: 10.7554/eLife.13214.018

Figure 7.

Figure 7—figure supplement 1. Correlation coefficients between individual brain samples for input and output distributions.

Figure 7—figure supplement 1.

(A) Input. (B) Output. Note the higher CCs within the boxes along the diagonal (between samples of the same cell type) than those outside of the boxes (between samples of different cell types). The CCs exactly along the diagonal (each brain sample with itself, CC = 1) were excluded from analysis.
DOI: http://dx.doi.org/10.7554/eLife.13214.019
Open in a new tab

For output distribution, the CCs between individual samples of the same cell type were also high (0.86 ± 0.05; Figure 7—figure supplement 1B), indicating reproducibility of the mapping. However, most of the CCs between cell types (Figure 7B, computed after averaging across samples of the same cell type) were much lower than those for input distribution. The two lowest CCs (ChAT+ vs. VGLUT2+ and PV+ neurons) reflect the fact that while the cholinergic neurons project strongly to structures within the cerebral cortex (including olfactory areas, isocortex, hippocampus, and cortical subplate) and weakly to the brain stem structures (thalamus, hypothalamus, and midbrain), glutamatergic and PV+ neurons (with output distributions highly similar to each other) showed complementary projection patterns (Figure 6).

Finally, we computed the CC between the input and output distributions of each cell type (Figure 7C). The highest CC was found for SOM+ neurons, reflecting their strong reciprocal connections with a number of brain structures, including the hypothalamus, striatum, pallidum, and olfactory areas (Figure 7D, lower right, Figure 7—source data 1. For glutamatergic neurons, the high CC reflects their strong reciprocal connections with the hypothalamus and striatum (Figure 7D, upper right). For cholinergic and PV+ GABAergic neurons, the CCs between input and output distributions were much lower, reflecting the facts that while both cell types receive strong input from the striatum, cholinergic neurons project strongly to the cerebral cortex, and PV+ neurons to the pallidum and hypothalamus (Figure 7D, upper and lower left).

Discussion

Using virus-mediated circuit mapping, we have characterized the whole-brain distributions of BF long-range connections, available in an open-access online database (http://sleepcircuits.org/bf/). Our experiments confirmed many previously demonstrated connections, but with cell-type specificity and quantitative analyses at multiple spatial scales. For example, we found that cortical inputs (Mesulam and Mufson, 1984; Steriade and McCarley, 2005) to all four BF cell types originate primarily from the agranular insular and orbital areas of the prefrontal cortex (Figure 2). While a previous ultrastructural study failed to detect convincing synaptic contact between prefrontal axons and BF cholinergic neurons (Zaborszky et al., 1997), our RV-mediated transsynaptic tracing demonstrated extensive monosynaptic innervation, which was also validated by electrophysiological recordings (Figure 4B and C). These findings have important implications on how the prefrontal cortex may exert top-down control of neural processing through its projection to the BF (Sarter et al., 2001). A recent study showed that cholinergic neurons in the BF are strongly activated by reinforcement signals during an auditory detection task (Hangya et al., 2015). Our whole-brain mapping of their inputs provides a list of candidate neurons through which the reinforcement signals are conveyed to the BF cholinergic neurons.

Regarding the outputs, we found striking differences across cell types (Figure 6). A recent study has shown that cholinergic, glutamatergic, and PV+ neurons all promote wakefulness, while SOM+ neurons promote sleep (Xu et al., 2015). The distinct projection patterns between cholinergic and glutamatergic/PV+ neurons (Figure 6) suggest that they preferentially regulate different brain functions during wakeful states. In a recent study, optogenetic activation of BF PV+ neurons was shown to enhance cortical gamma band oscillations (Kim et al., 2015). In addition to direct projections to the cortex, our study showed extensive subcortical projections of PV neurons, which may also contribute to the regulation of cortical gamma oscillations. The output distribution of SOM+ neurons, on the other hand, was highly correlated with the input distributions of all BF cell types (Figure 7C, bottom row); the broad GABAergic inhibition of these input areas by SOM+ neurons may be important for the sleep-promoting effect. Thus, while the highly convergent inputs from multiple brain areas allow a variety of sensory, motor, cognitive, and emotional signals to be integrated within the BF, the distinct projections by different cell types may enhance the versatility of the BF in coordinating diverse functions of multiple brain networks.

Materials and methods

Virus preparation

Transsynaptic retrograde tracing

To construct AAV2-EF1α-FLEX-eGFP-2a-TVA and AAV2-EF1α-FLEX-RG, TVA and eGFP linked by the 2A ‘self-cleaving’ peptide or rabies glycoprotein was respectively cloned into pAAV-MCS (Stratagene, La Jolla, CA) in an antisense direction flanked by a pair of canonical loxP sites and a pair of lox2272 sites. AAV particles (AAV2/2) were produced by co-transfection of packaging plasmids into HEK293T cells, and cell lysates were fractionated by iodixanol gradient ultracentrifugation. Viral particles were further purified from the crude fraction by heparin affinity column (HiTrap Heparin HP Columns; GE Healthcare, Pittsburgh, PA), desalted and concentrated with Amicon Ultra Centrifugal Filter (100 K, Millipore, Bellerica, MA). The genomic titer of AAV2-EF1α-FLEX-eGFP-2a-TVA (4.4 × 1013 gc/ml) and AAV2-EF1α-FLEX-RG (2.2 × 1012 gc/ml) was estimated by quantitative PCR. eGFP-2a-TVA and rabies glycoprotein were subcloned from the AAV-TRE-HTG plasmid from L. Luo.

RV-ΔG-tdTomato was amplified in B7GG cells and pseudotyped using BHK-EnvA cells. EnvA pseudotyped rabies virus was titered (1.5 × 109 IU/ml) by infecting the 293T-TVA8000 (Narayan et al., 2003) cell line with serial dilutions of the stock virus. RV-ΔG-tdTomato was a gift from B. Lim. B7GG cells, BHK-EnvA cells (Wickersham et al., 2007), and 293T-TVA8000 cells were gifts from E. Callaway.

Anterograde axon tracing

AAV2-EF1α-FLEX-mCherry was purchased from the UNC Vector Core (Chapel Hill, NC) and the titer was estimated to be ~1012 gc/ml.

Surgery and viral injections

All experimental procedures were approved by the Animal Care and Use Committee at the University of California, Berkeley. For the current study, we targeted the caudal portion of the BF (including the horizontal limb of the diagonal band of Broca, magnocellular preoptic nucleus, and substantia innominata) rather than the rostral nuclei (medial septum and the vertical limb of the diagonal band of Broca). For virus injection, adult (>P40) Chattm2(cre)Lowl (ChAT-Cre, JAX#006410), Slc17a6tm2(cre)Lowl (Vglut2-Cre, JAX#016963), Pvalbtm1(cre)Arbr (PV-Cre, JAX#008069), and Ssttm2.1(cre)Zjh (SOM-Cre, JAX#013044) mice were anesthetized with ~1.5% isoflurane in oxygen (flow rate of 1L/min). A craniotomy (~0.5 mm diameter) was made at 0.1 mm posterior to bregma, 1.3 mm lateral to midline. For anterograde axon tracing, 300–400 nL of AAV (serotype 2) expressing Cre-dependent mCherry (AAV2-EF1α-FLEX-mCherry) was stereotactically injected into the BF (5.2 mm from brain surface) using Nanoject II (Drummond Scientific, Broomall, PA) via a micro pipette. The following steps were taken to minimize virus leaking into the injection track: (1) The pipette opening was minimized (<20 μm); (2) The injector was mounted onto a motorized manipulator to ensure slow and smooth retraction; (3) The injection started 5 min after pipette insertion, and multiple 23 or 40 nl injections (13 nl/s) were made at 15–30 s intervals. The pipette was retracted 10 min after injection.

For transsynaptic retrograde tracing, 200–300 nl of helper AAV (AAV2-EF1α-FLEX-eGFP-2a-TVA and AAV2-EF1α-FLEX-RG mixed at 1:1 ratio of viral particles) was injected into the BF using the same procedure as described above. Two to three weeks after helper AAV injection, RVΔG-tdTomato+EnvA was injected into the same location. To further ensure localized virus expression, the helper AAV injection pipette was tilted at 20 degrees from vertical while RV injection pipette was inserted vertically in the majority of experiments.

Tissue processing

Brain tissue was processed according to standard procedures. In brief, two to three weeks after AAV injection (for anterograde tracing) or one week after RV injection (for retrograde tracing), mice were deeply anesthetized with isoflurane and immediately perfused intracardially with ~15 ml of phosphate-buffered saline (PBS) (pH 7.2) followed by ~15 ml of 4% paraformaldehyde (PFA) in PBS. Brain tissue was carefully removed, post-fixed in 4% PFA in PBS at 4°C overnight, dehydrated in 30% sucrose in PBS for 48 hr, and embedded in Tissue Freezing Medium (Triangle Biomedical Sciences, Cincinnati, OH). Brains were cut in 30 or 50 µm coronal sections using a cryostat (Thermo Scientific, Waltham, MA) and mounted with VECTASHIELD mounting medium with DAPI (Vector Laboratories, Burlingame, CA) or DAPI Fluoromount-G (Southern Biotech, Birmingham, AL). One out of every three sections were imaged using 20X/0.75 objective in a high-throughput slide scanner (Nanozoomer-2.0RS, Hamamatsu, Japan) for further processing. We also imaged selected brain regions (Figures 2 and 3) using a Zeiss (Germany) inverted AxioObserver Z1 fully motorized microscope with LSM 710 confocal scanhead, 10X/0.3 EC Plan Neofluar M277 objective or a 20X/0.8 Plan Apochromat M27 objective.

Immunostaining

To check for cell-type specific expression of rabies glycoprotein, tdTomato transgenic reporter mice (JAX#007914) were crossed to Cre-transgenic mice for each cell type and double transgenic offspring were injected with AAV2-EF1α-FLEX-RG. Alternatively, Cre-transgenic mice were injected with AAV2-EF1α-FLEX-mCherry and AAV2-EF1α-FLEX-RG.

After making coronal sections, brain slices were washed in PBS (3 x 10 min., room temperature), blocked with mouse IgG blocking reagent (Mouse on Mouse (M.O.M.) Kit, Vector Laboratories, Burlingame, CA) for 2 hr at room temperature, incubated with mouse anti-rabies glycoprotein (clone 24-3F-10, EMD Millipore, Billerica, MA) with M.O.M. protein concentrate in PBST (PBS + 0.3% Triton-X100) for 18 hr at room temperature, washed in PBST (3 x 20 min., room temperature), incubated with Alexa-Fluor 488 or 647 donkey anti-mouse (ThermoFisher Scientific, Waltham, MA) with M.O.M protein concentrate in PBST for 3 hr at room temperature, and finally washed in PBST (3 x 10 min, room temperature) prior to mounting the slides.

3D reconstruction and quantification

A software package was developed in Matlab to analyze the digitized brain images. The analysis software consists of three modules: image registration, signal detection, and quantification/visualization.

Registration module

The registration module is a reference point-based image alignment software used to align images of brain sections to the Allen Mouse Brain Atlas for further quantification and 3D reconstruction. First, we manually selected a set of reference points in both the atlas and the brain image. The module then applied several geometric transformations (translation, rotation and scaling) of the brain section to optimize the match of the reference points between the brain image and the atlas. Since histological sectioning can sometimes cause tissue compression, we allowed the scaling factors along the dorsal-ventral and medial-lateral axes to be optimized independently. Following the transformation, the match between the image and the atlas was inspected, and further adjustments were made manually if necessary. The main purpose of the manual adjustment was to correct errors generated by the registration procedure due the imperfect brain slice preparations, and it was mostly performed by research assistants not involved in the research design and unaware of the final conclusion of the study.

Detection module

The detection module has two independent sub-modules designed for counting RV-labeled cells and detecting axons, respectively. The cell counting module records the position of manually identified tdTomato-labeled neurons in each digitized brain sections image.

For axon detection, the ridge detection method was used (http://en.wikipedia.org/wiki/Ridge_detection). The following steps were taken to maximize the detection accuracy: (1) Image ridges were computed across multiple scales to extract all possible axon-like signals from each image. In the resulting binary ‘ridge image’, the number of pixels occupied by each detected axon depends on the length but not the thickness of the axon. In addition to valid axons, the ridge image also contains many noise pixels. (2) To remove the noise pixels due to the general background in the fluorescence image, we set a threshold based on the intensity distribution of the original image, and use this as a mask to remove the noise pixels in the ridge image obtained from step (1). (3) To remove the discrete noise pixels with fluorescence intensities higher than the general background (thus not removed by step 2), we first identified pixels that are spatially contiguous in the ridge image, computed the size of each contiguous region, and removed the regions below a threshold size. Steps 2 and 3 were repeated until satisfactory detection results were achieved. (4) The results were then visually inspected and the remaining noise pixels, which were mostly artifacts introduced during brain tissue processing, were rejected manually.

Quantification/visualization module

After detection and registration, signals were quantified across the whole brain and projected to the 3D reference atlas for better visualization. The 3D viewer plug-in of the ImageJ software was used to animate the final 3D model.

The atlas, 3D reference mouse brain, quantification ontology, and layouts for sunburst plot were obtained from the open online resource of Allen Institute for Brain Science, licensed under the Apache License (Version 2.0). Since the number of labeled neurons or axons varied across brains, the input from each region was quantified by dividing the number of labeled neurons found in that region by the total number of labeled neurons detected >850 μm from the injection site (see Figure 1—figure supplement 2). The output (axon projection) to each region was quantified as the number of pixels occupied by detected axons in the cleaned ridge image (Oh et al., 2014) (see Detection module above) divided by the total number of axon-occupied pixels found in the entire brain (after excluding the injection site and locations with known major fiber tracks).

Starter cell mapping

Starter cells were manually identified from the colocalization of tdTomato and eGFP signals using Nanozoomer images scanned at multiple focal planes. The starter cells were marked using the Cell Counter ImageJ plug-in, and registered to the Allen Brain Reference Atlas as described above. A starter cell heat map was generated in Python by calculating the normalized starter cell density for all samples from each cell type and applying bicubic interpolation. For each coronal section image, the cell density was binned from an anterior-posterior range of 0.24 mm, centered at the listed brain slice coordinate (Figure 1—figure supplement 3).

Slice recording

To validate the synaptic input from the prefrontal cortex to BF cholinergic neurons (as shown by RV-mediated input tracing), ChR2 was expressed in excitatory neurons in the prefrontal cortex of ChAT-eGFP mice (JAX#007902, P16-P18) by injecting ~500 nl of AAV-DJ-CaMKIIα-hChR2-eYFP (~1013 gc/ml, Stanford Gene Vector and Virus Core, Stanford, CA) into the orbital and agranular insular areas of the PFC (2.0 mm anterior to bregma, 1.5 mm lateral, 2.0 mm from brain surface) and recording from eGFP+ BF neurons. To validate the synaptic input from ACB, Cre-inducible ChR2-eYFP was expressed in GAD2-Cre mice (JAX#010802, P16-P18) by injecting 300–400 nl of AAV-DJ-EF1α-FLEX-ChR2-eYFP (~1013 gc/ml, Stanford Gene Vector and Virus Core, Stanford, CA) into the ACB (1.5 mm anterior to bregma, 0.8 mm lateral, 3.6 mm from brain surface) and recordings were made (one week after virus injection) from unlabeled BF neurons, which were identified after each recording via single-cell gene-expression analysis. Slice preparation, recording procedure, and single-cell gene-expression analysis were the same as described in a recent study (Xu et al., 2015).

Acknowledgements

We thank K-S Chen, T Lei, D Jeong for the help with histological processing and data analysis, and B Lim and E Callaway for viral vectors and cell lines.

Funding Statement

No external funding was received for this work

Additional information

Competing interests

LL: Reviewing editor, eLife.

The other authors declare that no competing interests exist.

Author contributions

JPD, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article.

MX, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article.

S-HL, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article.

W-CC, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article.

SZ, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article.

SC, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article.

TJY, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article.

JLF, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article.

KM, Analysis and interpretation of data, Drafting or revising the article, Contributed unpublished essential data or reagents.

LL, Analysis and interpretation of data, Drafting or revising the article, Contributed unpublished essential data or reagents.

YD, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article.

Ethics

Animal experimentation: All surgical and experimental procedures were in accordance with the Care and Use of Laboratory Animals of the National Institutes of Health Guide and approved (#R229) by the Animal Care and Use Committees of University of California, Berkeley.

References

  1. Asanuma C. Axonal arborizations of a magnocellular basal nucleus input and their relation to the neurons in the thalamic reticular nucleus of rats. PNAS. 1989;86:4746–4750. doi: 10.1073/pnas.86.12.4746. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Avila I, Lin SC. Distinct neuronal populations in the basal forebrain encode motivational salience and movement. Frontiers in Behavioral Neuroscience. 2014;8:421. doi: 10.3389/fnbeh.2014.00421. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bakin JS, Weinberger NM. Induction of a physiological memory in the cerebral cortex by stimulation of the nucleus basalis. PNAS. 1996;93:11219–11224. doi: 10.1073/pnas.93.20.11219. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Beier KT, Steinberg EE, DeLoach KE, Xie S, Miyamichi K, Schwarz L, Gao XJ, Kremer EJ, Malenka RC, Luo L. Circuit architecture of VTA dopamine neurons revealed by systematic input-output mapping. Cell. 2015;162:622–634. doi: 10.1016/j.cell.2015.07.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Brown RE, Basheer R, McKenna JT, Strecker RE, McCarley RW. Control of sleep and wakefulness. Physiological Reviews. 2012;92:1087–1187. doi: 10.1152/physrev.00032.2011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Callaway EM, Luo L. Monosynaptic circuit tracing with glycoprotein-deleted rabies viruses. Journal of Neuroscience. 2015;35:8979–8985. doi: 10.1523/JNEUROSCI.0409-15.2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Conner JM, Culberson A, Packowski C, Chiba AA, Tuszynski MH. Lesions of the Basal forebrain cholinergic system impair task acquisition and abolish cortical plasticity associated with motor skill learning. Neuron. 2003;38:819–829. doi: 10.1016/S0896-6273(03)00288-5. [DOI] [PubMed] [Google Scholar]
  8. Cullinan WE, Záborszky L. Organization of ascending hypothalamic projections to the rostral forebrain with special reference to the innervation of cholinergic projection neurons. Journal of Comparative Neurology. 1991;306:631–667. doi: 10.1002/cne.903060408. [DOI] [PubMed] [Google Scholar]
  9. DeNardo LA, Berns DS, DeLoach K, Luo L. Connectivity of mouse somatosensory and prefrontal cortex examined with trans-synaptic tracing. Nature Neuroscience. 2015;18:1687–1697. doi: 10.1038/nn.4131. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Everitt BJ, Robbins TW. Central cholinergic systems and cognition. Annual Review of Psychology. 1997;48:649–684. doi: 10.1146/annurev.psych.48.1.649. [DOI] [PubMed] [Google Scholar]
  11. Franklin KBJ, Paxinos G. The Mouse Brain in Stereotaxic Coordinates Compact. 3rd. Academic Press; 2008. [Google Scholar]
  12. Freund TF, Meskenaite V. gamma-Aminobutyric acid-containing basal forebrain neurons innervate inhibitory interneurons in the neocortex. PNAS. 1992;89:738–742. doi: 10.1073/pnas.89.2.738. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Froemke RC, Carcea I, Barker AJ, Yuan K, Seybold BA, Martins AR, Zaika N, Bernstein H, Wachs M, Levis PA, Polley DB, Merzenich MM, Schreiner CE. Long-term modification of cortical synapses improves sensory perception. Nature Neuroscience. 2013;16:79–88. doi: 10.1038/nn.3274. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Grove EA. Efferent connections of the substantia innominata in the rat. Journal of Comparative Neurology. 1988;277:347–364. doi: 10.1002/cne.902770303. [DOI] [PubMed] [Google Scholar]
  15. Grove EA. Neural associations of the substantia innominata in the rat: Afferent connections. Journal of Comparative Neurology. 1988;277:315–346. doi: 10.1002/cne.902770302. [DOI] [PubMed] [Google Scholar]
  16. Hangya B, Ranade SP, Lorenc M, Kepecs A. Central cholinergic neurons are rapidly recruited by reinforcement feedback. Cell. 2015;162:1155–1168. doi: 10.1016/j.cell.2015.07.057. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Hasselmo ME, Sarter M. Modes and models of forebrain cholinergic neuromodulation of cognition. Neuropsychopharmacology. 2011;36:52–73. doi: 10.1038/npp.2010.104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Henny P, Jones BE. Vesicular glutamate (VGlut), GABA (VGAT), and acetylcholine (VACht) transporters in basal forebrain axon terminals innervating the lateral hypothalamus. Journal of Comparative Neurology. 2006;496:453–467. doi: 10.1002/cne.20928. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Hikosaka O. The habenula: from stress evasion to value-based decision-making. Nature Reviews Neuroscience. 2010;11:503–513. doi: 10.1038/nrn2866. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Huang ZJ, Zeng H. Genetic approaches to neural circuits in the mouse. Annual Review of Neuroscience. 2013;36:183–215. doi: 10.1146/annurev-neuro-062012-170307. [DOI] [PubMed] [Google Scholar]
  21. Jones BE, Cuello AC. Afferents to the basal forebrain cholinergic cell area from pontomesencephalic--catecholamine, serotonin, and acetylcholine--neurons. Neuroscience. 1989;31:37–61. doi: 10.1016/0306-4522(89)90029-8. [DOI] [PubMed] [Google Scholar]
  22. Jones BE. Neurobiology of waking and sleeping. Handbook of Clinical Neurology. 2011;98:131–149. doi: 10.1016/B978-0-444-52006-7.00009-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Kim T, Thankachan S, McKenna JT, McNally JM, Yang C, Choi JH, Chen L, Kocsis B, Deisseroth K, Strecker RE, Basheer R, Brown RE, McCarley RW. Cortically projecting basal forebrain parvalbumin neurons regulate cortical gamma band oscillations. PNAS. 2015;112:3535–3540. doi: 10.1073/pnas.1413625112. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Lin SC, Brown RE, Hussain Shuler MG, Petersen CC, Kepecs A. Optogenetic dissection of the basal forebrain neuromodulatory control of cortical activation, plasticity, and cognition. Journal of Neuroscience. 2015;35:13896–13903. doi: 10.1523/JNEUROSCI.2590-15.2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Lin SC, Nicolelis MA. Neuronal ensemble bursting in the basal forebrain encodes salience irrespective of valence. Neuron. 2008;59:138–149. doi: 10.1016/j.neuron.2008.04.031. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Manns ID, Mainville L, Jones BE. Evidence for glutamate, in addition to acetylcholine and GABA, neurotransmitter synthesis in basal forebrain neurons projecting to the entorhinal cortex. Neuroscience. 2001;107:249–263. doi: 10.1016/S0306-4522(01)00302-5. [DOI] [PubMed] [Google Scholar]
  27. McGaughy J, Kaiser T, Sarter M. Behavioral vigilance following infusions of 192 IgG-saporin into the basal forebrain: selectivity of the behavioral impairment and relation to cortical AChE-positive fiber density. Behavioral Neuroscience. 1996;110:247–265. doi: 10.1037/0735-7044.110.2.247. [DOI] [PubMed] [Google Scholar]
  28. Menegas W, Bergan JF, Ogawa SK, Isogai Y, Umadevi Venkataraju K, Osten P, Uchida N, Watabe-Uchida M. Dopamine neurons projecting to the posterior striatum form an anatomically distinct subclass. eLife. 2015;4:e10032. doi: 10.7554/eLife.10032. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Mesulam MM, Mufson EJ. Neural inputs into the nucleus basalis of the substantia innominata (Ch4) in the rhesus monkey. Brain. 1984;107 (Pt 1):253–274. doi: 10.1093/brain/107.1.253. [DOI] [PubMed] [Google Scholar]
  30. Miyamichi K, Amat F, Moussavi F, Wang C, Wickersham I, Wall NR, Taniguchi H, Tasic B, Huang ZJ, He Z, Callaway EM, Horowitz MA, Luo L. Cortical representations of olfactory input by trans-synaptic tracing. Nature. 2011;472:191–196. doi: 10.1038/nature09714. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Miyamichi K, Shlomai-Fuchs Y, Shu M, Weissbourd BC, Luo L, Mizrahi A. Dissecting local circuits: parvalbumin interneurons underlie broad feedback control of olfactory bulb output. Neuron. 2013;80:1232–1245. doi: 10.1016/j.neuron.2013.08.027. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Narayan S, Barnard RJ, Young JA. Two retroviral entry pathways distinguished by lipid raft association of the viral receptor and differences in viral infectivity. Journal of Virology. 2003;77:1977–1983. doi: 10.1128/JVI.77.3.1977-1983.2003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Nguyen DP, Lin SC. A frontal cortex event-related potential driven by the basal forebrain. eLife. 2014;3:e02148. doi: 10.7554/eLife.02148. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Ogawa SK, Cohen JY, Hwang D, Uchida N, Watabe-Uchida M. Organization of monosynaptic inputs to the serotonin and dopamine neuromodulatory systems. Cell Reports. 2014;8:1105–1118. doi: 10.1016/j.celrep.2014.06.042. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Oh SW, Harris JA, Ng L, Winslow B, Cain N, Mihalas S, Wang Q, Lau C, Kuan L, Henry AM, Mortrud MT, Ouellette B, Nguyen TN, Sorensen SA, Slaughterbeck CR, Wakeman W, Li Y, Feng D, Ho A, Nicholas E, Hirokawa KE, Bohn P, Joines KM, Peng H, Hawrylycz MJ, Phillips JW, Hohmann JG, Wohnoutka P, Gerfen CR, Koch C, Bernard A, Dang C, Jones AR, Zeng H. A mesoscale connectome of the mouse brain. Nature. 2014;508:207–214. doi: 10.1038/nature13186. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Osten P, Margrie TW. Mapping brain circuitry with a light microscope. Nature Methods. 2013;10:515–523. doi: 10.1038/nmeth.2477. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Parent A, Paré D, Smith Y, Steriade M. Basal forebrain cholinergic and noncholinergic projections to the thalamus and brainstem in cats and monkeys. Journal of Comparative Neurology. 1988;277:281–301. doi: 10.1002/cne.902770209. [DOI] [PubMed] [Google Scholar]
  38. Paré D, Smith Y. GABAergic projection from the intercalated cell masses of the amygdala to the basal forebrain in cats. Journal of Comparative Neurology. 1994;344:33–49. doi: 10.1002/cne.903440104. [DOI] [PubMed] [Google Scholar]
  39. Pollak Dorocic I, Fürth D, Xuan Y, Johansson Y, Pozzi L, Silberberg G, Carlén M, Meletis K. A whole-brain atlas of inputs to serotonergic neurons of the dorsal and median raphe nuclei. Neuron. 2014;83:663–678. doi: 10.1016/j.neuron.2014.07.002. [DOI] [PubMed] [Google Scholar]
  40. Rye DB, Wainer BH, Mesulam MM, Mufson EJ, Saper CB. Cortical projections arising from the basal forebrain: a study of cholinergic and noncholinergic components employing combined retrograde tracing and immunohistochemical localization of choline acetyltransferase. Neuroscience. 1984;13:627–643. doi: 10.1016/0306-4522(84)90083-6. [DOI] [PubMed] [Google Scholar]
  41. Saper CB, Fuller PM, Pedersen NP, Lu J, Scammell TE. Sleep state switching. Neuron. 2010;68:1023–1042. doi: 10.1016/j.neuron.2010.11.032. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Sarter M, Givens B, Bruno JP. The cognitive neuroscience of sustained attention: where top-down meets bottom-up. Brain Research Reviews. 2001;35:146–160. doi: 10.1016/S0165-0173(01)00044-3. [DOI] [PubMed] [Google Scholar]
  43. Schliebs R, Arendt T. The cholinergic system in aging and neuronal degeneration. Behavioural Brain Research. 2011;221:555–563. doi: 10.1016/j.bbr.2010.11.058. [DOI] [PubMed] [Google Scholar]
  44. Semba K, Reiner PB, McGeer EG, Fibiger HC. Brainstem afferents to the magnocellular basal forebrain studied by axonal transport, immunohistochemistry, and electrophysiology in the rat. Journal of Comparative Neurology. 1988;267:433–453. doi: 10.1002/cne.902670311. [DOI] [PubMed] [Google Scholar]
  45. Semba K. Multiple output pathways of the basal forebrain: organization, chemical heterogeneity, and roles in vigilance. Behavioural Brain Research. 2000;115:117–141. doi: 10.1016/S0166-4328(00)00254-0. [DOI] [PubMed] [Google Scholar]
  46. Steriade M, McCarley RW. Brain Control of Wakefulness and Sleep. 2nd ed. New York: Springer; 2005. [DOI] [Google Scholar]
  47. Wall NR, De La Parra M, Callaway EM, Kreitzer AC. Differential innervation of direct- and indirect-pathway striatal projection neurons. Neuron. 2013;79:347–360. doi: 10.1016/j.neuron.2013.05.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Watabe-Uchida M, Zhu L, Ogawa SK, Vamanrao A, Uchida N. Whole-brain mapping of direct inputs to midbrain dopamine neurons. Neuron. 2012;74:858–873. doi: 10.1016/j.neuron.2012.03.017. [DOI] [PubMed] [Google Scholar]
  49. Weissbourd B, Ren J, DeLoach KE, Guenthner CJ, Miyamichi K, Luo L. Presynaptic partners of dorsal raphe serotonergic and GABAergic neurons. Neuron. 2014;83:645–662. doi: 10.1016/j.neuron.2014.06.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Wenk GL. The nucleus basalis magnocellularis cholinergic system: one hundred years of progress. Neurobiology of Learning and Memory. 1997;67:85–95. doi: 10.1006/nlme.1996.3757. [DOI] [PubMed] [Google Scholar]
  51. Whitehouse PJ, Price DL, Struble RG, Clark AW, Coyle JT, Delon MR. Alzheimer's disease and senile dementia: loss of neurons in the basal forebrain. Science. 1982;215:1237–1239. doi: 10.1126/science.7058341. [DOI] [PubMed] [Google Scholar]
  52. Wickersham IR, Lyon DC, Barnard RJ, Mori T, Finke S, Conzelmann KK, Young JA, Callaway EM. Monosynaptic restriction of transsynaptic tracing from single, genetically targeted neurons. Neuron. 2007;53:639–647. doi: 10.1016/j.neuron.2007.01.033. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Xu M, Chung S, Zhang S, Zhong P, Ma C, Chang WC, Weissbourd B, Sakai N, Luo L, Nishino S, Dan Y. Basal forebrain circuit for sleep-wake control. Nature Neuroscience. 2015;18:1641–1647. doi: 10.1038/nn.4143. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Yang C, McKenna JT, Zant JC, Winston S, Basheer R, Brown RE. Cholinergic neurons excite cortically projecting basal forebrain GABAergic neurons. Journal of Neuroscience. 2014;34:2832–2844. doi: 10.1523/JNEUROSCI.3235-13.2014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Zaborszky L, Cullinan WE. Projections from the nucleus accumbens to cholinergic neurons of the ventral pallidum: a correlated light and electron microscopic double-immunolabeling study in rat. Brain Research. 1992;570:92–101. doi: 10.1016/0006-8993(92)90568-T. [DOI] [PubMed] [Google Scholar]
  56. Zaborszky L, Duque A. Local synaptic connections of basal forebrain neurons. Behavioural Brain Research. 2000;115:143–158. doi: 10.1016/S0166-4328(00)00255-2. [DOI] [PubMed] [Google Scholar]
  57. Zaborszky L, Gaykema RP, Swanson DJ, Cullinan WE. Cortical input to the basal forebrain. Neuroscience. 1997;79:1051–1078. doi: 10.1016/S0306-4522(97)00049-3. [DOI] [PubMed] [Google Scholar]
  58. Zaborszky L, van den Pol A, Gyengesi E. Mouse Nervous System. 2012. The basal forebrain cholinergic projection system in mice; pp. 684–718. [DOI] [Google Scholar]
eLife. 2016 Sep 19;5:e13214. doi: 10.7554/eLife.13214.020

Decision letter

Editor: Gary L Westbrook1

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

[Editors’ note: this article was originally rejected after discussions between the reviewers, but the authors were invited to resubmit after an appeal against the decision.]

Thank you for submitting your work entitled "Cell Type-Specific Long-Range Connections of Basal Forebrain Circuit" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Gary Westbrook as the Senior Editor. Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

The reviewers appreciated the importance of the work but raised important issues regarding the specificity of the viral approach used. The main issue raised, as seen in Figure 1—figure supplement 1, is that the virus(es) infected cells without Cre, i.e. they were leaky. Therefore, it is unclear how much of the mapping is specific. Besides quantification and better discussion of the extent of the leakage, the reviewers suggest that the study may be more suitable for publication if the mapping could be done with a more specific virus. Alternatively, or in addition, each proposed projection could be validated by expressing ChR2 in the presynaptic cell, and recording from the genetically defined postsynaptic cell. Either of these experiments would require a substantial amount of work in our opinion, and therefore, as per eLife's policy, the study cannot be accepted.

Reviewer #1:

Do and colleagues perform a whole-brain mapping of both the inputs and outputs of the basal forebrain across its four major cell types using rabies virus. Understanding the connectivity of the basal forebrain is important and the new viral techniques used provide an unprecedented opportunity for revealing the cell-type-specific input-output logic of basal forebrain. These experiments required a lot of careful work and data analysis. The major take home message is that the inputs to all cell types are qualitatively similar but the output projections showed marked differences.

I was very enthusiastic about this paper until I got to supplementary figure 1, the control experiment. What it shows is that the expression of rabiesdG-tdTomato+EnvA can occur without TVA or there is leaky expression of TVA without Cre (Figure 1B). Figure 1C shows that about half of the rabiesdG-tdTomato+EnvA is not colocalized with AAV-TVA-GFP. These results indicate a severely leaky expression that causes non-specific infection of rabies in all cell types. The authors are to be commended for showing these controls. Clearly they realize their importance and try to focus on the relative fraction of inputs between cell-types. Unfortunately, however, having a mixed starter population can alone explain why the inputs of different cell types look the same, bringing the entire study into question.

Other comments:

1) The authors report that subsets of ChAT cells in the basal forebrain may project to or receive projections from different brain regions. Figure 1A showed the targeting injection site is a bit posterior and lateral compared to the usual coordinates of HDB, which may lead to only partial infection of HDB. In addition, the infected ChAT and PV neurons shown in Figure 1A are significantly fewer in number than the total number of ChAT and PV neurons in HDB. The author should stain for ChAT, PV, SOM, vGlut2 and quantify whether the infection includes at least most of the HDB neurons or just part of it.

2) A statement should be made with regard to whether the morphological analyses were made with the experimenter blinded to the treatment condition. This is especially important since all the quantification procedures from registration and detection module to cell counting are all manually performed or adjusted.

3) One particularly important piece of data in the paper is in Figure 2—figure supplement 1 showing physiological evidence for direct prefrontal inputs to cholinergic neurons. I think the authors should expand on these data and make into a full figure. There were not enough details provided to evaluate the experiment fully. The legend states the input is from "prefrontal cortex". Do the authors mean mPFC? Their rabies tracing data shows surprisingly little input from mPFC.

Reviewer #2:

In this study, the authors perform a detailed analysis of brain regions that input onto and receive outputs from four genetically defined cell-types in the basal forebrain. Using rabies-based labeling of monosynaptic inputs, they conclude that four major cell-types in the basal forebrain – PV+, SOM+, ChAT+, and VGLUT2+ neurons – receive qualitatively similar inputs, with striatum providing the main source of input. In addition, they report divergent outputs of each of these four cell-types as assessed by labeling of axons with an mCherry-expressing AAV, with ChAT+ neurons primarily projecting to the cortex, olfactory bulb, and hippocampus, PV+ neurons projecting to hypothalamus and pallidum, and both VGLUT2+ and SOM+ neurons forming reciprocal connections with the striatum and hypothalamus.

The analysis described in this study appear carefully done, and appropriate controls are shown. While the authors offer little in the way of interpretation, they have created a valuable, interactive resource that will be of great interest to other researchers interested in the basal forebrain.

There are several issues that should be addressed prior to publication. First, while the website is a valuable resource and welcome interactive supplement to the findings of the paper, there are some apparent discrepancies between the stated percentage values and the associated sunburst plot that either need to be addressed in the code or explained with better documentation or associated legends. For example, in the graphic showing inputs to ChAT cells, the "Basic cell groups and regions" lists 98.55% total inputs (it is understood from the legend of Figure 3—figure supplement 1 that some inputs come from unnamed sub-regions, though this could be mentioned in the main text). However, The 2nd level of the sunburst plot only includes "Brain stem" at 28.97% and "Cerebrum" at 64.80% – which sum to only 93.77%, not 98.44. As result, subsequent rings represent much more of the sunburst plot than the percentage would indicate – e.g. Inputs to ChAT neurons from Cerebral nuclei make up only 42.03% of the inputs, but well more than half of the sunburst plot. At a minimum this discrepancy needs to be better explained. Ideally, the sunburst plot would include an "un-named region" category, so that the graphic matches the percentage.

Perhaps the most striking finding of the paper is the major input to all cell-types from the striatum. To my knowledge, a major input from the striatum to the basal forebrain has not been previously described. Could this be the result of hitting starter cells in the pallidum during the rabies-assisted retrograde tracing? Confirming functional connectivity between striatal projection neurons and basal forebrain neurons of all types, as in Figure 2—figure supplement 1) would be very interesting and provide a necessary control to show that neurons of the BF proper also receive striatal input. This point needs to be addressed with electrophysiology and complete reconstruction as it would represent a radical change in our understanding of the BG wiring.

There is a great degree of variability in the number of labeled presynaptic neurons in different brains. Does this correlate in any way with the number of starter cells in each mouse? Greater analysis of the number and location of starter cells for each experiment would be useful.

Finally, the manuscript needs a greater discussion on the functional implications of the connectivity patterns discovered here, and the potential ways they do or do not meet expectations based on previous functional data. For example, the authors cite a recent paper showing PV+ BF neurons entrain cortical γ oscillations (Kim et al., 2015), but they show very little projection to the cortex from this population. The authors cite a recent paper (Hangya et al., 2015) that describes a recruitment of BF ChAT+ neurons by reinforcement with very long latency (~50 ms). Can the authors identify from their data an input that might explain this?

Reviewer #3:

Do et al. revealed input and output neural pathway to four different types of neurons (cholinergic, glutamatergic, parvalbumin (PV) and somatostain (SOM)) in the basal forebrain (BF) using Cre mice and transsynaptic retrograde tracing. They claimed that input pathway to four different cell types was similar but output pathway was different. Although the results were clear and informative to understand the physiological role of BF neurons, the following points need to be revised before publication.

Major points:

1) The author used cell type specific Cre recombinase-expressing mice (ChAT-Cre, VGLUT2-Cre, PV-Cre and SOM-Cre) to restrict starter cell for retrograde transsynaptic tracing. However, there is no information about specificity of Cre expression in each mouse line. How much Cre expressing neurons and which area of Cre expressing neurons were infected by rabies virus as starter cell should be provided.

2) Figure 2—figure supplement 1, the authors showed functional connection between prefrontal cortical (PFC) neurons and cholinergic neurons in BF. However, there is no information about feature of neurons in PFC labeled by retrograde transsynaptic tracing. In spite of this, the author used CaMKIIa promoter, which drives gene expression in glutamatergic neurons. Infected area of ChR2 expressing AAV in the PFC should be provided as well. Additionally, it is also important to show that brain area which is not innervate BF neurons from transsynaptic tracing is not functionally connected by expressing ChR2. This figure is referred only in the Discussion.

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for resubmitting your work entitled "Cell Type-Specific Long-Range Connections of Basal Forebrain Circuit" for further consideration at eLife. We apologize that this review took longer than is our goal. Your revised article has been favorably evaluated by Gary Westbrook (Senior editor), a Reviewing editor, and two reviewers, a reviewer from the original version of your manuscript and a new reviewer. The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below.

Summary:

Overall this is nice study showing differential outputs, but largely similar inputs, from/to four different neurotransmitter-defined cell groups in the basal forebrain. The authors use optogenetics in slice recordings to confirm some of the connections identified using monosynaptic tracing with rabies virus, including a connection (from PFC to cholinergic BF neurons) that was not reported in a previous study using electron microscopy. The authors have done a good job of making the data available as a resource for subsequent investigations. The authors have made changes to the manuscript and it is definitely improved. The data showing direct cortical projections to all basal forebrain cell-types (Figure 4) is important and nice.

However there are still concerns about the controls for cell-type specificity that need to be addressed before a decision on publication can be made. Note that points 2 and 3 are overlapping and if the control experiment works it is one experiment that shows specific expression of RG along with the cell-type-specific markers as we already know TVA is leaky.

1) The control experiments described in Supplementary figure 1 are good controls to do, and the ones in wild-type mice are fine (and important). For the controls in Cre mice, however, the authors use the wrong mouse line, for no apparent reason. The authors state that the use of GAD2-Cre mice "is likely to cause an overestimate of the exclusion zone"; this is plausible in the case of the PV-Cre and SOM-Cre lines, but not necessarily in the case of the Vglut2-Cre ones. It is known that EnvA-enveloped rabies virus can directly (non-transsynaptically) retrogradely infect TVA-expressing neurons projecting to an injection site (Huang & Hantman 2013). In the worst case in the present study, direct retrograde infection of neurons projecting to the injection site by the TVA AAV could allow direct retrograde infection of them by the RV, in the absence of G expression. The authors should redo these controls (e.g., omitting only the RV G AAV) in the four Cre lines used for the rest of the paper. This would be very easy to do and take little time but provide a much better set of controls.

2) Regarding cell-type specificity of their starter population, the references cited show nearly 100% co-localization of starter cell (TVA+ & RVG+) with Cre. For instance, Beier et al. (Cell 2015) Figure 1B and 1P shows that nearly all starter cells (yellow) are colocalized with TH. Those authors also used an anti-rabies glycoprotein to show specificity for RG expression. Watanabe-Uchida (Neuron 2012) also show that TVA expression overlaps with TH ~97%. The authors could could prove cell-type-specificity, for instance by performing immunos against one cell-type and rabies. They could show that RG expression is not leaky by injecting into wildtype mice. Perhaps better would be to inject both TC, RG and RVG into WT mice to demonstrate that they find no long-range projections.

3) To prove the cell type specificity of starter cells, as shown in all the references they mention, the authors need to do the same standard experiment. Immunostain for Chat in Chat-Cre transgenic mouse which is injected by AAV-FLEX-eGFP-2a-TVA, AAV-FLEX-RG and Rabies-dG-tdTomato+EnvA in HDB, then shows the percentage of co-localization of starter cell (yellow, eGFP-TVA positive & Rabies-tdTomato positive) with Chat immuno signal. Similar strategy for other cre lines. If the author can show a high percentage of colocalization, then it is done. But if the severe leaky of TVA cause a mismatch between starter cells and Chat immunosignal, then they need to prove the cell type specificity of glycoprotein G. To prove the cell type specificity of glycoprotein G it would be good to show that immunostaining signal for G is not in other cell types, but specifically in Chat neurons of Chat-cre mice and no signal in WT mice. Similar strategy for other Cre lines.

4) The concern about whether the variability of presynaptic neurons is correlated with the number of starter cells is not solved yet. There isn't much that can be done about this but it should be mentioned in the paper.

eLife. 2016 Sep 19;5:e13214. doi: 10.7554/eLife.13214.021

Author response

[Editors’ note: the author responses to the first round of peer review follow.]

Do and colleagues perform a whole-brain mapping of both the inputs and outputs of the basal forebrain across its four major cell types using rabies virus. Understanding the connectivity of the basal forebrain is important and the new viral techniques used provide an unprecedented opportunity for revealing the cell-type-specific input-output logic of basal forebrain. These experiments required a lot of careful work and data analysis. The major take home message is that the inputs to all cell types are qualitatively similar but the output projections showed marked differences.

I was very enthusiastic about this paper until I got to supplementary figure 1, the control experiment. What it shows is that the expression of rabiesdG-tdTomato+EnvA can occur without TVA or there is leaky expression of TVA without Cre (Figure 1B). Figure 1C shows that about half of the rabiesdG-tdTomato+EnvA is not colocalized with AAV-TVA-GFP. These results indicate a severely leaky expression that causes non-specific infection of rabies in all cell types. The authors are to be commended for showing these controls. Clearly they realize their importance and try to focus on the relative fraction of inputs between cell-types. Unfortunately, however, having a mixed starter population can alone explain why the inputs of different cell types look the same, bringing the entire study into question.

We appreciate the reviewer’s comment and apologize for not explaining this point clearly in our previous submission, which gave rise to the current confusion. The major concern is whether our input mapping was specific to each targeted cell type, given rabies infection of cells without Cre at the injection site (Figure 1—figure supplement 1B). In fact, this issue is common to most of the published studies using rabies (RVdG)-based input tracing (Beier et al., 2015; Menegas et al., 2015; Miyamichi et al., 2013; Ogawa et al., 2014; Pollak Dorocic et al., 2014; Wall et al., 2013; Watabe-Uchida et al., 2012; Weissbourd et al., 2014), but the potential contamination of traced input neurons can be minimized as long as the results are interpreted based on proper control experiments, which are shown in our Figure 1—figure supplement 1. As demonstrated in previous studies, specificity of the starter cells (which must express both TVA and RG to allow rabies virus infection and transsynaptic spread) is ensured by the highly specific Cre/loxP recombination system (Watabe-Uchida et al., 2012). However, due to the extremely efficient interaction between TVA and EnvA-pseudotyped rabies virus, the very low-level expression of TVA in non-Cre-expressing cell types (not detectable based on fluorescent protein markers) allows rabies virus to infect these cells and label them with tdTomato at the injection site. Importantly, this issue does not compromise the mapping of long-range inputs because, due to the highly specific Cre/loxP recombination system, RG (required for transsynaptic spread of RV) is not expressed in any non-Cre-expressing cells at sufficient levels for trans-complementation of RVdG to allow transsynaptic spread of RV(Callaway and Luo, 2015; Miyamichi et al., 2013). Thus, while this rabies-based method should not be used for tracing local inputs near the injection site (excluded in our study), the standard practice for tracing long-range inputs is to determine the spatial extent of the local contamination using the control experiment without RG and exclude this region when analyzing the long-range inputs. This is exactly the purpose of the control experiments shown in our Figure 1—figure supplement 1, which are similar to or more rigorous than those performed by previous studies using the same approach (Beier et al., 2015; Pollak Dorocic et al., 2014; Wall et al., 2013; Weissbourd et al., 2014). We have now modified the text to explain this issue clearly (third paragraph of Results section).

Other comments:

1) The authors report that subsets of ChAT cells in the basal forebrain may project to or receive projections from different brain regions. Figure 1A showed the targeting injection site is a bit posterior and lateral compared to the usual coordinates of HDB, which may lead to only partial infection of HDB. In addition, the infected ChAT and PV neurons shown in Figure 1A are significantly fewer in number than the total number of ChAT and PV neurons in HDB. The author should stain for ChAT, PV, SOM, vGlut2 and quantify whether the infection includes at least most of the HDB neurons or just part of it.

For all the tracing experiments, we targeted the caudal portion of the BF, including the HDB, magnocellular preoptic nucleus, and substantia innominata (Materials and methods, subsection “Surgery and viral injections”). To address the reviewer’s question, we computed the average number of starter cells per brain sample for each cell type and compared it with the total cell number of that type within these BF nuclei (based on the in situ hybridization results from the Allen Mouse Brain Atlas). We found that the starter cell percentages are similar across the four cell types (ChAT, 16.3 ± 6.8%, mean ± s.e.m.; VGLUT2, 13.8 ± 3.1%; PV, 15.4 ± 1.7%; SOM, 17.6 ± 5.5%). As shown in the new Figure 1 —figure supplement 2, the spatial distribution of starter cells was also similar across cell types. Note that although the infection does not include most of the BF neurons of the given type, which is common in RV tracing studies (see, e.g., Ogawa et al., 2014; Watabe-Uchida et al., 2012), it should not affect our main conclusion, which is based on the relative percentage of input from each brain area rather than the absolute number of inputs.

2) A statement should be made with regard to whether the morphological analyses were made with the experimenter blinded to the treatment condition. This is especially important since all the quantification procedures from registration and detection module to cell counting are all manually performed or adjusted.

The purpose of manual adjustment was to eliminate the errors produced by the registration module due to imperfect brain slice conditions. The majority of these analyses were performed by undergraduate research assistants who did not participate in the experimental design, had no knowledge of the whole-brain distributions of the traced inputs and outputs, and unaware of the conclusion of the study. In addition, signal extraction and manual adjustment were performed before the final quantification step (completely automatic), which counts the number of cells found within each brain structure, making it difficult to bias the result deliberately during the manual steps.

We have now included a statement on this in the Materials and methods (subsection “Registration module”).

3) One particularly important piece of data in the paper is in Figure 2—supplement 1 showing physiological evidence for direct prefrontal inputs to cholinergic neurons. I think the authors should expand on these data and make into a full figure. There were not enough details provided to evaluate the experiment fully. The legend states the input is from "prefrontal cortex". Do the authors mean mPFC? Their rabies tracing data shows surprisingly little input from mPFC.

We apologize for omitting some of the details in the original figure legend. The injection coordinate was 2.0 mm anterior to bregma, 1.5 mm lateral to midline and 2.0 mm from brain surface (Materials and methods, subsection “Slice recording”), which corresponds to orbital frontal/ insular areas of the PFC; this location was chosen based on our retrograde tracing results (Figure 2). We have now moved this result to the new Figure 4 and added more electrophysiology experiments.

Reviewer #2:

In this study, the authors perform a detailed analysis of brain regions that input onto and receive outputs from four genetically defined cell-types in the basal forebrain. Using rabies-based labeling of monosynaptic inputs, they conclude that four major cell-types in the basal forebrain – PV+, SOM+, ChAT+, and VGLUT2+ neurons – receive qualitatively similar inputs, with striatum providing the main source of input. In addition, they report divergent outputs of each of these four cell-types as assessed by labeling of axons with an mCherry-expressing AAV, with ChAT+ neurons primarily projecting to the cortex, olfactory bulb, and hippocampus, PV+ neurons projecting to hypothalamus and pallidum, and both VGLUT2+ and SOM+ neurons forming reciprocal connections with the striatum and hypothalamus.

The analysis described in this study appear carefully done, and appropriate controls are shown. While the authors offer little in the way of interpretation, they have created a valuable, interactive resource that will be of great interest to other researchers interested in the basal forebrain.

There are several issue that should be addressed prior to publication. First, while the website is a valuable resource and welcome interactive supplement to the findings of the paper, there are some apparent discrepancies between the stated percentage values and the associated sunburst plot that either need to be addressed in the code or explained with better documentation or associated legends. For example, in the graphic showing inputs to ChAT cells, the "Basic cell groups and regions" lists 98.55% total inputs (it is understood from the legend of Figure 3—figure supplement 1 that some inputs come from unnamed sub-regions, though this could be mentioned in the main text). However, The 2nd level of the sunburst plot only includes "Brain stem" at 28.97% and "Cerebrum" at 64.80% – which sum to only 93.77%, not 98.44. As result, subsequent rings represent much more of the sunburst plot than the percentage would indicate – e.g. Inputs to ChAT neurons from Cerebral nuclei make up only 42.03% of the inputs, but well more than half of the sunburst plot. At a minimum this discrepancy needs to be better explained. Ideally, the sunburst plot would include an "un-named region" category, so that the graphic matches the percentage.

We thank the reviewer for pointing out this apparent discrepancy. The visualization part of code was from the open source project of the Allen Institute for Brain Science. The mismatch between the section size and the actual value was due to the fact that there are unnamed regions in multiple brain structures. As the reviewer pointed out, the existence of unnamed region caused the percentage in parent structure to be larger than the sum of percentages from children structures in some cases (this was explained in the table legend of Figure 3 – source data 1). The original visualization code calculated the sector size by summing the signals only from the named children structures and ignoring the un-named regions, which causes a mismatch in the actual percentage and the sector size. Since this representation is likely to cause confusion and apparent discrepancy as the reviewer pointed out, we have modified the code such that a white sector was included in each level of the sunburst to represent signals from un-named regions, and the sector size matches the percentage of each structure in the new version.

Perhaps the most striking finding of the paper is the major input to all cell-types from the striatum. To my knowledge, a major input from the striatum to the basal forebrain has not been previously described. Could this be the result of hitting starter cells in the pallidum during the rabies-assisted retrograde tracing? Confirming functional connectivity between striatal projection neurons and basal forebrain neurons of all types, as in Figure 2 – supplement 1) would be very interesting and provide a necessary control to show that neurons of the BF proper also receive striatal input. This point needs to be addressed with electrophysiology and complete reconstruction as it would represent a radical change in our understanding of the BG wiring.

Our finding that BF cholinergic neurons receive inputs from the ventral part of the striatum (ACB) is consistent with a previous ultrastructural tracing study (Zaborszky and Cullinan, 1992). To further confirm that the ACB indeed provides input to the BF rather than other pallidum regions (due to virus leakage into these regions in the rabies experiment), we injected Cre-dependent AAV expressing ChR2 into ACB of GAD2-Cre mice and measured light-evoked postsynaptic response in the caudal portion of the BF (including the HDB, magnocellular preoptic nucleus, and substantia innominata). We found that all four BF cell types receive GABAergic ACB inputs, and these results are now included in Figure 4.

There is a great degree of variability in the number of labeled presynaptic neurons in different brains. Does this correlate in any way with the number of starter cells in each mouse? Greater analysis of the number and location of starter cells for each experiment would be useful.

We have now performed further analyses of the number and location of starter cells in each sample (Figure 1—figure supplement 2, Figure 1—figure supplement 3A; we excluded one of the VGLUT2 brain samples because many starter cells were found outside of the BF). We have also calculated the convergence index (the ratio between the number of inputs and starters, Figure 1—figure supplement 3B), a value related to the transynaptic efficiency of the virus (Callaway et al., 2015). The variability in the convergence index was similar to other rabies tracing studies (Miyamichi et al., 2011; Miyamichi et al., 2014; DeNardo et al., 2015).

Finally, the manuscript needs a greater discussion on the functional implications of the connectivity patterns discovered here, and the potential ways they do or do not meet expectations based on previous functional data. For example, the authors cite a recent paper showing PV+ BF neurons entrain cortical γ oscillations (Kim et al., 2015), but they show very little projection to the cortex from this population. The authors cite a recent paper (Hangya et al., 2015) that describes a recruitment of BF ChAT+ neurons by reinforcement with very long latency (~50 ms). Can the authors identify from their data an input that might explain this?

The projection from BF PV+ GABAergic neurons to the cortex overall is indeed not as dense as the projections to the hypothalamus (Figure 3). In addition to the direct cortical projections, the entrainment of cortical γ oscillations shown by Kim et al. (2015) could also be partly mediated by PV projections to subcortical structures, which may in turn regulate cortical activity. We have now discussed this possibility in the revised text (Second paragraph of Discussion section).

We have also added a discussion on the functional implication of the inputs mapped in this study. In particular, Hangya et al. (2015) reported that BF cholinergic neurons exhibited strong responses to reinforcement signals in an auditory detection task. Our whole-brain mapping of the inputs to BF cholinergic neurons provides a list of candidate regions through which the reinforcement signals may be conveyed to these neurons. We have now added a discussion of this point in the revision (First paragraph of Discussion section).

Reviewer #3:

Do et al. revealed input and output neural pathway to four different types of neurons (cholinergic, glutamatergic, parvalbumin (PV) and somatostain (SOM)) in the basal forebrain (BF) using Cre mice and transsynaptic retrograde tracing. They claimed that input pathway to four different cell types was similar but output pathway was different. Although the results were clear and informative to understand the physiological role of BF neurons, the following points need to be revised before publication.

Major points,

1) The author used cell type specific Cre recombinase-expressing mice (ChAT-Cre, VGLUT2-Cre, PV-Cre and SOM-Cre) to restrict starter cell for retrograde transsynaptic tracing. However, there is no information about specificity of Cre expression in each mouse line. How much Cre expressing neurons and which area of Cre expressing neurons were infected by rabies virus as starter cell should be provided.

The four Cre mouse lines we used in the current experiment were shown to label each BF cell types with high specificity in a previous study (Xu et al., 2015), and this information was described in 1st paragraph of our Results section (p. 5, lines 90-92). As shown in the control experiment (Figure 1—figure supplement 1A), starter cells were found only in Cre mouse lines and not in wild-type controls. Together, these results indicate that the starter cells were restricted to each of the four cell types.

Based on the reviewer’s suggestion, we have provided a new figure summarizing the locations of the starter cells for each cell type (Figure 1—figure supplement 2), showing that the spatial distribution of starter cells was similar across cell types. We also computed the mean number of starter cells for each cell type and compared it with the total cell number of that type within these BF nuclei (based on the in situ hybridization results from the Allen Mouse Brain Atlas). We found that the starter cell percentages are similar across the four cell types (ChAT, 16.3 ± 6.8%, mean ± s.e.m.; VGLUT2, 13.8 ± 3.1%; PV, 15.4 ± 1.7%; SOM, 17.6 ± 5.5%). Note that although the infection does not include most of the BF neurons of the given type, which is common in RV tracing studies (Ogawa et al., 2014; Watabe-Uchida et al., 2012), it should not affect our main conclusion, which is based on the relative percentage of input from each brain area rather than the absolute number of inputs.

2) Figure 2—figure supplement 1, the authors showed functional connection between prefrontal cortical (PFC) neurons and cholinergic neurons in BF. However, there is no information about feature of neurons in PFC labeled by retrograde transsynaptic tracing. In spite of this, the author used CaMKIIa promoter, which drives gene expression in glutamatergic neurons. Infected area of ChR2 expressing AAV in the PFC should be provided as well. Additionally, it is also important to show that brain area which is not innervate BF neurons from transsynaptic tracing is not functionally connected by expressing ChR2. This figure is referred only in the Discussion.

The PFC neurons labeled by retrograde transsynaptic tracing from the BF are most likely to be glutamatergic (expressing CaMKIIα), since the great majority of cortical projection neurons are glutamatergic and this issue was specifically addressed in a previous study (Weissbourd et al., 2014). Based on the reviewer’s suggestion we have also included a figure showing a ChR2- EYFP expressing PFC region (Figure 4—figure supplement 1).

Regarding the brain areas not labeled with transsynaptic tracing, some of them may in fact provide inputs to the BF, but may be missed by RV-mediated tracing, as the RV tracing technique used in the current study does not label all presynaptic neurons (Marshel et al., 2010). While in future studies it would be interesting to determine whether some of the RV-negative brain areas provide inputs to the BF, it is beyond the scope of the current revision, given the vast number of brain areas that could be tested.

[Editors’ note: the author responses to the re-review follow.]

1) The control experiments described in supplementary figure 1 are good controls to do, and the ones in wild-type mice are fine (and important). For the controls in Cre mice, however, the authors use the wrong mouse line, for no apparent reason. The authors state that the use of GAD2-Cre mice "is likely to cause an overestimate of the exclusion zone"; this is plausible in the case of the PV-Cre and SOM-Cre lines, but not necessarily in the case of the Vglut2-Cre ones. It is known that EnvA-enveloped rabies virus can directly (non-transsynaptically) retrogradely infect TVA-expressing neurons projecting to an injection site (Huang…& Hantman 2013). In the worst case in the present study, direct retrograde infection of neurons projecting to the injection site by the TVA AAV could allow direct retrograde infection of them by the RV, in the absence of G expression. The authors should redo these controls (e.g., omitting only the RV G AAV) in the four Cre lines used for the rest of the paper. This would be very easy to do and take little time but provide a much better set of controls.

We thank the reviewer for pointing out the issue with the mouse line used in the control experiments. We initially chose to use GAD2-Cre mice because GAD2+ neurons are the most abundant cell type in the BF, with a much higher density than all four cell types we used in the current study, which we thought would give a safer exclusion zone. We appreciate the reviewer’s concern and have added new control experiments using each of the four Cre lines and omitting only the RG in the injection. We found that non-specific labeling across all lines were largely restricted within the exclusion zone initially set using the GAD2-Cre mice. This result is now included in revised Figure 1—figure supplement 2D.

2) Regarding cell-type specificity of their starter population, the references cited show nearly 100% co-localization of starter cell (TVA+ & RVG+) with Cre. For instance, Beier et al. (Cell 2015) Figure 1B and 1P shows that nearly all starter cells (yellow) are colocalized with TH. Those authors also used an anti-rabies glycoprotein to show specificity for RG expression. Watanabe-Uchida (Neuron 2012) also show that TVA expression overlaps with TH ~97%. The authors could could prove cell-type-specificity, for instance by performing immunos against one cell-type and rabies. They could show that RG expression is not leaky by injecting into wildtype mice. Perhaps better would be to inject both TC, RG and RVG into WT mice to demonstrate that they find no long-range projections

3) To prove the cell type specificity of starter cells, as shown in all the references they mention, the authors need to do the same standard experiment. Immunostain for Chat in Chat-Cre transgenic mouse which is injected by AAV-FLEX-eGFP-2a-TVA, AAV-FLEX-RG and Rabies-dG-tdTomato+EnvA in HDB, then shows the percentage of co-localization of starter cell (yellow, eGFP-TVA positive & Rabies-tdTomato positive) with Chat immuno signal. Similar strategy for other cre lines. If the author can show a high percentage of colocalization, then it is done. But if the severe leaky of TVA cause a mismatch between starter cells and Chat immunosignal, then they need to prove the cell type specificity of glycoprotein G. To prove the cell type specificity of glycoprotein G it would be good to show that immunostaining signal for G is not in other cell types, but specifically in Chat neurons of Chat-cre mice and no signal in WT mice. Similar strategy for other Cre lines.

We thank the reviewer for the excellent suggestions. To test the specificity of RG expression, we performed immunostaining of RG in mice expressing tdTomato or mCherry in Cre+ cells (by crossing tdTomato reporter line with a Cre driver line or by injecting AAV2-EF1α-FLEX-mCherry into a Cre line). We found high colocalization between RG and tdTomato/mCherry expression (Figure 1— figure supplement 1). In contrast, in wild type mice lacking Cre-recombinase expression, we found no RG labeling. Since RG is required for transsynaptic labeling, the high specificity of RG expression ensures specificity of the starter cell population. Furthermore, wild type mice injected with TVA, RG and RV showed no long-range projections (Figure 1—figure supplement 2C). Collectively, these experiments demonstrate the cell type specificity of the starter population.

4) The concern about whether the variability of presynaptic neurons is correlated with the number of starter cells is not solved yet. There isn't much that can be done about this but it should be mentioned in the paper.

We measured the convergence index (ratio between the number of input cells and starter cells) and found it to range between 4.3 and 77.7 for the four cell types (Figure 1—figure supplement 4B). Such a level of variability was also found in some other tracing studies using similar methods (Miyamichi et al., 2011, DeNardo et al., 2015). We have mentioned this in the revised manuscript (third paragraph of Results).

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    Figure 3—source data 1. Distribution of input cells in 53 brain areas for ChAT+, VGLUT2+, PV+, and SOM+ BF neurons.

    Shown are the mean ± s.e.m. of the percentage of inputs from each area for individual cell types. Note that within each of the 12 brain structures, there are unnamed sub-regions outside of the 53 areas listed in the table; thus the percentages of inputs in the listed areas do not always add up to the total percentage in the given brain region.

    DOI: http://dx.doi.org/10.7554/eLife.13214.009

    elife-13214-fig3-data1.xlsx (15KB, xlsx)
    DOI: 10.7554/eLife.13214.009
    Figure 6—source data 1. Distribution of axonal projections to 53 brain areas from ChAT+, VGLUT2+, PV+, and SOM+ BF neurons.

    Shown are the mean ± s.e.m. of the percentage of projections to each area for individual cell types.

    DOI: http://dx.doi.org/10.7554/eLife.13214.015

    elife-13214-fig6-data1.xlsx (14.9KB, xlsx)
    DOI: 10.7554/eLife.13214.015
    Figure 7—source data 1. Distribution of BF input and output from 12 major brain subdivisions across cell-type.

    Shown are the mean ± s.e.m. of the inputs (A) and outputs (B) for each color-coded brain region, for individual cell types.

    DOI: http://dx.doi.org/10.7554/eLife.13214.018

    elife-13214-fig7-data1.xlsx (12.3KB, xlsx)
    DOI: 10.7554/eLife.13214.018

    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES