Using the ABA, we conducted a comprehensive analysis of available neuromodulator receptor gene expression (see Table 1) in areas regarded as sources of neuromodulation, as well as the amygdala (see Fig. 1). The expression energy for all receptor subtypes, in which data was available in the ABA, was examined in all the brain regions of interest.
Total expression and individual receptor subtypes
In the examined brain regions, expression energy of cholinergic receptors was much higher and expression energy of adrenergic receptors was much lower than that for dopaminergic and serotonergic receptors. Figure 2 shows the total expression energy for available adrenergic, cholinergic, dopaminergic, and serotonergic receptors from the ABA across the 13 brain regions examined (note the different scale on the x-axes of Fig. 2). Each bar in Fig. 2 represents gene expression energy when combining all receptor subtypes per region. Brain regions were ranked and arranged based on total expression in Fig. 2, with the brain region having the highest expression energy at the top bar of each plot.
The SI of the basal forebrain, amygdala (AAA, CEA, and MEA), and the VTA had relatively high levels of receptor expression energy. The SI had the highest receptor expression energy of all neuromodulatory regions tested, implying that this region of the basal forebrain is strongly innervated by all neuromodulatory systems (see Fig. 2). The amygdala closely followed SI in terms of overall neuromodulatory receptor expression energy, but expression energy in the amygdala differed based on neuromodulatory receptor type and amygdala subregions. For example, MEA had the highest adrenergic, cholinergic, and serotonergic receptor expression energy among the amygdala regions. However, the CEA had the most dopaminergic receptor expression energy. Similar to the SI, the VTA, which contains dopaminergic neurons, displayed high expression energy for all neuromodulatory receptors.
Somewhat surprisingly, the LC and raphe nuclei (DR, CS, CLI, and RPO), which are sources of norepinephrine and serotonin, respectively, did not have high expression energy of neuromodulatory receptors relative to the other regions examined (see Fig. 2). Because the expression energy was normalized by area, this difference should not be due to the smaller size of these brain regions.
Different brain areas had distinct patterns of receptor subtype expression. Expression energy for individual receptor subtypes across all neuromodulatory systems are shown in Fig. 3. Subtypes were sorted by expression per neuromodulatory system with the top charts having the highest expression. Within each neuromodulatory system, the arrangement of brain regions from left to right on each chart was based on their overall expression as in Fig. 2. It is apparent that the distribution of gene expression per subtype from one brain region to another was not uniform (Fig. 3). However, looking at individual expression energy helps identify receptor subtypes that contribute to the total expression of a particular brain region being described in Fig. 2.
The expression profile of SI, for example, which has the highest receptor expression energy among all for neuromodulatory regions (Fig. 2), may be influenced by select subtypes within neuromodulatory systems. For instance, within the adrenergic receptors, Adra1d and Adrb2 made up for a large portion of the expression energy found in SI, while the remaining four adrenergic receptors did not contribute nearly as much (Fig. 3a). The cholinergic system, which had the most receptor subtypes, was dominated by the expression of the muscarinic subtypes Chrm1, Chrm2 and Chrm4, and the nicotinic Chrna1 (Fig. 3b). Even the dopaminergic system, having the fewest receptor subtypes, had differing receptor expression, with Drd1a and Drd2 having much higher expression value in SI than Drd3 (Fig. 3c). Lastly, serotonergic receptors Htr2c, Htr1f, Htr1a, and Htr1b described most of the total expression energy in SI with comparatively lower contribution from the other subtypes (Fig. 3d).
Among the neuromodulatory sources, VTA also displayed higher overall receptor expression energy compared to other regions. In general, many of the subtypes that have noticeably high expression energy in the SI also have high energy in the VTA (Fig. 3). The main difference we observed was the muscarinic receptor (Chrm2), the nicotinic (Chrna4, Chrna6, Chrnb3), and the dopaminergic Drd2 receptor expression was higher in VTA than in SI (compare Fig. 3b with c).
Different regions of the amygdala have distinct patterns of neuromodulatory receptor expression energy. The neuromodulatory receptor expression energy found in the amygdala, which was among the highest of the brain regions studied in this survey, differed based on the neuromodulatory system (Fig. 2), amygdalar subregion, and by receptor subtypes (Fig. 3). For ease of visualization, pie charts were used to illustrate how receptor subtypes were distributed within the different amygdala areas (Fig. 4). Figure 4 revealed a similar distribution set of prominent gene expression across the amygdala areas with similar proportions. In the adrenergic system, Adra1a was highly expressed in the CEA and AAA, but lower in the MEA. In contrast, Adrb2 had higher expression energy in MEA than in AAA or CAE (Fig. 4, first row). The nicotinic receptor Chrna1 and the muscarinic receptor Chrm1 were more highly expressed across all the amygdala areas in comparison to other nicotinic and muscarinic receptors, though it is interesting to note that Chrm2 had relatively higher expression in the AAA as compared to CEA and MEA (Fig. 4, second row). Dopamine and serotonin receptors also showed differences in receptor expression energy across the amygdala. Drd2 and Htr1f contributed most strongly to the expression found in the CEA, whereas Drd1a and Htr2c contributed most strongly to the expression found in the AAA and MEA regions (Fig. 4, third and fourth row).
Hierarchical clustering analysis
To illustrate the relationship between neuromodulatory receptor expression energy and brain region, hierarchical cluster analyses were performed for expression energy and anatomical location. A hierarchical clustering analysis is a commonly used exploratory technique to handle a large set of data whose interrelationships are elusive and not fully understood. The cluster analysis assigned subsets of gene expression data into groups based on the similarity in their expression patterns (Fig. 5a), and based on the location of the brain regions examined (Fig. 5b). A hierarchy of groupings can emerge using this methodology, and such analyses have previously shown relationships between biological function and anatomical location (Gerstein and Jansen 2000).
To perform the receptor expression energy cluster analysis, a vector of the total expression across the 38 genes was constructed for each of the 13 brain regions. The pairwise distance between these vectors were calculated using Euclidean distance. To create the dendrogram in Fig. 5a, an Unweighted Pair Group Method with Arithmetic Mean (UPGMA) was calculated based on the Euclidean distance metric. Threshold values in Fig. 5a represented the computed distance and linkage between brain regions. The cutoff for determining clusters was set to a threshold of 0.19 to yield three separate clusters, denoted by their different coloring scheme in Fig. 5a.
To examine the relationship between gene expression and anatomical location, a separate hierarchical cluster analysis was conducted using the centroid location for all of the 13 brain regions (Fig. 5b). The procedure was identical to the gene expression hierarchical cluster shown in Fig. 5a, except that a vector of the (x, y, z) coordinates from the reference atlas file (AtlasAnnotation200.sva) was used for clustering instead of gene expression data. The threshold for determining clusters was set to 0.02 to yield four clusters, as in Fig. 5b.
The clusters shown in Fig. 5 suggest several relationships between neuromodulatory receptor expression and anatomical location. The amygdala (AAA, MEA, CEA) and the SI formed a tight cluster (Fig. 5a, blue) in gene expression, as well as anatomically (Fig. 5b, cyan and purple). The SI and basal forebrain are located near the amygdala (see Fig. 1) and like the amygdala contain high overall neuromodulatory receptor expression energy (see Figs. 2, 3). LC and NTS, which contain noradrenergic neurons (McGaugh 2004; Samuels and Szabadi 2008), formed a tight cluster both in terms of gene expression and to a slightly lesser extent anatomically [Fig. 5a (green), b (green and red)]. There was also tight clustering among the raphe nuclei, the source of serotonin in the CNS [Fig. 5a (red), b (red and green)].
There were a few receptor expression energy clusters that did not match their anatomical cluster counterpart or did not form a strong cluster based on expression. For instance, the cholinergic sources SI and MA (Dani and Bertrand 2007; Ishii and Kurachi 2006; Nicholas et al. 1996) did not cluster together based on expression energy, though their distance apart from each other is still relatively small (Fig. 5a, blue and green). However, they are found in neighboring regions of the brain (see Fig. 1) and thus clustered together based on their centroid location (Fig. 5b, cyan). Perhaps the SI and MA not clustering together may be due in part to their proportionally higher expression energy across all four neuromodulatory systems in the SI as compared to MA (see Figs. 2, 3). The dopaminergic region (VTA) and the CLI of the raphe nucleus brain region did not fall within a cluster below the threshold when analyzing gene expression (Fig. 5a). However, in the anatomical cluster analysis, the VTA clustered together with all the raphe regions, PPN, and NTS (see Fig. 5, green and red), as they are located beside each other (Fig. 1).
GABA and glutamate receptor distribution across SI and LC
One of our main findings was that the SI of the basal forebrain had high receptor expression energy for all four neuromodulatory systems (Fig. 2). In contrast, the LC had the lowest overall expression energy across the receptors examined (Fig. 2).
To see if high expression energy in SI and low expression energy in LC existed for receptors other than neuromodulators, we measured the expression energy of GABA and glutamate receptors in the SI and LC (see “GABA and glutamate genes”). We performed the same analysis as before for acquiring expression energy and generating the total expression of neuromodulatory systems found in the SI and LC (see “Total expression and individual receptor subtypes”) with these GABA and glutamate receptors.
We found that, similar to the profile of neuromodulatory receptors, the SI had very high expression energy of GABA and glutamate receptors, while the LC was low. Figure 6 shows the total expression energy for GABA and glutamate across the SI (top) and LC (bottom). For ease of visualization, in Fig. 6 we also included the total expression energy of adrenergic, cholinergic, dopaminergic, and serotonergic receptors from Fig. 2. The values in each bar in Fig. 6 represent the accumulated amount of expression energy when combining all subtypes per region. Note that there is a much higher order of magnitude in expression found in the SI compared to LC (Fig. 6).
Interestingly, we noticed a proportional relationship between the receptor expression found in the SI and LC. Though receptor expression in SI was much higher than in LC, the relative distribution of expression between GABA, glutamate, adrenergic, cholinergic, dopaminergic, and serotonergic receptors had very similar profiles to the LC, with glutamate receptors displaying the highest amount of expression, followed by GABA, acetylcholine, and serotonin (Fig. 6). This implies that the LC region has proportionally lower receptor expression energy when compared to SI, and other brain regions in this study. Since the receptor expression energy was normalized over region size (see “Methods”), this lower overall receptor expression energy level reflects a unique property of the LC region.
Contrast between ABA data and prior in situ hybridization mRNA rat experiments
Although not exhaustive, neuroinformatics web resources such as the Gene Expression Nervous System Atlas (GENSAT) and the Neuroscience Information Framework (NIF) provide an accessible way to obtain gene expression data from various experiments (Heintz 2004; Gardner et al. 2008; Müller et al. 2008). We compared and contrasted data reported from the ABA to results from studies retrieved from these resources.
Table 2 shows the relative expression level in the brain regions of interest per receptor subtype. This was accomplished by first querying NIF using all genes listed in Table 1. NIF returned results from GENSAT that contained gene expression information from the mouse brain based on bacterial artificial chromosomes (BACs) experiments. However, because BAC experiments measure the relative rates of transcription for each gene, it is thereby not a direct measurement of mRNA accumulation. As such, in addition to the BAC expression data, GENSAT provides background literature, primarily from rat experiments, that measure localized mRNA using ISH, which GENSAT uses to correlate with their results. We utilized this feature to collect prior literature on gene receptor expression localization and intensities.
Table 2 Comparison between expression levels in the brain regions of interest per subtype found in rat literature and ABA
Altogether, twenty-six papers were retrieved from GENSAT to compare and contrast gene receptor expression with the ABA in Table 2. With the exception of two receptors (Htr3a and Htr3b) coming from mouse literature, and six not having any prior literature found in GENSAT (Chrna1, Chrna7, Chrna9, Chrnb1, Chrnb3, Htr2b), all remaining receptors from Table 1 were available in GENSAT and feature rat brain in their experiment. As such, it should be noted that Table 2 is an indirect comparison of species to species receptor expression. However, all experiments retrieved from GENSAT document localization of mRNA using ISH.
Once literature was acquired, classification of expression level in the prior studies was taken directly from the wording in the corresponding reference. For example, some studies stated relative values (high, moderate, low), while others created tables using symbols (−, +, ++, +++) to denote the density of expression from in situ hybridization analysis. Classification of expression level in the present ABA study was based on the relative expression energy within a brain category. Expression energy less than the 33rd percentile was classified as low expression, moderate expression was between the 33rd and 66th percentiles, and above the 66th percentile was considered highly expressed. The 13 brain regions were condensed into 5 categories: Amygdala (AAA, MEA, CEA), Dopaminergic (VTA), Serotonergic (DR, RPO, CLI, CS), Cholinergic (SI, MA, PPN), and Adrenergic (LC, NTS) neuron regions. To determine the energy of expression, the average expression across these categorized brain regions was computed, and then percentiles were calculated across each gene in each category. If the expression of a gene (row) in a brain category (column) from the ABA coincided with previous work, then we considered the comparison to be in agreement, and the green entries in Table 2 denoted this. If the expression in the ABA was classified higher than in prior experiments, the table entry was colored red. Blue denoted lower expression in the ABA than in prior studies. Gray entries in the table represent expression data not found in previous studies, while yellow entries represent experiments not conducted in the literature. In the case where there was no expression found, but experiments conducted, in both the literature and ABA, entries were flagged in orange. Black entries represent a unique case where, for a given gene, no data was found in the ABA, but no experiment conducted in the literature.
In general, the comprehensiveness of the ABA revealed information that was previously unreported (Table 2, gray and yellow entries), and reported higher receptor expression in the amygdala and basal forebrain across all neuromodulatory systems than in previously reported studies (Table 2, red entries).
Network visualization and connectivity
In order to better analyze complex systems of interaction, Pajek, a software package designed for examining large networks (Batagelj and Mrvar 1998), was used to visualize potential connectivity relationships between brain regions based on expression data from the ABA. We make the assumption that given a neuromodulatory source, such as VTA, we can infer the strength of a projection to a target area from that source based on the receptor expression energy (e.g., by looking at the overall dopaminergic expression energy in a target region).
Figure 7 shows the overall relationship among the neuromodulatory systems along with its interactions with the amygdala. Nodes corresponded to either a class of neurotransmitter source (e.g., ACh from SI, MA, and PPN) or the different regions of the amygdala, which were recipients of neuromodulation. Directional arcs represented inferred projections from a neuromodulatory system to a target brain area. The thickness of each arc was proportional to the amount of receptor expression energy found in the target region. The diameter of each node represented the total amount of receptor expression energy in that brain region. For example, the cholinergic receptor expression energy in MEA was much higher than serotonergic, as can be seen in Fig. 7 by the thickness of the arc (compare the arc extending from green node to MEA with the arc extending from the red node to MEA). All networks from Pajek were rendered using the circular layout; all other parameters were set to default. For ease of visualization, the amounts of receptor expression energy were scaled down, dividing the amount of receptor energy expression by 100.
Expression energy emanating from the cholinergic system is overwhelmingly the highest, followed by serotonergic, dopaminergic, and adrenergic (Fig. 7). All neuromodulatory systems project heavily to the cholinergic system as compared to other brain regions (Fig. 7 green node). The rest of the projections remained relatively low, though there may be an indication that serotonin projects more heavily to AAA compared to other amygdala areas (Fig. 7).
In addition to looking at the overall neuromodulatory connectivity network, we examined the influence of receptor subtypes on the different brain regions. Families of receptors were categorized in the following way: α (Adra1a, Adra1b, Adra2a, Adra2c) versus β (Adrb1, Adrb2) adrenergic receptors; muscarinic (Chrm1, Chrm2, Chrm3, Chrm4, Chrm5) versus nicotinic (Chrna1, Chrna2, Chrna3, Chrna4, Chrna5, Chrna6, Chrna7, Chrna9, Chrnb1, Chrnb2, Chrnb3) cholinergic receptors; D1 (Drd1a) versus D2 (Drd2, Drd3) dopaminergic receptors; and serotonin receptors that produce an inhibitory response (Htr1a, Htr1b, Htr1d, Htr1f, Htr5a, Htr5b) versus serotonin receptors that produce an excitatory response (Htr2b, Htr2c, Htr3a, Htr3b, Htr4, Htr6, Htr7).
In general, different families of receptors had noticeable differences in how they are distributed across different brain regions (see Figs. 8, 9, 10, 11). For comparison purposes, the layout, arc thickness, and node diameter proportions were scaled down, dividing the amount of receptor energy expression by 1,000 for Figs. 8, 9, 10, 11. Expression energy from α-adrenergic receptors (Fig. 8, top) was more prevalent in cholinergic regions, as well as in the anterior amygdalar area (AAA), and within itself compared to β-adrenergic receptors (Fig. 8, bottom), which had a stronger influence on dopaminergic areas. Both the D1 and D2 dopamine families had a strong influence on the regions associated with acetylcholine and the CEA (Fig. 9); however, the D2 family of receptors expressed more within dopaminergic sources compared to D1 (Fig. 9, bottom). Muscarinic acetylcholine expression (Fig. 10, top) was higher than nicotinic expression in the amygdala (MEA, CEA), while nicotinic receptors (Fig. 10, bottom) were more strongly expressed in the dopaminergic areas. As for the serotonergic receptors, the amount of expression was roughly the same for the inhibitory and excitatory HTR families (Fig. 11).