Adenosine A1 Receptor mRNA Expression by Neurons and Glia in the Auditory Forebrain.

In the brain, purines such as ATP and adenosine can function as neurotransmitters and co-transmitters, or serve as signals in neuron-glial interactions. In thalamocortical (TC) projections to sensory cortex, adenosine functions as a negative regulator of glutamate release via activation of the presynaptic adenosine A1 receptor (A1 R). In the auditory forebrain, restriction of A1 R-adenosine signaling in medial geniculate (MG) neurons is sufficient to extend LTP, LTD, and tonotopic map plasticity in adult mice for months beyond the critical period. Interfering with adenosine signaling in primary auditory cortex (A1) does not contribute to these forms of plasticity, suggesting regional differences in the roles of A1 R-mediated adenosine signaling in the forebrain. To advance understanding of the circuitry, in situ hybridization was used to localize neuronal and glial cell types in the auditory forebrain that express A1 R transcripts (Adora1), based on co-expression with cell-specific markers for neuronal and glial subtypes. In A1, Adora1 transcripts were concentrated in L3/4 and L6 of glutamatergic neurons. Subpopulations of GABAergic neurons, astrocytes, oligodendrocytes, and microglia expressed lower levels of Adora1. In MG, Adora1 was expressed by glutamatergic neurons in all divisions, and subpopulations of all glial classes. The collective findings imply that A1 R-mediated signaling broadly extends to all subdivisions of auditory cortex and MG. Selective expression by neuronal and glial subpopulations suggests that experimental manipulations of A1 R-adenosine signaling could impact several cell types, depending on their location. Strategies to target Adora1 in specific cell types can be developed from the data generated here. Anat Rec, 301:1882-1905, 2018.

primary auditory cortex

adenosine 1a receptor

adenosine kinase

adenosine

adenosine diphosphate

adenosine monophosphate

aquaporin‐4

adenosine triphosphate

dorsal/ventral auditory areas

hippocampus subdivisons

corticocortical projection

corticotectal projection

corticothalamic projection

dorsal division (of MG)

ectonucleotidase

entorhinal cortex

equilibrative nucleoside transporter

glyceraldehyde 3‐phosphate dehydrogenase

hippocampus

integrin subunit alpha M

medial division (of MG)

myelin basic protein

medial geniculate body

marginal zone (of MGv)

ecto‐5′‐nucleotidase (CD73)

pontine gray

peri interlaminar n.

peripeduncular n.

retrosplenial area

superior colliculus

thalamocortical projection

ventral division (of MG)

vesicular GABA transporter

vesicular glutamate transporter 1

vesicular glutamate transporter 2

white matter

INTRODUCTION

Purines and pyrimidines are nucleotides found in all cells, contributing to multiple processes, such as DNA/RNA construction, energy homeostasis, and cell signaling. In the peripheral and central nervous systems, purinergic signaling is a widespread mechanism for communication between neurons, glia, and vascular cells. Purines such as adenosine and adenosine triphosphate (ATP) can function as neurotransmitters, co‐transmitters, neuromodulators, and as signals in neuro–glial interactions (Burnstock, 2007; Boison, 2008a; Abbracchio et al., 2009; Housley et al., 2009; Koles et al., 2016). Accordingly, their functional impact on the nervous system is broad, including key roles in development, behavior, regeneration, plasticity, and pathology (Dale, 2008; Burnstock et al., 2011; Wei et al., 2011; Zimmermann, 2011; Del Puerto et al., 2013; Dias et al., 2013; Sebastiao and Ribeiro, 2015; Krugel, 2016; Pedata et al., 2016). In the brain, ATP and adenosine are released in an activity dependent manner from the presynaptic terminals of neurons by vesicle‐mediated exocytosis, postsynaptic membranes (dendrites), and axons (Fredholm et al., 2005; Wall and Dale, 2008; Abbracchio et al., 2009; Burnstock et al., 2011; Cunha, 2016). They may also be released by glia, especially astrocytes, via vesicular secretion, transport proteins, and membrane channels (Boison et al., 2010; Koles et al., 2016).

Their actions are mediated by purinergic receptors at pre‐, post‐, and nonsynaptic sites on neurons and glia. Adenosine is bound by members of the P1 receptor class (G‐protein‐coupled), and ATP by the larger P2 receptor family (P2X, ligand‐gated ion channel; P2Y, G‐protein‐coupled). The P1 and P2 receptor subtypes have varied levels of expression in neurons and glia of the cortex, hippocampus, cerebellum, striatum thalamus, brainstem, and spinal cord (Fields and Burnstock, 2006; Wei et al., 2011). For the P1 class, distributions of its four subtypes (A1, A2A, A2B, and A3) are generally distinct, with overlap in some brain regions, and may be colocalized in some cells. Likewise, the P2 receptors are widely expressed by neurons and glia in the brain, where the localization of specific subtypes can be distinct (Burnstock and Knight, 2004). Cell type‐specific transcriptome profiling has revealed that the expression of P1 and P2 receptors differs substantially among several major neuronal and glial subtypes (Cahoy et al., 2008; Zhang et al., 2014; Zeisel et al., 2015). In the auditory forebrain, transcriptome profiling indicated that several members of the P1 and P2 receptor classes are expressed in A1 and MG (Hackett et al., 2015). P1‐class expression was dominated by the A1 subtype, and P2‐class expression was highest for P2rx4, P2rx7, P2ry1, and P2ry12 (Fig. 1). Additionally, transcript levels tended to increase or decrease during the maturational period studied (P7 through adult), especially between P7 and P14 (before and after hearing onset).

Figure 1

Purinergic receptor transcript expression in the auditory forebrain (A1, MG) during postnatal development. For each gene, mean normalized read counts derived from RNA sequencing of A1 and MG are plotted at 4 postnatal ages (P7, P14, P21, adult). Adora1 expression (top panels) is plotted separately from the P2 receptors (bottom panels) due to differences in scale. *P < 0.05; **P < 0.01. Data derived from Hackett et al. (2015) database.

Studies in hippocampus and other brain regions reveal that adenosine can regulate synaptic transmission and plasticity through A1 and A2a receptors located at pre‐ and postsynaptic sites on neurons and astrocytes (Ochiishi et al., 1999; Rebola et al., 2005; Sebastiao and Ribeiro, 2009; Sperlagh and Vizi, 2011; Ota et al., 2013; Chen et al., 2014; Sebastiao and Ribeiro, 2015). Importantly, neuromodulation by adenosine is not limited to glutamatergic transmission, but also other modulators (e.g., GABA, acetylcholine, BDNF, cannabinoids). The actions of adenosine via A1 and A2a receptors are most often associated with inhibition and facilitation, respectively. The inhibitory effects of A1 receptors have primarily been observed at glutamatergic and cholinergic endings (hippocampus), but not GABAergic (Cunha and Ribeiro, 2000; Cristovao‐Ferreira et al., 2009). Instead, the A1 receptor appears to have indirect influence in particular interneuron subpopulations (Klausberger et al., 2005; Rombo et al., 2016), indicating cell‐type specificity in expression (Rivkees et al., 1995; Ochiishi et al., 1999) and mechanisms of action.

Consistent with the foregoing, recent studies of thalamocortical (TC) inputs to sensory cortex have revealed that adenosine functions as a negative regulator of glutamate release via activation of presynaptic A1 receptors (Fontanez and Porter, 2006; Bayazitov et al., 2007; Blundon et al., 2011; Chun et al., 2013; Ferrati et al., 2016). In the auditory forebrain, restriction of A1‐mediated adenosine signaling in medial geniculate (MG) neurons is sufficient to extend LTP, LTD, and tonotopic map plasticity in adult mice for months beyond the critical period (Blundon et al., 2011; Chun et al., 2013; Blundon et al., 2017). In contrast, restricting adenosine signaling in neurons of the primary auditory cortex (A1) does not contribute to these forms of plasticity, suggesting that the effects are confined to presynaptic TC projections. Thus, adenosine signaling through presynaptic A1 receptors on TC terminals appears to act as a gate that regulates glutamate release and plasticity at TC synapses in A1.

Adora1 Expression by Neurons and Glia in the Auditory Forebrain

The focus of this study is the adenosine A1 receptor subtype (hereafter, A1R). The purpose was to advance understanding of the circuitry involved in A1R‐mediated adenosine signaling in the auditory forebrain by cells that express A1R mRNA transcripts (i.e., Adora1). Specific goals were to identify the neuronal and glial cell subtypes that express Adora1, and map their locations across laminae in primary auditory cortex (A1) and nuclear subdivisions of the medial geniculate body (MG), based on co‐expression with cell‐type specific markers.

Rationale for this study includes several observations. First, as indicated above, adenosine receptor expression is widespread, but can vary by brain region, cell class, and receptor subtype. The A1R subtype is the most highly expressed purinergic receptor in the auditory forebrain (Fig. 1), but its regional‐ and cell‐type‐specificity have not been profiled. Thus, the structural foundation for studies of A1R function has not been established.

Second, glia and neuron–glia interactions are involved in a vast array of brain functions that often involve adenosine signaling (Stevens et al., 2002; Fields and Burnstock, 2006; Boison, 2008a; Pelligrino et al., 2011; Lovatt et al., 2012; Del Puerto et al., 2013; Domercq et al., 2013; Bynoe et al., 2015; Coppi et al., 2015; Koles et al., 2016). Whether Adora1 expression is restricted to specific neuronal and glial subpopulations or subdivisions of the auditory forebrain is unknown, but this knowledge is essential for the understanding and modeling of adenosine signaling in forebrain circuits.

Third, cell‐specific expression patterns have important implications for the design of experimental studies expecting to target Adora1 (e.g., optogenetics, gene editing, pharmacology, behavior). Target specificity can significantly impact experimental outcomes and must be incorporated into experimental designs. For example, altering A1R expression at the mRNA or protein levels would be expected to produce different results if expression was common to all cells, as opposed to particular neuronal or glial subpopulations.

With these goals in mind, in situ hybridization (ISH) was used to profile the expression of Adora1 transcripts in subclasses of neurons (glutamatergic, GABAergic) and glia (astrocytes, oligodendrocytes, microglia) in the auditory forebrain of adult mice. The results indicate that Adora1 is widely expressed by neurons and glia in A1 and MG, but in a manner that varies by cell type and specific location. Thus, A1R‐mediated adenosine signaling would be expected to differentially impact subpopulations of neurons and glia in the auditory forebrain. Strategies to target Adora1 in specific cell types can be developed from the data generated here.

MATERIALS AND METHODS

Tissue Acquisition

All procedures were approved by the Animal Care and Use Committee at Vanderbilt University and followed the guidelines established by the National Institutes of Health for the care and use of laboratory animals. Brains collected from adult (8–10 weeks) C57BL/6J mice (Jackson Labs 000664), 8 of each sex. Animals were anesthetized with isoflurane in an isolation chamber until areflexive, and decapitated. Brains were immediately removed, blocked in the coronal plane, embedded in OCT compound (Tissue‐Tek, Torrance, CA), then flash frozen in liquid‐nitrogen‐cooled isopentane. Blocks containing 4 brains each (2 male and 2 female) were stored at −80°C. Blocks were sectioned coronally at 10 μm, and collected onto Superfrost Plus slides (Thermo Fisher Scientific, Waltham, MA), and stored at −80°C prior to histological assays.

Selection of Neuronal and Glial Markers

Adora1 mRNA expression was studied in three subtypes of neurons and glia, based on co‐expression for cell‐specific markers (Table 1). Subpopulations of glutamatergic neurons were distinguished by expression or co‐expression of vesicular glutamate transporters 1 and 2 (VGluT1, VGluT2), which have a partially complementary distribution in the brain (Kaneko and Fujiyama, 2002; Fremeau et al., 2004). The expression of the vesicular GABA/glycine transporter (VGAT) is specific to GABAergic and glycinergic neurons (Chaudhry et al., 1998; Dumoulin et al., 1999). In a prior study, we mapped the spatial distributions of glutamatergic and GABAergic neurons in A1 and MG of adult and developing mice, based on expression of these markers (Hackett et al., 2016). VGluT1 was preferentially expressed in A1, and VGluT2 was co‐expressed with VGluT1 in a small subset of those neurons. In MG, nearly all neurons expressed VGluT2 or co‐expressed VGluT1 and VGluT2. The expression of VGAT was restricted to neurons in A1, as GABAergic neurons are rare in the rodent MG. The VGluT1, VGluT2, and VGAT transcripts are highly specific to glutamatergic and GABAergic neuronal subpopulations in the forebrain (Zeisel et al., 2015).

Table 1

Details of riboprobes for Adora1 and markers of neuronal and glial subtypes targeted by in situ hybridization

Cell type	Protein	Gene	Accession No.	Position
Glutamatergic neuron	Vesicular glutamate transporter 1	Slc17a7 (VGluT1)	NM_182993.2	464–1,415
Glutamatergic neuron	Vesicular glutamate transporter 2	Slc17a6 (VGluT2)	NM_080853.3	1986–2,998
GABAergic neuron	Vesicular GABA/glycine transporter	Slc32a1 (VGAT)	NM_009508.2	894–2037
Astrocyte	Aquaporin‐4	Aqp4	NM_009700.2	1,646–2,644
Oligodendrocyte	Myelin basic protein	MBP	NM_001025259.2	648–2,107
Microglia	Integrin subunit Alpha M	Itgam (Cd11b)	NM_001082960.1	538–1,528
General marker, positive control	Glyceraldehyde 3‐phosphate dehydrogenase	Gapdh	NM_008084.2	21–935
All	Adenosine A1 receptor	Adora1	NM_001008533.2	4,018–4,999

Glial cell distributions have not been previously mapped in the auditory forebrain. Markers of each glial class were selected to achieve specific expression for that cell type, as well as the broadest expression across subpopulations of each class. For myelinating oligodendrocytes, myelin basic protein (MBP) was chosen for ubiquitous expression in myelinating oligodendrocytes in mature animals (Cahoy et al., 2008; Zeisel et al., 2015; Holt and Olsen, 2016). For astrocytes, aquaporin‐4 (Aqp4) was chosen over the more commonly used marker, glial fibrillary astrocytic protein (GFAP). Aqp4 is highly specific to astrocytes (Venero et al., 1999; Badaut et al., 2002; Nicchia et al., 2004; Saadoun et al., 2005; Hsu et al., 2011; Aoyama et al., 2012; Papadopoulos and Verkman, 2013; Smith et al., 2014; Hubbard et al., 2015), and transcript expression levels are higher than GFAP for both major astrocyte subtypes (Cahoy et al., 2008; Zhang et al., 2014; Zeisel et al., 2015). Some evidence also suggests that GFAP is nominally expressed by some astrocyte subpopulations (Cahoy et al., 2008; Zeisel et al., 2015), and may be co‐expressed with chondroitin sulfate proteoglycan 4 (aka: Cspg4, NG2), a marker of oligodendrocyte precursor cells (Zhang et al., 2014; Alghamdi and Fern, 2015). For microglia, several specific markers of microglia are known (e.g., Aif1, Cx3cr1, Hexb, Tlr4), but the cell adhesion molecule, integrin alpha‐M (Itgam) (aka Cd11b) was chosen because it is more widely used as a marker and expressed by all subtypes (Zhang et al., 2014; Zeisel et al., 2015; Holt and Olsen, 2016).

In situ Hybridization (ISH)

Chromogenic and multiplex fluorescence assays were conducted using custom riboprobes designed and manufactured by Advanced Cell Diagnostics (ACD, Hayward, CA) (Wang et al., 2012) (Table 1). The vesicular transporter probes were previously used in studies of the mouse auditory forebrain (Hackett et al., 2016). These proprietary probes and assays (RNAscope) utilize a unique signal amplification and background suppression methodology that consistently yielded exceptionally high sensitivity and specificity with low background. Briefly, after a 30‐min protease permeabilization step, two independent probes (double Z probe) were hybridized for 2 hr to each target sequence (~20 probe pairs per target molecule). The lower region of each probe is complementary to the target sequence, and the upper region is a 14‐base tail sequence. Together, the dual probe construct provides a 28‐base binding site for the preamplifiers, which were built up during a multistage amplification cycle. In the final steps, labeled probes containing the fluorescent or chromogen conjugates were bound to each of the 20 binding sites on each preamplifier. All incubation steps were performed at 40°C in a hybridization oven (HybEZ, ACD, Hayward CA) using the RNAscope Multiplex Fluorescent or Duplex Reagent kits, according to the manufacturer's instructions for fresh frozen brain tissue.

Single colorimetric ISH assays were conducted for Adora1, VGluT1, VGluT2, VGAT, MBP, Aqp4, and Itgam as a baseline reference for the locations of labeled cells. Then, duplex chromogenic assays paired each cell marker with the housekeeping gene, Gapdh (glyceraldehyde‐3‐phosphate dehydrogenase) as a positive control for a second riboprobe. Additional series of duplex chromogenic assays paired each cell marker with Adora1. In duplex assays, the red and blue–green substrates were fast‐red and fast‐green, respectively. As a negative control, sections were incubated in buffer solution that did not contain any riboprobes. These assays revealed no significant labeling in any channel. Multiplex fluorescence ISH (FISH) was performed in two separate series of sections. One combined all three neuronal markers with Adora1 in separate channels (VGluT1, VGluT2, VGAT, Adora1), the other combined all three glial markers with Adora1 (MBP, Aqp4, Itgam, Adora1). These assays were done to permit visualization of all genes in a single tissue section, and as a positive control for marker specificity.

All sections were counterstained to improve identification of brain areas and provide a focal point for cytoplasmic probe labeling. FISH‐reacted sections were counterstained with DAPI (30 sec), and chromogen‐reacted sections were counterstained with 50% hematoxylin (30 sec).

Antibody Selection and Immunohistochemistry (IHC)

Four commercially available primary antibodies for A1R were evaluated for staining quality and consistency (Table 2). Antibody specificity was tested by incubating each antibody with a 10× concentration of the control protein provided by the manufacturer, when available. Negative controls, omitting the primary antibody, were used in the testing of all antibodies to assess nonspecific staining. Optimal primary antibody concentrations were determined from these tests, as listed in Table 2. The Protein‐Tech antibody consistently produced the strongest punctate signals in chromogenic assays of fresh frozen and fixed tissues. This A1R antibody was combined with the neuron‐specific marker, NeuN (Fox3) in dual fluorescence IHC assays as a marker of neuronal nuclei and somata, as detailed previously (Hackett et al., 2016). Secondary antibodies used for IHC and FIHC are listed in Table 2.

Table 2

Primary and secondary antibodies used for IHC

Antibody	Species	Supplier	Part Number	Dilution	References
Primary antibodies
Adora1 A1R	rb	Protein‐Tech	55,026‐1‐AP	1:250	Burke et al., 2015
Adora1 A1R	rb	Novus Biologicals	NB300‐549 RRID:AB_10002337	1:250	Kimura et al., 2003; Yamaguchi et al., 2014
Adora1 A1R	gp	Alamone Labs	AAR‐006 RRID:AB_2039705	1:250	Garcia et al., 2013; Vindeirinho et al., 2013
Adora1 A1R	rb	Aviva Systems	OABF01723 RRID: not listed	1:250	‐‐‐
Fox3/NeuN	ms	Covance	SIG39860 RRID:AB_11220035	1:1000	Wimmer et al., 2010

Abbreviations: ms, mouse; rb, rabbit; G/g, goat, Ch, chicken.

Single chromogenic IHC and dual fluorescence IHC (FIHC) assays were performed on slide‐mounted fresh‐frozen sections (10 μm) and fixed floating sections (40 μm). Sections were rinsed for 30 min in 0.01 M PBS‐Tx (phosphate buffered saline, 0.1% triton), followed by three changes of 0.01 M PBS for 10 min (standard rinse procedure). In FIHC assays, nonspecific labeling of myelin by fluorescent secondary antibodies was blocked by incubation in IT‐Fx (Life Tech) for 60 min at RT, followed by a standard rinse. In IHC and FIHC, sections were incubated for 5 min at RT in a single purified glyocoprotein blocking solution (Superblock, ScyTek Laboratories, Logan, UT), followed by a single rinse in PBS. Superblock reagent reduces nonspecific antibody binding, and was used in lieu of species‐specific sera or bovine serum albumin (Evans et al., 1996; Buttini et al., 2002; Turtzo et al., 2014). IHC sections were incubated for 48 hr in primary antibodies at 4°C, 2 hr in the biotinylated secondary antibody at RT, 1 hr in avidin–biotin solution (Vectastain ABC Kit, PK‐6100, Vector Laboratories, Burlingame, CA), then 5 min in the chromogen solution (ImmPACT VIP Kit, SK‐4605, Vector Laboratories). FIHC sections were then incubated for 48 hr in the primary antibodies at 4°C, then 2 hr in the secondary antibody cocktail (A1R + NeuN) at RT. Standard rinsing steps were performed between all incubations. All incubations and rinsing steps were performed on a laboratory shaker with constant agitation.

Image Acquisition and Analyses

Bright‐field images of single colorimetric ISH and IHC tissue sections were obtained with a Nikon 80i microscope controlled by Neurolucida 10 software (MBF Bioscience). Fluorescence wide‐field images and image montages of FISH and FIHC sections were obtained with a Nikon 90i epiflourescence microscope and Hamamatsu Orca 4.0 CCD camera, controlled by Nikon Elements AR software using 20×, 40×, and 100× objectives. Exposure parameters for each color channel were maintained at the same levels across all samples to standardize signal intensity. Images were assembled into figures using Adobe Illustrator CS6 (Adobe Systems, Inc.). 40× and 100× images were created from z‐plane image stacks collapsed to two dimensions, using an Extended Depth of Focus plugin to the Nikon Elements software.

Identification of auditory cortex and medial geniculate was based on previously established architectonic criteria (Hackett et al., 2011a; Hackett et al., 2016), with reference to standard reference atlases (Franklin and Paxinos, 2007; Lein et al., 2007). Laminar density profiles of Adora1 expression from auditory cortex (areas A1, AuD, AuV) were derived from images of colorimetric ISH sections converted to 8‐bit grayscale, then thresholded at 50% using ImageJ software (NIH, http://nih.gov). Raw grayscale intensity was measured using rectangular selection boxes (1,000 pixel width) from layer 1 (L1) just below the pial boundary through the white matter below L6. The average grayscale intensity value was averaged across 1,000 pixels of each row of pixels separately for A1. The resulting functions were smoothed using a 200‐point moving average.

Cell counts were performed manually from A1 and MG in 4 different brains using diffusion interference contrast (DIC) microscopy (40×, 100×) on chromogen‐reacted sections. For each marker, a cell was considered positively labeled (e.g., Adora1 ) if three or more labeled puncta were contained within the somatic cytoplasm (see Discussion for rationale). The average number of labeled and nonlabeled cells was tabulated for each layer of A1 and division of MG, along with their standard deviations. The X–Y locations of each cell were plotted on schematic diagrams of each brain region. The division between MGm and the suprageniculate nucleus (Sg) was approximated in figures, but their cell counts were combined, as definitive borders between the two divisions were ambiguous in these preparations, and because transcript distributions were typically uniform across the region. The difficulties in delineation of Sg in mouse have been noted previously, (Jones, 2007; Watson et al., 2012), and identification varies across studies (Cruikshank et al., 2001; Anderson et al., 2009; Lu et al., 2009; Anderson and Linden, 2011; Marquez‐Legorreta et al., 2016).

RESULTS

Regional and Local Distributions of Adora1 mRNA

Adora1 mRNA expression was widespread in the adult mouse brain, but concentrations of Adora1 puncta varied by regional and subcellular location. Figure 2A,B shows a coronal section through the auditory forebrain stained for Adora1 (red) in a single‐probe ISH assay, counterstained with hematoxylin (light blue). Adora1 signal was most conspicuous in middle and deep layers of cerebral cortex, cell‐dense domains of the hippocampus, MG, and superior colliculus.

Figure 2

Adora1 mRNA expression in a coronal section of the mouse brain at the rostral‐caudal level of A1 and MG. (A) Adora1 signal (red) is prominent in the middle and deep layers of cerebral cortex, cellular regions of the hippocampus, MG, superior colliculus (SC), and pontine gray (PG). Hematoxylin counterstain. (B) Schematic reference. Scale bar, 1 mm. (C) Adora1 expression across layers of auditory cortex (A1, primary auditory cortex; AuD, dorsal area; AuV, ventral area; TeA, temporal area). (D) Adora1, 50% thresholded grayscale image from panel C. Gray masked rectangles indicate the portion of each area measured in calculation of grayscale intensity. (E) Plots of average grayscale intensity for A1 (black), AuD (blue), AuV (red). Gray rectangles denote boundaries of L4 and L6, where the main signal peaks were centered. Scale, 250 μm.

Laminar intensity profiles in A1 and adjacent fields (AuD, AuV) (Fig. 2C–E) highlight the prominent peaks in L6 and secondary peaks centered on L4. Note that L6 peak magnitude was comparable for all three fields, while the L4 peak was higher in A1. Intensity profiles were not measured in MG, where Adora1 expression was relatively homogeneous across divisions.

In Figures 3 and 4, duplex ISH labeling of Adora1 (red) and Gapdh (green) (positive control) is illustrated at higher magnification to show labeling at the cellular level. Adora1 puncta were concentrated within the somata of neurons and glia in A1 and MG (Figs. 3 and 4). As shown in the high‐magnification images (panels B–E), puncta numbers varied between cells. Some had low numbers (<5), while high‐density expression in many other cells merged and could not be individually resolved. Cells with high levels of Gapdh (mostly neurons) had higher concentrations of Adora1 than cells with low Gapdh levels (mostly glia).

Figure 3

Duplex chromogenic ISH in A1. Adora1 (red) + Gapdh (aqua). Hematoxylin counterstain (pale blue). (A) Adora1/Gapdh expression across layers of A1. Insets correspond to locations of images in panels B–E. Scale bar, 250 μm. (B–E) 100× DIC images of inset locations in panel A. Adora1 puncta (red dots) are concentrated within the somatic cytoplasm, mainly in Gapdh cells. Some cells, often glia (asterisk, panel D) contained no puncta. Adora1 puncta were also located along fine peripheral processes (arrowheads). Scale bar, 20 μm.

Figure 4

Duplex chromogenic ISH in MG. Adora1 (red) + Gapdh (aqua). Hematoxylin counterstain (pale blue). (A) Adora1/Gapdh expression across MG subdivisions (v, ventral; d, dorsal; m, medial/magnocellular) and suprageniculate nucleus (Sg). Insets correspond to panels B–E. Scale bar, 250 μm. (B–E) 100× DIC images of inset locations in panel A. Adora1 puncta (red dots) are concentrated within the somatic cytoplasm, mainly in Gapdh cells. Some cells, often glia (asterisk, panel D) contained few or no puncta. Adora1 puncta were also located along fine peripheral processes (arrowheads). Scale bar, 20 μm.

Adora1 puncta were also sparsely distributed in the neuropil, outside of somatic boundaries in A1 and MG (Figs. 3 and 4, arrowheads). DIC microscopy in panels B–E localized these along thin peripheral processes in the dense neuropil matrix, but the cells of origin could not usually be determined. The numbers of puncta not visibly localized to a cellular compartment (i.e., noise) appeared negligible. Assay specificity for Adora1 was very high, as evidenced by the near total absence of puncta in cell‐poor zones (Figs. 3B and 4B,C). Transcript expression in peripheral processes was also observed for Aqp4 and MBP, described below. See Discussion for more on this subject.

Co‐Expression of Adora1 with Neuronal and Glial Markers

Multiplex fluorescence ISH (FISH) was used to demonstrate co‐expression of markers within cells, and as a positive control to test probe specificity. Figure 5 shows the co‐expression for neuronal and glial markers in auditory cortex (L3). In Figure 5A, Adora1 (red) was combined with markers for VGluT1 (white), VGluT2 (yellow), and VGAT (green). In these neurons, typical of A1 and MG, Adora1 was co‐expressed with VGluT1 and VGluT2, but not VGAT. In Figure 5B, Adora1 (red) was combined with markers for MBP (white), Aqp4 (yellow) and Itgam (green). The results indicate very low (nominal) expression of Adora1 within glia, consistent with findings in colorimetric ISH assays (see below). In all panels, the absence of co‐localized signals within puncta (“dots”) demonstrates that the markers of each gene were specific to their targets.

Figure 5

Fluorescence FISH in layer 3 of A1, combining Adora1 with neuronal or glial markers. (A) Adora1 (red) with neuronal markers: VGluT1 (white), VGluT2 (yellow), and VGAT (green). (B) Adora1 (red) with glial markers: MBP/oligodendrocytes (white), Aqp4/astrocytes (yellow), and Itgam/microglia (green). Reactive puncta (dots) of different markers do not colocalize, but may be co‐expressed within the same cell (e.g., Adora1 + VGluT1). Scale bar, 20 μm.

Adora1 Expression by Neurons and Glia

In the remaining assays described below, neuronal and glial markers were paired with Adora1 in duplex chromogenic assays. Data summaries for each combination are illustrated in Figures 6, 7, 8, 9, 10, 11 (neurons) and Figures 12, 13, 14, 15, 16, 17 (glia). Cell counts for all combinations are compiled in Figure 18.

Figure 6

Duplex chromogenic ISH in A1. Adora1 (red) + VGluT1 (aqua). Hematoxylin counterstain (light blue). (A) Bright‐field image of A1. Insets correspond to panels i–viii. (B) VGluT1 cells plotted by cortical layer. Filled circles, co‐expression of Adora1 + VGluT1. Open circles, VGluT1 only. Scale bar, 250 μm. (Panels i–viii) 100× DIC images of inset locations in panel A. Solid arrows, dual‐labeled; gray arrows, nominal Adora1 puncta. See text for details. Scale bar, 20 μm.

Figure 7

Duplex chromogenic ISH in MG. Adora1 (red) + VGluT1 (aqua). Hematoxylin counterstain (light blue). (A) Bright‐field image of MG. Insets correspond to panels i–viii. (B) VGluT1 cells plotted by MG subdivision. Filled circles, co‐expression of Adora1 + VGluT1. Open circles, VGluT1 only. Scale bar, 250 μm. (Panels i–viii) 100× DIC images of inset locations in panel A. Solid arrows, dual‐labeled cells; gray arrows, nominal VGluT1 puncta in Adora1 cell; white arrows, no VGluT1 detected. Gray arrowheads in panels i–v mark examples of small and out‐of‐focus VGluT1 puncta. See text for details. Scale bar, 20 μm.

Figure 8

Duplex chromogenic ISH in A1. Adora1 (red) + VGluT2 (aqua). Hematoxylin counterstain (light blue). (A) Bright‐field image of A1. Insets correspond to panels i–viii. (B) VGluT2 cells plotted by cortical layer. Filled circles, co‐expression of Adora1 + VGluT2. Scale bar, 250 μm. (Panels i–viii) 100× DIC images of inset locations in panel A. Solid arrows, dual‐labeled. Note low VGluT2 expression levels. See text for details. Scale bar, 20 μm.

Figure 9

Duplex chromogenic ISH in MG. Adora1 (red) + VGluT2 (aqua). Hematoxylin counterstain (light blue). (A) Bright‐field image of MG. Insets correspond to panels i–viii. (B) VGluT2 cells plotted by MG subdivision. Filled circles, co‐expression of Adora1 + VGluT2. Open circles, VGluT2 only. Scale bar, 250 μm. (Panels i–viii) 100× DIC images of inset locations in panel A. Solid arrows, dual‐labeled; gray arrows, nominal Adora1 puncta. See text for details. Scale bar, 20 μm.

Figure 10

Duplex chromogenic ISH in A1. Adora1 (red) + VGAT (aqua). Hematoxylin counterstain (light blue). (A) Bright‐field image of A1. Insets correspond to panels –i–viii. (B) VGAT cells plotted by cortical layer. Filled circles, co‐expression of Adora1 + VGAT. Open circles, VGAT only. Scale bar, 250 μm. (Panels i–viii) 100× DIC images of inset locations in panel A. Solid arrows, dual‐labeled; gray arrows, nominal Adora1 puncta. See text for details. Scale bar, 20 μm.

Figure 11

Duplex chromogenic ISH in MG. Adora1 (red) + VGAT (aqua). Hematoxylin counterstain (light blue). (A) Bright‐field image of MG. Insets correspond to panels –i–viii. (B) VGAT cells plotted by MG subdivision. Filled circles, co‐expression of Adora1 + VGAT. Open circles, VGAT only. Scale bar, 250 μm. (Panels i–viii) 100× DIC images of inset locations in panel A. Solid arrows, dual‐labeled; gray arrows, nominal Adora1 puncta. See text for details. Scale bar, 20 μm.

Figure 12

Duplex chromogenic ISH in A1. Adora1 (red) + MBP (aqua). Hematoxylin counterstain (light blue). (A) Bright‐field image of A1. Insets correspond to panels –i–viii. (B) MBP cells plotted by cortical layer. Filled circles, co‐expression of Adora1 + MBP. Open circles, MBP only. White matter, wm. Scale bar, 250 μm. (Panels i–viii) 100× DIC images of inset locations in panel A. Solid arrows, dual‐labeled; gray arrows, nominal Adora1 puncta. Arrowheads, extrasomatic MBP puncta (black) and Adora1 puncta (gray). See text for details. Scale bar, 20 μm.

Figure 13

Duplex chromogenic ISH in MG. Adora1 (red) + MBP (aqua). Hematoxylin counterstain (light blue). (A) Bright‐field image of MG. Insets correspond to panels –i–viii. (B) MBP cells plotted by MG subdivision. Filled circles, co‐expression of Adora1 + MBP. Open circles, MBP only. Scale bar, 250 μm. (Panels i–viii) 100× DIC images of inset locations in panel A. Solid arrows, dual‐labeled; gray arrows, nominal Adora1 puncta. See text for details. Scale bar, 20 μm.

Figure 14

Duplex chromogenic ISH in A1. Adora1 (red) + Aqp4 (aqua). Hematoxylin counterstain (light blue). (A) Bright‐field image of A1. Insets correspond to panels –i–viii. (B) Aqp4 cells plotted by cortical layer. Filled circles, co‐expression of Adora1 + Aqp4. Open circles, Aqp4 only. White matter, wm. Scale bar, 250 μm. (Panels i–viii) 100× DIC images of inset locations in panel A. Solid arrows, dual‐labeled; gray arrows, nominal Adora1 puncta. Arrowheads, perivascular Aqp4 puncta (gray). See text for details. Scale bar, 20 μm.

Figure 15

Duplex chromogenic ISH in MG. Adora1 (red) + Aqp4 (aqua). Hematoxylin counterstain (light blue). (A) Bright‐field image of MG. Insets correspond to panels –i–viii. (B) Aqp4 cells plotted by MG subdivision. Filled circles, co‐expression of Adora1 + Aqp4. Open circles, Aqp4 only. Scale bar, 250 μm. (Panels i–viii) 100× DIC images of inset locations in panel A. Solid arrows, dual‐labeled; gray arrows, nominal Adora1 puncta. Arrowhead (panel v), single Adora1 puncta within an Aqp4 cell. See text for details. Scale bar, 20 μm.

Figure 16

Duplex chromogenic ISH in A1. Adora1 (red) + Itgam (aqua). Hematoxylin counterstain (light blue). (A) Bright‐field image of A1. Insets correspond to panels –i–viii. (B) Itgam cells plotted by cortical layer. Filled circles, co‐expression of Adora1 + Itgam. Open circles, Itgam only. White matter, wm. Scale bar, 250 μm. (Panels i–viii) 100× DIC images of inset locations in panel A. Solid arrows, dual‐labeled; gray arrows, nominal Adora1 puncta. Arrowhead (panels v, viii), single Adora1 puncta within an Adora1 cell. See text for details. Scale bar, 20 μm.

Figure 17

Duplex chromogenic ISH in MG. Adora1 (red) + Itgam (aqua). Hematoxylin counterstain (light blue). (A) Bright‐field image of MG. Insets correspond to panels –i–viii. (B) Itgam cells plotted by MG subdivision. Filled circles, co‐expression of Adora1 + Itgam. Open circles, Itgam only. Scale bar, 250 μm. (Panels i–viii) 100× DIC images of inset locations in panel A. Solid arrows, dual‐labeled; gray arrows, nominal Adora1 puncta. See text for details. Scale bar, 20 μm.

Figure 18

Cell counts in A1 and MG for each marker combination. Histograms reflect the mean number of single‐ (blue bars) and dual‐labeled cells (red bars) for each cell type marker. (A–C) Neuronal subtypes + Adora1 in A1 by cortical layer. (D–E) Glial subtypes + Adora1 in A1. (G–I) Neuronal subtypes + Adora1 in MG by division. (J–L) Glial subtypes + Adora1 in MG. Error bars, standard deviation. Note variations in y‐axis scale between panels.

Note in figures illustrating duplex labeling, DIC microscopy sometimes caused a slight z‐plane distortion in the green color spectrum, affecting the appearance (size or color) of fast‐green‐labeled puncta. Depending on precise z‐location, labeled puncta could range in color from yellow‐green to blue, and could be quite small. Arrowheads were used to mark illustrative examples of small or out‐of‐focus puncta in some panels, where copy number was low for a given marker.

Adora1 Expression by VGluT1 Neurons in A1 and MG

As previously established (Ito and Oliver, 2010; Hackett et al., 2011b; Hackett et al., 2016), VGluT1 (Slc17a7) was expressed by likely all glutamatergic neurons in A1, and a majority of neurons in MGv and MGd (Figs. 6 and 7). In A1 Adora1 was coexpressed by the majority of VGluT1 neurons in L3–6 (Figs. 6ii–viii and 18). Most neurons in L2 had nominal Adora1 expression, as did a minor subpopulation of VGluT1 cells in other layers (Figs. 6i,vi and 18). Adora1 transcripts were most heavily concentrated in VGluT1 neurons of L3b/4 and L6 (Fig. 6iii,iv,vii,viii). Strongly‐labeled cells were present in other layers, as well (Fig. 6iii), indicating cell‐specific variations in Adora1 expression, independent of cortical layer.

In MG, VGluT1 expression was largely restricted to MGv and MGd, as relatively few neurons were VGluT1 in MGm (Fig. 7). VGluT1 transcript abundance was generally lower in MG neurons, compared to VGluT2 neurons in MG (below) or VGluT1 neurons in A1 (see examples in Fig. 7i,ii,v–vii). These distributions are consistent with previous findings (Barroso‐Chinea et al., 2007; Ito et al., 2011; Storace et al., 2012). Adora1 was expressed in nearly all VGluT1 cells in MGv and MGd, while in MGm, the majority of cells expressed Adora1, but not VGluT1 (Figs. 7iii,iv and 18).

Adora1 Expression by VGluT2 Neurons in A1 and MG

In A1, VGluT2 (Slc17a6) was expressed by a minor subpopulation of neurons that co‐express VGluT1, mainly in L3/4, as previously reported (Ito and Oliver, 2010, Hackett et al., 2011b, Hackett et al., 2016). VGluT2 transcript abundance was relatively low in most cells, compared with VGluT1 (Fig. 8i–viii). All of the VGluT2 neurons identified in A1 coexpressed Adora1, regardless of layer (Figs. 8 and 18).

In MG, VGluT2 expression was strong in nearly all neurons in all divisions (Figs. 9 and 18). The vast majority co‐expressed Adora1. VGluT2 cells with nominal Adora1 expression were rare in MGv and MGd, while slightly more were found in MGm/Sg (Figs. 9iii and 18).

Adora1 Expression by VGAT Neurons in A1 and MG

In A1, VGAT (Slc32a1) was expressed by GABAergic neurons in all layers, with occasional labeling of cells in the white matter (Figs. 10 and 18). Adora1 expression by the majority of VGAT neurons was absent or nominal (Fig. 10i,ii,v,vi–viii), and low‐abundance co‐expression was observed in a minority of GABAergic neurons scattered across L3–L6 (Fig. 10ii,iv). While the somata of many VGAT cells were isolated from other cell types by neuropil, others were closely opposed to one or more unlabeled cells (likely glia) (Fig. 10i–iv,vii). At the border between these cells, the clustering of Adora1 transcripts was common. For plotting, inspection at 100×, DIC was usually required to determine whether the transcripts were located within the VGAT cell or the apposed cell. Most often it was the latter.

In the MG of rodents, GABAergic cells (and therefore VGAT neurons) are uncommon (Winer and Larue, 1996; Ito et al., 2011; Yuge et al., 2011; Hackett et al., 2016). Consistent with this prior research, the MG contained very few VGAT cells (Figs. 11 and 18) (note the reduced y‐axis scale in Fig. 18). There was a slight tendency for VGAT cells to more often be found in MGm/Sg. Overall, Adora1 was expressed at low levels by roughly half of the sparse population of VGAT cells identified anywhere within the MG, with comparable low expression in adjacent nuclei, where VGAT cells were abundant.

Adora1 Expression by MBP Cells in A1 and MG

In A1, MBP was expressed in the somata of myelinating oligodendrocytes in L2–L6, and in very high numbers in the underlying white matter (Figs. 12 and 18). Adora1 was coexpressed by roughly half of the MBP cells in L2–L5 and 66% in L6. In the white matter, only 34% of MBP cells coexpressed Adora1 (Figs. 12viii and 18). Overall, Adora1 transcript abundance in the cytoplasm of MBP somata was modest, as evident in Figure 12i–viii (black arrows). Adora1 puncta were typically localized near the cell membrane and at the base of peripheral processes extending away from the soma (Fig. 12iv,vi, gray arrowheads).

In addition to dense concentrations in the somatic cytoplasm, MBP puncta were also located in peripheral processes, distant from cell somata, visible in all panels in Figure 12. The DIC images (Fig. 12i–viii, arrowheads) reveal that puncta were typically located in thin processes, often distributed along the trajectory of a single process (Fig. 12iii,v, black arrowheads). Such processes were found in L1–L6, and especially dense in white matter tracts (Fig. 12vii,viii, arrowheads). As discussed above, extrasomatic Adora1 transcripts could be located along peripheral processes. Typically, extrasomatic MBP and Adora1 puncta were not contiguous within processes, but closely apposed puncta were sometimes observed (Fig. 12ii), suggesting that local translation of both proteins may sometimes occur in contiguous subcellular compartments.

In MG, MBP cells were evenly distributed across divisions (Fig. 13A). Transcript abundance in the somatic cytoplasm was typically high (Fig. 13i–viii). As in cortex, MBP puncta were also abundant within extrasomatic processes, often forming continuous trajectories that could be followed for 300 μm, or more, visible as long striations in Figure 13A,v,vii. These were especially conspicuous in MGv. Adora1 expression was nominal in the majority of MBP cells in all divisions (Figs. 13B and 18). These tended to be located at or near the somatic periphery, often at the base of peripheral processes (Fig. 13I,iii,iv,vi–viii). In the neuropil, Adora1 puncta were occasionally located among MBP puncta within the same peripheral processes.

Adora1 Expression by Aqp4 Cells in A1 and MG

In A1, Aqp4 was expressed by astrocytes within all cortical layers, white matter, pial membranes, and adjacent to blood vessels (Figs. 14 and 18). Adora1 was co‐expressed in subpopulations of Aqp4 cells in L2–L6 and white matter. The highest concentration of co‐labeled cells was in L6 (Fig. 14B). As for the other glial cells, Adora1 transcript abundance was low in astrocytes, compared to neurons (Fig. 14i–viii). The perivascular (Fig. 14ii) and pial astrocytes (Fig. 14i) typically had little to no detectable expression of Adora1. Aqp4 puncta were contained within the somatic cytoplasm of most cells, but some were located along fine processes in the extracellular neuropil in all cortical layers.

In MG, Aqp4 cells were distributed evenly across divisions (Figs. 15 and 18). Adora1 was expressed by most of the astrocytes detected by Aqp4. As in A1, Adora1 transcript abundance was typically low in Aqp4 cells within the main body of the MG, and nominal in most perivascular and pial astrocytes (Fig. 15i–viii). This distinction signals a potential difference in adenosine signaling between astrocytes in these domains in both A1 and MG. Aqp4 puncta were also common in fine peripheral processes in the neuropil, and outside of somata along pial and vascular membranes.

Adora1 Expression by Itgam Cells in A1 and MG

In A1, Itgam cells were relatively evenly distributed across cortical layers, with higher numbers in white matter (Figs. 16 and 18). Microglia were often spatially isolated from other cells (Fig. 16v), but many were found closely opposed to neurons or other cells (Fig. 16i–iv), consistent with recent studies (Baalman et al., 2015). The majority of Itgam cells did not co‐express Adora1, but often contained 1 or 2 Adora1 puncta (gray arrowheads in Fig. 16i–viii). Itgam puncta were largely confined to somatic cytoplasm, and rarely found in the neuropil along processes.

In MG, Itgam cells were sparsely distributed across divisions, compared to astrocytes and oligodendrocytes (Figs. 17 and 18). As in cortex, the majority did not co‐express Adora1, and Adora1 transcript abundance in cells identified as co‐labeled was generally very low (near the cutoff of 3 Adora1 puncta).

A1R Immunohistochemistry (IHC, FIHC)

Chromogenic (IHC) and fluorescence (FIHC) assays revealed strong punctate A1R immunoreactivity in A1 and MG. Reactivity was concentrated in the neuropil, characterized by dense perisomatic labeling (Fig. 19A,B) around most cells (Fig. 19A,B). Sparse labeling was also observed in the somatic cytoplasm, mainly in neurons (DAPI+/NeuN+), but not presumptive glia (DAPI+/NeuN−) (Fig. 19D–F). In A1, immunoreactivity was strong across L2–L6, but the density of reactive puncta was relatively high in L4 (Fig. 19A), compared to other layers (e.g., L6) (Fig. 19B) (qualitative impression). In addition to dense perisomatic and neuropil labeling, immunoreactive puncta sometimes formed orderly distributions (strings) along peripheral processes. Examples are shown in Figure 19D,E. These striations were evident in the densely reactive neuropil of L3/L4 (Fig. 19D), and also in L5/L6, where puncta were sometimes localized along apical dendrites (Fig. 19E, arrowheads). These patterns are similar to an earlier study in other forebrain areas using a custom antibody (Ochiishi et al., 1999), in which A1Rs were localized to pre‐ and postsynaptic compartments of neurons.

Figure 19

Adora1 (A1R) IHC. (A) A1R (green) and NeuN (red) FIHC. Low‐magnification view of auditory forebrain and surrounding structures. Insets correspond to panels D–G. (B, C) Bright‐field images of single chromogen A1R IHC in A1 layer 4 (B) and layer 6 (C). (D, E) FIHC showing A1R (green), NeuN (red), and DAPI (blue). (D) A1 layer 4. (E) A1 layer 6. Arrowheads show strings of A1R‐immunoreactive puncta. (F) MGv. (G) CA3 region of hippocampus. Scale bars: (A) 500 μm; (B–F) 20 μm; (G) 50 μm.

DISCUSSION

The main purpose of this study was to map the distributions of neuronal and glial cell types that express Adora1 transcripts in the auditory forebrain of adult mice. A general finding was that Adora1 is strongly expressed by the vast majority of glutamatergic neurons in A1 and MG, and expressed at lower levels by subpopulations of GABAergic neurons, astrocytes, myelinating oligodendrocytes, and microglia. These patterns reveal regional and cell‐type specificity in Adora1 expression, and suggest functional differences in adenosine signaling mediated by the A1R receptor.

The discussion begins with a brief overview of adenosine signaling, which depends heavily on studies in hippocampus, striatum, and other brain regions. This is followed by discussion of A1R‐mediated contributions to synaptic plasticity in the auditory forebrain. The balance of the discussion is devoted to integrating the findings of this study with existing models of auditory forebrain circuitry, along with directions for future study.

Sources of Adenosine and A1R‐Mediated Signaling in the Forebrain

Adenosine plays important roles in the modulation of neuronal activity and neuro–glial interactions in the brain (Burnstock, 2007; Boison, 2008a; Abbracchio et al., 2009; Housley et al., 2009; Koles et al., 2016). Multiple sources of extracellular adenosine have been identified, and their respective contributions in this complex system continue to be studied. The two main categories are direct release by neurons and an indirect process involving catabolism of ATP released by neurons and glia (Fig. 20C) (Fredholm et al., 2005; Wall and Dale, 2013; Sebastiao and Ribeiro, 2015; Cunha, 2016). Direct neuronal release of adenosine, which may be pre‐ or postsynaptic, appears to be the main source and is achieved by vesicle‐mediated excocytosis or translocation through bidirectional equilibrative nucleoside transporters (ENTs) (Fredholm and Hedqvist, 1980; Mitchell et al., 1993; Manzoni et al., 1994; Brundege and Dunwiddie, 1996; Wall and Dale, 2008; Klyuch et al., 2012; Lovatt et al., 2012; Melani et al., 2012). ATP released by glia and neurons may be converted to AMP and adenosine by ectonucleotidases (ENase), including ecto‐5′‐nucleotidase (Nt5e) (Lloyd et al., 1993; Zhang et al., 2003; Pascual et al., 2005; Lovatt et al., 2012; Zimmermann et al., 2012; Lalo et al., 2014). Basal levels of extracellular adenosine are typically maintained at low concentrations (Cunha, 2001; Boison, 2006), sufficient to affect tonic activation of A1Rs and inhibitory tone (Fredholm, 2007, 2012; Lalo et al., 2014). ENTs on neurons and glia (especially astrocytes) at the synapse help to maintain basal levels and restore balance after activity‐related increases in extracellular adenosine levels, which primarily result from neuronal release (Fredholm et al., 2005; Dale, 2008; Boison et al., 2010; Lovatt et al., 2012). ENT activity is regulated by adenosine receptor activation through intracellular protein kinase pathways (e.g., adenosine kinase, ADK) responsible for adenosine metabolism (Boison, 2013).

Figure 20

Schematic summaries of Adora1 mRNA expression in the auditory forebrain (left) and adenosine signaling at a glutamatergic synapse (right). Symbol keys for A–C below each panel. (A) Adora1 expression by glutamatergic (triangles) and GABAergic (circles) neurons in A1 and MG. Filled symbols denote Adora1 cells. White fill, Adora1 . Arrows summarize thalamocortical (TC), corticocortical (CC), corticothalamic (CT), and corticotectal (Ct) projections of VGluT1/A1R and VGluT2/A1R containing neurons in A1 and MG. Line thickness corresponds to projection strength. Background shading in A1 layers represents relative density of VGluT2/A1R TC inputs from MG (Hackett et al., 2016). (B) Adora1 expression by astrocytes (circles), oligodendrocytes (ovals), and microglia (triangles) in A1 and MG. Filled symbols denote Adora1 cells. White fill, Adora1 . (C) Schematic highlighting basic elements of adenosine signaling at a glutamatergic synapse. Adenosine (ADO, red circles) and ATP (blue circles) are released mainly by neurons, astrocytes, and/or microglia by several mechanisms, including exocytosis and translocation through equilibrative nucleoside transporters (ENTs) and connexin/pannexin hemichannels. Intracellular adenosine levels are controlled by a cycle involving adenosine kinase (ADK) and ecto‐5′‐nucleotidase (Nt5E). Extracellular adenosine levels are regulated via catabolism of ATP by ectonucleotidases (ENase, Nt5E). Adenosine (A1R) and ATP (P2 class) receptors occupy pre‐ and postsynaptic neuronal membranes, astrocytes, microglia, and oligodendrocytes. Cholinergic binding of presynaptic M1 acetylcholine receptors interferes with A1R signaling, permitting sustained glutamate release. Group 1 mGluRs mediate LTP and LTD in TC projections. Model compiled from present and prior studies reviewed in the Discussion. For simplicity, only the most relevant signaling cascades and receptors are included. Some details may vary by brain region and cell type, and are subject to later revision.

Adenosine signaling is mainly achieved through G‐protein‐coupled A1Rs, the most highly expressed of the four receptor subtypes. In neurons, A1Rs occupy pre‐ and postsynaptic sites, but abundance appears to be highest in presynaptic terminals (Tetzlaff et al., 1987; Ochiishi et al., 1999; Rebola et al., 2003; Rebola et al., 2005). The primary action of A1R‐mediated adenosine signaling at excitatory synapses is presynaptic inhibition of glutamate release, which is increased by interfering with A1R activation (Fredholm and Dunwiddie, 1988; Dunwiddie and Masino, 2001; Fredholm et al., 2005). Postsynaptic effects involving A1Rs include decreased activation of ionotropic NMDA receptors and ion channel conductances (Proctor and Dunwiddie, 1987; Klishin et al., 1995a; Klishin et al., 1995b; Wetherington and Lambert, 2002; Chung et al., 2009; Kim and Johnston, 2015). Through these mechanisms, activity‐dependent adenosine signaling through A1Rs contributes to adaptive changes in neurotransmission associated with various processes, such as neuroprotection and plasticity.

Adenosine Signaling and Synaptic Plasticity in the Auditory Forebrain

The mechanisms that contribute to synaptic plasticity in the auditory forebrain remain the subject of intensive research, as their elucidation is essential for improved understanding of learning and memory (Weinberger, 2007), reorganization/augmentation of stimulus representations (Fritz et al., 2007; Sanes and Bao, 2009; Schreiner and Polley, 2014), and critical periods (Bavelier et al., 2010; Froemke and Jones, 2011; Takesian and Hensch, 2013). Critical periods form as synaptic machinery stabilizes during postnatal development, and plasticity at both corticocortical and thalamocortical synapses is dampened, limiting the changes in activity that can be induced by experience (Zhang and Poo, 2001; Sanes and Woolley, 2011). Thalamocortical synapses become especially resistant to change with maturation (Crair and Malenka, 1995; Barkat et al., 2011; Chun et al., 2013), whereas prior to critical period closure, even passive exposure to a sound is sufficient to induce changes (Zhang et al., 2001; de Villers‐Sidani et al., 2007; Insanally et al., 2009; Barkat et al., 2011). Yet, plasticity can still be induced in mature TC and CC synapses by an active process, such as attention to behaviorally relevant sounds (Recanzone et al., 1993; Blake et al., 2002; Edeline, 2003; Keeling et al., 2008; Sarro et al., 2015), electrical stimulation of MG (Jafari et al., 2007; Ma and Suga, 2009; Zhu et al., 2014), or other forms of neuromodulation (Edeline, 2003; Metherate, 2004; Metherate and Hsieh, 2004; Weinberger, 2004; Froemke and Jones, 2011; Metherate, 2011; Edeline, 2012; Blundon and Zakharenko, 2013; Takesian and Hensch, 2013). For example, merely pairing the release of acetylcholine with a sound is sufficient to increase neuronal responsiveness to that stimulus (Bao et al., 2003).

Taken together, these findings imply that plasticity becomes limited during maturation by the establishment of gating mechanisms, which can subsequently be altered under various conditions to permit plasticity. Although many factors are very well known (e.g., acetylcholine, GABA), the roles of other players, such as adenosine, continue to be identified and refined. In hippocampus and other brain regions, adenosine can regulate synaptic transmission and plasticity through A1 and A2a receptors, located at pre‐ and postsynaptic sites on neurons and astrocytes (Sebastiao and Ribeiro, 2009; Sperlagh and Vizi, 2011; Dias et al., 2013; Ota et al., 2013; Chen et al., 2014; Sebastiao and Ribeiro, 2015). In somatosensory cortex, for example, adenosine application downregulated glutamate release from TC synapses, while A1R inhibition prevented this effect, enabling the modulation of short‐term plasticity (Ferrati et al., 2016).

In auditory cortex, a series of studies revealed that adenosine signaling through A1Rs is a major factor in gating thalamocortical (TC) plasticity after the early critical period. Blundon et al. (2011) showed that postsynaptic LTD depends on group 1 metabotropic glutamate receptors (mGluR1), but is gated presynaptically at mature synapses in A1 by mechanisms that involve adenosine. Briefly, they found that activation of presynaptic M1 acetylcholine receptors (M1Rs) interfered with adenosine‐A1R signaling downstream, permitting sustained glutamate release and LTD induction. Further, inhibition or deletion of A1Rs was sufficient to release this gating. Chun et al. (2013) showed that postsynaptic LTP in TC synapses in A1 also depends on Group 1 mGluRs (mGluR1, mGluR5), but not NMDARs, and becomes gated during maturation. In A1 of adult animals, TC LTP could be induced by combining cholinergic release with disinhibition, but not by either mechanism alone. As in the LTD study, cholinergic activation triggered sustained glutamate release by negative regulation of adenosine‐A1R signaling, unmasking LTP. In A1R−/− knockout mice, LTP could be induced even when M1Rs were blocked, indicating that the cholinergic inputs are not required when A1Rs have been removed. Most recently, this line of work was extended by this group to receptive field plasticity, in vivo (Blundon et al., 2017). As noted above, passive exposure to a pure tone in an enriched acoustic environment is sufficient to alter receptive fields and expand map representation of that tone in A1 neurons of immature animals, but not adults. In this study, interfering with A1Rs in the MG of adult animals by any of several methods (deletion, knockdown, and inhibition) restored map plasticity in A1 and improved tone discrimination after pairing with passive exposure to a pure tone probe. A1R agonists inhibited map plasticity. Thus, the maturation of adenosine‐A1R signaling in TC synapses acts as a gate, limiting the expression of TC plasticity in adults. Removal of that gating mechanism enables TC synaptic plasticity (Fig. 20c).

The time course of these effects parallels the maturational trajectory of adenosine receptor expression in the mouse auditory forebrain. In previous work, we generated whole genome transcriptome profiles of A1 and MG, highlighting expression trajectories for over 200 gene families at key stages during postnatal development in mice (P7, P14, P21, adult) (Hackett et al., 2015; Guo et al., 2016). For the adenosine receptor family, Adora1 transcript expression was dominant in A1 and MG, compared to very low expression of Adora2a and Adora2b (Fig. 1), consistent with studies in other brain regions (Cahoy et al., 2008; Zhang et al., 2014; Zeisel et al., 2015). We also found that Adora1 expression increased sharply from P7 to P14 in A1 and MG, and then remained elevated through adulthood. This upregulation is notable, as the P7–P14 interval spans the period before and after hearing onset, and the closing of the early critical period (de Villers‐Sidani et al., 2007, Insanally et al., 2009, Barkat et al., 2011). Rapid upregulation during this period has been noted in other brain regions (Rivkees, 1995). In addition, we found that expression of several neurotransmitter receptors (e.g., M1R) and numerous coding and noncoding genes known to be involved in critical periods plasticity were in flux during this time interval, suggesting that activity‐dependent mechanisms may alter expression among networks of genes (Hackett et al., 2015; Guo et al., 2016; Hackett et al., 2016).

Although not a focus of this study, it may be important context for future purposes that several P2 ATP receptors were significantly expressed in the auditory forebrain, and that levels changed from P7 to adult (Fig. 1). For example, P2rx2 was expressed at low levels at P7, then dropped to nominal levels by P14; P2rx4 expression steadily increased from P7 to adult; and P2ry12 increased sharply from P7 to P14 in A1 and MG. The roles of P2 receptors are poorly understood in sensory forebrain regions, but intensive studies elsewhere in the brain reveal important roles in plasticity and development (Zimmermann, 2011; Sebastiao and Ribeiro, 2015). Thus, there is likely much more to learn about purinergic and other signaling mechanisms with respect to synaptic plasticity and the establishment of critical periods. These data can serve to guide focused inquiries.

Adora1 Expression by Neurons in the MG

In the MG, Adora1 expression by neurons was almost entirely restricted to glutamatergic neurons, as GABAergic neurons are rare in the rodent MG. Expression was uniform within and across subdivisions, implying that adenosine utilization is comparable across the cell populations within parallel pathways represented in the MG (Fig. 20A). These results indicate that A1R‐adenosine signaling could impact the presynaptic activity of TC projections MG to all areas of auditory cortex (primary and secondary), as well as descending projections to auditory brainstem (e.g., inferior colliculus). These patterns are consistent with autoradiographic binding studies of A1Rs in several species (Fastbom et al., 1987), and also support recent observations (discussed above) that A1Rs located on TC terminals in L3/L4 of A1 modulate presynaptic glutamate release with profound effects on excitatory transmission and synaptic plasticity (Blundon and Zakharenko, 2013; Chun et al., 2013; Blundon et al., 2017).

With this functionality in mind, the expression patterns in MG raise several questions for consideration and future study. First, do presynaptic A1Rs on TC projections modulate glutamate release in other thalamorecipient layers of auditory cortex (e.g., L1, L6)? This is a reasonable hypothesis, given that MG neurons co‐express VGluT2 and Adora1 transcripts, and immunoreactive terminals are present in all thalamorecipient layers (Fig. 19) (Hackett et al., 2011b; Hackett et al., 2016). Second, does adenosine signaling mediate the same functions in auditory cortical areas other than A1? Strong Adora1 expression in all MG subdivisions suggests that presynaptic A1 receptors are localized on TC inputs to all areas of auditory cortex, not just A1 (Fig. 20A). Third, do the descending projections of MG neurons to the inferior colliculus and other subcortical targets bear presynaptic A1 receptors? If so, are they also involved in the modulation of glutamate release? This seems likely, given that A1R‐adenosine signaling has similar effects across brain regions (Cunha, 2016).

Adora1 Expression by Neurons in A1

As summarized in Figure 20A, inputs to A1 include laminar‐specific thalamic and cortical projections, mainly from MG and from other auditory cortical fields. The downstream targets of A1 projection neurons vary by layer of origin, and include both cortical and subcortical targets in both hemispheres. The findings of this study indicate that Adora1 was expressed by the majority of glutamatergic neurons in L3–L6 of A1, based on co‐expression with VGluT1 and/or VGluT2. In contrast, Adora1 was expressed with relatively low abundance in a small subpopulation of GABAergic neurons. Expression varied by layer, characterized by the highest levels in L6, intermediate in L3–L5, and nominal in L1–L2. These laminar patterns were maintained in the belt areas surrounding A1, as well as in frontal and other sensory cortical regions (data not shown), suggesting common organizational principles across cortical fields.

A1R immunohistochemistry revealed dense punctate expression in the neuropil and perisomatic domains, suggesting utilization of A1Rs at pre‐ and postsynaptic sites, as observed in other brain regions (Tetzlaff et al., 1987; Ochiishi et al., 1999; Rebola et al., 2003; Rebola et al., 2005). Ochiishi et al. (1999) found A1R expression mainly in layers 2–6 in most areas of cerebral cortex. Light and electron microscopy revealed expression in presynaptic terminals, postsynaptic membranes, and some expression in the somato‐dendritic cytoplasm. They suggested that the latter might reflect active transport of A1Rs to synaptic membranes. Thus, the combined ISH and IHC results suggest that Adora1 transcripts are concentrated in the somatic cytoplasm, while the translated receptor proteins are transported to cell membranes (Fig. 20). These expression patterns raise several questions about A1R‐adenosine signaling in A1.

First, what are the roles of A1Rs in the cortico‐cortical and cortico‐fugal projections of A1 neurons? Neurons in L3/L5/L6 project to a wide range of cortical and subcortical targets (Fig. 20). Briefly, neurons in L6, which had the highest Adora1 expression levels, project to MG and IC, and receive inputs from the MG and cortical neurons in other locations. Neurons in L5 have widespread subcortical, corticocortical, or callosal projections, depending on the subpopulation. Neurons in L4 receive dense projections from the MG and have largely local projections. L3 neurons also receive thalamic inputs, and have local and longer range corticocortical and callosal projections. Is presynaptic A1R‐adenosine signaling also involved in regulating glutamate release in all of these projections? It may be significant that while plasticity becomes gated at mature TC synapses, cortico‐cortical synapses remain plastic after critical periods close (Amitai, 2001; Froemke et al., 2007; Jiang et al., 2007; Oswald and Reyes, 2008, 2011; Huang et al., 2012), suggesting different regulatory mechanisms for TC and CC projections. Indeed, whereas interference with A1R‐adenosine signaling in MG was sufficient to restore cortical map plasticity, the same manipulations at corticocortical synapses in A1 were not (Blundon et al., 2017). The reasons remain unknown. Thus, presynaptic A1R‐adenosine gating may be specific to TC projections, but have other functions in CC or corticofugal projections.

Second, what is the significance of different Adora1 expression levels between neurons or across cortical layers? Adora1 was widely expressed by glutamatergic neurons in L3–L6, but levels varied significantly between layers in A1, the belt areas bordering A1 (AuV, AuD), and in the cortical domains of other sensory modalities (Fig. 2), suggesting functional homologies among sensory cortical areas. Unfortunately, the precise nature of the relationship between transcript and receptor density in different cellular compartments is not straightforward. In addition to variation in the regulation of transcription or translation, mRNA transcripts and proteins may be localized to different subcellular compartments. For example, studies in hippocampus identified a mismatch between Adora1 mRNA and A1R protein expression, largely explained by the fact that A1Rs are concentrated in presynaptic terminals located in brain regions distant from their mRNA‐containing somata (Johansson et al., 1993; Swanson et al., 1995). In this study, Adora1 mRNA expression was highest in neuronal somata of L6, followed by L3/L4, whereas punctate A1R immunoreactivity was highest in L3/L4, followed by L6, consistent with findings in several species (Fastbom et al., 1987). This mismatched laminar pattern partly reflects the high density of immunoreactive TC terminals in L3/L4 from Adora1 projection neurons in the MG, and naturally parallels VGluT2 expression patterns in the auditory forebrain (Hackett et al., 2016). That is, VGluT2 immunoreactive terminals are most highly expressed in L3/4, followed by L6 (similar to A1R), and VGluT2 and Adora1 mRNA are co‐expressed by perhaps all glutamatergic neurons in MG. Thus, there is a general correspondence in the expression patterns of VGluT2 and Adora1 mRNA and their associated proteins in the TC circuit, reflecting the functional link between presynaptic adenosine and glutamate (Fig. 20).

By extension, such anatomical relationships may be expected to apply to other projections. For example, VGluT1 neurons in L6, which have the highest Adora1 mRNA expression in A1, are the source of a dense network of descending projections to multiple subcortical nuclei in thalamus and brainstem (Bajo et al., 1995; Winer, 2005; Bajo and King, 2012; Crandall et al., 2015; Guo et al., 2017) (Fig. 20). A reasonable prediction is that presynaptic A1R‐adenosine signaling is important in the modulatory environment of those projections, as well. Conversely, A1R‐adenosine signaling could be considered relatively less important for neurons in superficial cortical layers (e.g., L2), where Adora1 expression is nominal. These relationships are reflected in the summary diagram of Figure 20A, where A1R expression is contained in VGluT1 or VGluT2 neurons in corticocortical (CC), thalamocortical (TC), corticothalamic (CT), and corticotectal (Ct) projection systems in the auditory forebrain.

Third, in addition to laminar differences in expression, there was diversity among neurons within the same layers. For example, Adora1 puncta numbers were very high in some VGluT1 neurons, but not others (Fig. 6). Because the projections of neurons within the same cortical layer can vary significantly (i.e., cortico‐cortical, cortico‐thalamic, cortico‐tectal), differences in A1R utilization may also vary by neuronal subpopulation, perhaps in a manner related to connectivity in the network. Indeed, the anatomical and physiological characteristics of neurons in the same cortical layer are often distinct, and those differences linked to their transcriptional heterogeneity (Chevee et al., 2018). Further, Adora1 expression was relatively low in the majority of GABAergic neurons. Roughly 20% of VGAT + neurons in A1 were identified as Adora1 , but most contained just a few puncta, and higher numbers were restricted to a minor subpopulation. The identity of that subpopulation is not yet known. Functionally, the implications are not clear, as adenosine signaling has not been studied in GABAergic neurons in A1. Elsewhere in the brain (e.g., hippocampus), A1R expression is strong in most glutatamatergic neurons, but tends to be significant only in subpopulations of GABAergic interneurons (Rivkees et al., 1995; Ochiishi et al., 1999; Zeisel et al., 2015). Accordingly, GABA release is often not influenced by activation of A1Rs (Cunha and Ribeiro, 2000; Cristovao‐Ferreira et al., 2009). However, in a specific CA1 interneuron subtype, downregulation of extrasynaptic GABAA receptors by adenosine disinhibits those GABAergic interneurons (Klausberger et al., 2005; Rombo et al., 2016). Alternatively, adenosine signaling can regulate synaptic GABA levels via astrocytic transport initiated by activation of A1–A2a heteromers on astrocytes (Cristovao‐Ferreira et al., 2013). Thus, the impact of A1R‐mediated adenosine signaling on GABAergic inhibition in auditory forebrain may operate through such indirect mechanisms and involve particular neuronal classes.

Last, what are the impacts of postsynaptic A1R‐adenosine signaling on neurons in A1? Are the effects different for A1Rs located on somata or dendrites? The expression patterns identified here imply that adenosine signaling via postsynaptic A1 receptors could influence the activity of glutamatergic neurons in L3–L6 of auditory cortex, and perhaps a subpopulation of GABAergic cells, consistent with findings in other brain regions (Abbracchio et al., 2009; Wei et al., 2011; Lovatt et al., 2012; Dias et al., 2013). Functionally, studies conducted primarily in hippocampus indicate that the postsynaptic effects of A1R activation typically include decreased activation of ionotropic NMDA receptors and associated cation channel conductances, while A1R blockade has the opposite effect (Proctor and Dunwiddie, 1987; Klishin et al., 1995a; Klishin et al., 1995b; Wetherington and Lambert, 2002; Chung et al., 2009; Kim and Johnston, 2015). Moreover, these effects can vary regionally by neuronal subpopulation and A1R expression density (Lee et al., 1983; Kim and Johnston, 2015), indicating that properties may vary by specific location or cell type. As mentioned above, induction of LTD, LTP and receptive field plasticity in auditory cortex was distinct for pre‐ and postsynaptic A1Rs (Blundon et al., 2011; Chun et al., 2013; Blundon et al., 2017), highlighting potential differences in pre‐ or postsynaptic A1R‐adenosine signaling.

Adora1 Expression by Glia in the Auditory Forebrain

Historically, studies of brain structure and function have emphasized characterization of neurons and neuronal networks, but interest in the roles of glia has rapidly increased (Kettenmann and Ransom, 2013). In part, this stems from evidence that glial and neuro‐glial interactions are involved in a vast array of brain functions, such as regulation of synaptic transmission, plasticity, information processing, myelination, neuronal development, and response to injury. Several outstanding reviews cover various aspects of this rapidly growing field (Stevens et al., 2002; Fields and Burnstock, 2006; Boison, 2008a; Pelligrino et al., 2011; Lovatt et al., 2012; Del Puerto et al., 2013; Domercq et al., 2013; Boison and Aronica, 2015; Bynoe et al., 2015; Coppi et al., 2015; Croft et al., 2015; Koles et al., 2016). The various roles played by glia often involve bidirectional signaling between neurons and glia, or between glia (Parpura and Zorec, 2010; Araque et al., 2014; Perea et al., 2014; Croft et al., 2015). Astrocytes, in particular, can detect synaptic activity via ionotropic and metabotropic receptors, initiating changes in signaling that modulate synaptic transmission and plasticity (Ota et al., 2013; Bernardinelli et al., 2014; Chung et al., 2015; De Pitta et al., 2016). For example, Ca2+ signaling in astrocytes can be invoked by sensory stimulation (Wang et al., 2006; Winship et al., 2007), initiating the vesicular exocytosis of “gliotransmitters” (e.g., glutamate, ATP, adenosine, peptides) that modulate the activity of neurons and other glia. Oligodendrocytes are responsible for myelination during development, but even in mature cells, activity‐dependent changes in myelin production may contribute to the regulation of critical periods for plasticity in sensory systems (Takesian and Hensch, 2013). Some of these interactions are mediated by neurotransmitter receptors localized on oligodendrocytes (Garcia‐Barcina and Matute, 1996; Karadottir et al., 2005; Micu et al., 2007; Bagayogo and Dreyfus, 2009), paired with release mechanisms that permit signaling back to neurons (Fruhbeis et al., 2013). Microglia are activated in response to traumatic injury to the brain, but in healthy tissue (not activated), their processes move about in the neuropil, surveying local conditions (Wake et al., 2013). Like astrocytes, microglia make contacts with synapses, with effects on structure and function (Wake et al., 2009; Tremblay et al., 2010), neuronal activity (Moriguchi et al., 2003; Hayashi et al., 2006), synaptic plasticity and learning (Tremblay and Majewska, 2011; Tremblay et al., 2012), and many other functions (Kettenmann et al., 2013; Morris et al., 2013). These interactions are achieved, in part, through the expression of many types of neurotransmitter receptors in several families (Pocock and Kettenmann, 2007; Kettenmann et al., 2011; Gertig and Hanisch, 2014).

Among the signaling pathways identified, purinergic signaling is a widespread mechanism for communication between neurons, glia, and vascular cells in the brain (Burnstock and Knight, 2004; Fields and Burnstock, 2006; Burnstock et al., 2011). Both ATP and adenosine can function as signals in these interactions, subserving key roles in development, behavior, regeneration, plasticity, and pathology (Stevens et al., 2002; Dale, 2008; Boison et al., 2010; Burnstock et al., 2011; Wei et al., 2011; Zimmermann, 2011; Krugel, 2016; Pedata et al., 2016). Astrocytes (and neurons) release ATP, which is converted to adenosine in the extracellular space, with impacts on both neurons and glia (Pascual et al., 2005; Lovatt et al., 2012). A major role of A1Rs in astrocytes is to trigger the removal of excess adenosine from the synapse via ENTs (Boison, 2008b; Lovatt et al., 2012; Hines and Haydon, 2014), with effects on glutamatergic transmission (Masino et al., 2002) and plasticity (Navarrete et al., 2012). A1R signaling can also mediate protective effects on astrocytes during insults such as hypoxia (Ciccarelli et al., 2007; Bjorklund et al., 2008). In oligodendrocytes, A1Rs promote myelination during axonal maturation (Stevens et al., 2002), and contribute to white matter loss in pathology (Kim et al., 2005). In microglia, A1R activation on microglia can regulate proliferation (Haselkorn et al., 2010), limit the inflammatory response (Tsutsui et al., 2004), contribute to the neuroprotective effects associated with adenosine delivery (Lauro et al., 2010; Pedata et al., 2016), and modulate short‐term synaptic transmission and plasticity (George et al., 2016). Many other functions are mediated by purinergic signaling through other receptors and pathways, as well.

A general finding of this study was that Adora1 is expressed by subpopulations of all three glial classes in A1 and MG (Fig. 20). Astrocytes and oligodendrocytes were more likely to express Adora1 puncta than microglia. Transcript abundance was much lower in glia than neurons, regardless of location, and was uniform across cortical layers in A1 or between divisions in MG (except perivascular and pial astrocytes, where Adora1 expression was usually nominal). This pattern contrasts sharply with relatively strong Adora1 expression by the majority of neurons.

The reasons for the differences between neurons and glia are not entirely clear, but presumably related to the specific functions mediated by A1Rs in each cell type. Certainly, expression levels in glia are sufficient to meet functional demands, and the uniformly low transcript abundance suggests that those needs do not fluctuate widely over time or location. That is, high receptor numbers are not necessary to maintain functionality, in general, nor are high numbers needed to achieve a specialized function that might require elevated sensitivity.

For other functions, A1R‐adenosine signaling in glia may not be as important as those mediated by other purine receptors (Fields and Burnstock, 2006), several of which are significantly expressed in A1 and MG (Fig. 1). As one example, the P2 ATP receptor, P2ry12, is highly expressed in microglia compared to neurons and other glia (Zhang et al., 2014; Zeisel et al., 2015), with specialized roles in the response to local brain injury (Davalos et al., 2005; Amadio et al., 2006). Much remains to be learned about purinergic signaling, in general, as this receptor family may contribute to neuron–glia interactions, and many other functions in auditory forebrain.

Subcellular Localization of Transcripts

Transcripts for the seven mRNA markers used in this study were primarily concentrated in the somatic cytoplasm, but for some markers (MBP, Aqp4, Adora1), labeled puncta were consistently found in the neuropil. DIC microscopy indicated that most were localized to peripheral processes. High numbers of MBP transcripts were located in the peripheral processes of oligodendrocytes (presumably), consistent with numerous previous studies in other brain regions (Colman et al., 1982; Trapp et al., 1987; Verity and Campagnoni, 1988; Shiota et al., 1989; Barbarese et al., 1995; Ainger et al., 1997; Muller et al., 2013). Although some instances of labeled puncta were possibly nonspecific labeling (noise), the majority was localized to peripheral processes. Internal evidence for this conclusion derives from the high specificity of the probes, and the low levels of noise in areas with few cells (e.g., cortical layer 1).

In addition, it is well established that the transcripts of perhaps thousands of protein coding genes are localized in the neuropil, where local translation may occur (Steward and Levy, 1982; Tiruchinapalli et al., 2003; Bockers et al., 2004; Poon et al., 2006; Zhong et al., 2006; Lein et al., 2007; Giuditta et al., 2008; Cajigas et al., 2012; Sakers et al., 2017). Among the genes with significant neuropil expression in hippocampus are Adora1, MBP, Aqp4, VGluT1, and numerous neurotransmitter receptors (Cajigas et al., 2012). In a recent study focused on astrocytes, Sakers et al. (2017) showed that the machinery for local translation of transcripts is localized in their peripheral processes. RNAseq analyses of peripheral astrocyte fractions revealed modest levels of Adora1, Aqp4, and GFAP, suggesting some degree of transcript expression outside somata. This could be functionally significant, as a single astrocyte may contact thousands of synapses (Oberheim et al., 2006), and numerous functions may benefit from local translation. Altogether, these findings are consistent with present observations in the auditory forebrain, and suggest that peripheral transcript localization is a general property of some genes.

Transcript Abundance

Relevant to the subject of transcript localization is the relationship between transcript and protein abundance. Within a cellular compartment, such as the somatic cytoplasm or peripheral process, how many transcripts constitute a functionally significant number? For somata, we took a conservative approach in setting the threshold to three or more copies. This threshold was somewhat arbitrary, and may under‐ or over‐represent cells that in fact produce significant amounts of functional protein. Moreover, our approach excluded transcripts located beyond somata within peripheral processes. Indeed, the presence of single transport granules containing mRNA (along with ribosome and other molecules) appears to be sufficient for local translation in the peripheral processes of oligodendrocytes (e.g., myelin basic protein) (Muller et al., 2013).

Unfortunately, correlations between transcript and protein abundance are variable, and tend to be modest on average (Ghazalpour et al., 2011; Vogel and Marcotte, 2012). Factors contributing to the variability include regulation of transcription and translation, transcript stability, and protein degradation, as well as methodological limitations. Thus, the relationship between transcript copy number in the cytoplasm and protein expression levels in the cytoplasm or elsewhere in the cell is not clear for a given target. As mentioned above, the relationship between transcripts and translated proteins may be further complicated in neurons for which transcripts are concentration in somatic cytoplasm, but the proteins are expressed in a different subcellular compartment, such as axon terminals. This is the case for VGluT1, VGluT2, and VGAT (Hackett et al., 2016), as well as A1Rs.

Summary

The findings of this study are summarized in Figure 20. The main findings for neurons were that Adora1 mRNA is strongly expressed by the majority of glutamatergic neurons in layers 3–6 of A1 and all divisions of the MG, compared to modest expression by subpopulations of GABAergic neurons in A1. A1R immunoreactivity was also found across layers, but peak expression corresponded with bands of elevated VGluT2 expression in L3/L4 and L6. These patterns indicate that A1Rs are likely co‐expressed with VGluT1 or VGluT2 in axon terminals of respective glutamatergic projections. For glia, Adora1 mRNA is expressed by subpopulations of all three glial subtypes in L3–L6 of A1 and all divisions of MG, but transcript abundance is low compared to neurons.

The widespread Adora1 expression by MG neurons implies that the presynaptic modulation of glutamate release by adenosine may be a general property of all MG projections to auditory cortex, and therefore potentially impact excitatory neurotransmission and plasticity in all areas of auditory cortex. As A1R functionality in auditory cortical neurons appears to be distinct from MG neurons, focused studies will be needed to uncover the roles of pre‐ and postsynaptic receptors in the networks of corticocortical (CC), corticothalamic (CT), and corticotectal (Ct) projections. The expression of Adora1 by glial subpopulations is consistent with recognized roles in development, behavior, regeneration, plasticity, and pathology, but specific roles in the auditory forebrain are unknown.

An added application of the data from this study concerns the targeting of specific cell types for future functional studies, as the properties of A1R‐adenosine signaling in the auditory forebrain are currently limited to a subset of the TC projections from MGv to L4 of A1. Gene editing, transgenic, optogenetic, and pharmacologic manipulations of A1R‐mediated signaling would be expected to differentially impact subpopulations of neurons and glia in A1 and MG. Strategies to target Adora1 or A1Rs in specific cell populations for experimental study of function can be developed from the data generated here.