Distinct Anatomical Connectivity Patterns Differentiate Subdivisions of the Nonlemniscal Auditory Thalamus in Mice

Abstract

Systematic examination of the inputs and outputs of the nonlemniscal auditory thalamus will facilitate the functional elucidation of this complex structure in the central auditory system. In mice, comprehensive tracing studies that reveal the long-range connectivity of the nonlemniscal auditory thalamus are lacking. To this end, we used Cre-inducible anterograde and monosynaptic retrograde viruses in Calbindin-2A-dgCre-D and Calretinin-IRES-Cre mice, focusing on the differences across subdivisions of the nonlemniscal auditory thalamus. We found that, 1) the dorsal and medial parts of the auditory thalamus were predominantly connected to sensory processing centers, whereas the posterior intralaminar (PIN) and peripeduncular nucleus (PP) were additionally connected to emotion and motivation modulation centers; 2) ventral auditory cortical areas were the major source of cortical inputs for all subdivisions, and the PIN/PP received more inputs from cortical layer 5 than other subdivisions did; 3) deep layers of the superior colliculus and rostral part of the nonlemniscal inferior colliculus preferentially projected to the PIN/PP; and 4) compared with the dorsal auditory thalamus, the PIN/PP mainly innervated association cortices. In addition, new brain areas connected to the nonlemniscal auditory thalamus, mostly the PIN/PP, were identified. Our results suggested subdivision-specific function of the nonlemniscal auditory thalamus in sound processing.

calcium-binding proteins, connectivity, mouse, nonlemniscal auditory thalamus, virus-assisted tracing

The auditory thalamus consists of lemniscal and nonlemniscal relays (Guillery and Sherman 2002; Lu et al. 2009). The lemniscal relay is the ventral division of the medial geniculate body (MGBv) (Winer 1984; Rouiller et al. 1989), which plays a critical role in transferring acoustic information from the central nucleus of the inferior colliculus (ICc) to the primary auditory cortex (Au1) (McMullen and de Venecia 1993; de la Mothe et al. 2006; Lee and Winer 2008a; Llano and Sherman 2008; Razak and Fuzessery 2010; Hackett et al. 2011; Smith et al. 2012). The nonlemniscal relay surrounds the MGBv and can be subdivided into the dorsal division of the MGB (MGBd), suprageniculate nucleus (SG), medial division of the MGB (MGBm), posterior intralaminar thalamic nucleus (PIN), and peripeduncular nucleus (PP) (Bordi and LeDoux 1994b, 1994a; Linke 1999a; Lu et al. 2009; Lee 2015). Compared with the MGBv, the functions of different subdivisions of the nonlemniscal auditory thalamus are less well understood.

Considerable efforts have been made to reveal the connectivity patterns of subdivisions of the nonlemniscal auditory thalamus, with the hope of shedding light upon their functions. It has been reported that the MGBd receives projections from both the IC shell and Au1 and mainly projects to the secondary auditory cortices (Morest 1965; Oliver and Hall 1978; Linke 1999a, 1999b; Huang and Winer 2000; Llano and Sherman 2008; Smith et al. 2012). Regarding the MGBm, PIN, and PP, these subdivisions receive inputs not only from the IC (Calford and Aitkin 1983) but also the spinal cord (Oliver and Hall 1978; Bordi and LeDoux 1994b) and superior colliculus (SC) (Winer and Morest 1983a; Linke 1999a), although they are more likely to respond to auditory stimuli (Bordi and LeDoux 1994b). In terms of outputs, these subdivisions project to a wide range of auditory and association cortical areas (Lee and Winer 2008b) and to subcortical centers critically involved in emotional behavior and motion control (Kudo et al. 1989; Clugnet et al. 1990; LeDoux et al. 1991).

However, the above studies were all conducted on cats or rats. In the past 10 years, the mouse has evolved as a powerful animal model for mapping neural circuits due to the availability of various transgenic mouse lines and viral tools, which enable the selective labeling of specific cell types and connected cells. Furthermore, appropriate combinations of Cre mouse lines with optogenetic manipulation have greatly facilitated the functional elucidation of specific neural circuits in both information processing and animal behavior. Thus, to increase our understanding of the nonlemniscal auditory thalamus, systematic examination of the inputs and outputs of this complex structure in mice is necessary. Calcium-binding proteins calbindin (CB) and calretinin (CR) are abundantly expressed in the nonlemniscal auditory thalamus of humans, primates, and rodents (Morel et al. 1997; Jones 1998, 2001;Cruikshank et al. 2001; Lu et al. 2009); however, their expression is much weaker in the MGBv (Bartlett 2013). In addition, the distribution pattern of CR+ neurons is generally similar to that of CB+ neurons (Fortin et al. 1998; Munkle et al. 2000; Lu et al. 2009). These distributions are not unique to the auditory thalamus and have been used to formulate models of the whole thalamic organization (Jones 2001). Therefore, both CB and CR can be considered as high-affinity protein markers for the nonlemniscal thalamus of sensory modalities. However, whether and how CB+ and CR+ neurons are connected with other brain areas and whether these 2 cell types show similar connectivity patterns remain unclear. Here, we applied virus-assisted retrograde and anterograde tracing in transgenic C57BL/6 mice. We quantitatively characterized the inputs and qualitatively depicted the outputs of CB+/CR+ neurons. Our results indicated that most glutamatergic neurons in the nonlemniscal auditory thalamus expressed both CB and CR proteins, giving rise to similar connectivity patterns between CB+ and CR+ neurons. In addition, the connectivity patterns of different subdivisions of the nonlemniscal auditory thalamus were distinct, suggesting their differential roles in sound processing and sound-associated behaviors. Our results further suggested that both CB- and CR-Cre mice could be very useful for the functional dissection of the nonlemniscal auditory thalamus.

Materials and Methods

Mice

All animal care procedures and experiments were approved by the Institutional Animal Care and Use Committee at Tsinghua University, Beijing, China. Both before and after surgery, all mice were raised in the Laboratory Animal Research Center of Tsinghua University (SYXK, 2014-0024) with standard temperature, humidity, and light/dark conditions. Adult mice (2–5 months old) of both sexes were used. All transgenic mice were obtained from the Jackson Lab and were maintained in a C57BL/6 background (CB-2A-dgCre-D, Stock No: 023531; CR-IRES-Cre, Stock No: 010774, vesicular glutamate transporter 2 (VGluT2)-IRES-Cre, Stock No: 016963). The CB-2A-dgCre-D mice received trimethoprim (TMP, Sigma-Aldrich, USA, T7883, 0.3 mg/g body weight) over 3 consecutive days via oral gavage (p.o.) within one week after virus injection, inducing high levels of Cre recombinase activity in CB+ neurons (Sando et al. 2013; Harris et al. 2014). Wild-type C57BL/6 control mice were purchased from Wei Tong Li Hua Experimental Animal Co., Ltd (Beijing, China).

Surgery and Stereotaxic Injection of Virus

Mice were first anesthetized with pentobarbital (i.p. 80 mg/kg) and mounted in a stereotaxic apparatus (RWD, 68001, Shenzhen, China). Erythromycin eye ointment was used to prevent eye drying and an electric heating pad was applied to maintain the body temperature of mice (~37 °C). The skin over the skull midline was cut using sterilized scissors and forceps were used to expose the bregma, lambda, and skull surface over the MGB. A small hole was drilled in the skull to load the injection pipette (4878, WPI, USA).

To reveal the input neurons of CB+/CR+ neurons in the nonlemniscal auditory thalamus, a widely used rabies virus (RV)-mediated retrograde monosynaptic tracing strategy was adopted (Callaway and Luo 2015). Briefly, mixed Cre-dependent adeno-associated virus (AAV) helper viruses (AAV-DIO-EGFP-TVA and AAV-DIO-RG, BrainVTA, Wuhan, China) were used to selectively express TVA and RG in the CB+/CR+ neurons (Fig. 1A, upper left). To locally express the viruses in each subdivision of the nonlemniscal auditory thalamus, less than 50 nL of the 1:1 mixed AAV helper virus was injected into the target area (anterior–posterior/medial–lateral/dorsal–ventral coordinates (mm): MGBd, −3.1/±2.15/2.8; MGBm, −3/±1.9/2.9; PIN/PP, -3/±2.1/3.2) with a microsyringe pump (UMP3 and Micro4, WPI, USA) in either hemisphere (Fig. 1B, left). Three weeks later, following biosafety level-2 lab procedures (IACUC, 15-YKX1), 100 nL of pseudotyped RV equipped with the TVA selective avian ASLV type A (EnvA) was injected into the same area to enable cell type-specific retrograde tracing. The rabies RG gene required for transsynaptic spreading beyond initially infected neurons was replaced by the red fluorescent DsRed transgene (SADΔG-DsRed RV, BrainVTA, Wuhan, China). As a result, DsRed was introduced to label the monosynaptic input neurons of the starter neurons, which were both green and red fluorescence-positive (Fig. 1B, upper right). RV was allowed to express and retrogradely spread for another 7 days before the mice were sacrificed (Fig. 1A, upper right).

Figure 1.

Experimental procedures for cell type-specific tracing in the nonlemniscal auditory thalamus. (A) Upper panel, RV-mediated monosynaptic retrograde tracing. Left, construction of AAV helper viruses (AAV-DIO-EGFP-TVA and AAV-DIO-RG) and pseudotyped RV; Right, time course of virus injection and histology. Lower panel, AAV-mediated anterograde tracing. Left, construction of AAV; Right, time course of virus injection and histology. (B) Left panel, coronal diagram of a brain section containing the auditory thalamus (red box). Inset, enlarged view of the auditory thalamus. Right panel, representative fluorescent images of the auditory thalamus showing localized infection of RV and AAV in the MGBd/SG. Scale bars, 100 μm. Insets, enlarged view of boxed areas. White arrows in inset, starter neurons expressing both EGFP and DsRed. Scale bars in insets, 20 μm.

For anterograde tracing, less than 50 nL of AAV-DIO-EYFP (Upenn Viral Core, USA) was stereotaxically injected into subdivisions of the nonlemniscal auditory thalamus following a procedure similar to that used for retrograde tracing (Fig. 1A, lower left). Mice survived for 28 days after surgery and virus injection to achieve strong expression of fluorescence (Fig. 1A, lower right; Fig. 1B, lower right).

To verify the specificity of Cre mouse lines and examine the coexpression levels of VGluT2, CB, and CR, 200 nL of AAV-DIO-tdTomato (UNC Vector Core Facility, USA) was injected into the nonlemniscal auditory thalamus to label putative VGluT2+, CB+, and CR+ neurons. After 10 days of virus expression, brains were sliced and immunostained (see below for details).

To test the level of Cre-dependency of AAVs used in the present study, the viruses were also injected into the nonlemniscal auditory thalamus of wild-type mice, which showed no fluorescence-positive neurons in the injected areas following the same virus expression and tissue processing procedures (Supplementary Fig. S1A). Regarding RV, omission of AAV-DIO-RG prevented the labeling of any distant input neurons even though many starter neurons were found in the nonlemniscal auditory thalamus (Supplementary Fig. S1B), suggesting the necessity of RG for the transsynaptic spreading of RV.

Histology, Immunostaining, and Fluorescence In-Situ Hybridization

In the histological experiments, mice were deeply anesthetized with an overdose of chloral hydrate (i.p., 10% W/V, 300 mg/kg body weight) and perfused transcardially with 0.9% saline followed by 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS). After postfixation overnight, brains were cryoprotected in 20% and 30% sucrose in 0.1 M PBS successively till they sank. Brains were then frozen in optimal cutting temperature (OCT) compound (Sakura, Tokyo, Japan), and coronally sliced with a cryostat (CM1950, Leica Biosystems, Germany) at a thickness of 50 μm. Every other section was collected and mounted on gelatin-coated slides. The sections were then further washed with PBS for 10 min and coverslipped with 50% glycerol mounting medium. In some slices, 1:5000 DAPI (D1306, Invitrogen, USA) was added to visualize nuclei.

To determine cortical layers, Nissl staining was performed for 2 mice (3 months old). Frozen 35 μm-thick sections were washed with 0.1 M PBS (pH 7.4) for at least 40 min, then treated with PBS containing 0.3% Triton^® X-100 for 20 min. NeuroTrace stain (1:200, N-21479, Molecular Probes, USA) diluted with PBS was added to the buffer and the sections were incubated for another 20 min. Finally, the sections were immersed in PBS overnight at 4 °C and mounted on gelatin-coated slides.

For immunofluorescence histochemical (IHC) staining of CB, the mouse brains were first postfixed for 4 h in 4% PFA (RT) after perfusion, then cryoprotected by infiltration with 30% sucrose overnight. Brains were frozen with cryoembedding media at −20 °C and sectioned at 30 μm using a freezing microtome. The sections were collected in 6-well plates containing 0.1 M PBS (precooled) and washed in the same buffer for 30 min. Sections were immersed in PBS containing 0.3% Triton-X for 30 min to increase membrane permeability and prepared for later immunostaining. The sections were blocked with 3% BSA in PBS containing 0.3% Triton-X for 60 min, then incubated with the primary antibody for 24–48 h at 4 °C, washed with PBS, and incubated with corresponding secondary antibodies (1:500) for 2 h. Finally, sections were washed thoroughly with PBS, mounted on gelatin-coated slides, and coverslipped using mounting solution (50% glycerol containing 0.2% DAPI).

The primary antibodies used in the present study were monoclonal anti-CB-D28k antibody (1:1000, C9848, Sigma-Aldrich, USA), mouse anti-CB-D28k (1:1000, 300, Swant, Switzerland), and rabbit anti-CR (1:1000, 7697, Swant, Switzerland). The monoclonal anti-CB-D28k antibody was generated against purified bovine kidney CB-D28k. Per the manufacturer, this monoclonal antibody does not react with other members of the EF-hand family such as CB-D9K, calretinin, myosin light chain, parvalbumin, S-100a, S-100b, S100A2 (S100L), or S100A6 (calcyclin). Every product was tested by immunoblotting (which used an antigen extracted from bovine kidney cells and yielded a single band at 28 kDa). The monoclonal antibody has been used on mouse tissues in many studies (Sadakata et al. 2007; Lu et al. 2009; Zhao et al. 2013). Specificities of primary monoclonal anti-CB-D28k antibody (300, Swant, Switzerland) and rabbit anti-CR antibody (7697, Swant, Switzerland) have been validated by the manufacturer using CB-D28k and CR-knockout mouse lines. More details can be found on the website of the manufacturer (https://www.sigmaaldrich.com). In the present study, no staining of CB+ cells was found in CB-Cre mice due to an unexpected effect of TMP on the CB protein, but both primary antibodies worked well in wild-type C57BL/6, CR-, and VGluT2-Cre mice. The secondary antibodies were anti-rabbit antibody (1:500, 111-545-045, Jackson ImmunoResearch, USA) for the CR primary antibody and anti-mouse antibody (1:500, A11001, Invitrogen, USA) for the CB primary antibody. Control experiments were run by eliminating the primary antibody, with no signal observed.

Fluorescence in-situ hybridization (FISH) was applied to verify the VGluT2 specificity of VGluT2-Cre mice. To create FISH probes, the 5′-overhang of forward primer was modified with a t7 promoter. DNA fragments of the target gene were obtained by PCR from whole brain cDNA of the mice. Afterwards, DNA was transcribed by DIG-RNA Labeling Mix (11277073910, Roche, Switzerland) to produce the VGluT2 probes (accession number: NM_080853-3; probe region: 720–1635). The FISH experiment followed previous study (Xiu et al. 2014). Sample preparations were similar to IHC staining, except for the usage of DEPC-treated buffers to prevent the degradation of RNA during the perfusion and cryoprotection procedures. The next step was to incubate the sections with prehybridization buffer (50% formamide, 5× SSC, 5 mM EDTA pH 8.0, 0.1% Tween 20, 1% CHAPS) for 2 h, with the probe then diluted in hybridization solution (50% formamide, 5× SSC, 5 mM EDTA pH 8.0, 0.1% Tween 20, 1% CHAPS, 300 μg/mL tRNA, 1 × Denhalt’s solution, 1% Heparin) to 1 μg/mL and crossed with brain sections for 20 h at 65 °C. After hybridization, sections were incubated with prehybridization buffer at 65 °C for 30 min, with the same operation carried out in a mixture of TBST (Tris 1 M + Nacl 2.5 M + 10% TWEEN + Ultrapure Water) and prehybridization buffer at a ratio of 1:1, then twice washed with TBST. After thrice washing with Tris-Acetate EDTA (TAE) buffer, the sections were then washed once with a mixture of TBST and TAE (1:1). Electrophoresis was used to remove foreign material from the sections. Afterwards, the sections were incubated with anti-digoxigenin-POD (10520200, 1:500, Roche, Switzerland) at 4 °C for 30 h. After washing with TNT, TSA-Plus Cyanine 5 (NEL745B001KT, 1:100, PerkinElmer, CA, USA) was used to detect the primary antibody. Anti-DsRed antibody (1:250, 632496, Clontech, USA) was incubated overnight at 4 °C to detect the fluorescence of AAV (tdTomato), which had been quenched by high temperature during hybridization. Finally, sections were incubated with Alexa Fluor 594 conjugated secondary antibody (1:500, 111-545-045, Jackson ImmunoResearch, USA).

Imaging

Fluorescence images for retrograde and anterograde tracing were obtained with an automated slide scanner (VS120 Virtual Slide, Olympus, Japan), while IHC and FISH images were captured by a multiphoton confocal microscope (A1RMPSi, Nikon, Japan). During the processing of tracing data, Image J (NIH) software was used to export high-resolution images for further analysis. To unify the number of brain sections to be analyzed for each mouse, only sections corresponding to AP positions in The Mouse Brain in Stereotaxic Coordinates were used to count input neurons and calculate the fiber fluorescence intensity-index for fibers.

Cell-Counting to Determine Coexpression Levels of CB, CR, and VGluT2

Given the physical proximity between the MGBd and SG and between the PIN and PP, viral infection was usually not restricted to a single subdivision, even if the injection was targeted at a specific subdivision. Therefore, we combined data from the MGBd and SG and from the PIN and PP, respectively, for data analysis, and use the terms MGBd/SG and PIN/PP for later data presentation. Cell number was averaged across all mice in the same set of experiments. To determine the coexpression levels of CB, CR, and VGluT2, 2 rules were applied for cell-counting (Lu et al. 2009; Bae et al. 2015): 1) for each subdivision, the section with the densest virus labeling was selected, and only cells in a 100 × 100-μm² box, centered in the subdivision, were counted; and 2) only “complete” cells with a visible nucleus were counted. We used a different data analysis approach for retrogradely labeled neurons, which is shown as follows.

Data Acquisition and Analysis for Retrograde and Anterograde Tracing

For analysis of retrograde and anterograde tracing results, a program was developed in MATLAB (MathWorks, USA) to process the digitized brain images. The data analysis procedure consisted of module-image registration, signal detection, and signal registration, which were implemented by the following 5 steps (Supplementary Fig. S2): 1) the brain atlas (Paxinos and Franklin 2001) that matched the section image to be analyzed was selected; 2) reference points in the section image and selected atlas were marked. Landmarks of brain sections, such as the rhinal fissure and ventricle, are often used as reference points (commonly used reference points are shown in Supplementary Fig. S2); 3) errors caused by brain sectioning and handling, such as compression and stretching, were corrected, and reference points-based triangular segmentation of section images was performed, followed by a nonlinear affine transformation. As a result, each pixel of the original image was mapped onto the standard brain atlas; 4) 10 equidistant detection thresholds were set in the program to detect input neurons fluorescently labeled by RV. Thus, pixels that exceeded each threshold and formed a connected area in the size range of a neuron were extracted. This multiple-threshold approach yielded more precise and complete signal detection than using the single-threshold approach. For starter neurons, only those that showed suprathreshold signals for both red and green fluorescence were counted. For neurons too dense to be discriminated by the program, quantification was conducted manually. For anterograde tracing, an optimal threshold was determined by several trials to allow the extraction of fluorescent fibers without false positive signals; and 5) all signals detected in each section were mapped onto the standard brain atlas. To evaluate the similarity in the number of retrogradely labeled neurons among input sources, we used Pearson’s correlation coefficient of normalized inputs, calculated as the total number of input neurons from a specific brain area divided by the total number of input neurons. Based on the correlation coefficients, we constructed cluster trees using the linkage function.For statistical analysis, two-sided non-parametric test was performed using GraphPad Prism 6 (GraphPad Software, Inc.). Data with error bars are presented as mean ± S.E.M. (standard error of the mean).

Abbreviations

Most brain region abbreviations were from The Mouse Brain in Stereotaxic Coordinates, second edition, Franklin, K. B. J. and Paxinos, G. Please see Table 1 for a detailed list of abbreviations. All brain outlines in this paper were also adapted from this atlas.

Table 1

Abbreviations

Abbreviation	Definition
AC	Auditory cortex
Au1	Primary auditory cortex
AuD	Secondary auditory cortex, dorsal area
AuV	Secondary auditory cortex, ventral area
CnF	Cuneiform nucleus
CPu	Caudate putamen (striatum)
DpG	Deep gray layer of superior colliculus
DpWh	Deep white layer of superior colliculus
DR	Dorsal raphe nucleus
Ect	Ectorhinal cortex
HTH	Hypothalamus
ic	Internal capsule
IC	Inferior colliculus
ICc	Central nucleus of inferior colliculus
ICd	Dorsal nucleus of inferior colliculus
ICl	Lateral nucleus of inferior colliculus
ICr	Rostral pole of inferior colliculus
InG	Intermediate gray layer of superior colliculus
InWh	Intermediate white layer of superior colliculus
La	Lateral amygdaloid nucleus
LC	Locus coeruleus
LDTg	Laterodorsal tegmental nucleus
LGP	Lateral globus pallidus
LP	Lateral posterior thalamic nucleus
MGBd/m/v	Dorsal/medial/ventral division of medial geniculate body
NLL	Nuclei of lateral lemniscus
PAG	Periaqueductal gray
PIN	Posterior intralaminar nucleus
PP	Peripeduncular nucleus
PPTg	Pedunculopontine tegmental nucleus
PSTN	Parasubthalamic nucleus
PT	Pretectal nucleus
S1BF	Primary somatosensory cortex barrel field
S2	Secondary somatosensory cortex
SC	Superior colliculus
SG	Suprageniculate thalamic nucleus
SOC	Superior olivary complex
SPF	Subparafascicular thalamic nucleus
Rt	Reticular thalamic nucleus
TeA	Temporal association cortex
V2L	Secondary visual cortex, lateral area
Va	Ventral amygdaloid nucleus
VMH	Ventromedial hypothalamic nucleus
ZI	Zona incerta

Results

Distribution and Colocalization of CB/CR/VGluT2 in the Nonlemniscal Auditory Thalamus

Prior to further experiments and data analysis, the CB-, CR-, and VGluT2-Cre mouse lines were validated to confirm specific expression of Cre in cells containing the CB, CR, and VGluT2 proteins, respectively. To this end, a volume of Cre-dependent AAV-DIO-EGFP was injected into Cre-mice targeting the MGB. This yielded strong labeling of many neurons surrounding the MGBv of both CB- and CR-Cre mice (Fig. 2A, left and middle). According to the brain atlas, most of the fluorescently labeled neurons resided in the nonlemniscal part of the auditory thalamus, including the MGBd, SG, PIN, and PP, consistent with the reported distribution pattern of the CB and CR proteins in the MGB (Cruikshank et al. 2001; Lu et al. 2009). CR-Cre mice were further validated by 99.24% (±2%) double labeling of CR+ neurons by AAV-DIO-tdTomato-mediated fluorescence expression (red) and fluorescent immunostaining (green) (Supplementary Fig. S3A, top row, and Fig. 2B). However, we were unable to further validate CB-Cre mice using the same procedure due to an unexpected effect of TMP (see Materials and Methods) on the CB protein. In addition, we found that the MGBm neurons were not strongly labeled unless the virus was directly deposited into the MGBm (Fig. 2A, right). The density of labeled neurons in the MGBm was relatively low compared with that in other subdivisions.

Figure 2.

Distribution of CB+/CR+ neurons in the nonlemniscal auditory thalamus. (A) Selective fluorescence expression in the nonlemniscal subdivisions surrounding the MGBv of CB-/CR-Cre mice. Note the difference in the density of labeled MGBm neurons between MGBd-targeting (left and middle) and MGBm-targeting cases (right). Green, AAV-DIO-EGFP. Scale bars, 50 μm. (B) Nonselective fluorescence expression in both the lemniscal and nonlemniscal auditory thalamus of VGluT2-Cre mice. Red, AAV-DIO-tdTomato. Scale bar, 50 μm. (C) Expression level of CB (CB-IHC) in CR+/VGluT2+ neurons (CR-/VGluT2-Cre) and CR (CR-IHC) in CB+/VGluT2+ neurons (CB-/VGluT2-Cre) for each subdivision. Sample size: CB-IHC/CR-Cre, n = 8; CR-IHC/CB-Cre, n = 8; CB-IHC/VGluT2-Cre, n = 7; CR-IHC/VGluT2-Cre, n = 4. Data are means ± S.E.M. Statistical analysis by Kruskal–Wallis test (***P < 0.001; n.s. P > 0.05).

AAV-DIO-tdTomato injection targeting the MGBv of VGluT2-Cre mice led to neuronal labeling in both the MGBv and nonlemniscal subdivisions (Fig. 2B), consistent with reports that VGluT2 is strongly expressed in all divisions of the MGB (Ito et al. 2011; Hackett et al. 2016). In addition, re-staining (green) of AAV-mediated tdTomato and FISH staining (red) yielded 92.3% (±3.5%) double labeling of VGluT2+ neurons, largely validating the specific expression of Cre in VGluT2-expressing neurons (Supplementary Fig. S3A, bottom row, and S3B).

To examine the colocalization levels of CB, CR, and VGluT2 in different subdivisions of the nonlemniscal auditory thalamus, AAV-DIO-tdTomato was first injected into these subdivisions to allow sufficient fluorescence expression. Immunostaining was then performed for CB in CR- and VGluT2-Cre mice and for CR in CB- and VGluT2-Cre mice. Example images demonstrating double-labeled neurons are shown in Supplementary Fig. S3C (MGBd/SG, top row; MGBm, middle row; PIN/PP, bottom row). There were significantly more CR+ neurons that coexpressed CB in the MGBd/SG than in the MGBm (MGBd/SG, 83%; MGBm, 71%; PIN/PP, 77%; P = 0.0038, Kruskal–Wallis test followed by Dunn’s test to correct for multiple comparisons; Fig. 2C, left most group of histograms), whereas no significant differences were observed for the proportions of CB+ neurons coexpressing CR across the 3 subdivisions (Fig. 2C, second group). The proportion of VGluT2+ neurons that coexpressed CB or CR was considerable, but no significant differences were observed across the 3 subdivisions (Fig. 2C; CB-IHC, third group; CR-IHC, right most group), suggesting that the majority of neurons expressing CB and CR in the nonlemniscal auditory thalamus were excitatory neurons.

Major Sensory Inputs of the Nonlemniscal Auditory Thalamus

As expected, CB+ and CR+ neurons indeed received direct inputs from other brain areas. Supplementary Fig. S4A demonstrates the distribution of starter neurons in the MGBd/SG of 10 mice (7 CB-Cre, 3 CR-Cre), PIN/PP of 10 mice (4 CB-Cre, 6 CR-Cre), and MGBm of 2 mice (CB-Cre). In these cases, although some starter neurons were in other thalamic areas due to virus spill-over, most starter neurons were located well within the range of the nonlemniscal auditory thalamus. For one representative CB-Cre mouse (Fig. 3A), the input neurons (red) of starter neurons (yellow) in the PIN/PP were mapped onto the brain atlas at selected AP positions, and were detected from the section images exhibited in Figure 3B (middle row).

Figure 3.

Major sensory inputs of the nonlemniscal auditory thalamus. (A) Distribution of input neurons (red dots) mapped onto the coronal brain atlas at 3 selected AP positions. Blue boxed areas highlight input neurons in the AC, SC, and IC. (B) Example images showing distribution of starter neurons (left most column) and input neurons in the boxed areas shown in (A) (last 3 columns). Insets, enlarged view of the dotted box. Scale bars, 100 μm.

Many interesting differences in the distribution of input neurons were observed across the 3 subdivisions. In the cortex, a large number of pyramidal neurons in deep layers were retrogradely labeled by RV injections into the MGBd/SG and PIN/PP but not the MGBm (Fig. 3B, second column), suggesting that the MGBm may not be involved in the generation of thalamocortical oscillations (Crunelli et al. 2015). Input neurons of both the MGBd/SG and PIN/PP were observed in a wide range of auditory and association cortical areas, including the dorsal auditory cortex (AuD), Au1, ventral auditory cortex (AuV), temporal association cortex (TeA), and ectorhinal cortex (Ect) (Figs 3 and 4A). Nevertheless, the AuV and TeA collectively contained the most neurons projecting to both the MGBd/SG and PIN/PP (MGBd/SG, 76.0%; PIN/PP, 70.0%), followed by the Au1 (MGBd/SG, 11.4%; PIN/PP, 8.8%) (Fig. 4B). The major distinction between the MGBd/SG and PIN/PP was that, while the MGBd/SG received its inputs primarily from L6 (AuD, 97.1%; Au1, 95.7%; AuV/TeA, 85.9%; Ect, 92%), a large proportion of input neurons of the PIN/PP were observed in L5 (AuD, 38.6%; Au1, 36.7%; AuV/TeA, 34%; Ect, 52.8%) (Fig. 4C and Supplementary Fig. S5). This difference in cortical inputs suggested that the MGBd/SG mainly received modulatory inputs from the cortex, whereas the PIN/PP might receive much stronger driver inputs. Although the MGBd/SG is proposed to be the main target of L5 neurons in the Au1 (Sherman 2016), very few studies have made comparisons between the MGBd/SG and PIN/PP. In addition, compared with the MGBd/SG, the PIN/PP received much stronger inputs from the more anterior part of the Au1, AuV/TeA, and Ect, suggesting spatial separation of cortical input neurons between these 2 subdivisions (Fig. 4D). It is still not clear whether these observations are species-specific, and their functional implications remain to be explored.

Figure 4.

Distribution patterns of cortical neurons projecting to the nonlemniscal auditory thalamus. (A) Cortical input neurons at 6 selected AP positions. Data were from 2 representative CB-Cre mice. Green dots, input neurons of the MGBd/SG; red dots, input neurons of the PIN/PP. Light purple, AuD; light orange, Au1; light green, AuV/TeA; light blue, Ect. (B) Quantified distribution of input neurons in different cortical areas. (C) Laminar-dependent distribution of input neurons in the L5 and L6 of different cortical areas. Horizontal lines, mean values. Gray solid circle/rectangle, L5/6 neurons projecting to the MGBd/SG; Black hollow diamond/triangle, L5/6 neurons projecting to the PIN/PP. (D) Distribution pattern of input neurons along the AP direction in different cortical areas. Data are means ± S.E.M. Sample size (B–D): MGBd/SG, n = 10; PIN/PP, n = 7. Data from CB- and CR-Cre mice were analyzed collectively.

In the midbrain, RV injection in the MGBm yielded fewer labeled neurons in the SC but dense labeling in the IC (Fig. 3B, last 2 columns), suggesting that auditory input was predominant for the MGBm. Comparing the MGBd/SG with PIN/PP, although strong retrograde labeling was observed in both the SC and IC, the distribution pattern of input neurons differed. In the SC, neurons projecting to the MGBd/SG were predominantly found in the optic nerve (OP) layer (83.3%), whereas those projecting to the PIN/PP were located in both the OP layer (48.3%) and deeper layers, including the intermediate gray layer (InG), intermediate white layer (InWh), deep gray layer (DpG), and deep white layer (DpWh) (Fig. 3B, third column, and Fig. 5A,B). This difference suggests that the MGBd/SG and PIN/PP might integrate information from 2 functionally distinct systems in the SC, which are specialized for the processing of visual information (OP layer) and visually guided actions (deeper layers), respectively (Freedman et al. 1996; Wang et al. 2010). In addition, we noted that the emergence of input neurons in the OP layer was significantly more anterior than that in deeper layers for both MGBd/SG and PIN/PP-injection cases (Fig. 5C; MGBd/SG, P = 0.0207; PIN/PP, P = 0.026; Mann–Whitney test).

Figure 5.

Distribution patterns of input neurons in the SC and IC. (A) Input neurons in the SC at 5 selected AP positions. Green dots, input neurons of the MGBd/SG; red dots, input neurons of the PIN/PP. (B) Laminar-dependent distribution of input neurons in the SC. Sample size: MGBd/SG, n = 8; PIN/PP, n = 9. Data are means ± S.E.M. Statistical analysis by Mann–Whitney test (***P < 0.001). (C) Distribution patterns of input neurons in the SC along the AP direction. Dots, AP positions of the weighted median of normalized inputs from an individual mouse. Short vertical line, median of AP positions. Colored lines, distribution of input neurons in the OP and deep layers of the SC in the AP direction. (D) Input neurons in the IC at 6 selected AP positions. Colors of dots have the same meaning as those in (A). (E) Subdivision-dependent distribution of input neurons in the IC. Sample size: MGBd/SG, n = 7; PIN/PP, n = 7. Data are means ± S.E.M. Statistical analysis by Mann–Whitney test (**P < 0.01; *P < 0.05). (F) Distribution patterns of input neurons in the IC along the AP direction. Three types of lines, distribution of input neurons in 3 subdivisions of the IC. Green, inputs of the MGBd/SG; red, inputs of the PIN/PP. Data from CB- and CR-Cre mice were analyzed collectively for (B and C) and (E and F). Input neurons in (A) and (D) were from the same CB-Cre mice used in Figure 4A.

In the IC, although neurons projecting to the MGBd/SG and PIN/PP were found in all 3 subdivisions of the nonlemniscal IC (rostral pole of the IC, ICr; dorsal IC, ICd; lateral IC, ICl) (Fig. 3, last column and Fig. 5D), which surrounded the ICc, significantly more neurons in the ICr were retrogradely labeled following RV injection in the PIN/PP (Fig. 5E; ICr = 49.8%, P = 0.002, Kruskal–Wallis test followed by Dunn’s test). There were no significant differences across the 3 subdivisions of the nonlemniscal IC following injection in the MGBd/SG. Interestingly, the ICr also showed substantial projections to the intermediate layers of the SC (Harting and Van Lieshout 2000), which, in turn, projected to the PIN/PP (Fig. 5A). As the ICr receives direct inputs from the dorsal cochlear nucleus (Oliver 1984), which itself is multimodal (Burian and Gstoettner 1988; Wright and Ryugo 1996; Ohlrogge et al. 2001), as well as inputs from the medial and lateral superior olives (Henkel and Spangler 1983; Shneiderman and Henkel 1987), the direct and indirect ICr inputs to the PIN/PP could play important roles in determining the auditory latencies, receptive field properties, and multimodal characteristics of PIN/PP neurons (Bordi and LeDoux 1994b; Anderson and Linden 2011). In addition, in all 3 subdivisions, input neurons of the MGBd/SG and PIN/PP appeared and ended at similar anterior–posterior coordinates (Fig. 5F).

Nonsensory Inputs of the Nonlemniscal Auditory Thalamus

Besides sensory inputs, nonsensory inputs from more than 10 sources were also identified. Figure 6 shows PIN/PP projecting neurons in selected nuclei associated with action initiation (top row), including the caudate-putamen (CPu) (Grahn et al. 2008), lateral globus pallidus (LGP) (Hegeman et al. 2016), and cuneiform nucleus (CnF) (Kiehn 2016), and neurons in nuclei associated with different types of arousal (bottom row), including the subparafascicular thalamic nucleus (SPF) (Coolen, Veening, Petersen, et al. 2003; Coolen, Veening, Wells, et al. 2003), ventromedial hypothalamus (VMH) (King 2006; Kunwar et al. 2015), and dorsal raphe nucleus (DR) (Li et al. 2016). In addition, the PIN/PP also received projections from the locus coeruleus (LC) and pedunculopontine tegmental nucleus (PPTg), which are critically involved in the sleep-awake cycle (Benarroch 2009; Scammell et al. 2017), and from the parasubthalamic nucleus (PSTN), which plays a role in feeding behavior (Goto and Swanson 2004). Compared with the PIN/PP, the connectivity between these nuclei and the MGBd/SG and MGBm were either very weak or nonexistent.

Figure 6.

Example images showing nonsensory inputs of the CB+ neurons in the PIN/PP. Top row, nuclei associated with locomotion modulation; bottom row, nuclei associated with different types of arousal. Scale bars, 100 μm.

Distinct Input Patterns Across Subdivisions of the Nonlemniscal Auditory Thalamus

Regarding the inputs, one interesting question was if the 3 subdivisions exhibited any patterned preference. To this end, the inputs of each subdivision were normalized as the number of input neurons in each individual source divided by the total number of input neurons of that subdivision. Figure 7A shows the distribution of inputs from selected sources after normalization. Notably, although the IC made strong projections to all 3 subdivisions, its contribution to inputs of the MGBm was highest (MGBm, 47.8%; MGBd/SG, 20.7%; PIN/PP, 21.3%), suggesting that ascending information played a major role in determining the response properties of MGBm neurons. Interestingly, the SC projected strongly to all 3 subdivisions (MGBm, 14.6%; MGBd/SG, 32.1%; PIN/PP, 23%), indicating non-negligible vision-audition integration in the nonlemniscal auditory thalamus. Supposedly, mice use this integration to enhance cortical representation of sound made by visual targets (Komura et al. 2005). Among cortical input sources, the AuV/TeA provided the most inputs for all 3 subdivisions, and the MGBd/SG seemed to be a preferable target, suggesting stronger corticofugal innervation from the secondary auditory cortices and association cortical areas. In addition, the PIN/PP and MGBm received higher proportions of inputs from the nonsensory nuclei associated with motion initiation and arousal, including the CPu, LGP, CnF, SPF, hypothalamus (HTH), DR, and laterodorsal tegmental nuclei (LDTg), suggesting their tighter functional association with behavioral states.

Figure 7.

Distinct input patterns across subdivisions of the nonlemniscal auditory thalamus. (A) Whole-brain distribution of normalized inputs of CB+ neurons in the MGBd/SG (green), PIN/PP (red), and MGBm (light blue). Sample size: MGBd/SG^CB, n = 7; PIN/PP^CB, n = 4; MGBm^CB, n = 2. Data are means ± S.E.M. (B) Cell type-specific connectivity preference between subdivisions of the nonlemniscal auditory thalamus and other brain areas. Filled circle, significant projection preference of AuV/TeA for CR+ neurons in the MGBd/SG (P < 0.05). Hollow circles and diamonds, brain areas not showing significant projection preference for either CB+ or CR+ neurons. Statistical analysis by Mann–Whitney test. Dashed line, y = x. (C) Division-specific connectivity preference. Inset, enlarged view of the black boxed area. Filled green and red circles, brain areas showing significant projection preference for the MGBd/SG and PIN/PP, respectively (P < 0.05). Black hollow circles, brain areas not showing significant projection preference for either the MGBd/SG or PIN/PP. Statistical analysis by Mann–Whitney test. Dashed line y = x. (D) and (E), Left column, cluster trees of dissimilarity among all input brain areas. Brain areas grouped by the same color demonstrated high levels of similarity in projection preference across difference cases. Right column, matrix of Pearson’s correlation coefficients between each pair of input brain areas. Higher correlation is represented by warmer color. Sample size in (B–E): MGBd/SG^CB, n = 7; PIN/PP^CB, n = 4. MGBd/SG^CR, n = 3; PIN/PP^CR, n = 6. Data from CB- and CR-Cre mice were analyzed collectively for (C), (D), and (E).

It should be noted that most of the data presented thus far were from CB-Cre mice, and possible differences in the connectivity patterns between CB+ and CR+ neurons remained unclear. To address this question, we constructed averaged-input vectors, which were composed of normalized inputs from all input sources, for both cell types, then performed correlation analysis based on the vectors. As we failed to achieve localized virus injection in the MGBm of CR-Cre mice, the following data analyses were only performed for the MGBd/SG and PIN/PP. We found that the connectivity preferences for input sources were highly correlated between the 2 cell types in both subdivisions (MGBd/SG: r = 0.86; PIN/PP: r = 0.95) (Fig. 7B). The only significant difference was that CR+ neurons in the MGBd/SG received stronger inputs from the AuV/TeA (Supplementary Fig. S6A, the second column and Fig. 7B) (P = 0.0286, Mann–Whitney test). The reason for this difference remains unknown. Given the similarity between the 2 cell types in connectivity, data from CB- and CR-Cre mice were considered collectively to reveal the distinctions, if any, between the MGBd/SG and PIN/PP in connectivity patterns.

To identify subdivision-specific connectivity patterns, we plotted normalized inputs of the PIN/PP against those of the MGBd/SG (Fig. 7C). The weight of sensory inputs was clearly skewed toward the MGBd/SG because a significantly higher proportion of input neurons was observed in the auditory cortices and visual areas (solid green dots), including the secondary visual cortex (V2L), SC, and pretectal nucleus (PT). In contrast, the PIN/PP seemed to be under significantly stronger modulation of the animals’ behavioral states because of the considerable amount of inputs from brain areas associated with motion and arousal control (solid red dots). Interestingly, the MGBd/SG and PIN/PP were preferentially innervated by 2 thalamic GABAergic centers, the reticular thalamic nucleus (Rt/ic) and zona incerta (ZI), respectively, suggesting different mechanisms of circuit operation. Therefore, the MGBd/SG may be more relevant to the processing of sensory stimulation itself, whereas the PIN/PP may play an important role in thalamocortical modulation depending on the behavioral state of the mouse.

Although data from the MGBd and SG and from the PIN and PP were collectively considered due to virus spill-over, the potential differences between the combined subdivisions are still of interest. To this end, we examined the correlation between each pair of projecting brain areas using data from all cases (Fig. 7D,E, right) and constructed cluster trees of dissimilarity (Fig. 7D,E, left), as our conditions conformed to the assumption that each injection only covered a subset of neurons (Weissbourd et al. 2014; Menegas et al. 2015). For the cases of injection in the PIN/PP, 2 large clusters were identified. Cluster-1 mainly consisted of nuclei associated with brain states (Fig. 7D, left, red) and cluster-2 was predominantly composed of nuclei associated with sensory processing (Fig. 7D, left, light blue). Cluster-2 could be further divided into 2 subclusters. Subcluster-1 included the S2, NLL, and IC, and subcluster-2 included the remaining nuclei of cluster-2. Interestingly, relatively higher proportions of inputs were from cluster-1 and subcluster-1 nuclei when starter neurons were mainly distributed in the PIN (Supplementary Fig. S7A). Conversely, focused PP injection tended to yield a higher (or the same) proportion of input neurons in subcluster-2 nuclei. However, these comparisons were not statistically tested due to the limited number of cases. Nonetheless, the data trend suggested that the PP might be more closely associated with vision-audition integration and under stronger corticofugal control, whereas the PIN might be strongly modulated by brain states and play a more important role in relaying ascending acoustic information. Regarding the cases of injection in the MGBd/SG, input clustering was much less organized (Fig. 7E), which was in accordance with the very similar distribution pattern of inputs when starter neurons were mainly located in either the MGBd or SG (Supplementary Fig. S7B). Given the predominance of sensory inputs, these data suggest that the MGBd and SG may be functionally similar in sensory processing.

In the brain, most nuclei receive projections from the contralateral hemisphere. For the nonlemniscal auditory thalamus, contralateral inputs from many brain areas were observed. Figure 8A shows retrogradely labeled neurons on both sides of the selected sections when RV was injected in the right MGB of a CB-Cre mouse. Although labeling in the contralateral hemisphere was much sparser, the number of neurons was not negligible. Calculation of normalized inputs revealed that the SC made the largest contribution to the contralateral inputs of both the MGB/SG and PIN/PP, the PT only projected to the contralateral MGBd/SG, and only the PIN/PP received contralateral inputs from the MGB, ZI, SOC, and HTH (Fig. 8B). Further analysis demonstrated that the superior olivary complex (SOC) on the 2 sides sent comparable projections to the PIN/PP, the CnF and PT made notable projections to the contralateral MGBd/SG, and the contralateral inputs from the SC, IC, SPF, and ZI were relatively weak (Fig. 8C). These results suggest that the MGBd/SG and PIN/PP may be under strong modulation of certain contralateral nuclei, but the functional implications of these contralateral inputs remain unknown.

Figure 8.

Contralateral inputs of the nonlemniscal auditory thalamus. (A) Representative images showing bilateral inputs of CB+ neurons in the PIN/PP. (B) Normalized contralateral inputs of the MGBd/SG (n = 5) and PIN/PP (n = 6). (C) Contralateral (contra)-ipsilateral (ipsi) input ratio of the MGBd/SG and PIN/PP. Data are means ± S.E.M.

Based on retrograde tracing data, we constructed a schematic diagram to illustrate the projection preference of other brain areas for the MGBd/SG and PIN/PP (Fig. 9). In this diagram, brain areas responsible for auditory, visual, and multimodal processing contributed predominantly to the inputs of the MGBd/SG. In contrast, the PIN/PP received inputs not only from sensory modalities, but also from multiple brain areas involved in emotion, motivation, and locomotion control.

Figure 9.

Schematic diagram illustrating the projection preference of other brain areas (ipsilateral) for the MGBd/SG and PIN/PP. Red, PIN/PP is preferred target; green, MGBd/SG is preferred target. Left, input brain areas; right, diagram of auditory thalamus. Darker color indicates stronger preference.

Outputs of the Nonlemniscal Auditory Thalamus

As a distinct input pattern was observed for the subdivisions of the nonlemniscal auditory thalamus, we speculated whether the output pattern would be different as well. Figure 10A shows images of axonal fibers in selected nuclei following AAV-DIO-EYFP injection into the PIN/PP of CB-Cre mice. Fibers in the VMH, lateral amygdala (La), SPF, dorsolateral periaqueductal gray (DLPAG), SC, and LGP were highly branched and exhibited numerous varicosities, suggesting that they were predominantly axonal terminals rather than fibers-of-passage. In the Rt/ic and ZI, most fibers were coarser smooth and fasciculated, indicating that they were mainly fibers-of-passage (Hunnicutt et al. 2016). In addition, both axonal terminals and fibers-of-passage were found in the CPu and auditory cortex (AC).

Figure 10.

Outputs of the nonlemniscal auditory thalamus. (A) Example images showing morphological characteristics of axonal fibers of the MGBd/SG and PIN/PP in several target brain areas. Scale bars, 50 μm. (B) Distribution of axonal fibers at 6 selected AP positions following AAV injections in the MGBd/SG (n = 3), PIN/PP (n = 3), and MGBm (n = 1). Darker color indicates larger value of fluorescent intensity index. Data from CB- and CR-Cre mice were analyzed collectively.

To make comparisons across the 3 subdivisions of the nonlemniscal auditory thalamus, axonal fibers detected from the images were mapped onto the brain atlas, which demonstrated distinctions in output patterns (Fig. 10B). Compared with the other 2 subdivisions, the PIN/PP sent their axons to many extra brain areas mostly associated with emotion, including the medial preoptic area (MPA), VMH, anterior/posteromedial part of the cortical amygdaloid nucleus (ACo and PMCo), basomedial amygdaloid nucleus (BMA), medial amygdaloid nucleus (Me), SPF, DLPAG, and IC (Fig. 10B, middle row). The CPu, La, and AC received innervation from all 3 subdivisions, and the SC received sparse inputs from the MGBd/SG and PIN/PP but not from the MGBm. Interestingly, AAV injection in the MGBd/SG of CB-Cre mice yielded highly branched axonal fibers and varicosities in almost all divisions of the MGB (Supplementary Fig. S8, top-left corner), including the MGBv, but this was not the case for the PIN/PP and MGBm. For brain areas outside the auditory thalamus, the overall distribution patterns of axonal fibers in CB- and CR-Cre mice were similar. It should be noted that the distribution of axonal terminals was not quantified in detail because our signal detection method was not sensitive enough to nonambiguously differentiate fluorescence signals from axon terminals and fibers-of-passage.

As we were particularly interested in the innervation pattern of the AC by the nonlemniscal auditory thalamus, we investigated the areal and laminar distribution of axonal fibers from different subdivisions. The MGBd/SG projected to a wide range of auditory cortices, with more fibers in the AuV/TeA and fibers observed in almost all layers (Fig. 11, top row). Regarding the PIN/PP, although its fibers were mostly found in the TeA and Ect, it also projected to the L1 of all auditory cortices, except the AuD (Fig. 11, middle row). Remarkably, axonal fibers from the MGBm were found in the superficial, middle, and deep layers of the Au1 and AuV (Fig. 11, bottom row). Thus, both the MGBd/SG and MGBm targeted the middle layers of the Au1, and all 3 subdivisions innervated the L1 of a wide range of cortices.

Figure 11.

Distribution of axonal fibers in the AC at 6 selected AP positions following AAV injection in the MGBd/SG (n = 3), PIN/PP (n = 3), and MGBm (n = 1). Darker color indicates larger value of fluorescent intensity index. Data from CB- and CR-Cre mice were analyzed collectively.

Discussion

By combining transgenic CB-/CR-Cre mice with virus-assisted retrograde/anterograde tracing, we revisited the connectivity profile of the nonlemniscal auditory thalamus. Despite the potential differences between mice and other species, our results suggest that CB-/CR-Cre mice could be a useful animal model for functional investigation of the nonlemniscal auditory thalamus. This conclusion is supported by a recent diffusion tensor imaging (DTI) study, which showed that the mouse and human medial geniculate nucleus connectivity is homologous (Keifer OP, Jr. et al. 2015). As relatively little difference in connectivity patterns was observed between CB+ and CR+ neurons, our following discussion focuses on the differences between our and previous studies, and on the connectivity differences across the subdivisions of the nonlemniscal auditory thalamus.

In addition to brain areas reported to be connected with the nonlemniscal auditory thalamus, several new ones were revealed by our tracing study. Retrogradely, the PT, which is involved in pupillary light reflex (Clarke and Ikeda 1985), the DR, which encodes reward signals (Li et al. 2016), and the CnF, which is important for the initiation of gait (Jordan et al. 2008; Alam et al. 2011), were identified following RV injections in the MGBd/SG. In terms of anterograde tracing, branched axonal fibers were detected in the DLPAG, which plays a critical role in animal flight behavior (Di Scala et al. 1987; Castilho et al. 2002; Kim et al. 2013). As the nonlemniscal auditory thalamus receives direct inputs from the cochlea nucleus and SOC (Schofield, Mellott, et al. 2014; Schofield, Motts, et al. 2014), our data suggested that this thalamic region might drive rapid behavioral responses to acoustic stimuli.

Although the connectivity between the nonlemniscal auditory thalamus and other brain areas was the major focus of this study, we also noted interesting connectivity within the auditory thalamus (Supplementary Fig. S8, top-left corner). Our results showed that the MGBd/SG projected diffusely to almost all other divisions of the auditory thalamus, including the MGBv. A previous study on cats showed that the MGBd receives diffuse terminals from the MGBv (Winer 1985). Although it is still unclear whether this projection pattern is specific to mice, it would not be surprising if this observation also applies to other species because the secondary auditory cortices, which are the main cortical targets of the nonlemniscal auditory thalamus, also project to the primary auditory cortices, which are the major cortical targets of the lemniscal auditory thalamus (Lee and Winer 2008a, 2008b). This finding suggests that, at least starting from the level of the thalamus, information transmission between the lemniscal and nonlemniscal auditory pathway is already bidirectional, and that the MGBd/SG may be able to transfer multimodal information to the MGBv to modulate neuronal activity in the Au1.

The connectivity between the nonlemniscal auditory thalamus and cortex was a focus of our research. A recent study also used CR-IRES-Cre mice to characterize the nonlemniscal auditory thalamic inputs to the AC (Vasquez-Lopez et al. 2017). Although our conclusion was generally in agreement, in that inputs from the nonlemniscal auditory thalamus were mostly restricted to the ventrally located auditory cortices (AuV/TeA in our study), 2 discrepancies were observed. First, Vasquez-Lopez et al. (2017) reported that projections from the MGBm/PIN provided only extremely sparse inputs to the AC. However, our data demonstrated that the MGBm sent locally dense axonal fibers to the superficial, middle, and deep layers of both the Au1 and AuV, and fibers from the PIN/PP were densely distributed in the ventral auditory cortical areas (Fig. 11). Second, Vasquez-Lopez et al. (2017) reported that the PP did not project to the AC. Although we were not able to achieve local AAV injections in the PP, PP-dominant retrograde tracing revealed strong feedback projections from L6 of the auditory cortices (Supplementary Fig. S7A), suggesting feedforward projections from the PP to AC, which has also been observed in rats (Linke and Schwegler 2000). The discrepancies discussed above may have arisen from the previous research using highly diluted viruses, suggesting that virus volume and titer may need to be considered if the focus of research is to characterize the pattern of connectivity. However, it is unsurprising that the AuV/TeA rather than other auditory cortical areas provided the strongest corticofugal (mostly feedback) inputs to each individual subdivision of the nonlemniscal auditory thalamus because the AuV/TeA was the major cortical target of these subdivisions.

A critical role of the nonlemniscal thalamus in transferring information between cortical areas was proposed recently (Theyel et al. 2010; Sherman 2016). Specifically, neurons in the L5 of primary sensory cortices may send driver inputs to the nonlemniscal thalamus, which may then transfer the information to the secondary cortices. In the auditory system, the MGBd is suggested to be the central component of this cortico-thalamo-cortical (CTC) circuit (Llano and Sherman 2008). In our study, however, very few L5 neurons in the AC were observed following injection of retrograde monosynaptic tracing virus into the MGBd/SG of both CB- and CR-Cre mice (Figs 3B and 4). As many L6 neurons were labeled, it is unlikely that the sparseness of L5 labeling was due to the inefficiency of viral infection. In fact, sparse retrogradely labeled L5 neurons have also been reported in Llano and Sherman (2008). Interestingly, injection of the same virus into the PIN/PP yielded the labeling of many L5 neurons in a wide range of auditory cortical areas, including the Au1. Could the low numbers of L5 neurons projecting to the MGBd/SG simply be sufficient for a CTC circuit? Could the PIN/PP play a more important role than the MGBd in this circuit? Or, does the CTC information transfer theory need to be revisited, at least in the auditory system? Functional studies are needed to address these important questions.

GABAergic inputs to the nonlemniscal auditory thalamus may arise from 3 sources. Our data showed that the MGBd/SG and PIN/PP were preferentially connected with 2 GABAergic centers, that is, the Rt and ZI. The difference between the Rt and ZI is that the Rt is innervated topographically by corticothalamic (L6) and thalamocortical fibers (Pinault 2004), whereas the ZI receives peripheral and L5 cortical inputs but no thalamic feedback (Bartho et al. 2002). Furthermore, the ZI preferentially innervates nonprimary thalamic relays (Power et al. 1999; Bartho et al. 2002). In addition, the nonlemniscal IC sent substantial inputs to each individual subdivision of the nonlemniscal auditory thalamus. Previous studies have shown that the IC is another major source of GABAergic inputs to the MGB (Winer and Larue 1996; Peruzzi et al. 1997; Mellott et al. 2014). Ito et al. (2009) demonstrated that GABAergic tectothalamic neurons in rats are usually large and surrounded by VGluT2+ axosomatic endings. These GABAergic inputs may be important for the regulation of firing patterns in thalamocortical neurons (Peruzzi et al. 1997). Further immunohistological experiments are necessary to determine to what extent CB+/CR+ neurons in the nonlemniscal auditory thalamus are innervated by GABAergic inputs.

There are several caveats in the interpretation of our tracing results. First, the diffusion of the virus in brain tissue may have compromised the spatial specificity of tracing. We initially injected only small volumes of the virus or diluted virus to avoid leakage; however, this approach resulted in very low efficiency of viral infection, leading to failure or incomplete tracing (please see our discussion about the tracing results of Vasquez-Lopez et al. 2017). Therefore, we injected larger volumes, which led to difficulty in distinguishing the input neurons and output fibers of the MGBd from those of the SG or those of the PIN from those of the PP. Thus, we had to combine data from adjacent subdivisions of the nonlemniscal auditory thalamus for further analysis. Secondly, due to potential virus leakage, we cannot exclude the possibility that some contribution from the lateral posterior thalamic nucleus (LP) may have been included due to its proximity to the SG and enrichment of CB/CR expression. Our retrograde tracing results showed that many neurons in the OP layer of the SC were labeled following RV injection in the MGBd/SG. As these neurons project strongly to the LP (Zingg et al. 2017), it is possible that the retrograde labeling in the SC was simply due to virus leakage into the LP. However, another tracing study, which locally deposited Miniruby into the SG of rats, observed distribution patterns of retrogradely labeled neurons in the SC similar to ours (Linke 1999a). In addition, AAV injection in the mouse SC yielded strong fluorescent labeling of fibers, thus demonstrating numerous varicosities in the SG (Allen Brain Atlas; http://connectivity.brain-map.org). Therefore, although the LP may have contributed to the SC labeling in our study, the MGBd/SG, which was the target of our injection, should have contributed as well. Thirdly, it is known, at least in the cat, that the MGBd itself may be comprised of several subdivisions, including the superficial dorsal nucleus, dorsal nucleus, deep dorsal nucleus, and even the SG (Morest 1964, 1965; Winer and Morest 1983b), and neurons in these subdivisions demonstrate differential neuronal morphologies and connection patterns. Although subdivisions in the mouse MGBd have not been reported, it is conceivable that such subdivisions exist. In the present study, neurons at different depths in the MGBd were not always evenly infected as the distribution of infected neurons was rather superficial in some cases. The inconsistency in infection depth may have led to variations in the distribution of axonal fibers in anterograde tracing and input neurons in retrograde tracing, thereby compromising the accuracy of related conclusions. More localized injections may be needed to adequately differentiate the distinction between potential MGBd subdivisions in connection patterns in mice. Finally, our conclusions regarding thalamocortical projections would be more informative and precise if the cortical areas in the present study were defined functionally. Studies using the optical approach, such as flavoprotein fluorescence imaging (Tsukano et al. 2016) and intrinsic imaging (Storace et al. 2012), have produced physiological maps that are more precise than the brain atlas for the auditory cortex in rodents. In addition, Storace et al. showed that the rostral MGBv, in which vesicular glutamate transporter 1 (VGluT1) is highly expressed, predominantly projects to the Au1 in rats. Theoretically, AAV injections into the rostral MGBv of VGluT1-Cre mice would help to precisely locate the Au1. However, both the optical approach and VGluT1-Cre mice were unavailable to us in this study. The adoption of these methods in future tracing studies would be of merit.

Funding

National Natural Science Foundation of China (31371115 and 81527901), Center for Brain-Inspired Computing Research, and Tsinghua IDG/McGovern Institute for Brain Research.

Notes

We thank C.E. Schreiner and D.T. Larue for critical comments and suggestions, and P. Cao, C. Chen, M. Cheng, Y. Li, Q. Feng, C. Zhan, J. Zhang, X. Zhang, and C. Zhu for technical assistance. Conflict of Interest: The authors declare no competing financial interests.

Author notes

Dongqin Cai and Yin Yue contributed equally to this work.